593
390
# unchanged, carry over.
594
391
ie.reference_revision = parent_entry.reference_revision
595
392
ie.revision = parent_entry.revision
596
return self._get_delta(ie, basis_inv, path), False, None
393
return self._get_delta(ie, basis_inv, path), False
597
394
ie.reference_revision = content_summary[3]
598
if ie.reference_revision is None:
599
raise AssertionError("invalid content_summary for nested tree: %r"
600
% (content_summary,))
601
self._add_text_to_weave(ie.file_id, '', heads, None)
396
self._add_text_to_weave(ie.file_id, lines, heads, None)
603
398
raise NotImplementedError('unknown kind')
604
399
ie.revision = self._new_revision_id
605
self._any_changes = True
606
return self._get_delta(ie, basis_inv, path), True, fingerprint
608
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
609
_entry_factory=entry_factory):
610
"""Record a new tree via iter_changes.
612
:param tree: The tree to obtain text contents from for changed objects.
613
:param basis_revision_id: The revision id of the tree the iter_changes
614
has been generated against. Currently assumed to be the same
615
as self.parents[0] - if it is not, errors may occur.
616
:param iter_changes: An iter_changes iterator with the changes to apply
617
to basis_revision_id. The iterator must not include any items with
618
a current kind of None - missing items must be either filtered out
619
or errored-on beefore record_iter_changes sees the item.
620
:param _entry_factory: Private method to bind entry_factory locally for
622
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
625
# Create an inventory delta based on deltas between all the parents and
626
# deltas between all the parent inventories. We use inventory delta's
627
# between the inventory objects because iter_changes masks
628
# last-changed-field only changes.
630
# file_id -> change map, change is fileid, paths, changed, versioneds,
631
# parents, names, kinds, executables
633
# {file_id -> revision_id -> inventory entry, for entries in parent
634
# trees that are not parents[0]
400
return self._get_delta(ie, basis_inv, path), True
402
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
403
versionedfile = self.repository.weave_store.get_weave_or_empty(
404
file_id, self.repository.get_transaction())
405
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
406
# than add_lines, and allows committing when a parent is ghosted for
408
# Note: as we read the content directly from the tree, we know its not
409
# been turned into unicode or badly split - but a broken tree
410
# implementation could give us bad output from readlines() so this is
411
# not a guarantee of safety. What would be better is always checking
412
# the content during test suite execution. RBC 20070912
638
revtrees = list(self.repository.revision_trees(self.parents))
639
except errors.NoSuchRevision:
640
# one or more ghosts, slow path.
642
for revision_id in self.parents:
644
revtrees.append(self.repository.revision_tree(revision_id))
645
except errors.NoSuchRevision:
647
basis_revision_id = _mod_revision.NULL_REVISION
649
revtrees.append(self.repository.revision_tree(
650
_mod_revision.NULL_REVISION))
651
# The basis inventory from a repository
653
basis_inv = revtrees[0].inventory
655
basis_inv = self.repository.revision_tree(
656
_mod_revision.NULL_REVISION).inventory
657
if len(self.parents) > 0:
658
if basis_revision_id != self.parents[0] and not ghost_basis:
660
"arbitrary basis parents not yet supported with merges")
661
for revtree in revtrees[1:]:
662
for change in revtree.inventory._make_delta(basis_inv):
663
if change[1] is None:
664
# Not present in this parent.
666
if change[2] not in merged_ids:
667
if change[0] is not None:
668
basis_entry = basis_inv[change[2]]
669
merged_ids[change[2]] = [
671
basis_entry.revision,
674
parent_entries[change[2]] = {
676
basis_entry.revision:basis_entry,
678
change[3].revision:change[3],
681
merged_ids[change[2]] = [change[3].revision]
682
parent_entries[change[2]] = {change[3].revision:change[3]}
684
merged_ids[change[2]].append(change[3].revision)
685
parent_entries[change[2]][change[3].revision] = change[3]
688
# Setup the changes from the tree:
689
# changes maps file_id -> (change, [parent revision_ids])
691
for change in iter_changes:
692
# This probably looks up in basis_inv way to much.
693
if change[1][0] is not None:
694
head_candidate = [basis_inv[change[0]].revision]
697
changes[change[0]] = change, merged_ids.get(change[0],
699
unchanged_merged = set(merged_ids) - set(changes)
700
# Extend the changes dict with synthetic changes to record merges of
702
for file_id in unchanged_merged:
703
# Record a merged version of these items that did not change vs the
704
# basis. This can be either identical parallel changes, or a revert
705
# of a specific file after a merge. The recorded content will be
706
# that of the current tree (which is the same as the basis), but
707
# the per-file graph will reflect a merge.
708
# NB:XXX: We are reconstructing path information we had, this
709
# should be preserved instead.
710
# inv delta change: (file_id, (path_in_source, path_in_target),
711
# changed_content, versioned, parent, name, kind,
714
basis_entry = basis_inv[file_id]
715
except errors.NoSuchId:
716
# a change from basis->some_parents but file_id isn't in basis
717
# so was new in the merge, which means it must have changed
718
# from basis -> current, and as it hasn't the add was reverted
719
# by the user. So we discard this change.
723
(basis_inv.id2path(file_id), tree.id2path(file_id)),
725
(basis_entry.parent_id, basis_entry.parent_id),
726
(basis_entry.name, basis_entry.name),
727
(basis_entry.kind, basis_entry.kind),
728
(basis_entry.executable, basis_entry.executable))
729
changes[file_id] = (change, merged_ids[file_id])
730
# changes contains tuples with the change and a set of inventory
731
# candidates for the file.
733
# old_path, new_path, file_id, new_inventory_entry
734
seen_root = False # Is the root in the basis delta?
735
inv_delta = self._basis_delta
736
modified_rev = self._new_revision_id
737
for change, head_candidates in changes.values():
738
if change[3][1]: # versioned in target.
739
# Several things may be happening here:
740
# We may have a fork in the per-file graph
741
# - record a change with the content from tree
742
# We may have a change against < all trees
743
# - carry over the tree that hasn't changed
744
# We may have a change against all trees
745
# - record the change with the content from tree
748
entry = _entry_factory[kind](file_id, change[5][1],
750
head_set = self._heads(change[0], set(head_candidates))
753
for head_candidate in head_candidates:
754
if head_candidate in head_set:
755
heads.append(head_candidate)
756
head_set.remove(head_candidate)
759
# Could be a carry-over situation:
760
parent_entry_revs = parent_entries.get(file_id, None)
761
if parent_entry_revs:
762
parent_entry = parent_entry_revs.get(heads[0], None)
765
if parent_entry is None:
766
# The parent iter_changes was called against is the one
767
# that is the per-file head, so any change is relevant
768
# iter_changes is valid.
769
carry_over_possible = False
771
# could be a carry over situation
772
# A change against the basis may just indicate a merge,
773
# we need to check the content against the source of the
774
# merge to determine if it was changed after the merge
776
if (parent_entry.kind != entry.kind or
777
parent_entry.parent_id != entry.parent_id or
778
parent_entry.name != entry.name):
779
# Metadata common to all entries has changed
780
# against per-file parent
781
carry_over_possible = False
783
carry_over_possible = True
784
# per-type checks for changes against the parent_entry
787
# Cannot be a carry-over situation
788
carry_over_possible = False
789
# Populate the entry in the delta
791
# XXX: There is still a small race here: If someone reverts the content of a file
792
# after iter_changes examines and decides it has changed,
793
# we will unconditionally record a new version even if some
794
# other process reverts it while commit is running (with
795
# the revert happening after iter_changes did its
798
entry.executable = True
800
entry.executable = False
801
if (carry_over_possible and
802
parent_entry.executable == entry.executable):
803
# Check the file length, content hash after reading
805
nostore_sha = parent_entry.text_sha1
808
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
810
text = file_obj.read()
814
entry.text_sha1, entry.text_size = self._add_text_to_weave(
815
file_id, text, heads, nostore_sha)
816
yield file_id, change[1][1], (entry.text_sha1, stat_value)
817
except errors.ExistingContent:
818
# No content change against a carry_over parent
819
# Perhaps this should also yield a fs hash update?
821
entry.text_size = parent_entry.text_size
822
entry.text_sha1 = parent_entry.text_sha1
823
elif kind == 'symlink':
825
entry.symlink_target = tree.get_symlink_target(file_id)
826
if (carry_over_possible and
827
parent_entry.symlink_target == entry.symlink_target):
830
self._add_text_to_weave(change[0], '', heads, None)
831
elif kind == 'directory':
832
if carry_over_possible:
835
# Nothing to set on the entry.
836
# XXX: split into the Root and nonRoot versions.
837
if change[1][1] != '' or self.repository.supports_rich_root():
838
self._add_text_to_weave(change[0], '', heads, None)
839
elif kind == 'tree-reference':
840
if not self.repository._format.supports_tree_reference:
841
# This isn't quite sane as an error, but we shouldn't
842
# ever see this code path in practice: tree's don't
843
# permit references when the repo doesn't support tree
845
raise errors.UnsupportedOperation(tree.add_reference,
847
reference_revision = tree.get_reference_revision(change[0])
848
entry.reference_revision = reference_revision
849
if (carry_over_possible and
850
parent_entry.reference_revision == reference_revision):
853
self._add_text_to_weave(change[0], '', heads, None)
855
raise AssertionError('unknown kind %r' % kind)
857
entry.revision = modified_rev
859
entry.revision = parent_entry.revision
862
new_path = change[1][1]
863
inv_delta.append((change[1][0], new_path, change[0], entry))
866
self.new_inventory = None
868
# This should perhaps be guarded by a check that the basis we
869
# commit against is the basis for the commit and if not do a delta
871
self._any_changes = True
873
# housekeeping root entry changes do not affect no-change commits.
874
self._require_root_change(tree)
875
self.basis_delta_revision = basis_revision_id
877
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
878
parent_keys = tuple([(file_id, parent) for parent in parents])
879
return self.repository.texts._add_text(
880
(file_id, self._new_revision_id), parent_keys, new_text,
881
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
414
return versionedfile.add_lines_with_ghosts(
415
self._new_revision_id, parents, new_lines,
416
nostore_sha=nostore_sha, random_id=self.random_revid,
417
check_content=False)[0:2]
419
versionedfile.clear_cache()
884
422
class RootCommitBuilder(CommitBuilder):
885
423
"""This commitbuilder actually records the root id"""
887
425
# the root entry gets versioned properly by this builder.
888
426
_versioned_root = True
1032
def add_fallback_repository(self, repository):
1033
"""Add a repository to use for looking up data not held locally.
1035
:param repository: A repository.
1037
if not self._format.supports_external_lookups:
1038
raise errors.UnstackableRepositoryFormat(self._format, self.base)
1039
if self.is_locked():
1040
# This repository will call fallback.unlock() when we transition to
1041
# the unlocked state, so we make sure to increment the lock count
1042
repository.lock_read()
1043
self._check_fallback_repository(repository)
1044
self._fallback_repositories.append(repository)
1045
self.texts.add_fallback_versioned_files(repository.texts)
1046
self.inventories.add_fallback_versioned_files(repository.inventories)
1047
self.revisions.add_fallback_versioned_files(repository.revisions)
1048
self.signatures.add_fallback_versioned_files(repository.signatures)
1049
if self.chk_bytes is not None:
1050
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
1052
def _check_fallback_repository(self, repository):
1053
"""Check that this repository can fallback to repository safely.
1055
Raise an error if not.
1057
:param repository: A repository to fallback to.
1059
return InterRepository._assert_same_model(self, repository)
1061
489
def add_inventory(self, revision_id, inv, parents):
1062
490
"""Add the inventory inv to the repository as revision_id.
1064
492
:param parents: The revision ids of the parents that revision_id
1065
493
is known to have and are in the repository already.
1067
:returns: The validator(which is a sha1 digest, though what is sha'd is
1068
repository format specific) of the serialized inventory.
495
returns the sha1 of the serialized inventory.
1070
if not self.is_in_write_group():
1071
raise AssertionError("%r not in write group" % (self,))
1072
497
_mod_revision.check_not_reserved_id(revision_id)
1073
if not (inv.revision_id is None or inv.revision_id == revision_id):
1074
raise AssertionError(
1075
"Mismatch between inventory revision"
1076
" id and insertion revid (%r, %r)"
1077
% (inv.revision_id, revision_id))
1078
if inv.root is None:
1079
raise errors.RootMissing()
1080
return self._add_inventory_checked(revision_id, inv, parents)
1082
def _add_inventory_checked(self, revision_id, inv, parents):
1083
"""Add inv to the repository after checking the inputs.
1085
This function can be overridden to allow different inventory styles.
1087
:seealso: add_inventory, for the contract.
1089
inv_lines = self._serializer.write_inventory_to_lines(inv)
1090
return self._inventory_add_lines(revision_id, parents,
498
assert inv.revision_id is None or inv.revision_id == revision_id, \
499
"Mismatch between inventory revision" \
500
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
501
assert inv.root is not None
502
inv_lines = self._serialise_inventory_to_lines(inv)
503
inv_vf = self.get_inventory_weave()
504
return self._inventory_add_lines(inv_vf, revision_id, parents,
1091
505
inv_lines, check_content=False)
1093
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1094
parents, basis_inv=None, propagate_caches=False):
1095
"""Add a new inventory expressed as a delta against another revision.
1097
See the inventory developers documentation for the theory behind
1100
:param basis_revision_id: The inventory id the delta was created
1101
against. (This does not have to be a direct parent.)
1102
:param delta: The inventory delta (see Inventory.apply_delta for
1104
:param new_revision_id: The revision id that the inventory is being
1106
:param parents: The revision ids of the parents that revision_id is
1107
known to have and are in the repository already. These are supplied
1108
for repositories that depend on the inventory graph for revision
1109
graph access, as well as for those that pun ancestry with delta
1111
:param basis_inv: The basis inventory if it is already known,
1113
:param propagate_caches: If True, the caches for this inventory are
1114
copied to and updated for the result if possible.
1116
:returns: (validator, new_inv)
1117
The validator(which is a sha1 digest, though what is sha'd is
1118
repository format specific) of the serialized inventory, and the
1119
resulting inventory.
1121
if not self.is_in_write_group():
1122
raise AssertionError("%r not in write group" % (self,))
1123
_mod_revision.check_not_reserved_id(new_revision_id)
1124
basis_tree = self.revision_tree(basis_revision_id)
1125
basis_tree.lock_read()
1127
# Note that this mutates the inventory of basis_tree, which not all
1128
# inventory implementations may support: A better idiom would be to
1129
# return a new inventory, but as there is no revision tree cache in
1130
# repository this is safe for now - RBC 20081013
1131
if basis_inv is None:
1132
basis_inv = basis_tree.inventory
1133
basis_inv.apply_delta(delta)
1134
basis_inv.revision_id = new_revision_id
1135
return (self.add_inventory(new_revision_id, basis_inv, parents),
1140
def _inventory_add_lines(self, revision_id, parents, lines,
507
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
1141
508
check_content=True):
1142
509
"""Store lines in inv_vf and return the sha1 of the inventory."""
