2010-06-03 21:24:43 +08:00
|
|
|
#include <stdio.h>
|
|
|
|
#include <stdlib.h>
|
|
|
|
#include <netcdf.h>
|
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
void
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
check_err(const int stat, int line, const char* file, const char* func) {
|
2010-06-03 21:24:43 +08:00
|
|
|
if (stat != NC_NOERR) {
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
(void)fprintf(stderr,"line %d of %s.%s: %s\n", line, file, func, nc_strerror(stat));
|
2017-11-12 08:33:51 +08:00
|
|
|
fflush(stderr);
|
2010-06-03 21:24:43 +08:00
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
#define CHECK_ERR(err) check_err(err, __LINE__, __FILE__, __func__)
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
int
|
2017-11-12 08:33:51 +08:00
|
|
|
main() {/* create ctest0_64.nc */
|
|
|
|
|
|
|
|
int stat; /* return status */
|
|
|
|
int ncid; /* netCDF id */
|
|
|
|
|
|
|
|
/* dimension ids */
|
|
|
|
int Dr_dim;
|
|
|
|
int D1_dim;
|
|
|
|
int D2_dim;
|
|
|
|
int D3_dim;
|
|
|
|
int dim_MINUS_name_MINUS_dashes_dim;
|
|
|
|
int dim_PERIOD_name_PERIOD_dots_dim;
|
|
|
|
int dim_PLUS_name_PLUS_plusses_dim;
|
|
|
|
int dim_ATSIGN_name_ATSIGN_ats_dim;
|
|
|
|
|
|
|
|
/* dimension lengths */
|
|
|
|
size_t Dr_len = NC_UNLIMITED;
|
|
|
|
size_t D1_len = 1;
|
|
|
|
size_t D2_len = 2;
|
|
|
|
size_t D3_len = 3;
|
|
|
|
size_t dim_MINUS_name_MINUS_dashes_len = 4;
|
|
|
|
size_t dim_PERIOD_name_PERIOD_dots_len = 5;
|
|
|
|
size_t dim_PLUS_name_PLUS_plusses_len = 6;
|
|
|
|
size_t dim_ATSIGN_name_ATSIGN_ats_len = 7;
|
|
|
|
|
|
|
|
/* variable ids */
|
|
|
|
int c_id;
|
|
|
|
int b_id;
|
|
|
|
int s_id;
|
|
|
|
int i_id;
|
|
|
|
int f_id;
|
|
|
|
int d_id;
|
|
|
|
int cr_id;
|
|
|
|
int br_id;
|
|
|
|
int sr_id;
|
|
|
|
int ir_id;
|
|
|
|
int fr_id;
|
|
|
|
int dr_id;
|
|
|
|
int c1_id;
|
|
|
|
int b1_id;
|
|
|
|
int s1_id;
|
|
|
|
int i1_id;
|
|
|
|
int f1_id;
|
|
|
|
int d1_id;
|
|
|
|
int c2_id;
|
|
|
|
int b2_id;
|
|
|
|
int s2_id;
|
|
|
|
int i2_id;
|
|
|
|
int f2_id;
|
|
|
|
int d2_id;
|
|
|
|
int c3_id;
|
|
|
|
int b3_id;
|
|
|
|
int s3_id;
|
|
|
|
int i3_id;
|
|
|
|
int f3_id;
|
|
|
|
int d3_id;
|
|
|
|
int cr1_id;
|
|
|
|
int br2_id;
|
|
|
|
int sr3_id;
|
|
|
|
int f11_id;
|
|
|
|
int d12_id;
|
|
|
|
int c13_id;
|
|
|
|
int s21_id;
|
|
|
|
int i22_id;
|
|
|
|
int f23_id;
|
|
|
|
int c31_id;
|
|
|
|
int b32_id;
|
|
|
|
int s33_id;
|
|
|
|
int sr11_id;
|
|
|
|
int ir12_id;
|
|
|
|
int fr13_id;
|
|
|
|
int cr21_id;
|
|
|
|
int br22_id;
|
|
|
|
int sr23_id;
|
|
|
|
int fr31_id;
|
|
|
|
int dr32_id;
|
|
|
|
int cr33_id;
|
|
|
|
int c111_id;
|
|
|
|
int b112_id;
|
|
|
|
int s113_id;
|
|
|
|
int f121_id;
|
|
|
|
int d122_id;
|
|
|
|
int c123_id;
|
|
|
|
int s131_id;
|
|
|
|
int i132_id;
|
|
|
|
int f133_id;
|
|
|
|
int f211_id;
|
|
|
|
int d212_id;
|
|
|
|
int c213_id;
|
|
|
|
int s221_id;
|
|
|
|
int i222_id;
|
|
|
|
int f223_id;
|
|
|
|
int c231_id;
|
|
|
|
int b232_id;
|
|
|
|
int s233_id;
|
|
|
|
int s311_id;
|
|
|
|
int i312_id;
|
|
|
|
int f313_id;
|
|
|
|
int var_MINUS_name_MINUS_dashes_id;
|
|
|
|
int var_PERIOD_name_PERIOD_dots_id;
|
|
|
|
int var_PLUS_name_PLUS_plusses_id;
|
|
|
|
int var_ATSIGN_name_ATSIGN_ats_id;
|
|
|
|
|
|
|
|
/* rank (number of dimensions) for each variable */
|
|
|
|
# define RANK_c 0
|
|
|
|
# define RANK_b 0
|
|
|
|
# define RANK_s 0
|
|
|
|
# define RANK_i 0
|
|
|
|
# define RANK_f 0
|
|
|
|
# define RANK_d 0
|
|
|
|
# define RANK_cr 1
|
|
|
|
# define RANK_br 1
|
|
|
|
# define RANK_sr 1
|
|
|
|
# define RANK_ir 1
|
|
|
|
# define RANK_fr 1
|
|
|
|
# define RANK_dr 1
|
|
|
|
# define RANK_c1 1
|
|
|
|
# define RANK_b1 1
|
|
|
|
# define RANK_s1 1
|
|
|
|
# define RANK_i1 1
|
|
|
|
# define RANK_f1 1
|
|
|
|
# define RANK_d1 1
|
|
|
|
# define RANK_c2 1
|
|
|
|
# define RANK_b2 1
|
|
|
|
# define RANK_s2 1
|
|
|
|
# define RANK_i2 1
|
|
|
|
# define RANK_f2 1
|
|
|
|
# define RANK_d2 1
|
|
|
|
# define RANK_c3 1
|
|
|
|
# define RANK_b3 1
|
|
|
|
# define RANK_s3 1
|
|
|
|
# define RANK_i3 1
|
|
|
|
# define RANK_f3 1
|
|
|
|
# define RANK_d3 1
|
|
|
|
# define RANK_cr1 2
|
|
|
|
# define RANK_br2 2
|
|
|
|
# define RANK_sr3 2
|
|
|
|
# define RANK_f11 2
|
|
|
|
# define RANK_d12 2
|
|
|
|
# define RANK_c13 2
|
|
|
|
# define RANK_s21 2
|
|
|
|
# define RANK_i22 2
|
|
|
|
# define RANK_f23 2
|
|
|
|
# define RANK_c31 2
|
|
|
|
# define RANK_b32 2
|
|
|
|
# define RANK_s33 2
|
|
|
|
# define RANK_sr11 3
|
|
|
|
# define RANK_ir12 3
|
|
|
|
# define RANK_fr13 3
|
|
|
|
# define RANK_cr21 3
|
|
|
|
# define RANK_br22 3
|
|
|
|
# define RANK_sr23 3
|
|
|
|
# define RANK_fr31 3
|
|
|
|
# define RANK_dr32 3
|
|
|
|
# define RANK_cr33 3
|
|
|
|
# define RANK_c111 3
|
|
|
|
# define RANK_b112 3
|
|
|
|
# define RANK_s113 3
|
|
|
|
# define RANK_f121 3
|
|
|
|
# define RANK_d122 3
|
|
|
|
# define RANK_c123 3
|
|
|
|
# define RANK_s131 3
|
|
|
|
# define RANK_i132 3
|
|
|
|
# define RANK_f133 3
|
|
|
|
# define RANK_f211 3
|
|
|
|
# define RANK_d212 3
|
|
|
|
# define RANK_c213 3
|
|
|
|
# define RANK_s221 3
|
|
|
|
# define RANK_i222 3
|
|
|
|
# define RANK_f223 3
|
|
|
|
# define RANK_c231 3
|
|
|
|
# define RANK_b232 3
|
|
|
|
# define RANK_s233 3
|
|
|
|
# define RANK_s311 3
|
|
|
|
# define RANK_i312 3
|
|
|
|
# define RANK_f313 3
|
|
|
|
# define RANK_var_MINUS_name_MINUS_dashes 0
|
|
|
|
# define RANK_var_PERIOD_name_PERIOD_dots 0
|
|
|
|
# define RANK_var_PLUS_name_PLUS_plusses 0
|
|
|
|
# define RANK_var_ATSIGN_name_ATSIGN_ats 0
|
|
|
|
|
|
|
|
/* variable shapes */
|
|
|
|
int cr_dims[RANK_cr];
|
|
|
|
int br_dims[RANK_br];
|
|
|
|
int sr_dims[RANK_sr];
|
|
|
|
int ir_dims[RANK_ir];
|
|
|
|
int fr_dims[RANK_fr];
|
|
|
|
int dr_dims[RANK_dr];
|
|
|
|
int c1_dims[RANK_c1];
|
|
|
|
int b1_dims[RANK_b1];
|
|
|
|
int s1_dims[RANK_s1];
|
|
|
|
int i1_dims[RANK_i1];
|
|
|
|
int f1_dims[RANK_f1];
|
|
|
|
int d1_dims[RANK_d1];
|
|
|
|
int c2_dims[RANK_c2];
|
|
|
|
int b2_dims[RANK_b2];
|
|
|
|
int s2_dims[RANK_s2];
|
|
|
|
int i2_dims[RANK_i2];
|
|
|
|
int f2_dims[RANK_f2];
|
|
|
|
int d2_dims[RANK_d2];
|
|
|
|
int c3_dims[RANK_c3];
|
|
|
|
int b3_dims[RANK_b3];
|
|
|
|
int s3_dims[RANK_s3];
|
|
|
|
int i3_dims[RANK_i3];
|
|
|
|
int f3_dims[RANK_f3];
|
|
|
|
int d3_dims[RANK_d3];
|
|
|
|
int cr1_dims[RANK_cr1];
|
|
|
|
int br2_dims[RANK_br2];
|
|
|
|
int sr3_dims[RANK_sr3];
|
|
|
|
int f11_dims[RANK_f11];
|
|
|
|
int d12_dims[RANK_d12];
|
|
|
|
int c13_dims[RANK_c13];
|
|
|
|
int s21_dims[RANK_s21];
|
|
|
|
int i22_dims[RANK_i22];
|
|
|
|
int f23_dims[RANK_f23];
|
|
|
|
int c31_dims[RANK_c31];
|
|
|
|
int b32_dims[RANK_b32];
|
|
|
|
int s33_dims[RANK_s33];
|
|
|
|
int sr11_dims[RANK_sr11];
|
|
|
|
int ir12_dims[RANK_ir12];
|
|
|
|
int fr13_dims[RANK_fr13];
|
|
|
|
int cr21_dims[RANK_cr21];
|
|
|
|
int br22_dims[RANK_br22];
|
|
|
|
int sr23_dims[RANK_sr23];
|
|
|
|
int fr31_dims[RANK_fr31];
|
|
|
|
int dr32_dims[RANK_dr32];
|
|
|
|
int cr33_dims[RANK_cr33];
|
|
|
|
int c111_dims[RANK_c111];
|
|
|
|
int b112_dims[RANK_b112];
|
|
|
|
int s113_dims[RANK_s113];
|
|
|
|
int f121_dims[RANK_f121];
|
|
|
|
int d122_dims[RANK_d122];
|
|
|
|
int c123_dims[RANK_c123];
|
|
|
|
