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(), """
40
revision as _mod_revision,
46
from bzrlib.bundle import serializer
47
from bzrlib.revisiontree import RevisionTree
48
from bzrlib.store.versioned import VersionedFileStore
49
from bzrlib.store.text import TextStore
50
from bzrlib.testament import Testament
51
from bzrlib.util import bencode
54
from bzrlib.decorators import needs_read_lock, needs_write_lock
55
from bzrlib.inter import InterObject
56
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
57
from bzrlib.symbol_versioning import (
60
from bzrlib.trace import mutter, mutter_callsite, note, warning
63
# Old formats display a warning, but only once
64
_deprecation_warning_done = False
67
class CommitBuilder(object):
68
"""Provides an interface to build up a commit.
70
This allows describing a tree to be committed without needing to
71
know the internals of the format of the repository.
74
# all clients should supply tree roots.
75
record_root_entry = True
76
# the default CommitBuilder does not manage trees whose root is versioned.
77
_versioned_root = False
79
def __init__(self, repository, parents, config, timestamp=None,
80
timezone=None, committer=None, revprops=None,
82
"""Initiate a CommitBuilder.
84
:param repository: Repository to commit to.
85
:param parents: Revision ids of the parents of the new revision.
86
:param config: Configuration to use.
87
:param timestamp: Optional timestamp recorded for commit.
88
:param timezone: Optional timezone for timestamp.
89
:param committer: Optional committer to set for commit.
90
:param revprops: Optional dictionary of revision properties.
91
:param revision_id: Optional revision id.
96
self._committer = self._config.username()
98
assert isinstance(committer, basestring), type(committer)
99
self._committer = committer
101
self.new_inventory = Inventory(None)
102
self._new_revision_id = revision_id
103
self.parents = parents
104
self.repository = repository
107
if revprops is not None:
108
self._revprops.update(revprops)
110
if timestamp is None:
111
timestamp = time.time()
112
# Restrict resolution to 1ms
113
self._timestamp = round(timestamp, 3)
116
self._timezone = osutils.local_time_offset()
118
self._timezone = int(timezone)
120
self._generate_revision_if_needed()
121
self.__heads = graph.HeadsCache(repository.get_graph()).heads
123
def commit(self, message):
124
"""Make the actual commit.
126
:return: The revision id of the recorded revision.
128
rev = _mod_revision.Revision(
129
timestamp=self._timestamp,
130
timezone=self._timezone,
131
committer=self._committer,
133
inventory_sha1=self.inv_sha1,
134
revision_id=self._new_revision_id,
135
properties=self._revprops)
136
rev.parent_ids = self.parents
137
self.repository.add_revision(self._new_revision_id, rev,
138
self.new_inventory, self._config)
139
self.repository.commit_write_group()
140
return self._new_revision_id
143
"""Abort the commit that is being built.
145
self.repository.abort_write_group()
147
def revision_tree(self):
148
"""Return the tree that was just committed.
150
After calling commit() this can be called to get a RevisionTree
151
representing the newly committed tree. This is preferred to
152
calling Repository.revision_tree() because that may require
153
deserializing the inventory, while we already have a copy in
156
return RevisionTree(self.repository, self.new_inventory,
157
self._new_revision_id)
159
def finish_inventory(self):
160
"""Tell the builder that the inventory is finished."""
161
if self.new_inventory.root is None:
162
raise AssertionError('Root entry should be supplied to'
163
' record_entry_contents, as of bzr 0.10.',
164
DeprecationWarning, stacklevel=2)
165
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
166
self.new_inventory.revision_id = self._new_revision_id
167
self.inv_sha1 = self.repository.add_inventory(
168
self._new_revision_id,
173
def _gen_revision_id(self):
174
"""Return new revision-id."""
175
return generate_ids.gen_revision_id(self._config.username(),
178
def _generate_revision_if_needed(self):
179
"""Create a revision id if None was supplied.
181
If the repository can not support user-specified revision ids
182
they should override this function and raise CannotSetRevisionId
183
if _new_revision_id is not None.
185
:raises: CannotSetRevisionId
187
if self._new_revision_id is None:
188
self._new_revision_id = self._gen_revision_id()
189
self.random_revid = True
191
self.random_revid = False
193
def _heads(self, file_id, revision_ids):
194
"""Calculate the graph heads for revision_ids in the graph of file_id.
196
This can use either a per-file graph or a global revision graph as we
197
have an identity relationship between the two graphs.
199
return self.__heads(revision_ids)
201
def _check_root(self, ie, parent_invs, tree):
202
"""Helper for record_entry_contents.
204
:param ie: An entry being added.
205
:param parent_invs: The inventories of the parent revisions of the
207
:param tree: The tree that is being committed.
209
# In this revision format, root entries have no knit or weave When
210
# serializing out to disk and back in root.revision is always
212
ie.revision = self._new_revision_id
214
def _get_delta(self, ie, basis_inv, path):
215
"""Get a delta against the basis inventory for ie."""
216
if ie.file_id not in basis_inv:
218
return (None, path, ie.file_id, ie)
219
elif ie != basis_inv[ie.file_id]:
221
# TODO: avoid tis id2path call.
222
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
227
def record_entry_contents(self, ie, parent_invs, path, tree,
229
"""Record the content of ie from tree into the commit if needed.
231
Side effect: sets ie.revision when unchanged
233
:param ie: An inventory entry present in the commit.
234
:param parent_invs: The inventories of the parent revisions of the
236
:param path: The path the entry is at in the tree.
237
:param tree: The tree which contains this entry and should be used to
239
:param content_summary: Summary data from the tree about the paths
240
content - stat, length, exec, sha/link target. This is only
241
accessed when the entry has a revision of None - that is when it is
242
a candidate to commit.
243
:return: A tuple (change_delta, version_recorded). change_delta is
244
an inventory_delta change for this entry against the basis tree of
245
the commit, or None if no change occured against the basis tree.
246
version_recorded is True if a new version of the entry has been
247
recorded. For instance, committing a merge where a file was only
248
changed on the other side will return (delta, False).
250
if self.new_inventory.root is None:
251
if ie.parent_id is not None:
252
raise errors.RootMissing()
253
self._check_root(ie, parent_invs, tree)
254
if ie.revision is None:
255
kind = content_summary[0]
257
# ie is carried over from a prior commit
259
# XXX: repository specific check for nested tree support goes here - if
260
# the repo doesn't want nested trees we skip it ?
261
if (kind == 'tree-reference' and
262
not self.repository._format.supports_tree_reference):
263
# mismatch between commit builder logic and repository:
264
# this needs the entry creation pushed down into the builder.
265
raise NotImplementedError('Missing repository subtree support.')
266
self.new_inventory.add(ie)
268
# TODO: slow, take it out of the inner loop.
270
basis_inv = parent_invs[0]
272
basis_inv = Inventory(root_id=None)
274
# ie.revision is always None if the InventoryEntry is considered
275
# for committing. We may record the previous parents revision if the
276
# content is actually unchanged against a sole head.
277
if ie.revision is not None:
278
if not self._versioned_root and path == '':
279
# repositories that do not version the root set the root's
280
# revision to the new commit even when no change occurs, and
281
# this masks when a change may have occurred against the basis,
282
# so calculate if one happened.
283
if ie.file_id in basis_inv:
284
delta = (basis_inv.id2path(ie.file_id), path,
288
delta = (None, path, ie.file_id, ie)
291
# we don't need to commit this, because the caller already
292
# determined that an existing revision of this file is
294
return None, (ie.revision == self._new_revision_id)
295
# XXX: Friction: parent_candidates should return a list not a dict
296
# so that we don't have to walk the inventories again.
297
parent_candiate_entries = ie.parent_candidates(parent_invs)
298
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
300
for inv in parent_invs:
301
if ie.file_id in inv:
302
old_rev = inv[ie.file_id].revision
303
if old_rev in head_set:
304
heads.append(inv[ie.file_id].revision)
305
head_set.remove(inv[ie.file_id].revision)
308
# now we check to see if we need to write a new record to the
310
# We write a new entry unless there is one head to the ancestors, and
311
# the kind-derived content is unchanged.
313
# Cheapest check first: no ancestors, or more the one head in the
314
# ancestors, we write a new node.
318
# There is a single head, look it up for comparison
319
parent_entry = parent_candiate_entries[heads[0]]
320
# if the non-content specific data has changed, we'll be writing a
322
if (parent_entry.parent_id != ie.parent_id or
323
parent_entry.name != ie.name):
325
# now we need to do content specific checks:
327
# if the kind changed the content obviously has
328
if kind != parent_entry.kind:
331
assert content_summary[2] is not None, \
332
"Files must not have executable = None"
334
if (# if the file length changed we have to store:
335
parent_entry.text_size != content_summary[1] or
336
# if the exec bit has changed we have to store:
337
parent_entry.executable != content_summary[2]):
339
elif parent_entry.text_sha1 == content_summary[3]:
340
# all meta and content is unchanged (using a hash cache
341
# hit to check the sha)
342
ie.revision = parent_entry.revision
343
ie.text_size = parent_entry.text_size
344
ie.text_sha1 = parent_entry.text_sha1
345
ie.executable = parent_entry.executable
346
return self._get_delta(ie, basis_inv, path), False
348
# Either there is only a hash change(no hash cache entry,
349
# or same size content change), or there is no change on
351
# Provide the parent's hash to the store layer, so that the
352
# content is unchanged we will not store a new node.
353
nostore_sha = parent_entry.text_sha1
355
# We want to record a new node regardless of the presence or
356
# absence of a content change in the file.
358
ie.executable = content_summary[2]
359
lines = tree.get_file(ie.file_id, path).readlines()
361
ie.text_sha1, ie.text_size = self._add_text_to_weave(
362
ie.file_id, lines, heads, nostore_sha)
363
except errors.ExistingContent:
364
# Turns out that the file content was unchanged, and we were
365
# only going to store a new node if it was changed. Carry over
367
ie.revision = parent_entry.revision
368
ie.text_size = parent_entry.text_size
369
ie.text_sha1 = parent_entry.text_sha1
370
ie.executable = parent_entry.executable
371
return self._get_delta(ie, basis_inv, path), False
372
elif kind == 'directory':
374
# all data is meta here, nothing specific to directory, so
376
ie.revision = parent_entry.revision
377
return self._get_delta(ie, basis_inv, path), False
379
self._add_text_to_weave(ie.file_id, lines, heads, None)
380
elif kind == 'symlink':
381
current_link_target = content_summary[3]
383
# symlink target is not generic metadata, check if it has
385
if current_link_target != parent_entry.symlink_target:
388
# unchanged, carry over.
389
ie.revision = parent_entry.revision
390
ie.symlink_target = parent_entry.symlink_target
391
return self._get_delta(ie, basis_inv, path), False
392
ie.symlink_target = current_link_target
394
self._add_text_to_weave(ie.file_id, lines, heads, None)
395
elif kind == 'tree-reference':
397
if content_summary[3] != parent_entry.reference_revision:
400
# unchanged, carry over.
401
ie.reference_revision = parent_entry.reference_revision
402
ie.revision = parent_entry.revision
403
return self._get_delta(ie, basis_inv, path), False
404
ie.reference_revision = content_summary[3]
406
self._add_text_to_weave(ie.file_id, lines, heads, None)
408
raise NotImplementedError('unknown kind')
409
ie.revision = self._new_revision_id
410
return self._get_delta(ie, basis_inv, path), True
412
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
413
versionedfile = self.repository.weave_store.get_weave_or_empty(
414
file_id, self.repository.get_transaction())
415
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
416
# than add_lines, and allows committing when a parent is ghosted for
418
# Note: as we read the content directly from the tree, we know its not
419
# been turned into unicode or badly split - but a broken tree
420
# implementation could give us bad output from readlines() so this is
421
# not a guarantee of safety. What would be better is always checking
422
# the content during test suite execution. RBC 20070912
424
return versionedfile.add_lines_with_ghosts(
425
self._new_revision_id, parents, new_lines,
426
nostore_sha=nostore_sha, random_id=self.random_revid,
427
check_content=False)[0:2]
429
versionedfile.clear_cache()
432
class RootCommitBuilder(CommitBuilder):
433
"""This commitbuilder actually records the root id"""
435
# the root entry gets versioned properly by this builder.
436
_versioned_root = True
438
def _check_root(self, ie, parent_invs, tree):
439
"""Helper for record_entry_contents.
