1
# Copyright (C) 2005, 2006, 2007, 2008, 2009 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from bzrlib.lazy_import import lazy_import
18
lazy_import(globals(), """
37
revision as _mod_revision,
43
from bzrlib.bundle import serializer
44
from bzrlib.revisiontree import RevisionTree
45
from bzrlib.store.versioned import VersionedFileStore
46
from bzrlib.testament import Testament
49
from bzrlib.decorators import needs_read_lock, needs_write_lock
50
from bzrlib.inter import InterObject
51
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
52
from bzrlib import registry
53
from bzrlib.symbol_versioning import (
59
from bzrlib.trace import (
60
log_exception_quietly, note, mutter, mutter_callsite, warning)
63
# Old formats display a warning, but only once
64
_deprecation_warning_done = False
67
class CommitBuilder(object):
68
"""Provides an interface to build up a commit.
70
This allows describing a tree to be committed without needing to
71
know the internals of the format of the repository.
74
# all clients should supply tree roots.
75
record_root_entry = True
76
# the default CommitBuilder does not manage trees whose root is versioned.
77
_versioned_root = False
79
def __init__(self, repository, parents, config, timestamp=None,
80
timezone=None, committer=None, revprops=None,
82
"""Initiate a CommitBuilder.
84
:param repository: Repository to commit to.
85
:param parents: Revision ids of the parents of the new revision.
86
:param config: Configuration to use.
87
:param timestamp: Optional timestamp recorded for commit.
88
:param timezone: Optional timezone for timestamp.
89
:param committer: Optional committer to set for commit.
90
:param revprops: Optional dictionary of revision properties.
91
:param revision_id: Optional revision id.
96
self._committer = self._config.username()
98
self._committer = committer
100
self.new_inventory = Inventory(None)
101
self._new_revision_id = revision_id
102
self.parents = parents
103
self.repository = repository
106
if revprops is not None:
107
self._validate_revprops(revprops)
108
self._revprops.update(revprops)
110
if timestamp is None:
111
timestamp = time.time()
112
# Restrict resolution to 1ms
113
self._timestamp = round(timestamp, 3)
116
self._timezone = osutils.local_time_offset()
118
self._timezone = int(timezone)
120
self._generate_revision_if_needed()
121
self.__heads = graph.HeadsCache(repository.get_graph()).heads
122
self._basis_delta = []
123
# API compatibility, older code that used CommitBuilder did not call
124
# .record_delete(), which means the delta that is computed would not be
125
# valid. Callers that will call record_delete() should call
126
# .will_record_deletes() to indicate that.
127
self._recording_deletes = False
129
def _validate_unicode_text(self, text, context):
130
"""Verify things like commit messages don't have bogus characters."""
132
raise ValueError('Invalid value for %s: %r' % (context, text))
134
def _validate_revprops(self, revprops):
135
for key, value in revprops.iteritems():
136
# We know that the XML serializers do not round trip '\r'
137
# correctly, so refuse to accept them
138
if not isinstance(value, basestring):
139
raise ValueError('revision property (%s) is not a valid'
140
' (unicode) string: %r' % (key, value))
141
self._validate_unicode_text(value,
142
'revision property (%s)' % (key,))
144
def commit(self, message):
145
"""Make the actual commit.
147
:return: The revision id of the recorded revision.
149
self._validate_unicode_text(message, 'commit message')
150
rev = _mod_revision.Revision(
151
timestamp=self._timestamp,
152
timezone=self._timezone,
153
committer=self._committer,
155
inventory_sha1=self.inv_sha1,
156
revision_id=self._new_revision_id,
157
properties=self._revprops)
158
rev.parent_ids = self.parents
159
self.repository.add_revision(self._new_revision_id, rev,
160
self.new_inventory, self._config)
161
self.repository.commit_write_group()
162
return self._new_revision_id
165
"""Abort the commit that is being built.
167
self.repository.abort_write_group()
169
def revision_tree(self):
170
"""Return the tree that was just committed.
172
After calling commit() this can be called to get a RevisionTree
173
representing the newly committed tree. This is preferred to
174
calling Repository.revision_tree() because that may require
175
deserializing the inventory, while we already have a copy in
178
return RevisionTree(self.repository, self.new_inventory,
179
self._new_revision_id)
181
def finish_inventory(self):
182
"""Tell the builder that the inventory is finished."""
183
if self.new_inventory.root is None:
184
raise AssertionError('Root entry should be supplied to'
185
' record_entry_contents, as of bzr 0.10.')
186
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
187
self.new_inventory.revision_id = self._new_revision_id
188
self.inv_sha1 = self.repository.add_inventory(
189
self._new_revision_id,
194
def _gen_revision_id(self):
195
"""Return new revision-id."""
196
return generate_ids.gen_revision_id(self._config.username(),
199
def _generate_revision_if_needed(self):
200
"""Create a revision id if None was supplied.
202
If the repository can not support user-specified revision ids
203
they should override this function and raise CannotSetRevisionId
204
if _new_revision_id is not None.
206
:raises: CannotSetRevisionId
208
if self._new_revision_id is None:
209
self._new_revision_id = self._gen_revision_id()
210
self.random_revid = True
212
self.random_revid = False
214
def _heads(self, file_id, revision_ids):
215
"""Calculate the graph heads for revision_ids in the graph of file_id.
217
This can use either a per-file graph or a global revision graph as we
218
have an identity relationship between the two graphs.
220
return self.__heads(revision_ids)
222
def _check_root(self, ie, parent_invs, tree):
223
"""Helper for record_entry_contents.
225
:param ie: An entry being added.
226
:param parent_invs: The inventories of the parent revisions of the
228
:param tree: The tree that is being committed.
230
# In this revision format, root entries have no knit or weave When
231
# serializing out to disk and back in root.revision is always
233
ie.revision = self._new_revision_id
235
def _get_delta(self, ie, basis_inv, path):
236
"""Get a delta against the basis inventory for ie."""
237
if ie.file_id not in basis_inv:
239
result = (None, path, ie.file_id, ie)
240
self._basis_delta.append(result)
242
elif ie != basis_inv[ie.file_id]:
244
# TODO: avoid tis id2path call.
245
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
246
self._basis_delta.append(result)
252
def get_basis_delta(self):
253
"""Return the complete inventory delta versus the basis inventory.
255
This has been built up with the calls to record_delete and
256
record_entry_contents. The client must have already called
257
will_record_deletes() to indicate that they will be generating a
260
:return: An inventory delta, suitable for use with apply_delta, or
261
Repository.add_inventory_by_delta, etc.
263
if not self._recording_deletes:
264
raise AssertionError("recording deletes not activated.")
265
return self._basis_delta
267
def record_delete(self, path, file_id):
268
"""Record that a delete occured against a basis tree.
270
This is an optional API - when used it adds items to the basis_delta
271
being accumulated by the commit builder. It cannot be called unless the
272
method will_record_deletes() has been called to inform the builder that
273
a delta is being supplied.
275
:param path: The path of the thing deleted.
276
:param file_id: The file id that was deleted.
278
if not self._recording_deletes:
279
raise AssertionError("recording deletes not activated.")
280
delta = (path, None, file_id, None)
281
self._basis_delta.append(delta)
284
def will_record_deletes(self):
285
"""Tell the commit builder that deletes are being notified.
287
This enables the accumulation of an inventory delta; for the resulting
288
commit to be valid, deletes against the basis MUST be recorded via
289
builder.record_delete().
291
self._recording_deletes = True
293
def record_entry_contents(self, ie, parent_invs, path, tree,
295
"""Record the content of ie from tree into the commit if needed.
297
Side effect: sets ie.revision when unchanged
299
:param ie: An inventory entry present in the commit.
300
:param parent_invs: The inventories of the parent revisions of the
302
:param path: The path the entry is at in the tree.
303
:param tree: The tree which contains this entry and should be used to
305
:param content_summary: Summary data from the tree about the paths
306
content - stat, length, exec, sha/link target. This is only
307
accessed when the entry has a revision of None - that is when it is
308
a candidate to commit.
309
:return: A tuple (change_delta, version_recorded, fs_hash).
310
change_delta is an inventory_delta change for this entry against
311
the basis tree of the commit, or None if no change occured against
313
version_recorded is True if a new version of the entry has been
314
recorded. For instance, committing a merge where a file was only
315
changed on the other side will return (delta, False).
316
fs_hash is either None, or the hash details for the path (currently
317
a tuple of the contents sha1 and the statvalue returned by
318
tree.get_file_with_stat()).
320
if self.new_inventory.root is None:
321
if ie.parent_id is not None:
322
raise errors.RootMissing()
323
self._check_root(ie, parent_invs, tree)
324
if ie.revision is None:
325
kind = content_summary[0]
327
# ie is carried over from a prior commit
329
# XXX: repository specific check for nested tree support goes here - if
330
# the repo doesn't want nested trees we skip it ?
331
if (kind == 'tree-reference' and
332
not self.repository._format.supports_tree_reference):
333
# mismatch between commit builder logic and repository:
334
# this needs the entry creation pushed down into the builder.
335
raise NotImplementedError('Missing repository subtree support.')
336
self.new_inventory.add(ie)
338
# TODO: slow, take it out of the inner loop.
340
basis_inv = parent_invs[0]
342
basis_inv = Inventory(root_id=None)
344
# ie.revision is always None if the InventoryEntry is considered
345
# for committing. We may record the previous parents revision if the
346
# content is actually unchanged against a sole head.
347
if ie.revision is not None:
348
if not self._versioned_root and path == '':
349
# repositories that do not version the root set the root's
350
# revision to the new commit even when no change occurs (more
351
# specifically, they do not record a revision on the root; and
352
# the rev id is assigned to the root during deserialisation -
353
# this masks when a change may have occurred against the basis.
354
# To match this we always issue a delta, because the revision
355
# of the root will always be changing.
356
if ie.file_id in basis_inv:
357
delta = (basis_inv.id2path(ie.file_id), path,
361
delta = (None, path, ie.file_id, ie)
362
self._basis_delta.append(delta)
363
return delta, False, None
365
# we don't need to commit this, because the caller already
366
# determined that an existing revision of this file is
367
# appropriate. If its not being considered for committing then
368
# it and all its parents to the root must be unaltered so
369
# no-change against the basis.
370
if ie.revision == self._new_revision_id:
371
raise AssertionError("Impossible situation, a skipped "
372
"inventory entry (%r) claims to be modified in this "
373
"commit (%r).", (ie, self._new_revision_id))
374
return None, False, None
375
# XXX: Friction: parent_candidates should return a list not a dict
376
# so that we don't have to walk the inventories again.
377
parent_candiate_entries = ie.parent_candidates(parent_invs)
378
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
380
for inv in parent_invs:
381
if ie.file_id in inv:
382
old_rev = inv[ie.file_id].revision
383
if old_rev in head_set:
384
heads.append(inv[ie.file_id].revision)
385
head_set.remove(inv[ie.file_id].revision)
388
# now we check to see if we need to write a new record to the
390
# We write a new entry unless there is one head to the ancestors, and
391
# the kind-derived content is unchanged.
393
# Cheapest check first: no ancestors, or more the one head in the
394
# ancestors, we write a new node.
398
# There is a single head, look it up for comparison
399
parent_entry = parent_candiate_entries[heads[0]]
400
# if the non-content specific data has changed, we'll be writing a
402
if (parent_entry.parent_id != ie.parent_id or
403
parent_entry.name != ie.name):
405
# now we need to do content specific checks:
407
# if the kind changed the content obviously has
408
if kind != parent_entry.kind:
410
# Stat cache fingerprint feedback for the caller - None as we usually
411
# don't generate one.
414
if content_summary[2] is None:
415
raise ValueError("Files must not have executable = None")
417
if (# if the file length changed we have to store:
418
parent_entry.text_size != content_summary[1] or
419
# if the exec bit has changed we have to store:
420
parent_entry.executable != content_summary[2]):
422
elif parent_entry.text_sha1 == content_summary[3]:
423
# all meta and content is unchanged (using a hash cache
424
# hit to check the sha)
425
ie.revision = parent_entry.revision
426
ie.text_size = parent_entry.text_size
427
ie.text_sha1 = parent_entry.text_sha1
428
ie.executable = parent_entry.executable
429
return self._get_delta(ie, basis_inv, path), False, None
431
# Either there is only a hash change(no hash cache entry,
432
# or same size content change), or there is no change on
434
# Provide the parent's hash to the store layer, so that the
435
# content is unchanged we will not store a new node.
436
nostore_sha = parent_entry.text_sha1
438
# We want to record a new node regardless of the presence or
439
# absence of a content change in the file.
441
ie.executable = content_summary[2]
442
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
444
lines = file_obj.readlines()
448
ie.text_sha1, ie.text_size = self._add_text_to_weave(
449
ie.file_id, lines, heads, nostore_sha)
450
# Let the caller know we generated a stat fingerprint.
451
fingerprint = (ie.text_sha1, stat_value)
452
except errors.ExistingContent:
453
# Turns out that the file content was unchanged, and we were
454
# only going to store a new node if it was changed. Carry over
456
ie.revision = parent_entry.revision
457
ie.text_size = parent_entry.text_size
458
ie.text_sha1 = parent_entry.text_sha1
459
ie.executable = parent_entry.executable
460
return self._get_delta(ie, basis_inv, path), False, None
461
elif kind == 'directory':
463
# all data is meta here, nothing specific to directory, so
465
ie.revision = parent_entry.revision
466
return self._get_delta(ie, basis_inv, path), False, None
468
self._add_text_to_weave(ie.file_id, lines, heads, None)
469
elif kind == 'symlink':
470
current_link_target = content_summary[3]
472
# symlink target is not generic metadata, check if it has
474
if current_link_target != parent_entry.symlink_target:
477
# unchanged, carry over.
478
ie.revision = parent_entry.revision
479
ie.symlink_target = parent_entry.symlink_target
480
return self._get_delta(ie, basis_inv, path), False, None
481
ie.symlink_target = current_link_target
483
self._add_text_to_weave(ie.file_id, lines, heads, None)
484
elif kind == 'tree-reference':
486
if content_summary[3] != parent_entry.reference_revision:
489
# unchanged, carry over.
490
ie.reference_revision = parent_entry.reference_revision
491
ie.revision = parent_entry.revision
492
return self._get_delta(ie, basis_inv, path), False, None
493
ie.reference_revision = content_summary[3]
495
self._add_text_to_weave(ie.file_id, lines, heads, None)
497
raise NotImplementedError('unknown kind')
498
ie.revision = self._new_revision_id
499
return self._get_delta(ie, basis_inv, path), True, fingerprint
501
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
502
# Note: as we read the content directly from the tree, we know its not
503
# been turned into unicode or badly split - but a broken tree
504
# implementation could give us bad output from readlines() so this is
505
# not a guarantee of safety. What would be better is always checking
506
# the content during test suite execution. RBC 20070912
507
parent_keys = tuple((file_id, parent) for parent in parents)
508
return self.repository.texts.add_lines(
509
(file_id, self._new_revision_id), parent_keys, new_lines,
510
nostore_sha=nostore_sha, random_id=self.random_revid,
511
check_content=False)[0:2]
514
class RootCommitBuilder(CommitBuilder):
515
"""This commitbuilder actually records the root id"""
517
# the root entry gets versioned properly by this builder.
518
_versioned_root = True
520
def _check_root(self, ie, parent_invs, tree):
521
"""Helper for record_entry_contents.
