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 is_write_locked(self):
660
"""Return True if this object is write locked."""
661
return self.is_locked() and self.control_files._lock_mode == 'w'
663
def lock_write(self, token=None):
664
"""Lock this repository for writing.
666
This causes caching within the repository obejct to start accumlating
667
data during reads, and allows a 'write_group' to be obtained. Write
668
groups must be used for actual data insertion.
670
:param token: if this is already locked, then lock_write will fail
671
unless the token matches the existing lock.
672
:returns: a token if this instance supports tokens, otherwise None.
673
:raises TokenLockingNotSupported: when a token is given but this
674
instance doesn't support using token locks.
675
:raises MismatchedToken: if the specified token doesn't match the token
676
of the existing lock.
677
:seealso: start_write_group.
679
A token should be passed in if you know that you have locked the object
680
some other way, and need to synchronise this object's state with that
683
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
685
result = self.control_files.lock_write(token=token)
690
self.control_files.lock_read()
693
def get_physical_lock_status(self):
694
return self.control_files.get_physical_lock_status()
696
def leave_lock_in_place(self):
697
"""Tell this repository not to release the physical lock when this
700
If lock_write doesn't return a token, then this method is not supported.
702
self.control_files.leave_in_place()
704
def dont_leave_lock_in_place(self):
705
"""Tell this repository to release the physical lock when this
706
object is unlocked, even if it didn't originally acquire it.
708
If lock_write doesn't return a token, then this method is not supported.
710
self.control_files.dont_leave_in_place()
713
def gather_stats(self, revid=None, committers=None):
714
"""Gather statistics from a revision id.
716
:param revid: The revision id to gather statistics from, if None, then
717
no revision specific statistics are gathered.
718
:param committers: Optional parameter controlling whether to grab
719
a count of committers from the revision specific statistics.
720
:return: A dictionary of statistics. Currently this contains:
721
committers: The number of committers if requested.
722
firstrev: A tuple with timestamp, timezone for the penultimate left
723
most ancestor of revid, if revid is not the NULL_REVISION.
724
latestrev: A tuple with timestamp, timezone for revid, if revid is
725
not the NULL_REVISION.
726
revisions: The total revision count in the repository.
727
size: An estimate disk size of the repository in bytes.
730
if revid and committers:
731
result['committers'] = 0
732
if revid and revid != _mod_revision.NULL_REVISION:
734
all_committers = set()
735
revisions = self.get_ancestry(revid)
736
# pop the leading None
738
first_revision = None
740
# ignore the revisions in the middle - just grab first and last
741
revisions = revisions[0], revisions[-1]
742
for revision in self.get_revisions(revisions):
743
if not first_revision:
744
first_revision = revision
746
all_committers.add(revision.committer)
747
last_revision = revision
749
result['committers'] = len(all_committers)
750
result['firstrev'] = (first_revision.timestamp,
751
first_revision.timezone)
752
result['latestrev'] = (last_revision.timestamp,
753
last_revision.timezone)
755
# now gather global repository information
756
if self.bzrdir.root_transport.listable():
757
c, t = self._revision_store.total_size(self.get_transaction())
758
result['revisions'] = c
763
def missing_revision_ids(self, other, revision_id=None):
764
"""Return the revision ids that other has that this does not.
766
These are returned in topological order.
768
revision_id: only return revision ids included by revision_id.
770
return InterRepository.get(other, self).missing_revision_ids(revision_id)
774
"""Open the repository rooted at base.
776
For instance, if the repository is at URL/.bzr/repository,
777
Repository.open(URL) -> a Repository instance.
779
control = bzrdir.BzrDir.open(base)
780
return control.open_repository()
782
def copy_content_into(self, destination, revision_id=None):
783
"""Make a complete copy of the content in self into destination.
785
This is a destructive operation! Do not use it on existing
788
return InterRepository.get(self, destination).copy_content(revision_id)
790
def commit_write_group(self):
791
"""Commit the contents accrued within the current write group.
793
:seealso: start_write_group.
795
if self._write_group is not self.get_transaction():
796
# has an unlock or relock occured ?
797
raise errors.BzrError('mismatched lock context %r and '
799
(self.get_transaction(), self._write_group))
800
self._commit_write_group()
801
self._write_group = None
803
def _commit_write_group(self):
804
"""Template method for per-repository write group cleanup.
806
This is called before the write group is considered to be
807
finished and should ensure that all data handed to the repository
808
for writing during the write group is safely committed (to the
809
extent possible considering file system caching etc).
812
def fetch(self, source, revision_id=None, pb=None):
813
"""Fetch the content required to construct revision_id from source.
815
If revision_id is None all content is copied.
817
# fast path same-url fetch operations
818
if self.has_same_location(source):
819
# check that last_revision is in 'from' and then return a
821
if (revision_id is not None and
822
not _mod_revision.is_null(revision_id)):
823
self.get_revision(revision_id)
825
inter = InterRepository.get(source, self)
827
return inter.fetch(revision_id=revision_id, pb=pb)
828
except NotImplementedError:
829
raise errors.IncompatibleRepositories(source, self)
831
def create_bundle(self, target, base, fileobj, format=None):
832
return serializer.write_bundle(self, target, base, fileobj, format)
834
def get_commit_builder(self, branch, parents, config, timestamp=None,
835
timezone=None, committer=None, revprops=None,
837
"""Obtain a CommitBuilder for this repository.
839
:param branch: Branch to commit to.
840
:param parents: Revision ids of the parents of the new revision.
841
:param config: Configuration to use.
842
:param timestamp: Optional timestamp recorded for commit.
843
:param timezone: Optional timezone for timestamp.
844
:param committer: Optional committer to set for commit.
845
:param revprops: Optional dictionary of revision properties.
846
:param revision_id: Optional revision id.
848
result = self._commit_builder_class(self, parents, config,
849
timestamp, timezone, committer, revprops, revision_id)
850
self.start_write_group()
854
if (self.control_files._lock_count == 1 and
855
self.control_files._lock_mode == 'w'):
856
if self._write_group is not None:
857
raise errors.BzrError(
858
'Must end write groups before releasing write locks.')
859
self.control_files.unlock()
862
def clone(self, a_bzrdir, revision_id=None):
863
"""Clone this repository into a_bzrdir using the current format.
865
Currently no check is made that the format of this repository and
866
the bzrdir format are compatible. FIXME RBC 20060201.
868
:return: The newly created destination repository.
870
# TODO: deprecate after 0.16; cloning this with all its settings is
871
# probably not very useful -- mbp 20070423
872
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
873
self.copy_content_into(dest_repo, revision_id)
876
def start_write_group(self):
877
"""Start a write group in the repository.