441
:param ie: An entry being added.
442
:param parent_invs: The inventories of the parent revisions of the
444
:param tree: The tree that is being committed.
448
######################################################################
451
class Repository(object):
452
"""Repository holding history for one or more branches.
454
The repository holds and retrieves historical information including
455
revisions and file history. It's normally accessed only by the Branch,
456
which views a particular line of development through that history.
458
The Repository builds on top of Stores and a Transport, which respectively
459
describe the disk data format and the way of accessing the (possibly
463
# What class to use for a CommitBuilder. Often its simpler to change this
464
# in a Repository class subclass rather than to override
465
# get_commit_builder.
466
_commit_builder_class = CommitBuilder
467
# The search regex used by xml based repositories to determine what things
468
# where changed in a single commit.
469
_file_ids_altered_regex = lazy_regex.lazy_compile(
470
r'file_id="(?P<file_id>[^"]+)"'
471
r'.* revision="(?P<revision_id>[^"]+)"'
474
def abort_write_group(self):
475
"""Commit the contents accrued within the current write group.
477
:seealso: start_write_group.
479
if self._write_group is not self.get_transaction():
480
# has an unlock or relock occured ?
481
raise errors.BzrError('mismatched lock context and write group.')
482
self._abort_write_group()
483
self._write_group = None
485
def _abort_write_group(self):
486
"""Template method for per-repository write group cleanup.
488
This is called during abort before the write group is considered to be
489
finished and should cleanup any internal state accrued during the write
490
group. There is no requirement that data handed to the repository be
491
*not* made available - this is not a rollback - but neither should any
492
attempt be made to ensure that data added is fully commited. Abort is
493
invoked when an error has occured so futher disk or network operations
494
may not be possible or may error and if possible should not be
499
def add_inventory(self, revision_id, inv, parents):
500
"""Add the inventory inv to the repository as revision_id.
502
:param parents: The revision ids of the parents that revision_id
503
is known to have and are in the repository already.
505
:returns: The validator(which is a sha1 digest, though what is sha'd is
506
repository format specific) of the serialized inventory.
508
assert self.is_in_write_group()
509
_mod_revision.check_not_reserved_id(revision_id)
510
assert inv.revision_id is None or inv.revision_id == revision_id, \
511
"Mismatch between inventory revision" \
512
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
513
assert inv.root is not None
514
inv_lines = self._serialise_inventory_to_lines(inv)
515
inv_vf = self.get_inventory_weave()
516
return self._inventory_add_lines(inv_vf, revision_id, parents,
517
inv_lines, check_content=False)
519
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
521
"""Store lines in inv_vf and return the sha1 of the inventory."""
523
for parent in parents:
525
final_parents.append(parent)
526
return inv_vf.add_lines(revision_id, final_parents, lines,
527
check_content=check_content)[0]
530
def add_revision(self, revision_id, rev, inv=None, config=None):
531
"""Add rev to the revision store as revision_id.
533
:param revision_id: the revision id to use.
534
:param rev: The revision object.
535
:param inv: The inventory for the revision. if None, it will be looked
536
up in the inventory storer
537
:param config: If None no digital signature will be created.
538
If supplied its signature_needed method will be used
539
to determine if a signature should be made.
541
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
543
_mod_revision.check_not_reserved_id(revision_id)
544
if config is not None and config.signature_needed():
546
inv = self.get_inventory(revision_id)
547
plaintext = Testament(rev, inv).as_short_text()
548
self.store_revision_signature(
549
gpg.GPGStrategy(config), plaintext, revision_id)
550
if not revision_id in self.get_inventory_weave():
552
raise errors.WeaveRevisionNotPresent(revision_id,
553
self.get_inventory_weave())
555
# yes, this is not suitable for adding with ghosts.
556
self.add_inventory(revision_id, inv, rev.parent_ids)
557
self._revision_store.add_revision(rev, self.get_transaction())
559
def _add_revision_text(self, revision_id, text):
560
revision = self._revision_store._serializer.read_revision_from_string(
562
self._revision_store._add_revision(revision, StringIO(text),
563
self.get_transaction())
565
def all_revision_ids(self):
566
"""Returns a list of all the revision ids in the repository.
568
This is deprecated because code should generally work on the graph
569
reachable from a particular revision, and ignore any other revisions
570
that might be present. There is no direct replacement method.
572
if 'evil' in debug.debug_flags:
573
mutter_callsite(2, "all_revision_ids is linear with history.")
574
return self._all_revision_ids()
576
def _all_revision_ids(self):
577
"""Returns a list of all the revision ids in the repository.
579
These are in as much topological order as the underlying store can
582
raise NotImplementedError(self._all_revision_ids)
584
def break_lock(self):
585
"""Break a lock if one is present from another instance.
587
Uses the ui factory to ask for confirmation if the lock may be from
590
self.control_files.break_lock()
593
def _eliminate_revisions_not_present(self, revision_ids):
594
"""Check every revision id in revision_ids to see if we have it.
596
Returns a set of the present revisions.
599
for id in revision_ids:
600
if self.has_revision(id):
605
def create(a_bzrdir):
606
"""Construct the current default format repository in a_bzrdir."""
607
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
609
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
610
"""instantiate a Repository.
612
:param _format: The format of the repository on disk.
613
:param a_bzrdir: The BzrDir of the repository.
615
In the future we will have a single api for all stores for
616
getting file texts, inventories and revisions, then
617
this construct will accept instances of those things.
619
super(Repository, self).__init__()
620
self._format = _format
621
# the following are part of the public API for Repository:
622
self.bzrdir = a_bzrdir
623
self.control_files = control_files
624
self._revision_store = _revision_store
625
# backwards compatibility
626
self.weave_store = text_store
628
self._reconcile_does_inventory_gc = True
629
self._reconcile_fixes_text_parents = False
630
self._reconcile_backsup_inventory = True
631
# not right yet - should be more semantically clear ?
633
self.control_store = control_store
634
self.control_weaves = control_store
635
# TODO: make sure to construct the right store classes, etc, depending
636
# on whether escaping is required.
637
self._warn_if_deprecated()
638
self._write_group = None
639
self.base = control_files._transport.base
642
return '%s(%r)' % (self.__class__.__name__,
645
def has_same_location(self, other):
646
"""Returns a boolean indicating if this repository is at the same
647
location as another repository.
649
This might return False even when two repository objects are accessing
650
the same physical repository via different URLs.
652
if self.__class__ is not other.__class__:
654
return (self.control_files._transport.base ==
655
other.control_files._transport.base)
657
def is_in_write_group(self):
658
"""Return True if there is an open write group.
660
:seealso: start_write_group.
662
return self._write_group is not None
665
return self.control_files.is_locked()
667
def is_write_locked(self):
668
"""Return True if this object is write locked."""
669
return self.is_locked() and self.control_files._lock_mode == 'w'
671
def lock_write(self, token=None):
672
"""Lock this repository for writing.
674
This causes caching within the repository obejct to start accumlating
675
data during reads, and allows a 'write_group' to be obtained. Write
676
groups must be used for actual data insertion.
678
:param token: if this is already locked, then lock_write will fail
679
unless the token matches the existing lock.
680
:returns: a token if this instance supports tokens, otherwise None.
681
:raises TokenLockingNotSupported: when a token is given but this
682
instance doesn't support using token locks.
683
:raises MismatchedToken: if the specified token doesn't match the token
684
of the existing lock.
685
:seealso: start_write_group.
687
A token should be passed in if you know that you have locked the object
688
some other way, and need to synchronise this object's state with that
691
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
693
result = self.control_files.lock_write(token=token)
698
self.control_files.lock_read()
701
def get_physical_lock_status(self):
702
return self.control_files.get_physical_lock_status()
704
def leave_lock_in_place(self):
705
"""Tell this repository not to release the physical lock when this
708
If lock_write doesn't return a token, then this method is not supported.
710
self.control_files.leave_in_place()
712
def dont_leave_lock_in_place(self):
713
"""Tell this repository to release the physical lock when this
714
object is unlocked, even if it didn't originally acquire it.
716
If lock_write doesn't return a token, then this method is not supported.
718
self.control_files.dont_leave_in_place()
721
def gather_stats(self, revid=None, committers=None):
722
"""Gather statistics from a revision id.
724
:param revid: The revision id to gather statistics from, if None, then
725
no revision specific statistics are gathered.
726
:param committers: Optional parameter controlling whether to grab
727
a count of committers from the revision specific statistics.
728
:return: A dictionary of statistics. Currently this contains:
729
committers: The number of committers if requested.
730
firstrev: A tuple with timestamp, timezone for the penultimate left
731
most ancestor of revid, if revid is not the NULL_REVISION.
732
latestrev: A tuple with timestamp, timezone for revid, if revid is
733
not the NULL_REVISION.
734
revisions: The total revision count in the repository.
735
size: An estimate disk size of the repository in bytes.
738
if revid and committers:
739
result['committers'] = 0
740
if revid and revid != _mod_revision.NULL_REVISION:
742
all_committers = set()
743
revisions = self.get_ancestry(revid)
744
# pop the leading None
746
first_revision = None
748
# ignore the revisions in the middle - just grab first and last
749
revisions = revisions[0], revisions[-1]
750
for revision in self.get_revisions(revisions):
751
if not first_revision:
752
first_revision = revision
754
all_committers.add(revision.committer)
755
last_revision = revision
757
result['committers'] = len(all_committers)
758
result['firstrev'] = (first_revision.timestamp,
759
first_revision.timezone)
760
result['latestrev'] = (last_revision.timestamp,
761
last_revision.timezone)
763
# now gather global repository information
764
if self.bzrdir.root_transport.listable():
765
c, t = self._revision_store.total_size(self.get_transaction())
766
result['revisions'] = c
770
def find_branches(self, using=False):
771
"""Find branches underneath this repository.
773
This will include branches inside other branches.
775
:param using: If True, list only branches using this repository.
777
if using and not self.is_shared():
779
return [self.bzrdir.open_branch()]
780
except errors.NotBranchError:
782
class Evaluator(object):
785
self.first_call = True
787
def __call__(self, bzrdir):
788
# On the first call, the parameter is always the bzrdir
789
# containing the current repo.
790
if not self.first_call:
792
repository = bzrdir.open_repository()
793
except errors.NoRepositoryPresent:
796
return False, (None, repository)
797
self.first_call = False
799
value = (bzrdir.open_branch(), None)
800
except errors.NotBranchError:
805
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
806
self.bzrdir.root_transport, evaluate=Evaluator()):
807
if branch is not None:
808
branches.append(branch)
809
if not using and repository is not None:
810
branches.extend(repository.find_branches())
813
def get_data_stream(self, revision_ids):
814
raise NotImplementedError(self.get_data_stream)
816
def get_data_stream_for_search(self, search_result):
817
"""Get a data stream that can be inserted to a repository.
