Files
test2/source/blender/blenkernel/intern/lib_id_delete.cc

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

437 lines
13 KiB
C++
Raw Normal View History

/* SPDX-FileCopyrightText: 2023 Blender Authors
*
* SPDX-License-Identifier: GPL-2.0-or-later */
/** \file
* \ingroup bke
*
* Contains management of ID's for freeing & deletion.
*/
#include "MEM_guardedalloc.h"
/* all types are needed here, in order to do memory operations */
2020-04-21 14:56:02 +02:00
#include "DNA_ID.h"
#include "DNA_key_types.h"
#include "BLI_utildefines.h"
#include "BLI_listbase.h"
#include "BLI_set.hh"
#include "BLI_vector.hh"
#include "BKE_anim_data.hh"
#include "BKE_asset.hh"
2024-03-26 12:57:30 -04:00
#include "BKE_idprop.hh"
#include "BKE_idtype.hh"
2024-01-30 14:42:07 -05:00
#include "BKE_key.hh"
2024-01-23 15:18:09 -05:00
#include "BKE_layer.hh"
2024-01-15 12:44:04 -05:00
#include "BKE_lib_id.hh"
#include "BKE_lib_override.hh"
#include "BKE_lib_remap.hh"
#include "BKE_library.hh"
#include "BKE_main.hh"
#include "BKE_main_namemap.hh"
#include "BLO_readfile.hh"
#include "lib_intern.hh"
#include "DEG_depsgraph.hh"
#ifdef WITH_PYTHON
# include "BPY_extern.hh"
#endif
using namespace blender::bke::id;
void BKE_libblock_free_data(ID *id, const bool do_id_user)
{
if (id->properties) {
IDP_FreePropertyContent_ex(id->properties, do_id_user);
MEM_freeN(id->properties);
id->properties = nullptr;
}
if (id->system_properties) {
IDP_FreePropertyContent_ex(id->system_properties, do_id_user);
MEM_freeN(id->system_properties);
id->system_properties = nullptr;
}
if (id->override_library) {
BKE_lib_override_library_free(&id->override_library, do_id_user);
id->override_library = nullptr;
}
if (id->asset_data) {
BKE_asset_metadata_free(&id->asset_data);
}
if (id->library_weak_reference != nullptr) {
MEM_freeN(id->library_weak_reference);
}
BKE_animdata_free(id, do_id_user);
BKE_libblock_free_runtime_data(id);
}
void BKE_libblock_free_datablock(ID *id, const int /*flag*/)
{
const IDTypeInfo *idtype_info = BKE_idtype_get_info_from_id(id);
if (idtype_info != nullptr) {
if (idtype_info->free_data != nullptr) {
idtype_info->free_data(id);
}
return;
}
BLI_assert_msg(0, "IDType Missing IDTypeInfo");
}
void BKE_libblock_free_runtime_data(ID *id)
{
if (id->runtime) {
/* During "normal" file loading this data is released when versioning ends. Some versioning
* code also deletes IDs, though. For example, in the startup blend file, brushes that were
* replaced by assets are deleted. This means that the regular "delete this ID" flow (aka this
* code here) also needs to free this data. */
BLO_readfile_id_runtime_data_free(*id);
MEM_SAFE_DELETE(id->runtime);
}
}
static int id_free(Main *bmain, void *idv, int flag, const bool use_flag_from_idtag)
{
ID *id = static_cast<ID *>(idv);
if (use_flag_from_idtag) {
if ((id->tag & ID_TAG_NO_MAIN) != 0) {
flag |= LIB_ID_FREE_NO_MAIN | LIB_ID_FREE_NO_UI_USER | LIB_ID_FREE_NO_DEG_TAG;
}
else {
flag &= ~LIB_ID_FREE_NO_MAIN;
}
if ((id->tag & ID_TAG_NO_USER_REFCOUNT) != 0) {
flag |= LIB_ID_FREE_NO_USER_REFCOUNT;
}
else {
flag &= ~LIB_ID_FREE_NO_USER_REFCOUNT;
}
if ((id->tag & ID_TAG_NOT_ALLOCATED) != 0) {
flag |= LIB_ID_FREE_NOT_ALLOCATED;
}
else {
flag &= ~LIB_ID_FREE_NOT_ALLOCATED;
}
}
BLI_assert((flag & LIB_ID_FREE_NO_MAIN) != 0 || bmain != nullptr);
BLI_assert((flag & LIB_ID_FREE_NO_MAIN) != 0 || (flag & LIB_ID_FREE_NOT_ALLOCATED) == 0);
BLI_assert((flag & LIB_ID_FREE_NO_MAIN) != 0 || (flag & LIB_ID_FREE_NO_USER_REFCOUNT) == 0);
