414
411
def _copy_inventory_texts(self):
415
412
source_vf, target_vf = self._build_vfs('inventory', True, True)
416
# It is not sufficient to just use self.revision_keys, as stacked
417
# repositories can have more inventories than they have revisions.
418
# One alternative would be to do something with
419
# get_parent_map(self.revision_keys), but that shouldn't be any faster
421
inventory_keys = source_vf.keys()
422
missing_inventories = set(self.revision_keys).difference(inventory_keys)
423
if missing_inventories:
424
missing_inventories = sorted(missing_inventories)
425
raise ValueError('We are missing inventories for revisions: %s'
426
% (missing_inventories,))
427
self._copy_stream(source_vf, target_vf, inventory_keys,
413
self._copy_stream(source_vf, target_vf, self.revision_keys,
428
414
'inventories', self._get_filtered_inv_stream, 2)
430
416
def _copy_chk_texts(self):
585
562
pack_factory = GCPack
586
563
resumed_pack_factory = ResumedGCPack
588
def _check_new_inventories(self):
589
"""Detect missing inventories or chk root entries for the new revisions
592
:returns: list of strs, summarising any problems found. If the list is
593
empty no problems were found.
595
# Ensure that all revisions added in this write group have:
596
# - corresponding inventories,
597
# - chk root entries for those inventories,
598
# - and any present parent inventories have their chk root
600
# And all this should be independent of any fallback repository.
602
key_deps = self.repo.revisions._index._key_dependencies
603
new_revisions_keys = key_deps.get_new_keys()
604
no_fallback_inv_index = self.repo.inventories._index
605
no_fallback_chk_bytes_index = self.repo.chk_bytes._index
606
no_fallback_texts_index = self.repo.texts._index
607
inv_parent_map = no_fallback_inv_index.get_parent_map(
609
# Are any inventories for corresponding to the new revisions missing?
610
corresponding_invs = set(inv_parent_map)
611
missing_corresponding = set(new_revisions_keys)
612
missing_corresponding.difference_update(corresponding_invs)
613
if missing_corresponding:
614
problems.append("inventories missing for revisions %s" %
615
(sorted(missing_corresponding),))
617
# Are any chk root entries missing for any inventories? This includes
618
# any present parent inventories, which may be used when calculating
619
# deltas for streaming.
620
all_inv_keys = set(corresponding_invs)
621
for parent_inv_keys in inv_parent_map.itervalues():
622
all_inv_keys.update(parent_inv_keys)
623
# Filter out ghost parents.
624
all_inv_keys.intersection_update(
625
no_fallback_inv_index.get_parent_map(all_inv_keys))
626
parent_invs_only_keys = all_inv_keys.symmetric_difference(
629
inv_ids = [key[-1] for key in all_inv_keys]
630
parent_invs_only_ids = [key[-1] for key in parent_invs_only_keys]
631
root_key_info = _build_interesting_key_sets(
632
self.repo, inv_ids, parent_invs_only_ids)
633
expected_chk_roots = root_key_info.all_keys()
634
present_chk_roots = no_fallback_chk_bytes_index.get_parent_map(
636
missing_chk_roots = expected_chk_roots.difference(present_chk_roots)
637
if missing_chk_roots:
638
problems.append("missing referenced chk root keys: %s"
639
% (sorted(missing_chk_roots),))
640
# Don't bother checking any further.
642
# Find all interesting chk_bytes records, and make sure they are
643
# present, as well as the text keys they reference.
644
chk_bytes_no_fallbacks = self.repo.chk_bytes.without_fallbacks()
645
chk_bytes_no_fallbacks._search_key_func = \
646
self.repo.chk_bytes._search_key_func
647
chk_diff = chk_map.iter_interesting_nodes(
648
chk_bytes_no_fallbacks, root_key_info.interesting_root_keys,
649
root_key_info.uninteresting_root_keys)
650
bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
653
for record in _filter_text_keys(chk_diff, text_keys, bytes_to_info):
655
except errors.NoSuchRevision, e:
656
# XXX: It would be nice if we could give a more precise error here.
657
problems.append("missing chk node(s) for id_to_entry maps")
658
chk_diff = chk_map.iter_interesting_nodes(
659
chk_bytes_no_fallbacks, root_key_info.interesting_pid_root_keys,
660
root_key_info.uninteresting_pid_root_keys)
662
for interesting_rec, interesting_map in chk_diff:
664
except errors.NoSuchRevision, e:
666
"missing chk node(s) for parent_id_basename_to_file_id maps")
667
present_text_keys = no_fallback_texts_index.get_parent_map(text_keys)
668
missing_text_keys = text_keys.difference(present_text_keys)
669
if missing_text_keys:
670
problems.append("missing text keys: %r"
671
% (sorted(missing_text_keys),))
674
565
def _execute_pack_operations(self, pack_operations,
675
566
_packer_class=GCCHKPacker,
676
567
reload_func=None):
698
589
if packer.new_pack is not None:
699
590
packer.new_pack.abort()
703
592
for pack in packs:
704
593
self._remove_pack_from_memory(pack)
705
594
# record the newly available packs and stop advertising the old
708
for _, packs in pack_operations:
709
to_be_obsoleted.extend(packs)
710
result = self._save_pack_names(clear_obsolete_packs=True,
711
obsolete_packs=to_be_obsoleted)
596
self._save_pack_names(clear_obsolete_packs=True)