int s131_dims[RANK_s131];
|
|
|
|
int i132_dims[RANK_i132];
|
|
|
|
int f133_dims[RANK_f133];
|
|
|
|
int f211_dims[RANK_f211];
|
|
|
|
int d212_dims[RANK_d212];
|
|
|
|
int c213_dims[RANK_c213];
|
|
|
|
int s221_dims[RANK_s221];
|
|
|
|
int i222_dims[RANK_i222];
|
|
|
|
int f223_dims[RANK_f223];
|
|
|
|
int c231_dims[RANK_c231];
|
|
|
|
int b232_dims[RANK_b232];
|
|
|
|
int s233_dims[RANK_s233];
|
|
|
|
int s311_dims[RANK_s311];
|
|
|
|
int i312_dims[RANK_i312];
|
|
|
|
int f313_dims[RANK_f313];
|
|
|
|
|
|
|
|
/* enter define mode */
|
|
|
|
stat = nc_create("ctest0_64.nc", NC_CLOBBER|NC_64BIT_OFFSET, &ncid);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
/* define dimensions */
|
|
|
|
stat = nc_def_dim(ncid, "Dr", Dr_len, &Dr_dim);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_def_dim(ncid, "D1", D1_len, &D1_dim);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_def_dim(ncid, "D2", D2_len, &D2_dim);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_def_dim(ncid, "D3", D3_len, &D3_dim);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_def_dim(ncid, "dim-name-dashes", dim_MINUS_name_MINUS_dashes_len, &dim_MINUS_name_MINUS_dashes_dim);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_def_dim(ncid, "dim.name.dots", dim_PERIOD_name_PERIOD_dots_len, &dim_PERIOD_name_PERIOD_dots_dim);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_def_dim(ncid, "dim+name+plusses", dim_PLUS_name_PLUS_plusses_len, &dim_PLUS_name_PLUS_plusses_dim);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_def_dim(ncid, "dim@name@ats", dim_ATSIGN_name_ATSIGN_ats_len, &dim_ATSIGN_name_ATSIGN_ats_dim);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
/* define variables */
|
|
|
|
|
|
|
|
stat = nc_def_var(ncid, "c", NC_CHAR, RANK_c, 0, &c_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
stat = nc_def_var(ncid, "b", NC_BYTE, RANK_b, 0, &b_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
stat = nc_def_var(ncid, "s", NC_SHORT, RANK_s, 0, &s_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
stat = nc_def_var(ncid, "i", NC_INT, RANK_i, 0, &i_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
stat = nc_def_var(ncid, "f", NC_FLOAT, RANK_f, 0, &f_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
stat = nc_def_var(ncid, "d", NC_DOUBLE, RANK_d, 0, &d_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
cr_dims[0] = Dr_dim;
|
|
|
|
stat = nc_def_var(ncid, "cr", NC_CHAR, RANK_cr, cr_dims, &cr_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
br_dims[0] = Dr_dim;
|
|
|
|
stat = nc_def_var(ncid, "br", NC_BYTE, RANK_br, br_dims, &br_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
sr_dims[0] = Dr_dim;
|
|
|
|
stat = nc_def_var(ncid, "sr", NC_SHORT, RANK_sr, sr_dims, &sr_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
ir_dims[0] = Dr_dim;
|
|
|
|
stat = nc_def_var(ncid, "ir", NC_INT, RANK_ir, ir_dims, &ir_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
fr_dims[0] = Dr_dim;
|
|
|
|
stat = nc_def_var(ncid, "fr", NC_FLOAT, RANK_fr, fr_dims, &fr_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
dr_dims[0] = Dr_dim;
|
|
|
|
stat = nc_def_var(ncid, "dr", NC_DOUBLE, RANK_dr, dr_dims, &dr_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
c1_dims[0] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "c1", NC_CHAR, RANK_c1, c1_dims, &c1_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
b1_dims[0] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "b1", NC_BYTE, RANK_b1, b1_dims, &b1_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
s1_dims[0] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "s1", NC_SHORT, RANK_s1, s1_dims, &s1_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
i1_dims[0] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "i1", NC_INT, RANK_i1, i1_dims, &i1_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
f1_dims[0] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "f1", NC_FLOAT, RANK_f1, f1_dims, &f1_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
d1_dims[0] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "d1", NC_DOUBLE, RANK_d1, d1_dims, &d1_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
c2_dims[0] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "c2", NC_CHAR, RANK_c2, c2_dims, &c2_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
b2_dims[0] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "b2", NC_BYTE, RANK_b2, b2_dims, &b2_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
s2_dims[0] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "s2", NC_SHORT, RANK_s2, s2_dims, &s2_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
i2_dims[0] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "i2", NC_INT, RANK_i2, i2_dims, &i2_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
f2_dims[0] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "f2", NC_FLOAT, RANK_f2, f2_dims, &f2_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
d2_dims[0] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "d2", NC_DOUBLE, RANK_d2, d2_dims, &d2_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
c3_dims[0] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "c3", NC_CHAR, RANK_c3, c3_dims, &c3_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
b3_dims[0] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "b3", NC_BYTE, RANK_b3, b3_dims, &b3_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
s3_dims[0] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "s3", NC_SHORT, RANK_s3, s3_dims, &s3_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
i3_dims[0] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "i3", NC_INT, RANK_i3, i3_dims, &i3_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
f3_dims[0] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "f3", NC_FLOAT, RANK_f3, f3_dims, &f3_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
d3_dims[0] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "d3", NC_DOUBLE, RANK_d3, d3_dims, &d3_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
cr1_dims[0] = Dr_dim;
|
|
|
|
cr1_dims[1] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "cr1", NC_CHAR, RANK_cr1, cr1_dims, &cr1_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
br2_dims[0] = Dr_dim;
|
|
|
|
br2_dims[1] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "br2", NC_BYTE, RANK_br2, br2_dims, &br2_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
sr3_dims[0] = Dr_dim;
|
|
|
|
sr3_dims[1] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "sr3", NC_SHORT, RANK_sr3, sr3_dims, &sr3_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
f11_dims[0] = D1_dim;
|
|
|
|
f11_dims[1] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "f11", NC_FLOAT, RANK_f11, f11_dims, &f11_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
d12_dims[0] = D1_dim;
|
|
|
|
d12_dims[1] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "d12", NC_DOUBLE, RANK_d12, d12_dims, &d12_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
c13_dims[0] = D1_dim;
|
|
|
|
c13_dims[1] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "c13", NC_CHAR, RANK_c13, c13_dims, &c13_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
s21_dims[0] = D2_dim;
