netcdf-c/libdispatch/dinstance.c

555 lines
16 KiB
C
Raw Normal View History

Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
/*
Copyright (c) 1998-2018 University Corporation for Atmospheric Research/Unidata
See COPYRIGHT for license information.
*/
/*
This file contains various instance operations that operate
on a deep level rather than the shallow level of e.g. nc_free_vlen_t.
Currently two operations are defined:
1. reclaim a vector of instances
2. copy a vector of instances
*/
#include "config.h"
#include <stdlib.h>
#include <string.h>
#include <assert.h>
#include "netcdf.h"
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
#include "netcdf_aux.h"
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
#include "nc4internal.h"
#include "nc4dispatch.h"
#include "ncoffsets.h"
#include "ncbytes.h"
#undef REPORT
#undef DEBUG
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
/* DAP2 and DAP4 currently defer most of their API to a substrate NC
that hold the true metadata. So there is a level of indirection
necessary in order to get to the right NC* instance.
*/
#if defined(ENABLE_DAP4) || defined(ENABLE_DAP2)
EXTERNL NC* NCD4_get_substrate(NC* nc);
EXTERNL NC* NCD2_get_substrate(NC* nc);
static NC*
DAPSUBSTRATE(NC* nc)
{
if(USED2INFO(nc) != 0)
return NCD2_get_substrate(nc);
else if(USED4INFO(nc) != 0)
return NCD4_get_substrate(nc);
return nc;
}
#else
#define DAPSUBSTRATE(nc) (nc)
#endif
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
/* It is helpful to have a structure that contains memory and an offset */
typedef struct Position{char* memory; ptrdiff_t offset;} Position;
/* Forward */
#ifdef USE_NETCDF4
static int dump_datar(int ncid, nc_type xtype, Position*, NCbytes* buf);
static int dump_compound(int ncid, nc_type xtype, size_t size, size_t nfields, Position* offset, NCbytes* buf);
static int dump_vlen(int ncid, nc_type xtype, nc_type basetype, Position* offset, NCbytes* buf);
static int dump_enum(int ncid, nc_type xtype, nc_type basetype, Position* offset, NCbytes* buf);
static int dump_opaque(int ncid, nc_type xtype, size_t size, Position* offset, NCbytes* buf);
#endif
/**
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
\ingroup user_types
Reclaim an array of instances of an arbitrary type.
This function should be used when the other simpler functions
such as *nc_free_vlens* or *nc_free_string* cannot be used.
This recursively walks the top-level instances to reclaim
any nested data such as vlen or strings or such.
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
Assumes it is passed a pointer to count instances of xtype.
Reclaims any nested data.
WARNING: This needs access to the type metadata of the file, so
a valid ncid and typeid must be available, which means the file
must not have been closed or aborted.
WARNING: DOES NOT RECLAIM THE TOP-LEVEL MEMORY (see the
nc_reclaim_data_all function). The reason is that we do not
know how it was allocated (e.g. static vs dynamic); only the
caller can know that. But note that it assumes all memory
blocks other than the top were dynamically allocated, so they
will be free'd.
Should work for any netcdf format.
@param ncid root id
@param xtype type id
@param memory ptr to top-level memory to reclaim
@param count number of instances of the type in memory block
@return error code
*/
int
nc_reclaim_data(int ncid, nc_type xtype, void* memory, size_t count)
{
int stat = NC_NOERR;
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
NC* nc = NULL;
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
if(ncid < 0 || xtype <= 0)
{stat = NC_EINVAL; goto done;}
if(memory == NULL && count > 0)
{stat = NC_EINVAL; goto done;}
if(memory == NULL || count == 0)
goto done; /* ok, do nothing */
#ifdef REPORT
fprintf(stderr,">>> reclaim: memory=%p count=%lu ncid=%d xtype=%d\n",memory,(unsigned long)count,ncid,xtype);
#endif
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
if((stat = NC_check_id(ncid,&nc))) goto done;
nc = DAPSUBSTRATE(nc);
/* Call internal version */
stat = NC_reclaim_data(nc,xtype,memory,count);
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
#if 0
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
#ifdef USE_NETCDF4
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
NC_FILE_INFO_T* file = NULL;
/* Find info for this file and group and var, and set pointer to each. */
if ((stat = nc4_find_grp_h5(ncid, NULL, &file))) return stat;
/* Call internal version */
stat = NC_reclaim_data(file,xtype,memory,count);
#endif
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
#endif
done:
return stat;
}
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
/**
\ingroup user_types
Reclaim the memory allocated for an array of instances of an arbitrary type.
