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
symbol_versioning.warn('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
# transitional assert only, will remove before release.
256
assert ie.kind == kind
257
self.new_inventory.add(ie)
259
# TODO: slow, take it out of the inner loop.
261
basis_inv = parent_invs[0]
263
basis_inv = Inventory(root_id=None)
265
# ie.revision is always None if the InventoryEntry is considered
266
# for committing. We may record the previous parents revision if the
267
# content is actually unchanged against a sole head.
268
if ie.revision is not None:
269
if self._versioned_root or path != '':
270
# not considered for commit
273
# repositories that do not version the root set the root's
274
# revision to the new commit even when no change occurs, and
275
# this masks when a change may have occurred against the basis,
276
# so calculate if one happened.
277
if ie.file_id not in basis_inv:
279
delta = (None, path, ie.file_id, ie)
281
basis_id = basis_inv[ie.file_id]
282
if basis_id.name != '':
284
delta = (basis_inv.id2path(ie.file_id), path,
289
# not considered for commit, OR, for non-rich-root
290
return delta, ie.revision == self._new_revision_id and (path != '' or
291
self._versioned_root)
293
# XXX: Friction: parent_candidates should return a list not a dict
294
# so that we don't have to walk the inventories again.
295
parent_candiate_entries = ie.parent_candidates(parent_invs)
296
head_set = self._repo_graph.heads(parent_candiate_entries.keys())
298
for inv in parent_invs:
299
if ie.file_id in inv:
300
old_rev = inv[ie.file_id].revision
301
if old_rev in head_set:
302
heads.append(inv[ie.file_id].revision)
303
head_set.remove(inv[ie.file_id].revision)
306
# now we check to see if we need to write a new record to the
308
# We write a new entry unless there is one head to the ancestors, and
309
# the kind-derived content is unchanged.
311
# Cheapest check first: no ancestors, or more the one head in the
312
# ancestors, we write a new node.
316
# There is a single head, look it up for comparison
317
parent_entry = parent_candiate_entries[heads[0]]
318
# if the non-content specific data has changed, we'll be writing a
320
if (parent_entry.parent_id != ie.parent_id or
321
parent_entry.name != ie.name):
323
# now we need to do content specific checks:
325
# if the kind changed the content obviously has
326
if kind != parent_entry.kind:
330
if (# if the file length changed we have to store:
331
parent_entry.text_size != content_summary[1] or
332
# if the exec bit has changed we have to store:
333
parent_entry.executable != content_summary[2]):
335
elif parent_entry.text_sha1 == content_summary[3]:
336
# all meta and content is unchanged (using a hash cache
337
# hit to check the sha)
338
ie.revision = parent_entry.revision
339
ie.text_size = parent_entry.text_size
340
ie.text_sha1 = parent_entry.text_sha1
341
ie.executable = parent_entry.executable
342
return self._get_delta(ie, basis_inv, path), False
344
# Either there is only a hash change(no hash cache entry,
345
# or same size content change), or there is no change on
347
# Provide the parent's hash to the store layer, so that the
348
# content is unchanged we will not store a new node.
349
nostore_sha = parent_entry.text_sha1
351
# We want to record a new node regardless of the presence or
352
# absence of a content change in the file.
354
ie.executable = content_summary[2]
355
lines = tree.get_file(ie.file_id, path).readlines()
357
ie.text_sha1, ie.text_size = self._add_text_to_weave(
358
ie.file_id, lines, heads, nostore_sha)
359
except errors.ExistingContent:
360
# Turns out that the file content was unchanged, and we were
361
# only going to store a new node if it was changed. Carry over
363
ie.revision = parent_entry.revision
364
ie.text_size = parent_entry.text_size
365
ie.text_sha1 = parent_entry.text_sha1
366
ie.executable = parent_entry.executable
367
return self._get_delta(ie, basis_inv, path), False
368
elif kind == 'directory':
370
# all data is meta here, nothing specific to directory, so
372
ie.revision = parent_entry.revision
373
return self._get_delta(ie, basis_inv, path), False
375
self._add_text_to_weave(ie.file_id, lines, heads, None)
376
elif kind == 'symlink':
377
current_link_target = content_summary[3]
379
# symlink target is not generic metadata, check if it has
381
if current_link_target != parent_entry.symlink_target:
384
# unchanged, carry over.
385
ie.revision = parent_entry.revision
386
ie.symlink_target = parent_entry.symlink_target
387
return self._get_delta(ie, basis_inv, path), False
388
ie.symlink_target = current_link_target
390
self._add_text_to_weave(ie.file_id, lines, heads, None)
391
elif kind == 'tree-reference':
393
if content_summary[3] != parent_entry.reference_revision:
396
# unchanged, carry over.
397
ie.reference_revision = parent_entry.reference_revision
398
ie.revision = parent_entry.revision
399
return self._get_delta(ie, basis_inv, path), False
400
ie.reference_revision = content_summary[3]
402
self._add_text_to_weave(ie.file_id, lines, heads, None)
404
raise NotImplementedError('unknown kind')
405
ie.revision = self._new_revision_id
406
return self._get_delta(ie, basis_inv, path), True
408
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
409
versionedfile = self.repository.weave_store.get_weave_or_empty(
410
file_id, self.repository.get_transaction())
411
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
412
# than add_lines, and allows committing when a parent is ghosted for
414
# Note: as we read the content directly from the tree, we know its not
415
# been turned into unicode or badly split - but a broken tree
416
# implementation could give us bad output from readlines() so this is
417
# not a guarantee of safety. What would be better is always checking
418
# the content during test suite execution. RBC 20070912
420
return versionedfile.add_lines_with_ghosts(
421
self._new_revision_id, parents, new_lines,
422
nostore_sha=nostore_sha, random_id=self.random_revid,
423
check_content=False)[0:2]
425
versionedfile.clear_cache()
428
class RootCommitBuilder(CommitBuilder):
429
"""This commitbuilder actually records the root id"""
431
# the root entry gets versioned properly by this builder.
432
_versioned_root = True
434
def _check_root(self, ie, parent_invs, tree):
435
"""Helper for record_entry_contents.
437
:param ie: An entry being added.
438
:param parent_invs: The inventories of the parent revisions of the
440
:param tree: The tree that is being committed.
444
######################################################################
447
class Repository(object):
448
"""Repository holding history for one or more branches.
450
The repository holds and retrieves historical information including
451
revisions and file history. It's normally accessed only by the Branch,
452
which views a particular line of development through that history.
454
The Repository builds on top of Stores and a Transport, which respectively
455
describe the disk data format and the way of accessing the (possibly
459
# What class to use for a CommitBuilder. Often its simpler to change this
460
# in a Repository class subclass rather than to override
461
# get_commit_builder.
462
_commit_builder_class = CommitBuilder
463
# The search regex used by xml based repositories to determine what things
464
# where changed in a single commit.
465
_file_ids_altered_regex = lazy_regex.lazy_compile(
466
r'file_id="(?P<file_id>[^"]+)"'
467
r'.* revision="(?P<revision_id>[^"]+)"'
470
def abort_write_group(self):
471
"""Commit the contents accrued within the current write group.
473
:seealso: start_write_group.
475
if self._write_group is not self.get_transaction():
476
# has an unlock or relock occured ?
477
raise errors.BzrError('mismatched lock context and write group.')
478
self._abort_write_group()
479
self._write_group = None
481
def _abort_write_group(self):
482
"""Template method for per-repository write group cleanup.
484
This is called during abort before the write group is considered to be
485
finished and should cleanup any internal state accrued during the write
486
group. There is no requirement that data handed to the repository be
487
*not* made available - this is not a rollback - but neither should any
488
attempt be made to ensure that data added is fully commited. Abort is
489
invoked when an error has occured so futher disk or network operations
490
may not be possible or may error and if possible should not be
495
def add_inventory(self, revision_id, inv, parents):
496
"""Add the inventory inv to the repository as revision_id.
498
:param parents: The revision ids of the parents that revision_id
499
is known to have and are in the repository already.
501
returns the sha1 of the serialized inventory.