523
:param ie: An entry being added.
524
:param parent_invs: The inventories of the parent revisions of the
526
:param tree: The tree that is being committed.
530
######################################################################
533
class Repository(object):
534
"""Repository holding history for one or more branches.
536
The repository holds and retrieves historical information including
537
revisions and file history. It's normally accessed only by the Branch,
538
which views a particular line of development through that history.
540
The Repository builds on top of some byte storage facilies (the revisions,
541
signatures, inventories and texts attributes) and a Transport, which
542
respectively provide byte storage and a means to access the (possibly
545
The byte storage facilities are addressed via tuples, which we refer to
546
as 'keys' throughout the code base. Revision_keys, inventory_keys and
547
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
548
(file_id, revision_id). We use this interface because it allows low
549
friction with the underlying code that implements disk indices, network
550
encoding and other parts of bzrlib.
552
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
553
the serialised revisions for the repository. This can be used to obtain
554
revision graph information or to access raw serialised revisions.
555
The result of trying to insert data into the repository via this store
556
is undefined: it should be considered read-only except for implementors
558
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
559
the serialised signatures for the repository. This can be used to
560
obtain access to raw serialised signatures. The result of trying to
561
insert data into the repository via this store is undefined: it should
562
be considered read-only except for implementors of repositories.
563
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
564
the serialised inventories for the repository. This can be used to
565
obtain unserialised inventories. The result of trying to insert data
566
into the repository via this store is undefined: it should be
567
considered read-only except for implementors of repositories.
568
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
569
texts of files and directories for the repository. This can be used to
570
obtain file texts or file graphs. Note that Repository.iter_file_bytes
571
is usually a better interface for accessing file texts.
572
The result of trying to insert data into the repository via this store
573
is undefined: it should be considered read-only except for implementors
575
:ivar _transport: Transport for file access to repository, typically
576
pointing to .bzr/repository.
579
# What class to use for a CommitBuilder. Often its simpler to change this
580
# in a Repository class subclass rather than to override
581
# get_commit_builder.
582
_commit_builder_class = CommitBuilder
583
# The search regex used by xml based repositories to determine what things
584
# where changed in a single commit.
585
_file_ids_altered_regex = lazy_regex.lazy_compile(
586
r'file_id="(?P<file_id>[^"]+)"'
587
r'.* revision="(?P<revision_id>[^"]+)"'
590
def abort_write_group(self, suppress_errors=False):
591
"""Commit the contents accrued within the current write group.
593
:param suppress_errors: if true, abort_write_group will catch and log
594
unexpected errors that happen during the abort, rather than
595
allowing them to propagate. Defaults to False.
597
:seealso: start_write_group.
599
if self._write_group is not self.get_transaction():
600
# has an unlock or relock occured ?
601
raise errors.BzrError('mismatched lock context and write group.')
603
self._abort_write_group()
604
except Exception, exc:
605
self._write_group = None
606
if not suppress_errors:
608
mutter('abort_write_group failed')
609
log_exception_quietly()
610
note('bzr: ERROR (ignored): %s', exc)
611
self._write_group = None
613
def _abort_write_group(self):
614
"""Template method for per-repository write group cleanup.
616
This is called during abort before the write group is considered to be
617
finished and should cleanup any internal state accrued during the write
618
group. There is no requirement that data handed to the repository be
619
*not* made available - this is not a rollback - but neither should any
620
attempt be made to ensure that data added is fully commited. Abort is
621
invoked when an error has occured so futher disk or network operations
622
may not be possible or may error and if possible should not be
626
def add_fallback_repository(self, repository):
627
"""Add a repository to use for looking up data not held locally.
629
:param repository: A repository.
631
if not self._format.supports_external_lookups:
632
raise errors.UnstackableRepositoryFormat(self._format, self.base)
633
self._check_fallback_repository(repository)
634
self._fallback_repositories.append(repository)
635
self.texts.add_fallback_versioned_files(repository.texts)
636
self.inventories.add_fallback_versioned_files(repository.inventories)
637
self.revisions.add_fallback_versioned_files(repository.revisions)
638
self.signatures.add_fallback_versioned_files(repository.signatures)
640
def _check_fallback_repository(self, repository):
641
"""Check that this repository can fallback to repository safely.
643
Raise an error if not.
645
:param repository: A repository to fallback to.
647
return InterRepository._assert_same_model(self, repository)
649
def add_inventory(self, revision_id, inv, parents):
650
"""Add the inventory inv to the repository as revision_id.
652
:param parents: The revision ids of the parents that revision_id
653
is known to have and are in the repository already.
655
:returns: The validator(which is a sha1 digest, though what is sha'd is
656
repository format specific) of the serialized inventory.
658
if not self.is_in_write_group():
659
raise AssertionError("%r not in write group" % (self,))
660
_mod_revision.check_not_reserved_id(revision_id)
661
if not (inv.revision_id is None or inv.revision_id == revision_id):
662
raise AssertionError(
663
"Mismatch between inventory revision"
664
" id and insertion revid (%r, %r)"
665
% (inv.revision_id, revision_id))
667
raise AssertionError()
668
inv_lines = self._serialise_inventory_to_lines(inv)
669
return self._inventory_add_lines(revision_id, parents,
670
inv_lines, check_content=False)
672
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
674
"""Add a new inventory expressed as a delta against another revision.
676
:param basis_revision_id: The inventory id the delta was created
677
against. (This does not have to be a direct parent.)
678
:param delta: The inventory delta (see Inventory.apply_delta for
680
:param new_revision_id: The revision id that the inventory is being
682
:param parents: The revision ids of the parents that revision_id is
683
known to have and are in the repository already. These are supplied
684
for repositories that depend on the inventory graph for revision
685
graph access, as well as for those that pun ancestry with delta
688
:returns: (validator, new_inv)
689
The validator(which is a sha1 digest, though what is sha'd is
690
repository format specific) of the serialized inventory, and the
693
if not self.is_in_write_group():
694
raise AssertionError("%r not in write group" % (self,))
695
_mod_revision.check_not_reserved_id(new_revision_id)
696
basis_tree = self.revision_tree(basis_revision_id)
697
basis_tree.lock_read()
699
# Note that this mutates the inventory of basis_tree, which not all
700
# inventory implementations may support: A better idiom would be to
701
# return a new inventory, but as there is no revision tree cache in
702
# repository this is safe for now - RBC 20081013
703
basis_inv = basis_tree.inventory
704
basis_inv.apply_delta(delta)
705
basis_inv.revision_id = new_revision_id
706
return (self.add_inventory(new_revision_id, basis_inv, parents),
711
def _inventory_add_lines(self, revision_id, parents, lines,
713
"""Store lines in inv_vf and return the sha1 of the inventory."""
714
parents = [(parent,) for parent in parents]
715
return self.inventories.add_lines((revision_id,), parents, lines,
716
check_content=check_content)[0]
718
def add_revision(self, revision_id, rev, inv=None, config=None):
719
"""Add rev to the revision store as revision_id.
721
:param revision_id: the revision id to use.
722
:param rev: The revision object.
723
:param inv: The inventory for the revision. if None, it will be looked
724
up in the inventory storer
725
:param config: If None no digital signature will be created.
726
If supplied its signature_needed method will be used
727
to determine if a signature should be made.
729
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
731
_mod_revision.check_not_reserved_id(revision_id)
732
if config is not None and config.signature_needed():
734
inv = self.get_inventory(revision_id)
735
plaintext = Testament(rev, inv).as_short_text()
736
self.store_revision_signature(
737
gpg.GPGStrategy(config), plaintext, revision_id)
738
# check inventory present
739
if not self.inventories.get_parent_map([(revision_id,)]):
741
raise errors.WeaveRevisionNotPresent(revision_id,
744
# yes, this is not suitable for adding with ghosts.
745
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
749
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
750
self._add_revision(rev)
752
def _add_revision(self, revision):
753
text = self._serializer.write_revision_to_string(revision)
754
key = (revision.revision_id,)
755
parents = tuple((parent,) for parent in revision.parent_ids)
756
self.revisions.add_lines(key, parents, osutils.split_lines(text))
758
def all_revision_ids(self):
759
"""Returns a list of all the revision ids in the repository.
761
This is conceptually deprecated because code should generally work on
762
the graph reachable from a particular revision, and ignore any other
763
revisions that might be present. There is no direct replacement
766
if 'evil' in debug.debug_flags:
767
mutter_callsite(2, "all_revision_ids is linear with history.")
768
return self._all_revision_ids()
770
def _all_revision_ids(self):
771
"""Returns a list of all the revision ids in the repository.
773
These are in as much topological order as the underlying store can
776
raise NotImplementedError(self._all_revision_ids)
778
def break_lock(self):
779
"""Break a lock if one is present from another instance.
781
Uses the ui factory to ask for confirmation if the lock may be from
784
self.control_files.break_lock()
787
def _eliminate_revisions_not_present(self, revision_ids):
788
"""Check every revision id in revision_ids to see if we have it.
790
Returns a set of the present revisions.
793
graph = self.get_graph()
794
parent_map = graph.get_parent_map(revision_ids)
795
# The old API returned a list, should this actually be a set?
796
return parent_map.keys()
799
def create(a_bzrdir):
800
"""Construct the current default format repository in a_bzrdir."""
801
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
803
def __init__(self, _format, a_bzrdir, control_files):
804
"""instantiate a Repository.
806
:param _format: The format of the repository on disk.
807
:param a_bzrdir: The BzrDir of the repository.
809
In the future we will have a single api for all stores for
810
getting file texts, inventories and revisions, then
811
this construct will accept instances of those things.
813
super(Repository, self).__init__()
814
self._format = _format
815
# the following are part of the public API for Repository:
816
self.bzrdir = a_bzrdir
817
self.control_files = control_files
818
self._transport = control_files._transport
819
self.base = self._transport.base
821
self._reconcile_does_inventory_gc = True
822
self._reconcile_fixes_text_parents = False
823
self._reconcile_backsup_inventory = True
824
# not right yet - should be more semantically clear ?
826
# TODO: make sure to construct the right store classes, etc, depending
827
# on whether escaping is required.
828
self._warn_if_deprecated()
829
self._write_group = None
830
# Additional places to query for data.
831
self._fallback_repositories = []
832
# An InventoryEntry cache, used during deserialization
833
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
836
return '%s(%r)' % (self.__class__.__name__,
839
def has_same_location(self, other):
840
"""Returns a boolean indicating if this repository is at the same
841
location as another repository.
843
This might return False even when two repository objects are accessing
844
the same physical repository via different URLs.
846
if self.__class__ is not other.__class__:
848
return (self._transport.base == other._transport.base)
850
def is_in_write_group(self):
851
"""Return True if there is an open write group.
853
:seealso: start_write_group.
855
return self._write_group is not None
858
return self.control_files.is_locked()
860
def is_write_locked(self):
861
"""Return True if this object is write locked."""
862
return self.is_locked() and self.control_files._lock_mode == 'w'
864
def lock_write(self, token=None):
865
"""Lock this repository for writing.
867
This causes caching within the repository obejct to start accumlating
868
data during reads, and allows a 'write_group' to be obtained. Write
869
groups must be used for actual data insertion.
871
:param token: if this is already locked, then lock_write will fail
872
unless the token matches the existing lock.
873
:returns: a token if this instance supports tokens, otherwise None.
874
:raises TokenLockingNotSupported: when a token is given but this
875
instance doesn't support using token locks.
876
:raises MismatchedToken: if the specified token doesn't match the token
877
of the existing lock.
878
:seealso: start_write_group.
880
A token should be passed in if you know that you have locked the object
881
some other way, and need to synchronise this object's state with that
884
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
886
locked = self.is_locked()
887
result = self.control_files.lock_write(token=token)
888
for repo in self._fallback_repositories:
889
# Writes don't affect fallback repos
896
locked = self.is_locked()
897
self.control_files.lock_read()
898
for repo in self._fallback_repositories:
903
def get_physical_lock_status(self):
904
return self.control_files.get_physical_lock_status()
906
def leave_lock_in_place(self):
907
"""Tell this repository not to release the physical lock when this
910
If lock_write doesn't return a token, then this method is not supported.
912
self.control_files.leave_in_place()
914
def dont_leave_lock_in_place(self):
915
"""Tell this repository to release the physical lock when this
916
object is unlocked, even if it didn't originally acquire it.
918
If lock_write doesn't return a token, then this method is not supported.
920
self.control_files.dont_leave_in_place()
923
def gather_stats(self, revid=None, committers=None):
924
"""Gather statistics from a revision id.
926
:param revid: The revision id to gather statistics from, if None, then
927
no revision specific statistics are gathered.
928
:param committers: Optional parameter controlling whether to grab
929
a count of committers from the revision specific statistics.
930
:return: A dictionary of statistics. Currently this contains:
931
committers: The number of committers if requested.
932
firstrev: A tuple with timestamp, timezone for the penultimate left
933
most ancestor of revid, if revid is not the NULL_REVISION.
934
latestrev: A tuple with timestamp, timezone for revid, if revid is
935
not the NULL_REVISION.
936
revisions: The total revision count in the repository.
937
size: An estimate disk size of the repository in bytes.
940
if revid and committers:
941
result['committers'] = 0
942
if revid and revid != _mod_revision.NULL_REVISION:
944
all_committers = set()
945
revisions = self.get_ancestry(revid)
946
# pop the leading None
948
first_revision = None
950
# ignore the revisions in the middle - just grab first and last
951
revisions = revisions[0], revisions[-1]
952
for revision in self.get_revisions(revisions):
953
if not first_revision:
954
first_revision = revision
956
all_committers.add(revision.committer)
957
last_revision = revision
959
result['committers'] = len(all_committers)
960
result['firstrev'] = (first_revision.timestamp,
961
first_revision.timezone)
962
result['latestrev'] = (last_revision.timestamp,
963
last_revision.timezone)
965
# now gather global repository information
966
# XXX: This is available for many repos regardless of listability.
967
if self.bzrdir.root_transport.listable():
968
# XXX: do we want to __define len__() ?
969
# Maybe the versionedfiles object should provide a different
970
# method to get the number of keys.
971
result['revisions'] = len(self.revisions.keys())
975
def find_branches(self, using=False):
976
"""Find branches underneath this repository.
978
This will include branches inside other branches.
980
:param using: If True, list only branches using this repository.
