1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from cStringIO import StringIO
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
39
revision as _mod_revision,
44
from bzrlib.bundle import serializer
45
from bzrlib.revisiontree import RevisionTree
46
from bzrlib.store.versioned import VersionedFileStore
47
from bzrlib.store.text import TextStore
48
from bzrlib.testament import Testament
51
from bzrlib.decorators import needs_read_lock, needs_write_lock
52
from bzrlib.inter import InterObject
53
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
54
from bzrlib.symbol_versioning import (
57
from bzrlib.trace import mutter, mutter_callsite, note, warning
60
# Old formats display a warning, but only once
61
_deprecation_warning_done = False
64
class CommitBuilder(object):
65
"""Provides an interface to build up a commit.
67
This allows describing a tree to be committed without needing to
68
know the internals of the format of the repository.
71
# all clients should supply tree roots.
72
record_root_entry = True
73
# the default CommitBuilder does not manage trees whose root is versioned.
74
_versioned_root = False
76
def __init__(self, repository, parents, config, timestamp=None,
77
timezone=None, committer=None, revprops=None,
79
"""Initiate a CommitBuilder.
81
:param repository: Repository to commit to.
82
:param parents: Revision ids of the parents of the new revision.
83
:param config: Configuration to use.
84
:param timestamp: Optional timestamp recorded for commit.
85
:param timezone: Optional timezone for timestamp.
86
:param committer: Optional committer to set for commit.
87
:param revprops: Optional dictionary of revision properties.
88
:param revision_id: Optional revision id.
93
self._committer = self._config.username()
95
assert isinstance(committer, basestring), type(committer)
96
self._committer = committer
98
self.new_inventory = Inventory(None)
99
self._new_revision_id = revision_id
100
self.parents = parents
101
self.repository = repository
104
if revprops is not None:
105
self._revprops.update(revprops)
107
if timestamp is None:
108
timestamp = time.time()
109
# Restrict resolution to 1ms
110
self._timestamp = round(timestamp, 3)
113
self._timezone = osutils.local_time_offset()
115
self._timezone = int(timezone)
117
self._generate_revision_if_needed()
118
self._repo_graph = repository.get_graph()
120
def commit(self, message):
121
"""Make the actual commit.
123
:return: The revision id of the recorded revision.
125
rev = _mod_revision.Revision(
126
timestamp=self._timestamp,
127
timezone=self._timezone,
128
committer=self._committer,
130
inventory_sha1=self.inv_sha1,
131
revision_id=self._new_revision_id,
132
properties=self._revprops)
133
rev.parent_ids = self.parents
134
self.repository.add_revision(self._new_revision_id, rev,
135
self.new_inventory, self._config)
136
self.repository.commit_write_group()
137
return self._new_revision_id
140
"""Abort the commit that is being built.
142
self.repository.abort_write_group()
144
def revision_tree(self):
145
"""Return the tree that was just committed.
147
After calling commit() this can be called to get a RevisionTree
148
representing the newly committed tree. This is preferred to
149
calling Repository.revision_tree() because that may require
150
deserializing the inventory, while we already have a copy in
153
return RevisionTree(self.repository, self.new_inventory,
154
self._new_revision_id)
156
def finish_inventory(self):
157
"""Tell the builder that the inventory is finished."""
158
if self.new_inventory.root is None:
159
raise AssertionError('Root entry should be supplied to'
160
' record_entry_contents, as of bzr 0.10.',
161
DeprecationWarning, stacklevel=2)
162
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
163
self.new_inventory.revision_id = self._new_revision_id
164
self.inv_sha1 = self.repository.add_inventory(
165
self._new_revision_id,
170
def _gen_revision_id(self):
171
"""Return new revision-id."""
172
return generate_ids.gen_revision_id(self._config.username(),
175
def _generate_revision_if_needed(self):
176
"""Create a revision id if None was supplied.
178
If the repository can not support user-specified revision ids
179
they should override this function and raise CannotSetRevisionId
180
if _new_revision_id is not None.
182
:raises: CannotSetRevisionId
184
if self._new_revision_id is None:
185
self._new_revision_id = self._gen_revision_id()
186
self.random_revid = True
188
self.random_revid = False
190
def _check_root(self, ie, parent_invs, tree):
191
"""Helper for record_entry_contents.
193
:param ie: An entry being added.
194
:param parent_invs: The inventories of the parent revisions of the
196
:param tree: The tree that is being committed.
198
# In this revision format, root entries have no knit or weave When
199
# serializing out to disk and back in root.revision is always
201
ie.revision = self._new_revision_id
203
def _get_delta(self, ie, basis_inv, path):
204
"""Get a delta against the basis inventory for ie."""
205
if ie.file_id not in basis_inv:
207
return (None, path, ie.file_id, ie)
208
elif ie != basis_inv[ie.file_id]:
210
# TODO: avoid tis id2path call.
211
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
216
def record_entry_contents(self, ie, parent_invs, path, tree,
218
"""Record the content of ie from tree into the commit if needed.
220
Side effect: sets ie.revision when unchanged
222
:param ie: An inventory entry present in the commit.
223
:param parent_invs: The inventories of the parent revisions of the
225
:param path: The path the entry is at in the tree.
226
:param tree: The tree which contains this entry and should be used to
228
:param content_summary: Summary data from the tree about the paths
229
content - stat, length, exec, sha/link target. This is only
230
accessed when the entry has a revision of None - that is when it is
231
a candidate to commit.
232
:return: A tuple (change_delta, version_recorded). change_delta is
233
an inventory_delta change for this entry against the basis tree of
234
the commit, or None if no change occured against the basis tree.
235
version_recorded is True if a new version of the entry has been
236
recorded. For instance, committing a merge where a file was only
237
changed on the other side will return (delta, False).
239
if self.new_inventory.root is None:
240
if ie.parent_id is not None:
241
raise errors.RootMissing()
242
self._check_root(ie, parent_invs, tree)
243
if ie.revision is None:
244
kind = content_summary[0]
246
# ie is carried over from a prior commit
248
# XXX: repository specific check for nested tree support goes here - if
249
# the repo doesn't want nested trees we skip it ?
250
if (kind == 'tree-reference' and
251
not self.repository._format.supports_tree_reference):
252
# mismatch between commit builder logic and repository:
253
# this needs the entry creation pushed down into the builder.
254
raise NotImplementedError('Missing repository subtree support.')
255
self.new_inventory.add(ie)
257
# TODO: slow, take it out of the inner loop.
259
basis_inv = parent_invs[0]
261
basis_inv = Inventory(root_id=None)
263
# ie.revision is always None if the InventoryEntry is considered
264
# for committing. We may record the previous parents revision if the
265
# content is actually unchanged against a sole head.
266
if ie.revision is not None:
267
if not self._versioned_root and path == '':
268
# repositories that do not version the root set the root's
269
# revision to the new commit even when no change occurs, and
270
# this masks when a change may have occurred against the basis,
271
# so calculate if one happened.
272
if ie.file_id in basis_inv:
273
delta = (basis_inv.id2path(ie.file_id), path,
277
delta = (None, path, ie.file_id, ie)
280
# we don't need to commit this, because the caller already
281
# determined that an existing revision of this file is
283
return None, (ie.revision == self._new_revision_id)
284
# XXX: Friction: parent_candidates should return a list not a dict
285
# so that we don't have to walk the inventories again.
286
parent_candiate_entries = ie.parent_candidates(parent_invs)
287
head_set = self._repo_graph.heads(parent_candiate_entries.keys())
289
for inv in parent_invs:
290
if ie.file_id in inv:
291
old_rev = inv[ie.file_id].revision
292
if old_rev in head_set:
293
heads.append(inv[ie.file_id].revision)
294
head_set.remove(inv[ie.file_id].revision)
297
# now we check to see if we need to write a new record to the
299
# We write a new entry unless there is one head to the ancestors, and
300
# the kind-derived content is unchanged.
302
# Cheapest check first: no ancestors, or more the one head in the
303
# ancestors, we write a new node.
307
# There is a single head, look it up for comparison
308
parent_entry = parent_candiate_entries[heads[0]]
309
# if the non-content specific data has changed, we'll be writing a
311
if (parent_entry.parent_id != ie.parent_id or
312
parent_entry.name != ie.name):
314
# now we need to do content specific checks:
316
# if the kind changed the content obviously has
317
if kind != parent_entry.kind:
321
if (# if the file length changed we have to store:
322
parent_entry.text_size != content_summary[1] or
323
# if the exec bit has changed we have to store:
324
parent_entry.executable != content_summary[2]):
326
elif parent_entry.text_sha1 == content_summary[3]:
327
# all meta and content is unchanged (using a hash cache
328
# hit to check the sha)
329
ie.revision = parent_entry.revision
330
ie.text_size = parent_entry.text_size
331
ie.text_sha1 = parent_entry.text_sha1
332
ie.executable = parent_entry.executable
333
return self._get_delta(ie, basis_inv, path), False
335
# Either there is only a hash change(no hash cache entry,
336
# or same size content change), or there is no change on
338
# Provide the parent's hash to the store layer, so that the
339
# content is unchanged we will not store a new node.
340
nostore_sha = parent_entry.text_sha1
342
# We want to record a new node regardless of the presence or
343
# absence of a content change in the file.
345
ie.executable = content_summary[2]
346
lines = tree.get_file(ie.file_id, path).readlines()
348
ie.text_sha1, ie.text_size = self._add_text_to_weave(
349
ie.file_id, lines, heads, nostore_sha)
350
except errors.ExistingContent:
351
# Turns out that the file content was unchanged, and we were
352
# only going to store a new node if it was changed. Carry over
354
ie.revision = parent_entry.revision
355
ie.text_size = parent_entry.text_size
356
ie.text_sha1 = parent_entry.text_sha1
357
ie.executable = parent_entry.executable
358
return self._get_delta(ie, basis_inv, path), False
359
elif kind == 'directory':
361
# all data is meta here, nothing specific to directory, so
363
ie.revision = parent_entry.revision
364
return self._get_delta(ie, basis_inv, path), False
366
self._add_text_to_weave(ie.file_id, lines, heads, None)
367
elif kind == 'symlink':
368
current_link_target = content_summary[3]
370
# symlink target is not generic metadata, check if it has
372
if current_link_target != parent_entry.symlink_target:
375
# unchanged, carry over.
376
ie.revision = parent_entry.revision
377
ie.symlink_target = parent_entry.symlink_target
378
return self._get_delta(ie, basis_inv, path), False
379
ie.symlink_target = current_link_target
381
self._add_text_to_weave(ie.file_id, lines, heads, None)
382
elif kind == 'tree-reference':
384
if content_summary[3] != parent_entry.reference_revision:
387
# unchanged, carry over.
388
ie.reference_revision = parent_entry.reference_revision
389
ie.revision = parent_entry.revision
390
return self._get_delta(ie, basis_inv, path), False
391
ie.reference_revision = content_summary[3]
393
self._add_text_to_weave(ie.file_id, lines, heads, None)
395
raise NotImplementedError('unknown kind')
396
ie.revision = self._new_revision_id
397
return self._get_delta(ie, basis_inv, path), True
399
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
400
versionedfile = self.repository.weave_store.get_weave_or_empty(
401
file_id, self.repository.get_transaction())
402
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
403
# than add_lines, and allows committing when a parent is ghosted for
405
# Note: as we read the content directly from the tree, we know its not
406
# been turned into unicode or badly split - but a broken tree
407
# implementation could give us bad output from readlines() so this is
408
# not a guarantee of safety. What would be better is always checking
409
# the content during test suite execution. RBC 20070912
411
return versionedfile.add_lines_with_ghosts(
412
self._new_revision_id, parents, new_lines,
413
nostore_sha=nostore_sha, random_id=self.random_revid,
414
check_content=False)[0:2]
416
versionedfile.clear_cache()
419
class RootCommitBuilder(CommitBuilder):
420
"""This commitbuilder actually records the root id"""
422
# the root entry gets versioned properly by this builder.
423
_versioned_root = True
425
def _check_root(self, ie, parent_invs, tree):
426
"""Helper for record_entry_contents.
428
:param ie: An entry being added.
429
:param parent_invs: The inventories of the parent revisions of the
431
:param tree: The tree that is being committed.
435
######################################################################
438
class Repository(object):
439
"""Repository holding history for one or more branches.
441
The repository holds and retrieves historical information including
442
revisions and file history. It's normally accessed only by the Branch,
443
which views a particular line of development through that history.
445
The Repository builds on top of Stores and a Transport, which respectively
446
describe the disk data format and the way of accessing the (possibly
450
# What class to use for a CommitBuilder. Often its simpler to change this
451
# in a Repository class subclass rather than to override
452
# get_commit_builder.
453
_commit_builder_class = CommitBuilder
454
# The search regex used by xml based repositories to determine what things
455
# where changed in a single commit.
456
_file_ids_altered_regex = lazy_regex.lazy_compile(
457
r'file_id="(?P<file_id>[^"]+)"'
458
r'.* revision="(?P<revision_id>[^"]+)"'
461
def abort_write_group(self):
462
"""Commit the contents accrued within the current write group.
464
:seealso: start_write_group.
466
if self._write_group is not self.get_transaction():
467
# has an unlock or relock occured ?
468
raise errors.BzrError('mismatched lock context and write group.')
469
self._abort_write_group()
470
self._write_group = None
472
def _abort_write_group(self):
473
"""Template method for per-repository write group cleanup.
475
This is called during abort before the write group is considered to be
476
finished and should cleanup any internal state accrued during the write
477
group. There is no requirement that data handed to the repository be
478
*not* made available - this is not a rollback - but neither should any
479
attempt be made to ensure that data added is fully commited. Abort is
480
invoked when an error has occured so futher disk or network operations
481
may not be possible or may error and if possible should not be
486
def add_inventory(self, revision_id, inv, parents):
487
"""Add the inventory inv to the repository as revision_id.
489
:param parents: The revision ids of the parents that revision_id
490
is known to have and are in the repository already.
492
returns the sha1 of the serialized inventory.
494
_mod_revision.check_not_reserved_id(revision_id)
495
assert inv.revision_id is None or inv.revision_id == revision_id, \
496
"Mismatch between inventory revision" \
497
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
498
assert inv.root is not None
499
inv_lines = self._serialise_inventory_to_lines(inv)
500
inv_vf = self.get_inventory_weave()
501
return self._inventory_add_lines(inv_vf, revision_id, parents,
502
inv_lines, check_content=False)
504
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
506
"""Store lines in inv_vf and return the sha1 of the inventory."""
508
for parent in parents:
510
final_parents.append(parent)
511
return inv_vf.add_lines(revision_id, final_parents, lines,
512
check_content=check_content)[0]
515
def add_revision(self, revision_id, rev, inv=None, config=None):
516
"""Add rev to the revision store as revision_id.
518
:param revision_id: the revision id to use.
519
:param rev: The revision object.
520
:param inv: The inventory for the revision. if None, it will be looked
521
up in the inventory storer
522
:param config: If None no digital signature will be created.
523
If supplied its signature_needed method will be used
524
to determine if a signature should be made.
526
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
528
_mod_revision.check_not_reserved_id(revision_id)
529
if config is not None and config.signature_needed():
531
inv = self.get_inventory(revision_id)
532
plaintext = Testament(rev, inv).as_short_text()
533
self.store_revision_signature(
534
gpg.GPGStrategy(config), plaintext, revision_id)
535
if not revision_id in self.get_inventory_weave():
537
raise errors.WeaveRevisionNotPresent(revision_id,
538
self.get_inventory_weave())
540
# yes, this is not suitable for adding with ghosts.
541
self.add_inventory(revision_id, inv, rev.parent_ids)
542
self._revision_store.add_revision(rev, self.get_transaction())
544
def _add_revision_text(self, revision_id, text):
545
revision = self._revision_store._serializer.read_revision_from_string(
547
self._revision_store._add_revision(revision, StringIO(text),
548
self.get_transaction())
550
def all_revision_ids(self):
551
"""Returns a list of all the revision ids in the repository.
553
This is deprecated because code should generally work on the graph
554
reachable from a particular revision, and ignore any other revisions
555
that might be present. There is no direct replacement method.
557
if 'evil' in debug.debug_flags:
558
mutter_callsite(2, "all_revision_ids is linear with history.")
559
return self._all_revision_ids()
561
def _all_revision_ids(self):
562
"""Returns a list of all the revision ids in the repository.
564
These are in as much topological order as the underlying store can
567
raise NotImplementedError(self._all_revision_ids)
569
def break_lock(self):
570
"""Break a lock if one is present from another instance.
572
Uses the ui factory to ask for confirmation if the lock may be from
575
self.control_files.break_lock()
578
def _eliminate_revisions_not_present(self, revision_ids):
579
"""Check every revision id in revision_ids to see if we have it.
581
Returns a set of the present revisions.
584
for id in revision_ids:
585
if self.has_revision(id):
590
def create(a_bzrdir):
591
"""Construct the current default format repository in a_bzrdir."""
592
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
594
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
595
"""instantiate a Repository.
597
:param _format: The format of the repository on disk.
598
:param a_bzrdir: The BzrDir of the repository.
600
In the future we will have a single api for all stores for
601
getting file texts, inventories and revisions, then
602
this construct will accept instances of those things.