503
_mod_revision.check_not_reserved_id(revision_id)
504
assert inv.revision_id is None or inv.revision_id == revision_id, \
505
"Mismatch between inventory revision" \
506
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
507
assert inv.root is not None
508
inv_lines = self._serialise_inventory_to_lines(inv)
509
inv_vf = self.get_inventory_weave()
510
return self._inventory_add_lines(inv_vf, revision_id, parents,
511
inv_lines, check_content=False)
513
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
515
"""Store lines in inv_vf and return the sha1 of the inventory."""
517
for parent in parents:
519
final_parents.append(parent)
520
return inv_vf.add_lines(revision_id, final_parents, lines,
521
check_content=check_content)[0]
524
def add_revision(self, revision_id, rev, inv=None, config=None):
525
"""Add rev to the revision store as revision_id.
527
:param revision_id: the revision id to use.
528
:param rev: The revision object.
529
:param inv: The inventory for the revision. if None, it will be looked
530
up in the inventory storer
531
:param config: If None no digital signature will be created.
532
If supplied its signature_needed method will be used
533
to determine if a signature should be made.
535
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
537
_mod_revision.check_not_reserved_id(revision_id)
538
if config is not None and config.signature_needed():
540
inv = self.get_inventory(revision_id)
541
plaintext = Testament(rev, inv).as_short_text()
542
self.store_revision_signature(
543
gpg.GPGStrategy(config), plaintext, revision_id)
544
if not revision_id in self.get_inventory_weave():
546
raise errors.WeaveRevisionNotPresent(revision_id,
547
self.get_inventory_weave())
549
# yes, this is not suitable for adding with ghosts.
550
self.add_inventory(revision_id, inv, rev.parent_ids)
551
self._revision_store.add_revision(rev, self.get_transaction())
553
def _add_revision_text(self, revision_id, text):
554
revision = self._revision_store._serializer.read_revision_from_string(
556
self._revision_store._add_revision(revision, StringIO(text),
557
self.get_transaction())
559
def all_revision_ids(self):
560
"""Returns a list of all the revision ids in the repository.
562
This is deprecated because code should generally work on the graph
563
reachable from a particular revision, and ignore any other revisions
564
that might be present. There is no direct replacement method.
566
if 'evil' in debug.debug_flags:
567
mutter_callsite(2, "all_revision_ids is linear with history.")
568
return self._all_revision_ids()
570
def _all_revision_ids(self):
571
"""Returns a list of all the revision ids in the repository.
573
These are in as much topological order as the underlying store can
576
raise NotImplementedError(self._all_revision_ids)
578
def break_lock(self):
579
"""Break a lock if one is present from another instance.
581
Uses the ui factory to ask for confirmation if the lock may be from
584
self.control_files.break_lock()
587
def _eliminate_revisions_not_present(self, revision_ids):
588
"""Check every revision id in revision_ids to see if we have it.
590
Returns a set of the present revisions.
593
for id in revision_ids:
594
if self.has_revision(id):
599
def create(a_bzrdir):
600
"""Construct the current default format repository in a_bzrdir."""
601
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
603
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
604
"""instantiate a Repository.
606
:param _format: The format of the repository on disk.
607
:param a_bzrdir: The BzrDir of the repository.
609
In the future we will have a single api for all stores for
610
getting file texts, inventories and revisions, then
611
this construct will accept instances of those things.
613
super(Repository, self).__init__()
614
self._format = _format
615
# the following are part of the public API for Repository:
616
self.bzrdir = a_bzrdir
617
self.control_files = control_files
618
self._revision_store = _revision_store
619
# backwards compatibility
620
self.weave_store = text_store
622
self._reconcile_does_inventory_gc = True
623
# not right yet - should be more semantically clear ?
625
self.control_store = control_store
626
self.control_weaves = control_store
627
# TODO: make sure to construct the right store classes, etc, depending
628
# on whether escaping is required.
629
self._warn_if_deprecated()
630
self._write_group = None
631
self.base = control_files._transport.base
634
return '%s(%r)' % (self.__class__.__name__,
637
def has_same_location(self, other):
638
"""Returns a boolean indicating if this repository is at the same
639
location as another repository.
641
This might return False even when two repository objects are accessing
642
the same physical repository via different URLs.
644
if self.__class__ is not other.__class__:
646
return (self.control_files._transport.base ==
647
other.control_files._transport.base)
649
def is_in_write_group(self):
650
"""Return True if there is an open write group.
652
:seealso: start_write_group.
654
return self._write_group is not None
657
return self.control_files.is_locked()
659
def lock_write(self, token=None):
660
"""Lock this repository for writing.
662
This causes caching within the repository obejct to start accumlating
663
data during reads, and allows a 'write_group' to be obtained. Write
664
groups must be used for actual data insertion.
666
:param token: if this is already locked, then lock_write will fail
667
unless the token matches the existing lock.
668
:returns: a token if this instance supports tokens, otherwise None.
669
:raises TokenLockingNotSupported: when a token is given but this
670
instance doesn't support using token locks.
671
:raises MismatchedToken: if the specified token doesn't match the token
672
of the existing lock.
673
:seealso: start_write_group.
675
A token should be passed in if you know that you have locked the object
676
some other way, and need to synchronise this object's state with that
679
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
681
result = self.control_files.lock_write(token=token)
686
self.control_files.lock_read()
689
def get_physical_lock_status(self):
690
return self.control_files.get_physical_lock_status()
692
def leave_lock_in_place(self):
693
"""Tell this repository not to release the physical lock when this
696
If lock_write doesn't return a token, then this method is not supported.
698
self.control_files.leave_in_place()
700
def dont_leave_lock_in_place(self):
701
"""Tell this repository to release the physical lock when this
702
object is unlocked, even if it didn't originally acquire it.
704
If lock_write doesn't return a token, then this method is not supported.
706
self.control_files.dont_leave_in_place()
709
def gather_stats(self, revid=None, committers=None):
710
"""Gather statistics from a revision id.
712
:param revid: The revision id to gather statistics from, if None, then
713
no revision specific statistics are gathered.
714
:param committers: Optional parameter controlling whether to grab
715
a count of committers from the revision specific statistics.
716
:return: A dictionary of statistics. Currently this contains:
717
committers: The number of committers if requested.
718
firstrev: A tuple with timestamp, timezone for the penultimate left
719
most ancestor of revid, if revid is not the NULL_REVISION.
720
latestrev: A tuple with timestamp, timezone for revid, if revid is
721
not the NULL_REVISION.
722
revisions: The total revision count in the repository.
723
size: An estimate disk size of the repository in bytes.
726
if revid and committers:
727
result['committers'] = 0
728
if revid and revid != _mod_revision.NULL_REVISION:
730
all_committers = set()
731
revisions = self.get_ancestry(revid)
732
# pop the leading None
734
first_revision = None
736
# ignore the revisions in the middle - just grab first and last
737
revisions = revisions[0], revisions[-1]
738
for revision in self.get_revisions(revisions):
739
if not first_revision:
740
first_revision = revision
742
all_committers.add(revision.committer)
743
last_revision = revision
745
result['committers'] = len(all_committers)
746
result['firstrev'] = (first_revision.timestamp,
747
first_revision.timezone)
748
result['latestrev'] = (last_revision.timestamp,
749
last_revision.timezone)
751
# now gather global repository information
752
if self.bzrdir.root_transport.listable():
753
c, t = self._revision_store.total_size(self.get_transaction())
754
result['revisions'] = c
759
def missing_revision_ids(self, other, revision_id=None):
760
"""Return the revision ids that other has that this does not.
762
These are returned in topological order.
764
revision_id: only return revision ids included by revision_id.
766
return InterRepository.get(other, self).missing_revision_ids(revision_id)
770
"""Open the repository rooted at base.
772
For instance, if the repository is at URL/.bzr/repository,
773
Repository.open(URL) -> a Repository instance.
775
control = bzrdir.BzrDir.open(base)
776
return control.open_repository()
778
def copy_content_into(self, destination, revision_id=None):
779
"""Make a complete copy of the content in self into destination.
781
This is a destructive operation! Do not use it on existing
784
return InterRepository.get(self, destination).copy_content(revision_id)
786
def commit_write_group(self):
787
"""Commit the contents accrued within the current write group.
