545
400
# unchanged, carry over.
546
401
ie.reference_revision = parent_entry.reference_revision
547
402
ie.revision = parent_entry.revision
548
return self._get_delta(ie, basis_inv, path), False, None
403
return self._get_delta(ie, basis_inv, path), False
549
404
ie.reference_revision = content_summary[3]
550
if ie.reference_revision is None:
551
raise AssertionError("invalid content_summary for nested tree: %r"
552
% (content_summary,))
553
self._add_text_to_weave(ie.file_id, '', heads, None)
406
self._add_text_to_weave(ie.file_id, lines, heads, None)
555
408
raise NotImplementedError('unknown kind')
556
409
ie.revision = self._new_revision_id
557
self._any_changes = True
558
return self._get_delta(ie, basis_inv, path), True, fingerprint
560
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
561
_entry_factory=entry_factory):
562
"""Record a new tree via iter_changes.
564
:param tree: The tree to obtain text contents from for changed objects.
565
:param basis_revision_id: The revision id of the tree the iter_changes
566
has been generated against. Currently assumed to be the same
567
as self.parents[0] - if it is not, errors may occur.
568
:param iter_changes: An iter_changes iterator with the changes to apply
569
to basis_revision_id. The iterator must not include any items with
570
a current kind of None - missing items must be either filtered out
571
or errored-on beefore record_iter_changes sees the item.
572
:param _entry_factory: Private method to bind entry_factory locally for
574
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
577
# Create an inventory delta based on deltas between all the parents and
578
# deltas between all the parent inventories. We use inventory delta's
579
# between the inventory objects because iter_changes masks
580
# last-changed-field only changes.
582
# file_id -> change map, change is fileid, paths, changed, versioneds,
583
# parents, names, kinds, executables
585
# {file_id -> revision_id -> inventory entry, for entries in parent
586
# trees that are not parents[0]
410
return self._get_delta(ie, basis_inv, path), True
412
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
413
versionedfile = self.repository.weave_store.get_weave_or_empty(
414
file_id, self.repository.get_transaction())
415
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
416
# than add_lines, and allows committing when a parent is ghosted for
418
# Note: as we read the content directly from the tree, we know its not
419
# been turned into unicode or badly split - but a broken tree
420
# implementation could give us bad output from readlines() so this is
421
# not a guarantee of safety. What would be better is always checking
422
# the content during test suite execution. RBC 20070912
590
revtrees = list(self.repository.revision_trees(self.parents))
591
except errors.NoSuchRevision:
592
# one or more ghosts, slow path.
594
for revision_id in self.parents:
596
revtrees.append(self.repository.revision_tree(revision_id))
597
except errors.NoSuchRevision:
599
basis_revision_id = _mod_revision.NULL_REVISION
601
revtrees.append(self.repository.revision_tree(
602
_mod_revision.NULL_REVISION))
603
# The basis inventory from a repository
605
basis_inv = revtrees[0].inventory
607
basis_inv = self.repository.revision_tree(
608
_mod_revision.NULL_REVISION).inventory
609
if len(self.parents) > 0:
610
if basis_revision_id != self.parents[0] and not ghost_basis:
612
"arbitrary basis parents not yet supported with merges")
613
for revtree in revtrees[1:]:
614
for change in revtree.inventory._make_delta(basis_inv):
615
if change[1] is None:
616
# Not present in this parent.
618
if change[2] not in merged_ids:
619
if change[0] is not None:
620
basis_entry = basis_inv[change[2]]
621
merged_ids[change[2]] = [
623
basis_entry.revision,
626
parent_entries[change[2]] = {
628
basis_entry.revision:basis_entry,
630
change[3].revision:change[3],
633
merged_ids[change[2]] = [change[3].revision]
634
parent_entries[change[2]] = {change[3].revision:change[3]}
636
merged_ids[change[2]].append(change[3].revision)
637
parent_entries[change[2]][change[3].revision] = change[3]
640
# Setup the changes from the tree:
641
# changes maps file_id -> (change, [parent revision_ids])
643
for change in iter_changes:
644
# This probably looks up in basis_inv way to much.
645
if change[1][0] is not None:
646
head_candidate = [basis_inv[change[0]].revision]
649
changes[change[0]] = change, merged_ids.get(change[0],
651
unchanged_merged = set(merged_ids) - set(changes)
652
# Extend the changes dict with synthetic changes to record merges of
654
for file_id in unchanged_merged:
655
# Record a merged version of these items that did not change vs the
656
# basis. This can be either identical parallel changes, or a revert
657
# of a specific file after a merge. The recorded content will be
658
# that of the current tree (which is the same as the basis), but
659
# the per-file graph will reflect a merge.
660
# NB:XXX: We are reconstructing path information we had, this
661
# should be preserved instead.
662
# inv delta change: (file_id, (path_in_source, path_in_target),
663
# changed_content, versioned, parent, name, kind,
666
basis_entry = basis_inv[file_id]
667
except errors.NoSuchId:
668
# a change from basis->some_parents but file_id isn't in basis
669
# so was new in the merge, which means it must have changed
670
# from basis -> current, and as it hasn't the add was reverted
671
# by the user. So we discard this change.
675
(basis_inv.id2path(file_id), tree.id2path(file_id)),
677
(basis_entry.parent_id, basis_entry.parent_id),
678
(basis_entry.name, basis_entry.name),
679
(basis_entry.kind, basis_entry.kind),
680
(basis_entry.executable, basis_entry.executable))
681
changes[file_id] = (change, merged_ids[file_id])
682
# changes contains tuples with the change and a set of inventory
683
# candidates for the file.
685
# old_path, new_path, file_id, new_inventory_entry
686
seen_root = False # Is the root in the basis delta?
687
inv_delta = self._basis_delta
688
modified_rev = self._new_revision_id
689
for change, head_candidates in changes.values():
690
if change[3][1]: # versioned in target.
691
# Several things may be happening here:
692
# We may have a fork in the per-file graph
693
# - record a change with the content from tree
694
# We may have a change against < all trees
695
# - carry over the tree that hasn't changed
696
# We may have a change against all trees
697
# - record the change with the content from tree
700
entry = _entry_factory[kind](file_id, change[5][1],
702
head_set = self._heads(change[0], set(head_candidates))
705
for head_candidate in head_candidates:
706
if head_candidate in head_set:
707
heads.append(head_candidate)
708
head_set.remove(head_candidate)
711
# Could be a carry-over situation:
712
parent_entry_revs = parent_entries.get(file_id, None)
713
if parent_entry_revs:
714
parent_entry = parent_entry_revs.get(heads[0], None)
717
if parent_entry is None:
718
# The parent iter_changes was called against is the one
719
# that is the per-file head, so any change is relevant
720
# iter_changes is valid.
721
carry_over_possible = False
723
# could be a carry over situation
724
# A change against the basis may just indicate a merge,
725
# we need to check the content against the source of the
726
# merge to determine if it was changed after the merge
728
if (parent_entry.kind != entry.kind or
729
parent_entry.parent_id != entry.parent_id or
730
parent_entry.name != entry.name):
731
# Metadata common to all entries has changed
732
# against per-file parent
733
carry_over_possible = False
735
carry_over_possible = True
736
# per-type checks for changes against the parent_entry
739
# Cannot be a carry-over situation
740
carry_over_possible = False
741
# Populate the entry in the delta
743
# XXX: There is still a small race here: If someone reverts the content of a file
744
# after iter_changes examines and decides it has changed,
745
# we will unconditionally record a new version even if some
746
# other process reverts it while commit is running (with
747
# the revert happening after iter_changes did it's
750
entry.executable = True
752
entry.executable = False
753
if (carry_over_possible and
754
parent_entry.executable == entry.executable):
755
# Check the file length, content hash after reading
757
nostore_sha = parent_entry.text_sha1
760
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
762
text = file_obj.read()
766
entry.text_sha1, entry.text_size = self._add_text_to_weave(
767
file_id, text, heads, nostore_sha)
768
yield file_id, change[1][1], (entry.text_sha1, stat_value)
769
except errors.ExistingContent:
770
# No content change against a carry_over parent
771
# Perhaps this should also yield a fs hash update?
773
entry.text_size = parent_entry.text_size
774
entry.text_sha1 = parent_entry.text_sha1
775
elif kind == 'symlink':
777
entry.symlink_target = tree.get_symlink_target(file_id)
778
if (carry_over_possible and
779
parent_entry.symlink_target == entry.symlink_target):
782
self._add_text_to_weave(change[0], '', heads, None)
783
elif kind == 'directory':
784
if carry_over_possible:
787
# Nothing to set on the entry.
788
# XXX: split into the Root and nonRoot versions.
789
if change[1][1] != '' or self.repository.supports_rich_root():
790
self._add_text_to_weave(change[0], '', heads, None)
791
elif kind == 'tree-reference':
792
if not self.repository._format.supports_tree_reference:
793
# This isn't quite sane as an error, but we shouldn't
794
# ever see this code path in practice: tree's don't
795
# permit references when the repo doesn't support tree
797
raise errors.UnsupportedOperation(tree.add_reference,
799
reference_revision = tree.get_reference_revision(change[0])
800
entry.reference_revision = reference_revision
801
if (carry_over_possible and
802
parent_entry.reference_revision == reference_revision):
805
self._add_text_to_weave(change[0], '', heads, None)
807
raise AssertionError('unknown kind %r' % kind)
809
entry.revision = modified_rev
811
entry.revision = parent_entry.revision
814
new_path = change[1][1]
815
inv_delta.append((change[1][0], new_path, change[0], entry))
818
self.new_inventory = None
820
# This should perhaps be guarded by a check that the basis we
821
# commit against is the basis for the commit and if not do a delta
823
self._any_changes = True
825
# housekeeping root entry changes do not affect no-change commits.
826
self._require_root_change(tree)
827
self.basis_delta_revision = basis_revision_id
829
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
830
parent_keys = tuple([(file_id, parent) for parent in parents])
831
return self.repository.texts._add_text(
832
(file_id, self._new_revision_id), parent_keys, new_text,
833
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
424
return versionedfile.add_lines_with_ghosts(
425
self._new_revision_id, parents, new_lines,
426
nostore_sha=nostore_sha, random_id=self.random_revid,
427
check_content=False)[0:2]
429
versionedfile.clear_cache()
836
432
class RootCommitBuilder(CommitBuilder):
837
433
"""This commitbuilder actually records the root id"""
839
435
# the root entry gets versioned properly by this builder.
840
436
_versioned_root = True
973
def add_fallback_repository(self, repository):
974
"""Add a repository to use for looking up data not held locally.
976
:param repository: A repository.
978
if not self._format.supports_external_lookups:
979
raise errors.UnstackableRepositoryFormat(self._format, self.base)
981
# This repository will call fallback.unlock() when we transition to
982
# the unlocked state, so we make sure to increment the lock count
983
repository.lock_read()
984
self._check_fallback_repository(repository)
985
self._fallback_repositories.append(repository)
986
self.texts.add_fallback_versioned_files(repository.texts)
987
self.inventories.add_fallback_versioned_files(repository.inventories)
988
self.revisions.add_fallback_versioned_files(repository.revisions)
989
self.signatures.add_fallback_versioned_files(repository.signatures)
990
if self.chk_bytes is not None:
991
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
993
def _check_fallback_repository(self, repository):
994
"""Check that this repository can fallback to repository safely.
996
Raise an error if not.
998
:param repository: A repository to fallback to.
1000
return InterRepository._assert_same_model(self, repository)
1002
498
def add_inventory(self, revision_id, inv, parents):
1003
499
"""Add the inventory inv to the repository as revision_id.
1005
501
:param parents: The revision ids of the parents that revision_id
1006
502
is known to have and are in the repository already.
1008
504
:returns: The validator(which is a sha1 digest, though what is sha'd is
1009
505
repository format specific) of the serialized inventory.
