1
# Copyright (C) 2005, 2006, 2007, 2008 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from bzrlib.lazy_import import lazy_import
18
lazy_import(globals(), """
38
revision as _mod_revision,
44
from bzrlib.bundle import serializer
45
from bzrlib.revisiontree import RevisionTree
46
from bzrlib.store.versioned import VersionedFileStore
47
from bzrlib.testament import Testament
50
from bzrlib.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
result = self.control_files.lock_write(token=token)
888
for repo in self._fallback_repositories:
889
# Writes don't affect fallback repos
895
self.control_files.lock_read()
896
for repo in self._fallback_repositories:
900
def get_physical_lock_status(self):
901
return self.control_files.get_physical_lock_status()
903
def leave_lock_in_place(self):
904
"""Tell this repository not to release the physical lock when this
907
If lock_write doesn't return a token, then this method is not supported.
909
self.control_files.leave_in_place()
911
def dont_leave_lock_in_place(self):
912
"""Tell this repository to release the physical lock when this
913
object is unlocked, even if it didn't originally acquire it.
915
If lock_write doesn't return a token, then this method is not supported.
917
self.control_files.dont_leave_in_place()
920
def gather_stats(self, revid=None, committers=None):
921
"""Gather statistics from a revision id.
923
:param revid: The revision id to gather statistics from, if None, then
924
no revision specific statistics are gathered.
925
:param committers: Optional parameter controlling whether to grab
926
a count of committers from the revision specific statistics.
927
:return: A dictionary of statistics. Currently this contains:
928
committers: The number of committers if requested.
929
firstrev: A tuple with timestamp, timezone for the penultimate left
930
most ancestor of revid, if revid is not the NULL_REVISION.
931
latestrev: A tuple with timestamp, timezone for revid, if revid is
932
not the NULL_REVISION.
933
revisions: The total revision count in the repository.
934
size: An estimate disk size of the repository in bytes.
937
if revid and committers:
938
result['committers'] = 0
939
if revid and revid != _mod_revision.NULL_REVISION:
941
all_committers = set()
942
revisions = self.get_ancestry(revid)
943
# pop the leading None
945
first_revision = None
947
# ignore the revisions in the middle - just grab first and last
948
revisions = revisions[0], revisions[-1]
949
for revision in self.get_revisions(revisions):
950
if not first_revision:
951
first_revision = revision
953
all_committers.add(revision.committer)
954
last_revision = revision
956
result['committers'] = len(all_committers)
957
result['firstrev'] = (first_revision.timestamp,
958
first_revision.timezone)
959
result['latestrev'] = (last_revision.timestamp,
960
last_revision.timezone)
962
# now gather global repository information
963
# XXX: This is available for many repos regardless of listability.
964
if self.bzrdir.root_transport.listable():
965
# XXX: do we want to __define len__() ?
966
# Maybe the versionedfiles object should provide a different
967
# method to get the number of keys.
968
result['revisions'] = len(self.revisions.keys())
972
def find_branches(self, using=False):
973
"""Find branches underneath this repository.
975
This will include branches inside other branches.
977
:param using: If True, list only branches using this repository.
979
if using and not self.is_shared():
981
return [self.bzrdir.open_branch()]
982
except errors.NotBranchError:
984
class Evaluator(object):
987
self.first_call = True
989
def __call__(self, bzrdir):
990
# On the first call, the parameter is always the bzrdir
991
# containing the current repo.
992
if not self.first_call:
994
repository = bzrdir.open_repository()
995
except errors.NoRepositoryPresent:
998
return False, (None, repository)
999
self.first_call = False
1001
value = (bzrdir.open_branch(), None)
1002
except errors.NotBranchError:
1003
value = (None, None)
1007
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1008
self.bzrdir.root_transport, evaluate=Evaluator()):
1009
if branch is not None:
1010
branches.append(branch)
1011
if not using and repository is not None:
1012
branches.extend(repository.find_branches())
1016
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1017
"""Return the revision ids that other has that this does not.
1019
These are returned in topological order.
1021
revision_id: only return revision ids included by revision_id.
1023
return InterRepository.get(other, self).search_missing_revision_ids(
1024
revision_id, find_ghosts)
1026
@deprecated_method(one_two)
1028
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1029
"""Return the revision ids that other has that this does not.
1031
These are returned in topological order.
1033
revision_id: only return revision ids included by revision_id.
1035
keys = self.search_missing_revision_ids(
1036
other, revision_id, find_ghosts).get_keys()
1039
parents = other.get_graph().get_parent_map(keys)
1042
return tsort.topo_sort(parents)
1046
"""Open the repository rooted at base.
1048
For instance, if the repository is at URL/.bzr/repository,
1049
Repository.open(URL) -> a Repository instance.
1051
control = bzrdir.BzrDir.open(base)
1052
return control.open_repository()
1054
def copy_content_into(self, destination, revision_id=None):
1055
"""Make a complete copy of the content in self into destination.
1057
This is a destructive operation! Do not use it on existing
1060
return InterRepository.get(self, destination).copy_content(revision_id)
1062
def commit_write_group(self):
1063
"""Commit the contents accrued within the current write group.
1065
:seealso: start_write_group.
1067
if self._write_group is not self.get_transaction():
1068
# has an unlock or relock occured ?
1069
raise errors.BzrError('mismatched lock context %r and '
1071
(self.get_transaction(), self._write_group))
1072
self._commit_write_group()
1073
self._write_group = None
1075
def _commit_write_group(self):
1076
"""Template method for per-repository write group cleanup.
1078
This is called before the write group is considered to be
1079
finished and should ensure that all data handed to the repository
1080
for writing during the write group is safely committed (to the
1081
extent possible considering file system caching etc).
1084
def suspend_write_group(self):
1085
raise errors.UnsuspendableWriteGroup(self)
1087
def resume_write_group(self, tokens):
1088
if not self.is_write_locked():
1089
raise errors.NotWriteLocked(self)
1090
if self._write_group:
1091
raise errors.BzrError('already in a write group')
1092
self._resume_write_group(tokens)
1093
# so we can detect unlock/relock - the write group is now entered.
1094
self._write_group = self.get_transaction()
1096
def _resume_write_group(self, tokens):
1097
raise errors.UnsuspendableWriteGroup(self)
1099
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1100
"""Fetch the content required to construct revision_id from source.
1102
If revision_id is None all content is copied.
1103
:param find_ghosts: Find and copy revisions in the source that are
1104
ghosts in the target (and not reachable directly by walking out to
1105
the first-present revision in target from revision_id).
1107
# fast path same-url fetch operations
1108
if self.has_same_location(source):
1109
# check that last_revision is in 'from' and then return a
1111
if (revision_id is not None and
1112
not _mod_revision.is_null(revision_id)):
1113
self.get_revision(revision_id)
1115
# if there is no specific appropriate InterRepository, this will get
1116
# the InterRepository base class, which raises an
1117
# IncompatibleRepositories when asked to fetch.
1118
inter = InterRepository.get(source, self)
1119
return inter.fetch(revision_id=revision_id, pb=pb,
1120
find_ghosts=find_ghosts)
1122
def create_bundle(self, target, base, fileobj, format=None):
1123
return serializer.write_bundle(self, target, base, fileobj, format)
1125
def get_commit_builder(self, branch, parents, config, timestamp=None,
1126
timezone=None, committer=None, revprops=None,
1128
"""Obtain a CommitBuilder for this repository.
1130
:param branch: Branch to commit to.
1131
:param parents: Revision ids of the parents of the new revision.
1132
:param config: Configuration to use.
1133
:param timestamp: Optional timestamp recorded for commit.
1134
:param timezone: Optional timezone for timestamp.
1135
:param committer: Optional committer to set for commit.
1136
:param revprops: Optional dictionary of revision properties.
1137
:param revision_id: Optional revision id.
1139
result = self._commit_builder_class(self, parents, config,
1140
timestamp, timezone, committer, revprops, revision_id)
1141
self.start_write_group()
1145
if (self.control_files._lock_count == 1 and
1146
self.control_files._lock_mode == 'w'):
1147
if self._write_group is not None:
1148
self.abort_write_group()
1149
self.control_files.unlock()
1150
raise errors.BzrError(
1151
'Must end write groups before releasing write locks.')
1152
self.control_files.unlock()
1153
if self.control_files._lock_count == 0:
1154
self._inventory_entry_cache.clear()
1155
for repo in self._fallback_repositories:
1159
def clone(self, a_bzrdir, revision_id=None):
1160
"""Clone this repository into a_bzrdir using the current format.
1162
Currently no check is made that the format of this repository and
1163
the bzrdir format are compatible. FIXME RBC 20060201.
1165
:return: The newly created destination repository.
1167
# TODO: deprecate after 0.16; cloning this with all its settings is
1168
# probably not very useful -- mbp 20070423
1169
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1170
self.copy_content_into(dest_repo, revision_id)
1173
def start_write_group(self):
1174
"""Start a write group in the repository.
1176
Write groups are used by repositories which do not have a 1:1 mapping
1177
between file ids and backend store to manage the insertion of data from
1178
both fetch and commit operations.
1180
A write lock is required around the start_write_group/commit_write_group
1181
for the support of lock-requiring repository formats.
1183
One can only insert data into a repository inside a write group.
1187
if not self.is_write_locked():
1188
raise errors.NotWriteLocked(self)
1189
if self._write_group:
1190
raise errors.BzrError('already in a write group')
1191
self._start_write_group()
1192
# so we can detect unlock/relock - the write group is now entered.
1193
self._write_group = self.get_transaction()
1195
def _start_write_group(self):
1196
"""Template method for per-repository write group startup.
1198
This is called before the write group is considered to be
1203
def sprout(self, to_bzrdir, revision_id=None):
1204
"""Create a descendent repository for new development.
1206
Unlike clone, this does not copy the settings of the repository.
1208
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1209
dest_repo.fetch(self, revision_id=revision_id)
1212
def _create_sprouting_repo(self, a_bzrdir, shared):
1213
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1214
# use target default format.
1215
dest_repo = a_bzrdir.create_repository()
1217
# Most control formats need the repository to be specifically
1218
# created, but on some old all-in-one formats it's not needed
1220
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1221
except errors.UninitializableFormat:
1222
dest_repo = a_bzrdir.open_repository()
1225
def _get_sink(self):
1226
"""Return a sink for streaming into this repository."""
1227
return StreamSink(self)
1229
def _get_source(self, to_format):
1230
"""Return a source for streaming from this repository."""
1231
return StreamSource(self, to_format)
1234
def has_revision(self, revision_id):
1235
"""True if this repository has a copy of the revision."""
1236
return revision_id in self.has_revisions((revision_id,))
1239
def has_revisions(self, revision_ids):
1240
"""Probe to find out the presence of multiple revisions.
1242
:param revision_ids: An iterable of revision_ids.
1243
:return: A set of the revision_ids that were present.
1245
parent_map = self.revisions.get_parent_map(
1246
[(rev_id,) for rev_id in revision_ids])
1248
if _mod_revision.NULL_REVISION in revision_ids:
1249
result.add(_mod_revision.NULL_REVISION)
1250
result.update([key[0] for key in parent_map])
1254
def get_revision(self, revision_id):
1255
"""Return the Revision object for a named revision."""
1256
return self.get_revisions([revision_id])[0]
1259
def get_revision_reconcile(self, revision_id):
1260
"""'reconcile' helper routine that allows access to a revision always.
1262
This variant of get_revision does not cross check the weave graph
1263
against the revision one as get_revision does: but it should only
1264
be used by reconcile, or reconcile-alike commands that are correcting
1265
or testing the revision graph.
