1
# Copyright (C) 2005, 2006, 2007, 2008 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from bzrlib.lazy_import import lazy_import
18
lazy_import(globals(), """
38
revision as _mod_revision,
44
from bzrlib.bundle import serializer
45
from bzrlib.revisiontree import RevisionTree
46
from bzrlib.store.versioned import VersionedFileStore
47
from bzrlib.testament import Testament
50
from bzrlib.decorators import needs_read_lock, needs_write_lock
51
from bzrlib.inter import InterObject
52
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
53
from bzrlib import registry
54
from bzrlib.symbol_versioning import (
60
from bzrlib.trace import (
61
log_exception_quietly, note, mutter, mutter_callsite, warning)
64
# Old formats display a warning, but only once
65
_deprecation_warning_done = False
68
class CommitBuilder(object):
69
"""Provides an interface to build up a commit.
71
This allows describing a tree to be committed without needing to
72
know the internals of the format of the repository.
75
# all clients should supply tree roots.
76
record_root_entry = True
77
# the default CommitBuilder does not manage trees whose root is versioned.
78
_versioned_root = False
80
def __init__(self, repository, parents, config, timestamp=None,
81
timezone=None, committer=None, revprops=None,
83
"""Initiate a CommitBuilder.
85
:param repository: Repository to commit to.
86
:param parents: Revision ids of the parents of the new revision.
87
:param config: Configuration to use.
88
:param timestamp: Optional timestamp recorded for commit.
89
:param timezone: Optional timezone for timestamp.
90
:param committer: Optional committer to set for commit.
91
:param revprops: Optional dictionary of revision properties.
92
:param revision_id: Optional revision id.
97
self._committer = self._config.username()
99
self._committer = committer
101
self.new_inventory = Inventory(None)
102
self._new_revision_id = revision_id
103
self.parents = parents
104
self.repository = repository
107
if revprops is not None:
108
self._validate_revprops(revprops)
109
self._revprops.update(revprops)
111
if timestamp is None:
112
timestamp = time.time()
113
# Restrict resolution to 1ms
114
self._timestamp = round(timestamp, 3)
117
self._timezone = osutils.local_time_offset()
119
self._timezone = int(timezone)
121
self._generate_revision_if_needed()
122
self.__heads = graph.HeadsCache(repository.get_graph()).heads
123
self._basis_delta = []
124
# API compatibility, older code that used CommitBuilder did not call
125
# .record_delete(), which means the delta that is computed would not be
126
# valid. Callers that will call record_delete() should call
127
# .will_record_deletes() to indicate that.
128
self._recording_deletes = False
130
def _validate_unicode_text(self, text, context):
131
"""Verify things like commit messages don't have bogus characters."""
133
raise ValueError('Invalid value for %s: %r' % (context, text))
135
def _validate_revprops(self, revprops):
136
for key, value in revprops.iteritems():
137
# We know that the XML serializers do not round trip '\r'
138
# correctly, so refuse to accept them
139
if not isinstance(value, basestring):
140
raise ValueError('revision property (%s) is not a valid'
141
' (unicode) string: %r' % (key, value))
142
self._validate_unicode_text(value,
143
'revision property (%s)' % (key,))
145
def commit(self, message):
146
"""Make the actual commit.
148
:return: The revision id of the recorded revision.
150
self._validate_unicode_text(message, 'commit message')
151
rev = _mod_revision.Revision(
152
timestamp=self._timestamp,
153
timezone=self._timezone,
154
committer=self._committer,
156
inventory_sha1=self.inv_sha1,
157
revision_id=self._new_revision_id,
158
properties=self._revprops)
159
rev.parent_ids = self.parents
160
self.repository.add_revision(self._new_revision_id, rev,
161
self.new_inventory, self._config)
162
self.repository.commit_write_group()
163
return self._new_revision_id
166
"""Abort the commit that is being built.
168
self.repository.abort_write_group()
170
def revision_tree(self):
171
"""Return the tree that was just committed.
173
After calling commit() this can be called to get a RevisionTree
174
representing the newly committed tree. This is preferred to
175
calling Repository.revision_tree() because that may require
176
deserializing the inventory, while we already have a copy in
179
return RevisionTree(self.repository, self.new_inventory,
180
self._new_revision_id)
182
def finish_inventory(self):
183
"""Tell the builder that the inventory is finished."""
184
if self.new_inventory.root is None:
185
raise AssertionError('Root entry should be supplied to'
186
' record_entry_contents, as of bzr 0.10.')
187
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
188
self.new_inventory.revision_id = self._new_revision_id
189
self.inv_sha1 = self.repository.add_inventory(
190
self._new_revision_id,
195
def _gen_revision_id(self):
196
"""Return new revision-id."""
197
return generate_ids.gen_revision_id(self._config.username(),
200
def _generate_revision_if_needed(self):
201
"""Create a revision id if None was supplied.
203
If the repository can not support user-specified revision ids
204
they should override this function and raise CannotSetRevisionId
205
if _new_revision_id is not None.
207
:raises: CannotSetRevisionId
209
if self._new_revision_id is None:
210
self._new_revision_id = self._gen_revision_id()
211
self.random_revid = True
213
self.random_revid = False
215
def _heads(self, file_id, revision_ids):
216
"""Calculate the graph heads for revision_ids in the graph of file_id.
218
This can use either a per-file graph or a global revision graph as we
219
have an identity relationship between the two graphs.
221
return self.__heads(revision_ids)
223
def _check_root(self, ie, parent_invs, tree):
224
"""Helper for record_entry_contents.
226
:param ie: An entry being added.
227
:param parent_invs: The inventories of the parent revisions of the
229
:param tree: The tree that is being committed.
231
# In this revision format, root entries have no knit or weave When
232
# serializing out to disk and back in root.revision is always
234
ie.revision = self._new_revision_id
236
def _get_delta(self, ie, basis_inv, path):
237
"""Get a delta against the basis inventory for ie."""
238
if ie.file_id not in basis_inv:
240
result = (None, path, ie.file_id, ie)
241
self._basis_delta.append(result)
243
elif ie != basis_inv[ie.file_id]:
245
# TODO: avoid tis id2path call.
246
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
247
self._basis_delta.append(result)
253
def get_basis_delta(self):
254
"""Return the complete inventory delta versus the basis inventory.
256
This has been built up with the calls to record_delete and
257
record_entry_contents. The client must have already called
258
will_record_deletes() to indicate that they will be generating a
261
:return: An inventory delta, suitable for use with apply_delta, or
262
Repository.add_inventory_by_delta, etc.
264
if not self._recording_deletes:
265
raise AssertionError("recording deletes not activated.")
266
return self._basis_delta
268
def record_delete(self, path, file_id):
269
"""Record that a delete occured against a basis tree.
271
This is an optional API - when used it adds items to the basis_delta
272
being accumulated by the commit builder. It cannot be called unless the
273
method will_record_deletes() has been called to inform the builder that
274
a delta is being supplied.
276
:param path: The path of the thing deleted.
277
:param file_id: The file id that was deleted.
279
if not self._recording_deletes:
280
raise AssertionError("recording deletes not activated.")
281
delta = (path, None, file_id, None)
282
self._basis_delta.append(delta)
285
def will_record_deletes(self):
286
"""Tell the commit builder that deletes are being notified.
288
This enables the accumulation of an inventory delta; for the resulting
289
commit to be valid, deletes against the basis MUST be recorded via
290
builder.record_delete().
292
self._recording_deletes = True
294
def record_entry_contents(self, ie, parent_invs, path, tree,
296
"""Record the content of ie from tree into the commit if needed.
298
Side effect: sets ie.revision when unchanged
300
:param ie: An inventory entry present in the commit.
301
:param parent_invs: The inventories of the parent revisions of the
303
:param path: The path the entry is at in the tree.
304
:param tree: The tree which contains this entry and should be used to
306
:param content_summary: Summary data from the tree about the paths
307
content - stat, length, exec, sha/link target. This is only
308
accessed when the entry has a revision of None - that is when it is
309
a candidate to commit.
310
:return: A tuple (change_delta, version_recorded, fs_hash).
311
change_delta is an inventory_delta change for this entry against
312
the basis tree of the commit, or None if no change occured against
314
version_recorded is True if a new version of the entry has been
315
recorded. For instance, committing a merge where a file was only
316
changed on the other side will return (delta, False).
317
fs_hash is either None, or the hash details for the path (currently
318
a tuple of the contents sha1 and the statvalue returned by
319
tree.get_file_with_stat()).
321
if self.new_inventory.root is None:
322
if ie.parent_id is not None:
323
raise errors.RootMissing()
324
self._check_root(ie, parent_invs, tree)
325
if ie.revision is None:
326
kind = content_summary[0]
328
# ie is carried over from a prior commit
330
# XXX: repository specific check for nested tree support goes here - if
331
# the repo doesn't want nested trees we skip it ?
332
if (kind == 'tree-reference' and
333
not self.repository._format.supports_tree_reference):
334
# mismatch between commit builder logic and repository:
335
# this needs the entry creation pushed down into the builder.
336
raise NotImplementedError('Missing repository subtree support.')
337
self.new_inventory.add(ie)
339
# TODO: slow, take it out of the inner loop.
341
basis_inv = parent_invs[0]
343
basis_inv = Inventory(root_id=None)
345
# ie.revision is always None if the InventoryEntry is considered
346
# for committing. We may record the previous parents revision if the
347
# content is actually unchanged against a sole head.
348
if ie.revision is not None:
349
if not self._versioned_root and path == '':
350
# repositories that do not version the root set the root's
351
# revision to the new commit even when no change occurs (more
352
# specifically, they do not record a revision on the root; and
353
# the rev id is assigned to the root during deserialisation -
354
# this masks when a change may have occurred against the basis.
355
# To match this we always issue a delta, because the revision
356
# of the root will always be changing.
357
if ie.file_id in basis_inv:
358
delta = (basis_inv.id2path(ie.file_id), path,
362
delta = (None, path, ie.file_id, ie)
363
self._basis_delta.append(delta)
364
return delta, False, None
366
# we don't need to commit this, because the caller already
367
# determined that an existing revision of this file is
368
# appropriate. If its not being considered for committing then
369
# it and all its parents to the root must be unaltered so
370
# no-change against the basis.
371
if ie.revision == self._new_revision_id:
372
raise AssertionError("Impossible situation, a skipped "
373
"inventory entry (%r) claims to be modified in this "
374
"commit (%r).", (ie, self._new_revision_id))
375
return None, False, None
376
# XXX: Friction: parent_candidates should return a list not a dict
377
# so that we don't have to walk the inventories again.
378
parent_candiate_entries = ie.parent_candidates(parent_invs)
379
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
381
for inv in parent_invs:
382
if ie.file_id in inv:
383
old_rev = inv[ie.file_id].revision
384
if old_rev in head_set:
385
heads.append(inv[ie.file_id].revision)
386
head_set.remove(inv[ie.file_id].revision)
389
# now we check to see if we need to write a new record to the
391
# We write a new entry unless there is one head to the ancestors, and
392
# the kind-derived content is unchanged.
394
# Cheapest check first: no ancestors, or more the one head in the
395
# ancestors, we write a new node.
399
# There is a single head, look it up for comparison
400
parent_entry = parent_candiate_entries[heads[0]]
401
# if the non-content specific data has changed, we'll be writing a
403
if (parent_entry.parent_id != ie.parent_id or
404
parent_entry.name != ie.name):
406
# now we need to do content specific checks:
408
# if the kind changed the content obviously has
409
if kind != parent_entry.kind:
411
# Stat cache fingerprint feedback for the caller - None as we usually
412
# don't generate one.
415
if content_summary[2] is None:
416
raise ValueError("Files must not have executable = None")
418
if (# if the file length changed we have to store:
419
parent_entry.text_size != content_summary[1] or
420
# if the exec bit has changed we have to store:
421
parent_entry.executable != content_summary[2]):
423
elif parent_entry.text_sha1 == content_summary[3]:
424
# all meta and content is unchanged (using a hash cache
425
# hit to check the sha)
426
ie.revision = parent_entry.revision
427
ie.text_size = parent_entry.text_size
428
ie.text_sha1 = parent_entry.text_sha1
429
ie.executable = parent_entry.executable
430
return self._get_delta(ie, basis_inv, path), False, None
432
# Either there is only a hash change(no hash cache entry,
433
# or same size content change), or there is no change on
435
# Provide the parent's hash to the store layer, so that the
436
# content is unchanged we will not store a new node.
437
nostore_sha = parent_entry.text_sha1
439
# We want to record a new node regardless of the presence or
440
# absence of a content change in the file.
442
ie.executable = content_summary[2]
443
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
445
lines = file_obj.readlines()
449
ie.text_sha1, ie.text_size = self._add_text_to_weave(
450
ie.file_id, lines, heads, nostore_sha)
451
# Let the caller know we generated a stat fingerprint.
452
fingerprint = (ie.text_sha1, stat_value)
453
except errors.ExistingContent:
454
# Turns out that the file content was unchanged, and we were
455
# only going to store a new node if it was changed. Carry over
457
ie.revision = parent_entry.revision
458
ie.text_size = parent_entry.text_size
459
ie.text_sha1 = parent_entry.text_sha1
460
ie.executable = parent_entry.executable
461
return self._get_delta(ie, basis_inv, path), False, None
462
elif kind == 'directory':
464
# all data is meta here, nothing specific to directory, so
466
ie.revision = parent_entry.revision
467
return self._get_delta(ie, basis_inv, path), False, None
469
self._add_text_to_weave(ie.file_id, lines, heads, None)
470
elif kind == 'symlink':
471
current_link_target = content_summary[3]
473
# symlink target is not generic metadata, check if it has
475
if current_link_target != parent_entry.symlink_target:
478
# unchanged, carry over.
479
ie.revision = parent_entry.revision
480
ie.symlink_target = parent_entry.symlink_target
481
return self._get_delta(ie, basis_inv, path), False, None
482
ie.symlink_target = current_link_target
484
self._add_text_to_weave(ie.file_id, lines, heads, None)
485
elif kind == 'tree-reference':
487
if content_summary[3] != parent_entry.reference_revision:
490
# unchanged, carry over.
491
ie.reference_revision = parent_entry.reference_revision
492
ie.revision = parent_entry.revision
493
return self._get_delta(ie, basis_inv, path), False, None
494
ie.reference_revision = content_summary[3]
496
self._add_text_to_weave(ie.file_id, lines, heads, None)
498
raise NotImplementedError('unknown kind')
499
ie.revision = self._new_revision_id
500
return self._get_delta(ie, basis_inv, path), True, fingerprint
502
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
503
# Note: as we read the content directly from the tree, we know its not
504
# been turned into unicode or badly split - but a broken tree
505
# implementation could give us bad output from readlines() so this is
506
# not a guarantee of safety. What would be better is always checking
507
# the content during test suite execution. RBC 20070912
508
parent_keys = tuple((file_id, parent) for parent in parents)
509
return self.repository.texts.add_lines(
510
(file_id, self._new_revision_id), parent_keys, new_lines,
511
nostore_sha=nostore_sha, random_id=self.random_revid,
512
check_content=False)[0:2]
515
class RootCommitBuilder(CommitBuilder):
516
"""This commitbuilder actually records the root id"""
518
# the root entry gets versioned properly by this builder.
519
_versioned_root = True
521
def _check_root(self, ie, parent_invs, tree):
522
"""Helper for record_entry_contents.
524
:param ie: An entry being added.
525
:param parent_invs: The inventories of the parent revisions of the
527
:param tree: The tree that is being committed.
531
######################################################################
534
class Repository(object):
535
"""Repository holding history for one or more branches.
537
The repository holds and retrieves historical information including
538
revisions and file history. It's normally accessed only by the Branch,
539
which views a particular line of development through that history.
541
The Repository builds on top of some byte storage facilies (the revisions,
542
signatures, inventories and texts attributes) and a Transport, which
543
respectively provide byte storage and a means to access the (possibly
546
The byte storage facilities are addressed via tuples, which we refer to
547
as 'keys' throughout the code base. Revision_keys, inventory_keys and
548
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
549
(file_id, revision_id). We use this interface because it allows low
550
friction with the underlying code that implements disk indices, network
551
encoding and other parts of bzrlib.
553
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
554
the serialised revisions for the repository. This can be used to obtain
555
revision graph information or to access raw serialised revisions.
556
The result of trying to insert data into the repository via this store
557
is undefined: it should be considered read-only except for implementors
559
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
560
the serialised signatures for the repository. This can be used to
561
obtain access to raw serialised signatures. The result of trying to
562
insert data into the repository via this store is undefined: it should
563
be considered read-only except for implementors of repositories.
564
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
565
the serialised inventories for the repository. This can be used to
566
obtain unserialised inventories. The result of trying to insert data
567
into the repository via this store is undefined: it should be
568
considered read-only except for implementors of repositories.
569
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
570
texts of files and directories for the repository. This can be used to
571
obtain file texts or file graphs. Note that Repository.iter_file_bytes
572
is usually a better interface for accessing file texts.
573
The result of trying to insert data into the repository via this store
574
is undefined: it should be considered read-only except for implementors
576
:ivar _transport: Transport for file access to repository, typically
577
pointing to .bzr/repository.
580
# What class to use for a CommitBuilder. Often its simpler to change this
581
# in a Repository class subclass rather than to override
582
# get_commit_builder.
583
_commit_builder_class = CommitBuilder
584
# The search regex used by xml based repositories to determine what things
585
# where changed in a single commit.
586
_file_ids_altered_regex = lazy_regex.lazy_compile(
587
r'file_id="(?P<file_id>[^"]+)"'
588
r'.* revision="(?P<revision_id>[^"]+)"'
591
def abort_write_group(self, suppress_errors=False):
592
"""Commit the contents accrued within the current write group.
594
:param suppress_errors: if true, abort_write_group will catch and log
595
unexpected errors that happen during the abort, rather than
596
allowing them to propagate. Defaults to False.
598
:seealso: start_write_group.
600
if self._write_group is not self.get_transaction():
601
# has an unlock or relock occured ?
602
raise errors.BzrError('mismatched lock context and write group.')
604
self._abort_write_group()
605
except Exception, exc:
606
self._write_group = None
607
if not suppress_errors:
609
mutter('abort_write_group failed')
610
log_exception_quietly()
611
note('bzr: ERROR (ignored): %s', exc)
612
self._write_group = None
614
def _abort_write_group(self):
615
"""Template method for per-repository write group cleanup.
617
This is called during abort before the write group is considered to be
618
finished and should cleanup any internal state accrued during the write
619
group. There is no requirement that data handed to the repository be
620
*not* made available - this is not a rollback - but neither should any
621
attempt be made to ensure that data added is fully commited. Abort is
622
invoked when an error has occured so futher disk or network operations
623
may not be possible or may error and if possible should not be
627
def add_fallback_repository(self, repository):
628
"""Add a repository to use for looking up data not held locally.
630
:param repository: A repository.
632
if not self._format.supports_external_lookups:
633
raise errors.UnstackableRepositoryFormat(self._format, self.base)
634
self._check_fallback_repository(repository)
635
self._fallback_repositories.append(repository)
636
self.texts.add_fallback_versioned_files(repository.texts)
637
self.inventories.add_fallback_versioned_files(repository.inventories)
638
self.revisions.add_fallback_versioned_files(repository.revisions)
639
self.signatures.add_fallback_versioned_files(repository.signatures)
641
def _check_fallback_repository(self, repository):
642
"""Check that this repository can fallback to repository safely.
644
Raise an error if not.
646
:param repository: A repository to fallback to.
648
return InterRepository._assert_same_model(self, repository)
650
def add_inventory(self, revision_id, inv, parents):
651
"""Add the inventory inv to the repository as revision_id.
653
:param parents: The revision ids of the parents that revision_id
654
is known to have and are in the repository already.
656
:returns: The validator(which is a sha1 digest, though what is sha'd is
657
repository format specific) of the serialized inventory.
659
if not self.is_in_write_group():
660
raise AssertionError("%r not in write group" % (self,))
661
_mod_revision.check_not_reserved_id(revision_id)
662
if not (inv.revision_id is None or inv.revision_id == revision_id):
663
raise AssertionError(
664
"Mismatch between inventory revision"
665
" id and insertion revid (%r, %r)"
666
% (inv.revision_id, revision_id))
668
raise AssertionError()
669
inv_lines = self._serialise_inventory_to_lines(inv)
670
return self._inventory_add_lines(revision_id, parents,
671
inv_lines, check_content=False)
673
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
675
"""Add a new inventory expressed as a delta against another revision.
677
:param basis_revision_id: The inventory id the delta was created
678
against. (This does not have to be a direct parent.)
679
:param delta: The inventory delta (see Inventory.apply_delta for
681
:param new_revision_id: The revision id that the inventory is being
683
:param parents: The revision ids of the parents that revision_id is
684
known to have and are in the repository already. These are supplied
685
for repositories that depend on the inventory graph for revision
686
graph access, as well as for those that pun ancestry with delta
689
:returns: (validator, new_inv)
690
The validator(which is a sha1 digest, though what is sha'd is
691
repository format specific) of the serialized inventory, and the
694
if not self.is_in_write_group():
695
raise AssertionError("%r not in write group" % (self,))
696
_mod_revision.check_not_reserved_id(new_revision_id)
697
basis_tree = self.revision_tree(basis_revision_id)
698
basis_tree.lock_read()
700
# Note that this mutates the inventory of basis_tree, which not all
701
# inventory implementations may support: A better idiom would be to
702
# return a new inventory, but as there is no revision tree cache in
703
# repository this is safe for now - RBC 20081013
704
basis_inv = basis_tree.inventory
705
basis_inv.apply_delta(delta)
706
basis_inv.revision_id = new_revision_id
707
return (self.add_inventory(new_revision_id, basis_inv, parents),
712
def _inventory_add_lines(self, revision_id, parents, lines,
714
"""Store lines in inv_vf and return the sha1 of the inventory."""
715
parents = [(parent,) for parent in parents]
716
return self.inventories.add_lines((revision_id,), parents, lines,
717
check_content=check_content)[0]
719
def add_revision(self, revision_id, rev, inv=None, config=None):
720
"""Add rev to the revision store as revision_id.
722
:param revision_id: the revision id to use.
723
:param rev: The revision object.
724
:param inv: The inventory for the revision. if None, it will be looked
725
up in the inventory storer
726
:param config: If None no digital signature will be created.
727
If supplied its signature_needed method will be used
728
to determine if a signature should be made.
730
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
732
_mod_revision.check_not_reserved_id(revision_id)
733
if config is not None and config.signature_needed():
735
inv = self.get_inventory(revision_id)
736
plaintext = Testament(rev, inv).as_short_text()
737
self.store_revision_signature(
738
gpg.GPGStrategy(config), plaintext, revision_id)
739
# check inventory present
740
if not self.inventories.get_parent_map([(revision_id,)]):
742
raise errors.WeaveRevisionNotPresent(revision_id,
745
# yes, this is not suitable for adding with ghosts.
746
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
750
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
751
self._add_revision(rev)
753
def _add_revision(self, revision):
754
text = self._serializer.write_revision_to_string(revision)
755
key = (revision.revision_id,)
756
parents = tuple((parent,) for parent in revision.parent_ids)
757
self.revisions.add_lines(key, parents, osutils.split_lines(text))
759
def all_revision_ids(self):
760
"""Returns a list of all the revision ids in the repository.
762
This is conceptually deprecated because code should generally work on
763
the graph reachable from a particular revision, and ignore any other
764
revisions that might be present. There is no direct replacement
767
if 'evil' in debug.debug_flags:
768
mutter_callsite(2, "all_revision_ids is linear with history.")
769
return self._all_revision_ids()
771
def _all_revision_ids(self):
772
"""Returns a list of all the revision ids in the repository.
774
These are in as much topological order as the underlying store can
777
raise NotImplementedError(self._all_revision_ids)
779
def break_lock(self):
780
"""Break a lock if one is present from another instance.
782
Uses the ui factory to ask for confirmation if the lock may be from
785
self.control_files.break_lock()
788
def _eliminate_revisions_not_present(self, revision_ids):
789
"""Check every revision id in revision_ids to see if we have it.
791
Returns a set of the present revisions.
794
graph = self.get_graph()
795
parent_map = graph.get_parent_map(revision_ids)
796
# The old API returned a list, should this actually be a set?
797
return parent_map.keys()
800
def create(a_bzrdir):
801
"""Construct the current default format repository in a_bzrdir."""
802
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
804
def __init__(self, _format, a_bzrdir, control_files):
805
"""instantiate a Repository.
807
:param _format: The format of the repository on disk.
808
:param a_bzrdir: The BzrDir of the repository.
810
In the future we will have a single api for all stores for
811
getting file texts, inventories and revisions, then
812
this construct will accept instances of those things.