This recursively walks the top-level instances to reclaim any nested data such as vlen or strings or such.
This function differs from *nc_reclaim_data* in that it also reclaims the top-level memory.
Assumes it is passed a count and a pointer to the top-level memory.
Should work for any netcdf format.
@param ncid root id
@param xtype type id
@param memory ptr to top-level memory to reclaim
@param count number of instances of the type in memory block
@return error code
*/
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
int
nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count)
{
int stat = NC_NOERR;
stat = nc_reclaim_data(ncid,xtypeid,memory,count);
if(stat == NC_NOERR && memory != NULL)
free(memory);
return stat;
}
/**************************************************/
/**
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
\ingroup user_types
Copy an array of instances of an arbitrary type. This recursively walks
the top-level instances to copy any nested data such as vlen or strings or such.
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
Assumes it is passed a pointer to count instances of xtype and a
space into which to copy the instance. Copys any nested data.
WARNING: This needs access to the type metadata of the file, so
a valid ncid and typeid must be available, which means the file
must not have been closed or aborted.
WARNING: DOES NOT ALLOCATE THE TOP-LEVEL MEMORY (see the
nc_copy_data_all function). Note that all memory blocks other
than the top are dynamically allocated.
Should work for any netcdf format.
@param ncid root id
@param xtype type id
@param memory ptr to top-level memory to copy
@param count number of instances of the type in memory block
@param copy top-level space into which to copy the instance
@return error code
*/
int
nc_copy_data(int ncid, nc_type xtype, const void* memory, size_t count, void* copy)
{
int stat = NC_NOERR;
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
NC* nc = NULL;
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
if(ncid < 0 || xtype <= 0)
{stat = NC_EINVAL; goto done;}
if(memory == NULL && count > 0)
{stat = NC_EINVAL; goto done;}
if(copy == NULL && count > 0)
{stat = NC_EINVAL; goto done;}
if(memory == NULL || count == 0)
goto done; /* ok, do nothing */
#ifdef REPORT
fprintf(stderr,">>> copy : copy =%p memory=%p count=%lu ncid=%d xtype=%d\n",copy,memory,(unsigned long)count,ncid,xtype);
#endif
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
if((stat = NC_check_id(ncid,&nc))) goto done;
nc = DAPSUBSTRATE(nc);
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
/* Call internal version */
stat = NC_copy_data(nc,xtype,memory,count,copy);
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
done:
return stat;
}
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
/**
\ingroup user_types
Copy an array of instances of an arbitrary type. This recursively walks
the top-level instances to copy any nested data such as vlen or strings or such.
This function differs from *nc_copy_data* in that it also allocates
the top-level memory.
Assumes it is passed a pointer to count instances of xtype and a pointer
into which the top-level allocated space is stored.
@param ncid root id
@param xtype type id
@param memory ptr to top-level memory to copy
@param count number of instances of the type in memory block
@param copyp pointer into which the allocated top-level space is stored.
@return error code
*/
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
int
nc_copy_data_all(int ncid, nc_type xtype, const void* memory, size_t count, void** copyp)
{
int stat = NC_NOERR;
size_t xsize = 0;
void* copy = NULL;
/* Get type size */
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
if((stat = ncaux_inq_any_type(ncid,xtype,NULL,&xsize,NULL,NULL,NULL))) goto done;
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
/* allocate the top-level */
if(count > 0) {
if((copy = calloc(xsize,count))==NULL)
{stat = NC_ENOMEM; goto done;}
}
stat = nc_copy_data(ncid,xtype,memory,count,copy);
if(copyp) {*copyp = copy; copy = NULL;}
done:
if(copy)
stat = nc_reclaim_data_all(ncid,xtype,copy,count);
return stat;
}
/**************************************************/
/* Alignment functions */
#ifdef USE_NETCDF4
/**
@param ncid - only needed for a compound type
@param xtype - type for which alignment is requested
@return 0 if not found
*/
int
NC_type_alignment(int ncid, nc_type xtype, size_t* alignp)
{
int stat = NC_NOERR;
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
NC* nc = NULL;
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