982
if using and not self.is_shared():
984
return [self.bzrdir.open_branch()]
985
except errors.NotBranchError:
987
class Evaluator(object):
990
self.first_call = True
992
def __call__(self, bzrdir):
993
# On the first call, the parameter is always the bzrdir
994
# containing the current repo.
995
if not self.first_call:
997
repository = bzrdir.open_repository()
998
except errors.NoRepositoryPresent:
1001
return False, (None, repository)
1002
self.first_call = False
1004
value = (bzrdir.open_branch(), None)
1005
except errors.NotBranchError:
1006
value = (None, None)
1010
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1011
self.bzrdir.root_transport, evaluate=Evaluator()):
1012
if branch is not None:
1013
branches.append(branch)
1014
if not using and repository is not None:
1015
branches.extend(repository.find_branches())
1019
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1020
"""Return the revision ids that other has that this does not.
1022
These are returned in topological order.
1024
revision_id: only return revision ids included by revision_id.
1026
return InterRepository.get(other, self).search_missing_revision_ids(
1027
revision_id, find_ghosts)
1029
@deprecated_method(one_two)
1031
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1032
"""Return the revision ids that other has that this does not.
1034
These are returned in topological order.
1036
revision_id: only return revision ids included by revision_id.
1038
keys = self.search_missing_revision_ids(
1039
other, revision_id, find_ghosts).get_keys()
1042
parents = other.get_graph().get_parent_map(keys)
1045
return tsort.topo_sort(parents)
1049
"""Open the repository rooted at base.
1051
For instance, if the repository is at URL/.bzr/repository,
1052
Repository.open(URL) -> a Repository instance.
1054
control = bzrdir.BzrDir.open(base)
1055
return control.open_repository()
1057
def copy_content_into(self, destination, revision_id=None):
1058
"""Make a complete copy of the content in self into destination.
1060
This is a destructive operation! Do not use it on existing
1063
return InterRepository.get(self, destination).copy_content(revision_id)
1065
def commit_write_group(self):
1066
"""Commit the contents accrued within the current write group.
1068
:seealso: start_write_group.
1070
if self._write_group is not self.get_transaction():
1071
# has an unlock or relock occured ?
1072
raise errors.BzrError('mismatched lock context %r and '
1074
(self.get_transaction(), self._write_group))
1075
self._commit_write_group()
1076
self._write_group = None
1078
def _commit_write_group(self):
1079
"""Template method for per-repository write group cleanup.
1081
This is called before the write group is considered to be
1082
finished and should ensure that all data handed to the repository
1083
for writing during the write group is safely committed (to the
1084
extent possible considering file system caching etc).
1087
def suspend_write_group(self):
1088
raise errors.UnsuspendableWriteGroup(self)
1090
def refresh_data(self):
1091
"""Re-read any data needed to to synchronise with disk.
1093
This method is intended to be called after another repository instance
1094
(such as one used by a smart server) has inserted data into the
1095
repository. It may not be called during a write group, but may be
1096
called at any other time.
1098
if self.is_in_write_group():
1099
raise errors.InternalBzrError(
1100
"May not refresh_data while in a write group.")
1101
self._refresh_data()
1103
def resume_write_group(self, tokens):
1104
if not self.is_write_locked():
1105
raise errors.NotWriteLocked(self)
1106
if self._write_group:
1107
raise errors.BzrError('already in a write group')
1108
self._resume_write_group(tokens)
1109
# so we can detect unlock/relock - the write group is now entered.
1110
self._write_group = self.get_transaction()
1112
def _resume_write_group(self, tokens):
1113
raise errors.UnsuspendableWriteGroup(self)
1115
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
1117
"""Fetch the content required to construct revision_id from source.
1119
If revision_id is None and fetch_spec is None, then all content is
1122
fetch() may not be used when the repository is in a write group -
1123
either finish the current write group before using fetch, or use
1124
fetch before starting the write group.
1126
:param find_ghosts: Find and copy revisions in the source that are
1127
ghosts in the target (and not reachable directly by walking out to
1128
the first-present revision in target from revision_id).
1129
:param revision_id: If specified, all the content needed for this
1130
revision ID will be copied to the target. Fetch will determine for
1131
itself which content needs to be copied.
1132
:param fetch_spec: If specified, a SearchResult or
1133
PendingAncestryResult that describes which revisions to copy. This
1134
allows copying multiple heads at once. Mutually exclusive with
1137
if fetch_spec is not None and revision_id is not None:
1138
raise AssertionError(
1139
"fetch_spec and revision_id are mutually exclusive.")
1140
if self.is_in_write_group():
1141
raise errors.InternalBzrError(
1142
"May not fetch while in a write group.")
1143
# fast path same-url fetch operations
1144
if self.has_same_location(source) and fetch_spec is None:
1145
# check that last_revision is in 'from' and then return a
1147
if (revision_id is not None and
1148
not _mod_revision.is_null(revision_id)):
1149
self.get_revision(revision_id)
1151
# if there is no specific appropriate InterRepository, this will get
1152
# the InterRepository base class, which raises an
1153
# IncompatibleRepositories when asked to fetch.
1154
inter = InterRepository.get(source, self)
1155
return inter.fetch(revision_id=revision_id, pb=pb,
1156
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
1158
def create_bundle(self, target, base, fileobj, format=None):
1159
return serializer.write_bundle(self, target, base, fileobj, format)
1161
def get_commit_builder(self, branch, parents, config, timestamp=None,
1162
timezone=None, committer=None, revprops=None,
1164
"""Obtain a CommitBuilder for this repository.
1166
:param branch: Branch to commit to.
1167
:param parents: Revision ids of the parents of the new revision.
1168
:param config: Configuration to use.
1169
:param timestamp: Optional timestamp recorded for commit.
1170
:param timezone: Optional timezone for timestamp.
1171
:param committer: Optional committer to set for commit.
1172
:param revprops: Optional dictionary of revision properties.
1173
:param revision_id: Optional revision id.
1175
result = self._commit_builder_class(self, parents, config,
1176
timestamp, timezone, committer, revprops, revision_id)
1177
self.start_write_group()
1181
if (self.control_files._lock_count == 1 and
1182
self.control_files._lock_mode == 'w'):
1183
if self._write_group is not None:
1184
self.abort_write_group()
1185
self.control_files.unlock()
1186
raise errors.BzrError(
1187
'Must end write groups before releasing write locks.')
1188
self.control_files.unlock()
1189
if self.control_files._lock_count == 0:
1190
self._inventory_entry_cache.clear()
1191
for repo in self._fallback_repositories:
1195
def clone(self, a_bzrdir, revision_id=None):
1196
"""Clone this repository into a_bzrdir using the current format.
1198
Currently no check is made that the format of this repository and
1199
the bzrdir format are compatible. FIXME RBC 20060201.
1201
:return: The newly created destination repository.
1203
# TODO: deprecate after 0.16; cloning this with all its settings is
1204
# probably not very useful -- mbp 20070423
1205
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1206
self.copy_content_into(dest_repo, revision_id)
1209
def start_write_group(self):
1210
"""Start a write group in the repository.
1212
Write groups are used by repositories which do not have a 1:1 mapping
1213
between file ids and backend store to manage the insertion of data from
1214
both fetch and commit operations.
1216
A write lock is required around the start_write_group/commit_write_group
1217
for the support of lock-requiring repository formats.
1219
One can only insert data into a repository inside a write group.
1223
if not self.is_write_locked():
1224
raise errors.NotWriteLocked(self)
1225
if self._write_group:
1226
raise errors.BzrError('already in a write group')
1227
self._start_write_group()
1228
# so we can detect unlock/relock - the write group is now entered.
1229
self._write_group = self.get_transaction()
1231
def _start_write_group(self):
1232
"""Template method for per-repository write group startup.
1234
This is called before the write group is considered to be
1239
def sprout(self, to_bzrdir, revision_id=None):
1240
"""Create a descendent repository for new development.
1242
Unlike clone, this does not copy the settings of the repository.
1244
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1245
dest_repo.fetch(self, revision_id=revision_id)
1248
def _create_sprouting_repo(self, a_bzrdir, shared):
1249
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1250
# use target default format.
1251
dest_repo = a_bzrdir.create_repository()
1253
# Most control formats need the repository to be specifically
1254
# created, but on some old all-in-one formats it's not needed
1256
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1257
except errors.UninitializableFormat:
1258
dest_repo = a_bzrdir.open_repository()
1261
def _get_sink(self):
1262
"""Return a sink for streaming into this repository."""
1263
return StreamSink(self)
1265
def _get_source(self, to_format):
1266
"""Return a source for streaming from this repository."""
1267
return StreamSource(self, to_format)
1270
def has_revision(self, revision_id):
1271
"""True if this repository has a copy of the revision."""
1272
return revision_id in self.has_revisions((revision_id,))
1275
def has_revisions(self, revision_ids):
1276
"""Probe to find out the presence of multiple revisions.
1278
:param revision_ids: An iterable of revision_ids.
1279
:return: A set of the revision_ids that were present.
1281
parent_map = self.revisions.get_parent_map(
1282
[(rev_id,) for rev_id in revision_ids])
1284
if _mod_revision.NULL_REVISION in revision_ids:
1285
result.add(_mod_revision.NULL_REVISION)
1286
result.update([key[0] for key in parent_map])
1290
def get_revision(self, revision_id):
1291
"""Return the Revision object for a named revision."""
1292
return self.get_revisions([revision_id])[0]
1295
def get_revision_reconcile(self, revision_id):
1296
"""'reconcile' helper routine that allows access to a revision always.
1298
This variant of get_revision does not cross check the weave graph
1299
against the revision one as get_revision does: but it should only
1300
be used by reconcile, or reconcile-alike commands that are correcting
1301
or testing the revision graph.
1303
return self._get_revisions([revision_id])[0]
1306
def get_revisions(self, revision_ids):
1307
"""Get many revisions at once."""
1308
return self._get_revisions(revision_ids)
1311
def _get_revisions(self, revision_ids):
1312
"""Core work logic to get many revisions without sanity checks."""
1313
for rev_id in revision_ids:
1314
if not rev_id or not isinstance(rev_id, basestring):
1315
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1316
keys = [(key,) for key in revision_ids]
1317
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1319
for record in stream:
1320
if record.storage_kind == 'absent':
1321
raise errors.NoSuchRevision(self, record.key[0])
1322
text = record.get_bytes_as('fulltext')
1323
rev = self._serializer.read_revision_from_string(text)
1324
revs[record.key[0]] = rev
1325
return [revs[revid] for revid in revision_ids]
1328
def get_revision_xml(self, revision_id):
1329
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1330
# would have already do it.
1331
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1332
rev = self.get_revision(revision_id)
1333
rev_tmp = cStringIO.StringIO()
1334
# the current serializer..
1335
self._serializer.write_revision(rev, rev_tmp)
1337
return rev_tmp.getvalue()
1339
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1340
"""Produce a generator of revision deltas.
1342
Note that the input is a sequence of REVISIONS, not revision_ids.
1343
Trees will be held in memory until the generator exits.
1344
Each delta is relative to the revision's lefthand predecessor.
1346
:param specific_fileids: if not None, the result is filtered
1347
so that only those file-ids, their parents and their
1348
children are included.
1350
# Get the revision-ids of interest
1351
required_trees = set()
1352
for revision in revisions:
1353
required_trees.add(revision.revision_id)
1354
required_trees.update(revision.parent_ids[:1])
1356
# Get the matching filtered trees. Note that it's more
1357
# efficient to pass filtered trees to changes_from() rather
1358
# than doing the filtering afterwards. changes_from() could
1359
# arguably do the filtering itself but it's path-based, not
1360
# file-id based, so filtering before or afterwards is
1362
if specific_fileids is None:
1363
trees = dict((t.get_revision_id(), t) for
1364
t in self.revision_trees(required_trees))
1366
trees = dict((t.get_revision_id(), t) for
1367
t in self._filtered_revision_trees(required_trees,
1370
# Calculate the deltas
1371
for revision in revisions:
1372
if not revision.parent_ids:
1373
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1375
old_tree = trees[revision.parent_ids[0]]
1376
yield trees[revision.revision_id].changes_from(old_tree)
1379
def get_revision_delta(self, revision_id, specific_fileids=None):
1380
"""Return the delta for one revision.
1382
The delta is relative to the left-hand predecessor of the
1385
:param specific_fileids: if not None, the result is filtered
1386
so that only those file-ids, their parents and their
1387
children are included.
1389
r = self.get_revision(revision_id)
1390
return list(self.get_deltas_for_revisions([r],
1391
specific_fileids=specific_fileids))[0]
1394
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1395
signature = gpg_strategy.sign(plaintext)
1396
self.add_signature_text(revision_id, signature)
1399
def add_signature_text(self, revision_id, signature):
1400
self.signatures.add_lines((revision_id,), (),
1401
osutils.split_lines(signature))
1403
def find_text_key_references(self):
1404
"""Find the text key references within the repository.
1406
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1407
to whether they were referred to by the inventory of the
1408
revision_id that they contain. The inventory texts from all present
1409
revision ids are assessed to generate this report.
1411
revision_keys = self.revisions.keys()
1412
w = self.inventories
1413
pb = ui.ui_factory.nested_progress_bar()
1415
return self._find_text_key_references_from_xml_inventory_lines(
1416
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1420
def _find_text_key_references_from_xml_inventory_lines(self,
1422
"""Core routine for extracting references to texts from inventories.
1424
This performs the translation of xml lines to revision ids.
1426
:param line_iterator: An iterator of lines, origin_version_id
1427
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1428
to whether they were referred to by the inventory of the
1429
revision_id that they contain. Note that if that revision_id was
1430
not part of the line_iterator's output then False will be given -
1431
even though it may actually refer to that key.
