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
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
# XXX: It looks like this is only hit when _check_root decided
269
# to set a new revision on the root. We seem to be overriding
270
# ie.revision being set at this point to mean either its an
271
# unversioned root, or that it's an unchanged file.
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 in basis_inv:
278
delta = (basis_inv.id2path(ie.file_id), path,
282
delta = (None, path, ie.file_id, ie)
285
# we don't need to commit this, because the caller already
286
# determined that an existing revision of this file is
289
return delta, ie.revision == self._new_revision_id
290
# XXX: Friction: parent_candidates should return a list not a dict
291
# so that we don't have to walk the inventories again.
292
parent_candiate_entries = ie.parent_candidates(parent_invs)
293
head_set = self._repo_graph.heads(parent_candiate_entries.keys())
295
for inv in parent_invs:
296
if ie.file_id in inv:
297
old_rev = inv[ie.file_id].revision
298
if old_rev in head_set:
299
heads.append(inv[ie.file_id].revision)
300
head_set.remove(inv[ie.file_id].revision)
303
# now we check to see if we need to write a new record to the
305
# We write a new entry unless there is one head to the ancestors, and
306
# the kind-derived content is unchanged.
308
# Cheapest check first: no ancestors, or more the one head in the
309
# ancestors, we write a new node.
313
# There is a single head, look it up for comparison
314
parent_entry = parent_candiate_entries[heads[0]]
315
# if the non-content specific data has changed, we'll be writing a
317
if (parent_entry.parent_id != ie.parent_id or
318
parent_entry.name != ie.name):
320
# now we need to do content specific checks:
322
# if the kind changed the content obviously has
323
if kind != parent_entry.kind:
327
if (# if the file length changed we have to store:
328
parent_entry.text_size != content_summary[1] or
329
# if the exec bit has changed we have to store:
330
parent_entry.executable != content_summary[2]):
332
elif parent_entry.text_sha1 == content_summary[3]:
333
# all meta and content is unchanged (using a hash cache
334
# hit to check the sha)
335
ie.revision = parent_entry.revision
336
ie.text_size = parent_entry.text_size
337
ie.text_sha1 = parent_entry.text_sha1
338
ie.executable = parent_entry.executable
339
return self._get_delta(ie, basis_inv, path), False
341
# Either there is only a hash change(no hash cache entry,
342
# or same size content change), or there is no change on
344
# Provide the parent's hash to the store layer, so that the
345
# content is unchanged we will not store a new node.
346
nostore_sha = parent_entry.text_sha1
348
# We want to record a new node regardless of the presence or
349
# absence of a content change in the file.
351
ie.executable = content_summary[2]
352
lines = tree.get_file(ie.file_id, path).readlines()
354
ie.text_sha1, ie.text_size = self._add_text_to_weave(
355
ie.file_id, lines, heads, nostore_sha)
356
except errors.ExistingContent:
357
# Turns out that the file content was unchanged, and we were
358
# only going to store a new node if it was changed. Carry over
360
ie.revision = parent_entry.revision
361
ie.text_size = parent_entry.text_size
362
ie.text_sha1 = parent_entry.text_sha1
363
ie.executable = parent_entry.executable
364
return self._get_delta(ie, basis_inv, path), False
365
elif kind == 'directory':
367
# all data is meta here, nothing specific to directory, so
369
ie.revision = parent_entry.revision
370
return self._get_delta(ie, basis_inv, path), False
372
self._add_text_to_weave(ie.file_id, lines, heads, None)
373
elif kind == 'symlink':
374
current_link_target = content_summary[3]
376
# symlink target is not generic metadata, check if it has
378
if current_link_target != parent_entry.symlink_target:
381
# unchanged, carry over.
382
ie.revision = parent_entry.revision
383
ie.symlink_target = parent_entry.symlink_target
384
return self._get_delta(ie, basis_inv, path), False
385
ie.symlink_target = current_link_target
387
self._add_text_to_weave(ie.file_id, lines, heads, None)
388
elif kind == 'tree-reference':
390
if content_summary[3] != parent_entry.reference_revision:
393
# unchanged, carry over.
394
ie.reference_revision = parent_entry.reference_revision
395
ie.revision = parent_entry.revision
396
return self._get_delta(ie, basis_inv, path), False
397
ie.reference_revision = content_summary[3]
399
self._add_text_to_weave(ie.file_id, lines, heads, None)
401
raise NotImplementedError('unknown kind')
402
ie.revision = self._new_revision_id
403
return self._get_delta(ie, basis_inv, path), True
405
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
406
versionedfile = self.repository.weave_store.get_weave_or_empty(
407
file_id, self.repository.get_transaction())
408
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
409
# than add_lines, and allows committing when a parent is ghosted for
411
# Note: as we read the content directly from the tree, we know its not
412
# been turned into unicode or badly split - but a broken tree
413
# implementation could give us bad output from readlines() so this is
414
# not a guarantee of safety. What would be better is always checking
415
# the content during test suite execution. RBC 20070912
417
return versionedfile.add_lines_with_ghosts(
418
self._new_revision_id, parents, new_lines,
419
nostore_sha=nostore_sha, random_id=self.random_revid,
420
check_content=False)[0:2]
422
versionedfile.clear_cache()
425
class RootCommitBuilder(CommitBuilder):
426
"""This commitbuilder actually records the root id"""
428
# the root entry gets versioned properly by this builder.
429
_versioned_root = True
431
def _check_root(self, ie, parent_invs, tree):
432
"""Helper for record_entry_contents.
434
:param ie: An entry being added.
435
:param parent_invs: The inventories of the parent revisions of the
437
:param tree: The tree that is being committed.
441
######################################################################
444
class Repository(object):
445
"""Repository holding history for one or more branches.
447
The repository holds and retrieves historical information including
448
revisions and file history. It's normally accessed only by the Branch,
449
which views a particular line of development through that history.
451
The Repository builds on top of Stores and a Transport, which respectively
452
describe the disk data format and the way of accessing the (possibly
456
# What class to use for a CommitBuilder. Often its simpler to change this
457
# in a Repository class subclass rather than to override
458
# get_commit_builder.
459
_commit_builder_class = CommitBuilder
460
# The search regex used by xml based repositories to determine what things
461
# where changed in a single commit.
462
_file_ids_altered_regex = lazy_regex.lazy_compile(
463
r'file_id="(?P<file_id>[^"]+)"'
464
r'.* revision="(?P<revision_id>[^"]+)"'
467
def abort_write_group(self):
468
"""Commit the contents accrued within the current write group.
470
:seealso: start_write_group.
472
if self._write_group is not self.get_transaction():
473
# has an unlock or relock occured ?
474
raise errors.BzrError('mismatched lock context and write group.')
475
self._abort_write_group()
476
self._write_group = None
478
def _abort_write_group(self):
479
"""Template method for per-repository write group cleanup.
481
This is called during abort before the write group is considered to be
482
finished and should cleanup any internal state accrued during the write
483
group. There is no requirement that data handed to the repository be
484
*not* made available - this is not a rollback - but neither should any
485
attempt be made to ensure that data added is fully commited. Abort is
486
invoked when an error has occured so futher disk or network operations
487
may not be possible or may error and if possible should not be
492
def add_inventory(self, revision_id, inv, parents):
493
"""Add the inventory inv to the repository as revision_id.
495
:param parents: The revision ids of the parents that revision_id
496
is known to have and are in the repository already.
498
returns the sha1 of the serialized inventory.
500
_mod_revision.check_not_reserved_id(revision_id)
501
assert inv.revision_id is None or inv.revision_id == revision_id, \
502
"Mismatch between inventory revision" \
503
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
504
assert inv.root is not None
505
inv_lines = self._serialise_inventory_to_lines(inv)
506
inv_vf = self.get_inventory_weave()
507
return self._inventory_add_lines(inv_vf, revision_id, parents,
508
inv_lines, check_content=False)
510
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
512
"""Store lines in inv_vf and return the sha1 of the inventory."""