1267
return self._get_revisions([revision_id])[0]
1270
def get_revisions(self, revision_ids):
1271
"""Get many revisions at once."""
1272
return self._get_revisions(revision_ids)
1275
def _get_revisions(self, revision_ids):
1276
"""Core work logic to get many revisions without sanity checks."""
1277
for rev_id in revision_ids:
1278
if not rev_id or not isinstance(rev_id, basestring):
1279
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1280
keys = [(key,) for key in revision_ids]
1281
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1283
for record in stream:
1284
if record.storage_kind == 'absent':
1285
raise errors.NoSuchRevision(self, record.key[0])
1286
text = record.get_bytes_as('fulltext')
1287
rev = self._serializer.read_revision_from_string(text)
1288
revs[record.key[0]] = rev
1289
return [revs[revid] for revid in revision_ids]
1292
def get_revision_xml(self, revision_id):
1293
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1294
# would have already do it.
1295
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1296
rev = self.get_revision(revision_id)
1297
rev_tmp = cStringIO.StringIO()
1298
# the current serializer..
1299
self._serializer.write_revision(rev, rev_tmp)
1301
return rev_tmp.getvalue()
1303
def get_deltas_for_revisions(self, revisions):
1304
"""Produce a generator of revision deltas.
1306
Note that the input is a sequence of REVISIONS, not revision_ids.
1307
Trees will be held in memory until the generator exits.
1308
Each delta is relative to the revision's lefthand predecessor.
1310
required_trees = set()
1311
for revision in revisions:
1312
required_trees.add(revision.revision_id)
1313
required_trees.update(revision.parent_ids[:1])
1314
trees = dict((t.get_revision_id(), t) for
1315
t in self.revision_trees(required_trees))
1316
for revision in revisions:
1317
if not revision.parent_ids:
1318
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1320
old_tree = trees[revision.parent_ids[0]]
1321
yield trees[revision.revision_id].changes_from(old_tree)
1324
def get_revision_delta(self, revision_id):
1325
"""Return the delta for one revision.
1327
The delta is relative to the left-hand predecessor of the
1330
r = self.get_revision(revision_id)
1331
return list(self.get_deltas_for_revisions([r]))[0]
1334
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1335
signature = gpg_strategy.sign(plaintext)
1336
self.add_signature_text(revision_id, signature)
1339
def add_signature_text(self, revision_id, signature):
1340
self.signatures.add_lines((revision_id,), (),
1341
osutils.split_lines(signature))
1343
def find_text_key_references(self):
1344
"""Find the text key references within the repository.
1346
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1347
revision_ids. Each altered file-ids has the exact revision_ids that
1348
altered it listed explicitly.
1349
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1350
to whether they were referred to by the inventory of the
1351
revision_id that they contain. The inventory texts from all present
1352
revision ids are assessed to generate this report.
1354
revision_keys = self.revisions.keys()
1355
w = self.inventories
1356
pb = ui.ui_factory.nested_progress_bar()
1358
return self._find_text_key_references_from_xml_inventory_lines(
1359
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1363
def _find_text_key_references_from_xml_inventory_lines(self,
1365
"""Core routine for extracting references to texts from inventories.
1367
This performs the translation of xml lines to revision ids.
1369
:param line_iterator: An iterator of lines, origin_version_id
1370
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1371
to whether they were referred to by the inventory of the
1372
revision_id that they contain. Note that if that revision_id was
1373
not part of the line_iterator's output then False will be given -
1374
even though it may actually refer to that key.
1376
if not self._serializer.support_altered_by_hack:
1377
raise AssertionError(
1378
"_find_text_key_references_from_xml_inventory_lines only "
1379
"supported for branches which store inventory as unnested xml"
1380
", not on %r" % self)
1383
# this code needs to read every new line in every inventory for the
1384
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1385
# not present in one of those inventories is unnecessary but not
1386
# harmful because we are filtering by the revision id marker in the
1387
# inventory lines : we only select file ids altered in one of those
1388
# revisions. We don't need to see all lines in the inventory because
1389
# only those added in an inventory in rev X can contain a revision=X
1391
unescape_revid_cache = {}
1392
unescape_fileid_cache = {}
1394
# jam 20061218 In a big fetch, this handles hundreds of thousands
1395
# of lines, so it has had a lot of inlining and optimizing done.
1396
# Sorry that it is a little bit messy.
1397
# Move several functions to be local variables, since this is a long
1399
search = self._file_ids_altered_regex.search
1400
unescape = _unescape_xml
1401
setdefault = result.setdefault
1402
for line, line_key in line_iterator:
1403
match = search(line)
1406
# One call to match.group() returning multiple items is quite a
1407
# bit faster than 2 calls to match.group() each returning 1
1408
file_id, revision_id = match.group('file_id', 'revision_id')
1410
# Inlining the cache lookups helps a lot when you make 170,000
1411
# lines and 350k ids, versus 8.4 unique ids.
1412
# Using a cache helps in 2 ways:
1413
# 1) Avoids unnecessary decoding calls
1414
# 2) Re-uses cached strings, which helps in future set and
1416
# (2) is enough that removing encoding entirely along with
1417
# the cache (so we are using plain strings) results in no
1418
# performance improvement.
1420
revision_id = unescape_revid_cache[revision_id]
1422
unescaped = unescape(revision_id)
1423
unescape_revid_cache[revision_id] = unescaped
1424
revision_id = unescaped
1426
# Note that unconditionally unescaping means that we deserialise
1427
# every fileid, which for general 'pull' is not great, but we don't
1428
# really want to have some many fulltexts that this matters anyway.
1431
file_id = unescape_fileid_cache[file_id]
1433
unescaped = unescape(file_id)
1434
unescape_fileid_cache[file_id] = unescaped
1437
key = (file_id, revision_id)
1438
setdefault(key, False)
1439
if revision_id == line_key[-1]:
1443
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1445
"""Helper routine for fileids_altered_by_revision_ids.
1447
This performs the translation of xml lines to revision ids.
1449
:param line_iterator: An iterator of lines, origin_version_id
1450
:param revision_ids: The revision ids to filter for. This should be a
1451
set or other type which supports efficient __contains__ lookups, as
1452
the revision id from each parsed line will be looked up in the
1453
revision_ids filter.
1454
:return: a dictionary mapping altered file-ids to an iterable of
1455
revision_ids. Each altered file-ids has the exact revision_ids that
1456
altered it listed explicitly.
1459
setdefault = result.setdefault
1461
self._find_text_key_references_from_xml_inventory_lines(
1462
line_iterator).iterkeys():
1463
# once data is all ensured-consistent; then this is
1464
# if revision_id == version_id
1465
if key[-1:] in revision_ids:
1466
setdefault(key[0], set()).add(key[-1])
1469
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1470
"""Find the file ids and versions affected by revisions.
1472
:param revisions: an iterable containing revision ids.
1473
:param _inv_weave: The inventory weave from this repository or None.
1474
If None, the inventory weave will be opened automatically.
1475
:return: a dictionary mapping altered file-ids to an iterable of
1476
revision_ids. Each altered file-ids has the exact revision_ids that
1477
altered it listed explicitly.
1479
selected_keys = set((revid,) for revid in revision_ids)
1480
w = _inv_weave or self.inventories
1481
pb = ui.ui_factory.nested_progress_bar()
1483
return self._find_file_ids_from_xml_inventory_lines(
1484
w.iter_lines_added_or_present_in_keys(
1485
selected_keys, pb=pb),
1490
def iter_files_bytes(self, desired_files):
1491
"""Iterate through file versions.
1493
Files will not necessarily be returned in the order they occur in
1494
desired_files. No specific order is guaranteed.
1496
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1497
value supplied by the caller as part of desired_files. It should
1498
uniquely identify the file version in the caller's context. (Examples:
1499
an index number or a TreeTransform trans_id.)
1501
bytes_iterator is an iterable of bytestrings for the file. The
1502
kind of iterable and length of the bytestrings are unspecified, but for
1503
this implementation, it is a list of bytes produced by
1504
VersionedFile.get_record_stream().
1506
:param desired_files: a list of (file_id, revision_id, identifier)
1510
for file_id, revision_id, callable_data in desired_files:
1511
text_keys[(file_id, revision_id)] = callable_data
1512
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1513
if record.storage_kind == 'absent':
1514
raise errors.RevisionNotPresent(record.key, self)
1515
yield text_keys[record.key], record.get_bytes_as('fulltext')
1517
def _generate_text_key_index(self, text_key_references=None,
1519
"""Generate a new text key index for the repository.
1521
This is an expensive function that will take considerable time to run.
1523
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1524
list of parents, also text keys. When a given key has no parents,
1525
the parents list will be [NULL_REVISION].
1527
# All revisions, to find inventory parents.
1528
if ancestors is None:
1529
graph = self.get_graph()
1530
ancestors = graph.get_parent_map(self.all_revision_ids())
1531
if text_key_references is None:
1532
text_key_references = self.find_text_key_references()
1533
pb = ui.ui_factory.nested_progress_bar()
1535
return self._do_generate_text_key_index(ancestors,
1536
text_key_references, pb)
1540
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1541
"""Helper for _generate_text_key_index to avoid deep nesting."""
1542
revision_order = tsort.topo_sort(ancestors)
1543
invalid_keys = set()
1545
for revision_id in revision_order:
1546
revision_keys[revision_id] = set()
1547
text_count = len(text_key_references)
1548
# a cache of the text keys to allow reuse; costs a dict of all the
1549
# keys, but saves a 2-tuple for every child of a given key.
1551
for text_key, valid in text_key_references.iteritems():
1553
invalid_keys.add(text_key)
1555
revision_keys[text_key[1]].add(text_key)
1556
text_key_cache[text_key] = text_key
1557
del text_key_references
1559
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1560
NULL_REVISION = _mod_revision.NULL_REVISION
1561
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1562
# too small for large or very branchy trees. However, for 55K path
1563
# trees, it would be easy to use too much memory trivially. Ideally we
1564
# could gauge this by looking at available real memory etc, but this is
1565
# always a tricky proposition.
1566
inventory_cache = lru_cache.LRUCache(10)
1567
batch_size = 10 # should be ~150MB on a 55K path tree
1568
batch_count = len(revision_order) / batch_size + 1
1570
pb.update("Calculating text parents.", processed_texts, text_count)
1571
for offset in xrange(batch_count):
1572
to_query = revision_order[offset * batch_size:(offset + 1) *
1576
for rev_tree in self.revision_trees(to_query):
1577
revision_id = rev_tree.get_revision_id()
1578
parent_ids = ancestors[revision_id]
1579
for text_key in revision_keys[revision_id]:
1580
pb.update("Calculating text parents.", processed_texts)
1581
processed_texts += 1
1582
candidate_parents = []
1583
for parent_id in parent_ids:
1584
parent_text_key = (text_key[0], parent_id)
1586
check_parent = parent_text_key not in \
1587
revision_keys[parent_id]
1589
# the parent parent_id is a ghost:
1590
check_parent = False
1591
# truncate the derived graph against this ghost.
1592
parent_text_key = None
1594
# look at the parent commit details inventories to
1595
# determine possible candidates in the per file graph.
1598
inv = inventory_cache[parent_id]
1600
inv = self.revision_tree(parent_id).inventory
1601
inventory_cache[parent_id] = inv
1602
parent_entry = inv._byid.get(text_key[0], None)
1603
if parent_entry is not None:
1605
text_key[0], parent_entry.revision)
1607
parent_text_key = None
1608
if parent_text_key is not None:
1609
candidate_parents.append(
1610
text_key_cache[parent_text_key])
1611
parent_heads = text_graph.heads(candidate_parents)
1612
new_parents = list(parent_heads)
1613
new_parents.sort(key=lambda x:candidate_parents.index(x))
1614
if new_parents == []:
1615
new_parents = [NULL_REVISION]
1616
text_index[text_key] = new_parents
1618
for text_key in invalid_keys:
1619
text_index[text_key] = [NULL_REVISION]
1622
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1623
"""Get an iterable listing the keys of all the data introduced by a set
1626
The keys will be ordered so that the corresponding items can be safely
1627
fetched and inserted in that order.
