1
# Copyright (C) 2005, 2006, 2007, 2008, 2009 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.decorators import needs_read_lock, needs_write_lock
51
from bzrlib.inter import InterObject
52
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
53
from bzrlib import registry
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)
641
def _check_fallback_repository(self, repository):
642
"""Check that this repository can fallback to repository safely.
644
Raise an error if not.
646
:param repository: A repository to fallback to.
648
return InterRepository._assert_same_model(self, repository)
650
def add_inventory(self, revision_id, inv, parents):
651
"""Add the inventory inv to the repository as revision_id.
653
:param parents: The revision ids of the parents that revision_id
654
is known to have and are in the repository already.
656
:returns: The validator(which is a sha1 digest, though what is sha'd is
657
repository format specific) of the serialized inventory.
659
if not self.is_in_write_group():
660
raise AssertionError("%r not in write group" % (self,))
661
_mod_revision.check_not_reserved_id(revision_id)
662
if not (inv.revision_id is None or inv.revision_id == revision_id):
663
raise AssertionError(
664
"Mismatch between inventory revision"
665
" id and insertion revid (%r, %r)"
666
% (inv.revision_id, revision_id))
668
raise AssertionError()
669
inv_lines = self._serialise_inventory_to_lines(inv)
670
return self._inventory_add_lines(revision_id, parents,
671
inv_lines, check_content=False)
673
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
675
"""Add a new inventory expressed as a delta against another revision.
677
:param basis_revision_id: The inventory id the delta was created
678
against. (This does not have to be a direct parent.)
679
:param delta: The inventory delta (see Inventory.apply_delta for
681
:param new_revision_id: The revision id that the inventory is being
683
:param parents: The revision ids of the parents that revision_id is
684
known to have and are in the repository already. These are supplied
685
for repositories that depend on the inventory graph for revision
686
graph access, as well as for those that pun ancestry with delta
689
:returns: (validator, new_inv)
690
The validator(which is a sha1 digest, though what is sha'd is
691
repository format specific) of the serialized inventory, and the
694
if not self.is_in_write_group():
695
raise AssertionError("%r not in write group" % (self,))
696
_mod_revision.check_not_reserved_id(new_revision_id)
697
basis_tree = self.revision_tree(basis_revision_id)
698
basis_tree.lock_read()
700
# Note that this mutates the inventory of basis_tree, which not all
701
# inventory implementations may support: A better idiom would be to
702
# return a new inventory, but as there is no revision tree cache in
703
# repository this is safe for now - RBC 20081013
704
basis_inv = basis_tree.inventory
705
basis_inv.apply_delta(delta)
706
basis_inv.revision_id = new_revision_id
707
return (self.add_inventory(new_revision_id, basis_inv, parents),
712
def _inventory_add_lines(self, revision_id, parents, lines,
714
"""Store lines in inv_vf and return the sha1 of the inventory."""
715
parents = [(parent,) for parent in parents]
716
return self.inventories.add_lines((revision_id,), parents, lines,
717
check_content=check_content)[0]
719
def add_revision(self, revision_id, rev, inv=None, config=None):
720
"""Add rev to the revision store as revision_id.
722
:param revision_id: the revision id to use.
723
:param rev: The revision object.
724
:param inv: The inventory for the revision. if None, it will be looked
725
up in the inventory storer
726
:param config: If None no digital signature will be created.
727
If supplied its signature_needed method will be used
728
to determine if a signature should be made.
730
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
732
_mod_revision.check_not_reserved_id(revision_id)
733
if config is not None and config.signature_needed():
735
inv = self.get_inventory(revision_id)
736
plaintext = Testament(rev, inv).as_short_text()
737
self.store_revision_signature(
738
gpg.GPGStrategy(config), plaintext, revision_id)
739
# check inventory present
740
if not self.inventories.get_parent_map([(revision_id,)]):
742
raise errors.WeaveRevisionNotPresent(revision_id,
745
# yes, this is not suitable for adding with ghosts.
746
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
750
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
751
self._add_revision(rev)
753
def _add_revision(self, revision):
754
text = self._serializer.write_revision_to_string(revision)
755
key = (revision.revision_id,)
756
parents = tuple((parent,) for parent in revision.parent_ids)
757
self.revisions.add_lines(key, parents, osutils.split_lines(text))
759
def all_revision_ids(self):
760
"""Returns a list of all the revision ids in the repository.
762
This is conceptually deprecated because code should generally work on
763
the graph reachable from a particular revision, and ignore any other
764
revisions that might be present. There is no direct replacement
767
if 'evil' in debug.debug_flags:
768
mutter_callsite(2, "all_revision_ids is linear with history.")
769
return self._all_revision_ids()
771
def _all_revision_ids(self):
772
"""Returns a list of all the revision ids in the repository.
774
These are in as much topological order as the underlying store can
777
raise NotImplementedError(self._all_revision_ids)
779
def break_lock(self):
780
"""Break a lock if one is present from another instance.
782
Uses the ui factory to ask for confirmation if the lock may be from
785
self.control_files.break_lock()
788
def _eliminate_revisions_not_present(self, revision_ids):
789
"""Check every revision id in revision_ids to see if we have it.
791
Returns a set of the present revisions.
794
graph = self.get_graph()
795
parent_map = graph.get_parent_map(revision_ids)
796
# The old API returned a list, should this actually be a set?
797
return parent_map.keys()
800
def create(a_bzrdir):
801
"""Construct the current default format repository in a_bzrdir."""
802
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
804
def __init__(self, _format, a_bzrdir, control_files):
805
"""instantiate a Repository.
807
:param _format: The format of the repository on disk.
808
:param a_bzrdir: The BzrDir of the repository.
810
In the future we will have a single api for all stores for
811
getting file texts, inventories and revisions, then
812
this construct will accept instances of those things.
814
super(Repository, self).__init__()
815
self._format = _format
816
# the following are part of the public API for Repository:
817
self.bzrdir = a_bzrdir
818
self.control_files = control_files
819
self._transport = control_files._transport
820
self.base = self._transport.base
822
self._reconcile_does_inventory_gc = True
823
self._reconcile_fixes_text_parents = False
824
self._reconcile_backsup_inventory = True
825
# not right yet - should be more semantically clear ?
827
# TODO: make sure to construct the right store classes, etc, depending
828
# on whether escaping is required.
829
self._warn_if_deprecated()
830
self._write_group = None
831
# Additional places to query for data.
832
self._fallback_repositories = []
833
# An InventoryEntry cache, used during deserialization
834
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
837
return '%s(%r)' % (self.__class__.__name__,
840
def has_same_location(self, other):
841
"""Returns a boolean indicating if this repository is at the same
842
location as another repository.
844
This might return False even when two repository objects are accessing
845
the same physical repository via different URLs.
847
if self.__class__ is not other.__class__:
849
return (self._transport.base == other._transport.base)
851
def is_in_write_group(self):
852
"""Return True if there is an open write group.
854
:seealso: start_write_group.
856
return self._write_group is not None
859
return self.control_files.is_locked()
861
def is_write_locked(self):
862
"""Return True if this object is write locked."""
863
return self.is_locked() and self.control_files._lock_mode == 'w'
865
def lock_write(self, token=None):
866
"""Lock this repository for writing.
868
This causes caching within the repository obejct to start accumlating
869
data during reads, and allows a 'write_group' to be obtained. Write
870
groups must be used for actual data insertion.
872
:param token: if this is already locked, then lock_write will fail
873
unless the token matches the existing lock.
874
:returns: a token if this instance supports tokens, otherwise None.
875
:raises TokenLockingNotSupported: when a token is given but this
876
instance doesn't support using token locks.
877
:raises MismatchedToken: if the specified token doesn't match the token
878
of the existing lock.
879
:seealso: start_write_group.
881
A token should be passed in if you know that you have locked the object
882
some other way, and need to synchronise this object's state with that
885
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
887
locked = self.is_locked()
888
result = self.control_files.lock_write(token=token)
889
for repo in self._fallback_repositories:
890
# Writes don't affect fallback repos
897
locked = self.is_locked()
898
self.control_files.lock_read()
899
for repo in self._fallback_repositories:
904
def get_physical_lock_status(self):
905
return self.control_files.get_physical_lock_status()
907
def leave_lock_in_place(self):
908
"""Tell this repository not to release the physical lock when this
911
If lock_write doesn't return a token, then this method is not supported.
913
self.control_files.leave_in_place()
915
def dont_leave_lock_in_place(self):
916
"""Tell this repository to release the physical lock when this
917
object is unlocked, even if it didn't originally acquire it.
919
If lock_write doesn't return a token, then this method is not supported.
921
self.control_files.dont_leave_in_place()
924
def gather_stats(self, revid=None, committers=None):
925
"""Gather statistics from a revision id.
927
:param revid: The revision id to gather statistics from, if None, then
928
no revision specific statistics are gathered.
929
:param committers: Optional parameter controlling whether to grab
930
a count of committers from the revision specific statistics.
931
:return: A dictionary of statistics. Currently this contains:
932
committers: The number of committers if requested.
933
firstrev: A tuple with timestamp, timezone for the penultimate left
934
most ancestor of revid, if revid is not the NULL_REVISION.
935
latestrev: A tuple with timestamp, timezone for revid, if revid is
936
not the NULL_REVISION.
937
revisions: The total revision count in the repository.
938
size: An estimate disk size of the repository in bytes.
941
if revid and committers:
942
result['committers'] = 0
943
if revid and revid != _mod_revision.NULL_REVISION:
945
all_committers = set()
946
revisions = self.get_ancestry(revid)
947
# pop the leading None
949
first_revision = None
951
# ignore the revisions in the middle - just grab first and last
952
revisions = revisions[0], revisions[-1]
953
for revision in self.get_revisions(revisions):
954
if not first_revision:
955
first_revision = revision
957
all_committers.add(revision.committer)
958
last_revision = revision
960
result['committers'] = len(all_committers)
961
result['firstrev'] = (first_revision.timestamp,
962
first_revision.timezone)
963
result['latestrev'] = (last_revision.timestamp,
964
last_revision.timezone)
966
# now gather global repository information
967
# XXX: This is available for many repos regardless of listability.
968
if self.bzrdir.root_transport.listable():
969
# XXX: do we want to __define len__() ?
970
# Maybe the versionedfiles object should provide a different
971
# method to get the number of keys.
972
result['revisions'] = len(self.revisions.keys())
976
def find_branches(self, using=False):
977
"""Find branches underneath this repository.
979
This will include branches inside other branches.
981
:param using: If True, list only branches using this repository.
983
if using and not self.is_shared():
985
return [self.bzrdir.open_branch()]
986
except errors.NotBranchError:
988
class Evaluator(object):
991
self.first_call = True
993
def __call__(self, bzrdir):
994
# On the first call, the parameter is always the bzrdir
995
# containing the current repo.
996
if not self.first_call:
998
repository = bzrdir.open_repository()
999
except errors.NoRepositoryPresent:
1002
return False, (None, repository)
1003
self.first_call = False
1005
value = (bzrdir.open_branch(), None)
1006
except errors.NotBranchError:
1007
value = (None, None)
1011
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1012
self.bzrdir.root_transport, evaluate=Evaluator()):
1013
if branch is not None:
1014
branches.append(branch)
1015
if not using and repository is not None:
1016
branches.extend(repository.find_branches())
1020
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1021
"""Return the revision ids that other has that this does not.
1023
These are returned in topological order.
1025
revision_id: only return revision ids included by revision_id.
1027
return InterRepository.get(other, self).search_missing_revision_ids(
1028
revision_id, find_ghosts)
1030
@deprecated_method(one_two)
1032
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1033
"""Return the revision ids that other has that this does not.
1035
These are returned in topological order.
1037
revision_id: only return revision ids included by revision_id.
1039
keys = self.search_missing_revision_ids(
1040
other, revision_id, find_ghosts).get_keys()
1043
parents = other.get_graph().get_parent_map(keys)
1046
return tsort.topo_sort(parents)
1050
"""Open the repository rooted at base.
1052
For instance, if the repository is at URL/.bzr/repository,
1053
Repository.open(URL) -> a Repository instance.
1055
control = bzrdir.BzrDir.open(base)
1056
return control.open_repository()
1058
def copy_content_into(self, destination, revision_id=None):
1059
"""Make a complete copy of the content in self into destination.
1061
This is a destructive operation! Do not use it on existing
1064
return InterRepository.get(self, destination).copy_content(revision_id)
1066
def commit_write_group(self):
1067
"""Commit the contents accrued within the current write group.
1069
:seealso: start_write_group.
1071
if self._write_group is not self.get_transaction():
1072
# has an unlock or relock occured ?
1073
raise errors.BzrError('mismatched lock context %r and '
1075
(self.get_transaction(), self._write_group))
1076
self._commit_write_group()
1077
self._write_group = None
1079
def _commit_write_group(self):
1080
"""Template method for per-repository write group cleanup.