789
:seealso: start_write_group.
791
if self._write_group is not self.get_transaction():
792
# has an unlock or relock occured ?
793
raise errors.BzrError('mismatched lock context %r and '
795
(self.get_transaction(), self._write_group))
796
self._commit_write_group()
797
self._write_group = None
799
def _commit_write_group(self):
800
"""Template method for per-repository write group cleanup.
802
This is called before the write group is considered to be
803
finished and should ensure that all data handed to the repository
804
for writing during the write group is safely committed (to the
805
extent possible considering file system caching etc).
808
def fetch(self, source, revision_id=None, pb=None):
809
"""Fetch the content required to construct revision_id from source.
811
If revision_id is None all content is copied.
813
# fast path same-url fetch operations
814
if self.has_same_location(source):
815
# check that last_revision is in 'from' and then return a
817
if (revision_id is not None and
818
not _mod_revision.is_null(revision_id)):
819
self.get_revision(revision_id)
821
inter = InterRepository.get(source, self)
823
return inter.fetch(revision_id=revision_id, pb=pb)
824
except NotImplementedError:
825
raise errors.IncompatibleRepositories(source, self)
827
def create_bundle(self, target, base, fileobj, format=None):
828
return serializer.write_bundle(self, target, base, fileobj, format)
830
def get_commit_builder(self, branch, parents, config, timestamp=None,
831
timezone=None, committer=None, revprops=None,
833
"""Obtain a CommitBuilder for this repository.
835
:param branch: Branch to commit to.
836
:param parents: Revision ids of the parents of the new revision.
837
:param config: Configuration to use.
838
:param timestamp: Optional timestamp recorded for commit.
839
:param timezone: Optional timezone for timestamp.
840
:param committer: Optional committer to set for commit.
841
:param revprops: Optional dictionary of revision properties.
842
:param revision_id: Optional revision id.
844
result = self._commit_builder_class(self, parents, config,
845
timestamp, timezone, committer, revprops, revision_id)
846
self.start_write_group()
850
if (self.control_files._lock_count == 1 and
851
self.control_files._lock_mode == 'w'):
852
if self._write_group is not None:
853
raise errors.BzrError(
854
'Must end write groups before releasing write locks.')
855
self.control_files.unlock()
858
def clone(self, a_bzrdir, revision_id=None):
859
"""Clone this repository into a_bzrdir using the current format.
861
Currently no check is made that the format of this repository and
862
the bzrdir format are compatible. FIXME RBC 20060201.
864
:return: The newly created destination repository.
866
# TODO: deprecate after 0.16; cloning this with all its settings is
867
# probably not very useful -- mbp 20070423
868
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
869
self.copy_content_into(dest_repo, revision_id)
872
def start_write_group(self):
873
"""Start a write group in the repository.
875
Write groups are used by repositories which do not have a 1:1 mapping
876
between file ids and backend store to manage the insertion of data from
877
both fetch and commit operations.
879
A write lock is required around the start_write_group/commit_write_group
880
for the support of lock-requiring repository formats.
882
One can only insert data into a repository inside a write group.
886
if not self.is_locked() or self.control_files._lock_mode != 'w':
887
raise errors.NotWriteLocked(self)
888
if self._write_group:
889
raise errors.BzrError('already in a write group')
890
self._start_write_group()
891
# so we can detect unlock/relock - the write group is now entered.
892
self._write_group = self.get_transaction()
894
def _start_write_group(self):
895
"""Template method for per-repository write group startup.
897
This is called before the write group is considered to be
902
def sprout(self, to_bzrdir, revision_id=None):
903
"""Create a descendent repository for new development.
905
Unlike clone, this does not copy the settings of the repository.
907
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
908
dest_repo.fetch(self, revision_id=revision_id)
911
def _create_sprouting_repo(self, a_bzrdir, shared):
912
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
913
# use target default format.
914
dest_repo = a_bzrdir.create_repository()
916
# Most control formats need the repository to be specifically
917
# created, but on some old all-in-one formats it's not needed
919
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
920
except errors.UninitializableFormat:
921
dest_repo = a_bzrdir.open_repository()
925
def has_revision(self, revision_id):
926
"""True if this repository has a copy of the revision."""
927
if 'evil' in debug.debug_flags:
928
mutter_callsite(3, "has_revision is a LBYL symptom.")
929
return self._revision_store.has_revision_id(revision_id,
930
self.get_transaction())
933
def get_revision(self, revision_id):
934
"""Return the Revision object for a named revision."""
935
return self.get_revisions([revision_id])[0]
938
def get_revision_reconcile(self, revision_id):
939
"""'reconcile' helper routine that allows access to a revision always.
941
This variant of get_revision does not cross check the weave graph
942
against the revision one as get_revision does: but it should only
943
be used by reconcile, or reconcile-alike commands that are correcting
944
or testing the revision graph.
946
return self._get_revisions([revision_id])[0]
949
def get_revisions(self, revision_ids):
950
"""Get many revisions at once."""
951
return self._get_revisions(revision_ids)
954
def _get_revisions(self, revision_ids):
955
"""Core work logic to get many revisions without sanity checks."""
956
for rev_id in revision_ids:
957
if not rev_id or not isinstance(rev_id, basestring):
958
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
959
revs = self._revision_store.get_revisions(revision_ids,
960
self.get_transaction())
962
assert not isinstance(rev.revision_id, unicode)
963
for parent_id in rev.parent_ids:
964
assert not isinstance(parent_id, unicode)
968
def get_revision_xml(self, revision_id):
969
# TODO: jam 20070210 This shouldn't be necessary since get_revision
970
# would have already do it.
971
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
972
rev = self.get_revision(revision_id)
974
# the current serializer..
975
self._revision_store._serializer.write_revision(rev, rev_tmp)
977
return rev_tmp.getvalue()
980
def get_deltas_for_revisions(self, revisions):
981
"""Produce a generator of revision deltas.
983
Note that the input is a sequence of REVISIONS, not revision_ids.
984
Trees will be held in memory until the generator exits.
985
Each delta is relative to the revision's lefthand predecessor.
987
required_trees = set()
988
for revision in revisions:
989
required_trees.add(revision.revision_id)
990
required_trees.update(revision.parent_ids[:1])
991
trees = dict((t.get_revision_id(), t) for
992
t in self.revision_trees(required_trees))
993
for revision in revisions:
994
if not revision.parent_ids:
995
old_tree = self.revision_tree(None)
997
old_tree = trees[revision.parent_ids[0]]
998
yield trees[revision.revision_id].changes_from(old_tree)
1001
def get_revision_delta(self, revision_id):
1002
"""Return the delta for one revision.
1004
The delta is relative to the left-hand predecessor of the
1007
r = self.get_revision(revision_id)
1008
return list(self.get_deltas_for_revisions([r]))[0]
1011
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1012
signature = gpg_strategy.sign(plaintext)
1013
self._revision_store.add_revision_signature_text(revision_id,
1015
self.get_transaction())
1017
def fileids_altered_by_revision_ids(self, revision_ids):
1018
"""Find the file ids and versions affected by revisions.
1020
:param revisions: an iterable containing revision ids.
1021
:return: a dictionary mapping altered file-ids to an iterable of
1022
revision_ids. Each altered file-ids has the exact revision_ids that
1023
altered it listed explicitly.