1629
:returns: An iterable producing tuples of (knit-kind, file-id,
1630
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1631
'revisions'. file-id is None unless knit-kind is 'file'.
1633
# XXX: it's a bit weird to control the inventory weave caching in this
1634
# generator. Ideally the caching would be done in fetch.py I think. Or
1635
# maybe this generator should explicitly have the contract that it
1636
# should not be iterated until the previously yielded item has been
1638
inv_w = self.inventories
1640
# file ids that changed
1641
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1643
num_file_ids = len(file_ids)
1644
for file_id, altered_versions in file_ids.iteritems():
1645
if _files_pb is not None:
1646
_files_pb.update("fetch texts", count, num_file_ids)
1648
yield ("file", file_id, altered_versions)
1649
# We're done with the files_pb. Note that it finished by the caller,
1650
# just as it was created by the caller.
1654
yield ("inventory", None, revision_ids)
1657
# XXX: Note ATM no callers actually pay attention to this return
1658
# instead they just use the list of revision ids and ignore
1659
# missing sigs. Consider removing this work entirely
1660
revisions_with_signatures = set(self.signatures.get_parent_map(
1661
[(r,) for r in revision_ids]))
1662
revisions_with_signatures = set(
1663
[r for (r,) in revisions_with_signatures])
1664
revisions_with_signatures.intersection_update(revision_ids)
1665
yield ("signatures", None, revisions_with_signatures)
1668
yield ("revisions", None, revision_ids)
1671
def get_inventory(self, revision_id):
1672
"""Get Inventory object by revision id."""
1673
return self.iter_inventories([revision_id]).next()
1675
def iter_inventories(self, revision_ids):
1676
"""Get many inventories by revision_ids.
1678
This will buffer some or all of the texts used in constructing the
1679
inventories in memory, but will only parse a single inventory at a
1682
:return: An iterator of inventories.
1684
if ((None in revision_ids)
1685
or (_mod_revision.NULL_REVISION in revision_ids)):
1686
raise ValueError('cannot get null revision inventory')
1687
return self._iter_inventories(revision_ids)
1689
def _iter_inventories(self, revision_ids):
1690
"""single-document based inventory iteration."""
1691
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1692
yield self.deserialise_inventory(revision_id, text)
1694
def _iter_inventory_xmls(self, revision_ids):
1695
keys = [(revision_id,) for revision_id in revision_ids]
1696
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1698
for record in stream:
1699
if record.storage_kind != 'absent':
1700
text_chunks[record.key] = record.get_bytes_as('chunked')
1702
raise errors.NoSuchRevision(self, record.key)
1704
chunks = text_chunks.pop(key)
1705
yield ''.join(chunks), key[-1]
1707
def deserialise_inventory(self, revision_id, xml):
1708
"""Transform the xml into an inventory object.
1710
:param revision_id: The expected revision id of the inventory.
1711
:param xml: A serialised inventory.
1713
result = self._serializer.read_inventory_from_string(xml, revision_id,
1714
entry_cache=self._inventory_entry_cache)
1715
if result.revision_id != revision_id:
1716
raise AssertionError('revision id mismatch %s != %s' % (
1717
result.revision_id, revision_id))
1720
def serialise_inventory(self, inv):
1721
return self._serializer.write_inventory_to_string(inv)
1723
def _serialise_inventory_to_lines(self, inv):
1724
return self._serializer.write_inventory_to_lines(inv)
1726
def get_serializer_format(self):
1727
return self._serializer.format_num
1730
def get_inventory_xml(self, revision_id):
1731
"""Get inventory XML as a file object."""
1732
texts = self._iter_inventory_xmls([revision_id])
1734
text, revision_id = texts.next()
1735
except StopIteration:
1736
raise errors.HistoryMissing(self, 'inventory', revision_id)
1740
def get_inventory_sha1(self, revision_id):
1741
"""Return the sha1 hash of the inventory entry
1743
return self.get_revision(revision_id).inventory_sha1
1745
def iter_reverse_revision_history(self, revision_id):
1746
"""Iterate backwards through revision ids in the lefthand history
1748
:param revision_id: The revision id to start with. All its lefthand
1749
ancestors will be traversed.
1751
graph = self.get_graph()
1752
next_id = revision_id
1754
if next_id in (None, _mod_revision.NULL_REVISION):
1757
# Note: The following line may raise KeyError in the event of
1758
# truncated history. We decided not to have a try:except:raise
1759
# RevisionNotPresent here until we see a use for it, because of the
1760
# cost in an inner loop that is by its very nature O(history).
1761
# Robert Collins 20080326
1762
parents = graph.get_parent_map([next_id])[next_id]
1763
if len(parents) == 0:
1766
next_id = parents[0]
1769
def get_revision_inventory(self, revision_id):
1770
"""Return inventory of a past revision."""
1771
# TODO: Unify this with get_inventory()
1772
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1773
# must be the same as its revision, so this is trivial.
1774
if revision_id is None:
1775
# This does not make sense: if there is no revision,
1776
# then it is the current tree inventory surely ?!
1777
# and thus get_root_id() is something that looks at the last
1778
# commit on the branch, and the get_root_id is an inventory check.
1779
raise NotImplementedError
1780
# return Inventory(self.get_root_id())
1782
return self.get_inventory(revision_id)
1784
def is_shared(self):
1785
"""Return True if this repository is flagged as a shared repository."""
1786
raise NotImplementedError(self.is_shared)
1789
def reconcile(self, other=None, thorough=False):
1790
"""Reconcile this repository."""
1791
from bzrlib.reconcile import RepoReconciler
1792
reconciler = RepoReconciler(self, thorough=thorough)
1793
reconciler.reconcile()
1796
def _refresh_data(self):
1797
"""Helper called from lock_* to ensure coherency with disk.
1799
The default implementation does nothing; it is however possible
1800
for repositories to maintain loaded indices across multiple locks
1801
by checking inside their implementation of this method to see
1802
whether their indices are still valid. This depends of course on
1803
the disk format being validatable in this manner.
1807
def revision_tree(self, revision_id):
1808
"""Return Tree for a revision on this branch.
1810
`revision_id` may be NULL_REVISION for the empty tree revision.
1812
revision_id = _mod_revision.ensure_null(revision_id)
1813
# TODO: refactor this to use an existing revision object
1814
# so we don't need to read it in twice.
1815
if revision_id == _mod_revision.NULL_REVISION:
1816
return RevisionTree(self, Inventory(root_id=None),
1817
_mod_revision.NULL_REVISION)
1819
inv = self.get_revision_inventory(revision_id)
1820
return RevisionTree(self, inv, revision_id)
1822
def revision_trees(self, revision_ids):
1823
"""Return Tree for a revision on this branch.
1825
`revision_id` may not be None or 'null:'"""
1826
inventories = self.iter_inventories(revision_ids)
1827
for inv in inventories:
1828
yield RevisionTree(self, inv, inv.revision_id)
1831
def get_ancestry(self, revision_id, topo_sorted=True):
1832
"""Return a list of revision-ids integrated by a revision.
1834
The first element of the list is always None, indicating the origin
1835
revision. This might change when we have history horizons, or
1836
perhaps we should have a new API.
1838
This is topologically sorted.
1840
if _mod_revision.is_null(revision_id):
1842
if not self.has_revision(revision_id):
1843
raise errors.NoSuchRevision(self, revision_id)
1844
graph = self.get_graph()
1846
search = graph._make_breadth_first_searcher([revision_id])
1849
found, ghosts = search.next_with_ghosts()
1850
except StopIteration:
1853
if _mod_revision.NULL_REVISION in keys:
1854
keys.remove(_mod_revision.NULL_REVISION)
1856
parent_map = graph.get_parent_map(keys)
1857
keys = tsort.topo_sort(parent_map)
1858
return [None] + list(keys)
1861
"""Compress the data within the repository.
1863
This operation only makes sense for some repository types. For other
1864
types it should be a no-op that just returns.
1866
This stub method does not require a lock, but subclasses should use
1867
@needs_write_lock as this is a long running call its reasonable to
1868
implicitly lock for the user.
1872
@deprecated_method(one_six)
1873
def print_file(self, file, revision_id):
1874
"""Print `file` to stdout.
1876
FIXME RBC 20060125 as John Meinel points out this is a bad api
1877
- it writes to stdout, it assumes that that is valid etc. Fix
1878
by creating a new more flexible convenience function.
1880
tree = self.revision_tree(revision_id)
1881
# use inventory as it was in that revision
1882
file_id = tree.inventory.path2id(file)
1884
# TODO: jam 20060427 Write a test for this code path
1885
# it had a bug in it, and was raising the wrong
1887
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1888
tree.print_file(file_id)
1890
def get_transaction(self):
1891
return self.control_files.get_transaction()
1893
@deprecated_method(one_one)
1894
def get_parents(self, revision_ids):
1895
"""See StackedParentsProvider.get_parents"""
1896
parent_map = self.get_parent_map(revision_ids)
1897
return [parent_map.get(r, None) for r in revision_ids]
1899
def get_parent_map(self, revision_ids):
1900
"""See graph._StackedParentsProvider.get_parent_map"""
1901
# revisions index works in keys; this just works in revisions
1902
# therefore wrap and unwrap
1905
for revision_id in revision_ids:
1906
if revision_id == _mod_revision.NULL_REVISION:
1907
result[revision_id] = ()
1908
elif revision_id is None:
1909
raise ValueError('get_parent_map(None) is not valid')
1911
query_keys.append((revision_id ,))
1912
for ((revision_id,), parent_keys) in \
1913
self.revisions.get_parent_map(query_keys).iteritems():
1915
result[revision_id] = tuple(parent_revid
1916
for (parent_revid,) in parent_keys)
1918
result[revision_id] = (_mod_revision.NULL_REVISION,)
1921
def _make_parents_provider(self):
1924
def get_graph(self, other_repository=None):
1925
"""Return the graph walker for this repository format"""
1926
parents_provider = self._make_parents_provider()
1927
if (other_repository is not None and
1928
not self.has_same_location(other_repository)):
1929
parents_provider = graph._StackedParentsProvider(
1930
[parents_provider, other_repository._make_parents_provider()])
1931
return graph.Graph(parents_provider)
1933
def _get_versioned_file_checker(self):
1934
"""Return an object suitable for checking versioned files."""
1935
return _VersionedFileChecker(self)
1937
def revision_ids_to_search_result(self, result_set):
1938
"""Convert a set of revision ids to a graph SearchResult."""
1939
result_parents = set()
1940
for parents in self.get_graph().get_parent_map(
1941
result_set).itervalues():
1942
result_parents.update(parents)
1943
included_keys = result_set.intersection(result_parents)
1944
start_keys = result_set.difference(included_keys)
1945
exclude_keys = result_parents.difference(result_set)
1946
result = graph.SearchResult(start_keys, exclude_keys,
1947
len(result_set), result_set)
1951
def set_make_working_trees(self, new_value):
1952
"""Set the policy flag for making working trees when creating branches.
1954
This only applies to branches that use this repository.
1956
The default is 'True'.
1957
:param new_value: True to restore the default, False to disable making
1960
raise NotImplementedError(self.set_make_working_trees)
1962
def make_working_trees(self):
1963
"""Returns the policy for making working trees on new branches."""
