1
# Copyright (C) 2005, 2006, 2007, 2008 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 bzrlib.lazy_import import lazy_import
18
lazy_import(globals(), """
38
revision as _mod_revision,
44
from bzrlib.bundle import serializer
45
from bzrlib.revisiontree import RevisionTree
46
from bzrlib.store.versioned import VersionedFileStore
47
from bzrlib.testament import Testament
50
from bzrlib import registry
51
from bzrlib.decorators import needs_read_lock, needs_write_lock
52
from bzrlib.inter import InterObject
53
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
54
from bzrlib.symbol_versioning import (
60
from bzrlib.trace import (
61
log_exception_quietly, note, mutter, mutter_callsite, warning)
64
# Old formats display a warning, but only once
65
_deprecation_warning_done = False
68
class CommitBuilder(object):
69
"""Provides an interface to build up a commit.
71
This allows describing a tree to be committed without needing to
72
know the internals of the format of the repository.
75
# all clients should supply tree roots.
76
record_root_entry = True
77
# the default CommitBuilder does not manage trees whose root is versioned.
78
_versioned_root = False
80
def __init__(self, repository, parents, config, timestamp=None,
81
timezone=None, committer=None, revprops=None,
83
"""Initiate a CommitBuilder.
85
:param repository: Repository to commit to.
86
:param parents: Revision ids of the parents of the new revision.
87
:param config: Configuration to use.
88
:param timestamp: Optional timestamp recorded for commit.
89
:param timezone: Optional timezone for timestamp.
90
:param committer: Optional committer to set for commit.
91
:param revprops: Optional dictionary of revision properties.
92
:param revision_id: Optional revision id.
97
self._committer = self._config.username()
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._validate_revprops(revprops)
109
self._revprops.update(revprops)
111
if timestamp is None:
112
timestamp = time.time()
113
# Restrict resolution to 1ms
114
self._timestamp = round(timestamp, 3)
117
self._timezone = osutils.local_time_offset()
119
self._timezone = int(timezone)
121
self._generate_revision_if_needed()
122
self.__heads = graph.HeadsCache(repository.get_graph()).heads
123
self._basis_delta = []
124
# API compatibility, older code that used CommitBuilder did not call
125
# .record_delete(), which means the delta that is computed would not be
126
# valid. Callers that will call record_delete() should call
127
# .will_record_deletes() to indicate that.
128
self._recording_deletes = False
130
def _validate_unicode_text(self, text, context):
131
"""Verify things like commit messages don't have bogus characters."""
133
raise ValueError('Invalid value for %s: %r' % (context, text))
135
def _validate_revprops(self, revprops):
136
for key, value in revprops.iteritems():
137
# We know that the XML serializers do not round trip '\r'
138
# correctly, so refuse to accept them
139
if not isinstance(value, basestring):
140
raise ValueError('revision property (%s) is not a valid'
141
' (unicode) string: %r' % (key, value))
142
self._validate_unicode_text(value,
143
'revision property (%s)' % (key,))
145
def commit(self, message):
146
"""Make the actual commit.
148
:return: The revision id of the recorded revision.
150
self._validate_unicode_text(message, 'commit message')
151
rev = _mod_revision.Revision(
152
timestamp=self._timestamp,
153
timezone=self._timezone,
154
committer=self._committer,
156
inventory_sha1=self.inv_sha1,
157
revision_id=self._new_revision_id,
158
properties=self._revprops)
159
rev.parent_ids = self.parents
160
self.repository.add_revision(self._new_revision_id, rev,
161
self.new_inventory, self._config)
162
self.repository.commit_write_group()
163
return self._new_revision_id
166
"""Abort the commit that is being built.
168
self.repository.abort_write_group()
170
def revision_tree(self):
171
"""Return the tree that was just committed.
173
After calling commit() this can be called to get a RevisionTree
174
representing the newly committed tree. This is preferred to
175
calling Repository.revision_tree() because that may require
176
deserializing the inventory, while we already have a copy in
179
return RevisionTree(self.repository, self.new_inventory,
180
self._new_revision_id)
182
def finish_inventory(self):
183
"""Tell the builder that the inventory is finished."""
184
if self.new_inventory.root is None:
185
raise AssertionError('Root entry should be supplied to'
186
' record_entry_contents, as of bzr 0.10.')
187
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
188
self.new_inventory.revision_id = self._new_revision_id
189
self.inv_sha1 = self.repository.add_inventory(
190
self._new_revision_id,
195
def _gen_revision_id(self):
196
"""Return new revision-id."""
197
return generate_ids.gen_revision_id(self._config.username(),
200
def _generate_revision_if_needed(self):
201
"""Create a revision id if None was supplied.
203
If the repository can not support user-specified revision ids
204
they should override this function and raise CannotSetRevisionId
205
if _new_revision_id is not None.
207
:raises: CannotSetRevisionId
209
if self._new_revision_id is None:
210
self._new_revision_id = self._gen_revision_id()
211
self.random_revid = True
213
self.random_revid = False
215
def _heads(self, file_id, revision_ids):
216
"""Calculate the graph heads for revision_ids in the graph of file_id.
218
This can use either a per-file graph or a global revision graph as we
219
have an identity relationship between the two graphs.
221
return self.__heads(revision_ids)
223
def _check_root(self, ie, parent_invs, tree):
224
"""Helper for record_entry_contents.
226
:param ie: An entry being added.
227
:param parent_invs: The inventories of the parent revisions of the
229
:param tree: The tree that is being committed.
231
# In this revision format, root entries have no knit or weave When
232
# serializing out to disk and back in root.revision is always
234
ie.revision = self._new_revision_id
236
def _get_delta(self, ie, basis_inv, path):
237
"""Get a delta against the basis inventory for ie."""
238
if ie.file_id not in basis_inv:
240
result = (None, path, ie.file_id, ie)
241
self._basis_delta.append(result)
243
elif ie != basis_inv[ie.file_id]:
245
# TODO: avoid tis id2path call.
246
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
247
self._basis_delta.append(result)
253
def get_basis_delta(self):
254
"""Return the complete inventory delta versus the basis inventory.
256
This has been built up with the calls to record_delete and
257
record_entry_contents. The client must have already called
258
will_record_deletes() to indicate that they will be generating a
261
:return: An inventory delta, suitable for use with apply_delta, or
262
Repository.add_inventory_by_delta, etc.
264
if not self._recording_deletes:
265
raise AssertionError("recording deletes not activated.")
266
return self._basis_delta
268
def record_delete(self, path, file_id):
269
"""Record that a delete occured against a basis tree.
271
This is an optional API - when used it adds items to the basis_delta
272
being accumulated by the commit builder. It cannot be called unless the
273
method will_record_deletes() has been called to inform the builder that
274
a delta is being supplied.
276
:param path: The path of the thing deleted.
277
:param file_id: The file id that was deleted.
279
if not self._recording_deletes:
280
raise AssertionError("recording deletes not activated.")
281
delta = (path, None, file_id, None)
282
self._basis_delta.append(delta)
285
def will_record_deletes(self):
286
"""Tell the commit builder that deletes are being notified.
288
This enables the accumulation of an inventory delta; for the resulting
289
commit to be valid, deletes against the basis MUST be recorded via
290
builder.record_delete().
292
self._recording_deletes = True
294
def record_entry_contents(self, ie, parent_invs, path, tree,
296
"""Record the content of ie from tree into the commit if needed.
298
Side effect: sets ie.revision when unchanged
300
:param ie: An inventory entry present in the commit.
301
:param parent_invs: The inventories of the parent revisions of the
303
:param path: The path the entry is at in the tree.
304
:param tree: The tree which contains this entry and should be used to
306
:param content_summary: Summary data from the tree about the paths
307
content - stat, length, exec, sha/link target. This is only
308
accessed when the entry has a revision of None - that is when it is
309
a candidate to commit.
310
:return: A tuple (change_delta, version_recorded, fs_hash).
311
change_delta is an inventory_delta change for this entry against
312
the basis tree of the commit, or None if no change occured against
314
version_recorded is True if a new version of the entry has been
315
recorded. For instance, committing a merge where a file was only
316
changed on the other side will return (delta, False).
317
fs_hash is either None, or the hash details for the path (currently
318
a tuple of the contents sha1 and the statvalue returned by
319
tree.get_file_with_stat()).
321
if self.new_inventory.root is None:
322
if ie.parent_id is not None:
323
raise errors.RootMissing()
324
self._check_root(ie, parent_invs, tree)
325
if ie.revision is None:
326
kind = content_summary[0]
328
# ie is carried over from a prior commit
330
# XXX: repository specific check for nested tree support goes here - if
331
# the repo doesn't want nested trees we skip it ?
332
if (kind == 'tree-reference' and
333
not self.repository._format.supports_tree_reference):
334
# mismatch between commit builder logic and repository:
335
# this needs the entry creation pushed down into the builder.
336
raise NotImplementedError('Missing repository subtree support.')
337
self.new_inventory.add(ie)
339
# TODO: slow, take it out of the inner loop.
341
basis_inv = parent_invs[0]
343
basis_inv = Inventory(root_id=None)
345
# ie.revision is always None if the InventoryEntry is considered
346
# for committing. We may record the previous parents revision if the
347
# content is actually unchanged against a sole head.
348
if ie.revision is not None:
349
if not self._versioned_root and path == '':
350
# repositories that do not version the root set the root's
351
# revision to the new commit even when no change occurs (more
352
# specifically, they do not record a revision on the root; and
353
# the rev id is assigned to the root during deserialisation -
354
# this masks when a change may have occurred against the basis.
355
# To match this we always issue a delta, because the revision
356
# of the root will always be changing.
357
if ie.file_id in basis_inv:
358
delta = (basis_inv.id2path(ie.file_id), path,
362
delta = (None, path, ie.file_id, ie)
363
self._basis_delta.append(delta)
364
return delta, False, None
366
# we don't need to commit this, because the caller already
367
# determined that an existing revision of this file is
368
# appropriate. If its not being considered for committing then
369
# it and all its parents to the root must be unaltered so
370
# no-change against the basis.
371
if ie.revision == self._new_revision_id:
372
raise AssertionError("Impossible situation, a skipped "
373
"inventory entry (%r) claims to be modified in this "
374
"commit (%r).", (ie, self._new_revision_id))
375
return None, False, None
376
# XXX: Friction: parent_candidates should return a list not a dict
377
# so that we don't have to walk the inventories again.
378
parent_candiate_entries = ie.parent_candidates(parent_invs)
379
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
381
for inv in parent_invs:
382
if ie.file_id in inv:
383
old_rev = inv[ie.file_id].revision
384
if old_rev in head_set:
385
heads.append(inv[ie.file_id].revision)
386
head_set.remove(inv[ie.file_id].revision)
389
# now we check to see if we need to write a new record to the
391
# We write a new entry unless there is one head to the ancestors, and
392
# the kind-derived content is unchanged.
394
# Cheapest check first: no ancestors, or more the one head in the
395
# ancestors, we write a new node.
399
# There is a single head, look it up for comparison
400
parent_entry = parent_candiate_entries[heads[0]]
401
# if the non-content specific data has changed, we'll be writing a
403
if (parent_entry.parent_id != ie.parent_id or
404
parent_entry.name != ie.name):
406
# now we need to do content specific checks:
408
# if the kind changed the content obviously has
409
if kind != parent_entry.kind:
411
# Stat cache fingerprint feedback for the caller - None as we usually
412
# don't generate one.
415
if content_summary[2] is None:
416
raise ValueError("Files must not have executable = None")
418
if (# if the file length changed we have to store:
419
parent_entry.text_size != content_summary[1] or
420
# if the exec bit has changed we have to store:
421
parent_entry.executable != content_summary[2]):
423
elif parent_entry.text_sha1 == content_summary[3]:
424
# all meta and content is unchanged (using a hash cache
425
# hit to check the sha)
426
ie.revision = parent_entry.revision
427
ie.text_size = parent_entry.text_size
428
ie.text_sha1 = parent_entry.text_sha1
429
ie.executable = parent_entry.executable
430
return self._get_delta(ie, basis_inv, path), False, None
432
# Either there is only a hash change(no hash cache entry,
433
# or same size content change), or there is no change on
435
# Provide the parent's hash to the store layer, so that the
436
# content is unchanged we will not store a new node.
437
nostore_sha = parent_entry.text_sha1
439
# We want to record a new node regardless of the presence or
440
# absence of a content change in the file.
442
ie.executable = content_summary[2]
443
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
445
lines = file_obj.readlines()
449
ie.text_sha1, ie.text_size = self._add_text_to_weave(
450
ie.file_id, lines, heads, nostore_sha)
451
# Let the caller know we generated a stat fingerprint.
452
fingerprint = (ie.text_sha1, stat_value)
453
except errors.ExistingContent:
454
# Turns out that the file content was unchanged, and we were
455
# only going to store a new node if it was changed. Carry over
457
ie.revision = parent_entry.revision
458
ie.text_size = parent_entry.text_size
459
ie.text_sha1 = parent_entry.text_sha1
460
ie.executable = parent_entry.executable
461
return self._get_delta(ie, basis_inv, path), False, None
462
elif kind == 'directory':
464
# all data is meta here, nothing specific to directory, so
466
ie.revision = parent_entry.revision
467
return self._get_delta(ie, basis_inv, path), False, None
469
self._add_text_to_weave(ie.file_id, lines, heads, None)
470
elif kind == 'symlink':
471
current_link_target = content_summary[3]
473
# symlink target is not generic metadata, check if it has
475
if current_link_target != parent_entry.symlink_target:
478
# unchanged, carry over.
479
ie.revision = parent_entry.revision
480
ie.symlink_target = parent_entry.symlink_target
481
return self._get_delta(ie, basis_inv, path), False, None
482
ie.symlink_target = current_link_target
484
self._add_text_to_weave(ie.file_id, lines, heads, None)
485
elif kind == 'tree-reference':
487
if content_summary[3] != parent_entry.reference_revision:
490
# unchanged, carry over.
491
ie.reference_revision = parent_entry.reference_revision
492
ie.revision = parent_entry.revision
493
return self._get_delta(ie, basis_inv, path), False, None
494
ie.reference_revision = content_summary[3]
496
self._add_text_to_weave(ie.file_id, lines, heads, None)
498
raise NotImplementedError('unknown kind')
499
ie.revision = self._new_revision_id
500
return self._get_delta(ie, basis_inv, path), True, fingerprint
502
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
503
# Note: as we read the content directly from the tree, we know its not
504
# been turned into unicode or badly split - but a broken tree
505
# implementation could give us bad output from readlines() so this is
506
# not a guarantee of safety. What would be better is always checking
507
# the content during test suite execution. RBC 20070912
508
parent_keys = tuple((file_id, parent) for parent in parents)
509
return self.repository.texts.add_lines(
510
(file_id, self._new_revision_id), parent_keys, new_lines,
511
nostore_sha=nostore_sha, random_id=self.random_revid,
512
check_content=False)[0:2]
515
class RootCommitBuilder(CommitBuilder):
516
"""This commitbuilder actually records the root id"""
518
# the root entry gets versioned properly by this builder.
519
_versioned_root = True
521
def _check_root(self, ie, parent_invs, tree):
522
"""Helper for record_entry_contents.