|
|
|
|
s21_dims[1] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "s21", NC_SHORT, RANK_s21, s21_dims, &s21_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
i22_dims[0] = D2_dim;
|
|
|
|
i22_dims[1] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "i22", NC_INT, RANK_i22, i22_dims, &i22_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
f23_dims[0] = D2_dim;
|
|
|
|
f23_dims[1] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "f23", NC_FLOAT, RANK_f23, f23_dims, &f23_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
c31_dims[0] = D3_dim;
|
|
|
|
c31_dims[1] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "c31", NC_CHAR, RANK_c31, c31_dims, &c31_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
b32_dims[0] = D3_dim;
|
|
|
|
b32_dims[1] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "b32", NC_BYTE, RANK_b32, b32_dims, &b32_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
s33_dims[0] = D3_dim;
|
|
|
|
s33_dims[1] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "s33", NC_SHORT, RANK_s33, s33_dims, &s33_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
sr11_dims[0] = Dr_dim;
|
|
|
|
sr11_dims[1] = D1_dim;
|
|
|
|
sr11_dims[2] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "sr11", NC_SHORT, RANK_sr11, sr11_dims, &sr11_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
ir12_dims[0] = Dr_dim;
|
|
|
|
ir12_dims[1] = D1_dim;
|
|
|
|
ir12_dims[2] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "ir12", NC_INT, RANK_ir12, ir12_dims, &ir12_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
fr13_dims[0] = Dr_dim;
|
|
|
|
fr13_dims[1] = D1_dim;
|
|
|
|
fr13_dims[2] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "fr13", NC_FLOAT, RANK_fr13, fr13_dims, &fr13_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
cr21_dims[0] = Dr_dim;
|
|
|
|
cr21_dims[1] = D2_dim;
|
|
|
|
cr21_dims[2] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "cr21", NC_CHAR, RANK_cr21, cr21_dims, &cr21_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
br22_dims[0] = Dr_dim;
|
|
|
|
br22_dims[1] = D2_dim;
|
|
|
|
br22_dims[2] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "br22", NC_BYTE, RANK_br22, br22_dims, &br22_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
sr23_dims[0] = Dr_dim;
|
|
|
|
sr23_dims[1] = D2_dim;
|
|
|
|
sr23_dims[2] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "sr23", NC_SHORT, RANK_sr23, sr23_dims, &sr23_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
fr31_dims[0] = Dr_dim;
|
|
|
|
fr31_dims[1] = D3_dim;
|
|
|
|
fr31_dims[2] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "fr31", NC_FLOAT, RANK_fr31, fr31_dims, &fr31_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
dr32_dims[0] = Dr_dim;
|
|
|
|
dr32_dims[1] = D3_dim;
|
|
|
|
dr32_dims[2] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "dr32", NC_DOUBLE, RANK_dr32, dr32_dims, &dr32_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
cr33_dims[0] = Dr_dim;
|
|
|
|
cr33_dims[1] = D3_dim;
|
|
|
|
cr33_dims[2] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "cr33", NC_CHAR, RANK_cr33, cr33_dims, &cr33_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
c111_dims[0] = D1_dim;
|
|
|
|
c111_dims[1] = D1_dim;
|
|
|
|
c111_dims[2] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "c111", NC_CHAR, RANK_c111, c111_dims, &c111_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
b112_dims[0] = D1_dim;
|
|
|
|
b112_dims[1] = D1_dim;
|
|
|
|
b112_dims[2] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "b112", NC_BYTE, RANK_b112, b112_dims, &b112_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
s113_dims[0] = D1_dim;
|
|
|
|
s113_dims[1] = D1_dim;
|
|
|
|
s113_dims[2] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "s113", NC_SHORT, RANK_s113, s113_dims, &s113_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
f121_dims[0] = D1_dim;
|
|
|
|
f121_dims[1] = D2_dim;
|
|
|
|
f121_dims[2] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "f121", NC_FLOAT, RANK_f121, f121_dims, &f121_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
d122_dims[0] = D1_dim;
|
|
|
|
d122_dims[1] = D2_dim;
|
|
|
|
d122_dims[2] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "d122", NC_DOUBLE, RANK_d122, d122_dims, &d122_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
c123_dims[0] = D1_dim;
|
|
|
|
c123_dims[1] = D2_dim;
|
|
|
|
c123_dims[2] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "c123", NC_CHAR, RANK_c123, c123_dims, &c123_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
s131_dims[0] = D1_dim;
|
|
|
|
s131_dims[1] = D3_dim;
|
|
|
|
s131_dims[2] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "s131", NC_SHORT, RANK_s131, s131_dims, &s131_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
i132_dims[0] = D1_dim;
|
|
|
|
i132_dims[1] = D3_dim;
|
|
|
|
i132_dims[2] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "i132", NC_INT, RANK_i132, i132_dims, &i132_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
f133_dims[0] = D1_dim;
|
|
|
|
f133_dims[1] = D3_dim;
|
|
|
|
f133_dims[2] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "f133", NC_FLOAT, RANK_f133, f133_dims, &f133_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
f211_dims[0] = D2_dim;
|
|
|
|
f211_dims[1] = D1_dim;
|
|
|
|
f211_dims[2] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "f211", NC_FLOAT, RANK_f211, f211_dims, &f211_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
d212_dims[0] = D2_dim;
|
|
|
|
d212_dims[1] = D1_dim;
|
|
|
|
d212_dims[2] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "d212", NC_DOUBLE, RANK_d212, d212_dims, &d212_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
c213_dims[0] = D2_dim;
|
|
|
|
c213_dims[1] = D1_dim;
|
|
|
|
c213_dims[2] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "c213", NC_CHAR, RANK_c213, c213_dims, &c213_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
s221_dims[0] = D2_dim;
|
|
|
|
s221_dims[1] = D2_dim;
|
|
|
|
s221_dims[2] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "s221", NC_SHORT, RANK_s221, s221_dims, &s221_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
i222_dims[0] = D2_dim;
|
|
|
|
i222_dims[1] = D2_dim;
|
|
|
|
i222_dims[2] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "i222", NC_INT, RANK_i222, i222_dims, &i222_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
f223_dims[0] = D2_dim;
|
|
|
|
f223_dims[1] = D2_dim;
|
|
|
|
f223_dims[2] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "f223", NC_FLOAT, RANK_f223, f223_dims, &f223_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
c231_dims[0] = D2_dim;
|
|
|
|
c231_dims[1] = D3_dim;
|
|
|
|
c231_dims[2] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "c231", NC_CHAR, RANK_c231, c231_dims, &c231_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
b232_dims[0] = D2_dim;
|
|
|
|
b232_dims[1] = D3_dim;
|
|
|
|
b232_dims[2] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "b232", NC_BYTE, RANK_b232, b232_dims, &b232_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
s233_dims[0] = D2_dim;
|
|
|
|
s233_dims[1] = D3_dim;
|
|
|
|
s233_dims[2] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "s233", NC_SHORT, RANK_s233, s233_dims, &s233_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