const short type = GS(id->name);
if (bmain && (flag & LIB_ID_FREE_NO_DEG_TAG) == 0) {
BLI_assert(bmain->is_locked_for_linking == false);
DEG_id_type_tag(bmain, type);
}
BKE_libblock_free_data_py(id);
Key *key = ((flag & LIB_ID_FREE_NO_MAIN) == 0) ? BKE_key_from_id(id) : nullptr;
if ((flag & LIB_ID_FREE_NO_USER_REFCOUNT) == 0) {
BKE_libblock_relink_ex(bmain, id, nullptr, nullptr, ID_REMAP_SKIP_USER_CLEAR);
}
if ((flag & LIB_ID_FREE_NO_MAIN) == 0 && key != nullptr) {
id_free(bmain, &key->id, flag, use_flag_from_idtag);
}
BKE_libblock_free_datablock(id, flag);
/* avoid notifying on removed data */
if ((flag & LIB_ID_FREE_NO_MAIN) == 0) {
BKE_main_lock(bmain);
}
if ((flag & LIB_ID_FREE_NO_UI_USER) == 0) {
if (free_notifier_reference_cb) {
free_notifier_reference_cb(id);
}
if (remap_editor_id_reference_cb) {
IDRemapper remapper;
remapper.add(id, nullptr);
Performance: Remap multiple items in UI During sprite fright loading of complex scenes would spend a long time in remapping ID's The remapping process is done on a per ID instance that resulted in a very time consuming process that goes over every possible ID reference to find out if it needs to be updated. If there are N of references to ID blocks and there are M ID blocks that needed to be remapped it would take N*M checks. These checks are scattered around the place and memory. Each reference would only be updated at most once, but most of the time no update is needed at all. Idea: By grouping the changes together will reduce the number of checks resulting in improved performance. This would only require N checks. Additional benefits is improved data locality as data is only loaded once in the L2 cache. It has be implemented for the resyncing process and UI editors. On an Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz 16Gig the resyncing process went from 170 seconds to 145 seconds (during hotspot recording). After this patch has been applied we could add similar approach to references (references between data blocks) and functionality (tagged deletion). In my understanding this could reduce the resyncing process to less than a second. Opening the village production file between 10 and 20 seconds. Flame graphs showing that UI remapping isn't visible anymore (`WM_main_remap_editor_id_reference`) * Master {F12769210 size=full} * This patch {F12769211 size=full} Reviewed By: mont29 Maniphest Tasks: T94185 Differential Revision: https://developer.blender.org/D13615
2022-01-25 14:51:35 +01:00
remap_editor_id_reference_cb(remapper);
}
}
if ((flag & LIB_ID_FREE_NO_MAIN) == 0) {
ListBase *lb = which_libbase(bmain, type);
BLI_remlink(lb, id);
if ((flag & LIB_ID_FREE_NO_NAMEMAP_REMOVE) == 0) {
Refactor: Core: ID's 'namemap' used to generate unique ID names. Note: This commit is essentially non-behavioral change, expect in some fairly rare edge cases. This commit does a few things: * Move the whole BKE_main_namemap code to modern C++. * Split API calls to work with the global namemap, or the local ones. * Simplify and make the code easier to follow and understand. * Reduce 'default' memory usage by using growing BitVector for numeric suffix management, instead of a fixed 1K items. * Fix inconsistent handling of 'same base name and numeric suffix, different name' issues (e.g. 