604
super(Repository, self).__init__()
605
self._format = _format
606
# the following are part of the public API for Repository:
607
self.bzrdir = a_bzrdir
608
self.control_files = control_files
609
self._revision_store = _revision_store
610
# backwards compatibility
611
self.weave_store = text_store
613
self._reconcile_does_inventory_gc = True
614
# not right yet - should be more semantically clear ?
616
self.control_store = control_store
617
self.control_weaves = control_store
618
# TODO: make sure to construct the right store classes, etc, depending
619
# on whether escaping is required.
620
self._warn_if_deprecated()
621
self._write_group = None
622
self.base = control_files._transport.base
625
return '%s(%r)' % (self.__class__.__name__,
628
def has_same_location(self, other):
629
"""Returns a boolean indicating if this repository is at the same
630
location as another repository.
632
This might return False even when two repository objects are accessing
633
the same physical repository via different URLs.
635
if self.__class__ is not other.__class__:
637
return (self.control_files._transport.base ==
638
other.control_files._transport.base)
640
def is_in_write_group(self):
641
"""Return True if there is an open write group.
643
:seealso: start_write_group.
645
return self._write_group is not None
648
return self.control_files.is_locked()
650
def lock_write(self, token=None):
651
"""Lock this repository for writing.
653
This causes caching within the repository obejct to start accumlating
654
data during reads, and allows a 'write_group' to be obtained. Write
655
groups must be used for actual data insertion.
657
:param token: if this is already locked, then lock_write will fail
658
unless the token matches the existing lock.
659
:returns: a token if this instance supports tokens, otherwise None.
660
:raises TokenLockingNotSupported: when a token is given but this
661
instance doesn't support using token locks.
662
:raises MismatchedToken: if the specified token doesn't match the token
663
of the existing lock.
664
:seealso: start_write_group.
666
A token should be passed in if you know that you have locked the object
667
some other way, and need to synchronise this object's state with that
670
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
672
result = self.control_files.lock_write(token=token)
677
self.control_files.lock_read()
680
def get_physical_lock_status(self):
681
return self.control_files.get_physical_lock_status()
683
def leave_lock_in_place(self):
684
"""Tell this repository not to release the physical lock when this
687
If lock_write doesn't return a token, then this method is not supported.
689
self.control_files.leave_in_place()
691
def dont_leave_lock_in_place(self):
692
"""Tell this repository to release the physical lock when this
693
object is unlocked, even if it didn't originally acquire it.
695
If lock_write doesn't return a token, then this method is not supported.
697
self.control_files.dont_leave_in_place()
700
def gather_stats(self, revid=None, committers=None):
701
"""Gather statistics from a revision id.
703
:param revid: The revision id to gather statistics from, if None, then
704
no revision specific statistics are gathered.
705
:param committers: Optional parameter controlling whether to grab
706
a count of committers from the revision specific statistics.
707
:return: A dictionary of statistics. Currently this contains:
708
committers: The number of committers if requested.
709
firstrev: A tuple with timestamp, timezone for the penultimate left
710
most ancestor of revid, if revid is not the NULL_REVISION.
711
latestrev: A tuple with timestamp, timezone for revid, if revid is
712
not the NULL_REVISION.
713
revisions: The total revision count in the repository.
714
size: An estimate disk size of the repository in bytes.
717
if revid and committers:
718
result['committers'] = 0
719
if revid and revid != _mod_revision.NULL_REVISION:
721
all_committers = set()
722
revisions = self.get_ancestry(revid)
723
# pop the leading None
725
first_revision = None
727
# ignore the revisions in the middle - just grab first and last
728
revisions = revisions[0], revisions[-1]
729
for revision in self.get_revisions(revisions):
730
if not first_revision:
731
first_revision = revision
733
all_committers.add(revision.committer)
734
last_revision = revision
736
result['committers'] = len(all_committers)
737
result['firstrev'] = (first_revision.timestamp,
738
first_revision.timezone)
739
result['latestrev'] = (last_revision.timestamp,
740
last_revision.timezone)
742
# now gather global repository information
743
if self.bzrdir.root_transport.listable():
744
c, t = self._revision_store.total_size(self.get_transaction())
745
result['revisions'] = c
750
def missing_revision_ids(self, other, revision_id=None):
751
"""Return the revision ids that other has that this does not.
753
These are returned in topological order.
755
revision_id: only return revision ids included by revision_id.
757
return InterRepository.get(other, self).missing_revision_ids(revision_id)
761
"""Open the repository rooted at base.
763
For instance, if the repository is at URL/.bzr/repository,
764
Repository.open(URL) -> a Repository instance.
766
control = bzrdir.BzrDir.open(base)
767
return control.open_repository()
769
def copy_content_into(self, destination, revision_id=None):
770
"""Make a complete copy of the content in self into destination.
772
This is a destructive operation! Do not use it on existing
775
return InterRepository.get(self, destination).copy_content(revision_id)
777
def commit_write_group(self):
778
"""Commit the contents accrued within the current write group.
780
:seealso: start_write_group.
782
if self._write_group is not self.get_transaction():
783
# has an unlock or relock occured ?
784
raise errors.BzrError('mismatched lock context %r and '
786
(self.get_transaction(), self._write_group))
787
self._commit_write_group()
788
self._write_group = None
790
def _commit_write_group(self):
791
"""Template method for per-repository write group cleanup.
793
This is called before the write group is considered to be
794
finished and should ensure that all data handed to the repository
795
for writing during the write group is safely committed (to the
796
extent possible considering file system caching etc).
799
def fetch(self, source, revision_id=None, pb=None):
800
"""Fetch the content required to construct revision_id from source.
802
If revision_id is None all content is copied.
804
# fast path same-url fetch operations
805
if self.has_same_location(source):
806
# check that last_revision is in 'from' and then return a
808
if (revision_id is not None and
809
not _mod_revision.is_null(revision_id)):
810
self.get_revision(revision_id)
812
inter = InterRepository.get(source, self)
814
return inter.fetch(revision_id=revision_id, pb=pb)
815
except NotImplementedError:
816
raise errors.IncompatibleRepositories(source, self)
818
def create_bundle(self, target, base, fileobj, format=None):
819
return serializer.write_bundle(self, target, base, fileobj, format)
821
def get_commit_builder(self, branch, parents, config, timestamp=None,
822
timezone=None, committer=None, revprops=None,
824
"""Obtain a CommitBuilder for this repository.
826
:param branch: Branch to commit to.
827
:param parents: Revision ids of the parents of the new revision.
828
:param config: Configuration to use.
829
:param timestamp: Optional timestamp recorded for commit.
830
:param timezone: Optional timezone for timestamp.
831
:param committer: Optional committer to set for commit.
832
:param revprops: Optional dictionary of revision properties.
833
:param revision_id: Optional revision id.
835
result = self._commit_builder_class(self, parents, config,
836
timestamp, timezone, committer, revprops, revision_id)
837
self.start_write_group()
841
if (self.control_files._lock_count == 1 and
842
self.control_files._lock_mode == 'w'):
843
if self._write_group is not None:
844
raise errors.BzrError(
845
'Must end write groups before releasing write locks.')
846
self.control_files.unlock()
849
def clone(self, a_bzrdir, revision_id=None):
850
"""Clone this repository into a_bzrdir using the current format.
852
Currently no check is made that the format of this repository and
853
the bzrdir format are compatible. FIXME RBC 20060201.
855
:return: The newly created destination repository.
857
# TODO: deprecate after 0.16; cloning this with all its settings is
858
# probably not very useful -- mbp 20070423
859
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
860
self.copy_content_into(dest_repo, revision_id)
863
def start_write_group(self):
864
"""Start a write group in the repository.
866
Write groups are used by repositories which do not have a 1:1 mapping
867
between file ids and backend store to manage the insertion of data from
868
both fetch and commit operations.
870
A write lock is required around the start_write_group/commit_write_group
871
for the support of lock-requiring repository formats.
873
One can only insert data into a repository inside a write group.
877
if not self.is_locked() or self.control_files._lock_mode != 'w':
878
raise errors.NotWriteLocked(self)
879
if self._write_group:
880
raise errors.BzrError('already in a write group')
881
self._start_write_group()
882
# so we can detect unlock/relock - the write group is now entered.
883
self._write_group = self.get_transaction()
885
def _start_write_group(self):
886
"""Template method for per-repository write group startup.
888
This is called before the write group is considered to be
893
def sprout(self, to_bzrdir, revision_id=None):
894
"""Create a descendent repository for new development.
896
Unlike clone, this does not copy the settings of the repository.
898
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
899
dest_repo.fetch(self, revision_id=revision_id)
902
def _create_sprouting_repo(self, a_bzrdir, shared):
903
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
904
# use target default format.
905
dest_repo = a_bzrdir.create_repository()
907
# Most control formats need the repository to be specifically
908
# created, but on some old all-in-one formats it's not needed
910
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
911
except errors.UninitializableFormat:
912
dest_repo = a_bzrdir.open_repository()
916
def has_revision(self, revision_id):
917
"""True if this repository has a copy of the revision."""
918
if 'evil' in debug.debug_flags:
919
mutter_callsite(3, "has_revision is a LBYL symptom.")