524
:param ie: An entry being added.
525
:param parent_invs: The inventories of the parent revisions of the
527
:param tree: The tree that is being committed.
531
######################################################################
534
class Repository(object):
535
"""Repository holding history for one or more branches.
537
The repository holds and retrieves historical information including
538
revisions and file history. It's normally accessed only by the Branch,
539
which views a particular line of development through that history.
541
The Repository builds on top of some byte storage facilies (the revisions,
542
signatures, inventories and texts attributes) and a Transport, which
543
respectively provide byte storage and a means to access the (possibly
546
The byte storage facilities are addressed via tuples, which we refer to
547
as 'keys' throughout the code base. Revision_keys, inventory_keys and
548
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
549
(file_id, revision_id). We use this interface because it allows low
550
friction with the underlying code that implements disk indices, network
551
encoding and other parts of bzrlib.
553
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
554
the serialised revisions for the repository. This can be used to obtain
555
revision graph information or to access raw serialised revisions.
556
The result of trying to insert data into the repository via this store
557
is undefined: it should be considered read-only except for implementors
559
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
560
the serialised signatures for the repository. This can be used to
561
obtain access to raw serialised signatures. The result of trying to
562
insert data into the repository via this store is undefined: it should
563
be considered read-only except for implementors of repositories.
564
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
565
the serialised inventories for the repository. This can be used to
566
obtain unserialised inventories. The result of trying to insert data
567
into the repository via this store is undefined: it should be
568
considered read-only except for implementors of repositories.
569
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
570
texts of files and directories for the repository. This can be used to
571
obtain file texts or file graphs. Note that Repository.iter_file_bytes
572
is usually a better interface for accessing file texts.
573
The result of trying to insert data into the repository via this store
574
is undefined: it should be considered read-only except for implementors
576
:ivar _transport: Transport for file access to repository, typically
577
pointing to .bzr/repository.
580
# What class to use for a CommitBuilder. Often its simpler to change this
581
# in a Repository class subclass rather than to override
582
# get_commit_builder.
583
_commit_builder_class = CommitBuilder
584
# The search regex used by xml based repositories to determine what things
585
# where changed in a single commit.
586
_file_ids_altered_regex = lazy_regex.lazy_compile(
587
r'file_id="(?P<file_id>[^"]+)"'
588
r'.* revision="(?P<revision_id>[^"]+)"'
591
def abort_write_group(self, suppress_errors=False):
592
"""Commit the contents accrued within the current write group.
594
:param suppress_errors: if true, abort_write_group will catch and log
595
unexpected errors that happen during the abort, rather than
596
allowing them to propagate. Defaults to False.
598
:seealso: start_write_group.
600
if self._write_group is not self.get_transaction():
601
# has an unlock or relock occured ?
602
raise errors.BzrError('mismatched lock context and write group.')
604
self._abort_write_group()
605
except Exception, exc:
606
self._write_group = None
607
if not suppress_errors:
609
mutter('abort_write_group failed')
610
log_exception_quietly()
611
note('bzr: ERROR (ignored): %s', exc)
612
self._write_group = None
614
def _abort_write_group(self):
615
"""Template method for per-repository write group cleanup.
617
This is called during abort before the write group is considered to be
618
finished and should cleanup any internal state accrued during the write
619
group. There is no requirement that data handed to the repository be
620
*not* made available - this is not a rollback - but neither should any
621
attempt be made to ensure that data added is fully commited. Abort is
622
invoked when an error has occured so futher disk or network operations
623
may not be possible or may error and if possible should not be
627
def add_fallback_repository(self, repository):
628
"""Add a repository to use for looking up data not held locally.
630
:param repository: A repository.
632
if not self._format.supports_external_lookups:
633
raise errors.UnstackableRepositoryFormat(self._format, self.base)
634
self._check_fallback_repository(repository)
635
self._fallback_repositories.append(repository)
636
self.texts.add_fallback_versioned_files(repository.texts)
637
self.inventories.add_fallback_versioned_files(repository.inventories)
638
self.revisions.add_fallback_versioned_files(repository.revisions)
639
self.signatures.add_fallback_versioned_files(repository.signatures)
640
self._fetch_order = 'topological'
642
def _check_fallback_repository(self, repository):
643
"""Check that this repository can fallback to repository safely.
645
Raise an error if not.
647
:param repository: A repository to fallback to.
649
return InterRepository._assert_same_model(self, repository)
651
def add_inventory(self, revision_id, inv, parents):
652
"""Add the inventory inv to the repository as revision_id.
654
:param parents: The revision ids of the parents that revision_id
655
is known to have and are in the repository already.
657
:returns: The validator(which is a sha1 digest, though what is sha'd is
658
repository format specific) of the serialized inventory.
660
if not self.is_in_write_group():
661
raise AssertionError("%r not in write group" % (self,))
662
_mod_revision.check_not_reserved_id(revision_id)
663
if not (inv.revision_id is None or inv.revision_id == revision_id):
664
raise AssertionError(
665
"Mismatch between inventory revision"
666
" id and insertion revid (%r, %r)"
667
% (inv.revision_id, revision_id))
669
raise AssertionError()
670
inv_lines = self._serialise_inventory_to_lines(inv)
671
return self._inventory_add_lines(revision_id, parents,
672
inv_lines, check_content=False)
674
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
676
"""Add a new inventory expressed as a delta against another revision.
678
:param basis_revision_id: The inventory id the delta was created
679
against. (This does not have to be a direct parent.)
680
:param delta: The inventory delta (see Inventory.apply_delta for
682
:param new_revision_id: The revision id that the inventory is being
684
:param parents: The revision ids of the parents that revision_id is
685
known to have and are in the repository already. These are supplied
686
for repositories that depend on the inventory graph for revision
687
graph access, as well as for those that pun ancestry with delta
690
:returns: (validator, new_inv)
691
The validator(which is a sha1 digest, though what is sha'd is
692
repository format specific) of the serialized inventory, and the
695
if not self.is_in_write_group():
696
raise AssertionError("%r not in write group" % (self,))
697
_mod_revision.check_not_reserved_id(new_revision_id)
698
basis_tree = self.revision_tree(basis_revision_id)
699
basis_tree.lock_read()
701
# Note that this mutates the inventory of basis_tree, which not all
702
# inventory implementations may support: A better idiom would be to
703
# return a new inventory, but as there is no revision tree cache in
704
# repository this is safe for now - RBC 20081013
705
basis_inv = basis_tree.inventory
706
basis_inv.apply_delta(delta)
707
basis_inv.revision_id = new_revision_id
708
return (self.add_inventory(new_revision_id, basis_inv, parents),
713
def _inventory_add_lines(self, revision_id, parents, lines,
715
"""Store lines in inv_vf and return the sha1 of the inventory."""
716
parents = [(parent,) for parent in parents]
717
return self.inventories.add_lines((revision_id,), parents, lines,
718
check_content=check_content)[0]
720
def add_revision(self, revision_id, rev, inv=None, config=None):
721
"""Add rev to the revision store as revision_id.
723
:param revision_id: the revision id to use.
724
:param rev: The revision object.
725
:param inv: The inventory for the revision. if None, it will be looked
726
up in the inventory storer
727
:param config: If None no digital signature will be created.
728
If supplied its signature_needed method will be used
729
to determine if a signature should be made.
731
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
733
_mod_revision.check_not_reserved_id(revision_id)
734
if config is not None and config.signature_needed():
736
inv = self.get_inventory(revision_id)
737
plaintext = Testament(rev, inv).as_short_text()
738
self.store_revision_signature(
739
gpg.GPGStrategy(config), plaintext, revision_id)
740
# check inventory present
741
if not self.inventories.get_parent_map([(revision_id,)]):
743
raise errors.WeaveRevisionNotPresent(revision_id,
746
# yes, this is not suitable for adding with ghosts.
747
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
751
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
752
self._add_revision(rev)
754
def _add_revision(self, revision):
755
text = self._serializer.write_revision_to_string(revision)
756
key = (revision.revision_id,)
757
parents = tuple((parent,) for parent in revision.parent_ids)
758
self.revisions.add_lines(key, parents, osutils.split_lines(text))
760
def all_revision_ids(self):
761
"""Returns a list of all the revision ids in the repository.
763
This is conceptually deprecated because code should generally work on
764
the graph reachable from a particular revision, and ignore any other
765
revisions that might be present. There is no direct replacement
768
if 'evil' in debug.debug_flags:
769
mutter_callsite(2, "all_revision_ids is linear with history.")
770
return self._all_revision_ids()
772
def _all_revision_ids(self):
773
"""Returns a list of all the revision ids in the repository.
775
These are in as much topological order as the underlying store can
778
raise NotImplementedError(self._all_revision_ids)
780
def break_lock(self):
781
"""Break a lock if one is present from another instance.
783
Uses the ui factory to ask for confirmation if the lock may be from
786
self.control_files.break_lock()
789
def _eliminate_revisions_not_present(self, revision_ids):
790
"""Check every revision id in revision_ids to see if we have it.
792
Returns a set of the present revisions.
795
graph = self.get_graph()
796
parent_map = graph.get_parent_map(revision_ids)
797
# The old API returned a list, should this actually be a set?
798
return parent_map.keys()
801
def create(a_bzrdir):
802
"""Construct the current default format repository in a_bzrdir."""
803
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
805
def __init__(self, _format, a_bzrdir, control_files):
806
"""instantiate a Repository.
808
:param _format: The format of the repository on disk.
809
:param a_bzrdir: The BzrDir of the repository.
811
In the future we will have a single api for all stores for
812
getting file texts, inventories and revisions, then
813
this construct will accept instances of those things.
815
super(Repository, self).__init__()
816
self._format = _format
817
# the following are part of the public API for Repository:
818
self.bzrdir = a_bzrdir
819
self.control_files = control_files
820
self._transport = control_files._transport
821
self.base = self._transport.base
823
self._reconcile_does_inventory_gc = True
824
self._reconcile_fixes_text_parents = False
825
self._reconcile_backsup_inventory = True
826
# not right yet - should be more semantically clear ?
828
# TODO: make sure to construct the right store classes, etc, depending
829
# on whether escaping is required.
830
self._warn_if_deprecated()
831
self._write_group = None
832
# Additional places to query for data.
833
self._fallback_repositories = []
834
# What order should fetch operations request streams in?
835
# The default is unordered as that is the cheapest for an origin to
837
self._fetch_order = 'unordered'
838
# Does this repository use deltas that can be fetched as-deltas ?
839
# (E.g. knits, where the knit deltas can be transplanted intact.
840
# We default to False, which will ensure that enough data to get
841
# a full text out of any fetch stream will be grabbed.
842
self._fetch_uses_deltas = False
843
# Should fetch trigger a reconcile after the fetch? Only needed for
844
# some repository formats that can suffer internal inconsistencies.
845
self._fetch_reconcile = False
846
# An InventoryEntry cache, used during deserialization
847
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
850
return '%s(%r)' % (self.__class__.__name__,
853
def has_same_location(self, other):
854
"""Returns a boolean indicating if this repository is at the same
855
location as another repository.
857
This might return False even when two repository objects are accessing
858
the same physical repository via different URLs.
860
if self.__class__ is not other.__class__:
862
return (self._transport.base == other._transport.base)
864
def is_in_write_group(self):
865
"""Return True if there is an open write group.
867
:seealso: start_write_group.
869
return self._write_group is not None
872
return self.control_files.is_locked()
874
def is_write_locked(self):
875
"""Return True if this object is write locked."""
876
return self.is_locked() and self.control_files._lock_mode == 'w'
878
def lock_write(self, token=None):
879
"""Lock this repository for writing.
881
This causes caching within the repository obejct to start accumlating
882
data during reads, and allows a 'write_group' to be obtained. Write
883
groups must be used for actual data insertion.
885
:param token: if this is already locked, then lock_write will fail
886
unless the token matches the existing lock.
887
:returns: a token if this instance supports tokens, otherwise None.
888
:raises TokenLockingNotSupported: when a token is given but this
889
instance doesn't support using token locks.
890
:raises MismatchedToken: if the specified token doesn't match the token
891
of the existing lock.
892
:seealso: start_write_group.
894
A token should be passed in if you know that you have locked the object
895
some other way, and need to synchronise this object's state with that
898
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
900
result = self.control_files.lock_write(token=token)
901
for repo in self._fallback_repositories:
902
# Writes don't affect fallback repos
908
self.control_files.lock_read()
909
for repo in self._fallback_repositories:
913
def get_physical_lock_status(self):
914
return self.control_files.get_physical_lock_status()
916
def leave_lock_in_place(self):
917
"""Tell this repository not to release the physical lock when this
920
If lock_write doesn't return a token, then this method is not supported.
922
self.control_files.leave_in_place()
924
def dont_leave_lock_in_place(self):
925
"""Tell this repository to release the physical lock when this
926
object is unlocked, even if it didn't originally acquire it.
928
If lock_write doesn't return a token, then this method is not supported.
930
self.control_files.dont_leave_in_place()
933
def gather_stats(self, revid=None, committers=None):
934
"""Gather statistics from a revision id.
936
:param revid: The revision id to gather statistics from, if None, then
937
no revision specific statistics are gathered.
938
:param committers: Optional parameter controlling whether to grab
939
a count of committers from the revision specific statistics.
940
:return: A dictionary of statistics. Currently this contains:
941
committers: The number of committers if requested.
942
firstrev: A tuple with timestamp, timezone for the penultimate left
943
most ancestor of revid, if revid is not the NULL_REVISION.
944
latestrev: A tuple with timestamp, timezone for revid, if revid is
945
not the NULL_REVISION.
946
revisions: The total revision count in the repository.
947
size: An estimate disk size of the repository in bytes.