'Foo.1' and 'Foo.001'), see `re_create_equivalent_numeric_suffixes` new unittest. * Fix completely broken handling of `global` namemaps. This was (probably!) OK so far because of their currently very limited use-cases. It also adds a few minor improvements to existing behavior (essentially in exotic rare edge cases): * Names that get too long are now only shortened by one char at a time, trying to modify the requested base name as little as possible. * Names that are short, but for which all the manageable numeric suffixes are already in use, are extended with an (increasing) number, instead of being shortened. This work also allowed to detect a few (apparently harmless?) bugs in existing code, which have been fixed already in 4.4 and main, or in this commit as well when they depend on changes in namemap code itself. About performances: This commit introduces a minor slow-down. Some tests heavily relying on this code (like `bl_id_management` and `blendkernel` e.g.) get slightly slower (resp. about 1% and 5%). This seems to come mostly from the added complexity to handle correctly multiple different names with the same base and numeric suffix value ('Foo.1' and 'Foo.001', but also in the global namemap context where IDs from different libraries can have the same name). Pull Request: https://projects.blender.org/blender/blender/pulls/135199
2025-02-28 17:58:58 +01:00
BKE_main_namemap_remove_id(*bmain, *id);
}
}
BKE_libblock_free_data(id, (flag & LIB_ID_FREE_NO_USER_REFCOUNT) == 0);
if ((flag & LIB_ID_FREE_NO_MAIN) == 0) {
BKE_main_unlock(bmain);
}
if ((flag & LIB_ID_FREE_NOT_ALLOCATED) == 0) {
MEM_freeN(id);
}
return flag;
}
void BKE_id_free_ex(Main *bmain, void *idv, const int flag_orig, const bool use_flag_from_idtag)
{
/* ViewLayer resync needs to be delayed during Scene freeing, since internal relationships
* between the Scene's master collection and its view_layers become invalid
* (due to remapping). */
if (bmain && (flag_orig & LIB_ID_FREE_NO_MAIN) == 0) {
BKE_layer_collection_resync_forbid();
}
const ID_Type id_type = GS(static_cast<ID *>(idv)->name);
int flag_final = id_free(bmain, idv, flag_orig, use_flag_from_idtag);
if (bmain) {
if ((flag_orig & LIB_ID_FREE_NO_MAIN) == 0) {
BKE_layer_collection_resync_allow();
}
if ((flag_final & LIB_ID_FREE_NO_MAIN) == 0) {
if (ELEM(id_type, ID_SCE, ID_GR, ID_OB)) {
BKE_main_collection_sync_remap(bmain);
}
}
}
}
void BKE_id_free(Main *bmain, void *idv)
{
BKE_id_free_ex(bmain, idv, 0, true);
}
void BKE_id_free_us(Main *bmain, void *idv) /* test users */
{
ID *id = static_cast<ID *>(idv);
id_us_min(id);
/* XXX This is a temp (2.77) hack so that we keep same behavior as in 2.76 regarding collections
* when deleting an object. Since only 'user_one' usage of objects is collections,
* and only 'real user' usage of objects is scenes, removing that 'user_one' tag when there
* is no more real (scene) users of an object ensures it gets fully unlinked.
* But only for local objects, not linked ones!
* Otherwise, there is no real way to get rid of an object anymore -
* better handling of this is TODO.
*/
if ((GS(id->name) == ID_OB) && (id->us == 1) && !ID_IS_LINKED(id)) {
id_us_clear_real(id);
}
if (id->us == 0) {
const bool is_lib = GS(id->name) == ID_LI;
BKE_libblock_unlink(bmain, id, false);
BKE_id_free(bmain, id);
if (is_lib) {
BKE_library_main_rebuild_hierarchy(bmain);
}
}
}
static size_t id_delete(Main *bmain,
blender::Set<ID *> &ids_to_delete,
const int extra_remapping_flags)
{
bool has_deleted_library = false;
/* Used by batch tagged deletion, when we call BKE_id_free then, id is no more in Main database,
* and has already properly unlinked its other IDs usages.