814
super(Repository, self).__init__()
815
self._format = _format
816
# the following are part of the public API for Repository:
817
self.bzrdir = a_bzrdir
818
self.control_files = control_files
819
self._transport = control_files._transport
820
self.base = self._transport.base
822
self._reconcile_does_inventory_gc = True
823
self._reconcile_fixes_text_parents = False
824
self._reconcile_backsup_inventory = True
825
# not right yet - should be more semantically clear ?
827
# TODO: make sure to construct the right store classes, etc, depending
828
# on whether escaping is required.
829
self._warn_if_deprecated()
830
self._write_group = None
831
# Additional places to query for data.
832
self._fallback_repositories = []
833
# What order should fetch operations request streams in?
834
# The default is unordered as that is the cheapest for an origin to
836
self._fetch_order = 'unordered'
837
# Does this repository use deltas that can be fetched as-deltas ?
838
# (E.g. knits, where the knit deltas can be transplanted intact.
839
# We default to False, which will ensure that enough data to get
840
# a full text out of any fetch stream will be grabbed.
841
self._fetch_uses_deltas = False
842
# Should fetch trigger a reconcile after the fetch? Only needed for
843
# some repository formats that can suffer internal inconsistencies.
844
self._fetch_reconcile = False
845
# An InventoryEntry cache, used during deserialization
846
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
849
return '%s(%r)' % (self.__class__.__name__,
852
def has_same_location(self, other):
853
"""Returns a boolean indicating if this repository is at the same
854
location as another repository.
856
This might return False even when two repository objects are accessing
857
the same physical repository via different URLs.
859
if self.__class__ is not other.__class__:
861
return (self._transport.base == other._transport.base)
863
def is_in_write_group(self):
864
"""Return True if there is an open write group.
866
:seealso: start_write_group.
868
return self._write_group is not None
871
return self.control_files.is_locked()
873
def is_write_locked(self):
874
"""Return True if this object is write locked."""
875
return self.is_locked() and self.control_files._lock_mode == 'w'
877
def lock_write(self, token=None):
878
"""Lock this repository for writing.
880
This causes caching within the repository obejct to start accumlating
881
data during reads, and allows a 'write_group' to be obtained. Write
882
groups must be used for actual data insertion.
884
:param token: if this is already locked, then lock_write will fail
885
unless the token matches the existing lock.
886
:returns: a token if this instance supports tokens, otherwise None.
887
:raises TokenLockingNotSupported: when a token is given but this
888
instance doesn't support using token locks.
889
:raises MismatchedToken: if the specified token doesn't match the token
890
of the existing lock.
891
:seealso: start_write_group.
893
A token should be passed in if you know that you have locked the object
894
some other way, and need to synchronise this object's state with that
897
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
899
result = self.control_files.lock_write(token=token)
900
for repo in self._fallback_repositories:
901
# Writes don't affect fallback repos
907
self.control_files.lock_read()
908
for repo in self._fallback_repositories:
912
def get_physical_lock_status(self):
913
return self.control_files.get_physical_lock_status()
915
def leave_lock_in_place(self):
916
"""Tell this repository not to release the physical lock when this
919
If lock_write doesn't return a token, then this method is not supported.
921
self.control_files.leave_in_place()
923
def dont_leave_lock_in_place(self):
924
"""Tell this repository to release the physical lock when this
925
object is unlocked, even if it didn't originally acquire it.
927
If lock_write doesn't return a token, then this method is not supported.
929
self.control_files.dont_leave_in_place()
932
def gather_stats(self, revid=None, committers=None):
933
"""Gather statistics from a revision id.
935
:param revid: The revision id to gather statistics from, if None, then
936
no revision specific statistics are gathered.
937
:param committers: Optional parameter controlling whether to grab
938
a count of committers from the revision specific statistics.
939
:return: A dictionary of statistics. Currently this contains:
940
committers: The number of committers if requested.
941
firstrev: A tuple with timestamp, timezone for the penultimate left
942
most ancestor of revid, if revid is not the NULL_REVISION.
943
latestrev: A tuple with timestamp, timezone for revid, if revid is
944
not the NULL_REVISION.
945
revisions: The total revision count in the repository.
946
size: An estimate disk size of the repository in bytes.
949
if revid and committers:
950
result['committers'] = 0
951
if revid and revid != _mod_revision.NULL_REVISION:
953
all_committers = set()
954
revisions = self.get_ancestry(revid)
955
# pop the leading None
957
first_revision = None
959
# ignore the revisions in the middle - just grab first and last
960
revisions = revisions[0], revisions[-1]
961
for revision in self.get_revisions(revisions):
962
if not first_revision:
963
first_revision = revision
965
all_committers.add(revision.committer)
966
last_revision = revision
968
result['committers'] = len(all_committers)
969
result['firstrev'] = (first_revision.timestamp,
970
first_revision.timezone)
971
result['latestrev'] = (last_revision.timestamp,
972
last_revision.timezone)
974
# now gather global repository information
975
# XXX: This is available for many repos regardless of listability.
976
if self.bzrdir.root_transport.listable():
977
# XXX: do we want to __define len__() ?
978
# Maybe the versionedfiles object should provide a different
979
# method to get the number of keys.
980
result['revisions'] = len(self.revisions.keys())
984
def find_branches(self, using=False):
985
"""Find branches underneath this repository.
987
This will include branches inside other branches.
989
:param using: If True, list only branches using this repository.
991
if using and not self.is_shared():
993
return [self.bzrdir.open_branch()]
994
except errors.NotBranchError:
996
class Evaluator(object):
999
self.first_call = True
1001
def __call__(self, bzrdir):
1002
# On the first call, the parameter is always the bzrdir
1003
# containing the current repo.
1004
if not self.first_call:
1006
repository = bzrdir.open_repository()
1007
except errors.NoRepositoryPresent:
1010
return False, (None, repository)
1011
self.first_call = False
1013
value = (bzrdir.open_branch(), None)
1014
except errors.NotBranchError:
1015
value = (None, None)
1019
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1020
self.bzrdir.root_transport, evaluate=Evaluator()):
1021
if branch is not None:
1022
branches.append(branch)
1023
if not using and repository is not None:
1024
branches.extend(repository.find_branches())
1028
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1029
"""Return the revision ids that other has that this does not.
1031
These are returned in topological order.
1033
revision_id: only return revision ids included by revision_id.
1035
return InterRepository.get(other, self).search_missing_revision_ids(
1036
revision_id, find_ghosts)
1038
@deprecated_method(one_two)
1040
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1041
"""Return the revision ids that other has that this does not.
1043
These are returned in topological order.
1045
revision_id: only return revision ids included by revision_id.
1047
keys = self.search_missing_revision_ids(
1048
other, revision_id, find_ghosts).get_keys()
1051
parents = other.get_graph().get_parent_map(keys)
1054
return tsort.topo_sort(parents)
1058
"""Open the repository rooted at base.
1060
For instance, if the repository is at URL/.bzr/repository,
1061
Repository.open(URL) -> a Repository instance.
1063
control = bzrdir.BzrDir.open(base)
1064
return control.open_repository()
1066
def copy_content_into(self, destination, revision_id=None):
1067
"""Make a complete copy of the content in self into destination.
1069
This is a destructive operation! Do not use it on existing
1072
return InterRepository.get(self, destination).copy_content(revision_id)
1074
def commit_write_group(self):
1075
"""Commit the contents accrued within the current write group.
1077
:seealso: start_write_group.
1079
if self._write_group is not self.get_transaction():
1080
# has an unlock or relock occured ?
1081
raise errors.BzrError('mismatched lock context %r and '
1083
(self.get_transaction(), self._write_group))
1084
self._commit_write_group()
1085
self._write_group = None
1087
def _commit_write_group(self):
1088
"""Template method for per-repository write group cleanup.
1090
This is called before the write group is considered to be
1091
finished and should ensure that all data handed to the repository
1092
for writing during the write group is safely committed (to the
1093
extent possible considering file system caching etc).
1096
def suspend_write_group(self):
1097
raise errors.UnsuspendableWriteGroup(self)
1099
def resume_write_group(self, tokens):
1100
if not self.is_write_locked():
1101
raise errors.NotWriteLocked(self)
1102
if self._write_group:
1103
raise errors.BzrError('already in a write group')
1104
self._resume_write_group(tokens)
1105
# so we can detect unlock/relock - the write group is now entered.
1106
self._write_group = self.get_transaction()
1108
def _resume_write_group(self, tokens):
1109
raise errors.UnsuspendableWriteGroup(self)
1111
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1112
"""Fetch the content required to construct revision_id from source.
1114
If revision_id is None all content is copied.
1115
:param find_ghosts: Find and copy revisions in the source that are
1116
ghosts in the target (and not reachable directly by walking out to
1117
the first-present revision in target from revision_id).
1119
# fast path same-url fetch operations
1120
if self.has_same_location(source):
1121
# check that last_revision is in 'from' and then return a
1123
if (revision_id is not None and
1124
not _mod_revision.is_null(revision_id)):
1125
self.get_revision(revision_id)
1127
# if there is no specific appropriate InterRepository, this will get
1128
# the InterRepository base class, which raises an
1129
# IncompatibleRepositories when asked to fetch.
1130
inter = InterRepository.get(source, self)
1131
return inter.fetch(revision_id=revision_id, pb=pb,
1132
find_ghosts=find_ghosts)
1134
def create_bundle(self, target, base, fileobj, format=None):
1135
return serializer.write_bundle(self, target, base, fileobj, format)
1137
def get_commit_builder(self, branch, parents, config, timestamp=None,
1138
timezone=None, committer=None, revprops=None,
1140
"""Obtain a CommitBuilder for this repository.
1142
:param branch: Branch to commit to.
1143
:param parents: Revision ids of the parents of the new revision.
1144
:param config: Configuration to use.
1145
:param timestamp: Optional timestamp recorded for commit.
1146
:param timezone: Optional timezone for timestamp.
1147
:param committer: Optional committer to set for commit.
1148
:param revprops: Optional dictionary of revision properties.
1149
:param revision_id: Optional revision id.
1151
result = self._commit_builder_class(self, parents, config,
1152
timestamp, timezone, committer, revprops, revision_id)
1153
self.start_write_group()
1157
if (self.control_files._lock_count == 1 and
1158
self.control_files._lock_mode == 'w'):
1159
if self._write_group is not None:
1160
self.abort_write_group()
1161
self.control_files.unlock()
1162
raise errors.BzrError(
1163
'Must end write groups before releasing write locks.')
1164
self.control_files.unlock()
1165
if self.control_files._lock_count == 0:
1166
self._inventory_entry_cache.clear()
1167
for repo in self._fallback_repositories:
1171
def clone(self, a_bzrdir, revision_id=None):
1172
"""Clone this repository into a_bzrdir using the current format.
1174
Currently no check is made that the format of this repository and
1175
the bzrdir format are compatible. FIXME RBC 20060201.
1177
:return: The newly created destination repository.
1179
# TODO: deprecate after 0.16; cloning this with all its settings is
1180
# probably not very useful -- mbp 20070423
1181
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1182
self.copy_content_into(dest_repo, revision_id)
1185
def start_write_group(self):
1186
"""Start a write group in the repository.
1188
Write groups are used by repositories which do not have a 1:1 mapping
1189
between file ids and backend store to manage the insertion of data from
1190
both fetch and commit operations.
1192
A write lock is required around the start_write_group/commit_write_group
1193
for the support of lock-requiring repository formats.
1195
One can only insert data into a repository inside a write group.
1199
if not self.is_write_locked():
1200
raise errors.NotWriteLocked(self)
1201
if self._write_group:
1202
raise errors.BzrError('already in a write group')
1203
self._start_write_group()
1204
# so we can detect unlock/relock - the write group is now entered.
1205
self._write_group = self.get_transaction()
1207
def _start_write_group(self):
1208
"""Template method for per-repository write group startup.
1210
This is called before the write group is considered to be
1215
def sprout(self, to_bzrdir, revision_id=None):
1216
"""Create a descendent repository for new development.
1218
Unlike clone, this does not copy the settings of the repository.
1220
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1221
dest_repo.fetch(self, revision_id=revision_id)
1224
def _create_sprouting_repo(self, a_bzrdir, shared):
1225
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1226
# use target default format.
1227
dest_repo = a_bzrdir.create_repository()
1229
# Most control formats need the repository to be specifically
1230
# created, but on some old all-in-one formats it's not needed
1232
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1233
except errors.UninitializableFormat:
1234
dest_repo = a_bzrdir.open_repository()
1237
def _get_sink(self):
1238
"""Return a sink for streaming into this repository."""