s311_dims[0] = D3_dim;
|
|
|
|
s311_dims[1] = D1_dim;
|
|
|
|
s311_dims[2] = D1_dim;
|
|
|
|
stat = nc_def_var(ncid, "s311", NC_SHORT, RANK_s311, s311_dims, &s311_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
i312_dims[0] = D3_dim;
|
|
|
|
i312_dims[1] = D1_dim;
|
|
|
|
i312_dims[2] = D2_dim;
|
|
|
|
stat = nc_def_var(ncid, "i312", NC_INT, RANK_i312, i312_dims, &i312_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
f313_dims[0] = D3_dim;
|
|
|
|
f313_dims[1] = D1_dim;
|
|
|
|
f313_dims[2] = D3_dim;
|
|
|
|
stat = nc_def_var(ncid, "f313", NC_FLOAT, RANK_f313, f313_dims, &f313_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
stat = nc_def_var(ncid, "var-name-dashes", NC_DOUBLE, RANK_var_MINUS_name_MINUS_dashes, 0, &var_MINUS_name_MINUS_dashes_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
stat = nc_def_var(ncid, "var.name.dots", NC_DOUBLE, RANK_var_PERIOD_name_PERIOD_dots, 0, &var_PERIOD_name_PERIOD_dots_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
stat = nc_def_var(ncid, "var+name+plusses", NC_DOUBLE, RANK_var_PLUS_name_PLUS_plusses, 0, &var_PLUS_name_PLUS_plusses_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
stat = nc_def_var(ncid, "var@name@ats", NC_DOUBLE, RANK_var_ATSIGN_name_ATSIGN_ats, 0, &var_ATSIGN_name_ATSIGN_ats_id);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
/* assign global attributes */
|
|
|
|
|
|
|
|
{
|
|
|
|
stat = nc_put_att_text(ncid, NC_GLOBAL, "Gc", 1, "");
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const signed char c0_Gb_att[2] = {-128, 127} ;
|
|
|
|
stat = nc_put_att_schar(ncid, NC_GLOBAL, "Gb", NC_BYTE, 2, c0_Gb_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const short c0_Gs_att[3] = {-32768, 0, 32767} ;
|
|
|
|
stat = nc_put_att_short(ncid, NC_GLOBAL, "Gs", NC_SHORT, 3, c0_Gs_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const int c0_Gi_att[3] = {-2147483647, 0, 2147483647} ;
|
|
|
|
stat = nc_put_att_int(ncid, NC_GLOBAL, "Gi", NC_INT, 3, c0_Gi_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const float c0_Gf_att[3] = {((float)-9.9999996e+35), ((float)0), ((float)9.9999996e+35)} ;
|
|
|
|
stat = nc_put_att_float(ncid, NC_GLOBAL, "Gf", NC_FLOAT, 3, c0_Gf_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const double c0_Gd_att[3] = {((double)-1e+308), ((double)0), ((double)1e+308)} ;
|
|
|
|
stat = nc_put_att_double(ncid, NC_GLOBAL, "Gd", NC_DOUBLE, 3, c0_Gd_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const int c0_Gatt_MINUS_name_MINUS_dashes_att[1] = {-1} ;
|
|
|
|
stat = nc_put_att_int(ncid, NC_GLOBAL, "Gatt-name-dashes", NC_INT, 1, c0_Gatt_MINUS_name_MINUS_dashes_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const int c0_Gatt_DOT_name_DOT_dots_att[1] = {-2} ;
|
|
|
|
stat = nc_put_att_int(ncid, NC_GLOBAL, "Gatt.name.dots", NC_INT, 1, c0_Gatt_DOT_name_DOT_dots_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const int c0_Gatt_PLUS_name_PLUS_plusses_att[1] = {-3} ;
|
|
|
|
stat = nc_put_att_int(ncid, NC_GLOBAL, "Gatt+name+plusses", NC_INT, 1, c0_Gatt_PLUS_name_PLUS_plusses_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const int c0_Gatt_ATSIGN_name_ATSIGN_ats_att[1] = {-4} ;
|
|
|
|
stat = nc_put_att_int(ncid, NC_GLOBAL, "Gatt@name@ats", NC_INT, 1, c0_Gatt_ATSIGN_name_ATSIGN_ats_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/* assign per-variable attributes */
|
|
|
|
|
|
|
|
{
|
|
|
|
static const int c0_att_MINUS_name_MINUS_dashes_att[1] = {4} ;
|
|
|
|
stat = nc_put_att_int(ncid, c_id, "att-name-dashes", NC_INT, 1, c0_att_MINUS_name_MINUS_dashes_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const int c0_att_DOT_name_DOT_dots_att[1] = {5} ;
|
|
|
|
stat = nc_put_att_int(ncid, c_id, "att.name.dots", NC_INT, 1, c0_att_DOT_name_DOT_dots_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const int c0_att_PLUS_name_PLUS_plusses_att[1] = {6} ;
|
|
|
|
stat = nc_put_att_int(ncid, c_id, "att+name+plusses", NC_INT, 1, c0_att_PLUS_name_PLUS_plusses_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const int c0_att_ATSIGN_name_ATSIGN_ats_att[1] = {7} ;
|
|
|
|
stat = nc_put_att_int(ncid, c_id, "att@name@ats", NC_INT, 1, c0_att_ATSIGN_name_ATSIGN_ats_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
stat = nc_put_att_text(ncid, b_id, "c", 1, "");
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const signed char c0_b_att[4] = {0, 127, -128, -1} ;
|
|
|
|
stat = nc_put_att_schar(ncid, s_id, "b", NC_BYTE, 4, c0_b_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const short c0_s_att[3] = {-32768, 0, 32767} ;
|
|
|
|
stat = nc_put_att_short(ncid, s_id, "s", NC_SHORT, 3, c0_s_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const int c0_i_att[3] = {-2147483647, 0, 2147483647} ;
|
|
|
|
stat = nc_put_att_int(ncid, i_id, "i", NC_INT, 3, c0_i_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const float c0_f_att[3] = {((float)-9.9999996e+35), ((float)0), ((float)9.9999996e+35)} ;
|
|
|
|
stat = nc_put_att_float(ncid, i_id, "f", NC_FLOAT, 3, c0_f_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
static const double c0_d_att[3] = {((double)-1e+308), ((double)0), ((double)1e+308)} ;
|
|
|
|
stat = nc_put_att_double(ncid, i_id, "d", NC_DOUBLE, 3, c0_d_att);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
stat = nc_put_att_text(ncid, f_id, "c", 1, "x");
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
stat = nc_put_att_text(ncid, d_id, "c", 8, "abcd\tZ$&");
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/* leave define mode */
|
|
|
|
stat = nc_enddef (ncid);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
/* assign variable data */
|
|
|
|
|
|
|
|
{
|
|
|
|
size_t count = 0;
|
|
|
|
static char c_data[1] = {'2'};
|
|
|
|
stat = nc_put_var1(ncid, c_id, &count, c_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
size_t count = 0;
|
|
|
|
static signed char b_data[1] = {-2};
|
|
|
|
stat = nc_put_var1(ncid, b_id, &count, b_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
size_t count = 0;
|
|
|
|
static short s_data[1] = {-5};
|
|
|
|
stat = nc_put_var1(ncid, s_id, &count, s_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
size_t count = 0;
|
|
|
|
static int i_data[1] = {-20};
|
|
|
|
stat = nc_put_var1(ncid, i_id, &count, i_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
size_t count = 0;
|
|
|
|
static float f_data[1] = {((float)-9)};
|
|
|
|
stat = nc_put_var1(ncid, f_id, &count, f_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
size_t count = 0;
|
|
|
|
static double d_data[1] = {((double)-10)};
|
|
|
|
stat = nc_put_var1(ncid, d_id, &count, d_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
char* cr_data = "ab" ;
|
|
|
|
size_t cr_startset[1] = {0} ;
|
|
|
|
size_t cr_countset[1] = {2};
|
|
|
|
stat = nc_put_vara(ncid, cr_id, cr_startset, cr_countset, cr_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
signed char br_data[2] = {-128, 127} ;
|
|
|
|
size_t br_startset[1] = {0} ;
|
|
|
|
size_t br_countset[1] = {2};
|
|
|
|
stat = nc_put_vara(ncid, br_id, br_startset, br_countset, br_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