1011
if not self.is_in_write_group():
1012
raise AssertionError("%r not in write group" % (self,))
507
assert self.is_in_write_group()
1013
508
_mod_revision.check_not_reserved_id(revision_id)
1014
if not (inv.revision_id is None or inv.revision_id == revision_id):
1015
raise AssertionError(
1016
"Mismatch between inventory revision"
1017
" id and insertion revid (%r, %r)"
1018
% (inv.revision_id, revision_id))
1019
if inv.root is None:
1020
raise AssertionError()
1021
return self._add_inventory_checked(revision_id, inv, parents)
1023
def _add_inventory_checked(self, revision_id, inv, parents):
1024
"""Add inv to the repository after checking the inputs.
1026
This function can be overridden to allow different inventory styles.
1028
:seealso: add_inventory, for the contract.
509
assert inv.revision_id is None or inv.revision_id == revision_id, \
510
"Mismatch between inventory revision" \
511
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
512
assert inv.root is not None
1030
513
inv_lines = self._serialise_inventory_to_lines(inv)
1031
return self._inventory_add_lines(revision_id, parents,
514
inv_vf = self.get_inventory_weave()
515
return self._inventory_add_lines(inv_vf, revision_id, parents,
1032
516
inv_lines, check_content=False)
1034
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1035
parents, basis_inv=None, propagate_caches=False):
1036
"""Add a new inventory expressed as a delta against another revision.
1038
See the inventory developers documentation for the theory behind
1041
:param basis_revision_id: The inventory id the delta was created
1042
against. (This does not have to be a direct parent.)
1043
:param delta: The inventory delta (see Inventory.apply_delta for
1045
:param new_revision_id: The revision id that the inventory is being
1047
:param parents: The revision ids of the parents that revision_id is
1048
known to have and are in the repository already. These are supplied
1049
for repositories that depend on the inventory graph for revision
1050
graph access, as well as for those that pun ancestry with delta
1052
:param basis_inv: The basis inventory if it is already known,
1054
:param propagate_caches: If True, the caches for this inventory are
1055
copied to and updated for the result if possible.
1057
:returns: (validator, new_inv)
1058
The validator(which is a sha1 digest, though what is sha'd is
1059
repository format specific) of the serialized inventory, and the
1060
resulting inventory.
1062
if not self.is_in_write_group():
1063
raise AssertionError("%r not in write group" % (self,))
1064
_mod_revision.check_not_reserved_id(new_revision_id)
1065
basis_tree = self.revision_tree(basis_revision_id)
1066
basis_tree.lock_read()
1068
# Note that this mutates the inventory of basis_tree, which not all
1069
# inventory implementations may support: A better idiom would be to
1070
# return a new inventory, but as there is no revision tree cache in
1071
# repository this is safe for now - RBC 20081013
1072
if basis_inv is None:
1073
basis_inv = basis_tree.inventory
1074
basis_inv.apply_delta(delta)
1075
basis_inv.revision_id = new_revision_id
1076
return (self.add_inventory(new_revision_id, basis_inv, parents),
1081
def _inventory_add_lines(self, revision_id, parents, lines,
518
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
1082
519
check_content=True):
1083
520
"""Store lines in inv_vf and return the sha1 of the inventory."""
1084
parents = [(parent,) for parent in parents]
1085
result = self.inventories.add_lines((revision_id,), parents, lines,
522
for parent in parents:
524
final_parents.append(parent)
525
return inv_vf.add_lines(revision_id, final_parents, lines,
1086
526
check_content=check_content)[0]
1087
self.inventories._access.flush()
1090
528
def add_revision(self, revision_id, rev, inv=None, config=None):
1091
529
"""Add rev to the revision store as revision_id.
1162
594
Returns a set of the present revisions.
1165
graph = self.get_graph()
1166
parent_map = graph.get_parent_map(revision_ids)
1167
# The old API returned a list, should this actually be a set?
1168
return parent_map.keys()
1170
def _check_inventories(self, checker):
1171
"""Check the inventories found from the revision scan.
1173
This is responsible for verifying the sha1 of inventories and
1174
creating a pending_keys set that covers data referenced by inventories.
1176
bar = ui.ui_factory.nested_progress_bar()
1178
self._do_check_inventories(checker, bar)
1182
def _do_check_inventories(self, checker, bar):
1183
"""Helper for _check_inventories."""
1185
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1186
kinds = ['chk_bytes', 'texts']
1187
count = len(checker.pending_keys)
1188
bar.update("inventories", 0, 2)
1189
current_keys = checker.pending_keys
1190
checker.pending_keys = {}
1191
# Accumulate current checks.
1192
for key in current_keys:
1193
if key[0] != 'inventories' and key[0] not in kinds:
1194
checker._report_items.append('unknown key type %r' % (key,))
1195
keys[key[0]].add(key[1:])
1196
if keys['inventories']:
1197
# NB: output order *should* be roughly sorted - topo or
1198
# inverse topo depending on repository - either way decent
1199
# to just delta against. However, pre-CHK formats didn't
1200
# try to optimise inventory layout on disk. As such the
1201
# pre-CHK code path does not use inventory deltas.
1203
for record in self.inventories.check(keys=keys['inventories']):
1204
if record.storage_kind == 'absent':
1205
checker._report_items.append(
1206
'Missing inventory {%s}' % (record.key,))
1208
last_object = self._check_record('inventories', record,
1209
checker, last_object,
1210
current_keys[('inventories',) + record.key])
1211
del keys['inventories']
1214
bar.update("texts", 1)
1215
while (checker.pending_keys or keys['chk_bytes']
1217
# Something to check.
1218
current_keys = checker.pending_keys
1219
checker.pending_keys = {}
1220
# Accumulate current checks.
1221
for key in current_keys:
1222
if key[0] not in kinds:
1223
checker._report_items.append('unknown key type %r' % (key,))
1224
keys[key[0]].add(key[1:])
1225
# Check the outermost kind only - inventories || chk_bytes || texts
1229
for record in getattr(self, kind).check(keys=keys[kind]):
1230
if record.storage_kind == 'absent':
1231
checker._report_items.append(
1232
'Missing %s {%s}' % (kind, record.key,))
1234
last_object = self._check_record(kind, record,
1235
checker, last_object, current_keys[(kind,) + record.key])
1239
def _check_record(self, kind, record, checker, last_object, item_data):
1240
"""Check a single text from this repository."""
1241
if kind == 'inventories':
1242
rev_id = record.key[0]
1243
inv = self._deserialise_inventory(rev_id,
1244
record.get_bytes_as('fulltext'))
1245
if last_object is not None:
1246
delta = inv._make_delta(last_object)
1247
for old_path, path, file_id, ie in delta:
1250
ie.check(checker, rev_id, inv)
1252
for path, ie in inv.iter_entries():
1253
ie.check(checker, rev_id, inv)
1254
if self._format.fast_deltas:
1256
elif kind == 'chk_bytes':
1257
# No code written to check chk_bytes for this repo format.
1258
checker._report_items.append(
1259
'unsupported key type chk_bytes for %s' % (record.key,))
1260
elif kind == 'texts':
1261
self._check_text(record, checker, item_data)
1263
checker._report_items.append(
1264
'unknown key type %s for %s' % (kind, record.key))
1266
def _check_text(self, record, checker, item_data):
1267
"""Check a single text."""
1268
# Check it is extractable.
1269
# TODO: check length.
1270
if record.storage_kind == 'chunked':
1271
chunks = record.get_bytes_as(record.storage_kind)
1272
sha1 = osutils.sha_strings(chunks)
1273
length = sum(map(len, chunks))
1275
content = record.get_bytes_as('fulltext')
1276
sha1 = osutils.sha_string(content)
1277
length = len(content)
1278
if item_data and sha1 != item_data[1]:
1279
checker._report_items.append(
1280
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1281
(record.key, sha1, item_data[1], item_data[2]))
597
for id in revision_ids:
598
if self.has_revision(id):
1284
603
def create(a_bzrdir):
1285
604
"""Construct the current default format repository in a_bzrdir."""
1286
605
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
1288
def __init__(self, _format, a_bzrdir, control_files):
607
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1289
608
"""instantiate a Repository.
1291
610
:param _format: The format of the repository on disk.
1546
910
"""Commit the contents accrued within the current write group.
1548
912
:seealso: start_write_group.
1550
:return: it may return an opaque hint that can be passed to 'pack'.
1552
914
if self._write_group is not self.get_transaction():
1553
915
# has an unlock or relock occured ?
1554
916
raise errors.BzrError('mismatched lock context %r and '
1555
917
'write group %r.' %
1556
918
(self.get_transaction(), self._write_group))
1557
result = self._commit_write_group()
919
self._commit_write_group()
1558
920
self._write_group = None
1561
922
def _commit_write_group(self):
1562
923
"""Template method for per-repository write group cleanup.
1564
This is called before the write group is considered to be
925
This is called before the write group is considered to be
1565
926
finished and should ensure that all data handed to the repository
1566
for writing during the write group is safely committed (to the
927
for writing during the write group is safely committed (to the
1567
928
extent possible considering file system caching etc).
1570
def suspend_write_group(self):
1571
raise errors.UnsuspendableWriteGroup(self)
1573
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1574
"""Return the keys of missing inventory parents for revisions added in
1577
A revision is not complete if the inventory delta for that revision
1578
cannot be calculated. Therefore if the parent inventories of a
1579
revision are not present, the revision is incomplete, and e.g. cannot
1580
be streamed by a smart server. This method finds missing inventory
1581
parents for revisions added in this write group.
1583
if not self._format.supports_external_lookups:
1584
# This is only an issue for stacked repositories
1586
if not self.is_in_write_group():
1587
raise AssertionError('not in a write group')
1589
# XXX: We assume that every added revision already has its
1590
# corresponding inventory, so we only check for parent inventories that
1591
# might be missing, rather than all inventories.
1592
parents = set(self.revisions._index.get_missing_parents())
1593
parents.discard(_mod_revision.NULL_REVISION)
1594
unstacked_inventories = self.inventories._index
1595
present_inventories = unstacked_inventories.get_parent_map(
1596
key[-1:] for key in parents)
1597
parents.difference_update(present_inventories)
1598
if len(parents) == 0:
1599
# No missing parent inventories.
1601
if not check_for_missing_texts:
1602
return set(('inventories', rev_id) for (rev_id,) in parents)
1603
# Ok, now we have a list of missing inventories. But these only matter
1604
# if the inventories that reference them are missing some texts they
1605
# appear to introduce.
1606
# XXX: Texts referenced by all added inventories need to be present,
1607
# but at the moment we're only checking for texts referenced by
1608
# inventories at the graph's edge.
1609
key_deps = self.revisions._index._key_dependencies
1610
key_deps.satisfy_refs_for_keys(present_inventories)
1611
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1612
file_ids = self.fileids_altered_by_revision_ids(referrers)
1613
missing_texts = set()
1614
for file_id, version_ids in file_ids.iteritems():
1615
missing_texts.update(
1616
(file_id, version_id) for version_id in version_ids)
1617
present_texts = self.texts.get_parent_map(missing_texts)
1618
missing_texts.difference_update(present_texts)
1619
if not missing_texts:
1620
# No texts are missing, so all revisions and their deltas are
1623
# Alternatively the text versions could be returned as the missing
1624
# keys, but this is likely to be less data.
1625
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1628
def refresh_data(self):
1629
"""Re-read any data needed to to synchronise with disk.
1631
This method is intended to be called after another repository instance
1632
(such as one used by a smart server) has inserted data into the
1633
repository. It may not be called during a write group, but may be
1634
called at any other time.
1636
if self.is_in_write_group():
1637
raise errors.InternalBzrError(
1638
"May not refresh_data while in a write group.")
