163
163
have deltas based on a fallback repository.
164
164
(See <https://bugs.launchpad.net/bzr/+bug/288751>)
166
# Groupcompress packs don't have any external references
166
# Groupcompress packs don't have any external references, arguably CHK
167
# pages have external references, but we cannot 'cheaply' determine
168
# them without actually walking all of the chk pages.
171
class ResumedGCPack(ResumedPack):
173
def _check_references(self):
174
"""Make sure our external compression parents are present."""
175
# See GCPack._check_references for why this is empty
177
def _get_external_refs(self, index):
178
# GC repositories don't have compression parents external to a given
169
183
class GCCHKPacker(Packer):
604
620
self.inventories = GroupCompressVersionedFiles(
605
621
_GCGraphIndex(self._pack_collection.inventory_index.combined_index,
606
622
add_callback=self._pack_collection.inventory_index.add_callback,
607
parents=True, is_locked=self.is_locked),
623
parents=True, is_locked=self.is_locked,
624
inconsistency_fatal=False),
608
625
access=self._pack_collection.inventory_index.data_access)
609
626
self.revisions = GroupCompressVersionedFiles(
610
627
_GCGraphIndex(self._pack_collection.revision_index.combined_index,
611
628
add_callback=self._pack_collection.revision_index.add_callback,
612
parents=True, is_locked=self.is_locked),
629
parents=True, is_locked=self.is_locked,
630
track_external_parent_refs=True),
613
631
access=self._pack_collection.revision_index.data_access,
615
633
self.signatures = GroupCompressVersionedFiles(
616
634
_GCGraphIndex(self._pack_collection.signature_index.combined_index,
617
635
add_callback=self._pack_collection.signature_index.add_callback,
618
parents=False, is_locked=self.is_locked),
636
parents=False, is_locked=self.is_locked,
637
inconsistency_fatal=False),
619
638
access=self._pack_collection.signature_index.data_access,
621
640
self.texts = GroupCompressVersionedFiles(
622
641
_GCGraphIndex(self._pack_collection.text_index.combined_index,
623
642
add_callback=self._pack_collection.text_index.add_callback,
624
parents=True, is_locked=self.is_locked),
643
parents=True, is_locked=self.is_locked,
644
inconsistency_fatal=False),
625
645
access=self._pack_collection.text_index.data_access)
626
646
# No parents, individual CHK pages don't have specific ancestry
627
647
self.chk_bytes = GroupCompressVersionedFiles(
628
648
_GCGraphIndex(self._pack_collection.chk_index.combined_index,
629
649
add_callback=self._pack_collection.chk_index.add_callback,
630
parents=False, is_locked=self.is_locked),
650
parents=False, is_locked=self.is_locked,
651
inconsistency_fatal=False),
631
652
access=self._pack_collection.chk_index.data_access)
632
653
# True when the repository object is 'write locked' (as opposed to the
633
654
# physical lock only taken out around changes to the pack-names list.)
657
678
return self._inventory_add_lines(revision_id, parents,
658
679
inv_lines, check_content=False)
681
def _create_inv_from_null(self, delta, revision_id):
682
"""This will mutate new_inv directly.
684
This is a simplified form of create_by_apply_delta which knows that all
685
the old values must be None, so everything is a create.
687
serializer = self._format._serializer
688
new_inv = inventory.CHKInventory(serializer.search_key_name)
689
new_inv.revision_id = revision_id
690
entry_to_bytes = new_inv._entry_to_bytes
691
id_to_entry_dict = {}
692
parent_id_basename_dict = {}
693
for old_path, new_path, file_id, entry in delta:
694
if old_path is not None:
695
raise ValueError('Invalid delta, somebody tried to delete %r'
696
' from the NULL_REVISION'
697
% ((old_path, file_id),))
699
raise ValueError('Invalid delta, delta from NULL_REVISION has'
700
' no new_path %r' % (file_id,))
702
new_inv.root_id = file_id
703
parent_id_basename_key = ('', '')
705
utf8_entry_name = entry.name.encode('utf-8')
706
parent_id_basename_key = (entry.parent_id, utf8_entry_name)
707
new_value = entry_to_bytes(entry)
709
# new_inv._path_to_fileid_cache[new_path] = file_id
710
id_to_entry_dict[(file_id,)] = new_value
711
parent_id_basename_dict[parent_id_basename_key] = file_id
713
new_inv._populate_from_dicts(self.chk_bytes, id_to_entry_dict,
714
parent_id_basename_dict, maximum_size=serializer.maximum_size)
660
717
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
661
718
parents, basis_inv=None, propagate_caches=False):
662
719
"""Add a new inventory expressed as a delta against another revision.