597
# Move the old packs out of the way now they are no longer referenced.
598
for revision_count, packs in pack_operations:
599
self._obsolete_packs(packs)
715
602
class CHKInventoryRepository(KnitPackRepository):
816
703
' no new_path %r' % (file_id,))
817
704
if new_path == '':
818
705
new_inv.root_id = file_id
819
parent_id_basename_key = StaticTuple('', '').intern()
706
parent_id_basename_key = ('', '')
821
708
utf8_entry_name = entry.name.encode('utf-8')
822
parent_id_basename_key = StaticTuple(entry.parent_id,
823
utf8_entry_name).intern()
709
parent_id_basename_key = (entry.parent_id, utf8_entry_name)
824
710
new_value = entry_to_bytes(entry)
825
711
# Populate Caches?
826
712
# new_inv._path_to_fileid_cache[new_path] = file_id
827
key = StaticTuple(file_id).intern()
828
id_to_entry_dict[key] = new_value
713
id_to_entry_dict[(file_id,)] = new_value
829
714
parent_id_basename_dict[parent_id_basename_key] = file_id
831
716
new_inv._populate_from_dicts(self.chk_bytes, id_to_entry_dict,
881
766
if basis_tree is not None:
882
767
basis_tree.unlock()
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):
769
def _iter_inventories(self, revision_ids):
889
770
"""Iterate over many inventory objects."""
891
ordering = 'unordered'
892
771
keys = [(revision_id,) for revision_id in revision_ids]
893
stream = self.inventories.get_record_stream(keys, ordering, True)
772
stream = self.inventories.get_record_stream(keys, 'unordered', True)
895
774
for record in stream:
896
775
if record.storage_kind != 'absent':
901
780
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
782
def _iter_inventory_xmls(self, revision_ids):
783
# Without a native 'xml' inventory, this method doesn't make sense, so
784
# make it raise to trap naughty direct users.
785
raise NotImplementedError(self._iter_inventory_xmls)
914
787
def _find_present_inventory_keys(self, revision_keys):
915
788
parent_map = self.inventories.get_parent_map(revision_keys)
942
815
present_parent_inv_ids = set(
943
816
[k[-1] for k in present_parent_inv_keys])
817
uninteresting_root_keys = set()
818
interesting_root_keys = set()
944
819
inventories_to_read = set(revision_ids)
945
820
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
821
for inv in self.iter_inventories(inventories_to_read):
822
entry_chk_root_key = inv.id_to_entry.key()
823
if inv.revision_id in present_parent_inv_ids:
824
uninteresting_root_keys.add(entry_chk_root_key)
826
interesting_root_keys.add(entry_chk_root_key)
950
828
chk_bytes = self.chk_bytes
951
829
for record, items in chk_map.iter_interesting_nodes(chk_bytes,
952
830
interesting_root_keys, uninteresting_root_keys,
954
832
for name, bytes in items:
955
833
(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
834
if not rich_root and name_utf8 == '':
1006
880
def _get_source(self, to_format):
1007
881
"""Return a source for streaming from this repository."""
1008
if self._format._serializer == to_format._serializer:
882
if isinstance(to_format, remote.RemoteRepositoryFormat):
883
# Can't just check attributes on to_format with the current code,
885
to_format._ensure_real()
886
to_format = to_format._custom_format
887
if to_format.__class__ is self._format.__class__:
1009
888
# 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.
889
# page layout might be different
1013
890
return GroupCHKStreamSource(self, to_format)
1014
891
return super(CHKInventoryRepository, self)._get_source(to_format)
1090
967
bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
1091
968
chk_bytes = self.from_repository.chk_bytes
1092
969
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,
970
for record, items in chk_map.iter_interesting_nodes(chk_bytes,
971
self._chk_id_roots, uninteresting_root_keys):
972
for name, bytes in items:
973
# Note: we don't care about name_utf8, because we are always
975
_, file_id, revision_id = bytes_to_info(bytes)
976
self._text_keys.add((file_id, revision_id))
1097
977
if record is not None:
1155
1030
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
1033
class RepositoryFormatCHK1(RepositoryFormatPack):
1209
1034
"""A hashed CHK+group compress pack repository."""
1249
1074
return ("Development repository format - rich roots, group compression"
1250
1075
" and chk inventories")
1077
def check_conversion_target(self, target_format):
1078
if not target_format.rich_root_data:
1079
raise errors.BadConversionTarget(
1080
'Does not support rich root data.', target_format)
1081
if (self.supports_tree_reference and
1082
not getattr(target_format, 'supports_tree_reference', False)):
1083
raise errors.BadConversionTarget(
1084
'Does not support nested trees', target_format)
1253
1088
class RepositoryFormatCHK2(RepositoryFormatCHK1):
1254
1089
"""A CHK repository that uses the bencode revision serializer."""
added
removed
bzrlib/repofmt/groupcompress_repo.py