1433
if not self._serializer.support_altered_by_hack:
1434
raise AssertionError(
1435
"_find_text_key_references_from_xml_inventory_lines only "
1436
"supported for branches which store inventory as unnested xml"
1437
", not on %r" % self)
1440
# this code needs to read every new line in every inventory for the
1441
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1442
# not present in one of those inventories is unnecessary but not
1443
# harmful because we are filtering by the revision id marker in the
1444
# inventory lines : we only select file ids altered in one of those
1445
# revisions. We don't need to see all lines in the inventory because
1446
# only those added in an inventory in rev X can contain a revision=X
1448
unescape_revid_cache = {}
1449
unescape_fileid_cache = {}
1451
# jam 20061218 In a big fetch, this handles hundreds of thousands
1452
# of lines, so it has had a lot of inlining and optimizing done.
1453
# Sorry that it is a little bit messy.
1454
# Move several functions to be local variables, since this is a long
1456
search = self._file_ids_altered_regex.search
1457
unescape = _unescape_xml
1458
setdefault = result.setdefault
1459
for line, line_key in line_iterator:
1460
match = search(line)
1463
# One call to match.group() returning multiple items is quite a
1464
# bit faster than 2 calls to match.group() each returning 1
1465
file_id, revision_id = match.group('file_id', 'revision_id')
1467
# Inlining the cache lookups helps a lot when you make 170,000
1468
# lines and 350k ids, versus 8.4 unique ids.
1469
# Using a cache helps in 2 ways:
1470
# 1) Avoids unnecessary decoding calls
1471
# 2) Re-uses cached strings, which helps in future set and
1473
# (2) is enough that removing encoding entirely along with
1474
# the cache (so we are using plain strings) results in no
1475
# performance improvement.
1477
revision_id = unescape_revid_cache[revision_id]
1479
unescaped = unescape(revision_id)
1480
unescape_revid_cache[revision_id] = unescaped
1481
revision_id = unescaped
1483
# Note that unconditionally unescaping means that we deserialise
1484
# every fileid, which for general 'pull' is not great, but we don't
1485
# really want to have some many fulltexts that this matters anyway.
1488
file_id = unescape_fileid_cache[file_id]
1490
unescaped = unescape(file_id)
1491
unescape_fileid_cache[file_id] = unescaped
1494
key = (file_id, revision_id)
1495
setdefault(key, False)
1496
if revision_id == line_key[-1]:
1500
def _inventory_xml_lines_for_keys(self, keys):
1501
"""Get a line iterator of the sort needed for findind references.
1503
Not relevant for non-xml inventory repositories.
1505
Ghosts in revision_keys are ignored.
1507
:param revision_keys: The revision keys for the inventories to inspect.
1508
:return: An iterator over (inventory line, revid) for the fulltexts of
1509
all of the xml inventories specified by revision_keys.
1511
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1512
for record in stream:
1513
if record.storage_kind != 'absent':
1514
chunks = record.get_bytes_as('chunked')
1515
revid = record.key[-1]
1516
lines = osutils.chunks_to_lines(chunks)
1520
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1522
"""Helper routine for fileids_altered_by_revision_ids.
1524
This performs the translation of xml lines to revision ids.
1526
:param line_iterator: An iterator of lines, origin_version_id
1527
:param revision_ids: The revision ids to filter for. This should be a
1528
set or other type which supports efficient __contains__ lookups, as
1529
the revision id from each parsed line will be looked up in the
1530
revision_ids filter.
1531
:return: a dictionary mapping altered file-ids to an iterable of
1532
revision_ids. Each altered file-ids has the exact revision_ids that
1533
altered it listed explicitly.
1535
seen = set(self._find_text_key_references_from_xml_inventory_lines(
1536
line_iterator).iterkeys())
1537
# Note that revision_ids are revision keys.
1538
parent_maps = self.revisions.get_parent_map(revision_ids)
1540
map(parents.update, parent_maps.itervalues())
1541
parents.difference_update(revision_ids)
1542
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
1543
self._inventory_xml_lines_for_keys(parents)))
1544
new_keys = seen - parent_seen
1546
setdefault = result.setdefault
1547
for key in new_keys:
1548
setdefault(key[0], set()).add(key[-1])
1551
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1552
"""Find the file ids and versions affected by revisions.
1554
:param revisions: an iterable containing revision ids.
1555
:param _inv_weave: The inventory weave from this repository or None.
1556
If None, the inventory weave will be opened automatically.
1557
:return: a dictionary mapping altered file-ids to an iterable of
1558
revision_ids. Each altered file-ids has the exact revision_ids that
1559
altered it listed explicitly.
1561
selected_keys = set((revid,) for revid in revision_ids)
1562
w = _inv_weave or self.inventories
1563
pb = ui.ui_factory.nested_progress_bar()
1565
return self._find_file_ids_from_xml_inventory_lines(
1566
w.iter_lines_added_or_present_in_keys(
1567
selected_keys, pb=pb),
1572
def iter_files_bytes(self, desired_files):
1573
"""Iterate through file versions.
1575
Files will not necessarily be returned in the order they occur in
1576
desired_files. No specific order is guaranteed.
1578
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1579
value supplied by the caller as part of desired_files. It should
1580
uniquely identify the file version in the caller's context. (Examples:
1581
an index number or a TreeTransform trans_id.)
1583
bytes_iterator is an iterable of bytestrings for the file. The
1584
kind of iterable and length of the bytestrings are unspecified, but for
1585
this implementation, it is a list of bytes produced by
1586
VersionedFile.get_record_stream().
1588
:param desired_files: a list of (file_id, revision_id, identifier)
1592
for file_id, revision_id, callable_data in desired_files:
1593
text_keys[(file_id, revision_id)] = callable_data
1594
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1595
if record.storage_kind == 'absent':
1596
raise errors.RevisionNotPresent(record.key, self)
1597
yield text_keys[record.key], record.get_bytes_as('fulltext')
1599
def _generate_text_key_index(self, text_key_references=None,
1601
"""Generate a new text key index for the repository.
1603
This is an expensive function that will take considerable time to run.
1605
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1606
list of parents, also text keys. When a given key has no parents,
1607
the parents list will be [NULL_REVISION].
1609
# All revisions, to find inventory parents.
1610
if ancestors is None:
1611
graph = self.get_graph()
1612
ancestors = graph.get_parent_map(self.all_revision_ids())
1613
if text_key_references is None:
1614
text_key_references = self.find_text_key_references()
1615
pb = ui.ui_factory.nested_progress_bar()
1617
return self._do_generate_text_key_index(ancestors,
1618
text_key_references, pb)
1622
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1623
"""Helper for _generate_text_key_index to avoid deep nesting."""
1624
revision_order = tsort.topo_sort(ancestors)
1625
invalid_keys = set()
1627
for revision_id in revision_order:
1628
revision_keys[revision_id] = set()
1629
text_count = len(text_key_references)
1630
# a cache of the text keys to allow reuse; costs a dict of all the
1631
# keys, but saves a 2-tuple for every child of a given key.
1633
for text_key, valid in text_key_references.iteritems():
1635
invalid_keys.add(text_key)
1637
revision_keys[text_key[1]].add(text_key)
1638
text_key_cache[text_key] = text_key
1639
del text_key_references
1641
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1642
NULL_REVISION = _mod_revision.NULL_REVISION
1643
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1644
# too small for large or very branchy trees. However, for 55K path
1645
# trees, it would be easy to use too much memory trivially. Ideally we
1646
# could gauge this by looking at available real memory etc, but this is
1647
# always a tricky proposition.
1648
inventory_cache = lru_cache.LRUCache(10)
1649
batch_size = 10 # should be ~150MB on a 55K path tree
1650
batch_count = len(revision_order) / batch_size + 1
1652
pb.update("Calculating text parents", processed_texts, text_count)
1653
for offset in xrange(batch_count):
1654
to_query = revision_order[offset * batch_size:(offset + 1) *
1658
for rev_tree in self.revision_trees(to_query):
1659
revision_id = rev_tree.get_revision_id()
1660
parent_ids = ancestors[revision_id]
1661
for text_key in revision_keys[revision_id]:
1662
pb.update("Calculating text parents", processed_texts)
1663
processed_texts += 1
1664
candidate_parents = []
1665
for parent_id in parent_ids:
1666
parent_text_key = (text_key[0], parent_id)
1668
check_parent = parent_text_key not in \
1669
revision_keys[parent_id]
1671
# the parent parent_id is a ghost:
1672
check_parent = False
1673
# truncate the derived graph against this ghost.
1674
parent_text_key = None
1676
# look at the parent commit details inventories to
1677
# determine possible candidates in the per file graph.
1680
inv = inventory_cache[parent_id]
1682
inv = self.revision_tree(parent_id).inventory
1683
inventory_cache[parent_id] = inv
1684
parent_entry = inv._byid.get(text_key[0], None)
1685
if parent_entry is not None:
1687
text_key[0], parent_entry.revision)
1689
parent_text_key = None
1690
if parent_text_key is not None:
1691
candidate_parents.append(
1692
text_key_cache[parent_text_key])
1693
parent_heads = text_graph.heads(candidate_parents)
1694
new_parents = list(parent_heads)
1695
new_parents.sort(key=lambda x:candidate_parents.index(x))
1696
if new_parents == []:
1697
new_parents = [NULL_REVISION]
1698
text_index[text_key] = new_parents
1700
for text_key in invalid_keys:
1701
text_index[text_key] = [NULL_REVISION]
1704
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1705
"""Get an iterable listing the keys of all the data introduced by a set
1708
The keys will be ordered so that the corresponding items can be safely
1709
fetched and inserted in that order.
1711
:returns: An iterable producing tuples of (knit-kind, file-id,
1712
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1713
'revisions'. file-id is None unless knit-kind is 'file'.
1715
# XXX: it's a bit weird to control the inventory weave caching in this
1716
# generator. Ideally the caching would be done in fetch.py I think. Or
1717
# maybe this generator should explicitly have the contract that it
1718
# should not be iterated until the previously yielded item has been
1720
inv_w = self.inventories
1722
# file ids that changed
1723
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1725
num_file_ids = len(file_ids)
1726
for file_id, altered_versions in file_ids.iteritems():
1727
if _files_pb is not None:
1728
_files_pb.update("fetch texts", count, num_file_ids)
1730
yield ("file", file_id, altered_versions)
1731
# We're done with the files_pb. Note that it finished by the caller,
1732
# just as it was created by the caller.
1736
yield ("inventory", None, revision_ids)
1739
# XXX: Note ATM no callers actually pay attention to this return
1740
# instead they just use the list of revision ids and ignore
1741
# missing sigs. Consider removing this work entirely
1742
revisions_with_signatures = set(self.signatures.get_parent_map(
1743
[(r,) for r in revision_ids]))
1744
revisions_with_signatures = set(
1745
[r for (r,) in revisions_with_signatures])
1746
revisions_with_signatures.intersection_update(revision_ids)
1747
yield ("signatures", None, revisions_with_signatures)
1750
yield ("revisions", None, revision_ids)
1753
def get_inventory(self, revision_id):
1754
"""Get Inventory object by revision id."""
1755
return self.iter_inventories([revision_id]).next()
1757
def iter_inventories(self, revision_ids):
1758
"""Get many inventories by revision_ids.
1760
This will buffer some or all of the texts used in constructing the
1761
inventories in memory, but will only parse a single inventory at a
1764
:return: An iterator of inventories.
1766
if ((None in revision_ids)
1767
or (_mod_revision.NULL_REVISION in revision_ids)):
1768
raise ValueError('cannot get null revision inventory')
1769
return self._iter_inventories(revision_ids)
1771
def _iter_inventories(self, revision_ids):
1772
"""single-document based inventory iteration."""
1773
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1774
yield self.deserialise_inventory(revision_id, text)
1776
def _iter_inventory_xmls(self, revision_ids):
1777
keys = [(revision_id,) for revision_id in revision_ids]
1778
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1780
for record in stream:
1781
if record.storage_kind != 'absent':
1782
text_chunks[record.key] = record.get_bytes_as('chunked')
1784
raise errors.NoSuchRevision(self, record.key)
1786
chunks = text_chunks.pop(key)
1787
yield ''.join(chunks), key[-1]
1789
def deserialise_inventory(self, revision_id, xml):
1790
"""Transform the xml into an inventory object.
1792
:param revision_id: The expected revision id of the inventory.
1793
:param xml: A serialised inventory.
1795
result = self._serializer.read_inventory_from_string(xml, revision_id,
1796
entry_cache=self._inventory_entry_cache)
1797
if result.revision_id != revision_id:
1798
raise AssertionError('revision id mismatch %s != %s' % (
1799
result.revision_id, revision_id))
1802
def serialise_inventory(self, inv):
1803
return self._serializer.write_inventory_to_string(inv)
1805
def _serialise_inventory_to_lines(self, inv):
1806
return self._serializer.write_inventory_to_lines(inv)
1808
def get_serializer_format(self):
1809
return self._serializer.format_num
1812
def get_inventory_xml(self, revision_id):
1813
"""Get inventory XML as a file object."""
1814
texts = self._iter_inventory_xmls([revision_id])
1816
text, revision_id = texts.next()
1817
except StopIteration:
1818
raise errors.HistoryMissing(self, 'inventory', revision_id)
1822
def get_inventory_sha1(self, revision_id):
1823
"""Return the sha1 hash of the inventory entry
1825
return self.get_revision(revision_id).inventory_sha1
1827
def iter_reverse_revision_history(self, revision_id):
1828
"""Iterate backwards through revision ids in the lefthand history
1830
:param revision_id: The revision id to start with. All its lefthand
1831
ancestors will be traversed.
1833
graph = self.get_graph()
1834
next_id = revision_id
1836
if next_id in (None, _mod_revision.NULL_REVISION):
1839
# Note: The following line may raise KeyError in the event of
1840
# truncated history. We decided not to have a try:except:raise
1841
# RevisionNotPresent here until we see a use for it, because of the
1842
# cost in an inner loop that is by its very nature O(history).
1843
# Robert Collins 20080326
1844
parents = graph.get_parent_map([next_id])[next_id]
1845
if len(parents) == 0:
1848
next_id = parents[0]
1851
def get_revision_inventory(self, revision_id):
1852
"""Return inventory of a past revision."""
1853
# TODO: Unify this with get_inventory()
1854
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1855
# must be the same as its revision, so this is trivial.
1856
if revision_id is None:
1857
# This does not make sense: if there is no revision,
1858
# then it is the current tree inventory surely ?!
1859
# and thus get_root_id() is something that looks at the last
1860
# commit on the branch, and the get_root_id is an inventory check.
1861
raise NotImplementedError
1862
# return Inventory(self.get_root_id())
1864
return self.get_inventory(revision_id)
1866
def is_shared(self):
1867
"""Return True if this repository is flagged as a shared repository."""
1868
raise NotImplementedError(self.is_shared)
1871
def reconcile(self, other=None, thorough=False):
1872
"""Reconcile this repository."""
1873
from bzrlib.reconcile import RepoReconciler
1874
reconciler = RepoReconciler(self, thorough=thorough)
1875
reconciler.reconcile()
1878
def _refresh_data(self):
1879
"""Helper called from lock_* to ensure coherency with disk.
1881
The default implementation does nothing; it is however possible
1882
for repositories to maintain loaded indices across multiple locks
1883
by checking inside their implementation of this method to see
1884
whether their indices are still valid. This depends of course on
1885
the disk format being validatable in this manner. This method is
1886
also called by the refresh_data() public interface to cause a refresh
1887
to occur while in a write lock so that data inserted by a smart server
1888
push operation is visible on the client's instance of the physical
1893
def revision_tree(self, revision_id):
1894
"""Return Tree for a revision on this branch.
1896
`revision_id` may be NULL_REVISION for the empty tree revision.
1898
revision_id = _mod_revision.ensure_null(revision_id)
1899
# TODO: refactor this to use an existing revision object
1900
# so we don't need to read it in twice.
1901
if revision_id == _mod_revision.NULL_REVISION:
1902
return RevisionTree(self, Inventory(root_id=None),
1903
_mod_revision.NULL_REVISION)
1905
inv = self.get_revision_inventory(revision_id)
1906
return RevisionTree(self, inv, revision_id)
1908
def revision_trees(self, revision_ids):
1909
"""Return Trees for revisions in this repository.