1639
self._refresh_data()
1641
def resume_write_group(self, tokens):
1642
if not self.is_write_locked():
1643
raise errors.NotWriteLocked(self)
1644
if self._write_group:
1645
raise errors.BzrError('already in a write group')
1646
self._resume_write_group(tokens)
1647
# so we can detect unlock/relock - the write group is now entered.
1648
self._write_group = self.get_transaction()
1650
def _resume_write_group(self, tokens):
1651
raise errors.UnsuspendableWriteGroup(self)
1653
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
931
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1655
932
"""Fetch the content required to construct revision_id from source.
1657
If revision_id is None and fetch_spec is None, then all content is
1660
fetch() may not be used when the repository is in a write group -
1661
either finish the current write group before using fetch, or use
1662
fetch before starting the write group.
934
If revision_id is None all content is copied.
1664
935
:param find_ghosts: Find and copy revisions in the source that are
1665
936
ghosts in the target (and not reachable directly by walking out to
1666
937
the first-present revision in target from revision_id).
1667
:param revision_id: If specified, all the content needed for this
1668
revision ID will be copied to the target. Fetch will determine for
1669
itself which content needs to be copied.
1670
:param fetch_spec: If specified, a SearchResult or
1671
PendingAncestryResult that describes which revisions to copy. This
1672
allows copying multiple heads at once. Mutually exclusive with
1675
if fetch_spec is not None and revision_id is not None:
1676
raise AssertionError(
1677
"fetch_spec and revision_id are mutually exclusive.")
1678
if self.is_in_write_group():
1679
raise errors.InternalBzrError(
1680
"May not fetch while in a write group.")
1681
939
# fast path same-url fetch operations
1682
# TODO: lift out to somewhere common with RemoteRepository
1683
# <https://bugs.edge.launchpad.net/bzr/+bug/401646>
1684
if (self.has_same_location(source)
1685
and fetch_spec is None
1686
and self._has_same_fallbacks(source)):
940
if self.has_same_location(source):
1687
941
# check that last_revision is in 'from' and then return a
1689
943
if (revision_id is not None and
1690
944
not _mod_revision.is_null(revision_id)):
1691
945
self.get_revision(revision_id)
1693
# if there is no specific appropriate InterRepository, this will get
1694
# the InterRepository base class, which raises an
1695
# IncompatibleRepositories when asked to fetch.
1696
947
inter = InterRepository.get(source, self)
1697
return inter.fetch(revision_id=revision_id, pb=pb,
1698
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
949
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
950
except NotImplementedError:
951
raise errors.IncompatibleRepositories(source, self)
1700
953
def create_bundle(self, target, base, fileobj, format=None):
1701
954
return serializer.write_bundle(self, target, base, fileobj, format)
1852
1084
@needs_read_lock
1853
1085
def get_revisions(self, revision_ids):
1854
"""Get many revisions at once.
1856
Repositories that need to check data on every revision read should
1857
subclass this method.
1086
"""Get many revisions at once."""
1859
1087
return self._get_revisions(revision_ids)
1861
1089
@needs_read_lock
1862
1090
def _get_revisions(self, revision_ids):
1863
1091
"""Core work logic to get many revisions without sanity checks."""
1865
for revid, rev in self._iter_revisions(revision_ids):
1867
raise errors.NoSuchRevision(self, revid)
1869
return [revs[revid] for revid in revision_ids]
1871
def _iter_revisions(self, revision_ids):
1872
"""Iterate over revision objects.
1874
:param revision_ids: An iterable of revisions to examine. None may be
1875
passed to request all revisions known to the repository. Note that
1876
not all repositories can find unreferenced revisions; for those
1877
repositories only referenced ones will be returned.
1878
:return: An iterator of (revid, revision) tuples. Absent revisions (
1879
those asked for but not available) are returned as (revid, None).
1881
if revision_ids is None:
1882
revision_ids = self.all_revision_ids()
1884
for rev_id in revision_ids:
1885
if not rev_id or not isinstance(rev_id, basestring):
1886
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1887
keys = [(key,) for key in revision_ids]
1888
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1889
for record in stream:
1890
revid = record.key[0]
1891
if record.storage_kind == 'absent':
1894
text = record.get_bytes_as('fulltext')
1895
rev = self._serializer.read_revision_from_string(text)
1092
for rev_id in revision_ids:
1093
if not rev_id or not isinstance(rev_id, basestring):
1094
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1095
revs = self._revision_store.get_revisions(revision_ids,
1096
self.get_transaction())
1098
assert not isinstance(rev.revision_id, unicode)
1099
for parent_id in rev.parent_ids:
1100
assert not isinstance(parent_id, unicode)
1898
1103
@needs_read_lock
1899
1104
def get_revision_xml(self, revision_id):
1900
1105
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1901
1106
# would have already do it.
1902
1107
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1903
# TODO: this can't just be replaced by:
1904
# return self._serializer.write_revision_to_string(
1905
# self.get_revision(revision_id))
1906
# as cStringIO preservers the encoding unlike write_revision_to_string
1907
# or some other call down the path.
1908
1108
rev = self.get_revision(revision_id)
1909
rev_tmp = cStringIO.StringIO()
1109
rev_tmp = StringIO()
1910
1110
# the current serializer..
1911
self._serializer.write_revision(rev, rev_tmp)
1111
self._revision_store._serializer.write_revision(rev, rev_tmp)
1912
1112
rev_tmp.seek(0)
1913
1113
return rev_tmp.getvalue()
1915
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1116
def get_deltas_for_revisions(self, revisions):
1916
1117
"""Produce a generator of revision deltas.
1918
1119
Note that the input is a sequence of REVISIONS, not revision_ids.
1919
1120
Trees will be held in memory until the generator exits.
1920
1121
Each delta is relative to the revision's lefthand predecessor.
1922
:param specific_fileids: if not None, the result is filtered
1923
so that only those file-ids, their parents and their
1924
children are included.
1926
# Get the revision-ids of interest
1927
1123
required_trees = set()
1928
1124
for revision in revisions:
1929
1125
required_trees.add(revision.revision_id)
1930
1126
required_trees.update(revision.parent_ids[:1])
1932
# Get the matching filtered trees. Note that it's more
1933
# efficient to pass filtered trees to changes_from() rather
1934
# than doing the filtering afterwards. changes_from() could
1935
# arguably do the filtering itself but it's path-based, not
1936
# file-id based, so filtering before or afterwards is
1938
if specific_fileids is None:
1939
trees = dict((t.get_revision_id(), t) for
1940
t in self.revision_trees(required_trees))
1942
trees = dict((t.get_revision_id(), t) for
1943
t in self._filtered_revision_trees(required_trees,
1946
# Calculate the deltas
1127
trees = dict((t.get_revision_id(), t) for
1128
t in self.revision_trees(required_trees))
1947
1129
for revision in revisions:
1948
1130
if not revision.parent_ids:
1949
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1131
old_tree = self.revision_tree(None)
1951
1133
old_tree = trees[revision.parent_ids[0]]
1952
1134
yield trees[revision.revision_id].changes_from(old_tree)
1954
1136
@needs_read_lock
1955
def get_revision_delta(self, revision_id, specific_fileids=None):
1137
def get_revision_delta(self, revision_id):
1956
1138
"""Return the delta for one revision.
1958
1140
The delta is relative to the left-hand predecessor of the
1961
:param specific_fileids: if not None, the result is filtered
1962
so that only those file-ids, their parents and their
1963
children are included.
1965
1143
r = self.get_revision(revision_id)
1966
return list(self.get_deltas_for_revisions([r],
1967
specific_fileids=specific_fileids))[0]
1144
return list(self.get_deltas_for_revisions([r]))[0]
1969
1146
@needs_write_lock
1970
1147
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
2070
1251
key = (file_id, revision_id)
2071
1252
setdefault(key, False)
2072
if revision_id == line_key[-1]:
1253
if revision_id == version_id:
2073
1254
result[key] = True
2076
def _inventory_xml_lines_for_keys(self, keys):
2077
"""Get a line iterator of the sort needed for findind references.
2079
Not relevant for non-xml inventory repositories.
2081
Ghosts in revision_keys are ignored.
2083
:param revision_keys: The revision keys for the inventories to inspect.
2084
:return: An iterator over (inventory line, revid) for the fulltexts of
2085
all of the xml inventories specified by revision_keys.
2087
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2088
for record in stream:
2089
if record.storage_kind != 'absent':
2090
chunks = record.get_bytes_as('chunked')
2091
revid = record.key[-1]
2092
lines = osutils.chunks_to_lines(chunks)
2096
1257
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2098
1259
"""Helper routine for fileids_altered_by_revision_ids.
2100
1261
This performs the translation of xml lines to revision ids.
2102
1263
:param line_iterator: An iterator of lines, origin_version_id
2103
:param revision_keys: The revision ids to filter for. This should be a
1264
:param revision_ids: The revision ids to filter for. This should be a
2104
1265
set or other type which supports efficient __contains__ lookups, as
2105
the revision key from each parsed line will be looked up in the
2106
revision_keys filter.
1266
the revision id from each parsed line will be looked up in the
1267
revision_ids filter.
2107
1268
:return: a dictionary mapping altered file-ids to an iterable of
2108
1269
revision_ids. Each altered file-ids has the exact revision_ids that
2109
1270
altered it listed explicitly.
2111
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2112
line_iterator).iterkeys())
2113
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2114
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2115
self._inventory_xml_lines_for_keys(parent_keys)))
2116
new_keys = seen - parent_seen
2118
1273
setdefault = result.setdefault
2119
for key in new_keys:
2120
setdefault(key[0], set()).add(key[-1])
1274
for file_id, revision_id in \
1275
self._find_text_key_references_from_xml_inventory_lines(
1276
line_iterator).iterkeys():
1277
# once data is all ensured-consistent; then this is
1278
# if revision_id == version_id
1279
if revision_id in revision_ids:
1280
setdefault(file_id, set()).add(revision_id)
2123
def _find_parent_ids_of_revisions(self, revision_ids):
2124
"""Find all parent ids that are mentioned in the revision graph.
2126
:return: set of revisions that are parents of revision_ids which are
2127
not part of revision_ids themselves
2129
parent_map = self.get_parent_map(revision_ids)
2131
map(parent_ids.update, parent_map.itervalues())
2132
parent_ids.difference_update(revision_ids)
2133
parent_ids.discard(_mod_revision.NULL_REVISION)
2136
def _find_parent_keys_of_revisions(self, revision_keys):
2137
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2139
:param revision_keys: An iterable of revision_keys.
2140
:return: The parents of all revision_keys that are not already in
2143
parent_map = self.revisions.get_parent_map(revision_keys)
2145
map(parent_keys.update, parent_map.itervalues())
2146
parent_keys.difference_update(revision_keys)
2147
parent_keys.discard(_mod_revision.NULL_REVISION)
2150
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1283
def fileids_altered_by_revision_ids(self, revision_ids):
2151
1284
"""Find the file ids and versions affected by revisions.
2153
1286
:param revisions: an iterable containing revision ids.
2154
:param _inv_weave: The inventory weave from this repository or None.
2155
If None, the inventory weave will be opened automatically.
2156
1287
:return: a dictionary mapping altered file-ids to an iterable of
2157
1288
revision_ids. Each altered file-ids has the exact revision_ids that
2158
1289
altered it listed explicitly.
2160
selected_keys = set((revid,) for revid in revision_ids)
2161
w = _inv_weave or self.inventories
2162
return self._find_file_ids_from_xml_inventory_lines(
2163
w.iter_lines_added_or_present_in_keys(
2164
selected_keys, pb=None),
1291
selected_revision_ids = set(revision_ids)
1292
w = self.get_inventory_weave()
1293
pb = ui.ui_factory.nested_progress_bar()
1295
return self._find_file_ids_from_xml_inventory_lines(
1296
w.iter_lines_added_or_present_in_versions(
1297
selected_revision_ids, pb=pb),
1298
selected_revision_ids)
2167
1302
def iter_files_bytes(self, desired_files):
2168
1303
"""Iterate through file versions.
