594
537
return new_pack.data_inserted() and self._data_changed
597
class GCCHKCanonicalizingPacker(GCCHKPacker):
598
"""A packer that ensures inventories have canonical-form CHK maps.
600
Ideally this would be part of reconcile, but it's very slow and rarely
601
needed. (It repairs repositories affected by
602
https://bugs.launchpad.net/bzr/+bug/522637).
605
def __init__(self, *args, **kwargs):
606
super(GCCHKCanonicalizingPacker, self).__init__(*args, **kwargs)
607
self._data_changed = False
609
def _exhaust_stream(self, source_vf, keys, message, vf_to_stream, pb_offset):
610
"""Create and exhaust a stream, but don't insert it.
612
This is useful to get the side-effects of generating a stream.
614
self.pb.update('scanning %s' % (message,), pb_offset)
615
child_pb = ui.ui_factory.nested_progress_bar()
617
list(vf_to_stream(source_vf, keys, message, child_pb))
621
def _copy_inventory_texts(self):
622
source_vf, target_vf = self._build_vfs('inventory', True, True)
623
source_chk_vf, target_chk_vf = self._get_chk_vfs_for_copy()
624
inventory_keys = source_vf.keys()
625
# First, copy the existing CHKs on the assumption that most of them
626
# will be correct. This will save us from having to reinsert (and
627
# recompress) these records later at the cost of perhaps preserving a
629
# (Iterate but don't insert _get_filtered_inv_stream to populate the
630
# variables needed by GCCHKPacker._copy_chk_texts.)
631
self._exhaust_stream(source_vf, inventory_keys, 'inventories',
632
self._get_filtered_inv_stream, 2)
633
GCCHKPacker._copy_chk_texts(self)
634
# Now copy and fix the inventories, and any regenerated CHKs.
635
def chk_canonicalizing_inv_stream(source_vf, keys, message, pb=None):
636
return self._get_filtered_canonicalizing_inv_stream(
637
source_vf, keys, message, pb, source_chk_vf, target_chk_vf)
638
self._copy_stream(source_vf, target_vf, inventory_keys,
639
'inventories', chk_canonicalizing_inv_stream, 4)
641
def _copy_chk_texts(self):
642
# No-op; in this class this happens during _copy_inventory_texts.
645
def _get_filtered_canonicalizing_inv_stream(self, source_vf, keys, message,
646
pb=None, source_chk_vf=None, target_chk_vf=None):
647
"""Filter the texts of inventories, regenerating CHKs to make sure they
650
total_keys = len(keys)
651
target_chk_vf = versionedfile.NoDupeAddLinesDecorator(target_chk_vf)
652
def _filtered_inv_stream():
653
stream = source_vf.get_record_stream(keys, 'groupcompress', True)
654
search_key_name = None
655
for idx, record in enumerate(stream):
656
# Inventories should always be with revisions; assume success.
657
bytes = record.get_bytes_as('fulltext')
658
chk_inv = inventory.CHKInventory.deserialise(
659
source_chk_vf, bytes, record.key)
661
pb.update('inv', idx, total_keys)
662
chk_inv.id_to_entry._ensure_root()
663
if search_key_name is None:
664
# Find the name corresponding to the search_key_func
665
search_key_reg = chk_map.search_key_registry
666
for search_key_name, func in search_key_reg.iteritems():
667
if func == chk_inv.id_to_entry._search_key_func:
669
canonical_inv = inventory.CHKInventory.from_inventory(
670
target_chk_vf, chk_inv,
671
maximum_size=chk_inv.id_to_entry._root_node._maximum_size,
672
search_key_name=search_key_name)
673
if chk_inv.id_to_entry.key() != canonical_inv.id_to_entry.key():
675
'Non-canonical CHK map for id_to_entry of inv: %s '
676
'(root is %s, should be %s)' % (chk_inv.revision_id,
677
chk_inv.id_to_entry.key()[0],
678
canonical_inv.id_to_entry.key()[0]))
679
self._data_changed = True
680
p_id_map = chk_inv.parent_id_basename_to_file_id
681
p_id_map._ensure_root()
682
canon_p_id_map = canonical_inv.parent_id_basename_to_file_id
683
if p_id_map.key() != canon_p_id_map.key():
685
'Non-canonical CHK map for parent_id_to_basename of '
686
'inv: %s (root is %s, should be %s)'
687
% (chk_inv.revision_id, p_id_map.key()[0],
688
canon_p_id_map.key()[0]))
689
self._data_changed = True
690
yield versionedfile.ChunkedContentFactory(record.key,
691
record.parents, record.sha1,
692
canonical_inv.to_lines())
693
# We have finished processing all of the inventory records, we
694
# don't need these sets anymore
695
return _filtered_inv_stream()
697
def _use_pack(self, new_pack):
698
"""Override _use_pack to check for reconcile having changed content."""