1143
parents = [(parent,) for parent in parents]
1144
result = self.inventories.add_lines((revision_id,), parents, lines,
511
for parent in parents:
513
final_parents.append(parent)
514
return inv_vf.add_lines(revision_id, final_parents, lines,
1145
515
check_content=check_content)[0]
1146
self.inventories._access.flush()
1149
518
def add_revision(self, revision_id, rev, inv=None, config=None):
1150
519
"""Add rev to the revision store as revision_id.
1223
584
Returns a set of the present revisions.
1226
graph = self.get_graph()
1227
parent_map = graph.get_parent_map(revision_ids)
1228
# The old API returned a list, should this actually be a set?
1229
return parent_map.keys()
1231
def _check_inventories(self, checker):
1232
"""Check the inventories found from the revision scan.
1234
This is responsible for verifying the sha1 of inventories and
1235
creating a pending_keys set that covers data referenced by inventories.
1237
bar = ui.ui_factory.nested_progress_bar()
1239
self._do_check_inventories(checker, bar)
1243
def _do_check_inventories(self, checker, bar):
1244
"""Helper for _check_inventories."""
1246
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1247
kinds = ['chk_bytes', 'texts']
1248
count = len(checker.pending_keys)
1249
bar.update("inventories", 0, 2)
1250
current_keys = checker.pending_keys
1251
checker.pending_keys = {}
1252
# Accumulate current checks.
1253
for key in current_keys:
1254
if key[0] != 'inventories' and key[0] not in kinds:
1255
checker._report_items.append('unknown key type %r' % (key,))
1256
keys[key[0]].add(key[1:])
1257
if keys['inventories']:
1258
# NB: output order *should* be roughly sorted - topo or
1259
# inverse topo depending on repository - either way decent
1260
# to just delta against. However, pre-CHK formats didn't
1261
# try to optimise inventory layout on disk. As such the
1262
# pre-CHK code path does not use inventory deltas.
1264
for record in self.inventories.check(keys=keys['inventories']):
1265
if record.storage_kind == 'absent':
1266
checker._report_items.append(
1267
'Missing inventory {%s}' % (record.key,))
1269
last_object = self._check_record('inventories', record,
1270
checker, last_object,
1271
current_keys[('inventories',) + record.key])
1272
del keys['inventories']
1275
bar.update("texts", 1)
1276
while (checker.pending_keys or keys['chk_bytes']
1278
# Something to check.
1279
current_keys = checker.pending_keys
1280
checker.pending_keys = {}
1281
# Accumulate current checks.
1282
for key in current_keys:
1283
if key[0] not in kinds:
1284
checker._report_items.append('unknown key type %r' % (key,))
1285
keys[key[0]].add(key[1:])
1286
# Check the outermost kind only - inventories || chk_bytes || texts
1290
for record in getattr(self, kind).check(keys=keys[kind]):
1291
if record.storage_kind == 'absent':
1292
checker._report_items.append(
1293
'Missing %s {%s}' % (kind, record.key,))
1295
last_object = self._check_record(kind, record,
1296
checker, last_object, current_keys[(kind,) + record.key])
1300
def _check_record(self, kind, record, checker, last_object, item_data):
1301
"""Check a single text from this repository."""
1302
if kind == 'inventories':
1303
rev_id = record.key[0]
1304
inv = self._deserialise_inventory(rev_id,
1305
record.get_bytes_as('fulltext'))
1306
if last_object is not None:
1307
delta = inv._make_delta(last_object)
1308
for old_path, path, file_id, ie in delta:
1311
ie.check(checker, rev_id, inv)
1313
for path, ie in inv.iter_entries():
1314
ie.check(checker, rev_id, inv)
1315
if self._format.fast_deltas:
1317
elif kind == 'chk_bytes':
1318
# No code written to check chk_bytes for this repo format.
1319
checker._report_items.append(
1320
'unsupported key type chk_bytes for %s' % (record.key,))
1321
elif kind == 'texts':
1322
self._check_text(record, checker, item_data)
1324
checker._report_items.append(
1325
'unknown key type %s for %s' % (kind, record.key))
1327
def _check_text(self, record, checker, item_data):
1328
"""Check a single text."""
1329
# Check it is extractable.
1330
# TODO: check length.
1331
if record.storage_kind == 'chunked':
1332
chunks = record.get_bytes_as(record.storage_kind)
1333
sha1 = osutils.sha_strings(chunks)
1334
length = sum(map(len, chunks))
1336
content = record.get_bytes_as('fulltext')
1337
sha1 = osutils.sha_string(content)
1338
length = len(content)
1339
if item_data and sha1 != item_data[1]:
1340
checker._report_items.append(
1341
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1342
(record.key, sha1, item_data[1], item_data[2]))
587
for id in revision_ids:
588
if self.has_revision(id):
1345
593
def create(a_bzrdir):
1346
594
"""Construct the current default format repository in a_bzrdir."""
1347
595
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
1349
def __init__(self, _format, a_bzrdir, control_files):
597
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1350
598
"""instantiate a Repository.
1352
600
:param _format: The format of the repository on disk.
1353
601
:param a_bzrdir: The BzrDir of the repository.
603
In the future we will have a single api for all stores for
604
getting file texts, inventories and revisions, then
605
this construct will accept instances of those things.
1355
# In the future we will have a single api for all stores for
1356
# getting file texts, inventories and revisions, then
1357
# this construct will accept instances of those things.
1358
607
super(Repository, self).__init__()
1359
608
self._format = _format
1360
609
# the following are part of the public API for Repository:
1361
610
self.bzrdir = a_bzrdir
1362
611
self.control_files = control_files
1363
self._transport = control_files._transport
1364
self.base = self._transport.base
612
self._revision_store = _revision_store
613
# backwards compatibility
614
self.weave_store = text_store
1366
616
self._reconcile_does_inventory_gc = True
1367
617
self._reconcile_fixes_text_parents = False
1368
self._reconcile_backsup_inventory = True
618
# not right yet - should be more semantically clear ?
620
self.control_store = control_store
621
self.control_weaves = control_store
622
# TODO: make sure to construct the right store classes, etc, depending
623
# on whether escaping is required.
624
self._warn_if_deprecated()
1369
625
self._write_group = None
1370
# Additional places to query for data.
1371
self._fallback_repositories = []
1372
# An InventoryEntry cache, used during deserialization
1373
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1374
# Is it safe to return inventory entries directly from the entry cache,
1375
# rather copying them?
1376
self._safe_to_return_from_cache = False
1379
def user_transport(self):
1380
return self.bzrdir.user_transport
1383
def control_transport(self):
1384
return self._transport
626
self.base = control_files._transport.base
1386
628
def __repr__(self):
1387
if self._fallback_repositories:
1388
return '%s(%r, fallback_repositories=%r)' % (
1389
self.__class__.__name__,
1391
self._fallback_repositories)
1393
return '%s(%r)' % (self.__class__.__name__,
1396
def _has_same_fallbacks(self, other_repo):
1397
"""Returns true if the repositories have the same fallbacks."""
1398
my_fb = self._fallback_repositories
1399
other_fb = other_repo._fallback_repositories
1400
if len(my_fb) != len(other_fb):
1402
for f, g in zip(my_fb, other_fb):
1403
if not f.has_same_location(g):
629
return '%s(%r)' % (self.__class__.__name__,
1407
632
def has_same_location(self, other):
1408
633
"""Returns a boolean indicating if this repository is at the same
1634
816
"""Commit the contents accrued within the current write group.
1636
818
:seealso: start_write_group.
1638
:return: it may return an opaque hint that can be passed to 'pack'.
1640
820
if self._write_group is not self.get_transaction():
1641
821
# has an unlock or relock occured ?
1642
822
raise errors.BzrError('mismatched lock context %r and '
1643
823
'write group %r.' %
1644
824
(self.get_transaction(), self._write_group))
1645
result = self._commit_write_group()
825
self._commit_write_group()
1646
826
self._write_group = None
1649
828
def _commit_write_group(self):
1650
829
"""Template method for per-repository write group cleanup.
1652
This is called before the write group is considered to be
831
This is called before the write group is considered to be
1653
832
finished and should ensure that all data handed to the repository
1654
for writing during the write group is safely committed (to the
833
for writing during the write group is safely committed (to the
1655
834
extent possible considering file system caching etc).
1658
def suspend_write_group(self):
1659
raise errors.UnsuspendableWriteGroup(self)
1661
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1662
"""Return the keys of missing inventory parents for revisions added in
1665
A revision is not complete if the inventory delta for that revision
1666
cannot be calculated. Therefore if the parent inventories of a
1667
revision are not present, the revision is incomplete, and e.g. cannot
1668
be streamed by a smart server. This method finds missing inventory
1669
parents for revisions added in this write group.
1671
if not self._format.supports_external_lookups:
1672
# This is only an issue for stacked repositories
1674
if not self.is_in_write_group():
1675
raise AssertionError('not in a write group')
1677
# XXX: We assume that every added revision already has its
1678
# corresponding inventory, so we only check for parent inventories that
1679
# might be missing, rather than all inventories.
1680
parents = set(self.revisions._index.get_missing_parents())
1681
parents.discard(_mod_revision.NULL_REVISION)
1682
unstacked_inventories = self.inventories._index
1683
present_inventories = unstacked_inventories.get_parent_map(
1684
key[-1:] for key in parents)
1685
parents.difference_update(present_inventories)
1686
if len(parents) == 0:
1687
# No missing parent inventories.
1689
if not check_for_missing_texts:
1690
return set(('inventories', rev_id) for (rev_id,) in parents)
1691
# Ok, now we have a list of missing inventories. But these only matter
1692
# if the inventories that reference them are missing some texts they
1693
# appear to introduce.
1694
# XXX: Texts referenced by all added inventories need to be present,
1695
# but at the moment we're only checking for texts referenced by
1696
# inventories at the graph's edge.
1697
key_deps = self.revisions._index._key_dependencies
1698
key_deps.satisfy_refs_for_keys(present_inventories)
1699
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1700
file_ids = self.fileids_altered_by_revision_ids(referrers)
1701
missing_texts = set()
1702
for file_id, version_ids in file_ids.iteritems():
1703
missing_texts.update(
1704
(file_id, version_id) for version_id in version_ids)
1705
present_texts = self.texts.get_parent_map(missing_texts)
1706
missing_texts.difference_update(present_texts)
1707
if not missing_texts:
1708
# No texts are missing, so all revisions and their deltas are
1711
# Alternatively the text versions could be returned as the missing
1712
# keys, but this is likely to be less data.
1713
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1716
def refresh_data(self):
1717
"""Re-read any data needed to synchronise with disk.
1719
This method is intended to be called after another repository instance
1720
(such as one used by a smart server) has inserted data into the
1721
repository. On all repositories this will work outside of write groups.
1722
Some repository formats (pack and newer for bzrlib native formats)
1723
support refresh_data inside write groups. If called inside a write
1724
group on a repository that does not support refreshing in a write group
1725
IsInWriteGroupError will be raised.
1727
self._refresh_data()
1729
def resume_write_group(self, tokens):
1730
if not self.is_write_locked():
1731
raise errors.NotWriteLocked(self)
1732
if self._write_group:
1733
raise errors.BzrError('already in a write group')
1734
self._resume_write_group(tokens)
1735
# so we can detect unlock/relock - the write group is now entered.
1736
self._write_group = self.get_transaction()
1738
def _resume_write_group(self, tokens):
1739
raise errors.UnsuspendableWriteGroup(self)
1741
def fetch(self, source, revision_id=None, find_ghosts=False,
837
def fetch(self, source, revision_id=None, pb=None):
1743
838
"""Fetch the content required to construct revision_id from source.
1745
If revision_id is None and fetch_spec is None, then all content is
1748
fetch() may not be used when the repository is in a write group -
1749
either finish the current write group before using fetch, or use
1750
fetch before starting the write group.
1752
:param find_ghosts: Find and copy revisions in the source that are
1753
ghosts in the target (and not reachable directly by walking out to
1754
the first-present revision in target from revision_id).
1755
:param revision_id: If specified, all the content needed for this
1756
revision ID will be copied to the target. Fetch will determine for
1757
itself which content needs to be copied.
1758
:param fetch_spec: If specified, a SearchResult or
1759
PendingAncestryResult that describes which revisions to copy. This
1760
allows copying multiple heads at once. Mutually exclusive with
840
If revision_id is None all content is copied.
1763
if fetch_spec is not None and revision_id is not None:
1764
raise AssertionError(
1765
"fetch_spec and revision_id are mutually exclusive.")
1766
if self.is_in_write_group():
1767
raise errors.InternalBzrError(
1768
"May not fetch while in a write group.")
1769
842
# fast path same-url fetch operations
1770
# TODO: lift out to somewhere common with RemoteRepository
1771
# <https://bugs.launchpad.net/bzr/+bug/401646>
1772
if (self.has_same_location(source)
1773
and fetch_spec is None
1774
and self._has_same_fallbacks(source)):
843
if self.has_same_location(source):
1775
844
# check that last_revision is in 'from' and then return a
1777
846
if (revision_id is not None and
1940
977
@needs_read_lock
1941
978
def get_revisions(self, revision_ids):
1942
"""Get many revisions at once.
1944
Repositories that need to check data on every revision read should
1945
subclass this method.
979
"""Get many revisions at once."""
1947
980
return self._get_revisions(revision_ids)
1949
982
@needs_read_lock
1950
983
def _get_revisions(self, revision_ids):
1951
984
"""Core work logic to get many revisions without sanity checks."""
1953
for revid, rev in self._iter_revisions(revision_ids):
1955
raise errors.NoSuchRevision(self, revid)
1957
return [revs[revid] for revid in revision_ids]
1959
def _iter_revisions(self, revision_ids):
1960
"""Iterate over revision objects.
1962
:param revision_ids: An iterable of revisions to examine. None may be
1963
passed to request all revisions known to the repository. Note that
1964
not all repositories can find unreferenced revisions; for those
1965
repositories only referenced ones will be returned.
1966
:return: An iterator of (revid, revision) tuples. Absent revisions (
1967
those asked for but not available) are returned as (revid, None).