950
if revid and committers:
951
result['committers'] = 0
952
if revid and revid != _mod_revision.NULL_REVISION:
954
all_committers = set()
955
revisions = self.get_ancestry(revid)
956
# pop the leading None
958
first_revision = None
960
# ignore the revisions in the middle - just grab first and last
961
revisions = revisions[0], revisions[-1]
962
for revision in self.get_revisions(revisions):
963
if not first_revision:
964
first_revision = revision
966
all_committers.add(revision.committer)
967
last_revision = revision
969
result['committers'] = len(all_committers)
970
result['firstrev'] = (first_revision.timestamp,
971
first_revision.timezone)
972
result['latestrev'] = (last_revision.timestamp,
973
last_revision.timezone)
975
# now gather global repository information
976
# XXX: This is available for many repos regardless of listability.
977
if self.bzrdir.root_transport.listable():
978
# XXX: do we want to __define len__() ?
979
# Maybe the versionedfiles object should provide a different
980
# method to get the number of keys.
981
result['revisions'] = len(self.revisions.keys())
985
def find_branches(self, using=False):
986
"""Find branches underneath this repository.
988
This will include branches inside other branches.
990
:param using: If True, list only branches using this repository.
992
if using and not self.is_shared():
994
return [self.bzrdir.open_branch()]
995
except errors.NotBranchError:
997
class Evaluator(object):
1000
self.first_call = True
1002
def __call__(self, bzrdir):
1003
# On the first call, the parameter is always the bzrdir
1004
# containing the current repo.
1005
if not self.first_call:
1007
repository = bzrdir.open_repository()
1008
except errors.NoRepositoryPresent:
1011
return False, (None, repository)
1012
self.first_call = False
1014
value = (bzrdir.open_branch(), None)
1015
except errors.NotBranchError:
1016
value = (None, None)
1020
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1021
self.bzrdir.root_transport, evaluate=Evaluator()):
1022
if branch is not None:
1023
branches.append(branch)
1024
if not using and repository is not None:
1025
branches.extend(repository.find_branches())
1029
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1030
"""Return the revision ids that other has that this does not.
1032
These are returned in topological order.
1034
revision_id: only return revision ids included by revision_id.
1036
return InterRepository.get(other, self).search_missing_revision_ids(
1037
revision_id, find_ghosts)
1039
@deprecated_method(one_two)
1041
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1042
"""Return the revision ids that other has that this does not.
1044
These are returned in topological order.
1046
revision_id: only return revision ids included by revision_id.
1048
keys = self.search_missing_revision_ids(
1049
other, revision_id, find_ghosts).get_keys()
1052
parents = other.get_graph().get_parent_map(keys)
1055
return tsort.topo_sort(parents)
1059
"""Open the repository rooted at base.
1061
For instance, if the repository is at URL/.bzr/repository,
1062
Repository.open(URL) -> a Repository instance.
1064
control = bzrdir.BzrDir.open(base)
1065
return control.open_repository()
1067
def copy_content_into(self, destination, revision_id=None):
1068
"""Make a complete copy of the content in self into destination.
1070
This is a destructive operation! Do not use it on existing
1073
return InterRepository.get(self, destination).copy_content(revision_id)
1075
def commit_write_group(self):
1076
"""Commit the contents accrued within the current write group.
1078
:seealso: start_write_group.
1080
if self._write_group is not self.get_transaction():
1081
# has an unlock or relock occured ?
1082
raise errors.BzrError('mismatched lock context %r and '
1084
(self.get_transaction(), self._write_group))
1085
self._commit_write_group()
1086
self._write_group = None
1088
def _commit_write_group(self):
1089
"""Template method for per-repository write group cleanup.
1091
This is called before the write group is considered to be
1092
finished and should ensure that all data handed to the repository
1093
for writing during the write group is safely committed (to the
1094
extent possible considering file system caching etc).
1097
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1098
"""Fetch the content required to construct revision_id from source.
1100
If revision_id is None all content is copied.
1101
:param find_ghosts: Find and copy revisions in the source that are
1102
ghosts in the target (and not reachable directly by walking out to
1103
the first-present revision in target from revision_id).
1105
# fast path same-url fetch operations
1106
if self.has_same_location(source):
1107
# check that last_revision is in 'from' and then return a
1109
if (revision_id is not None and
1110
not _mod_revision.is_null(revision_id)):
1111
self.get_revision(revision_id)
1113
# if there is no specific appropriate InterRepository, this will get
1114
# the InterRepository base class, which raises an
1115
# IncompatibleRepositories when asked to fetch.
1116
inter = InterRepository.get(source, self)
1117
return inter.fetch(revision_id=revision_id, pb=pb,
1118
find_ghosts=find_ghosts)
1120
def create_bundle(self, target, base, fileobj, format=None):
1121
return serializer.write_bundle(self, target, base, fileobj, format)
1123
def get_commit_builder(self, branch, parents, config, timestamp=None,
1124
timezone=None, committer=None, revprops=None,
1126
"""Obtain a CommitBuilder for this repository.
1128
:param branch: Branch to commit to.
1129
:param parents: Revision ids of the parents of the new revision.
1130
:param config: Configuration to use.
1131
:param timestamp: Optional timestamp recorded for commit.
1132
:param timezone: Optional timezone for timestamp.
1133
:param committer: Optional committer to set for commit.
1134
:param revprops: Optional dictionary of revision properties.
1135
:param revision_id: Optional revision id.
1137
result = self._commit_builder_class(self, parents, config,
1138
timestamp, timezone, committer, revprops, revision_id)
1139
self.start_write_group()
1143
if (self.control_files._lock_count == 1 and
1144
self.control_files._lock_mode == 'w'):
1145
if self._write_group is not None:
1146
self.abort_write_group()
1147
self.control_files.unlock()
1148
raise errors.BzrError(
1149
'Must end write groups before releasing write locks.')
1150
self.control_files.unlock()
1151
if self.control_files._lock_count == 0:
1152
self._inventory_entry_cache.clear()
1153
for repo in self._fallback_repositories:
1157
def clone(self, a_bzrdir, revision_id=None):
1158
"""Clone this repository into a_bzrdir using the current format.
1160
Currently no check is made that the format of this repository and
1161
the bzrdir format are compatible. FIXME RBC 20060201.
1163
:return: The newly created destination repository.
1165
# TODO: deprecate after 0.16; cloning this with all its settings is
1166
# probably not very useful -- mbp 20070423
1167
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1168
self.copy_content_into(dest_repo, revision_id)
1171
def start_write_group(self):
1172
"""Start a write group in the repository.
1174
Write groups are used by repositories which do not have a 1:1 mapping
1175
between file ids and backend store to manage the insertion of data from
1176
both fetch and commit operations.
1178
A write lock is required around the start_write_group/commit_write_group
1179
for the support of lock-requiring repository formats.
1181
One can only insert data into a repository inside a write group.
1185
if not self.is_write_locked():
1186
raise errors.NotWriteLocked(self)
1187
if self._write_group:
1188
raise errors.BzrError('already in a write group')
1189
self._start_write_group()
1190
# so we can detect unlock/relock - the write group is now entered.
1191
self._write_group = self.get_transaction()
1193
def _start_write_group(self):
1194
"""Template method for per-repository write group startup.
1196
This is called before the write group is considered to be
1201
def sprout(self, to_bzrdir, revision_id=None):
1202
"""Create a descendent repository for new development.
1204
Unlike clone, this does not copy the settings of the repository.
1206
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1207
dest_repo.fetch(self, revision_id=revision_id)
1210
def _create_sprouting_repo(self, a_bzrdir, shared):
1211
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1212
# use target default format.
1213
dest_repo = a_bzrdir.create_repository()
1215
# Most control formats need the repository to be specifically
1216
# created, but on some old all-in-one formats it's not needed
1218
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1219
except errors.UninitializableFormat:
1220
dest_repo = a_bzrdir.open_repository()
1223
def _get_sink(self):
1224
"""Return a sink for streaming into this repository."""
1225
return StreamSink(self)
1228
def has_revision(self, revision_id):
1229
"""True if this repository has a copy of the revision."""
1230
return revision_id in self.has_revisions((revision_id,))
1233
def has_revisions(self, revision_ids):
1234
"""Probe to find out the presence of multiple revisions.
1236
:param revision_ids: An iterable of revision_ids.
1237
:return: A set of the revision_ids that were present.
1239
parent_map = self.revisions.get_parent_map(
1240
[(rev_id,) for rev_id in revision_ids])
1242
if _mod_revision.NULL_REVISION in revision_ids:
1243
result.add(_mod_revision.NULL_REVISION)
1244
result.update([key[0] for key in parent_map])
1248
def get_revision(self, revision_id):
1249
"""Return the Revision object for a named revision."""
1250
return self.get_revisions([revision_id])[0]
1253
def get_revision_reconcile(self, revision_id):
1254
"""'reconcile' helper routine that allows access to a revision always.
1256
This variant of get_revision does not cross check the weave graph
1257
against the revision one as get_revision does: but it should only
1258
be used by reconcile, or reconcile-alike commands that are correcting
1259
or testing the revision graph.
1261
return self._get_revisions([revision_id])[0]
1264
def get_revisions(self, revision_ids):
1265
"""Get many revisions at once."""
1266
return self._get_revisions(revision_ids)
1269
def _get_revisions(self, revision_ids):
1270
"""Core work logic to get many revisions without sanity checks."""
1271
for rev_id in revision_ids:
1272
if not rev_id or not isinstance(rev_id, basestring):
1273
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1274
keys = [(key,) for key in revision_ids]
1275
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1277
for record in stream:
1278
if record.storage_kind == 'absent':
1279
raise errors.NoSuchRevision(self, record.key[0])
1280
text = record.get_bytes_as('fulltext')
1281
rev = self._serializer.read_revision_from_string(text)
1282
revs[record.key[0]] = rev
1283
return [revs[revid] for revid in revision_ids]
1286
def get_revision_xml(self, revision_id):
1287
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1288
# would have already do it.
1289
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1290
rev = self.get_revision(revision_id)
1291
rev_tmp = cStringIO.StringIO()
1292
# the current serializer..
1293
self._serializer.write_revision(rev, rev_tmp)
1295
return rev_tmp.getvalue()
1297
def get_deltas_for_revisions(self, revisions):
1298
"""Produce a generator of revision deltas.
1300
Note that the input is a sequence of REVISIONS, not revision_ids.
1301
Trees will be held in memory until the generator exits.
1302
Each delta is relative to the revision's lefthand predecessor.
1304
required_trees = set()
1305
for revision in revisions:
1306
required_trees.add(revision.revision_id)
1307
required_trees.update(revision.parent_ids[:1])
1308
trees = dict((t.get_revision_id(), t) for
1309
t in self.revision_trees(required_trees))
1310
for revision in revisions:
1311
if not revision.parent_ids:
1312
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1314
old_tree = trees[revision.parent_ids[0]]
1315
yield trees[revision.revision_id].changes_from(old_tree)
1318
def get_revision_delta(self, revision_id):
1319
"""Return the delta for one revision.
1321
The delta is relative to the left-hand predecessor of the
1324
r = self.get_revision(revision_id)
1325
return list(self.get_deltas_for_revisions([r]))[0]
1328
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1329
signature = gpg_strategy.sign(plaintext)
1330
self.add_signature_text(revision_id, signature)
1333
def add_signature_text(self, revision_id, signature):
1334
self.signatures.add_lines((revision_id,), (),
1335
osutils.split_lines(signature))
1337
def find_text_key_references(self):
1338
"""Find the text key references within the repository.
1340
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1341
revision_ids. Each altered file-ids has the exact revision_ids that
1342
altered it listed explicitly.
1343
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1344
to whether they were referred to by the inventory of the
1345
revision_id that they contain. The inventory texts from all present
1346
revision ids are assessed to generate this report.
1348
revision_keys = self.revisions.keys()
1349
w = self.inventories
1350
pb = ui.ui_factory.nested_progress_bar()
1352
return self._find_text_key_references_from_xml_inventory_lines(
1353
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1357
def _find_text_key_references_from_xml_inventory_lines(self,
1359
"""Core routine for extracting references to texts from inventories.
1361
This performs the translation of xml lines to revision ids.
1363
:param line_iterator: An iterator of lines, origin_version_id
1364
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1365
to whether they were referred to by the inventory of the
1366
revision_id that they contain. Note that if that revision_id was
1367
not part of the line_iterator's output then False will be given -
1368
even though it may actually refer to that key.
1370
if not self._serializer.support_altered_by_hack:
1371
raise AssertionError(
1372
"_find_text_key_references_from_xml_inventory_lines only "
1373
"supported for branches which store inventory as unnested xml"
1374
", not on %r" % self)
1377
# this code needs to read every new line in every inventory for the
1378
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1379
# not present in one of those inventories is unnecessary but not
1380
# harmful because we are filtering by the revision id marker in the
1381
# inventory lines : we only select file ids altered in one of those
1382
# revisions. We don't need to see all lines in the inventory because
1383
# only those added in an inventory in rev X can contain a revision=X
1385
unescape_revid_cache = {}
1386
unescape_fileid_cache = {}
1388
# jam 20061218 In a big fetch, this handles hundreds of thousands
1389
# of lines, so it has had a lot of inlining and optimizing done.
1390
# Sorry that it is a little bit messy.
1391
# Move several functions to be local variables, since this is a long
1393
search = self._file_ids_altered_regex.search
1394
unescape = _unescape_xml
1395
setdefault = result.setdefault
1396
for line, line_key in line_iterator:
1397
match = search(line)
1400
# One call to match.group() returning multiple items is quite a
1401
# bit faster than 2 calls to match.group() each returning 1
1402
file_id, revision_id = match.group('file_id', 'revision_id')
1404
# Inlining the cache lookups helps a lot when you make 170,000
1405
# lines and 350k ids, versus 8.4 unique ids.
1406
# Using a cache helps in 2 ways:
1407
# 1) Avoids unnecessary decoding calls
1408
# 2) Re-uses cached strings, which helps in future set and
1410
# (2) is enough that removing encoding entirely along with
1411
# the cache (so we are using plain strings) results in no
1412
# performance improvement.
1414
revision_id = unescape_revid_cache[revision_id]
1416
unescaped = unescape(revision_id)
1417
unescape_revid_cache[revision_id] = unescaped
1418
revision_id = unescaped
1420
# Note that unconditionally unescaping means that we deserialise
1421
# every fileid, which for general 'pull' is not great, but we don't
1422
# really want to have some many fulltexts that this matters anyway.
1425
file_id = unescape_fileid_cache[file_id]
1427
unescaped = unescape(file_id)
1428
unescape_fileid_cache[file_id] = unescaped
1431
key = (file_id, revision_id)
1432
setdefault(key, False)
1433
if revision_id == line_key[-1]:
1437
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1439
"""Helper routine for fileids_altered_by_revision_ids.
1441
This performs the translation of xml lines to revision ids.