699
return new_pack.data_inserted() and self._data_changed
702
540
class GCRepositoryPackCollection(RepositoryPackCollection):
704
542
pack_factory = GCPack
705
resumed_pack_factory = ResumedGCPack
706
normal_packer_class = GCCHKPacker
707
optimising_packer_class = GCCHKPacker
709
def _check_new_inventories(self):
710
"""Detect missing inventories or chk root entries for the new revisions
713
:returns: list of strs, summarising any problems found. If the list is
714
empty no problems were found.
544
def _already_packed(self):
545
"""Is the collection already packed?"""
546
# Always repack GC repositories for now
549
def _execute_pack_operations(self, pack_operations,
550
_packer_class=GCCHKPacker,
552
"""Execute a series of pack operations.
554
:param pack_operations: A list of [revision_count, packs_to_combine].
555
:param _packer_class: The class of packer to use (default: Packer).
716
# Ensure that all revisions added in this write group have:
717
# - corresponding inventories,
718
# - chk root entries for those inventories,
719
# - and any present parent inventories have their chk root
721
# And all this should be independent of any fallback repository.
723
key_deps = self.repo.revisions._index._key_dependencies
724
new_revisions_keys = key_deps.get_new_keys()
725
no_fallback_inv_index = self.repo.inventories._index
726
no_fallback_chk_bytes_index = self.repo.chk_bytes._index
727
no_fallback_texts_index = self.repo.texts._index
728
inv_parent_map = no_fallback_inv_index.get_parent_map(
730
# Are any inventories for corresponding to the new revisions missing?
731
corresponding_invs = set(inv_parent_map)
732
missing_corresponding = set(new_revisions_keys)
733
missing_corresponding.difference_update(corresponding_invs)
734
if missing_corresponding:
735
problems.append("inventories missing for revisions %s" %
736
(sorted(missing_corresponding),))
738
# Are any chk root entries missing for any inventories? This includes
739
# any present parent inventories, which may be used when calculating
740
# deltas for streaming.
741
all_inv_keys = set(corresponding_invs)
742
for parent_inv_keys in inv_parent_map.itervalues():
743
all_inv_keys.update(parent_inv_keys)
744
# Filter out ghost parents.
745
all_inv_keys.intersection_update(
746
no_fallback_inv_index.get_parent_map(all_inv_keys))
747
parent_invs_only_keys = all_inv_keys.symmetric_difference(
750
inv_ids = [key[-1] for key in all_inv_keys]
751
parent_invs_only_ids = [key[-1] for key in parent_invs_only_keys]
752
root_key_info = _build_interesting_key_sets(
753
self.repo, inv_ids, parent_invs_only_ids)
754
expected_chk_roots = root_key_info.all_keys()
755
present_chk_roots = no_fallback_chk_bytes_index.get_parent_map(
757
missing_chk_roots = expected_chk_roots.difference(present_chk_roots)
758
if missing_chk_roots:
759
problems.append("missing referenced chk root keys: %s"
760
% (sorted(missing_chk_roots),))
761
# Don't bother checking any further.
763
# Find all interesting chk_bytes records, and make sure they are
764
# present, as well as the text keys they reference.
765
chk_bytes_no_fallbacks = self.repo.chk_bytes.without_fallbacks()
766
chk_bytes_no_fallbacks._search_key_func = \
767
self.repo.chk_bytes._search_key_func
768
chk_diff = chk_map.iter_interesting_nodes(
769
chk_bytes_no_fallbacks, root_key_info.interesting_root_keys,
770
root_key_info.uninteresting_root_keys)
773
for record in _filter_text_keys(chk_diff, text_keys,
774
chk_map._bytes_to_text_key):
776
except errors.NoSuchRevision, e:
777
# XXX: It would be nice if we could give a more precise error here.
778
problems.append("missing chk node(s) for id_to_entry maps")
779
chk_diff = chk_map.iter_interesting_nodes(
780
chk_bytes_no_fallbacks, root_key_info.interesting_pid_root_keys,
781
root_key_info.uninteresting_pid_root_keys)
783
for interesting_rec, interesting_map in chk_diff:
785
except errors.NoSuchRevision, e:
787
"missing chk node(s) for parent_id_basename_to_file_id maps")
788
present_text_keys = no_fallback_texts_index.get_parent_map(text_keys)
789
missing_text_keys = text_keys.difference(present_text_keys)
790
if missing_text_keys:
791
problems.append("missing text keys: %r"
792
% (sorted(missing_text_keys),))
796
class CHKInventoryRepository(PackRepository):
797
"""subclass of PackRepository that uses CHK based inventories."""
558
# XXX: Copied across from RepositoryPackCollection simply because we
559
# want to override the _packer_class ... :(
560
for revision_count, packs in pack_operations:
561
# we may have no-ops from the setup logic
564
packer = GCCHKPacker(self, packs, '.autopack',
565
reload_func=reload_func)
568
except errors.RetryWithNewPacks:
569
# An exception is propagating out of this context, make sure
570
# this packer has cleaned up. Packer() doesn't set its new_pack
571
# state into the RepositoryPackCollection object, so we only
572
# have access to it directly here.
573
if packer.new_pack is not None:
574
packer.new_pack.abort()
577
self._remove_pack_from_memory(pack)
578
# record the newly available packs and stop advertising the old
580
self._save_pack_names(clear_obsolete_packs=True)
581
# Move the old packs out of the way now they are no longer referenced.
582
for revision_count, packs in pack_operations:
583
self._obsolete_packs(packs)
586
class CHKInventoryRepository(KnitPackRepository):
587
"""subclass of KnitPackRepository that uses CHK based inventories."""
799
589
def __init__(self, _format, a_bzrdir, control_files, _commit_builder_class,
801
591
"""Overridden to change pack collection class."""
802
super(CHKInventoryRepository, self).__init__(_format, a_bzrdir,
803
control_files, _commit_builder_class, _serializer)
592
KnitPackRepository.__init__(self, _format, a_bzrdir, control_files,
593
_commit_builder_class, _serializer)
594
# and now replace everything it did :)
804
595
index_transport = self._transport.clone('indices')
805
596
self._pack_collection = GCRepositoryPackCollection(self,
806
597
self._transport, index_transport,
813
604
self.inventories = GroupCompressVersionedFiles(
814
605
_GCGraphIndex(self._pack_collection.inventory_index.combined_index,
815
606
add_callback=self._pack_collection.inventory_index.add_callback,
816
parents=True, is_locked=self.is_locked,
817
inconsistency_fatal=False),
607
parents=True, is_locked=self.is_locked),
818
608
access=self._pack_collection.inventory_index.data_access)
819
609
self.revisions = GroupCompressVersionedFiles(
820
610
_GCGraphIndex(self._pack_collection.revision_index.combined_index,
821
611
add_callback=self._pack_collection.revision_index.add_callback,
822
parents=True, is_locked=self.is_locked,
823
track_external_parent_refs=True, track_new_keys=True),
612
parents=True, is_locked=self.is_locked),
824
613
access=self._pack_collection.revision_index.data_access,
826
615
self.signatures = GroupCompressVersionedFiles(
827
616
_GCGraphIndex(self._pack_collection.signature_index.combined_index,
828
617
add_callback=self._pack_collection.signature_index.add_callback,
829
parents=False, is_locked=self.is_locked,
830
inconsistency_fatal=False),
618
parents=False, is_locked=self.is_locked),
831
619
access=self._pack_collection.signature_index.data_access,
833
621
self.texts = GroupCompressVersionedFiles(
834
622
_GCGraphIndex(self._pack_collection.text_index.combined_index,
835
623
add_callback=self._pack_collection.text_index.add_callback,
836
parents=True, is_locked=self.is_locked,
837
inconsistency_fatal=False),
624
parents=True, is_locked=self.is_locked),
838
625
access=self._pack_collection.text_index.data_access)
839
626
# No parents, individual CHK pages don't have specific ancestry
840
627
self.chk_bytes = GroupCompressVersionedFiles(
841
628
_GCGraphIndex(self._pack_collection.chk_index.combined_index,
842
629
add_callback=self._pack_collection.chk_index.add_callback,
843
parents=False, is_locked=self.is_locked,
844
inconsistency_fatal=False),
630
parents=False, is_locked=self.is_locked),
845
631
access=self._pack_collection.chk_index.data_access)
846
search_key_name = self._format._serializer.search_key_name
847
search_key_func = chk_map.search_key_registry.get(search_key_name)