1969
if revision_ids is None:
1970
revision_ids = self.all_revision_ids()
1972
for rev_id in revision_ids:
1973
if not rev_id or not isinstance(rev_id, basestring):
1974
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1975
keys = [(key,) for key in revision_ids]
1976
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1977
for record in stream:
1978
revid = record.key[0]
1979
if record.storage_kind == 'absent':
1982
text = record.get_bytes_as('fulltext')
1983
rev = self._serializer.read_revision_from_string(text)
1986
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
985
for rev_id in revision_ids:
986
if not rev_id or not isinstance(rev_id, basestring):
987
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
988
revs = self._revision_store.get_revisions(revision_ids,
989
self.get_transaction())
991
assert not isinstance(rev.revision_id, unicode)
992
for parent_id in rev.parent_ids:
993
assert not isinstance(parent_id, unicode)
997
def get_revision_xml(self, revision_id):
998
# TODO: jam 20070210 This shouldn't be necessary since get_revision
999
# would have already do it.
1000
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1001
rev = self.get_revision(revision_id)
1002
rev_tmp = StringIO()
1003
# the current serializer..
1004
self._revision_store._serializer.write_revision(rev, rev_tmp)
1006
return rev_tmp.getvalue()
1009
def get_deltas_for_revisions(self, revisions):
1987
1010
"""Produce a generator of revision deltas.
1989
1012
Note that the input is a sequence of REVISIONS, not revision_ids.
1990
1013
Trees will be held in memory until the generator exits.
1991
1014
Each delta is relative to the revision's lefthand predecessor.
1993
:param specific_fileids: if not None, the result is filtered
1994
so that only those file-ids, their parents and their
1995
children are included.
1997
# Get the revision-ids of interest
1998
1016
required_trees = set()
1999
1017
for revision in revisions:
2000
1018
required_trees.add(revision.revision_id)
2001
1019
required_trees.update(revision.parent_ids[:1])
2003
# Get the matching filtered trees. Note that it's more
2004
# efficient to pass filtered trees to changes_from() rather
2005
# than doing the filtering afterwards. changes_from() could
2006
# arguably do the filtering itself but it's path-based, not
2007
# file-id based, so filtering before or afterwards is
2009
if specific_fileids is None:
2010
trees = dict((t.get_revision_id(), t) for
2011
t in self.revision_trees(required_trees))
2013
trees = dict((t.get_revision_id(), t) for
2014
t in self._filtered_revision_trees(required_trees,
2017
# Calculate the deltas
1020
trees = dict((t.get_revision_id(), t) for
1021
t in self.revision_trees(required_trees))
2018
1022
for revision in revisions:
2019
1023
if not revision.parent_ids:
2020
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1024
old_tree = self.revision_tree(None)
2022
1026
old_tree = trees[revision.parent_ids[0]]
2023
1027
yield trees[revision.revision_id].changes_from(old_tree)
2025
1029
@needs_read_lock
2026
def get_revision_delta(self, revision_id, specific_fileids=None):
1030
def get_revision_delta(self, revision_id):
2027
1031
"""Return the delta for one revision.
2029
1033
The delta is relative to the left-hand predecessor of the
2032
:param specific_fileids: if not None, the result is filtered
2033
so that only those file-ids, their parents and their
2034
children are included.
2036
1036
r = self.get_revision(revision_id)
2037
return list(self.get_deltas_for_revisions([r],
2038
specific_fileids=specific_fileids))[0]
1037
return list(self.get_deltas_for_revisions([r]))[0]
2040
1039
@needs_write_lock
2041
1040
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
2042
1041
signature = gpg_strategy.sign(plaintext)
2043
self.add_signature_text(revision_id, signature)
2046
def add_signature_text(self, revision_id, signature):
2047
self.signatures.add_lines((revision_id,), (),
2048
osutils.split_lines(signature))
2050
def find_text_key_references(self):
2051
"""Find the text key references within the repository.
2053
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2054
to whether they were referred to by the inventory of the
2055
revision_id that they contain. The inventory texts from all present
2056
revision ids are assessed to generate this report.
2058
revision_keys = self.revisions.keys()
2059
w = self.inventories
2060
pb = ui.ui_factory.nested_progress_bar()
2062
return self._serializer._find_text_key_references(
2063
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
2067
def _inventory_xml_lines_for_keys(self, keys):
2068
"""Get a line iterator of the sort needed for findind references.
2070
Not relevant for non-xml inventory repositories.
2072
Ghosts in revision_keys are ignored.
2074
:param revision_keys: The revision keys for the inventories to inspect.
2075
:return: An iterator over (inventory line, revid) for the fulltexts of
2076
all of the xml inventories specified by revision_keys.
2078
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2079
for record in stream:
2080
if record.storage_kind != 'absent':
2081
chunks = record.get_bytes_as('chunked')
2082
revid = record.key[-1]
2083
lines = osutils.chunks_to_lines(chunks)
2087
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2089
"""Helper routine for fileids_altered_by_revision_ids.
2091
This performs the translation of xml lines to revision ids.
2093
:param line_iterator: An iterator of lines, origin_version_id
2094
:param revision_keys: The revision ids to filter for. This should be a
2095
set or other type which supports efficient __contains__ lookups, as
2096
the revision key from each parsed line will be looked up in the
2097
revision_keys filter.
2098
:return: a dictionary mapping altered file-ids to an iterable of
2099
revision_ids. Each altered file-ids has the exact revision_ids that
2100
altered it listed explicitly.
2102
seen = set(self._serializer._find_text_key_references(
2103
line_iterator).iterkeys())
2104
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2105
parent_seen = set(self._serializer._find_text_key_references(
2106
self._inventory_xml_lines_for_keys(parent_keys)))
2107
new_keys = seen - parent_seen
2109
setdefault = result.setdefault
2110
for key in new_keys:
2111
setdefault(key[0], set()).add(key[-1])
2114
def _find_parent_ids_of_revisions(self, revision_ids):
2115
"""Find all parent ids that are mentioned in the revision graph.
2117
:return: set of revisions that are parents of revision_ids which are
2118
not part of revision_ids themselves
2120
parent_map = self.get_parent_map(revision_ids)
2122
map(parent_ids.update, parent_map.itervalues())
2123
parent_ids.difference_update(revision_ids)
2124
parent_ids.discard(_mod_revision.NULL_REVISION)
2127
def _find_parent_keys_of_revisions(self, revision_keys):
2128
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2130
:param revision_keys: An iterable of revision_keys.
2131
:return: The parents of all revision_keys that are not already in
2134
parent_map = self.revisions.get_parent_map(revision_keys)
2136
map(parent_keys.update, parent_map.itervalues())
2137
parent_keys.difference_update(revision_keys)
2138
parent_keys.discard(_mod_revision.NULL_REVISION)
2141
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1042
self._revision_store.add_revision_signature_text(revision_id,
1044
self.get_transaction())
1046
def fileids_altered_by_revision_ids(self, revision_ids):
2142
1047
"""Find the file ids and versions affected by revisions.
2144
1049
:param revisions: an iterable containing revision ids.
2145
:param _inv_weave: The inventory weave from this repository or None.
2146
If None, the inventory weave will be opened automatically.
2147
1050
:return: a dictionary mapping altered file-ids to an iterable of
2148
1051
revision_ids. Each altered file-ids has the exact revision_ids that
2149
1052
altered it listed explicitly.
2151
selected_keys = set((revid,) for revid in revision_ids)
2152
w = _inv_weave or self.inventories
2153
return self._find_file_ids_from_xml_inventory_lines(
2154
w.iter_lines_added_or_present_in_keys(
2155
selected_keys, pb=None),
1054
assert self._serializer.support_altered_by_hack, \
1055
("fileids_altered_by_revision_ids only supported for branches "
1056
"which store inventory as unnested xml, not on %r" % self)
1057
selected_revision_ids = set(revision_ids)
1058
w = self.get_inventory_weave()
1061
# this code needs to read every new line in every inventory for the
1062
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1063
# not present in one of those inventories is unnecessary but not
1064
# harmful because we are filtering by the revision id marker in the
1065
# inventory lines : we only select file ids altered in one of those
1066
# revisions. We don't need to see all lines in the inventory because
1067
# only those added in an inventory in rev X can contain a revision=X
1069
unescape_revid_cache = {}
1070
unescape_fileid_cache = {}
1072
# jam 20061218 In a big fetch, this handles hundreds of thousands
1073
# of lines, so it has had a lot of inlining and optimizing done.
1074
# Sorry that it is a little bit messy.
1075
# Move several functions to be local variables, since this is a long
1077
search = self._file_ids_altered_regex.search
1078
unescape = _unescape_xml
1079
setdefault = result.setdefault
1080
pb = ui.ui_factory.nested_progress_bar()
1082
for line in w.iter_lines_added_or_present_in_versions(
1083
selected_revision_ids, pb=pb):
1084
match = search(line)
1087
# One call to match.group() returning multiple items is quite a
1088
# bit faster than 2 calls to match.group() each returning 1
1089
file_id, revision_id = match.group('file_id', 'revision_id')
1091
# Inlining the cache lookups helps a lot when you make 170,000
1092
# lines and 350k ids, versus 8.4 unique ids.
1093
# Using a cache helps in 2 ways:
1094
# 1) Avoids unnecessary decoding calls
1095
# 2) Re-uses cached strings, which helps in future set and
1097
# (2) is enough that removing encoding entirely along with
1098
# the cache (so we are using plain strings) results in no
1099
# performance improvement.
1101
revision_id = unescape_revid_cache[revision_id]
1103
unescaped = unescape(revision_id)
1104
unescape_revid_cache[revision_id] = unescaped
1105
revision_id = unescaped
1107
if revision_id in selected_revision_ids:
1109
file_id = unescape_fileid_cache[file_id]
1111
unescaped = unescape(file_id)
1112
unescape_fileid_cache[file_id] = unescaped
1114
setdefault(file_id, set()).add(revision_id)
2158
1119
def iter_files_bytes(self, desired_files):
2159
1120
"""Iterate through file versions.
2300
1154
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2301
1155
'revisions'. file-id is None unless knit-kind is 'file'.
2303
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2306
for result in self._find_non_file_keys_to_fetch(revision_ids):
2309
def _find_file_keys_to_fetch(self, revision_ids, pb):
2310
1157
# XXX: it's a bit weird to control the inventory weave caching in this
2311
1158
# generator. Ideally the caching would be done in fetch.py I think. Or
2312
1159
# maybe this generator should explicitly have the contract that it
2313
1160
# should not be iterated until the previously yielded item has been
2315
inv_w = self.inventories
1163
inv_w = self.get_inventory_weave()
1164
inv_w.enable_cache()
2317
1166
# file ids that changed
2318
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1167
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
2320
1169
num_file_ids = len(file_ids)
2321
1170
for file_id, altered_versions in file_ids.iteritems():
2323
pb.update("Fetch texts", count, num_file_ids)
1171
if _files_pb is not None:
1172
_files_pb.update("fetch texts", count, num_file_ids)
2325
1174
yield ("file", file_id, altered_versions)
1175
# We're done with the files_pb. Note that it finished by the caller,
1176
# just as it was created by the caller.
2327
def _find_non_file_keys_to_fetch(self, revision_ids):
2329
1180
yield ("inventory", None, revision_ids)
2332
# XXX: Note ATM no callers actually pay attention to this return
2333
# instead they just use the list of revision ids and ignore
2334
# missing sigs. Consider removing this work entirely
2335
revisions_with_signatures = set(self.signatures.get_parent_map(
2336
[(r,) for r in revision_ids]))
2337
revisions_with_signatures = set(
2338
[r for (r,) in revisions_with_signatures])
2339
revisions_with_signatures.intersection_update(revision_ids)
1184
revisions_with_signatures = set()
1185
for rev_id in revision_ids:
1187
self.get_signature_text(rev_id)
1188
except errors.NoSuchRevision:
1192
revisions_with_signatures.add(rev_id)
2340
1194
yield ("signatures", None, revisions_with_signatures)
2343
1197
yield ("revisions", None, revision_ids)
2345
1199
@needs_read_lock
1200
def get_inventory_weave(self):
1201
return self.control_weaves.get_weave('inventory',
1202
self.get_transaction())
2346
1205
def get_inventory(self, revision_id):
2347
"""Get Inventory object by revision id."""
2348
return self.iter_inventories([revision_id]).next()
2350
def iter_inventories(self, revision_ids, ordering=None):
2351
"""Get many inventories by revision_ids.
2353
This will buffer some or all of the texts used in constructing the
2354
inventories in memory, but will only parse a single inventory at a
2357
:param revision_ids: The expected revision ids of the inventories.
2358
:param ordering: optional ordering, e.g. 'topological'. If not
2359
specified, the order of revision_ids will be preserved (by
2360
buffering if necessary).
2361
:return: An iterator of inventories.
2363
if ((None in revision_ids)
2364
or (_mod_revision.NULL_REVISION in revision_ids)):
2365
raise ValueError('cannot get null revision inventory')
2366
return self._iter_inventories(revision_ids, ordering)
2368
def _iter_inventories(self, revision_ids, ordering):
2369
"""single-document based inventory iteration."""
2370
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2371
for text, revision_id in inv_xmls:
2372
yield self._deserialise_inventory(revision_id, text)
2374
def _iter_inventory_xmls(self, revision_ids, ordering):
2375
if ordering is None:
2376
order_as_requested = True
2377
ordering = 'unordered'
2379
order_as_requested = False
2380
keys = [(revision_id,) for revision_id in revision_ids]
2383
if order_as_requested:
2384
key_iter = iter(keys)
2385
next_key = key_iter.next()
2386
stream = self.inventories.get_record_stream(keys, ordering, True)
2388
for record in stream:
2389
if record.storage_kind != 'absent':
2390
chunks = record.get_bytes_as('chunked')
2391
if order_as_requested:
2392
text_chunks[record.key] = chunks
2394
yield ''.join(chunks), record.key[-1]
2396
raise errors.NoSuchRevision(self, record.key)
2397
if order_as_requested:
2398
# Yield as many results as we can while preserving order.
2399
while next_key in text_chunks:
2400
chunks = text_chunks.pop(next_key)
2401
yield ''.join(chunks), next_key[-1]
2403
next_key = key_iter.next()
2404
except StopIteration:
2405
# We still want to fully consume the get_record_stream,
2406
# just in case it is not actually finished at this point
2410
def _deserialise_inventory(self, revision_id, xml):
2411
"""Transform the xml into an inventory object.
1206
"""Get Inventory object by hash."""
1207
return self.deserialise_inventory(
1208
revision_id, self.get_inventory_xml(revision_id))
1210
def deserialise_inventory(self, revision_id, xml):
1211
"""Transform the xml into an inventory object.
2413
1213
:param revision_id: The expected revision id of the inventory.
2414
1214
:param xml: A serialised inventory.