1082
This is called before the write group is considered to be
1083
finished and should ensure that all data handed to the repository
1084
for writing during the write group is safely committed (to the
1085
extent possible considering file system caching etc).
1088
def suspend_write_group(self):
1089
raise errors.UnsuspendableWriteGroup(self)
1091
def refresh_data(self):
1092
"""Re-read any data needed to to synchronise with disk.
1094
This method is intended to be called after another repository instance
1095
(such as one used by a smart server) has inserted data into the
1096
repository. It may not be called during a write group, but may be
1097
called at any other time.
1099
if self.is_in_write_group():
1100
raise errors.InternalBzrError(
1101
"May not refresh_data while in a write group.")
1102
self._refresh_data()
1104
def resume_write_group(self, tokens):
1105
if not self.is_write_locked():
1106
raise errors.NotWriteLocked(self)
1107
if self._write_group:
1108
raise errors.BzrError('already in a write group')
1109
self._resume_write_group(tokens)
1110
# so we can detect unlock/relock - the write group is now entered.
1111
self._write_group = self.get_transaction()
1113
def _resume_write_group(self, tokens):
1114
raise errors.UnsuspendableWriteGroup(self)
1116
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
1118
"""Fetch the content required to construct revision_id from source.
1120
If revision_id is None and fetch_spec is None, then all content is
1123
fetch() may not be used when the repository is in a write group -
1124
either finish the current write group before using fetch, or use
1125
fetch before starting the write group.
1127
:param find_ghosts: Find and copy revisions in the source that are
1128
ghosts in the target (and not reachable directly by walking out to
1129
the first-present revision in target from revision_id).
1130
:param revision_id: If specified, all the content needed for this
1131
revision ID will be copied to the target. Fetch will determine for
1132
itself which content needs to be copied.
1133
:param fetch_spec: If specified, a SearchResult or
1134
PendingAncestryResult that describes which revisions to copy. This
1135
allows copying multiple heads at once. Mutually exclusive with
1138
if fetch_spec is not None and revision_id is not None:
1139
raise AssertionError(
1140
"fetch_spec and revision_id are mutually exclusive.")
1141
if self.is_in_write_group():
1142
raise errors.BzrError("May not fetch while in a write group.")
1143
# fast path same-url fetch operations
1144
if self.has_same_location(source) and fetch_spec is None:
1145
# check that last_revision is in 'from' and then return a
1147
if (revision_id is not None and
1148
not _mod_revision.is_null(revision_id)):
1149
self.get_revision(revision_id)
1151
# if there is no specific appropriate InterRepository, this will get
1152
# the InterRepository base class, which raises an
1153
# IncompatibleRepositories when asked to fetch.
1154
inter = InterRepository.get(source, self)
1155
return inter.fetch(revision_id=revision_id, pb=pb,
1156
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
1158
def create_bundle(self, target, base, fileobj, format=None):
1159
return serializer.write_bundle(self, target, base, fileobj, format)
1161
def get_commit_builder(self, branch, parents, config, timestamp=None,
1162
timezone=None, committer=None, revprops=None,
1164
"""Obtain a CommitBuilder for this repository.
1166
:param branch: Branch to commit to.
1167
:param parents: Revision ids of the parents of the new revision.
1168
:param config: Configuration to use.
1169
:param timestamp: Optional timestamp recorded for commit.
1170
:param timezone: Optional timezone for timestamp.
1171
:param committer: Optional committer to set for commit.
1172
:param revprops: Optional dictionary of revision properties.
1173
:param revision_id: Optional revision id.
1175
result = self._commit_builder_class(self, parents, config,
1176
timestamp, timezone, committer, revprops, revision_id)
1177
self.start_write_group()
1181
if (self.control_files._lock_count == 1 and
1182
self.control_files._lock_mode == 'w'):
1183
if self._write_group is not None:
1184
self.abort_write_group()
1185
self.control_files.unlock()
1186
raise errors.BzrError(
1187
'Must end write groups before releasing write locks.')
1188
self.control_files.unlock()
1189
if self.control_files._lock_count == 0:
1190
self._inventory_entry_cache.clear()
1191
for repo in self._fallback_repositories:
1195
def clone(self, a_bzrdir, revision_id=None):
1196
"""Clone this repository into a_bzrdir using the current format.
1198
Currently no check is made that the format of this repository and
1199
the bzrdir format are compatible. FIXME RBC 20060201.
1201
:return: The newly created destination repository.
1203
# TODO: deprecate after 0.16; cloning this with all its settings is
1204
# probably not very useful -- mbp 20070423
1205
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1206
self.copy_content_into(dest_repo, revision_id)
1209
def start_write_group(self):
1210
"""Start a write group in the repository.
1212
Write groups are used by repositories which do not have a 1:1 mapping
1213
between file ids and backend store to manage the insertion of data from
1214
both fetch and commit operations.
1216
A write lock is required around the start_write_group/commit_write_group
1217
for the support of lock-requiring repository formats.
1219
One can only insert data into a repository inside a write group.
1223
if not self.is_write_locked():
1224
raise errors.NotWriteLocked(self)
1225
if self._write_group:
1226
raise errors.BzrError('already in a write group')
1227
self._start_write_group()
1228
# so we can detect unlock/relock - the write group is now entered.
1229
self._write_group = self.get_transaction()
1231
def _start_write_group(self):
1232
"""Template method for per-repository write group startup.
1234
This is called before the write group is considered to be
1239
def sprout(self, to_bzrdir, revision_id=None):
1240
"""Create a descendent repository for new development.
1242
Unlike clone, this does not copy the settings of the repository.
1244
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1245
dest_repo.fetch(self, revision_id=revision_id)
1248
def _create_sprouting_repo(self, a_bzrdir, shared):
1249
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1250
# use target default format.
1251
dest_repo = a_bzrdir.create_repository()
1253
# Most control formats need the repository to be specifically
1254
# created, but on some old all-in-one formats it's not needed
1256
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1257
except errors.UninitializableFormat:
1258
dest_repo = a_bzrdir.open_repository()
1261
def _get_sink(self):
1262
"""Return a sink for streaming into this repository."""
1263
return StreamSink(self)
1265
def _get_source(self, to_format):
1266
"""Return a source for streaming from this repository."""
1267
return StreamSource(self, to_format)
1270
def has_revision(self, revision_id):
1271
"""True if this repository has a copy of the revision."""
1272
return revision_id in self.has_revisions((revision_id,))
1275
def has_revisions(self, revision_ids):
1276
"""Probe to find out the presence of multiple revisions.
1278
:param revision_ids: An iterable of revision_ids.
1279
:return: A set of the revision_ids that were present.
1281
parent_map = self.revisions.get_parent_map(
1282
[(rev_id,) for rev_id in revision_ids])
1284
if _mod_revision.NULL_REVISION in revision_ids:
1285
result.add(_mod_revision.NULL_REVISION)
1286
result.update([key[0] for key in parent_map])
1290
def get_revision(self, revision_id):
1291
"""Return the Revision object for a named revision."""
1292
return self.get_revisions([revision_id])[0]
1295
def get_revision_reconcile(self, revision_id):
1296
"""'reconcile' helper routine that allows access to a revision always.
1298
This variant of get_revision does not cross check the weave graph
1299
against the revision one as get_revision does: but it should only
1300
be used by reconcile, or reconcile-alike commands that are correcting
1301
or testing the revision graph.
1303
return self._get_revisions([revision_id])[0]
1306
def get_revisions(self, revision_ids):
1307
"""Get many revisions at once."""
1308
return self._get_revisions(revision_ids)
1311
def _get_revisions(self, revision_ids):
1312
"""Core work logic to get many revisions without sanity checks."""
1313
for rev_id in revision_ids:
1314
if not rev_id or not isinstance(rev_id, basestring):
1315
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1316
keys = [(key,) for key in revision_ids]
1317
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1319
for record in stream:
1320
if record.storage_kind == 'absent':
1321
raise errors.NoSuchRevision(self, record.key[0])
1322
text = record.get_bytes_as('fulltext')
1323
rev = self._serializer.read_revision_from_string(text)
1324
revs[record.key[0]] = rev
1325
return [revs[revid] for revid in revision_ids]
1328
def get_revision_xml(self, revision_id):
1329
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1330
# would have already do it.
1331
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1332
rev = self.get_revision(revision_id)
1333
rev_tmp = cStringIO.StringIO()
1334
# the current serializer..
1335
self._serializer.write_revision(rev, rev_tmp)
1337
return rev_tmp.getvalue()
1339
def get_deltas_for_revisions(self, revisions):
1340
"""Produce a generator of revision deltas.
1342
Note that the input is a sequence of REVISIONS, not revision_ids.
1343
Trees will be held in memory until the generator exits.
1344
Each delta is relative to the revision's lefthand predecessor.
1346
required_trees = set()
1347
for revision in revisions:
1348
required_trees.add(revision.revision_id)
1349
required_trees.update(revision.parent_ids[:1])
1350
trees = dict((t.get_revision_id(), t) for
1351
t in self.revision_trees(required_trees))
1352
for revision in revisions:
1353
if not revision.parent_ids:
1354
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1356
old_tree = trees[revision.parent_ids[0]]
1357
yield trees[revision.revision_id].changes_from(old_tree)
1360
def get_revision_delta(self, revision_id):
1361
"""Return the delta for one revision.
1363
The delta is relative to the left-hand predecessor of the
1366
r = self.get_revision(revision_id)
1367
return list(self.get_deltas_for_revisions([r]))[0]
1370
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1371
signature = gpg_strategy.sign(plaintext)
1372
self.add_signature_text(revision_id, signature)
1375
def add_signature_text(self, revision_id, signature):
1376
self.signatures.add_lines((revision_id,), (),
1377
osutils.split_lines(signature))
1379
def find_text_key_references(self):
1380
"""Find the text key references within the repository.
1382
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1383
revision_ids. Each altered file-ids has the exact revision_ids that
1384
altered it listed explicitly.
1385
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1386
to whether they were referred to by the inventory of the
1387
revision_id that they contain. The inventory texts from all present
1388
revision ids are assessed to generate this report.
1390
revision_keys = self.revisions.keys()
1391
w = self.inventories
1392
pb = ui.ui_factory.nested_progress_bar()
1394
return self._find_text_key_references_from_xml_inventory_lines(
1395
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1399
def _find_text_key_references_from_xml_inventory_lines(self,
1401
"""Core routine for extracting references to texts from inventories.
1403
This performs the translation of xml lines to revision ids.
1405
:param line_iterator: An iterator of lines, origin_version_id
1406
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1407
to whether they were referred to by the inventory of the
1408
revision_id that they contain. Note that if that revision_id was
1409
not part of the line_iterator's output then False will be given -
1410
even though it may actually refer to that key.
1412
if not self._serializer.support_altered_by_hack:
1413
raise AssertionError(
1414
"_find_text_key_references_from_xml_inventory_lines only "
1415
"supported for branches which store inventory as unnested xml"
1416
", not on %r" % self)
1419
# this code needs to read every new line in every inventory for the
1420
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1421
# not present in one of those inventories is unnecessary but not
1422
# harmful because we are filtering by the revision id marker in the
1423
# inventory lines : we only select file ids altered in one of those
1424
# revisions. We don't need to see all lines in the inventory because
1425
# only those added in an inventory in rev X can contain a revision=X
1427
unescape_revid_cache = {}
1428
unescape_fileid_cache = {}
1430
# jam 20061218 In a big fetch, this handles hundreds of thousands
1431
# of lines, so it has had a lot of inlining and optimizing done.
1432
# Sorry that it is a little bit messy.
1433
# Move several functions to be local variables, since this is a long
1435
search = self._file_ids_altered_regex.search
1436
unescape = _unescape_xml
1437
setdefault = result.setdefault
1438
for line, line_key in line_iterator:
1439
match = search(line)
1442
# One call to match.group() returning multiple items is quite a
1443
# bit faster than 2 calls to match.group() each returning 1
1444
file_id, revision_id = match.group('file_id', 'revision_id')
1446
# Inlining the cache lookups helps a lot when you make 170,000
1447
# lines and 350k ids, versus 8.4 unique ids.
1448
# Using a cache helps in 2 ways:
1449
# 1) Avoids unnecessary decoding calls
1450
# 2) Re-uses cached strings, which helps in future set and
1452
# (2) is enough that removing encoding entirely along with
1453
# the cache (so we are using plain strings) results in no
1454
# performance improvement.
1456
revision_id = unescape_revid_cache[revision_id]
1458
unescaped = unescape(revision_id)
1459
unescape_revid_cache[revision_id] = unescaped
1460
revision_id = unescaped
1462
# Note that unconditionally unescaping means that we deserialise
1463
# every fileid, which for general 'pull' is not great, but we don't
1464
# really want to have some many fulltexts that this matters anyway.
1467
file_id = unescape_fileid_cache[file_id]
1469
unescaped = unescape(file_id)
1470
unescape_fileid_cache[file_id] = unescaped
1473
key = (file_id, revision_id)
1474
setdefault(key, False)
1475
if revision_id == line_key[-1]:
1479
def _inventory_xml_lines_for_keys(self, keys):
1480
"""Get a line iterator of the sort needed for findind references.