848
self.chk_bytes._search_key_func = search_key_func
849
632
# True when the repository object is 'write locked' (as opposed to the
850
633
# physical lock only taken out around changes to the pack-names list.)
851
634
# Another way to represent this would be a decorator around the control
937
682
repository format specific) of the serialized inventory, and the
938
683
resulting inventory.
685
if basis_revision_id == _mod_revision.NULL_REVISION:
686
return KnitPackRepository.add_inventory_by_delta(self,
687
basis_revision_id, delta, new_revision_id, parents)
940
688
if not self.is_in_write_group():
941
689
raise AssertionError("%r not in write group" % (self,))
942
690
_mod_revision.check_not_reserved_id(new_revision_id)
944
if basis_inv is None:
945
if basis_revision_id == _mod_revision.NULL_REVISION:
946
new_inv = self._create_inv_from_null(delta, new_revision_id)
947
if new_inv.root_id is None:
948
raise errors.RootMissing()
949
inv_lines = new_inv.to_lines()
950
return self._inventory_add_lines(new_revision_id, parents,
951
inv_lines, check_content=False), new_inv
953
basis_tree = self.revision_tree(basis_revision_id)
954
basis_tree.lock_read()
691
basis_tree = self.revision_tree(basis_revision_id)
692
basis_tree.lock_read()
694
if basis_inv is None:
955
695
basis_inv = basis_tree.inventory
957
696
result = basis_inv.create_by_apply_delta(delta, new_revision_id,
958
697
propagate_caches=propagate_caches)
959
698
inv_lines = result.to_lines()
960
699
return self._inventory_add_lines(new_revision_id, parents,
961
700
inv_lines, check_content=False), result
963
if basis_tree is not None:
966
def _deserialise_inventory(self, revision_id, bytes):
967
return inventory.CHKInventory.deserialise(self.chk_bytes, bytes,
970
def _iter_inventories(self, revision_ids, ordering):
704
def _iter_inventories(self, revision_ids):
971
705
"""Iterate over many inventory objects."""
973
ordering = 'unordered'
974
706
keys = [(revision_id,) for revision_id in revision_ids]
975
stream = self.inventories.get_record_stream(keys, ordering, True)
707
stream = self.inventories.get_record_stream(keys, 'unordered', True)
977
709
for record in stream:
978
710
if record.storage_kind != 'absent':
983
715
yield inventory.CHKInventory.deserialise(self.chk_bytes, texts[key], key)
985
def _iter_inventory_xmls(self, revision_ids, ordering):
986
# Without a native 'xml' inventory, this method doesn't make sense.
987
# However older working trees, and older bundles want it - so we supply
988
# it allowing _get_inventory_xml to work. Bundles currently use the
989
# serializer directly; this also isn't ideal, but there isn't an xml
990
# iteration interface offered at all for repositories. We could make
991
# _iter_inventory_xmls be part of the contract, even if kept private.
992
inv_to_str = self._serializer.write_inventory_to_string
993
for inv in self.iter_inventories(revision_ids, ordering=ordering):
994
yield inv_to_str(inv), inv.revision_id
996
def _find_present_inventory_keys(self, revision_keys):
997
parent_map = self.inventories.get_parent_map(revision_keys)
998
present_inventory_keys = set(k for k in parent_map)
999
return present_inventory_keys
717
def _iter_inventory_xmls(self, revision_ids):