2416
result = self._serializer.read_inventory_from_string(xml, revision_id,
2417
entry_cache=self._inventory_entry_cache,
2418
return_from_cache=self._safe_to_return_from_cache)
2419
if result.revision_id != revision_id:
2420
raise AssertionError('revision id mismatch %s != %s' % (
2421
result.revision_id, revision_id))
1216
return self._serializer.read_inventory_from_string(xml, revision_id)
1218
def serialise_inventory(self, inv):
1219
return self._serializer.write_inventory_to_string(inv)
1221
def _serialise_inventory_to_lines(self, inv):
1222
return self._serializer.write_inventory_to_lines(inv)
2424
1224
def get_serializer_format(self):
2425
1225
return self._serializer.format_num
2427
1227
@needs_read_lock
2428
def _get_inventory_xml(self, revision_id):
2429
"""Get serialized inventory as a string."""
2430
texts = self._iter_inventory_xmls([revision_id], 'unordered')
1228
def get_inventory_xml(self, revision_id):
1229
"""Get inventory XML as a file object."""
2432
text, revision_id = texts.next()
2433
except StopIteration:
1231
assert isinstance(revision_id, str), type(revision_id)
1232
iw = self.get_inventory_weave()
1233
return iw.get_text(revision_id)
2434
1235
raise errors.HistoryMissing(self, 'inventory', revision_id)
2437
def get_rev_id_for_revno(self, revno, known_pair):
2438
"""Return the revision id of a revno, given a later (revno, revid)
2439
pair in the same history.
2441
:return: if found (True, revid). If the available history ran out
2442
before reaching the revno, then this returns
2443
(False, (closest_revno, closest_revid)).
2445
known_revno, known_revid = known_pair
2446
partial_history = [known_revid]
2447
distance_from_known = known_revno - revno
2448
if distance_from_known < 0:
2450
'requested revno (%d) is later than given known revno (%d)'
2451
% (revno, known_revno))
2454
self, partial_history, stop_index=distance_from_known)
2455
except errors.RevisionNotPresent, err:
2456
if err.revision_id == known_revid:
2457
# The start revision (known_revid) wasn't found.
2459
# This is a stacked repository with no fallbacks, or a there's a
2460
# left-hand ghost. Either way, even though the revision named in
2461
# the error isn't in this repo, we know it's the next step in this
2462
# left-hand history.
2463
partial_history.append(err.revision_id)
2464
if len(partial_history) <= distance_from_known:
2465
# Didn't find enough history to get a revid for the revno.
2466
earliest_revno = known_revno - len(partial_history) + 1
2467
return (False, (earliest_revno, partial_history[-1]))
2468
if len(partial_history) - 1 > distance_from_known:
2469
raise AssertionError('_iter_for_revno returned too much history')
2470
return (True, partial_history[-1])
1238
def get_inventory_sha1(self, revision_id):
1239
"""Return the sha1 hash of the inventory entry
1241
return self.get_revision(revision_id).inventory_sha1
1244
def get_revision_graph(self, revision_id=None):
1245
"""Return a dictionary containing the revision graph.
1247
NB: This method should not be used as it accesses the entire graph all
1248
at once, which is much more data than most operations should require.
1250
:param revision_id: The revision_id to get a graph from. If None, then
1251
the entire revision graph is returned. This is a deprecated mode of
1252
operation and will be removed in the future.
1253
:return: a dictionary of revision_id->revision_parents_list.
1255
raise NotImplementedError(self.get_revision_graph)
1258
def get_revision_graph_with_ghosts(self, revision_ids=None):
1259
"""Return a graph of the revisions with ghosts marked as applicable.
1261
:param revision_ids: an iterable of revisions to graph or None for all.
1262
:return: a Graph object with the graph reachable from revision_ids.
1264
if 'evil' in debug.debug_flags:
1266
"get_revision_graph_with_ghosts scales with size of history.")
1267
result = deprecated_graph.Graph()
1268
if not revision_ids:
1269
pending = set(self.all_revision_ids())
1272
pending = set(revision_ids)
1273
# special case NULL_REVISION
1274
if _mod_revision.NULL_REVISION in pending:
1275
pending.remove(_mod_revision.NULL_REVISION)
1276
required = set(pending)
1279
revision_id = pending.pop()
1281
rev = self.get_revision(revision_id)
1282
except errors.NoSuchRevision:
1283
if revision_id in required:
1286
result.add_ghost(revision_id)
1288
for parent_id in rev.parent_ids:
1289
# is this queued or done ?
1290
if (parent_id not in pending and
1291
parent_id not in done):
1293
pending.add(parent_id)
1294
result.add_node(revision_id, rev.parent_ids)
1295
done.add(revision_id)
1298
def _get_history_vf(self):
1299
"""Get a versionedfile whose history graph reflects all revisions.
1301
For weave repositories, this is the inventory weave.
1303
return self.get_inventory_weave()
2472
1305
def iter_reverse_revision_history(self, revision_id):
2473
1306
"""Iterate backwards through revision ids in the lefthand history
2498
1356
for repositories to maintain loaded indices across multiple locks
2499
1357
by checking inside their implementation of this method to see
2500
1358
whether their indices are still valid. This depends of course on
2501
the disk format being validatable in this manner. This method is
2502
also called by the refresh_data() public interface to cause a refresh
2503
to occur while in a write lock so that data inserted by a smart server
2504
push operation is visible on the client's instance of the physical
1359
the disk format being validatable in this manner.
2508
1362
@needs_read_lock
2509
1363
def revision_tree(self, revision_id):
2510
1364
"""Return Tree for a revision on this branch.
2512
`revision_id` may be NULL_REVISION for the empty tree revision.
1366
`revision_id` may be None for the empty tree revision.
2514
revision_id = _mod_revision.ensure_null(revision_id)
2515
1368
# TODO: refactor this to use an existing revision object
2516
1369
# so we don't need to read it in twice.
2517
if revision_id == _mod_revision.NULL_REVISION:
2518
return InventoryRevisionTree(self,
2519
Inventory(root_id=None), _mod_revision.NULL_REVISION)
1370
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1371
return RevisionTree(self, Inventory(root_id=None),
1372
_mod_revision.NULL_REVISION)
2521
inv = self.get_inventory(revision_id)
2522
return InventoryRevisionTree(self, inv, revision_id)
1374
inv = self.get_revision_inventory(revision_id)
1375
return RevisionTree(self, inv, revision_id)
2524
1378
def revision_trees(self, revision_ids):
2525
"""Return Trees for revisions in this repository.
2527
:param revision_ids: a sequence of revision-ids;
2528
a revision-id may not be None or 'null:'
2530
inventories = self.iter_inventories(revision_ids)
2531
for inv in inventories:
2532
yield InventoryRevisionTree(self, inv, inv.revision_id)
2534
def _filtered_revision_trees(self, revision_ids, file_ids):
2535
"""Return Tree for a revision on this branch with only some files.
2537
:param revision_ids: a sequence of revision-ids;
2538
a revision-id may not be None or 'null:'
2539
:param file_ids: if not None, the result is filtered
2540
so that only those file-ids, their parents and their
2541
children are included.
2543
inventories = self.iter_inventories(revision_ids)
2544
for inv in inventories:
2545
# Should we introduce a FilteredRevisionTree class rather
2546
# than pre-filter the inventory here?
2547
filtered_inv = inv.filter(file_ids)
2548
yield InventoryRevisionTree(self, filtered_inv, filtered_inv.revision_id)
1379
"""Return Tree for a revision on this branch.
1381
`revision_id` may not be None or 'null:'"""
1382
assert None not in revision_ids
1383
assert _mod_revision.NULL_REVISION not in revision_ids
1384
texts = self.get_inventory_weave().get_texts(revision_ids)
1385
for text, revision_id in zip(texts, revision_ids):
1386
inv = self.deserialise_inventory(revision_id, text)
1387
yield RevisionTree(self, inv, revision_id)
2550
1389
@needs_read_lock
2551
1390
def get_ancestry(self, revision_id, topo_sorted=True):
2552
1391
"""Return a list of revision-ids integrated by a revision.
2554
The first element of the list is always None, indicating the origin
2555
revision. This might change when we have history horizons, or
1393
The first element of the list is always None, indicating the origin
1394
revision. This might change when we have history horizons, or
2556
1395
perhaps we should have a new API.
2558
1397
This is topologically sorted.
2560
1399
if _mod_revision.is_null(revision_id):
2562
1401
if not self.has_revision(revision_id):
2563
1402
raise errors.NoSuchRevision(self, revision_id)
2564
graph = self.get_graph()
2566
search = graph._make_breadth_first_searcher([revision_id])
2569
found, ghosts = search.next_with_ghosts()
2570
except StopIteration:
2573
if _mod_revision.NULL_REVISION in keys:
2574
keys.remove(_mod_revision.NULL_REVISION)
2576
parent_map = graph.get_parent_map(keys)
2577
keys = tsort.topo_sort(parent_map)
2578
return [None] + list(keys)
1403
w = self.get_inventory_weave()
1404
candidates = w.get_ancestry(revision_id, topo_sorted)
1405
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
2580
def pack(self, hint=None, clean_obsolete_packs=False):
2581
1408
"""Compress the data within the repository.
2583
1410
This operation only makes sense for some repository types. For other
2584
1411
types it should be a no-op that just returns.
2586
1413
This stub method does not require a lock, but subclasses should use
2587
@needs_write_lock as this is a long running call it's reasonable to
1414
@needs_write_lock as this is a long running call its reasonable to
2588
1415
implicitly lock for the user.
2590
:param hint: If not supplied, the whole repository is packed.
2591
If supplied, the repository may use the hint parameter as a
2592
hint for the parts of the repository to pack. A hint can be
2593
obtained from the result of commit_write_group(). Out of
2594
date hints are simply ignored, because concurrent operations
2595
can obsolete them rapidly.
2597
:param clean_obsolete_packs: Clean obsolete packs immediately after
1419
def print_file(self, file, revision_id):
1420
"""Print `file` to stdout.
1422
FIXME RBC 20060125 as John Meinel points out this is a bad api
1423
- it writes to stdout, it assumes that that is valid etc. Fix
1424
by creating a new more flexible convenience function.
1426
tree = self.revision_tree(revision_id)
1427
# use inventory as it was in that revision
1428
file_id = tree.inventory.path2id(file)
1430
# TODO: jam 20060427 Write a test for this code path
1431
# it had a bug in it, and was raising the wrong
1433
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1434
tree.print_file(file_id)
2601
1436
def get_transaction(self):
2602
1437
return self.control_files.get_transaction()
2604
def get_parent_map(self, revision_ids):
2605
"""See graph.StackedParentsProvider.get_parent_map"""
2606
# revisions index works in keys; this just works in revisions
2607
# therefore wrap and unwrap
1439
def revision_parents(self, revision_id):
1440
return self.get_inventory_weave().parent_names(revision_id)
1442
def get_parents(self, revision_ids):
1443
"""See StackedParentsProvider.get_parents"""
2610
1445
for revision_id in revision_ids:
2611
1446
if revision_id == _mod_revision.NULL_REVISION:
2612
result[revision_id] = ()
2613
elif revision_id is None:
2614
raise ValueError('get_parent_map(None) is not valid')
2616
query_keys.append((revision_id ,))
2617
for ((revision_id,), parent_keys) in \
2618
self.revisions.get_parent_map(query_keys).iteritems():
2620
result[revision_id] = tuple([parent_revid
2621
for (parent_revid,) in parent_keys])
2623
result[revision_id] = (_mod_revision.NULL_REVISION,)
1450
parents = self.get_revision(revision_id).parent_ids
1451
except errors.NoSuchRevision:
1454
if len(parents) == 0:
1455
parents = [_mod_revision.NULL_REVISION]
1456
parents_list.append(parents)
2626
1459
def _make_parents_provider(self):
2630
def get_known_graph_ancestry(self, revision_ids):
2631
"""Return the known graph for a set of revision ids and their ancestors.
2633
st = static_tuple.StaticTuple
2634
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2635
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2636
return graph.GraphThunkIdsToKeys(known_graph)
2638
1462
def get_graph(self, other_repository=None):
2639
1463
"""Return the graph walker for this repository format"""
2640
1464
parents_provider = self._make_parents_provider()
2641
1465
if (other_repository is not None and
2642
not self.has_same_location(other_repository)):
2643
parents_provider = graph.StackedParentsProvider(
1466
other_repository.bzrdir.transport.base !=
1467
self.bzrdir.transport.base):
1468
parents_provider = graph._StackedParentsProvider(
2644
1469
[parents_provider, other_repository._make_parents_provider()])
2645
1470
return graph.Graph(parents_provider)
2647
def _get_versioned_file_checker(self, text_key_references=None,
2649
"""Return an object suitable for checking versioned files.
2651
:param text_key_references: if non-None, an already built
2652
dictionary mapping text keys ((fileid, revision_id) tuples)
2653
to whether they were referred to by the inventory of the
2654
revision_id that they contain. If None, this will be
2656
:param ancestors: Optional result from
2657
self.get_graph().get_parent_map(self.all_revision_ids()) if already
2660
return _VersionedFileChecker(self,
2661
text_key_references=text_key_references, ancestors=ancestors)
2663
def revision_ids_to_search_result(self, result_set):
2664
"""Convert a set of revision ids to a graph SearchResult."""
2665
result_parents = set()
2666
for parents in self.get_graph().get_parent_map(
2667
result_set).itervalues():
2668
result_parents.update(parents)
2669
included_keys = result_set.intersection(result_parents)
2670
start_keys = result_set.difference(included_keys)
2671
exclude_keys = result_parents.difference(result_set)
2672
result = graph.SearchResult(start_keys, exclude_keys,
2673
len(result_set), result_set)
1472
def get_versioned_file_checker(self, revisions, revision_versions_cache):
1473
return VersionedFileChecker(revisions, revision_versions_cache, self)
2676
1475
@needs_write_lock
2677
1476
def set_make_working_trees(self, new_value):
3345
1947
InterRepository.get(other).method_name(parameters).
3348
_walk_to_common_revisions_batch_size = 50
3349
1950
_optimisers = []
3350
1951
"""The available optimised InterRepository types."""
3353
1953
def copy_content(self, revision_id=None):
3354
"""Make a complete copy of the content in self into destination.
3356
This is a destructive operation! Do not use it on existing
3359
:param revision_id: Only copy the content needed to construct
3360
revision_id and its parents.
3363
self.target.set_make_working_trees(self.source.make_working_trees())
3364
except NotImplementedError:
3366
self.target.fetch(self.source, revision_id=revision_id)
3369
def fetch(self, revision_id=None, find_ghosts=False,
1954
raise NotImplementedError(self.copy_content)
1956
def fetch(self, revision_id=None, pb=None):
3371
1957
"""Fetch the content required to construct revision_id.