920
return self._revision_store.has_revision_id(revision_id,
921
self.get_transaction())
924
def get_revision(self, revision_id):
925
"""Return the Revision object for a named revision."""
926
return self.get_revisions([revision_id])[0]
929
def get_revision_reconcile(self, revision_id):
930
"""'reconcile' helper routine that allows access to a revision always.
932
This variant of get_revision does not cross check the weave graph
933
against the revision one as get_revision does: but it should only
934
be used by reconcile, or reconcile-alike commands that are correcting
935
or testing the revision graph.
937
return self._get_revisions([revision_id])[0]
940
def get_revisions(self, revision_ids):
941
"""Get many revisions at once."""
942
return self._get_revisions(revision_ids)
945
def _get_revisions(self, revision_ids):
946
"""Core work logic to get many revisions without sanity checks."""
947
for rev_id in revision_ids:
948
if not rev_id or not isinstance(rev_id, basestring):
949
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
950
revs = self._revision_store.get_revisions(revision_ids,
951
self.get_transaction())
953
assert not isinstance(rev.revision_id, unicode)
954
for parent_id in rev.parent_ids:
955
assert not isinstance(parent_id, unicode)
959
def get_revision_xml(self, revision_id):
960
# TODO: jam 20070210 This shouldn't be necessary since get_revision
961
# would have already do it.
962
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
963
rev = self.get_revision(revision_id)
965
# the current serializer..
966
self._revision_store._serializer.write_revision(rev, rev_tmp)
968
return rev_tmp.getvalue()
971
def get_deltas_for_revisions(self, revisions):
972
"""Produce a generator of revision deltas.
974
Note that the input is a sequence of REVISIONS, not revision_ids.
975
Trees will be held in memory until the generator exits.
976
Each delta is relative to the revision's lefthand predecessor.
978
required_trees = set()
979
for revision in revisions:
980
required_trees.add(revision.revision_id)
981
required_trees.update(revision.parent_ids[:1])
982
trees = dict((t.get_revision_id(), t) for
983
t in self.revision_trees(required_trees))
984
for revision in revisions:
985
if not revision.parent_ids:
986
old_tree = self.revision_tree(None)
988
old_tree = trees[revision.parent_ids[0]]
989
yield trees[revision.revision_id].changes_from(old_tree)
992
def get_revision_delta(self, revision_id):
993
"""Return the delta for one revision.
995
The delta is relative to the left-hand predecessor of the
998
r = self.get_revision(revision_id)
999
return list(self.get_deltas_for_revisions([r]))[0]
1002
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1003
signature = gpg_strategy.sign(plaintext)
1004
self._revision_store.add_revision_signature_text(revision_id,
1006
self.get_transaction())
1008
def fileids_altered_by_revision_ids(self, revision_ids):
1009
"""Find the file ids and versions affected by revisions.
1011
:param revisions: an iterable containing revision ids.
1012
:return: a dictionary mapping altered file-ids to an iterable of
1013
revision_ids. Each altered file-ids has the exact revision_ids that
1014
altered it listed explicitly.
1016
assert self._serializer.support_altered_by_hack, \
1017
("fileids_altered_by_revision_ids only supported for branches "
1018
"which store inventory as unnested xml, not on %r" % self)
1019
selected_revision_ids = set(revision_ids)
1020
w = self.get_inventory_weave()
1023
# this code needs to read every new line in every inventory for the
1024
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1025
# not present in one of those inventories is unnecessary but not
1026
# harmful because we are filtering by the revision id marker in the
1027
# inventory lines : we only select file ids altered in one of those
1028
# revisions. We don't need to see all lines in the inventory because
1029
# only those added in an inventory in rev X can contain a revision=X
1031
unescape_revid_cache = {}
1032
unescape_fileid_cache = {}
1034
# jam 20061218 In a big fetch, this handles hundreds of thousands
1035
# of lines, so it has had a lot of inlining and optimizing done.
1036
# Sorry that it is a little bit messy.
1037
# Move several functions to be local variables, since this is a long
1039
search = self._file_ids_altered_regex.search
1040
unescape = _unescape_xml
1041
setdefault = result.setdefault
1042
pb = ui.ui_factory.nested_progress_bar()
1044
for line in w.iter_lines_added_or_present_in_versions(
1045
selected_revision_ids, pb=pb):
1046
match = search(line)
1049
# One call to match.group() returning multiple items is quite a
1050
# bit faster than 2 calls to match.group() each returning 1
1051
file_id, revision_id = match.group('file_id', 'revision_id')
1053
# Inlining the cache lookups helps a lot when you make 170,000
1054
# lines and 350k ids, versus 8.4 unique ids.
1055
# Using a cache helps in 2 ways:
1056
# 1) Avoids unnecessary decoding calls
1057
# 2) Re-uses cached strings, which helps in future set and
1059
# (2) is enough that removing encoding entirely along with
1060
# the cache (so we are using plain strings) results in no
1061
# performance improvement.
1063
revision_id = unescape_revid_cache[revision_id]
1065
unescaped = unescape(revision_id)
1066
unescape_revid_cache[revision_id] = unescaped
1067
revision_id = unescaped
1069
if revision_id in selected_revision_ids:
1071
file_id = unescape_fileid_cache[file_id]
1073
unescaped = unescape(file_id)
1074
unescape_fileid_cache[file_id] = unescaped
1076
setdefault(file_id, set()).add(revision_id)
1081
def iter_files_bytes(self, desired_files):
1082
"""Iterate through file versions.
1084
Files will not necessarily be returned in the order they occur in
1085
desired_files. No specific order is guaranteed.
1087
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1088
value supplied by the caller as part of desired_files. It should
1089
uniquely identify the file version in the caller's context. (Examples:
1090
an index number or a TreeTransform trans_id.)
1092
bytes_iterator is an iterable of bytestrings for the file. The
1093
kind of iterable and length of the bytestrings are unspecified, but for
1094
this implementation, it is a list of lines produced by
1095
VersionedFile.get_lines().
1097
:param desired_files: a list of (file_id, revision_id, identifier)
1100
transaction = self.get_transaction()
1101
for file_id, revision_id, callable_data in desired_files:
1103
weave = self.weave_store.get_weave(file_id, transaction)
1104
except errors.NoSuchFile:
1105
raise errors.NoSuchIdInRepository(self, file_id)
1106
yield callable_data, weave.get_lines(revision_id)
1108
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1109
"""Get an iterable listing the keys of all the data introduced by a set
1112
The keys will be ordered so that the corresponding items can be safely
1113
fetched and inserted in that order.
1115
:returns: An iterable producing tuples of (knit-kind, file-id,
1116
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1117
'revisions'. file-id is None unless knit-kind is 'file'.
1119
# XXX: it's a bit weird to control the inventory weave caching in this
1120
# generator. Ideally the caching would be done in fetch.py I think. Or
1121
# maybe this generator should explicitly have the contract that it
1122
# should not be iterated until the previously yielded item has been
1124
inv_w = self.get_inventory_weave()
1125
inv_w.enable_cache()
1127
# file ids that changed
1128
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1130
num_file_ids = len(file_ids)
1131
for file_id, altered_versions in file_ids.iteritems():
1132
if _files_pb is not None:
1133
_files_pb.update("fetch texts", count, num_file_ids)
1135
yield ("file", file_id, altered_versions)
1136
# We're done with the files_pb. Note that it finished by the caller,
1137
# just as it was created by the caller.
1141
yield ("inventory", None, revision_ids)
1145
revisions_with_signatures = set()
1146
for rev_id in revision_ids:
1148
self.get_signature_text(rev_id)
1149
except errors.NoSuchRevision:
1153
revisions_with_signatures.add(rev_id)
1154
yield ("signatures", None, revisions_with_signatures)
1157
yield ("revisions", None, revision_ids)
1160
def get_inventory_weave(self):
1161
return self.control_weaves.get_weave('inventory',
1162
self.get_transaction())
1165
def get_inventory(self, revision_id):
1166
"""Get Inventory object by hash."""
1167
return self.deserialise_inventory(
1168
revision_id, self.get_inventory_xml(revision_id))
1170
def deserialise_inventory(self, revision_id, xml):
1171
"""Transform the xml into an inventory object.
1173
:param revision_id: The expected revision id of the inventory.
1174
:param xml: A serialised inventory.
1176
return self._serializer.read_inventory_from_string(xml, revision_id)
1178
def serialise_inventory(self, inv):
1179
return self._serializer.write_inventory_to_string(inv)
1181
def _serialise_inventory_to_lines(self, inv):
1182
return self._serializer.write_inventory_to_lines(inv)
1184
def get_serializer_format(self):
1185
return self._serializer.format_num
1188
def get_inventory_xml(self, revision_id):
1189
"""Get inventory XML as a file object."""