718
# Without a native 'xml' inventory, this method doesn't make sense, so
719
# make it raise to trap naughty direct users.
720
raise NotImplementedError(self._iter_inventory_xmls)
722
def _find_revision_outside_set(self, revision_ids):
723
revision_set = frozenset(revision_ids)
724
for revid in revision_ids:
725
parent_ids = self.get_parent_map([revid]).get(revid, ())
726
for parent in parent_ids:
727
if parent in revision_set:
728
# Parent is not outside the set
730
if parent not in self.get_parent_map([parent]):
734
return _mod_revision.NULL_REVISION
736
def _find_file_keys_to_fetch(self, revision_ids, pb):
737
rich_root = self.supports_rich_root()
738
revision_outside_set = self._find_revision_outside_set(revision_ids)
739
if revision_outside_set == _mod_revision.NULL_REVISION:
740
uninteresting_root_keys = set()
742
uninteresting_inv = self.get_inventory(revision_outside_set)
743
uninteresting_root_keys = set([uninteresting_inv.id_to_entry.key()])
744
interesting_root_keys = set()
745
for idx, inv in enumerate(self.iter_inventories(revision_ids)):
746
interesting_root_keys.add(inv.id_to_entry.key())
747
revision_ids = frozenset(revision_ids)
748
file_id_revisions = {}
749
bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
750
for record, items in chk_map.iter_interesting_nodes(self.chk_bytes,
751
interesting_root_keys, uninteresting_root_keys,
753
# This is cheating a bit to use the last grabbed 'inv', but it
755
for name, bytes in items:
756
(name_utf8, file_id, revision_id) = bytes_to_info(bytes)
757
if not rich_root and name_utf8 == '':
759
if revision_id in revision_ids:
760
# Would we rather build this up into file_id => revision
763
file_id_revisions[file_id].add(revision_id)
765
file_id_revisions[file_id] = set([revision_id])
766
for file_id, revisions in file_id_revisions.iteritems():
767
yield ('file', file_id, revisions)
1001
769
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1002
770
"""Find the file ids and versions affected by revisions.
1008
776
revision_ids. Each altered file-ids has the exact revision_ids that
1009
777
altered it listed explicitly.
1011
rich_root = self.supports_rich_root()
1012
bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
1013
file_id_revisions = {}
779
rich_roots = self.supports_rich_root()
1014
781
pb = ui.ui_factory.nested_progress_bar()
1016
revision_keys = [(r,) for r in revision_ids]
1017
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
1018
# TODO: instead of using _find_present_inventory_keys, change the
1019
# code paths to allow missing inventories to be tolerated.
1020
# However, we only want to tolerate missing parent
1021
# inventories, not missing inventories for revision_ids
1022
present_parent_inv_keys = self._find_present_inventory_keys(
1024
present_parent_inv_ids = set(
1025
[k[-1] for k in present_parent_inv_keys])
1026
inventories_to_read = set(revision_ids)
1027
inventories_to_read.update(present_parent_inv_ids)
1028
root_key_info = _build_interesting_key_sets(
1029
self, inventories_to_read, present_parent_inv_ids)
1030
interesting_root_keys = root_key_info.interesting_root_keys
1031
uninteresting_root_keys = root_key_info.uninteresting_root_keys
1032
chk_bytes = self.chk_bytes
1033
for record, items in chk_map.iter_interesting_nodes(chk_bytes,
1034
interesting_root_keys, uninteresting_root_keys,
1036
for name, bytes in items:
1037
(name_utf8, file_id, revision_id) = bytes_to_info(bytes)
1038
# TODO: consider interning file_id, revision_id here, or
1039
# pushing that intern() into bytes_to_info()
1040
# TODO: rich_root should always be True here, for all
1041
# repositories that support chk_bytes
1042
if not rich_root and name_utf8 == '':
1045
file_id_revisions[file_id].add(revision_id)
1047
file_id_revisions[file_id] = set([revision_id])
783
total = len(revision_ids)
784
for pos, inv in enumerate(self.iter_inventories(revision_ids)):
785
pb.update("Finding text references", pos, total)
786
for entry in inv.iter_just_entries():
787
if entry.revision != inv.revision_id:
789
if not rich_roots and entry.file_id == inv.root_id:
791
alterations = result.setdefault(entry.file_id, set([]))
792
alterations.add(entry.revision)
1050
return file_id_revisions
1052
797
def find_text_key_references(self):
1053
798
"""Find the text key references within the repository.