3373
1959
The content is copied from self.source to self.target.
3375
1961
:param revision_id: if None all content is copied, if NULL_REVISION no
3376
1962
content is copied.
3379
ui.ui_factory.warn_experimental_format_fetch(self)
3380
from bzrlib.fetch import RepoFetcher
3381
# See <https://launchpad.net/bugs/456077> asking for a warning here
3382
if self.source._format.network_name() != self.target._format.network_name():
3383
ui.ui_factory.show_user_warning('cross_format_fetch',
3384
from_format=self.source._format,
3385
to_format=self.target._format)
3386
f = RepoFetcher(to_repository=self.target,
3387
from_repository=self.source,
3388
last_revision=revision_id,
3389
fetch_spec=fetch_spec,
3390
find_ghosts=find_ghosts)
3392
def _walk_to_common_revisions(self, revision_ids, if_present_ids=None):
3393
"""Walk out from revision_ids in source to revisions target has.
3395
:param revision_ids: The start point for the search.
3396
:return: A set of revision ids.
3398
target_graph = self.target.get_graph()
3399
revision_ids = frozenset(revision_ids)
3401
all_wanted_revs = revision_ids.union(if_present_ids)
3403
all_wanted_revs = revision_ids
3404
missing_revs = set()
3405
source_graph = self.source.get_graph()
3406
# ensure we don't pay silly lookup costs.
3407
searcher = source_graph._make_breadth_first_searcher(all_wanted_revs)
3408
null_set = frozenset([_mod_revision.NULL_REVISION])
3409
searcher_exhausted = False
3413
# Iterate the searcher until we have enough next_revs
3414
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3416
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3417
next_revs.update(next_revs_part)
3418
ghosts.update(ghosts_part)
3419
except StopIteration:
3420
searcher_exhausted = True
3422
# If there are ghosts in the source graph, and the caller asked for
3423
# them, make sure that they are present in the target.
3424
# We don't care about other ghosts as we can't fetch them and
3425
# haven't been asked to.
3426
ghosts_to_check = set(revision_ids.intersection(ghosts))
3427
revs_to_get = set(next_revs).union(ghosts_to_check)
3429
have_revs = set(target_graph.get_parent_map(revs_to_get))
3430
# we always have NULL_REVISION present.
3431
have_revs = have_revs.union(null_set)
3432
# Check if the target is missing any ghosts we need.
3433
ghosts_to_check.difference_update(have_revs)
3435
# One of the caller's revision_ids is a ghost in both the
3436
# source and the target.
3437
raise errors.NoSuchRevision(
3438
self.source, ghosts_to_check.pop())
3439
missing_revs.update(next_revs - have_revs)
3440
# Because we may have walked past the original stop point, make
3441
# sure everything is stopped
3442
stop_revs = searcher.find_seen_ancestors(have_revs)
3443
searcher.stop_searching_any(stop_revs)
3444
if searcher_exhausted:
3446
return searcher.get_result()
1963
:param pb: optional progress bar to use for progress reports. If not
1964
provided a default one will be created.
1966
Returns the copied revision count and the failed revisions in a tuple:
1969
raise NotImplementedError(self.fetch)
3448
1971
@needs_read_lock
3449
def search_missing_revision_ids(self,
3450
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
3451
find_ghosts=True, revision_ids=None, if_present_ids=None):
1972
def missing_revision_ids(self, revision_id=None):
3452
1973
"""Return the revision ids that source has that target does not.
1975
These are returned in topological order.
3454
1977
:param revision_id: only return revision ids included by this
3456
:param revision_ids: return revision ids included by these
3457
revision_ids. NoSuchRevision will be raised if any of these
3458
revisions are not present.
3459
:param if_present_ids: like revision_ids, but will not cause
3460
NoSuchRevision if any of these are absent, instead they will simply
3461
not be in the result. This is useful for e.g. finding revisions
3462
to fetch for tags, which may reference absent revisions.
3463
:param find_ghosts: If True find missing revisions in deep history
3464
rather than just finding the surface difference.
3465
:return: A bzrlib.graph.SearchResult.
3467
if symbol_versioning.deprecated_passed(revision_id):
3468
symbol_versioning.warn(
3469
'search_missing_revision_ids(revision_id=...) was '
3470
'deprecated in 2.4. Use revision_ids=[...] instead.',
3471
DeprecationWarning, stacklevel=2)
3472
if revision_ids is not None:
3473
raise AssertionError(
3474
'revision_ids is mutually exclusive with revision_id')
3475
if revision_id is not None:
3476
revision_ids = [revision_id]
3478
# stop searching at found target revisions.
3479
if not find_ghosts and (revision_ids is not None or if_present_ids is
3481
return self._walk_to_common_revisions(revision_ids,
3482
if_present_ids=if_present_ids)
3483
1980
# generic, possibly worst case, slow code path.
3484
1981
target_ids = set(self.target.all_revision_ids())
3485
source_ids = self._present_source_revisions_for(
3486
revision_ids, if_present_ids)
1982
if revision_id is not None:
1983
source_ids = self.source.get_ancestry(revision_id)
1984
assert source_ids[0] is None
1987
source_ids = self.source.all_revision_ids()
3487
1988
result_set = set(source_ids).difference(target_ids)
3488
return self.source.revision_ids_to_search_result(result_set)
3490
def _present_source_revisions_for(self, revision_ids, if_present_ids=None):
3491
"""Returns set of all revisions in ancestry of revision_ids present in
3494
:param revision_ids: if None, all revisions in source are returned.
3495
:param if_present_ids: like revision_ids, but if any/all of these are
3496
absent no error is raised.
3498
if revision_ids is not None or if_present_ids is not None:
3499
# First, ensure all specified revisions exist. Callers expect
3500
# NoSuchRevision when they pass absent revision_ids here.
3501
if revision_ids is None:
3502
revision_ids = set()
3503
if if_present_ids is None:
3504
if_present_ids = set()
3505
revision_ids = set(revision_ids)
3506
if_present_ids = set(if_present_ids)
3507
all_wanted_ids = revision_ids.union(if_present_ids)
3508
graph = self.source.get_graph()
3509
present_revs = set(graph.get_parent_map(all_wanted_ids))
3510
missing = revision_ids.difference(present_revs)
3512
raise errors.NoSuchRevision(self.source, missing.pop())
3513
found_ids = all_wanted_ids.intersection(present_revs)
3514
source_ids = [rev_id for (rev_id, parents) in
3515
graph.iter_ancestry(found_ids)
3516
if rev_id != _mod_revision.NULL_REVISION
3517
and parents is not None]
3519
source_ids = self.source.all_revision_ids()
3520
return set(source_ids)
1989
# this may look like a no-op: its not. It preserves the ordering
1990
# other_ids had while only returning the members from other_ids
1991
# that we've decided we need.
1992
return [rev_id for rev_id in source_ids if rev_id in result_set]
3523
1995
def _same_model(source, target):
3524
"""True if source and target have the same data representation.
3526
Note: this is always called on the base class; overriding it in a
3527
subclass will have no effect.
3530
InterRepository._assert_same_model(source, target)
3532
except errors.IncompatibleRepositories, e:
3536
def _assert_same_model(source, target):
3537
"""Raise an exception if two repositories do not use the same model.
1996
"""True if source and target have the same data representation."""
3539
1997
if source.supports_rich_root() != target.supports_rich_root():
3540
raise errors.IncompatibleRepositories(source, target,
3541
"different rich-root support")
3542
1999
if source._serializer != target._serializer:
3543
raise errors.IncompatibleRepositories(source, target,
3544
"different serializers")
3547
2004
class InterSameDataRepository(InterRepository):
3548
2005
"""Code for converting between repositories that represent the same data.
3550
2007
Data format and model must match for this to work.
3554
2011
def _get_repo_format_to_test(self):
3555
2012
"""Repository format for testing with.
3557
2014
InterSameData can pull from subtree to subtree and from non-subtree to
3558
2015
non-subtree, so we test this with the richest repository format.
3564
2021
def is_compatible(source, target):
3565
2022
return InterRepository._same_model(source, target)
3568
class InterDifferingSerializer(InterRepository):
3571
def _get_repo_format_to_test(self):
3575
def is_compatible(source, target):
3576
"""Be compatible with Knit2 source and Knit3 target"""
3577
# This is redundant with format.check_conversion_target(), however that
3578
# raises an exception, and we just want to say "False" as in we won't
3579
# support converting between these formats.
3580
if 'IDS_never' in debug.debug_flags:
3582
if source.supports_rich_root() and not target.supports_rich_root():
3584
if (source._format.supports_tree_reference
3585
and not target._format.supports_tree_reference):
3587
if target._fallback_repositories and target._format.supports_chks:
3588
# IDS doesn't know how to copy CHKs for the parent inventories it
3589
# adds to stacked repos.
3591
if 'IDS_always' in debug.debug_flags:
3593
# Only use this code path for local source and target. IDS does far
3594
# too much IO (both bandwidth and roundtrips) over a network.
3595
if not source.bzrdir.transport.base.startswith('file:///'):
3597
if not target.bzrdir.transport.base.startswith('file:///'):
3601
def _get_trees(self, revision_ids, cache):
3603
for rev_id in revision_ids:
3605
possible_trees.append((rev_id, cache[rev_id]))
3607
# Not cached, but inventory might be present anyway.
3609
tree = self.source.revision_tree(rev_id)
3610
except errors.NoSuchRevision:
3611
# Nope, parent is ghost.
3614
cache[rev_id] = tree
3615
possible_trees.append((rev_id, tree))
3616
return possible_trees
3618
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3619
"""Get the best delta and base for this revision.
3621
:return: (basis_id, delta)
3624
# Generate deltas against each tree, to find the shortest.
3625
texts_possibly_new_in_tree = set()
3626
for basis_id, basis_tree in possible_trees:
3627
delta = tree.inventory._make_delta(basis_tree.inventory)
3628
for old_path, new_path, file_id, new_entry in delta:
3629
if new_path is None:
3630
# This file_id isn't present in the new rev, so we don't
3634
# Rich roots are handled elsewhere...
3636
kind = new_entry.kind
3637
if kind != 'directory' and kind != 'file':
3638
# No text record associated with this inventory entry.
3640
# This is a directory or file that has changed somehow.
3641
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3642
deltas.append((len(delta), basis_id, delta))
3644
return deltas[0][1:]
3646
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3647
"""Find all parent revisions that are absent, but for which the
3648
inventory is present, and copy those inventories.
3650
This is necessary to preserve correctness when the source is stacked
3651
without fallbacks configured. (Note that in cases like upgrade the
3652
source may be not have _fallback_repositories even though it is
3656
for parents in parent_map.values():
3657
parent_revs.update(parents)
3658
present_parents = self.source.get_parent_map(parent_revs)
3659
absent_parents = set(parent_revs).difference(present_parents)
3660
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3661
(rev_id,) for rev_id in absent_parents)
3662
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3663
for parent_tree in self.source.revision_trees(parent_inv_ids):
3664
current_revision_id = parent_tree.get_revision_id()
3665
parents_parents_keys = parent_invs_keys_for_stacking[
3666
(current_revision_id,)]
3667
parents_parents = [key[-1] for key in parents_parents_keys]
3668
basis_id = _mod_revision.NULL_REVISION
3669
basis_tree = self.source.revision_tree(basis_id)
3670
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3671
self.target.add_inventory_by_delta(
3672
basis_id, delta, current_revision_id, parents_parents)
3673
cache[current_revision_id] = parent_tree
3675
def _fetch_batch(self, revision_ids, basis_id, cache):
3676
"""Fetch across a few revisions.
3678
:param revision_ids: The revisions to copy
3679
:param basis_id: The revision_id of a tree that must be in cache, used
3680
as a basis for delta when no other base is available
3681
:param cache: A cache of RevisionTrees that we can use.
3682
:return: The revision_id of the last converted tree. The RevisionTree
3683
for it will be in cache
3685
# Walk though all revisions; get inventory deltas, copy referenced
3686
# texts that delta references, insert the delta, revision and
3688
root_keys_to_create = set()
3691
pending_revisions = []
3692
parent_map = self.source.get_parent_map(revision_ids)
3693
self._fetch_parent_invs_for_stacking(parent_map, cache)
3694
self.source._safe_to_return_from_cache = True
3695
for tree in self.source.revision_trees(revision_ids):
3696
# Find a inventory delta for this revision.
3697
# Find text entries that need to be copied, too.
3698
current_revision_id = tree.get_revision_id()
3699
parent_ids = parent_map.get(current_revision_id, ())
3700
parent_trees = self._get_trees(parent_ids, cache)
3701
possible_trees = list(parent_trees)
3702
if len(possible_trees) == 0:
3703
# There either aren't any parents, or the parents are ghosts,
3704
# so just use the last converted tree.
3705
possible_trees.append((basis_id, cache[basis_id]))
3706
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3708
revision = self.source.get_revision(current_revision_id)
3709
pending_deltas.append((basis_id, delta,
3710
current_revision_id, revision.parent_ids))
3711
if self._converting_to_rich_root:
3712
self._revision_id_to_root_id[current_revision_id] = \
3714
# Determine which texts are in present in this revision but not in
3715
# any of the available parents.
3716
texts_possibly_new_in_tree = set()
3717
for old_path, new_path, file_id, entry in delta:
3718
if new_path is None:
3719
# This file_id isn't present in the new rev
3723
if not self.target.supports_rich_root():
3724
# The target doesn't support rich root, so we don't
3727
if self._converting_to_rich_root:
3728
# This can't be copied normally, we have to insert
3730
root_keys_to_create.add((file_id, entry.revision))
3733
texts_possibly_new_in_tree.add((file_id, entry.revision))
3734
for basis_id, basis_tree in possible_trees:
3735
basis_inv = basis_tree.inventory
3736
for file_key in list(texts_possibly_new_in_tree):
3737
file_id, file_revision = file_key
3739
entry = basis_inv[file_id]
3740
except errors.NoSuchId:
3742
if entry.revision == file_revision:
3743
texts_possibly_new_in_tree.remove(file_key)
3744
text_keys.update(texts_possibly_new_in_tree)
3745
pending_revisions.append(revision)
3746
cache[current_revision_id] = tree
3747
basis_id = current_revision_id
3748
self.source._safe_to_return_from_cache = False
3750
from_texts = self.source.texts
3751
to_texts = self.target.texts
3752
if root_keys_to_create:
3753
root_stream = _mod_fetch._new_root_data_stream(
3754
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3756
to_texts.insert_record_stream(root_stream)
3757
to_texts.insert_record_stream(from_texts.get_record_stream(
3758
text_keys, self.target._format._fetch_order,
3759
not self.target._format._fetch_uses_deltas))
3760
# insert inventory deltas
3761
for delta in pending_deltas:
3762
self.target.add_inventory_by_delta(*delta)
3763
if self.target._fallback_repositories:
3764
# Make sure this stacked repository has all the parent inventories
3765
# for the new revisions that we are about to insert. We do this
3766
# before adding the revisions so that no revision is added until
3767
# all the inventories it may depend on are added.
3768
# Note that this is overzealous, as we may have fetched these in an
3771
revision_ids = set()
3772
for revision in pending_revisions:
3773
revision_ids.add(revision.revision_id)
3774
parent_ids.update(revision.parent_ids)
3775
parent_ids.difference_update(revision_ids)
3776
parent_ids.discard(_mod_revision.NULL_REVISION)
3777
parent_map = self.source.get_parent_map(parent_ids)
3778
# we iterate over parent_map and not parent_ids because we don't
3779
# want to try copying any revision which is a ghost
3780
for parent_tree in self.source.revision_trees(parent_map):
3781
current_revision_id = parent_tree.get_revision_id()
3782
parents_parents = parent_map[current_revision_id]
3783
possible_trees = self._get_trees(parents_parents, cache)
3784
if len(possible_trees) == 0:
3785
# There either aren't any parents, or the parents are
3786
# ghosts, so just use the last converted tree.
3787
possible_trees.append((basis_id, cache[basis_id]))
3788
basis_id, delta = self._get_delta_for_revision(parent_tree,
3789
parents_parents, possible_trees)
3790
self.target.add_inventory_by_delta(
3791
basis_id, delta, current_revision_id, parents_parents)
3792
# insert signatures and revisions
3793
for revision in pending_revisions:
3795
signature = self.source.get_signature_text(
3796
revision.revision_id)
3797
self.target.add_signature_text(revision.revision_id,
3799
except errors.NoSuchRevision:
3801
self.target.add_revision(revision.revision_id, revision)
3804
def _fetch_all_revisions(self, revision_ids, pb):
3805
"""Fetch everything for the list of revisions.