short sr_data[2] = {-32768, 32767} ;
|
|
|
|
size_t sr_startset[1] = {0} ;
|
|
|
|
size_t sr_countset[1] = {2};
|
|
|
|
stat = nc_put_vara(ncid, sr_id, sr_startset, sr_countset, sr_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
int ir_data[2] = {-2147483646, 2147483647} ;
|
|
|
|
size_t ir_startset[1] = {0} ;
|
|
|
|
size_t ir_countset[1] = {2};
|
|
|
|
stat = nc_put_vara(ncid, ir_id, ir_startset, ir_countset, ir_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
float fr_data[2] = {((float)-9.9999996e+35), ((float)9.9999996e+35)} ;
|
|
|
|
size_t fr_startset[1] = {0} ;
|
|
|
|
size_t fr_countset[1] = {2};
|
|
|
|
stat = nc_put_vara(ncid, fr_id, fr_startset, fr_countset, fr_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
double dr_data[2] = {((double)-1e+308), ((double)1e+308)} ;
|
|
|
|
size_t dr_startset[1] = {0} ;
|
|
|
|
size_t dr_countset[1] = {2};
|
|
|
|
stat = nc_put_vara(ncid, dr_id, dr_startset, dr_countset, dr_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
char* c1_data = "\000" ;
|
|
|
|
size_t c1_startset[1] = {0} ;
|
|
|
|
size_t c1_countset[1] = {1};
|
|
|
|
stat = nc_put_vara(ncid, c1_id, c1_startset, c1_countset, c1_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
signed char b1_data[1] = {-128} ;
|
|
|
|
size_t b1_startset[1] = {0} ;
|
|
|
|
size_t b1_countset[1] = {1};
|
|
|
|
stat = nc_put_vara(ncid, b1_id, b1_startset, b1_countset, b1_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
short s1_data[1] = {-32768} ;
|
|
|
|
size_t s1_startset[1] = {0} ;
|
|
|
|
size_t s1_countset[1] = {1};
|
|
|
|
stat = nc_put_vara(ncid, s1_id, s1_startset, s1_countset, s1_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
int i1_data[1] = {-2147483646} ;
|
|
|
|
size_t i1_startset[1] = {0} ;
|
|
|
|
size_t i1_countset[1] = {1};
|
|
|
|
stat = nc_put_vara(ncid, i1_id, i1_startset, i1_countset, i1_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
float f1_data[1] = {((float)-9.9999996e+35)} ;
|
|
|
|
size_t f1_startset[1] = {0} ;
|
|
|
|
size_t f1_countset[1] = {1};
|
|
|
|
stat = nc_put_vara(ncid, f1_id, f1_startset, f1_countset, f1_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
double d1_data[1] = {((double)-1e+308)} ;
|
|
|
|
size_t d1_startset[1] = {0} ;
|
|
|
|
size_t d1_countset[1] = {1};
|
|
|
|
stat = nc_put_vara(ncid, d1_id, d1_startset, d1_countset, d1_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
char* c2_data = "ab" ;
|
|
|
|
size_t c2_startset[1] = {0} ;
|
|
|
|
size_t c2_countset[1] = {2};
|
|
|
|
stat = nc_put_vara(ncid, c2_id, c2_startset, c2_countset, c2_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
signed char b2_data[2] = {-128, 127} ;
|
|
|
|
size_t b2_startset[1] = {0} ;
|
|
|
|
size_t b2_countset[1] = {2};
|
|
|
|
stat = nc_put_vara(ncid, b2_id, b2_startset, b2_countset, b2_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
short s2_data[2] = {-32768, 32767} ;
|
|
|
|
size_t s2_startset[1] = {0} ;
|
|
|
|
size_t s2_countset[1] = {2};
|
|
|
|
stat = nc_put_vara(ncid, s2_id, s2_startset, s2_countset, s2_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
int i2_data[2] = {-2147483646, 2147483647} ;
|
|
|
|
size_t i2_startset[1] = {0} ;
|
|
|
|
size_t i2_countset[1] = {2};
|
|
|
|
stat = nc_put_vara(ncid, i2_id, i2_startset, i2_countset, i2_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
float f2_data[2] = {((float)-9.9999996e+35), ((float)9.9999996e+35)} ;
|
|
|
|
size_t f2_startset[1] = {0} ;
|
|
|
|
size_t f2_countset[1] = {2};
|
|
|
|
stat = nc_put_vara(ncid, f2_id, f2_startset, f2_countset, f2_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
double d2_data[2] = {((double)-1e+308), ((double)1e+308)} ;
|
|
|
|
size_t d2_startset[1] = {0} ;
|
|
|
|
size_t d2_countset[1] = {2};
|
|
|
|
stat = nc_put_vara(ncid, d2_id, d2_startset, d2_countset, d2_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
char* c3_data = "\001\177." ;
|
|
|
|
size_t c3_startset[1] = {0} ;
|
|
|
|
size_t c3_countset[1] = {3};
|
|
|
|
stat = nc_put_vara(ncid, c3_id, c3_startset, c3_countset, c3_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
signed char b3_data[3] = {-128, 127, -1} ;
|
|
|
|
size_t b3_startset[1] = {0} ;
|
|
|
|
size_t b3_countset[1] = {3};
|
|
|
|
stat = nc_put_vara(ncid, b3_id, b3_startset, b3_countset, b3_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
short s3_data[3] = {-32768, 0, 32767} ;
|
|
|
|
size_t s3_startset[1] = {0} ;
|
|
|
|
size_t s3_countset[1] = {3};
|
|
|
|
stat = nc_put_vara(ncid, s3_id, s3_startset, s3_countset, s3_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
int i3_data[3] = {-2147483646, 0, 2147483647} ;
|
|
|
|
size_t i3_startset[1] = {0} ;
|
|
|
|
size_t i3_countset[1] = {3};
|
|
|
|
stat = nc_put_vara(ncid, i3_id, i3_startset, i3_countset, i3_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
float f3_data[3] = {((float)-9.9999996e+35), ((float)0), ((float)9.9999996e+35)} ;
|
|
|
|
size_t f3_startset[1] = {0} ;
|
|
|
|
size_t f3_countset[1] = {3};
|
|
|
|
stat = nc_put_vara(ncid, f3_id, f3_startset, f3_countset, f3_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
double d3_data[3] = {((double)-1e+308), ((double)0), ((double)1e+308)} ;
|
|
|
|
size_t d3_startset[1] = {0} ;
|
|
|
|
size_t d3_countset[1] = {3};
|
|
|
|
stat = nc_put_vara(ncid, d3_id, d3_startset, d3_countset, d3_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
char* cr1_data = "xy" ;
|
|
|
|
size_t cr1_startset[2] = {0, 0} ;
|
|
|
|
size_t cr1_countset[2] = {2, 1};
|
|
|
|
stat = nc_put_vara(ncid, cr1_id, cr1_startset, cr1_countset, cr1_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
signed char br2_data[2] = {-24, -26} ;
|
2017-11-12 08:33:51 +08:00
|
|
|
size_t br2_startset[2] = {0, 0} ;
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
size_t br2_countset[2] = {1, 2};
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_put_vara(ncid, br2_id, br2_startset, br2_countset, br2_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
signed char br2_data[2] = {-20, -22} ;
|
|
|
|
size_t br2_startset[2] = {1, 0} ;
|
|
|
|
size_t br2_countset[2] = {1, 2};
|
|
|
|
stat = nc_put_vara(ncid, br2_id, br2_startset, br2_countset, br2_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
short sr3_data[3] = {-375, -380, -385} ;
|
2017-11-12 08:33:51 +08:00
|
|
|
size_t sr3_startset[2] = {0, 0} ;
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
size_t sr3_countset[2] = {1, 3};
|
|
|
|
stat = nc_put_vara(ncid, sr3_id, sr3_startset, sr3_countset, sr3_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
short sr3_data[3] = {-350, -355, -360} ;
|
|
|
|
size_t sr3_startset[2] = {1, 0} ;
|
|
|
|
size_t sr3_countset[2] = {1, 3};
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_put_vara(ncid, sr3_id, sr3_startset, sr3_countset, sr3_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
float f11_data[1] = {((float)-2187)} ;
|
|
|
|
size_t f11_startset[2] = {0, 0} ;
|
|
|
|
size_t f11_countset[2] = {1, 1};
|
|
|
|
stat = nc_put_vara(ncid, f11_id, f11_startset, f11_countset, f11_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
double d12_data[2] = {((double)-3000), ((double)-3010)} ;
|
|
|
|
size_t d12_startset[2] = {0, 0} ;