1191
assert isinstance(revision_id, str), type(revision_id)
1192
iw = self.get_inventory_weave()
1193
return iw.get_text(revision_id)
1195
raise errors.HistoryMissing(self, 'inventory', revision_id)
1198
def get_inventory_sha1(self, revision_id):
1199
"""Return the sha1 hash of the inventory entry
1201
return self.get_revision(revision_id).inventory_sha1
1204
def get_revision_graph(self, revision_id=None):
1205
"""Return a dictionary containing the revision graph.
1207
NB: This method should not be used as it accesses the entire graph all
1208
at once, which is much more data than most operations should require.
1210
:param revision_id: The revision_id to get a graph from. If None, then
1211
the entire revision graph is returned. This is a deprecated mode of
1212
operation and will be removed in the future.
1213
:return: a dictionary of revision_id->revision_parents_list.
1215
raise NotImplementedError(self.get_revision_graph)
1218
def get_revision_graph_with_ghosts(self, revision_ids=None):
1219
"""Return a graph of the revisions with ghosts marked as applicable.
1221
:param revision_ids: an iterable of revisions to graph or None for all.
1222
:return: a Graph object with the graph reachable from revision_ids.
1224
if 'evil' in debug.debug_flags:
1226
"get_revision_graph_with_ghosts scales with size of history.")
1227
result = deprecated_graph.Graph()
1228
if not revision_ids:
1229
pending = set(self.all_revision_ids())
1232
pending = set(revision_ids)
1233
# special case NULL_REVISION
1234
if _mod_revision.NULL_REVISION in pending:
1235
pending.remove(_mod_revision.NULL_REVISION)
1236
required = set(pending)
1239
revision_id = pending.pop()
1241
rev = self.get_revision(revision_id)
1242
except errors.NoSuchRevision:
1243
if revision_id in required:
1246
result.add_ghost(revision_id)
1248
for parent_id in rev.parent_ids:
1249
# is this queued or done ?
1250
if (parent_id not in pending and
1251
parent_id not in done):
1253
pending.add(parent_id)
1254
result.add_node(revision_id, rev.parent_ids)
1255
done.add(revision_id)
1258
def _get_history_vf(self):
1259
"""Get a versionedfile whose history graph reflects all revisions.
1261
For weave repositories, this is the inventory weave.
1263
return self.get_inventory_weave()
1265
def iter_reverse_revision_history(self, revision_id):
1266
"""Iterate backwards through revision ids in the lefthand history
1268
:param revision_id: The revision id to start with. All its lefthand
1269
ancestors will be traversed.
1271
if revision_id in (None, _mod_revision.NULL_REVISION):
1273
next_id = revision_id
1274
versionedfile = self._get_history_vf()
1277
parents = versionedfile.get_parents(next_id)
1278
if len(parents) == 0:
1281
next_id = parents[0]
1284
def get_revision_inventory(self, revision_id):
1285
"""Return inventory of a past revision."""
1286
# TODO: Unify this with get_inventory()
1287
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1288
# must be the same as its revision, so this is trivial.
1289
if revision_id is None:
1290
# This does not make sense: if there is no revision,
1291
# then it is the current tree inventory surely ?!
1292
# and thus get_root_id() is something that looks at the last
1293
# commit on the branch, and the get_root_id is an inventory check.
1294
raise NotImplementedError
1295
# return Inventory(self.get_root_id())
1297
return self.get_inventory(revision_id)
1300
def is_shared(self):
1301
"""Return True if this repository is flagged as a shared repository."""
1302
raise NotImplementedError(self.is_shared)
1305
def reconcile(self, other=None, thorough=False):
1306
"""Reconcile this repository."""
1307
from bzrlib.reconcile import RepoReconciler
1308
reconciler = RepoReconciler(self, thorough=thorough)
1309
reconciler.reconcile()
1312
def _refresh_data(self):
1313
"""Helper called from lock_* to ensure coherency with disk.
1315
The default implementation does nothing; it is however possible
1316
for repositories to maintain loaded indices across multiple locks
1317
by checking inside their implementation of this method to see
1318
whether their indices are still valid. This depends of course on
1319
the disk format being validatable in this manner.
1323
def revision_tree(self, revision_id):
1324
"""Return Tree for a revision on this branch.
1326
`revision_id` may be None for the empty tree revision.
1328
# TODO: refactor this to use an existing revision object
1329
# so we don't need to read it in twice.
1330
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1331
return RevisionTree(self, Inventory(root_id=None),
1332
_mod_revision.NULL_REVISION)
1334
inv = self.get_revision_inventory(revision_id)
1335
return RevisionTree(self, inv, revision_id)
1338
def revision_trees(self, revision_ids):
1339
"""Return Tree for a revision on this branch.
1341
`revision_id` may not be None or 'null:'"""
1342
assert None not in revision_ids
1343
assert _mod_revision.NULL_REVISION not in revision_ids
1344
texts = self.get_inventory_weave().get_texts(revision_ids)
1345
for text, revision_id in zip(texts, revision_ids):
1346
inv = self.deserialise_inventory(revision_id, text)
1347
yield RevisionTree(self, inv, revision_id)
1350
def get_ancestry(self, revision_id, topo_sorted=True):
1351
"""Return a list of revision-ids integrated by a revision.
1353
The first element of the list is always None, indicating the origin
1354
revision. This might change when we have history horizons, or
1355
perhaps we should have a new API.
1357
This is topologically sorted.
1359
if _mod_revision.is_null(revision_id):
1361
if not self.has_revision(revision_id):
1362
raise errors.NoSuchRevision(self, revision_id)
1363
w = self.get_inventory_weave()
1364
candidates = w.get_ancestry(revision_id, topo_sorted)
1365
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1368
"""Compress the data within the repository.
1370
This operation only makes sense for some repository types. For other
1371
types it should be a no-op that just returns.
1373
This stub method does not require a lock, but subclasses should use
1374
@needs_write_lock as this is a long running call its reasonable to
1375
implicitly lock for the user.
1379
def print_file(self, file, revision_id):
1380
"""Print `file` to stdout.
1382
FIXME RBC 20060125 as John Meinel points out this is a bad api
1383
- it writes to stdout, it assumes that that is valid etc. Fix
1384
by creating a new more flexible convenience function.
1386
tree = self.revision_tree(revision_id)
1387
# use inventory as it was in that revision
1388
file_id = tree.inventory.path2id(file)
1390
# TODO: jam 20060427 Write a test for this code path
1391
# it had a bug in it, and was raising the wrong
1393
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1394
tree.print_file(file_id)
1396
def get_transaction(self):
1397
return self.control_files.get_transaction()
1399
def revision_parents(self, revision_id):
1400
return self.get_inventory_weave().parent_names(revision_id)
1402
def get_parents(self, revision_ids):
1403
"""See StackedParentsProvider.get_parents"""
1405
for revision_id in revision_ids:
1406
if revision_id == _mod_revision.NULL_REVISION:
1410
parents = self.get_revision(revision_id).parent_ids
1411
except errors.NoSuchRevision:
1414
if len(parents) == 0:
1415
parents = [_mod_revision.NULL_REVISION]
1416
parents_list.append(parents)
1419
def _make_parents_provider(self):
1422
def get_graph(self, other_repository=None):
1423
"""Return the graph walker for this repository format"""
1424
parents_provider = self._make_parents_provider()
1425
if (other_repository is not None and
1426
other_repository.bzrdir.transport.base !=
1427
self.bzrdir.transport.base):
1428
parents_provider = graph._StackedParentsProvider(
1429
[parents_provider, other_repository._make_parents_provider()])
1430
return graph.Graph(parents_provider)
1433
def set_make_working_trees(self, new_value):
1434
"""Set the policy flag for making working trees when creating branches.
1436
This only applies to branches that use this repository.
1438
The default is 'True'.
1439
:param new_value: True to restore the default, False to disable making
1442
raise NotImplementedError(self.set_make_working_trees)
1444
def make_working_trees(self):
1445
"""Returns the policy for making working trees on new branches."""
1446
raise NotImplementedError(self.make_working_trees)
1449
def sign_revision(self, revision_id, gpg_strategy):
1450
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1451
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1454
def has_signature_for_revision_id(self, revision_id):
1455
"""Query for a revision signature for revision_id in the repository."""
1456
return self._revision_store.has_signature(revision_id,
1457
self.get_transaction())
1460
def get_signature_text(self, revision_id):
1461
"""Return the text for a signature."""
1462
return self._revision_store.get_signature_text(revision_id,
1463
self.get_transaction())
1466
def check(self, revision_ids):
1467
"""Check consistency of all history of given revision_ids.
1469
Different repository implementations should override _check().
1471
:param revision_ids: A non-empty list of revision_ids whose ancestry
1472
will be checked. Typically the last revision_id of a branch.