1482
Not relevant for non-xml inventory repositories.
1484
Ghosts in revision_keys are ignored.
1486
:param revision_keys: The revision keys for the inventories to inspect.
1487
:return: An iterator over (inventory line, revid) for the fulltexts of
1488
all of the xml inventories specified by revision_keys.
1490
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1491
for record in stream:
1492
if record.storage_kind != 'absent':
1493
chunks = record.get_bytes_as('chunked')
1494
revid = record.key[-1]
1495
lines = osutils.chunks_to_lines(chunks)
1499
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1501
"""Helper routine for fileids_altered_by_revision_ids.
1503
This performs the translation of xml lines to revision ids.
1505
:param line_iterator: An iterator of lines, origin_version_id
1506
:param revision_ids: The revision ids to filter for. This should be a
1507
set or other type which supports efficient __contains__ lookups, as
1508
the revision id from each parsed line will be looked up in the
1509
revision_ids filter.
1510
:return: a dictionary mapping altered file-ids to an iterable of
1511
revision_ids. Each altered file-ids has the exact revision_ids that
1512
altered it listed explicitly.
1514
seen = set(self._find_text_key_references_from_xml_inventory_lines(
1515
line_iterator).iterkeys())
1516
# Note that revision_ids are revision keys.
1517
parent_maps = self.revisions.get_parent_map(revision_ids)
1519
map(parents.update, parent_maps.itervalues())
1520
parents.difference_update(revision_ids)
1521
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
1522
self._inventory_xml_lines_for_keys(parents)))
1523
new_keys = seen - parent_seen
1525
setdefault = result.setdefault
1526
for key in new_keys:
1527
setdefault(key[0], set()).add(key[-1])
1530
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1531
"""Find the file ids and versions affected by revisions.
1533
:param revisions: an iterable containing revision ids.
1534
:param _inv_weave: The inventory weave from this repository or None.
1535
If None, the inventory weave will be opened automatically.
1536
:return: a dictionary mapping altered file-ids to an iterable of
1537
revision_ids. Each altered file-ids has the exact revision_ids that
1538
altered it listed explicitly.
1540
selected_keys = set((revid,) for revid in revision_ids)
1541
w = _inv_weave or self.inventories
1542
pb = ui.ui_factory.nested_progress_bar()
1544
return self._find_file_ids_from_xml_inventory_lines(
1545
w.iter_lines_added_or_present_in_keys(
1546
selected_keys, pb=pb),
1551
def iter_files_bytes(self, desired_files):
1552
"""Iterate through file versions.
1554
Files will not necessarily be returned in the order they occur in
1555
desired_files. No specific order is guaranteed.
1557
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1558
value supplied by the caller as part of desired_files. It should
1559
uniquely identify the file version in the caller's context. (Examples:
1560
an index number or a TreeTransform trans_id.)
1562
bytes_iterator is an iterable of bytestrings for the file. The
1563
kind of iterable and length of the bytestrings are unspecified, but for
1564
this implementation, it is a list of bytes produced by
1565
VersionedFile.get_record_stream().
1567
:param desired_files: a list of (file_id, revision_id, identifier)
1571
for file_id, revision_id, callable_data in desired_files:
1572
text_keys[(file_id, revision_id)] = callable_data
1573
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1574
if record.storage_kind == 'absent':
1575
raise errors.RevisionNotPresent(record.key, self)
1576
yield text_keys[record.key], record.get_bytes_as('fulltext')
1578
def _generate_text_key_index(self, text_key_references=None,
1580
"""Generate a new text key index for the repository.
1582
This is an expensive function that will take considerable time to run.
1584
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1585
list of parents, also text keys. When a given key has no parents,
1586
the parents list will be [NULL_REVISION].
1588
# All revisions, to find inventory parents.
1589
if ancestors is None:
1590
graph = self.get_graph()
1591
ancestors = graph.get_parent_map(self.all_revision_ids())
1592
if text_key_references is None:
1593
text_key_references = self.find_text_key_references()
1594
pb = ui.ui_factory.nested_progress_bar()
1596
return self._do_generate_text_key_index(ancestors,
1597
text_key_references, pb)
1601
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1602
"""Helper for _generate_text_key_index to avoid deep nesting."""
1603
revision_order = tsort.topo_sort(ancestors)
1604
invalid_keys = set()
1606
for revision_id in revision_order:
1607
revision_keys[revision_id] = set()
1608
text_count = len(text_key_references)
1609
# a cache of the text keys to allow reuse; costs a dict of all the
1610
# keys, but saves a 2-tuple for every child of a given key.
1612
for text_key, valid in text_key_references.iteritems():
1614
invalid_keys.add(text_key)
1616
revision_keys[text_key[1]].add(text_key)
1617
text_key_cache[text_key] = text_key
1618
del text_key_references
1620
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1621
NULL_REVISION = _mod_revision.NULL_REVISION
1622
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1623
# too small for large or very branchy trees. However, for 55K path
1624
# trees, it would be easy to use too much memory trivially. Ideally we
1625
# could gauge this by looking at available real memory etc, but this is
1626
# always a tricky proposition.
1627
inventory_cache = lru_cache.LRUCache(10)
1628
batch_size = 10 # should be ~150MB on a 55K path tree
1629
batch_count = len(revision_order) / batch_size + 1
1631
pb.update("Calculating text parents", processed_texts, text_count)
1632
for offset in xrange(batch_count):
1633
to_query = revision_order[offset * batch_size:(offset + 1) *
1637
for rev_tree in self.revision_trees(to_query):
1638
revision_id = rev_tree.get_revision_id()
1639
parent_ids = ancestors[revision_id]
1640
for text_key in revision_keys[revision_id]:
1641
pb.update("Calculating text parents", processed_texts)
1642
processed_texts += 1
1643
candidate_parents = []
1644
for parent_id in parent_ids:
1645
parent_text_key = (text_key[0], parent_id)
1647
check_parent = parent_text_key not in \
1648
revision_keys[parent_id]
1650
# the parent parent_id is a ghost:
1651
check_parent = False
1652
# truncate the derived graph against this ghost.
1653
parent_text_key = None
1655
# look at the parent commit details inventories to
1656
# determine possible candidates in the per file graph.
1659
inv = inventory_cache[parent_id]
1661
inv = self.revision_tree(parent_id).inventory
1662
inventory_cache[parent_id] = inv
1663
parent_entry = inv._byid.get(text_key[0], None)
1664
if parent_entry is not None:
1666
text_key[0], parent_entry.revision)
1668
parent_text_key = None
1669
if parent_text_key is not None:
1670
candidate_parents.append(
1671
text_key_cache[parent_text_key])
1672
parent_heads = text_graph.heads(candidate_parents)
1673
new_parents = list(parent_heads)
1674
new_parents.sort(key=lambda x:candidate_parents.index(x))
1675
if new_parents == []:
1676
new_parents = [NULL_REVISION]
1677
text_index[text_key] = new_parents
1679
for text_key in invalid_keys:
1680
text_index[text_key] = [NULL_REVISION]
1683
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1684
"""Get an iterable listing the keys of all the data introduced by a set
1687
The keys will be ordered so that the corresponding items can be safely
1688
fetched and inserted in that order.
1690
:returns: An iterable producing tuples of (knit-kind, file-id,
1691
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1692
'revisions'. file-id is None unless knit-kind is 'file'.
1694
# XXX: it's a bit weird to control the inventory weave caching in this
1695
# generator. Ideally the caching would be done in fetch.py I think. Or
1696
# maybe this generator should explicitly have the contract that it
1697
# should not be iterated until the previously yielded item has been
1699
inv_w = self.inventories
1701
# file ids that changed
1702
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1704
num_file_ids = len(file_ids)
1705
for file_id, altered_versions in file_ids.iteritems():
1706
if _files_pb is not None:
1707
_files_pb.update("fetch texts", count, num_file_ids)
1709
yield ("file", file_id, altered_versions)
1710
# We're done with the files_pb. Note that it finished by the caller,
1711
# just as it was created by the caller.
1715
yield ("inventory", None, revision_ids)
1718
# XXX: Note ATM no callers actually pay attention to this return
1719
# instead they just use the list of revision ids and ignore
1720
# missing sigs. Consider removing this work entirely
1721
revisions_with_signatures = set(self.signatures.get_parent_map(
1722
[(r,) for r in revision_ids]))
1723
revisions_with_signatures = set(
1724
[r for (r,) in revisions_with_signatures])
1725
revisions_with_signatures.intersection_update(revision_ids)
1726
yield ("signatures", None, revisions_with_signatures)
1729
yield ("revisions", None, revision_ids)
1732
def get_inventory(self, revision_id):
1733
"""Get Inventory object by revision id."""
1734
return self.iter_inventories([revision_id]).next()
1736
def iter_inventories(self, revision_ids):
1737
"""Get many inventories by revision_ids.
1739
This will buffer some or all of the texts used in constructing the
1740
inventories in memory, but will only parse a single inventory at a
1743
:return: An iterator of inventories.
1745
if ((None in revision_ids)
1746
or (_mod_revision.NULL_REVISION in revision_ids)):
1747
raise ValueError('cannot get null revision inventory')
1748
return self._iter_inventories(revision_ids)
1750
def _iter_inventories(self, revision_ids):
1751
"""single-document based inventory iteration."""
1752
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1753
yield self.deserialise_inventory(revision_id, text)
1755
def _iter_inventory_xmls(self, revision_ids):
1756
keys = [(revision_id,) for revision_id in revision_ids]
1757
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1759
for record in stream:
1760
if record.storage_kind != 'absent':
1761
text_chunks[record.key] = record.get_bytes_as('chunked')
1763
raise errors.NoSuchRevision(self, record.key)
1765
chunks = text_chunks.pop(key)
1766
yield ''.join(chunks), key[-1]
1768
def deserialise_inventory(self, revision_id, xml):
1769
"""Transform the xml into an inventory object.
1771
:param revision_id: The expected revision id of the inventory.
1772
:param xml: A serialised inventory.
1774
result = self._serializer.read_inventory_from_string(xml, revision_id,
1775
entry_cache=self._inventory_entry_cache)
1776
if result.revision_id != revision_id:
1777
raise AssertionError('revision id mismatch %s != %s' % (
1778
result.revision_id, revision_id))
1781
def serialise_inventory(self, inv):
1782
return self._serializer.write_inventory_to_string(inv)
1784
def _serialise_inventory_to_lines(self, inv):
1785
return self._serializer.write_inventory_to_lines(inv)
1787
def get_serializer_format(self):
1788
return self._serializer.format_num
1791
def get_inventory_xml(self, revision_id):
1792
"""Get inventory XML as a file object."""
1793
texts = self._iter_inventory_xmls([revision_id])
1795
text, revision_id = texts.next()
1796
except StopIteration:
1797
raise errors.HistoryMissing(self, 'inventory', revision_id)
1801
def get_inventory_sha1(self, revision_id):
1802
"""Return the sha1 hash of the inventory entry
1804
return self.get_revision(revision_id).inventory_sha1
1806
def iter_reverse_revision_history(self, revision_id):
1807
"""Iterate backwards through revision ids in the lefthand history
1809
:param revision_id: The revision id to start with. All its lefthand
1810
ancestors will be traversed.
1812
graph = self.get_graph()
1813
next_id = revision_id
1815
if next_id in (None, _mod_revision.NULL_REVISION):
1818
# Note: The following line may raise KeyError in the event of
1819
# truncated history. We decided not to have a try:except:raise
1820
# RevisionNotPresent here until we see a use for it, because of the
1821
# cost in an inner loop that is by its very nature O(history).
1822
# Robert Collins 20080326
1823
parents = graph.get_parent_map([next_id])[next_id]
1824
if len(parents) == 0:
1827
next_id = parents[0]
1830
def get_revision_inventory(self, revision_id):
1831
"""Return inventory of a past revision."""
1832
# TODO: Unify this with get_inventory()
1833
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1834
# must be the same as its revision, so this is trivial.
1835
if revision_id is None:
1836
# This does not make sense: if there is no revision,
1837
# then it is the current tree inventory surely ?!
1838
# and thus get_root_id() is something that looks at the last
1839
# commit on the branch, and the get_root_id is an inventory check.
1840
raise NotImplementedError
1841
# return Inventory(self.get_root_id())
1843
return self.get_inventory(revision_id)
1845
def is_shared(self):
1846
"""Return True if this repository is flagged as a shared repository."""