1964
raise NotImplementedError(self.make_working_trees)
1967
def sign_revision(self, revision_id, gpg_strategy):
1968
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1969
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1972
def has_signature_for_revision_id(self, revision_id):
1973
"""Query for a revision signature for revision_id in the repository."""
1974
if not self.has_revision(revision_id):
1975
raise errors.NoSuchRevision(self, revision_id)
1976
sig_present = (1 == len(
1977
self.signatures.get_parent_map([(revision_id,)])))
1981
def get_signature_text(self, revision_id):
1982
"""Return the text for a signature."""
1983
stream = self.signatures.get_record_stream([(revision_id,)],
1985
record = stream.next()
1986
if record.storage_kind == 'absent':
1987
raise errors.NoSuchRevision(self, revision_id)
1988
return record.get_bytes_as('fulltext')
1991
def check(self, revision_ids=None):
1992
"""Check consistency of all history of given revision_ids.
1994
Different repository implementations should override _check().
1996
:param revision_ids: A non-empty list of revision_ids whose ancestry
1997
will be checked. Typically the last revision_id of a branch.
1999
return self._check(revision_ids)
2001
def _check(self, revision_ids):
2002
result = check.Check(self)
2006
def _warn_if_deprecated(self):
2007
global _deprecation_warning_done
2008
if _deprecation_warning_done:
2010
_deprecation_warning_done = True
2011
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2012
% (self._format, self.bzrdir.transport.base))
2014
def supports_rich_root(self):
2015
return self._format.rich_root_data
2017
def _check_ascii_revisionid(self, revision_id, method):
2018
"""Private helper for ascii-only repositories."""
2019
# weave repositories refuse to store revisionids that are non-ascii.
2020
if revision_id is not None:
2021
# weaves require ascii revision ids.
2022
if isinstance(revision_id, unicode):
2024
revision_id.encode('ascii')
2025
except UnicodeEncodeError:
2026
raise errors.NonAsciiRevisionId(method, self)
2029
revision_id.decode('ascii')
2030
except UnicodeDecodeError:
2031
raise errors.NonAsciiRevisionId(method, self)
2033
def revision_graph_can_have_wrong_parents(self):
2034
"""Is it possible for this repository to have a revision graph with
2037
If True, then this repository must also implement
2038
_find_inconsistent_revision_parents so that check and reconcile can
2039
check for inconsistencies before proceeding with other checks that may
2040
depend on the revision index being consistent.
2042
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2045
# remove these delegates a while after bzr 0.15
2046
def __make_delegated(name, from_module):
2047
def _deprecated_repository_forwarder():
2048
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2049
% (name, from_module),
2052
m = __import__(from_module, globals(), locals(), [name])
2054
return getattr(m, name)
2055
except AttributeError:
2056
raise AttributeError('module %s has no name %s'
2058
globals()[name] = _deprecated_repository_forwarder
2061
'AllInOneRepository',
2062
'WeaveMetaDirRepository',
2063
'PreSplitOutRepositoryFormat',
2064
'RepositoryFormat4',
2065
'RepositoryFormat5',
2066
'RepositoryFormat6',
2067
'RepositoryFormat7',
2069
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2073
'RepositoryFormatKnit',
2074
'RepositoryFormatKnit1',
2076
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2079
def install_revision(repository, rev, revision_tree):
2080
"""Install all revision data into a repository."""
2081
install_revisions(repository, [(rev, revision_tree, None)])
2084
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2085
"""Install all revision data into a repository.
2087
Accepts an iterable of revision, tree, signature tuples. The signature
2090
repository.start_write_group()
2092
for n, (revision, revision_tree, signature) in enumerate(iterable):
2093
_install_revision(repository, revision, revision_tree, signature)
2095
pb.update('Transferring revisions', n + 1, num_revisions)
2097
repository.abort_write_group()
2100
repository.commit_write_group()
2103
def _install_revision(repository, rev, revision_tree, signature):
2104
"""Install all revision data into a repository."""
2105
present_parents = []
2107
for p_id in rev.parent_ids:
2108
if repository.has_revision(p_id):
2109
present_parents.append(p_id)
2110
parent_trees[p_id] = repository.revision_tree(p_id)
2112
parent_trees[p_id] = repository.revision_tree(
2113
_mod_revision.NULL_REVISION)
2115
inv = revision_tree.inventory
2116
entries = inv.iter_entries()
2117
# backwards compatibility hack: skip the root id.
2118
if not repository.supports_rich_root():
2119
path, root = entries.next()
2120
if root.revision != rev.revision_id:
2121
raise errors.IncompatibleRevision(repr(repository))
2123
for path, ie in entries:
2124
text_keys[(ie.file_id, ie.revision)] = ie
2125
text_parent_map = repository.texts.get_parent_map(text_keys)
2126
missing_texts = set(text_keys) - set(text_parent_map)
2127
# Add the texts that are not already present
2128
for text_key in missing_texts:
2129
ie = text_keys[text_key]
2131
# FIXME: TODO: The following loop overlaps/duplicates that done by
2132
# commit to determine parents. There is a latent/real bug here where
2133
# the parents inserted are not those commit would do - in particular
2134
# they are not filtered by heads(). RBC, AB
2135
for revision, tree in parent_trees.iteritems():
2136
if ie.file_id not in tree:
2138
parent_id = tree.inventory[ie.file_id].revision
2139
if parent_id in text_parents:
2141
text_parents.append((ie.file_id, parent_id))
2142
lines = revision_tree.get_file(ie.file_id).readlines()
2143
repository.texts.add_lines(text_key, text_parents, lines)
2145
# install the inventory
2146
repository.add_inventory(rev.revision_id, inv, present_parents)
2147
except errors.RevisionAlreadyPresent:
2149
if signature is not None:
2150
repository.add_signature_text(rev.revision_id, signature)
2151
repository.add_revision(rev.revision_id, rev, inv)
2154
class MetaDirRepository(Repository):
2155
"""Repositories in the new meta-dir layout.
2157
:ivar _transport: Transport for access to repository control files,
2158
typically pointing to .bzr/repository.
2161
def __init__(self, _format, a_bzrdir, control_files):
2162
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2163
self._transport = control_files._transport
2165
def is_shared(self):
2166
"""Return True if this repository is flagged as a shared repository."""
2167
return self._transport.has('shared-storage')
2170
def set_make_working_trees(self, new_value):
2171
"""Set the policy flag for making working trees when creating branches.
2173
This only applies to branches that use this repository.
2175
The default is 'True'.
2176
:param new_value: True to restore the default, False to disable making
2181
self._transport.delete('no-working-trees')
2182
except errors.NoSuchFile:
2185
self._transport.put_bytes('no-working-trees', '',
2186
mode=self.bzrdir._get_file_mode())
2188
def make_working_trees(self):
2189
"""Returns the policy for making working trees on new branches."""
2190
return not self._transport.has('no-working-trees')
2193
class MetaDirVersionedFileRepository(MetaDirRepository):
2194
"""Repositories in a meta-dir, that work via versioned file objects."""
2196
def __init__(self, _format, a_bzrdir, control_files):
2197
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2201
network_format_registry = registry.FormatRegistry()
2202
"""Registry of formats indexed by their network name.
2204
The network name for a repository format is an identifier that can be used when
2205
referring to formats with smart server operations. See
2206
RepositoryFormat.network_name() for more detail.
2210
format_registry = registry.FormatRegistry(network_format_registry)
2211
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2213
This can contain either format instances themselves, or classes/factories that
2214
can be called to obtain one.
2218
#####################################################################
2219
# Repository Formats
2221
class RepositoryFormat(object):
2222
"""A repository format.
2224
Formats provide four things:
2225
* An initialization routine to construct repository data on disk.
2226
* a optional format string which is used when the BzrDir supports
2228
* an open routine which returns a Repository instance.
2229
* A network name for referring to the format in smart server RPC
2232
There is one and only one Format subclass for each on-disk format. But
2233
there can be one Repository subclass that is used for several different
2234
formats. The _format attribute on a Repository instance can be used to
2235
determine the disk format.
2237
Formats are placed in a registry by their format string for reference
2238
during opening. These should be subclasses of RepositoryFormat for
2241
Once a format is deprecated, just deprecate the initialize and open
2242
methods on the format class. Do not deprecate the object, as the
2243
object may be created even when a repository instnace hasn't been
2246
Common instance attributes:
2247
_matchingbzrdir - the bzrdir format that the repository format was
2248
originally written to work with. This can be used if manually
2249
constructing a bzrdir and repository, or more commonly for test suite
2253
# Set to True or False in derived classes. True indicates that the format
2254
# supports ghosts gracefully.
2255
supports_ghosts = None
2256
# Can this repository be given external locations to lookup additional
2257
# data. Set to True or False in derived classes.
2258
supports_external_lookups = None
2259
# What order should fetch operations request streams in?
2260
# The default is unordered as that is the cheapest for an origin to
2262
_fetch_order = 'unordered'
2263
# Does this repository format use deltas that can be fetched as-deltas ?
2264
# (E.g. knits, where the knit deltas can be transplanted intact.
2265
# We default to False, which will ensure that enough data to get
2266
# a full text out of any fetch stream will be grabbed.
2267
_fetch_uses_deltas = False
2268
# Should fetch trigger a reconcile after the fetch? Only needed for
2269
# some repository formats that can suffer internal inconsistencies.
2270
_fetch_reconcile = False
2273
return "<%s>" % self.__class__.__name__
2275
def __eq__(self, other):
2276
# format objects are generally stateless
2277
return isinstance(other, self.__class__)
2279
def __ne__(self, other):
2280
return not self == other
2283
def find_format(klass, a_bzrdir):
2284
"""Return the format for the repository object in a_bzrdir.
2286
This is used by bzr native formats that have a "format" file in
2287
the repository. Other methods may be used by different types of
2291
transport = a_bzrdir.get_repository_transport(None)
2292
format_string = transport.get("format").read()
2293
return format_registry.get(format_string)
2294
except errors.NoSuchFile:
2295
raise errors.NoRepositoryPresent(a_bzrdir)
2297
raise errors.UnknownFormatError(format=format_string,
2301
def register_format(klass, format):
2302
format_registry.register(format.get_format_string(), format)
2305
def unregister_format(klass, format):
2306
format_registry.remove(format.get_format_string())
2309
def get_default_format(klass):
2310
"""Return the current default format."""
2311
from bzrlib import bzrdir
2312
return bzrdir.format_registry.make_bzrdir('default').repository_format
2314
def get_format_string(self):
2315
"""Return the ASCII format string that identifies this format.
2317
Note that in pre format ?? repositories the format string is
2318
not permitted nor written to disk.
2320
raise NotImplementedError(self.get_format_string)
2322
def get_format_description(self):
2323
"""Return the short description for this format."""
2324
raise NotImplementedError(self.get_format_description)
2326
# TODO: this shouldn't be in the base class, it's specific to things that
2327
# use weaves or knits -- mbp 20070207
2328
def _get_versioned_file_store(self,
2333
versionedfile_class=None,
2334
versionedfile_kwargs={},
2336
if versionedfile_class is None:
2337
versionedfile_class = self._versionedfile_class
2338
weave_transport = control_files._transport.clone(name)
2339
dir_mode = control_files._dir_mode
2340
file_mode = control_files._file_mode
2341
return VersionedFileStore(weave_transport, prefixed=prefixed,
2343
file_mode=file_mode,
2344
versionedfile_class=versionedfile_class,
2345
versionedfile_kwargs=versionedfile_kwargs,
2348
def initialize(self, a_bzrdir, shared=False):
2349
"""Initialize a repository of this format in a_bzrdir.