1443
:param line_iterator: An iterator of lines, origin_version_id
1444
:param revision_ids: The revision ids to filter for. This should be a
1445
set or other type which supports efficient __contains__ lookups, as
1446
the revision id from each parsed line will be looked up in the
1447
revision_ids filter.
1448
:return: a dictionary mapping altered file-ids to an iterable of
1449
revision_ids. Each altered file-ids has the exact revision_ids that
1450
altered it listed explicitly.
1453
setdefault = result.setdefault
1455
self._find_text_key_references_from_xml_inventory_lines(
1456
line_iterator).iterkeys():
1457
# once data is all ensured-consistent; then this is
1458
# if revision_id == version_id
1459
if key[-1:] in revision_ids:
1460
setdefault(key[0], set()).add(key[-1])
1463
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1464
"""Find the file ids and versions affected by revisions.
1466
:param revisions: an iterable containing revision ids.
1467
:param _inv_weave: The inventory weave from this repository or None.
1468
If None, the inventory weave will be opened automatically.
1469
:return: a dictionary mapping altered file-ids to an iterable of
1470
revision_ids. Each altered file-ids has the exact revision_ids that
1471
altered it listed explicitly.
1473
selected_keys = set((revid,) for revid in revision_ids)
1474
w = _inv_weave or self.inventories
1475
pb = ui.ui_factory.nested_progress_bar()
1477
return self._find_file_ids_from_xml_inventory_lines(
1478
w.iter_lines_added_or_present_in_keys(
1479
selected_keys, pb=pb),
1484
def iter_files_bytes(self, desired_files):
1485
"""Iterate through file versions.
1487
Files will not necessarily be returned in the order they occur in
1488
desired_files. No specific order is guaranteed.
1490
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1491
value supplied by the caller as part of desired_files. It should
1492
uniquely identify the file version in the caller's context. (Examples:
1493
an index number or a TreeTransform trans_id.)
1495
bytes_iterator is an iterable of bytestrings for the file. The
1496
kind of iterable and length of the bytestrings are unspecified, but for
1497
this implementation, it is a list of bytes produced by
1498
VersionedFile.get_record_stream().
1500
:param desired_files: a list of (file_id, revision_id, identifier)
1504
for file_id, revision_id, callable_data in desired_files:
1505
text_keys[(file_id, revision_id)] = callable_data
1506
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1507
if record.storage_kind == 'absent':
1508
raise errors.RevisionNotPresent(record.key, self)
1509
yield text_keys[record.key], record.get_bytes_as('fulltext')
1511
def _generate_text_key_index(self, text_key_references=None,
1513
"""Generate a new text key index for the repository.
1515
This is an expensive function that will take considerable time to run.
1517
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1518
list of parents, also text keys. When a given key has no parents,
1519
the parents list will be [NULL_REVISION].
1521
# All revisions, to find inventory parents.
1522
if ancestors is None:
1523
graph = self.get_graph()
1524
ancestors = graph.get_parent_map(self.all_revision_ids())
1525
if text_key_references is None:
1526
text_key_references = self.find_text_key_references()
1527
pb = ui.ui_factory.nested_progress_bar()
1529
return self._do_generate_text_key_index(ancestors,
1530
text_key_references, pb)
1534
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1535
"""Helper for _generate_text_key_index to avoid deep nesting."""
1536
revision_order = tsort.topo_sort(ancestors)
1537
invalid_keys = set()
1539
for revision_id in revision_order:
1540
revision_keys[revision_id] = set()
1541
text_count = len(text_key_references)
1542
# a cache of the text keys to allow reuse; costs a dict of all the
1543
# keys, but saves a 2-tuple for every child of a given key.
1545
for text_key, valid in text_key_references.iteritems():
1547
invalid_keys.add(text_key)
1549
revision_keys[text_key[1]].add(text_key)
1550
text_key_cache[text_key] = text_key
1551
del text_key_references
1553
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1554
NULL_REVISION = _mod_revision.NULL_REVISION
1555
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1556
# too small for large or very branchy trees. However, for 55K path
1557
# trees, it would be easy to use too much memory trivially. Ideally we
1558
# could gauge this by looking at available real memory etc, but this is
1559
# always a tricky proposition.
1560
inventory_cache = lru_cache.LRUCache(10)
1561
batch_size = 10 # should be ~150MB on a 55K path tree
1562
batch_count = len(revision_order) / batch_size + 1
1564
pb.update("Calculating text parents.", processed_texts, text_count)
1565
for offset in xrange(batch_count):
1566
to_query = revision_order[offset * batch_size:(offset + 1) *
1570
for rev_tree in self.revision_trees(to_query):
1571
revision_id = rev_tree.get_revision_id()
1572
parent_ids = ancestors[revision_id]
1573
for text_key in revision_keys[revision_id]:
1574
pb.update("Calculating text parents.", processed_texts)
1575
processed_texts += 1
1576
candidate_parents = []
1577
for parent_id in parent_ids:
1578
parent_text_key = (text_key[0], parent_id)
1580
check_parent = parent_text_key not in \
1581
revision_keys[parent_id]
1583
# the parent parent_id is a ghost:
1584
check_parent = False
1585
# truncate the derived graph against this ghost.
1586
parent_text_key = None
1588
# look at the parent commit details inventories to
1589
# determine possible candidates in the per file graph.
1592
inv = inventory_cache[parent_id]
1594
inv = self.revision_tree(parent_id).inventory
1595
inventory_cache[parent_id] = inv
1596
parent_entry = inv._byid.get(text_key[0], None)
1597
if parent_entry is not None:
1599
text_key[0], parent_entry.revision)
1601
parent_text_key = None
1602
if parent_text_key is not None:
1603
candidate_parents.append(
1604
text_key_cache[parent_text_key])
1605
parent_heads = text_graph.heads(candidate_parents)
1606
new_parents = list(parent_heads)
1607
new_parents.sort(key=lambda x:candidate_parents.index(x))
1608
if new_parents == []:
1609
new_parents = [NULL_REVISION]
1610
text_index[text_key] = new_parents
1612
for text_key in invalid_keys:
1613
text_index[text_key] = [NULL_REVISION]
1616
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1617
"""Get an iterable listing the keys of all the data introduced by a set
1620
The keys will be ordered so that the corresponding items can be safely
1621
fetched and inserted in that order.
1623
:returns: An iterable producing tuples of (knit-kind, file-id,
1624
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1625
'revisions'. file-id is None unless knit-kind is 'file'.
1627
# XXX: it's a bit weird to control the inventory weave caching in this
1628
# generator. Ideally the caching would be done in fetch.py I think. Or
1629
# maybe this generator should explicitly have the contract that it
1630
# should not be iterated until the previously yielded item has been
1632
inv_w = self.inventories
1634
# file ids that changed
1635
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1637
num_file_ids = len(file_ids)
1638
for file_id, altered_versions in file_ids.iteritems():
1639
if _files_pb is not None:
1640
_files_pb.update("fetch texts", count, num_file_ids)
1642
yield ("file", file_id, altered_versions)
1643
# We're done with the files_pb. Note that it finished by the caller,
1644
# just as it was created by the caller.
1648
yield ("inventory", None, revision_ids)
1651
# XXX: Note ATM no callers actually pay attention to this return
1652
# instead they just use the list of revision ids and ignore
1653
# missing sigs. Consider removing this work entirely
1654
revisions_with_signatures = set(self.signatures.get_parent_map(
1655
[(r,) for r in revision_ids]))
1656
revisions_with_signatures = set(
1657
[r for (r,) in revisions_with_signatures])
1658
revisions_with_signatures.intersection_update(revision_ids)
1659
yield ("signatures", None, revisions_with_signatures)
1662
yield ("revisions", None, revision_ids)
1665
def get_inventory(self, revision_id):
1666
"""Get Inventory object by revision id."""
1667
return self.iter_inventories([revision_id]).next()
1669
def iter_inventories(self, revision_ids):
1670
"""Get many inventories by revision_ids.
1672
This will buffer some or all of the texts used in constructing the
1673
inventories in memory, but will only parse a single inventory at a
1676
:return: An iterator of inventories.
1678
if ((None in revision_ids)
1679
or (_mod_revision.NULL_REVISION in revision_ids)):
1680
raise ValueError('cannot get null revision inventory')
1681
return self._iter_inventories(revision_ids)
1683
def _iter_inventories(self, revision_ids):
1684
"""single-document based inventory iteration."""
1685
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1686
yield self.deserialise_inventory(revision_id, text)
1688
def _iter_inventory_xmls(self, revision_ids):
1689
keys = [(revision_id,) for revision_id in revision_ids]
1690
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1692
for record in stream:
1693
if record.storage_kind != 'absent':
1694
text_chunks[record.key] = record.get_bytes_as('chunked')
1696
raise errors.NoSuchRevision(self, record.key)
1698
chunks = text_chunks.pop(key)
1699
yield ''.join(chunks), key[-1]
1701
def deserialise_inventory(self, revision_id, xml):
1702
"""Transform the xml into an inventory object.
1704
:param revision_id: The expected revision id of the inventory.
1705
:param xml: A serialised inventory.
1707
result = self._serializer.read_inventory_from_string(xml, revision_id,
1708
entry_cache=self._inventory_entry_cache)
1709
if result.revision_id != revision_id:
1710
raise AssertionError('revision id mismatch %s != %s' % (
1711
result.revision_id, revision_id))
1714
def serialise_inventory(self, inv):
1715
return self._serializer.write_inventory_to_string(inv)
1717
def _serialise_inventory_to_lines(self, inv):
1718
return self._serializer.write_inventory_to_lines(inv)
1720
def get_serializer_format(self):
1721
return self._serializer.format_num
1724
def get_inventory_xml(self, revision_id):
1725
"""Get inventory XML as a file object."""
1726
texts = self._iter_inventory_xmls([revision_id])
1728
text, revision_id = texts.next()
1729
except StopIteration:
1730
raise errors.HistoryMissing(self, 'inventory', revision_id)
1734
def get_inventory_sha1(self, revision_id):
1735
"""Return the sha1 hash of the inventory entry
1737
return self.get_revision(revision_id).inventory_sha1
1739
def iter_reverse_revision_history(self, revision_id):
1740
"""Iterate backwards through revision ids in the lefthand history
1742
:param revision_id: The revision id to start with. All its lefthand
1743
ancestors will be traversed.
1745
graph = self.get_graph()
1746
next_id = revision_id
1748
if next_id in (None, _mod_revision.NULL_REVISION):
1751
# Note: The following line may raise KeyError in the event of
1752
# truncated history. We decided not to have a try:except:raise
1753
# RevisionNotPresent here until we see a use for it, because of the
1754
# cost in an inner loop that is by its very nature O(history).
1755
# Robert Collins 20080326
1756
parents = graph.get_parent_map([next_id])[next_id]
1757
if len(parents) == 0:
1760
next_id = parents[0]
1763
def get_revision_inventory(self, revision_id):
1764
"""Return inventory of a past revision."""
1765
# TODO: Unify this with get_inventory()
1766
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1767
# must be the same as its revision, so this is trivial.
1768
if revision_id is None:
1769
# This does not make sense: if there is no revision,
1770
# then it is the current tree inventory surely ?!
1771
# and thus get_root_id() is something that looks at the last
1772
# commit on the branch, and the get_root_id is an inventory check.
1773
raise NotImplementedError
1774
# return Inventory(self.get_root_id())
1776
return self.get_inventory(revision_id)
1778
def is_shared(self):
1779
"""Return True if this repository is flagged as a shared repository."""
1780
raise NotImplementedError(self.is_shared)
1783
def reconcile(self, other=None, thorough=False):
1784
"""Reconcile this repository."""
1785
from bzrlib.reconcile import RepoReconciler
1786
reconciler = RepoReconciler(self, thorough=thorough)
1787
reconciler.reconcile()
1790
def _refresh_data(self):
1791
"""Helper called from lock_* to ensure coherency with disk.
1793
The default implementation does nothing; it is however possible
1794
for repositories to maintain loaded indices across multiple locks
1795
by checking inside their implementation of this method to see
1796
whether their indices are still valid. This depends of course on
1797
the disk format being validatable in this manner.
1801
def revision_tree(self, revision_id):
1802
"""Return Tree for a revision on this branch.
1804
`revision_id` may be NULL_REVISION for the empty tree revision.
1806
revision_id = _mod_revision.ensure_null(revision_id)
1807
# TODO: refactor this to use an existing revision object
1808
# so we don't need to read it in twice.
1809
if revision_id == _mod_revision.NULL_REVISION:
1810
return RevisionTree(self, Inventory(root_id=None),
1811
_mod_revision.NULL_REVISION)
1813
inv = self.get_revision_inventory(revision_id)
1814
return RevisionTree(self, inv, revision_id)
1816
def revision_trees(self, revision_ids):
1817
"""Return Tree for a revision on this branch.
1819
`revision_id` may not be None or 'null:'"""
1820
inventories = self.iter_inventories(revision_ids)
1821
for inv in inventories:
1822
yield RevisionTree(self, inv, inv.revision_id)
1825
def get_ancestry(self, revision_id, topo_sorted=True):
1826
"""Return a list of revision-ids integrated by a revision.
1828
The first element of the list is always None, indicating the origin
1829
revision. This might change when we have history horizons, or
1830
perhaps we should have a new API.
1832
This is topologically sorted.
1834
if _mod_revision.is_null(revision_id):
1836
if not self.has_revision(revision_id):
1837
raise errors.NoSuchRevision(self, revision_id)
1838
graph = self.get_graph()
1840
search = graph._make_breadth_first_searcher([revision_id])
1843
found, ghosts = search.next_with_ghosts()
1844
except StopIteration:
1847
if _mod_revision.NULL_REVISION in keys:
1848
keys.remove(_mod_revision.NULL_REVISION)
1850
parent_map = graph.get_parent_map(keys)
1851
keys = tsort.topo_sort(parent_map)
1852
return [None] + list(keys)
1855
"""Compress the data within the repository.
1857
This operation only makes sense for some repository types. For other
1858
types it should be a no-op that just returns.
1860
This stub method does not require a lock, but subclasses should use
1861
@needs_write_lock as this is a long running call its reasonable to
1862
implicitly lock for the user.
1866
@deprecated_method(one_six)
1867
def print_file(self, file, revision_id):
1868
"""Print `file` to stdout.
1870
FIXME RBC 20060125 as John Meinel points out this is a bad api
1871
- it writes to stdout, it assumes that that is valid etc. Fix
1872
by creating a new more flexible convenience function.