|
|
|
|
size_t d12_countset[2] = {1, 2};
|
|
|
|
stat = nc_put_vara(ncid, d12_id, d12_startset, d12_countset, d12_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
char* c13_data = "\tb\177" ;
|
|
|
|
size_t c13_startset[2] = {0, 0} ;
|
|
|
|
size_t c13_countset[2] = {1, 3};
|
|
|
|
stat = nc_put_vara(ncid, c13_id, c13_startset, c13_countset, c13_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
short s21_data[2] = {-375, -350} ;
|
|
|
|
size_t s21_startset[2] = {0, 0} ;
|
|
|
|
size_t s21_countset[2] = {2, 1};
|
|
|
|
stat = nc_put_vara(ncid, s21_id, s21_startset, s21_countset, s21_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
int i22_data[4] = {-24000, -24020, -23600, -23620} ;
|
|
|
|
size_t i22_startset[2] = {0, 0} ;
|
|
|
|
size_t i22_countset[2] = {2, 2};
|
|
|
|
stat = nc_put_vara(ncid, i22_id, i22_startset, i22_countset, i22_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
float f23_data[6] = {((float)-2187), ((float)-2196), ((float)-2205), ((float)-2106), ((float)-2115), ((float)-2124)} ;
|
|
|
|
size_t f23_startset[2] = {0, 0} ;
|
|
|
|
size_t f23_countset[2] = {2, 3};
|
|
|
|
stat = nc_put_vara(ncid, f23_id, f23_startset, f23_countset, f23_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
char* c31_data = "+- " ;
|
|
|
|
size_t c31_startset[2] = {0, 0} ;
|
|
|
|
size_t c31_countset[2] = {3, 1};
|
|
|
|
stat = nc_put_vara(ncid, c31_id, c31_startset, c31_countset, c31_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
signed char b32_data[6] = {-24, -26, -20, -22, -16, -18} ;
|
|
|
|
size_t b32_startset[2] = {0, 0} ;
|
|
|
|
size_t b32_countset[2] = {3, 2};
|
|
|
|
stat = nc_put_vara(ncid, b32_id, b32_startset, b32_countset, b32_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
short s33_data[9] = {-375, -380, -385, -350, -355, -360, -325, -330, -335} ;
|
|
|
|
size_t s33_startset[2] = {0, 0} ;
|
|
|
|
size_t s33_countset[2] = {3, 3};
|
|
|
|
stat = nc_put_vara(ncid, s33_id, s33_startset, s33_countset, s33_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
short sr11_data[1] = {2500} ;
|
2017-11-12 08:33:51 +08:00
|
|
|
size_t sr11_startset[3] = {0, 0, 0} ;
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
size_t sr11_countset[3] = {1, 1, 1};
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_put_vara(ncid, sr11_id, sr11_startset, sr11_countset, sr11_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
short sr11_data[1] = {2375} ;
|
|
|
|
size_t sr11_startset[3] = {1, 0, 0} ;
|
|
|
|
size_t sr11_countset[3] = {1, 1, 1};
|
|
|
|
stat = nc_put_vara(ncid, sr11_id, sr11_startset, sr11_countset, sr11_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
int ir12_data[2] = {640000, 639980} ;
|
2017-11-12 08:33:51 +08:00
|
|
|
size_t ir12_startset[3] = {0, 0, 0} ;
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
size_t ir12_countset[3] = {1, 1, 2};
|
|
|
|
stat = nc_put_vara(ncid, ir12_id, ir12_startset, ir12_countset, ir12_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
int ir12_data[2] = {632000, 631980} ;
|
|
|
|
size_t ir12_startset[3] = {1, 0, 0} ;
|
|
|
|
size_t ir12_countset[3] = {1, 1, 2};
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_put_vara(ncid, ir12_id, ir12_startset, ir12_countset, ir12_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
float fr13_data[3] = {((float)26244), ((float)26235), ((float)26226)} ;
|
2017-11-12 08:33:51 +08:00
|
|
|
size_t fr13_startset[3] = {0, 0, 0} ;
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
size_t fr13_countset[3] = {1, 1, 3};
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_put_vara(ncid, fr13_id, fr13_startset, fr13_countset, fr13_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
float fr13_data[3] = {((float)25515), ((float)25506), ((float)25497)} ;
|
|
|
|
size_t fr13_startset[3] = {1, 0, 0} ;
|
|
|
|
size_t fr13_countset[3] = {1, 1, 3};
|
|
|
|
stat = nc_put_vara(ncid, fr13_id, fr13_startset, fr13_countset, fr13_data);
|
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
char* cr21_data = "@DHL" ;
|
|
|
|
size_t cr21_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t cr21_countset[3] = {2, 2, 1};
|
|
|
|
stat = nc_put_vara(ncid, cr21_id, cr21_startset, cr21_countset, cr21_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
signed char br22_data[2] = {64, 62} ;
|
2017-11-12 08:33:51 +08:00
|
|
|
size_t br22_startset[3] = {0, 0, 0} ;
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
size_t br22_countset[3] = {1, 1, 2};
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_put_vara(ncid, br22_id, br22_startset, br22_countset, br22_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
signed char br22_data[2] = {68, 66} ;
|
|
|
|
size_t br22_startset[3] = {0, 1, 0} ;
|
|
|
|
size_t br22_countset[3] = {1, 1, 2};
|
|
|
|
stat = nc_put_vara(ncid, br22_id, br22_startset, br22_countset, br22_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
signed char br22_data[2] = {56, 54} ;
|
|
|
|
size_t br22_startset[3] = {1, 0, 0} ;
|
|
|
|
size_t br22_countset[3] = {1, 1, 2};
|
|
|
|
stat = nc_put_vara(ncid, br22_id, br22_startset, br22_countset, br22_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
signed char br22_data[2] = {60, 58} ;
|
|
|
|
size_t br22_startset[3] = {1, 1, 0} ;
|
|
|
|
size_t br22_countset[3] = {1, 1, 2};
|
|
|
|
stat = nc_put_vara(ncid, br22_id, br22_startset, br22_countset, br22_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
short sr23_data[3] = {2500, 2495, 2490} ;
|
2017-11-12 08:33:51 +08:00
|
|
|
size_t sr23_startset[3] = {0, 0, 0} ;
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
size_t sr23_countset[3] = {1, 1, 3};
|
|
|
|
stat = nc_put_vara(ncid, sr23_id, sr23_startset, sr23_countset, sr23_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
short sr23_data[3] = {2525, 2520, 2515} ;
|
|
|
|
size_t sr23_startset[3] = {0, 1, 0} ;
|
|
|
|
size_t sr23_countset[3] = {1, 1, 3};
|
|
|
|
stat = nc_put_vara(ncid, sr23_id, sr23_startset, sr23_countset, sr23_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
short sr23_data[3] = {2375, 2370, 2365} ;
|
|
|
|
size_t sr23_startset[3] = {1, 0, 0} ;
|
|
|
|
size_t sr23_countset[3] = {1, 1, 3};
|
|
|
|
stat = nc_put_vara(ncid, sr23_id, sr23_startset, sr23_countset, sr23_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
short sr23_data[3] = {2400, 2395, 2390} ;
|
|
|
|
size_t sr23_startset[3] = {1, 1, 0} ;
|
|
|
|
size_t sr23_countset[3] = {1, 1, 3};
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_put_vara(ncid, sr23_id, sr23_startset, sr23_countset, sr23_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
float fr31_data[1] = {((float)26244)} ;
|
2017-11-12 08:33:51 +08:00
|
|
|
size_t fr31_startset[3] = {0, 0, 0} ;
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
size_t fr31_countset[3] = {1, 1, 1};
|
|
|
|
stat = nc_put_vara(ncid, fr31_id, fr31_startset, fr31_countset, fr31_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
float fr31_data[1] = {((float)26325)} ;
|
|
|
|
size_t fr31_startset[3] = {0, 1, 0} ;
|
|
|
|
size_t fr31_countset[3] = {1, 1, 1};
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_put_vara(ncid, fr31_id, fr31_startset, fr31_countset, fr31_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