1085
def reconcile_canonicalize_chks(self):
1086
"""Reconcile this repository to make sure all CHKs are in canonical
1089
from bzrlib.reconcile import PackReconciler
1090
reconciler = PackReconciler(self, thorough=True, canonicalize_chks=True)
1091
reconciler.reconcile()
1094
829
def _reconcile_pack(self, collection, packs, extension, revs, pb):
1095
830
packer = GCCHKReconcilePacker(collection, packs, extension)
1096
831
return packer.pack(pb)
1098
def _canonicalize_chks_pack(self, collection, packs, extension, revs, pb):
1099
packer = GCCHKCanonicalizingPacker(collection, packs, extension, revs)
1100
return packer.pack(pb)
1102
833
def _get_source(self, to_format):
1103
834
"""Return a source for streaming from this repository."""
1104
if self._format._serializer == to_format._serializer:
835
if isinstance(to_format, remote.RemoteRepositoryFormat):
836
# Can't just check attributes on to_format with the current code,
838
to_format._ensure_real()
839
to_format = to_format._custom_format
840
if to_format.__class__ is self._format.__class__:
1105
841
# We must be exactly the same format, otherwise stuff like the chk
1106
# page layout might be different.
1107
# Actually, this test is just slightly looser than exact so that
1108
# CHK2 <-> 2a transfers will work.
842
# page layout might be different
1109
843
return GroupCHKStreamSource(self, to_format)
1110
844
return super(CHKInventoryRepository, self)._get_source(to_format)
1112
def _find_inconsistent_revision_parents(self, revisions_iterator=None):
1113
"""Find revisions with different parent lists in the revision object
1114
and in the index graph.
1116
:param revisions_iterator: None, or an iterator of (revid,
1117
Revision-or-None). This iterator controls the revisions checked.
1118
:returns: an iterator yielding tuples of (revison-id, parents-in-index,
1119
parents-in-revision).
1121
if not self.is_locked():
1122
raise AssertionError()
1124
if revisions_iterator is None:
1125
revisions_iterator = self._iter_revisions(None)
1126
for revid, revision in revisions_iterator:
1127
if revision is None:
1129
parent_map = vf.get_parent_map([(revid,)])
1130
parents_according_to_index = tuple(parent[-1] for parent in
1131
parent_map[(revid,)])
1132
parents_according_to_revision = tuple(revision.parent_ids)
1133
if parents_according_to_index != parents_according_to_revision:
1134
yield (revid, parents_according_to_index,
1135
parents_according_to_revision)
1137
def _check_for_inconsistent_revision_parents(self):
1138
inconsistencies = list(self._find_inconsistent_revision_parents())
1140
raise errors.BzrCheckError(
1141
"Revision index has inconsistent parents.")
1144
class GroupCHKStreamSource(StreamSource):
846
def suspend_write_group(self):
847
raise errors.UnsuspendableWriteGroup(self)
849
def _resume_write_group(self, tokens):
850
raise errors.UnsuspendableWriteGroup(self)
853
class GroupCHKStreamSource(repository.StreamSource):
1145
854
"""Used when both the source and target repo are GroupCHK repos."""