2309
1442
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2310
1443
'revisions'. file-id is None unless knit-kind is 'file'.
2312
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2315
for result in self._find_non_file_keys_to_fetch(revision_ids):
2318
def _find_file_keys_to_fetch(self, revision_ids, pb):
2319
1445
# XXX: it's a bit weird to control the inventory weave caching in this
2320
1446
# generator. Ideally the caching would be done in fetch.py I think. Or
2321
1447
# maybe this generator should explicitly have the contract that it
2322
1448
# should not be iterated until the previously yielded item has been
2324
inv_w = self.inventories
1451
inv_w = self.get_inventory_weave()
1452
inv_w.enable_cache()
2326
1454
# file ids that changed
2327
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1455
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
2329
1457
num_file_ids = len(file_ids)
2330
1458
for file_id, altered_versions in file_ids.iteritems():
2332
pb.update("Fetch texts", count, num_file_ids)
1459
if _files_pb is not None:
1460
_files_pb.update("fetch texts", count, num_file_ids)
2334
1462
yield ("file", file_id, altered_versions)
1463
# We're done with the files_pb. Note that it finished by the caller,
1464
# just as it was created by the caller.
2336
def _find_non_file_keys_to_fetch(self, revision_ids):
2338
1468
yield ("inventory", None, revision_ids)
2341
# XXX: Note ATM no callers actually pay attention to this return
2342
# instead they just use the list of revision ids and ignore
2343
# missing sigs. Consider removing this work entirely
2344
revisions_with_signatures = set(self.signatures.get_parent_map(
2345
[(r,) for r in revision_ids]))
2346
revisions_with_signatures = set(
2347
[r for (r,) in revisions_with_signatures])
2348
revisions_with_signatures.intersection_update(revision_ids)
1472
revisions_with_signatures = set()
1473
for rev_id in revision_ids:
1475
self.get_signature_text(rev_id)
1476
except errors.NoSuchRevision:
1480
revisions_with_signatures.add(rev_id)
2349
1482
yield ("signatures", None, revisions_with_signatures)
2352
1485
yield ("revisions", None, revision_ids)
2354
1487
@needs_read_lock
1488
def get_inventory_weave(self):
1489
return self.control_weaves.get_weave('inventory',
1490
self.get_transaction())
2355
1493
def get_inventory(self, revision_id):
2356
1494
"""Get Inventory object by revision id."""
2357
1495
return self.iter_inventories([revision_id]).next()
2359
def iter_inventories(self, revision_ids, ordering=None):
1497
def iter_inventories(self, revision_ids):
2360
1498
"""Get many inventories by revision_ids.
2362
1500
This will buffer some or all of the texts used in constructing the
2363
1501
inventories in memory, but will only parse a single inventory at a
2366
:param revision_ids: The expected revision ids of the inventories.
2367
:param ordering: optional ordering, e.g. 'topological'. If not
2368
specified, the order of revision_ids will be preserved (by
2369
buffering if necessary).
2370
1504
:return: An iterator of inventories.
2372
if ((None in revision_ids)
2373
or (_mod_revision.NULL_REVISION in revision_ids)):
2374
raise ValueError('cannot get null revision inventory')
2375
return self._iter_inventories(revision_ids, ordering)
1506
assert None not in revision_ids
1507
assert _mod_revision.NULL_REVISION not in revision_ids
1508
return self._iter_inventories(revision_ids)
2377
def _iter_inventories(self, revision_ids, ordering):
1510
def _iter_inventories(self, revision_ids):
2378
1511
"""single-document based inventory iteration."""
2379
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2380
for text, revision_id in inv_xmls:
2381
yield self._deserialise_inventory(revision_id, text)
2383
def _iter_inventory_xmls(self, revision_ids, ordering):
2384
if ordering is None:
2385
order_as_requested = True
2386
ordering = 'unordered'
2388
order_as_requested = False
2389
keys = [(revision_id,) for revision_id in revision_ids]
2392
if order_as_requested:
2393
key_iter = iter(keys)
2394
next_key = key_iter.next()
2395
stream = self.inventories.get_record_stream(keys, ordering, True)
2397
for record in stream:
2398
if record.storage_kind != 'absent':
2399
chunks = record.get_bytes_as('chunked')
2400
if order_as_requested:
2401
text_chunks[record.key] = chunks
2403
yield ''.join(chunks), record.key[-1]
2405
raise errors.NoSuchRevision(self, record.key)
2406
if order_as_requested:
2407
# Yield as many results as we can while preserving order.
2408
while next_key in text_chunks:
2409
chunks = text_chunks.pop(next_key)
2410
yield ''.join(chunks), next_key[-1]
2412
next_key = key_iter.next()
2413
except StopIteration:
2414
# We still want to fully consume the get_record_stream,
2415
# just in case it is not actually finished at this point
2419
def _deserialise_inventory(self, revision_id, xml):
2420
"""Transform the xml into an inventory object.
1512
texts = self.get_inventory_weave().get_texts(revision_ids)
1513
for text, revision_id in zip(texts, revision_ids):
1514
yield self.deserialise_inventory(revision_id, text)
1516
def deserialise_inventory(self, revision_id, xml):
1517
"""Transform the xml into an inventory object.
2422
1519
:param revision_id: The expected revision id of the inventory.
2423
1520
:param xml: A serialised inventory.
2425
result = self._serializer.read_inventory_from_string(xml, revision_id,
2426
entry_cache=self._inventory_entry_cache,
2427
return_from_cache=self._safe_to_return_from_cache)
1522
result = self._serializer.read_inventory_from_string(xml, revision_id)
2428
1523
if result.revision_id != revision_id:
2429
1524
raise AssertionError('revision id mismatch %s != %s' % (
2430
1525
result.revision_id, revision_id))
2433
def _serialise_inventory(self, inv):
1528
def serialise_inventory(self, inv):
2434
1529
return self._serializer.write_inventory_to_string(inv)
2436
1531
def _serialise_inventory_to_lines(self, inv):
2440
1535
return self._serializer.format_num
2442
1537
@needs_read_lock
2443
def _get_inventory_xml(self, revision_id):
2444
"""Get serialized inventory as a string."""
2445
texts = self._iter_inventory_xmls([revision_id], 'unordered')
1538
def get_inventory_xml(self, revision_id):
1539
"""Get inventory XML as a file object."""
2447
text, revision_id = texts.next()
2448
except StopIteration:
1541
assert isinstance(revision_id, str), type(revision_id)
1542
iw = self.get_inventory_weave()
1543
return iw.get_text(revision_id)
2449
1545
raise errors.HistoryMissing(self, 'inventory', revision_id)
2452
def get_rev_id_for_revno(self, revno, known_pair):
2453
"""Return the revision id of a revno, given a later (revno, revid)
2454
pair in the same history.
2456
:return: if found (True, revid). If the available history ran out
2457
before reaching the revno, then this returns
2458
(False, (closest_revno, closest_revid)).
2460
known_revno, known_revid = known_pair
2461
partial_history = [known_revid]
2462
distance_from_known = known_revno - revno
2463
if distance_from_known < 0:
2465
'requested revno (%d) is later than given known revno (%d)'
2466
% (revno, known_revno))
2469
self, partial_history, stop_index=distance_from_known)
2470
except errors.RevisionNotPresent, err:
2471
if err.revision_id == known_revid:
2472
# The start revision (known_revid) wasn't found.
2474
# This is a stacked repository with no fallbacks, or a there's a
2475
# left-hand ghost. Either way, even though the revision named in
2476
# the error isn't in this repo, we know it's the next step in this
2477
# left-hand history.
2478
partial_history.append(err.revision_id)
2479
if len(partial_history) <= distance_from_known:
2480
# Didn't find enough history to get a revid for the revno.
2481
earliest_revno = known_revno - len(partial_history) + 1
2482
return (False, (earliest_revno, partial_history[-1]))
2483
if len(partial_history) - 1 > distance_from_known:
2484
raise AssertionError('_iter_for_revno returned too much history')
2485
return (True, partial_history[-1])
1548
def get_inventory_sha1(self, revision_id):
1549
"""Return the sha1 hash of the inventory entry
1551
return self.get_revision(revision_id).inventory_sha1
1554
def get_revision_graph(self, revision_id=None):
1555
"""Return a dictionary containing the revision graph.
1557
NB: This method should not be used as it accesses the entire graph all
1558
at once, which is much more data than most operations should require.
1560
:param revision_id: The revision_id to get a graph from. If None, then
1561
the entire revision graph is returned. This is a deprecated mode of
1562
operation and will be removed in the future.
1563
:return: a dictionary of revision_id->revision_parents_list.
1565
raise NotImplementedError(self.get_revision_graph)
1568
def get_revision_graph_with_ghosts(self, revision_ids=None):
1569
"""Return a graph of the revisions with ghosts marked as applicable.
1571
:param revision_ids: an iterable of revisions to graph or None for all.
1572
:return: a Graph object with the graph reachable from revision_ids.
1574
if 'evil' in debug.debug_flags:
1576
"get_revision_graph_with_ghosts scales with size of history.")
1577
result = deprecated_graph.Graph()
1578
if not revision_ids:
1579
pending = set(self.all_revision_ids())
1582
pending = set(revision_ids)
1583
# special case NULL_REVISION
1584
if _mod_revision.NULL_REVISION in pending:
1585
pending.remove(_mod_revision.NULL_REVISION)
1586
required = set(pending)
1589
revision_id = pending.pop()
1591
rev = self.get_revision(revision_id)
1592
except errors.NoSuchRevision:
1593
if revision_id in required:
1596
result.add_ghost(revision_id)
1598
for parent_id in rev.parent_ids:
1599
# is this queued or done ?
1600
if (parent_id not in pending and
1601
parent_id not in done):
1603
pending.add(parent_id)
1604
result.add_node(revision_id, rev.parent_ids)
1605
done.add(revision_id)
1608
def _get_history_vf(self):
1609
"""Get a versionedfile whose history graph reflects all revisions.
1611
For weave repositories, this is the inventory weave.
1613
return self.get_inventory_weave()
2487
1615
def iter_reverse_revision_history(self, revision_id):
2488
1616
"""Iterate backwards through revision ids in the lefthand history
2540
1666
for repositories to maintain loaded indices across multiple locks
2541
1667
by checking inside their implementation of this method to see
2542
1668
whether their indices are still valid. This depends of course on
2543
the disk format being validatable in this manner. This method is
2544
also called by the refresh_data() public interface to cause a refresh
2545
to occur while in a write lock so that data inserted by a smart server
2546
push operation is visible on the client's instance of the physical
1669
the disk format being validatable in this manner.
2550
1672
@needs_read_lock
2551
1673
def revision_tree(self, revision_id):
2552
1674
"""Return Tree for a revision on this branch.
2554
`revision_id` may be NULL_REVISION for the empty tree revision.
1676
`revision_id` may be None for the empty tree revision.
2556
revision_id = _mod_revision.ensure_null(revision_id)
2557
1678
# TODO: refactor this to use an existing revision object
2558
1679
# so we don't need to read it in twice.
2559
if revision_id == _mod_revision.NULL_REVISION:
2560
return RevisionTree(self, Inventory(root_id=None),
1680
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1681
return RevisionTree(self, Inventory(root_id=None),
2561
1682
_mod_revision.NULL_REVISION)
2563
1684
inv = self.get_revision_inventory(revision_id)
2564
1685
return RevisionTree(self, inv, revision_id)
2566
1688
def revision_trees(self, revision_ids):
2567
"""Return Trees for revisions in this repository.
1689
"""Return Tree for a revision on this branch.