1911
:param revision_ids: a sequence of revision-ids;
1912
a revision-id may not be None or 'null:'
1914
inventories = self.iter_inventories(revision_ids)
1915
for inv in inventories:
1916
yield RevisionTree(self, inv, inv.revision_id)
1918
def _filtered_revision_trees(self, revision_ids, file_ids):
1919
"""Return Tree for a revision on this branch with only some files.
1921
:param revision_ids: a sequence of revision-ids;
1922
a revision-id may not be None or 'null:'
1923
:param file_ids: if not None, the result is filtered
1924
so that only those file-ids, their parents and their
1925
children are included.
1927
inventories = self.iter_inventories(revision_ids)
1928
for inv in inventories:
1929
# Should we introduce a FilteredRevisionTree class rather
1930
# than pre-filter the inventory here?
1931
filtered_inv = inv.filter(file_ids)
1932
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
1935
def get_ancestry(self, revision_id, topo_sorted=True):
1936
"""Return a list of revision-ids integrated by a revision.
1938
The first element of the list is always None, indicating the origin
1939
revision. This might change when we have history horizons, or
1940
perhaps we should have a new API.
1942
This is topologically sorted.
1944
if _mod_revision.is_null(revision_id):
1946
if not self.has_revision(revision_id):
1947
raise errors.NoSuchRevision(self, revision_id)
1948
graph = self.get_graph()
1950
search = graph._make_breadth_first_searcher([revision_id])
1953
found, ghosts = search.next_with_ghosts()
1954
except StopIteration:
1957
if _mod_revision.NULL_REVISION in keys:
1958
keys.remove(_mod_revision.NULL_REVISION)
1960
parent_map = graph.get_parent_map(keys)
1961
keys = tsort.topo_sort(parent_map)
1962
return [None] + list(keys)
1965
"""Compress the data within the repository.
1967
This operation only makes sense for some repository types. For other
1968
types it should be a no-op that just returns.
1970
This stub method does not require a lock, but subclasses should use
1971
@needs_write_lock as this is a long running call its reasonable to
1972
implicitly lock for the user.
1976
@deprecated_method(one_six)
1977
def print_file(self, file, revision_id):
1978
"""Print `file` to stdout.
1980
FIXME RBC 20060125 as John Meinel points out this is a bad api
1981
- it writes to stdout, it assumes that that is valid etc. Fix
1982
by creating a new more flexible convenience function.
1984
tree = self.revision_tree(revision_id)
1985
# use inventory as it was in that revision
1986
file_id = tree.inventory.path2id(file)
1988
# TODO: jam 20060427 Write a test for this code path
1989
# it had a bug in it, and was raising the wrong
1991
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1992
tree.print_file(file_id)
1994
def get_transaction(self):
1995
return self.control_files.get_transaction()
1997
@deprecated_method(one_one)
1998
def get_parents(self, revision_ids):
1999
"""See StackedParentsProvider.get_parents"""
2000
parent_map = self.get_parent_map(revision_ids)
2001
return [parent_map.get(r, None) for r in revision_ids]
2003
def get_parent_map(self, revision_ids):
2004
"""See graph._StackedParentsProvider.get_parent_map"""
2005
# revisions index works in keys; this just works in revisions
2006
# therefore wrap and unwrap
2009
for revision_id in revision_ids:
2010
if revision_id == _mod_revision.NULL_REVISION:
2011
result[revision_id] = ()
2012
elif revision_id is None:
2013
raise ValueError('get_parent_map(None) is not valid')
2015
query_keys.append((revision_id ,))
2016
for ((revision_id,), parent_keys) in \
2017
self.revisions.get_parent_map(query_keys).iteritems():
2019
result[revision_id] = tuple(parent_revid
2020
for (parent_revid,) in parent_keys)
2022
result[revision_id] = (_mod_revision.NULL_REVISION,)
2025
def _make_parents_provider(self):
2028
def get_graph(self, other_repository=None):
2029
"""Return the graph walker for this repository format"""
2030
parents_provider = self._make_parents_provider()
2031
if (other_repository is not None and
2032
not self.has_same_location(other_repository)):
2033
parents_provider = graph._StackedParentsProvider(
2034
[parents_provider, other_repository._make_parents_provider()])
2035
return graph.Graph(parents_provider)
2037
def _get_versioned_file_checker(self, text_key_references=None):
2038
"""Return an object suitable for checking versioned files.
2040
:param text_key_references: if non-None, an already built
2041
dictionary mapping text keys ((fileid, revision_id) tuples)
2042
to whether they were referred to by the inventory of the
2043
revision_id that they contain. If None, this will be
2046
return _VersionedFileChecker(self,
2047
text_key_references=text_key_references)
2049
def revision_ids_to_search_result(self, result_set):
2050
"""Convert a set of revision ids to a graph SearchResult."""
2051
result_parents = set()
2052
for parents in self.get_graph().get_parent_map(
2053
result_set).itervalues():
2054
result_parents.update(parents)
2055
included_keys = result_set.intersection(result_parents)
2056
start_keys = result_set.difference(included_keys)
2057
exclude_keys = result_parents.difference(result_set)
2058
result = graph.SearchResult(start_keys, exclude_keys,
2059
len(result_set), result_set)
2063
def set_make_working_trees(self, new_value):
2064
"""Set the policy flag for making working trees when creating branches.
2066
This only applies to branches that use this repository.
2068
The default is 'True'.
2069
:param new_value: True to restore the default, False to disable making
2072
raise NotImplementedError(self.set_make_working_trees)
2074
def make_working_trees(self):
2075
"""Returns the policy for making working trees on new branches."""
2076
raise NotImplementedError(self.make_working_trees)
2079
def sign_revision(self, revision_id, gpg_strategy):
2080
plaintext = Testament.from_revision(self, revision_id).as_short_text()
2081
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2084
def has_signature_for_revision_id(self, revision_id):
2085
"""Query for a revision signature for revision_id in the repository."""
2086
if not self.has_revision(revision_id):
2087
raise errors.NoSuchRevision(self, revision_id)
2088
sig_present = (1 == len(
2089
self.signatures.get_parent_map([(revision_id,)])))
2093
def get_signature_text(self, revision_id):
2094
"""Return the text for a signature."""
2095
stream = self.signatures.get_record_stream([(revision_id,)],
2097
record = stream.next()
2098
if record.storage_kind == 'absent':
2099
raise errors.NoSuchRevision(self, revision_id)
2100
return record.get_bytes_as('fulltext')
2103
def check(self, revision_ids=None):
2104
"""Check consistency of all history of given revision_ids.
2106
Different repository implementations should override _check().
2108
:param revision_ids: A non-empty list of revision_ids whose ancestry
2109
will be checked. Typically the last revision_id of a branch.
2111
return self._check(revision_ids)
2113
def _check(self, revision_ids):
2114
result = check.Check(self)
2118
def _warn_if_deprecated(self):
2119
global _deprecation_warning_done
2120
if _deprecation_warning_done:
2122
_deprecation_warning_done = True
2123
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2124
% (self._format, self.bzrdir.transport.base))
2126
def supports_rich_root(self):
2127
return self._format.rich_root_data
2129
def _check_ascii_revisionid(self, revision_id, method):
2130
"""Private helper for ascii-only repositories."""
2131
# weave repositories refuse to store revisionids that are non-ascii.
2132
if revision_id is not None:
2133
# weaves require ascii revision ids.
2134
if isinstance(revision_id, unicode):
2136
revision_id.encode('ascii')
2137
except UnicodeEncodeError:
2138
raise errors.NonAsciiRevisionId(method, self)
2141
revision_id.decode('ascii')
2142
except UnicodeDecodeError:
2143
raise errors.NonAsciiRevisionId(method, self)
2145
def revision_graph_can_have_wrong_parents(self):
2146
"""Is it possible for this repository to have a revision graph with
2149
If True, then this repository must also implement
2150
_find_inconsistent_revision_parents so that check and reconcile can
2151
check for inconsistencies before proceeding with other checks that may
2152
depend on the revision index being consistent.
2154
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2157
# remove these delegates a while after bzr 0.15
2158
def __make_delegated(name, from_module):
2159
def _deprecated_repository_forwarder():
2160
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2161
% (name, from_module),
2164
m = __import__(from_module, globals(), locals(), [name])
2166
return getattr(m, name)
2167
except AttributeError:
2168
raise AttributeError('module %s has no name %s'
2170
globals()[name] = _deprecated_repository_forwarder
2173
'AllInOneRepository',
2174
'WeaveMetaDirRepository',
2175
'PreSplitOutRepositoryFormat',
2176
'RepositoryFormat4',
2177
'RepositoryFormat5',
2178
'RepositoryFormat6',
2179
'RepositoryFormat7',
2181
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2185
'RepositoryFormatKnit',
2186
'RepositoryFormatKnit1',
2188
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2191
def install_revision(repository, rev, revision_tree):
2192
"""Install all revision data into a repository."""
2193
install_revisions(repository, [(rev, revision_tree, None)])
2196
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2197
"""Install all revision data into a repository.
2199
Accepts an iterable of revision, tree, signature tuples. The signature
2202
repository.start_write_group()
2204
for n, (revision, revision_tree, signature) in enumerate(iterable):
2205
_install_revision(repository, revision, revision_tree, signature)
2207
pb.update('Transferring revisions', n + 1, num_revisions)
2209
repository.abort_write_group()
2212
repository.commit_write_group()
2215
def _install_revision(repository, rev, revision_tree, signature):
2216
"""Install all revision data into a repository."""
2217
present_parents = []
2219
for p_id in rev.parent_ids:
2220
if repository.has_revision(p_id):
2221
present_parents.append(p_id)
2222
parent_trees[p_id] = repository.revision_tree(p_id)
2224
parent_trees[p_id] = repository.revision_tree(
2225
_mod_revision.NULL_REVISION)
2227
inv = revision_tree.inventory
2228
entries = inv.iter_entries()
2229
# backwards compatibility hack: skip the root id.
2230
if not repository.supports_rich_root():
2231
path, root = entries.next()
2232
if root.revision != rev.revision_id:
2233
raise errors.IncompatibleRevision(repr(repository))
2235
for path, ie in entries:
2236
text_keys[(ie.file_id, ie.revision)] = ie
2237
text_parent_map = repository.texts.get_parent_map(text_keys)
2238
missing_texts = set(text_keys) - set(text_parent_map)
2239
# Add the texts that are not already present
2240
for text_key in missing_texts:
2241
ie = text_keys[text_key]
2243
# FIXME: TODO: The following loop overlaps/duplicates that done by
2244
# commit to determine parents. There is a latent/real bug here where
2245
# the parents inserted are not those commit would do - in particular
2246
# they are not filtered by heads(). RBC, AB
2247
for revision, tree in parent_trees.iteritems():
2248
if ie.file_id not in tree:
2250
parent_id = tree.inventory[ie.file_id].revision
2251
if parent_id in text_parents:
2253
text_parents.append((ie.file_id, parent_id))
2254
lines = revision_tree.get_file(ie.file_id).readlines()
2255
repository.texts.add_lines(text_key, text_parents, lines)
2257
# install the inventory
2258
repository.add_inventory(rev.revision_id, inv, present_parents)
2259
except errors.RevisionAlreadyPresent:
2261
if signature is not None:
2262
repository.add_signature_text(rev.revision_id, signature)
2263
repository.add_revision(rev.revision_id, rev, inv)
2266
class MetaDirRepository(Repository):
2267
"""Repositories in the new meta-dir layout.
2269
:ivar _transport: Transport for access to repository control files,
2270
typically pointing to .bzr/repository.
2273
def __init__(self, _format, a_bzrdir, control_files):
2274
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2275
self._transport = control_files._transport
2277
def is_shared(self):
2278
"""Return True if this repository is flagged as a shared repository."""
2279
return self._transport.has('shared-storage')
2282
def set_make_working_trees(self, new_value):
2283
"""Set the policy flag for making working trees when creating branches.
2285
This only applies to branches that use this repository.
2287
The default is 'True'.
2288
:param new_value: True to restore the default, False to disable making
2293
self._transport.delete('no-working-trees')
2294
except errors.NoSuchFile:
2297
self._transport.put_bytes('no-working-trees', '',
2298
mode=self.bzrdir._get_file_mode())
2300
def make_working_trees(self):
2301
"""Returns the policy for making working trees on new branches."""
2302
return not self._transport.has('no-working-trees')
2305
class MetaDirVersionedFileRepository(MetaDirRepository):
2306
"""Repositories in a meta-dir, that work via versioned file objects."""
2308
def __init__(self, _format, a_bzrdir, control_files):
2309
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2313
network_format_registry = registry.FormatRegistry()
2314
"""Registry of formats indexed by their network name.
2316
The network name for a repository format is an identifier that can be used when
2317
referring to formats with smart server operations. See
2318
RepositoryFormat.network_name() for more detail.
2322
format_registry = registry.FormatRegistry(network_format_registry)
2323
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2325
This can contain either format instances themselves, or classes/factories that
2326
can be called to obtain one.
2330
#####################################################################
2331
# Repository Formats
2333
class RepositoryFormat(object):
2334
"""A repository format.
2336
Formats provide four things:
2337
* An initialization routine to construct repository data on disk.
2338
* a optional format string which is used when the BzrDir supports
2340
* an open routine which returns a Repository instance.
2341
* A network name for referring to the format in smart server RPC
2344
There is one and only one Format subclass for each on-disk format. But
2345
there can be one Repository subclass that is used for several different
2346
formats. The _format attribute on a Repository instance can be used to
2347
determine the disk format.
2349
Formats are placed in a registry by their format string for reference
2350
during opening. These should be subclasses of RepositoryFormat for
2353
Once a format is deprecated, just deprecate the initialize and open
2354
methods on the format class. Do not deprecate the object, as the
2355
object may be created even when a repository instnace hasn't been
2358
Common instance attributes:
2359
_matchingbzrdir - the bzrdir format that the repository format was
2360
originally written to work with. This can be used if manually
2361
constructing a bzrdir and repository, or more commonly for test suite
2365
# Set to True or False in derived classes. True indicates that the format
2366
# supports ghosts gracefully.
2367
supports_ghosts = None
2368
# Can this repository be given external locations to lookup additional
2369
# data. Set to True or False in derived classes.
2370
supports_external_lookups = None
2371
# What order should fetch operations request streams in?
2372
# The default is unordered as that is the cheapest for an origin to
2374
_fetch_order = 'unordered'
2375
# Does this repository format use deltas that can be fetched as-deltas ?
2376
# (E.g. knits, where the knit deltas can be transplanted intact.
2377
# We default to False, which will ensure that enough data to get
2378
# a full text out of any fetch stream will be grabbed.
2379
_fetch_uses_deltas = False
2380
# Should fetch trigger a reconcile after the fetch? Only needed for
2381
# some repository formats that can suffer internal inconsistencies.
2382
_fetch_reconcile = False
2385
return "<%s>" % self.__class__.__name__
2387
def __eq__(self, other):
2388
# format objects are generally stateless
2389
return isinstance(other, self.__class__)
2391
def __ne__(self, other):
2392
return not self == other
2395
def find_format(klass, a_bzrdir):
2396
"""Return the format for the repository object in a_bzrdir.