682
739
repository format specific) of the serialized inventory, and the
683
740
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)
688
742
if not self.is_in_write_group():
689
743
raise AssertionError("%r not in write group" % (self,))
690
744
_mod_revision.check_not_reserved_id(new_revision_id)
691
basis_tree = self.revision_tree(basis_revision_id)
692
basis_tree.lock_read()
694
if basis_inv is None:
746
if basis_inv is None:
747
if basis_revision_id == _mod_revision.NULL_REVISION:
748
new_inv = self._create_inv_from_null(delta, new_revision_id)
749
inv_lines = new_inv.to_lines()
750
return self._inventory_add_lines(new_revision_id, parents,
751
inv_lines, check_content=False), new_inv
753
basis_tree = self.revision_tree(basis_revision_id)
754
basis_tree.lock_read()
695
755
basis_inv = basis_tree.inventory
696
757
result = basis_inv.create_by_apply_delta(delta, new_revision_id,
697
758
propagate_caches=propagate_caches)
698
759
inv_lines = result.to_lines()
699
760
return self._inventory_add_lines(new_revision_id, parents,
700
761
inv_lines, check_content=False), result
763
if basis_tree is not None:
704
766
def _iter_inventories(self, revision_ids):
705
767
"""Iterate over many inventory objects."""
719
781
# make it raise to trap naughty direct users.
720
782
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)
784
def _find_present_inventory_keys(self, revision_keys):
785
parent_map = self.inventories.get_parent_map(revision_keys)
786
present_inventory_keys = set(k for k in parent_map)
787
return present_inventory_keys
769
789
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
770
790
"""Find the file ids and versions affected by revisions.
776
796
revision_ids. Each altered file-ids has the exact revision_ids that
777
797
altered it listed explicitly.
779
rich_roots = self.supports_rich_root()
799
rich_root = self.supports_rich_root()
800
bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
801
file_id_revisions = {}
781
802
pb = ui.ui_factory.nested_progress_bar()
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)
804
revision_keys = [(r,) for r in revision_ids]
805
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
806
# TODO: instead of using _find_present_inventory_keys, change the
807
# code paths to allow missing inventories to be tolerated.
808
# However, we only want to tolerate missing parent
809
# inventories, not missing inventories for revision_ids
810
present_parent_inv_keys = self._find_present_inventory_keys(
812
present_parent_inv_ids = set(
813
[k[-1] for k in present_parent_inv_keys])
814
uninteresting_root_keys = set()
815
interesting_root_keys = set()
816
inventories_to_read = set(revision_ids)
817
inventories_to_read.update(present_parent_inv_ids)
818
for inv in self.iter_inventories(inventories_to_read):
819
entry_chk_root_key = inv.id_to_entry.key()
820
if inv.revision_id in present_parent_inv_ids:
821
uninteresting_root_keys.add(entry_chk_root_key)
823
interesting_root_keys.add(entry_chk_root_key)
825
chk_bytes = self.chk_bytes
826
for record, items in chk_map.iter_interesting_nodes(chk_bytes,
827
interesting_root_keys, uninteresting_root_keys,
829
for name, bytes in items:
830
(name_utf8, file_id, revision_id) = bytes_to_info(bytes)
831
if not rich_root and name_utf8 == '':
834
file_id_revisions[file_id].add(revision_id)
836
file_id_revisions[file_id] = set([revision_id])
839
return file_id_revisions
797
841
def find_text_key_references(self):
798
842
"""Find the text key references within the repository.
843
887
return GroupCHKStreamSource(self, to_format)
844
888
return super(CHKInventoryRepository, self)._get_source(to_format)
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):
891
class GroupCHKStreamSource(KnitPackStreamSource):
854
892
"""Used when both the source and target repo are GroupCHK repos."""
856
894
def __init__(self, from_repository, to_format):
858
896
super(GroupCHKStreamSource, self).__init__(from_repository, to_format)
859
897
self._revision_keys = None
860
898
self._text_keys = None
899
self._text_fetch_order = 'groupcompress'
861
900
self._chk_id_roots = None
862
901
self._chk_p_id_roots = None
864
def _get_filtered_inv_stream(self):
903
def _get_inventory_stream(self, inventory_keys, allow_absent=False):
865
904
"""Get a stream of inventory texts.