2351
:param a_bzrdir: The bzrdir to put the new repository in it.
2352
:param shared: The repository should be initialized as a sharable one.
2353
:returns: The new repository object.
2355
This may raise UninitializableFormat if shared repository are not
2356
compatible the a_bzrdir.
2358
raise NotImplementedError(self.initialize)
2360
def is_supported(self):
2361
"""Is this format supported?
2363
Supported formats must be initializable and openable.
2364
Unsupported formats may not support initialization or committing or
2365
some other features depending on the reason for not being supported.
2369
def network_name(self):
2370
"""A simple byte string uniquely identifying this format for RPC calls.
2372
MetaDir repository formats use their disk format string to identify the
2373
repository over the wire. All in one formats such as bzr < 0.8, and
2374
foreign formats like svn/git and hg should use some marker which is
2375
unique and immutable.
2377
raise NotImplementedError(self.network_name)
2379
def check_conversion_target(self, target_format):
2380
raise NotImplementedError(self.check_conversion_target)
2382
def open(self, a_bzrdir, _found=False):
2383
"""Return an instance of this format for the bzrdir a_bzrdir.
2385
_found is a private parameter, do not use it.
2387
raise NotImplementedError(self.open)
2390
class MetaDirRepositoryFormat(RepositoryFormat):
2391
"""Common base class for the new repositories using the metadir layout."""
2393
rich_root_data = False
2394
supports_tree_reference = False
2395
supports_external_lookups = False
2398
def _matchingbzrdir(self):
2399
matching = bzrdir.BzrDirMetaFormat1()
2400
matching.repository_format = self
2404
super(MetaDirRepositoryFormat, self).__init__()
2406
def _create_control_files(self, a_bzrdir):
2407
"""Create the required files and the initial control_files object."""
2408
# FIXME: RBC 20060125 don't peek under the covers
2409
# NB: no need to escape relative paths that are url safe.
2410
repository_transport = a_bzrdir.get_repository_transport(self)
2411
control_files = lockable_files.LockableFiles(repository_transport,
2412
'lock', lockdir.LockDir)
2413
control_files.create_lock()
2414
return control_files
2416
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2417
"""Upload the initial blank content."""
2418
control_files = self._create_control_files(a_bzrdir)
2419
control_files.lock_write()
2420
transport = control_files._transport
2422
utf8_files += [('shared-storage', '')]
2424
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2425
for (filename, content_stream) in files:
2426
transport.put_file(filename, content_stream,
2427
mode=a_bzrdir._get_file_mode())
2428
for (filename, content_bytes) in utf8_files:
2429
transport.put_bytes_non_atomic(filename, content_bytes,
2430
mode=a_bzrdir._get_file_mode())
2432
control_files.unlock()
2434
def network_name(self):
2435
"""Metadir formats have matching disk and network format strings."""
2436
return self.get_format_string()
2439
# Pre-0.8 formats that don't have a disk format string (because they are
2440
# versioned by the matching control directory). We use the control directories
2441
# disk format string as a key for the network_name because they meet the
2442
# constraints (simple string, unique, immmutable).
2443
network_format_registry.register_lazy(
2444
"Bazaar-NG branch, format 5\n",
2445
'bzrlib.repofmt.weaverepo',
2446
'RepositoryFormat5',
2448
network_format_registry.register_lazy(
2449
"Bazaar-NG branch, format 6\n",
2450
'bzrlib.repofmt.weaverepo',
2451
'RepositoryFormat6',
2454
# formats which have no format string are not discoverable or independently
2455
# creatable on disk, so are not registered in format_registry. They're
2456
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2457
# needed, it's constructed directly by the BzrDir. Non-native formats where
2458
# the repository is not separately opened are similar.
2460
format_registry.register_lazy(
2461
'Bazaar-NG Repository format 7',
2462
'bzrlib.repofmt.weaverepo',
2466
format_registry.register_lazy(
2467
'Bazaar-NG Knit Repository Format 1',
2468
'bzrlib.repofmt.knitrepo',
2469
'RepositoryFormatKnit1',
2472
format_registry.register_lazy(
2473
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2474
'bzrlib.repofmt.knitrepo',
2475
'RepositoryFormatKnit3',
2478
format_registry.register_lazy(
2479
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2480
'bzrlib.repofmt.knitrepo',
2481
'RepositoryFormatKnit4',
2484
# Pack-based formats. There is one format for pre-subtrees, and one for
2485
# post-subtrees to allow ease of testing.
2486
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2487
format_registry.register_lazy(
2488
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2489
'bzrlib.repofmt.pack_repo',
2490
'RepositoryFormatKnitPack1',
2492
format_registry.register_lazy(
2493
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2494
'bzrlib.repofmt.pack_repo',
2495
'RepositoryFormatKnitPack3',
2497
format_registry.register_lazy(
2498
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2499
'bzrlib.repofmt.pack_repo',
2500
'RepositoryFormatKnitPack4',
2502
format_registry.register_lazy(
2503
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2504
'bzrlib.repofmt.pack_repo',
2505
'RepositoryFormatKnitPack5',
2507
format_registry.register_lazy(
2508
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2509
'bzrlib.repofmt.pack_repo',
2510
'RepositoryFormatKnitPack5RichRoot',
2512
format_registry.register_lazy(
2513
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2514
'bzrlib.repofmt.pack_repo',
2515
'RepositoryFormatKnitPack5RichRootBroken',
2517
format_registry.register_lazy(
2518
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2519
'bzrlib.repofmt.pack_repo',
2520
'RepositoryFormatKnitPack6',
2522
format_registry.register_lazy(
2523
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2524
'bzrlib.repofmt.pack_repo',
2525
'RepositoryFormatKnitPack6RichRoot',
2528
# Development formats.
2529
# 1.7->1.8 go below here
2530
format_registry.register_lazy(
2531
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2532
'bzrlib.repofmt.pack_repo',
2533
'RepositoryFormatPackDevelopment2',
2535
format_registry.register_lazy(
2536
("Bazaar development format 2 with subtree support "
2537
"(needs bzr.dev from before 1.8)\n"),
2538
'bzrlib.repofmt.pack_repo',
2539
'RepositoryFormatPackDevelopment2Subtree',
2543
class InterRepository(InterObject):
2544
"""This class represents operations taking place between two repositories.
2546
Its instances have methods like copy_content and fetch, and contain
2547
references to the source and target repositories these operations can be
2550
Often we will provide convenience methods on 'repository' which carry out
2551
operations with another repository - they will always forward to
2552
InterRepository.get(other).method_name(parameters).
2555
_walk_to_common_revisions_batch_size = 1
2557
"""The available optimised InterRepository types."""
2559
def __init__(self, source, target):
2560
InterObject.__init__(self, source, target)
2561
# These two attributes may be overridden by e.g. InterOtherToRemote to
2562
# provide a faster implementation.
2563
self.target_get_graph = self.target.get_graph
2564
self.target_get_parent_map = self.target.get_parent_map
2567
def copy_content(self, revision_id=None):
2568
"""Make a complete copy of the content in self into destination.
2570
This is a destructive operation! Do not use it on existing
2573
:param revision_id: Only copy the content needed to construct
2574
revision_id and its parents.
2577
self.target.set_make_working_trees(self.source.make_working_trees())
2578
except NotImplementedError:
2580
self.target.fetch(self.source, revision_id=revision_id)
2582
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2583
"""Fetch the content required to construct revision_id.
2585
The content is copied from self.source to self.target.
2587
:param revision_id: if None all content is copied, if NULL_REVISION no
2589
:param pb: optional progress bar to use for progress reports. If not
2590
provided a default one will be created.
2593
from bzrlib.fetch import RepoFetcher
2594
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2595
self.source, self.source._format, self.target,
2596
self.target._format)
2597
f = RepoFetcher(to_repository=self.target,
2598
from_repository=self.source,
2599
last_revision=revision_id,
2600
pb=pb, find_ghosts=find_ghosts)
2602
def _walk_to_common_revisions(self, revision_ids):
2603
"""Walk out from revision_ids in source to revisions target has.
2605
:param revision_ids: The start point for the search.
2606
:return: A set of revision ids.
2608
target_graph = self.target_get_graph()
2609
revision_ids = frozenset(revision_ids)
2610
# Fast path for the case where all the revisions are already in the
2612
# (Although this does incur an extra round trip for the
2613
# fairly common case where the target doesn't already have the revision
2615
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2616
return graph.SearchResult(revision_ids, set(), 0, set())
2617
missing_revs = set()
2618
source_graph = self.source.get_graph()
2619
# ensure we don't pay silly lookup costs.
2620
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2621
null_set = frozenset([_mod_revision.NULL_REVISION])
2622
searcher_exhausted = False
2626
# Iterate the searcher until we have enough next_revs
2627
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2629
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2630
next_revs.update(next_revs_part)
2631
ghosts.update(ghosts_part)
2632
except StopIteration:
2633
searcher_exhausted = True
2635
# If there are ghosts in the source graph, and the caller asked for
2636
# them, make sure that they are present in the target.
2637
# We don't care about other ghosts as we can't fetch them and
2638
# haven't been asked to.
2639
ghosts_to_check = set(revision_ids.intersection(ghosts))
2640
revs_to_get = set(next_revs).union(ghosts_to_check)
2642
have_revs = set(target_graph.get_parent_map(revs_to_get))
2643
# we always have NULL_REVISION present.
2644
have_revs = have_revs.union(null_set)
2645
# Check if the target is missing any ghosts we need.
2646
ghosts_to_check.difference_update(have_revs)
2648
# One of the caller's revision_ids is a ghost in both the
2649
# source and the target.
2650
raise errors.NoSuchRevision(
2651
self.source, ghosts_to_check.pop())
2652
missing_revs.update(next_revs - have_revs)
2653
# Because we may have walked past the original stop point, make
2654
# sure everything is stopped
2655
stop_revs = searcher.find_seen_ancestors(have_revs)
2656
searcher.stop_searching_any(stop_revs)
2657
if searcher_exhausted:
2659
return searcher.get_result()
2661
@deprecated_method(one_two)
2663
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2664
"""Return the revision ids that source has that target does not.
2666
These are returned in topological order.
2668
:param revision_id: only return revision ids included by this
2670
:param find_ghosts: If True find missing revisions in deep history
2671
rather than just finding the surface difference.
2673
return list(self.search_missing_revision_ids(
2674
revision_id, find_ghosts).get_keys())
2677
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2678
"""Return the revision ids that source has that target does not.
2680
:param revision_id: only return revision ids included by this
2682
:param find_ghosts: If True find missing revisions in deep history
2683
rather than just finding the surface difference.
2684
:return: A bzrlib.graph.SearchResult.
2686
# stop searching at found target revisions.
2687
if not find_ghosts and revision_id is not None:
2688
return self._walk_to_common_revisions([revision_id])
2689
# generic, possibly worst case, slow code path.
2690
target_ids = set(self.target.all_revision_ids())
2691
if revision_id is not None:
2692
source_ids = self.source.get_ancestry(revision_id)
2693
if source_ids[0] is not None:
2694
raise AssertionError()
2697
source_ids = self.source.all_revision_ids()
2698
result_set = set(source_ids).difference(target_ids)
2699
return self.source.revision_ids_to_search_result(result_set)
2702
def _same_model(source, target):
2703
"""True if source and target have the same data representation.
2705
Note: this is always called on the base class; overriding it in a
2706
subclass will have no effect.
2709
InterRepository._assert_same_model(source, target)
2711
except errors.IncompatibleRepositories, e:
2715
def _assert_same_model(source, target):
2716
"""Raise an exception if two repositories do not use the same model.