* UI users are always cleared in BKE_libblock_remap_locked() call, so we can always skip it. */
const int free_flag = LIB_ID_FREE_NO_UI_USER | LIB_ID_FREE_NO_MAIN |
LIB_ID_FREE_NO_USER_REFCOUNT;
const int remapping_flags = (ID_REMAP_STORE_NEVER_NULL_USAGE | ID_REMAP_FORCE_NEVER_NULL_USAGE |
ID_REMAP_FORCE_INTERNAL_RUNTIME_POINTERS | extra_remapping_flags);
MainListsArray lbarray = BKE_main_lists_get(*bmain);
const int base_count = lbarray.size();
BKE_main_lock(bmain);
BKE_layer_collection_resync_forbid();
IDRemapper id_remapper;
/* Main idea of batch deletion is to remove all IDs to be deleted from Main database.
* This means that we won't have to loop over all deleted IDs to remove usages
* of other deleted IDs.
* This gives tremendous speed-up when deleting a large amount of IDs from a Main
* containing thousands of these.
* This also means that we have to be very careful here, as we bypass many 'common'
* processing, hence risking to 'corrupt' at least user counts, if not IDs themselves. */
bool keep_looping = true;
while (keep_looping) {
keep_looping = false;
/* First tag and remove from Main all datablocks directly from target lib.
* Note that we go forward here, since we want to check dependencies before users
* (e.g. meshes before objects). Reduces the chances to have to loop many times in the
* `while (keep_looking)` outer loop. */
for (int i = 0; i < base_count; i++) {
ListBase *lb = lbarray[i];
ID *id_iter;
FOREACH_MAIN_LISTBASE_ID_BEGIN (lb, id_iter) {
/* NOTE: in case we delete a library, we also delete all its datablocks! */
if (ids_to_delete.contains(id_iter) ||
(ID_IS_LINKED(id_iter) && ids_to_delete.contains(&id_iter->lib->id)))
{
BLI_remlink(lb, id_iter);
Refactor: Core: ID's 'namemap' used to generate unique ID names. Note: This commit is essentially non-behavioral change, expect in some fairly rare edge cases. This commit does a few things: * Move the whole BKE_main_namemap code to modern C++. * Split API calls to work with the global namemap, or the local ones. * Simplify and make the code easier to follow and understand. * Reduce 'default' memory usage by using growing BitVector for numeric suffix management, instead of a fixed 1K items. * Fix inconsistent handling of 'same base name and numeric suffix, different name' issues (e.g. 'Foo.1' and 'Foo.001'), see `re_create_equivalent_numeric_suffixes` new unittest. * Fix completely broken handling of `global` namemaps. This was (probably!) OK so far because of their currently very limited use-cases. It also adds a few minor improvements to existing behavior (essentially in exotic rare edge cases): * Names that get too long are now only shortened by one char at a time, trying to modify the requested base name as little as possible. * Names that are short, but for which all the manageable numeric suffixes are already in use, are extended with an (increasing) number, instead of being shortened. This work also allowed to detect a few (apparently harmless?) bugs in existing code, which have been fixed already in 4.4 and main, or in this commit as well when they depend on changes in namemap code itself. About performances: This commit introduces a minor slow-down. Some tests heavily relying on this code (like `bl_id_management` and `blendkernel` e.g.) get slightly slower (resp. about 1% and 5%). This seems to come mostly from the added complexity to handle correctly multiple different names with the same base and numeric suffix value ('Foo.1' and 'Foo.001', but also in the global namemap context where IDs from different libraries can have the same name). Pull Request: https://projects.blender.org/blender/blender/pulls/135199
2025-02-28 17:58:58 +01:00
BKE_main_namemap_remove_id(*bmain, *id_iter);
ids_to_delete.add(id_iter);
id_remapper.add(id_iter, nullptr);
/* Do not tag as no_main now, we want to unlink it first (lower-level ID management