514
for parent in parents:
516
final_parents.append(parent)
517
return inv_vf.add_lines(revision_id, final_parents, lines,
518
check_content=check_content)[0]
521
def add_revision(self, revision_id, rev, inv=None, config=None):
522
"""Add rev to the revision store as revision_id.
524
:param revision_id: the revision id to use.
525
:param rev: The revision object.
526
:param inv: The inventory for the revision. if None, it will be looked
527
up in the inventory storer
528
:param config: If None no digital signature will be created.
529
If supplied its signature_needed method will be used
530
to determine if a signature should be made.
532
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
534
_mod_revision.check_not_reserved_id(revision_id)
535
if config is not None and config.signature_needed():
537
inv = self.get_inventory(revision_id)
538
plaintext = Testament(rev, inv).as_short_text()
539
self.store_revision_signature(
540
gpg.GPGStrategy(config), plaintext, revision_id)
541
if not revision_id in self.get_inventory_weave():
543
raise errors.WeaveRevisionNotPresent(revision_id,
544
self.get_inventory_weave())
546
# yes, this is not suitable for adding with ghosts.
547
self.add_inventory(revision_id, inv, rev.parent_ids)
548
self._revision_store.add_revision(rev, self.get_transaction())
550
def _add_revision_text(self, revision_id, text):
551
revision = self._revision_store._serializer.read_revision_from_string(
553
self._revision_store._add_revision(revision, StringIO(text),
554
self.get_transaction())
556
def all_revision_ids(self):
557
"""Returns a list of all the revision ids in the repository.
559
This is deprecated because code should generally work on the graph
560
reachable from a particular revision, and ignore any other revisions
561
that might be present. There is no direct replacement method.
563
if 'evil' in debug.debug_flags:
564
mutter_callsite(2, "all_revision_ids is linear with history.")
565
return self._all_revision_ids()
567
def _all_revision_ids(self):
568
"""Returns a list of all the revision ids in the repository.
570
These are in as much topological order as the underlying store can
573
raise NotImplementedError(self._all_revision_ids)
575
def break_lock(self):
576
"""Break a lock if one is present from another instance.
578
Uses the ui factory to ask for confirmation if the lock may be from
581
self.control_files.break_lock()
584
def _eliminate_revisions_not_present(self, revision_ids):
585
"""Check every revision id in revision_ids to see if we have it.
587
Returns a set of the present revisions.
590
for id in revision_ids:
591
if self.has_revision(id):
596
def create(a_bzrdir):
597
"""Construct the current default format repository in a_bzrdir."""
598
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
600
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
601
"""instantiate a Repository.
603
:param _format: The format of the repository on disk.
604
:param a_bzrdir: The BzrDir of the repository.
606
In the future we will have a single api for all stores for
607
getting file texts, inventories and revisions, then
608
this construct will accept instances of those things.
610
super(Repository, self).__init__()
611
self._format = _format
612
# the following are part of the public API for Repository:
613
self.bzrdir = a_bzrdir
614
self.control_files = control_files
615
self._revision_store = _revision_store
616
# backwards compatibility
617
self.weave_store = text_store
619
self._reconcile_does_inventory_gc = True
620
# not right yet - should be more semantically clear ?
622
self.control_store = control_store
623
self.control_weaves = control_store
624
# TODO: make sure to construct the right store classes, etc, depending
625
# on whether escaping is required.
626
self._warn_if_deprecated()
627
self._write_group = None
628
self.base = control_files._transport.base
631
return '%s(%r)' % (self.__class__.__name__,
634
def has_same_location(self, other):
635
"""Returns a boolean indicating if this repository is at the same
636
location as another repository.
638
This might return False even when two repository objects are accessing
639
the same physical repository via different URLs.
641
if self.__class__ is not other.__class__:
643
return (self.control_files._transport.base ==
644
other.control_files._transport.base)
646
def is_in_write_group(self):
647
"""Return True if there is an open write group.
649
:seealso: start_write_group.
651
return self._write_group is not None
654
return self.control_files.is_locked()
656
def lock_write(self, token=None):
657
"""Lock this repository for writing.
659
This causes caching within the repository obejct to start accumlating
660
data during reads, and allows a 'write_group' to be obtained. Write
661
groups must be used for actual data insertion.
663
:param token: if this is already locked, then lock_write will fail
664
unless the token matches the existing lock.
665
:returns: a token if this instance supports tokens, otherwise None.
666
:raises TokenLockingNotSupported: when a token is given but this
667
instance doesn't support using token locks.
668
:raises MismatchedToken: if the specified token doesn't match the token
669
of the existing lock.
670
:seealso: start_write_group.
672
A token should be passed in if you know that you have locked the object
673
some other way, and need to synchronise this object's state with that
676
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
678
result = self.control_files.lock_write(token=token)
683
self.control_files.lock_read()
686
def get_physical_lock_status(self):
687
return self.control_files.get_physical_lock_status()
689
def leave_lock_in_place(self):
690
"""Tell this repository not to release the physical lock when this
693
If lock_write doesn't return a token, then this method is not supported.
695
self.control_files.leave_in_place()
697
def dont_leave_lock_in_place(self):
698
"""Tell this repository to release the physical lock when this
699
object is unlocked, even if it didn't originally acquire it.
701
If lock_write doesn't return a token, then this method is not supported.
703
self.control_files.dont_leave_in_place()
706
def gather_stats(self, revid=None, committers=None):
707
"""Gather statistics from a revision id.
709
:param revid: The revision id to gather statistics from, if None, then
710
no revision specific statistics are gathered.
711
:param committers: Optional parameter controlling whether to grab
712
a count of committers from the revision specific statistics.
713
:return: A dictionary of statistics. Currently this contains:
714
committers: The number of committers if requested.
715
firstrev: A tuple with timestamp, timezone for the penultimate left
716
most ancestor of revid, if revid is not the NULL_REVISION.
717
latestrev: A tuple with timestamp, timezone for revid, if revid is
718
not the NULL_REVISION.
719
revisions: The total revision count in the repository.
720
size: An estimate disk size of the repository in bytes.
723
if revid and committers:
724
result['committers'] = 0
725
if revid and revid != _mod_revision.NULL_REVISION:
727
all_committers = set()
728
revisions = self.get_ancestry(revid)
729
# pop the leading None
731
first_revision = None
733
# ignore the revisions in the middle - just grab first and last
734
revisions = revisions[0], revisions[-1]
735
for revision in self.get_revisions(revisions):
736
if not first_revision:
737
first_revision = revision
739
all_committers.add(revision.committer)
740
last_revision = revision
742
result['committers'] = len(all_committers)
743
result['firstrev'] = (first_revision.timestamp,
744
first_revision.timezone)
745
result['latestrev'] = (last_revision.timestamp,
746
last_revision.timezone)
748
# now gather global repository information
749
if self.bzrdir.root_transport.listable():
750
c, t = self._revision_store.total_size(self.get_transaction())
751
result['revisions'] = c
756
def missing_revision_ids(self, other, revision_id=None):
757
"""Return the revision ids that other has that this does not.
759
These are returned in topological order.
761
revision_id: only return revision ids included by revision_id.
763
return InterRepository.get(other, self).missing_revision_ids(revision_id)
767
"""Open the repository rooted at base.
769
For instance, if the repository is at URL/.bzr/repository,
770
Repository.open(URL) -> a Repository instance.
772
control = bzrdir.BzrDir.open(base)
773
return control.open_repository()
775
def copy_content_into(self, destination, revision_id=None):
776
"""Make a complete copy of the content in self into destination.
778
This is a destructive operation! Do not use it on existing
781
return InterRepository.get(self, destination).copy_content(revision_id)
783
def commit_write_group(self):
784
"""Commit the contents accrued within the current write group.
786
:seealso: start_write_group.
788
if self._write_group is not self.get_transaction():
789
# has an unlock or relock occured ?
790
raise errors.BzrError('mismatched lock context %r and '
792
(self.get_transaction(), self._write_group))
793
self._commit_write_group()
794
self._write_group = None
796
def _commit_write_group(self):
797
"""Template method for per-repository write group cleanup.