1847
raise NotImplementedError(self.is_shared)
1850
def reconcile(self, other=None, thorough=False):
1851
"""Reconcile this repository."""
1852
from bzrlib.reconcile import RepoReconciler
1853
reconciler = RepoReconciler(self, thorough=thorough)
1854
reconciler.reconcile()
1857
def _refresh_data(self):
1858
"""Helper called from lock_* to ensure coherency with disk.
1860
The default implementation does nothing; it is however possible
1861
for repositories to maintain loaded indices across multiple locks
1862
by checking inside their implementation of this method to see
1863
whether their indices are still valid. This depends of course on
1864
the disk format being validatable in this manner. This method is
1865
also called by the refresh_data() public interface to cause a refresh
1866
to occur while in a write lock so that data inserted by a smart server
1867
push operation is visible on the client's instance of the physical
1872
def revision_tree(self, revision_id):
1873
"""Return Tree for a revision on this branch.
1875
`revision_id` may be NULL_REVISION for the empty tree revision.
1877
revision_id = _mod_revision.ensure_null(revision_id)
1878
# TODO: refactor this to use an existing revision object
1879
# so we don't need to read it in twice.
1880
if revision_id == _mod_revision.NULL_REVISION:
1881
return RevisionTree(self, Inventory(root_id=None),
1882
_mod_revision.NULL_REVISION)
1884
inv = self.get_revision_inventory(revision_id)
1885
return RevisionTree(self, inv, revision_id)
1887
def revision_trees(self, revision_ids):
1888
"""Return Tree for a revision on this branch.
1890
`revision_id` may not be None or 'null:'"""
1891
inventories = self.iter_inventories(revision_ids)
1892
for inv in inventories:
1893
yield RevisionTree(self, inv, inv.revision_id)
1896
def get_ancestry(self, revision_id, topo_sorted=True):
1897
"""Return a list of revision-ids integrated by a revision.
1899
The first element of the list is always None, indicating the origin
1900
revision. This might change when we have history horizons, or
1901
perhaps we should have a new API.
1903
This is topologically sorted.
1905
if _mod_revision.is_null(revision_id):
1907
if not self.has_revision(revision_id):
1908
raise errors.NoSuchRevision(self, revision_id)
1909
graph = self.get_graph()
1911
search = graph._make_breadth_first_searcher([revision_id])
1914
found, ghosts = search.next_with_ghosts()
1915
except StopIteration:
1918
if _mod_revision.NULL_REVISION in keys:
1919
keys.remove(_mod_revision.NULL_REVISION)
1921
parent_map = graph.get_parent_map(keys)
1922
keys = tsort.topo_sort(parent_map)
1923
return [None] + list(keys)
1926
"""Compress the data within the repository.
1928
This operation only makes sense for some repository types. For other
1929
types it should be a no-op that just returns.
1931
This stub method does not require a lock, but subclasses should use
1932
@needs_write_lock as this is a long running call its reasonable to
1933
implicitly lock for the user.
1937
@deprecated_method(one_six)
1938
def print_file(self, file, revision_id):
1939
"""Print `file` to stdout.
1941
FIXME RBC 20060125 as John Meinel points out this is a bad api
1942
- it writes to stdout, it assumes that that is valid etc. Fix
1943
by creating a new more flexible convenience function.
1945
tree = self.revision_tree(revision_id)
1946
# use inventory as it was in that revision
1947
file_id = tree.inventory.path2id(file)
1949
# TODO: jam 20060427 Write a test for this code path
1950
# it had a bug in it, and was raising the wrong
1952
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1953
tree.print_file(file_id)
1955
def get_transaction(self):
1956
return self.control_files.get_transaction()
1958
@deprecated_method(one_one)
1959
def get_parents(self, revision_ids):
1960
"""See StackedParentsProvider.get_parents"""
1961
parent_map = self.get_parent_map(revision_ids)
1962
return [parent_map.get(r, None) for r in revision_ids]
1964
def get_parent_map(self, revision_ids):
1965
"""See graph._StackedParentsProvider.get_parent_map"""
1966
# revisions index works in keys; this just works in revisions
1967
# therefore wrap and unwrap
1970
for revision_id in revision_ids:
1971
if revision_id == _mod_revision.NULL_REVISION:
1972
result[revision_id] = ()
1973
elif revision_id is None:
1974
raise ValueError('get_parent_map(None) is not valid')
1976
query_keys.append((revision_id ,))
1977
for ((revision_id,), parent_keys) in \
1978
self.revisions.get_parent_map(query_keys).iteritems():
1980
result[revision_id] = tuple(parent_revid
1981
for (parent_revid,) in parent_keys)
1983
result[revision_id] = (_mod_revision.NULL_REVISION,)
1986
def _make_parents_provider(self):
1989
def get_graph(self, other_repository=None):
1990
"""Return the graph walker for this repository format"""
1991
parents_provider = self._make_parents_provider()
1992
if (other_repository is not None and
1993
not self.has_same_location(other_repository)):
1994
parents_provider = graph._StackedParentsProvider(
1995
[parents_provider, other_repository._make_parents_provider()])
1996
return graph.Graph(parents_provider)
1998
def _get_versioned_file_checker(self):
1999
"""Return an object suitable for checking versioned files."""
2000
return _VersionedFileChecker(self)
2002
def revision_ids_to_search_result(self, result_set):
2003
"""Convert a set of revision ids to a graph SearchResult."""
2004
result_parents = set()
2005
for parents in self.get_graph().get_parent_map(
2006
result_set).itervalues():
2007
result_parents.update(parents)
2008
included_keys = result_set.intersection(result_parents)
2009
start_keys = result_set.difference(included_keys)
2010
exclude_keys = result_parents.difference(result_set)
2011
result = graph.SearchResult(start_keys, exclude_keys,
2012
len(result_set), result_set)
2016
def set_make_working_trees(self, new_value):
2017
"""Set the policy flag for making working trees when creating branches.
2019
This only applies to branches that use this repository.
2021
The default is 'True'.
2022
:param new_value: True to restore the default, False to disable making
2025
raise NotImplementedError(self.set_make_working_trees)
2027
def make_working_trees(self):
2028
"""Returns the policy for making working trees on new branches."""
2029
raise NotImplementedError(self.make_working_trees)
2032
def sign_revision(self, revision_id, gpg_strategy):
2033
plaintext = Testament.from_revision(self, revision_id).as_short_text()
2034
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2037
def has_signature_for_revision_id(self, revision_id):
2038
"""Query for a revision signature for revision_id in the repository."""
2039
if not self.has_revision(revision_id):
2040
raise errors.NoSuchRevision(self, revision_id)
2041
sig_present = (1 == len(
2042
self.signatures.get_parent_map([(revision_id,)])))
2046
def get_signature_text(self, revision_id):
2047
"""Return the text for a signature."""
2048
stream = self.signatures.get_record_stream([(revision_id,)],
2050
record = stream.next()
2051
if record.storage_kind == 'absent':
2052
raise errors.NoSuchRevision(self, revision_id)
2053
return record.get_bytes_as('fulltext')
2056
def check(self, revision_ids=None):
2057
"""Check consistency of all history of given revision_ids.
2059
Different repository implementations should override _check().
2061
:param revision_ids: A non-empty list of revision_ids whose ancestry
2062
will be checked. Typically the last revision_id of a branch.
2064
return self._check(revision_ids)
2066
def _check(self, revision_ids):
2067
result = check.Check(self)
2071
def _warn_if_deprecated(self):
2072
global _deprecation_warning_done
2073
if _deprecation_warning_done:
2075
_deprecation_warning_done = True
2076
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2077
% (self._format, self.bzrdir.transport.base))
2079
def supports_rich_root(self):
2080
return self._format.rich_root_data
2082
def _check_ascii_revisionid(self, revision_id, method):
2083
"""Private helper for ascii-only repositories."""
2084
# weave repositories refuse to store revisionids that are non-ascii.
2085
if revision_id is not None:
2086
# weaves require ascii revision ids.
2087
if isinstance(revision_id, unicode):
2089
revision_id.encode('ascii')
2090
except UnicodeEncodeError:
2091
raise errors.NonAsciiRevisionId(method, self)
2094
revision_id.decode('ascii')
2095
except UnicodeDecodeError:
2096
raise errors.NonAsciiRevisionId(method, self)
2098
def revision_graph_can_have_wrong_parents(self):
2099
"""Is it possible for this repository to have a revision graph with
2102
If True, then this repository must also implement
2103
_find_inconsistent_revision_parents so that check and reconcile can
2104
check for inconsistencies before proceeding with other checks that may
2105
depend on the revision index being consistent.
2107
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2110
# remove these delegates a while after bzr 0.15
2111
def __make_delegated(name, from_module):
2112
def _deprecated_repository_forwarder():
2113
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2114
% (name, from_module),
2117
m = __import__(from_module, globals(), locals(), [name])
2119
return getattr(m, name)
2120
except AttributeError:
2121
raise AttributeError('module %s has no name %s'
2123
globals()[name] = _deprecated_repository_forwarder
2126
'AllInOneRepository',
2127
'WeaveMetaDirRepository',
2128
'PreSplitOutRepositoryFormat',
2129
'RepositoryFormat4',
2130
'RepositoryFormat5',
2131
'RepositoryFormat6',
2132
'RepositoryFormat7',
2134
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2138
'RepositoryFormatKnit',
2139
'RepositoryFormatKnit1',
2141
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2144
def install_revision(repository, rev, revision_tree):
2145
"""Install all revision data into a repository."""
2146
install_revisions(repository, [(rev, revision_tree, None)])
2149
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2150
"""Install all revision data into a repository.
2152
Accepts an iterable of revision, tree, signature tuples. The signature
2155
repository.start_write_group()
2157
for n, (revision, revision_tree, signature) in enumerate(iterable):
2158
_install_revision(repository, revision, revision_tree, signature)
2160
pb.update('Transferring revisions', n + 1, num_revisions)
2162
repository.abort_write_group()
2165
repository.commit_write_group()
2168
def _install_revision(repository, rev, revision_tree, signature):
2169
"""Install all revision data into a repository."""
2170
present_parents = []
2172
for p_id in rev.parent_ids:
2173
if repository.has_revision(p_id):
2174
present_parents.append(p_id)
2175
parent_trees[p_id] = repository.revision_tree(p_id)
2177
parent_trees[p_id] = repository.revision_tree(
2178
_mod_revision.NULL_REVISION)
2180
inv = revision_tree.inventory
2181
entries = inv.iter_entries()
2182
# backwards compatibility hack: skip the root id.
2183
if not repository.supports_rich_root():
2184
path, root = entries.next()
2185
if root.revision != rev.revision_id:
2186
raise errors.IncompatibleRevision(repr(repository))
2188
for path, ie in entries:
2189
text_keys[(ie.file_id, ie.revision)] = ie
2190
text_parent_map = repository.texts.get_parent_map(text_keys)
2191
missing_texts = set(text_keys) - set(text_parent_map)
2192
# Add the texts that are not already present
2193
for text_key in missing_texts:
2194
ie = text_keys[text_key]
2196
# FIXME: TODO: The following loop overlaps/duplicates that done by
2197
# commit to determine parents. There is a latent/real bug here where
2198
# the parents inserted are not those commit would do - in particular
2199
# they are not filtered by heads(). RBC, AB
2200
for revision, tree in parent_trees.iteritems():
2201
if ie.file_id not in tree:
2203
parent_id = tree.inventory[ie.file_id].revision
2204
if parent_id in text_parents:
2206
text_parents.append((ie.file_id, parent_id))
2207
lines = revision_tree.get_file(ie.file_id).readlines()
2208
repository.texts.add_lines(text_key, text_parents, lines)
2210
# install the inventory
2211
repository.add_inventory(rev.revision_id, inv, present_parents)
2212
except errors.RevisionAlreadyPresent:
2214
if signature is not None:
2215
repository.add_signature_text(rev.revision_id, signature)
2216
repository.add_revision(rev.revision_id, rev, inv)
2219
class MetaDirRepository(Repository):
2220
"""Repositories in the new meta-dir layout.
2222
:ivar _transport: Transport for access to repository control files,
2223
typically pointing to .bzr/repository.
2226
def __init__(self, _format, a_bzrdir, control_files):
2227
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2228
self._transport = control_files._transport
2230
def is_shared(self):
2231
"""Return True if this repository is flagged as a shared repository."""
2232
return self._transport.has('shared-storage')
2235
def set_make_working_trees(self, new_value):
2236
"""Set the policy flag for making working trees when creating branches.
2238
This only applies to branches that use this repository.
2240
The default is 'True'.
2241
:param new_value: True to restore the default, False to disable making
2246
self._transport.delete('no-working-trees')
2247
except errors.NoSuchFile:
2250
self._transport.put_bytes('no-working-trees', '',
2251
mode=self.bzrdir._get_file_mode())
2253
def make_working_trees(self):
2254
"""Returns the policy for making working trees on new branches."""