1025
assert self._serializer.support_altered_by_hack, \
1026
("fileids_altered_by_revision_ids only supported for branches "
1027
"which store inventory as unnested xml, not on %r" % self)
1028
selected_revision_ids = set(revision_ids)
1029
w = self.get_inventory_weave()
1032
# this code needs to read every new line in every inventory for the
1033
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1034
# not present in one of those inventories is unnecessary but not
1035
# harmful because we are filtering by the revision id marker in the
1036
# inventory lines : we only select file ids altered in one of those
1037
# revisions. We don't need to see all lines in the inventory because
1038
# only those added in an inventory in rev X can contain a revision=X
1040
unescape_revid_cache = {}
1041
unescape_fileid_cache = {}
1043
# jam 20061218 In a big fetch, this handles hundreds of thousands
1044
# of lines, so it has had a lot of inlining and optimizing done.
1045
# Sorry that it is a little bit messy.
1046
# Move several functions to be local variables, since this is a long
1048
search = self._file_ids_altered_regex.search
1049
unescape = _unescape_xml
1050
setdefault = result.setdefault
1051
pb = ui.ui_factory.nested_progress_bar()
1053
for line in w.iter_lines_added_or_present_in_versions(
1054
selected_revision_ids, pb=pb):
1055
match = search(line)
1058
# One call to match.group() returning multiple items is quite a
1059
# bit faster than 2 calls to match.group() each returning 1
1060
file_id, revision_id = match.group('file_id', 'revision_id')
1062
# Inlining the cache lookups helps a lot when you make 170,000
1063
# lines and 350k ids, versus 8.4 unique ids.
1064
# Using a cache helps in 2 ways:
1065
# 1) Avoids unnecessary decoding calls
1066
# 2) Re-uses cached strings, which helps in future set and
1068
# (2) is enough that removing encoding entirely along with
1069
# the cache (so we are using plain strings) results in no
1070
# performance improvement.
1072
revision_id = unescape_revid_cache[revision_id]
1074
unescaped = unescape(revision_id)
1075
unescape_revid_cache[revision_id] = unescaped
1076
revision_id = unescaped
1078
if revision_id in selected_revision_ids:
1080
file_id = unescape_fileid_cache[file_id]
1082
unescaped = unescape(file_id)
1083
unescape_fileid_cache[file_id] = unescaped
1085
setdefault(file_id, set()).add(revision_id)
1090
def iter_files_bytes(self, desired_files):
1091
"""Iterate through file versions.
1093
Files will not necessarily be returned in the order they occur in
1094
desired_files. No specific order is guaranteed.
1096
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1097
value supplied by the caller as part of desired_files. It should
1098
uniquely identify the file version in the caller's context. (Examples:
1099
an index number or a TreeTransform trans_id.)
1101
bytes_iterator is an iterable of bytestrings for the file. The
1102
kind of iterable and length of the bytestrings are unspecified, but for
1103
this implementation, it is a list of lines produced by
1104
VersionedFile.get_lines().
1106
:param desired_files: a list of (file_id, revision_id, identifier)
1109
transaction = self.get_transaction()
1110
for file_id, revision_id, callable_data in desired_files:
1112
weave = self.weave_store.get_weave(file_id, transaction)
1113
except errors.NoSuchFile:
1114
raise errors.NoSuchIdInRepository(self, file_id)
1115
yield callable_data, weave.get_lines(revision_id)
1117
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1118
"""Get an iterable listing the keys of all the data introduced by a set
1121
The keys will be ordered so that the corresponding items can be safely
1122
fetched and inserted in that order.
1124
:returns: An iterable producing tuples of (knit-kind, file-id,
1125
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1126
'revisions'. file-id is None unless knit-kind is 'file'.
1128
# XXX: it's a bit weird to control the inventory weave caching in this
1129
# generator. Ideally the caching would be done in fetch.py I think. Or
1130
# maybe this generator should explicitly have the contract that it
1131
# should not be iterated until the previously yielded item has been
1133
inv_w = self.get_inventory_weave()
1134
inv_w.enable_cache()
1136
# file ids that changed
1137
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1139
num_file_ids = len(file_ids)
1140
for file_id, altered_versions in file_ids.iteritems():
1141
if _files_pb is not None:
1142
_files_pb.update("fetch texts", count, num_file_ids)
1144
yield ("file", file_id, altered_versions)
1145
# We're done with the files_pb. Note that it finished by the caller,
1146
# just as it was created by the caller.
1150
yield ("inventory", None, revision_ids)
1154
revisions_with_signatures = set()
1155
for rev_id in revision_ids:
1157
self.get_signature_text(rev_id)
1158
except errors.NoSuchRevision:
1162
revisions_with_signatures.add(rev_id)
1163
yield ("signatures", None, revisions_with_signatures)
1166
yield ("revisions", None, revision_ids)
1169
def get_inventory_weave(self):
1170
return self.control_weaves.get_weave('inventory',
1171
self.get_transaction())
1174
def get_inventory(self, revision_id):
1175
"""Get Inventory object by hash."""
1176
return self.deserialise_inventory(
1177
revision_id, self.get_inventory_xml(revision_id))
1179
def deserialise_inventory(self, revision_id, xml):
1180
"""Transform the xml into an inventory object.
1182
:param revision_id: The expected revision id of the inventory.
1183
:param xml: A serialised inventory.
1185
return self._serializer.read_inventory_from_string(xml, revision_id)
1187
def serialise_inventory(self, inv):
1188
return self._serializer.write_inventory_to_string(inv)
1190
def _serialise_inventory_to_lines(self, inv):
1191
return self._serializer.write_inventory_to_lines(inv)
1193
def get_serializer_format(self):
1194
return self._serializer.format_num
1197
def get_inventory_xml(self, revision_id):
1198
"""Get inventory XML as a file object."""
1200
assert isinstance(revision_id, str), type(revision_id)
1201
iw = self.get_inventory_weave()
1202
return iw.get_text(revision_id)
1204
raise errors.HistoryMissing(self, 'inventory', revision_id)
1207
def get_inventory_sha1(self, revision_id):
1208
"""Return the sha1 hash of the inventory entry
1210
return self.get_revision(revision_id).inventory_sha1
1213
def get_revision_graph(self, revision_id=None):
1214
"""Return a dictionary containing the revision graph.
1216
NB: This method should not be used as it accesses the entire graph all
1217
at once, which is much more data than most operations should require.
1219
:param revision_id: The revision_id to get a graph from. If None, then
1220
the entire revision graph is returned. This is a deprecated mode of
1221
operation and will be removed in the future.
1222
:return: a dictionary of revision_id->revision_parents_list.
1224
raise NotImplementedError(self.get_revision_graph)
1227
def get_revision_graph_with_ghosts(self, revision_ids=None):
1228
"""Return a graph of the revisions with ghosts marked as applicable.
1230
:param revision_ids: an iterable of revisions to graph or None for all.
1231
:return: a Graph object with the graph reachable from revision_ids.
1233
if 'evil' in debug.debug_flags:
1235
"get_revision_graph_with_ghosts scales with size of history.")
1236
result = deprecated_graph.Graph()
1237
if not revision_ids:
1238
pending = set(self.all_revision_ids())
1241
pending = set(revision_ids)
1242
# special case NULL_REVISION
1243
if _mod_revision.NULL_REVISION in pending:
1244
pending.remove(_mod_revision.NULL_REVISION)
1245
required = set(pending)
1248
revision_id = pending.pop()
1250
rev = self.get_revision(revision_id)
1251
except errors.NoSuchRevision:
1252
if revision_id in required:
1255
result.add_ghost(revision_id)
1257
for parent_id in rev.parent_ids:
1258
# is this queued or done ?
1259
if (parent_id not in pending and
1260
parent_id not in done):
1262
pending.add(parent_id)
1263
result.add_node(revision_id, rev.parent_ids)
1264
done.add(revision_id)
1267
def _get_history_vf(self):
1268
"""Get a versionedfile whose history graph reflects all revisions.
1270
For weave repositories, this is the inventory weave.
1272
return self.get_inventory_weave()
1274
def iter_reverse_revision_history(self, revision_id):
1275
"""Iterate backwards through revision ids in the lefthand history
1277
:param revision_id: The revision id to start with. All its lefthand
1278
ancestors will be traversed.
1280
if revision_id in (None, _mod_revision.NULL_REVISION):
1282
next_id = revision_id
1283
versionedfile = self._get_history_vf()
1286
parents = versionedfile.get_parents(next_id)
1287
if len(parents) == 0:
1290
next_id = parents[0]
1293
def get_revision_inventory(self, revision_id):
1294
"""Return inventory of a past revision."""