1239
return StreamSink(self)
1242
def has_revision(self, revision_id):
1243
"""True if this repository has a copy of the revision."""
1244
return revision_id in self.has_revisions((revision_id,))
1247
def has_revisions(self, revision_ids):
1248
"""Probe to find out the presence of multiple revisions.
1250
:param revision_ids: An iterable of revision_ids.
1251
:return: A set of the revision_ids that were present.
1253
parent_map = self.revisions.get_parent_map(
1254
[(rev_id,) for rev_id in revision_ids])
1256
if _mod_revision.NULL_REVISION in revision_ids:
1257
result.add(_mod_revision.NULL_REVISION)
1258
result.update([key[0] for key in parent_map])
1262
def get_revision(self, revision_id):
1263
"""Return the Revision object for a named revision."""
1264
return self.get_revisions([revision_id])[0]
1267
def get_revision_reconcile(self, revision_id):
1268
"""'reconcile' helper routine that allows access to a revision always.
1270
This variant of get_revision does not cross check the weave graph
1271
against the revision one as get_revision does: but it should only
1272
be used by reconcile, or reconcile-alike commands that are correcting
1273
or testing the revision graph.
1275
return self._get_revisions([revision_id])[0]
1278
def get_revisions(self, revision_ids):
1279
"""Get many revisions at once."""
1280
return self._get_revisions(revision_ids)
1283
def _get_revisions(self, revision_ids):
1284
"""Core work logic to get many revisions without sanity checks."""
1285
for rev_id in revision_ids:
1286
if not rev_id or not isinstance(rev_id, basestring):
1287
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1288
keys = [(key,) for key in revision_ids]
1289
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1291
for record in stream:
1292
if record.storage_kind == 'absent':
1293
raise errors.NoSuchRevision(self, record.key[0])
1294
text = record.get_bytes_as('fulltext')
1295
rev = self._serializer.read_revision_from_string(text)
1296
revs[record.key[0]] = rev
1297
return [revs[revid] for revid in revision_ids]
1300
def get_revision_xml(self, revision_id):
1301
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1302
# would have already do it.
1303
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1304
rev = self.get_revision(revision_id)
1305
rev_tmp = cStringIO.StringIO()
1306
# the current serializer..
1307
self._serializer.write_revision(rev, rev_tmp)
1309
return rev_tmp.getvalue()
1311
def get_deltas_for_revisions(self, revisions):
1312
"""Produce a generator of revision deltas.
1314
Note that the input is a sequence of REVISIONS, not revision_ids.
1315
Trees will be held in memory until the generator exits.
1316
Each delta is relative to the revision's lefthand predecessor.
1318
required_trees = set()
1319
for revision in revisions:
1320
required_trees.add(revision.revision_id)
1321
required_trees.update(revision.parent_ids[:1])
1322
trees = dict((t.get_revision_id(), t) for
1323
t in self.revision_trees(required_trees))
1324
for revision in revisions:
1325
if not revision.parent_ids:
1326
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1328
old_tree = trees[revision.parent_ids[0]]
1329
yield trees[revision.revision_id].changes_from(old_tree)
1332
def get_revision_delta(self, revision_id):
1333
"""Return the delta for one revision.
1335
The delta is relative to the left-hand predecessor of the
1338
r = self.get_revision(revision_id)
1339
return list(self.get_deltas_for_revisions([r]))[0]
1342
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1343
signature = gpg_strategy.sign(plaintext)
1344
self.add_signature_text(revision_id, signature)
1347
def add_signature_text(self, revision_id, signature):
1348
self.signatures.add_lines((revision_id,), (),
1349
osutils.split_lines(signature))
1351
def find_text_key_references(self):
1352
"""Find the text key references within the repository.
1354
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1355
revision_ids. Each altered file-ids has the exact revision_ids that
1356
altered it listed explicitly.
1357
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1358
to whether they were referred to by the inventory of the
1359
revision_id that they contain. The inventory texts from all present
1360
revision ids are assessed to generate this report.
1362
revision_keys = self.revisions.keys()
1363
w = self.inventories
1364
pb = ui.ui_factory.nested_progress_bar()
1366
return self._find_text_key_references_from_xml_inventory_lines(
1367
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1371
def _find_text_key_references_from_xml_inventory_lines(self,
1373
"""Core routine for extracting references to texts from inventories.
1375
This performs the translation of xml lines to revision ids.
1377
:param line_iterator: An iterator of lines, origin_version_id
1378
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1379
to whether they were referred to by the inventory of the
1380
revision_id that they contain. Note that if that revision_id was
1381
not part of the line_iterator's output then False will be given -
1382
even though it may actually refer to that key.
1384
if not self._serializer.support_altered_by_hack:
1385
raise AssertionError(
1386
"_find_text_key_references_from_xml_inventory_lines only "
1387
"supported for branches which store inventory as unnested xml"
1388
", not on %r" % self)
1391
# this code needs to read every new line in every inventory for the
1392
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1393
# not present in one of those inventories is unnecessary but not
1394
# harmful because we are filtering by the revision id marker in the
1395
# inventory lines : we only select file ids altered in one of those
1396
# revisions. We don't need to see all lines in the inventory because
1397
# only those added in an inventory in rev X can contain a revision=X
1399
unescape_revid_cache = {}
1400
unescape_fileid_cache = {}
1402
# jam 20061218 In a big fetch, this handles hundreds of thousands
1403
# of lines, so it has had a lot of inlining and optimizing done.
1404
# Sorry that it is a little bit messy.
1405
# Move several functions to be local variables, since this is a long
1407
search = self._file_ids_altered_regex.search
1408
unescape = _unescape_xml
1409
setdefault = result.setdefault
1410
for line, line_key in line_iterator:
1411
match = search(line)
1414
# One call to match.group() returning multiple items is quite a
1415
# bit faster than 2 calls to match.group() each returning 1
1416
file_id, revision_id = match.group('file_id', 'revision_id')
1418
# Inlining the cache lookups helps a lot when you make 170,000
1419
# lines and 350k ids, versus 8.4 unique ids.
1420
# Using a cache helps in 2 ways:
1421
# 1) Avoids unnecessary decoding calls
1422
# 2) Re-uses cached strings, which helps in future set and
1424
# (2) is enough that removing encoding entirely along with
1425
# the cache (so we are using plain strings) results in no
1426
# performance improvement.
1428
revision_id = unescape_revid_cache[revision_id]
1430
unescaped = unescape(revision_id)
1431
unescape_revid_cache[revision_id] = unescaped
1432
revision_id = unescaped
1434
# Note that unconditionally unescaping means that we deserialise
1435
# every fileid, which for general 'pull' is not great, but we don't
1436
# really want to have some many fulltexts that this matters anyway.
1439
file_id = unescape_fileid_cache[file_id]
1441
unescaped = unescape(file_id)
1442
unescape_fileid_cache[file_id] = unescaped
1445
key = (file_id, revision_id)
1446
setdefault(key, False)
1447
if revision_id == line_key[-1]:
1451
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1453
"""Helper routine for fileids_altered_by_revision_ids.
1455
This performs the translation of xml lines to revision ids.
1457
:param line_iterator: An iterator of lines, origin_version_id
1458
:param revision_ids: The revision ids to filter for. This should be a
1459
set or other type which supports efficient __contains__ lookups, as
1460
the revision id from each parsed line will be looked up in the
1461
revision_ids filter.
1462
:return: a dictionary mapping altered file-ids to an iterable of
1463
revision_ids. Each altered file-ids has the exact revision_ids that
1464
altered it listed explicitly.
1467
setdefault = result.setdefault
1469
self._find_text_key_references_from_xml_inventory_lines(
1470
line_iterator).iterkeys():
1471
# once data is all ensured-consistent; then this is
1472
# if revision_id == version_id
1473
if key[-1:] in revision_ids:
1474
setdefault(key[0], set()).add(key[-1])
1477
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1478
"""Find the file ids and versions affected by revisions.
1480
:param revisions: an iterable containing revision ids.
1481
:param _inv_weave: The inventory weave from this repository or None.
1482
If None, the inventory weave will be opened automatically.
1483
:return: a dictionary mapping altered file-ids to an iterable of
1484
revision_ids. Each altered file-ids has the exact revision_ids that
1485
altered it listed explicitly.
1487
selected_keys = set((revid,) for revid in revision_ids)
1488
w = _inv_weave or self.inventories
1489
pb = ui.ui_factory.nested_progress_bar()
1491
return self._find_file_ids_from_xml_inventory_lines(
1492
w.iter_lines_added_or_present_in_keys(
1493
selected_keys, pb=pb),
1498
def iter_files_bytes(self, desired_files):
1499
"""Iterate through file versions.
1501
Files will not necessarily be returned in the order they occur in
1502
desired_files. No specific order is guaranteed.
1504
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1505
value supplied by the caller as part of desired_files. It should
1506
uniquely identify the file version in the caller's context. (Examples:
1507
an index number or a TreeTransform trans_id.)
1509
bytes_iterator is an iterable of bytestrings for the file. The
1510
kind of iterable and length of the bytestrings are unspecified, but for
1511
this implementation, it is a list of bytes produced by
1512
VersionedFile.get_record_stream().
1514
:param desired_files: a list of (file_id, revision_id, identifier)
1518
for file_id, revision_id, callable_data in desired_files:
1519
text_keys[(file_id, revision_id)] = callable_data
1520
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1521
if record.storage_kind == 'absent':
1522
raise errors.RevisionNotPresent(record.key, self)
1523
yield text_keys[record.key], record.get_bytes_as('fulltext')
1525
def _generate_text_key_index(self, text_key_references=None,
1527
"""Generate a new text key index for the repository.
1529
This is an expensive function that will take considerable time to run.
1531
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1532
list of parents, also text keys. When a given key has no parents,
1533
the parents list will be [NULL_REVISION].
1535
# All revisions, to find inventory parents.
1536
if ancestors is None:
1537
graph = self.get_graph()
1538
ancestors = graph.get_parent_map(self.all_revision_ids())
1539
if text_key_references is None:
1540
text_key_references = self.find_text_key_references()
1541
pb = ui.ui_factory.nested_progress_bar()
1543
return self._do_generate_text_key_index(ancestors,
1544
text_key_references, pb)
1548
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1549
"""Helper for _generate_text_key_index to avoid deep nesting."""
1550
revision_order = tsort.topo_sort(ancestors)
1551
invalid_keys = set()
1553
for revision_id in revision_order:
1554
revision_keys[revision_id] = set()
1555
text_count = len(text_key_references)
1556
# a cache of the text keys to allow reuse; costs a dict of all the
1557
# keys, but saves a 2-tuple for every child of a given key.
1559
for text_key, valid in text_key_references.iteritems():
1561
invalid_keys.add(text_key)
1563
revision_keys[text_key[1]].add(text_key)
1564
text_key_cache[text_key] = text_key
1565
del text_key_references
1567
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1568
NULL_REVISION = _mod_revision.NULL_REVISION
1569
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1570
# too small for large or very branchy trees. However, for 55K path
1571
# trees, it would be easy to use too much memory trivially. Ideally we
1572
# could gauge this by looking at available real memory etc, but this is
1573
# always a tricky proposition.
1574
inventory_cache = lru_cache.LRUCache(10)
1575
batch_size = 10 # should be ~150MB on a 55K path tree
1576
batch_count = len(revision_order) / batch_size + 1
1578
pb.update("Calculating text parents.", processed_texts, text_count)
1579
for offset in xrange(batch_count):
1580
to_query = revision_order[offset * batch_size:(offset + 1) *
1584
for rev_tree in self.revision_trees(to_query):
1585
revision_id = rev_tree.get_revision_id()
1586
parent_ids = ancestors[revision_id]
1587
for text_key in revision_keys[revision_id]:
1588
pb.update("Calculating text parents.", processed_texts)
1589
processed_texts += 1
1590
candidate_parents = []
1591
for parent_id in parent_ids:
1592
parent_text_key = (text_key[0], parent_id)
1594
check_parent = parent_text_key not in \
1595
revision_keys[parent_id]
1597
# the parent parent_id is a ghost:
1598
check_parent = False
1599
# truncate the derived graph against this ghost.
1600
parent_text_key = None
1602
# look at the parent commit details inventories to
1603
# determine possible candidates in the per file graph.
1606
inv = inventory_cache[parent_id]
1608
inv = self.revision_tree(parent_id).inventory
1609
inventory_cache[parent_id] = inv
1610
parent_entry = inv._byid.get(text_key[0], None)
1611
if parent_entry is not None:
1613
text_key[0], parent_entry.revision)
1615
parent_text_key = None
1616
if parent_text_key is not None:
1617
candidate_parents.append(
1618
text_key_cache[parent_text_key])
1619
parent_heads = text_graph.heads(candidate_parents)
1620
new_parents = list(parent_heads)
1621
new_parents.sort(key=lambda x:candidate_parents.index(x))
1622
if new_parents == []:
1623
new_parents = [NULL_REVISION]
1624
text_index[text_key] = new_parents
1626
for text_key in invalid_keys:
1627
text_index[text_key] = [NULL_REVISION]
1630
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1631
"""Get an iterable listing the keys of all the data introduced by a set
1634
The keys will be ordered so that the corresponding items can be safely
1635
fetched and inserted in that order.