2569
:param revision_ids: a sequence of revision-ids;
2570
a revision-id may not be None or 'null:'
1691
`revision_id` may not be None or 'null:'"""
2572
1692
inventories = self.iter_inventories(revision_ids)
2573
1693
for inv in inventories:
2574
1694
yield RevisionTree(self, inv, inv.revision_id)
2576
def _filtered_revision_trees(self, revision_ids, file_ids):
2577
"""Return Tree for a revision on this branch with only some files.
2579
:param revision_ids: a sequence of revision-ids;
2580
a revision-id may not be None or 'null:'
2581
:param file_ids: if not None, the result is filtered
2582
so that only those file-ids, their parents and their
2583
children are included.
2585
inventories = self.iter_inventories(revision_ids)
2586
for inv in inventories:
2587
# Should we introduce a FilteredRevisionTree class rather
2588
# than pre-filter the inventory here?
2589
filtered_inv = inv.filter(file_ids)
2590
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
2592
1696
@needs_read_lock
2593
1697
def get_ancestry(self, revision_id, topo_sorted=True):
2594
1698
"""Return a list of revision-ids integrated by a revision.
2596
The first element of the list is always None, indicating the origin
2597
revision. This might change when we have history horizons, or
1700
The first element of the list is always None, indicating the origin
1701
revision. This might change when we have history horizons, or
2598
1702
perhaps we should have a new API.
2600
1704
This is topologically sorted.
2602
1706
if _mod_revision.is_null(revision_id):
2604
1708
if not self.has_revision(revision_id):
2605
1709
raise errors.NoSuchRevision(self, revision_id)
2606
graph = self.get_graph()
2608
search = graph._make_breadth_first_searcher([revision_id])
2611
found, ghosts = search.next_with_ghosts()
2612
except StopIteration:
2615
if _mod_revision.NULL_REVISION in keys:
2616
keys.remove(_mod_revision.NULL_REVISION)
2618
parent_map = graph.get_parent_map(keys)
2619
keys = tsort.topo_sort(parent_map)
2620
return [None] + list(keys)
1710
w = self.get_inventory_weave()
1711
candidates = w.get_ancestry(revision_id, topo_sorted)
1712
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
2622
def pack(self, hint=None):
2623
1715
"""Compress the data within the repository.
2625
1717
This operation only makes sense for some repository types. For other
2626
1718
types it should be a no-op that just returns.
2628
1720
This stub method does not require a lock, but subclasses should use
2629
@needs_write_lock as this is a long running call its reasonable to
1721
@needs_write_lock as this is a long running call its reasonable to
2630
1722
implicitly lock for the user.
2632
:param hint: If not supplied, the whole repository is packed.
2633
If supplied, the repository may use the hint parameter as a
2634
hint for the parts of the repository to pack. A hint can be
2635
obtained from the result of commit_write_group(). Out of
2636
date hints are simply ignored, because concurrent operations
2637
can obsolete them rapidly.
1726
def print_file(self, file, revision_id):
1727
"""Print `file` to stdout.
1729
FIXME RBC 20060125 as John Meinel points out this is a bad api
1730
- it writes to stdout, it assumes that that is valid etc. Fix
1731
by creating a new more flexible convenience function.
1733
tree = self.revision_tree(revision_id)
1734
# use inventory as it was in that revision
1735
file_id = tree.inventory.path2id(file)
1737
# TODO: jam 20060427 Write a test for this code path
1738
# it had a bug in it, and was raising the wrong
1740
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1741
tree.print_file(file_id)
2640
1743
def get_transaction(self):
2641
1744
return self.control_files.get_transaction()
2643
def get_parent_map(self, revision_ids):
2644
"""See graph.StackedParentsProvider.get_parent_map"""
2645
# revisions index works in keys; this just works in revisions
2646
# therefore wrap and unwrap
2649
for revision_id in revision_ids:
1746
def revision_parents(self, revision_id):
1747
return self.get_inventory_weave().parent_names(revision_id)
1749
@deprecated_method(symbol_versioning.one_one)
1750
def get_parents(self, revision_ids):
1751
"""See StackedParentsProvider.get_parents"""
1752
parent_map = self.get_parent_map(revision_ids)
1753
return [parent_map.get(r, None) for r in revision_ids]
1755
def get_parent_map(self, keys):
1756
"""See graph._StackedParentsProvider.get_parent_map"""
1758
for revision_id in keys:
2650
1759
if revision_id == _mod_revision.NULL_REVISION:
2651
result[revision_id] = ()
2652
elif revision_id is None:
2653
raise ValueError('get_parent_map(None) is not valid')
2655
query_keys.append((revision_id ,))
2656
for ((revision_id,), parent_keys) in \
2657
self.revisions.get_parent_map(query_keys).iteritems():
2659
result[revision_id] = tuple([parent_revid
2660
for (parent_revid,) in parent_keys])
2662
result[revision_id] = (_mod_revision.NULL_REVISION,)
1760
parent_map[revision_id] = ()
1763
parent_id_list = self.get_revision(revision_id).parent_ids
1764
except errors.NoSuchRevision:
1767
if len(parent_id_list) == 0:
1768
parent_ids = (_mod_revision.NULL_REVISION,)
1770
parent_ids = tuple(parent_id_list)
1771
parent_map[revision_id] = parent_ids
2665
1774
def _make_parents_provider(self):
3710
2697
return self.source.revision_ids_to_search_result(result_set)
3713
class InterDifferingSerializer(InterRepository):
2700
class InterPackRepo(InterSameDataRepository):
2701
"""Optimised code paths between Pack based repositories."""
2704
def _get_repo_format_to_test(self):
2705
from bzrlib.repofmt import pack_repo
2706
return pack_repo.RepositoryFormatKnitPack1()
2709
def is_compatible(source, target):
2710
"""Be compatible with known Pack formats.
2712
We don't test for the stores being of specific types because that
2713
could lead to confusing results, and there is no need to be
2716
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2718
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2719
isinstance(target._format, RepositoryFormatPack))
2720
except AttributeError:
2722
return are_packs and InterRepository._same_model(source, target)
2725
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2726
"""See InterRepository.fetch()."""
2727
from bzrlib.repofmt.pack_repo import Packer
2728
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2729
self.source, self.source._format, self.target, self.target._format)
2730
self.count_copied = 0
2731
if revision_id is None:
2733
# everything to do - use pack logic
2734
# to fetch from all packs to one without
2735
# inventory parsing etc, IFF nothing to be copied is in the target.
2737
revision_ids = self.source.all_revision_ids()
2738
# implementing the TODO will involve:
2739
# - detecting when all of a pack is selected
2740
# - avoiding as much as possible pre-selection, so the
2741
# more-core routines such as create_pack_from_packs can filter in
2742
# a just-in-time fashion. (though having a HEADS list on a
2743
# repository might make this a lot easier, because we could
2744
# sensibly detect 'new revisions' without doing a full index scan.
2745
elif _mod_revision.is_null(revision_id):
2750
revision_ids = self.search_missing_revision_ids(revision_id,
2751
find_ghosts=find_ghosts).get_keys()
2752
except errors.NoSuchRevision:
2753
raise errors.InstallFailed([revision_id])
2754
packs = self.source._pack_collection.all_packs()
2755
pack = Packer(self.target._pack_collection, packs, '.fetch',
2756
revision_ids).pack()
2757
if pack is not None:
2758
self.target._pack_collection._save_pack_names()
2759
# Trigger an autopack. This may duplicate effort as we've just done
2760
# a pack creation, but for now it is simpler to think about as
2761
# 'upload data, then repack if needed'.
2762
self.target._pack_collection.autopack()
2763
return (pack.get_revision_count(), [])
2768
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2769
"""See InterRepository.missing_revision_ids().
2771
:param find_ghosts: Find ghosts throughout the ancestry of
2774
if not find_ghosts and revision_id is not None:
2775
return self._walk_to_common_revisions([revision_id])
2776
elif revision_id is not None:
2777
source_ids = self.source.get_ancestry(revision_id)
2778
assert source_ids[0] is None
2781
source_ids = self.source.all_revision_ids()
2782
# source_ids is the worst possible case we may need to pull.
2783
# now we want to filter source_ids against what we actually
2784
# have in target, but don't try to check for existence where we know
2785
# we do not have a revision as that would be pointless.
2786
target_ids = set(self.target.all_revision_ids())
2787
result_set = set(source_ids).difference(target_ids)
2788
return self.source.revision_ids_to_search_result(result_set)
2791
class InterModel1and2(InterRepository):
2794
def _get_repo_format_to_test(self):
2798
def is_compatible(source, target):
2799
if not source.supports_rich_root() and target.supports_rich_root():
2805
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2806
"""See InterRepository.fetch()."""
2807
from bzrlib.fetch import Model1toKnit2Fetcher
2808
f = Model1toKnit2Fetcher(to_repository=self.target,
2809
from_repository=self.source,
2810
last_revision=revision_id,
2811
pb=pb, find_ghosts=find_ghosts)
2812
return f.count_copied, f.failed_revisions
2815
def copy_content(self, revision_id=None):
2816
"""Make a complete copy of the content in self into destination.
2818
This is a destructive operation! Do not use it on existing
2821
:param revision_id: Only copy the content needed to construct
2822
revision_id and its parents.
2825
self.target.set_make_working_trees(self.source.make_working_trees())
2826
except NotImplementedError:
2828
# but don't bother fetching if we have the needed data now.
2829
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2830
self.target.has_revision(revision_id)):
2832
self.target.fetch(self.source, revision_id=revision_id)
2835
class InterKnit1and2(InterKnitRepo):
2838
def _get_repo_format_to_test(self):
2842
def is_compatible(source, target):
2843
"""Be compatible with Knit1 source and Knit3 target"""
2844
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2846
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2847
RepositoryFormatKnit3)
2848
from bzrlib.repofmt.pack_repo import (
2849
RepositoryFormatKnitPack1,
2850
RepositoryFormatKnitPack3,
2851
RepositoryFormatPackDevelopment0,
2852
RepositoryFormatPackDevelopment0Subtree,
2855
RepositoryFormatKnit1,
2856
RepositoryFormatKnitPack1,
2857
RepositoryFormatPackDevelopment0,
2860
RepositoryFormatKnit3,
2861
RepositoryFormatKnitPack3,
2862
RepositoryFormatPackDevelopment0Subtree,
2864
return (isinstance(source._format, nosubtrees) and
2865
isinstance(target._format, subtrees))
2866
except AttributeError:
2870
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2871
"""See InterRepository.fetch()."""
2872
from bzrlib.fetch import Knit1to2Fetcher
2873
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2874
self.source, self.source._format, self.target,
2875
self.target._format)
2876
f = Knit1to2Fetcher(to_repository=self.target,
2877
from_repository=self.source,
2878
last_revision=revision_id,
2879
pb=pb, find_ghosts=find_ghosts)
2880
return f.count_copied, f.failed_revisions
2883
class InterDifferingSerializer(InterKnitRepo):
3716
2886
def _get_repo_format_to_test(self):
3720
2890
def is_compatible(source, target):
3721
2891
"""Be compatible with Knit2 source and Knit3 target"""
3722
# This is redundant with format.check_conversion_target(), however that
3723
# raises an exception, and we just want to say "False" as in we won't
3724
# support converting between these formats.
3725
if 'IDS_never' in debug.debug_flags:
3727
if source.supports_rich_root() and not target.supports_rich_root():
3729
if (source._format.supports_tree_reference
3730
and not target._format.supports_tree_reference):
3732
if target._fallback_repositories and target._format.supports_chks:
3733
# IDS doesn't know how to copy CHKs for the parent inventories it
3734
# adds to stacked repos.
3736
if 'IDS_always' in debug.debug_flags:
3738
# Only use this code path for local source and target. IDS does far
3739
# too much IO (both bandwidth and roundtrips) over a network.
3740
if not source.bzrdir.transport.base.startswith('file:///'):
3742
if not target.bzrdir.transport.base.startswith('file:///'):
2892
if source.supports_rich_root() != target.supports_rich_root():
2894
# Ideally, we'd support fetching if the source had no tree references
2895
# even if it supported them...
2896
if (getattr(source, '_format.supports_tree_reference', False) and
2897
not getattr(target, '_format.supports_tree_reference', False)):
3746
def _get_trees(self, revision_ids, cache):
3748
for rev_id in revision_ids:
3750
possible_trees.append((rev_id, cache[rev_id]))
3752
# Not cached, but inventory might be present anyway.
3754
tree = self.source.revision_tree(rev_id)
3755
except errors.NoSuchRevision:
3756
# Nope, parent is ghost.
3759
cache[rev_id] = tree
3760
possible_trees.append((rev_id, tree))
3761
return possible_trees
3763
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3764
"""Get the best delta and base for this revision.