* code has some specific handling of 'no main' IDs that would be a problem in that
* case). */
/* Forcefully also delete shapekeys of the deleted ID if any, 'orphaned' shapekeys are
* not allowed in Blender and will cause lots of problem in modern code (liboverrides,
* warning on write & read, etc.). */
Key *shape_key = BKE_key_from_id(id_iter);
if (shape_key && !ids_to_delete.contains(&shape_key->id)) {
BLI_remlink(&bmain->shapekeys, &shape_key->id);
Refactor: Core: ID's 'namemap' used to generate unique ID names. Note: This commit is essentially non-behavioral change, expect in some fairly rare edge cases. This commit does a few things: * Move the whole BKE_main_namemap code to modern C++. * Split API calls to work with the global namemap, or the local ones. * Simplify and make the code easier to follow and understand. * Reduce 'default' memory usage by using growing BitVector for numeric suffix management, instead of a fixed 1K items. * Fix inconsistent handling of 'same base name and numeric suffix, different name' issues (e.g. 'Foo.1' and 'Foo.001'), see `re_create_equivalent_numeric_suffixes` new unittest. * Fix completely broken handling of `global` namemaps. This was (probably!) OK so far because of their currently very limited use-cases. It also adds a few minor improvements to existing behavior (essentially in exotic rare edge cases): * Names that get too long are now only shortened by one char at a time, trying to modify the requested base name as little as possible. * Names that are short, but for which all the manageable numeric suffixes are already in use, are extended with an (increasing) number, instead of being shortened. This work also allowed to detect a few (apparently harmless?) bugs in existing code, which have been fixed already in 4.4 and main, or in this commit as well when they depend on changes in namemap code itself. About performances: This commit introduces a minor slow-down. Some tests heavily relying on this code (like `bl_id_management` and `blendkernel` e.g.) get slightly slower (resp. about 1% and 5%). This seems to come mostly from the added complexity to handle correctly multiple different names with the same base and numeric suffix value ('Foo.1' and 'Foo.001', but also in the global namemap context where IDs from different libraries can have the same name). Pull Request: https://projects.blender.org/blender/blender/pulls/135199
2025-02-28 17:58:58 +01:00
BKE_main_namemap_remove_id(*bmain, shape_key->id);
ids_to_delete.add(&shape_key->id);
id_remapper.add(&shape_key->id, nullptr);
}
Performance: Remap multiple items in UI During sprite fright loading of complex scenes would spend a long time in remapping ID's The remapping process is done on a per ID instance that resulted in a very time consuming process that goes over every possible ID reference to find out if it needs to be updated. If there are N of references to ID blocks and there are M ID blocks that needed to be remapped it would take N*M checks. These checks are scattered around the place and memory. Each reference would only be updated at most once, but most of the time no update is needed at all. Idea: By grouping the changes together will reduce the number of checks resulting in improved performance. This would only require N checks. Additional benefits is improved data locality as data is only loaded once in the L2 cache. It has be implemented for the resyncing process and UI editors. On an Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz 16Gig the resyncing process went from 170 seconds to 145 seconds (during hotspot recording). After this patch has been applied we could add similar approach to references (references between data blocks) and functionality (tagged deletion). In my understanding this could reduce the resyncing process to less than a second. Opening the village production file between 10 and 20 seconds. Flame graphs showing that UI remapping isn't visible anymore (`WM_main_remap_editor_id_reference`) * Master {F12769210 size=full} * This patch {F12769211 size=full} Reviewed By: mont29 Maniphest Tasks: T94185 Differential Revision: https://developer.blender.org/D13615
