415
406
def _copy_inventory_texts(self):
416
407
source_vf, target_vf = self._build_vfs('inventory', True, True)
417
# It is not sufficient to just use self.revision_keys, as stacked
418
# repositories can have more inventories than they have revisions.
419
# One alternative would be to do something with
420
# get_parent_map(self.revision_keys), but that shouldn't be any faster
422
inventory_keys = source_vf.keys()
423
missing_inventories = set(self.revision_keys).difference(inventory_keys)
424
if missing_inventories:
425
missing_inventories = sorted(missing_inventories)
426
raise ValueError('We are missing inventories for revisions: %s'
427
% (missing_inventories,))
428
self._copy_stream(source_vf, target_vf, inventory_keys,
408
self._copy_stream(source_vf, target_vf, self.revision_keys,
429
409
'inventories', self._get_filtered_inv_stream, 2)
431
411
def _copy_chk_texts(self):
586
557
pack_factory = GCPack
587
558
resumed_pack_factory = ResumedGCPack
589
def _check_new_inventories(self):
590
"""Detect missing inventories or chk root entries for the new revisions
593
:returns: list of strs, summarising any problems found. If the list is
594
empty no problems were found.
596
# Ensure that all revisions added in this write group have:
597
# - corresponding inventories,
598
# - chk root entries for those inventories,
599
# - and any present parent inventories have their chk root
601
# And all this should be independent of any fallback repository.
603
key_deps = self.repo.revisions._index._key_dependencies
604
new_revisions_keys = key_deps.get_new_keys()
605
no_fallback_inv_index = self.repo.inventories._index
606
no_fallback_chk_bytes_index = self.repo.chk_bytes._index
607
no_fallback_texts_index = self.repo.texts._index
608
inv_parent_map = no_fallback_inv_index.get_parent_map(
610
# Are any inventories for corresponding to the new revisions missing?
611
corresponding_invs = set(inv_parent_map)
612
missing_corresponding = set(new_revisions_keys)
613
missing_corresponding.difference_update(corresponding_invs)
614
if missing_corresponding:
615
problems.append("inventories missing for revisions %s" %
616
(sorted(missing_corresponding),))
618
# Are any chk root entries missing for any inventories? This includes
619
# any present parent inventories, which may be used when calculating
620
# deltas for streaming.
621
all_inv_keys = set(corresponding_invs)
622
for parent_inv_keys in inv_parent_map.itervalues():
623
all_inv_keys.update(parent_inv_keys)
624
# Filter out ghost parents.
625
all_inv_keys.intersection_update(
626
no_fallback_inv_index.get_parent_map(all_inv_keys))
627
parent_invs_only_keys = all_inv_keys.symmetric_difference(
630
inv_ids = [key[-1] for key in all_inv_keys]
631
parent_invs_only_ids = [key[-1] for key in parent_invs_only_keys]
632
root_key_info = _build_interesting_key_sets(
633
self.repo, inv_ids, parent_invs_only_ids)
634
expected_chk_roots = root_key_info.all_keys()
635
present_chk_roots = no_fallback_chk_bytes_index.get_parent_map(
637
missing_chk_roots = expected_chk_roots.difference(present_chk_roots)
638
if missing_chk_roots:
639
problems.append("missing referenced chk root keys: %s"
640
% (sorted(missing_chk_roots),))
641
# Don't bother checking any further.
643
# Find all interesting chk_bytes records, and make sure they are
644
# present, as well as the text keys they reference.
645
chk_bytes_no_fallbacks = self.repo.chk_bytes.without_fallbacks()
646
chk_bytes_no_fallbacks._search_key_func = \
647
self.repo.chk_bytes._search_key_func
648
chk_diff = chk_map.iter_interesting_nodes(
649
chk_bytes_no_fallbacks, root_key_info.interesting_root_keys,
650
root_key_info.uninteresting_root_keys)
651
bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
654
for record in _filter_text_keys(chk_diff, text_keys, bytes_to_info):
656
except errors.NoSuchRevision, e:
657
# XXX: It would be nice if we could give a more precise error here.
658
problems.append("missing chk node(s) for id_to_entry maps")
659
chk_diff = chk_map.iter_interesting_nodes(
660
chk_bytes_no_fallbacks, root_key_info.interesting_pid_root_keys,
661
root_key_info.uninteresting_pid_root_keys)
663
for interesting_rec, interesting_map in chk_diff:
665
except errors.NoSuchRevision, e:
667
"missing chk node(s) for parent_id_basename_to_file_id maps")
668
present_text_keys = no_fallback_texts_index.get_parent_map(text_keys)
669
missing_text_keys = text_keys.difference(present_text_keys)
670
if missing_text_keys:
671
problems.append("missing text keys: %r"
672
% (sorted(missing_text_keys),))
560
def _already_packed(self):
561
"""Is the collection already packed?"""
562
# Always repack GC repositories for now
675
565
def _execute_pack_operations(self, pack_operations,
676
566
_packer_class=GCCHKPacker,
733
620
self.inventories = GroupCompressVersionedFiles(
734
621
_GCGraphIndex(self._pack_collection.inventory_index.combined_index,
735
622
add_callback=self._pack_collection.inventory_index.add_callback,
736
parents=True, is_locked=self.is_locked,
737
inconsistency_fatal=False),
623
parents=True, is_locked=self.is_locked),
738
624
access=self._pack_collection.inventory_index.data_access)
739
625
self.revisions = GroupCompressVersionedFiles(
740
626
_GCGraphIndex(self._pack_collection.revision_index.combined_index,
741
627
add_callback=self._pack_collection.revision_index.add_callback,
742
628
parents=True, is_locked=self.is_locked,
743
track_external_parent_refs=True, track_new_keys=True),
629
track_external_parent_refs=True),
744
630
access=self._pack_collection.revision_index.data_access,
746
632
self.signatures = GroupCompressVersionedFiles(
747
633
_GCGraphIndex(self._pack_collection.signature_index.combined_index,
748
634
add_callback=self._pack_collection.signature_index.add_callback,
749
parents=False, is_locked=self.is_locked,
750
inconsistency_fatal=False),
635
parents=False, is_locked=self.is_locked),
751
636
access=self._pack_collection.signature_index.data_access,
753
638
self.texts = GroupCompressVersionedFiles(
754
639
_GCGraphIndex(self._pack_collection.text_index.combined_index,
755
640
add_callback=self._pack_collection.text_index.add_callback,
756
parents=True, is_locked=self.is_locked,
757
inconsistency_fatal=False),
641
parents=True, is_locked=self.is_locked),
758
642
access=self._pack_collection.text_index.data_access)
759
643
# No parents, individual CHK pages don't have specific ancestry
760
644
self.chk_bytes = GroupCompressVersionedFiles(
761
645
_GCGraphIndex(self._pack_collection.chk_index.combined_index,
762
646
add_callback=self._pack_collection.chk_index.add_callback,
763
parents=False, is_locked=self.is_locked,
764
inconsistency_fatal=False),
647
parents=False, is_locked=self.is_locked),
765
648
access=self._pack_collection.chk_index.data_access)
766
search_key_name = self._format._serializer.search_key_name
767
search_key_func = chk_map.search_key_registry.get(search_key_name)
768
self.chk_bytes._search_key_func = search_key_func
769
649
# True when the repository object is 'write locked' (as opposed to the
770
650
# physical lock only taken out around changes to the pack-names list.)
771
651
# Another way to represent this would be a decorator around the control
794
674
return self._inventory_add_lines(revision_id, parents,
795
675
inv_lines, check_content=False)
797
def _create_inv_from_null(self, delta, revision_id):
798
"""This will mutate new_inv directly.
800
This is a simplified form of create_by_apply_delta which knows that all
801
the old values must be None, so everything is a create.