3766
:return: (basis_id, delta)
3769
# Generate deltas against each tree, to find the shortest.
3770
texts_possibly_new_in_tree = set()
3771
for basis_id, basis_tree in possible_trees:
3772
delta = tree.inventory._make_delta(basis_tree.inventory)
3773
for old_path, new_path, file_id, new_entry in delta:
3774
if new_path is None:
3775
# This file_id isn't present in the new rev, so we don't
3779
# Rich roots are handled elsewhere...
3781
kind = new_entry.kind
3782
if kind != 'directory' and kind != 'file':
3783
# No text record associated with this inventory entry.
3785
# This is a directory or file that has changed somehow.
3786
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3787
deltas.append((len(delta), basis_id, delta))
3789
return deltas[0][1:]
3791
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3792
"""Find all parent revisions that are absent, but for which the
3793
inventory is present, and copy those inventories.
3795
This is necessary to preserve correctness when the source is stacked
3796
without fallbacks configured. (Note that in cases like upgrade the
3797
source may be not have _fallback_repositories even though it is
3801
for parents in parent_map.values():
3802
parent_revs.update(parents)
3803
present_parents = self.source.get_parent_map(parent_revs)
3804
absent_parents = set(parent_revs).difference(present_parents)
3805
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3806
(rev_id,) for rev_id in absent_parents)
3807
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3808
for parent_tree in self.source.revision_trees(parent_inv_ids):
3809
current_revision_id = parent_tree.get_revision_id()
3810
parents_parents_keys = parent_invs_keys_for_stacking[
3811
(current_revision_id,)]
3812
parents_parents = [key[-1] for key in parents_parents_keys]
3813
basis_id = _mod_revision.NULL_REVISION
3814
basis_tree = self.source.revision_tree(basis_id)
3815
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3816
self.target.add_inventory_by_delta(
3817
basis_id, delta, current_revision_id, parents_parents)
3818
cache[current_revision_id] = parent_tree
3820
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3821
"""Fetch across a few revisions.
3823
:param revision_ids: The revisions to copy
3824
:param basis_id: The revision_id of a tree that must be in cache, used
3825
as a basis for delta when no other base is available
3826
:param cache: A cache of RevisionTrees that we can use.
3827
:param a_graph: A Graph object to determine the heads() of the
3828
rich-root data stream.
3829
:return: The revision_id of the last converted tree. The RevisionTree
3830
for it will be in cache
3832
# Walk though all revisions; get inventory deltas, copy referenced
3833
# texts that delta references, insert the delta, revision and
3835
root_keys_to_create = set()
3838
pending_revisions = []
3839
parent_map = self.source.get_parent_map(revision_ids)
3840
self._fetch_parent_invs_for_stacking(parent_map, cache)
3841
self.source._safe_to_return_from_cache = True
3842
for tree in self.source.revision_trees(revision_ids):
3843
# Find a inventory delta for this revision.
3844
# Find text entries that need to be copied, too.
3845
current_revision_id = tree.get_revision_id()
3846
parent_ids = parent_map.get(current_revision_id, ())
3847
parent_trees = self._get_trees(parent_ids, cache)
3848
possible_trees = list(parent_trees)
3849
if len(possible_trees) == 0:
3850
# There either aren't any parents, or the parents are ghosts,
3851
# so just use the last converted tree.
3852
possible_trees.append((basis_id, cache[basis_id]))
3853
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3855
revision = self.source.get_revision(current_revision_id)
3856
pending_deltas.append((basis_id, delta,
3857
current_revision_id, revision.parent_ids))
3858
if self._converting_to_rich_root:
3859
self._revision_id_to_root_id[current_revision_id] = \
3861
# Determine which texts are in present in this revision but not in
3862
# any of the available parents.
3863
texts_possibly_new_in_tree = set()
3864
for old_path, new_path, file_id, entry in delta:
3865
if new_path is None:
3866
# This file_id isn't present in the new rev
3870
if not self.target.supports_rich_root():
3871
# The target doesn't support rich root, so we don't
3874
if self._converting_to_rich_root:
3875
# This can't be copied normally, we have to insert
3877
root_keys_to_create.add((file_id, entry.revision))
3880
texts_possibly_new_in_tree.add((file_id, entry.revision))
3881
for basis_id, basis_tree in possible_trees:
3882
basis_inv = basis_tree.inventory
3883
for file_key in list(texts_possibly_new_in_tree):
3884
file_id, file_revision = file_key
3886
entry = basis_inv[file_id]
3887
except errors.NoSuchId:
3889
if entry.revision == file_revision:
3890
texts_possibly_new_in_tree.remove(file_key)
3891
text_keys.update(texts_possibly_new_in_tree)
3892
pending_revisions.append(revision)
3893
cache[current_revision_id] = tree
3894
basis_id = current_revision_id
3895
self.source._safe_to_return_from_cache = False
3897
from_texts = self.source.texts
3898
to_texts = self.target.texts
3899
if root_keys_to_create:
3900
root_stream = _mod_fetch._new_root_data_stream(
3901
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3902
self.source, graph=a_graph)
3903
to_texts.insert_record_stream(root_stream)
3904
to_texts.insert_record_stream(from_texts.get_record_stream(
3905
text_keys, self.target._format._fetch_order,
3906
not self.target._format._fetch_uses_deltas))
3907
# insert inventory deltas
3908
for delta in pending_deltas:
3909
self.target.add_inventory_by_delta(*delta)
3910
if self.target._fallback_repositories:
3911
# Make sure this stacked repository has all the parent inventories
3912
# for the new revisions that we are about to insert. We do this
3913
# before adding the revisions so that no revision is added until
3914
# all the inventories it may depend on are added.
3915
# Note that this is overzealous, as we may have fetched these in an
3918
revision_ids = set()
3919
for revision in pending_revisions:
3920
revision_ids.add(revision.revision_id)
3921
parent_ids.update(revision.parent_ids)
3922
parent_ids.difference_update(revision_ids)
3923
parent_ids.discard(_mod_revision.NULL_REVISION)
3924
parent_map = self.source.get_parent_map(parent_ids)
3925
# we iterate over parent_map and not parent_ids because we don't
3926
# want to try copying any revision which is a ghost
3927
for parent_tree in self.source.revision_trees(parent_map):
3928
current_revision_id = parent_tree.get_revision_id()
3929
parents_parents = parent_map[current_revision_id]
3930
possible_trees = self._get_trees(parents_parents, cache)
3931
if len(possible_trees) == 0:
3932
# There either aren't any parents, or the parents are
3933
# ghosts, so just use the last converted tree.
3934
possible_trees.append((basis_id, cache[basis_id]))
3935
basis_id, delta = self._get_delta_for_revision(parent_tree,
3936
parents_parents, possible_trees)
3937
self.target.add_inventory_by_delta(
3938
basis_id, delta, current_revision_id, parents_parents)
3939
# insert signatures and revisions
3940
for revision in pending_revisions:
3942
signature = self.source.get_signature_text(
3943
revision.revision_id)
3944
self.target.add_signature_text(revision.revision_id,
3946
except errors.NoSuchRevision:
3948
self.target.add_revision(revision.revision_id, revision)
3951
def _fetch_all_revisions(self, revision_ids, pb):
3952
"""Fetch everything for the list of revisions.
3954
:param revision_ids: The list of revisions to fetch. Must be in
3956
:param pb: A ProgressTask
3959
basis_id, basis_tree = self._get_basis(revision_ids[0])
3961
cache = lru_cache.LRUCache(100)
3962
cache[basis_id] = basis_tree
3963
del basis_tree # We don't want to hang on to it here
3965
if self._converting_to_rich_root and len(revision_ids) > 100:
3966
a_graph = _mod_fetch._get_rich_root_heads_graph(self.source,
3971
for offset in range(0, len(revision_ids), batch_size):
3972
self.target.start_write_group()
3974
pb.update('Transferring revisions', offset,
3976
batch = revision_ids[offset:offset+batch_size]
3977
basis_id = self._fetch_batch(batch, basis_id, cache,
3980
self.source._safe_to_return_from_cache = False
3981
self.target.abort_write_group()
3984
hint = self.target.commit_write_group()
3987
if hints and self.target._format.pack_compresses:
3988
self.target.pack(hint=hints)
3989
pb.update('Transferring revisions', len(revision_ids),
3992
2901
@needs_write_lock
3993
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2902
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3995
2903
"""See InterRepository.fetch()."""
3996
if fetch_spec is not None:
3997
raise AssertionError("Not implemented yet...")
3998
# See <https://launchpad.net/bugs/456077> asking for a warning here
4000
# nb this is only active for local-local fetches; other things using
4002
ui.ui_factory.warn_cross_format_fetch(self.source._format,
4003
self.target._format)
4004
ui.ui_factory.warn_experimental_format_fetch(self)
4005
if (not self.source.supports_rich_root()
4006
and self.target.supports_rich_root()):
4007
self._converting_to_rich_root = True
4008
self._revision_id_to_root_id = {}
4010
self._converting_to_rich_root = False
4011
2904
revision_ids = self.target.search_missing_revision_ids(self.source,
4012
2905
revision_id, find_ghosts=find_ghosts).get_keys()
4013
if not revision_ids:
4015
2906
revision_ids = tsort.topo_sort(
4016
2907
self.source.get_graph().get_parent_map(revision_ids))
4017
if not revision_ids:
4019
# Walk though all revisions; get inventory deltas, copy referenced
4020
# texts that delta references, insert the delta, revision and
2908
def revisions_iterator():
2909
for current_revision_id in revision_ids:
2910
revision = self.source.get_revision(current_revision_id)
2911
tree = self.source.revision_tree(current_revision_id)
2913
signature = self.source.get_signature_text(
2914
current_revision_id)
2915
except errors.NoSuchRevision:
2917
yield revision, tree, signature
4023
2919
my_pb = ui.ui_factory.nested_progress_bar()
4026
symbol_versioning.warn(
4027
symbol_versioning.deprecated_in((1, 14, 0))
4028
% "pb parameter to fetch()")
4031
self._fetch_all_revisions(revision_ids, pb)
2924
install_revisions(self.target, revisions_iterator(),
2925
len(revision_ids), pb)
4033
2927
if my_pb is not None:
4034
2928
my_pb.finished()
4035
2929
return len(revision_ids), 0
4037
def _get_basis(self, first_revision_id):
4038
"""Get a revision and tree which exists in the target.
4040
This assumes that first_revision_id is selected for transmission
4041
because all other ancestors are already present. If we can't find an
4042
ancestor we fall back to NULL_REVISION since we know that is safe.