879
Write groups are used by repositories which do not have a 1:1 mapping
880
between file ids and backend store to manage the insertion of data from
881
both fetch and commit operations.
883
A write lock is required around the start_write_group/commit_write_group
884
for the support of lock-requiring repository formats.
886
One can only insert data into a repository inside a write group.
890
if not self.is_write_locked():
891
raise errors.NotWriteLocked(self)
892
if self._write_group:
893
raise errors.BzrError('already in a write group')
894
self._start_write_group()
895
# so we can detect unlock/relock - the write group is now entered.
896
self._write_group = self.get_transaction()
898
def _start_write_group(self):
899
"""Template method for per-repository write group startup.
901
This is called before the write group is considered to be
906
def sprout(self, to_bzrdir, revision_id=None):
907
"""Create a descendent repository for new development.
909
Unlike clone, this does not copy the settings of the repository.
911
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
912
dest_repo.fetch(self, revision_id=revision_id)
915
def _create_sprouting_repo(self, a_bzrdir, shared):
916
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
917
# use target default format.
918
dest_repo = a_bzrdir.create_repository()
920
# Most control formats need the repository to be specifically
921
# created, but on some old all-in-one formats it's not needed
923
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
924
except errors.UninitializableFormat:
925
dest_repo = a_bzrdir.open_repository()
929
def has_revision(self, revision_id):
930
"""True if this repository has a copy of the revision."""
931
if 'evil' in debug.debug_flags:
932
mutter_callsite(3, "has_revision is a LBYL symptom.")
933
return self._revision_store.has_revision_id(revision_id,
934
self.get_transaction())
937
def get_revision(self, revision_id):
938
"""Return the Revision object for a named revision."""
939
return self.get_revisions([revision_id])[0]
942
def get_revision_reconcile(self, revision_id):
943
"""'reconcile' helper routine that allows access to a revision always.
945
This variant of get_revision does not cross check the weave graph
946
against the revision one as get_revision does: but it should only
947
be used by reconcile, or reconcile-alike commands that are correcting
948
or testing the revision graph.
950
return self._get_revisions([revision_id])[0]
953
def get_revisions(self, revision_ids):
954
"""Get many revisions at once."""
955
return self._get_revisions(revision_ids)
958
def _get_revisions(self, revision_ids):
959
"""Core work logic to get many revisions without sanity checks."""
960
for rev_id in revision_ids:
961
if not rev_id or not isinstance(rev_id, basestring):
962
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
963
revs = self._revision_store.get_revisions(revision_ids,
964
self.get_transaction())
966
assert not isinstance(rev.revision_id, unicode)
967
for parent_id in rev.parent_ids:
968
assert not isinstance(parent_id, unicode)
972
def get_revision_xml(self, revision_id):
973
# TODO: jam 20070210 This shouldn't be necessary since get_revision
974
# would have already do it.
975
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
976
rev = self.get_revision(revision_id)
978
# the current serializer..
979
self._revision_store._serializer.write_revision(rev, rev_tmp)
981
return rev_tmp.getvalue()
984
def get_deltas_for_revisions(self, revisions):
985
"""Produce a generator of revision deltas.
987
Note that the input is a sequence of REVISIONS, not revision_ids.
988
Trees will be held in memory until the generator exits.
989
Each delta is relative to the revision's lefthand predecessor.
991
required_trees = set()
992
for revision in revisions:
993
required_trees.add(revision.revision_id)
994
required_trees.update(revision.parent_ids[:1])
995
trees = dict((t.get_revision_id(), t) for
996
t in self.revision_trees(required_trees))
997
for revision in revisions:
998
if not revision.parent_ids:
999
old_tree = self.revision_tree(None)
1001
old_tree = trees[revision.parent_ids[0]]
1002
yield trees[revision.revision_id].changes_from(old_tree)
1005
def get_revision_delta(self, revision_id):
1006
"""Return the delta for one revision.
1008
The delta is relative to the left-hand predecessor of the
1011
r = self.get_revision(revision_id)
1012
return list(self.get_deltas_for_revisions([r]))[0]
1015
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1016
signature = gpg_strategy.sign(plaintext)
1017
self._revision_store.add_revision_signature_text(revision_id,
1019
self.get_transaction())
1021
def fileids_altered_by_revision_ids(self, revision_ids):
1022
"""Find the file ids and versions affected by revisions.
1024
:param revisions: an iterable containing revision ids.
1025
:return: a dictionary mapping altered file-ids to an iterable of
1026
revision_ids. Each altered file-ids has the exact revision_ids that
1027
altered it listed explicitly.
1029
assert self._serializer.support_altered_by_hack, \
1030
("fileids_altered_by_revision_ids only supported for branches "
1031
"which store inventory as unnested xml, not on %r" % self)
1032
selected_revision_ids = set(revision_ids)
1033
w = self.get_inventory_weave()
1036
# this code needs to read every new line in every inventory for the
1037
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1038
# not present in one of those inventories is unnecessary but not
1039
# harmful because we are filtering by the revision id marker in the
1040
# inventory lines : we only select file ids altered in one of those
1041
# revisions. We don't need to see all lines in the inventory because
1042
# only those added in an inventory in rev X can contain a revision=X
1044
unescape_revid_cache = {}
1045
unescape_fileid_cache = {}
1047
# jam 20061218 In a big fetch, this handles hundreds of thousands
1048
# of lines, so it has had a lot of inlining and optimizing done.
1049
# Sorry that it is a little bit messy.
1050
# Move several functions to be local variables, since this is a long
1052
search = self._file_ids_altered_regex.search
1053
unescape = _unescape_xml
1054
setdefault = result.setdefault
1055
pb = ui.ui_factory.nested_progress_bar()
1057
for line in w.iter_lines_added_or_present_in_versions(
1058
selected_revision_ids, pb=pb):
1059
match = search(line)
1062
# One call to match.group() returning multiple items is quite a
1063
# bit faster than 2 calls to match.group() each returning 1
1064
file_id, revision_id = match.group('file_id', 'revision_id')
1066
# Inlining the cache lookups helps a lot when you make 170,000
1067
# lines and 350k ids, versus 8.4 unique ids.
1068
# Using a cache helps in 2 ways:
1069
# 1) Avoids unnecessary decoding calls
1070
# 2) Re-uses cached strings, which helps in future set and
1072
# (2) is enough that removing encoding entirely along with
1073
# the cache (so we are using plain strings) results in no
1074
# performance improvement.
1076
revision_id = unescape_revid_cache[revision_id]
1078
unescaped = unescape(revision_id)
1079
unescape_revid_cache[revision_id] = unescaped
1080
revision_id = unescaped
1082
if revision_id in selected_revision_ids:
1084
file_id = unescape_fileid_cache[file_id]
1086
unescaped = unescape(file_id)
1087
unescape_fileid_cache[file_id] = unescaped
1089
setdefault(file_id, set()).add(revision_id)
1094
def iter_files_bytes(self, desired_files):
1095
"""Iterate through file versions.