2398
This is used by bzr native formats that have a "format" file in
2399
the repository. Other methods may be used by different types of
2403
transport = a_bzrdir.get_repository_transport(None)
2404
format_string = transport.get("format").read()
2405
return format_registry.get(format_string)
2406
except errors.NoSuchFile:
2407
raise errors.NoRepositoryPresent(a_bzrdir)
2409
raise errors.UnknownFormatError(format=format_string,
2413
def register_format(klass, format):
2414
format_registry.register(format.get_format_string(), format)
2417
def unregister_format(klass, format):
2418
format_registry.remove(format.get_format_string())
2421
def get_default_format(klass):
2422
"""Return the current default format."""
2423
from bzrlib import bzrdir
2424
return bzrdir.format_registry.make_bzrdir('default').repository_format
2426
def get_format_string(self):
2427
"""Return the ASCII format string that identifies this format.
2429
Note that in pre format ?? repositories the format string is
2430
not permitted nor written to disk.
2432
raise NotImplementedError(self.get_format_string)
2434
def get_format_description(self):
2435
"""Return the short description for this format."""
2436
raise NotImplementedError(self.get_format_description)
2438
# TODO: this shouldn't be in the base class, it's specific to things that
2439
# use weaves or knits -- mbp 20070207
2440
def _get_versioned_file_store(self,
2445
versionedfile_class=None,
2446
versionedfile_kwargs={},
2448
if versionedfile_class is None:
2449
versionedfile_class = self._versionedfile_class
2450
weave_transport = control_files._transport.clone(name)
2451
dir_mode = control_files._dir_mode
2452
file_mode = control_files._file_mode
2453
return VersionedFileStore(weave_transport, prefixed=prefixed,
2455
file_mode=file_mode,
2456
versionedfile_class=versionedfile_class,
2457
versionedfile_kwargs=versionedfile_kwargs,
2460
def initialize(self, a_bzrdir, shared=False):
2461
"""Initialize a repository of this format in a_bzrdir.
2463
:param a_bzrdir: The bzrdir to put the new repository in it.
2464
:param shared: The repository should be initialized as a sharable one.
2465
:returns: The new repository object.
2467
This may raise UninitializableFormat if shared repository are not
2468
compatible the a_bzrdir.
2470
raise NotImplementedError(self.initialize)
2472
def is_supported(self):
2473
"""Is this format supported?
2475
Supported formats must be initializable and openable.
2476
Unsupported formats may not support initialization or committing or
2477
some other features depending on the reason for not being supported.
2481
def network_name(self):
2482
"""A simple byte string uniquely identifying this format for RPC calls.
2484
MetaDir repository formats use their disk format string to identify the
2485
repository over the wire. All in one formats such as bzr < 0.8, and
2486
foreign formats like svn/git and hg should use some marker which is
2487
unique and immutable.
2489
raise NotImplementedError(self.network_name)
2491
def check_conversion_target(self, target_format):
2492
raise NotImplementedError(self.check_conversion_target)
2494
def open(self, a_bzrdir, _found=False):
2495
"""Return an instance of this format for the bzrdir a_bzrdir.
2497
_found is a private parameter, do not use it.
2499
raise NotImplementedError(self.open)
2502
class MetaDirRepositoryFormat(RepositoryFormat):
2503
"""Common base class for the new repositories using the metadir layout."""
2505
rich_root_data = False
2506
supports_tree_reference = False
2507
supports_external_lookups = False
2510
def _matchingbzrdir(self):
2511
matching = bzrdir.BzrDirMetaFormat1()
2512
matching.repository_format = self
2516
super(MetaDirRepositoryFormat, self).__init__()
2518
def _create_control_files(self, a_bzrdir):
2519
"""Create the required files and the initial control_files object."""
2520
# FIXME: RBC 20060125 don't peek under the covers
2521
# NB: no need to escape relative paths that are url safe.
2522
repository_transport = a_bzrdir.get_repository_transport(self)
2523
control_files = lockable_files.LockableFiles(repository_transport,
2524
'lock', lockdir.LockDir)
2525
control_files.create_lock()
2526
return control_files
2528
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2529
"""Upload the initial blank content."""
2530
control_files = self._create_control_files(a_bzrdir)
2531
control_files.lock_write()
2532
transport = control_files._transport
2534
utf8_files += [('shared-storage', '')]
2536
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2537
for (filename, content_stream) in files:
2538
transport.put_file(filename, content_stream,
2539
mode=a_bzrdir._get_file_mode())
2540
for (filename, content_bytes) in utf8_files:
2541
transport.put_bytes_non_atomic(filename, content_bytes,
2542
mode=a_bzrdir._get_file_mode())
2544
control_files.unlock()
2546
def network_name(self):
2547
"""Metadir formats have matching disk and network format strings."""
2548
return self.get_format_string()
2551
# Pre-0.8 formats that don't have a disk format string (because they are
2552
# versioned by the matching control directory). We use the control directories
2553
# disk format string as a key for the network_name because they meet the
2554
# constraints (simple string, unique, immmutable).
2555
network_format_registry.register_lazy(
2556
"Bazaar-NG branch, format 5\n",
2557
'bzrlib.repofmt.weaverepo',
2558
'RepositoryFormat5',
2560
network_format_registry.register_lazy(
2561
"Bazaar-NG branch, format 6\n",
2562
'bzrlib.repofmt.weaverepo',
2563
'RepositoryFormat6',
2566
# formats which have no format string are not discoverable or independently
2567
# creatable on disk, so are not registered in format_registry. They're
2568
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2569
# needed, it's constructed directly by the BzrDir. Non-native formats where
2570
# the repository is not separately opened are similar.
2572
format_registry.register_lazy(
2573
'Bazaar-NG Repository format 7',
2574
'bzrlib.repofmt.weaverepo',
2578
format_registry.register_lazy(
2579
'Bazaar-NG Knit Repository Format 1',
2580
'bzrlib.repofmt.knitrepo',
2581
'RepositoryFormatKnit1',
2584
format_registry.register_lazy(
2585
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2586
'bzrlib.repofmt.knitrepo',
2587
'RepositoryFormatKnit3',
2590
format_registry.register_lazy(
2591
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2592
'bzrlib.repofmt.knitrepo',
2593
'RepositoryFormatKnit4',
2596
# Pack-based formats. There is one format for pre-subtrees, and one for
2597
# post-subtrees to allow ease of testing.
2598
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2599
format_registry.register_lazy(
2600
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2601
'bzrlib.repofmt.pack_repo',
2602
'RepositoryFormatKnitPack1',
2604
format_registry.register_lazy(
2605
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2606
'bzrlib.repofmt.pack_repo',
2607
'RepositoryFormatKnitPack3',
2609
format_registry.register_lazy(
2610
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2611
'bzrlib.repofmt.pack_repo',
2612
'RepositoryFormatKnitPack4',
2614
format_registry.register_lazy(
2615
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2616
'bzrlib.repofmt.pack_repo',
2617
'RepositoryFormatKnitPack5',
2619
format_registry.register_lazy(
2620
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2621
'bzrlib.repofmt.pack_repo',
2622
'RepositoryFormatKnitPack5RichRoot',
2624
format_registry.register_lazy(
2625
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2626
'bzrlib.repofmt.pack_repo',
2627
'RepositoryFormatKnitPack5RichRootBroken',
2629
format_registry.register_lazy(
2630
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2631
'bzrlib.repofmt.pack_repo',
2632
'RepositoryFormatKnitPack6',
2634
format_registry.register_lazy(
2635
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2636
'bzrlib.repofmt.pack_repo',
2637
'RepositoryFormatKnitPack6RichRoot',
2640
# Development formats.
2641
# 1.7->1.8 go below here
2642
format_registry.register_lazy(
2643
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2644
'bzrlib.repofmt.pack_repo',
2645
'RepositoryFormatPackDevelopment2',
2647
format_registry.register_lazy(
2648
("Bazaar development format 2 with subtree support "
2649
"(needs bzr.dev from before 1.8)\n"),
2650
'bzrlib.repofmt.pack_repo',
2651
'RepositoryFormatPackDevelopment2Subtree',
2655
class InterRepository(InterObject):
2656
"""This class represents operations taking place between two repositories.
2658
Its instances have methods like copy_content and fetch, and contain
2659
references to the source and target repositories these operations can be
2662
Often we will provide convenience methods on 'repository' which carry out
2663
operations with another repository - they will always forward to
2664
InterRepository.get(other).method_name(parameters).
2667
_walk_to_common_revisions_batch_size = 50
2669
"""The available optimised InterRepository types."""
2672
def copy_content(self, revision_id=None):
2673
"""Make a complete copy of the content in self into destination.
2675
This is a destructive operation! Do not use it on existing
2678
:param revision_id: Only copy the content needed to construct
2679
revision_id and its parents.
2682
self.target.set_make_working_trees(self.source.make_working_trees())
2683
except NotImplementedError:
2685
self.target.fetch(self.source, revision_id=revision_id)
2688
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2690
"""Fetch the content required to construct revision_id.
2692
The content is copied from self.source to self.target.
2694
:param revision_id: if None all content is copied, if NULL_REVISION no
2696
:param pb: optional progress bar to use for progress reports. If not
2697
provided a default one will be created.
2700
from bzrlib.fetch import RepoFetcher
2701
f = RepoFetcher(to_repository=self.target,
2702
from_repository=self.source,
2703
last_revision=revision_id,
2704
fetch_spec=fetch_spec,
2705
pb=pb, find_ghosts=find_ghosts)
2707
def _walk_to_common_revisions(self, revision_ids):
2708
"""Walk out from revision_ids in source to revisions target has.
2710
:param revision_ids: The start point for the search.
2711
:return: A set of revision ids.
2713
target_graph = self.target.get_graph()
2714
revision_ids = frozenset(revision_ids)
2715
# Fast path for the case where all the revisions are already in the
2717
# (Although this does incur an extra round trip for the
2718
# fairly common case where the target doesn't already have the revision
2720
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2721
return graph.SearchResult(revision_ids, set(), 0, set())
2722
missing_revs = set()
2723
source_graph = self.source.get_graph()
2724
# ensure we don't pay silly lookup costs.
2725
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2726
null_set = frozenset([_mod_revision.NULL_REVISION])
2727
searcher_exhausted = False
2731
# Iterate the searcher until we have enough next_revs
2732
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2734
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2735
next_revs.update(next_revs_part)
2736
ghosts.update(ghosts_part)
2737
except StopIteration:
2738
searcher_exhausted = True
2740
# If there are ghosts in the source graph, and the caller asked for
2741
# them, make sure that they are present in the target.
2742
# We don't care about other ghosts as we can't fetch them and
2743
# haven't been asked to.
2744
ghosts_to_check = set(revision_ids.intersection(ghosts))
2745
revs_to_get = set(next_revs).union(ghosts_to_check)
2747
have_revs = set(target_graph.get_parent_map(revs_to_get))
2748
# we always have NULL_REVISION present.
2749
have_revs = have_revs.union(null_set)
2750
# Check if the target is missing any ghosts we need.
2751
ghosts_to_check.difference_update(have_revs)
2753
# One of the caller's revision_ids is a ghost in both the
2754
# source and the target.
2755
raise errors.NoSuchRevision(
2756
self.source, ghosts_to_check.pop())
2757
missing_revs.update(next_revs - have_revs)
2758
# Because we may have walked past the original stop point, make
2759
# sure everything is stopped
2760
stop_revs = searcher.find_seen_ancestors(have_revs)
2761
searcher.stop_searching_any(stop_revs)
2762
if searcher_exhausted:
2764
return searcher.get_result()
2766
@deprecated_method(one_two)
2768
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2769
"""Return the revision ids that source has that target does not.
2771
These are returned in topological order.
2773
:param revision_id: only return revision ids included by this
2775
:param find_ghosts: If True find missing revisions in deep history
2776
rather than just finding the surface difference.
2778
return list(self.search_missing_revision_ids(
2779
revision_id, find_ghosts).get_keys())
2782
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2783
"""Return the revision ids that source has that target does not.
2785
:param revision_id: only return revision ids included by this
2787
:param find_ghosts: If True find missing revisions in deep history
2788
rather than just finding the surface difference.
2789
:return: A bzrlib.graph.SearchResult.
2791
# stop searching at found target revisions.
2792
if not find_ghosts and revision_id is not None:
2793
return self._walk_to_common_revisions([revision_id])
2794
# generic, possibly worst case, slow code path.
2795
target_ids = set(self.target.all_revision_ids())
2796
if revision_id is not None:
2797
source_ids = self.source.get_ancestry(revision_id)
2798
if source_ids[0] is not None:
2799
raise AssertionError()
2802
source_ids = self.source.all_revision_ids()
2803
result_set = set(source_ids).difference(target_ids)
2804
return self.source.revision_ids_to_search_result(result_set)
2807
def _same_model(source, target):
2808
"""True if source and target have the same data representation.
2810
Note: this is always called on the base class; overriding it in a
2811
subclass will have no effect.
2814
InterRepository._assert_same_model(source, target)
2816
except errors.IncompatibleRepositories, e:
2820
def _assert_same_model(source, target):
2821
"""Raise an exception if two repositories do not use the same model.
2823
if source.supports_rich_root() != target.supports_rich_root():
2824
raise errors.IncompatibleRepositories(source, target,
2825
"different rich-root support")
2826
if source._serializer != target._serializer:
2827
raise errors.IncompatibleRepositories(source, target,
2828
"different serializers")
2831
class InterSameDataRepository(InterRepository):
2832
"""Code for converting between repositories that represent the same data.
2834
Data format and model must match for this to work.
2838
def _get_repo_format_to_test(self):
2839
"""Repository format for testing with.
2841
InterSameData can pull from subtree to subtree and from non-subtree to
2842
non-subtree, so we test this with the richest repository format.
2844
from bzrlib.repofmt import knitrepo
2845
return knitrepo.RepositoryFormatKnit3()
2848
def is_compatible(source, target):
2849
return InterRepository._same_model(source, target)
2852
class InterWeaveRepo(InterSameDataRepository):
2853
"""Optimised code paths between Weave based repositories.
2855
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2856
implemented lazy inter-object optimisation.
2860
def _get_repo_format_to_test(self):
2861
from bzrlib.repofmt import weaverepo
2862
return weaverepo.RepositoryFormat7()
2865
def is_compatible(source, target):
2866
"""Be compatible with known Weave formats.