2022-01-25 14:51:35 +01:00
keep_looping = true;
}
Performance: Remap multiple items in UI During sprite fright loading of complex scenes would spend a long time in remapping ID's The remapping process is done on a per ID instance that resulted in a very time consuming process that goes over every possible ID reference to find out if it needs to be updated. If there are N of references to ID blocks and there are M ID blocks that needed to be remapped it would take N*M checks. These checks are scattered around the place and memory. Each reference would only be updated at most once, but most of the time no update is needed at all. Idea: By grouping the changes together will reduce the number of checks resulting in improved performance. This would only require N checks. Additional benefits is improved data locality as data is only loaded once in the L2 cache. It has be implemented for the resyncing process and UI editors. On an Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz 16Gig the resyncing process went from 170 seconds to 145 seconds (during hotspot recording). After this patch has been applied we could add similar approach to references (references between data blocks) and functionality (tagged deletion). In my understanding this could reduce the resyncing process to less than a second. Opening the village production file between 10 and 20 seconds. Flame graphs showing that UI remapping isn't visible anymore (`WM_main_remap_editor_id_reference`) * Master {F12769210 size=full} * This patch {F12769211 size=full} Reviewed By: mont29 Maniphest Tasks: T94185 Differential Revision: https://developer.blender.org/D13615
2022-01-25 14:51:35 +01:00
}
FOREACH_MAIN_LISTBASE_ID_END;
}
Performance: Remap multiple items in UI During sprite fright loading of complex scenes would spend a long time in remapping ID's The remapping process is done on a per ID instance that resulted in a very time consuming process that goes over every possible ID reference to find out if it needs to be updated. If there are N of references to ID blocks and there are M ID blocks that needed to be remapped it would take N*M checks. These checks are scattered around the place and memory. Each reference would only be updated at most once, but most of the time no update is needed at all. Idea: By grouping the changes together will reduce the number of checks resulting in improved performance. This would only require N checks. Additional benefits is improved data locality as data is only loaded once in the L2 cache. It has be implemented for the resyncing process and UI editors. On an Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz 16Gig the resyncing process went from 170 seconds to 145 seconds (during hotspot recording). After this patch has been applied we could add similar approach to references (references between data blocks) and functionality (tagged deletion). In my understanding this could reduce the resyncing process to less than a second. Opening the village production file between 10 and 20 seconds. Flame graphs showing that UI remapping isn't visible anymore (`WM_main_remap_editor_id_reference`) * Master {F12769210 size=full} * This patch {F12769211 size=full} Reviewed By: mont29 Maniphest Tasks: T94185 Differential Revision: https://developer.blender.org/D13615
2022-01-25 14:51:35 +01:00
/* Will tag 'never nullptr' users of this ID too.
*
* NOTE: #BKE_libblock_unlink() cannot be used here, since it would ignore indirect
* links, this can lead to nasty crashing here in second, actual deleting loop.