1874
tree = self.revision_tree(revision_id)
1875
# use inventory as it was in that revision
1876
file_id = tree.inventory.path2id(file)
1878
# TODO: jam 20060427 Write a test for this code path
1879
# it had a bug in it, and was raising the wrong
1881
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1882
tree.print_file(file_id)
1884
def get_transaction(self):
1885
return self.control_files.get_transaction()
1887
@deprecated_method(one_one)
1888
def get_parents(self, revision_ids):
1889
"""See StackedParentsProvider.get_parents"""
1890
parent_map = self.get_parent_map(revision_ids)
1891
return [parent_map.get(r, None) for r in revision_ids]
1893
def get_parent_map(self, revision_ids):
1894
"""See graph._StackedParentsProvider.get_parent_map"""
1895
# revisions index works in keys; this just works in revisions
1896
# therefore wrap and unwrap
1899
for revision_id in revision_ids:
1900
if revision_id == _mod_revision.NULL_REVISION:
1901
result[revision_id] = ()
1902
elif revision_id is None:
1903
raise ValueError('get_parent_map(None) is not valid')
1905
query_keys.append((revision_id ,))
1906
for ((revision_id,), parent_keys) in \
1907
self.revisions.get_parent_map(query_keys).iteritems():
1909
result[revision_id] = tuple(parent_revid
1910
for (parent_revid,) in parent_keys)
1912
result[revision_id] = (_mod_revision.NULL_REVISION,)
1915
def _make_parents_provider(self):
1918
def get_graph(self, other_repository=None):
1919
"""Return the graph walker for this repository format"""
1920
parents_provider = self._make_parents_provider()
1921
if (other_repository is not None and
1922
not self.has_same_location(other_repository)):
1923
parents_provider = graph._StackedParentsProvider(
1924
[parents_provider, other_repository._make_parents_provider()])
1925
return graph.Graph(parents_provider)
1927
def _get_versioned_file_checker(self):
1928
"""Return an object suitable for checking versioned files."""
1929
return _VersionedFileChecker(self)
1931
def revision_ids_to_search_result(self, result_set):
1932
"""Convert a set of revision ids to a graph SearchResult."""
1933
result_parents = set()
1934
for parents in self.get_graph().get_parent_map(
1935
result_set).itervalues():
1936
result_parents.update(parents)
1937
included_keys = result_set.intersection(result_parents)
1938
start_keys = result_set.difference(included_keys)
1939
exclude_keys = result_parents.difference(result_set)
1940
result = graph.SearchResult(start_keys, exclude_keys,
1941
len(result_set), result_set)
1945
def set_make_working_trees(self, new_value):
1946
"""Set the policy flag for making working trees when creating branches.
1948
This only applies to branches that use this repository.
1950
The default is 'True'.
1951
:param new_value: True to restore the default, False to disable making
1954
raise NotImplementedError(self.set_make_working_trees)
1956
def make_working_trees(self):
1957
"""Returns the policy for making working trees on new branches."""
1958
raise NotImplementedError(self.make_working_trees)
1961
def sign_revision(self, revision_id, gpg_strategy):
1962
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1963
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1966
def has_signature_for_revision_id(self, revision_id):
1967
"""Query for a revision signature for revision_id in the repository."""
1968
if not self.has_revision(revision_id):
1969
raise errors.NoSuchRevision(self, revision_id)
1970
sig_present = (1 == len(
1971
self.signatures.get_parent_map([(revision_id,)])))
1975
def get_signature_text(self, revision_id):
1976
"""Return the text for a signature."""
1977
stream = self.signatures.get_record_stream([(revision_id,)],
1979
record = stream.next()
1980
if record.storage_kind == 'absent':
1981
raise errors.NoSuchRevision(self, revision_id)
1982
return record.get_bytes_as('fulltext')
1985
def check(self, revision_ids=None):
1986
"""Check consistency of all history of given revision_ids.
1988
Different repository implementations should override _check().
1990
:param revision_ids: A non-empty list of revision_ids whose ancestry
1991
will be checked. Typically the last revision_id of a branch.
1993
return self._check(revision_ids)
1995
def _check(self, revision_ids):
1996
result = check.Check(self)
2000
def _warn_if_deprecated(self):
2001
global _deprecation_warning_done
2002
if _deprecation_warning_done:
2004
_deprecation_warning_done = True
2005
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2006
% (self._format, self.bzrdir.transport.base))
2008
def supports_rich_root(self):
2009
return self._format.rich_root_data
2011
def _check_ascii_revisionid(self, revision_id, method):
2012
"""Private helper for ascii-only repositories."""
2013
# weave repositories refuse to store revisionids that are non-ascii.
2014
if revision_id is not None:
2015
# weaves require ascii revision ids.
2016
if isinstance(revision_id, unicode):
2018
revision_id.encode('ascii')
2019
except UnicodeEncodeError:
2020
raise errors.NonAsciiRevisionId(method, self)
2023
revision_id.decode('ascii')
2024
except UnicodeDecodeError:
2025
raise errors.NonAsciiRevisionId(method, self)
2027
def revision_graph_can_have_wrong_parents(self):
2028
"""Is it possible for this repository to have a revision graph with
2031
If True, then this repository must also implement
2032
_find_inconsistent_revision_parents so that check and reconcile can
2033
check for inconsistencies before proceeding with other checks that may
2034
depend on the revision index being consistent.
2036
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2039
# remove these delegates a while after bzr 0.15
2040
def __make_delegated(name, from_module):
2041
def _deprecated_repository_forwarder():
2042
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2043
% (name, from_module),
2046
m = __import__(from_module, globals(), locals(), [name])
2048
return getattr(m, name)
2049
except AttributeError:
2050
raise AttributeError('module %s has no name %s'
2052
globals()[name] = _deprecated_repository_forwarder
2055
'AllInOneRepository',
2056
'WeaveMetaDirRepository',
2057
'PreSplitOutRepositoryFormat',
2058
'RepositoryFormat4',
2059
'RepositoryFormat5',
2060
'RepositoryFormat6',
2061
'RepositoryFormat7',
2063
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2067
'RepositoryFormatKnit',
2068
'RepositoryFormatKnit1',
2070
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2073
def install_revision(repository, rev, revision_tree):
2074
"""Install all revision data into a repository."""
2075
install_revisions(repository, [(rev, revision_tree, None)])
2078
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2079
"""Install all revision data into a repository.
2081
Accepts an iterable of revision, tree, signature tuples. The signature
2084
repository.start_write_group()
2086
for n, (revision, revision_tree, signature) in enumerate(iterable):
2087
_install_revision(repository, revision, revision_tree, signature)
2089
pb.update('Transferring revisions', n + 1, num_revisions)
2091
repository.abort_write_group()
2094
repository.commit_write_group()
2097
def _install_revision(repository, rev, revision_tree, signature):
2098
"""Install all revision data into a repository."""
2099
present_parents = []
2101
for p_id in rev.parent_ids:
2102
if repository.has_revision(p_id):
2103
present_parents.append(p_id)
2104
parent_trees[p_id] = repository.revision_tree(p_id)
2106
parent_trees[p_id] = repository.revision_tree(
2107
_mod_revision.NULL_REVISION)
2109
inv = revision_tree.inventory
2110
entries = inv.iter_entries()
2111
# backwards compatibility hack: skip the root id.
2112
if not repository.supports_rich_root():
2113
path, root = entries.next()
2114
if root.revision != rev.revision_id:
2115
raise errors.IncompatibleRevision(repr(repository))
2117
for path, ie in entries:
2118
text_keys[(ie.file_id, ie.revision)] = ie
2119
text_parent_map = repository.texts.get_parent_map(text_keys)
2120
missing_texts = set(text_keys) - set(text_parent_map)
2121
# Add the texts that are not already present
2122
for text_key in missing_texts:
2123
ie = text_keys[text_key]
2125
# FIXME: TODO: The following loop overlaps/duplicates that done by
2126
# commit to determine parents. There is a latent/real bug here where
2127
# the parents inserted are not those commit would do - in particular
2128
# they are not filtered by heads(). RBC, AB
2129
for revision, tree in parent_trees.iteritems():
2130
if ie.file_id not in tree:
2132
parent_id = tree.inventory[ie.file_id].revision
2133
if parent_id in text_parents:
2135
text_parents.append((ie.file_id, parent_id))
2136
lines = revision_tree.get_file(ie.file_id).readlines()
2137
repository.texts.add_lines(text_key, text_parents, lines)
2139
# install the inventory
2140
repository.add_inventory(rev.revision_id, inv, present_parents)
2141
except errors.RevisionAlreadyPresent:
2143
if signature is not None:
2144
repository.add_signature_text(rev.revision_id, signature)
2145
repository.add_revision(rev.revision_id, rev, inv)
2148
class MetaDirRepository(Repository):
2149
"""Repositories in the new meta-dir layout.
2151
:ivar _transport: Transport for access to repository control files,
2152
typically pointing to .bzr/repository.
2155
def __init__(self, _format, a_bzrdir, control_files):
2156
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2157
self._transport = control_files._transport
2159
def is_shared(self):
2160
"""Return True if this repository is flagged as a shared repository."""
2161
return self._transport.has('shared-storage')
2164
def set_make_working_trees(self, new_value):
2165
"""Set the policy flag for making working trees when creating branches.
2167
This only applies to branches that use this repository.
2169
The default is 'True'.
2170
:param new_value: True to restore the default, False to disable making
2175
self._transport.delete('no-working-trees')
2176
except errors.NoSuchFile:
2179
self._transport.put_bytes('no-working-trees', '',
2180
mode=self.bzrdir._get_file_mode())
2182
def make_working_trees(self):
2183
"""Returns the policy for making working trees on new branches."""
2184
return not self._transport.has('no-working-trees')
2187
class MetaDirVersionedFileRepository(MetaDirRepository):
2188
"""Repositories in a meta-dir, that work via versioned file objects."""
2190
def __init__(self, _format, a_bzrdir, control_files):
2191
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2195
class RepositoryFormatRegistry(registry.Registry):
2196
"""Registry of RepositoryFormats."""
2198
def __init__(self, other_registry=None):
2199
registry.Registry.__init__(self)
2200
self._other_registry = other_registry
2202
def register_lazy(self, key, module_name, member_name,
2203
help=None, info=None,
2204
override_existing=False):
2205
# Overridden to allow capturing registrations to two seperate
2206
# registries in a single call.
2207
registry.Registry.register_lazy(self, key, module_name, member_name,
2208
help=help, info=info, override_existing=override_existing)
2209
if self._other_registry is not None:
2210
self._other_registry.register_lazy(key, module_name, member_name,
2211
help=help, info=info, override_existing=override_existing)
2213
def get(self, format_string):
2214
r = registry.Registry.get(self, format_string)
2220
network_format_registry = RepositoryFormatRegistry()
2221
"""Registry of formats indexed by their network name.
2223
The network name for a repository format is an identifier that can be used when
2224
referring to formats with smart server operations. See
2225
RepositoryFormat.network_name() for more detail.
2229
format_registry = RepositoryFormatRegistry(network_format_registry)
2230
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2232
This can contain either format instances themselves, or classes/factories that
2233
can be called to obtain one.
2237
#####################################################################
2238
# Repository Formats
2240
class RepositoryFormat(object):
2241
"""A repository format.
2243
Formats provide four things:
2244
* An initialization routine to construct repository data on disk.
2245
* a optional format string which is used when the BzrDir supports
2247
* an open routine which returns a Repository instance.
2248
* A network name for referring to the format in smart server RPC
2251
There is one and only one Format subclass for each on-disk format. But
2252
there can be one Repository subclass that is used for several different
2253
formats. The _format attribute on a Repository instance can be used to
2254
determine the disk format.
2256
Formats are placed in a registry by their format string for reference
2257
during opening. These should be subclasses of RepositoryFormat for
2260
Once a format is deprecated, just deprecate the initialize and open
2261
methods on the format class. Do not deprecate the object, as the
2262
object may be created even when a repository instnace hasn't been
2265
Common instance attributes:
2266
_matchingbzrdir - the bzrdir format that the repository format was
2267
originally written to work with. This can be used if manually
2268
constructing a bzrdir and repository, or more commonly for test suite
2272
# Set to True or False in derived classes. True indicates that the format
2273
# supports ghosts gracefully.
2274
supports_ghosts = None
2275
# Can this repository be given external locations to lookup additional
2276
# data. Set to True or False in derived classes.
2277
supports_external_lookups = None
2280
return "<%s>" % self.__class__.__name__
2282
def __eq__(self, other):
2283
# format objects are generally stateless
2284
return isinstance(other, self.__class__)
2286
def __ne__(self, other):
2287
return not self == other
2290
def find_format(klass, a_bzrdir):
2291
"""Return the format for the repository object in a_bzrdir.
2293
This is used by bzr native formats that have a "format" file in
2294
the repository. Other methods may be used by different types of
2298
transport = a_bzrdir.get_repository_transport(None)
2299
format_string = transport.get("format").read()
2300
return format_registry.get(format_string)
2301
except errors.NoSuchFile:
2302
raise errors.NoRepositoryPresent(a_bzrdir)
2304
raise errors.UnknownFormatError(format=format_string,
2308
def register_format(klass, format):
2309
format_registry.register(format.get_format_string(), format)
2312
def unregister_format(klass, format):
2313
format_registry.remove(format.get_format_string())
2316
def get_default_format(klass):
2317
"""Return the current default format."""
2318
from bzrlib import bzrdir
2319
return bzrdir.format_registry.make_bzrdir('default').repository_format
2321
def get_format_string(self):
2322
"""Return the ASCII format string that identifies this format.
2324
Note that in pre format ?? repositories the format string is
2325
not permitted nor written to disk.
2327
raise NotImplementedError(self.get_format_string)
2329
def get_format_description(self):
2330
"""Return the short description for this format."""
2331
raise NotImplementedError(self.get_format_description)
2333
# TODO: this shouldn't be in the base class, it's specific to things that
2334
# use weaves or knits -- mbp 20070207
2335
def _get_versioned_file_store(self,
2340
versionedfile_class=None,
2341
versionedfile_kwargs={},
2343
if versionedfile_class is None:
2344
versionedfile_class = self._versionedfile_class
2345
weave_transport = control_files._transport.clone(name)
2346
dir_mode = control_files._dir_mode
2347
file_mode = control_files._file_mode
2348
return VersionedFileStore(weave_transport, prefixed=prefixed,
2350
file_mode=file_mode,
2351
versionedfile_class=versionedfile_class,
2352
versionedfile_kwargs=versionedfile_kwargs,
2355
def initialize(self, a_bzrdir, shared=False):
2356
"""Initialize a repository of this format in a_bzrdir.
2358
:param a_bzrdir: The bzrdir to put the new repository in it.
2359
:param shared: The repository should be initialized as a sharable one.
2360
:returns: The new repository object.