3807
:param revision_ids: The list of revisions to fetch. Must be in
3809
:param pb: A ProgressTask
3812
basis_id, basis_tree = self._get_basis(revision_ids[0])
3814
cache = lru_cache.LRUCache(100)
3815
cache[basis_id] = basis_tree
3816
del basis_tree # We don't want to hang on to it here
3820
for offset in range(0, len(revision_ids), batch_size):
3821
self.target.start_write_group()
3823
pb.update('Transferring revisions', offset,
3825
batch = revision_ids[offset:offset+batch_size]
3826
basis_id = self._fetch_batch(batch, basis_id, cache)
3828
self.source._safe_to_return_from_cache = False
3829
self.target.abort_write_group()
3832
hint = self.target.commit_write_group()
3835
if hints and self.target._format.pack_compresses:
3836
self.target.pack(hint=hints)
3837
pb.update('Transferring revisions', len(revision_ids),
3841
def fetch(self, revision_id=None, find_ghosts=False,
3843
"""See InterRepository.fetch()."""
3844
if fetch_spec is not None:
3845
revision_ids = fetch_spec.get_keys()
3848
ui.ui_factory.warn_experimental_format_fetch(self)
3849
if (not self.source.supports_rich_root()
3850
and self.target.supports_rich_root()):
3851
self._converting_to_rich_root = True
3852
self._revision_id_to_root_id = {}
3854
self._converting_to_rich_root = False
3855
# See <https://launchpad.net/bugs/456077> asking for a warning here
3856
if self.source._format.network_name() != self.target._format.network_name():
3857
ui.ui_factory.show_user_warning('cross_format_fetch',
3858
from_format=self.source._format,
3859
to_format=self.target._format)
3860
if revision_ids is None:
3862
search_revision_ids = [revision_id]
3864
search_revision_ids = None
3865
revision_ids = self.target.search_missing_revision_ids(self.source,
3866
revision_ids=search_revision_ids,
3867
find_ghosts=find_ghosts).get_keys()
3868
if not revision_ids:
3870
revision_ids = tsort.topo_sort(
3871
self.source.get_graph().get_parent_map(revision_ids))
3872
if not revision_ids:
3874
# Walk though all revisions; get inventory deltas, copy referenced
3875
# texts that delta references, insert the delta, revision and
3877
pb = ui.ui_factory.nested_progress_bar()
3879
self._fetch_all_revisions(revision_ids, pb)
3882
return len(revision_ids), 0
3884
def _get_basis(self, first_revision_id):
3885
"""Get a revision and tree which exists in the target.
3887
This assumes that first_revision_id is selected for transmission
3888
because all other ancestors are already present. If we can't find an
3889
ancestor we fall back to NULL_REVISION since we know that is safe.
3891
:return: (basis_id, basis_tree)
3893
first_rev = self.source.get_revision(first_revision_id)
3895
basis_id = first_rev.parent_ids[0]
3896
# only valid as a basis if the target has it
3897
self.target.get_revision(basis_id)
3898
# Try to get a basis tree - if it's a ghost it will hit the
3899
# NoSuchRevision case.
3900
basis_tree = self.source.revision_tree(basis_id)
3901
except (IndexError, errors.NoSuchRevision):
3902
basis_id = _mod_revision.NULL_REVISION
3903
basis_tree = self.source.revision_tree(basis_id)
3904
return basis_id, basis_tree
3907
InterRepository.register_optimiser(InterDifferingSerializer)
2025
def copy_content(self, revision_id=None):
2026
"""Make a complete copy of the content in self into destination.
2028
This copies both the repository's revision data, and configuration information
2029
such as the make_working_trees setting.
2031
This is a destructive operation! Do not use it on existing
2034
:param revision_id: Only copy the content needed to construct
2035
revision_id and its parents.
2038
self.target.set_make_working_trees(self.source.make_working_trees())
2039
except NotImplementedError:
2041
# but don't bother fetching if we have the needed data now.
2042
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2043
self.target.has_revision(revision_id)):
2045
self.target.fetch(self.source, revision_id=revision_id)
2048
def fetch(self, revision_id=None, pb=None):
2049
"""See InterRepository.fetch()."""
2050
from bzrlib.fetch import GenericRepoFetcher
2051
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2052
self.source, self.source._format, self.target,
2053
self.target._format)
2054
f = GenericRepoFetcher(to_repository=self.target,
2055
from_repository=self.source,
2056
last_revision=revision_id,
2058
return f.count_copied, f.failed_revisions
2061
class InterWeaveRepo(InterSameDataRepository):
2062
"""Optimised code paths between Weave based repositories.
2064
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2065
implemented lazy inter-object optimisation.
2069
def _get_repo_format_to_test(self):
2070
from bzrlib.repofmt import weaverepo
2071
return weaverepo.RepositoryFormat7()
2074
def is_compatible(source, target):
2075
"""Be compatible with known Weave formats.
2077
We don't test for the stores being of specific types because that
2078
could lead to confusing results, and there is no need to be
2081
from bzrlib.repofmt.weaverepo import (
2087
return (isinstance(source._format, (RepositoryFormat5,
2089
RepositoryFormat7)) and
2090
isinstance(target._format, (RepositoryFormat5,
2092
RepositoryFormat7)))
2093
except AttributeError:
2097
def copy_content(self, revision_id=None):
2098
"""See InterRepository.copy_content()."""
2099
# weave specific optimised path:
2101
self.target.set_make_working_trees(self.source.make_working_trees())
2102
except NotImplementedError:
2104
# FIXME do not peek!
2105
if self.source.control_files._transport.listable():
2106
pb = ui.ui_factory.nested_progress_bar()
2108
self.target.weave_store.copy_all_ids(
2109
self.source.weave_store,
2111
from_transaction=self.source.get_transaction(),
2112
to_transaction=self.target.get_transaction())
2113
pb.update('copying inventory', 0, 1)
2114
self.target.control_weaves.copy_multi(
2115
self.source.control_weaves, ['inventory'],
2116
from_transaction=self.source.get_transaction(),
2117
to_transaction=self.target.get_transaction())
2118
self.target._revision_store.text_store.copy_all_ids(
2119
self.source._revision_store.text_store,
2124
self.target.fetch(self.source, revision_id=revision_id)
2127
def fetch(self, revision_id=None, pb=None):
2128
"""See InterRepository.fetch()."""
2129
from bzrlib.fetch import GenericRepoFetcher
2130
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2131
self.source, self.source._format, self.target, self.target._format)
2132
f = GenericRepoFetcher(to_repository=self.target,
2133
from_repository=self.source,
2134
last_revision=revision_id,
2136
return f.count_copied, f.failed_revisions
2139
def missing_revision_ids(self, revision_id=None):
2140
"""See InterRepository.missing_revision_ids()."""
2141
# we want all revisions to satisfy revision_id in source.
2142
# but we don't want to stat every file here and there.
2143
# we want then, all revisions other needs to satisfy revision_id
2144
# checked, but not those that we have locally.
2145
# so the first thing is to get a subset of the revisions to
2146
# satisfy revision_id in source, and then eliminate those that
2147
# we do already have.
2148
# this is slow on high latency connection to self, but as as this
2149
# disk format scales terribly for push anyway due to rewriting
2150
# inventory.weave, this is considered acceptable.
2152
if revision_id is not None:
2153
source_ids = self.source.get_ancestry(revision_id)
2154
assert source_ids[0] is None
2157
source_ids = self.source._all_possible_ids()
2158
source_ids_set = set(source_ids)
2159
# source_ids is the worst possible case we may need to pull.
2160
# now we want to filter source_ids against what we actually
2161
# have in target, but don't try to check for existence where we know
2162
# we do not have a revision as that would be pointless.
2163
target_ids = set(self.target._all_possible_ids())
2164
possibly_present_revisions = target_ids.intersection(source_ids_set)
2165
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2166
required_revisions = source_ids_set.difference(actually_present_revisions)
2167
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2168
if revision_id is not None:
2169
# we used get_ancestry to determine source_ids then we are assured all
2170
# revisions referenced are present as they are installed in topological order.
2171
# and the tip revision was validated by get_ancestry.
2172
return required_topo_revisions
2174
# if we just grabbed the possibly available ids, then
2175
# we only have an estimate of whats available and need to validate
2176
# that against the revision records.
2177
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2180
class InterKnitRepo(InterSameDataRepository):
2181
"""Optimised code paths between Knit based repositories."""
2184
def _get_repo_format_to_test(self):
2185
from bzrlib.repofmt import knitrepo
2186
return knitrepo.RepositoryFormatKnit1()
2189
def is_compatible(source, target):
2190
"""Be compatible with known Knit formats.
2192
We don't test for the stores being of specific types because that
2193
could lead to confusing results, and there is no need to be
2196
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2198
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2199
isinstance(target._format, RepositoryFormatKnit))
2200
except AttributeError:
2202
return are_knits and InterRepository._same_model(source, target)
2205
def fetch(self, revision_id=None, pb=None):
2206
"""See InterRepository.fetch()."""
2207
from bzrlib.fetch import KnitRepoFetcher
2208
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2209
self.source, self.source._format, self.target, self.target._format)
2210
f = KnitRepoFetcher(to_repository=self.target,
2211
from_repository=self.source,
2212
last_revision=revision_id,
2214
return f.count_copied, f.failed_revisions
2217
def missing_revision_ids(self, revision_id=None):
2218
"""See InterRepository.missing_revision_ids()."""
2219
if revision_id is not None:
2220
source_ids = self.source.get_ancestry(revision_id)
2221
assert source_ids[0] is None
2224
source_ids = self.source.all_revision_ids()
2225
source_ids_set = set(source_ids)
2226
# source_ids is the worst possible case we may need to pull.
2227
# now we want to filter source_ids against what we actually
2228
# have in target, but don't try to check for existence where we know
2229
# we do not have a revision as that would be pointless.
2230
target_ids = set(self.target.all_revision_ids())
2231
possibly_present_revisions = target_ids.intersection(source_ids_set)
2232
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2233
required_revisions = source_ids_set.difference(actually_present_revisions)
2234
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2235
if revision_id is not None:
2236
# we used get_ancestry to determine source_ids then we are assured all
2237
# revisions referenced are present as they are installed in topological order.
2238
# and the tip revision was validated by get_ancestry.
2239
return required_topo_revisions
2241
# if we just grabbed the possibly available ids, then
2242
# we only have an estimate of whats available and need to validate
2243
# that against the revision records.
2244
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2247
class InterModel1and2(InterRepository):
2250
def _get_repo_format_to_test(self):
2254
def is_compatible(source, target):
2255
if not source.supports_rich_root() and target.supports_rich_root():
2261
def fetch(self, revision_id=None, pb=None):
2262
"""See InterRepository.fetch()."""
2263
from bzrlib.fetch import Model1toKnit2Fetcher
2264
f = Model1toKnit2Fetcher(to_repository=self.target,
2265
from_repository=self.source,
2266
last_revision=revision_id,
2268
return f.count_copied, f.failed_revisions
2271
def copy_content(self, revision_id=None):
2272
"""Make a complete copy of the content in self into destination.
2274
This is a destructive operation! Do not use it on existing
2277
:param revision_id: Only copy the content needed to construct
2278
revision_id and its parents.
2281
self.target.set_make_working_trees(self.source.make_working_trees())
2282
except NotImplementedError:
2284
# but don't bother fetching if we have the needed data now.
2285
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2286
self.target.has_revision(revision_id)):
2288
self.target.fetch(self.source, revision_id=revision_id)
2291
class InterKnit1and2(InterKnitRepo):
2294
def _get_repo_format_to_test(self):
2298
def is_compatible(source, target):
2299
"""Be compatible with Knit1 source and Knit3 target"""
2300
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2302
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit1, \
2303
RepositoryFormatKnit3
2304
return (isinstance(source._format, (RepositoryFormatKnit1)) and
2305
isinstance(target._format, (RepositoryFormatKnit3)))
2306
except AttributeError:
2310
def fetch(self, revision_id=None, pb=None):
2311
"""See InterRepository.fetch()."""
2312
from bzrlib.fetch import Knit1to2Fetcher
2313
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2314
self.source, self.source._format, self.target,
2315
self.target._format)
2316
f = Knit1to2Fetcher(to_repository=self.target,
2317
from_repository=self.source,
2318
last_revision=revision_id,
2320
return f.count_copied, f.failed_revisions
2323
class InterRemoteToOther(InterRepository):
2325
def __init__(self, source, target):
2326
InterRepository.__init__(self, source, target)
2327
self._real_inter = None
2330
def is_compatible(source, target):
2331
if not isinstance(source, remote.RemoteRepository):
2333
source._ensure_real()
2334
real_source = source._real_repository
2335
# Is source's model compatible with target's model, and are they the
2336
# same format? Currently we can only optimise fetching from an
2337
# identical model & format repo.
2338
assert not isinstance(real_source, remote.RemoteRepository), (
2339
"We don't support remote repos backed by remote repos yet.")
