583
537
return new_pack.data_inserted() and self._data_changed
586
class GCCHKCanonicalizingPacker(GCCHKPacker):
587
"""A packer that ensures inventories have canonical-form CHK maps.
589
Ideally this would be part of reconcile, but it's very slow and rarely
590
needed. (It repairs repositories affected by
591
https://bugs.launchpad.net/bzr/+bug/522637).
594
def __init__(self, *args, **kwargs):
595
super(GCCHKCanonicalizingPacker, self).__init__(*args, **kwargs)
596
self._data_changed = False
598
def _exhaust_stream(self, source_vf, keys, message, vf_to_stream, pb_offset):
599
"""Create and exhaust a stream, but don't insert it.
601
This is useful to get the side-effects of generating a stream.
603
self.pb.update('scanning %s' % (message,), pb_offset)
604
child_pb = ui.ui_factory.nested_progress_bar()
606
list(vf_to_stream(source_vf, keys, message, child_pb))
610
def _copy_inventory_texts(self):
611
source_vf, target_vf = self._build_vfs('inventory', True, True)
612
source_chk_vf, target_chk_vf = self._get_chk_vfs_for_copy()
613
inventory_keys = source_vf.keys()
614
# First, copy the existing CHKs on the assumption that most of them
615
# will be correct. This will save us from having to reinsert (and
616
# recompress) these records later at the cost of perhaps preserving a
618
# (Iterate but don't insert _get_filtered_inv_stream to populate the
619
# variables needed by GCCHKPacker._copy_chk_texts.)
620
self._exhaust_stream(source_vf, inventory_keys, 'inventories',
621
self._get_filtered_inv_stream, 2)
622
GCCHKPacker._copy_chk_texts(self)
623
# Now copy and fix the inventories, and any regenerated CHKs.
624
def chk_canonicalizing_inv_stream(source_vf, keys, message, pb=None):
625
return self._get_filtered_canonicalizing_inv_stream(
626
source_vf, keys, message, pb, source_chk_vf, target_chk_vf)
627
self._copy_stream(source_vf, target_vf, inventory_keys,
628
'inventories', chk_canonicalizing_inv_stream, 4)
630
def _copy_chk_texts(self):
631
# No-op; in this class this happens during _copy_inventory_texts.
634
def _get_filtered_canonicalizing_inv_stream(self, source_vf, keys, message,
635
pb=None, source_chk_vf=None, target_chk_vf=None):
636
"""Filter the texts of inventories, regenerating CHKs to make sure they
639
total_keys = len(keys)
640
target_chk_vf = versionedfile.NoDupeAddLinesDecorator(target_chk_vf)
641
def _filtered_inv_stream():
642
stream = source_vf.get_record_stream(keys, 'groupcompress', True)
643
search_key_name = None
644
for idx, record in enumerate(stream):
645
# Inventories should always be with revisions; assume success.
646
bytes = record.get_bytes_as('fulltext')
647
chk_inv = inventory.CHKInventory.deserialise(
648
source_chk_vf, bytes, record.key)
650
pb.update('inv', idx, total_keys)
651
chk_inv.id_to_entry._ensure_root()
652
if search_key_name is None:
653
# Find the name corresponding to the search_key_func
654
search_key_reg = chk_map.search_key_registry
655
for search_key_name, func in search_key_reg.iteritems():
656
if func == chk_inv.id_to_entry._search_key_func:
658
canonical_inv = inventory.CHKInventory.from_inventory(
659
target_chk_vf, chk_inv,
660
maximum_size=chk_inv.id_to_entry._root_node._maximum_size,
661
search_key_name=search_key_name)
662
if chk_inv.id_to_entry.key() != canonical_inv.id_to_entry.key():
664
'Non-canonical CHK map for id_to_entry of inv: %s '
665
'(root is %s, should be %s)' % (chk_inv.revision_id,
666
chk_inv.id_to_entry.key()[0],
667
canonical_inv.id_to_entry.key()[0]))
668
self._data_changed = True
669
p_id_map = chk_inv.parent_id_basename_to_file_id
670
p_id_map._ensure_root()
671
canon_p_id_map = canonical_inv.parent_id_basename_to_file_id
672
if p_id_map.key() != canon_p_id_map.key():
674
'Non-canonical CHK map for parent_id_to_basename of '
675
'inv: %s (root is %s, should be %s)'
676
% (chk_inv.revision_id, p_id_map.key()[0],
677
canon_p_id_map.key()[0]))
678
self._data_changed = True
679
yield versionedfile.ChunkedContentFactory(record.key,
680
record.parents, record.sha1,
681
canonical_inv.to_lines())
682
# We have finished processing all of the inventory records, we
683
# don't need these sets anymore
684
return _filtered_inv_stream()
686
def _use_pack(self, new_pack):
687
"""Override _use_pack to check for reconcile having changed content."""
