netcdf-c/libnczarr/zfile.c
Dennis Heimbigner 11fe00ea05 Add filter support to NCZarr
Filter support has three goals:

1. Use the existing HDF5 filter implementations,
2. Allow filter metadata to be stored in the NumCodecs metadata format used by Zarr,
3. Allow filters to be used even when HDF5 is disabled

Detailed usage directions are define in docs/filters.md.

For now, the existing filter API is left in place. So filters
are defined using ''nc_def_var_filter'' using the HDF5 style
where the id and parameters are unsigned integers.

This is a big change since filters affect many parts of the code.

In the following, the terms "compressor" and "filter" and "codec" are generally
used synonomously.

### Filter-Related Changes:
* In order to support dynamic loading of shared filter libraries, a new library was added in the libncpoco directory; it helps to isolate dynamic loading across multiple platforms.
* Provide a json parsing library for use by plugins; this is created by merging libdispatch/ncjson.c with include/ncjson.h.
* Add a new _Codecs attribute to allow clients to see what codecs are being used; let ncdump -s print it out.
* Provide special headers to help support compilation of HDF5 filters when HDF5 is not enabled: netcdf_filter_hdf5_build.h and netcdf_filter_build.h.
* Add a number of new test to test the new nczarr filters.
* Let ncgen parse _Codecs attribute, although it is ignored.

### Plugin directory changes:
* Add support for the Blosc compressor; this is essential because it is the most common compressor used in Zarr datasets. This also necessitated adding a CMake FindBlosc.cmake file
* Add NCZarr support for the big-four filters provided by HDF5: shuffle, fletcher32, deflate (zlib), and szip
* Add a Codec defaulter (see docs/filters.md) for the big four filters.
* Make plugins work with windows by properly adding __declspec declaration.

### Misc. Non-Filter Changes
* Replace most uses of USE_NETCDF4 (deprecated) with USE_HDF5.
* Improve support for caching
* More fixes for path conversion code
* Fix misc. memory leaks
* Add new utility -- ncdump/ncpathcvt -- that does more or less the same thing as cygpath.
* Add a number of new test to test the non-filter fixes.
* Update the parsers
* Convert most instances of '#ifdef _MSC_VER' to '#ifdef _WIN32'
2021-09-02 17:04:26 -06:00