819
:param search_result: A bzrlib.graph.SearchResult selecting the
821
:return: A data stream that can be inserted into a repository using
824
raise NotImplementedError(self.get_data_stream_for_search)
826
def insert_data_stream(self, stream):
827
"""XXX What does this really do?
829
Is it a substitute for fetch?
830
Should it manage its own write group ?
832
for item_key, bytes in stream:
833
if item_key[0] == 'file':
834
(file_id,) = item_key[1:]
835
knit = self.weave_store.get_weave_or_empty(
836
file_id, self.get_transaction())
837
elif item_key == ('inventory',):
838
knit = self.get_inventory_weave()
839
elif item_key == ('revisions',):
840
knit = self._revision_store.get_revision_file(
841
self.get_transaction())
842
elif item_key == ('signatures',):
843
knit = self._revision_store.get_signature_file(
844
self.get_transaction())
846
raise RepositoryDataStreamError(
847
"Unrecognised data stream key '%s'" % (item_key,))
848
decoded_list = bencode.bdecode(bytes)
849
format = decoded_list.pop(0)
852
for version, options, parents, some_bytes in decoded_list:
853
data_list.append((version, options, len(some_bytes), parents))
854
knit_bytes += some_bytes
855
buffer = StringIO(knit_bytes)
856
def reader_func(count):
860
return buffer.read(count)
861
knit.insert_data_stream(
862
(format, data_list, reader_func))
865
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
866
"""Return the revision ids that other has that this does not.
868
These are returned in topological order.
870
revision_id: only return revision ids included by revision_id.
872
return InterRepository.get(other, self).search_missing_revision_ids(
873
revision_id, find_ghosts)
875
@deprecated_method(symbol_versioning.one_two)
877
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
878
"""Return the revision ids that other has that this does not.
880
These are returned in topological order.
882
revision_id: only return revision ids included by revision_id.
884
keys = self.search_missing_revision_ids(
885
other, revision_id, find_ghosts).get_keys()
888
parents = other.get_graph().get_parent_map(keys)
891
return tsort.topo_sort(parents)
895
"""Open the repository rooted at base.
897
For instance, if the repository is at URL/.bzr/repository,
898
Repository.open(URL) -> a Repository instance.
900
control = bzrdir.BzrDir.open(base)
901
return control.open_repository()
903
def copy_content_into(self, destination, revision_id=None):
904
"""Make a complete copy of the content in self into destination.
906
This is a destructive operation! Do not use it on existing
909
return InterRepository.get(self, destination).copy_content(revision_id)
911
def commit_write_group(self):
912
"""Commit the contents accrued within the current write group.
914
:seealso: start_write_group.
916
if self._write_group is not self.get_transaction():
917
# has an unlock or relock occured ?
918
raise errors.BzrError('mismatched lock context %r and '
920
(self.get_transaction(), self._write_group))
921
self._commit_write_group()
922
self._write_group = None
924
def _commit_write_group(self):
925
"""Template method for per-repository write group cleanup.
927
This is called before the write group is considered to be
928
finished and should ensure that all data handed to the repository
929
for writing during the write group is safely committed (to the
930
extent possible considering file system caching etc).
933
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
934
"""Fetch the content required to construct revision_id from source.
936
If revision_id is None all content is copied.
937
:param find_ghosts: Find and copy revisions in the source that are
938
ghosts in the target (and not reachable directly by walking out to
939
the first-present revision in target from revision_id).
941
# fast path same-url fetch operations
942
if self.has_same_location(source):
943
# check that last_revision is in 'from' and then return a
945
if (revision_id is not None and
946
not _mod_revision.is_null(revision_id)):
947
self.get_revision(revision_id)
949
inter = InterRepository.get(source, self)
951
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
952
except NotImplementedError:
953
raise errors.IncompatibleRepositories(source, self)
955
def create_bundle(self, target, base, fileobj, format=None):
956
return serializer.write_bundle(self, target, base, fileobj, format)
958
def get_commit_builder(self, branch, parents, config, timestamp=None,
959
timezone=None, committer=None, revprops=None,
961
"""Obtain a CommitBuilder for this repository.
963
:param branch: Branch to commit to.
964
:param parents: Revision ids of the parents of the new revision.
965
:param config: Configuration to use.
966
:param timestamp: Optional timestamp recorded for commit.
967
:param timezone: Optional timezone for timestamp.
968
:param committer: Optional committer to set for commit.
969
:param revprops: Optional dictionary of revision properties.
970
:param revision_id: Optional revision id.
972
result = self._commit_builder_class(self, parents, config,
973
timestamp, timezone, committer, revprops, revision_id)
974
self.start_write_group()
978
if (self.control_files._lock_count == 1 and
979
self.control_files._lock_mode == 'w'):
980
if self._write_group is not None:
981
self.abort_write_group()
982
self.control_files.unlock()
983
raise errors.BzrError(
984
'Must end write groups before releasing write locks.')
985
self.control_files.unlock()
988
def clone(self, a_bzrdir, revision_id=None):
989
"""Clone this repository into a_bzrdir using the current format.
991
Currently no check is made that the format of this repository and
992
the bzrdir format are compatible. FIXME RBC 20060201.
994
:return: The newly created destination repository.
996
# TODO: deprecate after 0.16; cloning this with all its settings is
997
# probably not very useful -- mbp 20070423
998
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
999
self.copy_content_into(dest_repo, revision_id)
1002
def start_write_group(self):
1003
"""Start a write group in the repository.
1005
Write groups are used by repositories which do not have a 1:1 mapping
1006
between file ids and backend store to manage the insertion of data from
1007
both fetch and commit operations.
1009
A write lock is required around the start_write_group/commit_write_group
1010
for the support of lock-requiring repository formats.
1012
One can only insert data into a repository inside a write group.
1016
if not self.is_write_locked():
1017
raise errors.NotWriteLocked(self)
1018
if self._write_group:
1019
raise errors.BzrError('already in a write group')
1020
self._start_write_group()
1021
# so we can detect unlock/relock - the write group is now entered.
1022
self._write_group = self.get_transaction()
1024
def _start_write_group(self):
1025
"""Template method for per-repository write group startup.
1027
This is called before the write group is considered to be
1032
def sprout(self, to_bzrdir, revision_id=None):
1033
"""Create a descendent repository for new development.
1035
Unlike clone, this does not copy the settings of the repository.
1037
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1038
dest_repo.fetch(self, revision_id=revision_id)
1041
def _create_sprouting_repo(self, a_bzrdir, shared):
1042
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1043
# use target default format.
1044
dest_repo = a_bzrdir.create_repository()
1046
# Most control formats need the repository to be specifically
1047
# created, but on some old all-in-one formats it's not needed
1049
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1050
except errors.UninitializableFormat:
1051
dest_repo = a_bzrdir.open_repository()
1055
def has_revision(self, revision_id):
1056
"""True if this repository has a copy of the revision."""
1057
return revision_id in self.has_revisions((revision_id,))
1059
def has_revisions(self, revision_ids):
1060
"""Probe to find out the presence of multiple revisions.
1062
:param revision_ids: An iterable of revision_ids.
1063
:return: A set of the revision_ids that were present.
1065
raise NotImplementedError(self.has_revisions)
1067
return self._revision_store.has_revision_id(revision_id,
1068
self.get_transaction())
1071
def get_revision(self, revision_id):
1072
"""Return the Revision object for a named revision."""
1073
return self.get_revisions([revision_id])[0]
1076
def get_revision_reconcile(self, revision_id):
1077
"""'reconcile' helper routine that allows access to a revision always.
1079
This variant of get_revision does not cross check the weave graph
1080
against the revision one as get_revision does: but it should only
1081
be used by reconcile, or reconcile-alike commands that are correcting
1082
or testing the revision graph.
1084
return self._get_revisions([revision_id])[0]
1087
def get_revisions(self, revision_ids):
1088
"""Get many revisions at once."""
1089
return self._get_revisions(revision_ids)
1092
def _get_revisions(self, revision_ids):
1093
"""Core work logic to get many revisions without sanity checks."""
1094
for rev_id in revision_ids:
1095
if not rev_id or not isinstance(rev_id, basestring):
1096
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1097
revs = self._revision_store.get_revisions(revision_ids,
1098
self.get_transaction())
1100
assert not isinstance(rev.revision_id, unicode)
1101
for parent_id in rev.parent_ids:
1102
assert not isinstance(parent_id, unicode)
1106
def get_revision_xml(self, revision_id):
1107
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1108
# would have already do it.
1109
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1110
rev = self.get_revision(revision_id)
1111
rev_tmp = StringIO()
1112
# the current serializer..
1113
self._revision_store._serializer.write_revision(rev, rev_tmp)
1115
return rev_tmp.getvalue()
1118
def get_deltas_for_revisions(self, revisions):
1119
"""Produce a generator of revision deltas.
1121
Note that the input is a sequence of REVISIONS, not revision_ids.
1122
Trees will be held in memory until the generator exits.
1123
Each delta is relative to the revision's lefthand predecessor.
1125
required_trees = set()
1126
for revision in revisions:
1127
required_trees.add(revision.revision_id)
1128
required_trees.update(revision.parent_ids[:1])
1129
trees = dict((t.get_revision_id(), t) for
1130
t in self.revision_trees(required_trees))
1131
for revision in revisions:
1132
if not revision.parent_ids:
1133
old_tree = self.revision_tree(None)
1135
old_tree = trees[revision.parent_ids[0]]
1136
yield trees[revision.revision_id].changes_from(old_tree)
1139
def get_revision_delta(self, revision_id):
1140
"""Return the delta for one revision.
1142
The delta is relative to the left-hand predecessor of the
1145
r = self.get_revision(revision_id)
1146
return list(self.get_deltas_for_revisions([r]))[0]
1149
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1150
signature = gpg_strategy.sign(plaintext)
1151
self.add_signature_text(revision_id, signature)
1154
def add_signature_text(self, revision_id, signature):
1155
self._revision_store.add_revision_signature_text(revision_id,
1157
self.get_transaction())
1159
def find_text_key_references(self):
1160
"""Find the text key references within the repository.
1162
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1163
revision_ids. Each altered file-ids has the exact revision_ids that
1164
altered it listed explicitly.
1165
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1166
to whether they were referred to by the inventory of the
1167
revision_id that they contain. The inventory texts from all present
1168
revision ids are assessed to generate this report.
1170
revision_ids = self.all_revision_ids()
1171
w = self.get_inventory_weave()
1172
pb = ui.ui_factory.nested_progress_bar()
1174
return self._find_text_key_references_from_xml_inventory_lines(
1175
w.iter_lines_added_or_present_in_versions(revision_ids, pb=pb))
1179
def _find_text_key_references_from_xml_inventory_lines(self,
1181
"""Core routine for extracting references to texts from inventories.
1183
This performs the translation of xml lines to revision ids.
1185
:param line_iterator: An iterator of lines, origin_version_id
1186
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1187
to whether they were referred to by the inventory of the
1188
revision_id that they contain. Note that if that revision_id was
1189
not part of the line_iterator's output then False will be given -
1190
even though it may actually refer to that key.