1474
if not revision_ids:
1475
raise ValueError("revision_ids must be non-empty in %s.check"
1477
return self._check(revision_ids)
1479
def _check(self, revision_ids):
1480
result = check.Check(self)
1484
def _warn_if_deprecated(self):
1485
global _deprecation_warning_done
1486
if _deprecation_warning_done:
1488
_deprecation_warning_done = True
1489
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1490
% (self._format, self.bzrdir.transport.base))
1492
def supports_rich_root(self):
1493
return self._format.rich_root_data
1495
def _check_ascii_revisionid(self, revision_id, method):
1496
"""Private helper for ascii-only repositories."""
1497
# weave repositories refuse to store revisionids that are non-ascii.
1498
if revision_id is not None:
1499
# weaves require ascii revision ids.
1500
if isinstance(revision_id, unicode):
1502
revision_id.encode('ascii')
1503
except UnicodeEncodeError:
1504
raise errors.NonAsciiRevisionId(method, self)
1507
revision_id.decode('ascii')
1508
except UnicodeDecodeError:
1509
raise errors.NonAsciiRevisionId(method, self)
1513
# remove these delegates a while after bzr 0.15
1514
def __make_delegated(name, from_module):
1515
def _deprecated_repository_forwarder():
1516
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1517
% (name, from_module),
1520
m = __import__(from_module, globals(), locals(), [name])
1522
return getattr(m, name)
1523
except AttributeError:
1524
raise AttributeError('module %s has no name %s'
1526
globals()[name] = _deprecated_repository_forwarder
1529
'AllInOneRepository',
1530
'WeaveMetaDirRepository',
1531
'PreSplitOutRepositoryFormat',
1532
'RepositoryFormat4',
1533
'RepositoryFormat5',
1534
'RepositoryFormat6',
1535
'RepositoryFormat7',
1537
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1541
'RepositoryFormatKnit',
1542
'RepositoryFormatKnit1',
1544
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1547
def install_revision(repository, rev, revision_tree):
1548
"""Install all revision data into a repository."""
1549
present_parents = []
1551
for p_id in rev.parent_ids:
1552
if repository.has_revision(p_id):
1553
present_parents.append(p_id)
1554
parent_trees[p_id] = repository.revision_tree(p_id)
1556
parent_trees[p_id] = repository.revision_tree(None)
1558
inv = revision_tree.inventory
1559
entries = inv.iter_entries()
1560
# backwards compatibility hack: skip the root id.
1561
if not repository.supports_rich_root():
1562
path, root = entries.next()
1563
if root.revision != rev.revision_id:
1564
raise errors.IncompatibleRevision(repr(repository))
1565
# Add the texts that are not already present
1566
for path, ie in entries:
1567
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1568
repository.get_transaction())
1569
if ie.revision not in w:
1571
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1572
# with InventoryEntry.find_previous_heads(). if it is, then there
1573
# is a latent bug here where the parents may have ancestors of each
1575
for revision, tree in parent_trees.iteritems():
1576
if ie.file_id not in tree:
1578
parent_id = tree.inventory[ie.file_id].revision
1579
if parent_id in text_parents:
1581
text_parents.append(parent_id)
1583
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1584
repository.get_transaction())
1585
lines = revision_tree.get_file(ie.file_id).readlines()
1586
vfile.add_lines(rev.revision_id, text_parents, lines)
1588
# install the inventory
1589
repository.add_inventory(rev.revision_id, inv, present_parents)
1590
except errors.RevisionAlreadyPresent:
1592
repository.add_revision(rev.revision_id, rev, inv)
1595
class MetaDirRepository(Repository):
1596
"""Repositories in the new meta-dir layout."""
1598
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1599
super(MetaDirRepository, self).__init__(_format,
1605
dir_mode = self.control_files._dir_mode
1606
file_mode = self.control_files._file_mode
1609
def is_shared(self):
1610
"""Return True if this repository is flagged as a shared repository."""
1611
return self.control_files._transport.has('shared-storage')
1614
def set_make_working_trees(self, new_value):
1615
"""Set the policy flag for making working trees when creating branches.
1617
This only applies to branches that use this repository.
1619
The default is 'True'.
1620
:param new_value: True to restore the default, False to disable making
1625
self.control_files._transport.delete('no-working-trees')
1626
except errors.NoSuchFile:
1629
self.control_files.put_utf8('no-working-trees', '')
1631
def make_working_trees(self):
1632
"""Returns the policy for making working trees on new branches."""
1633
return not self.control_files._transport.has('no-working-trees')
1636
class RepositoryFormatRegistry(registry.Registry):
1637
"""Registry of RepositoryFormats."""
1639
def get(self, format_string):
1640
r = registry.Registry.get(self, format_string)
1646
format_registry = RepositoryFormatRegistry()
1647
"""Registry of formats, indexed by their identifying format string.
1649
This can contain either format instances themselves, or classes/factories that
1650
can be called to obtain one.
1654
#####################################################################
1655
# Repository Formats
1657
class RepositoryFormat(object):
1658
"""A repository format.
1660
Formats provide three things:
1661
* An initialization routine to construct repository data on disk.
1662
* a format string which is used when the BzrDir supports versioned
1664
* an open routine which returns a Repository instance.
1666
There is one and only one Format subclass for each on-disk format. But
1667
there can be one Repository subclass that is used for several different
1668
formats. The _format attribute on a Repository instance can be used to
1669
determine the disk format.
1671
Formats are placed in an dict by their format string for reference
1672
during opening. These should be subclasses of RepositoryFormat
1675
Once a format is deprecated, just deprecate the initialize and open
1676
methods on the format class. Do not deprecate the object, as the
1677
object will be created every system load.
1679
Common instance attributes:
1680
_matchingbzrdir - the bzrdir format that the repository format was
1681
originally written to work with. This can be used if manually
1682
constructing a bzrdir and repository, or more commonly for test suite
1687
return "<%s>" % self.__class__.__name__
1689
def __eq__(self, other):
1690
# format objects are generally stateless
1691
return isinstance(other, self.__class__)
1693
def __ne__(self, other):
1694
return not self == other
1697
def find_format(klass, a_bzrdir):
1698
"""Return the format for the repository object in a_bzrdir.
1700
This is used by bzr native formats that have a "format" file in
1701
the repository. Other methods may be used by different types of
1705
transport = a_bzrdir.get_repository_transport(None)
1706
format_string = transport.get("format").read()
1707
return format_registry.get(format_string)
1708
except errors.NoSuchFile:
1709
raise errors.NoRepositoryPresent(a_bzrdir)
1711
raise errors.UnknownFormatError(format=format_string)
1714
def register_format(klass, format):
1715
format_registry.register(format.get_format_string(), format)
1718
def unregister_format(klass, format):
1719
format_registry.remove(format.get_format_string())
1722
def get_default_format(klass):
1723
"""Return the current default format."""
1724
from bzrlib import bzrdir
1725
return bzrdir.format_registry.make_bzrdir('default').repository_format
1727
def _get_control_store(self, repo_transport, control_files):
1728
"""Return the control store for this repository."""
1729
raise NotImplementedError(self._get_control_store)
1731
def get_format_string(self):
1732
"""Return the ASCII format string that identifies this format.
1734
Note that in pre format ?? repositories the format string is
1735
not permitted nor written to disk.
1737
raise NotImplementedError(self.get_format_string)
1739
def get_format_description(self):
1740
"""Return the short description for this format."""
1741
raise NotImplementedError(self.get_format_description)
1743
def _get_revision_store(self, repo_transport, control_files):
1744
"""Return the revision store object for this a_bzrdir."""
1745
raise NotImplementedError(self._get_revision_store)
1747
def _get_text_rev_store(self,
1754
"""Common logic for getting a revision store for a repository.
1756
see self._get_revision_store for the subclass-overridable method to
1757
get the store for a repository.
1759
from bzrlib.store.revision.text import TextRevisionStore
1760
dir_mode = control_files._dir_mode
1761
file_mode = control_files._file_mode
1762
text_store = TextStore(transport.clone(name),
1764
compressed=compressed,
1766
file_mode=file_mode)
1767
_revision_store = TextRevisionStore(text_store, serializer)
1768
return _revision_store
1770
# TODO: this shouldn't be in the base class, it's specific to things that
1771
# use weaves or knits -- mbp 20070207
1772
def _get_versioned_file_store(self,
1777
versionedfile_class=None,
1778
versionedfile_kwargs={},
1780
if versionedfile_class is None:
1781
versionedfile_class = self._versionedfile_class
1782
weave_transport = control_files._transport.clone(name)
1783
dir_mode = control_files._dir_mode
1784
file_mode = control_files._file_mode
1785
return VersionedFileStore(weave_transport, prefixed=prefixed,
1787
file_mode=file_mode,
1788
versionedfile_class=versionedfile_class,
1789
versionedfile_kwargs=versionedfile_kwargs,
1792
def initialize(self, a_bzrdir, shared=False):
1793
"""Initialize a repository of this format in a_bzrdir.
1795
:param a_bzrdir: The bzrdir to put the new repository in it.