1097
Files will not necessarily be returned in the order they occur in
1098
desired_files. No specific order is guaranteed.
1100
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1101
value supplied by the caller as part of desired_files. It should
1102
uniquely identify the file version in the caller's context. (Examples:
1103
an index number or a TreeTransform trans_id.)
1105
bytes_iterator is an iterable of bytestrings for the file. The
1106
kind of iterable and length of the bytestrings are unspecified, but for
1107
this implementation, it is a list of lines produced by
1108
VersionedFile.get_lines().
1110
:param desired_files: a list of (file_id, revision_id, identifier)
1113
transaction = self.get_transaction()
1114
for file_id, revision_id, callable_data in desired_files:
1116
weave = self.weave_store.get_weave(file_id, transaction)
1117
except errors.NoSuchFile:
1118
raise errors.NoSuchIdInRepository(self, file_id)
1119
yield callable_data, weave.get_lines(revision_id)
1121
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1122
"""Get an iterable listing the keys of all the data introduced by a set
1125
The keys will be ordered so that the corresponding items can be safely
1126
fetched and inserted in that order.
1128
:returns: An iterable producing tuples of (knit-kind, file-id,
1129
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1130
'revisions'. file-id is None unless knit-kind is 'file'.
1132
# XXX: it's a bit weird to control the inventory weave caching in this
1133
# generator. Ideally the caching would be done in fetch.py I think. Or
1134
# maybe this generator should explicitly have the contract that it
1135
# should not be iterated until the previously yielded item has been
1137
inv_w = self.get_inventory_weave()
1138
inv_w.enable_cache()
1140
# file ids that changed
1141
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1143
num_file_ids = len(file_ids)
1144
for file_id, altered_versions in file_ids.iteritems():
1145
if _files_pb is not None:
1146
_files_pb.update("fetch texts", count, num_file_ids)
1148
yield ("file", file_id, altered_versions)
1149
# We're done with the files_pb. Note that it finished by the caller,
1150
# just as it was created by the caller.
1154
yield ("inventory", None, revision_ids)
1158
revisions_with_signatures = set()
1159
for rev_id in revision_ids:
1161
self.get_signature_text(rev_id)
1162
except errors.NoSuchRevision:
1166
revisions_with_signatures.add(rev_id)
1167
yield ("signatures", None, revisions_with_signatures)
1170
yield ("revisions", None, revision_ids)
1173
def get_inventory_weave(self):
1174
return self.control_weaves.get_weave('inventory',
1175
self.get_transaction())
1178
def get_inventory(self, revision_id):
1179
"""Get Inventory object by hash."""
1180
return self.deserialise_inventory(
1181
revision_id, self.get_inventory_xml(revision_id))
1183
def deserialise_inventory(self, revision_id, xml):
1184
"""Transform the xml into an inventory object.
1186
:param revision_id: The expected revision id of the inventory.
1187
:param xml: A serialised inventory.
1189
return self._serializer.read_inventory_from_string(xml, revision_id)
1191
def serialise_inventory(self, inv):
1192
return self._serializer.write_inventory_to_string(inv)
1194
def _serialise_inventory_to_lines(self, inv):
1195
return self._serializer.write_inventory_to_lines(inv)
1197
def get_serializer_format(self):
1198
return self._serializer.format_num
1201
def get_inventory_xml(self, revision_id):
1202
"""Get inventory XML as a file object."""
1204
assert isinstance(revision_id, str), type(revision_id)
1205
iw = self.get_inventory_weave()
1206
return iw.get_text(revision_id)
1208
raise errors.HistoryMissing(self, 'inventory', revision_id)
1211
def get_inventory_sha1(self, revision_id):
1212
"""Return the sha1 hash of the inventory entry
1214
return self.get_revision(revision_id).inventory_sha1
1217
def get_revision_graph(self, revision_id=None):
1218
"""Return a dictionary containing the revision graph.
1220
NB: This method should not be used as it accesses the entire graph all
1221
at once, which is much more data than most operations should require.
1223
:param revision_id: The revision_id to get a graph from. If None, then
1224
the entire revision graph is returned. This is a deprecated mode of
1225
operation and will be removed in the future.
1226
:return: a dictionary of revision_id->revision_parents_list.
1228
raise NotImplementedError(self.get_revision_graph)
1231
def get_revision_graph_with_ghosts(self, revision_ids=None):
1232
"""Return a graph of the revisions with ghosts marked as applicable.
1234
:param revision_ids: an iterable of revisions to graph or None for all.
1235
:return: a Graph object with the graph reachable from revision_ids.
1237
if 'evil' in debug.debug_flags:
1239
"get_revision_graph_with_ghosts scales with size of history.")
1240
result = deprecated_graph.Graph()
1241
if not revision_ids:
1242
pending = set(self.all_revision_ids())
1245
pending = set(revision_ids)
1246
# special case NULL_REVISION
1247
if _mod_revision.NULL_REVISION in pending:
1248
pending.remove(_mod_revision.NULL_REVISION)
1249
required = set(pending)
1252
revision_id = pending.pop()
1254
rev = self.get_revision(revision_id)
1255
except errors.NoSuchRevision:
1256
if revision_id in required:
1259
result.add_ghost(revision_id)
1261
for parent_id in rev.parent_ids:
1262
# is this queued or done ?
1263
if (parent_id not in pending and
1264
parent_id not in done):
1266
pending.add(parent_id)
1267
result.add_node(revision_id, rev.parent_ids)
1268
done.add(revision_id)
1271
def _get_history_vf(self):
1272
"""Get a versionedfile whose history graph reflects all revisions.
1274
For weave repositories, this is the inventory weave.
1276
return self.get_inventory_weave()
1278
def iter_reverse_revision_history(self, revision_id):
1279
"""Iterate backwards through revision ids in the lefthand history
1281
:param revision_id: The revision id to start with. All its lefthand
1282
ancestors will be traversed.
1284
if revision_id in (None, _mod_revision.NULL_REVISION):
1286
next_id = revision_id
1287
versionedfile = self._get_history_vf()
1290
parents = versionedfile.get_parents(next_id)
1291
if len(parents) == 0:
1294
next_id = parents[0]
1297
def get_revision_inventory(self, revision_id):
1298
"""Return inventory of a past revision."""