1637
:returns: An iterable producing tuples of (knit-kind, file-id,
1638
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1639
'revisions'. file-id is None unless knit-kind is 'file'.
1641
# XXX: it's a bit weird to control the inventory weave caching in this
1642
# generator. Ideally the caching would be done in fetch.py I think. Or
1643
# maybe this generator should explicitly have the contract that it
1644
# should not be iterated until the previously yielded item has been
1646
inv_w = self.inventories
1648
# file ids that changed
1649
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1651
num_file_ids = len(file_ids)
1652
for file_id, altered_versions in file_ids.iteritems():
1653
if _files_pb is not None:
1654
_files_pb.update("fetch texts", count, num_file_ids)
1656
yield ("file", file_id, altered_versions)
1657
# We're done with the files_pb. Note that it finished by the caller,
1658
# just as it was created by the caller.
1662
yield ("inventory", None, revision_ids)
1665
# XXX: Note ATM no callers actually pay attention to this return
1666
# instead they just use the list of revision ids and ignore
1667
# missing sigs. Consider removing this work entirely
1668
revisions_with_signatures = set(self.signatures.get_parent_map(
1669
[(r,) for r in revision_ids]))
1670
revisions_with_signatures = set(
1671
[r for (r,) in revisions_with_signatures])
1672
revisions_with_signatures.intersection_update(revision_ids)
1673
yield ("signatures", None, revisions_with_signatures)
1676
yield ("revisions", None, revision_ids)
1679
def get_inventory(self, revision_id):
1680
"""Get Inventory object by revision id."""
1681
return self.iter_inventories([revision_id]).next()
1683
def iter_inventories(self, revision_ids):
1684
"""Get many inventories by revision_ids.
1686
This will buffer some or all of the texts used in constructing the
1687
inventories in memory, but will only parse a single inventory at a
1690
:return: An iterator of inventories.
1692
if ((None in revision_ids)
1693
or (_mod_revision.NULL_REVISION in revision_ids)):
1694
raise ValueError('cannot get null revision inventory')
1695
return self._iter_inventories(revision_ids)
1697
def _iter_inventories(self, revision_ids):
1698
"""single-document based inventory iteration."""
1699
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1700
yield self.deserialise_inventory(revision_id, text)
1702
def _iter_inventory_xmls(self, revision_ids):
1703
keys = [(revision_id,) for revision_id in revision_ids]
1704
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1706
for record in stream:
1707
if record.storage_kind != 'absent':
1708
text_chunks[record.key] = record.get_bytes_as('chunked')
1710
raise errors.NoSuchRevision(self, record.key)
1712
chunks = text_chunks.pop(key)
1713
yield ''.join(chunks), key[-1]
1715
def deserialise_inventory(self, revision_id, xml):
1716
"""Transform the xml into an inventory object.
1718
:param revision_id: The expected revision id of the inventory.
1719
:param xml: A serialised inventory.
1721
result = self._serializer.read_inventory_from_string(xml, revision_id,
1722
entry_cache=self._inventory_entry_cache)
1723
if result.revision_id != revision_id:
1724
raise AssertionError('revision id mismatch %s != %s' % (
1725
result.revision_id, revision_id))
1728
def serialise_inventory(self, inv):
1729
return self._serializer.write_inventory_to_string(inv)
1731
def _serialise_inventory_to_lines(self, inv):
1732
return self._serializer.write_inventory_to_lines(inv)
1734
def get_serializer_format(self):
1735
return self._serializer.format_num
1738
def get_inventory_xml(self, revision_id):
1739
"""Get inventory XML as a file object."""
1740
texts = self._iter_inventory_xmls([revision_id])
1742
text, revision_id = texts.next()
1743
except StopIteration:
1744
raise errors.HistoryMissing(self, 'inventory', revision_id)
1748
def get_inventory_sha1(self, revision_id):
1749
"""Return the sha1 hash of the inventory entry
1751
return self.get_revision(revision_id).inventory_sha1
1753
def iter_reverse_revision_history(self, revision_id):
1754
"""Iterate backwards through revision ids in the lefthand history
1756
:param revision_id: The revision id to start with. All its lefthand
1757
ancestors will be traversed.
1759
graph = self.get_graph()
1760
next_id = revision_id
1762
if next_id in (None, _mod_revision.NULL_REVISION):
1765
# Note: The following line may raise KeyError in the event of
1766
# truncated history. We decided not to have a try:except:raise
1767
# RevisionNotPresent here until we see a use for it, because of the
1768
# cost in an inner loop that is by its very nature O(history).
1769
# Robert Collins 20080326
1770
parents = graph.get_parent_map([next_id])[next_id]
1771
if len(parents) == 0:
1774
next_id = parents[0]
1777
def get_revision_inventory(self, revision_id):
1778
"""Return inventory of a past revision."""
1779
# TODO: Unify this with get_inventory()
1780
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1781
# must be the same as its revision, so this is trivial.
1782
if revision_id is None:
1783
# This does not make sense: if there is no revision,
1784
# then it is the current tree inventory surely ?!
1785
# and thus get_root_id() is something that looks at the last
1786
# commit on the branch, and the get_root_id is an inventory check.
1787
raise NotImplementedError
1788
# return Inventory(self.get_root_id())
1790
return self.get_inventory(revision_id)
1792
def is_shared(self):
1793
"""Return True if this repository is flagged as a shared repository."""
1794
raise NotImplementedError(self.is_shared)
1797
def reconcile(self, other=None, thorough=False):
1798
"""Reconcile this repository."""
1799
from bzrlib.reconcile import RepoReconciler
1800
reconciler = RepoReconciler(self, thorough=thorough)
1801
reconciler.reconcile()
1804
def _refresh_data(self):
1805
"""Helper called from lock_* to ensure coherency with disk.
1807
The default implementation does nothing; it is however possible
1808
for repositories to maintain loaded indices across multiple locks
1809
by checking inside their implementation of this method to see
1810
whether their indices are still valid. This depends of course on
1811
the disk format being validatable in this manner.
1815
def revision_tree(self, revision_id):
1816
"""Return Tree for a revision on this branch.
1818
`revision_id` may be NULL_REVISION for the empty tree revision.
1820
revision_id = _mod_revision.ensure_null(revision_id)
1821
# TODO: refactor this to use an existing revision object
1822
# so we don't need to read it in twice.
1823
if revision_id == _mod_revision.NULL_REVISION:
1824
return RevisionTree(self, Inventory(root_id=None),
1825
_mod_revision.NULL_REVISION)
1827
inv = self.get_revision_inventory(revision_id)
1828
return RevisionTree(self, inv, revision_id)
1830
def revision_trees(self, revision_ids):
1831
"""Return Tree for a revision on this branch.
1833
`revision_id` may not be None or 'null:'"""
1834
inventories = self.iter_inventories(revision_ids)
1835
for inv in inventories:
1836
yield RevisionTree(self, inv, inv.revision_id)
1839
def get_ancestry(self, revision_id, topo_sorted=True):
1840
"""Return a list of revision-ids integrated by a revision.
1842
The first element of the list is always None, indicating the origin
1843
revision. This might change when we have history horizons, or
1844
perhaps we should have a new API.
1846
This is topologically sorted.
1848
if _mod_revision.is_null(revision_id):
1850
if not self.has_revision(revision_id):
1851
raise errors.NoSuchRevision(self, revision_id)
1852
graph = self.get_graph()
1854
search = graph._make_breadth_first_searcher([revision_id])
1857
found, ghosts = search.next_with_ghosts()
1858
except StopIteration:
1861
if _mod_revision.NULL_REVISION in keys:
1862
keys.remove(_mod_revision.NULL_REVISION)
1864
parent_map = graph.get_parent_map(keys)
1865
keys = tsort.topo_sort(parent_map)
1866
return [None] + list(keys)
1869
"""Compress the data within the repository.
1871
This operation only makes sense for some repository types. For other
1872
types it should be a no-op that just returns.
1874
This stub method does not require a lock, but subclasses should use
1875
@needs_write_lock as this is a long running call its reasonable to
1876
implicitly lock for the user.
1880
@deprecated_method(one_six)
1881
def print_file(self, file, revision_id):
1882
"""Print `file` to stdout.
1884
FIXME RBC 20060125 as John Meinel points out this is a bad api
1885
- it writes to stdout, it assumes that that is valid etc. Fix
1886
by creating a new more flexible convenience function.
1888
tree = self.revision_tree(revision_id)
1889
# use inventory as it was in that revision
1890
file_id = tree.inventory.path2id(file)
1892
# TODO: jam 20060427 Write a test for this code path
1893
# it had a bug in it, and was raising the wrong
1895
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1896
tree.print_file(file_id)
1898
def get_transaction(self):
1899
return self.control_files.get_transaction()
1901
@deprecated_method(one_one)
1902
def get_parents(self, revision_ids):
1903
"""See StackedParentsProvider.get_parents"""
1904
parent_map = self.get_parent_map(revision_ids)
1905
return [parent_map.get(r, None) for r in revision_ids]
1907
def get_parent_map(self, revision_ids):
1908
"""See graph._StackedParentsProvider.get_parent_map"""
1909
# revisions index works in keys; this just works in revisions
1910
# therefore wrap and unwrap
1913
for revision_id in revision_ids:
1914
if revision_id == _mod_revision.NULL_REVISION:
1915
result[revision_id] = ()
1916
elif revision_id is None:
1917
raise ValueError('get_parent_map(None) is not valid')
1919
query_keys.append((revision_id ,))
1920
for ((revision_id,), parent_keys) in \
1921
self.revisions.get_parent_map(query_keys).iteritems():
1923
result[revision_id] = tuple(parent_revid
1924
for (parent_revid,) in parent_keys)
1926
result[revision_id] = (_mod_revision.NULL_REVISION,)
1929
def _make_parents_provider(self):
1932
def get_graph(self, other_repository=None):
1933
"""Return the graph walker for this repository format"""
1934
parents_provider = self._make_parents_provider()
1935
if (other_repository is not None and
1936
not self.has_same_location(other_repository)):
1937
parents_provider = graph._StackedParentsProvider(
1938
[parents_provider, other_repository._make_parents_provider()])
1939
return graph.Graph(parents_provider)
1941
def _get_versioned_file_checker(self):
1942
"""Return an object suitable for checking versioned files."""
1943
return _VersionedFileChecker(self)
1945
def revision_ids_to_search_result(self, result_set):
1946
"""Convert a set of revision ids to a graph SearchResult."""
1947
result_parents = set()
1948
for parents in self.get_graph().get_parent_map(
1949
result_set).itervalues():
1950
result_parents.update(parents)
1951
included_keys = result_set.intersection(result_parents)
1952
start_keys = result_set.difference(included_keys)
1953
exclude_keys = result_parents.difference(result_set)
1954
result = graph.SearchResult(start_keys, exclude_keys,
1955
len(result_set), result_set)
1959
def set_make_working_trees(self, new_value):
1960
"""Set the policy flag for making working trees when creating branches.
1962
This only applies to branches that use this repository.
1964
The default is 'True'.
1965
:param new_value: True to restore the default, False to disable making
1968
raise NotImplementedError(self.set_make_working_trees)
1970
def make_working_trees(self):
1971
"""Returns the policy for making working trees on new branches."""
1972
raise NotImplementedError(self.make_working_trees)
1975
def sign_revision(self, revision_id, gpg_strategy):
1976
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1977
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1980
def has_signature_for_revision_id(self, revision_id):
1981
"""Query for a revision signature for revision_id in the repository."""
1982
if not self.has_revision(revision_id):
1983
raise errors.NoSuchRevision(self, revision_id)
1984
sig_present = (1 == len(
1985
self.signatures.get_parent_map([(revision_id,)])))
1989
def get_signature_text(self, revision_id):
1990
"""Return the text for a signature."""