2868
We don't test for the stores being of specific types because that
2869
could lead to confusing results, and there is no need to be
2872
from bzrlib.repofmt.weaverepo import (
2878
return (isinstance(source._format, (RepositoryFormat5,
2880
RepositoryFormat7)) and
2881
isinstance(target._format, (RepositoryFormat5,
2883
RepositoryFormat7)))
2884
except AttributeError:
2888
def copy_content(self, revision_id=None):
2889
"""See InterRepository.copy_content()."""
2890
# weave specific optimised path:
2892
self.target.set_make_working_trees(self.source.make_working_trees())
2893
except (errors.RepositoryUpgradeRequired, NotImplemented):
2895
# FIXME do not peek!
2896
if self.source._transport.listable():
2897
pb = ui.ui_factory.nested_progress_bar()
2899
self.target.texts.insert_record_stream(
2900
self.source.texts.get_record_stream(
2901
self.source.texts.keys(), 'topological', False))
2902
pb.update('copying inventory', 0, 1)
2903
self.target.inventories.insert_record_stream(
2904
self.source.inventories.get_record_stream(
2905
self.source.inventories.keys(), 'topological', False))
2906
self.target.signatures.insert_record_stream(
2907
self.source.signatures.get_record_stream(
2908
self.source.signatures.keys(),
2910
self.target.revisions.insert_record_stream(
2911
self.source.revisions.get_record_stream(
2912
self.source.revisions.keys(),
2913
'topological', True))
2917
self.target.fetch(self.source, revision_id=revision_id)
2920
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2921
"""See InterRepository.missing_revision_ids()."""
2922
# we want all revisions to satisfy revision_id in source.
2923
# but we don't want to stat every file here and there.
2924
# we want then, all revisions other needs to satisfy revision_id
2925
# checked, but not those that we have locally.
2926
# so the first thing is to get a subset of the revisions to
2927
# satisfy revision_id in source, and then eliminate those that
2928
# we do already have.
2929
# this is slow on high latency connection to self, but as as this
2930
# disk format scales terribly for push anyway due to rewriting
2931
# inventory.weave, this is considered acceptable.
2933
if revision_id is not None:
2934
source_ids = self.source.get_ancestry(revision_id)
2935
if source_ids[0] is not None:
2936
raise AssertionError()
2939
source_ids = self.source._all_possible_ids()
2940
source_ids_set = set(source_ids)
2941
# source_ids is the worst possible case we may need to pull.
2942
# now we want to filter source_ids against what we actually
2943
# have in target, but don't try to check for existence where we know
2944
# we do not have a revision as that would be pointless.
2945
target_ids = set(self.target._all_possible_ids())
2946
possibly_present_revisions = target_ids.intersection(source_ids_set)
2947
actually_present_revisions = set(
2948
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2949
required_revisions = source_ids_set.difference(actually_present_revisions)
2950
if revision_id is not None:
2951
# we used get_ancestry to determine source_ids then we are assured all
2952
# revisions referenced are present as they are installed in topological order.
2953
# and the tip revision was validated by get_ancestry.
2954
result_set = required_revisions
2956
# if we just grabbed the possibly available ids, then
2957
# we only have an estimate of whats available and need to validate
2958
# that against the revision records.
2960
self.source._eliminate_revisions_not_present(required_revisions))
2961
return self.source.revision_ids_to_search_result(result_set)
2964
class InterKnitRepo(InterSameDataRepository):
2965
"""Optimised code paths between Knit based repositories."""
2968
def _get_repo_format_to_test(self):
2969
from bzrlib.repofmt import knitrepo
2970
return knitrepo.RepositoryFormatKnit1()
2973
def is_compatible(source, target):
2974
"""Be compatible with known Knit formats.
2976
We don't test for the stores being of specific types because that
2977
could lead to confusing results, and there is no need to be
2980
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2982
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2983
isinstance(target._format, RepositoryFormatKnit))
2984
except AttributeError:
2986
return are_knits and InterRepository._same_model(source, target)
2989
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2990
"""See InterRepository.missing_revision_ids()."""
2991
if revision_id is not None:
2992
source_ids = self.source.get_ancestry(revision_id)
2993
if source_ids[0] is not None:
2994
raise AssertionError()
2997
source_ids = self.source.all_revision_ids()
2998
source_ids_set = set(source_ids)
2999
# source_ids is the worst possible case we may need to pull.
3000
# now we want to filter source_ids against what we actually
3001
# have in target, but don't try to check for existence where we know
3002
# we do not have a revision as that would be pointless.
3003
target_ids = set(self.target.all_revision_ids())
3004
possibly_present_revisions = target_ids.intersection(source_ids_set)
3005
actually_present_revisions = set(
3006
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3007
required_revisions = source_ids_set.difference(actually_present_revisions)
3008
if revision_id is not None:
3009
# we used get_ancestry to determine source_ids then we are assured all
3010
# revisions referenced are present as they are installed in topological order.
3011
# and the tip revision was validated by get_ancestry.
3012
result_set = required_revisions
3014
# if we just grabbed the possibly available ids, then
3015
# we only have an estimate of whats available and need to validate
3016
# that against the revision records.
3018
self.source._eliminate_revisions_not_present(required_revisions))
3019
return self.source.revision_ids_to_search_result(result_set)
3022
class InterPackRepo(InterSameDataRepository):
3023
"""Optimised code paths between Pack based repositories."""
3026
def _get_repo_format_to_test(self):
3027
from bzrlib.repofmt import pack_repo
3028
return pack_repo.RepositoryFormatKnitPack1()
3031
def is_compatible(source, target):
3032
"""Be compatible with known Pack formats.
3034
We don't test for the stores being of specific types because that
3035
could lead to confusing results, and there is no need to be
3038
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3040
are_packs = (isinstance(source._format, RepositoryFormatPack) and
3041
isinstance(target._format, RepositoryFormatPack))
3042
except AttributeError:
3044
return are_packs and InterRepository._same_model(source, target)
3047
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3049
"""See InterRepository.fetch()."""
3050
if (len(self.source._fallback_repositories) > 0 or
3051
len(self.target._fallback_repositories) > 0):
3052
# The pack layer is not aware of fallback repositories, so when
3053
# fetching from a stacked repository or into a stacked repository
3054
# we use the generic fetch logic which uses the VersionedFiles
3055
# attributes on repository.
3056
from bzrlib.fetch import RepoFetcher
3057
fetcher = RepoFetcher(self.target, self.source, revision_id,
3058
pb, find_ghosts, fetch_spec=fetch_spec)
3059
if fetch_spec is not None:
3060
if len(list(fetch_spec.heads)) != 1:
3061
raise AssertionError(
3062
"InterPackRepo.fetch doesn't support "
3063
"fetching multiple heads yet.")
3064
revision_id = list(fetch_spec.heads)[0]
3066
if revision_id is None:
3068
# everything to do - use pack logic
3069
# to fetch from all packs to one without
3070
# inventory parsing etc, IFF nothing to be copied is in the target.
3072
source_revision_ids = frozenset(self.source.all_revision_ids())
3073
revision_ids = source_revision_ids - \
3074
frozenset(self.target.get_parent_map(source_revision_ids))
3075
revision_keys = [(revid,) for revid in revision_ids]
3076
index = self.target._pack_collection.revision_index.combined_index
3077
present_revision_ids = set(item[1][0] for item in
3078
index.iter_entries(revision_keys))
3079
revision_ids = set(revision_ids) - present_revision_ids
3080
# implementing the TODO will involve:
3081
# - detecting when all of a pack is selected
3082
# - avoiding as much as possible pre-selection, so the
3083
# more-core routines such as create_pack_from_packs can filter in
3084
# a just-in-time fashion. (though having a HEADS list on a
3085
# repository might make this a lot easier, because we could
3086
# sensibly detect 'new revisions' without doing a full index scan.
3087
elif _mod_revision.is_null(revision_id):
3092
revision_ids = self.search_missing_revision_ids(revision_id,
3093
find_ghosts=find_ghosts).get_keys()
3094
except errors.NoSuchRevision:
3095
raise errors.InstallFailed([revision_id])
3096
if len(revision_ids) == 0:
3098
return self._pack(self.source, self.target, revision_ids)
3100
def _pack(self, source, target, revision_ids):
3101
from bzrlib.repofmt.pack_repo import Packer
3102
packs = source._pack_collection.all_packs()
3103
pack = Packer(self.target._pack_collection, packs, '.fetch',
3104
revision_ids).pack()
3105
if pack is not None:
3106
self.target._pack_collection._save_pack_names()
3107
copied_revs = pack.get_revision_count()
3108
# Trigger an autopack. This may duplicate effort as we've just done
3109
# a pack creation, but for now it is simpler to think about as
3110
# 'upload data, then repack if needed'.
3111
self.target._pack_collection.autopack()
3112
return (copied_revs, [])
3117
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3118
"""See InterRepository.missing_revision_ids().
3120
:param find_ghosts: Find ghosts throughout the ancestry of
3123
if not find_ghosts and revision_id is not None:
3124
return self._walk_to_common_revisions([revision_id])
3125
elif revision_id is not None:
3126
# Find ghosts: search for revisions pointing from one repository to
3127
# the other, and vice versa, anywhere in the history of revision_id.
3128
graph = self.target.get_graph(other_repository=self.source)
3129
searcher = graph._make_breadth_first_searcher([revision_id])
3133
next_revs, ghosts = searcher.next_with_ghosts()
3134
except StopIteration:
3136
if revision_id in ghosts:
3137
raise errors.NoSuchRevision(self.source, revision_id)
3138
found_ids.update(next_revs)
3139
found_ids.update(ghosts)
3140
found_ids = frozenset(found_ids)
3141
# Double query here: should be able to avoid this by changing the
3142
# graph api further.
3143
result_set = found_ids - frozenset(
3144
self.target.get_parent_map(found_ids))
3146
source_ids = self.source.all_revision_ids()
3147
# source_ids is the worst possible case we may need to pull.
3148
# now we want to filter source_ids against what we actually
3149
# have in target, but don't try to check for existence where we know
3150
# we do not have a revision as that would be pointless.
3151
target_ids = set(self.target.all_revision_ids())
3152
result_set = set(source_ids).difference(target_ids)
3153
return self.source.revision_ids_to_search_result(result_set)
3156
class InterDifferingSerializer(InterKnitRepo):
3159
def _get_repo_format_to_test(self):
3163
def is_compatible(source, target):
3164
"""Be compatible with Knit2 source and Knit3 target"""
3165
if source.supports_rich_root() != target.supports_rich_root():
3167
# Ideally, we'd support fetching if the source had no tree references
3168
# even if it supported them...
3169
if (getattr(source, '_format.supports_tree_reference', False) and
3170
not getattr(target, '_format.supports_tree_reference', False)):
3174
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3175
"""Get the best delta and base for this revision.
3177
:return: (basis_id, delta)
3179
possible_trees = [(parent_id, cache[parent_id])
3180
for parent_id in parent_ids
3181
if parent_id in cache]
3182
if len(possible_trees) == 0:
3183
# There either aren't any parents, or the parents aren't in the
3184
# cache, so just use the last converted tree
3185
possible_trees.append((basis_id, cache[basis_id]))
3187
for basis_id, basis_tree in possible_trees:
3188
delta = tree.inventory._make_delta(basis_tree.inventory)
3189
deltas.append((len(delta), basis_id, delta))
3191
return deltas[0][1:]
3193
def _fetch_batch(self, revision_ids, basis_id, cache):
3194
"""Fetch across a few revisions.
3196
:param revision_ids: The revisions to copy
3197
:param basis_id: The revision_id of a tree that must be in cache, used
3198
as a basis for delta when no other base is available
3199
:param cache: A cache of RevisionTrees that we can use.
3200
:return: The revision_id of the last converted tree. The RevisionTree
3201
for it will be in cache
3203
# Walk though all revisions; get inventory deltas, copy referenced
3204
# texts that delta references, insert the delta, revision and
3208
pending_revisions = []
3209
parent_map = self.source.get_parent_map(revision_ids)
3210
for tree in self.source.revision_trees(revision_ids):
3211
current_revision_id = tree.get_revision_id()
3212
parent_ids = parent_map.get(current_revision_id, ())
3213
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3215
# Find text entries that need to be copied
3216
for old_path, new_path, file_id, entry in delta:
3217
if new_path is not None:
3218
if not (new_path or self.target.supports_rich_root()):
3219
# We don't copy the text for the root node unless the
3220
# target supports_rich_root.
3222
text_keys.add((file_id, entry.revision))
3223
revision = self.source.get_revision(current_revision_id)
3224
pending_deltas.append((basis_id, delta,
3225
current_revision_id, revision.parent_ids))
3226
pending_revisions.append(revision)
3227
cache[current_revision_id] = tree
3228
basis_id = current_revision_id
3230
from_texts = self.source.texts
3231
to_texts = self.target.texts
3232
to_texts.insert_record_stream(from_texts.get_record_stream(
3233
text_keys, self.target._format._fetch_order,
3234
not self.target._format._fetch_uses_deltas))
3236
for delta in pending_deltas:
3237
self.target.add_inventory_by_delta(*delta)
3238
# insert signatures and revisions
3239
for revision in pending_revisions:
3241
signature = self.source.get_signature_text(
3242
revision.revision_id)
3243
self.target.add_signature_text(revision.revision_id,
3245
except errors.NoSuchRevision:
3247
self.target.add_revision(revision.revision_id, revision)
3250
def _fetch_all_revisions(self, revision_ids, pb):
3251
"""Fetch everything for the list of revisions.
3253
:param revision_ids: The list of revisions to fetch. Must be in
3255
:param pb: A ProgressBar
3258
basis_id, basis_tree = self._get_basis(revision_ids[0])
3260
cache = lru_cache.LRUCache(100)
3261
cache[basis_id] = basis_tree
3262
del basis_tree # We don't want to hang on to it here
3263
for offset in range(0, len(revision_ids), batch_size):
3264
self.target.start_write_group()
3266
pb.update('Transferring revisions', offset,
3268
batch = revision_ids[offset:offset+batch_size]
3269
basis_id = self._fetch_batch(batch, basis_id, cache)
3271
self.target.abort_write_group()
3274
self.target.commit_write_group()
3275
pb.update('Transferring revisions', len(revision_ids),
3279
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3281
"""See InterRepository.fetch()."""
3282
if fetch_spec is not None:
3283
raise AssertionError("Not implemented yet...")
3284
revision_ids = self.target.search_missing_revision_ids(self.source,
3285
revision_id, find_ghosts=find_ghosts).get_keys()
3286
if not revision_ids:
3288
revision_ids = tsort.topo_sort(
3289
self.source.get_graph().get_parent_map(revision_ids))
3291
my_pb = ui.ui_factory.nested_progress_bar()
3294
symbol_versioning.warn(
3295
symbol_versioning.deprecated_in((1, 14, 0))
3296
% "pb parameter to fetch()")
3299
self._fetch_all_revisions(revision_ids, pb)
3301
if my_pb is not None:
3303
return len(revision_ids), 0
3305
def _get_basis(self, first_revision_id):
3306
"""Get a revision and tree which exists in the target.