4044
:return: (basis_id, basis_tree)
4046
first_rev = self.source.get_revision(first_revision_id)
4048
basis_id = first_rev.parent_ids[0]
4049
# only valid as a basis if the target has it
4050
self.target.get_revision(basis_id)
4051
# Try to get a basis tree - if its a ghost it will hit the
4052
# NoSuchRevision case.
4053
basis_tree = self.source.revision_tree(basis_id)
4054
except (IndexError, errors.NoSuchRevision):
4055
basis_id = _mod_revision.NULL_REVISION
4056
basis_tree = self.source.revision_tree(basis_id)
4057
return basis_id, basis_tree
2932
class InterRemoteToOther(InterRepository):
2934
def __init__(self, source, target):
2935
InterRepository.__init__(self, source, target)
2936
self._real_inter = None
2939
def is_compatible(source, target):
2940
if not isinstance(source, remote.RemoteRepository):
2942
# Is source's model compatible with target's model?
2943
source._ensure_real()
2944
real_source = source._real_repository
2945
assert not isinstance(real_source, remote.RemoteRepository), (
2946
"We don't support remote repos backed by remote repos yet.")
2947
return InterRepository._same_model(real_source, target)
2950
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2951
"""See InterRepository.fetch()."""
2952
from bzrlib.fetch import RemoteToOtherFetcher
2953
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2954
self.source, self.target, self.target._format)
2955
# TODO: jam 20070210 This should be an assert, not a translate
2956
revision_id = osutils.safe_revision_id(revision_id)
2957
f = RemoteToOtherFetcher(to_repository=self.target,
2958
from_repository=self.source,
2959
last_revision=revision_id,
2960
pb=pb, find_ghosts=find_ghosts)
2961
return f.count_copied, f.failed_revisions
2964
def _get_repo_format_to_test(self):
2968
class InterOtherToRemote(InterRepository):
2970
def __init__(self, source, target):
2971
InterRepository.__init__(self, source, target)
2972
self._real_inter = None
2975
def is_compatible(source, target):
2976
if isinstance(target, remote.RemoteRepository):
2980
def _ensure_real_inter(self):
2981
if self._real_inter is None:
2982
self.target._ensure_real()
2983
real_target = self.target._real_repository
2984
self._real_inter = InterRepository.get(self.source, real_target)
2986
def copy_content(self, revision_id=None):
2987
self._ensure_real_inter()
2988
self._real_inter.copy_content(revision_id=revision_id)
2990
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2991
self._ensure_real_inter()
2992
self._real_inter.fetch(revision_id=revision_id, pb=pb,
2993
find_ghosts=find_ghosts)
2996
def _get_repo_format_to_test(self):
4060
3000
InterRepository.register_optimiser(InterDifferingSerializer)
4061
3001
InterRepository.register_optimiser(InterSameDataRepository)
4062
3002
InterRepository.register_optimiser(InterWeaveRepo)
4063
3003
InterRepository.register_optimiser(InterKnitRepo)
3004
InterRepository.register_optimiser(InterModel1and2)
3005
InterRepository.register_optimiser(InterKnit1and2)
3006
InterRepository.register_optimiser(InterPackRepo)
3007
InterRepository.register_optimiser(InterRemoteToOther)
3008
InterRepository.register_optimiser(InterOtherToRemote)
4066
3011
class CopyConverter(object):
4067
3012
"""A repository conversion tool which just performs a copy of the content.
4069
3014
This is slow but quite reliable.
4170
3119
revision_id) tuples for versions that are present in this versioned
4171
3120
file, but not used by the corresponding inventory.
4173
local_progress = None
4174
if progress_bar is None:
4175
local_progress = ui.ui_factory.nested_progress_bar()
4176
progress_bar = local_progress
4178
return self._check_file_version_parents(texts, progress_bar)
4181
local_progress.finished()
4183
def _check_file_version_parents(self, texts, progress_bar):
4184
"""See check_file_version_parents."""
4185
3122
wrong_parents = {}
4186
self.file_ids = set([file_id for file_id, _ in
4187
self.text_index.iterkeys()])
4188
# text keys is now grouped by file_id
4189
n_versions = len(self.text_index)
4190
progress_bar.update('loading text store', 0, n_versions)
4191
parent_map = self.repository.texts.get_parent_map(self.text_index)
4192
# On unlistable transports this could well be empty/error...
4193
text_keys = self.repository.texts.keys()
4194
unused_keys = frozenset(text_keys) - set(self.text_index)
4195
for num, key in enumerate(self.text_index.iterkeys()):
4196
progress_bar.update('checking text graph', num, n_versions)
4197
correct_parents = self.calculate_file_version_parents(key)
4199
knit_parents = parent_map[key]
4200
except errors.RevisionNotPresent:
4203
if correct_parents != knit_parents:
4204
wrong_parents[key] = (knit_parents, correct_parents)
4205
return wrong_parents, unused_keys
4208
def _old_get_graph(repository, revision_id):
4209
"""DO NOT USE. That is all. I'm serious."""
4210
graph = repository.get_graph()
4211
revision_graph = dict(((key, value) for key, value in
4212
graph.iter_ancestry([revision_id]) if value is not None))
4213
return _strip_NULL_ghosts(revision_graph)
4216
def _strip_NULL_ghosts(revision_graph):
4217
"""Also don't use this. more compatibility code for unmigrated clients."""
4218
# Filter ghosts, and null:
4219
if _mod_revision.NULL_REVISION in revision_graph:
4220
del revision_graph[_mod_revision.NULL_REVISION]
4221
for key, parents in revision_graph.items():
4222
revision_graph[key] = tuple(parent for parent in parents if parent
4224
return revision_graph
4227
class StreamSink(object):
4228
"""An object that can insert a stream into a repository.
4230
This interface handles the complexity of reserialising inventories and
4231
revisions from different formats, and allows unidirectional insertion into
4232
stacked repositories without looking for the missing basis parents
4236
def __init__(self, target_repo):
4237
self.target_repo = target_repo
4239
def insert_stream(self, stream, src_format, resume_tokens):
4240
"""Insert a stream's content into the target repository.
4242
:param src_format: a bzr repository format.
4244
:return: a list of resume tokens and an iterable of keys additional
4245
items required before the insertion can be completed.
4247
self.target_repo.lock_write()
4250
self.target_repo.resume_write_group(resume_tokens)
4253
self.target_repo.start_write_group()
4256
# locked_insert_stream performs a commit|suspend.
4257
return self._locked_insert_stream(stream, src_format, is_resume)
4259
self.target_repo.abort_write_group(suppress_errors=True)
4262
self.target_repo.unlock()
4264
def _locked_insert_stream(self, stream, src_format, is_resume):
4265
to_serializer = self.target_repo._format._serializer
4266
src_serializer = src_format._serializer
4268
if to_serializer == src_serializer:
4269
# If serializers match and the target is a pack repository, set the
4270
# write cache size on the new pack. This avoids poor performance
4271
# on transports where append is unbuffered (such as
4272
# RemoteTransport). This is safe to do because nothing should read
4273
# back from the target repository while a stream with matching
4274
# serialization is being inserted.
4275
# The exception is that a delta record from the source that should
4276
# be a fulltext may need to be expanded by the target (see
4277
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4278
# explicitly flush any buffered writes first in that rare case.
4280
new_pack = self.target_repo._pack_collection._new_pack
4281
except AttributeError:
4282
# Not a pack repository
4285
new_pack.set_write_cache_size(1024*1024)
4286
for substream_type, substream in stream:
4287
if 'stream' in debug.debug_flags:
4288
mutter('inserting substream: %s', substream_type)
4289
if substream_type == 'texts':
4290
self.target_repo.texts.insert_record_stream(substream)
4291
elif substream_type == 'inventories':
4292
if src_serializer == to_serializer:
4293
self.target_repo.inventories.insert_record_stream(
4296
self._extract_and_insert_inventories(
4297
substream, src_serializer)
4298
elif substream_type == 'inventory-deltas':
4299
ui.ui_factory.warn_cross_format_fetch(src_format,
4300
self.target_repo._format)
4301
self._extract_and_insert_inventory_deltas(
4302
substream, src_serializer)
4303
elif substream_type == 'chk_bytes':
4304
# XXX: This doesn't support conversions, as it assumes the
4305
# conversion was done in the fetch code.
4306
self.target_repo.chk_bytes.insert_record_stream(substream)
4307
elif substream_type == 'revisions':
4308
# This may fallback to extract-and-insert more often than
4309
# required if the serializers are different only in terms of
4311
if src_serializer == to_serializer:
4312
self.target_repo.revisions.insert_record_stream(
4315
self._extract_and_insert_revisions(substream,
4317
elif substream_type == 'signatures':
4318
self.target_repo.signatures.insert_record_stream(substream)
4320
raise AssertionError('kaboom! %s' % (substream_type,))
4321
# Done inserting data, and the missing_keys calculations will try to
4322
# read back from the inserted data, so flush the writes to the new pack
4323
# (if this is pack format).
4324
if new_pack is not None:
4325
new_pack._write_data('', flush=True)
4326
# Find all the new revisions (including ones from resume_tokens)
4327
missing_keys = self.target_repo.get_missing_parent_inventories(
4328
check_for_missing_texts=is_resume)
4330
for prefix, versioned_file in (
4331
('texts', self.target_repo.texts),
4332
('inventories', self.target_repo.inventories),
4333
('revisions', self.target_repo.revisions),
4334
('signatures', self.target_repo.signatures),
4335
('chk_bytes', self.target_repo.chk_bytes),
4337
if versioned_file is None:
4339
# TODO: key is often going to be a StaticTuple object
4340
# I don't believe we can define a method by which
4341
# (prefix,) + StaticTuple will work, though we could
4342
# define a StaticTuple.sq_concat that would allow you to
4343
# pass in either a tuple or a StaticTuple as the second
4344
# object, so instead we could have:
4345
# StaticTuple(prefix) + key here...
4346
missing_keys.update((prefix,) + key for key in
4347
versioned_file.get_missing_compression_parent_keys())
4348
except NotImplementedError:
4349
# cannot even attempt suspending, and missing would have failed
4350
# during stream insertion.
4351
missing_keys = set()
4354
# suspend the write group and tell the caller what we is
4355
# missing. We know we can suspend or else we would not have
4356
# entered this code path. (All repositories that can handle
4357
# missing keys can handle suspending a write group).
4358
write_group_tokens = self.target_repo.suspend_write_group()
4359
return write_group_tokens, missing_keys
4360
hint = self.target_repo.commit_write_group()
4361
if (to_serializer != src_serializer and
4362
self.target_repo._format.pack_compresses):
4363
self.target_repo.pack(hint=hint)
4366
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4367
target_rich_root = self.target_repo._format.rich_root_data
4368
target_tree_refs = self.target_repo._format.supports_tree_reference
4369
for record in substream:
4370
# Insert the delta directly
4371
inventory_delta_bytes = record.get_bytes_as('fulltext')
4372
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4374
parse_result = deserialiser.parse_text_bytes(
4375
inventory_delta_bytes)
4376
except inventory_delta.IncompatibleInventoryDelta, err:
4377
trace.mutter("Incompatible delta: %s", err.msg)
4378
raise errors.IncompatibleRevision(self.target_repo._format)
4379
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4380
revision_id = new_id
4381
parents = [key[0] for key in record.parents]
4382
self.target_repo.add_inventory_by_delta(
4383
basis_id, inv_delta, revision_id, parents)
4385
def _extract_and_insert_inventories(self, substream, serializer,
4387
"""Generate a new inventory versionedfile in target, converting data.
4389
The inventory is retrieved from the source, (deserializing it), and
4390
stored in the target (reserializing it in a different format).
4392
target_rich_root = self.target_repo._format.rich_root_data
4393
target_tree_refs = self.target_repo._format.supports_tree_reference
4394
for record in substream:
4395
# It's not a delta, so it must be a fulltext in the source
4396
# serializer's format.
4397
bytes = record.get_bytes_as('fulltext')
4398
revision_id = record.key[0]
4399
inv = serializer.read_inventory_from_string(bytes, revision_id)
4400
parents = [key[0] for key in record.parents]
4401
self.target_repo.add_inventory(revision_id, inv, parents)
4402
# No need to keep holding this full inv in memory when the rest of
4403
# the substream is likely to be all deltas.
4406
def _extract_and_insert_revisions(self, substream, serializer):
4407
for record in substream:
4408
bytes = record.get_bytes_as('fulltext')
4409
revision_id = record.key[0]
4410
rev = serializer.read_revision_from_string(bytes)
4411
if rev.revision_id != revision_id:
4412
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4413
self.target_repo.add_revision(revision_id, rev)
4416
if self.target_repo._format._fetch_reconcile:
4417
self.target_repo.reconcile()
4420
class StreamSource(object):
4421
"""A source of a stream for fetching between repositories."""