1192
if not self._serializer.support_altered_by_hack:
1193
raise AssertionError(
1194
"_find_text_key_references_from_xml_inventory_lines only "
1195
"supported for branches which store inventory as unnested xml"
1196
", not on %r" % self)
1199
# this code needs to read every new line in every inventory for the
1200
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1201
# not present in one of those inventories is unnecessary but not
1202
# harmful because we are filtering by the revision id marker in the
1203
# inventory lines : we only select file ids altered in one of those
1204
# revisions. We don't need to see all lines in the inventory because
1205
# only those added in an inventory in rev X can contain a revision=X
1207
unescape_revid_cache = {}
1208
unescape_fileid_cache = {}
1210
# jam 20061218 In a big fetch, this handles hundreds of thousands
1211
# of lines, so it has had a lot of inlining and optimizing done.
1212
# Sorry that it is a little bit messy.
1213
# Move several functions to be local variables, since this is a long
1215
search = self._file_ids_altered_regex.search
1216
unescape = _unescape_xml
1217
setdefault = result.setdefault
1218
for line, version_id in line_iterator:
1219
match = search(line)
1222
# One call to match.group() returning multiple items is quite a
1223
# bit faster than 2 calls to match.group() each returning 1
1224
file_id, revision_id = match.group('file_id', 'revision_id')
1226
# Inlining the cache lookups helps a lot when you make 170,000
1227
# lines and 350k ids, versus 8.4 unique ids.
1228
# Using a cache helps in 2 ways:
1229
# 1) Avoids unnecessary decoding calls
1230
# 2) Re-uses cached strings, which helps in future set and
1232
# (2) is enough that removing encoding entirely along with
1233
# the cache (so we are using plain strings) results in no
1234
# performance improvement.
1236
revision_id = unescape_revid_cache[revision_id]
1238
unescaped = unescape(revision_id)
1239
unescape_revid_cache[revision_id] = unescaped
1240
revision_id = unescaped
1242
# Note that unconditionally unescaping means that we deserialise
1243
# every fileid, which for general 'pull' is not great, but we don't
1244
# really want to have some many fulltexts that this matters anyway.
1247
file_id = unescape_fileid_cache[file_id]
1249
unescaped = unescape(file_id)
1250
unescape_fileid_cache[file_id] = unescaped
1253
key = (file_id, revision_id)
1254
setdefault(key, False)
1255
if revision_id == version_id:
1259
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1261
"""Helper routine for fileids_altered_by_revision_ids.
1263
This performs the translation of xml lines to revision ids.
1265
:param line_iterator: An iterator of lines, origin_version_id
1266
:param revision_ids: The revision ids to filter for. This should be a
1267
set or other type which supports efficient __contains__ lookups, as
1268
the revision id from each parsed line will be looked up in the
1269
revision_ids filter.
1270
:return: a dictionary mapping altered file-ids to an iterable of
1271
revision_ids. Each altered file-ids has the exact revision_ids that
1272
altered it listed explicitly.
1275
setdefault = result.setdefault
1276
for file_id, revision_id in \
1277
self._find_text_key_references_from_xml_inventory_lines(
1278
line_iterator).iterkeys():
1279
# once data is all ensured-consistent; then this is
1280
# if revision_id == version_id
1281
if revision_id in revision_ids:
1282
setdefault(file_id, set()).add(revision_id)
1285
def fileids_altered_by_revision_ids(self, revision_ids):
1286
"""Find the file ids and versions affected by revisions.
1288
:param revisions: an iterable containing revision ids.
1289
:return: a dictionary mapping altered file-ids to an iterable of
1290
revision_ids. Each altered file-ids has the exact revision_ids that
1291
altered it listed explicitly.
1293
selected_revision_ids = set(revision_ids)
1294
w = self.get_inventory_weave()
1295
pb = ui.ui_factory.nested_progress_bar()
1297
return self._find_file_ids_from_xml_inventory_lines(
1298
w.iter_lines_added_or_present_in_versions(
1299
selected_revision_ids, pb=pb),
1300
selected_revision_ids)
1304
def iter_files_bytes(self, desired_files):
1305
"""Iterate through file versions.
1307
Files will not necessarily be returned in the order they occur in
1308
desired_files. No specific order is guaranteed.
1310
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1311
value supplied by the caller as part of desired_files. It should
1312
uniquely identify the file version in the caller's context. (Examples:
1313
an index number or a TreeTransform trans_id.)
1315
bytes_iterator is an iterable of bytestrings for the file. The
1316
kind of iterable and length of the bytestrings are unspecified, but for
1317
this implementation, it is a list of lines produced by
1318
VersionedFile.get_lines().
1320
:param desired_files: a list of (file_id, revision_id, identifier)
1323
transaction = self.get_transaction()
1324
for file_id, revision_id, callable_data in desired_files:
1326
weave = self.weave_store.get_weave(file_id, transaction)
1327
except errors.NoSuchFile:
1328
raise errors.NoSuchIdInRepository(self, file_id)
1329
yield callable_data, weave.get_lines(revision_id)
1331
def _generate_text_key_index(self, text_key_references=None,
1333
"""Generate a new text key index for the repository.
1335
This is an expensive function that will take considerable time to run.
1337
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1338
list of parents, also text keys. When a given key has no parents,
1339
the parents list will be [NULL_REVISION].
1341
# All revisions, to find inventory parents.
1342
if ancestors is None:
1343
revision_graph = self.get_revision_graph_with_ghosts()
1344
ancestors = revision_graph.get_ancestors()
1345
if text_key_references is None:
1346
text_key_references = self.find_text_key_references()
1347
pb = ui.ui_factory.nested_progress_bar()
1349
return self._do_generate_text_key_index(ancestors,
1350
text_key_references, pb)
1354
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1355
"""Helper for _generate_text_key_index to avoid deep nesting."""
1356
revision_order = tsort.topo_sort(ancestors)
1357
invalid_keys = set()
1359
for revision_id in revision_order:
1360
revision_keys[revision_id] = set()
1361
text_count = len(text_key_references)
1362
# a cache of the text keys to allow reuse; costs a dict of all the
1363
# keys, but saves a 2-tuple for every child of a given key.
1365
for text_key, valid in text_key_references.iteritems():
1367
invalid_keys.add(text_key)
1369
revision_keys[text_key[1]].add(text_key)
1370
text_key_cache[text_key] = text_key
1371
del text_key_references
1373
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1374
NULL_REVISION = _mod_revision.NULL_REVISION
1375
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1376
# too small for large or very branchy trees. However, for 55K path
1377
# trees, it would be easy to use too much memory trivially. Ideally we
1378
# could gauge this by looking at available real memory etc, but this is
1379
# always a tricky proposition.
1380
inventory_cache = lru_cache.LRUCache(10)
1381
batch_size = 10 # should be ~150MB on a 55K path tree
1382
batch_count = len(revision_order) / batch_size + 1
1384
pb.update("Calculating text parents.", processed_texts, text_count)
1385
for offset in xrange(batch_count):
1386
to_query = revision_order[offset * batch_size:(offset + 1) *
1390
for rev_tree in self.revision_trees(to_query):
1391
revision_id = rev_tree.get_revision_id()
1392
parent_ids = ancestors[revision_id]
1393
for text_key in revision_keys[revision_id]:
1394
pb.update("Calculating text parents.", processed_texts)
1395
processed_texts += 1
1396
candidate_parents = []
1397
for parent_id in parent_ids:
1398
parent_text_key = (text_key[0], parent_id)
1400
check_parent = parent_text_key not in \
1401
revision_keys[parent_id]
1403
# the parent parent_id is a ghost:
1404
check_parent = False
1405
# truncate the derived graph against this ghost.
1406
parent_text_key = None
1408
# look at the parent commit details inventories to
1409
# determine possible candidates in the per file graph.
1412
inv = inventory_cache[parent_id]
1414
inv = self.revision_tree(parent_id).inventory
1415
inventory_cache[parent_id] = inv
1416
parent_entry = inv._byid.get(text_key[0], None)
1417
if parent_entry is not None:
1419
text_key[0], parent_entry.revision)
1421
parent_text_key = None
1422
if parent_text_key is not None:
1423
candidate_parents.append(
1424
text_key_cache[parent_text_key])
1425
parent_heads = text_graph.heads(candidate_parents)
1426
new_parents = list(parent_heads)
1427
new_parents.sort(key=lambda x:candidate_parents.index(x))
1428
if new_parents == []:
1429
new_parents = [NULL_REVISION]
1430
text_index[text_key] = new_parents
1432
for text_key in invalid_keys:
1433
text_index[text_key] = [NULL_REVISION]
1436
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1437
"""Get an iterable listing the keys of all the data introduced by a set
1440
The keys will be ordered so that the corresponding items can be safely
1441
fetched and inserted in that order.
1443
:returns: An iterable producing tuples of (knit-kind, file-id,
1444
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1445
'revisions'. file-id is None unless knit-kind is 'file'.
1447
# XXX: it's a bit weird to control the inventory weave caching in this
1448
# generator. Ideally the caching would be done in fetch.py I think. Or
1449
# maybe this generator should explicitly have the contract that it
1450
# should not be iterated until the previously yielded item has been
1453
inv_w = self.get_inventory_weave()
1454
inv_w.enable_cache()
1456
# file ids that changed
1457
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1459
num_file_ids = len(file_ids)
1460
for file_id, altered_versions in file_ids.iteritems():
1461
if _files_pb is not None:
1462
_files_pb.update("fetch texts", count, num_file_ids)
1464
yield ("file", file_id, altered_versions)
1465
# We're done with the files_pb. Note that it finished by the caller,
1466
# just as it was created by the caller.
1470
yield ("inventory", None, revision_ids)
1474
revisions_with_signatures = set()
1475
for rev_id in revision_ids:
1477
self.get_signature_text(rev_id)
1478
except errors.NoSuchRevision:
1482
revisions_with_signatures.add(rev_id)
1484
yield ("signatures", None, revisions_with_signatures)
1487
yield ("revisions", None, revision_ids)
1490
def get_inventory_weave(self):
1491
return self.control_weaves.get_weave('inventory',
1492
self.get_transaction())
1495
def get_inventory(self, revision_id):
1496
"""Get Inventory object by revision id."""
1497
return self.iter_inventories([revision_id]).next()
1499
def iter_inventories(self, revision_ids):
1500
"""Get many inventories by revision_ids.
1502
This will buffer some or all of the texts used in constructing the
1503
inventories in memory, but will only parse a single inventory at a
1506
:return: An iterator of inventories.
1508
assert None not in revision_ids
1509
assert _mod_revision.NULL_REVISION not in revision_ids
1510
return self._iter_inventories(revision_ids)
1512
def _iter_inventories(self, revision_ids):
1513
"""single-document based inventory iteration."""
1514
texts = self.get_inventory_weave().get_texts(revision_ids)
1515
for text, revision_id in zip(texts, revision_ids):
1516
yield self.deserialise_inventory(revision_id, text)
1518
def deserialise_inventory(self, revision_id, xml):
1519
"""Transform the xml into an inventory object.
1521
:param revision_id: The expected revision id of the inventory.
1522
:param xml: A serialised inventory.
1524
result = self._serializer.read_inventory_from_string(xml, revision_id)
1525
if result.revision_id != revision_id:
1526
raise AssertionError('revision id mismatch %s != %s' % (
1527
result.revision_id, revision_id))
1530
def serialise_inventory(self, inv):
1531
return self._serializer.write_inventory_to_string(inv)
1533
def _serialise_inventory_to_lines(self, inv):
1534
return self._serializer.write_inventory_to_lines(inv)
1536
def get_serializer_format(self):
1537
return self._serializer.format_num
1540
def get_inventory_xml(self, revision_id):
1541
"""Get inventory XML as a file object."""
1543
assert isinstance(revision_id, str), type(revision_id)
1544
iw = self.get_inventory_weave()
1545
return iw.get_text(revision_id)
1547
raise errors.HistoryMissing(self, 'inventory', revision_id)
1550
def get_inventory_sha1(self, revision_id):
1551
"""Return the sha1 hash of the inventory entry
1553
return self.get_revision(revision_id).inventory_sha1
1556
def get_revision_graph(self, revision_id=None):
1557
"""Return a dictionary containing the revision graph.
1559
NB: This method should not be used as it accesses the entire graph all
1560
at once, which is much more data than most operations should require.