float fr31_data[1] = {((float)26406)} ;
|
|
|
|
size_t fr31_startset[3] = {0, 2, 0} ;
|
|
|
|
size_t fr31_countset[3] = {1, 1, 1};
|
|
|
|
stat = nc_put_vara(ncid, fr31_id, fr31_startset, fr31_countset, fr31_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
float fr31_data[1] = {((float)25515)} ;
|
|
|
|
size_t fr31_startset[3] = {1, 0, 0} ;
|
|
|
|
size_t fr31_countset[3] = {1, 1, 1};
|
|
|
|
stat = nc_put_vara(ncid, fr31_id, fr31_startset, fr31_countset, fr31_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
float fr31_data[1] = {((float)25596)} ;
|
|
|
|
size_t fr31_startset[3] = {1, 1, 0} ;
|
|
|
|
size_t fr31_countset[3] = {1, 1, 1};
|
|
|
|
stat = nc_put_vara(ncid, fr31_id, fr31_startset, fr31_countset, fr31_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
float fr31_data[1] = {((float)25677)} ;
|
|
|
|
size_t fr31_startset[3] = {1, 2, 0} ;
|
|
|
|
size_t fr31_countset[3] = {1, 1, 1};
|
|
|
|
stat = nc_put_vara(ncid, fr31_id, fr31_startset, fr31_countset, fr31_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
double dr32_data[2] = {((double)40000), ((double)39990)} ;
|
2017-11-12 08:33:51 +08:00
|
|
|
size_t dr32_startset[3] = {0, 0, 0} ;
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
size_t dr32_countset[3] = {1, 1, 2};
|
|
|
|
stat = nc_put_vara(ncid, dr32_id, dr32_startset, dr32_countset, dr32_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
double dr32_data[2] = {((double)40100), ((double)40090)} ;
|
|
|
|
size_t dr32_startset[3] = {0, 1, 0} ;
|
|
|
|
size_t dr32_countset[3] = {1, 1, 2};
|
|
|
|
stat = nc_put_vara(ncid, dr32_id, dr32_startset, dr32_countset, dr32_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
double dr32_data[2] = {((double)40200), ((double)40190)} ;
|
|
|
|
size_t dr32_startset[3] = {0, 2, 0} ;
|
|
|
|
size_t dr32_countset[3] = {1, 1, 2};
|
|
|
|
stat = nc_put_vara(ncid, dr32_id, dr32_startset, dr32_countset, dr32_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
double dr32_data[2] = {((double)39000), ((double)38990)} ;
|
|
|
|
size_t dr32_startset[3] = {1, 0, 0} ;
|
|
|
|
size_t dr32_countset[3] = {1, 1, 2};
|
|
|
|
stat = nc_put_vara(ncid, dr32_id, dr32_startset, dr32_countset, dr32_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
double dr32_data[2] = {((double)39100), ((double)39090)} ;
|
|
|
|
size_t dr32_startset[3] = {1, 1, 0} ;
|
|
|
|
size_t dr32_countset[3] = {1, 1, 2};
|
|
|
|
stat = nc_put_vara(ncid, dr32_id, dr32_startset, dr32_countset, dr32_data);
|
|
|
|
CHECK_ERR(stat);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
double dr32_data[2] = {((double)39200), ((double)39190)} ;
|
|
|
|
size_t dr32_startset[3] = {1, 2, 0} ;
|
|
|
|
size_t dr32_countset[3] = {1, 1, 2};
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_put_vara(ncid, dr32_id, dr32_startset, dr32_countset, dr32_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
char* cr33_data = "1\000\000two3\000\0004\000\0005\000\000six" ;
|
|
|
|
size_t cr33_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t cr33_countset[3] = {2, 3, 3};
|
|
|
|
stat = nc_put_vara(ncid, cr33_id, cr33_startset, cr33_countset, cr33_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
char* c111_data = "@" ;
|
|
|
|
size_t c111_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t c111_countset[3] = {1, 1, 1};
|
|
|
|
stat = nc_put_vara(ncid, c111_id, c111_startset, c111_countset, c111_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
signed char b112_data[2] = {64, 62} ;
|
|
|
|
size_t b112_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t b112_countset[3] = {1, 1, 2};
|
|
|
|
stat = nc_put_vara(ncid, b112_id, b112_startset, b112_countset, b112_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
short s113_data[3] = {2500, 2495, 2490} ;
|
|
|
|
size_t s113_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t s113_countset[3] = {1, 1, 3};
|
|
|
|
stat = nc_put_vara(ncid, s113_id, s113_startset, s113_countset, s113_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
float f121_data[2] = {((float)26244), ((float)26325)} ;
|
|
|
|
size_t f121_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t f121_countset[3] = {1, 2, 1};
|
|
|
|
stat = nc_put_vara(ncid, f121_id, f121_startset, f121_countset, f121_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
double d122_data[4] = {((double)40000), ((double)39990), ((double)40100), ((double)40090)} ;
|
|
|
|
size_t d122_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t d122_countset[3] = {1, 2, 2};
|
|
|
|
stat = nc_put_vara(ncid, d122_id, d122_startset, d122_countset, d122_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
char* c123_data = "one2\000\000" ;
|
|
|
|
size_t c123_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t c123_countset[3] = {1, 2, 3};
|
|
|
|
stat = nc_put_vara(ncid, c123_id, c123_startset, c123_countset, c123_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
short s131_data[3] = {2500, 2525, 2550} ;
|
|
|
|
size_t s131_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t s131_countset[3] = {1, 3, 1};
|
|
|
|
stat = nc_put_vara(ncid, s131_id, s131_startset, s131_countset, s131_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
int i132_data[6] = {640000, 639980, 640400, 640380, 640800, 640780} ;
|
|
|
|
size_t i132_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t i132_countset[3] = {1, 3, 2};
|
|
|
|
stat = nc_put_vara(ncid, i132_id, i132_startset, i132_countset, i132_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
float f133_data[9] = {((float)26244), ((float)26235), ((float)26226), ((float)26325), ((float)26316), ((float)26307), ((float)26406), ((float)26397), ((float)26388)} ;
|
|
|
|
size_t f133_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t f133_countset[3] = {1, 3, 3};
|
|
|
|
stat = nc_put_vara(ncid, f133_id, f133_startset, f133_countset, f133_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
float f211_data[2] = {((float)26244), ((float)25515)} ;
|
|
|
|
size_t f211_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t f211_countset[3] = {2, 1, 1};
|
|
|
|
stat = nc_put_vara(ncid, f211_id, f211_startset, f211_countset, f211_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
double d212_data[4] = {((double)40000), ((double)39990), ((double)39000), ((double)38990)} ;
|
|
|
|
size_t d212_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t d212_countset[3] = {2, 1, 2};
|
|
|
|
stat = nc_put_vara(ncid, d212_id, d212_startset, d212_countset, d212_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
char* c213_data = "\000\000\000\000\000\000" ;
|
|
|
|
size_t c213_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t c213_countset[3] = {2, 1, 3};
|
|
|
|
stat = nc_put_vara(ncid, c213_id, c213_startset, c213_countset, c213_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
short s221_data[4] = {2500, 2525, 2375, 2400} ;
|
|
|
|
size_t s221_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t s221_countset[3] = {2, 2, 1};
|
|
|
|
stat = nc_put_vara(ncid, s221_id, s221_startset, s221_countset, s221_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
int i222_data[8] = {640000, 639980, 640400, 640380, 632000, 631980, 632400, 632380} ;