if((stat = NC_check_id(ncid,&nc))) goto done;
nc = DAPSUBSTRATE(nc);
if(USENC3INFO(nc)) goto done;
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
/* Call internal version */
stat = NC_type_alignment_internal((NC_FILE_INFO_T*)nc->dispatchdata,xtype,NULL,alignp);
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
done:
return stat;
}
#endif
/**************************************************/
/* Dump an instance into a bytebuffer
@param ncid root id
@param xtype type id
@param memory ptr to top-level memory to dump
@param count number of instances of the type in memory block
@return error code
*/
int
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
nc_dump_data(int ncid, nc_type xtype, const void* memory, size_t count, char** bufp)
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
{
int stat = NC_NOERR;
size_t i;
Position offset;
NCbytes* buf = ncbytesnew();
if(ncid < 0 || xtype <= 0)
{stat = NC_EINVAL; goto done;}
if(memory == NULL && count > 0)
{stat = NC_EINVAL; goto done;}
if(memory == NULL || count == 0)
goto done; /* ok, do nothing */
#ifdef REPORT
fprintf(stderr,">>> dump: memory=%p count=%lu ncid=%d xtype=%d\n",memory,(unsigned long)count,ncid,xtype);
#endif
offset.memory = (char*)memory; /* use char* so we can do pointer arithmetic */
offset.offset = 0;
for(i=0;i<count;i++) {
if(i > 0) ncbytescat(buf," ");
if((stat=dump_datar(ncid,xtype,&offset,buf))) /* dump one instance */
break;
}
if(bufp) *bufp = ncbytesextract(buf);
done:
ncbytesfree(buf);
return stat;
}
int
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
nc_print_data(int ncid, nc_type xtype, const void* memory, size_t count)
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
{
char* s = NULL;
int stat = NC_NOERR;
if((stat=nc_dump_data(ncid,xtype,memory,count,&s))) return stat;
fprintf(stderr,"%s\n",s);
nullfree(s)
return stat;
}
/* Recursive type walker: dump a single instance */
static int
dump_datar(int ncid, nc_type xtype, Position* offset, NCbytes* buf)
{
int stat = NC_NOERR;
size_t xsize;
nc_type basetype;
size_t nfields;
int klass;
char s[128];
/* Get relevant type info */
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
if((stat = ncaux_inq_any_type(ncid,xtype,NULL,&xsize,&basetype,&nfields,&klass))) goto done;
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
switch (xtype) {
case NC_CHAR:
snprintf(s,sizeof(s),"'%c'",*(char*)(offset->memory+offset->offset));
ncbytescat(buf,s);
break;
case NC_BYTE:
snprintf(s,sizeof(s),"%d",*(char*)(offset->memory+offset->offset));
ncbytescat(buf,s);
break;
case NC_UBYTE:
snprintf(s,sizeof(s),"%u",*(unsigned char*)(offset->memory+offset->offset));
ncbytescat(buf,s);
break;
case NC_SHORT:
snprintf(s,sizeof(s),"%d",*(short*)(offset->memory+offset->offset));
ncbytescat(buf,s);
break;
case NC_USHORT:
snprintf(s,sizeof(s),"%d",*(unsigned short*)(offset->memory+offset->offset));
ncbytescat(buf,s);
break;
case NC_INT:
snprintf(s,sizeof(s),"%d",*(int*)(offset->memory+offset->offset));
ncbytescat(buf,s);
break;
case NC_UINT:
snprintf(s,sizeof(s),"%d",*(unsigned int*)(offset->memory+offset->offset));
ncbytescat(buf,s);
break;
case NC_FLOAT:
snprintf(s,sizeof(s),"%f",*(float*)(offset->memory+offset->offset));
ncbytescat(buf,s);
break;
case NC_INT64:
snprintf(s,sizeof(s),"%lld",*(long long*)(offset->memory+offset->offset));
ncbytescat(buf,s);
break;
case NC_UINT64:
snprintf(s,sizeof(s),"%llu",*(unsigned long long*)(offset->memory+offset->offset));
ncbytescat(buf,s);
break;
case NC_DOUBLE:
snprintf(s,sizeof(s),"%lf",*(double*)(offset->memory+offset->offset));
ncbytescat(buf,s);
break;
#ifdef USE_NETCDF4
case NC_STRING: {
char* s = *(char**)(offset->memory + offset->offset);
ncbytescat(buf,"\"");
ncbytescat(buf,s);
ncbytescat(buf,"\"");
} break;
#endif
default:
#ifdef USE_NETCDF4
/* dump a user type */
switch (klass) {
case NC_OPAQUE: stat = dump_opaque(ncid,xtype,xsize,offset,buf); break;
case NC_ENUM: stat = dump_enum(ncid,xtype,basetype,offset,buf); break;
case NC_COMPOUND: stat = dump_compound(ncid,xtype,xsize,nfields,offset,buf); break;
case NC_VLEN: stat = dump_vlen(ncid,xtype,basetype,offset,buf); break;
default: stat = NC_EBADTYPE; break;
}
#else
stat = NC_EBADTYPE;
#endif
break;
}
if(xtype <= NC_MAX_ATOMIC_TYPE)
offset->offset += xsize;
done:
return stat;
}
#ifdef USE_NETCDF4
static int
dump_vlen(int ncid, nc_type xtype, nc_type basetype, Position* offset, NCbytes* buf)
{
int stat = NC_NOERR;
size_t i;
nc_vlen_t* vl = (nc_vlen_t*)(offset->memory+offset->offset);
char s[128];
if(vl->len > 0 && vl->p == NULL)
{stat = NC_EINVAL; goto done;}