1147
856
def __init__(self, from_repository, to_format):
1194
897
p_id_roots_set.clear()
1195
898
return ('inventories', _filtered_inv_stream())
1197
def _get_filtered_chk_streams(self, excluded_revision_keys):
900
def _get_filtered_chk_streams(self, excluded_keys):
1198
901
self._text_keys = set()
1199
excluded_revision_keys.discard(_mod_revision.NULL_REVISION)
1200
if not excluded_revision_keys:
902
excluded_keys.discard(_mod_revision.NULL_REVISION)
903
if not excluded_keys:
1201
904
uninteresting_root_keys = set()
1202
905
uninteresting_pid_root_keys = set()
1204
# filter out any excluded revisions whose inventories are not
1206
# TODO: Update Repository.iter_inventories() to add
1207
# ignore_missing=True
1208
present_keys = self.from_repository._find_present_inventory_keys(
1209
excluded_revision_keys)
1210
present_ids = [k[-1] for k in present_keys]
1211
907
uninteresting_root_keys = set()
1212
908
uninteresting_pid_root_keys = set()
1213
for inv in self.from_repository.iter_inventories(present_ids):
909
for inv in self.from_repository.iter_inventories(excluded_keys):
1214
910
uninteresting_root_keys.add(inv.id_to_entry.key())
1215
911
uninteresting_pid_root_keys.add(
1216
912
inv.parent_id_basename_to_file_id.key())
913
bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
1217
914
chk_bytes = self.from_repository.chk_bytes
1218
915
def _filter_id_to_entry():
1219
interesting_nodes = chk_map.iter_interesting_nodes(chk_bytes,
1220
self._chk_id_roots, uninteresting_root_keys)
1221
for record in _filter_text_keys(interesting_nodes, self._text_keys,
1222
chk_map._bytes_to_text_key):
916
for record, items in chk_map.iter_interesting_nodes(chk_bytes,
917
self._chk_id_roots, uninteresting_root_keys):
918
for name, bytes in items:
919
# Note: we don't care about name_utf8, because we are always
921
_, file_id, revision_id = bytes_to_info(bytes)
922
self._text_keys.add((file_id, revision_id))
1223
923
if record is not None:
1226
self._chk_id_roots = None
1227
925
yield 'chk_bytes', _filter_id_to_entry()
1228
926
def _get_parent_id_basename_to_file_id_pages():
1229
927
for record, items in chk_map.iter_interesting_nodes(chk_bytes,
1230
928
self._chk_p_id_roots, uninteresting_pid_root_keys):
1231
929
if record is not None:
1234
self._chk_p_id_roots = None
1235
931
yield 'chk_bytes', _get_parent_id_basename_to_file_id_pages()
1237
933
def _get_text_stream(self):
1238
934
# Note: We know we don't have to handle adding root keys, because both
1239
# the source and target are the identical network name.
1240
text_stream = self.from_repository.texts.get_record_stream(
1241
self._text_keys, self._text_fetch_order, False)
1242
return ('texts', text_stream)
935
# the source and target are GCCHK, and those always support rich-roots
936
# We may want to request as 'unordered', in case the source has done a
938
return ('texts', self.from_repository.texts.get_record_stream(
939
self._text_keys, 'groupcompress', False))
1244
941
def get_stream(self, search):
1245
def wrap_and_count(pb, rc, stream):
1246
"""Yield records from stream while showing progress."""
1248
for record in stream:
1249
if count == rc.STEP:
1251
pb.update('Estimate', rc.current, rc.max)
1256
942
revision_ids = search.get_keys()
1257
pb = ui.ui_factory.nested_progress_bar()
1258
rc = self._record_counter
1259
self._record_counter.setup(len(revision_ids))
1260
943
for stream_info in self._fetch_revision_texts(revision_ids):
1261
yield (stream_info[0],
1262
wrap_and_count(pb, rc, stream_info[1]))
1263
945
self._revision_keys = [(rev_id,) for rev_id in revision_ids]
1264
# TODO: The keys to exclude might be part of the search recipe
1265
# For now, exclude all parents that are at the edge of ancestry, for
1266
# which we have inventories
1267
from_repo = self.from_repository
1268
parent_keys = from_repo._find_parent_keys_of_revisions(
1269
self._revision_keys)
1270
self.from_repository.revisions.clear_cache()
1271
self.from_repository.signatures.clear_cache()
1272
# Clear the repo's get_parent_map cache too.
1273
self.from_repository._unstacked_provider.disable_cache()
1274
self.from_repository._unstacked_provider.enable_cache()
1275
s = self._get_inventory_stream(self._revision_keys)
1276
yield (s[0], wrap_and_count(pb, rc, s[1]))
1277
self.from_repository.inventories.clear_cache()
1278
for stream_info in self._get_filtered_chk_streams(parent_keys):
1279
yield (stream_info[0], wrap_and_count(pb, rc, stream_info[1]))
1280
self.from_repository.chk_bytes.clear_cache()
1281
s = self._get_text_stream()
1282
yield (s[0], wrap_and_count(pb, rc, s[1]))
1283
self.from_repository.texts.clear_cache()
1284
pb.update('Done', rc.max, rc.max)
1287
def get_stream_for_missing_keys(self, missing_keys):