|
|
|
|
size_t i222_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t i222_countset[3] = {2, 2, 2};
|
|
|
|
stat = nc_put_vara(ncid, i222_id, i222_startset, i222_countset, i222_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
float f223_data[12] = {((float)26244), ((float)26235), ((float)26226), ((float)26325), ((float)26316), ((float)26307), ((float)25515), ((float)25506), ((float)25497), ((float)25596), ((float)25587), ((float)25578)} ;
|
|
|
|
size_t f223_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t f223_countset[3] = {2, 2, 3};
|
|
|
|
stat = nc_put_vara(ncid, f223_id, f223_startset, f223_countset, f223_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
char* c231_data = "@DHHLP" ;
|
|
|
|
size_t c231_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t c231_countset[3] = {2, 3, 1};
|
|
|
|
stat = nc_put_vara(ncid, c231_id, c231_startset, c231_countset, c231_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
signed char b232_data[12] = {64, 62, 68, 66, 72, 70, 56, 54, 60, 58, 64, 62} ;
|
|
|
|
size_t b232_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t b232_countset[3] = {2, 3, 2};
|
|
|
|
stat = nc_put_vara(ncid, b232_id, b232_startset, b232_countset, b232_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
short s233_data[18] = {2500, 2495, 2490, 2525, 2520, 2515, 2550, 2545, 2540, 2375, 2370, 2365, 2400, 2395, 2390, 2425, 2420, 2415} ;
|
|
|
|
size_t s233_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t s233_countset[3] = {2, 3, 3};
|
|
|
|
stat = nc_put_vara(ncid, s233_id, s233_startset, s233_countset, s233_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
short s311_data[3] = {2500, 2375, 2250} ;
|
|
|
|
size_t s311_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t s311_countset[3] = {3, 1, 1};
|
|
|
|
stat = nc_put_vara(ncid, s311_id, s311_startset, s311_countset, s311_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
int i312_data[6] = {640000, 639980, 632000, 631980, 624000, 623980} ;
|
|
|
|
size_t i312_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t i312_countset[3] = {3, 1, 2};
|
|
|
|
stat = nc_put_vara(ncid, i312_id, i312_startset, i312_countset, i312_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
float f313_data[9] = {((float)26244), ((float)26235), ((float)26226), ((float)25515), ((float)25506), ((float)25497), ((float)24786), ((float)24777), ((float)24768)} ;
|
|
|
|
size_t f313_startset[3] = {0, 0, 0} ;
|
|
|
|
size_t f313_countset[3] = {3, 1, 3};
|
|
|
|
stat = nc_put_vara(ncid, f313_id, f313_startset, f313_countset, f313_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
{
|
|
|
|
size_t count = 0;
|
|
|
|
static double var_MINUS_name_MINUS_dashes_data[1] = {((double)-1)};
|
|
|
|
stat = nc_put_var1(ncid, var_MINUS_name_MINUS_dashes_id, &count, var_MINUS_name_MINUS_dashes_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
size_t count = 0;
|
|
|
|
static double var_PERIOD_name_PERIOD_dots_data[1] = {((double)-2)};
|
|
|
|
stat = nc_put_var1(ncid, var_PERIOD_name_PERIOD_dots_id, &count, var_PERIOD_name_PERIOD_dots_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
size_t count = 0;
|
|
|
|
static double var_PLUS_name_PLUS_plusses_data[1] = {((double)9.969209968386869e+36)};
|
|
|
|
stat = nc_put_var1(ncid, var_PLUS_name_PLUS_plusses_id, &count, var_PLUS_name_PLUS_plusses_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
|
|
|
|
{
|
|
|
|
size_t count = 0;
|
|
|
|
static double var_ATSIGN_name_ATSIGN_ats_data[1] = {((double)9.969209968386869e+36)};
|
|
|
|
stat = nc_put_var1(ncid, var_ATSIGN_name_ATSIGN_ats_id, &count, var_ATSIGN_name_ATSIGN_ats_data);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
}
|
|
|
|
|
2010-06-03 21:24:43 +08:00
|
|
|
|
2017-11-12 08:33:51 +08:00
|
|
|
stat = nc_close(ncid);
|
This PR adds EXPERIMENTAL support for accessing data in the
cloud using a variant of the Zarr protocol and storage
format. This enhancement is generically referred to as "NCZarr".
The data model supported by NCZarr is netcdf-4 minus the user-defined
types and the String type. In this sense it is similar to the CDF-5
data model.
More detailed information about enabling and using NCZarr is
described in the document NUG/nczarr.md and in a
[Unidata Developer's blog entry](https://www.unidata.ucar.edu/blogs/developer/en/entry/overview-of-zarr-support-in).
WARNING: this code has had limited testing, so do use this version
for production work. Also, performance improvements are ongoing.
Note especially the following platform matrix of successful tests:
Platform | Build System | S3 support
------------------------------------
Linux+gcc | Automake | yes
Linux+gcc | CMake | yes
Visual Studio | CMake | no
Additionally, and as a consequence of the addition of NCZarr,
major changes have been made to the Filter API. NOTE: NCZarr
does not yet support filters, but these changes are enablers for
that support in the future. Note that it is possible
(probable?) that there will be some accidental reversions if the
changes here did not correctly mimic the existing filter testing.
In any case, previously filter ids and parameters were of type
unsigned int. In order to support the more general zarr filter
model, this was all converted to char*. The old HDF5-specific,
unsigned int operations are still supported but they are
wrappers around the new, char* based nc_filterx_XXX functions.
This entailed at least the following changes:
1. Added the files libdispatch/dfilterx.c and include/ncfilter.h
2. Some filterx utilities have been moved to libdispatch/daux.c
3. A new entry, "filter_actions" was added to the NCDispatch table
and the version bumped.
4. An overly complex set of structs was created to support funnelling
all of the filterx operations thru a single dispatch
"filter_actions" entry.
5. Move common code to from libhdf5 to libsrc4 so that it is accessible
to nczarr.
Changes directly related to Zarr:
1. Modified CMakeList.txt and configure.ac to support both C and C++
-- this is in support of S3 support via the awd-sdk libraries.
2. Define a size64_t type to support nczarr.
3. More reworking of libdispatch/dinfermodel.c to
support zarr and to regularize the structure of the fragments
section of a URL.
Changes not directly related to Zarr:
1. Make client-side filter registration be conditional, with default off.
2. Hack include/nc4internal.h to make some flags added by Ed be unique:
e.g. NC_CREAT, NC_INDEF, etc.
3. cleanup include/nchttp.h and libdispatch/dhttp.c.
4. Misc. changes to support compiling under Visual Studio including:
* Better testing under windows for dirent.h and opendir and closedir.
5. Misc. changes to the oc2 code to support various libcurl CURLOPT flags
and to centralize error reporting.
6. By default, suppress the vlen tests that have unfixed memory leaks; add option to enable them.
7. Make part of the nc_test/test_byterange.sh test be contingent on remotetest.unidata.ucar.edu being accessible.
Changes Left TO-DO:
1. fix provenance code, it is too HDF5 specific.
2020-06-29 08:02:47 +08:00
|
|
|
CHECK_ERR(stat);
|
2017-11-12 08:33:51 +08:00
|
|
|
return 0;
|
2010-06-03 21:24:43 +08:00
|
|
|
}
|