478 lines
12 KiB
C

/* Copyright 2003-2018, University Corporation for Atmospheric
* Research. See COPYRIGHT file for copying and redistribution
* conditions. */
/**
* @file
* @internal The netCDF-4 file functions.
*
* This file is part of netcdf-4, a netCDF-like interface for NCZ, or
* a ZARR backend for netCDF, depending on your point of view.
*
* @author Dennis Heimbigner, Ed Hartnett
*/
#include "zincludes.h"
#include "zfilter.h"
/* Forward */
static int NCZ_enddef(int ncid);
static int ncz_sync_netcdf4_file(NC_FILE_INFO_T* file, int isclose);
/**
* @internal Put the file back in redef mode. This is done
* automatically for netcdf-4 files, if the user forgets.
*
* @param ncid File and group ID.
*
* @return ::NC_NOERR No error.
* @author Dennis Heimbigner, Ed Hartnett
*/
int
NCZ_redef(int ncid)
{
NC_FILE_INFO_T* zinfo = NULL;
int stat = NC_NOERR;
LOG((1, "%s: ncid 0x%x", __func__, ncid));
/* Find this file's metadata. */
if ((stat = nc4_find_grp_h5(ncid, NULL, &zinfo)))
return stat;
assert(zinfo);
/* If we're already in define mode, return an error. */
if (zinfo->flags & NC_INDEF)
return NC_EINDEFINE;
/* If the file is read-only, return an error. */
if (zinfo->no_write)
return NC_EPERM;
/* Set define mode. */
zinfo->flags |= NC_INDEF;
/* For nc_abort, we need to remember if we're in define mode as a
redef. */
zinfo->redef = NC_TRUE;
return NC_NOERR;
}
/**
* @internal For netcdf-4 files, this just calls nc_enddef, ignoring
* the extra parameters.
*
* @param ncid File and group ID.
* @param h_minfree Ignored for netCDF-4 files.
* @param v_align Ignored for netCDF-4 files.
* @param v_minfree Ignored for netCDF-4 files.
* @param r_align Ignored for netCDF-4 files.
*
* @return ::NC_NOERR No error.
* @author Dennis Heimbigner, Ed Hartnett
*/
int
NCZ__enddef(int ncid, size_t h_minfree, size_t v_align,
size_t v_minfree, size_t r_align)
{
int stat = NC_NOERR;
ZTRACE(0,"ncid=%d",ncid);
stat = NCZ_enddef(ncid);
return ZUNTRACE(stat);
}
/**
* @internal Take the file out of define mode. This is called
* automatically for netcdf-4 files, if the user forgets.
*
* @param ncid File and group ID.
*
* @return ::NC_NOERR No error.
* @return ::NC_EBADID Bad ncid.
* @return ::NC_EBADGRPID Bad group ID.
* @author Dennis Heimbigner, Ed Hartnett
*/
static int
NCZ_enddef(int ncid)
{
NC_FILE_INFO_T* h5 = NULL;
NC_GRP_INFO_T *grp;
NC_VAR_INFO_T *var;
int i,j;
int stat = NC_NOERR;
ZTRACE(1,"ncid=%d",ncid);
/* Find pointer to group and zinfo. */
if ((stat = nc4_find_grp_h5(ncid, &grp, &h5)))
goto done;
/* When exiting define mode, process all variables */
for (i = 0; i < nclistlength(h5->allgroups); i++) {
NC_GRP_INFO_T* g = nclistget(h5->allgroups,i);
for (j = 0; j < ncindexsize(g->vars); j++) {
var = (NC_VAR_INFO_T *)ncindexith(g->vars, j);
assert(var);
/* set the fill value and _FillValue attribute */
if((stat = ncz_get_fill_value(h5,var,NULL))) goto done; /* ensure var->fill_value is set */
assert(var->fill_value != NULL);
var->written_to = NC_TRUE; /* mark it written */
/* rebuild the fill chunk */
if((stat = NCZ_adjust_var_cache(var))) goto done;
/* Build the filter working parameters for any filters */
if((stat = NCZ_filter_setup(var))) goto done;
}
}
stat = ncz_enddef_netcdf4_file(h5);
done:
return ZUNTRACE(stat);
}
/**
* @internal Flushes all buffers associated with the file, after
* writing all changed metadata. This may only be called in data mode.
*
* @param ncid File and group ID.
*
* @return ::NC_NOERR No error.
* @return ::NC_EBADID Bad ncid.
* @return ::NC_EINDEFINE Classic model file is in define mode.
* @author Dennis Heimbigner, Ed Hartnett
*/
int
NCZ_sync(int ncid)
{
int stat = NC_NOERR;
NC_FILE_INFO_T* file = NULL;
ZTRACE(0,"ncid=%d",ncid);
LOG((2, "%s: ncid 0x%x", __func__, ncid));
if ((stat = nc4_find_grp_h5(ncid, NULL, &file)))
return stat;
assert(file);