1288
# missing keys can only occur when we are byte copying and not
1289
# translating (because translation means we don't send
1290
# unreconstructable deltas ever).
1291
missing_inventory_keys = set()
1292
for key in missing_keys:
1293
if key[0] != 'inventories':
1294
raise AssertionError('The only missing keys we should'
1295
' be filling in are inventory keys, not %s'
1297
missing_inventory_keys.add(key[1:])
1298
if self._chk_id_roots or self._chk_p_id_roots:
1299
raise AssertionError('Cannot call get_stream_for_missing_keys'
1300
' until all of get_stream() has been consumed.')
1301
# Yield the inventory stream, so we can find the chk stream
1302
# Some of the missing_keys will be missing because they are ghosts.
1303
# As such, we can ignore them. The Sink is required to verify there are
1304
# no unavailable texts when the ghost inventories are not filled in.
1305
yield self._get_inventory_stream(missing_inventory_keys,
1307
# We use the empty set for excluded_revision_keys, to make it clear
1308
# that we want to transmit all referenced chk pages.
1309
for stream_info in self._get_filtered_chk_streams(set()):
946
yield self._get_filtered_inv_stream()
947
# The keys to exclude are part of the search recipe
948
_, _, exclude_keys, _ = search.get_recipe()
949
for stream_info in self._get_filtered_chk_streams(exclude_keys):
1310
950
yield stream_info
1313
class _InterestingKeyInfo(object):
1315
self.interesting_root_keys = set()
1316
self.interesting_pid_root_keys = set()
1317
self.uninteresting_root_keys = set()
1318
self.uninteresting_pid_root_keys = set()
1320
def all_interesting(self):
1321
return self.interesting_root_keys.union(self.interesting_pid_root_keys)
1323
def all_uninteresting(self):
1324
return self.uninteresting_root_keys.union(
1325
self.uninteresting_pid_root_keys)
1328
return self.all_interesting().union(self.all_uninteresting())
1331
def _build_interesting_key_sets(repo, inventory_ids, parent_only_inv_ids):
1332
result = _InterestingKeyInfo()
1333
for inv in repo.iter_inventories(inventory_ids, 'unordered'):
1334
root_key = inv.id_to_entry.key()
1335
pid_root_key = inv.parent_id_basename_to_file_id.key()
1336
if inv.revision_id in parent_only_inv_ids:
1337
result.uninteresting_root_keys.add(root_key)
1338
result.uninteresting_pid_root_keys.add(pid_root_key)
1340
result.interesting_root_keys.add(root_key)
1341
result.interesting_pid_root_keys.add(pid_root_key)
1345
def _filter_text_keys(interesting_nodes_iterable, text_keys, bytes_to_text_key):
1346
"""Iterate the result of iter_interesting_nodes, yielding the records
1347
and adding to text_keys.
1349
text_keys_update = text_keys.update
1350
for record, items in interesting_nodes_iterable:
1351
text_keys_update([bytes_to_text_key(b) for n,b in items])
1355
class RepositoryFormat2a(RepositoryFormatPack):
1356
"""A CHK repository that uses the bencode revision serializer."""
951
yield self._get_text_stream()
954
class RepositoryFormatCHK1(RepositoryFormatPack):
955
"""A hashed CHK+group compress pack repository."""
1358
957
repository_class = CHKInventoryRepository
1359
supports_external_lookups = True
1360
958
supports_chks = True
959
# For right now, setting this to True gives us InterModel1And2 rather
960
# than InterDifferingSerializer
1361
961
_commit_builder_class = PackRootCommitBuilder
1362
962
rich_root_data = True
1363
_serializer = chk_serializer.chk_bencode_serializer
963
_serializer = chk_serializer.chk_serializer_255_bigpage
1364
964
_commit_inv_deltas = True
1365
965
# What index classes to use
1366
966
index_builder_class = BTreeBuilder
added
removed
bzrlib/repofmt/groupcompress_repo.py