* Also, this will also flag users of deleted data that cannot be unlinked
* (object using deleted obdata, etc.), so that they also get deleted. */
BKE_libblock_remap_multiple_locked(bmain, id_remapper, remapping_flags);
for (ID *id_never_null_iter : id_remapper.never_null_users()) {
ids_to_delete.add(id_never_null_iter);
}
id_remapper.clear();
}
Performance: Remap multiple items in UI During sprite fright loading of complex scenes would spend a long time in remapping ID's The remapping process is done on a per ID instance that resulted in a very time consuming process that goes over every possible ID reference to find out if it needs to be updated. If there are N of references to ID blocks and there are M ID blocks that needed to be remapped it would take N*M checks. These checks are scattered around the place and memory. Each reference would only be updated at most once, but most of the time no update is needed at all. Idea: By grouping the changes together will reduce the number of checks resulting in improved performance. This would only require N checks. Additional benefits is improved data locality as data is only loaded once in the L2 cache. It has be implemented for the resyncing process and UI editors. On an Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz 16Gig the resyncing process went from 170 seconds to 145 seconds (during hotspot recording). After this patch has been applied we could add similar approach to references (references between data blocks) and functionality (tagged deletion). In my understanding this could reduce the resyncing process to less than a second. Opening the village production file between 10 and 20 seconds. Flame graphs showing that UI remapping isn't visible anymore (`WM_main_remap_editor_id_reference`) * Master {F12769210 size=full} * This patch {F12769211 size=full} Reviewed By: mont29 Maniphest Tasks: T94185 Differential Revision: https://developer.blender.org/D13615
2022-01-25 14:51:35 +01:00
/* Remapping above may have left some Library::runtime::archived_libraries items to nullptr,
* clean this up and shrink the vector accordingly. */
blender::bke::library::main_cleanup_parent_archives(*bmain);
/* Since we removed IDs from Main, their own other IDs usages need to be removed 'manually'. */
blender::Vector<ID *> cleanup_ids{ids_to_delete.begin(), ids_to_delete.end()};
BKE_libblock_relink_multiple(
bmain,
cleanup_ids,
ID_REMAP_TYPE_CLEANUP,
id_remapper,
(ID_REMAP_FORCE_INTERNAL_RUNTIME_POINTERS | ID_REMAP_SKIP_USER_CLEAR));
cleanup_ids.clear();
/* Now we can safely mark that ID as not being in Main database anymore. */
/* NOTE: This needs to be done in a separate loop than above, otherwise some user-counts of
* deleted IDs may not be properly decreased by the remappings (since `NO_MAIN` ID user-counts
* is never affected). */
for (ID *id : ids_to_delete) {
id->tag |= ID_TAG_NO_MAIN;
/* User-count needs to be reset artificially, since some usages may not be cleared in batch
* deletion (typically, if one deleted ID uses another deleted ID, this may not be cleared by
* remapping code, depending on order in which these are handled). */
id->us = ID_FAKE_USERS(id);
2024-02-22 19:56:59 +01:00
if (!has_deleted_library && GS(id->name) == ID_LI) {
has_deleted_library = true;
}
2024-02-22 19:56:59 +01:00
id_free(bmain, id, free_flag, false);
}
BKE_main_unlock(bmain);
BKE_layer_collection_resync_allow();
BKE_main_collection_sync_remap(bmain);
if (has_deleted_library) {
BKE_library_main_rebuild_hierarchy(bmain);
}
bmain->is_memfile_undo_written = false;
return size_t(ids_to_delete.size());
}
void BKE_id_delete_ex(Main *bmain, void *idv, const int extra_remapping_flags)
{
ID *id = static_cast<ID *>(idv);
BLI_assert_msg((id->tag & ID_TAG_NO_MAIN) == 0, "Cannot be used with IDs outside of Main");
blender::Set<ID *> ids_to_delete = {id};
id_delete(bmain, ids_to_delete, extra_remapping_flags);
}
void BKE_id_delete(Main *bmain, void *idv)
{
BKE_id_delete_ex(bmain, idv, 0);
}
size_t BKE_id_multi_tagged_delete(Main *bmain)
{
blender::Set<ID *> ids_to_delete;
ID *id_iter;
FOREACH_MAIN_ID_BEGIN (bmain, id_iter) {
if (id_iter->tag & ID_TAG_DOIT) {
ids_to_delete.add(id_iter);
}
}
FOREACH_MAIN_ID_END;
return id_delete(bmain, ids_to_delete, 0);
}
size_t BKE_id_multi_delete(Main *bmain, blender::Set<ID *> &ids_to_delete)
{
return id_delete(bmain, ids_to_delete, 0);
}
/* -------------------------------------------------------------------- */
/** \name Python Data Handling
* \{ */
void BKE_libblock_free_data_py(ID *id)
{
#ifdef WITH_PYTHON
# ifdef WITH_PYTHON_SAFETY
BPY_id_release(id);
# endif
if (id->py_instance) {
BPY_DECREF_RNA_INVALIDATE(id->py_instance);
}
#else
UNUSED_VARS(id);
#endif
}
/** \} */