2362
This may raise UninitializableFormat if shared repository are not
2363
compatible the a_bzrdir.
2365
raise NotImplementedError(self.initialize)
2367
def is_supported(self):
2368
"""Is this format supported?
2370
Supported formats must be initializable and openable.
2371
Unsupported formats may not support initialization or committing or
2372
some other features depending on the reason for not being supported.
2376
def network_name(self):
2377
"""A simple byte string uniquely identifying this format for RPC calls.
2379
MetaDir repository formats use their disk format string to identify the
2380
repository over the wire. All in one formats such as bzr < 0.8, and
2381
foreign formats like svn/git and hg should use some marker which is
2382
unique and immutable.
2384
raise NotImplementedError(self.network_name)
2386
def check_conversion_target(self, target_format):
2387
raise NotImplementedError(self.check_conversion_target)
2389
def open(self, a_bzrdir, _found=False):
2390
"""Return an instance of this format for the bzrdir a_bzrdir.
2392
_found is a private parameter, do not use it.
2394
raise NotImplementedError(self.open)
2397
class MetaDirRepositoryFormat(RepositoryFormat):
2398
"""Common base class for the new repositories using the metadir layout."""
2400
rich_root_data = False
2401
supports_tree_reference = False
2402
supports_external_lookups = False
2405
def _matchingbzrdir(self):
2406
matching = bzrdir.BzrDirMetaFormat1()
2407
matching.repository_format = self
2411
super(MetaDirRepositoryFormat, self).__init__()
2413
def _create_control_files(self, a_bzrdir):
2414
"""Create the required files and the initial control_files object."""
2415
# FIXME: RBC 20060125 don't peek under the covers
2416
# NB: no need to escape relative paths that are url safe.
2417
repository_transport = a_bzrdir.get_repository_transport(self)
2418
control_files = lockable_files.LockableFiles(repository_transport,
2419
'lock', lockdir.LockDir)
2420
control_files.create_lock()
2421
return control_files
2423
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2424
"""Upload the initial blank content."""
2425
control_files = self._create_control_files(a_bzrdir)
2426
control_files.lock_write()
2427
transport = control_files._transport
2429
utf8_files += [('shared-storage', '')]
2431
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2432
for (filename, content_stream) in files:
2433
transport.put_file(filename, content_stream,
2434
mode=a_bzrdir._get_file_mode())
2435
for (filename, content_bytes) in utf8_files:
2436
transport.put_bytes_non_atomic(filename, content_bytes,
2437
mode=a_bzrdir._get_file_mode())
2439
control_files.unlock()
2441
def network_name(self):
2442
"""Metadir formats have matching disk and network format strings."""
2443
return self.get_format_string()
2446
# Pre-0.8 formats that don't have a disk format string (because they are
2447
# versioned by the matching control directory). We use the control directories
2448
# disk format string as a key for the network_name because they meet the
2449
# constraints (simple string, unique, immmutable).
2450
network_format_registry.register_lazy(
2451
"Bazaar-NG branch, format 5\n",
2452
'bzrlib.repofmt.weaverepo',
2453
'RepositoryFormat5',
2455
network_format_registry.register_lazy(
2456
"Bazaar-NG branch, format 6\n",
2457
'bzrlib.repofmt.weaverepo',
2458
'RepositoryFormat6',
2461
# formats which have no format string are not discoverable or independently
2462
# creatable on disk, so are not registered in format_registry. They're
2463
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2464
# needed, it's constructed directly by the BzrDir. Non-native formats where
2465
# the repository is not separately opened are similar.
2467
format_registry.register_lazy(
2468
'Bazaar-NG Repository format 7',
2469
'bzrlib.repofmt.weaverepo',
2473
format_registry.register_lazy(
2474
'Bazaar-NG Knit Repository Format 1',
2475
'bzrlib.repofmt.knitrepo',
2476
'RepositoryFormatKnit1',
2479
format_registry.register_lazy(
2480
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2481
'bzrlib.repofmt.knitrepo',
2482
'RepositoryFormatKnit3',
2485
format_registry.register_lazy(
2486
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2487
'bzrlib.repofmt.knitrepo',
2488
'RepositoryFormatKnit4',
2491
# Pack-based formats. There is one format for pre-subtrees, and one for
2492
# post-subtrees to allow ease of testing.
2493
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2494
format_registry.register_lazy(
2495
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2496
'bzrlib.repofmt.pack_repo',
2497
'RepositoryFormatKnitPack1',
2499
format_registry.register_lazy(
2500
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2501
'bzrlib.repofmt.pack_repo',
2502
'RepositoryFormatKnitPack3',
2504
format_registry.register_lazy(
2505
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2506
'bzrlib.repofmt.pack_repo',
2507
'RepositoryFormatKnitPack4',
2509
format_registry.register_lazy(
2510
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2511
'bzrlib.repofmt.pack_repo',
2512
'RepositoryFormatKnitPack5',
2514
format_registry.register_lazy(
2515
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2516
'bzrlib.repofmt.pack_repo',
2517
'RepositoryFormatKnitPack5RichRoot',
2519
format_registry.register_lazy(
2520
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2521
'bzrlib.repofmt.pack_repo',
2522
'RepositoryFormatKnitPack5RichRootBroken',
2524
format_registry.register_lazy(
2525
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2526
'bzrlib.repofmt.pack_repo',
2527
'RepositoryFormatKnitPack6',
2529
format_registry.register_lazy(
2530
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2531
'bzrlib.repofmt.pack_repo',
2532
'RepositoryFormatKnitPack6RichRoot',
2535
# Development formats.
2536
# 1.7->1.8 go below here
2537
format_registry.register_lazy(
2538
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2539
'bzrlib.repofmt.pack_repo',
2540
'RepositoryFormatPackDevelopment2',
2542
format_registry.register_lazy(
2543
("Bazaar development format 2 with subtree support "
2544
"(needs bzr.dev from before 1.8)\n"),
2545
'bzrlib.repofmt.pack_repo',
2546
'RepositoryFormatPackDevelopment2Subtree',
2550
class InterRepository(InterObject):
2551
"""This class represents operations taking place between two repositories.
2553
Its instances have methods like copy_content and fetch, and contain
2554
references to the source and target repositories these operations can be
2557
Often we will provide convenience methods on 'repository' which carry out
2558
operations with another repository - they will always forward to
2559
InterRepository.get(other).method_name(parameters).
2562
_walk_to_common_revisions_batch_size = 1
2564
"""The available optimised InterRepository types."""
2566
def __init__(self, source, target):
2567
InterObject.__init__(self, source, target)
2568
# These two attributes may be overridden by e.g. InterOtherToRemote to
2569
# provide a faster implementation.
2570
self.target_get_graph = self.target.get_graph
2571
self.target_get_parent_map = self.target.get_parent_map
2573
def copy_content(self, revision_id=None):
2574
raise NotImplementedError(self.copy_content)
2576
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2577
"""Fetch the content required to construct revision_id.
2579
The content is copied from self.source to self.target.
2581
:param revision_id: if None all content is copied, if NULL_REVISION no
2583
:param pb: optional progress bar to use for progress reports. If not
2584
provided a default one will be created.
2586
:returns: (copied_revision_count, failures).
2588
# Normally we should find a specific InterRepository subclass to do
2589
# the fetch; if nothing else then at least InterSameDataRepository.
2590
# If none of them is suitable it looks like fetching is not possible;
2591
# we try to give a good message why. _assert_same_model will probably
2592
# give a helpful message; otherwise a generic one.
2593
self._assert_same_model(self.source, self.target)
2594
raise errors.IncompatibleRepositories(self.source, self.target,
2595
"no suitableInterRepository found")
2597
def _walk_to_common_revisions(self, revision_ids):
2598
"""Walk out from revision_ids in source to revisions target has.
2600
:param revision_ids: The start point for the search.
2601
:return: A set of revision ids.
2603
target_graph = self.target_get_graph()
2604
revision_ids = frozenset(revision_ids)
2605
# Fast path for the case where all the revisions are already in the
2607
# (Although this does incur an extra round trip for the
2608
# fairly common case where the target doesn't already have the revision
2610
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2611
return graph.SearchResult(revision_ids, set(), 0, set())
2612
missing_revs = set()
2613
source_graph = self.source.get_graph()
2614
# ensure we don't pay silly lookup costs.
2615
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2616
null_set = frozenset([_mod_revision.NULL_REVISION])
2617
searcher_exhausted = False
2621
# Iterate the searcher until we have enough next_revs
2622
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2624
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2625
next_revs.update(next_revs_part)
2626
ghosts.update(ghosts_part)
2627
except StopIteration:
2628
searcher_exhausted = True
2630
# If there are ghosts in the source graph, and the caller asked for
2631
# them, make sure that they are present in the target.
2632
# We don't care about other ghosts as we can't fetch them and
2633
# haven't been asked to.
2634
ghosts_to_check = set(revision_ids.intersection(ghosts))
2635
revs_to_get = set(next_revs).union(ghosts_to_check)
2637
have_revs = set(target_graph.get_parent_map(revs_to_get))
2638
# we always have NULL_REVISION present.
2639
have_revs = have_revs.union(null_set)
2640
# Check if the target is missing any ghosts we need.
2641
ghosts_to_check.difference_update(have_revs)
2643
# One of the caller's revision_ids is a ghost in both the
2644
# source and the target.
2645
raise errors.NoSuchRevision(
2646
self.source, ghosts_to_check.pop())
2647
missing_revs.update(next_revs - have_revs)
2648
# Because we may have walked past the original stop point, make
2649
# sure everything is stopped
2650
stop_revs = searcher.find_seen_ancestors(have_revs)
2651
searcher.stop_searching_any(stop_revs)
2652
if searcher_exhausted:
2654
return searcher.get_result()
2656
@deprecated_method(one_two)
2658
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2659
"""Return the revision ids that source has that target does not.
2661
These are returned in topological order.
2663
:param revision_id: only return revision ids included by this
2665
:param find_ghosts: If True find missing revisions in deep history
2666
rather than just finding the surface difference.
2668
return list(self.search_missing_revision_ids(
2669
revision_id, find_ghosts).get_keys())
2672
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2673
"""Return the revision ids that source has that target does not.
2675
:param revision_id: only return revision ids included by this
2677
:param find_ghosts: If True find missing revisions in deep history
2678
rather than just finding the surface difference.
2679
:return: A bzrlib.graph.SearchResult.
2681
# stop searching at found target revisions.
2682
if not find_ghosts and revision_id is not None:
2683
return self._walk_to_common_revisions([revision_id])
2684
# generic, possibly worst case, slow code path.
2685
target_ids = set(self.target.all_revision_ids())
2686
if revision_id is not None:
2687
source_ids = self.source.get_ancestry(revision_id)
2688
if source_ids[0] is not None:
2689
raise AssertionError()
2692
source_ids = self.source.all_revision_ids()
2693
result_set = set(source_ids).difference(target_ids)
2694
return self.source.revision_ids_to_search_result(result_set)
2697
def _same_model(source, target):
2698
"""True if source and target have the same data representation.
2700
Note: this is always called on the base class; overriding it in a
2701
subclass will have no effect.
2704
InterRepository._assert_same_model(source, target)
2706
except errors.IncompatibleRepositories, e:
2710
def _assert_same_model(source, target):
2711
"""Raise an exception if two repositories do not use the same model.
2713
if source.supports_rich_root() != target.supports_rich_root():
2714
raise errors.IncompatibleRepositories(source, target,
2715
"different rich-root support")
2716
if source._serializer != target._serializer:
2717
raise errors.IncompatibleRepositories(source, target,
2718
"different serializers")
2721
class InterSameDataRepository(InterRepository):
2722
"""Code for converting between repositories that represent the same data.
2724
Data format and model must match for this to work.
2728
def _get_repo_format_to_test(self):
2729
"""Repository format for testing with.
2731
InterSameData can pull from subtree to subtree and from non-subtree to
2732
non-subtree, so we test this with the richest repository format.
2734
from bzrlib.repofmt import knitrepo
2735
return knitrepo.RepositoryFormatKnit3()
2738
def is_compatible(source, target):
2739
return InterRepository._same_model(source, target)
2742
def copy_content(self, revision_id=None):
2743
"""Make a complete copy of the content in self into destination.
2745
This copies both the repository's revision data, and configuration information
2746
such as the make_working_trees setting.
2748
This is a destructive operation! Do not use it on existing
2751
:param revision_id: Only copy the content needed to construct
2752
revision_id and its parents.
2755
self.target.set_make_working_trees(self.source.make_working_trees())
2756
except NotImplementedError:
2758
# but don't bother fetching if we have the needed data now.
2759
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2760
self.target.has_revision(revision_id)):
2762
self.target.fetch(self.source, revision_id=revision_id)
2765
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2766
"""See InterRepository.fetch()."""
2767
from bzrlib.fetch import RepoFetcher
2768
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2769
self.source, self.source._format, self.target,
2770
self.target._format)
2771
f = RepoFetcher(to_repository=self.target,
2772
from_repository=self.source,
2773
last_revision=revision_id,
2774
pb=pb, find_ghosts=find_ghosts)
2775
return f.count_copied, f.failed_revisions
2778
class InterWeaveRepo(InterSameDataRepository):
2779
"""Optimised code paths between Weave based repositories.
2781
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2782
implemented lazy inter-object optimisation.
2786
def _get_repo_format_to_test(self):
2787
from bzrlib.repofmt import weaverepo
2788
return weaverepo.RepositoryFormat7()
2791
def is_compatible(source, target):
2792
"""Be compatible with known Weave formats.
2794
We don't test for the stores being of specific types because that
2795
could lead to confusing results, and there is no need to be
2798
from bzrlib.repofmt.weaverepo import (
2804
return (isinstance(source._format, (RepositoryFormat5,
2806
RepositoryFormat7)) and
2807
isinstance(target._format, (RepositoryFormat5,
2809
RepositoryFormat7)))
2810
except AttributeError:
2814
def copy_content(self, revision_id=None):
2815
"""See InterRepository.copy_content()."""
2816
# weave specific optimised path:
2818
self.target.set_make_working_trees(self.source.make_working_trees())
2819
except (errors.RepositoryUpgradeRequired, NotImplemented):
2821
# FIXME do not peek!
2822
if self.source._transport.listable():
2823
pb = ui.ui_factory.nested_progress_bar()
2825
self.target.texts.insert_record_stream(
2826
self.source.texts.get_record_stream(
2827
self.source.texts.keys(), 'topological', False))
2828
pb.update('copying inventory', 0, 1)
2829
self.target.inventories.insert_record_stream(
2830
self.source.inventories.get_record_stream(
2831
self.source.inventories.keys(), 'topological', False))
2832
self.target.signatures.insert_record_stream(
2833
self.source.signatures.get_record_stream(
2834
self.source.signatures.keys(),
2836
self.target.revisions.insert_record_stream(
2837
self.source.revisions.get_record_stream(
2838
self.source.revisions.keys(),
2839
'topological', True))
2843
self.target.fetch(self.source, revision_id=revision_id)
2846
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2847
"""See InterRepository.fetch()."""