799
This is called before the write group is considered to be
800
finished and should ensure that all data handed to the repository
801
for writing during the write group is safely committed (to the
802
extent possible considering file system caching etc).
805
def fetch(self, source, revision_id=None, pb=None):
806
"""Fetch the content required to construct revision_id from source.
808
If revision_id is None all content is copied.
810
# fast path same-url fetch operations
811
if self.has_same_location(source):
812
# check that last_revision is in 'from' and then return a
814
if (revision_id is not None and
815
not _mod_revision.is_null(revision_id)):
816
self.get_revision(revision_id)
818
inter = InterRepository.get(source, self)
820
return inter.fetch(revision_id=revision_id, pb=pb)
821
except NotImplementedError:
822
raise errors.IncompatibleRepositories(source, self)
824
def create_bundle(self, target, base, fileobj, format=None):
825
return serializer.write_bundle(self, target, base, fileobj, format)
827
def get_commit_builder(self, branch, parents, config, timestamp=None,
828
timezone=None, committer=None, revprops=None,
830
"""Obtain a CommitBuilder for this repository.
832
:param branch: Branch to commit to.
833
:param parents: Revision ids of the parents of the new revision.
834
:param config: Configuration to use.
835
:param timestamp: Optional timestamp recorded for commit.
836
:param timezone: Optional timezone for timestamp.
837
:param committer: Optional committer to set for commit.
838
:param revprops: Optional dictionary of revision properties.
839
:param revision_id: Optional revision id.
841
result = self._commit_builder_class(self, parents, config,
842
timestamp, timezone, committer, revprops, revision_id)
843
self.start_write_group()
847
if (self.control_files._lock_count == 1 and
848
self.control_files._lock_mode == 'w'):
849
if self._write_group is not None:
850
raise errors.BzrError(
851
'Must end write groups before releasing write locks.')
852
self.control_files.unlock()
855
def clone(self, a_bzrdir, revision_id=None):
856
"""Clone this repository into a_bzrdir using the current format.
858
Currently no check is made that the format of this repository and
859
the bzrdir format are compatible. FIXME RBC 20060201.
861
:return: The newly created destination repository.
863
# TODO: deprecate after 0.16; cloning this with all its settings is
864
# probably not very useful -- mbp 20070423
865
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
866
self.copy_content_into(dest_repo, revision_id)
869
def start_write_group(self):
870
"""Start a write group in the repository.
872
Write groups are used by repositories which do not have a 1:1 mapping
873
between file ids and backend store to manage the insertion of data from
874
both fetch and commit operations.
876
A write lock is required around the start_write_group/commit_write_group
877
for the support of lock-requiring repository formats.
879
One can only insert data into a repository inside a write group.
883
if not self.is_locked() or self.control_files._lock_mode != 'w':
884
raise errors.NotWriteLocked(self)
885
if self._write_group:
886
raise errors.BzrError('already in a write group')
887
self._start_write_group()
888
# so we can detect unlock/relock - the write group is now entered.
889
self._write_group = self.get_transaction()
891
def _start_write_group(self):
892
"""Template method for per-repository write group startup.
894
This is called before the write group is considered to be
899
def sprout(self, to_bzrdir, revision_id=None):
900
"""Create a descendent repository for new development.
902
Unlike clone, this does not copy the settings of the repository.
904
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
905
dest_repo.fetch(self, revision_id=revision_id)
908
def _create_sprouting_repo(self, a_bzrdir, shared):
909
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
910
# use target default format.
911
dest_repo = a_bzrdir.create_repository()
913
# Most control formats need the repository to be specifically
914
# created, but on some old all-in-one formats it's not needed
916
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
917
except errors.UninitializableFormat:
918
dest_repo = a_bzrdir.open_repository()
922
def has_revision(self, revision_id):
923
"""True if this repository has a copy of the revision."""
924
if 'evil' in debug.debug_flags:
925
mutter_callsite(3, "has_revision is a LBYL symptom.")
926
return self._revision_store.has_revision_id(revision_id,
927
self.get_transaction())
930
def get_revision(self, revision_id):
931
"""Return the Revision object for a named revision."""
932
return self.get_revisions([revision_id])[0]
935
def get_revision_reconcile(self, revision_id):
936
"""'reconcile' helper routine that allows access to a revision always.
938
This variant of get_revision does not cross check the weave graph
939
against the revision one as get_revision does: but it should only
940
be used by reconcile, or reconcile-alike commands that are correcting
941
or testing the revision graph.
943
return self._get_revisions([revision_id])[0]
946
def get_revisions(self, revision_ids):
947
"""Get many revisions at once."""
948
return self._get_revisions(revision_ids)
951
def _get_revisions(self, revision_ids):
952
"""Core work logic to get many revisions without sanity checks."""
953
for rev_id in revision_ids:
954
if not rev_id or not isinstance(rev_id, basestring):
955
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
956
revs = self._revision_store.get_revisions(revision_ids,
957
self.get_transaction())
959
assert not isinstance(rev.revision_id, unicode)
960
for parent_id in rev.parent_ids:
961
assert not isinstance(parent_id, unicode)
965
def get_revision_xml(self, revision_id):
966
# TODO: jam 20070210 This shouldn't be necessary since get_revision
967
# would have already do it.
968
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
969
rev = self.get_revision(revision_id)
971
# the current serializer..
972
self._revision_store._serializer.write_revision(rev, rev_tmp)
974
return rev_tmp.getvalue()
977
def get_deltas_for_revisions(self, revisions):
978
"""Produce a generator of revision deltas.
980
Note that the input is a sequence of REVISIONS, not revision_ids.
981
Trees will be held in memory until the generator exits.
982
Each delta is relative to the revision's lefthand predecessor.
984
required_trees = set()
985
for revision in revisions:
986
required_trees.add(revision.revision_id)
987
required_trees.update(revision.parent_ids[:1])
988
trees = dict((t.get_revision_id(), t) for
989
t in self.revision_trees(required_trees))
990
for revision in revisions:
991
if not revision.parent_ids:
992
old_tree = self.revision_tree(None)
994
old_tree = trees[revision.parent_ids[0]]
995
yield trees[revision.revision_id].changes_from(old_tree)
998
def get_revision_delta(self, revision_id):
999
"""Return the delta for one revision.
1001
The delta is relative to the left-hand predecessor of the
1004
r = self.get_revision(revision_id)
1005
return list(self.get_deltas_for_revisions([r]))[0]
1008
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1009
signature = gpg_strategy.sign(plaintext)
1010
self._revision_store.add_revision_signature_text(revision_id,
1012
self.get_transaction())
1014
def fileids_altered_by_revision_ids(self, revision_ids):
1015
"""Find the file ids and versions affected by revisions.
1017
:param revisions: an iterable containing revision ids.
1018
:return: a dictionary mapping altered file-ids to an iterable of
1019
revision_ids. Each altered file-ids has the exact revision_ids that
1020
altered it listed explicitly.