803
serializer = self._format._serializer
804
new_inv = inventory.CHKInventory(serializer.search_key_name)
805
new_inv.revision_id = revision_id
806
entry_to_bytes = new_inv._entry_to_bytes
807
id_to_entry_dict = {}
808
parent_id_basename_dict = {}
809
for old_path, new_path, file_id, entry in delta:
810
if old_path is not None:
811
raise ValueError('Invalid delta, somebody tried to delete %r'
812
' from the NULL_REVISION'
813
% ((old_path, file_id),))
815
raise ValueError('Invalid delta, delta from NULL_REVISION has'
816
' no new_path %r' % (file_id,))
818
new_inv.root_id = file_id
819
parent_id_basename_key = StaticTuple('', '').intern()
821
utf8_entry_name = entry.name.encode('utf-8')
822
parent_id_basename_key = StaticTuple(entry.parent_id,
823
utf8_entry_name).intern()
824
new_value = entry_to_bytes(entry)
826
# new_inv._path_to_fileid_cache[new_path] = file_id
827
key = StaticTuple(file_id).intern()
828
id_to_entry_dict[key] = new_value
829
parent_id_basename_dict[parent_id_basename_key] = file_id
831
new_inv._populate_from_dicts(self.chk_bytes, id_to_entry_dict,
832
parent_id_basename_dict, maximum_size=serializer.maximum_size)
835
677
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
836
678
parents, basis_inv=None, propagate_caches=False):
837
679
"""Add a new inventory expressed as a delta against another revision.
857
699
repository format specific) of the serialized inventory, and the
858
700
resulting inventory.
702
if basis_revision_id == _mod_revision.NULL_REVISION:
703
return KnitPackRepository.add_inventory_by_delta(self,
704
basis_revision_id, delta, new_revision_id, parents)
860
705
if not self.is_in_write_group():
861
706
raise AssertionError("%r not in write group" % (self,))
862
707
_mod_revision.check_not_reserved_id(new_revision_id)
864
if basis_inv is None:
865
if basis_revision_id == _mod_revision.NULL_REVISION:
866
new_inv = self._create_inv_from_null(delta, new_revision_id)
867
inv_lines = new_inv.to_lines()
868
return self._inventory_add_lines(new_revision_id, parents,
869
inv_lines, check_content=False), new_inv
871
basis_tree = self.revision_tree(basis_revision_id)
872
basis_tree.lock_read()
708
basis_tree = self.revision_tree(basis_revision_id)
709
basis_tree.lock_read()
711
if basis_inv is None:
873
712
basis_inv = basis_tree.inventory
875
713
result = basis_inv.create_by_apply_delta(delta, new_revision_id,
876
714
propagate_caches=propagate_caches)
877
715
inv_lines = result.to_lines()
878
716
return self._inventory_add_lines(new_revision_id, parents,
879
717
inv_lines, check_content=False), result
881
if basis_tree is not None:
884
def deserialise_inventory(self, revision_id, bytes):
885
return inventory.CHKInventory.deserialise(self.chk_bytes, bytes,
888
def _iter_inventories(self, revision_ids, ordering):
721
def _iter_inventories(self, revision_ids):
889
722
"""Iterate over many inventory objects."""
891
ordering = 'unordered'
892
723
keys = [(revision_id,) for revision_id in revision_ids]
893
stream = self.inventories.get_record_stream(keys, ordering, True)
724
stream = self.inventories.get_record_stream(keys, 'unordered', True)
895
726
for record in stream:
896
727
if record.storage_kind != 'absent':
901
732
yield inventory.CHKInventory.deserialise(self.chk_bytes, texts[key], key)
903
def _iter_inventory_xmls(self, revision_ids, ordering):