688
return new_pack.data_inserted() and self._data_changed
691
540
class GCRepositoryPackCollection(RepositoryPackCollection):
693
542
pack_factory = GCPack
694
resumed_pack_factory = ResumedGCPack
696
def _check_new_inventories(self):
697
"""Detect missing inventories or chk root entries for the new revisions
700
:returns: list of strs, summarising any problems found. If the list is
701
empty no problems were found.
703
# Ensure that all revisions added in this write group have:
704
# - corresponding inventories,
705
# - chk root entries for those inventories,
706
# - and any present parent inventories have their chk root
708
# And all this should be independent of any fallback repository.
710
key_deps = self.repo.revisions._index._key_dependencies
711
new_revisions_keys = key_deps.get_new_keys()
712
no_fallback_inv_index = self.repo.inventories._index
713
no_fallback_chk_bytes_index = self.repo.chk_bytes._index
714
no_fallback_texts_index = self.repo.texts._index
715
inv_parent_map = no_fallback_inv_index.get_parent_map(
717
# Are any inventories for corresponding to the new revisions missing?
718
corresponding_invs = set(inv_parent_map)
719
missing_corresponding = set(new_revisions_keys)
720
missing_corresponding.difference_update(corresponding_invs)
721
if missing_corresponding:
722
problems.append("inventories missing for revisions %s" %
723
(sorted(missing_corresponding),))
725
# Are any chk root entries missing for any inventories? This includes
726
# any present parent inventories, which may be used when calculating
727
# deltas for streaming.
728
all_inv_keys = set(corresponding_invs)
729
for parent_inv_keys in inv_parent_map.itervalues():
730
all_inv_keys.update(parent_inv_keys)
731
# Filter out ghost parents.
732
all_inv_keys.intersection_update(
733
no_fallback_inv_index.get_parent_map(all_inv_keys))
734
parent_invs_only_keys = all_inv_keys.symmetric_difference(
737
inv_ids = [key[-1] for key in all_inv_keys]
738
parent_invs_only_ids = [key[-1] for key in parent_invs_only_keys]
739
root_key_info = _build_interesting_key_sets(
740
self.repo, inv_ids, parent_invs_only_ids)
741
expected_chk_roots = root_key_info.all_keys()
742
present_chk_roots = no_fallback_chk_bytes_index.get_parent_map(
744
missing_chk_roots = expected_chk_roots.difference(present_chk_roots)
745
if missing_chk_roots:
746
problems.append("missing referenced chk root keys: %s"
747
% (sorted(missing_chk_roots),))
748
# Don't bother checking any further.
750
# Find all interesting chk_bytes records, and make sure they are
751
# present, as well as the text keys they reference.
752
chk_bytes_no_fallbacks = self.repo.chk_bytes.without_fallbacks()
753
chk_bytes_no_fallbacks._search_key_func = \
754
self.repo.chk_bytes._search_key_func
755
chk_diff = chk_map.iter_interesting_nodes(
756
chk_bytes_no_fallbacks, root_key_info.interesting_root_keys,
757
root_key_info.uninteresting_root_keys)
760
for record in _filter_text_keys(chk_diff, text_keys,
761
chk_map._bytes_to_text_key):
763
except errors.NoSuchRevision, e:
764
# XXX: It would be nice if we could give a more precise error here.
765
problems.append("missing chk node(s) for id_to_entry maps")
766
chk_diff = chk_map.iter_interesting_nodes(
767
chk_bytes_no_fallbacks, root_key_info.interesting_pid_root_keys,
768
root_key_info.uninteresting_pid_root_keys)
770
for interesting_rec, interesting_map in chk_diff:
772
except errors.NoSuchRevision, e:
774
"missing chk node(s) for parent_id_basename_to_file_id maps")
775
present_text_keys = no_fallback_texts_index.get_parent_map(text_keys)
776
missing_text_keys = text_keys.difference(present_text_keys)
777
if missing_text_keys:
778
problems.append("missing text keys: %r"
779
% (sorted(missing_text_keys),))
544
def _already_packed(self):
545
"""Is the collection already packed?"""