3308
This assumes that first_revision_id is selected for transmission
3309
because all other ancestors are already present. If we can't find an
3310
ancestor we fall back to NULL_REVISION since we know that is safe.
3312
:return: (basis_id, basis_tree)
3314
first_rev = self.source.get_revision(first_revision_id)
3316
basis_id = first_rev.parent_ids[0]
3317
# only valid as a basis if the target has it
3318
self.target.get_revision(basis_id)
3319
# Try to get a basis tree - if its a ghost it will hit the
3320
# NoSuchRevision case.
3321
basis_tree = self.source.revision_tree(basis_id)
3322
except (IndexError, errors.NoSuchRevision):
3323
basis_id = _mod_revision.NULL_REVISION
3324
basis_tree = self.source.revision_tree(basis_id)
3325
return basis_id, basis_tree
3328
InterRepository.register_optimiser(InterDifferingSerializer)
3329
InterRepository.register_optimiser(InterSameDataRepository)
3330
InterRepository.register_optimiser(InterWeaveRepo)
3331
InterRepository.register_optimiser(InterKnitRepo)
3332
InterRepository.register_optimiser(InterPackRepo)
3335
class CopyConverter(object):
3336
"""A repository conversion tool which just performs a copy of the content.
3338
This is slow but quite reliable.
3341
def __init__(self, target_format):
3342
"""Create a CopyConverter.
3344
:param target_format: The format the resulting repository should be.
3346
self.target_format = target_format
3348
def convert(self, repo, pb):
3349
"""Perform the conversion of to_convert, giving feedback via pb.
3351
:param to_convert: The disk object to convert.
3352
:param pb: a progress bar to use for progress information.
3357
# this is only useful with metadir layouts - separated repo content.
3358
# trigger an assertion if not such
3359
repo._format.get_format_string()
3360
self.repo_dir = repo.bzrdir
3361
self.step('Moving repository to repository.backup')
3362
self.repo_dir.transport.move('repository', 'repository.backup')
3363
backup_transport = self.repo_dir.transport.clone('repository.backup')
3364
repo._format.check_conversion_target(self.target_format)
3365
self.source_repo = repo._format.open(self.repo_dir,
3367
_override_transport=backup_transport)
3368
self.step('Creating new repository')
3369
converted = self.target_format.initialize(self.repo_dir,
3370
self.source_repo.is_shared())
3371
converted.lock_write()
3373
self.step('Copying content into repository.')
3374
self.source_repo.copy_content_into(converted)
3377
self.step('Deleting old repository content.')
3378
self.repo_dir.transport.delete_tree('repository.backup')
3379
self.pb.note('repository converted')
3381
def step(self, message):
3382
"""Update the pb by a step."""
3384
self.pb.update(message, self.count, self.total)
3396
def _unescaper(match, _map=_unescape_map):
3397
code = match.group(1)
3401
if not code.startswith('#'):
3403
return unichr(int(code[1:])).encode('utf8')
3409
def _unescape_xml(data):
3410
"""Unescape predefined XML entities in a string of data."""
3412
if _unescape_re is None:
3413
_unescape_re = re.compile('\&([^;]*);')
3414
return _unescape_re.sub(_unescaper, data)
3417
class _VersionedFileChecker(object):
3419
def __init__(self, repository, text_key_references=None):
3420
self.repository = repository
3421
self.text_index = self.repository._generate_text_key_index(
3422
text_key_references=text_key_references)
3424
def calculate_file_version_parents(self, text_key):
3425
"""Calculate the correct parents for a file version according to
3428
parent_keys = self.text_index[text_key]
3429
if parent_keys == [_mod_revision.NULL_REVISION]:
3431
return tuple(parent_keys)
3433
def check_file_version_parents(self, texts, progress_bar=None):
3434
"""Check the parents stored in a versioned file are correct.
3436
It also detects file versions that are not referenced by their
3437
corresponding revision's inventory.
3439
:returns: A tuple of (wrong_parents, dangling_file_versions).
3440
wrong_parents is a dict mapping {revision_id: (stored_parents,
3441
correct_parents)} for each revision_id where the stored parents
3442
are not correct. dangling_file_versions is a set of (file_id,
3443
revision_id) tuples for versions that are present in this versioned
3444
file, but not used by the corresponding inventory.
3447
self.file_ids = set([file_id for file_id, _ in
3448
self.text_index.iterkeys()])
3449
# text keys is now grouped by file_id
3450
n_weaves = len(self.file_ids)
3451
files_in_revisions = {}
3452
revisions_of_files = {}
3453
n_versions = len(self.text_index)
3454
progress_bar.update('loading text store', 0, n_versions)
3455
parent_map = self.repository.texts.get_parent_map(self.text_index)
3456
# On unlistable transports this could well be empty/error...
3457
text_keys = self.repository.texts.keys()
3458
unused_keys = frozenset(text_keys) - set(self.text_index)
3459
for num, key in enumerate(self.text_index.iterkeys()):
3460
if progress_bar is not None:
3461
progress_bar.update('checking text graph', num, n_versions)
3462
correct_parents = self.calculate_file_version_parents(key)
3464
knit_parents = parent_map[key]
3465
except errors.RevisionNotPresent:
3468
if correct_parents != knit_parents:
3469
wrong_parents[key] = (knit_parents, correct_parents)
3470
return wrong_parents, unused_keys
3473
def _old_get_graph(repository, revision_id):
3474
"""DO NOT USE. That is all. I'm serious."""
3475
graph = repository.get_graph()
3476
revision_graph = dict(((key, value) for key, value in
3477
graph.iter_ancestry([revision_id]) if value is not None))
3478
return _strip_NULL_ghosts(revision_graph)
3481
def _strip_NULL_ghosts(revision_graph):
3482
"""Also don't use this. more compatibility code for unmigrated clients."""
3483
# Filter ghosts, and null:
3484
if _mod_revision.NULL_REVISION in revision_graph:
3485
del revision_graph[_mod_revision.NULL_REVISION]
3486
for key, parents in revision_graph.items():
3487
revision_graph[key] = tuple(parent for parent in parents if parent
3489
return revision_graph
3492
class StreamSink(object):
3493
"""An object that can insert a stream into a repository.
3495
This interface handles the complexity of reserialising inventories and
3496
revisions from different formats, and allows unidirectional insertion into
3497
stacked repositories without looking for the missing basis parents
3501
def __init__(self, target_repo):
3502
self.target_repo = target_repo
3504
def insert_stream(self, stream, src_format, resume_tokens):
3505
"""Insert a stream's content into the target repository.
3507
:param src_format: a bzr repository format.
3509
:return: a list of resume tokens and an iterable of keys additional
3510
items required before the insertion can be completed.
3512
self.target_repo.lock_write()
3515
self.target_repo.resume_write_group(resume_tokens)
3517
self.target_repo.start_write_group()
3519
# locked_insert_stream performs a commit|suspend.
3520
return self._locked_insert_stream(stream, src_format)
3522
self.target_repo.abort_write_group(suppress_errors=True)
3525
self.target_repo.unlock()
3527
def _locked_insert_stream(self, stream, src_format):
3528
to_serializer = self.target_repo._format._serializer
3529
src_serializer = src_format._serializer
3530
for substream_type, substream in stream:
3531
if substream_type == 'texts':
3532
self.target_repo.texts.insert_record_stream(substream)
3533
elif substream_type == 'inventories':
3534
if src_serializer == to_serializer:
3535
self.target_repo.inventories.insert_record_stream(
3538
self._extract_and_insert_inventories(
3539
substream, src_serializer)
3540
elif substream_type == 'revisions':
3541
# This may fallback to extract-and-insert more often than
3542
# required if the serializers are different only in terms of
3544
if src_serializer == to_serializer:
3545
self.target_repo.revisions.insert_record_stream(
3548
self._extract_and_insert_revisions(substream,
3550
elif substream_type == 'signatures':
3551
self.target_repo.signatures.insert_record_stream(substream)
3553
raise AssertionError('kaboom! %s' % (substream_type,))
3555
missing_keys = set()
3556
for prefix, versioned_file in (
3557
('texts', self.target_repo.texts),
3558
('inventories', self.target_repo.inventories),
3559
('revisions', self.target_repo.revisions),
3560
('signatures', self.target_repo.signatures),
3562
missing_keys.update((prefix,) + key for key in
3563
versioned_file.get_missing_compression_parent_keys())
3564
except NotImplementedError:
3565
# cannot even attempt suspending, and missing would have failed
3566
# during stream insertion.
3567
missing_keys = set()
3570
# suspend the write group and tell the caller what we is
3571
# missing. We know we can suspend or else we would not have
3572
# entered this code path. (All repositories that can handle
3573
# missing keys can handle suspending a write group).
3574
write_group_tokens = self.target_repo.suspend_write_group()
3575
return write_group_tokens, missing_keys
3576
self.target_repo.commit_write_group()
3579
def _extract_and_insert_inventories(self, substream, serializer):
3580
"""Generate a new inventory versionedfile in target, converting data.
3582
The inventory is retrieved from the source, (deserializing it), and
3583
stored in the target (reserializing it in a different format).
3585
for record in substream:
3586
bytes = record.get_bytes_as('fulltext')
3587
revision_id = record.key[0]
3588
inv = serializer.read_inventory_from_string(bytes, revision_id)
3589
parents = [key[0] for key in record.parents]
3590
self.target_repo.add_inventory(revision_id, inv, parents)
3592
def _extract_and_insert_revisions(self, substream, serializer):
3593
for record in substream:
3594
bytes = record.get_bytes_as('fulltext')
3595
revision_id = record.key[0]
3596
rev = serializer.read_revision_from_string(bytes)
3597
if rev.revision_id != revision_id:
3598
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
3599
self.target_repo.add_revision(revision_id, rev)
3602
if self.target_repo._format._fetch_reconcile:
3603
self.target_repo.reconcile()
3606
class StreamSource(object):
3607
"""A source of a stream for fetching between repositories."""
3609
def __init__(self, from_repository, to_format):
3610
"""Create a StreamSource streaming from from_repository."""
3611
self.from_repository = from_repository
3612
self.to_format = to_format
3614
def delta_on_metadata(self):
3615
"""Return True if delta's are permitted on metadata streams.
3617
That is on revisions and signatures.
3619
src_serializer = self.from_repository._format._serializer
3620
target_serializer = self.to_format._serializer
3621
return (self.to_format._fetch_uses_deltas and
3622
src_serializer == target_serializer)
3624
def _fetch_revision_texts(self, revs):
3625
# fetch signatures first and then the revision texts
3626
# may need to be a InterRevisionStore call here.
3627
from_sf = self.from_repository.signatures
3628
# A missing signature is just skipped.
3629
keys = [(rev_id,) for rev_id in revs]
3630
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
3632
self.to_format._fetch_order,
3633
not self.to_format._fetch_uses_deltas))
3634
# If a revision has a delta, this is actually expanded inside the
3635
# insert_record_stream code now, which is an alternate fix for
3637
from_rf = self.from_repository.revisions
3638
revisions = from_rf.get_record_stream(
3640
self.to_format._fetch_order,
3641
not self.delta_on_metadata())
3642
return [('signatures', signatures), ('revisions', revisions)]
3644
def _generate_root_texts(self, revs):
3645
"""This will be called by __fetch between fetching weave texts and
3646
fetching the inventory weave.
3648
Subclasses should override this if they need to generate root texts
3649
after fetching weave texts.
3651
if self._rich_root_upgrade():
3653
return bzrlib.fetch.Inter1and2Helper(
3654
self.from_repository).generate_root_texts(revs)
3658
def get_stream(self, search):
3660
revs = search.get_keys()
3661
graph = self.from_repository.get_graph()
3662
revs = list(graph.iter_topo_order(revs))
3663
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
3665
for knit_kind, file_id, revisions in data_to_fetch:
3666
if knit_kind != phase:
3668
# Make a new progress bar for this phase
3669
if knit_kind == "file":
3670
# Accumulate file texts
3671
text_keys.extend([(file_id, revision) for revision in
3673
elif knit_kind == "inventory":
3674
# Now copy the file texts.
3675
from_texts = self.from_repository.texts
3676
yield ('texts', from_texts.get_record_stream(
3677
text_keys, self.to_format._fetch_order,
3678
not self.to_format._fetch_uses_deltas))
3679
# Cause an error if a text occurs after we have done the
3682
# Before we process the inventory we generate the root
3683
# texts (if necessary) so that the inventories references
3685
for _ in self._generate_root_texts(revs):
3687
# NB: This currently reopens the inventory weave in source;
3688
# using a single stream interface instead would avoid this.
3689
from_weave = self.from_repository.inventories
3690
# we fetch only the referenced inventories because we do not
3691
# know for unselected inventories whether all their required
3692
# texts are present in the other repository - it could be
3694
yield ('inventories', from_weave.get_record_stream(
3695
[(rev_id,) for rev_id in revs],
3696
self.inventory_fetch_order(),
3697
not self.delta_on_metadata()))
3698
elif knit_kind == "signatures":
3699
# Nothing to do here; this will be taken care of when
3700
# _fetch_revision_texts happens.
3702
elif knit_kind == "revisions":
3703
for record in self._fetch_revision_texts(revs):
3706
raise AssertionError("Unknown knit kind %r" % knit_kind)
3708
def get_stream_for_missing_keys(self, missing_keys):
3709
# missing keys can only occur when we are byte copying and not
3710
# translating (because translation means we don't send
3711
# unreconstructable deltas ever).
3713
keys['texts'] = set()
3714
keys['revisions'] = set()
3715
keys['inventories'] = set()
3716
keys['signatures'] = set()
3717
for key in missing_keys:
3718
keys[key[0]].add(key[1:])
3719
if len(keys['revisions']):
3720
# If we allowed copying revisions at this point, we could end up
3721
# copying a revision without copying its required texts: a
3722
# violation of the requirements for repository integrity.
3723
raise AssertionError(
3724
'cannot copy revisions to fill in missing deltas %s' % (
3725
keys['revisions'],))
3726
for substream_kind, keys in keys.iteritems():
3727
vf = getattr(self.from_repository, substream_kind)
3728
# Ask for full texts always so that we don't need more round trips
3729
# after this stream.
3730
stream = vf.get_record_stream(keys,
3731
self.to_format._fetch_order, True)
3732
yield substream_kind, stream
3734
def inventory_fetch_order(self):
3735
if self._rich_root_upgrade():
3736
return 'topological'
3738
return self.to_format._fetch_order
3740
def _rich_root_upgrade(self):
3741
return (not self.from_repository._format.rich_root_data and
3742
self.to_format.rich_root_data)
added
removed
bzrlib/repository.py