867
906
When this function returns, self._chk_id_roots and self._chk_p_id_roots
897
941
p_id_roots_set.clear()
898
942
return ('inventories', _filtered_inv_stream())
900
def _get_filtered_chk_streams(self, excluded_keys):
944
def _get_filtered_chk_streams(self, excluded_revision_keys):
901
945
self._text_keys = set()
902
excluded_keys.discard(_mod_revision.NULL_REVISION)
903
if not excluded_keys:
946
excluded_revision_keys.discard(_mod_revision.NULL_REVISION)
947
if not excluded_revision_keys:
904
948
uninteresting_root_keys = set()
905
949
uninteresting_pid_root_keys = set()
951
# filter out any excluded revisions whose inventories are not
953
# TODO: Update Repository.iter_inventories() to add
954
# ignore_missing=True
955
present_keys = self.from_repository._find_present_inventory_keys(
956
excluded_revision_keys)
957
present_ids = [k[-1] for k in present_keys]
907
958
uninteresting_root_keys = set()
908
959
uninteresting_pid_root_keys = set()
909
for inv in self.from_repository.iter_inventories(excluded_keys):
960
for inv in self.from_repository.iter_inventories(present_ids):
910
961
uninteresting_root_keys.add(inv.id_to_entry.key())
911
962
uninteresting_pid_root_keys.add(
912
963
inv.parent_id_basename_to_file_id.key())
922
973
self._text_keys.add((file_id, revision_id))
923
974
if record is not None:
977
self._chk_id_roots = None
925
978
yield 'chk_bytes', _filter_id_to_entry()
926
979
def _get_parent_id_basename_to_file_id_pages():
927
980
for record, items in chk_map.iter_interesting_nodes(chk_bytes,
928
981
self._chk_p_id_roots, uninteresting_pid_root_keys):
929
982
if record is not None:
985
self._chk_p_id_roots = None
931
986
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))
941
988
def get_stream(self, search):
942
989
revision_ids = search.get_keys()
943
990
for stream_info in self._fetch_revision_texts(revision_ids):
944
991
yield stream_info
945
992
self._revision_keys = [(rev_id,) for rev_id in revision_ids]
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):
993
yield self._get_inventory_stream(self._revision_keys)
994
# TODO: The keys to exclude might be part of the search recipe
995
# For now, exclude all parents that are at the edge of ancestry, for
996
# which we have inventories
997
from_repo = self.from_repository
998
parent_keys = from_repo._find_parent_keys_of_revisions(
1000
for stream_info in self._get_filtered_chk_streams(parent_keys):
950
1001
yield stream_info
951
1002
yield self._get_text_stream()
1004
def get_stream_for_missing_keys(self, missing_keys):
1005
# missing keys can only occur when we are byte copying and not
1006
# translating (because translation means we don't send
1007
# unreconstructable deltas ever).
1008
missing_inventory_keys = set()
1009
for key in missing_keys:
1010
if key[0] != 'inventories':
1011
raise AssertionError('The only missing keys we should'
1012
' be filling in are inventory keys, not %s'
1014
missing_inventory_keys.add(key[1:])
1015
if self._chk_id_roots or self._chk_p_id_roots:
1016
raise AssertionError('Cannot call get_stream_for_missing_keys'
1017
' untill all of get_stream() has been consumed.')
1018
# Yield the inventory stream, so we can find the chk stream
1019
# Some of the missing_keys will be missing because they are ghosts.
1020
# As such, we can ignore them. The Sink is required to verify there are
1021
# no unavailable texts when the ghost inventories are not filled in.
1022
yield self._get_inventory_stream(missing_inventory_keys,
1024
# We use the empty set for excluded_revision_keys, to make it clear
1025
# that we want to transmit all referenced chk pages.
1026
for stream_info in self._get_filtered_chk_streams(set()):
954
1030
class RepositoryFormatCHK1(RepositoryFormatPack):
955
1031
"""A hashed CHK+group compress pack repository."""
957
1033
repository_class = CHKInventoryRepository
1034
supports_external_lookups = True
958
1035
supports_chks = True
959
1036
# For right now, setting this to True gives us InterModel1And2 rather
960
1037
# than InterDifferingSerializer
997
1075
if not target_format.rich_root_data:
998
1076
raise errors.BadConversionTarget(
999
1077
'Does not support rich root data.', target_format)
1000
if not getattr(target_format, 'supports_tree_reference', False):
1078
if (self.supports_tree_reference and
1079
not getattr(target_format, 'supports_tree_reference', False)):
1001
1080
raise errors.BadConversionTarget(
1002
1081
'Does not support nested trees', target_format)
1085
class RepositoryFormatCHK2(RepositoryFormatCHK1):
1086
"""A CHK repository that uses the bencode revision serializer."""
1088
_serializer = chk_serializer.chk_bencode_serializer
1090
def _get_matching_bzrdir(self):
1091
return bzrdir.format_registry.make_bzrdir('development7-rich-root')
1093
def _ignore_setting_bzrdir(self, format):
1096
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
1098
def get_format_string(self):
1099
"""See RepositoryFormat.get_format_string()."""
1100
return ('Bazaar development format - chk repository with bencode '
1101
'revision serialization (needs bzr.dev from 1.16)\n')
1104
class RepositoryFormat2a(RepositoryFormatCHK2):
1105
"""A CHK repository that uses the bencode revision serializer.
1107
This is the same as RepositoryFormatCHK2 but with a public name.
1110
_serializer = chk_serializer.chk_bencode_serializer
1112
def _get_matching_bzrdir(self):
1113
return bzrdir.format_registry.make_bzrdir('2a')
1115
def _ignore_setting_bzrdir(self, format):
1118
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
1120
def get_format_string(self):
1121
return ('Bazaar repository format 2a (needs bzr 1.16 or later)\n')
added
removed
bzrlib/repofmt/groupcompress_repo.py