1562
:param revision_id: The revision_id to get a graph from. If None, then
1563
the entire revision graph is returned. This is a deprecated mode of
1564
operation and will be removed in the future.
1565
:return: a dictionary of revision_id->revision_parents_list.
1567
raise NotImplementedError(self.get_revision_graph)
1570
def get_revision_graph_with_ghosts(self, revision_ids=None):
1571
"""Return a graph of the revisions with ghosts marked as applicable.
1573
:param revision_ids: an iterable of revisions to graph or None for all.
1574
:return: a Graph object with the graph reachable from revision_ids.
1576
if 'evil' in debug.debug_flags:
1578
"get_revision_graph_with_ghosts scales with size of history.")
1579
result = deprecated_graph.Graph()
1580
if not revision_ids:
1581
pending = set(self.all_revision_ids())
1584
pending = set(revision_ids)
1585
# special case NULL_REVISION
1586
if _mod_revision.NULL_REVISION in pending:
1587
pending.remove(_mod_revision.NULL_REVISION)
1588
required = set(pending)
1591
revision_id = pending.pop()
1593
rev = self.get_revision(revision_id)
1594
except errors.NoSuchRevision:
1595
if revision_id in required:
1598
result.add_ghost(revision_id)
1600
for parent_id in rev.parent_ids:
1601
# is this queued or done ?
1602
if (parent_id not in pending and
1603
parent_id not in done):
1605
pending.add(parent_id)
1606
result.add_node(revision_id, rev.parent_ids)
1607
done.add(revision_id)
1610
def _get_history_vf(self):
1611
"""Get a versionedfile whose history graph reflects all revisions.
1613
For weave repositories, this is the inventory weave.
1615
return self.get_inventory_weave()
1617
def iter_reverse_revision_history(self, revision_id):
1618
"""Iterate backwards through revision ids in the lefthand history
1620
:param revision_id: The revision id to start with. All its lefthand
1621
ancestors will be traversed.
1623
if revision_id in (None, _mod_revision.NULL_REVISION):
1625
next_id = revision_id
1626
versionedfile = self._get_history_vf()
1629
parents = versionedfile.get_parents(next_id)
1630
if len(parents) == 0:
1633
next_id = parents[0]
1636
def get_revision_inventory(self, revision_id):
1637
"""Return inventory of a past revision."""
1638
# TODO: Unify this with get_inventory()
1639
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1640
# must be the same as its revision, so this is trivial.
1641
if revision_id is None:
1642
# This does not make sense: if there is no revision,
1643
# then it is the current tree inventory surely ?!
1644
# and thus get_root_id() is something that looks at the last
1645
# commit on the branch, and the get_root_id is an inventory check.
1646
raise NotImplementedError
1647
# return Inventory(self.get_root_id())
1649
return self.get_inventory(revision_id)
1652
def is_shared(self):
1653
"""Return True if this repository is flagged as a shared repository."""
1654
raise NotImplementedError(self.is_shared)
1657
def reconcile(self, other=None, thorough=False):
1658
"""Reconcile this repository."""
1659
from bzrlib.reconcile import RepoReconciler
1660
reconciler = RepoReconciler(self, thorough=thorough)
1661
reconciler.reconcile()
1664
def _refresh_data(self):
1665
"""Helper called from lock_* to ensure coherency with disk.
1667
The default implementation does nothing; it is however possible
1668
for repositories to maintain loaded indices across multiple locks
1669
by checking inside their implementation of this method to see
1670
whether their indices are still valid. This depends of course on
1671
the disk format being validatable in this manner.
1675
def revision_tree(self, revision_id):
1676
"""Return Tree for a revision on this branch.
1678
`revision_id` may be None for the empty tree revision.
1680
# TODO: refactor this to use an existing revision object
1681
# so we don't need to read it in twice.
1682
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1683
return RevisionTree(self, Inventory(root_id=None),
1684
_mod_revision.NULL_REVISION)
1686
inv = self.get_revision_inventory(revision_id)
1687
return RevisionTree(self, inv, revision_id)
1690
def revision_trees(self, revision_ids):
1691
"""Return Tree for a revision on this branch.
1693
`revision_id` may not be None or 'null:'"""
1694
inventories = self.iter_inventories(revision_ids)
1695
for inv in inventories:
1696
yield RevisionTree(self, inv, inv.revision_id)
1699
def get_ancestry(self, revision_id, topo_sorted=True):
1700
"""Return a list of revision-ids integrated by a revision.
1702
The first element of the list is always None, indicating the origin
1703
revision. This might change when we have history horizons, or
1704
perhaps we should have a new API.
1706
This is topologically sorted.
1708
if _mod_revision.is_null(revision_id):
1710
if not self.has_revision(revision_id):
1711
raise errors.NoSuchRevision(self, revision_id)
1712
w = self.get_inventory_weave()
1713
candidates = w.get_ancestry(revision_id, topo_sorted)
1714
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1717
"""Compress the data within the repository.
1719
This operation only makes sense for some repository types. For other
1720
types it should be a no-op that just returns.
1722
This stub method does not require a lock, but subclasses should use
1723
@needs_write_lock as this is a long running call its reasonable to
1724
implicitly lock for the user.
1728
def print_file(self, file, revision_id):
1729
"""Print `file` to stdout.
1731
FIXME RBC 20060125 as John Meinel points out this is a bad api
1732
- it writes to stdout, it assumes that that is valid etc. Fix
1733
by creating a new more flexible convenience function.
1735
tree = self.revision_tree(revision_id)
1736
# use inventory as it was in that revision
1737
file_id = tree.inventory.path2id(file)
1739
# TODO: jam 20060427 Write a test for this code path
1740
# it had a bug in it, and was raising the wrong
1742
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1743
tree.print_file(file_id)
1745
def get_transaction(self):
1746
return self.control_files.get_transaction()
1748
def revision_parents(self, revision_id):
1749
return self.get_inventory_weave().parent_names(revision_id)
1751
@deprecated_method(symbol_versioning.one_one)
1752
def get_parents(self, revision_ids):
1753
"""See StackedParentsProvider.get_parents"""
1754
parent_map = self.get_parent_map(revision_ids)
1755
return [parent_map.get(r, None) for r in revision_ids]
1757
def get_parent_map(self, keys):
1758
"""See graph._StackedParentsProvider.get_parent_map"""
1760
for revision_id in keys:
1761
if revision_id == _mod_revision.NULL_REVISION:
1762
parent_map[revision_id] = ()
1765
parent_id_list = self.get_revision(revision_id).parent_ids
1766
except errors.NoSuchRevision:
1769
if len(parent_id_list) == 0:
1770
parent_ids = (_mod_revision.NULL_REVISION,)
1772
parent_ids = tuple(parent_id_list)
1773
parent_map[revision_id] = parent_ids
1776
def _make_parents_provider(self):
1779
def get_graph(self, other_repository=None):
1780
"""Return the graph walker for this repository format"""
1781
parents_provider = self._make_parents_provider()
1782
if (other_repository is not None and
1783
other_repository.bzrdir.transport.base !=
1784
self.bzrdir.transport.base):
1785
parents_provider = graph._StackedParentsProvider(
1786
[parents_provider, other_repository._make_parents_provider()])
1787
return graph.Graph(parents_provider)
1789
def _get_versioned_file_checker(self):
1790
"""Return an object suitable for checking versioned files."""
1791
return _VersionedFileChecker(self)
1793
def revision_ids_to_search_result(self, result_set):
1794
"""Convert a set of revision ids to a graph SearchResult."""
1795
result_parents = set()
1796
for parents in self.get_graph().get_parent_map(
1797
result_set).itervalues():
1798
result_parents.update(parents)
1799
included_keys = result_set.intersection(result_parents)
1800
start_keys = result_set.difference(included_keys)
1801
exclude_keys = result_parents.difference(result_set)
1802
result = graph.SearchResult(start_keys, exclude_keys,
1803
len(result_set), result_set)
1807
def set_make_working_trees(self, new_value):
1808
"""Set the policy flag for making working trees when creating branches.
1810
This only applies to branches that use this repository.
1812
The default is 'True'.
1813
:param new_value: True to restore the default, False to disable making
1816
raise NotImplementedError(self.set_make_working_trees)
1818
def make_working_trees(self):
1819
"""Returns the policy for making working trees on new branches."""
1820
raise NotImplementedError(self.make_working_trees)
1823
def sign_revision(self, revision_id, gpg_strategy):
1824
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1825
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1828
def has_signature_for_revision_id(self, revision_id):
1829
"""Query for a revision signature for revision_id in the repository."""
1830
return self._revision_store.has_signature(revision_id,
1831
self.get_transaction())
1834
def get_signature_text(self, revision_id):
1835
"""Return the text for a signature."""
1836
return self._revision_store.get_signature_text(revision_id,
1837
self.get_transaction())
1840
def check(self, revision_ids=None):
1841
"""Check consistency of all history of given revision_ids.
1843
Different repository implementations should override _check().
1845
:param revision_ids: A non-empty list of revision_ids whose ancestry
1846
will be checked. Typically the last revision_id of a branch.
1848
return self._check(revision_ids)
1850
def _check(self, revision_ids):
1851
result = check.Check(self)
1855
def _warn_if_deprecated(self):
1856
global _deprecation_warning_done
1857
if _deprecation_warning_done:
1859
_deprecation_warning_done = True
1860
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1861
% (self._format, self.bzrdir.transport.base))
1863
def supports_rich_root(self):
1864
return self._format.rich_root_data
1866
def _check_ascii_revisionid(self, revision_id, method):
1867
"""Private helper for ascii-only repositories."""
1868
# weave repositories refuse to store revisionids that are non-ascii.
1869
if revision_id is not None:
1870
# weaves require ascii revision ids.
1871
if isinstance(revision_id, unicode):
1873
revision_id.encode('ascii')
1874
except UnicodeEncodeError:
1875
raise errors.NonAsciiRevisionId(method, self)
1878
revision_id.decode('ascii')
1879
except UnicodeDecodeError:
1880
raise errors.NonAsciiRevisionId(method, self)
1882
def revision_graph_can_have_wrong_parents(self):
1883
"""Is it possible for this repository to have a revision graph with
1886
If True, then this repository must also implement
1887
_find_inconsistent_revision_parents so that check and reconcile can
1888
check for inconsistencies before proceeding with other checks that may
1889
depend on the revision index being consistent.
1891
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1894
# remove these delegates a while after bzr 0.15
1895
def __make_delegated(name, from_module):
1896
def _deprecated_repository_forwarder():
1897
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1898
% (name, from_module),
1901
m = __import__(from_module, globals(), locals(), [name])
1903
return getattr(m, name)
1904
except AttributeError:
1905
raise AttributeError('module %s has no name %s'
1907
globals()[name] = _deprecated_repository_forwarder
1910
'AllInOneRepository',
1911
'WeaveMetaDirRepository',
1912
'PreSplitOutRepositoryFormat',
1913
'RepositoryFormat4',
1914
'RepositoryFormat5',
1915
'RepositoryFormat6',
1916
'RepositoryFormat7',
1918
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1922
'RepositoryFormatKnit',
1923
'RepositoryFormatKnit1',
1925
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1928
def install_revision(repository, rev, revision_tree):
1929
"""Install all revision data into a repository."""
1930
install_revisions(repository, [(rev, revision_tree, None)])
1933
def install_revisions(repository, iterable, num_revisions=None, pb=None):
1934
"""Install all revision data into a repository.
1936
Accepts an iterable of revision, tree, signature tuples. The signature
1939
repository.start_write_group()
1941
for n, (revision, revision_tree, signature) in enumerate(iterable):
1942
_install_revision(repository, revision, revision_tree, signature)
1944
pb.update('Transferring revisions', n + 1, num_revisions)
1946
repository.abort_write_group()
1949
repository.commit_write_group()
1952
def _install_revision(repository, rev, revision_tree, signature):
1953
"""Install all revision data into a repository."""