1295
# TODO: Unify this with get_inventory()
1296
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1297
# must be the same as its revision, so this is trivial.
1298
if revision_id is None:
1299
# This does not make sense: if there is no revision,
1300
# then it is the current tree inventory surely ?!
1301
# and thus get_root_id() is something that looks at the last
1302
# commit on the branch, and the get_root_id is an inventory check.
1303
raise NotImplementedError
1304
# return Inventory(self.get_root_id())
1306
return self.get_inventory(revision_id)
1309
def is_shared(self):
1310
"""Return True if this repository is flagged as a shared repository."""
1311
raise NotImplementedError(self.is_shared)
1314
def reconcile(self, other=None, thorough=False):
1315
"""Reconcile this repository."""
1316
from bzrlib.reconcile import RepoReconciler
1317
reconciler = RepoReconciler(self, thorough=thorough)
1318
reconciler.reconcile()
1321
def _refresh_data(self):
1322
"""Helper called from lock_* to ensure coherency with disk.
1324
The default implementation does nothing; it is however possible
1325
for repositories to maintain loaded indices across multiple locks
1326
by checking inside their implementation of this method to see
1327
whether their indices are still valid. This depends of course on
1328
the disk format being validatable in this manner.
1332
def revision_tree(self, revision_id):
1333
"""Return Tree for a revision on this branch.
1335
`revision_id` may be None for the empty tree revision.
1337
# TODO: refactor this to use an existing revision object
1338
# so we don't need to read it in twice.
1339
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1340
return RevisionTree(self, Inventory(root_id=None),
1341
_mod_revision.NULL_REVISION)
1343
inv = self.get_revision_inventory(revision_id)
1344
return RevisionTree(self, inv, revision_id)
1347
def revision_trees(self, revision_ids):
1348
"""Return Tree for a revision on this branch.
1350
`revision_id` may not be None or 'null:'"""
1351
assert None not in revision_ids
1352
assert _mod_revision.NULL_REVISION not in revision_ids
1353
texts = self.get_inventory_weave().get_texts(revision_ids)
1354
for text, revision_id in zip(texts, revision_ids):
1355
inv = self.deserialise_inventory(revision_id, text)
1356
yield RevisionTree(self, inv, revision_id)
1359
def get_ancestry(self, revision_id, topo_sorted=True):
1360
"""Return a list of revision-ids integrated by a revision.
1362
The first element of the list is always None, indicating the origin
1363
revision. This might change when we have history horizons, or
1364
perhaps we should have a new API.
1366
This is topologically sorted.
1368
if _mod_revision.is_null(revision_id):
1370
if not self.has_revision(revision_id):
1371
raise errors.NoSuchRevision(self, revision_id)
1372
w = self.get_inventory_weave()
1373
candidates = w.get_ancestry(revision_id, topo_sorted)
1374
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1377
"""Compress the data within the repository.
1379
This operation only makes sense for some repository types. For other
1380
types it should be a no-op that just returns.
1382
This stub method does not require a lock, but subclasses should use
1383
@needs_write_lock as this is a long running call its reasonable to
1384
implicitly lock for the user.
1388
def print_file(self, file, revision_id):
1389
"""Print `file` to stdout.
1391
FIXME RBC 20060125 as John Meinel points out this is a bad api
1392
- it writes to stdout, it assumes that that is valid etc. Fix
1393
by creating a new more flexible convenience function.
1395
tree = self.revision_tree(revision_id)
1396
# use inventory as it was in that revision
1397
file_id = tree.inventory.path2id(file)
1399
# TODO: jam 20060427 Write a test for this code path
1400
# it had a bug in it, and was raising the wrong
1402
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1403
tree.print_file(file_id)
1405
def get_transaction(self):
1406
return self.control_files.get_transaction()
1408
def revision_parents(self, revision_id):
1409
return self.get_inventory_weave().parent_names(revision_id)
1411
def get_parents(self, revision_ids):
1412
"""See StackedParentsProvider.get_parents"""
1414
for revision_id in revision_ids:
1415
if revision_id == _mod_revision.NULL_REVISION:
1419
parents = self.get_revision(revision_id).parent_ids
1420
except errors.NoSuchRevision:
1423
if len(parents) == 0:
1424
parents = [_mod_revision.NULL_REVISION]
1425
parents_list.append(parents)
1428
def _make_parents_provider(self):
1431
def get_graph(self, other_repository=None):
1432
"""Return the graph walker for this repository format"""
1433
parents_provider = self._make_parents_provider()
1434
if (other_repository is not None and
1435
other_repository.bzrdir.transport.base !=
1436
self.bzrdir.transport.base):
1437
parents_provider = graph._StackedParentsProvider(
1438
[parents_provider, other_repository._make_parents_provider()])
1439
return graph.Graph(parents_provider)
1442
def set_make_working_trees(self, new_value):
1443
"""Set the policy flag for making working trees when creating branches.
1445
This only applies to branches that use this repository.
1447
The default is 'True'.
1448
:param new_value: True to restore the default, False to disable making
1451
raise NotImplementedError(self.set_make_working_trees)
1453
def make_working_trees(self):
1454
"""Returns the policy for making working trees on new branches."""
1455
raise NotImplementedError(self.make_working_trees)
1458
def sign_revision(self, revision_id, gpg_strategy):
1459
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1460
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1463
def has_signature_for_revision_id(self, revision_id):
1464
"""Query for a revision signature for revision_id in the repository."""
1465
return self._revision_store.has_signature(revision_id,
1466
self.get_transaction())
1469
def get_signature_text(self, revision_id):
1470
"""Return the text for a signature."""
1471
return self._revision_store.get_signature_text(revision_id,
1472
self.get_transaction())
1475
def check(self, revision_ids):
1476
"""Check consistency of all history of given revision_ids.
1478
Different repository implementations should override _check().
1480
:param revision_ids: A non-empty list of revision_ids whose ancestry
1481
will be checked. Typically the last revision_id of a branch.
1483
if not revision_ids:
1484
raise ValueError("revision_ids must be non-empty in %s.check"
1486
return self._check(revision_ids)
1488
def _check(self, revision_ids):
1489
result = check.Check(self)
1493
def _warn_if_deprecated(self):
1494
global _deprecation_warning_done
1495
if _deprecation_warning_done:
1497
_deprecation_warning_done = True
1498
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1499
% (self._format, self.bzrdir.transport.base))
1501
def supports_rich_root(self):
1502
return self._format.rich_root_data
1504
def _check_ascii_revisionid(self, revision_id, method):
1505
"""Private helper for ascii-only repositories."""
1506
# weave repositories refuse to store revisionids that are non-ascii.
1507
if revision_id is not None:
1508
# weaves require ascii revision ids.
1509
if isinstance(revision_id, unicode):
1511
revision_id.encode('ascii')
1512
except UnicodeEncodeError:
1513
raise errors.NonAsciiRevisionId(method, self)
1516
revision_id.decode('ascii')
1517
except UnicodeDecodeError:
1518
raise errors.NonAsciiRevisionId(method, self)
1522
# remove these delegates a while after bzr 0.15
1523
def __make_delegated(name, from_module):
1524
def _deprecated_repository_forwarder():
1525
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1526
% (name, from_module),
1529
m = __import__(from_module, globals(), locals(), [name])
1531
return getattr(m, name)
1532
except AttributeError:
1533
raise AttributeError('module %s has no name %s'
1535
globals()[name] = _deprecated_repository_forwarder
1538
'AllInOneRepository',
1539
'WeaveMetaDirRepository',
1540
'PreSplitOutRepositoryFormat',
1541
'RepositoryFormat4',
1542
'RepositoryFormat5',
1543
'RepositoryFormat6',
1544
'RepositoryFormat7',
1546
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1550
'RepositoryFormatKnit',
1551
'RepositoryFormatKnit1',
1553
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1556
def install_revision(repository, rev, revision_tree):
1557
"""Install all revision data into a repository."""