1991
stream = self.signatures.get_record_stream([(revision_id,)],
1993
record = stream.next()
1994
if record.storage_kind == 'absent':
1995
raise errors.NoSuchRevision(self, revision_id)
1996
return record.get_bytes_as('fulltext')
1999
def check(self, revision_ids=None):
2000
"""Check consistency of all history of given revision_ids.
2002
Different repository implementations should override _check().
2004
:param revision_ids: A non-empty list of revision_ids whose ancestry
2005
will be checked. Typically the last revision_id of a branch.
2007
return self._check(revision_ids)
2009
def _check(self, revision_ids):
2010
result = check.Check(self)
2014
def _warn_if_deprecated(self):
2015
global _deprecation_warning_done
2016
if _deprecation_warning_done:
2018
_deprecation_warning_done = True
2019
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2020
% (self._format, self.bzrdir.transport.base))
2022
def supports_rich_root(self):
2023
return self._format.rich_root_data
2025
def _check_ascii_revisionid(self, revision_id, method):
2026
"""Private helper for ascii-only repositories."""
2027
# weave repositories refuse to store revisionids that are non-ascii.
2028
if revision_id is not None:
2029
# weaves require ascii revision ids.
2030
if isinstance(revision_id, unicode):
2032
revision_id.encode('ascii')
2033
except UnicodeEncodeError:
2034
raise errors.NonAsciiRevisionId(method, self)
2037
revision_id.decode('ascii')
2038
except UnicodeDecodeError:
2039
raise errors.NonAsciiRevisionId(method, self)
2041
def revision_graph_can_have_wrong_parents(self):
2042
"""Is it possible for this repository to have a revision graph with
2045
If True, then this repository must also implement
2046
_find_inconsistent_revision_parents so that check and reconcile can
2047
check for inconsistencies before proceeding with other checks that may
2048
depend on the revision index being consistent.
2050
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2053
# remove these delegates a while after bzr 0.15
2054
def __make_delegated(name, from_module):
2055
def _deprecated_repository_forwarder():
2056
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2057
% (name, from_module),
2060
m = __import__(from_module, globals(), locals(), [name])
2062
return getattr(m, name)
2063
except AttributeError:
2064
raise AttributeError('module %s has no name %s'
2066
globals()[name] = _deprecated_repository_forwarder
2069
'AllInOneRepository',
2070
'WeaveMetaDirRepository',
2071
'PreSplitOutRepositoryFormat',
2072
'RepositoryFormat4',
2073
'RepositoryFormat5',
2074
'RepositoryFormat6',
2075
'RepositoryFormat7',
2077
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2081
'RepositoryFormatKnit',
2082
'RepositoryFormatKnit1',
2084
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2087
def install_revision(repository, rev, revision_tree):
2088
"""Install all revision data into a repository."""
2089
install_revisions(repository, [(rev, revision_tree, None)])
2092
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2093
"""Install all revision data into a repository.
2095
Accepts an iterable of revision, tree, signature tuples. The signature
2098
repository.start_write_group()
2100
for n, (revision, revision_tree, signature) in enumerate(iterable):
2101
_install_revision(repository, revision, revision_tree, signature)
2103
pb.update('Transferring revisions', n + 1, num_revisions)
2105
repository.abort_write_group()
2108
repository.commit_write_group()
2111
def _install_revision(repository, rev, revision_tree, signature):
2112
"""Install all revision data into a repository."""
2113
present_parents = []
2115
for p_id in rev.parent_ids:
2116
if repository.has_revision(p_id):
2117
present_parents.append(p_id)
2118
parent_trees[p_id] = repository.revision_tree(p_id)
2120
parent_trees[p_id] = repository.revision_tree(
2121
_mod_revision.NULL_REVISION)
2123
inv = revision_tree.inventory
2124
entries = inv.iter_entries()
2125
# backwards compatibility hack: skip the root id.
2126
if not repository.supports_rich_root():
2127
path, root = entries.next()
2128
if root.revision != rev.revision_id:
2129
raise errors.IncompatibleRevision(repr(repository))
2131
for path, ie in entries:
2132
text_keys[(ie.file_id, ie.revision)] = ie
2133
text_parent_map = repository.texts.get_parent_map(text_keys)
2134
missing_texts = set(text_keys) - set(text_parent_map)
2135
# Add the texts that are not already present
2136
for text_key in missing_texts:
2137
ie = text_keys[text_key]
2139
# FIXME: TODO: The following loop overlaps/duplicates that done by
2140
# commit to determine parents. There is a latent/real bug here where
2141
# the parents inserted are not those commit would do - in particular
2142
# they are not filtered by heads(). RBC, AB
2143
for revision, tree in parent_trees.iteritems():
2144
if ie.file_id not in tree:
2146
parent_id = tree.inventory[ie.file_id].revision
2147
if parent_id in text_parents:
2149
text_parents.append((ie.file_id, parent_id))
2150
lines = revision_tree.get_file(ie.file_id).readlines()
2151
repository.texts.add_lines(text_key, text_parents, lines)
2153
# install the inventory
2154
repository.add_inventory(rev.revision_id, inv, present_parents)
2155
except errors.RevisionAlreadyPresent:
2157
if signature is not None:
2158
repository.add_signature_text(rev.revision_id, signature)
2159
repository.add_revision(rev.revision_id, rev, inv)
2162
class MetaDirRepository(Repository):
2163
"""Repositories in the new meta-dir layout.
2165
:ivar _transport: Transport for access to repository control files,
2166
typically pointing to .bzr/repository.
2169
def __init__(self, _format, a_bzrdir, control_files):
2170
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2171
self._transport = control_files._transport
2173
def is_shared(self):
2174
"""Return True if this repository is flagged as a shared repository."""
2175
return self._transport.has('shared-storage')
2178
def set_make_working_trees(self, new_value):
2179
"""Set the policy flag for making working trees when creating branches.
2181
This only applies to branches that use this repository.
2183
The default is 'True'.
2184
:param new_value: True to restore the default, False to disable making
2189
self._transport.delete('no-working-trees')
2190
except errors.NoSuchFile:
2193
self._transport.put_bytes('no-working-trees', '',
2194
mode=self.bzrdir._get_file_mode())
2196
def make_working_trees(self):
2197
"""Returns the policy for making working trees on new branches."""
2198
return not self._transport.has('no-working-trees')
2201
class MetaDirVersionedFileRepository(MetaDirRepository):
2202
"""Repositories in a meta-dir, that work via versioned file objects."""
2204
def __init__(self, _format, a_bzrdir, control_files):
2205
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2209
network_format_registry = registry.FormatRegistry()
2210
"""Registry of formats indexed by their network name.
2212
The network name for a repository format is an identifier that can be used when
2213
referring to formats with smart server operations. See
2214
RepositoryFormat.network_name() for more detail.
2218
format_registry = registry.FormatRegistry(network_format_registry)
2219
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2221
This can contain either format instances themselves, or classes/factories that
2222
can be called to obtain one.
2226
#####################################################################
2227
# Repository Formats
2229
class RepositoryFormat(object):
2230
"""A repository format.
2232
Formats provide four things:
2233
* An initialization routine to construct repository data on disk.
2234
* a optional format string which is used when the BzrDir supports
2236
* an open routine which returns a Repository instance.
2237
* A network name for referring to the format in smart server RPC
2240
There is one and only one Format subclass for each on-disk format. But
2241
there can be one Repository subclass that is used for several different
2242
formats. The _format attribute on a Repository instance can be used to
2243
determine the disk format.
2245
Formats are placed in a registry by their format string for reference
2246
during opening. These should be subclasses of RepositoryFormat for
2249
Once a format is deprecated, just deprecate the initialize and open
2250
methods on the format class. Do not deprecate the object, as the
2251
object may be created even when a repository instnace hasn't been
2254
Common instance attributes:
2255
_matchingbzrdir - the bzrdir format that the repository format was
2256
originally written to work with. This can be used if manually
2257
constructing a bzrdir and repository, or more commonly for test suite
2261
# Set to True or False in derived classes. True indicates that the format
2262
# supports ghosts gracefully.
2263
supports_ghosts = None
2264
# Can this repository be given external locations to lookup additional
2265
# data. Set to True or False in derived classes.
2266
supports_external_lookups = None
2269
return "<%s>" % self.__class__.__name__
2271
def __eq__(self, other):
2272
# format objects are generally stateless
2273
return isinstance(other, self.__class__)
2275
def __ne__(self, other):
2276
return not self == other
2279
def find_format(klass, a_bzrdir):
2280
"""Return the format for the repository object in a_bzrdir.
2282
This is used by bzr native formats that have a "format" file in
2283
the repository. Other methods may be used by different types of
2287
transport = a_bzrdir.get_repository_transport(None)
2288
format_string = transport.get("format").read()
2289
return format_registry.get(format_string)
2290
except errors.NoSuchFile:
2291
raise errors.NoRepositoryPresent(a_bzrdir)
2293
raise errors.UnknownFormatError(format=format_string,
2297
def register_format(klass, format):
2298
format_registry.register(format.get_format_string(), format)
2301
def unregister_format(klass, format):
2302
format_registry.remove(format.get_format_string())
2305
def get_default_format(klass):
2306
"""Return the current default format."""
2307
from bzrlib import bzrdir
2308
return bzrdir.format_registry.make_bzrdir('default').repository_format
2310
def get_format_string(self):
2311
"""Return the ASCII format string that identifies this format.
2313
Note that in pre format ?? repositories the format string is
2314
not permitted nor written to disk.
2316
raise NotImplementedError(self.get_format_string)
2318
def get_format_description(self):
2319
"""Return the short description for this format."""
2320
raise NotImplementedError(self.get_format_description)
2322
# TODO: this shouldn't be in the base class, it's specific to things that
2323
# use weaves or knits -- mbp 20070207
2324
def _get_versioned_file_store(self,
2329
versionedfile_class=None,
2330
versionedfile_kwargs={},
2332
if versionedfile_class is None:
2333
versionedfile_class = self._versionedfile_class
2334
weave_transport = control_files._transport.clone(name)
2335
dir_mode = control_files._dir_mode
2336
file_mode = control_files._file_mode
2337
return VersionedFileStore(weave_transport, prefixed=prefixed,
2339
file_mode=file_mode,
2340
versionedfile_class=versionedfile_class,
2341
versionedfile_kwargs=versionedfile_kwargs,
2344
def initialize(self, a_bzrdir, shared=False):
2345
"""Initialize a repository of this format in a_bzrdir.
2347
:param a_bzrdir: The bzrdir to put the new repository in it.
2348
:param shared: The repository should be initialized as a sharable one.
2349
:returns: The new repository object.
2351
This may raise UninitializableFormat if shared repository are not
2352
compatible the a_bzrdir.
2354
raise NotImplementedError(self.initialize)
2356
def is_supported(self):
2357
"""Is this format supported?
2359
Supported formats must be initializable and openable.
2360
Unsupported formats may not support initialization or committing or
2361
some other features depending on the reason for not being supported.
2365
def network_name(self):
2366
"""A simple byte string uniquely identifying this format for RPC calls.
2368
MetaDir repository formats use their disk format string to identify the
2369
repository over the wire. All in one formats such as bzr < 0.8, and
2370
foreign formats like svn/git and hg should use some marker which is
2371
unique and immutable.
2373
raise NotImplementedError(self.network_name)
2375
def check_conversion_target(self, target_format):
2376
raise NotImplementedError(self.check_conversion_target)
2378
def open(self, a_bzrdir, _found=False):
2379
"""Return an instance of this format for the bzrdir a_bzrdir.
2381
_found is a private parameter, do not use it.
2383
raise NotImplementedError(self.open)
2386
class MetaDirRepositoryFormat(RepositoryFormat):
2387
"""Common base class for the new repositories using the metadir layout."""
2389
rich_root_data = False
2390
supports_tree_reference = False
2391
supports_external_lookups = False
2394
def _matchingbzrdir(self):
2395
matching = bzrdir.BzrDirMetaFormat1()
2396
matching.repository_format = self
2400
super(MetaDirRepositoryFormat, self).__init__()
2402
def _create_control_files(self, a_bzrdir):
2403
"""Create the required files and the initial control_files object."""
2404
# FIXME: RBC 20060125 don't peek under the covers
2405
# NB: no need to escape relative paths that are url safe.
2406
repository_transport = a_bzrdir.get_repository_transport(self)
2407
control_files = lockable_files.LockableFiles(repository_transport,
2408
'lock', lockdir.LockDir)
2409
control_files.create_lock()
2410
return control_files
2412
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2413
"""Upload the initial blank content."""