1954
present_parents = []
1956
for p_id in rev.parent_ids:
1957
if repository.has_revision(p_id):
1958
present_parents.append(p_id)
1959
parent_trees[p_id] = repository.revision_tree(p_id)
1961
parent_trees[p_id] = repository.revision_tree(None)
1963
inv = revision_tree.inventory
1964
entries = inv.iter_entries()
1965
# backwards compatibility hack: skip the root id.
1966
if not repository.supports_rich_root():
1967
path, root = entries.next()
1968
if root.revision != rev.revision_id:
1969
raise errors.IncompatibleRevision(repr(repository))
1970
# Add the texts that are not already present
1971
for path, ie in entries:
1972
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1973
repository.get_transaction())
1974
if ie.revision not in w:
1976
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1977
# with InventoryEntry.find_previous_heads(). if it is, then there
1978
# is a latent bug here where the parents may have ancestors of each
1980
for revision, tree in parent_trees.iteritems():
1981
if ie.file_id not in tree:
1983
parent_id = tree.inventory[ie.file_id].revision
1984
if parent_id in text_parents:
1986
text_parents.append(parent_id)
1988
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1989
repository.get_transaction())
1990
lines = revision_tree.get_file(ie.file_id).readlines()
1991
vfile.add_lines(rev.revision_id, text_parents, lines)
1993
# install the inventory
1994
repository.add_inventory(rev.revision_id, inv, present_parents)
1995
except errors.RevisionAlreadyPresent:
1997
if signature is not None:
1998
repository.add_signature_text(rev.revision_id, signature)
1999
repository.add_revision(rev.revision_id, rev, inv)
2002
class MetaDirRepository(Repository):
2003
"""Repositories in the new meta-dir layout."""
2005
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
2006
super(MetaDirRepository, self).__init__(_format,
2012
dir_mode = self.control_files._dir_mode
2013
file_mode = self.control_files._file_mode
2016
def is_shared(self):
2017
"""Return True if this repository is flagged as a shared repository."""
2018
return self.control_files._transport.has('shared-storage')
2021
def set_make_working_trees(self, new_value):
2022
"""Set the policy flag for making working trees when creating branches.
2024
This only applies to branches that use this repository.
2026
The default is 'True'.
2027
:param new_value: True to restore the default, False to disable making
2032
self.control_files._transport.delete('no-working-trees')
2033
except errors.NoSuchFile:
2036
self.control_files.put_utf8('no-working-trees', '')
2038
def make_working_trees(self):
2039
"""Returns the policy for making working trees on new branches."""
2040
return not self.control_files._transport.has('no-working-trees')
2043
class RepositoryFormatRegistry(registry.Registry):
2044
"""Registry of RepositoryFormats."""
2046
def get(self, format_string):
2047
r = registry.Registry.get(self, format_string)
2053
format_registry = RepositoryFormatRegistry()
2054
"""Registry of formats, indexed by their identifying format string.
2056
This can contain either format instances themselves, or classes/factories that
2057
can be called to obtain one.
2061
#####################################################################
2062
# Repository Formats
2064
class RepositoryFormat(object):
2065
"""A repository format.
2067
Formats provide three things:
2068
* An initialization routine to construct repository data on disk.
2069
* a format string which is used when the BzrDir supports versioned
2071
* an open routine which returns a Repository instance.
2073
There is one and only one Format subclass for each on-disk format. But
2074
there can be one Repository subclass that is used for several different
2075
formats. The _format attribute on a Repository instance can be used to
2076
determine the disk format.
2078
Formats are placed in an dict by their format string for reference
2079
during opening. These should be subclasses of RepositoryFormat
2082
Once a format is deprecated, just deprecate the initialize and open
2083
methods on the format class. Do not deprecate the object, as the
2084
object will be created every system load.
2086
Common instance attributes:
2087
_matchingbzrdir - the bzrdir format that the repository format was
2088
originally written to work with. This can be used if manually
2089
constructing a bzrdir and repository, or more commonly for test suite
2093
# Set to True or False in derived classes. True indicates that the format
2094
# supports ghosts gracefully.
2095
supports_ghosts = None
2098
return "<%s>" % self.__class__.__name__
2100
def __eq__(self, other):
2101
# format objects are generally stateless
2102
return isinstance(other, self.__class__)
2104
def __ne__(self, other):
2105
return not self == other
2108
def find_format(klass, a_bzrdir):
2109
"""Return the format for the repository object in a_bzrdir.
2111
This is used by bzr native formats that have a "format" file in
2112
the repository. Other methods may be used by different types of
2116
transport = a_bzrdir.get_repository_transport(None)
2117
format_string = transport.get("format").read()
2118
return format_registry.get(format_string)
2119
except errors.NoSuchFile:
2120
raise errors.NoRepositoryPresent(a_bzrdir)
2122
raise errors.UnknownFormatError(format=format_string)
2125
def register_format(klass, format):
2126
format_registry.register(format.get_format_string(), format)
2129
def unregister_format(klass, format):
2130
format_registry.remove(format.get_format_string())
2133
def get_default_format(klass):
2134
"""Return the current default format."""
2135
from bzrlib import bzrdir
2136
return bzrdir.format_registry.make_bzrdir('default').repository_format
2138
def _get_control_store(self, repo_transport, control_files):
2139
"""Return the control store for this repository."""
2140
raise NotImplementedError(self._get_control_store)
2142
def get_format_string(self):
2143
"""Return the ASCII format string that identifies this format.
2145
Note that in pre format ?? repositories the format string is
2146
not permitted nor written to disk.
2148
raise NotImplementedError(self.get_format_string)
2150
def get_format_description(self):
2151
"""Return the short description for this format."""
2152
raise NotImplementedError(self.get_format_description)
2154
def _get_revision_store(self, repo_transport, control_files):
2155
"""Return the revision store object for this a_bzrdir."""
2156
raise NotImplementedError(self._get_revision_store)
2158
def _get_text_rev_store(self,
2165
"""Common logic for getting a revision store for a repository.
2167
see self._get_revision_store for the subclass-overridable method to
2168
get the store for a repository.
2170
from bzrlib.store.revision.text import TextRevisionStore
2171
dir_mode = control_files._dir_mode
2172
file_mode = control_files._file_mode
2173
text_store = TextStore(transport.clone(name),
2175
compressed=compressed,
2177
file_mode=file_mode)
2178
_revision_store = TextRevisionStore(text_store, serializer)
2179
return _revision_store
2181
# TODO: this shouldn't be in the base class, it's specific to things that
2182
# use weaves or knits -- mbp 20070207
2183
def _get_versioned_file_store(self,
2188
versionedfile_class=None,
2189
versionedfile_kwargs={},
2191
if versionedfile_class is None:
2192
versionedfile_class = self._versionedfile_class
2193
weave_transport = control_files._transport.clone(name)
2194
dir_mode = control_files._dir_mode
2195
file_mode = control_files._file_mode
2196
return VersionedFileStore(weave_transport, prefixed=prefixed,
2198
file_mode=file_mode,
2199
versionedfile_class=versionedfile_class,
2200
versionedfile_kwargs=versionedfile_kwargs,
2203
def initialize(self, a_bzrdir, shared=False):
2204
"""Initialize a repository of this format in a_bzrdir.
2206
:param a_bzrdir: The bzrdir to put the new repository in it.
2207
:param shared: The repository should be initialized as a sharable one.
2208
:returns: The new repository object.
2210
This may raise UninitializableFormat if shared repository are not
2211
compatible the a_bzrdir.
2213
raise NotImplementedError(self.initialize)
2215
def is_supported(self):
2216
"""Is this format supported?
2218
Supported formats must be initializable and openable.
2219
Unsupported formats may not support initialization or committing or
2220
some other features depending on the reason for not being supported.
2224
def check_conversion_target(self, target_format):
2225
raise NotImplementedError(self.check_conversion_target)
2227
def open(self, a_bzrdir, _found=False):
2228
"""Return an instance of this format for the bzrdir a_bzrdir.
2230
_found is a private parameter, do not use it.
2232
raise NotImplementedError(self.open)
2235
class MetaDirRepositoryFormat(RepositoryFormat):
2236
"""Common base class for the new repositories using the metadir layout."""
2238
rich_root_data = False
2239
supports_tree_reference = False
2240
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
2243
super(MetaDirRepositoryFormat, self).__init__()
2245
def _create_control_files(self, a_bzrdir):
2246
"""Create the required files and the initial control_files object."""
2247
# FIXME: RBC 20060125 don't peek under the covers
2248
# NB: no need to escape relative paths that are url safe.
2249
repository_transport = a_bzrdir.get_repository_transport(self)
2250
control_files = lockable_files.LockableFiles(repository_transport,
2251
'lock', lockdir.LockDir)
2252
control_files.create_lock()
2253
return control_files
2255
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2256
"""Upload the initial blank content."""
2257
control_files = self._create_control_files(a_bzrdir)
2258
control_files.lock_write()
2260
control_files._transport.mkdir_multi(dirs,
2261
mode=control_files._dir_mode)
2262
for file, content in files:
2263
control_files.put(file, content)
2264
for file, content in utf8_files:
2265
control_files.put_utf8(file, content)
2267
control_files.put_utf8('shared-storage', '')
2269
control_files.unlock()
2272
# formats which have no format string are not discoverable
2273
# and not independently creatable, so are not registered. They're
2274
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2275
# needed, it's constructed directly by the BzrDir. Non-native formats where
2276
# the repository is not separately opened are similar.
2278
format_registry.register_lazy(
2279
'Bazaar-NG Repository format 7',
2280
'bzrlib.repofmt.weaverepo',
2284
format_registry.register_lazy(
2285
'Bazaar-NG Knit Repository Format 1',
2286
'bzrlib.repofmt.knitrepo',
2287
'RepositoryFormatKnit1',
2290
format_registry.register_lazy(
2291
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2292
'bzrlib.repofmt.knitrepo',
2293
'RepositoryFormatKnit3',
2296
format_registry.register_lazy(
2297
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2298
'bzrlib.repofmt.knitrepo',
2299
'RepositoryFormatKnit4',
2302
# Pack-based formats. There is one format for pre-subtrees, and one for
2303
# post-subtrees to allow ease of testing.
2304
# NOTE: These are experimental in 0.92.
2305
format_registry.register_lazy(
2306
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2307
'bzrlib.repofmt.pack_repo',
2308
'RepositoryFormatKnitPack1',
2310
format_registry.register_lazy(
2311
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2312
'bzrlib.repofmt.pack_repo',
2313
'RepositoryFormatKnitPack3',
2315
format_registry.register_lazy(
2316
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2317
'bzrlib.repofmt.pack_repo',
2318
'RepositoryFormatKnitPack4',
2322
class InterRepository(InterObject):
2323
"""This class represents operations taking place between two repositories.
2325
Its instances have methods like copy_content and fetch, and contain
2326
references to the source and target repositories these operations can be
2329
Often we will provide convenience methods on 'repository' which carry out
2330
operations with another repository - they will always forward to
2331
InterRepository.get(other).method_name(parameters).
2335
"""The available optimised InterRepository types."""
2337
def copy_content(self, revision_id=None):
2338
raise NotImplementedError(self.copy_content)
2340
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2341
"""Fetch the content required to construct revision_id.
2343
The content is copied from self.source to self.target.
2345
:param revision_id: if None all content is copied, if NULL_REVISION no
2347
:param pb: optional progress bar to use for progress reports. If not
2348
provided a default one will be created.