snprintf(s,sizeof(s),"{len=%u,p=(",(unsigned)vl->len);
ncbytescat(buf,s);
/* dump each entry in the vlen list */
if(vl->len > 0) {
Position voffset;
size_t alignment = 0;
if((stat = NC_type_alignment(ncid,basetype,&alignment))) goto done;;
voffset.memory = vl->p;
voffset.offset = 0;
for(i=0;i<vl->len;i++) {
if(i > 0) ncbytescat(buf," ");
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
voffset.offset = NC_read_align(voffset.offset,alignment);
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
if((stat = dump_datar(ncid,basetype,&voffset,buf))) goto done;
}
}
ncbytescat(buf,")}");
offset->offset += sizeof(nc_vlen_t);
done:
return stat;
}
static int
dump_enum(int ncid, nc_type xtype, nc_type basetype, Position* offset, NCbytes* buf)
{
int stat = NC_NOERR;
/* basically same as an instance of the enum's integer basetype */
stat = dump_datar(ncid,basetype,offset,buf);
return stat;
}
static int
dump_opaque(int ncid, nc_type xtype, size_t size, Position* offset, NCbytes* buf)
{
size_t i;
char sx[16];
/* basically a fixed size sequence of bytes */
ncbytescat(buf,"|");
for(i=0;i<size;i++) {
unsigned char x = *(offset->memory+offset->offset+i);
snprintf(sx,sizeof(sx),"%2x",x);
ncbytescat(buf,sx);
}
ncbytescat(buf,"|");
offset->offset += size;
return NC_NOERR;
}
static int
dump_compound(int ncid, nc_type xtype, size_t size, size_t nfields, Position* offset, NCbytes* buf)
{
int stat = NC_NOERR;
size_t fid, i, arraycount;
ptrdiff_t saveoffset;
int ndims;
int dimsizes[NC_MAX_VAR_DIMS];
saveoffset = offset->offset;
ncbytescat(buf,"<");
/* Get info about each field in turn and dump it */
for(fid=0;fid<nfields;fid++) {
size_t fieldalignment;
nc_type fieldtype;
char name[NC_MAX_NAME];
char sd[128];
/* Get all relevant info about the field */
if((stat = nc_inq_compound_field(ncid,xtype,fid,name,&fieldalignment,&fieldtype,&ndims,dimsizes))) goto done;
if(fid > 0) ncbytescat(buf,";");
ncbytescat(buf,name);
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
ncbytescat(buf,"(");
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
if(ndims > 0) {
int j;
for(j=0;j<ndims;j++) {
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
snprintf(sd,sizeof(sd),"%s%d",(j==0?"":","),(int)dimsizes[j]);
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
ncbytescat(buf,sd);
}
}
Improve performance of the nc_reclaim_data and nc_copy_data functions. re: Issue https://github.com/Unidata/netcdf-c/issues/2685 re: PR https://github.com/Unidata/netcdf-c/pull/2179 As noted in PR https://github.com/Unidata/netcdf-c/pull/2179, the old code did not allow for reclaiming instances of types, nor for properly copying them. That PR provided new functions capable of reclaiming/copying instances of arbitrary types. However, as noted by Issue https://github.com/Unidata/netcdf-c/issues/2685, using these most general functions resulted in a significant performance degradation, even for common cases. This PR attempts to mitigate the cost of using the general reclaim/copy functions in two ways. First, the previous functions operating at the top level by using ncid and typeid arguments. These functions were augmented with equivalent versions that used the netcdf-c library internal data structures to allow direct access to needed information. These new functions are used internally to the library. The second mitigation involves optimizing the internal functions by providing early tests for common cases. This avoids unnecessary recursive function calls. The overall result is a significant improvement in speed by a factor of roughly twenty -- your mileage may vary. These optimized functions are still not as fast as the original (more limited) functions, but they are getting close. Additional optimizations are possible. But the cost is a significant "uglification" of the code that I deemed a step too far, at least for now. ## Misc. Changes 1. Added a test case to check the proper reclamation/copy of complex types. 2. Found and fixed some places where nc_reclaim/copy should have been used. 3. Replaced, in the netcdf-c library, (almost all) occurrences of nc_reclaim_copy with calls to NC_reclaim/copy. This plus the optimizations is the primary speed-up mechanism. 4. In DAP4, the metadata is held in a substrate in-memory file; this required some changes so that the reclaim/copy code accessed that substrate dispatcher rather than the DAP4 dispatcher. 5. Re-factored and isolated the code that computes if a type is (transitively) variable-sized or not. 6. Clean up the reclamation code in ncgen; adding the use of nc_reclaim exposed some memory problems.