546
# Always repack GC repositories for now
782
549
def _execute_pack_operations(self, pack_operations,
783
550
_packer_class=GCCHKPacker,
841
604
self.inventories = GroupCompressVersionedFiles(
842
605
_GCGraphIndex(self._pack_collection.inventory_index.combined_index,
843
606
add_callback=self._pack_collection.inventory_index.add_callback,
844
parents=True, is_locked=self.is_locked,
845
inconsistency_fatal=False),
607
parents=True, is_locked=self.is_locked),
846
608
access=self._pack_collection.inventory_index.data_access)
847
609
self.revisions = GroupCompressVersionedFiles(
848
610
_GCGraphIndex(self._pack_collection.revision_index.combined_index,
849
611
add_callback=self._pack_collection.revision_index.add_callback,
850
parents=True, is_locked=self.is_locked,
851
track_external_parent_refs=True, track_new_keys=True),
612
parents=True, is_locked=self.is_locked),
852
613
access=self._pack_collection.revision_index.data_access,
854
615
self.signatures = GroupCompressVersionedFiles(
855
616
_GCGraphIndex(self._pack_collection.signature_index.combined_index,
856
617
add_callback=self._pack_collection.signature_index.add_callback,
857
parents=False, is_locked=self.is_locked,
858
inconsistency_fatal=False),
618
parents=False, is_locked=self.is_locked),
859
619
access=self._pack_collection.signature_index.data_access,
861
621
self.texts = GroupCompressVersionedFiles(
862
622
_GCGraphIndex(self._pack_collection.text_index.combined_index,
863
623
add_callback=self._pack_collection.text_index.add_callback,
864
parents=True, is_locked=self.is_locked,
865
inconsistency_fatal=False),
624
parents=True, is_locked=self.is_locked),
866
625
access=self._pack_collection.text_index.data_access)
867
626
# No parents, individual CHK pages don't have specific ancestry
868
627
self.chk_bytes = GroupCompressVersionedFiles(
869
628
_GCGraphIndex(self._pack_collection.chk_index.combined_index,
870
629
add_callback=self._pack_collection.chk_index.add_callback,
871
parents=False, is_locked=self.is_locked,
872
inconsistency_fatal=False),
630
parents=False, is_locked=self.is_locked),
873
631
access=self._pack_collection.chk_index.data_access)
874
search_key_name = self._format._serializer.search_key_name
875
search_key_func = chk_map.search_key_registry.get(search_key_name)
876
self.chk_bytes._search_key_func = search_key_func
877
632
# True when the repository object is 'write locked' (as opposed to the
878
633
# physical lock only taken out around changes to the pack-names list.)
879
634
# Another way to represent this would be a decorator around the control
902
657
return self._inventory_add_lines(revision_id, parents,
903
658
inv_lines, check_content=False)
905
def _create_inv_from_null(self, delta, revision_id):
906
"""This will mutate new_inv directly.
908
This is a simplified form of create_by_apply_delta which knows that all
909
the old values must be None, so everything is a create.