2350
Returns the copied revision count and the failed revisions in a tuple:
2353
raise NotImplementedError(self.fetch)
2355
def _walk_to_common_revisions(self, revision_ids):
2356
"""Walk out from revision_ids in source to revisions target has.
2358
:param revision_ids: The start point for the search.
2359
:return: A set of revision ids.
2361
graph = self.source.get_graph()
2362
missing_revs = set()
2363
# ensure we don't pay silly lookup costs.
2364
revision_ids = frozenset(revision_ids)
2365
searcher = graph._make_breadth_first_searcher(revision_ids)
2366
null_set = frozenset([_mod_revision.NULL_REVISION])
2369
next_revs, ghosts = searcher.next_with_ghosts()
2370
except StopIteration:
2372
if revision_ids.intersection(ghosts):
2373
absent_ids = set(revision_ids.intersection(ghosts))
2374
# If all absent_ids are present in target, no error is needed.
2375
absent_ids.difference_update(
2376
self.target.has_revisions(absent_ids))
2378
raise errors.NoSuchRevision(self.source, absent_ids.pop())
2379
# we don't care about other ghosts as we can't fetch them and
2380
# haven't been asked to.
2381
next_revs = set(next_revs)
2382
# we always have NULL_REVISION present.
2383
have_revs = self.target.has_revisions(next_revs).union(null_set)
2384
missing_revs.update(next_revs - have_revs)
2385
searcher.stop_searching_any(have_revs)
2386
return searcher.get_result()
2388
@deprecated_method(symbol_versioning.one_two)
2390
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2391
"""Return the revision ids that source has that target does not.
2393
These are returned in topological order.
2395
:param revision_id: only return revision ids included by this
2397
:param find_ghosts: If True find missing revisions in deep history
2398
rather than just finding the surface difference.
2400
return list(self.search_missing_revision_ids(
2401
revision_id, find_ghosts).get_keys())
2404
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2405
"""Return the revision ids that source has that target does not.
2407
:param revision_id: only return revision ids included by this
2409
:param find_ghosts: If True find missing revisions in deep history
2410
rather than just finding the surface difference.
2411
:return: A bzrlib.graph.SearchResult.
2413
# stop searching at found target revisions.
2414
if not find_ghosts and revision_id is not None:
2415
return self._walk_to_common_revisions([revision_id])
2416
# generic, possibly worst case, slow code path.
2417
target_ids = set(self.target.all_revision_ids())
2418
if revision_id is not None:
2419
source_ids = self.source.get_ancestry(revision_id)
2420
assert source_ids[0] is None
2423
source_ids = self.source.all_revision_ids()
2424
result_set = set(source_ids).difference(target_ids)
2425
return self.source.revision_ids_to_search_result(result_set)
2428
def _same_model(source, target):
2429
"""True if source and target have the same data representation."""
2430
if source.supports_rich_root() != target.supports_rich_root():
2432
if source._serializer != target._serializer:
2437
class InterSameDataRepository(InterRepository):
2438
"""Code for converting between repositories that represent the same data.
2440
Data format and model must match for this to work.
2444
def _get_repo_format_to_test(self):
2445
"""Repository format for testing with.
2447
InterSameData can pull from subtree to subtree and from non-subtree to
2448
non-subtree, so we test this with the richest repository format.
2450
from bzrlib.repofmt import knitrepo
2451
return knitrepo.RepositoryFormatKnit3()
2454
def is_compatible(source, target):
2455
return InterRepository._same_model(source, target)
2458
def copy_content(self, revision_id=None):
2459
"""Make a complete copy of the content in self into destination.
2461
This copies both the repository's revision data, and configuration information
2462
such as the make_working_trees setting.
2464
This is a destructive operation! Do not use it on existing
2467
:param revision_id: Only copy the content needed to construct
2468
revision_id and its parents.
2471
self.target.set_make_working_trees(self.source.make_working_trees())
2472
except NotImplementedError:
2474
# but don't bother fetching if we have the needed data now.
2475
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2476
self.target.has_revision(revision_id)):
2478
self.target.fetch(self.source, revision_id=revision_id)
2481
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2482
"""See InterRepository.fetch()."""
2483
from bzrlib.fetch import GenericRepoFetcher
2484
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2485
self.source, self.source._format, self.target,
2486
self.target._format)
2487
f = GenericRepoFetcher(to_repository=self.target,
2488
from_repository=self.source,
2489
last_revision=revision_id,
2490
pb=pb, find_ghosts=find_ghosts)
2491
return f.count_copied, f.failed_revisions
2494
class InterWeaveRepo(InterSameDataRepository):
2495
"""Optimised code paths between Weave based repositories.
2497
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2498
implemented lazy inter-object optimisation.
2502
def _get_repo_format_to_test(self):
2503
from bzrlib.repofmt import weaverepo
2504
return weaverepo.RepositoryFormat7()
2507
def is_compatible(source, target):
2508
"""Be compatible with known Weave formats.
2510
We don't test for the stores being of specific types because that
2511
could lead to confusing results, and there is no need to be
2514
from bzrlib.repofmt.weaverepo import (
2520
return (isinstance(source._format, (RepositoryFormat5,
2522
RepositoryFormat7)) and
2523
isinstance(target._format, (RepositoryFormat5,
2525
RepositoryFormat7)))
2526
except AttributeError:
2530
def copy_content(self, revision_id=None):
2531
"""See InterRepository.copy_content()."""
2532
# weave specific optimised path:
2534
self.target.set_make_working_trees(self.source.make_working_trees())
2535
except NotImplementedError:
2537
# FIXME do not peek!
2538
if self.source.control_files._transport.listable():
2539
pb = ui.ui_factory.nested_progress_bar()
2541
self.target.weave_store.copy_all_ids(
2542
self.source.weave_store,
2544
from_transaction=self.source.get_transaction(),
2545
to_transaction=self.target.get_transaction())
2546
pb.update('copying inventory', 0, 1)
2547
self.target.control_weaves.copy_multi(
2548
self.source.control_weaves, ['inventory'],
2549
from_transaction=self.source.get_transaction(),
2550
to_transaction=self.target.get_transaction())
2551
self.target._revision_store.text_store.copy_all_ids(
2552
self.source._revision_store.text_store,
2557
self.target.fetch(self.source, revision_id=revision_id)
2560
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2561
"""See InterRepository.fetch()."""
2562
from bzrlib.fetch import GenericRepoFetcher
2563
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2564
self.source, self.source._format, self.target, self.target._format)
2565
f = GenericRepoFetcher(to_repository=self.target,
2566
from_repository=self.source,
2567
last_revision=revision_id,
2568
pb=pb, find_ghosts=find_ghosts)
2569
return f.count_copied, f.failed_revisions
2572
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2573
"""See InterRepository.missing_revision_ids()."""
2574
# we want all revisions to satisfy revision_id in source.
2575
# but we don't want to stat every file here and there.
2576
# we want then, all revisions other needs to satisfy revision_id
2577
# checked, but not those that we have locally.
2578
# so the first thing is to get a subset of the revisions to
2579
# satisfy revision_id in source, and then eliminate those that
2580
# we do already have.
2581
# this is slow on high latency connection to self, but as as this
2582
# disk format scales terribly for push anyway due to rewriting
2583
# inventory.weave, this is considered acceptable.
2585
if revision_id is not None:
2586
source_ids = self.source.get_ancestry(revision_id)
2587
assert source_ids[0] is None
2590
source_ids = self.source._all_possible_ids()
2591
source_ids_set = set(source_ids)
2592
# source_ids is the worst possible case we may need to pull.
2593
# now we want to filter source_ids against what we actually
2594
# have in target, but don't try to check for existence where we know
2595
# we do not have a revision as that would be pointless.
2596
target_ids = set(self.target._all_possible_ids())
2597
possibly_present_revisions = target_ids.intersection(source_ids_set)
2598
actually_present_revisions = set(
2599
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2600
required_revisions = source_ids_set.difference(actually_present_revisions)
2601
if revision_id is not None:
2602
# we used get_ancestry to determine source_ids then we are assured all
2603
# revisions referenced are present as they are installed in topological order.
2604
# and the tip revision was validated by get_ancestry.
2605
result_set = required_revisions
2607
# if we just grabbed the possibly available ids, then
2608
# we only have an estimate of whats available and need to validate
2609
# that against the revision records.
2611
self.source._eliminate_revisions_not_present(required_revisions))
2612
return self.source.revision_ids_to_search_result(result_set)
2615
class InterKnitRepo(InterSameDataRepository):
2616
"""Optimised code paths between Knit based repositories."""
2619
def _get_repo_format_to_test(self):
2620
from bzrlib.repofmt import knitrepo
2621
return knitrepo.RepositoryFormatKnit1()
2624
def is_compatible(source, target):
2625
"""Be compatible with known Knit formats.
2627
We don't test for the stores being of specific types because that
2628
could lead to confusing results, and there is no need to be
2631
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2633
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2634
isinstance(target._format, RepositoryFormatKnit))
2635
except AttributeError:
2637
return are_knits and InterRepository._same_model(source, target)
2640
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2641
"""See InterRepository.fetch()."""
2642
from bzrlib.fetch import KnitRepoFetcher
2643
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2644
self.source, self.source._format, self.target, self.target._format)
2645
f = KnitRepoFetcher(to_repository=self.target,
2646
from_repository=self.source,
2647
last_revision=revision_id,
2648
pb=pb, find_ghosts=find_ghosts)
2649
return f.count_copied, f.failed_revisions
2652
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2653
"""See InterRepository.missing_revision_ids()."""
2654
if revision_id is not None:
2655
source_ids = self.source.get_ancestry(revision_id)
2656
assert source_ids[0] is None
2659
source_ids = self.source.all_revision_ids()
2660
source_ids_set = set(source_ids)
2661
# source_ids is the worst possible case we may need to pull.
2662
# now we want to filter source_ids against what we actually
2663
# have in target, but don't try to check for existence where we know
2664
# we do not have a revision as that would be pointless.
2665
target_ids = set(self.target.all_revision_ids())
2666
possibly_present_revisions = target_ids.intersection(source_ids_set)
2667
actually_present_revisions = set(
2668
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2669
required_revisions = source_ids_set.difference(actually_present_revisions)
2670
if revision_id is not None:
2671
# we used get_ancestry to determine source_ids then we are assured all
2672
# revisions referenced are present as they are installed in topological order.
2673
# and the tip revision was validated by get_ancestry.
2674
result_set = required_revisions
2676
# if we just grabbed the possibly available ids, then
2677
# we only have an estimate of whats available and need to validate
2678
# that against the revision records.
2680
self.source._eliminate_revisions_not_present(required_revisions))
2681
return self.source.revision_ids_to_search_result(result_set)
2684
class InterPackRepo(InterSameDataRepository):
2685
"""Optimised code paths between Pack based repositories."""
2688
def _get_repo_format_to_test(self):
2689
from bzrlib.repofmt import pack_repo
2690
return pack_repo.RepositoryFormatKnitPack1()
2693
def is_compatible(source, target):
2694
"""Be compatible with known Pack formats.
2696
We don't test for the stores being of specific types because that
2697
could lead to confusing results, and there is no need to be
2700
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2702
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2703
isinstance(target._format, RepositoryFormatPack))
2704
except AttributeError:
2706
return are_packs and InterRepository._same_model(source, target)
2709
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2710
"""See InterRepository.fetch()."""