4423
def __init__(self, from_repository, to_format):
4424
"""Create a StreamSource streaming from from_repository."""
4425
self.from_repository = from_repository
4426
self.to_format = to_format
4428
def delta_on_metadata(self):
4429
"""Return True if delta's are permitted on metadata streams.
4431
That is on revisions and signatures.
4433
src_serializer = self.from_repository._format._serializer
4434
target_serializer = self.to_format._serializer
4435
return (self.to_format._fetch_uses_deltas and
4436
src_serializer == target_serializer)
4438
def _fetch_revision_texts(self, revs):
4439
# fetch signatures first and then the revision texts
4440
# may need to be a InterRevisionStore call here.
4441
from_sf = self.from_repository.signatures
4442
# A missing signature is just skipped.
4443
keys = [(rev_id,) for rev_id in revs]
4444
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4446
self.to_format._fetch_order,
4447
not self.to_format._fetch_uses_deltas))
4448
# If a revision has a delta, this is actually expanded inside the
4449
# insert_record_stream code now, which is an alternate fix for
4451
from_rf = self.from_repository.revisions
4452
revisions = from_rf.get_record_stream(
4454
self.to_format._fetch_order,
4455
not self.delta_on_metadata())
4456
return [('signatures', signatures), ('revisions', revisions)]
4458
def _generate_root_texts(self, revs):
4459
"""This will be called by get_stream between fetching weave texts and
4460
fetching the inventory weave.
4462
if self._rich_root_upgrade():
4463
return _mod_fetch.Inter1and2Helper(
4464
self.from_repository).generate_root_texts(revs)
4468
def get_stream(self, search):
4470
revs = search.get_keys()
4471
graph = self.from_repository.get_graph()
4472
revs = tsort.topo_sort(graph.get_parent_map(revs))
4473
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4475
for knit_kind, file_id, revisions in data_to_fetch:
4476
if knit_kind != phase:
4478
# Make a new progress bar for this phase
4479
if knit_kind == "file":
4480
# Accumulate file texts
4481
text_keys.extend([(file_id, revision) for revision in
4483
elif knit_kind == "inventory":
4484
# Now copy the file texts.
4485
from_texts = self.from_repository.texts
4486
yield ('texts', from_texts.get_record_stream(
4487
text_keys, self.to_format._fetch_order,
4488
not self.to_format._fetch_uses_deltas))
4489
# Cause an error if a text occurs after we have done the
4492
# Before we process the inventory we generate the root
4493
# texts (if necessary) so that the inventories references
4495
for _ in self._generate_root_texts(revs):
4497
# we fetch only the referenced inventories because we do not
4498
# know for unselected inventories whether all their required
4499
# texts are present in the other repository - it could be
4501
for info in self._get_inventory_stream(revs):
4503
elif knit_kind == "signatures":
4504
# Nothing to do here; this will be taken care of when
4505
# _fetch_revision_texts happens.
4507
elif knit_kind == "revisions":
4508
for record in self._fetch_revision_texts(revs):
4511
raise AssertionError("Unknown knit kind %r" % knit_kind)
4513
def get_stream_for_missing_keys(self, missing_keys):
4514
# missing keys can only occur when we are byte copying and not
4515
# translating (because translation means we don't send
4516
# unreconstructable deltas ever).
4518
keys['texts'] = set()
4519
keys['revisions'] = set()
4520
keys['inventories'] = set()
4521
keys['chk_bytes'] = set()
4522
keys['signatures'] = set()
4523
for key in missing_keys:
4524
keys[key[0]].add(key[1:])
4525
if len(keys['revisions']):
4526
# If we allowed copying revisions at this point, we could end up
4527
# copying a revision without copying its required texts: a
4528
# violation of the requirements for repository integrity.
4529
raise AssertionError(
4530
'cannot copy revisions to fill in missing deltas %s' % (
4531
keys['revisions'],))
4532
for substream_kind, keys in keys.iteritems():
4533
vf = getattr(self.from_repository, substream_kind)
4534
if vf is None and keys:
4535
raise AssertionError(
4536
"cannot fill in keys for a versioned file we don't"
4537
" have: %s needs %s" % (substream_kind, keys))
4539
# No need to stream something we don't have
4541
if substream_kind == 'inventories':
4542
# Some missing keys are genuinely ghosts, filter those out.
4543
present = self.from_repository.inventories.get_parent_map(keys)
4544
revs = [key[0] for key in present]
4545
# Get the inventory stream more-or-less as we do for the
4546
# original stream; there's no reason to assume that records
4547
# direct from the source will be suitable for the sink. (Think
4548
# e.g. 2a -> 1.9-rich-root).
4549
for info in self._get_inventory_stream(revs, missing=True):
4553
# Ask for full texts always so that we don't need more round trips
4554
# after this stream.
4555
# Some of the missing keys are genuinely ghosts, so filter absent
4556
# records. The Sink is responsible for doing another check to
4557
# ensure that ghosts don't introduce missing data for future
4559
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4560
self.to_format._fetch_order, True))
4561
yield substream_kind, stream
4563
def inventory_fetch_order(self):
4564
if self._rich_root_upgrade():
4565
return 'topological'
4567
return self.to_format._fetch_order
4569
def _rich_root_upgrade(self):
4570
return (not self.from_repository._format.rich_root_data and
4571
self.to_format.rich_root_data)
4573
def _get_inventory_stream(self, revision_ids, missing=False):
4574
from_format = self.from_repository._format
4575
if (from_format.supports_chks and self.to_format.supports_chks and
4576
from_format.network_name() == self.to_format.network_name()):
4577
raise AssertionError(
4578
"this case should be handled by GroupCHKStreamSource")
4579
elif 'forceinvdeltas' in debug.debug_flags:
4580
return self._get_convertable_inventory_stream(revision_ids,
4581
delta_versus_null=missing)
4582
elif from_format.network_name() == self.to_format.network_name():
4584
return self._get_simple_inventory_stream(revision_ids,
4586
elif (not from_format.supports_chks and not self.to_format.supports_chks
4587
and from_format._serializer == self.to_format._serializer):
4588
# Essentially the same format.
4589
return self._get_simple_inventory_stream(revision_ids,
4592
# Any time we switch serializations, we want to use an
4593
# inventory-delta based approach.
4594
return self._get_convertable_inventory_stream(revision_ids,
4595
delta_versus_null=missing)
4597
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4598
# NB: This currently reopens the inventory weave in source;
4599
# using a single stream interface instead would avoid this.
4600
from_weave = self.from_repository.inventories
4602
delta_closure = True
4604
delta_closure = not self.delta_on_metadata()
4605
yield ('inventories', from_weave.get_record_stream(
4606
[(rev_id,) for rev_id in revision_ids],
4607
self.inventory_fetch_order(), delta_closure))
4609
def _get_convertable_inventory_stream(self, revision_ids,
4610
delta_versus_null=False):
4611
# The two formats are sufficiently different that there is no fast
4612
# path, so we need to send just inventorydeltas, which any
4613
# sufficiently modern client can insert into any repository.
4614
# The StreamSink code expects to be able to
4615
# convert on the target, so we need to put bytes-on-the-wire that can
4616
# be converted. That means inventory deltas (if the remote is <1.19,
4617
# RemoteStreamSink will fallback to VFS to insert the deltas).
4618
yield ('inventory-deltas',
4619
self._stream_invs_as_deltas(revision_ids,
4620
delta_versus_null=delta_versus_null))
4622
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4623
"""Return a stream of inventory-deltas for the given rev ids.
4625
:param revision_ids: The list of inventories to transmit
4626
:param delta_versus_null: Don't try to find a minimal delta for this
4627
entry, instead compute the delta versus the NULL_REVISION. This
4628
effectively streams a complete inventory. Used for stuff like
4629
filling in missing parents, etc.
4631
from_repo = self.from_repository
4632
revision_keys = [(rev_id,) for rev_id in revision_ids]
4633
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4634
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4636
inventories = self.from_repository.iter_inventories(
4637
revision_ids, 'topological')
4638
format = from_repo._format
4639
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4640
inventory_cache = lru_cache.LRUCache(50)
4641
null_inventory = from_repo.revision_tree(
4642
_mod_revision.NULL_REVISION).inventory
4643
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4644
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4645
# repo back into a non-rich-root repo ought to be allowed)
4646
serializer = inventory_delta.InventoryDeltaSerializer(
4647
versioned_root=format.rich_root_data,
4648
tree_references=format.supports_tree_reference)
4649
for inv in inventories:
4650
key = (inv.revision_id,)
4651
parent_keys = parent_map.get(key, ())
4653
if not delta_versus_null and parent_keys:
4654
# The caller did not ask for complete inventories and we have
4655
# some parents that we can delta against. Make a delta against
4656
# each parent so that we can find the smallest.
4657
parent_ids = [parent_key[0] for parent_key in parent_keys]
4658
for parent_id in parent_ids:
4659
if parent_id not in invs_sent_so_far:
4660
# We don't know that the remote side has this basis, so
4663
if parent_id == _mod_revision.NULL_REVISION:
4664
parent_inv = null_inventory
4666
parent_inv = inventory_cache.get(parent_id, None)
4667
if parent_inv is None:
4668
parent_inv = from_repo.get_inventory(parent_id)
4669
candidate_delta = inv._make_delta(parent_inv)
4670
if (delta is None or
4671
len(delta) > len(candidate_delta)):
4672
delta = candidate_delta
4673
basis_id = parent_id
4675
# Either none of the parents ended up being suitable, or we
4676
# were asked to delta against NULL
4677
basis_id = _mod_revision.NULL_REVISION
4678
delta = inv._make_delta(null_inventory)
4679
invs_sent_so_far.add(inv.revision_id)
4680
inventory_cache[inv.revision_id] = inv
4681
delta_serialized = ''.join(
4682
serializer.delta_to_lines(basis_id, key[-1], delta))
4683
yield versionedfile.FulltextContentFactory(
4684
key, parent_keys, None, delta_serialized)
4687
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4688
stop_revision=None):
4689
"""Extend the partial history to include a given index
4691
If a stop_index is supplied, stop when that index has been reached.
4692
If a stop_revision is supplied, stop when that revision is
4693
encountered. Otherwise, stop when the beginning of history is
4696
:param stop_index: The index which should be present. When it is
4697
present, history extension will stop.
4698
:param stop_revision: The revision id which should be present. When
4699
it is encountered, history extension will stop.
4701
start_revision = partial_history_cache[-1]
4702
iterator = repo.iter_reverse_revision_history(start_revision)
4704
#skip the last revision in the list
4707
if (stop_index is not None and
4708
len(partial_history_cache) > stop_index):
4710
if partial_history_cache[-1] == stop_revision:
4712
revision_id = iterator.next()
4713
partial_history_cache.append(revision_id)
4714
except StopIteration:
3123
unused_versions = set()
3124
for num, revision_id in enumerate(weave.versions()):
3126
correct_parents = self.calculate_file_version_parents(
3127
revision_id, file_id)
3129
# The version is not part of the used keys.
3130
unused_versions.add(revision_id)
3133
knit_parents = tuple(weave.get_parents(revision_id))
3134
except errors.RevisionNotPresent:
3136
if correct_parents != knit_parents:
3137
wrong_parents[revision_id] = (knit_parents, correct_parents)
3138
return wrong_parents, unused_versions
added
removed
bzrlib/repository.py