1022
assert self._serializer.support_altered_by_hack, \
1023
("fileids_altered_by_revision_ids only supported for branches "
1024
"which store inventory as unnested xml, not on %r" % self)
1025
selected_revision_ids = set(revision_ids)
1026
w = self.get_inventory_weave()
1029
# this code needs to read every new line in every inventory for the
1030
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1031
# not present in one of those inventories is unnecessary but not
1032
# harmful because we are filtering by the revision id marker in the
1033
# inventory lines : we only select file ids altered in one of those
1034
# revisions. We don't need to see all lines in the inventory because
1035
# only those added in an inventory in rev X can contain a revision=X
1037
unescape_revid_cache = {}
1038
unescape_fileid_cache = {}
1040
# jam 20061218 In a big fetch, this handles hundreds of thousands
1041
# of lines, so it has had a lot of inlining and optimizing done.
1042
# Sorry that it is a little bit messy.
1043
# Move several functions to be local variables, since this is a long
1045
search = self._file_ids_altered_regex.search
1046
unescape = _unescape_xml
1047
setdefault = result.setdefault
1048
pb = ui.ui_factory.nested_progress_bar()
1050
for line in w.iter_lines_added_or_present_in_versions(
1051
selected_revision_ids, pb=pb):
1052
match = search(line)
1055
# One call to match.group() returning multiple items is quite a
1056
# bit faster than 2 calls to match.group() each returning 1
1057
file_id, revision_id = match.group('file_id', 'revision_id')
1059
# Inlining the cache lookups helps a lot when you make 170,000
1060
# lines and 350k ids, versus 8.4 unique ids.
1061
# Using a cache helps in 2 ways:
1062
# 1) Avoids unnecessary decoding calls
1063
# 2) Re-uses cached strings, which helps in future set and
1065
# (2) is enough that removing encoding entirely along with
1066
# the cache (so we are using plain strings) results in no
1067
# performance improvement.
1069
revision_id = unescape_revid_cache[revision_id]
1071
unescaped = unescape(revision_id)
1072
unescape_revid_cache[revision_id] = unescaped
1073
revision_id = unescaped
1075
if revision_id in selected_revision_ids:
1077
file_id = unescape_fileid_cache[file_id]
1079
unescaped = unescape(file_id)
1080
unescape_fileid_cache[file_id] = unescaped
1082
setdefault(file_id, set()).add(revision_id)
1087
def iter_files_bytes(self, desired_files):
1088
"""Iterate through file versions.
1090
Files will not necessarily be returned in the order they occur in
1091
desired_files. No specific order is guaranteed.
1093
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1094
value supplied by the caller as part of desired_files. It should
1095
uniquely identify the file version in the caller's context. (Examples:
1096
an index number or a TreeTransform trans_id.)
1098
bytes_iterator is an iterable of bytestrings for the file. The
1099
kind of iterable and length of the bytestrings are unspecified, but for
1100
this implementation, it is a list of lines produced by
1101
VersionedFile.get_lines().
1103
:param desired_files: a list of (file_id, revision_id, identifier)
1106
transaction = self.get_transaction()
1107
for file_id, revision_id, callable_data in desired_files:
1109
weave = self.weave_store.get_weave(file_id, transaction)
1110
except errors.NoSuchFile:
1111
raise errors.NoSuchIdInRepository(self, file_id)
1112
yield callable_data, weave.get_lines(revision_id)
1114
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1115
"""Get an iterable listing the keys of all the data introduced by a set
1118
The keys will be ordered so that the corresponding items can be safely
1119
fetched and inserted in that order.
1121
:returns: An iterable producing tuples of (knit-kind, file-id,
1122
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1123
'revisions'. file-id is None unless knit-kind is 'file'.
1125
# XXX: it's a bit weird to control the inventory weave caching in this
1126
# generator. Ideally the caching would be done in fetch.py I think. Or
1127
# maybe this generator should explicitly have the contract that it
1128
# should not be iterated until the previously yielded item has been
1130
inv_w = self.get_inventory_weave()
1131
inv_w.enable_cache()
1133
# file ids that changed
1134
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1136
num_file_ids = len(file_ids)
1137
for file_id, altered_versions in file_ids.iteritems():
1138
if _files_pb is not None:
1139
_files_pb.update("fetch texts", count, num_file_ids)
1141
yield ("file", file_id, altered_versions)
1142
# We're done with the files_pb. Note that it finished by the caller,
1143
# just as it was created by the caller.
1147
yield ("inventory", None, revision_ids)
1151
revisions_with_signatures = set()
1152
for rev_id in revision_ids:
1154
self.get_signature_text(rev_id)
1155
except errors.NoSuchRevision:
1159
revisions_with_signatures.add(rev_id)
1160
yield ("signatures", None, revisions_with_signatures)
1163
yield ("revisions", None, revision_ids)
1166
def get_inventory_weave(self):
1167
return self.control_weaves.get_weave('inventory',
1168
self.get_transaction())
1171
def get_inventory(self, revision_id):
1172
"""Get Inventory object by hash."""
1173
return self.deserialise_inventory(
1174
revision_id, self.get_inventory_xml(revision_id))
1176
def deserialise_inventory(self, revision_id, xml):
1177
"""Transform the xml into an inventory object.
1179
:param revision_id: The expected revision id of the inventory.
1180
:param xml: A serialised inventory.
1182
return self._serializer.read_inventory_from_string(xml, revision_id)
1184
def serialise_inventory(self, inv):
1185
return self._serializer.write_inventory_to_string(inv)
1187
def _serialise_inventory_to_lines(self, inv):
1188
return self._serializer.write_inventory_to_lines(inv)
1190
def get_serializer_format(self):
1191
return self._serializer.format_num
1194
def get_inventory_xml(self, revision_id):
1195
"""Get inventory XML as a file object."""
1197
assert isinstance(revision_id, str), type(revision_id)
1198
iw = self.get_inventory_weave()
1199
return iw.get_text(revision_id)
1201
raise errors.HistoryMissing(self, 'inventory', revision_id)
1204
def get_inventory_sha1(self, revision_id):
1205
"""Return the sha1 hash of the inventory entry
1207
return self.get_revision(revision_id).inventory_sha1
1210
def get_revision_graph(self, revision_id=None):
1211
"""Return a dictionary containing the revision graph.
1213
NB: This method should not be used as it accesses the entire graph all
1214
at once, which is much more data than most operations should require.
1216
:param revision_id: The revision_id to get a graph from. If None, then
1217
the entire revision graph is returned. This is a deprecated mode of
1218
operation and will be removed in the future.
1219
:return: a dictionary of revision_id->revision_parents_list.
1221
raise NotImplementedError(self.get_revision_graph)
1224
def get_revision_graph_with_ghosts(self, revision_ids=None):
1225
"""Return a graph of the revisions with ghosts marked as applicable.
1227
:param revision_ids: an iterable of revisions to graph or None for all.
1228
:return: a Graph object with the graph reachable from revision_ids.
1230
if 'evil' in debug.debug_flags:
1232
"get_revision_graph_with_ghosts scales with size of history.")
1233
result = deprecated_graph.Graph()
1234
if not revision_ids:
1235
pending = set(self.all_revision_ids())
1238
pending = set(revision_ids)
1239
# special case NULL_REVISION
1240
if _mod_revision.NULL_REVISION in pending:
1241
pending.remove(_mod_revision.NULL_REVISION)
1242
required = set(pending)
1245
revision_id = pending.pop()
1247
rev = self.get_revision(revision_id)
1248
except errors.NoSuchRevision:
1249
if revision_id in required:
1252
result.add_ghost(revision_id)
1254
for parent_id in rev.parent_ids:
1255
# is this queued or done ?
1256
if (parent_id not in pending and
1257
parent_id not in done):
1259
pending.add(parent_id)
1260
result.add_node(revision_id, rev.parent_ids)
1261
done.add(revision_id)
1264
def _get_history_vf(self):
1265
"""Get a versionedfile whose history graph reflects all revisions.
1267
For weave repositories, this is the inventory weave.
1269
return self.get_inventory_weave()
1271
def iter_reverse_revision_history(self, revision_id):
1272
"""Iterate backwards through revision ids in the lefthand history
1274
:param revision_id: The revision id to start with. All its lefthand
1275
ancestors will be traversed.
1277
if revision_id in (None, _mod_revision.NULL_REVISION):
1279
next_id = revision_id
1280
versionedfile = self._get_history_vf()
1283
parents = versionedfile.get_parents(next_id)
1284
if len(parents) == 0:
1287
next_id = parents[0]
1290
def get_revision_inventory(self, revision_id):
1291
"""Return inventory of a past revision."""
1292
# TODO: Unify this with get_inventory()
1293
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1294
# must be the same as its revision, so this is trivial.
1295
if revision_id is None:
1296
# This does not make sense: if there is no revision,
1297
# then it is the current tree inventory surely ?!
1298
# and thus get_root_id() is something that looks at the last
1299
# commit on the branch, and the get_root_id is an inventory check.
1300
raise NotImplementedError
1301
# return Inventory(self.get_root_id())
1303
return self.get_inventory(revision_id)
1306
def is_shared(self):
1307
"""Return True if this repository is flagged as a shared repository."""