911
serializer = self._format._serializer
912
new_inv = inventory.CHKInventory(serializer.search_key_name)
913
new_inv.revision_id = revision_id
914
entry_to_bytes = new_inv._entry_to_bytes
915
id_to_entry_dict = {}
916
parent_id_basename_dict = {}
917
for old_path, new_path, file_id, entry in delta:
918
if old_path is not None:
919
raise ValueError('Invalid delta, somebody tried to delete %r'
920
' from the NULL_REVISION'
921
% ((old_path, file_id),))
923
raise ValueError('Invalid delta, delta from NULL_REVISION has'
924
' no new_path %r' % (file_id,))
926
new_inv.root_id = file_id
927
parent_id_basename_key = StaticTuple('', '').intern()
929
utf8_entry_name = entry.name.encode('utf-8')
930
parent_id_basename_key = StaticTuple(entry.parent_id,
931
utf8_entry_name).intern()
932
new_value = entry_to_bytes(entry)
934
# new_inv._path_to_fileid_cache[new_path] = file_id
935
key = StaticTuple(file_id).intern()
936
id_to_entry_dict[key] = new_value
937
parent_id_basename_dict[parent_id_basename_key] = file_id
939
new_inv._populate_from_dicts(self.chk_bytes, id_to_entry_dict,
940
parent_id_basename_dict, maximum_size=serializer.maximum_size)
943
660
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
944
661
parents, basis_inv=None, propagate_caches=False):
945
662
"""Add a new inventory expressed as a delta against another revision.
965
682
repository format specific) of the serialized inventory, and the
966
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)
968
688
if not self.is_in_write_group():
969
689
raise AssertionError("%r not in write group" % (self,))
970
690
_mod_revision.check_not_reserved_id(new_revision_id)
972
if basis_inv is None:
973
if basis_revision_id == _mod_revision.NULL_REVISION:
974
new_inv = self._create_inv_from_null(delta, new_revision_id)
975
if new_inv.root_id is None:
976
raise errors.RootMissing()
977
inv_lines = new_inv.to_lines()
978
return self._inventory_add_lines(new_revision_id, parents,
979
inv_lines, check_content=False), new_inv
981
basis_tree = self.revision_tree(basis_revision_id)
982
basis_tree.lock_read()
691
basis_tree = self.revision_tree(basis_revision_id)
692
basis_tree.lock_read()
694
if basis_inv is None:
983
695
basis_inv = basis_tree.inventory
985
696
result = basis_inv.create_by_apply_delta(delta, new_revision_id,
986
697
propagate_caches=propagate_caches)
987
698
inv_lines = result.to_lines()
988
699
return self._inventory_add_lines(new_revision_id, parents,
989
700
inv_lines, check_content=False), result
991
if basis_tree is not None:
994
def _deserialise_inventory(self, revision_id, bytes):
995
return inventory.CHKInventory.deserialise(self.chk_bytes, bytes,
998
def _iter_inventories(self, revision_ids, ordering):
704
def _iter_inventories(self, revision_ids):
999
705
"""Iterate over many inventory objects."""
1000
if ordering is None:
1001
ordering = 'unordered'
1002
706
keys = [(revision_id,) for revision_id in revision_ids]
1003
stream = self.inventories.get_record_stream(keys, ordering, True)
707
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1005
709
for record in stream:
1006
710
if record.storage_kind != 'absent':
1010
714
for key in keys:
1011
715
yield inventory.CHKInventory.deserialise(self.chk_bytes, texts[key], key)
1013
def _iter_inventory_xmls(self, revision_ids, ordering):
1014
# Without a native 'xml' inventory, this method doesn't make sense.
1015
# However older working trees, and older bundles want it - so we supply
1016
# it allowing _get_inventory_xml to work. Bundles currently use the
1017
# serializer directly; this also isn't ideal, but there isn't an xml
1018
# iteration interface offered at all for repositories. We could make
1019
# _iter_inventory_xmls be part of the contract, even if kept private.
1020
inv_to_str = self._serializer.write_inventory_to_string
1021
for inv in self.iter_inventories(revision_ids, ordering=ordering):
1022
yield inv_to_str(inv), inv.revision_id
1024
def _find_present_inventory_keys(self, revision_keys):
1025
parent_map = self.inventories.get_parent_map(revision_keys)
1026
present_inventory_keys = set(k for k in parent_map)
1027
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)
1029
769
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1030
770
"""Find the file ids and versions affected by revisions.
1036
776
revision_ids. Each altered file-ids has the exact revision_ids that
1037
777
altered it listed explicitly.