2340
return real_source._format == target._format
2343
def fetch(self, revision_id=None, pb=None):
2344
"""See InterRepository.fetch()."""
2345
from bzrlib.fetch import RemoteToOtherFetcher
2346
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2347
self.source, self.target, self.target._format)
2348
# TODO: jam 20070210 This should be an assert, not a translate
2349
revision_id = osutils.safe_revision_id(revision_id)
2350
f = RemoteToOtherFetcher(to_repository=self.target,
2351
from_repository=self.source,
2352
last_revision=revision_id,
2354
return f.count_copied, f.failed_revisions
2357
def _get_repo_format_to_test(self):
2361
class InterOtherToRemote(InterRepository):
2363
def __init__(self, source, target):
2364
InterRepository.__init__(self, source, target)
2365
self._real_inter = None
2368
def is_compatible(source, target):
2369
if isinstance(target, remote.RemoteRepository):
2373
def _ensure_real_inter(self):
2374
if self._real_inter is None:
2375
self.target._ensure_real()
2376
real_target = self.target._real_repository
2377
self._real_inter = InterRepository.get(self.source, real_target)
2379
def copy_content(self, revision_id=None):
2380
self._ensure_real_inter()
2381
self._real_inter.copy_content(revision_id=revision_id)
2383
def fetch(self, revision_id=None, pb=None):
2384
self._ensure_real_inter()
2385
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2388
def _get_repo_format_to_test(self):
3908
2392
InterRepository.register_optimiser(InterSameDataRepository)
2393
InterRepository.register_optimiser(InterWeaveRepo)
2394
InterRepository.register_optimiser(InterKnitRepo)
2395
InterRepository.register_optimiser(InterModel1and2)
2396
InterRepository.register_optimiser(InterKnit1and2)
2397
InterRepository.register_optimiser(InterRemoteToOther)
2398
InterRepository.register_optimiser(InterOtherToRemote)
3911
2401
class CopyConverter(object):
3912
2402
"""A repository conversion tool which just performs a copy of the content.
3914
2404
This is slow but quite reliable.
3920
2410
:param target_format: The format the resulting repository should be.
3922
2412
self.target_format = target_format
3924
2414
def convert(self, repo, pb):
3925
2415
"""Perform the conversion of to_convert, giving feedback via pb.
3927
2417
:param to_convert: The disk object to convert.
3928
2418
:param pb: a progress bar to use for progress information.
3930
pb = ui.ui_factory.nested_progress_bar()
3933
2423
# this is only useful with metadir layouts - separated repo content.
3934
2424
# trigger an assertion if not such
3935
2425
repo._format.get_format_string()
3936
2426
self.repo_dir = repo.bzrdir
3937
pb.update('Moving repository to repository.backup')
2427
self.step('Moving repository to repository.backup')
3938
2428
self.repo_dir.transport.move('repository', 'repository.backup')
3939
2429
backup_transport = self.repo_dir.transport.clone('repository.backup')
3940
2430
repo._format.check_conversion_target(self.target_format)
3941
2431
self.source_repo = repo._format.open(self.repo_dir,
3943
2433
_override_transport=backup_transport)
3944
pb.update('Creating new repository')
2434
self.step('Creating new repository')
3945
2435
converted = self.target_format.initialize(self.repo_dir,
3946
2436
self.source_repo.is_shared())
3947
2437
converted.lock_write()
3949
pb.update('Copying content')
2439
self.step('Copying content into repository.')
3950
2440
self.source_repo.copy_content_into(converted)
3952
2442
converted.unlock()
3953
pb.update('Deleting old repository content')
2443
self.step('Deleting old repository content.')
3954
2444
self.repo_dir.transport.delete_tree('repository.backup')
3955
ui.ui_factory.note('repository converted')
3959
class _VersionedFileChecker(object):
3961
def __init__(self, repository, text_key_references=None, ancestors=None):
2445
self.pb.note('repository converted')
2447
def step(self, message):
2448
"""Update the pb by a step."""
2450
self.pb.update(message, self.count, self.total)
2462
def _unescaper(match, _map=_unescape_map):
2463
code = match.group(1)
2467
if not code.startswith('#'):
2469
return unichr(int(code[1:])).encode('utf8')
2475
def _unescape_xml(data):
2476
"""Unescape predefined XML entities in a string of data."""
2478
if _unescape_re is None:
2479
_unescape_re = re.compile('\&([^;]*);')
2480
return _unescape_re.sub(_unescaper, data)
2483
class _RevisionTextVersionCache(object):
2484
"""A cache of the versionedfile versions for revision and file-id."""
2486
def __init__(self, repository):
3962
2487
self.repository = repository
3963
self.text_index = self.repository._generate_text_key_index(
3964
text_key_references=text_key_references, ancestors=ancestors)
3966
def calculate_file_version_parents(self, text_key):
3967
"""Calculate the correct parents for a file version according to
3970
parent_keys = self.text_index[text_key]
3971
if parent_keys == [_mod_revision.NULL_REVISION]:
3973
return tuple(parent_keys)
3975
def check_file_version_parents(self, texts, progress_bar=None):
3976
"""Check the parents stored in a versioned file are correct.
3978
It also detects file versions that are not referenced by their
3979
corresponding revision's inventory.
3981
:returns: A tuple of (wrong_parents, dangling_file_versions).
3982
wrong_parents is a dict mapping {revision_id: (stored_parents,
3983
correct_parents)} for each revision_id where the stored parents
3984
are not correct. dangling_file_versions is a set of (file_id,
3985
revision_id) tuples for versions that are present in this versioned
3986
file, but not used by the corresponding inventory.
3988
local_progress = None
3989
if progress_bar is None:
3990
local_progress = ui.ui_factory.nested_progress_bar()
3991
progress_bar = local_progress
2488
self.revision_versions = {}
2489
self.revision_parents = {}
2490
self.repo_graph = self.repository.get_graph()
2491
# XXX: RBC: I haven't tracked down what uses this, but it would be
2492
# better to use the headscache directly I think.
2493
self.heads = graph.HeadsCache(self.repo_graph).heads
2495
def add_revision_text_versions(self, tree):
2496
"""Cache text version data from the supplied revision tree"""
2498
for path, entry in tree.iter_entries_by_dir():
2499
inv_revisions[entry.file_id] = entry.revision
2500
self.revision_versions[tree.get_revision_id()] = inv_revisions
2501
return inv_revisions
2503
def get_text_version(self, file_id, revision_id):
2504
"""Determine the text version for a given file-id and revision-id"""
3993
return self._check_file_version_parents(texts, progress_bar)
3996
local_progress.finished()
3998
def _check_file_version_parents(self, texts, progress_bar):
3999
"""See check_file_version_parents."""
4001
self.file_ids = set([file_id for file_id, _ in
4002
self.text_index.iterkeys()])
4003
# text keys is now grouped by file_id
4004
n_versions = len(self.text_index)
4005
progress_bar.update('loading text store', 0, n_versions)
4006
parent_map = self.repository.texts.get_parent_map(self.text_index)
4007
# On unlistable transports this could well be empty/error...
4008
text_keys = self.repository.texts.keys()
4009
unused_keys = frozenset(text_keys) - set(self.text_index)
4010
for num, key in enumerate(self.text_index.iterkeys()):
4011
progress_bar.update('checking text graph', num, n_versions)
4012
correct_parents = self.calculate_file_version_parents(key)
2506
inv_revisions = self.revision_versions[revision_id]
4014
knit_parents = parent_map[key]
2509
tree = self.repository.revision_tree(revision_id)
4015
2510
except errors.RevisionNotPresent:
2511
self.revision_versions[revision_id] = inv_revisions = {}
2513
inv_revisions = self.add_revision_text_versions(tree)
2514
return inv_revisions.get(file_id)
2516
def prepopulate_revs(self, revision_ids):
2517
# Filter out versions that we don't have an inventory for, so that the
2518
# revision_trees() call won't fail.
2519
inv_weave = self.repository.get_inventory_weave()
2520
revs = [r for r in revision_ids if inv_weave.has_version(r)]
2521
# XXX: this loop is very similar to
2522
# bzrlib.fetch.Inter1and2Helper.iter_rev_trees.
2524
for tree in self.repository.revision_trees(revs[:100]):
2525
if tree.inventory.revision_id is None:
2526
tree.inventory.revision_id = tree.get_revision_id()
2527
self.add_revision_text_versions(tree)
2530
def get_parents(self, revision_id):
2532
return self.revision_parents[revision_id]
2534
parents = self.repository.get_parents([revision_id])[0]
2535
self.revision_parents[revision_id] = parents
2539
class VersionedFileChecker(object):
2541
def __init__(self, planned_revisions, revision_versions, repository):
2542
self.planned_revisions = planned_revisions
2543
self.revision_versions = revision_versions
2544
self.repository = repository
2546
def calculate_file_version_parents(self, revision_id, file_id):
2547
text_revision = self.revision_versions.get_text_version(
2548
file_id, revision_id)
2549
if text_revision is None:
2551
parents_of_text_revision = self.revision_versions.get_parents(
2553
parents_from_inventories = []
2554
for parent in parents_of_text_revision:
2555
if parent == _mod_revision.NULL_REVISION:
2557
introduced_in = self.revision_versions.get_text_version(file_id,
2559
if introduced_in is not None:
2560
parents_from_inventories.append(introduced_in)
2561
heads = set(self.revision_versions.heads(parents_from_inventories))
2563
for parent in parents_from_inventories:
2564
if parent in heads and parent not in new_parents:
2565
new_parents.append(parent)
2568
def check_file_version_parents(self, weave, file_id):
2570
for num, revision_id in enumerate(self.planned_revisions):
2571
correct_parents = self.calculate_file_version_parents(
2572
revision_id, file_id)
2573
if correct_parents is None:
2575
text_revision = self.revision_versions.get_text_version(
2576
file_id, revision_id)
2577
knit_parents = weave.get_parents(text_revision)
4018
2578
if correct_parents != knit_parents:
4019
wrong_parents[key] = (knit_parents, correct_parents)
4020
return wrong_parents, unused_keys
4023
def _strip_NULL_ghosts(revision_graph):
4024
"""Also don't use this. more compatibility code for unmigrated clients."""
4025
# Filter ghosts, and null:
4026
if _mod_revision.NULL_REVISION in revision_graph:
4027
del revision_graph[_mod_revision.NULL_REVISION]
4028
for key, parents in revision_graph.items():
4029
revision_graph[key] = tuple(parent for parent in parents if parent
4031
return revision_graph
4034
class StreamSink(object):
4035
"""An object that can insert a stream into a repository.
4037
This interface handles the complexity of reserialising inventories and
4038
revisions from different formats, and allows unidirectional insertion into
4039
stacked repositories without looking for the missing basis parents
4043
def __init__(self, target_repo):
4044
self.target_repo = target_repo
4046
def insert_stream(self, stream, src_format, resume_tokens):
4047
"""Insert a stream's content into the target repository.
4049
:param src_format: a bzr repository format.
4051
:return: a list of resume tokens and an iterable of keys additional
4052
items required before the insertion can be completed.
4054
self.target_repo.lock_write()
4057
self.target_repo.resume_write_group(resume_tokens)
4060
self.target_repo.start_write_group()
4063
# locked_insert_stream performs a commit|suspend.
4064
missing_keys = self.insert_stream_without_locking(stream,
4065
src_format, is_resume)
4067
# suspend the write group and tell the caller what we is
4068
# missing. We know we can suspend or else we would not have
4069
# entered this code path. (All repositories that can handle
4070
# missing keys can handle suspending a write group).
4071
write_group_tokens = self.target_repo.suspend_write_group()
4072
return write_group_tokens, missing_keys
4073
hint = self.target_repo.commit_write_group()
4074
to_serializer = self.target_repo._format._serializer
4075
src_serializer = src_format._serializer
4076
if (to_serializer != src_serializer and
4077
self.target_repo._format.pack_compresses):
4078
self.target_repo.pack(hint=hint)
4081
self.target_repo.abort_write_group(suppress_errors=True)
4084
self.target_repo.unlock()
4086
def insert_stream_without_locking(self, stream, src_format,
4088
"""Insert a stream's content into the target repository.
4090
This assumes that you already have a locked repository and an active
4093
:param src_format: a bzr repository format.
4094
:param is_resume: Passed down to get_missing_parent_inventories to
4095
indicate if we should be checking for missing texts at the same
4098
:return: A set of keys that are missing.