1796
:param shared: The repository should be initialized as a sharable one.
1797
:returns: The new repository object.
1799
This may raise UninitializableFormat if shared repository are not
1800
compatible the a_bzrdir.
1802
raise NotImplementedError(self.initialize)
1804
def is_supported(self):
1805
"""Is this format supported?
1807
Supported formats must be initializable and openable.
1808
Unsupported formats may not support initialization or committing or
1809
some other features depending on the reason for not being supported.
1813
def check_conversion_target(self, target_format):
1814
raise NotImplementedError(self.check_conversion_target)
1816
def open(self, a_bzrdir, _found=False):
1817
"""Return an instance of this format for the bzrdir a_bzrdir.
1819
_found is a private parameter, do not use it.
1821
raise NotImplementedError(self.open)
1824
class MetaDirRepositoryFormat(RepositoryFormat):
1825
"""Common base class for the new repositories using the metadir layout."""
1827
rich_root_data = False
1828
supports_tree_reference = False
1829
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
1832
super(MetaDirRepositoryFormat, self).__init__()
1834
def _create_control_files(self, a_bzrdir):
1835
"""Create the required files and the initial control_files object."""
1836
# FIXME: RBC 20060125 don't peek under the covers
1837
# NB: no need to escape relative paths that are url safe.
1838
repository_transport = a_bzrdir.get_repository_transport(self)
1839
control_files = lockable_files.LockableFiles(repository_transport,
1840
'lock', lockdir.LockDir)
1841
control_files.create_lock()
1842
return control_files
1844
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
1845
"""Upload the initial blank content."""
1846
control_files = self._create_control_files(a_bzrdir)
1847
control_files.lock_write()
1849
control_files._transport.mkdir_multi(dirs,
1850
mode=control_files._dir_mode)
1851
for file, content in files:
1852
control_files.put(file, content)
1853
for file, content in utf8_files:
1854
control_files.put_utf8(file, content)
1856
control_files.put_utf8('shared-storage', '')
1858
control_files.unlock()
1861
# formats which have no format string are not discoverable
1862
# and not independently creatable, so are not registered. They're
1863
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1864
# needed, it's constructed directly by the BzrDir. Non-native formats where
1865
# the repository is not separately opened are similar.
1867
format_registry.register_lazy(
1868
'Bazaar-NG Repository format 7',
1869
'bzrlib.repofmt.weaverepo',
1873
# KEEP in sync with bzrdir.format_registry default, which controls the overall
1874
# default control directory format
1875
format_registry.register_lazy(
1876
'Bazaar-NG Knit Repository Format 1',
1877
'bzrlib.repofmt.knitrepo',
1878
'RepositoryFormatKnit1',
1880
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
1882
format_registry.register_lazy(
1883
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
1884
'bzrlib.repofmt.knitrepo',
1885
'RepositoryFormatKnit3',
1889
class InterRepository(InterObject):
1890
"""This class represents operations taking place between two repositories.
1892
Its instances have methods like copy_content and fetch, and contain
1893
references to the source and target repositories these operations can be
1896
Often we will provide convenience methods on 'repository' which carry out
1897
operations with another repository - they will always forward to
1898
InterRepository.get(other).method_name(parameters).
1902
"""The available optimised InterRepository types."""
1904
def copy_content(self, revision_id=None):
1905
raise NotImplementedError(self.copy_content)
1907
def fetch(self, revision_id=None, pb=None):
1908
"""Fetch the content required to construct revision_id.
1910
The content is copied from self.source to self.target.
1912
:param revision_id: if None all content is copied, if NULL_REVISION no
1914
:param pb: optional progress bar to use for progress reports. If not
1915
provided a default one will be created.
1917
Returns the copied revision count and the failed revisions in a tuple:
1920
raise NotImplementedError(self.fetch)
1923
def missing_revision_ids(self, revision_id=None):
1924
"""Return the revision ids that source has that target does not.
1926
These are returned in topological order.
1928
:param revision_id: only return revision ids included by this
1931
# generic, possibly worst case, slow code path.
1932
target_ids = set(self.target.all_revision_ids())
1933
if revision_id is not None:
1934
source_ids = self.source.get_ancestry(revision_id)
1935
assert source_ids[0] is None
1938
source_ids = self.source.all_revision_ids()
1939
result_set = set(source_ids).difference(target_ids)
1940
# this may look like a no-op: its not. It preserves the ordering
1941
# other_ids had while only returning the members from other_ids
1942
# that we've decided we need.
1943
return [rev_id for rev_id in source_ids if rev_id in result_set]
1946
def _same_model(source, target):
1947
"""True if source and target have the same data representation."""
1948
if source.supports_rich_root() != target.supports_rich_root():
1950
if source._serializer != target._serializer:
1955
class InterSameDataRepository(InterRepository):
1956
"""Code for converting between repositories that represent the same data.
1958
Data format and model must match for this to work.
1962
def _get_repo_format_to_test(self):
1963
"""Repository format for testing with.
1965
InterSameData can pull from subtree to subtree and from non-subtree to
1966
non-subtree, so we test this with the richest repository format.
1968
from bzrlib.repofmt import knitrepo
1969
return knitrepo.RepositoryFormatKnit3()
1972
def is_compatible(source, target):
1973
return InterRepository._same_model(source, target)
1976
def copy_content(self, revision_id=None):
1977
"""Make a complete copy of the content in self into destination.
1979
This copies both the repository's revision data, and configuration information
1980
such as the make_working_trees setting.
1982
This is a destructive operation! Do not use it on existing
1985
:param revision_id: Only copy the content needed to construct
1986
revision_id and its parents.
1989
self.target.set_make_working_trees(self.source.make_working_trees())
1990
except NotImplementedError:
1992
# but don't bother fetching if we have the needed data now.
1993
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
1994
self.target.has_revision(revision_id)):
1996
self.target.fetch(self.source, revision_id=revision_id)
1999
def fetch(self, revision_id=None, pb=None):
2000
"""See InterRepository.fetch()."""
2001
from bzrlib.fetch import GenericRepoFetcher
2002
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2003
self.source, self.source._format, self.target,
2004
self.target._format)
2005
f = GenericRepoFetcher(to_repository=self.target,
2006
from_repository=self.source,
2007
last_revision=revision_id,
2009
return f.count_copied, f.failed_revisions
2012
class InterWeaveRepo(InterSameDataRepository):
2013
"""Optimised code paths between Weave based repositories.
2015
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2016
implemented lazy inter-object optimisation.
2020
def _get_repo_format_to_test(self):
2021
from bzrlib.repofmt import weaverepo
2022
return weaverepo.RepositoryFormat7()
2025
def is_compatible(source, target):
2026
"""Be compatible with known Weave formats.
2028
We don't test for the stores being of specific types because that
2029
could lead to confusing results, and there is no need to be
2032
from bzrlib.repofmt.weaverepo import (
2038
return (isinstance(source._format, (RepositoryFormat5,
2040
RepositoryFormat7)) and
2041
isinstance(target._format, (RepositoryFormat5,
2043
RepositoryFormat7)))
2044
except AttributeError:
2048
def copy_content(self, revision_id=None):
2049
"""See InterRepository.copy_content()."""
2050
# weave specific optimised path:
2052
self.target.set_make_working_trees(self.source.make_working_trees())
2053
except NotImplementedError:
2055
# FIXME do not peek!
2056
if self.source.control_files._transport.listable():
2057
pb = ui.ui_factory.nested_progress_bar()
2059
self.target.weave_store.copy_all_ids(
2060
self.source.weave_store,
2062
from_transaction=self.source.get_transaction(),
2063
to_transaction=self.target.get_transaction())
2064
pb.update('copying inventory', 0, 1)
2065
self.target.control_weaves.copy_multi(
2066
self.source.control_weaves, ['inventory'],
2067
from_transaction=self.source.get_transaction(),
2068
to_transaction=self.target.get_transaction())
2069
self.target._revision_store.text_store.copy_all_ids(
2070
self.source._revision_store.text_store,
2075
self.target.fetch(self.source, revision_id=revision_id)
2078
def fetch(self, revision_id=None, pb=None):
2079
"""See InterRepository.fetch()."""
2080
from bzrlib.fetch import GenericRepoFetcher
2081
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2082
self.source, self.source._format, self.target, self.target._format)
2083
f = GenericRepoFetcher(to_repository=self.target,
2084
from_repository=self.source,
2085
last_revision=revision_id,
2087
return f.count_copied, f.failed_revisions
2090
def missing_revision_ids(self, revision_id=None):
2091
"""See InterRepository.missing_revision_ids()."""