2255
return not self._transport.has('no-working-trees')
2258
class MetaDirVersionedFileRepository(MetaDirRepository):
2259
"""Repositories in a meta-dir, that work via versioned file objects."""
2261
def __init__(self, _format, a_bzrdir, control_files):
2262
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2266
network_format_registry = registry.FormatRegistry()
2267
"""Registry of formats indexed by their network name.
2269
The network name for a repository format is an identifier that can be used when
2270
referring to formats with smart server operations. See
2271
RepositoryFormat.network_name() for more detail.
2275
format_registry = registry.FormatRegistry(network_format_registry)
2276
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2278
This can contain either format instances themselves, or classes/factories that
2279
can be called to obtain one.
2283
#####################################################################
2284
# Repository Formats
2286
class RepositoryFormat(object):
2287
"""A repository format.
2289
Formats provide four things:
2290
* An initialization routine to construct repository data on disk.
2291
* a optional format string which is used when the BzrDir supports
2293
* an open routine which returns a Repository instance.
2294
* A network name for referring to the format in smart server RPC
2297
There is one and only one Format subclass for each on-disk format. But
2298
there can be one Repository subclass that is used for several different
2299
formats. The _format attribute on a Repository instance can be used to
2300
determine the disk format.
2302
Formats are placed in a registry by their format string for reference
2303
during opening. These should be subclasses of RepositoryFormat for
2306
Once a format is deprecated, just deprecate the initialize and open
2307
methods on the format class. Do not deprecate the object, as the
2308
object may be created even when a repository instnace hasn't been
2311
Common instance attributes:
2312
_matchingbzrdir - the bzrdir format that the repository format was
2313
originally written to work with. This can be used if manually
2314
constructing a bzrdir and repository, or more commonly for test suite
2318
# Set to True or False in derived classes. True indicates that the format
2319
# supports ghosts gracefully.
2320
supports_ghosts = None
2321
# Can this repository be given external locations to lookup additional
2322
# data. Set to True or False in derived classes.
2323
supports_external_lookups = None
2324
# What order should fetch operations request streams in?
2325
# The default is unordered as that is the cheapest for an origin to
2327
_fetch_order = 'unordered'
2328
# Does this repository format use deltas that can be fetched as-deltas ?
2329
# (E.g. knits, where the knit deltas can be transplanted intact.
2330
# We default to False, which will ensure that enough data to get
2331
# a full text out of any fetch stream will be grabbed.
2332
_fetch_uses_deltas = False
2333
# Should fetch trigger a reconcile after the fetch? Only needed for
2334
# some repository formats that can suffer internal inconsistencies.
2335
_fetch_reconcile = False
2338
return "<%s>" % self.__class__.__name__
2340
def __eq__(self, other):
2341
# format objects are generally stateless
2342
return isinstance(other, self.__class__)
2344
def __ne__(self, other):
2345
return not self == other
2348
def find_format(klass, a_bzrdir):
2349
"""Return the format for the repository object in a_bzrdir.
2351
This is used by bzr native formats that have a "format" file in
2352
the repository. Other methods may be used by different types of
2356
transport = a_bzrdir.get_repository_transport(None)
2357
format_string = transport.get("format").read()
2358
return format_registry.get(format_string)
2359
except errors.NoSuchFile:
2360
raise errors.NoRepositoryPresent(a_bzrdir)
2362
raise errors.UnknownFormatError(format=format_string,
2366
def register_format(klass, format):
2367
format_registry.register(format.get_format_string(), format)
2370
def unregister_format(klass, format):
2371
format_registry.remove(format.get_format_string())
2374
def get_default_format(klass):
2375
"""Return the current default format."""
2376
from bzrlib import bzrdir
2377
return bzrdir.format_registry.make_bzrdir('default').repository_format
2379
def get_format_string(self):
2380
"""Return the ASCII format string that identifies this format.
2382
Note that in pre format ?? repositories the format string is
2383
not permitted nor written to disk.
2385
raise NotImplementedError(self.get_format_string)
2387
def get_format_description(self):
2388
"""Return the short description for this format."""
2389
raise NotImplementedError(self.get_format_description)
2391
# TODO: this shouldn't be in the base class, it's specific to things that
2392
# use weaves or knits -- mbp 20070207
2393
def _get_versioned_file_store(self,
2398
versionedfile_class=None,
2399
versionedfile_kwargs={},
2401
if versionedfile_class is None:
2402
versionedfile_class = self._versionedfile_class
2403
weave_transport = control_files._transport.clone(name)
2404
dir_mode = control_files._dir_mode
2405
file_mode = control_files._file_mode
2406
return VersionedFileStore(weave_transport, prefixed=prefixed,
2408
file_mode=file_mode,
2409
versionedfile_class=versionedfile_class,
2410
versionedfile_kwargs=versionedfile_kwargs,
2413
def initialize(self, a_bzrdir, shared=False):
2414
"""Initialize a repository of this format in a_bzrdir.
2416
:param a_bzrdir: The bzrdir to put the new repository in it.
2417
:param shared: The repository should be initialized as a sharable one.
2418
:returns: The new repository object.
2420
This may raise UninitializableFormat if shared repository are not
2421
compatible the a_bzrdir.
2423
raise NotImplementedError(self.initialize)
2425
def is_supported(self):
2426
"""Is this format supported?
2428
Supported formats must be initializable and openable.
2429
Unsupported formats may not support initialization or committing or
2430
some other features depending on the reason for not being supported.
2434
def network_name(self):
2435
"""A simple byte string uniquely identifying this format for RPC calls.
2437
MetaDir repository formats use their disk format string to identify the
2438
repository over the wire. All in one formats such as bzr < 0.8, and
2439
foreign formats like svn/git and hg should use some marker which is
2440
unique and immutable.
2442
raise NotImplementedError(self.network_name)
2444
def check_conversion_target(self, target_format):
2445
raise NotImplementedError(self.check_conversion_target)
2447
def open(self, a_bzrdir, _found=False):
2448
"""Return an instance of this format for the bzrdir a_bzrdir.
2450
_found is a private parameter, do not use it.
2452
raise NotImplementedError(self.open)
2455
class MetaDirRepositoryFormat(RepositoryFormat):
2456
"""Common base class for the new repositories using the metadir layout."""
2458
rich_root_data = False
2459
supports_tree_reference = False
2460
supports_external_lookups = False
2463
def _matchingbzrdir(self):
2464
matching = bzrdir.BzrDirMetaFormat1()
2465
matching.repository_format = self
2469
super(MetaDirRepositoryFormat, self).__init__()
2471
def _create_control_files(self, a_bzrdir):
2472
"""Create the required files and the initial control_files object."""
2473
# FIXME: RBC 20060125 don't peek under the covers
2474
# NB: no need to escape relative paths that are url safe.
2475
repository_transport = a_bzrdir.get_repository_transport(self)
2476
control_files = lockable_files.LockableFiles(repository_transport,
2477
'lock', lockdir.LockDir)
2478
control_files.create_lock()
2479
return control_files
2481
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2482
"""Upload the initial blank content."""
2483
control_files = self._create_control_files(a_bzrdir)
2484
control_files.lock_write()
2485
transport = control_files._transport
2487
utf8_files += [('shared-storage', '')]
2489
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2490
for (filename, content_stream) in files:
2491
transport.put_file(filename, content_stream,
2492
mode=a_bzrdir._get_file_mode())
2493
for (filename, content_bytes) in utf8_files:
2494
transport.put_bytes_non_atomic(filename, content_bytes,
2495
mode=a_bzrdir._get_file_mode())
2497
control_files.unlock()
2499
def network_name(self):
2500
"""Metadir formats have matching disk and network format strings."""
2501
return self.get_format_string()
2504
# Pre-0.8 formats that don't have a disk format string (because they are
2505
# versioned by the matching control directory). We use the control directories
2506
# disk format string as a key for the network_name because they meet the
2507
# constraints (simple string, unique, immmutable).
2508
network_format_registry.register_lazy(
2509
"Bazaar-NG branch, format 5\n",
2510
'bzrlib.repofmt.weaverepo',
2511
'RepositoryFormat5',
2513
network_format_registry.register_lazy(
2514
"Bazaar-NG branch, format 6\n",
2515
'bzrlib.repofmt.weaverepo',
2516
'RepositoryFormat6',
2519
# formats which have no format string are not discoverable or independently
2520
# creatable on disk, so are not registered in format_registry. They're
2521
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2522
# needed, it's constructed directly by the BzrDir. Non-native formats where
2523
# the repository is not separately opened are similar.
2525
format_registry.register_lazy(
2526
'Bazaar-NG Repository format 7',
2527
'bzrlib.repofmt.weaverepo',
2531
format_registry.register_lazy(
2532
'Bazaar-NG Knit Repository Format 1',
2533
'bzrlib.repofmt.knitrepo',
2534
'RepositoryFormatKnit1',
2537
format_registry.register_lazy(
2538
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2539
'bzrlib.repofmt.knitrepo',
2540
'RepositoryFormatKnit3',
2543
format_registry.register_lazy(
2544
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2545
'bzrlib.repofmt.knitrepo',
2546
'RepositoryFormatKnit4',
2549
# Pack-based formats. There is one format for pre-subtrees, and one for
2550
# post-subtrees to allow ease of testing.
2551
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2552
format_registry.register_lazy(
2553
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2554
'bzrlib.repofmt.pack_repo',
2555
'RepositoryFormatKnitPack1',
2557
format_registry.register_lazy(
2558
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2559
'bzrlib.repofmt.pack_repo',
2560
'RepositoryFormatKnitPack3',
2562
format_registry.register_lazy(
2563
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2564
'bzrlib.repofmt.pack_repo',
2565
'RepositoryFormatKnitPack4',
2567
format_registry.register_lazy(
2568
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2569
'bzrlib.repofmt.pack_repo',
2570
'RepositoryFormatKnitPack5',
2572
format_registry.register_lazy(
2573
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2574
'bzrlib.repofmt.pack_repo',
2575
'RepositoryFormatKnitPack5RichRoot',
2577
format_registry.register_lazy(
2578
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2579
'bzrlib.repofmt.pack_repo',
2580
'RepositoryFormatKnitPack5RichRootBroken',
2582
format_registry.register_lazy(
2583
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2584
'bzrlib.repofmt.pack_repo',
2585
'RepositoryFormatKnitPack6',
2587
format_registry.register_lazy(
2588
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2589
'bzrlib.repofmt.pack_repo',
2590
'RepositoryFormatKnitPack6RichRoot',
2593
# Development formats.
2594
# 1.7->1.8 go below here
2595
format_registry.register_lazy(
2596
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2597
'bzrlib.repofmt.pack_repo',
2598
'RepositoryFormatPackDevelopment2',
2600
format_registry.register_lazy(
2601
("Bazaar development format 2 with subtree support "
2602
"(needs bzr.dev from before 1.8)\n"),
2603
'bzrlib.repofmt.pack_repo',
2604
'RepositoryFormatPackDevelopment2Subtree',
2608
class InterRepository(InterObject):
2609
"""This class represents operations taking place between two repositories.
2611
Its instances have methods like copy_content and fetch, and contain
2612
references to the source and target repositories these operations can be
2615
Often we will provide convenience methods on 'repository' which carry out
2616
operations with another repository - they will always forward to
2617
InterRepository.get(other).method_name(parameters).
2620
_walk_to_common_revisions_batch_size = 1
2622
"""The available optimised InterRepository types."""
2624
def __init__(self, source, target):
2625
InterObject.__init__(self, source, target)
2626
# These two attributes may be overridden by e.g. InterOtherToRemote to
2627
# provide a faster implementation.
2628
self.target_get_graph = self.target.get_graph
2629
self.target_get_parent_map = self.target.get_parent_map
2632
def copy_content(self, revision_id=None):
2633
"""Make a complete copy of the content in self into destination.
2635
This is a destructive operation! Do not use it on existing
2638
:param revision_id: Only copy the content needed to construct
2639
revision_id and its parents.
2642
self.target.set_make_working_trees(self.source.make_working_trees())
2643
except NotImplementedError:
2645
self.target.fetch(self.source, revision_id=revision_id)
2648
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2650
"""Fetch the content required to construct revision_id.
2652
The content is copied from self.source to self.target.
2654
:param revision_id: if None all content is copied, if NULL_REVISION no
2656
:param pb: optional progress bar to use for progress reports. If not
2657
provided a default one will be created.
2660
from bzrlib.fetch import RepoFetcher
2661
f = RepoFetcher(to_repository=self.target,
2662
from_repository=self.source,
2663
last_revision=revision_id,
2664
fetch_spec=fetch_spec,
2665
pb=pb, find_ghosts=find_ghosts)
2667
def _walk_to_common_revisions(self, revision_ids):
2668
"""Walk out from revision_ids in source to revisions target has.