1308
raise NotImplementedError(self.is_shared)
1311
def reconcile(self, other=None, thorough=False):
1312
"""Reconcile this repository."""
1313
from bzrlib.reconcile import RepoReconciler
1314
reconciler = RepoReconciler(self, thorough=thorough)
1315
reconciler.reconcile()
1318
def _refresh_data(self):
1319
"""Helper called from lock_* to ensure coherency with disk.
1321
The default implementation does nothing; it is however possible
1322
for repositories to maintain loaded indices across multiple locks
1323
by checking inside their implementation of this method to see
1324
whether their indices are still valid. This depends of course on
1325
the disk format being validatable in this manner.
1329
def revision_tree(self, revision_id):
1330
"""Return Tree for a revision on this branch.
1332
`revision_id` may be None for the empty tree revision.
1334
# TODO: refactor this to use an existing revision object
1335
# so we don't need to read it in twice.
1336
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1337
return RevisionTree(self, Inventory(root_id=None),
1338
_mod_revision.NULL_REVISION)
1340
inv = self.get_revision_inventory(revision_id)
1341
return RevisionTree(self, inv, revision_id)
1344
def revision_trees(self, revision_ids):
1345
"""Return Tree for a revision on this branch.
1347
`revision_id` may not be None or 'null:'"""
1348
assert None not in revision_ids
1349
assert _mod_revision.NULL_REVISION not in revision_ids
1350
texts = self.get_inventory_weave().get_texts(revision_ids)
1351
for text, revision_id in zip(texts, revision_ids):
1352
inv = self.deserialise_inventory(revision_id, text)
1353
yield RevisionTree(self, inv, revision_id)
1356
def get_ancestry(self, revision_id, topo_sorted=True):
1357
"""Return a list of revision-ids integrated by a revision.
1359
The first element of the list is always None, indicating the origin
1360
revision. This might change when we have history horizons, or
1361
perhaps we should have a new API.
1363
This is topologically sorted.
1365
if _mod_revision.is_null(revision_id):
1367
if not self.has_revision(revision_id):
1368
raise errors.NoSuchRevision(self, revision_id)
1369
w = self.get_inventory_weave()
1370
candidates = w.get_ancestry(revision_id, topo_sorted)
1371
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1374
"""Compress the data within the repository.
1376
This operation only makes sense for some repository types. For other
1377
types it should be a no-op that just returns.
1379
This stub method does not require a lock, but subclasses should use
1380
@needs_write_lock as this is a long running call its reasonable to
1381
implicitly lock for the user.
1385
def print_file(self, file, revision_id):
1386
"""Print `file` to stdout.
1388
FIXME RBC 20060125 as John Meinel points out this is a bad api
1389
- it writes to stdout, it assumes that that is valid etc. Fix
1390
by creating a new more flexible convenience function.
1392
tree = self.revision_tree(revision_id)
1393
# use inventory as it was in that revision
1394
file_id = tree.inventory.path2id(file)
1396
# TODO: jam 20060427 Write a test for this code path
1397
# it had a bug in it, and was raising the wrong
1399
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1400
tree.print_file(file_id)
1402
def get_transaction(self):
1403
return self.control_files.get_transaction()
1405
def revision_parents(self, revision_id):
1406
return self.get_inventory_weave().parent_names(revision_id)
1408
def get_parents(self, revision_ids):
1409
"""See StackedParentsProvider.get_parents"""
1411
for revision_id in revision_ids:
1412
if revision_id == _mod_revision.NULL_REVISION:
1416
parents = self.get_revision(revision_id).parent_ids
1417
except errors.NoSuchRevision:
1420
if len(parents) == 0:
1421
parents = [_mod_revision.NULL_REVISION]
1422
parents_list.append(parents)
1425
def _make_parents_provider(self):
1428
def get_graph(self, other_repository=None):
1429
"""Return the graph walker for this repository format"""
1430
parents_provider = self._make_parents_provider()
1431
if (other_repository is not None and
1432
other_repository.bzrdir.transport.base !=
1433
self.bzrdir.transport.base):
1434
parents_provider = graph._StackedParentsProvider(
1435
[parents_provider, other_repository._make_parents_provider()])
1436
return graph.Graph(parents_provider)
1439
def set_make_working_trees(self, new_value):
1440
"""Set the policy flag for making working trees when creating branches.
1442
This only applies to branches that use this repository.
1444
The default is 'True'.
1445
:param new_value: True to restore the default, False to disable making
1448
raise NotImplementedError(self.set_make_working_trees)
1450
def make_working_trees(self):
1451
"""Returns the policy for making working trees on new branches."""
1452
raise NotImplementedError(self.make_working_trees)
1455
def sign_revision(self, revision_id, gpg_strategy):
1456
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1457
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1460
def has_signature_for_revision_id(self, revision_id):
1461
"""Query for a revision signature for revision_id in the repository."""
1462
return self._revision_store.has_signature(revision_id,
1463
self.get_transaction())
1466
def get_signature_text(self, revision_id):
1467
"""Return the text for a signature."""
1468
return self._revision_store.get_signature_text(revision_id,
1469
self.get_transaction())
1472
def check(self, revision_ids):
1473
"""Check consistency of all history of given revision_ids.
1475
Different repository implementations should override _check().
1477
:param revision_ids: A non-empty list of revision_ids whose ancestry
1478
will be checked. Typically the last revision_id of a branch.
1480
if not revision_ids:
1481
raise ValueError("revision_ids must be non-empty in %s.check"
1483
return self._check(revision_ids)
1485
def _check(self, revision_ids):
1486
result = check.Check(self)
1490
def _warn_if_deprecated(self):
1491
global _deprecation_warning_done
1492
if _deprecation_warning_done:
1494
_deprecation_warning_done = True
1495
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1496
% (self._format, self.bzrdir.transport.base))
1498
def supports_rich_root(self):
1499
return self._format.rich_root_data
1501
def _check_ascii_revisionid(self, revision_id, method):
1502
"""Private helper for ascii-only repositories."""
1503
# weave repositories refuse to store revisionids that are non-ascii.
1504
if revision_id is not None:
1505
# weaves require ascii revision ids.
1506
if isinstance(revision_id, unicode):
1508
revision_id.encode('ascii')
1509
except UnicodeEncodeError:
1510
raise errors.NonAsciiRevisionId(method, self)
1513
revision_id.decode('ascii')
1514
except UnicodeDecodeError:
1515
raise errors.NonAsciiRevisionId(method, self)
1519
# remove these delegates a while after bzr 0.15
1520
def __make_delegated(name, from_module):
1521
def _deprecated_repository_forwarder():
1522
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1523
% (name, from_module),
1526
m = __import__(from_module, globals(), locals(), [name])
1528
return getattr(m, name)
1529
except AttributeError:
1530
raise AttributeError('module %s has no name %s'
1532
globals()[name] = _deprecated_repository_forwarder
1535
'AllInOneRepository',
1536
'WeaveMetaDirRepository',
1537
'PreSplitOutRepositoryFormat',
1538
'RepositoryFormat4',
1539
'RepositoryFormat5',
1540
'RepositoryFormat6',
1541
'RepositoryFormat7',
1543
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1547
'RepositoryFormatKnit',
1548
'RepositoryFormatKnit1',
1550
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1553
def install_revision(repository, rev, revision_tree):
1554
"""Install all revision data into a repository."""