2848
from bzrlib.fetch import RepoFetcher
2849
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2850
self.source, self.source._format, self.target, self.target._format)
2851
f = RepoFetcher(to_repository=self.target,
2852
from_repository=self.source,
2853
last_revision=revision_id,
2854
pb=pb, find_ghosts=find_ghosts)
2855
return f.count_copied, f.failed_revisions
2858
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2859
"""See InterRepository.missing_revision_ids()."""
2860
# we want all revisions to satisfy revision_id in source.
2861
# but we don't want to stat every file here and there.
2862
# we want then, all revisions other needs to satisfy revision_id
2863
# checked, but not those that we have locally.
2864
# so the first thing is to get a subset of the revisions to
2865
# satisfy revision_id in source, and then eliminate those that
2866
# we do already have.
2867
# this is slow on high latency connection to self, but as as this
2868
# disk format scales terribly for push anyway due to rewriting
2869
# inventory.weave, this is considered acceptable.
2871
if revision_id is not None:
2872
source_ids = self.source.get_ancestry(revision_id)
2873
if source_ids[0] is not None:
2874
raise AssertionError()
2877
source_ids = self.source._all_possible_ids()
2878
source_ids_set = set(source_ids)
2879
# source_ids is the worst possible case we may need to pull.
2880
# now we want to filter source_ids against what we actually
2881
# have in target, but don't try to check for existence where we know
2882
# we do not have a revision as that would be pointless.
2883
target_ids = set(self.target._all_possible_ids())
2884
possibly_present_revisions = target_ids.intersection(source_ids_set)
2885
actually_present_revisions = set(
2886
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2887
required_revisions = source_ids_set.difference(actually_present_revisions)
2888
if revision_id is not None:
2889
# we used get_ancestry to determine source_ids then we are assured all
2890
# revisions referenced are present as they are installed in topological order.
2891
# and the tip revision was validated by get_ancestry.
2892
result_set = required_revisions
2894
# if we just grabbed the possibly available ids, then
2895
# we only have an estimate of whats available and need to validate
2896
# that against the revision records.
2898
self.source._eliminate_revisions_not_present(required_revisions))
2899
return self.source.revision_ids_to_search_result(result_set)
2902
class InterKnitRepo(InterSameDataRepository):
2903
"""Optimised code paths between Knit based repositories."""
2906
def _get_repo_format_to_test(self):
2907
from bzrlib.repofmt import knitrepo
2908
return knitrepo.RepositoryFormatKnit1()
2911
def is_compatible(source, target):
2912
"""Be compatible with known Knit formats.
2914
We don't test for the stores being of specific types because that
2915
could lead to confusing results, and there is no need to be
2918
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2920
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2921
isinstance(target._format, RepositoryFormatKnit))
2922
except AttributeError:
2924
return are_knits and InterRepository._same_model(source, target)
2927
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2928
"""See InterRepository.fetch()."""
2929
from bzrlib.fetch import RepoFetcher
2930
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2931
self.source, self.source._format, self.target, self.target._format)
2932
f = RepoFetcher(to_repository=self.target,
2933
from_repository=self.source,
2934
last_revision=revision_id,
2935
pb=pb, find_ghosts=find_ghosts)
2936
return f.count_copied, f.failed_revisions
2939
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2940
"""See InterRepository.missing_revision_ids()."""
2941
if revision_id is not None:
2942
source_ids = self.source.get_ancestry(revision_id)
2943
if source_ids[0] is not None:
2944
raise AssertionError()
2947
source_ids = self.source.all_revision_ids()
2948
source_ids_set = set(source_ids)
2949
# source_ids is the worst possible case we may need to pull.
2950
# now we want to filter source_ids against what we actually
2951
# have in target, but don't try to check for existence where we know
2952
# we do not have a revision as that would be pointless.
2953
target_ids = set(self.target.all_revision_ids())
2954
possibly_present_revisions = target_ids.intersection(source_ids_set)
2955
actually_present_revisions = set(
2956
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2957
required_revisions = source_ids_set.difference(actually_present_revisions)
2958
if revision_id is not None:
2959
# we used get_ancestry to determine source_ids then we are assured all
2960
# revisions referenced are present as they are installed in topological order.
2961
# and the tip revision was validated by get_ancestry.
2962
result_set = required_revisions
2964
# if we just grabbed the possibly available ids, then
2965
# we only have an estimate of whats available and need to validate
2966
# that against the revision records.
2968
self.source._eliminate_revisions_not_present(required_revisions))
2969
return self.source.revision_ids_to_search_result(result_set)
2972
class InterPackRepo(InterSameDataRepository):
2973
"""Optimised code paths between Pack based repositories."""
2976
def _get_repo_format_to_test(self):
2977
from bzrlib.repofmt import pack_repo
2978
return pack_repo.RepositoryFormatKnitPack1()
2981
def is_compatible(source, target):
2982
"""Be compatible with known Pack formats.
2984
We don't test for the stores being of specific types because that
2985
could lead to confusing results, and there is no need to be
2988
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2990
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2991
isinstance(target._format, RepositoryFormatPack))
2992
except AttributeError:
2994
return are_packs and InterRepository._same_model(source, target)
2997
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2998
"""See InterRepository.fetch()."""
2999
if (len(self.source._fallback_repositories) > 0 or
3000
len(self.target._fallback_repositories) > 0):
3001
# The pack layer is not aware of fallback repositories, so when
3002
# fetching from a stacked repository or into a stacked repository
3003
# we use the generic fetch logic which uses the VersionedFiles
3004
# attributes on repository.
3005
from bzrlib.fetch import RepoFetcher
3006
fetcher = RepoFetcher(self.target, self.source, revision_id,
3008
return fetcher.count_copied, fetcher.failed_revisions
3009
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3010
self.source, self.source._format, self.target, self.target._format)
3011
self.count_copied = 0
3012
if revision_id is None:
3014
# everything to do - use pack logic
3015
# to fetch from all packs to one without
3016
# inventory parsing etc, IFF nothing to be copied is in the target.
3018
source_revision_ids = frozenset(self.source.all_revision_ids())
3019
revision_ids = source_revision_ids - \
3020
frozenset(self.target_get_parent_map(source_revision_ids))
3021
revision_keys = [(revid,) for revid in revision_ids]
3022
target_pack_collection = self._get_target_pack_collection()
3023
index = target_pack_collection.revision_index.combined_index
3024
present_revision_ids = set(item[1][0] for item in
3025
index.iter_entries(revision_keys))
3026
revision_ids = set(revision_ids) - present_revision_ids
3027
# implementing the TODO will involve:
3028
# - detecting when all of a pack is selected
3029
# - avoiding as much as possible pre-selection, so the
3030
# more-core routines such as create_pack_from_packs can filter in
3031
# a just-in-time fashion. (though having a HEADS list on a
3032
# repository might make this a lot easier, because we could
3033
# sensibly detect 'new revisions' without doing a full index scan.
3034
elif _mod_revision.is_null(revision_id):
3039
revision_ids = self.search_missing_revision_ids(revision_id,
3040
find_ghosts=find_ghosts).get_keys()
3041
except errors.NoSuchRevision:
3042
raise errors.InstallFailed([revision_id])
3043
if len(revision_ids) == 0:
3045
return self._pack(self.source, self.target, revision_ids)
3047
def _pack(self, source, target, revision_ids):
3048
from bzrlib.repofmt.pack_repo import Packer
3049
target_pack_collection = self._get_target_pack_collection()
3050
packs = source._pack_collection.all_packs()
3051
pack = Packer(target_pack_collection, packs, '.fetch',
3052
revision_ids).pack()
3053
if pack is not None:
3054
target_pack_collection._save_pack_names()
3055
copied_revs = pack.get_revision_count()
3056
# Trigger an autopack. This may duplicate effort as we've just done
3057
# a pack creation, but for now it is simpler to think about as
3058
# 'upload data, then repack if needed'.
3060
return (copied_revs, [])
3064
def _autopack(self):
3065
self.target._pack_collection.autopack()
3067
def _get_target_pack_collection(self):
3068
return self.target._pack_collection
3071
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3072
"""See InterRepository.missing_revision_ids().
3074
:param find_ghosts: Find ghosts throughout the ancestry of
3077
if not find_ghosts and revision_id is not None:
3078
return self._walk_to_common_revisions([revision_id])
3079
elif revision_id is not None:
3080
# Find ghosts: search for revisions pointing from one repository to
3081
# the other, and vice versa, anywhere in the history of revision_id.
3082
graph = self.target_get_graph(other_repository=self.source)
3083
searcher = graph._make_breadth_first_searcher([revision_id])
3087
next_revs, ghosts = searcher.next_with_ghosts()
3088
except StopIteration:
3090
if revision_id in ghosts:
3091
raise errors.NoSuchRevision(self.source, revision_id)
3092
found_ids.update(next_revs)
3093
found_ids.update(ghosts)
3094
found_ids = frozenset(found_ids)
3095
# Double query here: should be able to avoid this by changing the
3096
# graph api further.
3097
result_set = found_ids - frozenset(
3098
self.target_get_parent_map(found_ids))
3100
source_ids = self.source.all_revision_ids()
3101
# source_ids is the worst possible case we may need to pull.
3102
# now we want to filter source_ids against what we actually
3103
# have in target, but don't try to check for existence where we know
3104
# we do not have a revision as that would be pointless.
3105
target_ids = set(self.target.all_revision_ids())
3106
result_set = set(source_ids).difference(target_ids)
3107
return self.source.revision_ids_to_search_result(result_set)
3110
class InterModel1and2(InterRepository):
3113
def _get_repo_format_to_test(self):
3117
def is_compatible(source, target):
3118
if not source.supports_rich_root() and target.supports_rich_root():
3124
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3125
"""See InterRepository.fetch()."""
3126
from bzrlib.fetch import Model1toKnit2Fetcher
3127
f = Model1toKnit2Fetcher(to_repository=self.target,
3128
from_repository=self.source,
3129
last_revision=revision_id,
3130
pb=pb, find_ghosts=find_ghosts)
3131
return f.count_copied, f.failed_revisions
3134
def copy_content(self, revision_id=None):
3135
"""Make a complete copy of the content in self into destination.
3137
This is a destructive operation! Do not use it on existing
3140
:param revision_id: Only copy the content needed to construct
3141
revision_id and its parents.
3144
self.target.set_make_working_trees(self.source.make_working_trees())
3145
except NotImplementedError:
3147
# but don't bother fetching if we have the needed data now.
3148
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
3149
self.target.has_revision(revision_id)):
3151
self.target.fetch(self.source, revision_id=revision_id)
3154
class InterKnit1and2(InterKnitRepo):
3157
def _get_repo_format_to_test(self):
3161
def is_compatible(source, target):
3162
"""Be compatible with Knit1 source and Knit3 target"""
3164
from bzrlib.repofmt.knitrepo import (
3165
RepositoryFormatKnit1,
3166
RepositoryFormatKnit3,
3168
from bzrlib.repofmt.pack_repo import (
3169
RepositoryFormatKnitPack1,
3170
RepositoryFormatKnitPack3,
3171
RepositoryFormatKnitPack4,
3172
RepositoryFormatKnitPack5,
3173
RepositoryFormatKnitPack5RichRoot,
3174
RepositoryFormatKnitPack6,
3175
RepositoryFormatKnitPack6RichRoot,
3176
RepositoryFormatPackDevelopment2,
3177
RepositoryFormatPackDevelopment2Subtree,
3180
RepositoryFormatKnit1, # no rr, no subtree
3181
RepositoryFormatKnitPack1, # no rr, no subtree
3182
RepositoryFormatPackDevelopment2, # no rr, no subtree
3183
RepositoryFormatKnitPack5, # no rr, no subtree
3184
RepositoryFormatKnitPack6, # no rr, no subtree
3187
RepositoryFormatKnit3, # rr, subtree
3188
RepositoryFormatKnitPack3, # rr, subtree
3189
RepositoryFormatKnitPack4, # rr, no subtree
3190
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
3191
RepositoryFormatKnitPack6RichRoot,# rr, no subtree
3192
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
3194
for format in norichroot:
3195
if format.rich_root_data:
3196
raise AssertionError('Format %s is a rich-root format'
3197
' but is included in the non-rich-root list'
3199
for format in richroot:
3200
if not format.rich_root_data:
3201
raise AssertionError('Format %s is not a rich-root format'
3202
' but is included in the rich-root list'
3204
# TODO: One alternative is to just check format.rich_root_data,
3205
# instead of keeping membership lists. However, the formats
3206
# *also* have to use the same 'Knit' style of storage
3207
# (line-deltas, fulltexts, etc.)
3208
return (isinstance(source._format, norichroot) and
3209
isinstance(target._format, richroot))
3210
except AttributeError:
3214
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3215
"""See InterRepository.fetch()."""
3216
from bzrlib.fetch import Knit1to2Fetcher
3217
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3218
self.source, self.source._format, self.target,
3219
self.target._format)
3220
f = Knit1to2Fetcher(to_repository=self.target,
3221
from_repository=self.source,
3222
last_revision=revision_id,
3223
pb=pb, find_ghosts=find_ghosts)
3224
return f.count_copied, f.failed_revisions
3227
class InterDifferingSerializer(InterKnitRepo):
3230
def _get_repo_format_to_test(self):
3234
def is_compatible(source, target):
3235
"""Be compatible with Knit2 source and Knit3 target"""
3236
if source.supports_rich_root() != target.supports_rich_root():
3238
# Ideally, we'd support fetching if the source had no tree references
3239
# even if it supported them...
3240
if (getattr(source, '_format.supports_tree_reference', False) and
3241
not getattr(target, '_format.supports_tree_reference', False)):
3245
def _fetch_batch(self, revision_ids, basis_id, basis_tree):
3246
"""Fetch across a few revisions.
3248
:param revision_ids: The revisions to copy
3249
:param basis_id: The revision_id of basis_tree
3250
:param basis_tree: A tree that is not in revision_ids which should
3251
already exist in the target.