1299
# TODO: Unify this with get_inventory()
1300
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1301
# must be the same as its revision, so this is trivial.
1302
if revision_id is None:
1303
# This does not make sense: if there is no revision,
1304
# then it is the current tree inventory surely ?!
1305
# and thus get_root_id() is something that looks at the last
1306
# commit on the branch, and the get_root_id is an inventory check.
1307
raise NotImplementedError
1308
# return Inventory(self.get_root_id())
1310
return self.get_inventory(revision_id)
1313
def is_shared(self):
1314
"""Return True if this repository is flagged as a shared repository."""
1315
raise NotImplementedError(self.is_shared)
1318
def reconcile(self, other=None, thorough=False):
1319
"""Reconcile this repository."""
1320
from bzrlib.reconcile import RepoReconciler
1321
reconciler = RepoReconciler(self, thorough=thorough)
1322
reconciler.reconcile()
1325
def _refresh_data(self):
1326
"""Helper called from lock_* to ensure coherency with disk.
1328
The default implementation does nothing; it is however possible
1329
for repositories to maintain loaded indices across multiple locks
1330
by checking inside their implementation of this method to see
1331
whether their indices are still valid. This depends of course on
1332
the disk format being validatable in this manner.
1336
def revision_tree(self, revision_id):
1337
"""Return Tree for a revision on this branch.
1339
`revision_id` may be None for the empty tree revision.
1341
# TODO: refactor this to use an existing revision object
1342
# so we don't need to read it in twice.
1343
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1344
return RevisionTree(self, Inventory(root_id=None),
1345
_mod_revision.NULL_REVISION)
1347
inv = self.get_revision_inventory(revision_id)
1348
return RevisionTree(self, inv, revision_id)
1351
def revision_trees(self, revision_ids):
1352
"""Return Tree for a revision on this branch.
1354
`revision_id` may not be None or 'null:'"""
1355
assert None not in revision_ids
1356
assert _mod_revision.NULL_REVISION not in revision_ids
1357
texts = self.get_inventory_weave().get_texts(revision_ids)
1358
for text, revision_id in zip(texts, revision_ids):
1359
inv = self.deserialise_inventory(revision_id, text)
1360
yield RevisionTree(self, inv, revision_id)
1363
def get_ancestry(self, revision_id, topo_sorted=True):
1364
"""Return a list of revision-ids integrated by a revision.
1366
The first element of the list is always None, indicating the origin
1367
revision. This might change when we have history horizons, or
1368
perhaps we should have a new API.
1370
This is topologically sorted.
1372
if _mod_revision.is_null(revision_id):
1374
if not self.has_revision(revision_id):
1375
raise errors.NoSuchRevision(self, revision_id)
1376
w = self.get_inventory_weave()
1377
candidates = w.get_ancestry(revision_id, topo_sorted)
1378
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1381
"""Compress the data within the repository.
1383
This operation only makes sense for some repository types. For other
1384
types it should be a no-op that just returns.
1386
This stub method does not require a lock, but subclasses should use
1387
@needs_write_lock as this is a long running call its reasonable to
1388
implicitly lock for the user.
1392
def print_file(self, file, revision_id):
1393
"""Print `file` to stdout.
1395
FIXME RBC 20060125 as John Meinel points out this is a bad api
1396
- it writes to stdout, it assumes that that is valid etc. Fix
1397
by creating a new more flexible convenience function.
1399
tree = self.revision_tree(revision_id)
1400
# use inventory as it was in that revision
1401
file_id = tree.inventory.path2id(file)
1403
# TODO: jam 20060427 Write a test for this code path
1404
# it had a bug in it, and was raising the wrong
1406
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1407
tree.print_file(file_id)
1409
def get_transaction(self):
1410
return self.control_files.get_transaction()
1412
def revision_parents(self, revision_id):
1413
return self.get_inventory_weave().parent_names(revision_id)
1415
def get_parents(self, revision_ids):
1416
"""See StackedParentsProvider.get_parents"""
1418
for revision_id in revision_ids:
1419
if revision_id == _mod_revision.NULL_REVISION:
1423
parents = self.get_revision(revision_id).parent_ids
1424
except errors.NoSuchRevision:
1427
if len(parents) == 0:
1428
parents = [_mod_revision.NULL_REVISION]
1429
parents_list.append(parents)
1432
def _make_parents_provider(self):
1435
def get_graph(self, other_repository=None):
1436
"""Return the graph walker for this repository format"""
1437
parents_provider = self._make_parents_provider()
1438
if (other_repository is not None and
1439
other_repository.bzrdir.transport.base !=
1440
self.bzrdir.transport.base):
1441
parents_provider = graph._StackedParentsProvider(
1442
[parents_provider, other_repository._make_parents_provider()])
1443
return graph.Graph(parents_provider)
1446
def set_make_working_trees(self, new_value):
1447
"""Set the policy flag for making working trees when creating branches.
1449
This only applies to branches that use this repository.
1451
The default is 'True'.
1452
:param new_value: True to restore the default, False to disable making
1455
raise NotImplementedError(self.set_make_working_trees)
1457
def make_working_trees(self):
1458
"""Returns the policy for making working trees on new branches."""
1459
raise NotImplementedError(self.make_working_trees)
1462
def sign_revision(self, revision_id, gpg_strategy):
1463
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1464
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1467
def has_signature_for_revision_id(self, revision_id):
1468
"""Query for a revision signature for revision_id in the repository."""
1469
return self._revision_store.has_signature(revision_id,
1470
self.get_transaction())
1473
def get_signature_text(self, revision_id):
1474
"""Return the text for a signature."""
1475
return self._revision_store.get_signature_text(revision_id,
1476
self.get_transaction())
1479
def check(self, revision_ids):
1480
"""Check consistency of all history of given revision_ids.
1482
Different repository implementations should override _check().
1484
:param revision_ids: A non-empty list of revision_ids whose ancestry
1485
will be checked. Typically the last revision_id of a branch.
1487
if not revision_ids:
1488
raise ValueError("revision_ids must be non-empty in %s.check"
1490
return self._check(revision_ids)
1492
def _check(self, revision_ids):
1493
result = check.Check(self)
1497
def _warn_if_deprecated(self):
1498
global _deprecation_warning_done
1499
if _deprecation_warning_done:
1501
_deprecation_warning_done = True
1502
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1503
% (self._format, self.bzrdir.transport.base))
1505
def supports_rich_root(self):
1506
return self._format.rich_root_data
1508
def _check_ascii_revisionid(self, revision_id, method):
1509
"""Private helper for ascii-only repositories."""