1555
present_parents = []
1557
for p_id in rev.parent_ids:
1558
if repository.has_revision(p_id):
1559
present_parents.append(p_id)
1560
parent_trees[p_id] = repository.revision_tree(p_id)
1562
parent_trees[p_id] = repository.revision_tree(None)
1564
inv = revision_tree.inventory
1565
entries = inv.iter_entries()
1566
# backwards compatibility hack: skip the root id.
1567
if not repository.supports_rich_root():
1568
path, root = entries.next()
1569
if root.revision != rev.revision_id:
1570
raise errors.IncompatibleRevision(repr(repository))
1571
# Add the texts that are not already present
1572
for path, ie in entries:
1573
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1574
repository.get_transaction())
1575
if ie.revision not in w:
1577
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1578
# with InventoryEntry.find_previous_heads(). if it is, then there
1579
# is a latent bug here where the parents may have ancestors of each
1581
for revision, tree in parent_trees.iteritems():
1582
if ie.file_id not in tree:
1584
parent_id = tree.inventory[ie.file_id].revision
1585
if parent_id in text_parents:
1587
text_parents.append(parent_id)
1589
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1590
repository.get_transaction())
1591
lines = revision_tree.get_file(ie.file_id).readlines()
1592
vfile.add_lines(rev.revision_id, text_parents, lines)
1594
# install the inventory
1595
repository.add_inventory(rev.revision_id, inv, present_parents)
1596
except errors.RevisionAlreadyPresent:
1598
repository.add_revision(rev.revision_id, rev, inv)
1601
class MetaDirRepository(Repository):
1602
"""Repositories in the new meta-dir layout."""
1604
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1605
super(MetaDirRepository, self).__init__(_format,
1611
dir_mode = self.control_files._dir_mode
1612
file_mode = self.control_files._file_mode
1615
def is_shared(self):
1616
"""Return True if this repository is flagged as a shared repository."""
1617
return self.control_files._transport.has('shared-storage')
1620
def set_make_working_trees(self, new_value):
1621
"""Set the policy flag for making working trees when creating branches.
1623
This only applies to branches that use this repository.
1625
The default is 'True'.
1626
:param new_value: True to restore the default, False to disable making
1631
self.control_files._transport.delete('no-working-trees')
1632
except errors.NoSuchFile:
1635
self.control_files.put_utf8('no-working-trees', '')
1637
def make_working_trees(self):
1638
"""Returns the policy for making working trees on new branches."""
1639
return not self.control_files._transport.has('no-working-trees')
1642
class RepositoryFormatRegistry(registry.Registry):
1643
"""Registry of RepositoryFormats."""
1645
def get(self, format_string):
1646
r = registry.Registry.get(self, format_string)
1652
format_registry = RepositoryFormatRegistry()
1653
"""Registry of formats, indexed by their identifying format string.
1655
This can contain either format instances themselves, or classes/factories that
1656
can be called to obtain one.
1660
#####################################################################
1661
# Repository Formats
1663
class RepositoryFormat(object):
1664
"""A repository format.
1666
Formats provide three things:
1667
* An initialization routine to construct repository data on disk.
1668
* a format string which is used when the BzrDir supports versioned
1670
* an open routine which returns a Repository instance.
1672
There is one and only one Format subclass for each on-disk format. But
1673
there can be one Repository subclass that is used for several different
1674
formats. The _format attribute on a Repository instance can be used to
1675
determine the disk format.
1677
Formats are placed in an dict by their format string for reference
1678
during opening. These should be subclasses of RepositoryFormat
1681
Once a format is deprecated, just deprecate the initialize and open
1682
methods on the format class. Do not deprecate the object, as the
1683
object will be created every system load.
1685
Common instance attributes:
1686
_matchingbzrdir - the bzrdir format that the repository format was
1687
originally written to work with. This can be used if manually
1688
constructing a bzrdir and repository, or more commonly for test suite
1693
return "<%s>" % self.__class__.__name__
1695
def __eq__(self, other):
1696
# format objects are generally stateless
1697
return isinstance(other, self.__class__)
1699
def __ne__(self, other):
1700
return not self == other
1703
def find_format(klass, a_bzrdir):
1704
"""Return the format for the repository object in a_bzrdir.
1706
This is used by bzr native formats that have a "format" file in
1707
the repository. Other methods may be used by different types of
1711
transport = a_bzrdir.get_repository_transport(None)
1712
format_string = transport.get("format").read()
1713
return format_registry.get(format_string)
1714
except errors.NoSuchFile:
1715
raise errors.NoRepositoryPresent(a_bzrdir)
1717
raise errors.UnknownFormatError(format=format_string)
1720
def register_format(klass, format):
1721
format_registry.register(format.get_format_string(), format)
1724
def unregister_format(klass, format):
1725
format_registry.remove(format.get_format_string())
1728
def get_default_format(klass):
1729
"""Return the current default format."""
1730
from bzrlib import bzrdir
1731
return bzrdir.format_registry.make_bzrdir('default').repository_format
1733
def _get_control_store(self, repo_transport, control_files):
1734
"""Return the control store for this repository."""
1735
raise NotImplementedError(self._get_control_store)
1737
def get_format_string(self):
1738
"""Return the ASCII format string that identifies this format.
1740
Note that in pre format ?? repositories the format string is
1741
not permitted nor written to disk.
1743
raise NotImplementedError(self.get_format_string)
1745
def get_format_description(self):
1746
"""Return the short description for this format."""
1747
raise NotImplementedError(self.get_format_description)
1749
def _get_revision_store(self, repo_transport, control_files):
1750
"""Return the revision store object for this a_bzrdir."""
1751
raise NotImplementedError(self._get_revision_store)
1753
def _get_text_rev_store(self,
1760
"""Common logic for getting a revision store for a repository.
1762
see self._get_revision_store for the subclass-overridable method to
1763
get the store for a repository.
1765
from bzrlib.store.revision.text import TextRevisionStore
1766
dir_mode = control_files._dir_mode
1767
file_mode = control_files._file_mode
1768
text_store = TextStore(transport.clone(name),
1770
compressed=compressed,
1772
file_mode=file_mode)
1773
_revision_store = TextRevisionStore(text_store, serializer)
1774
return _revision_store
1776
# TODO: this shouldn't be in the base class, it's specific to things that
1777
# use weaves or knits -- mbp 20070207
1778
def _get_versioned_file_store(self,
1783
versionedfile_class=None,
1784
versionedfile_kwargs={},
1786
if versionedfile_class is None:
1787
versionedfile_class = self._versionedfile_class
1788
weave_transport = control_files._transport.clone(name)
1789
dir_mode = control_files._dir_mode
1790
file_mode = control_files._file_mode
1791
return VersionedFileStore(weave_transport, prefixed=prefixed,
1793
file_mode=file_mode,
1794
versionedfile_class=versionedfile_class,
1795
versionedfile_kwargs=versionedfile_kwargs,
1798
def initialize(self, a_bzrdir, shared=False):
1799
"""Initialize a repository of this format in a_bzrdir.
1801
:param a_bzrdir: The bzrdir to put the new repository in it.
1802
:param shared: The repository should be initialized as a sharable one.
1803
:returns: The new repository object.
1805
This may raise UninitializableFormat if shared repository are not
1806
compatible the a_bzrdir.
1808
raise NotImplementedError(self.initialize)
1810
def is_supported(self):
1811
"""Is this format supported?
1813
Supported formats must be initializable and openable.
1814
Unsupported formats may not support initialization or committing or
1815
some other features depending on the reason for not being supported.
1819
def check_conversion_target(self, target_format):
1820
raise NotImplementedError(self.check_conversion_target)
1822
def open(self, a_bzrdir, _found=False):
1823
"""Return an instance of this format for the bzrdir a_bzrdir.
1825
_found is a private parameter, do not use it.
1827
raise NotImplementedError(self.open)
1830
class MetaDirRepositoryFormat(RepositoryFormat):
1831
"""Common base class for the new repositories using the metadir layout."""
1833
rich_root_data = False
1834
supports_tree_reference = False
1835
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
1838
super(MetaDirRepositoryFormat, self).__init__()
1840
def _create_control_files(self, a_bzrdir):
1841
"""Create the required files and the initial control_files object."""
1842
# FIXME: RBC 20060125 don't peek under the covers
1843
# NB: no need to escape relative paths that are url safe.
1844
repository_transport = a_bzrdir.get_repository_transport(self)
1845
control_files = lockable_files.LockableFiles(repository_transport,
1846
'lock', lockdir.LockDir)
1847
control_files.create_lock()
1848
return control_files
1850
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
1851
"""Upload the initial blank content."""