2023-05-21 07:11:25 +08:00
ncbytescat(buf,")");
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
if(ndims == 0) {ndims=1; dimsizes[0]=1;} /* fake the scalar case */
/* Align to this field */
offset->offset = saveoffset + fieldalignment;
/* compute the total number of elements in the field array */
arraycount = 1;
for(i=0;i<ndims;i++) arraycount *= dimsizes[i];
for(i=0;i<arraycount;i++) {
if(i > 0) ncbytescat(buf," ");
if((stat = dump_datar(ncid, fieldtype, offset,buf))) goto done;
}
}
ncbytescat(buf,">");
/* Return to beginning of the compound and move |compound| */
offset->offset = saveoffset;
offset->offset += size;
done:
return stat;
}
#endif
/* Extended version that can handle atomic typeids */
int
NC_inq_any_type(int ncid, nc_type typeid, char *name, size_t *size,
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
nc_type *basetypep, size_t *nfieldsp, int *classp)
{
int stat = NC_NOERR;
#ifdef USE_NETCDF4
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
if(typeid >= NC_FIRSTUSERTYPEID) {
stat = nc_inq_user_type(ncid,typeid,name,size,basetypep,nfieldsp,classp);
} else
#endif
if(typeid > NC_NAT && typeid <= NC_MAX_ATOMIC_TYPE) {
Fix various problem around VLEN's re: https://github.com/Unidata/netcdf-c/issues/541 re: https://github.com/Unidata/netcdf-c/issues/1208 re: https://github.com/Unidata/netcdf-c/issues/2078 re: https://github.com/Unidata/netcdf-c/issues/2041 re: https://github.com/Unidata/netcdf-c/issues/2143 For a long time, there have been known problems with the management of complex types containing VLENs. This also involves the string type because it is stored as a VLEN of chars. This PR (mostly) fixes this problem. But note that it adds new functions to netcdf.h (see below) and this may require bumping the .so number. These new functions can be removed, if desired, in favor of functions in netcdf_aux.h, but netcdf.h seems the better place for them because they are intended as alternatives to the nc_free_vlen and nc_free_string functions already in netcdf.h. The term complex type refers to any type that directly or transitively references a VLEN type. So an array of VLENS, a compound with a VLEN field, and so on. In order to properly handle instances of these complex types, it is necessary to have function that can recursively walk instances of such types to perform various actions on them. The term "deep" is also used to mean recursive. At the moment, the two operations needed by the netcdf library are: * free'ing an instance of the complex type * copying an instance of the complex type. The current library does only shallow free and shallow copy of complex types. This means that only the top level is properly free'd or copied, but deep internal blocks in the instance are not touched. Note that the term "vector" will be used to mean a contiguous (in memory) sequence of instances of some type. Given an array with, say, dimensions 2 X 3 X 4, this will be stored in memory as a vector of length 2*3*4=24 instances. The use cases are primarily these. ## nc_get_vars Suppose one is reading a vector of instances using nc_get_vars (or nc_get_vara or nc_get_var, etc.). These functions will return the vector in the top-level memory provided. All interior blocks (form nested VLEN or strings) will have been dynamically allocated. After using this vector of instances, it is necessary to free (aka reclaim) the dynamically allocated memory, otherwise a memory leak occurs. So, the recursive reclaim function is used to walk the returned instance vector and do a deep reclaim of the data. Currently functions are defined in netcdf.h that are supposed to handle this: nc_free_vlen(), nc_free_vlens(), and nc_free_string(). Unfortunately, these functions only do a shallow free, so deeply nested instances are not properly handled by them. Note that internally, the provided data is immediately written so there is no need to copy it. But the caller may need to reclaim the data it passed into the function. ## nc_put_att Suppose one is writing a vector of instances as the data of an attribute