1510
# weave repositories refuse to store revisionids that are non-ascii.
1511
if revision_id is not None:
1512
# weaves require ascii revision ids.
1513
if isinstance(revision_id, unicode):
1515
revision_id.encode('ascii')
1516
except UnicodeEncodeError:
1517
raise errors.NonAsciiRevisionId(method, self)
1520
revision_id.decode('ascii')
1521
except UnicodeDecodeError:
1522
raise errors.NonAsciiRevisionId(method, self)
1526
# remove these delegates a while after bzr 0.15
1527
def __make_delegated(name, from_module):
1528
def _deprecated_repository_forwarder():
1529
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1530
% (name, from_module),
1533
m = __import__(from_module, globals(), locals(), [name])
1535
return getattr(m, name)
1536
except AttributeError:
1537
raise AttributeError('module %s has no name %s'
1539
globals()[name] = _deprecated_repository_forwarder
1542
'AllInOneRepository',
1543
'WeaveMetaDirRepository',
1544
'PreSplitOutRepositoryFormat',
1545
'RepositoryFormat4',
1546
'RepositoryFormat5',
1547
'RepositoryFormat6',
1548
'RepositoryFormat7',
1550
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1554
'RepositoryFormatKnit',
1555
'RepositoryFormatKnit1',
1557
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1560
def install_revision(repository, rev, revision_tree):
1561
"""Install all revision data into a repository."""
1562
present_parents = []
1564
for p_id in rev.parent_ids:
1565
if repository.has_revision(p_id):
1566
present_parents.append(p_id)
1567
parent_trees[p_id] = repository.revision_tree(p_id)
1569
parent_trees[p_id] = repository.revision_tree(None)
1571
inv = revision_tree.inventory
1572
entries = inv.iter_entries()
1573
# backwards compatibility hack: skip the root id.
1574
if not repository.supports_rich_root():
1575
path, root = entries.next()
1576
if root.revision != rev.revision_id:
1577
raise errors.IncompatibleRevision(repr(repository))
1578
# Add the texts that are not already present
1579
for path, ie in entries:
1580
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1581
repository.get_transaction())
1582
if ie.revision not in w:
1584
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1585
# with InventoryEntry.find_previous_heads(). if it is, then there
1586
# is a latent bug here where the parents may have ancestors of each
1588
for revision, tree in parent_trees.iteritems():
1589
if ie.file_id not in tree:
1591
parent_id = tree.inventory[ie.file_id].revision
1592
if parent_id in text_parents:
1594
text_parents.append(parent_id)
1596
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1597
repository.get_transaction())
1598
lines = revision_tree.get_file(ie.file_id).readlines()
1599
vfile.add_lines(rev.revision_id, text_parents, lines)
1601
# install the inventory
1602
repository.add_inventory(rev.revision_id, inv, present_parents)
1603
except errors.RevisionAlreadyPresent:
1605
repository.add_revision(rev.revision_id, rev, inv)
1608
class MetaDirRepository(Repository):
1609
"""Repositories in the new meta-dir layout."""
1611
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1612
super(MetaDirRepository, self).__init__(_format,
1618
dir_mode = self.control_files._dir_mode
1619
file_mode = self.control_files._file_mode
1622
def is_shared(self):
1623
"""Return True if this repository is flagged as a shared repository."""
1624
return self.control_files._transport.has('shared-storage')
1627
def set_make_working_trees(self, new_value):
1628
"""Set the policy flag for making working trees when creating branches.
1630
This only applies to branches that use this repository.
1632
The default is 'True'.
1633
:param new_value: True to restore the default, False to disable making
1638
self.control_files._transport.delete('no-working-trees')
1639
except errors.NoSuchFile:
1642
self.control_files.put_utf8('no-working-trees', '')
1644
def make_working_trees(self):
1645
"""Returns the policy for making working trees on new branches."""
1646
return not self.control_files._transport.has('no-working-trees')
1649
class RepositoryFormatRegistry(registry.Registry):
1650
"""Registry of RepositoryFormats."""
1652
def get(self, format_string):
1653
r = registry.Registry.get(self, format_string)
1659
format_registry = RepositoryFormatRegistry()
1660
"""Registry of formats, indexed by their identifying format string.
1662
This can contain either format instances themselves, or classes/factories that
1663
can be called to obtain one.
1667
#####################################################################
1668
# Repository Formats
1670
class RepositoryFormat(object):
1671
"""A repository format.
1673
Formats provide three things:
1674
* An initialization routine to construct repository data on disk.
1675
* a format string which is used when the BzrDir supports versioned
1677
* an open routine which returns a Repository instance.
1679
There is one and only one Format subclass for each on-disk format. But
1680
there can be one Repository subclass that is used for several different
1681
formats. The _format attribute on a Repository instance can be used to
1682
determine the disk format.
1684
Formats are placed in an dict by their format string for reference
1685
during opening. These should be subclasses of RepositoryFormat
1688
Once a format is deprecated, just deprecate the initialize and open
1689
methods on the format class. Do not deprecate the object, as the
1690
object will be created every system load.
1692
Common instance attributes:
1693
_matchingbzrdir - the bzrdir format that the repository format was
1694
originally written to work with. This can be used if manually
1695
constructing a bzrdir and repository, or more commonly for test suite
1700
return "<%s>" % self.__class__.__name__
1702
def __eq__(self, other):
1703
# format objects are generally stateless
1704
return isinstance(other, self.__class__)
1706
def __ne__(self, other):
1707
return not self == other
1710
def find_format(klass, a_bzrdir):
1711
"""Return the format for the repository object in a_bzrdir.
1713
This is used by bzr native formats that have a "format" file in
1714
the repository. Other methods may be used by different types of
1718
transport = a_bzrdir.get_repository_transport(None)
1719
format_string = transport.get("format").read()
1720
return format_registry.get(format_string)
1721
except errors.NoSuchFile:
1722
raise errors.NoRepositoryPresent(a_bzrdir)
1724
raise errors.UnknownFormatError(format=format_string)
1727
def register_format(klass, format):
1728
format_registry.register(format.get_format_string(), format)
1731
def unregister_format(klass, format):
1732
format_registry.remove(format.get_format_string())
1735
def get_default_format(klass):
1736
"""Return the current default format."""
1737
from bzrlib import bzrdir
1738
return bzrdir.format_registry.make_bzrdir('default').repository_format
1740
def _get_control_store(self, repo_transport, control_files):
1741
"""Return the control store for this repository."""
1742
raise NotImplementedError(self._get_control_store)
1744
def get_format_string(self):
1745
"""Return the ASCII format string that identifies this format.