2718
if source.supports_rich_root() != target.supports_rich_root():
2719
raise errors.IncompatibleRepositories(source, target,
2720
"different rich-root support")
2721
if source._serializer != target._serializer:
2722
raise errors.IncompatibleRepositories(source, target,
2723
"different serializers")
2726
class InterSameDataRepository(InterRepository):
2727
"""Code for converting between repositories that represent the same data.
2729
Data format and model must match for this to work.
2733
def _get_repo_format_to_test(self):
2734
"""Repository format for testing with.
2736
InterSameData can pull from subtree to subtree and from non-subtree to
2737
non-subtree, so we test this with the richest repository format.
2739
from bzrlib.repofmt import knitrepo
2740
return knitrepo.RepositoryFormatKnit3()
2743
def is_compatible(source, target):
2744
return InterRepository._same_model(source, target)
2747
class InterWeaveRepo(InterSameDataRepository):
2748
"""Optimised code paths between Weave based repositories.
2750
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2751
implemented lazy inter-object optimisation.
2755
def _get_repo_format_to_test(self):
2756
from bzrlib.repofmt import weaverepo
2757
return weaverepo.RepositoryFormat7()
2760
def is_compatible(source, target):
2761
"""Be compatible with known Weave formats.
2763
We don't test for the stores being of specific types because that
2764
could lead to confusing results, and there is no need to be
2767
from bzrlib.repofmt.weaverepo import (
2773
return (isinstance(source._format, (RepositoryFormat5,
2775
RepositoryFormat7)) and
2776
isinstance(target._format, (RepositoryFormat5,
2778
RepositoryFormat7)))
2779
except AttributeError:
2783
def copy_content(self, revision_id=None):
2784
"""See InterRepository.copy_content()."""
2785
# weave specific optimised path:
2787
self.target.set_make_working_trees(self.source.make_working_trees())
2788
except (errors.RepositoryUpgradeRequired, NotImplemented):
2790
# FIXME do not peek!
2791
if self.source._transport.listable():
2792
pb = ui.ui_factory.nested_progress_bar()
2794
self.target.texts.insert_record_stream(
2795
self.source.texts.get_record_stream(
2796
self.source.texts.keys(), 'topological', False))
2797
pb.update('copying inventory', 0, 1)
2798
self.target.inventories.insert_record_stream(
2799
self.source.inventories.get_record_stream(
2800
self.source.inventories.keys(), 'topological', False))
2801
self.target.signatures.insert_record_stream(
2802
self.source.signatures.get_record_stream(
2803
self.source.signatures.keys(),
2805
self.target.revisions.insert_record_stream(
2806
self.source.revisions.get_record_stream(
2807
self.source.revisions.keys(),
2808
'topological', True))
2812
self.target.fetch(self.source, revision_id=revision_id)
2815
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2816
"""See InterRepository.fetch()."""
2817
from bzrlib.fetch import RepoFetcher
2818
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2819
self.source, self.source._format, self.target, self.target._format)
2820
f = RepoFetcher(to_repository=self.target,
2821
from_repository=self.source,
2822
last_revision=revision_id,
2823
pb=pb, find_ghosts=find_ghosts)
2826
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2827
"""See InterRepository.missing_revision_ids()."""
2828
# we want all revisions to satisfy revision_id in source.
2829
# but we don't want to stat every file here and there.
2830
# we want then, all revisions other needs to satisfy revision_id
2831
# checked, but not those that we have locally.
2832
# so the first thing is to get a subset of the revisions to
2833
# satisfy revision_id in source, and then eliminate those that
2834
# we do already have.
2835
# this is slow on high latency connection to self, but as as this
2836
# disk format scales terribly for push anyway due to rewriting
2837
# inventory.weave, this is considered acceptable.
2839
if revision_id is not None:
2840
source_ids = self.source.get_ancestry(revision_id)
2841
if source_ids[0] is not None:
2842
raise AssertionError()
2845
source_ids = self.source._all_possible_ids()
2846
source_ids_set = set(source_ids)
2847
# source_ids is the worst possible case we may need to pull.
2848
# now we want to filter source_ids against what we actually
2849
# have in target, but don't try to check for existence where we know
2850
# we do not have a revision as that would be pointless.
2851
target_ids = set(self.target._all_possible_ids())
2852
possibly_present_revisions = target_ids.intersection(source_ids_set)
2853
actually_present_revisions = set(
2854
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2855
required_revisions = source_ids_set.difference(actually_present_revisions)
2856
if revision_id is not None:
2857
# we used get_ancestry to determine source_ids then we are assured all
2858
# revisions referenced are present as they are installed in topological order.
2859
# and the tip revision was validated by get_ancestry.
2860
result_set = required_revisions
2862
# if we just grabbed the possibly available ids, then
2863
# we only have an estimate of whats available and need to validate
2864
# that against the revision records.
2866
self.source._eliminate_revisions_not_present(required_revisions))
2867
return self.source.revision_ids_to_search_result(result_set)
2870
class InterKnitRepo(InterSameDataRepository):
2871
"""Optimised code paths between Knit based repositories."""
2874
def _get_repo_format_to_test(self):
2875
from bzrlib.repofmt import knitrepo
2876
return knitrepo.RepositoryFormatKnit1()
2879
def is_compatible(source, target):
2880
"""Be compatible with known Knit formats.
2882
We don't test for the stores being of specific types because that
2883
could lead to confusing results, and there is no need to be
2886
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2888
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2889
isinstance(target._format, RepositoryFormatKnit))
2890
except AttributeError:
2892
return are_knits and InterRepository._same_model(source, target)
2895
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2896
"""See InterRepository.fetch()."""
2897
from bzrlib.fetch import RepoFetcher
2898
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2899
self.source, self.source._format, self.target, self.target._format)
2900
f = RepoFetcher(to_repository=self.target,
2901
from_repository=self.source,
2902
last_revision=revision_id,
2903
pb=pb, find_ghosts=find_ghosts)
2906
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2907
"""See InterRepository.missing_revision_ids()."""
2908
if revision_id is not None:
2909
source_ids = self.source.get_ancestry(revision_id)
2910
if source_ids[0] is not None:
2911
raise AssertionError()
2914
source_ids = self.source.all_revision_ids()
2915
source_ids_set = set(source_ids)
2916
# source_ids is the worst possible case we may need to pull.
2917
# now we want to filter source_ids against what we actually
2918
# have in target, but don't try to check for existence where we know
2919
# we do not have a revision as that would be pointless.
2920
target_ids = set(self.target.all_revision_ids())
2921
possibly_present_revisions = target_ids.intersection(source_ids_set)
2922
actually_present_revisions = set(
2923
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2924
required_revisions = source_ids_set.difference(actually_present_revisions)
2925
if revision_id is not None:
2926
# we used get_ancestry to determine source_ids then we are assured all
2927
# revisions referenced are present as they are installed in topological order.
2928
# and the tip revision was validated by get_ancestry.
2929
result_set = required_revisions
2931
# if we just grabbed the possibly available ids, then
2932
# we only have an estimate of whats available and need to validate
2933
# that against the revision records.
2935
self.source._eliminate_revisions_not_present(required_revisions))
2936
return self.source.revision_ids_to_search_result(result_set)
2939
class InterPackRepo(InterSameDataRepository):
2940
"""Optimised code paths between Pack based repositories."""
2943
def _get_repo_format_to_test(self):
2944
from bzrlib.repofmt import pack_repo
2945
return pack_repo.RepositoryFormatKnitPack1()
2948
def is_compatible(source, target):
2949
"""Be compatible with known Pack formats.
2951
We don't test for the stores being of specific types because that
2952
could lead to confusing results, and there is no need to be
2955
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2957
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2958
isinstance(target._format, RepositoryFormatPack))
2959
except AttributeError:
2961
return are_packs and InterRepository._same_model(source, target)
2964
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2965
"""See InterRepository.fetch()."""
2966
if (len(self.source._fallback_repositories) > 0 or
2967
len(self.target._fallback_repositories) > 0):
2968
# The pack layer is not aware of fallback repositories, so when
2969
# fetching from a stacked repository or into a stacked repository
2970
# we use the generic fetch logic which uses the VersionedFiles
2971
# attributes on repository.
2972
from bzrlib.fetch import RepoFetcher
2973
fetcher = RepoFetcher(self.target, self.source, revision_id,
2975
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2976
self.source, self.source._format, self.target, self.target._format)
2977
if revision_id is None:
2979
# everything to do - use pack logic
2980
# to fetch from all packs to one without
2981
# inventory parsing etc, IFF nothing to be copied is in the target.
2983
source_revision_ids = frozenset(self.source.all_revision_ids())
2984
revision_ids = source_revision_ids - \
2985
frozenset(self.target_get_parent_map(source_revision_ids))
2986
revision_keys = [(revid,) for revid in revision_ids]
2987
target_pack_collection = self._get_target_pack_collection()
2988
index = target_pack_collection.revision_index.combined_index
2989
present_revision_ids = set(item[1][0] for item in
2990
index.iter_entries(revision_keys))
2991
revision_ids = set(revision_ids) - present_revision_ids
2992
# implementing the TODO will involve:
2993
# - detecting when all of a pack is selected
2994
# - avoiding as much as possible pre-selection, so the
2995
# more-core routines such as create_pack_from_packs can filter in
2996
# a just-in-time fashion. (though having a HEADS list on a
2997
# repository might make this a lot easier, because we could
2998
# sensibly detect 'new revisions' without doing a full index scan.
2999
elif _mod_revision.is_null(revision_id):
3004
revision_ids = self.search_missing_revision_ids(revision_id,
3005
find_ghosts=find_ghosts).get_keys()
3006
except errors.NoSuchRevision:
3007
raise errors.InstallFailed([revision_id])
3008
if len(revision_ids) == 0:
3010
return self._pack(self.source, self.target, revision_ids)
3012
def _pack(self, source, target, revision_ids):
3013
from bzrlib.repofmt.pack_repo import Packer
3014
target_pack_collection = self._get_target_pack_collection()
3015
packs = source._pack_collection.all_packs()
3016
pack = Packer(target_pack_collection, packs, '.fetch',
3017
revision_ids).pack()
3018
if pack is not None:
3019
target_pack_collection._save_pack_names()
3020
copied_revs = pack.get_revision_count()
3021
# Trigger an autopack. This may duplicate effort as we've just done
3022
# a pack creation, but for now it is simpler to think about as
3023
# 'upload data, then repack if needed'.
3025
return (copied_revs, [])
3029
def _autopack(self):
3030
self.target._pack_collection.autopack()
3032
def _get_target_pack_collection(self):
3033
return self.target._pack_collection
3036
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3037
"""See InterRepository.missing_revision_ids().
3039
:param find_ghosts: Find ghosts throughout the ancestry of
3042
if not find_ghosts and revision_id is not None:
3043
return self._walk_to_common_revisions([revision_id])
3044
elif revision_id is not None:
3045
# Find ghosts: search for revisions pointing from one repository to
3046
# the other, and vice versa, anywhere in the history of revision_id.
3047
graph = self.target_get_graph(other_repository=self.source)
3048
searcher = graph._make_breadth_first_searcher([revision_id])
3052
next_revs, ghosts = searcher.next_with_ghosts()
3053
except StopIteration:
3055
if revision_id in ghosts:
3056
raise errors.NoSuchRevision(self.source, revision_id)
3057
found_ids.update(next_revs)
3058
found_ids.update(ghosts)
3059
found_ids = frozenset(found_ids)
3060
# Double query here: should be able to avoid this by changing the
3061
# graph api further.
3062
result_set = found_ids - frozenset(
3063
self.target_get_parent_map(found_ids))
3065
source_ids = self.source.all_revision_ids()
3066
# source_ids is the worst possible case we may need to pull.
3067
# now we want to filter source_ids against what we actually
3068
# have in target, but don't try to check for existence where we know
3069
# we do not have a revision as that would be pointless.
3070
target_ids = set(self.target.all_revision_ids())
3071
result_set = set(source_ids).difference(target_ids)
3072
return self.source.revision_ids_to_search_result(result_set)
3075
class InterDifferingSerializer(InterKnitRepo):
3078
def _get_repo_format_to_test(self):
3082
def is_compatible(source, target):
3083
"""Be compatible with Knit2 source and Knit3 target"""
3084
if source.supports_rich_root() != target.supports_rich_root():
3086
# Ideally, we'd support fetching if the source had no tree references
3087
# even if it supported them...
3088
if (getattr(source, '_format.supports_tree_reference', False) and
3089
not getattr(target, '_format.supports_tree_reference', False)):
3093
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3094
"""Get the best delta and base for this revision.