2414
control_files = self._create_control_files(a_bzrdir)
2415
control_files.lock_write()
2416
transport = control_files._transport
2418
utf8_files += [('shared-storage', '')]
2420
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2421
for (filename, content_stream) in files:
2422
transport.put_file(filename, content_stream,
2423
mode=a_bzrdir._get_file_mode())
2424
for (filename, content_bytes) in utf8_files:
2425
transport.put_bytes_non_atomic(filename, content_bytes,
2426
mode=a_bzrdir._get_file_mode())
2428
control_files.unlock()
2430
def network_name(self):
2431
"""Metadir formats have matching disk and network format strings."""
2432
return self.get_format_string()
2435
# Pre-0.8 formats that don't have a disk format string (because they are
2436
# versioned by the matching control directory). We use the control directories
2437
# disk format string as a key for the network_name because they meet the
2438
# constraints (simple string, unique, immmutable).
2439
network_format_registry.register_lazy(
2440
"Bazaar-NG branch, format 5\n",
2441
'bzrlib.repofmt.weaverepo',
2442
'RepositoryFormat5',
2444
network_format_registry.register_lazy(
2445
"Bazaar-NG branch, format 6\n",
2446
'bzrlib.repofmt.weaverepo',
2447
'RepositoryFormat6',
2450
# formats which have no format string are not discoverable or independently
2451
# creatable on disk, so are not registered in format_registry. They're
2452
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2453
# needed, it's constructed directly by the BzrDir. Non-native formats where
2454
# the repository is not separately opened are similar.
2456
format_registry.register_lazy(
2457
'Bazaar-NG Repository format 7',
2458
'bzrlib.repofmt.weaverepo',
2462
format_registry.register_lazy(
2463
'Bazaar-NG Knit Repository Format 1',
2464
'bzrlib.repofmt.knitrepo',
2465
'RepositoryFormatKnit1',
2468
format_registry.register_lazy(
2469
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2470
'bzrlib.repofmt.knitrepo',
2471
'RepositoryFormatKnit3',
2474
format_registry.register_lazy(
2475
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2476
'bzrlib.repofmt.knitrepo',
2477
'RepositoryFormatKnit4',
2480
# Pack-based formats. There is one format for pre-subtrees, and one for
2481
# post-subtrees to allow ease of testing.
2482
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2483
format_registry.register_lazy(
2484
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2485
'bzrlib.repofmt.pack_repo',
2486
'RepositoryFormatKnitPack1',
2488
format_registry.register_lazy(
2489
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2490
'bzrlib.repofmt.pack_repo',
2491
'RepositoryFormatKnitPack3',
2493
format_registry.register_lazy(
2494
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2495
'bzrlib.repofmt.pack_repo',
2496
'RepositoryFormatKnitPack4',
2498
format_registry.register_lazy(
2499
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2500
'bzrlib.repofmt.pack_repo',
2501
'RepositoryFormatKnitPack5',
2503
format_registry.register_lazy(
2504
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2505
'bzrlib.repofmt.pack_repo',
2506
'RepositoryFormatKnitPack5RichRoot',
2508
format_registry.register_lazy(
2509
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2510
'bzrlib.repofmt.pack_repo',
2511
'RepositoryFormatKnitPack5RichRootBroken',
2513
format_registry.register_lazy(
2514
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2515
'bzrlib.repofmt.pack_repo',
2516
'RepositoryFormatKnitPack6',
2518
format_registry.register_lazy(
2519
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2520
'bzrlib.repofmt.pack_repo',
2521
'RepositoryFormatKnitPack6RichRoot',
2524
# Development formats.
2525
# 1.7->1.8 go below here
2526
format_registry.register_lazy(
2527
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2528
'bzrlib.repofmt.pack_repo',
2529
'RepositoryFormatPackDevelopment2',
2531
format_registry.register_lazy(
2532
("Bazaar development format 2 with subtree support "
2533
"(needs bzr.dev from before 1.8)\n"),
2534
'bzrlib.repofmt.pack_repo',
2535
'RepositoryFormatPackDevelopment2Subtree',
2539
class InterRepository(InterObject):
2540
"""This class represents operations taking place between two repositories.
2542
Its instances have methods like copy_content and fetch, and contain
2543
references to the source and target repositories these operations can be
2546
Often we will provide convenience methods on 'repository' which carry out
2547
operations with another repository - they will always forward to
2548
InterRepository.get(other).method_name(parameters).
2551
_walk_to_common_revisions_batch_size = 1
2553
"""The available optimised InterRepository types."""
2555
def __init__(self, source, target):
2556
InterObject.__init__(self, source, target)
2557
# These two attributes may be overridden by e.g. InterOtherToRemote to
2558
# provide a faster implementation.
2559
self.target_get_graph = self.target.get_graph
2560
self.target_get_parent_map = self.target.get_parent_map
2562
def copy_content(self, revision_id=None):
2563
raise NotImplementedError(self.copy_content)
2565
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2566
"""Fetch the content required to construct revision_id.
2568
The content is copied from self.source to self.target.
2570
:param revision_id: if None all content is copied, if NULL_REVISION no
2572
:param pb: optional progress bar to use for progress reports. If not
2573
provided a default one will be created.
2575
:returns: (copied_revision_count, failures).
2577
# Normally we should find a specific InterRepository subclass to do
2578
# the fetch; if nothing else then at least InterSameDataRepository.
2579
# If none of them is suitable it looks like fetching is not possible;
2580
# we try to give a good message why. _assert_same_model will probably
2581
# give a helpful message; otherwise a generic one.
2582
self._assert_same_model(self.source, self.target)
2583
raise errors.IncompatibleRepositories(self.source, self.target,
2584
"no suitableInterRepository found")
2586
def _walk_to_common_revisions(self, revision_ids):
2587
"""Walk out from revision_ids in source to revisions target has.
2589
:param revision_ids: The start point for the search.
2590
:return: A set of revision ids.
2592
target_graph = self.target_get_graph()
2593
revision_ids = frozenset(revision_ids)
2594
# Fast path for the case where all the revisions are already in the
2596
# (Although this does incur an extra round trip for the
2597
# fairly common case where the target doesn't already have the revision
2599
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2600
return graph.SearchResult(revision_ids, set(), 0, set())
2601
missing_revs = set()
2602
source_graph = self.source.get_graph()
2603
# ensure we don't pay silly lookup costs.
2604
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2605
null_set = frozenset([_mod_revision.NULL_REVISION])
2606
searcher_exhausted = False
2610
# Iterate the searcher until we have enough next_revs
2611
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2613
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2614
next_revs.update(next_revs_part)
2615
ghosts.update(ghosts_part)
2616
except StopIteration:
2617
searcher_exhausted = True
2619
# If there are ghosts in the source graph, and the caller asked for
2620
# them, make sure that they are present in the target.
2621
# We don't care about other ghosts as we can't fetch them and
2622
# haven't been asked to.
2623
ghosts_to_check = set(revision_ids.intersection(ghosts))
2624
revs_to_get = set(next_revs).union(ghosts_to_check)
2626
have_revs = set(target_graph.get_parent_map(revs_to_get))
2627
# we always have NULL_REVISION present.
2628
have_revs = have_revs.union(null_set)
2629
# Check if the target is missing any ghosts we need.
2630
ghosts_to_check.difference_update(have_revs)
2632
# One of the caller's revision_ids is a ghost in both the
2633
# source and the target.
2634
raise errors.NoSuchRevision(
2635
self.source, ghosts_to_check.pop())
2636
missing_revs.update(next_revs - have_revs)
2637
# Because we may have walked past the original stop point, make
2638
# sure everything is stopped
2639
stop_revs = searcher.find_seen_ancestors(have_revs)
2640
searcher.stop_searching_any(stop_revs)
2641
if searcher_exhausted:
2643
return searcher.get_result()
2645
@deprecated_method(one_two)
2647
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2648
"""Return the revision ids that source has that target does not.
2650
These are returned in topological order.
2652
:param revision_id: only return revision ids included by this
2654
:param find_ghosts: If True find missing revisions in deep history
2655
rather than just finding the surface difference.
2657
return list(self.search_missing_revision_ids(
2658
revision_id, find_ghosts).get_keys())
2661
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2662
"""Return the revision ids that source has that target does not.
2664
:param revision_id: only return revision ids included by this
2666
:param find_ghosts: If True find missing revisions in deep history
2667
rather than just finding the surface difference.
2668
:return: A bzrlib.graph.SearchResult.
2670
# stop searching at found target revisions.
2671
if not find_ghosts and revision_id is not None:
2672
return self._walk_to_common_revisions([revision_id])
2673
# generic, possibly worst case, slow code path.
2674
target_ids = set(self.target.all_revision_ids())
2675
if revision_id is not None:
2676
source_ids = self.source.get_ancestry(revision_id)
2677
if source_ids[0] is not None:
2678
raise AssertionError()
2681
source_ids = self.source.all_revision_ids()
2682
result_set = set(source_ids).difference(target_ids)
2683
return self.source.revision_ids_to_search_result(result_set)
2686
def _same_model(source, target):
2687
"""True if source and target have the same data representation.
2689
Note: this is always called on the base class; overriding it in a
2690
subclass will have no effect.
2693
InterRepository._assert_same_model(source, target)
2695
except errors.IncompatibleRepositories, e:
2699
def _assert_same_model(source, target):
2700
"""Raise an exception if two repositories do not use the same model.
2702
if source.supports_rich_root() != target.supports_rich_root():
2703
raise errors.IncompatibleRepositories(source, target,
2704
"different rich-root support")
2705
if source._serializer != target._serializer:
2706
raise errors.IncompatibleRepositories(source, target,
2707
"different serializers")
2710
class InterSameDataRepository(InterRepository):
2711
"""Code for converting between repositories that represent the same data.
2713
Data format and model must match for this to work.
2717
def _get_repo_format_to_test(self):
2718
"""Repository format for testing with.
2720
InterSameData can pull from subtree to subtree and from non-subtree to
2721
non-subtree, so we test this with the richest repository format.
2723
from bzrlib.repofmt import knitrepo
2724
return knitrepo.RepositoryFormatKnit3()
2727
def is_compatible(source, target):
2728
return InterRepository._same_model(source, target)
2731
def copy_content(self, revision_id=None):
2732
"""Make a complete copy of the content in self into destination.
2734
This copies both the repository's revision data, and configuration information
2735
such as the make_working_trees setting.
2737
This is a destructive operation! Do not use it on existing
2740
:param revision_id: Only copy the content needed to construct
2741
revision_id and its parents.
2744
self.target.set_make_working_trees(self.source.make_working_trees())
2745
except NotImplementedError:
2747
# but don't bother fetching if we have the needed data now.
2748
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2749
self.target.has_revision(revision_id)):
2751
self.target.fetch(self.source, revision_id=revision_id)
2754
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2755
"""See InterRepository.fetch()."""
2756
from bzrlib.fetch import RepoFetcher
2757
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2758
self.source, self.source._format, self.target,
2759
self.target._format)
2760
f = RepoFetcher(to_repository=self.target,
2761
from_repository=self.source,
2762
last_revision=revision_id,
2763
pb=pb, find_ghosts=find_ghosts)
2764
return f.count_copied, f.failed_revisions
2767
class InterWeaveRepo(InterSameDataRepository):
2768
"""Optimised code paths between Weave based repositories.
2770
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2771
implemented lazy inter-object optimisation.
2775
def _get_repo_format_to_test(self):
2776
from bzrlib.repofmt import weaverepo
2777
return weaverepo.RepositoryFormat7()
2780
def is_compatible(source, target):
2781
"""Be compatible with known Weave formats.
2783
We don't test for the stores being of specific types because that
2784
could lead to confusing results, and there is no need to be
2787
from bzrlib.repofmt.weaverepo import (
2793
return (isinstance(source._format, (RepositoryFormat5,
2795
RepositoryFormat7)) and
2796
isinstance(target._format, (RepositoryFormat5,
2798
RepositoryFormat7)))
2799
except AttributeError:
2803
def copy_content(self, revision_id=None):
2804
"""See InterRepository.copy_content()."""