904
# Without a native 'xml' inventory, this method doesn't make sense.
905
# However older working trees, and older bundles want it - so we supply
906
# it allowing get_inventory_xml to work. Bundles currently use the
907
# serializer directly; this also isn't ideal, but there isn't an xml
908
# iteration interface offered at all for repositories. We could make
909
# _iter_inventory_xmls be part of the contract, even if kept private.
910
inv_to_str = self._serializer.write_inventory_to_string
911
for inv in self.iter_inventories(revision_ids, ordering=ordering):
912
yield inv_to_str(inv), inv.revision_id
914
def _find_present_inventory_keys(self, revision_keys):
915
parent_map = self.inventories.get_parent_map(revision_keys)
916
present_inventory_keys = set(k for k in parent_map)
917
return present_inventory_keys
734
def _iter_inventory_xmls(self, revision_ids):
735
# Without a native 'xml' inventory, this method doesn't make sense, so
736
# make it raise to trap naughty direct users.
737
raise NotImplementedError(self._iter_inventory_xmls)
739
def _find_parent_ids_of_revisions(self, revision_ids):
740
# TODO: we probably want to make this a helper that other code can get
742
parent_map = self.get_parent_map(revision_ids)
744
map(parents.update, parent_map.itervalues())
745
parents.difference_update(revision_ids)
746
parents.discard(_mod_revision.NULL_REVISION)
749
def _find_present_inventory_ids(self, revision_ids):
750
keys = [(r,) for r in revision_ids]
751
parent_map = self.inventories.get_parent_map(keys)
752
present_inventory_ids = set(k[-1] for k in parent_map)
753
return present_inventory_ids
919
755
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
920
756
"""Find the file ids and versions affected by revisions.
931
767
file_id_revisions = {}
932
768
pb = ui.ui_factory.nested_progress_bar()
934
revision_keys = [(r,) for r in revision_ids]
935
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
936
# TODO: instead of using _find_present_inventory_keys, change the
937
# code paths to allow missing inventories to be tolerated.
938
# However, we only want to tolerate missing parent
939
# inventories, not missing inventories for revision_ids
940
present_parent_inv_keys = self._find_present_inventory_keys(
942
present_parent_inv_ids = set(
943
[k[-1] for k in present_parent_inv_keys])
944
inventories_to_read = set(revision_ids)
945
inventories_to_read.update(present_parent_inv_ids)
946
root_key_info = _build_interesting_key_sets(
947
self, inventories_to_read, present_parent_inv_ids)
948
interesting_root_keys = root_key_info.interesting_root_keys
949
uninteresting_root_keys = root_key_info.uninteresting_root_keys
770
parent_ids = self._find_parent_ids_of_revisions(revision_ids)
771
present_parent_inv_ids = self._find_present_inventory_ids(parent_ids)
772
uninteresting_root_keys = set()
773
interesting_root_keys = set()
774
inventories_to_read = set(present_parent_inv_ids)
775
inventories_to_read.update(revision_ids)
776
for inv in self.iter_inventories(inventories_to_read):
777
entry_chk_root_key = inv.id_to_entry.key()
778
if inv.revision_id in present_parent_inv_ids:
779
uninteresting_root_keys.add(entry_chk_root_key)
781
interesting_root_keys.add(entry_chk_root_key)
950
783
chk_bytes = self.chk_bytes
951
784
for record, items in chk_map.iter_interesting_nodes(chk_bytes,
952
785
interesting_root_keys, uninteresting_root_keys,
954
787
for name, bytes in items:
955
788
(name_utf8, file_id, revision_id) = bytes_to_info(bytes)
956
# TODO: consider interning file_id, revision_id here, or
957
# pushing that intern() into bytes_to_info()
958
# TODO: rich_root should always be True here, for all
959
# repositories that support chk_bytes
960
789
if not rich_root and name_utf8 == '':
1006
835
def _get_source(self, to_format):
1007
836
"""Return a source for streaming from this repository."""
1008
if self._format._serializer == to_format._serializer:
837
if isinstance(to_format, remote.RemoteRepositoryFormat):
838
# Can't just check attributes on to_format with the current code,
840
to_format._ensure_real()
841
to_format = to_format._custom_format
842
if to_format.__class__ is self._format.__class__:
1009
843
# We must be exactly the same format, otherwise stuff like the chk
1010
# page layout might be different.
1011
# Actually, this test is just slightly looser than exact so that
1012
# CHK2 <-> 2a transfers will work.
844
# page layout might be different
1013
845
return GroupCHKStreamSource(self, to_format)
1014
846
return super(CHKInventoryRepository, self)._get_source(to_format)
1017
class GroupCHKStreamSource(KnitPackStreamSource):
849
class GroupCHKStreamSource(repository.StreamSource):
1018
850
"""Used when both the source and target repo are GroupCHK repos."""