/* If we're in define mode, we can't sync. */
if (file->flags & NC_INDEF)
{
if (file->cmode & NC_CLASSIC_MODEL)
return NC_EINDEFINE;
if ((stat = NCZ_enddef(ncid)))
return stat;
}
/* do not do this if file is writeonce */
stat = ncz_sync_netcdf4_file(file,!ZCLOSE);
return stat;
}
/**
* @internal From the netcdf-3 docs: The function nc_abort just closes
* the netCDF dataset, if not in define mode. If the dataset is being
* created and is still in define mode, the dataset is deleted. If
* define mode was entered by a call to nc_redef, the netCDF dataset
* is restored to its state before definition mode was entered and the
* dataset is closed.
*
* @param ncid File and group ID.
*
* @return ::NC_NOERR No error.
* @author Dennis Heimbigner, Ed Hartnett
*/
int
NCZ_abort(int ncid)
{
int stat = NC_NOERR;
NC_FILE_INFO_T* h5 = NULL;
ZTRACE(0,"ncid=%d",ncid);
LOG((2, "%s: ncid 0x%x", __func__, ncid));
/* Find metadata for this file. */
if ((stat = nc4_find_grp_h5(ncid, NULL, &h5)))
return stat;
assert(h5);
stat = ncz_closeorabort(h5, NULL, 1);
return ZUNTRACE(stat);
}
/**
* @internal Close the netcdf file, writing any changes first.
*
* @param ncid File and group ID.
* @param params any extra parameters in/out of close
*
* @return ::NC_NOERR No error.
* @author Dennis Heimbigner, Ed Hartnett
*/
int
NCZ_close(int ncid, void* params)
{
int stat = NC_NOERR;
NC_FILE_INFO_T* h5 = NULL;
ZTRACE(0,"ncid=%d",ncid);
LOG((1, "%s: ncid 0x%x", __func__, ncid));
/* Find metadata for this file. */
if ((stat = nc4_find_grp_h5(ncid, NULL, &h5)))
return stat;
assert(h5);
return ncz_closeorabort(h5, params, 0);
}
/**
* @internal From the netcdf-3 docs: The function nc_abort just closes
* the netCDF dataset, if not in define mode. If the dataset is being
* created and is still in define mode, the dataset is deleted. If
* define mode was entered by a call to nc_redef, the netCDF dataset
* is restored to its state before definition mode was entered and the
* dataset is closed.
*
* @param ncid File and group ID.
*
* @return ::NC_NOERR No error.
* @author Dennis Heimbigner, Ed Hartnett
*/
int
ncz_closeorabort(NC_FILE_INFO_T* h5, void* params, int abort)
{
int stat = NC_NOERR;
assert(h5);
NC_UNUSED(params);
ZTRACE(3,"file=%s abort=%d",h5->hdr.name,abort);
LOG((2, "%s: file: %p", __func__, h5));
/* If we're in define mode, but not redefing the file, delete it. */
if(!abort) {
/* Invoke enddef if needed, which mean sync first */
if(h5->flags & NC_INDEF) h5->flags ^= NC_INDEF;
/* Sync the file unless this is a read-only file. */
if(!h5->no_write) {
if((stat = ncz_sync_netcdf4_file(h5,ZCLOSE)))
goto done;
}
}
/* Reclaim memory */
/* Free any zarr-related data, including the map */
if ((stat = ncz_close_file(h5, abort)))
goto done;
/* Reclaim provenance info */
NCZ_clear_provenance(&h5->provenance);
/* Free the NC_FILE_INFO_T struct. */
if ((stat = nc4_nc4f_list_del(h5)))
return stat;
done:
return ZUNTRACE(stat);
}
/**************************************************/
/**
* @internal Learn number of dimensions, variables, global attributes,
* and the ID of the first unlimited dimension (if any).
*
* @note It's possible for any of these pointers to be NULL, in which
* case don't try to figure out that value.
*
* @param ncid File and group ID.
* @param ndimsp Pointer that gets number of dimensions.
* @param nvarsp Pointer that gets number of variables.
* @param nattsp Pointer that gets number of global attributes.
* @param unlimdimidp Pointer that gets first unlimited dimension ID,
* or -1 if there are no unlimied dimensions.
*
* @return ::NC_NOERR No error.
* @author Dennis Heimbigner, Ed Hartnett
*/
int
NCZ_inq(int ncid, int *ndimsp, int *nvarsp, int *nattsp, int *unlimdimidp)
{
NC *nc;
NC_FILE_INFO_T* file;
NC_GRP_INFO_T *grp;
int stat = NC_NOERR;
int i;
LOG((2, "%s: ncid 0x%x", __func__, ncid));