2711
from bzrlib.repofmt.pack_repo import Packer
2712
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2713
self.source, self.source._format, self.target, self.target._format)
2714
self.count_copied = 0
2715
if revision_id is None:
2717
# everything to do - use pack logic
2718
# to fetch from all packs to one without
2719
# inventory parsing etc, IFF nothing to be copied is in the target.
2721
revision_ids = self.source.all_revision_ids()
2722
# implementing the TODO will involve:
2723
# - detecting when all of a pack is selected
2724
# - avoiding as much as possible pre-selection, so the
2725
# more-core routines such as create_pack_from_packs can filter in
2726
# a just-in-time fashion. (though having a HEADS list on a
2727
# repository might make this a lot easier, because we could
2728
# sensibly detect 'new revisions' without doing a full index scan.
2729
elif _mod_revision.is_null(revision_id):
2734
revision_ids = self.search_missing_revision_ids(revision_id,
2735
find_ghosts=find_ghosts).get_keys()
2736
except errors.NoSuchRevision:
2737
raise errors.InstallFailed([revision_id])
2738
packs = self.source._pack_collection.all_packs()
2739
pack = Packer(self.target._pack_collection, packs, '.fetch',
2740
revision_ids).pack()
2741
if pack is not None:
2742
self.target._pack_collection._save_pack_names()
2743
# Trigger an autopack. This may duplicate effort as we've just done
2744
# a pack creation, but for now it is simpler to think about as
2745
# 'upload data, then repack if needed'.
2746
self.target._pack_collection.autopack()
2747
return (pack.get_revision_count(), [])
2752
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2753
"""See InterRepository.missing_revision_ids().
2755
:param find_ghosts: Find ghosts throughout the ancestry of
2758
if not find_ghosts and revision_id is not None:
2759
return self._walk_to_common_revisions([revision_id])
2760
elif revision_id is not None:
2761
source_ids = self.source.get_ancestry(revision_id)
2762
assert source_ids[0] is None
2765
source_ids = self.source.all_revision_ids()
2766
# source_ids is the worst possible case we may need to pull.
2767
# now we want to filter source_ids against what we actually
2768
# have in target, but don't try to check for existence where we know
2769
# we do not have a revision as that would be pointless.
2770
target_ids = set(self.target.all_revision_ids())
2771
result_set = set(source_ids).difference(target_ids)
2772
return self.source.revision_ids_to_search_result(result_set)
2775
class InterModel1and2(InterRepository):
2778
def _get_repo_format_to_test(self):
2782
def is_compatible(source, target):
2783
if not source.supports_rich_root() and target.supports_rich_root():
2789
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2790
"""See InterRepository.fetch()."""
2791
from bzrlib.fetch import Model1toKnit2Fetcher
2792
f = Model1toKnit2Fetcher(to_repository=self.target,
2793
from_repository=self.source,
2794
last_revision=revision_id,
2795
pb=pb, find_ghosts=find_ghosts)
2796
return f.count_copied, f.failed_revisions
2799
def copy_content(self, revision_id=None):
2800
"""Make a complete copy of the content in self into destination.
2802
This is a destructive operation! Do not use it on existing
2805
:param revision_id: Only copy the content needed to construct
2806
revision_id and its parents.
2809
self.target.set_make_working_trees(self.source.make_working_trees())
2810
except NotImplementedError:
2812
# but don't bother fetching if we have the needed data now.
2813
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2814
self.target.has_revision(revision_id)):
2816
self.target.fetch(self.source, revision_id=revision_id)
2819
class InterKnit1and2(InterKnitRepo):
2822
def _get_repo_format_to_test(self):
2826
def is_compatible(source, target):
2827
"""Be compatible with Knit1 source and Knit3 target"""
2828
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2830
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2831
RepositoryFormatKnit3)
2832
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2833
RepositoryFormatKnitPack3)
2834
return (isinstance(source._format,
2835
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2836
isinstance(target._format,
2837
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2839
except AttributeError:
2843
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2844
"""See InterRepository.fetch()."""
2845
from bzrlib.fetch import Knit1to2Fetcher
2846
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2847
self.source, self.source._format, self.target,
2848
self.target._format)
2849
f = Knit1to2Fetcher(to_repository=self.target,
2850
from_repository=self.source,
2851
last_revision=revision_id,
2852
pb=pb, find_ghosts=find_ghosts)
2853
return f.count_copied, f.failed_revisions
2856
class InterDifferingSerializer(InterKnitRepo):
2859
def _get_repo_format_to_test(self):
2863
def is_compatible(source, target):
2864
"""Be compatible with Knit2 source and Knit3 target"""
2865
if source.supports_rich_root() != target.supports_rich_root():
2867
# Ideally, we'd support fetching if the source had no tree references
2868
# even if it supported them...
2869
if (getattr(source, '_format.supports_tree_reference', False) and
2870
not getattr(target, '_format.supports_tree_reference', False)):
2875
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2876
"""See InterRepository.fetch()."""
2877
revision_ids = self.target.search_missing_revision_ids(self.source,
2878
revision_id, find_ghosts=find_ghosts).get_keys()
2879
revision_ids = tsort.topo_sort(
2880
self.source.get_graph().get_parent_map(revision_ids))
2881
def revisions_iterator():
2882
for current_revision_id in revision_ids:
2883
revision = self.source.get_revision(current_revision_id)
2884
tree = self.source.revision_tree(current_revision_id)
2886
signature = self.source.get_signature_text(
2887
current_revision_id)
2888
except errors.NoSuchRevision:
2890
yield revision, tree, signature
2892
my_pb = ui.ui_factory.nested_progress_bar()
2897
install_revisions(self.target, revisions_iterator(),
2898
len(revision_ids), pb)
2900
if my_pb is not None:
2902
return len(revision_ids), 0
2905
class InterRemoteToOther(InterRepository):
2907
def __init__(self, source, target):
2908
InterRepository.__init__(self, source, target)
2909
self._real_inter = None
2912
def is_compatible(source, target):
2913
if not isinstance(source, remote.RemoteRepository):
2915
# Is source's model compatible with target's model?
2916
source._ensure_real()
2917
real_source = source._real_repository
2918
assert not isinstance(real_source, remote.RemoteRepository), (
2919
"We don't support remote repos backed by remote repos yet.")
2920
return InterRepository._same_model(real_source, target)
2923
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2924
"""See InterRepository.fetch()."""
2925
from bzrlib.fetch import RemoteToOtherFetcher
2926
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2927
self.source, self.target, self.target._format)
2928
# TODO: jam 20070210 This should be an assert, not a translate
2929
revision_id = osutils.safe_revision_id(revision_id)
2930
f = RemoteToOtherFetcher(to_repository=self.target,
2931
from_repository=self.source,
2932
last_revision=revision_id,
2933
pb=pb, find_ghosts=find_ghosts)
2934
return f.count_copied, f.failed_revisions
2937
def _get_repo_format_to_test(self):
2941
class InterOtherToRemote(InterRepository):
2943
def __init__(self, source, target):
2944
InterRepository.__init__(self, source, target)
2945
self._real_inter = None
2948
def is_compatible(source, target):
2949
if isinstance(target, remote.RemoteRepository):
2953
def _ensure_real_inter(self):
2954
if self._real_inter is None:
2955
self.target._ensure_real()
2956
real_target = self.target._real_repository
2957
self._real_inter = InterRepository.get(self.source, real_target)
2959
def copy_content(self, revision_id=None):
2960
self._ensure_real_inter()
2961
self._real_inter.copy_content(revision_id=revision_id)
2963
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2964
self._ensure_real_inter()
2965
self._real_inter.fetch(revision_id=revision_id, pb=pb,
2966
find_ghosts=find_ghosts)
2969
def _get_repo_format_to_test(self):
2973
InterRepository.register_optimiser(InterDifferingSerializer)
2974
InterRepository.register_optimiser(InterSameDataRepository)
2975
InterRepository.register_optimiser(InterWeaveRepo)
2976
InterRepository.register_optimiser(InterKnitRepo)
2977
InterRepository.register_optimiser(InterModel1and2)
2978
InterRepository.register_optimiser(InterKnit1and2)
2979
InterRepository.register_optimiser(InterPackRepo)
2980
InterRepository.register_optimiser(InterRemoteToOther)
2981
InterRepository.register_optimiser(InterOtherToRemote)
2984
class CopyConverter(object):
2985
"""A repository conversion tool which just performs a copy of the content.
2987
This is slow but quite reliable.
2990
def __init__(self, target_format):
2991
"""Create a CopyConverter.
2993
:param target_format: The format the resulting repository should be.
2995
self.target_format = target_format
2997
def convert(self, repo, pb):
2998
"""Perform the conversion of to_convert, giving feedback via pb.
3000
:param to_convert: The disk object to convert.
3001
:param pb: a progress bar to use for progress information.
3006
# this is only useful with metadir layouts - separated repo content.
3007
# trigger an assertion if not such
3008
repo._format.get_format_string()
3009
self.repo_dir = repo.bzrdir
3010
self.step('Moving repository to repository.backup')
3011
self.repo_dir.transport.move('repository', 'repository.backup')
3012
backup_transport = self.repo_dir.transport.clone('repository.backup')
3013
repo._format.check_conversion_target(self.target_format)
3014
self.source_repo = repo._format.open(self.repo_dir,
3016
_override_transport=backup_transport)
3017
self.step('Creating new repository')
3018
converted = self.target_format.initialize(self.repo_dir,
3019
self.source_repo.is_shared())
3020
converted.lock_write()
3022
self.step('Copying content into repository.')
3023
self.source_repo.copy_content_into(converted)
3026
self.step('Deleting old repository content.')
3027
self.repo_dir.transport.delete_tree('repository.backup')
3028
self.pb.note('repository converted')
3030
def step(self, message):
3031
"""Update the pb by a step."""
3033
self.pb.update(message, self.count, self.total)
3045
def _unescaper(match, _map=_unescape_map):
3046
code = match.group(1)
3050
if not code.startswith('#'):
3052
return unichr(int(code[1:])).encode('utf8')
3058
def _unescape_xml(data):
3059
"""Unescape predefined XML entities in a string of data."""
3061
if _unescape_re is None:
3062
_unescape_re = re.compile('\&([^;]*);')
3063
return _unescape_re.sub(_unescaper, data)
3066
class _VersionedFileChecker(object):
3068
def __init__(self, repository):
3069
self.repository = repository
3070
self.text_index = self.repository._generate_text_key_index()
3072
def calculate_file_version_parents(self, revision_id, file_id):
3073
"""Calculate the correct parents for a file version according to
3076
parent_keys = self.text_index[(file_id, revision_id)]
3077
if parent_keys == [_mod_revision.NULL_REVISION]:
3079
# strip the file_id, for the weave api
3080
return tuple([revision_id for file_id, revision_id in parent_keys])
3082
def check_file_version_parents(self, weave, file_id):
3083
"""Check the parents stored in a versioned file are correct.
3085
It also detects file versions that are not referenced by their
3086
corresponding revision's inventory.
3088
:returns: A tuple of (wrong_parents, dangling_file_versions).
3089
wrong_parents is a dict mapping {revision_id: (stored_parents,
3090
correct_parents)} for each revision_id where the stored parents
3091
are not correct. dangling_file_versions is a set of (file_id,
3092
revision_id) tuples for versions that are present in this versioned
3093
file, but not used by the corresponding inventory.
3096
unused_versions = set()
3097
for num, revision_id in enumerate(weave.versions()):
3099
correct_parents = self.calculate_file_version_parents(
3100
revision_id, file_id)
3102
# The version is not part of the used keys.
3103
unused_versions.add(revision_id)
3106
knit_parents = tuple(weave.get_parents(revision_id))
3107
except errors.RevisionNotPresent:
3109
if correct_parents != knit_parents:
3110
wrong_parents[revision_id] = (knit_parents, correct_parents)
3111
return wrong_parents, unused_versions
added
removed
bzrlib/repository.py