1039
rich_root = self.supports_rich_root()
1040
bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
1041
file_id_revisions = {}
779
rich_roots = self.supports_rich_root()
1042
781
pb = ui.ui_factory.nested_progress_bar()
1044
revision_keys = [(r,) for r in revision_ids]
1045
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
1046
# TODO: instead of using _find_present_inventory_keys, change the
1047
# code paths to allow missing inventories to be tolerated.
1048
# However, we only want to tolerate missing parent
1049
# inventories, not missing inventories for revision_ids
1050
present_parent_inv_keys = self._find_present_inventory_keys(
1052
present_parent_inv_ids = set(
1053
[k[-1] for k in present_parent_inv_keys])
1054
inventories_to_read = set(revision_ids)
1055
inventories_to_read.update(present_parent_inv_ids)
1056
root_key_info = _build_interesting_key_sets(
1057
self, inventories_to_read, present_parent_inv_ids)
1058
interesting_root_keys = root_key_info.interesting_root_keys
1059
uninteresting_root_keys = root_key_info.uninteresting_root_keys
1060
chk_bytes = self.chk_bytes
1061
for record, items in chk_map.iter_interesting_nodes(chk_bytes,
1062
interesting_root_keys, uninteresting_root_keys,
1064
for name, bytes in items:
1065
(name_utf8, file_id, revision_id) = bytes_to_info(bytes)
1066
# TODO: consider interning file_id, revision_id here, or
1067
# pushing that intern() into bytes_to_info()
1068
# TODO: rich_root should always be True here, for all
1069
# repositories that support chk_bytes
1070
if not rich_root and name_utf8 == '':
1073
file_id_revisions[file_id].add(revision_id)
1075
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)
1078
return file_id_revisions
1080
797
def find_text_key_references(self):
1081
798
"""Find the text key references within the repository.
1113
def reconcile_canonicalize_chks(self):
1114
"""Reconcile this repository to make sure all CHKs are in canonical
1117
from bzrlib.reconcile import PackReconciler
1118
reconciler = PackReconciler(self, thorough=True, canonicalize_chks=True)
1119
reconciler.reconcile()
1122
829
def _reconcile_pack(self, collection, packs, extension, revs, pb):
1123
830
packer = GCCHKReconcilePacker(collection, packs, extension)
1124
831
return packer.pack(pb)
1126
def _canonicalize_chks_pack(self, collection, packs, extension, revs, pb):
1127
packer = GCCHKCanonicalizingPacker(collection, packs, extension, revs)
1128
return packer.pack(pb)
1130
833
def _get_source(self, to_format):
1131
834
"""Return a source for streaming from this repository."""
1132
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__:
1133
841
# We must be exactly the same format, otherwise stuff like the chk
1134
# page layout might be different.
1135
# Actually, this test is just slightly looser than exact so that
1136
# CHK2 <-> 2a transfers will work.
842
# page layout might be different
1137
843
return GroupCHKStreamSource(self, to_format)
1138
844
return super(CHKInventoryRepository, self)._get_source(to_format)
1141
class GroupCHKStreamSource(KnitPackStreamSource):
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):
1142
854
"""Used when both the source and target repo are GroupCHK repos."""
1144
856
def __init__(self, from_repository, to_format):
1191
897
p_id_roots_set.clear()
1192
898
return ('inventories', _filtered_inv_stream())
1194
def _get_filtered_chk_streams(self, excluded_revision_keys):
900
def _get_filtered_chk_streams(self, excluded_keys):
1195
901
self._text_keys = set()
1196
excluded_revision_keys.discard(_mod_revision.NULL_REVISION)
1197
if not excluded_revision_keys:
902
excluded_keys.discard(_mod_revision.NULL_REVISION)
903
if not excluded_keys:
1198
904
uninteresting_root_keys = set()
1199
905
uninteresting_pid_root_keys = set()
1201
# filter out any excluded revisions whose inventories are not
1203
# TODO: Update Repository.iter_inventories() to add
1204
# ignore_missing=True
1205
present_keys = self.from_repository._find_present_inventory_keys(
1206
excluded_revision_keys)
1207
present_ids = [k[-1] for k in present_keys]
1208
907
uninteresting_root_keys = set()
1209
908
uninteresting_pid_root_keys = set()
1210
for inv in self.from_repository.iter_inventories(present_ids):
909
for inv in self.from_repository.iter_inventories(excluded_keys):
1211
910
uninteresting_root_keys.add(inv.id_to_entry.key())
1212
911
uninteresting_pid_root_keys.add(
1213
912
inv.parent_id_basename_to_file_id.key())
913
bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
1214
914
chk_bytes = self.from_repository.chk_bytes
1215
915
def _filter_id_to_entry():
1216
interesting_nodes = chk_map.iter_interesting_nodes(chk_bytes,
1217
self._chk_id_roots, uninteresting_root_keys)
1218
for record in _filter_text_keys(interesting_nodes, self._text_keys,
1219
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))
1220
923
if record is not None:
1223
self._chk_id_roots = None
1224
925
yield 'chk_bytes', _filter_id_to_entry()
1225
926
def _get_parent_id_basename_to_file_id_pages():
1226
927
for record, items in chk_map.iter_interesting_nodes(chk_bytes,
1227
928
self._chk_p_id_roots, uninteresting_pid_root_keys):
1228
929
if record is not None:
1231
self._chk_p_id_roots = None
1232
931
yield 'chk_bytes', _get_parent_id_basename_to_file_id_pages()
933
def _get_text_stream(self):
934
# Note: We know we don't have to handle adding root keys, because both
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))
1234
941
def get_stream(self, search):
1235
def wrap_and_count(pb, rc, stream):
1236
"""Yield records from stream while showing progress."""