3252
:return: (basis_id, basis_tree) A new basis to use now that these trees
3255
# Walk though all revisions; get inventory deltas, copy referenced
3256
# texts that delta references, insert the delta, revision and
3260
pending_revisions = []
3261
for tree in self.source.revision_trees(revision_ids):
3262
current_revision_id = tree.get_revision_id()
3263
delta = tree.inventory._make_delta(basis_tree.inventory)
3264
for old_path, new_path, file_id, entry in delta:
3265
if new_path is not None:
3266
if not (new_path or self.target.supports_rich_root()):
3267
# We leave the inventory delta in, because that
3268
# will have the deserialised inventory root
3272
# "if entry.revision == current_revision_id" ?
3273
if entry.revision == current_revision_id:
3274
text_keys.add((file_id, entry.revision))
3275
revision = self.source.get_revision(current_revision_id)
3276
pending_deltas.append((basis_id, delta,
3277
current_revision_id, revision.parent_ids))
3278
pending_revisions.append(revision)
3279
basis_id = current_revision_id
3282
from_texts = self.source.texts
3283
to_texts = self.target.texts
3284
to_texts.insert_record_stream(from_texts.get_record_stream(
3285
text_keys, self.target._fetch_order,
3286
not self.target._fetch_uses_deltas))
3288
for delta in pending_deltas:
3289
self.target.add_inventory_by_delta(*delta)
3290
# insert signatures and revisions
3291
for revision in pending_revisions:
3293
signature = self.source.get_signature_text(
3294
revision.revision_id)
3295
self.target.add_signature_text(revision.revision_id,
3297
except errors.NoSuchRevision:
3299
self.target.add_revision(revision.revision_id, revision)
3300
return basis_id, basis_tree
3302
def _fetch_all_revisions(self, revision_ids, pb):
3303
"""Fetch everything for the list of revisions.
3305
:param revision_ids: The list of revisions to fetch. Must be in
3307
:param pb: A ProgressBar
3310
basis_id, basis_tree = self._get_basis(revision_ids[0])
3312
for offset in range(0, len(revision_ids), batch_size):
3313
self.target.start_write_group()
3315
pb.update('Transferring revisions', offset,
3317
batch = revision_ids[offset:offset+batch_size]
3318
basis_id, basis_tree = self._fetch_batch(batch,
3319
basis_id, basis_tree)
3321
self.target.abort_write_group()
3324
self.target.commit_write_group()
3325
pb.update('Transferring revisions', len(revision_ids),
3329
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3330
"""See InterRepository.fetch()."""
3331
revision_ids = self.target.search_missing_revision_ids(self.source,
3332
revision_id, find_ghosts=find_ghosts).get_keys()
3333
if not revision_ids:
3335
revision_ids = tsort.topo_sort(
3336
self.source.get_graph().get_parent_map(revision_ids))
3338
my_pb = ui.ui_factory.nested_progress_bar()
3343
self._fetch_all_revisions(revision_ids, pb)
3345
if my_pb is not None:
3347
return len(revision_ids), 0
3349
def _get_basis(self, first_revision_id):
3350
"""Get a revision and tree which exists in the target.
3352
This assumes that first_revision_id is selected for transmission
3353
because all other ancestors are already present. If we can't find an
3354
ancestor we fall back to NULL_REVISION since we know that is safe.
3356
:return: (basis_id, basis_tree)
3358
first_rev = self.source.get_revision(first_revision_id)
3360
basis_id = first_rev.parent_ids[0]
3361
# only valid as a basis if the target has it
3362
self.target.get_revision(basis_id)
3363
# Try to get a basis tree - if its a ghost it will hit the
3364
# NoSuchRevision case.
3365
basis_tree = self.source.revision_tree(basis_id)
3366
except (IndexError, errors.NoSuchRevision):
3367
basis_id = _mod_revision.NULL_REVISION
3368
basis_tree = self.source.revision_tree(basis_id)
3369
return basis_id, basis_tree
3372
class InterOtherToRemote(InterRepository):
3373
"""An InterRepository that simply delegates to the 'real' InterRepository
3374
calculated for (source, target._real_repository).
3377
_walk_to_common_revisions_batch_size = 50
3379
def __init__(self, source, target):
3380
InterRepository.__init__(self, source, target)
3381
self._real_inter = None
3384
def is_compatible(source, target):
3385
if isinstance(target, remote.RemoteRepository):
3389
def _ensure_real_inter(self):
3390
if self._real_inter is None:
3391
self.target._ensure_real()
3392
real_target = self.target._real_repository
3393
self._real_inter = InterRepository.get(self.source, real_target)
3394
# Make _real_inter use the RemoteRepository for get_parent_map
3395
self._real_inter.target_get_graph = self.target.get_graph
3396
self._real_inter.target_get_parent_map = self.target.get_parent_map
3398
def copy_content(self, revision_id=None):
3399
self._ensure_real_inter()
3400
self._real_inter.copy_content(revision_id=revision_id)
3402
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3403
self._ensure_real_inter()
3404
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3405
find_ghosts=find_ghosts)
3408
def _get_repo_format_to_test(self):
3412
class InterRemoteToOther(InterRepository):
3414
def __init__(self, source, target):
3415
InterRepository.__init__(self, source, target)
3416
self._real_inter = None
3419
def is_compatible(source, target):
3420
if not isinstance(source, remote.RemoteRepository):
3422
# Is source's model compatible with target's model?
3423
source._ensure_real()
3424
real_source = source._real_repository
3425
if isinstance(real_source, remote.RemoteRepository):
3426
raise NotImplementedError(
3427
"We don't support remote repos backed by remote repos yet.")
3428
return InterRepository._same_model(real_source, target)
3430
def _ensure_real_inter(self):
3431
if self._real_inter is None:
3432
self.source._ensure_real()
3433
real_source = self.source._real_repository
3434
self._real_inter = InterRepository.get(real_source, self.target)
3436
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3437
self._ensure_real_inter()
3438
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3439
find_ghosts=find_ghosts)
3441
def copy_content(self, revision_id=None):
3442
self._ensure_real_inter()
3443
self._real_inter.copy_content(revision_id=revision_id)
3446
def _get_repo_format_to_test(self):
3451
class InterPackToRemotePack(InterPackRepo):
3452
"""A specialisation of InterPackRepo for a target that is a
3455
This will use the get_parent_map RPC rather than plain readvs, and also
3456
uses an RPC for autopacking.
3459
_walk_to_common_revisions_batch_size = 50
3462
def is_compatible(source, target):
3463
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3464
if isinstance(source._format, RepositoryFormatPack):
3465
if isinstance(target, remote.RemoteRepository):
3466
target._ensure_real()
3467
if isinstance(target._real_repository._format,
3468
RepositoryFormatPack):
3469
if InterRepository._same_model(source, target):
3473
def _autopack(self):
3474
self.target.autopack()
3476
def _get_target_pack_collection(self):
3477
return self.target._real_repository._pack_collection
3480
def _get_repo_format_to_test(self):
3484
InterRepository.register_optimiser(InterDifferingSerializer)
3485
InterRepository.register_optimiser(InterSameDataRepository)
3486
InterRepository.register_optimiser(InterWeaveRepo)
3487
InterRepository.register_optimiser(InterKnitRepo)
3488
InterRepository.register_optimiser(InterModel1and2)
3489
InterRepository.register_optimiser(InterKnit1and2)
3490
InterRepository.register_optimiser(InterPackRepo)
3491
InterRepository.register_optimiser(InterOtherToRemote)
3492
InterRepository.register_optimiser(InterRemoteToOther)
3493
InterRepository.register_optimiser(InterPackToRemotePack)
3496
class CopyConverter(object):
3497
"""A repository conversion tool which just performs a copy of the content.
3499
This is slow but quite reliable.
3502
def __init__(self, target_format):
3503
"""Create a CopyConverter.
3505
:param target_format: The format the resulting repository should be.
3507
self.target_format = target_format
3509
def convert(self, repo, pb):
3510
"""Perform the conversion of to_convert, giving feedback via pb.
3512
:param to_convert: The disk object to convert.
3513
:param pb: a progress bar to use for progress information.
3518
# this is only useful with metadir layouts - separated repo content.
3519
# trigger an assertion if not such
3520
repo._format.get_format_string()
3521
self.repo_dir = repo.bzrdir
3522
self.step('Moving repository to repository.backup')
3523
self.repo_dir.transport.move('repository', 'repository.backup')
3524
backup_transport = self.repo_dir.transport.clone('repository.backup')
3525
repo._format.check_conversion_target(self.target_format)
3526
self.source_repo = repo._format.open(self.repo_dir,
3528
_override_transport=backup_transport)
3529
self.step('Creating new repository')
3530
converted = self.target_format.initialize(self.repo_dir,
3531
self.source_repo.is_shared())
3532
converted.lock_write()
3534
self.step('Copying content into repository.')
3535
self.source_repo.copy_content_into(converted)
3538
self.step('Deleting old repository content.')
3539
self.repo_dir.transport.delete_tree('repository.backup')
3540
self.pb.note('repository converted')
3542
def step(self, message):
3543
"""Update the pb by a step."""
3545
self.pb.update(message, self.count, self.total)
3557
def _unescaper(match, _map=_unescape_map):
3558
code = match.group(1)
3562
if not code.startswith('#'):
3564
return unichr(int(code[1:])).encode('utf8')
3570
def _unescape_xml(data):
3571
"""Unescape predefined XML entities in a string of data."""
3573
if _unescape_re is None:
3574
_unescape_re = re.compile('\&([^;]*);')
3575
return _unescape_re.sub(_unescaper, data)
3578
class _VersionedFileChecker(object):
3580
def __init__(self, repository):
3581
self.repository = repository
3582
self.text_index = self.repository._generate_text_key_index()
3584
def calculate_file_version_parents(self, text_key):
3585
"""Calculate the correct parents for a file version according to
3588
parent_keys = self.text_index[text_key]
3589
if parent_keys == [_mod_revision.NULL_REVISION]:
3591
return tuple(parent_keys)
3593
def check_file_version_parents(self, texts, progress_bar=None):
3594
"""Check the parents stored in a versioned file are correct.
3596
It also detects file versions that are not referenced by their
3597
corresponding revision's inventory.
3599
:returns: A tuple of (wrong_parents, dangling_file_versions).
3600
wrong_parents is a dict mapping {revision_id: (stored_parents,
3601
correct_parents)} for each revision_id where the stored parents
3602
are not correct. dangling_file_versions is a set of (file_id,
3603
revision_id) tuples for versions that are present in this versioned
3604
file, but not used by the corresponding inventory.
3607
self.file_ids = set([file_id for file_id, _ in
3608
self.text_index.iterkeys()])
3609
# text keys is now grouped by file_id
3610
n_weaves = len(self.file_ids)
3611
files_in_revisions = {}
3612
revisions_of_files = {}
3613
n_versions = len(self.text_index)
3614
progress_bar.update('loading text store', 0, n_versions)
3615
parent_map = self.repository.texts.get_parent_map(self.text_index)
3616
# On unlistable transports this could well be empty/error...
3617
text_keys = self.repository.texts.keys()
3618
unused_keys = frozenset(text_keys) - set(self.text_index)
3619
for num, key in enumerate(self.text_index.iterkeys()):
3620
if progress_bar is not None:
3621
progress_bar.update('checking text graph', num, n_versions)
3622
correct_parents = self.calculate_file_version_parents(key)
3624
knit_parents = parent_map[key]
3625
except errors.RevisionNotPresent:
3628
if correct_parents != knit_parents:
3629
wrong_parents[key] = (knit_parents, correct_parents)
3630
return wrong_parents, unused_keys
3633
def _old_get_graph(repository, revision_id):
3634
"""DO NOT USE. That is all. I'm serious."""
3635
graph = repository.get_graph()
3636
revision_graph = dict(((key, value) for key, value in
3637
graph.iter_ancestry([revision_id]) if value is not None))
3638
return _strip_NULL_ghosts(revision_graph)
3641
def _strip_NULL_ghosts(revision_graph):
3642
"""Also don't use this. more compatibility code for unmigrated clients."""
3643
# Filter ghosts, and null:
3644
if _mod_revision.NULL_REVISION in revision_graph:
3645
del revision_graph[_mod_revision.NULL_REVISION]
3646
for key, parents in revision_graph.items():
3647
revision_graph[key] = tuple(parent for parent in parents if parent
3649
return revision_graph
3652
class StreamSink(object):
3653
"""An object that can insert a stream into a repository.
3655
This interface handles the complexity of reserialising inventories and
3656
revisions from different formats, and allows unidirectional insertion into
3657
stacked repositories without looking for the missing basis parents
3661
def __init__(self, target_repo):
3662
self.target_repo = target_repo
3664
def insert_stream(self, stream, src_format):
3665
"""Insert a stream's content into the target repository.
3667
:param src_format: a bzr repository format.
3669
:return: an iterable of keys additional items required before the
3670
insertion can be completed.
3673
to_serializer = self.target_repo._format._serializer
3674
src_serializer = src_format._serializer
3675
for substream_type, substream in stream:
3676
if substream_type == 'texts':
3677
self.target_repo.texts.insert_record_stream(substream)
3678
elif substream_type == 'inventories':
3679
if src_serializer == to_serializer:
3680
self.target_repo.inventories.insert_record_stream(
3683
self._extract_and_insert_inventories(
3684
substream, src_serializer)
3685
elif substream_type == 'revisions':
3686
# This may fallback to extract-and-insert more often than
3687
# required if the serializers are different only in terms of
3689
if src_serializer == to_serializer:
3690
self.target_repo.revisions.insert_record_stream(
3693
self._extract_and_insert_revisions(substream,
3695
elif substream_type == 'signatures':
3696
self.target_repo.signatures.insert_record_stream(substream)
3698
raise AssertionError('kaboom! %s' % (substream_type,))
3701
def _extract_and_insert_inventories(self, substream, serializer):
3702
"""Generate a new inventory versionedfile in target, converting data.
3704
The inventory is retrieved from the source, (deserializing it), and
3705
stored in the target (reserializing it in a different format).
3707
for record in substream:
3708
bytes = record.get_bytes_as('fulltext')
3709
revision_id = record.key[0]
3710
inv = serializer.read_inventory_from_string(bytes, revision_id)
3711
parents = [key[0] for key in record.parents]
3712
self.target_repo.add_inventory(revision_id, inv, parents)
3714
def _extract_and_insert_revisions(self, substream, serializer):
3715
for record in substream:
3716
bytes = record.get_bytes_as('fulltext')
3717
revision_id = record.key[0]
3718
rev = serializer.read_revision_from_string(bytes)
3719
if rev.revision_id != revision_id:
3720
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
3721
self.target_repo.add_revision(revision_id, rev)
3724
if self.target_repo._fetch_reconcile:
3725
self.target_repo.reconcile()
added
removed
bzrlib/repository.py