1852
control_files = self._create_control_files(a_bzrdir)
1853
control_files.lock_write()
1855
control_files._transport.mkdir_multi(dirs,
1856
mode=control_files._dir_mode)
1857
for file, content in files:
1858
control_files.put(file, content)
1859
for file, content in utf8_files:
1860
control_files.put_utf8(file, content)
1862
control_files.put_utf8('shared-storage', '')
1864
control_files.unlock()
1867
# formats which have no format string are not discoverable
1868
# and not independently creatable, so are not registered. They're
1869
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1870
# needed, it's constructed directly by the BzrDir. Non-native formats where
1871
# the repository is not separately opened are similar.
1873
format_registry.register_lazy(
1874
'Bazaar-NG Repository format 7',
1875
'bzrlib.repofmt.weaverepo',
1879
# KEEP in sync with bzrdir.format_registry default, which controls the overall
1880
# default control directory format
1881
format_registry.register_lazy(
1882
'Bazaar-NG Knit Repository Format 1',
1883
'bzrlib.repofmt.knitrepo',
1884
'RepositoryFormatKnit1',
1886
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
1888
format_registry.register_lazy(
1889
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
1890
'bzrlib.repofmt.knitrepo',
1891
'RepositoryFormatKnit3',
1895
class InterRepository(InterObject):
1896
"""This class represents operations taking place between two repositories.
1898
Its instances have methods like copy_content and fetch, and contain
1899
references to the source and target repositories these operations can be
1902
Often we will provide convenience methods on 'repository' which carry out
1903
operations with another repository - they will always forward to
1904
InterRepository.get(other).method_name(parameters).
1908
"""The available optimised InterRepository types."""
1910
def copy_content(self, revision_id=None):
1911
raise NotImplementedError(self.copy_content)
1913
def fetch(self, revision_id=None, pb=None):
1914
"""Fetch the content required to construct revision_id.
1916
The content is copied from self.source to self.target.
1918
:param revision_id: if None all content is copied, if NULL_REVISION no
1920
:param pb: optional progress bar to use for progress reports. If not
1921
provided a default one will be created.
1923
Returns the copied revision count and the failed revisions in a tuple:
1926
raise NotImplementedError(self.fetch)
1929
def missing_revision_ids(self, revision_id=None):
1930
"""Return the revision ids that source has that target does not.
1932
These are returned in topological order.
1934
:param revision_id: only return revision ids included by this
1937
# generic, possibly worst case, slow code path.
1938
target_ids = set(self.target.all_revision_ids())
1939
if revision_id is not None:
1940
source_ids = self.source.get_ancestry(revision_id)
1941
assert source_ids[0] is None
1944
source_ids = self.source.all_revision_ids()
1945
result_set = set(source_ids).difference(target_ids)
1946
# this may look like a no-op: its not. It preserves the ordering
1947
# other_ids had while only returning the members from other_ids
1948
# that we've decided we need.
1949
return [rev_id for rev_id in source_ids if rev_id in result_set]
1952
def _same_model(source, target):
1953
"""True if source and target have the same data representation."""
1954
if source.supports_rich_root() != target.supports_rich_root():
1956
if source._serializer != target._serializer:
1961
class InterSameDataRepository(InterRepository):
1962
"""Code for converting between repositories that represent the same data.
1964
Data format and model must match for this to work.
1968
def _get_repo_format_to_test(self):
1969
"""Repository format for testing with.
1971
InterSameData can pull from subtree to subtree and from non-subtree to
1972
non-subtree, so we test this with the richest repository format.
1974
from bzrlib.repofmt import knitrepo
1975
return knitrepo.RepositoryFormatKnit3()
1978
def is_compatible(source, target):
1979
return InterRepository._same_model(source, target)
1982
def copy_content(self, revision_id=None):
1983
"""Make a complete copy of the content in self into destination.
1985
This copies both the repository's revision data, and configuration information
1986
such as the make_working_trees setting.
1988
This is a destructive operation! Do not use it on existing
1991
:param revision_id: Only copy the content needed to construct
1992
revision_id and its parents.
1995
self.target.set_make_working_trees(self.source.make_working_trees())
1996
except NotImplementedError:
1998
# but don't bother fetching if we have the needed data now.
1999
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2000
self.target.has_revision(revision_id)):
2002
self.target.fetch(self.source, revision_id=revision_id)
2005
def fetch(self, revision_id=None, pb=None):
2006
"""See InterRepository.fetch()."""
2007
from bzrlib.fetch import GenericRepoFetcher
2008
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2009
self.source, self.source._format, self.target,
2010
self.target._format)
2011
f = GenericRepoFetcher(to_repository=self.target,
2012
from_repository=self.source,
2013
last_revision=revision_id,
2015
return f.count_copied, f.failed_revisions
2018
class InterWeaveRepo(InterSameDataRepository):
2019
"""Optimised code paths between Weave based repositories.
2021
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2022
implemented lazy inter-object optimisation.
2026
def _get_repo_format_to_test(self):
2027
from bzrlib.repofmt import weaverepo
2028
return weaverepo.RepositoryFormat7()
2031
def is_compatible(source, target):
2032
"""Be compatible with known Weave formats.
2034
We don't test for the stores being of specific types because that
2035
could lead to confusing results, and there is no need to be
2038
from bzrlib.repofmt.weaverepo import (
2044
return (isinstance(source._format, (RepositoryFormat5,
2046
RepositoryFormat7)) and
2047
isinstance(target._format, (RepositoryFormat5,
2049
RepositoryFormat7)))
2050
except AttributeError:
2054
def copy_content(self, revision_id=None):
2055
"""See InterRepository.copy_content()."""
2056
# weave specific optimised path:
2058
self.target.set_make_working_trees(self.source.make_working_trees())
2059
except NotImplementedError:
2061
# FIXME do not peek!
2062
if self.source.control_files._transport.listable():
2063
pb = ui.ui_factory.nested_progress_bar()
2065
self.target.weave_store.copy_all_ids(
2066
self.source.weave_store,
2068
from_transaction=self.source.get_transaction(),
2069
to_transaction=self.target.get_transaction())
2070
pb.update('copying inventory', 0, 1)
2071
self.target.control_weaves.copy_multi(
2072
self.source.control_weaves, ['inventory'],
2073
from_transaction=self.source.get_transaction(),
2074
to_transaction=self.target.get_transaction())
2075
self.target._revision_store.text_store.copy_all_ids(
2076
self.source._revision_store.text_store,
2081
self.target.fetch(self.source, revision_id=revision_id)
2084
def fetch(self, revision_id=None, pb=None):
2085
"""See InterRepository.fetch()."""
2086
from bzrlib.fetch import GenericRepoFetcher
2087
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2088
self.source, self.source._format, self.target, self.target._format)
2089
f = GenericRepoFetcher(to_repository=self.target,
2090
from_repository=self.source,
2091
last_revision=revision_id,
2093
return f.count_copied, f.failed_revisions
2096
def missing_revision_ids(self, revision_id=None):
2097
"""See InterRepository.missing_revision_ids()."""
2098
# we want all revisions to satisfy revision_id in source.
2099
# but we don't want to stat every file here and there.
2100
# we want then, all revisions other needs to satisfy revision_id
2101
# checked, but not those that we have locally.
2102
# so the first thing is to get a subset of the revisions to
2103
# satisfy revision_id in source, and then eliminate those that
2104
# we do already have.
2105
# this is slow on high latency connection to self, but as as this
2106
# disk format scales terribly for push anyway due to rewriting
2107
# inventory.weave, this is considered acceptable.
2109
if revision_id is not None:
2110
source_ids = self.source.get_ancestry(revision_id)
2111
assert source_ids[0] is None
2114
source_ids = self.source._all_possible_ids()
2115
source_ids_set = set(source_ids)
2116
# source_ids is the worst possible case we may need to pull.
2117
# now we want to filter source_ids against what we actually
2118
# have in target, but don't try to check for existence where we know
2119
# we do not have a revision as that would be pointless.
2120
target_ids = set(self.target._all_possible_ids())
2121
possibly_present_revisions = target_ids.intersection(source_ids_set)
2122
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2123
required_revisions = source_ids_set.difference(actually_present_revisions)
2124
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2125
if revision_id is not None:
2126
# we used get_ancestry to determine source_ids then we are assured all
2127
# revisions referenced are present as they are installed in topological order.
2128
# and the tip revision was validated by get_ancestry.
2129
return required_topo_revisions
2131
# if we just grabbed the possibly available ids, then
2132
# we only have an estimate of whats available and need to validate
2133
# that against the revision records.
2134
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2137
class InterKnitRepo(InterSameDataRepository):
2138
"""Optimised code paths between Knit based repositories."""