1747
Note that in pre format ?? repositories the format string is
1748
not permitted nor written to disk.
1750
raise NotImplementedError(self.get_format_string)
1752
def get_format_description(self):
1753
"""Return the short description for this format."""
1754
raise NotImplementedError(self.get_format_description)
1756
def _get_revision_store(self, repo_transport, control_files):
1757
"""Return the revision store object for this a_bzrdir."""
1758
raise NotImplementedError(self._get_revision_store)
1760
def _get_text_rev_store(self,
1767
"""Common logic for getting a revision store for a repository.
1769
see self._get_revision_store for the subclass-overridable method to
1770
get the store for a repository.
1772
from bzrlib.store.revision.text import TextRevisionStore
1773
dir_mode = control_files._dir_mode
1774
file_mode = control_files._file_mode
1775
text_store = TextStore(transport.clone(name),
1777
compressed=compressed,
1779
file_mode=file_mode)
1780
_revision_store = TextRevisionStore(text_store, serializer)
1781
return _revision_store
1783
# TODO: this shouldn't be in the base class, it's specific to things that
1784
# use weaves or knits -- mbp 20070207
1785
def _get_versioned_file_store(self,
1790
versionedfile_class=None,
1791
versionedfile_kwargs={},
1793
if versionedfile_class is None:
1794
versionedfile_class = self._versionedfile_class
1795
weave_transport = control_files._transport.clone(name)
1796
dir_mode = control_files._dir_mode
1797
file_mode = control_files._file_mode
1798
return VersionedFileStore(weave_transport, prefixed=prefixed,
1800
file_mode=file_mode,
1801
versionedfile_class=versionedfile_class,
1802
versionedfile_kwargs=versionedfile_kwargs,
1805
def initialize(self, a_bzrdir, shared=False):
1806
"""Initialize a repository of this format in a_bzrdir.
1808
:param a_bzrdir: The bzrdir to put the new repository in it.
1809
:param shared: The repository should be initialized as a sharable one.
1810
:returns: The new repository object.
1812
This may raise UninitializableFormat if shared repository are not
1813
compatible the a_bzrdir.
1815
raise NotImplementedError(self.initialize)
1817
def is_supported(self):
1818
"""Is this format supported?
1820
Supported formats must be initializable and openable.
1821
Unsupported formats may not support initialization or committing or
1822
some other features depending on the reason for not being supported.
1826
def check_conversion_target(self, target_format):
1827
raise NotImplementedError(self.check_conversion_target)
1829
def open(self, a_bzrdir, _found=False):
1830
"""Return an instance of this format for the bzrdir a_bzrdir.
1832
_found is a private parameter, do not use it.
1834
raise NotImplementedError(self.open)
1837
class MetaDirRepositoryFormat(RepositoryFormat):
1838
"""Common base class for the new repositories using the metadir layout."""
1840
rich_root_data = False
1841
supports_tree_reference = False
1842
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
1845
super(MetaDirRepositoryFormat, self).__init__()
1847
def _create_control_files(self, a_bzrdir):
1848
"""Create the required files and the initial control_files object."""
1849
# FIXME: RBC 20060125 don't peek under the covers
1850
# NB: no need to escape relative paths that are url safe.
1851
repository_transport = a_bzrdir.get_repository_transport(self)
1852
control_files = lockable_files.LockableFiles(repository_transport,
1853
'lock', lockdir.LockDir)
1854
control_files.create_lock()
1855
return control_files
1857
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
1858
"""Upload the initial blank content."""
1859
control_files = self._create_control_files(a_bzrdir)
1860
control_files.lock_write()
1862
control_files._transport.mkdir_multi(dirs,
1863
mode=control_files._dir_mode)
1864
for file, content in files:
1865
control_files.put(file, content)
1866
for file, content in utf8_files:
1867
control_files.put_utf8(file, content)
1869
control_files.put_utf8('shared-storage', '')
1871
control_files.unlock()
1874
# formats which have no format string are not discoverable
1875
# and not independently creatable, so are not registered. They're
1876
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1877
# needed, it's constructed directly by the BzrDir. Non-native formats where
1878
# the repository is not separately opened are similar.
1880
format_registry.register_lazy(
1881
'Bazaar-NG Repository format 7',
1882
'bzrlib.repofmt.weaverepo',
1886
# KEEP in sync with bzrdir.format_registry default, which controls the overall
1887
# default control directory format
1888
format_registry.register_lazy(
1889
'Bazaar-NG Knit Repository Format 1',
1890
'bzrlib.repofmt.knitrepo',
1891
'RepositoryFormatKnit1',
1893
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
1895
format_registry.register_lazy(
1896
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
1897
'bzrlib.repofmt.knitrepo',
1898
'RepositoryFormatKnit3',
1902
class InterRepository(InterObject):
1903
"""This class represents operations taking place between two repositories.
1905
Its instances have methods like copy_content and fetch, and contain
1906
references to the source and target repositories these operations can be
1909
Often we will provide convenience methods on 'repository' which carry out
1910
operations with another repository - they will always forward to
1911
InterRepository.get(other).method_name(parameters).
1915
"""The available optimised InterRepository types."""
1917
def copy_content(self, revision_id=None):
1918
raise NotImplementedError(self.copy_content)
1920
def fetch(self, revision_id=None, pb=None):
1921
"""Fetch the content required to construct revision_id.
1923
The content is copied from self.source to self.target.
1925
:param revision_id: if None all content is copied, if NULL_REVISION no
1927
:param pb: optional progress bar to use for progress reports. If not
1928
provided a default one will be created.
1930
Returns the copied revision count and the failed revisions in a tuple:
1933
raise NotImplementedError(self.fetch)
1936
def missing_revision_ids(self, revision_id=None):
1937
"""Return the revision ids that source has that target does not.
1939
These are returned in topological order.
1941
:param revision_id: only return revision ids included by this
1944
# generic, possibly worst case, slow code path.
1945
target_ids = set(self.target.all_revision_ids())
1946
if revision_id is not None:
1947
source_ids = self.source.get_ancestry(revision_id)
1948
assert source_ids[0] is None
1951
source_ids = self.source.all_revision_ids()
1952
result_set = set(source_ids).difference(target_ids)
1953
# this may look like a no-op: its not. It preserves the ordering
1954
# other_ids had while only returning the members from other_ids
1955
# that we've decided we need.
1956
return [rev_id for rev_id in source_ids if rev_id in result_set]
1959
def _same_model(source, target):
1960
"""True if source and target have the same data representation."""