1020
852
def __init__(self, from_repository, to_format):
1067
898
p_id_roots_set.clear()
1068
899
return ('inventories', _filtered_inv_stream())
1070
def _get_filtered_chk_streams(self, excluded_revision_keys):
901
def _find_present_inventories(self, revision_ids):
902
revision_keys = [(r,) for r in revision_ids]
903
inventories = self.from_repository.inventories
904
present_inventories = inventories.get_parent_map(revision_keys)
905
return [p[-1] for p in present_inventories]
907
def _get_filtered_chk_streams(self, excluded_revision_ids):
1071
908
self._text_keys = set()
1072
excluded_revision_keys.discard(_mod_revision.NULL_REVISION)
1073
if not excluded_revision_keys:
909
excluded_revision_ids.discard(_mod_revision.NULL_REVISION)
910
if not excluded_revision_ids:
1074
911
uninteresting_root_keys = set()
1075
912
uninteresting_pid_root_keys = set()
1090
927
bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
1091
928
chk_bytes = self.from_repository.chk_bytes
1092
929
def _filter_id_to_entry():
1093
interesting_nodes = chk_map.iter_interesting_nodes(chk_bytes,
1094
self._chk_id_roots, uninteresting_root_keys)
1095
for record in _filter_text_keys(interesting_nodes, self._text_keys,
930
for record, items in chk_map.iter_interesting_nodes(chk_bytes,
931
self._chk_id_roots, uninteresting_root_keys):
932
for name, bytes in items:
933
# Note: we don't care about name_utf8, because we are always
935
_, file_id, revision_id = bytes_to_info(bytes)
936
self._text_keys.add((file_id, revision_id))
1097
937
if record is not None:
1108
948
self._chk_p_id_roots = None
1109
949
yield 'chk_bytes', _get_parent_id_basename_to_file_id_pages()
951
def _get_text_stream(self):
952
# Note: We know we don't have to handle adding root keys, because both
953
# the source and target are GCCHK, and those always support rich-roots
954
# We may want to request as 'unordered', in case the source has done a
956
return ('texts', self.from_repository.texts.get_record_stream(
957
self._text_keys, 'groupcompress', False))
1111
959
def get_stream(self, search):
1112
960
revision_ids = search.get_keys()
1113
961
for stream_info in self._fetch_revision_texts(revision_ids):
1114
962
yield stream_info
1115
963
self._revision_keys = [(rev_id,) for rev_id in revision_ids]
1116
self.from_repository.revisions.clear_cache()
1117
self.from_repository.signatures.clear_cache()
1118
964
yield self._get_inventory_stream(self._revision_keys)
1119
self.from_repository.inventories.clear_cache()
1120
965
# TODO: The keys to exclude might be part of the search recipe
1121
966
# For now, exclude all parents that are at the edge of ancestry, for
1122
967
# which we have inventories
1123
968
from_repo = self.from_repository
1124
parent_keys = from_repo._find_parent_keys_of_revisions(
1125
self._revision_keys)
1126
for stream_info in self._get_filtered_chk_streams(parent_keys):
969
parent_ids = from_repo._find_parent_ids_of_revisions(revision_ids)
970
for stream_info in self._get_filtered_chk_streams(parent_ids):
1127
971
yield stream_info
1128
self.from_repository.chk_bytes.clear_cache()
1129
972
yield self._get_text_stream()
1130
self.from_repository.texts.clear_cache()
1132
974
def get_stream_for_missing_keys(self, missing_keys):
1133
975
# missing keys can only occur when we are byte copying and not
1142
984
missing_inventory_keys.add(key[1:])
1143
985
if self._chk_id_roots or self._chk_p_id_roots:
1144
986
raise AssertionError('Cannot call get_stream_for_missing_keys'
1145
' until all of get_stream() has been consumed.')