2141
def _get_repo_format_to_test(self):
2142
from bzrlib.repofmt import knitrepo
2143
return knitrepo.RepositoryFormatKnit1()
2146
def is_compatible(source, target):
2147
"""Be compatible with known Knit formats.
2149
We don't test for the stores being of specific types because that
2150
could lead to confusing results, and there is no need to be
2153
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2155
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2156
isinstance(target._format, RepositoryFormatKnit))
2157
except AttributeError:
2159
return are_knits and InterRepository._same_model(source, target)
2162
def fetch(self, revision_id=None, pb=None):
2163
"""See InterRepository.fetch()."""
2164
from bzrlib.fetch import KnitRepoFetcher
2165
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2166
self.source, self.source._format, self.target, self.target._format)
2167
f = KnitRepoFetcher(to_repository=self.target,
2168
from_repository=self.source,
2169
last_revision=revision_id,
2171
return f.count_copied, f.failed_revisions
2174
def missing_revision_ids(self, revision_id=None):
2175
"""See InterRepository.missing_revision_ids()."""
2176
if revision_id is not None:
2177
source_ids = self.source.get_ancestry(revision_id)
2178
assert source_ids[0] is None
2181
source_ids = self.source.all_revision_ids()
2182
source_ids_set = set(source_ids)
2183
# source_ids is the worst possible case we may need to pull.
2184
# now we want to filter source_ids against what we actually
2185
# have in target, but don't try to check for existence where we know
2186
# we do not have a revision as that would be pointless.
2187
target_ids = set(self.target.all_revision_ids())
2188
possibly_present_revisions = target_ids.intersection(source_ids_set)
2189
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2190
required_revisions = source_ids_set.difference(actually_present_revisions)
2191
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2192
if revision_id is not None:
2193
# we used get_ancestry to determine source_ids then we are assured all
2194
# revisions referenced are present as they are installed in topological order.
2195
# and the tip revision was validated by get_ancestry.
2196
return required_topo_revisions
2198
# if we just grabbed the possibly available ids, then
2199
# we only have an estimate of whats available and need to validate
2200
# that against the revision records.
2201
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2204
class InterModel1and2(InterRepository):
2207
def _get_repo_format_to_test(self):
2211
def is_compatible(source, target):
2212
if not source.supports_rich_root() and target.supports_rich_root():
2218
def fetch(self, revision_id=None, pb=None):
2219
"""See InterRepository.fetch()."""
2220
from bzrlib.fetch import Model1toKnit2Fetcher
2221
f = Model1toKnit2Fetcher(to_repository=self.target,
2222
from_repository=self.source,
2223
last_revision=revision_id,
2225
return f.count_copied, f.failed_revisions
2228
def copy_content(self, revision_id=None):
2229
"""Make a complete copy of the content in self into destination.
2231
This is a destructive operation! Do not use it on existing
2234
:param revision_id: Only copy the content needed to construct
2235
revision_id and its parents.
2238
self.target.set_make_working_trees(self.source.make_working_trees())
2239
except NotImplementedError:
2241
# but don't bother fetching if we have the needed data now.
2242
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2243
self.target.has_revision(revision_id)):
2245
self.target.fetch(self.source, revision_id=revision_id)
2248
class InterKnit1and2(InterKnitRepo):
2251
def _get_repo_format_to_test(self):
2255
def is_compatible(source, target):
2256
"""Be compatible with Knit1 source and Knit3 target"""
2257
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2259
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit1, \
2260
RepositoryFormatKnit3
2261
return (isinstance(source._format, (RepositoryFormatKnit1)) and
2262
isinstance(target._format, (RepositoryFormatKnit3)))
2263
except AttributeError:
2267
def fetch(self, revision_id=None, pb=None):
2268
"""See InterRepository.fetch()."""
2269
from bzrlib.fetch import Knit1to2Fetcher
2270
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2271
self.source, self.source._format, self.target,
2272
self.target._format)
2273
f = Knit1to2Fetcher(to_repository=self.target,
2274
from_repository=self.source,
2275
last_revision=revision_id,
2277
return f.count_copied, f.failed_revisions
2280
class InterRemoteRepository(InterRepository):
2281
"""Code for converting between RemoteRepository objects.
2283
This just gets an non-remote repository from the RemoteRepository, and calls
2284
InterRepository.get again.
2287
def __init__(self, source, target):
2288
if isinstance(source, remote.RemoteRepository):
2289
source._ensure_real()
2290
real_source = source._real_repository
2292
real_source = source
2293
if isinstance(target, remote.RemoteRepository):
2294
target._ensure_real()
2295
real_target = target._real_repository
2297
real_target = target
2298
self.real_inter = InterRepository.get(real_source, real_target)
2301
def is_compatible(source, target):
2302
if isinstance(source, remote.RemoteRepository):
2304
if isinstance(target, remote.RemoteRepository):
2308
def copy_content(self, revision_id=None):
2309
self.real_inter.copy_content(revision_id=revision_id)
2311
def fetch(self, revision_id=None, pb=None):
2312
self.real_inter.fetch(revision_id=revision_id, pb=pb)
2315
def _get_repo_format_to_test(self):
2319
InterRepository.register_optimiser(InterSameDataRepository)
2320
InterRepository.register_optimiser(InterWeaveRepo)
2321
InterRepository.register_optimiser(InterKnitRepo)
2322
InterRepository.register_optimiser(InterModel1and2)
2323
InterRepository.register_optimiser(InterKnit1and2)
2324
InterRepository.register_optimiser(InterRemoteRepository)
2327
class CopyConverter(object):
2328
"""A repository conversion tool which just performs a copy of the content.
2330
This is slow but quite reliable.
2333
def __init__(self, target_format):
2334
"""Create a CopyConverter.
2336
:param target_format: The format the resulting repository should be.
2338
self.target_format = target_format
2340
def convert(self, repo, pb):
2341
"""Perform the conversion of to_convert, giving feedback via pb.
2343
:param to_convert: The disk object to convert.
2344
:param pb: a progress bar to use for progress information.
2349
# this is only useful with metadir layouts - separated repo content.
2350
# trigger an assertion if not such
2351
repo._format.get_format_string()
2352
self.repo_dir = repo.bzrdir
2353
self.step('Moving repository to repository.backup')
2354
self.repo_dir.transport.move('repository', 'repository.backup')
2355
backup_transport = self.repo_dir.transport.clone('repository.backup')
2356
repo._format.check_conversion_target(self.target_format)
2357
self.source_repo = repo._format.open(self.repo_dir,
2359
_override_transport=backup_transport)
2360
self.step('Creating new repository')
2361
converted = self.target_format.initialize(self.repo_dir,
2362
self.source_repo.is_shared())
2363
converted.lock_write()
2365
self.step('Copying content into repository.')
2366
self.source_repo.copy_content_into(converted)
2369
self.step('Deleting old repository content.')
2370
self.repo_dir.transport.delete_tree('repository.backup')
2371
self.pb.note('repository converted')
2373
def step(self, message):
2374
"""Update the pb by a step."""
2376
self.pb.update(message, self.count, self.total)
2388
def _unescaper(match, _map=_unescape_map):
2389
code = match.group(1)
2393
if not code.startswith('#'):
2395
return unichr(int(code[1:])).encode('utf8')
2401
def _unescape_xml(data):
2402
"""Unescape predefined XML entities in a string of data."""
2404
if _unescape_re is None:
2405
_unescape_re = re.compile('\&([^;]*);')
2406
return _unescape_re.sub(_unescaper, data)
added
removed
bzrlib/repository.py