3096
:return: (basis_id, delta)
3098
possible_trees = [(parent_id, cache[parent_id])
3099
for parent_id in parent_ids
3100
if parent_id in cache]
3101
if len(possible_trees) == 0:
3102
# There either aren't any parents, or the parents aren't in the
3103
# cache, so just use the last converted tree
3104
possible_trees.append((basis_id, cache[basis_id]))
3106
for basis_id, basis_tree in possible_trees:
3107
delta = tree.inventory._make_delta(basis_tree.inventory)
3108
deltas.append((len(delta), basis_id, delta))
3110
return deltas[0][1:]
3112
def _fetch_batch(self, revision_ids, basis_id, cache):
3113
"""Fetch across a few revisions.
3115
:param revision_ids: The revisions to copy
3116
:param basis_id: The revision_id of a tree that must be in cache, used
3117
as a basis for delta when no other base is available
3118
:param cache: A cache of RevisionTrees that we can use.
3119
:return: The revision_id of the last converted tree. The RevisionTree
3120
for it will be in cache
3122
# Walk though all revisions; get inventory deltas, copy referenced
3123
# texts that delta references, insert the delta, revision and
3127
pending_revisions = []
3128
parent_map = self.source.get_parent_map(revision_ids)
3129
for tree in self.source.revision_trees(revision_ids):
3130
current_revision_id = tree.get_revision_id()
3131
parent_ids = parent_map.get(current_revision_id, ())
3132
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3134
# Find text entries that need to be copied
3135
for old_path, new_path, file_id, entry in delta:
3136
if new_path is not None:
3137
if not (new_path or self.target.supports_rich_root()):
3138
# We don't copy the text for the root node unless the
3139
# target supports_rich_root.
3142
# "if entry.revision == current_revision_id" ?
3143
if entry.revision == current_revision_id:
3144
text_keys.add((file_id, entry.revision))
3145
revision = self.source.get_revision(current_revision_id)
3146
pending_deltas.append((basis_id, delta,
3147
current_revision_id, revision.parent_ids))
3148
pending_revisions.append(revision)
3149
cache[current_revision_id] = tree
3150
basis_id = current_revision_id
3152
from_texts = self.source.texts
3153
to_texts = self.target.texts
3154
to_texts.insert_record_stream(from_texts.get_record_stream(
3155
text_keys, self.target._format._fetch_order,
3156
not self.target._format._fetch_uses_deltas))
3158
for delta in pending_deltas:
3159
self.target.add_inventory_by_delta(*delta)
3160
# insert signatures and revisions
3161
for revision in pending_revisions:
3163
signature = self.source.get_signature_text(
3164
revision.revision_id)
3165
self.target.add_signature_text(revision.revision_id,
3167
except errors.NoSuchRevision:
3169
self.target.add_revision(revision.revision_id, revision)
3172
def _fetch_all_revisions(self, revision_ids, pb):
3173
"""Fetch everything for the list of revisions.
3175
:param revision_ids: The list of revisions to fetch. Must be in
3177
:param pb: A ProgressBar
3180
basis_id, basis_tree = self._get_basis(revision_ids[0])
3182
cache = lru_cache.LRUCache(100)
3183
cache[basis_id] = basis_tree
3184
del basis_tree # We don't want to hang on to it here
3185
for offset in range(0, len(revision_ids), batch_size):
3186
self.target.start_write_group()
3188
pb.update('Transferring revisions', offset,
3190
batch = revision_ids[offset:offset+batch_size]
3191
basis_id = self._fetch_batch(batch, basis_id, cache)
3193
self.target.abort_write_group()
3196
self.target.commit_write_group()
3197
pb.update('Transferring revisions', len(revision_ids),
3201
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3202
"""See InterRepository.fetch()."""
3203
revision_ids = self.target.search_missing_revision_ids(self.source,
3204
revision_id, find_ghosts=find_ghosts).get_keys()
3205
if not revision_ids:
3207
revision_ids = tsort.topo_sort(
3208
self.source.get_graph().get_parent_map(revision_ids))
3210
my_pb = ui.ui_factory.nested_progress_bar()
3215
self._fetch_all_revisions(revision_ids, pb)
3217
if my_pb is not None:
3219
return len(revision_ids), 0
3221
def _get_basis(self, first_revision_id):
3222
"""Get a revision and tree which exists in the target.
3224
This assumes that first_revision_id is selected for transmission
3225
because all other ancestors are already present. If we can't find an
3226
ancestor we fall back to NULL_REVISION since we know that is safe.
3228
:return: (basis_id, basis_tree)
3230
first_rev = self.source.get_revision(first_revision_id)
3232
basis_id = first_rev.parent_ids[0]
3233
# only valid as a basis if the target has it
3234
self.target.get_revision(basis_id)
3235
# Try to get a basis tree - if its a ghost it will hit the
3236
# NoSuchRevision case.
3237
basis_tree = self.source.revision_tree(basis_id)
3238
except (IndexError, errors.NoSuchRevision):
3239
basis_id = _mod_revision.NULL_REVISION
3240
basis_tree = self.source.revision_tree(basis_id)
3241
return basis_id, basis_tree
3244
class InterOtherToRemote(InterRepository):
3245
"""An InterRepository that simply delegates to the 'real' InterRepository
3246
calculated for (source, target._real_repository).
3249
_walk_to_common_revisions_batch_size = 50
3251
def __init__(self, source, target):
3252
InterRepository.__init__(self, source, target)
3253
self._real_inter = None
3256
def is_compatible(source, target):
3257
if isinstance(target, remote.RemoteRepository):
3261
def _ensure_real_inter(self):
3262
if self._real_inter is None:
3263
self.target._ensure_real()
3264
real_target = self.target._real_repository
3265
self._real_inter = InterRepository.get(self.source, real_target)
3266
# Make _real_inter use the RemoteRepository for get_parent_map
3267
self._real_inter.target_get_graph = self.target.get_graph
3268
self._real_inter.target_get_parent_map = self.target.get_parent_map
3270
def copy_content(self, revision_id=None):
3271
self._ensure_real_inter()
3272
self._real_inter.copy_content(revision_id=revision_id)
3274
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3275
self._ensure_real_inter()
3276
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3277
find_ghosts=find_ghosts)
3280
def _get_repo_format_to_test(self):
3284
class InterRemoteToOther(InterRepository):
3286
def __init__(self, source, target):
3287
InterRepository.__init__(self, source, target)
3288
self._real_inter = None
3291
def is_compatible(source, target):
3292
if not isinstance(source, remote.RemoteRepository):
3294
return InterRepository._same_model(source, target)
3296
def _ensure_real_inter(self):
3297
if self._real_inter is None:
3298
self.source._ensure_real()
3299
real_source = self.source._real_repository
3300
self._real_inter = InterRepository.get(real_source, self.target)
3303
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3304
"""See InterRepository.fetch()."""
3305
# Always fetch using the generic streaming fetch code, to allow
3306
# streaming fetching from remote servers.
3307
from bzrlib.fetch import RepoFetcher
3308
fetcher = RepoFetcher(self.target, self.source, revision_id,
3311
def copy_content(self, revision_id=None):
3312
self._ensure_real_inter()
3313
self._real_inter.copy_content(revision_id=revision_id)
3316
def _get_repo_format_to_test(self):
3321
class InterPackToRemotePack(InterPackRepo):
3322
"""A specialisation of InterPackRepo for a target that is a
3325
This will use the get_parent_map RPC rather than plain readvs, and also
3326
uses an RPC for autopacking.
3329
_walk_to_common_revisions_batch_size = 50
3332
def is_compatible(source, target):
3333
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3334
if isinstance(source._format, RepositoryFormatPack):
3335
if isinstance(target, remote.RemoteRepository):
3336
target._format._ensure_real()
3337
if isinstance(target._format._custom_format,
3338
RepositoryFormatPack):
3339
if InterRepository._same_model(source, target):
3343
def _autopack(self):
3344
self.target.autopack()
3347
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3348
"""See InterRepository.fetch()."""
3349
# Always fetch using the generic streaming fetch code, to allow
3350
# streaming fetching into remote servers.
3351
from bzrlib.fetch import RepoFetcher
3352
fetcher = RepoFetcher(self.target, self.source, revision_id,
3355
def _get_target_pack_collection(self):
3356
return self.target._real_repository._pack_collection
3359
def _get_repo_format_to_test(self):
3363
InterRepository.register_optimiser(InterDifferingSerializer)
3364
InterRepository.register_optimiser(InterSameDataRepository)
3365
InterRepository.register_optimiser(InterWeaveRepo)
3366
InterRepository.register_optimiser(InterKnitRepo)
3367
InterRepository.register_optimiser(InterPackRepo)
3368
InterRepository.register_optimiser(InterOtherToRemote)
3369
InterRepository.register_optimiser(InterRemoteToOther)
3370
InterRepository.register_optimiser(InterPackToRemotePack)
3373
class CopyConverter(object):
3374
"""A repository conversion tool which just performs a copy of the content.
3376
This is slow but quite reliable.
3379
def __init__(self, target_format):
3380
"""Create a CopyConverter.
3382
:param target_format: The format the resulting repository should be.
3384
self.target_format = target_format
3386
def convert(self, repo, pb):
3387
"""Perform the conversion of to_convert, giving feedback via pb.
3389
:param to_convert: The disk object to convert.
3390
:param pb: a progress bar to use for progress information.
3395
# this is only useful with metadir layouts - separated repo content.
3396
# trigger an assertion if not such
3397
repo._format.get_format_string()
3398
self.repo_dir = repo.bzrdir
3399
self.step('Moving repository to repository.backup')
3400
self.repo_dir.transport.move('repository', 'repository.backup')
3401
backup_transport = self.repo_dir.transport.clone('repository.backup')
3402
repo._format.check_conversion_target(self.target_format)
3403
self.source_repo = repo._format.open(self.repo_dir,
3405
_override_transport=backup_transport)
3406
self.step('Creating new repository')
3407
converted = self.target_format.initialize(self.repo_dir,
3408
self.source_repo.is_shared())
3409
converted.lock_write()
3411
self.step('Copying content into repository.')
3412
self.source_repo.copy_content_into(converted)
3415
self.step('Deleting old repository content.')
3416
self.repo_dir.transport.delete_tree('repository.backup')
3417
self.pb.note('repository converted')
3419
def step(self, message):
3420
"""Update the pb by a step."""
3422
self.pb.update(message, self.count, self.total)
3434
def _unescaper(match, _map=_unescape_map):
3435
code = match.group(1)
3439
if not code.startswith('#'):
3441
return unichr(int(code[1:])).encode('utf8')
3447
def _unescape_xml(data):
3448
"""Unescape predefined XML entities in a string of data."""