1238
for record in stream:
1239
if count == rc.STEP:
1241
pb.update('Estimate', rc.current, rc.max)
1246
942
revision_ids = search.get_keys()
1247
pb = ui.ui_factory.nested_progress_bar()
1248
rc = self._record_counter
1249
self._record_counter.setup(len(revision_ids))
1250
943
for stream_info in self._fetch_revision_texts(revision_ids):
1251
yield (stream_info[0],
1252
wrap_and_count(pb, rc, stream_info[1]))
1253
945
self._revision_keys = [(rev_id,) for rev_id in revision_ids]
1254
self.from_repository.revisions.clear_cache()
1255
self.from_repository.signatures.clear_cache()
1256
s = self._get_inventory_stream(self._revision_keys)
1257
yield (s[0], wrap_and_count(pb, rc, s[1]))
1258
self.from_repository.inventories.clear_cache()
1259
# TODO: The keys to exclude might be part of the search recipe
1260
# For now, exclude all parents that are at the edge of ancestry, for
1261
# which we have inventories
1262
from_repo = self.from_repository
1263
parent_keys = from_repo._find_parent_keys_of_revisions(
1264
self._revision_keys)
1265
for stream_info in self._get_filtered_chk_streams(parent_keys):
1266
yield (stream_info[0], wrap_and_count(pb, rc, stream_info[1]))
1267
self.from_repository.chk_bytes.clear_cache()
1268
s = self._get_text_stream()
1269
yield (s[0], wrap_and_count(pb, rc, s[1]))
1270
self.from_repository.texts.clear_cache()
1271
pb.update('Done', rc.max, rc.max)
1274
def get_stream_for_missing_keys(self, missing_keys):
1275
# missing keys can only occur when we are byte copying and not
1276
# translating (because translation means we don't send
1277
# unreconstructable deltas ever).
1278
missing_inventory_keys = set()
1279
for key in missing_keys:
1280
if key[0] != 'inventories':
1281
raise AssertionError('The only missing keys we should'
1282
' be filling in are inventory keys, not %s'
1284
missing_inventory_keys.add(key[1:])
1285
if self._chk_id_roots or self._chk_p_id_roots:
1286
raise AssertionError('Cannot call get_stream_for_missing_keys'
1287
' until all of get_stream() has been consumed.')