1961
if source.supports_rich_root() != target.supports_rich_root():
1963
if source._serializer != target._serializer:
1968
class InterSameDataRepository(InterRepository):
1969
"""Code for converting between repositories that represent the same data.
1971
Data format and model must match for this to work.
1975
def _get_repo_format_to_test(self):
1976
"""Repository format for testing with.
1978
InterSameData can pull from subtree to subtree and from non-subtree to
1979
non-subtree, so we test this with the richest repository format.
1981
from bzrlib.repofmt import knitrepo
1982
return knitrepo.RepositoryFormatKnit3()
1985
def is_compatible(source, target):
1986
return InterRepository._same_model(source, target)
1989
def copy_content(self, revision_id=None):
1990
"""Make a complete copy of the content in self into destination.
1992
This copies both the repository's revision data, and configuration information
1993
such as the make_working_trees setting.
1995
This is a destructive operation! Do not use it on existing
1998
:param revision_id: Only copy the content needed to construct
1999
revision_id and its parents.
2002
self.target.set_make_working_trees(self.source.make_working_trees())
2003
except NotImplementedError:
2005
# but don't bother fetching if we have the needed data now.
2006
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2007
self.target.has_revision(revision_id)):
2009
self.target.fetch(self.source, revision_id=revision_id)
2012
def fetch(self, revision_id=None, pb=None):
2013
"""See InterRepository.fetch()."""
2014
from bzrlib.fetch import GenericRepoFetcher
2015
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2016
self.source, self.source._format, self.target,
2017
self.target._format)
2018
f = GenericRepoFetcher(to_repository=self.target,
2019
from_repository=self.source,
2020
last_revision=revision_id,
2022
return f.count_copied, f.failed_revisions
2025
class InterWeaveRepo(InterSameDataRepository):
2026
"""Optimised code paths between Weave based repositories.
2028
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2029
implemented lazy inter-object optimisation.
2033
def _get_repo_format_to_test(self):
2034
from bzrlib.repofmt import weaverepo
2035
return weaverepo.RepositoryFormat7()
2038
def is_compatible(source, target):
2039
"""Be compatible with known Weave formats.
2041
We don't test for the stores being of specific types because that
2042
could lead to confusing results, and there is no need to be
2045
from bzrlib.repofmt.weaverepo import (
2051
return (isinstance(source._format, (RepositoryFormat5,
2053
RepositoryFormat7)) and
2054
isinstance(target._format, (RepositoryFormat5,
2056
RepositoryFormat7)))
2057
except AttributeError:
2061
def copy_content(self, revision_id=None):
2062
"""See InterRepository.copy_content()."""
2063
# weave specific optimised path:
2065
self.target.set_make_working_trees(self.source.make_working_trees())
2066
except NotImplementedError:
2068
# FIXME do not peek!
2069
if self.source.control_files._transport.listable():
2070
pb = ui.ui_factory.nested_progress_bar()
2072
self.target.weave_store.copy_all_ids(
2073
self.source.weave_store,
2075
from_transaction=self.source.get_transaction(),
2076
to_transaction=self.target.get_transaction())
2077
pb.update('copying inventory', 0, 1)
2078
self.target.control_weaves.copy_multi(
2079
self.source.control_weaves, ['inventory'],
2080
from_transaction=self.source.get_transaction(),
2081
to_transaction=self.target.get_transaction())
2082
self.target._revision_store.text_store.copy_all_ids(
2083
self.source._revision_store.text_store,
2088
self.target.fetch(self.source, revision_id=revision_id)
2091
def fetch(self, revision_id=None, pb=None):
2092
"""See InterRepository.fetch()."""
2093
from bzrlib.fetch import GenericRepoFetcher
2094
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2095
self.source, self.source._format, self.target, self.target._format)
2096
f = GenericRepoFetcher(to_repository=self.target,
2097
from_repository=self.source,
2098
last_revision=revision_id,
2100
return f.count_copied, f.failed_revisions
2103
def missing_revision_ids(self, revision_id=None):
2104
"""See InterRepository.missing_revision_ids()."""
2105
# we want all revisions to satisfy revision_id in source.
2106
# but we don't want to stat every file here and there.
2107
# we want then, all revisions other needs to satisfy revision_id
2108
# checked, but not those that we have locally.
2109
# so the first thing is to get a subset of the revisions to
2110
# satisfy revision_id in source, and then eliminate those that
2111
# we do already have.
2112
# this is slow on high latency connection to self, but as as this
2113
# disk format scales terribly for push anyway due to rewriting
2114
# inventory.weave, this is considered acceptable.
2116
if revision_id is not None:
2117
source_ids = self.source.get_ancestry(revision_id)
2118
assert source_ids[0] is None
2121
source_ids = self.source._all_possible_ids()
2122
source_ids_set = set(source_ids)
2123
# source_ids is the worst possible case we may need to pull.
2124
# now we want to filter source_ids against what we actually
2125
# have in target, but don't try to check for existence where we know
2126
# we do not have a revision as that would be pointless.
2127
target_ids = set(self.target._all_possible_ids())
2128
possibly_present_revisions = target_ids.intersection(source_ids_set)
2129
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2130
required_revisions = source_ids_set.difference(actually_present_revisions)
2131
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2132
if revision_id is not None:
2133
# we used get_ancestry to determine source_ids then we are assured all
2134
# revisions referenced are present as they are installed in topological order.
2135
# and the tip revision was validated by get_ancestry.
2136
return required_topo_revisions
2138
# if we just grabbed the possibly available ids, then
2139
# we only have an estimate of whats available and need to validate
2140
# that against the revision records.
2141
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2144
class InterKnitRepo(InterSameDataRepository):
2145
"""Optimised code paths between Knit based repositories."""
2148
def _get_repo_format_to_test(self):
2149
from bzrlib.repofmt import knitrepo
2150
return knitrepo.RepositoryFormatKnit1()
2153
def is_compatible(source, target):
2154
"""Be compatible with known Knit formats.
2156
We don't test for the stores being of specific types because that
2157
could lead to confusing results, and there is no need to be
2160
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2162
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2163
isinstance(target._format, RepositoryFormatKnit))
2164
except AttributeError:
2166
return are_knits and InterRepository._same_model(source, target)
2169
def fetch(self, revision_id=None, pb=None):
2170
"""See InterRepository.fetch()."""
2171
from bzrlib.fetch import KnitRepoFetcher
2172
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2173
self.source, self.source._format, self.target, self.target._format)
2174
f = KnitRepoFetcher(to_repository=self.target,
2175
from_repository=self.source,
2176
last_revision=revision_id,
2178
return f.count_copied, f.failed_revisions
2181
def missing_revision_ids(self, revision_id=None):
2182
"""See InterRepository.missing_revision_ids()."""