3450
if _unescape_re is None:
3451
_unescape_re = re.compile('\&([^;]*);')
3452
return _unescape_re.sub(_unescaper, data)
3455
class _VersionedFileChecker(object):
3457
def __init__(self, repository):
3458
self.repository = repository
3459
self.text_index = self.repository._generate_text_key_index()
3461
def calculate_file_version_parents(self, text_key):
3462
"""Calculate the correct parents for a file version according to
3465
parent_keys = self.text_index[text_key]
3466
if parent_keys == [_mod_revision.NULL_REVISION]:
3468
return tuple(parent_keys)
3470
def check_file_version_parents(self, texts, progress_bar=None):
3471
"""Check the parents stored in a versioned file are correct.
3473
It also detects file versions that are not referenced by their
3474
corresponding revision's inventory.
3476
:returns: A tuple of (wrong_parents, dangling_file_versions).
3477
wrong_parents is a dict mapping {revision_id: (stored_parents,
3478
correct_parents)} for each revision_id where the stored parents
3479
are not correct. dangling_file_versions is a set of (file_id,
3480
revision_id) tuples for versions that are present in this versioned
3481
file, but not used by the corresponding inventory.
3484
self.file_ids = set([file_id for file_id, _ in
3485
self.text_index.iterkeys()])
3486
# text keys is now grouped by file_id
3487
n_weaves = len(self.file_ids)
3488
files_in_revisions = {}
3489
revisions_of_files = {}
3490
n_versions = len(self.text_index)
3491
progress_bar.update('loading text store', 0, n_versions)
3492
parent_map = self.repository.texts.get_parent_map(self.text_index)
3493
# On unlistable transports this could well be empty/error...
3494
text_keys = self.repository.texts.keys()
3495
unused_keys = frozenset(text_keys) - set(self.text_index)
3496
for num, key in enumerate(self.text_index.iterkeys()):
3497
if progress_bar is not None:
3498
progress_bar.update('checking text graph', num, n_versions)
3499
correct_parents = self.calculate_file_version_parents(key)
3501
knit_parents = parent_map[key]
3502
except errors.RevisionNotPresent:
3505
if correct_parents != knit_parents:
3506
wrong_parents[key] = (knit_parents, correct_parents)
3507
return wrong_parents, unused_keys
3510
def _old_get_graph(repository, revision_id):
3511
"""DO NOT USE. That is all. I'm serious."""
3512
graph = repository.get_graph()
3513
revision_graph = dict(((key, value) for key, value in
3514
graph.iter_ancestry([revision_id]) if value is not None))
3515
return _strip_NULL_ghosts(revision_graph)
3518
def _strip_NULL_ghosts(revision_graph):
3519
"""Also don't use this. more compatibility code for unmigrated clients."""
3520
# Filter ghosts, and null:
3521
if _mod_revision.NULL_REVISION in revision_graph:
3522
del revision_graph[_mod_revision.NULL_REVISION]
3523
for key, parents in revision_graph.items():
3524
revision_graph[key] = tuple(parent for parent in parents if parent
3526
return revision_graph
3529
class StreamSink(object):
3530
"""An object that can insert a stream into a repository.
3532
This interface handles the complexity of reserialising inventories and
3533
revisions from different formats, and allows unidirectional insertion into
3534
stacked repositories without looking for the missing basis parents
3538
def __init__(self, target_repo):
3539
self.target_repo = target_repo
3541
def insert_stream(self, stream, src_format, resume_tokens):
3542
"""Insert a stream's content into the target repository.
3544
:param src_format: a bzr repository format.
3546
:return: a list of resume tokens and an iterable of keys additional
3547
items required before the insertion can be completed.
3549
self.target_repo.lock_write()
3552
self.target_repo.resume_write_group(resume_tokens)
3554
self.target_repo.start_write_group()
3556
# locked_insert_stream performs a commit|suspend.
3557
return self._locked_insert_stream(stream, src_format)
3559
self.target_repo.abort_write_group(suppress_errors=True)
3562
self.target_repo.unlock()
3564
def _locked_insert_stream(self, stream, src_format):
3565
to_serializer = self.target_repo._format._serializer
3566
src_serializer = src_format._serializer
3567
for substream_type, substream in stream:
3568
if substream_type == 'texts':
3569
self.target_repo.texts.insert_record_stream(substream)
3570
elif substream_type == 'inventories':
3571
if src_serializer == to_serializer:
3572
self.target_repo.inventories.insert_record_stream(
3575
self._extract_and_insert_inventories(
3576
substream, src_serializer)
3577
elif substream_type == 'revisions':
3578
# This may fallback to extract-and-insert more often than
3579
# required if the serializers are different only in terms of
3581
if src_serializer == to_serializer:
3582
self.target_repo.revisions.insert_record_stream(
3585
self._extract_and_insert_revisions(substream,
3587
elif substream_type == 'signatures':
3588
self.target_repo.signatures.insert_record_stream(substream)
3590
raise AssertionError('kaboom! %s' % (substream_type,))
3592
missing_keys = set()
3593
for prefix, versioned_file in (
3594
('texts', self.target_repo.texts),
3595
('inventories', self.target_repo.inventories),
3596
('revisions', self.target_repo.revisions),
3597
('signatures', self.target_repo.signatures),
3599
missing_keys.update((prefix,) + key for key in
3600
versioned_file.get_missing_compression_parent_keys())
3601
except NotImplementedError:
3602
# cannot even attempt suspending, and missing would have failed
3603
# during stream insertion.
3604
missing_keys = set()
3607
# suspend the write group and tell the caller what we is
3608
# missing. We know we can suspend or else we would not have
3609
# entered this code path. (All repositories that can handle
3610
# missing keys can handle suspending a write group).
3611
write_group_tokens = self.target_repo.suspend_write_group()
3612
return write_group_tokens, missing_keys
3613
self.target_repo.commit_write_group()
3616
def _extract_and_insert_inventories(self, substream, serializer):
3617
"""Generate a new inventory versionedfile in target, converting data.
3619
The inventory is retrieved from the source, (deserializing it), and
3620
stored in the target (reserializing it in a different format).
3622
for record in substream:
3623
bytes = record.get_bytes_as('fulltext')
3624
revision_id = record.key[0]
3625
inv = serializer.read_inventory_from_string(bytes, revision_id)
3626
parents = [key[0] for key in record.parents]
3627
self.target_repo.add_inventory(revision_id, inv, parents)
3629
def _extract_and_insert_revisions(self, substream, serializer):
3630
for record in substream:
3631
bytes = record.get_bytes_as('fulltext')
3632
revision_id = record.key[0]
3633
rev = serializer.read_revision_from_string(bytes)
3634
if rev.revision_id != revision_id:
3635
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
3636
self.target_repo.add_revision(revision_id, rev)
3639
if self.target_repo._format._fetch_reconcile:
3640
self.target_repo.reconcile()
3643
class StreamSource(object):
3644
"""A source of a stream for fetching between repositories."""
3646
def __init__(self, from_repository, to_format):
3647
"""Create a StreamSource streaming from from_repository."""
3648
self.from_repository = from_repository
3649
self.to_format = to_format
3651
def delta_on_metadata(self):
3652
"""Return True if delta's are permitted on metadata streams.
3654
That is on revisions and signatures.
3656
src_serializer = self.from_repository._format._serializer
3657
target_serializer = self.to_format._serializer
3658
return (self.to_format._fetch_uses_deltas and
3659
src_serializer == target_serializer)
3661
def _fetch_revision_texts(self, revs):
3662
# fetch signatures first and then the revision texts
3663
# may need to be a InterRevisionStore call here.
3664
from_sf = self.from_repository.signatures
3665
# A missing signature is just skipped.
3666
keys = [(rev_id,) for rev_id in revs]
3667
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
3669
self.to_format._fetch_order,
3670
not self.to_format._fetch_uses_deltas))
3671
# If a revision has a delta, this is actually expanded inside the
3672
# insert_record_stream code now, which is an alternate fix for
3674
from_rf = self.from_repository.revisions
3675
revisions = from_rf.get_record_stream(
3677
self.to_format._fetch_order,
3678
not self.delta_on_metadata())
3679
return [('signatures', signatures), ('revisions', revisions)]
3681
def _generate_root_texts(self, revs):
3682
"""This will be called by __fetch between fetching weave texts and
3683
fetching the inventory weave.
3685
Subclasses should override this if they need to generate root texts
3686
after fetching weave texts.
3688
if self._rich_root_upgrade():
3690
return bzrlib.fetch.Inter1and2Helper(
3691
self.from_repository).generate_root_texts(revs)
3695
def get_stream(self, search):
3697
revs = search.get_keys()
3698
graph = self.from_repository.get_graph()
3699
revs = list(graph.iter_topo_order(revs))
3700
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
3702
for knit_kind, file_id, revisions in data_to_fetch:
3703
if knit_kind != phase:
3705
# Make a new progress bar for this phase
3706
if knit_kind == "file":
3707
# Accumulate file texts
3708
text_keys.extend([(file_id, revision) for revision in
3710
elif knit_kind == "inventory":
3711
# Now copy the file texts.
3712
from_texts = self.from_repository.texts
3713
yield ('texts', from_texts.get_record_stream(
3714
text_keys, self.to_format._fetch_order,
3715
not self.to_format._fetch_uses_deltas))
3716
# Cause an error if a text occurs after we have done the
3719
# Before we process the inventory we generate the root
3720
# texts (if necessary) so that the inventories references
3722
for _ in self._generate_root_texts(revs):
3724
# NB: This currently reopens the inventory weave in source;
3725
# using a single stream interface instead would avoid this.
3726
from_weave = self.from_repository.inventories
3727
# we fetch only the referenced inventories because we do not
3728
# know for unselected inventories whether all their required
3729
# texts are present in the other repository - it could be
3731
yield ('inventories', from_weave.get_record_stream(
3732
[(rev_id,) for rev_id in revs],
3733
self.inventory_fetch_order(),
3734
not self.delta_on_metadata()))
3735
elif knit_kind == "signatures":
3736
# Nothing to do here; this will be taken care of when
3737
# _fetch_revision_texts happens.
3739
elif knit_kind == "revisions":
3740
for record in self._fetch_revision_texts(revs):
3743
raise AssertionError("Unknown knit kind %r" % knit_kind)
3745
def get_stream_for_missing_keys(self, missing_keys):
3746
# missing keys can only occur when we are byte copying and not
3747
# translating (because translation means we don't send
3748
# unreconstructable deltas ever).
3750
keys['texts'] = set()
3751
keys['revisions'] = set()
3752
keys['inventories'] = set()
3753
keys['signatures'] = set()
3754
for key in missing_keys:
3755
keys[key[0]].add(key[1:])
3756
if len(keys['revisions']):
3757
# If we allowed copying revisions at this point, we could end up
3758
# copying a revision without copying its required texts: a
3759
# violation of the requirements for repository integrity.
3760
raise AssertionError(
3761
'cannot copy revisions to fill in missing deltas %s' % (
3762
keys['revisions'],))
3763
for substream_kind, keys in keys.iteritems():
3764
vf = getattr(self.from_repository, substream_kind)
3765
# Ask for full texts always so that we don't need more round trips
3766
# after this stream.
3767
stream = vf.get_record_stream(keys,
3768
self.to_format._fetch_order, True)
3769
yield substream_kind, stream
3771
def inventory_fetch_order(self):
3772
if self._rich_root_upgrade():
3773
return 'topological'
3775
return self.to_format._fetch_order
3777
def _rich_root_upgrade(self):
3778
return (not self.from_repository._format.rich_root_data and
3779
self.to_format.rich_root_data)
added
removed
bzrlib/repository.py