2670
:param revision_ids: The start point for the search.
2671
:return: A set of revision ids.
2673
target_graph = self.target_get_graph()
2674
revision_ids = frozenset(revision_ids)
2675
# Fast path for the case where all the revisions are already in the
2677
# (Although this does incur an extra round trip for the
2678
# fairly common case where the target doesn't already have the revision
2680
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2681
return graph.SearchResult(revision_ids, set(), 0, set())
2682
missing_revs = set()
2683
source_graph = self.source.get_graph()
2684
# ensure we don't pay silly lookup costs.
2685
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2686
null_set = frozenset([_mod_revision.NULL_REVISION])
2687
searcher_exhausted = False
2691
# Iterate the searcher until we have enough next_revs
2692
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2694
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2695
next_revs.update(next_revs_part)
2696
ghosts.update(ghosts_part)
2697
except StopIteration:
2698
searcher_exhausted = True
2700
# If there are ghosts in the source graph, and the caller asked for
2701
# them, make sure that they are present in the target.
2702
# We don't care about other ghosts as we can't fetch them and
2703
# haven't been asked to.
2704
ghosts_to_check = set(revision_ids.intersection(ghosts))
2705
revs_to_get = set(next_revs).union(ghosts_to_check)
2707
have_revs = set(target_graph.get_parent_map(revs_to_get))
2708
# we always have NULL_REVISION present.
2709
have_revs = have_revs.union(null_set)
2710
# Check if the target is missing any ghosts we need.
2711
ghosts_to_check.difference_update(have_revs)
2713
# One of the caller's revision_ids is a ghost in both the
2714
# source and the target.
2715
raise errors.NoSuchRevision(
2716
self.source, ghosts_to_check.pop())
2717
missing_revs.update(next_revs - have_revs)
2718
# Because we may have walked past the original stop point, make
2719
# sure everything is stopped
2720
stop_revs = searcher.find_seen_ancestors(have_revs)
2721
searcher.stop_searching_any(stop_revs)
2722
if searcher_exhausted:
2724
return searcher.get_result()
2726
@deprecated_method(one_two)
2728
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2729
"""Return the revision ids that source has that target does not.
2731
These are returned in topological order.
2733
:param revision_id: only return revision ids included by this
2735
:param find_ghosts: If True find missing revisions in deep history
2736
rather than just finding the surface difference.
2738
return list(self.search_missing_revision_ids(
2739
revision_id, find_ghosts).get_keys())
2742
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2743
"""Return the revision ids that source has that target does not.
2745
:param revision_id: only return revision ids included by this
2747
:param find_ghosts: If True find missing revisions in deep history
2748
rather than just finding the surface difference.
2749
:return: A bzrlib.graph.SearchResult.
2751
# stop searching at found target revisions.
2752
if not find_ghosts and revision_id is not None:
2753
return self._walk_to_common_revisions([revision_id])
2754
# generic, possibly worst case, slow code path.
2755
target_ids = set(self.target.all_revision_ids())
2756
if revision_id is not None:
2757
source_ids = self.source.get_ancestry(revision_id)
2758
if source_ids[0] is not None:
2759
raise AssertionError()
2762
source_ids = self.source.all_revision_ids()
2763
result_set = set(source_ids).difference(target_ids)
2764
return self.source.revision_ids_to_search_result(result_set)
2767
def _same_model(source, target):
2768
"""True if source and target have the same data representation.
2770
Note: this is always called on the base class; overriding it in a
2771
subclass will have no effect.
2774
InterRepository._assert_same_model(source, target)
2776
except errors.IncompatibleRepositories, e:
2780
def _assert_same_model(source, target):
2781
"""Raise an exception if two repositories do not use the same model.
2783
if source.supports_rich_root() != target.supports_rich_root():
2784
raise errors.IncompatibleRepositories(source, target,
2785
"different rich-root support")
2786
if source._serializer != target._serializer:
2787
raise errors.IncompatibleRepositories(source, target,
2788
"different serializers")
2791
class InterSameDataRepository(InterRepository):
2792
"""Code for converting between repositories that represent the same data.
2794
Data format and model must match for this to work.
2798
def _get_repo_format_to_test(self):
2799
"""Repository format for testing with.
2801
InterSameData can pull from subtree to subtree and from non-subtree to
2802
non-subtree, so we test this with the richest repository format.
2804
from bzrlib.repofmt import knitrepo
2805
return knitrepo.RepositoryFormatKnit3()
2808
def is_compatible(source, target):
2809
return InterRepository._same_model(source, target)
2812
class InterWeaveRepo(InterSameDataRepository):
2813
"""Optimised code paths between Weave based repositories.
2815
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2816
implemented lazy inter-object optimisation.
2820
def _get_repo_format_to_test(self):
2821
from bzrlib.repofmt import weaverepo
2822
return weaverepo.RepositoryFormat7()
2825
def is_compatible(source, target):
2826
"""Be compatible with known Weave formats.
2828
We don't test for the stores being of specific types because that
2829
could lead to confusing results, and there is no need to be
2832
from bzrlib.repofmt.weaverepo import (
2838
return (isinstance(source._format, (RepositoryFormat5,
2840
RepositoryFormat7)) and
2841
isinstance(target._format, (RepositoryFormat5,
2843
RepositoryFormat7)))
2844
except AttributeError:
2848
def copy_content(self, revision_id=None):
2849
"""See InterRepository.copy_content()."""
2850
# weave specific optimised path:
2852
self.target.set_make_working_trees(self.source.make_working_trees())
2853
except (errors.RepositoryUpgradeRequired, NotImplemented):
2855
# FIXME do not peek!
2856
if self.source._transport.listable():
2857
pb = ui.ui_factory.nested_progress_bar()
2859
self.target.texts.insert_record_stream(
2860
self.source.texts.get_record_stream(
2861
self.source.texts.keys(), 'topological', False))
2862
pb.update('copying inventory', 0, 1)
2863
self.target.inventories.insert_record_stream(
2864
self.source.inventories.get_record_stream(
2865
self.source.inventories.keys(), 'topological', False))
2866
self.target.signatures.insert_record_stream(
2867
self.source.signatures.get_record_stream(
2868
self.source.signatures.keys(),
2870
self.target.revisions.insert_record_stream(
2871
self.source.revisions.get_record_stream(
2872
self.source.revisions.keys(),
2873
'topological', True))
2877
self.target.fetch(self.source, revision_id=revision_id)
2880
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2881
"""See InterRepository.missing_revision_ids()."""
2882
# we want all revisions to satisfy revision_id in source.
2883
# but we don't want to stat every file here and there.
2884
# we want then, all revisions other needs to satisfy revision_id
2885
# checked, but not those that we have locally.
2886
# so the first thing is to get a subset of the revisions to
2887
# satisfy revision_id in source, and then eliminate those that
2888
# we do already have.
2889
# this is slow on high latency connection to self, but as as this
2890
# disk format scales terribly for push anyway due to rewriting
2891
# inventory.weave, this is considered acceptable.
2893
if revision_id is not None:
2894
source_ids = self.source.get_ancestry(revision_id)
2895
if source_ids[0] is not None:
2896
raise AssertionError()
2899
source_ids = self.source._all_possible_ids()
2900
source_ids_set = set(source_ids)
2901
# source_ids is the worst possible case we may need to pull.
2902
# now we want to filter source_ids against what we actually
2903
# have in target, but don't try to check for existence where we know
2904
# we do not have a revision as that would be pointless.
2905
target_ids = set(self.target._all_possible_ids())
2906
possibly_present_revisions = target_ids.intersection(source_ids_set)
2907
actually_present_revisions = set(
2908
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2909
required_revisions = source_ids_set.difference(actually_present_revisions)
2910
if revision_id is not None:
2911
# we used get_ancestry to determine source_ids then we are assured all
2912
# revisions referenced are present as they are installed in topological order.
2913
# and the tip revision was validated by get_ancestry.
2914
result_set = required_revisions
2916
# if we just grabbed the possibly available ids, then
2917
# we only have an estimate of whats available and need to validate
2918
# that against the revision records.
2920
self.source._eliminate_revisions_not_present(required_revisions))
2921
return self.source.revision_ids_to_search_result(result_set)
2924
class InterKnitRepo(InterSameDataRepository):
2925
"""Optimised code paths between Knit based repositories."""
2928
def _get_repo_format_to_test(self):
2929
from bzrlib.repofmt import knitrepo
2930
return knitrepo.RepositoryFormatKnit1()
2933
def is_compatible(source, target):
2934
"""Be compatible with known Knit formats.
2936
We don't test for the stores being of specific types because that
2937
could lead to confusing results, and there is no need to be
2940
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2942
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2943
isinstance(target._format, RepositoryFormatKnit))
2944
except AttributeError:
2946
return are_knits and InterRepository._same_model(source, target)
2949
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2950
"""See InterRepository.missing_revision_ids()."""
2951
if revision_id is not None:
2952
source_ids = self.source.get_ancestry(revision_id)
2953
if source_ids[0] is not None:
2954
raise AssertionError()
2957
source_ids = self.source.all_revision_ids()
2958
source_ids_set = set(source_ids)
2959
# source_ids is the worst possible case we may need to pull.
2960
# now we want to filter source_ids against what we actually
2961
# have in target, but don't try to check for existence where we know
2962
# we do not have a revision as that would be pointless.
2963
target_ids = set(self.target.all_revision_ids())
2964
possibly_present_revisions = target_ids.intersection(source_ids_set)
2965
actually_present_revisions = set(
2966
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2967
required_revisions = source_ids_set.difference(actually_present_revisions)
2968
if revision_id is not None:
2969
# we used get_ancestry to determine source_ids then we are assured all
2970
# revisions referenced are present as they are installed in topological order.
2971
# and the tip revision was validated by get_ancestry.
2972
result_set = required_revisions
2974
# if we just grabbed the possibly available ids, then
2975
# we only have an estimate of whats available and need to validate
2976
# that against the revision records.
2978
self.source._eliminate_revisions_not_present(required_revisions))
2979
return self.source.revision_ids_to_search_result(result_set)
2982
class InterPackRepo(InterSameDataRepository):
2983
"""Optimised code paths between Pack based repositories."""
2986
def _get_repo_format_to_test(self):
2987
from bzrlib.repofmt import pack_repo
2988
return pack_repo.RepositoryFormatKnitPack1()
2991
def is_compatible(source, target):
2992
"""Be compatible with known Pack formats.
2994
We don't test for the stores being of specific types because that
2995
could lead to confusing results, and there is no need to be
2998
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3000
are_packs = (isinstance(source._format, RepositoryFormatPack) and
3001
isinstance(target._format, RepositoryFormatPack))
3002
except AttributeError:
3004
return are_packs and InterRepository._same_model(source, target)
3007
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3009
"""See InterRepository.fetch()."""
3010
if (len(self.source._fallback_repositories) > 0 or
3011
len(self.target._fallback_repositories) > 0):
3012
# The pack layer is not aware of fallback repositories, so when
3013
# fetching from a stacked repository or into a stacked repository
3014
# we use the generic fetch logic which uses the VersionedFiles
3015
# attributes on repository.
3016
from bzrlib.fetch import RepoFetcher
3017
fetcher = RepoFetcher(self.target, self.source, revision_id,
3018
pb, find_ghosts, fetch_spec=fetch_spec)
3019
if fetch_spec is not None:
3020
if len(list(fetch_spec.heads)) != 1:
3021
raise AssertionError(
3022
"InterPackRepo.fetch doesn't support "
3023
"fetching multiple heads yet.")
3024
revision_id = fetch_spec.heads[0]
3026
if revision_id is None:
3028
# everything to do - use pack logic
3029
# to fetch from all packs to one without
3030
# inventory parsing etc, IFF nothing to be copied is in the target.
3032
source_revision_ids = frozenset(self.source.all_revision_ids())
3033
revision_ids = source_revision_ids - \
3034
frozenset(self.target_get_parent_map(source_revision_ids))
3035
revision_keys = [(revid,) for revid in revision_ids]
3036
target_pack_collection = self._get_target_pack_collection()
3037
index = target_pack_collection.revision_index.combined_index
3038
present_revision_ids = set(item[1][0] for item in
3039
index.iter_entries(revision_keys))
3040
revision_ids = set(revision_ids) - present_revision_ids
3041
# implementing the TODO will involve:
3042
# - detecting when all of a pack is selected
3043
# - avoiding as much as possible pre-selection, so the
3044
# more-core routines such as create_pack_from_packs can filter in
3045
# a just-in-time fashion. (though having a HEADS list on a
3046
# repository might make this a lot easier, because we could
3047
# sensibly detect 'new revisions' without doing a full index scan.
3048
elif _mod_revision.is_null(revision_id):
3053
revision_ids = self.search_missing_revision_ids(revision_id,
3054
find_ghosts=find_ghosts).get_keys()
3055
except errors.NoSuchRevision:
3056
raise errors.InstallFailed([revision_id])
3057
if len(revision_ids) == 0:
3059
return self._pack(self.source, self.target, revision_ids)
3061
def _pack(self, source, target, revision_ids):
3062
from bzrlib.repofmt.pack_repo import Packer
3063
target_pack_collection = self._get_target_pack_collection()
3064
packs = source._pack_collection.all_packs()
3065
pack = Packer(target_pack_collection, packs, '.fetch',
3066
revision_ids).pack()
3067
if pack is not None:
3068
target_pack_collection._save_pack_names()
3069
copied_revs = pack.get_revision_count()
3070
# Trigger an autopack. This may duplicate effort as we've just done
3071
# a pack creation, but for now it is simpler to think about as
3072
# 'upload data, then repack if needed'.
3074
return (copied_revs, [])
3078
def _autopack(self):
3079
self.target._pack_collection.autopack()
3081
def _get_target_pack_collection(self):
3082
return self.target._pack_collection
3085
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3086
"""See InterRepository.missing_revision_ids().