using, say, nc_put_att. Internally, the incoming attribute data must be copied and stored so that changes/reclamation of the input data will not affect the attribute. Again, the code inside the netcdf library does only shallow copying rather than deep copy. As a result, one sees effects such as described in Github Issue https://github.com/Unidata/netcdf-c/issues/2143. Also, after defining the attribute, it may be necessary for the user to free the data that was provided as input to nc_put_att(). ## nc_get_att Suppose one is reading a vector of instances as the data of an attribute using, say, nc_get_att. Internally, the existing attribute data must be copied and returned to the caller, and the caller is responsible for reclaiming the returned data. Again, the code inside the netcdf library does only shallow copying rather than deep copy. So this can lead to memory leaks and errors because the deep data is shared between the library and the user. # Solution The solution is to build properly recursive reclaim and copy functions and use those as needed. These recursive functions are defined in libdispatch/dinstance.c and their signatures are defined in include/netcdf.h. For back compatibility, corresponding "ncaux_XXX" functions are defined in include/netcdf_aux.h. ```` int nc_reclaim_data(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_reclaim_data_all(int ncid, nc_type xtypeid, void* memory, size_t count); int nc_copy_data(int ncid, nc_type xtypeid, const void* memory, size_t count, void* copy); int nc_copy_data_all(int ncid, nc_type xtypeid, const void* memory, size_t count, void** copyp); ```` There are two variants. The first two, nc_reclaim_data() and nc_copy_data(), assume the top-level vector is managed by the caller. For reclaim, this is so the user can use, for example, a statically allocated vector. For copy, it assumes the user provides the space into which the copy is stored. The second two, nc_reclaim_data_all() and nc_copy_data_all(), allows the functions to manage the top-level. So for nc_reclaim_data_all, the top level is assumed to be dynamically allocated and will be free'd by nc_reclaim_data_all(). The nc_copy_data_all() function will allocate the top level and return a pointer to it to the user. The user can later pass that pointer to nc_reclaim_data_all() to reclaim the instance(s). # Internal Changes The netcdf-c library internals are changed to use the proper reclaim and copy functions. It turns out that the places where these functions are needed is quite pervasive in the netcdf-c library code. Using these functions also allows some simplification of the code since the stdata and vldata fields of NC_ATT_INFO are no longer needed. Currently this is commented out using the SEPDATA \#define macro. When any bugs are largely fixed, all this code will be removed. # Known Bugs 1. There is still one known failure that has not been solved. All the failures revolve around some variant of this .cdl file. The proximate cause of failure is the use of a VLEN FillValue. ```` netcdf x { types: float(*) row_of_floats ; dimensions: m = 5 ; variables: row_of_floats ragged_array(m) ; row_of_floats ragged_array:_FillValue = {-999} ; data: ragged_array = {10, 11, 12, 13, 14}, {20, 21, 22, 23}, {30, 31, 32}, {40, 41}, _ ; } ```` When a solution is found, I will either add it to this PR or post a new PR. # Related Changes * Mark nc_free_vlen(s) as deprecated in favor of ncaux_reclaim_data. * Remove the --enable-unfixed-memory-leaks option. * Remove the NC_VLENS_NOTEST code that suppresses some vlen tests. * Document this change in docs/internal.md * Disable the tst_vlen_data test in ncdump/tst_nccopy4.sh. * Mark types as fixed size or not (transitively) to optimize the reclaim and copy functions. # Misc. Changes * Make Doxygen process libdispatch/daux.c * Make sure the NC_ATT_INFO_T.container field is set.
2022-01-09 09:30:00 +08:00
if(basetypep) *basetypep = NC_NAT;
if(nfieldsp) *nfieldsp = 0;
if(classp) *classp = typeid;
stat = NC4_inq_atomic_type(typeid,name,size);
} else
stat = NC_EBADTYPE;
return stat;
}