1558
present_parents = []
1560
for p_id in rev.parent_ids:
1561
if repository.has_revision(p_id):
1562
present_parents.append(p_id)
1563
parent_trees[p_id] = repository.revision_tree(p_id)
1565
parent_trees[p_id] = repository.revision_tree(None)
1567
inv = revision_tree.inventory
1568
entries = inv.iter_entries()
1569
# backwards compatibility hack: skip the root id.
1570
if not repository.supports_rich_root():
1571
path, root = entries.next()
1572
if root.revision != rev.revision_id:
1573
raise errors.IncompatibleRevision(repr(repository))
1574
# Add the texts that are not already present
1575
for path, ie in entries:
1576
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1577
repository.get_transaction())
1578
if ie.revision not in w:
1580
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1581
# with InventoryEntry.find_previous_heads(). if it is, then there
1582
# is a latent bug here where the parents may have ancestors of each
1584
for revision, tree in parent_trees.iteritems():
1585
if ie.file_id not in tree:
1587
parent_id = tree.inventory[ie.file_id].revision
1588
if parent_id in text_parents:
1590
text_parents.append(parent_id)
1592
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1593
repository.get_transaction())
1594
lines = revision_tree.get_file(ie.file_id).readlines()
1595
vfile.add_lines(rev.revision_id, text_parents, lines)
1597
# install the inventory
1598
repository.add_inventory(rev.revision_id, inv, present_parents)
1599
except errors.RevisionAlreadyPresent:
1601
repository.add_revision(rev.revision_id, rev, inv)
1604
class MetaDirRepository(Repository):
1605
"""Repositories in the new meta-dir layout."""
1607
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1608
super(MetaDirRepository, self).__init__(_format,
1614
dir_mode = self.control_files._dir_mode
1615
file_mode = self.control_files._file_mode
1618
def is_shared(self):
1619
"""Return True if this repository is flagged as a shared repository."""
1620
return self.control_files._transport.has('shared-storage')
1623
def set_make_working_trees(self, new_value):
1624
"""Set the policy flag for making working trees when creating branches.
1626
This only applies to branches that use this repository.
1628
The default is 'True'.
1629
:param new_value: True to restore the default, False to disable making
1634
self.control_files._transport.delete('no-working-trees')
1635
except errors.NoSuchFile:
1638
self.control_files.put_utf8('no-working-trees', '')
1640
def make_working_trees(self):
1641
"""Returns the policy for making working trees on new branches."""
1642
return not self.control_files._transport.has('no-working-trees')
1645
class RepositoryFormatRegistry(registry.Registry):
1646
"""Registry of RepositoryFormats."""
1648
def get(self, format_string):
1649
r = registry.Registry.get(self, format_string)
1655
format_registry = RepositoryFormatRegistry()
1656
"""Registry of formats, indexed by their identifying format string.
1658
This can contain either format instances themselves, or classes/factories that
1659
can be called to obtain one.
1663
#####################################################################
1664
# Repository Formats
1666
class RepositoryFormat(object):
1667
"""A repository format.
1669
Formats provide three things:
1670
* An initialization routine to construct repository data on disk.
1671
* a format string which is used when the BzrDir supports versioned
1673
* an open routine which returns a Repository instance.
1675
There is one and only one Format subclass for each on-disk format. But
1676
there can be one Repository subclass that is used for several different
1677
formats. The _format attribute on a Repository instance can be used to
1678
determine the disk format.
1680
Formats are placed in an dict by their format string for reference
1681
during opening. These should be subclasses of RepositoryFormat
1684
Once a format is deprecated, just deprecate the initialize and open
1685
methods on the format class. Do not deprecate the object, as the
1686
object will be created every system load.
1688
Common instance attributes:
1689
_matchingbzrdir - the bzrdir format that the repository format was
1690
originally written to work with. This can be used if manually
1691
constructing a bzrdir and repository, or more commonly for test suite
1696
return "<%s>" % self.__class__.__name__
1698
def __eq__(self, other):
1699
# format objects are generally stateless
1700
return isinstance(other, self.__class__)
1702
def __ne__(self, other):
1703
return not self == other
1706
def find_format(klass, a_bzrdir):
1707
"""Return the format for the repository object in a_bzrdir.
1709
This is used by bzr native formats that have a "format" file in
1710
the repository. Other methods may be used by different types of
1714
transport = a_bzrdir.get_repository_transport(None)
1715
format_string = transport.get("format").read()
1716
return format_registry.get(format_string)
1717
except errors.NoSuchFile:
1718
raise errors.NoRepositoryPresent(a_bzrdir)
1720
raise errors.UnknownFormatError(format=format_string)
1723
def register_format(klass, format):
1724
format_registry.register(format.get_format_string(), format)
1727
def unregister_format(klass, format):
1728
format_registry.remove(format.get_format_string())
1731
def get_default_format(klass):
1732
"""Return the current default format."""
1733
from bzrlib import bzrdir
1734
return bzrdir.format_registry.make_bzrdir('default').repository_format
1736
def _get_control_store(self, repo_transport, control_files):
1737
"""Return the control store for this repository."""
1738
raise NotImplementedError(self._get_control_store)
1740
def get_format_string(self):
1741
"""Return the ASCII format string that identifies this format.
1743
Note that in pre format ?? repositories the format string is
1744
not permitted nor written to disk.
1746
raise NotImplementedError(self.get_format_string)
1748
def get_format_description(self):
1749
"""Return the short description for this format."""
1750
raise NotImplementedError(self.get_format_description)
1752
def _get_revision_store(self, repo_transport, control_files):
1753
"""Return the revision store object for this a_bzrdir."""
1754
raise NotImplementedError(self._get_revision_store)
1756
def _get_text_rev_store(self,
1763
"""Common logic for getting a revision store for a repository.
1765
see self._get_revision_store for the subclass-overridable method to
1766
get the store for a repository.
1768
from bzrlib.store.revision.text import TextRevisionStore
1769
dir_mode = control_files._dir_mode
1770
file_mode = control_files._file_mode
1771
text_store = TextStore(transport.clone(name),
1773
compressed=compressed,
1775
file_mode=file_mode)
1776
_revision_store = TextRevisionStore(text_store, serializer)
1777
return _revision_store
1779
# TODO: this shouldn't be in the base class, it's specific to things that
1780
# use weaves or knits -- mbp 20070207
1781
def _get_versioned_file_store(self,
1786
versionedfile_class=None,
1787
versionedfile_kwargs={},
1789
if versionedfile_class is None:
1790
versionedfile_class = self._versionedfile_class
1791
weave_transport = control_files._transport.clone(name)
1792
dir_mode = control_files._dir_mode
1793
file_mode = control_files._file_mode
1794
return VersionedFileStore(weave_transport, prefixed=prefixed,
1796
file_mode=file_mode,
1797
versionedfile_class=versionedfile_class,
1798
versionedfile_kwargs=versionedfile_kwargs,
1801
def initialize(self, a_bzrdir, shared=False):
1802
"""Initialize a repository of this format in a_bzrdir.
1804
:param a_bzrdir: The bzrdir to put the new repository in it.
1805
:param shared: The repository should be initialized as a sharable one.
1806
:returns: The new repository object.
1808
This may raise UninitializableFormat if shared repository are not
1809
compatible the a_bzrdir.
1811
raise NotImplementedError(self.initialize)
1813
def is_supported(self):
1814
"""Is this format supported?
1816
Supported formats must be initializable and openable.
1817
Unsupported formats may not support initialization or committing or
1818
some other features depending on the reason for not being supported.
1822
def check_conversion_target(self, target_format):
1823
raise NotImplementedError(self.check_conversion_target)
1825
def open(self, a_bzrdir, _found=False):
1826
"""Return an instance of this format for the bzrdir a_bzrdir.
1828
_found is a private parameter, do not use it.
1830
raise NotImplementedError(self.open)
1833
class MetaDirRepositoryFormat(RepositoryFormat):
1834
"""Common base class for the new repositories using the metadir layout."""
1836
rich_root_data = False
1837
supports_tree_reference = False
1838
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
1841
super(MetaDirRepositoryFormat, self).__init__()
1843
def _create_control_files(self, a_bzrdir):
1844
"""Create the required files and the initial control_files object."""
1845
# FIXME: RBC 20060125 don't peek under the covers
1846
# NB: no need to escape relative paths that are url safe.
1847
repository_transport = a_bzrdir.get_repository_transport(self)
1848
control_files = lockable_files.LockableFiles(repository_transport,
1849
'lock', lockdir.LockDir)
1850
control_files.create_lock()
1851
return control_files
1853
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
1854
"""Upload the initial blank content."""