987
' untill all of get_stream() has been consumed.')
1146
988
# Yield the inventory stream, so we can find the chk stream
1147
989
# Some of the missing_keys will be missing because they are ghosts.
1148
990
# As such, we can ignore them. The Sink is required to verify there are
1149
991
# no unavailable texts when the ghost inventories are not filled in.
1150
992
yield self._get_inventory_stream(missing_inventory_keys,
1151
993
allow_absent=True)
1152
# We use the empty set for excluded_revision_keys, to make it clear
1153
# that we want to transmit all referenced chk pages.
994
# We use the empty set for excluded_revision_ids, to make it clear that
995
# we want to transmit all referenced chk pages.
1154
996
for stream_info in self._get_filtered_chk_streams(set()):
1155
997
yield stream_info
1158
class _InterestingKeyInfo(object):
1160
self.interesting_root_keys = set()
1161
self.interesting_pid_root_keys = set()
1162
self.uninteresting_root_keys = set()
1163
self.uninteresting_pid_root_keys = set()
1165
def all_interesting(self):
1166
return self.interesting_root_keys.union(self.interesting_pid_root_keys)
1168
def all_uninteresting(self):
1169
return self.uninteresting_root_keys.union(
1170
self.uninteresting_pid_root_keys)
1173
return self.all_interesting().union(self.all_uninteresting())
1176
def _build_interesting_key_sets(repo, inventory_ids, parent_only_inv_ids):
1177
result = _InterestingKeyInfo()
1178
for inv in repo.iter_inventories(inventory_ids, 'unordered'):
1179
root_key = inv.id_to_entry.key()
1180
pid_root_key = inv.parent_id_basename_to_file_id.key()
1181
if inv.revision_id in parent_only_inv_ids:
1182
result.uninteresting_root_keys.add(root_key)
1183
result.uninteresting_pid_root_keys.add(pid_root_key)
1185
result.interesting_root_keys.add(root_key)
1186
result.interesting_pid_root_keys.add(pid_root_key)
1190
def _filter_text_keys(interesting_nodes_iterable, text_keys, bytes_to_info):
1191
"""Iterate the result of iter_interesting_nodes, yielding the records
1192
and adding to text_keys.
1194
for record, items in interesting_nodes_iterable:
1195
for name, bytes in items:
1196
# Note: we don't care about name_utf8, because groupcompress repos
1197
# are always rich-root, so there are no synthesised root records to
1199
_, file_id, revision_id = bytes_to_info(bytes)
1200
file_id = intern(file_id)
1201
revision_id = intern(revision_id)
1202
text_keys.add(StaticTuple(file_id, revision_id).intern())
1208
1000
class RepositoryFormatCHK1(RepositoryFormatPack):
1209
1001
"""A hashed CHK+group compress pack repository."""
1266
1066
def get_format_string(self):
1267
1067
"""See RepositoryFormat.get_format_string()."""
1268
1068
return ('Bazaar development format - chk repository with bencode '
1269
'revision serialization (needs bzr.dev from 1.16)\n')
1272
class RepositoryFormat2a(RepositoryFormatCHK2):
1273
"""A CHK repository that uses the bencode revision serializer.
1275
This is the same as RepositoryFormatCHK2 but with a public name.
1278
_serializer = chk_serializer.chk_bencode_serializer
1280
def _get_matching_bzrdir(self):
1281
return bzrdir.format_registry.make_bzrdir('2a')
1283
def _ignore_setting_bzrdir(self, format):
1286
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
1288
def get_format_string(self):
1289
return ('Bazaar repository format 2a (needs bzr 1.16 or later)\n')
1291
def get_format_description(self):
1292
"""See RepositoryFormat.get_format_description()."""
1293
return ("Repository format 2a - rich roots, group compression"
1294
" and chk inventories")
1069
'revision serialization (needs bzr.dev from 1.15)\n')
added
removed
bzrlib/repofmt/groupcompress_repo.py