2183
if revision_id is not None:
2184
source_ids = self.source.get_ancestry(revision_id)
2185
assert source_ids[0] is None
2188
source_ids = self.source.all_revision_ids()
2189
source_ids_set = set(source_ids)
2190
# source_ids is the worst possible case we may need to pull.
2191
# now we want to filter source_ids against what we actually
2192
# have in target, but don't try to check for existence where we know
2193
# we do not have a revision as that would be pointless.
2194
target_ids = set(self.target.all_revision_ids())
2195
possibly_present_revisions = target_ids.intersection(source_ids_set)
2196
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2197
required_revisions = source_ids_set.difference(actually_present_revisions)
2198
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2199
if revision_id is not None:
2200
# we used get_ancestry to determine source_ids then we are assured all
2201
# revisions referenced are present as they are installed in topological order.
2202
# and the tip revision was validated by get_ancestry.
2203
return required_topo_revisions
2205
# if we just grabbed the possibly available ids, then
2206
# we only have an estimate of whats available and need to validate
2207
# that against the revision records.
2208
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2211
class InterModel1and2(InterRepository):
2214
def _get_repo_format_to_test(self):
2218
def is_compatible(source, target):
2219
if not source.supports_rich_root() and target.supports_rich_root():
2225
def fetch(self, revision_id=None, pb=None):
2226
"""See InterRepository.fetch()."""
2227
from bzrlib.fetch import Model1toKnit2Fetcher
2228
f = Model1toKnit2Fetcher(to_repository=self.target,
2229
from_repository=self.source,
2230
last_revision=revision_id,
2232
return f.count_copied, f.failed_revisions
2235
def copy_content(self, revision_id=None):
2236
"""Make a complete copy of the content in self into destination.
2238
This is a destructive operation! Do not use it on existing
2241
:param revision_id: Only copy the content needed to construct
2242
revision_id and its parents.
2245
self.target.set_make_working_trees(self.source.make_working_trees())
2246
except NotImplementedError:
2248
# but don't bother fetching if we have the needed data now.
2249
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2250
self.target.has_revision(revision_id)):
2252
self.target.fetch(self.source, revision_id=revision_id)
2255
class InterKnit1and2(InterKnitRepo):
2258
def _get_repo_format_to_test(self):
2262
def is_compatible(source, target):
2263
"""Be compatible with Knit1 source and Knit3 target"""
2264
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2266
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit1, \
2267
RepositoryFormatKnit3
2268
return (isinstance(source._format, (RepositoryFormatKnit1)) and
2269
isinstance(target._format, (RepositoryFormatKnit3)))
2270
except AttributeError:
2274
def fetch(self, revision_id=None, pb=None):
2275
"""See InterRepository.fetch()."""
2276
from bzrlib.fetch import Knit1to2Fetcher
2277
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2278
self.source, self.source._format, self.target,
2279
self.target._format)
2280
f = Knit1to2Fetcher(to_repository=self.target,
2281
from_repository=self.source,
2282
last_revision=revision_id,
2284
return f.count_copied, f.failed_revisions
2287
class InterRemoteRepository(InterRepository):
2288
"""Code for converting between RemoteRepository objects.
2290
This just gets an non-remote repository from the RemoteRepository, and calls
2291
InterRepository.get again.
2294
def __init__(self, source, target):
2295
if isinstance(source, remote.RemoteRepository):
2296
source._ensure_real()
2297
real_source = source._real_repository
2299
real_source = source
2300
if isinstance(target, remote.RemoteRepository):
2301
target._ensure_real()
2302
real_target = target._real_repository
2304
real_target = target
2305
self.real_inter = InterRepository.get(real_source, real_target)
2308
def is_compatible(source, target):
2309
if isinstance(source, remote.RemoteRepository):
2311
if isinstance(target, remote.RemoteRepository):
2315
def copy_content(self, revision_id=None):
2316
self.real_inter.copy_content(revision_id=revision_id)
2318
def fetch(self, revision_id=None, pb=None):
2319
self.real_inter.fetch(revision_id=revision_id, pb=pb)
2322
def _get_repo_format_to_test(self):
2326
InterRepository.register_optimiser(InterSameDataRepository)
2327
InterRepository.register_optimiser(InterWeaveRepo)
2328
InterRepository.register_optimiser(InterKnitRepo)
2329
InterRepository.register_optimiser(InterModel1and2)
2330
InterRepository.register_optimiser(InterKnit1and2)
2331
InterRepository.register_optimiser(InterRemoteRepository)
2334
class CopyConverter(object):
2335
"""A repository conversion tool which just performs a copy of the content.
2337
This is slow but quite reliable.
2340
def __init__(self, target_format):
2341
"""Create a CopyConverter.
2343
:param target_format: The format the resulting repository should be.
2345
self.target_format = target_format
2347
def convert(self, repo, pb):
2348
"""Perform the conversion of to_convert, giving feedback via pb.
2350
:param to_convert: The disk object to convert.
2351
:param pb: a progress bar to use for progress information.
2356
# this is only useful with metadir layouts - separated repo content.
2357
# trigger an assertion if not such
2358
repo._format.get_format_string()
2359
self.repo_dir = repo.bzrdir
2360
self.step('Moving repository to repository.backup')
2361
self.repo_dir.transport.move('repository', 'repository.backup')
2362
backup_transport = self.repo_dir.transport.clone('repository.backup')
2363
repo._format.check_conversion_target(self.target_format)
2364
self.source_repo = repo._format.open(self.repo_dir,
2366
_override_transport=backup_transport)
2367
self.step('Creating new repository')
2368
converted = self.target_format.initialize(self.repo_dir,
2369
self.source_repo.is_shared())
2370
converted.lock_write()
2372
self.step('Copying content into repository.')
2373
self.source_repo.copy_content_into(converted)
2376
self.step('Deleting old repository content.')
2377
self.repo_dir.transport.delete_tree('repository.backup')
2378
self.pb.note('repository converted')
2380
def step(self, message):
2381
"""Update the pb by a step."""
2383
self.pb.update(message, self.count, self.total)
2395
def _unescaper(match, _map=_unescape_map):
2396
code = match.group(1)
2400
if not code.startswith('#'):
2402
return unichr(int(code[1:])).encode('utf8')
2408
def _unescape_xml(data):
2409
"""Unescape predefined XML entities in a string of data."""
2411
if _unescape_re is None:
2412
_unescape_re = re.compile('\&([^;]*);')
2413
return _unescape_re.sub(_unescaper, data)