1855
control_files = self._create_control_files(a_bzrdir)
1856
control_files.lock_write()
1858
control_files._transport.mkdir_multi(dirs,
1859
mode=control_files._dir_mode)
1860
for file, content in files:
1861
control_files.put(file, content)
1862
for file, content in utf8_files:
1863
control_files.put_utf8(file, content)
1865
control_files.put_utf8('shared-storage', '')
1867
control_files.unlock()
1870
# formats which have no format string are not discoverable
1871
# and not independently creatable, so are not registered. They're
1872
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1873
# needed, it's constructed directly by the BzrDir. Non-native formats where
1874
# the repository is not separately opened are similar.
1876
format_registry.register_lazy(
1877
'Bazaar-NG Repository format 7',
1878
'bzrlib.repofmt.weaverepo',
1882
# KEEP in sync with bzrdir.format_registry default, which controls the overall
1883
# default control directory format
1884
format_registry.register_lazy(
1885
'Bazaar-NG Knit Repository Format 1',
1886
'bzrlib.repofmt.knitrepo',
1887
'RepositoryFormatKnit1',
1889
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
1891
format_registry.register_lazy(
1892
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
1893
'bzrlib.repofmt.knitrepo',
1894
'RepositoryFormatKnit3',
1898
class InterRepository(InterObject):
1899
"""This class represents operations taking place between two repositories.
1901
Its instances have methods like copy_content and fetch, and contain
1902
references to the source and target repositories these operations can be
1905
Often we will provide convenience methods on 'repository' which carry out
1906
operations with another repository - they will always forward to
1907
InterRepository.get(other).method_name(parameters).
1911
"""The available optimised InterRepository types."""
1913
def copy_content(self, revision_id=None):
1914
raise NotImplementedError(self.copy_content)
1916
def fetch(self, revision_id=None, pb=None):
1917
"""Fetch the content required to construct revision_id.
1919
The content is copied from self.source to self.target.
1921
:param revision_id: if None all content is copied, if NULL_REVISION no
1923
:param pb: optional progress bar to use for progress reports. If not
1924
provided a default one will be created.
1926
Returns the copied revision count and the failed revisions in a tuple:
1929
raise NotImplementedError(self.fetch)
1932
def missing_revision_ids(self, revision_id=None):
1933
"""Return the revision ids that source has that target does not.
1935
These are returned in topological order.
1937
:param revision_id: only return revision ids included by this
1940
# generic, possibly worst case, slow code path.
1941
target_ids = set(self.target.all_revision_ids())
1942
if revision_id is not None:
1943
source_ids = self.source.get_ancestry(revision_id)
1944
assert source_ids[0] is None
1947
source_ids = self.source.all_revision_ids()
1948
result_set = set(source_ids).difference(target_ids)
1949
# this may look like a no-op: its not. It preserves the ordering
1950
# other_ids had while only returning the members from other_ids
1951
# that we've decided we need.
1952
return [rev_id for rev_id in source_ids if rev_id in result_set]
1955
def _same_model(source, target):
1956
"""True if source and target have the same data representation."""
1957
if source.supports_rich_root() != target.supports_rich_root():
1959
if source._serializer != target._serializer:
1964
class InterSameDataRepository(InterRepository):
1965
"""Code for converting between repositories that represent the same data.
1967
Data format and model must match for this to work.
1971
def _get_repo_format_to_test(self):
1972
"""Repository format for testing with.
1974
InterSameData can pull from subtree to subtree and from non-subtree to
1975
non-subtree, so we test this with the richest repository format.
1977
from bzrlib.repofmt import knitrepo
1978
return knitrepo.RepositoryFormatKnit3()
1981
def is_compatible(source, target):
1982
return InterRepository._same_model(source, target)
1985
def copy_content(self, revision_id=None):
1986
"""Make a complete copy of the content in self into destination.
1988
This copies both the repository's revision data, and configuration information
1989
such as the make_working_trees setting.
1991
This is a destructive operation! Do not use it on existing
1994
:param revision_id: Only copy the content needed to construct
1995
revision_id and its parents.
1998
self.target.set_make_working_trees(self.source.make_working_trees())
1999
except NotImplementedError:
2001
# but don't bother fetching if we have the needed data now.
2002
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2003
self.target.has_revision(revision_id)):
2005
self.target.fetch(self.source, revision_id=revision_id)
2008
def fetch(self, revision_id=None, pb=None):
2009
"""See InterRepository.fetch()."""
2010
from bzrlib.fetch import GenericRepoFetcher
2011
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2012
self.source, self.source._format, self.target,
2013
self.target._format)
2014
f = GenericRepoFetcher(to_repository=self.target,
2015
from_repository=self.source,
2016
last_revision=revision_id,
2018
return f.count_copied, f.failed_revisions
2021
class InterWeaveRepo(InterSameDataRepository):
2022
"""Optimised code paths between Weave based repositories.
2024
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2025
implemented lazy inter-object optimisation.
2029
def _get_repo_format_to_test(self):
2030
from bzrlib.repofmt import weaverepo
2031
return weaverepo.RepositoryFormat7()
2034
def is_compatible(source, target):
2035
"""Be compatible with known Weave formats.
2037
We don't test for the stores being of specific types because that
2038
could lead to confusing results, and there is no need to be
2041
from bzrlib.repofmt.weaverepo import (
2047
return (isinstance(source._format, (RepositoryFormat5,
2049
RepositoryFormat7)) and
2050
isinstance(target._format, (RepositoryFormat5,
2052
RepositoryFormat7)))
2053
except AttributeError:
2057
def copy_content(self, revision_id=None):
2058
"""See InterRepository.copy_content()."""
2059
# weave specific optimised path:
2061
self.target.set_make_working_trees(self.source.make_working_trees())
2062
except NotImplementedError:
2064
# FIXME do not peek!
2065
if self.source.control_files._transport.listable():
2066
pb = ui.ui_factory.nested_progress_bar()
2068
self.target.weave_store.copy_all_ids(
2069
self.source.weave_store,
2071
from_transaction=self.source.get_transaction(),
2072
to_transaction=self.target.get_transaction())
2073
pb.update('copying inventory', 0, 1)
2074
self.target.control_weaves.copy_multi(
2075
self.source.control_weaves, ['inventory'],
2076
from_transaction=self.source.get_transaction(),
2077
to_transaction=self.target.get_transaction())
2078
self.target._revision_store.text_store.copy_all_ids(
2079
self.source._revision_store.text_store,
2084
self.target.fetch(self.source, revision_id=revision_id)
2087
def fetch(self, revision_id=None, pb=None):
2088
"""See InterRepository.fetch()."""
2089
from bzrlib.fetch import GenericRepoFetcher
2090
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2091
self.source, self.source._format, self.target, self.target._format)
2092
f = GenericRepoFetcher(to_repository=self.target,
2093
from_repository=self.source,
2094
last_revision=revision_id,
2096
return f.count_copied, f.failed_revisions
2099
def missing_revision_ids(self, revision_id=None):
2100
"""See InterRepository.missing_revision_ids()."""
2101
# we want all revisions to satisfy revision_id in source.
2102
# but we don't want to stat every file here and there.
2103
# we want then, all revisions other needs to satisfy revision_id
2104
# checked, but not those that we have locally.
2105
# so the first thing is to get a subset of the revisions to
2106
# satisfy revision_id in source, and then eliminate those that
2107
# we do already have.
2108
# this is slow on high latency connection to self, but as as this
2109
# disk format scales terribly for push anyway due to rewriting
2110
# inventory.weave, this is considered acceptable.
2112
if revision_id is not None:
2113
source_ids = self.source.get_ancestry(revision_id)
2114
assert source_ids[0] is None
2117
source_ids = self.source._all_possible_ids()
2118
source_ids_set = set(source_ids)
2119
# source_ids is the worst possible case we may need to pull.
2120
# now we want to filter source_ids against what we actually
2121
# have in target, but don't try to check for existence where we know
2122
# we do not have a revision as that would be pointless.
2123
target_ids = set(self.target._all_possible_ids())
2124
possibly_present_revisions = target_ids.intersection(source_ids_set)
2125
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2126
required_revisions = source_ids_set.difference(actually_present_revisions)
2127
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2128
if revision_id is not None:
2129
# we used get_ancestry to determine source_ids then we are assured all
2130
# revisions referenced are present as they are installed in topological order.
2131
# and the tip revision was validated by get_ancestry.
2132
return required_topo_revisions
2134
# if we just grabbed the possibly available ids, then
2135
# we only have an estimate of whats available and need to validate
2136
# that against the revision records.
2137
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2140
class InterKnitRepo(InterSameDataRepository):
2141
"""Optimised code paths between Knit based repositories."""