3088
:param find_ghosts: Find ghosts throughout the ancestry of
3091
if not find_ghosts and revision_id is not None:
3092
return self._walk_to_common_revisions([revision_id])
3093
elif revision_id is not None:
3094
# Find ghosts: search for revisions pointing from one repository to
3095
# the other, and vice versa, anywhere in the history of revision_id.
3096
graph = self.target_get_graph(other_repository=self.source)
3097
searcher = graph._make_breadth_first_searcher([revision_id])
3101
next_revs, ghosts = searcher.next_with_ghosts()
3102
except StopIteration:
3104
if revision_id in ghosts:
3105
raise errors.NoSuchRevision(self.source, revision_id)
3106
found_ids.update(next_revs)
3107
found_ids.update(ghosts)
3108
found_ids = frozenset(found_ids)
3109
# Double query here: should be able to avoid this by changing the
3110
# graph api further.
3111
result_set = found_ids - frozenset(
3112
self.target_get_parent_map(found_ids))
3114
source_ids = self.source.all_revision_ids()
3115
# source_ids is the worst possible case we may need to pull.
3116
# now we want to filter source_ids against what we actually
3117
# have in target, but don't try to check for existence where we know
3118
# we do not have a revision as that would be pointless.
3119
target_ids = set(self.target.all_revision_ids())
3120
result_set = set(source_ids).difference(target_ids)
3121
return self.source.revision_ids_to_search_result(result_set)
3124
class InterDifferingSerializer(InterKnitRepo):
3127
def _get_repo_format_to_test(self):
3131
def is_compatible(source, target):
3132
"""Be compatible with Knit2 source and Knit3 target"""
3133
if source.supports_rich_root() != target.supports_rich_root():
3135
# Ideally, we'd support fetching if the source had no tree references
3136
# even if it supported them...
3137
if (getattr(source, '_format.supports_tree_reference', False) and
3138
not getattr(target, '_format.supports_tree_reference', False)):
3142
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3143
"""Get the best delta and base for this revision.
3145
:return: (basis_id, delta)
3147
possible_trees = [(parent_id, cache[parent_id])
3148
for parent_id in parent_ids
3149
if parent_id in cache]
3150
if len(possible_trees) == 0:
3151
# There either aren't any parents, or the parents aren't in the
3152
# cache, so just use the last converted tree
3153
possible_trees.append((basis_id, cache[basis_id]))
3155
for basis_id, basis_tree in possible_trees:
3156
delta = tree.inventory._make_delta(basis_tree.inventory)
3157
deltas.append((len(delta), basis_id, delta))
3159
return deltas[0][1:]
3161
def _fetch_batch(self, revision_ids, basis_id, cache):
3162
"""Fetch across a few revisions.
3164
:param revision_ids: The revisions to copy
3165
:param basis_id: The revision_id of a tree that must be in cache, used
3166
as a basis for delta when no other base is available
3167
:param cache: A cache of RevisionTrees that we can use.
3168
:return: The revision_id of the last converted tree. The RevisionTree
3169
for it will be in cache
3171
# Walk though all revisions; get inventory deltas, copy referenced
3172
# texts that delta references, insert the delta, revision and
3176
pending_revisions = []
3177
parent_map = self.source.get_parent_map(revision_ids)
3178
for tree in self.source.revision_trees(revision_ids):
3179
current_revision_id = tree.get_revision_id()
3180
parent_ids = parent_map.get(current_revision_id, ())
3181
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3183
# Find text entries that need to be copied
3184
for old_path, new_path, file_id, entry in delta:
3185
if new_path is not None:
3186
if not (new_path or self.target.supports_rich_root()):
3187
# We don't copy the text for the root node unless the
3188
# target supports_rich_root.
3190
text_keys.add((file_id, entry.revision))
3191
revision = self.source.get_revision(current_revision_id)
3192
pending_deltas.append((basis_id, delta,
3193
current_revision_id, revision.parent_ids))
3194
pending_revisions.append(revision)
3195
cache[current_revision_id] = tree
3196
basis_id = current_revision_id
3198
from_texts = self.source.texts
3199
to_texts = self.target.texts
3200
to_texts.insert_record_stream(from_texts.get_record_stream(
3201
text_keys, self.target._format._fetch_order,
3202
not self.target._format._fetch_uses_deltas))
3204
for delta in pending_deltas:
3205
self.target.add_inventory_by_delta(*delta)
3206
# insert signatures and revisions
3207
for revision in pending_revisions:
3209
signature = self.source.get_signature_text(
3210
revision.revision_id)
3211
self.target.add_signature_text(revision.revision_id,
3213
except errors.NoSuchRevision:
3215
self.target.add_revision(revision.revision_id, revision)
3218
def _fetch_all_revisions(self, revision_ids, pb):
3219
"""Fetch everything for the list of revisions.
3221
:param revision_ids: The list of revisions to fetch. Must be in
3223
:param pb: A ProgressBar
3226
basis_id, basis_tree = self._get_basis(revision_ids[0])
3228
cache = lru_cache.LRUCache(100)
3229
cache[basis_id] = basis_tree
3230
del basis_tree # We don't want to hang on to it here
3231
for offset in range(0, len(revision_ids), batch_size):
3232
self.target.start_write_group()
3234
pb.update('Transferring revisions', offset,
3236
batch = revision_ids[offset:offset+batch_size]
3237
basis_id = self._fetch_batch(batch, basis_id, cache)
3239
self.target.abort_write_group()
3242
self.target.commit_write_group()
3243
pb.update('Transferring revisions', len(revision_ids),
3247
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3249
"""See InterRepository.fetch()."""
3250
if fetch_spec is not None:
3251
raise AssertionError("Not implemented yet...")
3252
revision_ids = self.target.search_missing_revision_ids(self.source,
3253
revision_id, find_ghosts=find_ghosts).get_keys()
3254
if not revision_ids:
3256
revision_ids = tsort.topo_sort(
3257
self.source.get_graph().get_parent_map(revision_ids))
3259
my_pb = ui.ui_factory.nested_progress_bar()
3262
symbol_versioning.warn(
3263
symbol_versioning.deprecated_in((1, 14, 0))
3264
% "pb parameter to fetch()")
3267
self._fetch_all_revisions(revision_ids, pb)
3269
if my_pb is not None:
3271
return len(revision_ids), 0
3273
def _get_basis(self, first_revision_id):
3274
"""Get a revision and tree which exists in the target.
3276
This assumes that first_revision_id is selected for transmission
3277
because all other ancestors are already present. If we can't find an
3278
ancestor we fall back to NULL_REVISION since we know that is safe.
3280
:return: (basis_id, basis_tree)
3282
first_rev = self.source.get_revision(first_revision_id)
3284
basis_id = first_rev.parent_ids[0]
3285
# only valid as a basis if the target has it
3286
self.target.get_revision(basis_id)
3287
# Try to get a basis tree - if its a ghost it will hit the
3288
# NoSuchRevision case.
3289
basis_tree = self.source.revision_tree(basis_id)
3290
except (IndexError, errors.NoSuchRevision):
3291
basis_id = _mod_revision.NULL_REVISION
3292
basis_tree = self.source.revision_tree(basis_id)
3293
return basis_id, basis_tree
3296
class InterOtherToRemote(InterRepository):
3297
"""An InterRepository that simply delegates to the 'real' InterRepository
3298
calculated for (source, target._real_repository).
3301
_walk_to_common_revisions_batch_size = 50
3303
def __init__(self, source, target):
3304
InterRepository.__init__(self, source, target)
3305
self._real_inter = None
3308
def is_compatible(source, target):
3309
if isinstance(target, remote.RemoteRepository):
3313
def _ensure_real_inter(self):
3314
if self._real_inter is None:
3315
self.target._ensure_real()
3316
real_target = self.target._real_repository
3317
self._real_inter = InterRepository.get(self.source, real_target)
3318
# Make _real_inter use the RemoteRepository for get_parent_map
3319
self._real_inter.target_get_graph = self.target.get_graph
3320
self._real_inter.target_get_parent_map = self.target.get_parent_map
3322
def copy_content(self, revision_id=None):
3323
self._ensure_real_inter()
3324
self._real_inter.copy_content(revision_id=revision_id)
3326
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3328
self._ensure_real_inter()
3329
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3330
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
3333
def _get_repo_format_to_test(self):
3337
class InterRemoteToOther(InterRepository):
3339
def __init__(self, source, target):
3340
InterRepository.__init__(self, source, target)
3341
self._real_inter = None
3344
def is_compatible(source, target):
3345
if not isinstance(source, remote.RemoteRepository):
3347
return InterRepository._same_model(source, target)
3349
def _ensure_real_inter(self):
3350
if self._real_inter is None:
3351
self.source._ensure_real()
3352
real_source = self.source._real_repository
3353
self._real_inter = InterRepository.get(real_source, self.target)
3355
def copy_content(self, revision_id=None):
3356
self._ensure_real_inter()
3357
self._real_inter.copy_content(revision_id=revision_id)
3360
def _get_repo_format_to_test(self):
3365
class InterPackToRemotePack(InterPackRepo):
3366
"""A specialisation of InterPackRepo for a target that is a
3369
This will use the get_parent_map RPC rather than plain readvs, and also
3370
uses an RPC for autopacking.
3373
_walk_to_common_revisions_batch_size = 50
3376
def is_compatible(source, target):
3377
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3378
if isinstance(source._format, RepositoryFormatPack):
3379
if isinstance(target, remote.RemoteRepository):
3380
target._format._ensure_real()
3381
if isinstance(target._format._custom_format,
3382
RepositoryFormatPack):
3383
if InterRepository._same_model(source, target):
3387
def _autopack(self):
3388
self.target.autopack()
3391
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3393
"""See InterRepository.fetch()."""
3394
if self.target._client._medium._is_remote_before((1, 13)):
3395
# The server won't support the insert_stream RPC, so just use
3396
# regular InterPackRepo logic. This avoids a bug that causes many
3397
# round-trips for small append calls.
3398
return InterPackRepo.fetch(self, revision_id=revision_id, pb=pb,
3399
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
3400
# Always fetch using the generic streaming fetch code, to allow
3401
# streaming fetching into remote servers.
3402
from bzrlib.fetch import RepoFetcher
3403
fetcher = RepoFetcher(self.target, self.source, revision_id,
3404
pb, find_ghosts, fetch_spec=fetch_spec)
3406
def _get_target_pack_collection(self):
3407
return self.target._real_repository._pack_collection
3410
def _get_repo_format_to_test(self):
3414
InterRepository.register_optimiser(InterDifferingSerializer)
3415
InterRepository.register_optimiser(InterSameDataRepository)
3416
InterRepository.register_optimiser(InterWeaveRepo)
3417
InterRepository.register_optimiser(InterKnitRepo)
3418
InterRepository.register_optimiser(InterPackRepo)
3419
InterRepository.register_optimiser(InterOtherToRemote)
3420
InterRepository.register_optimiser(InterRemoteToOther)
3421
InterRepository.register_optimiser(InterPackToRemotePack)
3424
class CopyConverter(object):
3425
"""A repository conversion tool which just performs a copy of the content.
3427
This is slow but quite reliable.
3430
def __init__(self, target_format):
3431
"""Create a CopyConverter.
3433
:param target_format: The format the resulting repository should be.
3435
self.target_format = target_format
3437
def convert(self, repo, pb):
3438
"""Perform the conversion of to_convert, giving feedback via pb.
3440
:param to_convert: The disk object to convert.
3441
:param pb: a progress bar to use for progress information.