/* Find file metadata. */
if ((stat = nc4_find_nc_grp_h5(ncid, &nc, &grp, &file)))
return stat;
assert(file && grp && nc);
/* Count the number of dims, vars, and global atts; need to iterate
* because of possible nulls. */
if (ndimsp)
{
*ndimsp = ncindexcount(grp->dim);
}
if (nvarsp)
{
*nvarsp = ncindexcount(grp->vars);
}
if (nattsp)
{
/* Do we need to read the atts? */
if (!grp->atts_read)
if ((stat = ncz_read_atts(file,(NC_OBJ*)grp)))
return stat;
*nattsp = ncindexcount(grp->att);
}
if (unlimdimidp)
{
/* Default, no unlimited dimension */
*unlimdimidp = -1;
/* If there's more than one unlimited dim, which was not possible
with netcdf-3, then only the last unlimited one will be reported
back in xtendimp. */
/* Note that this code is inconsistent with nc_inq_unlimid() */
for(i=0;i<ncindexsize(grp->dim);i++) {
NC_DIM_INFO_T* d = (NC_DIM_INFO_T*)ncindexith(grp->dim,i);
if(d == NULL) continue;
if(d->unlimited) {
*unlimdimidp = d->hdr.id;
break;
}
}
}
return NC_NOERR;
}
/**
* @internal This function will write all changed metadata and flush
* ZARR file to disk.
*
* @param file Pointer to file info struct.
*
* @return ::NC_NOERR No error.
* @return ::NC_EINDEFINE Classic model file in define mode.
* @return ::NC_EHDFERR ZARR error.
* @author Dennis Heimbigner, Ed Hartnett
*/
static int
ncz_sync_netcdf4_file(NC_FILE_INFO_T* file, int isclose)
{
int stat = NC_NOERR;
assert(file && file->format_file_info);
LOG((3, "%s", __func__));
ZTRACE(2,"file=%s",file->hdr.name);
/* If we're in define mode, that's an error, for strict nc3 rules,
* otherwise, end define mode. */
if (file->flags & NC_INDEF)
{
if (file->cmode & NC_CLASSIC_MODEL)
return NC_EINDEFINE;
/* Turn define mode off. */
file->flags ^= NC_INDEF;
/* Redef mode needs to be tracked separately for nc_abort. */
file->redef = NC_FALSE;
}
#ifdef LOGGING
/* This will print out the names, types, lens, etc of the vars and
atts in the file, if the logging level is 2 or greater. */
log_metadata_nc(file);
#endif
/* Write any metadata that has changed. */
if (!file->no_write)
{
/* Write out provenance; will create _NCProperties */
if((stat = NCZ_write_provenance(file)))
goto done;
/* Write all the metadata. */
if((stat = ncz_sync_file(file,isclose)))
goto done;
}
done:
return ZUNTRACE(stat);
}
/**
* @internal This function will do the enddef stuff for a netcdf-4 file.
*
* @param file Pointer to ZARR file info struct.
*
* @return ::NC_NOERR No error.
* @return ::NC_ENOTINDEFINE Not in define mode.
* @author Dennis Heimbigner, Ed Hartnett
*/
int
ncz_enddef_netcdf4_file(NC_FILE_INFO_T* file)
{
assert(file);
LOG((3, "%s", __func__));
/* If we're not in define mode, return an error. */
if (!(file->flags & NC_INDEF))
return NC_ENOTINDEFINE;
/* Turn define mode off. */
file->flags ^= NC_INDEF;
/* Redef mode needs to be tracked separately for nc_abort. */
file->redef = NC_FALSE;
return ncz_sync_netcdf4_file(file,!ZCLOSE);
}
/**
* @internal IN netcdf, you first create
* the variable and then (optionally) specify the fill value.
*
* @param ncid File and group ID.
* @param fillmode File mode.
* @param old_modep Pointer that gets old mode. Ignored if NULL.
*
* @return ::NC_NOERR No error.
* @author Dennis Heimbigner
*/
int
NCZ_set_fill(int ncid, int fillmode, int *old_modep)
{
NC_FILE_INFO_T* h5 = NULL;
int stat = NC_NOERR;
LOG((2, "%s: ncid 0x%x fillmode %d", __func__, ncid, fillmode));
/* Get pointer to file info. */
if ((stat = nc4_find_grp_h5(ncid, NULL, &h5)))
return stat;
assert(h5);
/* Trying to set fill on a read-only file? You sicken me! */
if (h5->no_write)
return NC_EPERM;
/* Did you pass me some weird fillmode? */
if (fillmode != NC_FILL && fillmode != NC_NOFILL)
return NC_EINVAL;
/* If the user wants to know, tell him what the old mode was. */
if (old_modep)
*old_modep = h5->fill_mode;
h5->fill_mode = fillmode;
return NC_NOERR;
}