2092
# we want all revisions to satisfy revision_id in source.
2093
# but we don't want to stat every file here and there.
2094
# we want then, all revisions other needs to satisfy revision_id
2095
# checked, but not those that we have locally.
2096
# so the first thing is to get a subset of the revisions to
2097
# satisfy revision_id in source, and then eliminate those that
2098
# we do already have.
2099
# this is slow on high latency connection to self, but as as this
2100
# disk format scales terribly for push anyway due to rewriting
2101
# inventory.weave, this is considered acceptable.
2103
if revision_id is not None:
2104
source_ids = self.source.get_ancestry(revision_id)
2105
assert source_ids[0] is None
2108
source_ids = self.source._all_possible_ids()
2109
source_ids_set = set(source_ids)
2110
# source_ids is the worst possible case we may need to pull.
2111
# now we want to filter source_ids against what we actually
2112
# have in target, but don't try to check for existence where we know
2113
# we do not have a revision as that would be pointless.
2114
target_ids = set(self.target._all_possible_ids())
2115
possibly_present_revisions = target_ids.intersection(source_ids_set)
2116
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2117
required_revisions = source_ids_set.difference(actually_present_revisions)
2118
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2119
if revision_id is not None:
2120
# we used get_ancestry to determine source_ids then we are assured all
2121
# revisions referenced are present as they are installed in topological order.
2122
# and the tip revision was validated by get_ancestry.
2123
return required_topo_revisions
2125
# if we just grabbed the possibly available ids, then
2126
# we only have an estimate of whats available and need to validate
2127
# that against the revision records.
2128
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2131
class InterKnitRepo(InterSameDataRepository):
2132
"""Optimised code paths between Knit based repositories."""
2135
def _get_repo_format_to_test(self):
2136
from bzrlib.repofmt import knitrepo
2137
return knitrepo.RepositoryFormatKnit1()
2140
def is_compatible(source, target):
2141
"""Be compatible with known Knit formats.
2143
We don't test for the stores being of specific types because that
2144
could lead to confusing results, and there is no need to be
2147
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2149
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2150
isinstance(target._format, RepositoryFormatKnit))
2151
except AttributeError:
2153
return are_knits and InterRepository._same_model(source, target)
2156
def fetch(self, revision_id=None, pb=None):
2157
"""See InterRepository.fetch()."""
2158
from bzrlib.fetch import KnitRepoFetcher
2159
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2160
self.source, self.source._format, self.target, self.target._format)
2161
f = KnitRepoFetcher(to_repository=self.target,
2162
from_repository=self.source,
2163
last_revision=revision_id,
2165
return f.count_copied, f.failed_revisions
2168
def missing_revision_ids(self, revision_id=None):
2169
"""See InterRepository.missing_revision_ids()."""
2170
if revision_id is not None:
2171
source_ids = self.source.get_ancestry(revision_id)
2172
assert source_ids[0] is None
2175
source_ids = self.source.all_revision_ids()
2176
source_ids_set = set(source_ids)
2177
# source_ids is the worst possible case we may need to pull.
2178
# now we want to filter source_ids against what we actually
2179
# have in target, but don't try to check for existence where we know
2180
# we do not have a revision as that would be pointless.
2181
target_ids = set(self.target.all_revision_ids())
2182
possibly_present_revisions = target_ids.intersection(source_ids_set)
2183
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2184
required_revisions = source_ids_set.difference(actually_present_revisions)
2185
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2186
if revision_id is not None:
2187
# we used get_ancestry to determine source_ids then we are assured all
2188
# revisions referenced are present as they are installed in topological order.
2189
# and the tip revision was validated by get_ancestry.
2190
return required_topo_revisions
2192
# if we just grabbed the possibly available ids, then
2193
# we only have an estimate of whats available and need to validate
2194
# that against the revision records.
2195
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2198
class InterModel1and2(InterRepository):
2201
def _get_repo_format_to_test(self):
2205
def is_compatible(source, target):
2206
if not source.supports_rich_root() and target.supports_rich_root():
2212
def fetch(self, revision_id=None, pb=None):
2213
"""See InterRepository.fetch()."""
2214
from bzrlib.fetch import Model1toKnit2Fetcher
2215
f = Model1toKnit2Fetcher(to_repository=self.target,
2216
from_repository=self.source,
2217
last_revision=revision_id,
2219
return f.count_copied, f.failed_revisions
2222
def copy_content(self, revision_id=None):
2223
"""Make a complete copy of the content in self into destination.
2225
This is a destructive operation! Do not use it on existing
2228
:param revision_id: Only copy the content needed to construct
2229
revision_id and its parents.
2232
self.target.set_make_working_trees(self.source.make_working_trees())
2233
except NotImplementedError:
2235
# but don't bother fetching if we have the needed data now.
2236
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2237
self.target.has_revision(revision_id)):
2239
self.target.fetch(self.source, revision_id=revision_id)
2242
class InterKnit1and2(InterKnitRepo):
2245
def _get_repo_format_to_test(self):
2249
def is_compatible(source, target):
2250
"""Be compatible with Knit1 source and Knit3 target"""
2251
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2253
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit1, \
2254
RepositoryFormatKnit3
2255
return (isinstance(source._format, (RepositoryFormatKnit1)) and
2256
isinstance(target._format, (RepositoryFormatKnit3)))
2257
except AttributeError:
2261
def fetch(self, revision_id=None, pb=None):
2262
"""See InterRepository.fetch()."""
2263
from bzrlib.fetch import Knit1to2Fetcher
2264
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2265
self.source, self.source._format, self.target,
2266
self.target._format)
2267
f = Knit1to2Fetcher(to_repository=self.target,
2268
from_repository=self.source,
2269
last_revision=revision_id,
2271
return f.count_copied, f.failed_revisions
2274
class InterRemoteRepository(InterRepository):
2275
"""Code for converting between RemoteRepository objects.
2277
This just gets an non-remote repository from the RemoteRepository, and calls
2278
InterRepository.get again.
2281
def __init__(self, source, target):
2282
if isinstance(source, remote.RemoteRepository):
2283
source._ensure_real()
2284
real_source = source._real_repository
2286
real_source = source
2287
if isinstance(target, remote.RemoteRepository):
2288
target._ensure_real()
2289
real_target = target._real_repository
2291
real_target = target
2292
self.real_inter = InterRepository.get(real_source, real_target)
2295
def is_compatible(source, target):
2296
if isinstance(source, remote.RemoteRepository):
2298
if isinstance(target, remote.RemoteRepository):
2302
def copy_content(self, revision_id=None):
2303
self.real_inter.copy_content(revision_id=revision_id)
2305
def fetch(self, revision_id=None, pb=None):
2306
self.real_inter.fetch(revision_id=revision_id, pb=pb)
2309
def _get_repo_format_to_test(self):
2313
InterRepository.register_optimiser(InterSameDataRepository)
2314
InterRepository.register_optimiser(InterWeaveRepo)
2315
InterRepository.register_optimiser(InterKnitRepo)
2316
InterRepository.register_optimiser(InterModel1and2)
2317
InterRepository.register_optimiser(InterKnit1and2)
2318
InterRepository.register_optimiser(InterRemoteRepository)
2321
class CopyConverter(object):
2322
"""A repository conversion tool which just performs a copy of the content.
2324
This is slow but quite reliable.
2327
def __init__(self, target_format):
2328
"""Create a CopyConverter.
2330
:param target_format: The format the resulting repository should be.
2332
self.target_format = target_format
2334
def convert(self, repo, pb):
2335
"""Perform the conversion of to_convert, giving feedback via pb.
2337
:param to_convert: The disk object to convert.
2338
:param pb: a progress bar to use for progress information.
2343
# this is only useful with metadir layouts - separated repo content.
2344
# trigger an assertion if not such
2345
repo._format.get_format_string()
2346
self.repo_dir = repo.bzrdir
2347
self.step('Moving repository to repository.backup')
2348
self.repo_dir.transport.move('repository', 'repository.backup')
2349
backup_transport = self.repo_dir.transport.clone('repository.backup')
2350
repo._format.check_conversion_target(self.target_format)
2351
self.source_repo = repo._format.open(self.repo_dir,
2353
_override_transport=backup_transport)
2354
self.step('Creating new repository')
2355
converted = self.target_format.initialize(self.repo_dir,
2356
self.source_repo.is_shared())
2357
converted.lock_write()
2359
self.step('Copying content into repository.')
2360
self.source_repo.copy_content_into(converted)
2363
self.step('Deleting old repository content.')
2364
self.repo_dir.transport.delete_tree('repository.backup')
2365
self.pb.note('repository converted')
2367
def step(self, message):
2368
"""Update the pb by a step."""
2370
self.pb.update(message, self.count, self.total)
2382
def _unescaper(match, _map=_unescape_map):
2383
code = match.group(1)
2387
if not code.startswith('#'):
2389
return unichr(int(code[1:])).encode('utf8')
2395
def _unescape_xml(data):
2396
"""Unescape predefined XML entities in a string of data."""
2398
if _unescape_re is None:
2399
_unescape_re = re.compile('\&([^;]*);')
2400
return _unescape_re.sub(_unescaper, data)
added
removed
bzrlib/repository.py
