In non-classic netcdf-4 models, it is allowable to have
large numbers of dims and vars. In many operations, the
entire list of dims or vars is searched for a dim/var matching
a specific name which results in *lots* of strncmp or strcmp
calls.
If we add a hash field to the var and dim structs similar to what
has already been done for the netcdf-3 formats, then we can hash the
name being searched for and numerically compare that value with
the var/dim hash value. If they match, then do a more expensive
strncmp call to ensure that the names truly match.
It turns out that HDF4 supports chunking
(and compression). However the existing
HDF4 code does not support it.
So add HDF4 support for chunking.
Also add a test case.
error occurs after an "exit:" label.
Corrected a dozen Coverity errors (mainly allocation issues, along with a few
other things):
711711, 711802, 711803, 711905, 970825, 996123, 996124, 1025787,
1047274, 1130013, 1130014, 1139538
Refactored internal fill-value code to correctly handle string types, and
especially to allow NULL pointers and null strings (ie. "") to be
distinguished. The code now avoids partially aliasing the two together
(which only happened on the 'write' side of things and wasn't reflected on
the 'read' side, adding to the previous confusion).
Probably still weak on handling fill-values of variable-length and compound
datatypes.
Refactored the recursive metadata reads a bit more, to process HDF5 named
datatypes and datasets immediately, avoiding chewing up memory for those
types of objects, etc.
Finished uncommenting and updating the nc_test4/tst_fills2.c code (as I'm
proceeding alphabetically through the nc_test4 code files).
the rest of the dimension queries. Correct error in library where types used
in sub-group variables but that were added to the file after the sub-group was
created weren't available for sub-group variables to use. Start cleaning up
test suite and un-commenting tests that were commented out (got up to
nc_test4/tst_fills2.c, alphabetically) before running into an error in HDF5.
to clean up resources properly on failure.
Refactored doubly-linked list code for objects in the libsrc4 directory,
cleaning up the add/del routines, breaking out the common next/prev
pointers into a struct and extracting the add/del operations on them,
changed the list of dims to add new dims in the same order as the other
types, made all add routines able to optionally return a pointer to the
newly created object.
Removed some dead code (pg_var(), nc4_pg_var1(), nc4_pg_varm(), misc. small
routines, etc)
Fixed fill value handling for string types in nc4_get_vara().
Changed many malloc()+strcpy() pairs into calls to strdup().
Cleaned up misc. other minor Coverity issues.
many cleanups to fix compiler warnings, streamline iteration over objects
in HDF5 file when opening the file, and generally straightening out the code
to be cleaner and simpler.
Tested on Mac OS/X with gcc 4.8 and OpenMPI (which uses clang).
an unlimited-dimension variable, and different processes don't agree on the
whether to extend the underlying HDF5 dataset, or don't agree on the amount
to extend the dataset.
Also added ability capability for netCDF-4 to write and read NIL
values for string type attributes and variables, so these can be read
if used in HDF5 files.
Include are additions to CMakeLists files to reflect new tests.
unlimited dimension hanging. Extending the size of an unlimited
dimension in HDF5 must be a collective operation, so now an error is
returned if trying to extend in independent access mode.
Quincey's bug fixes for parallel build portability, particularly
OpenMPI on MacOS-X.
So, it turns out that just freeing
the nc4_info is not enough;
The root group must also be reclaimed.
So, it appears the best approach
is to invoke an abort on the
failed file.
coordinate variables and their associated dimensions occurs in any
subgroup, rather than just the root group. If this occurs, the
variable attribute "_Netcdf4Dimid" is created for every dimension
scale.
Also add a test for this bug fix in tst_dims3.nc, based on Pedro
Vicente's demo.
for JIRA issue NCF-241.
This is only temporary
until I can make pnetcdf
operate as a separate dispatch table.
Also, fix nc_test4/tst_pnetcdf
to open with nc_open_par;
this is necessary because a pnetcdf
created file cannot be opened
as a netcdf classic file.
Fix Jira issue NCF-29 (https://bugtracking.unidata.ucar.edu/browse/NCF-29):
making the netCDF-4 library ignore HDF5 datasets and attributes which have
datatypes (such as references) that it doesn't understand.
Tested on:
Mac OSX/64 10.8.2 (amazon) w/--enable-netcdf-4 --enable-extra-tests --enable-extra-example-tests
--disable-shared --enable-logging
CMake related changes in CMakeLists.txt files,
cmake_config.h.in. Other changes relate to
Windows-specific issues, and changes made
when regenerating generated source files.
with parallel-netcdf, to fully implement parallel-netcdf support for
other functions, and to prevent a hang in hdf5 from an eary return in
an nc4_put_vara() call. Also fixed an nccopy bug when
nc_inq_var_deflate() returns defalate_level of 0, but says the variable
is deflated.
Primarily focused on memory errors falling into a couple different types:
1) Static overrun errors.
2) Dereference uninitialized memory errors.
make distcheck works after applying these fixes, and coverity no longer sees an issue, so hopefully they are properly resolved.
contain as little file-type specific info as possible. It
modifies especially libsrc so that all of the netcdf-3 data
that used to be in struct NC is now kept in a separate chunk
of data pointed to by the struct NC. This makes all of
current protocols consistent: netcdf-3, netcdf-4, and dap.
The in-memory files can be made persistent if nc_create is called with
NC_DISKLESS|NC_WRITE flags set. Initial test case also included.
- Modified ncio mechanism to support
multiple ncio packages; this is so we
can have posixio and memio operating
at the same time.
- cleanup up a bunch of lint issues (unused variables, etc).
- Fix NCF-157 to modify DAP code to support
partial variable retrieval.
- Fix of NCF-154 to solve problem of ncgen
improperly processing data lists for variables
of size greater than 2**18 bytes.
- Fix ncgen processing of char variables that have
multiple unlimited dimensions.
- Partly fix Jira issue: NCF-145 (vlen issues).
- Benchmark program nc_test4/tst_ar4_*) requires arguments
and should only be invoked inside a shell
script; fixed so that they terminate cleanly
if invoked with no arguments.
- Fix the Doxygen processing so it will work
with make distcheck.
- Begin switchover to using an alternative to ncio.
- Begin support for in-memory (diskless) files.
NCF-42: _Format attribute sometimes being ignored
NCF-43: Fixed unsigned long long parsing.
NCF-47: Make opendap code properly handle illegal names like "x.y" by
supressing them
NCF-49: check for uint type
NCF-50: properly handle username:pwd embedded in urls.