4100
if not self.target_repo.is_write_locked():
4101
raise errors.ObjectNotLocked(self)
4102
if not self.target_repo.is_in_write_group():
4103
raise errors.BzrError('you must already be in a write group')
4104
to_serializer = self.target_repo._format._serializer
4105
src_serializer = src_format._serializer
4107
if to_serializer == src_serializer:
4108
# If serializers match and the target is a pack repository, set the
4109
# write cache size on the new pack. This avoids poor performance
4110
# on transports where append is unbuffered (such as
4111
# RemoteTransport). This is safe to do because nothing should read
4112
# back from the target repository while a stream with matching
4113
# serialization is being inserted.
4114
# The exception is that a delta record from the source that should
4115
# be a fulltext may need to be expanded by the target (see
4116
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4117
# explicitly flush any buffered writes first in that rare case.
4119
new_pack = self.target_repo._pack_collection._new_pack
4120
except AttributeError:
4121
# Not a pack repository
4124
new_pack.set_write_cache_size(1024*1024)
4125
for substream_type, substream in stream:
4126
if 'stream' in debug.debug_flags:
4127
mutter('inserting substream: %s', substream_type)
4128
if substream_type == 'texts':
4129
self.target_repo.texts.insert_record_stream(substream)
4130
elif substream_type == 'inventories':
4131
if src_serializer == to_serializer:
4132
self.target_repo.inventories.insert_record_stream(
4135
self._extract_and_insert_inventories(
4136
substream, src_serializer)
4137
elif substream_type == 'inventory-deltas':
4138
self._extract_and_insert_inventory_deltas(
4139
substream, src_serializer)
4140
elif substream_type == 'chk_bytes':
4141
# XXX: This doesn't support conversions, as it assumes the
4142
# conversion was done in the fetch code.
4143
self.target_repo.chk_bytes.insert_record_stream(substream)
4144
elif substream_type == 'revisions':
4145
# This may fallback to extract-and-insert more often than
4146
# required if the serializers are different only in terms of
4148
if src_serializer == to_serializer:
4149
self.target_repo.revisions.insert_record_stream(substream)
4151
self._extract_and_insert_revisions(substream,
4153
elif substream_type == 'signatures':
4154
self.target_repo.signatures.insert_record_stream(substream)
4156
raise AssertionError('kaboom! %s' % (substream_type,))
4157
# Done inserting data, and the missing_keys calculations will try to
4158
# read back from the inserted data, so flush the writes to the new pack
4159
# (if this is pack format).
4160
if new_pack is not None:
4161
new_pack._write_data('', flush=True)
4162
# Find all the new revisions (including ones from resume_tokens)
4163
missing_keys = self.target_repo.get_missing_parent_inventories(
4164
check_for_missing_texts=is_resume)
4166
for prefix, versioned_file in (
4167
('texts', self.target_repo.texts),
4168
('inventories', self.target_repo.inventories),
4169
('revisions', self.target_repo.revisions),
4170
('signatures', self.target_repo.signatures),
4171
('chk_bytes', self.target_repo.chk_bytes),
4173
if versioned_file is None:
4175
# TODO: key is often going to be a StaticTuple object
4176
# I don't believe we can define a method by which
4177
# (prefix,) + StaticTuple will work, though we could
4178
# define a StaticTuple.sq_concat that would allow you to
4179
# pass in either a tuple or a StaticTuple as the second
4180
# object, so instead we could have:
4181
# StaticTuple(prefix) + key here...
4182
missing_keys.update((prefix,) + key for key in
4183
versioned_file.get_missing_compression_parent_keys())
4184
except NotImplementedError:
4185
# cannot even attempt suspending, and missing would have failed
4186
# during stream insertion.
4187
missing_keys = set()
4190
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4191
target_rich_root = self.target_repo._format.rich_root_data
4192
target_tree_refs = self.target_repo._format.supports_tree_reference
4193
for record in substream:
4194
# Insert the delta directly
4195
inventory_delta_bytes = record.get_bytes_as('fulltext')
4196
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4198
parse_result = deserialiser.parse_text_bytes(
4199
inventory_delta_bytes)
4200
except inventory_delta.IncompatibleInventoryDelta, err:
4201
mutter("Incompatible delta: %s", err.msg)
4202
raise errors.IncompatibleRevision(self.target_repo._format)
4203
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4204
revision_id = new_id
4205
parents = [key[0] for key in record.parents]
4206
self.target_repo.add_inventory_by_delta(
4207
basis_id, inv_delta, revision_id, parents)
4209
def _extract_and_insert_inventories(self, substream, serializer,
4211
"""Generate a new inventory versionedfile in target, converting data.
4213
The inventory is retrieved from the source, (deserializing it), and
4214
stored in the target (reserializing it in a different format).
4216
target_rich_root = self.target_repo._format.rich_root_data
4217
target_tree_refs = self.target_repo._format.supports_tree_reference
4218
for record in substream:
4219
# It's not a delta, so it must be a fulltext in the source
4220
# serializer's format.
4221
bytes = record.get_bytes_as('fulltext')
4222
revision_id = record.key[0]
4223
inv = serializer.read_inventory_from_string(bytes, revision_id)
4224
parents = [key[0] for key in record.parents]
4225
self.target_repo.add_inventory(revision_id, inv, parents)
4226
# No need to keep holding this full inv in memory when the rest of
4227
# the substream is likely to be all deltas.
4230
def _extract_and_insert_revisions(self, substream, serializer):
4231
for record in substream:
4232
bytes = record.get_bytes_as('fulltext')
4233
revision_id = record.key[0]
4234
rev = serializer.read_revision_from_string(bytes)
4235
if rev.revision_id != revision_id:
4236
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4237
self.target_repo.add_revision(revision_id, rev)
4240
if self.target_repo._format._fetch_reconcile:
4241
self.target_repo.reconcile()
4244
class StreamSource(object):
4245
"""A source of a stream for fetching between repositories."""
4247
def __init__(self, from_repository, to_format):
4248
"""Create a StreamSource streaming from from_repository."""
4249
self.from_repository = from_repository
4250
self.to_format = to_format
4251
self._record_counter = RecordCounter()
4253
def delta_on_metadata(self):
4254
"""Return True if delta's are permitted on metadata streams.
4256
That is on revisions and signatures.
4258
src_serializer = self.from_repository._format._serializer
4259
target_serializer = self.to_format._serializer
4260
return (self.to_format._fetch_uses_deltas and
4261
src_serializer == target_serializer)
4263
def _fetch_revision_texts(self, revs):
4264
# fetch signatures first and then the revision texts
4265
# may need to be a InterRevisionStore call here.
4266
from_sf = self.from_repository.signatures
4267
# A missing signature is just skipped.
4268
keys = [(rev_id,) for rev_id in revs]
4269
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4271
self.to_format._fetch_order,
4272
not self.to_format._fetch_uses_deltas))
4273
# If a revision has a delta, this is actually expanded inside the
4274
# insert_record_stream code now, which is an alternate fix for
4276
from_rf = self.from_repository.revisions
4277
revisions = from_rf.get_record_stream(
4279
self.to_format._fetch_order,
4280
not self.delta_on_metadata())
4281
return [('signatures', signatures), ('revisions', revisions)]
4283
def _generate_root_texts(self, revs):
4284
"""This will be called by get_stream between fetching weave texts and
4285
fetching the inventory weave.
4287
if self._rich_root_upgrade():
4288
return _mod_fetch.Inter1and2Helper(
4289
self.from_repository).generate_root_texts(revs)
4293
def get_stream(self, search):
4295
revs = search.get_keys()
4296
graph = self.from_repository.get_graph()
4297
revs = tsort.topo_sort(graph.get_parent_map(revs))
4298
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4300
for knit_kind, file_id, revisions in data_to_fetch:
4301
if knit_kind != phase:
4303
# Make a new progress bar for this phase
4304
if knit_kind == "file":
4305
# Accumulate file texts
4306
text_keys.extend([(file_id, revision) for revision in
4308
elif knit_kind == "inventory":
4309
# Now copy the file texts.
4310
from_texts = self.from_repository.texts
4311
yield ('texts', from_texts.get_record_stream(
4312
text_keys, self.to_format._fetch_order,
4313
not self.to_format._fetch_uses_deltas))
4314
# Cause an error if a text occurs after we have done the
4317
# Before we process the inventory we generate the root
4318
# texts (if necessary) so that the inventories references
4320
for _ in self._generate_root_texts(revs):
4322
# we fetch only the referenced inventories because we do not
4323
# know for unselected inventories whether all their required
4324
# texts are present in the other repository - it could be
4326
for info in self._get_inventory_stream(revs):
4328
elif knit_kind == "signatures":
4329
# Nothing to do here; this will be taken care of when
4330
# _fetch_revision_texts happens.
4332
elif knit_kind == "revisions":
4333
for record in self._fetch_revision_texts(revs):
4336
raise AssertionError("Unknown knit kind %r" % knit_kind)
4338
def get_stream_for_missing_keys(self, missing_keys):
4339
# missing keys can only occur when we are byte copying and not
4340
# translating (because translation means we don't send
4341
# unreconstructable deltas ever).
4343
keys['texts'] = set()
4344
keys['revisions'] = set()
4345
keys['inventories'] = set()
4346
keys['chk_bytes'] = set()
4347
keys['signatures'] = set()
4348
for key in missing_keys:
4349
keys[key[0]].add(key[1:])
4350
if len(keys['revisions']):
4351
# If we allowed copying revisions at this point, we could end up
4352
# copying a revision without copying its required texts: a
4353
# violation of the requirements for repository integrity.
4354
raise AssertionError(
4355
'cannot copy revisions to fill in missing deltas %s' % (
4356
keys['revisions'],))
4357
for substream_kind, keys in keys.iteritems():
4358
vf = getattr(self.from_repository, substream_kind)
4359
if vf is None and keys:
4360
raise AssertionError(
4361
"cannot fill in keys for a versioned file we don't"
4362
" have: %s needs %s" % (substream_kind, keys))
4364
# No need to stream something we don't have
4366
if substream_kind == 'inventories':
4367
# Some missing keys are genuinely ghosts, filter those out.
4368
present = self.from_repository.inventories.get_parent_map(keys)
4369
revs = [key[0] for key in present]
4370
# Get the inventory stream more-or-less as we do for the
4371
# original stream; there's no reason to assume that records
4372
# direct from the source will be suitable for the sink. (Think
4373
# e.g. 2a -> 1.9-rich-root).
4374
for info in self._get_inventory_stream(revs, missing=True):
4378
# Ask for full texts always so that we don't need more round trips
4379
# after this stream.
4380
# Some of the missing keys are genuinely ghosts, so filter absent
4381
# records. The Sink is responsible for doing another check to
4382
# ensure that ghosts don't introduce missing data for future
4384
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4385
self.to_format._fetch_order, True))
4386
yield substream_kind, stream
4388
def inventory_fetch_order(self):
4389
if self._rich_root_upgrade():
4390
return 'topological'
4392
return self.to_format._fetch_order
4394
def _rich_root_upgrade(self):
4395
return (not self.from_repository._format.rich_root_data and
4396
self.to_format.rich_root_data)
4398
def _get_inventory_stream(self, revision_ids, missing=False):
4399
from_format = self.from_repository._format
4400
if (from_format.supports_chks and self.to_format.supports_chks and
4401
from_format.network_name() == self.to_format.network_name()):
4402
raise AssertionError(
4403
"this case should be handled by GroupCHKStreamSource")
4404
elif 'forceinvdeltas' in debug.debug_flags:
4405
return self._get_convertable_inventory_stream(revision_ids,
4406
delta_versus_null=missing)
4407
elif from_format.network_name() == self.to_format.network_name():
4409
return self._get_simple_inventory_stream(revision_ids,
4411
elif (not from_format.supports_chks and not self.to_format.supports_chks
4412
and from_format._serializer == self.to_format._serializer):
4413
# Essentially the same format.
4414
return self._get_simple_inventory_stream(revision_ids,
4417
# Any time we switch serializations, we want to use an
4418
# inventory-delta based approach.
4419
return self._get_convertable_inventory_stream(revision_ids,
4420
delta_versus_null=missing)
4422
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4423
# NB: This currently reopens the inventory weave in source;
4424
# using a single stream interface instead would avoid this.
4425
from_weave = self.from_repository.inventories
4427
delta_closure = True
4429
delta_closure = not self.delta_on_metadata()
4430
yield ('inventories', from_weave.get_record_stream(
4431
[(rev_id,) for rev_id in revision_ids],
4432
self.inventory_fetch_order(), delta_closure))
4434
def _get_convertable_inventory_stream(self, revision_ids,
4435
delta_versus_null=False):
4436
# The two formats are sufficiently different that there is no fast
4437
# path, so we need to send just inventorydeltas, which any
4438
# sufficiently modern client can insert into any repository.
4439
# The StreamSink code expects to be able to
4440
# convert on the target, so we need to put bytes-on-the-wire that can
4441
# be converted. That means inventory deltas (if the remote is <1.19,
4442
# RemoteStreamSink will fallback to VFS to insert the deltas).
4443
yield ('inventory-deltas',
4444
self._stream_invs_as_deltas(revision_ids,
4445
delta_versus_null=delta_versus_null))
4447
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4448
"""Return a stream of inventory-deltas for the given rev ids.
4450
:param revision_ids: The list of inventories to transmit
4451
:param delta_versus_null: Don't try to find a minimal delta for this
4452
entry, instead compute the delta versus the NULL_REVISION. This
4453
effectively streams a complete inventory. Used for stuff like
4454
filling in missing parents, etc.
4456
from_repo = self.from_repository
4457
revision_keys = [(rev_id,) for rev_id in revision_ids]
4458
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4459
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4461
inventories = self.from_repository.iter_inventories(
4462
revision_ids, 'topological')
4463
format = from_repo._format
4464
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4465
inventory_cache = lru_cache.LRUCache(50)
4466
null_inventory = from_repo.revision_tree(
4467
_mod_revision.NULL_REVISION).inventory
4468
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4469
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4470
# repo back into a non-rich-root repo ought to be allowed)
4471
serializer = inventory_delta.InventoryDeltaSerializer(
4472
versioned_root=format.rich_root_data,
4473
tree_references=format.supports_tree_reference)
4474
for inv in inventories:
4475
key = (inv.revision_id,)
4476
parent_keys = parent_map.get(key, ())
4478
if not delta_versus_null and parent_keys:
4479
# The caller did not ask for complete inventories and we have
4480
# some parents that we can delta against. Make a delta against
4481
# each parent so that we can find the smallest.
4482
parent_ids = [parent_key[0] for parent_key in parent_keys]
4483
for parent_id in parent_ids:
4484
if parent_id not in invs_sent_so_far:
4485
# We don't know that the remote side has this basis, so
4488
if parent_id == _mod_revision.NULL_REVISION:
4489
parent_inv = null_inventory
4491
parent_inv = inventory_cache.get(parent_id, None)
4492
if parent_inv is None:
4493
parent_inv = from_repo.get_inventory(parent_id)
4494
candidate_delta = inv._make_delta(parent_inv)
4495
if (delta is None or
4496
len(delta) > len(candidate_delta)):
4497
delta = candidate_delta
4498
basis_id = parent_id
4500
# Either none of the parents ended up being suitable, or we
4501
# were asked to delta against NULL
4502
basis_id = _mod_revision.NULL_REVISION
4503
delta = inv._make_delta(null_inventory)
4504
invs_sent_so_far.add(inv.revision_id)
4505
inventory_cache[inv.revision_id] = inv
4506
delta_serialized = ''.join(
4507
serializer.delta_to_lines(basis_id, key[-1], delta))
4508
yield versionedfile.FulltextContentFactory(
4509
key, parent_keys, None, delta_serialized)
4512
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4513
stop_revision=None):
4514
"""Extend the partial history to include a given index
4516
If a stop_index is supplied, stop when that index has been reached.
4517
If a stop_revision is supplied, stop when that revision is
4518
encountered. Otherwise, stop when the beginning of history is
4521
:param stop_index: The index which should be present. When it is
4522
present, history extension will stop.
4523
:param stop_revision: The revision id which should be present. When
4524
it is encountered, history extension will stop.
4526
start_revision = partial_history_cache[-1]
4527
iterator = repo.iter_reverse_revision_history(start_revision)
4529
#skip the last revision in the list
4532
if (stop_index is not None and
4533
len(partial_history_cache) > stop_index):
4535
if partial_history_cache[-1] == stop_revision:
4537
revision_id = iterator.next()
4538
partial_history_cache.append(revision_id)
4539
except StopIteration:
2579
result[revision_id] = (knit_parents, correct_parents)
added
removed
bzrlib/repository.py