1288
# Yield the inventory stream, so we can find the chk stream
1289
# Some of the missing_keys will be missing because they are ghosts.
1290
# As such, we can ignore them. The Sink is required to verify there are
1291
# no unavailable texts when the ghost inventories are not filled in.
1292
yield self._get_inventory_stream(missing_inventory_keys,
1294
# We use the empty set for excluded_revision_keys, to make it clear
1295
# that we want to transmit all referenced chk pages.
1296
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):
1297
950
yield stream_info
1300
class _InterestingKeyInfo(object):
1302
self.interesting_root_keys = set()
1303
self.interesting_pid_root_keys = set()
1304
self.uninteresting_root_keys = set()
1305
self.uninteresting_pid_root_keys = set()
1307
def all_interesting(self):
1308
return self.interesting_root_keys.union(self.interesting_pid_root_keys)
1310
def all_uninteresting(self):
1311
return self.uninteresting_root_keys.union(
1312
self.uninteresting_pid_root_keys)
1315
return self.all_interesting().union(self.all_uninteresting())
1318
def _build_interesting_key_sets(repo, inventory_ids, parent_only_inv_ids):
1319
result = _InterestingKeyInfo()
1320
for inv in repo.iter_inventories(inventory_ids, 'unordered'):
1321
root_key = inv.id_to_entry.key()
1322
pid_root_key = inv.parent_id_basename_to_file_id.key()
1323
if inv.revision_id in parent_only_inv_ids:
1324
result.uninteresting_root_keys.add(root_key)
1325
result.uninteresting_pid_root_keys.add(pid_root_key)
1327
result.interesting_root_keys.add(root_key)
1328
result.interesting_pid_root_keys.add(pid_root_key)
1332
def _filter_text_keys(interesting_nodes_iterable, text_keys, bytes_to_text_key):
1333
"""Iterate the result of iter_interesting_nodes, yielding the records
1334
and adding to text_keys.
1336
text_keys_update = text_keys.update
1337
for record, items in interesting_nodes_iterable:
1338
text_keys_update([bytes_to_text_key(b) for n,b in items])
951
yield self._get_text_stream()
1344
954
class RepositoryFormatCHK1(RepositoryFormatPack):
1345
955
"""A hashed CHK+group compress pack repository."""
1347
957
repository_class = CHKInventoryRepository
1348
supports_external_lookups = True
1349
958
supports_chks = True
1350
959
# For right now, setting this to True gives us InterModel1And2 rather
1351
960
# than InterDifferingSerializer
1385
993
return ("Development repository format - rich roots, group compression"
1386
994
" and chk inventories")
1389
class RepositoryFormatCHK2(RepositoryFormatCHK1):
1390
"""A CHK repository that uses the bencode revision serializer."""
1392
_serializer = chk_serializer.chk_bencode_serializer
1394
def _get_matching_bzrdir(self):
1395
return bzrdir.format_registry.make_bzrdir('development7-rich-root')
1397
def _ignore_setting_bzrdir(self, format):
1400
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
1402
def get_format_string(self):
1403
"""See RepositoryFormat.get_format_string()."""
1404
return ('Bazaar development format - chk repository with bencode '
1405
'revision serialization (needs bzr.dev from 1.16)\n')
1408
class RepositoryFormat2a(RepositoryFormatCHK2):
1409
"""A CHK repository that uses the bencode revision serializer.
1411
This is the same as RepositoryFormatCHK2 but with a public name.
1414
_serializer = chk_serializer.chk_bencode_serializer
1416
def _get_matching_bzrdir(self):
1417
return bzrdir.format_registry.make_bzrdir('2a')
1419
def _ignore_setting_bzrdir(self, format):
1422
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
1424
def get_format_string(self):
1425
return ('Bazaar repository format 2a (needs bzr 1.16 or later)\n')
1427
def get_format_description(self):
1428
"""See RepositoryFormat.get_format_description()."""
1429
return ("Repository format 2a - rich roots, group compression"
1430
" and chk inventories")
1433
class RepositoryFormat2aSubtree(RepositoryFormat2a):
1434
"""A 2a repository format that supports nested trees.
1438
def _get_matching_bzrdir(self):
1439
return bzrdir.format_registry.make_bzrdir('development-subtree')
1441
def _ignore_setting_bzrdir(self, format):
1444
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
1446
def get_format_string(self):
1447
return ('Bazaar development format 8\n')
1449
def get_format_description(self):
1450
"""See RepositoryFormat.get_format_description()."""
1451
return ("Development repository format 8 - nested trees, "
1452
"group compression and chk inventories")
1455
supports_tree_reference = True
996
def check_conversion_target(self, target_format):
997
if not target_format.rich_root_data:
998
raise errors.BadConversionTarget(
999
'Does not support rich root data.', target_format)
1000
if not getattr(target_format, 'supports_tree_reference', False):
1001
raise errors.BadConversionTarget(
1002
'Does not support nested trees', target_format)
added
removed
bzrlib/repofmt/groupcompress_repo.py