2144
def _get_repo_format_to_test(self):
2145
from bzrlib.repofmt import knitrepo
2146
return knitrepo.RepositoryFormatKnit1()
2149
def is_compatible(source, target):
2150
"""Be compatible with known Knit formats.
2152
We don't test for the stores being of specific types because that
2153
could lead to confusing results, and there is no need to be
2156
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2158
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2159
isinstance(target._format, RepositoryFormatKnit))
2160
except AttributeError:
2162
return are_knits and InterRepository._same_model(source, target)
2165
def fetch(self, revision_id=None, pb=None):
2166
"""See InterRepository.fetch()."""
2167
from bzrlib.fetch import KnitRepoFetcher
2168
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2169
self.source, self.source._format, self.target, self.target._format)
2170
f = KnitRepoFetcher(to_repository=self.target,
2171
from_repository=self.source,
2172
last_revision=revision_id,
2174
return f.count_copied, f.failed_revisions
2177
def missing_revision_ids(self, revision_id=None):
2178
"""See InterRepository.missing_revision_ids()."""
2179
if revision_id is not None:
2180
source_ids = self.source.get_ancestry(revision_id)
2181
assert source_ids[0] is None
2184
source_ids = self.source.all_revision_ids()
2185
source_ids_set = set(source_ids)
2186
# source_ids is the worst possible case we may need to pull.
2187
# now we want to filter source_ids against what we actually
2188
# have in target, but don't try to check for existence where we know
2189
# we do not have a revision as that would be pointless.
2190
target_ids = set(self.target.all_revision_ids())
2191
possibly_present_revisions = target_ids.intersection(source_ids_set)
2192
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2193
required_revisions = source_ids_set.difference(actually_present_revisions)
2194
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2195
if revision_id is not None:
2196
# we used get_ancestry to determine source_ids then we are assured all
2197
# revisions referenced are present as they are installed in topological order.
2198
# and the tip revision was validated by get_ancestry.
2199
return required_topo_revisions
2201
# if we just grabbed the possibly available ids, then
2202
# we only have an estimate of whats available and need to validate
2203
# that against the revision records.
2204
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2207
class InterModel1and2(InterRepository):
2210
def _get_repo_format_to_test(self):
2214
def is_compatible(source, target):
2215
if not source.supports_rich_root() and target.supports_rich_root():
2221
def fetch(self, revision_id=None, pb=None):
2222
"""See InterRepository.fetch()."""
2223
from bzrlib.fetch import Model1toKnit2Fetcher
2224
f = Model1toKnit2Fetcher(to_repository=self.target,
2225
from_repository=self.source,
2226
last_revision=revision_id,
2228
return f.count_copied, f.failed_revisions
2231
def copy_content(self, revision_id=None):
2232
"""Make a complete copy of the content in self into destination.
2234
This is a destructive operation! Do not use it on existing
2237
:param revision_id: Only copy the content needed to construct
2238
revision_id and its parents.
2241
self.target.set_make_working_trees(self.source.make_working_trees())
2242
except NotImplementedError:
2244
# but don't bother fetching if we have the needed data now.
2245
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2246
self.target.has_revision(revision_id)):
2248
self.target.fetch(self.source, revision_id=revision_id)
2251
class InterKnit1and2(InterKnitRepo):
2254
def _get_repo_format_to_test(self):
2258
def is_compatible(source, target):
2259
"""Be compatible with Knit1 source and Knit3 target"""
2260
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2262
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit1, \
2263
RepositoryFormatKnit3
2264
return (isinstance(source._format, (RepositoryFormatKnit1)) and
2265
isinstance(target._format, (RepositoryFormatKnit3)))
2266
except AttributeError:
2270
def fetch(self, revision_id=None, pb=None):
2271
"""See InterRepository.fetch()."""
2272
from bzrlib.fetch import Knit1to2Fetcher
2273
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2274
self.source, self.source._format, self.target,
2275
self.target._format)
2276
f = Knit1to2Fetcher(to_repository=self.target,
2277
from_repository=self.source,
2278
last_revision=revision_id,
2280
return f.count_copied, f.failed_revisions
2283
class InterRemoteRepository(InterRepository):
2284
"""Code for converting between RemoteRepository objects.
2286
This just gets an non-remote repository from the RemoteRepository, and calls
2287
InterRepository.get again.
2290
def __init__(self, source, target):
2291
if isinstance(source, remote.RemoteRepository):
2292
source._ensure_real()
2293
real_source = source._real_repository
2295
real_source = source
2296
if isinstance(target, remote.RemoteRepository):
2297
target._ensure_real()
2298
real_target = target._real_repository
2300
real_target = target
2301
self.real_inter = InterRepository.get(real_source, real_target)
2304
def is_compatible(source, target):
2305
if isinstance(source, remote.RemoteRepository):
2307
if isinstance(target, remote.RemoteRepository):
2311
def copy_content(self, revision_id=None):
2312
self.real_inter.copy_content(revision_id=revision_id)
2314
def fetch(self, revision_id=None, pb=None):
2315
self.real_inter.fetch(revision_id=revision_id, pb=pb)
2318
def _get_repo_format_to_test(self):
2322
InterRepository.register_optimiser(InterSameDataRepository)
2323
InterRepository.register_optimiser(InterWeaveRepo)
2324
InterRepository.register_optimiser(InterKnitRepo)
2325
InterRepository.register_optimiser(InterModel1and2)
2326
InterRepository.register_optimiser(InterKnit1and2)
2327
InterRepository.register_optimiser(InterRemoteRepository)
2330
class CopyConverter(object):
2331
"""A repository conversion tool which just performs a copy of the content.
2333
This is slow but quite reliable.
2336
def __init__(self, target_format):
2337
"""Create a CopyConverter.
2339
:param target_format: The format the resulting repository should be.
2341
self.target_format = target_format
2343
def convert(self, repo, pb):
2344
"""Perform the conversion of to_convert, giving feedback via pb.
2346
:param to_convert: The disk object to convert.
2347
:param pb: a progress bar to use for progress information.
2352
# this is only useful with metadir layouts - separated repo content.
2353
# trigger an assertion if not such
2354
repo._format.get_format_string()
2355
self.repo_dir = repo.bzrdir
2356
self.step('Moving repository to repository.backup')
2357
self.repo_dir.transport.move('repository', 'repository.backup')
2358
backup_transport = self.repo_dir.transport.clone('repository.backup')
2359
repo._format.check_conversion_target(self.target_format)
2360
self.source_repo = repo._format.open(self.repo_dir,
2362
_override_transport=backup_transport)
2363
self.step('Creating new repository')
2364
converted = self.target_format.initialize(self.repo_dir,
2365
self.source_repo.is_shared())
2366
converted.lock_write()
2368
self.step('Copying content into repository.')
2369
self.source_repo.copy_content_into(converted)
2372
self.step('Deleting old repository content.')
2373
self.repo_dir.transport.delete_tree('repository.backup')
2374
self.pb.note('repository converted')
2376
def step(self, message):
2377
"""Update the pb by a step."""
2379
self.pb.update(message, self.count, self.total)
2391
def _unescaper(match, _map=_unescape_map):
2392
code = match.group(1)
2396
if not code.startswith('#'):
2398
return unichr(int(code[1:])).encode('utf8')
2404
def _unescape_xml(data):
2405
"""Unescape predefined XML entities in a string of data."""
2407
if _unescape_re is None:
2408
_unescape_re = re.compile('\&([^;]*);')
2409
return _unescape_re.sub(_unescaper, data)
added
removed
bzrlib/repository.py