3446
# this is only useful with metadir layouts - separated repo content.
3447
# trigger an assertion if not such
3448
repo._format.get_format_string()
3449
self.repo_dir = repo.bzrdir
3450
self.step('Moving repository to repository.backup')
3451
self.repo_dir.transport.move('repository', 'repository.backup')
3452
backup_transport = self.repo_dir.transport.clone('repository.backup')
3453
repo._format.check_conversion_target(self.target_format)
3454
self.source_repo = repo._format.open(self.repo_dir,
3456
_override_transport=backup_transport)
3457
self.step('Creating new repository')
3458
converted = self.target_format.initialize(self.repo_dir,
3459
self.source_repo.is_shared())
3460
converted.lock_write()
3462
self.step('Copying content into repository.')
3463
self.source_repo.copy_content_into(converted)
3466
self.step('Deleting old repository content.')
3467
self.repo_dir.transport.delete_tree('repository.backup')
3468
self.pb.note('repository converted')
3470
def step(self, message):
3471
"""Update the pb by a step."""
3473
self.pb.update(message, self.count, self.total)
3485
def _unescaper(match, _map=_unescape_map):
3486
code = match.group(1)
3490
if not code.startswith('#'):
3492
return unichr(int(code[1:])).encode('utf8')
3498
def _unescape_xml(data):
3499
"""Unescape predefined XML entities in a string of data."""
3501
if _unescape_re is None:
3502
_unescape_re = re.compile('\&([^;]*);')
3503
return _unescape_re.sub(_unescaper, data)
3506
class _VersionedFileChecker(object):
3508
def __init__(self, repository):
3509
self.repository = repository
3510
self.text_index = self.repository._generate_text_key_index()
3512
def calculate_file_version_parents(self, text_key):
3513
"""Calculate the correct parents for a file version according to
3516
parent_keys = self.text_index[text_key]
3517
if parent_keys == [_mod_revision.NULL_REVISION]:
3519
return tuple(parent_keys)
3521
def check_file_version_parents(self, texts, progress_bar=None):
3522
"""Check the parents stored in a versioned file are correct.
3524
It also detects file versions that are not referenced by their
3525
corresponding revision's inventory.
3527
:returns: A tuple of (wrong_parents, dangling_file_versions).
3528
wrong_parents is a dict mapping {revision_id: (stored_parents,
3529
correct_parents)} for each revision_id where the stored parents
3530
are not correct. dangling_file_versions is a set of (file_id,
3531
revision_id) tuples for versions that are present in this versioned
3532
file, but not used by the corresponding inventory.
3535
self.file_ids = set([file_id for file_id, _ in
3536
self.text_index.iterkeys()])
3537
# text keys is now grouped by file_id
3538
n_weaves = len(self.file_ids)
3539
files_in_revisions = {}
3540
revisions_of_files = {}
3541
n_versions = len(self.text_index)
3542
progress_bar.update('loading text store', 0, n_versions)
3543
parent_map = self.repository.texts.get_parent_map(self.text_index)
3544
# On unlistable transports this could well be empty/error...
3545
text_keys = self.repository.texts.keys()
3546
unused_keys = frozenset(text_keys) - set(self.text_index)
3547
for num, key in enumerate(self.text_index.iterkeys()):
3548
if progress_bar is not None:
3549
progress_bar.update('checking text graph', num, n_versions)
3550
correct_parents = self.calculate_file_version_parents(key)
3552
knit_parents = parent_map[key]
3553
except errors.RevisionNotPresent:
3556
if correct_parents != knit_parents:
3557
wrong_parents[key] = (knit_parents, correct_parents)
3558
return wrong_parents, unused_keys
3561
def _old_get_graph(repository, revision_id):
3562
"""DO NOT USE. That is all. I'm serious."""
3563
graph = repository.get_graph()
3564
revision_graph = dict(((key, value) for key, value in
3565
graph.iter_ancestry([revision_id]) if value is not None))
3566
return _strip_NULL_ghosts(revision_graph)
3569
def _strip_NULL_ghosts(revision_graph):
3570
"""Also don't use this. more compatibility code for unmigrated clients."""
3571
# Filter ghosts, and null:
3572
if _mod_revision.NULL_REVISION in revision_graph:
3573
del revision_graph[_mod_revision.NULL_REVISION]
3574
for key, parents in revision_graph.items():
3575
revision_graph[key] = tuple(parent for parent in parents if parent
3577
return revision_graph
3580
class StreamSink(object):
3581
"""An object that can insert a stream into a repository.
3583
This interface handles the complexity of reserialising inventories and
3584
revisions from different formats, and allows unidirectional insertion into
3585
stacked repositories without looking for the missing basis parents
3589
def __init__(self, target_repo):
3590
self.target_repo = target_repo
3592
def insert_stream(self, stream, src_format, resume_tokens):
3593
"""Insert a stream's content into the target repository.
3595
:param src_format: a bzr repository format.
3597
:return: a list of resume tokens and an iterable of keys additional
3598
items required before the insertion can be completed.
3600
self.target_repo.lock_write()
3603
self.target_repo.resume_write_group(resume_tokens)
3605
self.target_repo.start_write_group()
3607
# locked_insert_stream performs a commit|suspend.
3608
return self._locked_insert_stream(stream, src_format)
3610
self.target_repo.abort_write_group(suppress_errors=True)
3613
self.target_repo.unlock()
3615
def _locked_insert_stream(self, stream, src_format):
3616
to_serializer = self.target_repo._format._serializer
3617
src_serializer = src_format._serializer
3618
for substream_type, substream in stream:
3619
if substream_type == 'texts':
3620
self.target_repo.texts.insert_record_stream(substream)
3621
elif substream_type == 'inventories':
3622
if src_serializer == to_serializer:
3623
self.target_repo.inventories.insert_record_stream(
3626
self._extract_and_insert_inventories(
3627
substream, src_serializer)
3628
elif substream_type == 'revisions':
3629
# This may fallback to extract-and-insert more often than
3630
# required if the serializers are different only in terms of
3632
if src_serializer == to_serializer:
3633
self.target_repo.revisions.insert_record_stream(
3636
self._extract_and_insert_revisions(substream,
3638
elif substream_type == 'signatures':
3639
self.target_repo.signatures.insert_record_stream(substream)
3641
raise AssertionError('kaboom! %s' % (substream_type,))
3643
missing_keys = set()
3644
for prefix, versioned_file in (
3645
('texts', self.target_repo.texts),
3646
('inventories', self.target_repo.inventories),
3647
('revisions', self.target_repo.revisions),
3648
('signatures', self.target_repo.signatures),
3650
missing_keys.update((prefix,) + key for key in
3651
versioned_file.get_missing_compression_parent_keys())
3652
except NotImplementedError:
3653
# cannot even attempt suspending, and missing would have failed
3654
# during stream insertion.
3655
missing_keys = set()
3658
# suspend the write group and tell the caller what we is
3659
# missing. We know we can suspend or else we would not have
3660
# entered this code path. (All repositories that can handle
3661
# missing keys can handle suspending a write group).
3662
write_group_tokens = self.target_repo.suspend_write_group()
3663
return write_group_tokens, missing_keys
3664
self.target_repo.commit_write_group()
3667
def _extract_and_insert_inventories(self, substream, serializer):
3668
"""Generate a new inventory versionedfile in target, converting data.
3670
The inventory is retrieved from the source, (deserializing it), and
3671
stored in the target (reserializing it in a different format).
3673
for record in substream:
3674
bytes = record.get_bytes_as('fulltext')
3675
revision_id = record.key[0]
3676
inv = serializer.read_inventory_from_string(bytes, revision_id)
3677
parents = [key[0] for key in record.parents]
3678
self.target_repo.add_inventory(revision_id, inv, parents)
3680
def _extract_and_insert_revisions(self, substream, serializer):
3681
for record in substream:
3682
bytes = record.get_bytes_as('fulltext')
3683
revision_id = record.key[0]
3684
rev = serializer.read_revision_from_string(bytes)
3685
if rev.revision_id != revision_id:
3686
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
3687
self.target_repo.add_revision(revision_id, rev)
3690
if self.target_repo._format._fetch_reconcile:
3691
self.target_repo.reconcile()
3694
class StreamSource(object):
3695
"""A source of a stream for fetching between repositories."""
3697
def __init__(self, from_repository, to_format):
3698
"""Create a StreamSource streaming from from_repository."""
3699
self.from_repository = from_repository
3700
self.to_format = to_format
3702
def delta_on_metadata(self):
3703
"""Return True if delta's are permitted on metadata streams.
3705
That is on revisions and signatures.
3707
src_serializer = self.from_repository._format._serializer
3708
target_serializer = self.to_format._serializer
3709
return (self.to_format._fetch_uses_deltas and
3710
src_serializer == target_serializer)
3712
def _fetch_revision_texts(self, revs):
3713
# fetch signatures first and then the revision texts
3714
# may need to be a InterRevisionStore call here.
3715
from_sf = self.from_repository.signatures
3716
# A missing signature is just skipped.
3717
keys = [(rev_id,) for rev_id in revs]
3718
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
3720
self.to_format._fetch_order,
3721
not self.to_format._fetch_uses_deltas))
3722
# If a revision has a delta, this is actually expanded inside the
3723
# insert_record_stream code now, which is an alternate fix for
3725
from_rf = self.from_repository.revisions
3726
revisions = from_rf.get_record_stream(
3728
self.to_format._fetch_order,
3729
not self.delta_on_metadata())
3730
return [('signatures', signatures), ('revisions', revisions)]
3732
def _generate_root_texts(self, revs):
3733
"""This will be called by __fetch between fetching weave texts and
3734
fetching the inventory weave.
3736
Subclasses should override this if they need to generate root texts
3737
after fetching weave texts.
3739
if self._rich_root_upgrade():
3741
return bzrlib.fetch.Inter1and2Helper(
3742
self.from_repository).generate_root_texts(revs)
3746
def get_stream(self, search):
3748
revs = search.get_keys()
3749
graph = self.from_repository.get_graph()
3750
revs = list(graph.iter_topo_order(revs))
3751
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
3753
for knit_kind, file_id, revisions in data_to_fetch:
3754
if knit_kind != phase:
3756
# Make a new progress bar for this phase
3757
if knit_kind == "file":
3758
# Accumulate file texts
3759
text_keys.extend([(file_id, revision) for revision in
3761
elif knit_kind == "inventory":
3762
# Now copy the file texts.
3763
from_texts = self.from_repository.texts
3764
yield ('texts', from_texts.get_record_stream(
3765
text_keys, self.to_format._fetch_order,
3766
not self.to_format._fetch_uses_deltas))
3767
# Cause an error if a text occurs after we have done the
3770
# Before we process the inventory we generate the root
3771
# texts (if necessary) so that the inventories references
3773
for _ in self._generate_root_texts(revs):
3775
# NB: This currently reopens the inventory weave in source;
3776
# using a single stream interface instead would avoid this.
3777
from_weave = self.from_repository.inventories
3778
# we fetch only the referenced inventories because we do not
3779
# know for unselected inventories whether all their required
3780
# texts are present in the other repository - it could be
3782
yield ('inventories', from_weave.get_record_stream(
3783
[(rev_id,) for rev_id in revs],
3784
self.inventory_fetch_order(),
3785
not self.delta_on_metadata()))
3786
elif knit_kind == "signatures":
3787
# Nothing to do here; this will be taken care of when
3788
# _fetch_revision_texts happens.
3790
elif knit_kind == "revisions":
3791
for record in self._fetch_revision_texts(revs):
3794
raise AssertionError("Unknown knit kind %r" % knit_kind)
3796
def get_stream_for_missing_keys(self, missing_keys):
3797
# missing keys can only occur when we are byte copying and not
3798
# translating (because translation means we don't send
3799
# unreconstructable deltas ever).
3801
keys['texts'] = set()
3802
keys['revisions'] = set()
3803
keys['inventories'] = set()
3804
keys['signatures'] = set()
3805
for key in missing_keys:
3806
keys[key[0]].add(key[1:])
3807
if len(keys['revisions']):
3808
# If we allowed copying revisions at this point, we could end up
3809
# copying a revision without copying its required texts: a
3810
# violation of the requirements for repository integrity.
3811
raise AssertionError(
3812
'cannot copy revisions to fill in missing deltas %s' % (
3813
keys['revisions'],))
3814
for substream_kind, keys in keys.iteritems():
3815
vf = getattr(self.from_repository, substream_kind)
3816
# Ask for full texts always so that we don't need more round trips
3817
# after this stream.
3818
stream = vf.get_record_stream(keys,
3819
self.to_format._fetch_order, True)
3820
yield substream_kind, stream
3822
def inventory_fetch_order(self):
3823
if self._rich_root_upgrade():
3824
return 'topological'
3826
return self.to_format._fetch_order
3828
def _rich_root_upgrade(self):
3829
return (not self.from_repository._format.rich_root_data and
3830
self.to_format.rich_root_data)
added
removed
bzrlib/repository.py