2805
# weave specific optimised path:
2807
self.target.set_make_working_trees(self.source.make_working_trees())
2808
except (errors.RepositoryUpgradeRequired, NotImplemented):
2810
# FIXME do not peek!
2811
if self.source._transport.listable():
2812
pb = ui.ui_factory.nested_progress_bar()
2814
self.target.texts.insert_record_stream(
2815
self.source.texts.get_record_stream(
2816
self.source.texts.keys(), 'topological', False))
2817
pb.update('copying inventory', 0, 1)
2818
self.target.inventories.insert_record_stream(
2819
self.source.inventories.get_record_stream(
2820
self.source.inventories.keys(), 'topological', False))
2821
self.target.signatures.insert_record_stream(
2822
self.source.signatures.get_record_stream(
2823
self.source.signatures.keys(),
2825
self.target.revisions.insert_record_stream(
2826
self.source.revisions.get_record_stream(
2827
self.source.revisions.keys(),
2828
'topological', True))
2832
self.target.fetch(self.source, revision_id=revision_id)
2835
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2836
"""See InterRepository.fetch()."""
2837
from bzrlib.fetch import RepoFetcher
2838
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2839
self.source, self.source._format, self.target, self.target._format)
2840
f = RepoFetcher(to_repository=self.target,
2841
from_repository=self.source,
2842
last_revision=revision_id,
2843
pb=pb, find_ghosts=find_ghosts)
2844
return f.count_copied, f.failed_revisions
2847
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2848
"""See InterRepository.missing_revision_ids()."""
2849
# we want all revisions to satisfy revision_id in source.
2850
# but we don't want to stat every file here and there.
2851
# we want then, all revisions other needs to satisfy revision_id
2852
# checked, but not those that we have locally.
2853
# so the first thing is to get a subset of the revisions to
2854
# satisfy revision_id in source, and then eliminate those that
2855
# we do already have.
2856
# this is slow on high latency connection to self, but as as this
2857
# disk format scales terribly for push anyway due to rewriting
2858
# inventory.weave, this is considered acceptable.
2860
if revision_id is not None:
2861
source_ids = self.source.get_ancestry(revision_id)
2862
if source_ids[0] is not None:
2863
raise AssertionError()
2866
source_ids = self.source._all_possible_ids()
2867
source_ids_set = set(source_ids)
2868
# source_ids is the worst possible case we may need to pull.
2869
# now we want to filter source_ids against what we actually
2870
# have in target, but don't try to check for existence where we know
2871
# we do not have a revision as that would be pointless.
2872
target_ids = set(self.target._all_possible_ids())
2873
possibly_present_revisions = target_ids.intersection(source_ids_set)
2874
actually_present_revisions = set(
2875
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2876
required_revisions = source_ids_set.difference(actually_present_revisions)
2877
if revision_id is not None:
2878
# we used get_ancestry to determine source_ids then we are assured all
2879
# revisions referenced are present as they are installed in topological order.
2880
# and the tip revision was validated by get_ancestry.
2881
result_set = required_revisions
2883
# if we just grabbed the possibly available ids, then
2884
# we only have an estimate of whats available and need to validate
2885
# that against the revision records.
2887
self.source._eliminate_revisions_not_present(required_revisions))
2888
return self.source.revision_ids_to_search_result(result_set)
2891
class InterKnitRepo(InterSameDataRepository):
2892
"""Optimised code paths between Knit based repositories."""
2895
def _get_repo_format_to_test(self):
2896
from bzrlib.repofmt import knitrepo
2897
return knitrepo.RepositoryFormatKnit1()
2900
def is_compatible(source, target):
2901
"""Be compatible with known Knit formats.
2903
We don't test for the stores being of specific types because that
2904
could lead to confusing results, and there is no need to be
2907
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2909
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2910
isinstance(target._format, RepositoryFormatKnit))
2911
except AttributeError:
2913
return are_knits and InterRepository._same_model(source, target)
2916
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2917
"""See InterRepository.fetch()."""
2918
from bzrlib.fetch import RepoFetcher
2919
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2920
self.source, self.source._format, self.target, self.target._format)
2921
f = RepoFetcher(to_repository=self.target,
2922
from_repository=self.source,
2923
last_revision=revision_id,
2924
pb=pb, find_ghosts=find_ghosts)
2925
return f.count_copied, f.failed_revisions
2928
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2929
"""See InterRepository.missing_revision_ids()."""
2930
if revision_id is not None:
2931
source_ids = self.source.get_ancestry(revision_id)
2932
if source_ids[0] is not None:
2933
raise AssertionError()
2936
source_ids = self.source.all_revision_ids()
2937
source_ids_set = set(source_ids)
2938
# source_ids is the worst possible case we may need to pull.
2939
# now we want to filter source_ids against what we actually
2940
# have in target, but don't try to check for existence where we know
2941
# we do not have a revision as that would be pointless.
2942
target_ids = set(self.target.all_revision_ids())
2943
possibly_present_revisions = target_ids.intersection(source_ids_set)
2944
actually_present_revisions = set(
2945
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2946
required_revisions = source_ids_set.difference(actually_present_revisions)
2947
if revision_id is not None:
2948
# we used get_ancestry to determine source_ids then we are assured all
2949
# revisions referenced are present as they are installed in topological order.
2950
# and the tip revision was validated by get_ancestry.
2951
result_set = required_revisions
2953
# if we just grabbed the possibly available ids, then
2954
# we only have an estimate of whats available and need to validate
2955
# that against the revision records.
2957
self.source._eliminate_revisions_not_present(required_revisions))
2958
return self.source.revision_ids_to_search_result(result_set)
2961
class InterPackRepo(InterSameDataRepository):
2962
"""Optimised code paths between Pack based repositories."""
2965
def _get_repo_format_to_test(self):
2966
from bzrlib.repofmt import pack_repo
2967
return pack_repo.RepositoryFormatKnitPack1()
2970
def is_compatible(source, target):
2971
"""Be compatible with known Pack formats.
2973
We don't test for the stores being of specific types because that
2974
could lead to confusing results, and there is no need to be
2977
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2979
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2980
isinstance(target._format, RepositoryFormatPack))
2981
except AttributeError:
2983
return are_packs and InterRepository._same_model(source, target)
2986
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2987
"""See InterRepository.fetch()."""
2988
if (len(self.source._fallback_repositories) > 0 or
2989
len(self.target._fallback_repositories) > 0):
2990
# The pack layer is not aware of fallback repositories, so when
2991
# fetching from a stacked repository or into a stacked repository
2992
# we use the generic fetch logic which uses the VersionedFiles
2993
# attributes on repository.
2994
from bzrlib.fetch import RepoFetcher
2995
fetcher = RepoFetcher(self.target, self.source, revision_id,
2997
return fetcher.count_copied, fetcher.failed_revisions
2998
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2999
self.source, self.source._format, self.target, self.target._format)
3000
self.count_copied = 0
3001
if revision_id is None:
3003
# everything to do - use pack logic
3004
# to fetch from all packs to one without
3005
# inventory parsing etc, IFF nothing to be copied is in the target.
3007
source_revision_ids = frozenset(self.source.all_revision_ids())
3008
revision_ids = source_revision_ids - \
3009
frozenset(self.target_get_parent_map(source_revision_ids))
3010
revision_keys = [(revid,) for revid in revision_ids]
3011
target_pack_collection = self._get_target_pack_collection()
3012
index = target_pack_collection.revision_index.combined_index
3013
present_revision_ids = set(item[1][0] for item in
3014
index.iter_entries(revision_keys))
3015
revision_ids = set(revision_ids) - present_revision_ids
3016
# implementing the TODO will involve:
3017
# - detecting when all of a pack is selected
3018
# - avoiding as much as possible pre-selection, so the
3019
# more-core routines such as create_pack_from_packs can filter in
3020
# a just-in-time fashion. (though having a HEADS list on a
3021
# repository might make this a lot easier, because we could
3022
# sensibly detect 'new revisions' without doing a full index scan.
3023
elif _mod_revision.is_null(revision_id):
3028
revision_ids = self.search_missing_revision_ids(revision_id,
3029
find_ghosts=find_ghosts).get_keys()
3030
except errors.NoSuchRevision:
3031
raise errors.InstallFailed([revision_id])
3032
if len(revision_ids) == 0:
3034
return self._pack(self.source, self.target, revision_ids)
3036
def _pack(self, source, target, revision_ids):
3037
from bzrlib.repofmt.pack_repo import Packer
3038
target_pack_collection = self._get_target_pack_collection()
3039
packs = source._pack_collection.all_packs()
3040
pack = Packer(target_pack_collection, packs, '.fetch',
3041
revision_ids).pack()
3042
if pack is not None:
3043
target_pack_collection._save_pack_names()
3044
copied_revs = pack.get_revision_count()
3045
# Trigger an autopack. This may duplicate effort as we've just done
3046
# a pack creation, but for now it is simpler to think about as
3047
# 'upload data, then repack if needed'.
3049
return (copied_revs, [])
3053
def _autopack(self):
3054
self.target._pack_collection.autopack()
3056
def _get_target_pack_collection(self):
3057
return self.target._pack_collection
3060
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3061
"""See InterRepository.missing_revision_ids().
3063
:param find_ghosts: Find ghosts throughout the ancestry of
3066
if not find_ghosts and revision_id is not None:
3067
return self._walk_to_common_revisions([revision_id])
3068
elif revision_id is not None:
3069
# Find ghosts: search for revisions pointing from one repository to
3070
# the other, and vice versa, anywhere in the history of revision_id.
3071
graph = self.target_get_graph(other_repository=self.source)
3072
searcher = graph._make_breadth_first_searcher([revision_id])
3076
next_revs, ghosts = searcher.next_with_ghosts()
3077
except StopIteration:
3079
if revision_id in ghosts:
3080
raise errors.NoSuchRevision(self.source, revision_id)
3081
found_ids.update(next_revs)
3082
found_ids.update(ghosts)
3083
found_ids = frozenset(found_ids)
3084
# Double query here: should be able to avoid this by changing the
3085
# graph api further.
3086
result_set = found_ids - frozenset(
3087
self.target_get_parent_map(found_ids))
3089
source_ids = self.source.all_revision_ids()
3090
# source_ids is the worst possible case we may need to pull.
3091
# now we want to filter source_ids against what we actually
3092
# have in target, but don't try to check for existence where we know
3093
# we do not have a revision as that would be pointless.
3094
target_ids = set(self.target.all_revision_ids())
3095
result_set = set(source_ids).difference(target_ids)
3096
return self.source.revision_ids_to_search_result(result_set)
3099
class InterModel1and2(InterRepository):
3102
def _get_repo_format_to_test(self):
3106
def is_compatible(source, target):
3107
if not source.supports_rich_root() and target.supports_rich_root():
3113
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3114
"""See InterRepository.fetch()."""
3115
from bzrlib.fetch import Model1toKnit2Fetcher
3116
f = Model1toKnit2Fetcher(to_repository=self.target,
3117
from_repository=self.source,
3118
last_revision=revision_id,
3119
pb=pb, find_ghosts=find_ghosts)
3120
return f.count_copied, f.failed_revisions
3123
def copy_content(self, revision_id=None):
3124
"""Make a complete copy of the content in self into destination.
3126
This is a destructive operation! Do not use it on existing
3129
:param revision_id: Only copy the content needed to construct
3130
revision_id and its parents.
3133
self.target.set_make_working_trees(self.source.make_working_trees())
3134
except NotImplementedError:
3136
# but don't bother fetching if we have the needed data now.
3137
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
3138
self.target.has_revision(revision_id)):
3140
self.target.fetch(self.source, revision_id=revision_id)
3143
class InterKnit1and2(InterKnitRepo):
3146
def _get_repo_format_to_test(self):
3150
def is_compatible(source, target):
3151
"""Be compatible with Knit1 source and Knit3 target"""
3153
from bzrlib.repofmt.knitrepo import (
3154
RepositoryFormatKnit1,
3155
RepositoryFormatKnit3,
3157
from bzrlib.repofmt.pack_repo import (
3158
RepositoryFormatKnitPack1,
3159
RepositoryFormatKnitPack3,
3160
RepositoryFormatKnitPack4,
3161
RepositoryFormatKnitPack5,
3162
RepositoryFormatKnitPack5RichRoot,
3163
RepositoryFormatKnitPack6,
3164
RepositoryFormatKnitPack6RichRoot,
3165
RepositoryFormatPackDevelopment2,
3166
RepositoryFormatPackDevelopment2Subtree,
3169
RepositoryFormatKnit1, # no rr, no subtree
3170
RepositoryFormatKnitPack1, # no rr, no subtree
3171
RepositoryFormatPackDevelopment2, # no rr, no subtree
3172
RepositoryFormatKnitPack5, # no rr, no subtree
3173
RepositoryFormatKnitPack6, # no rr, no subtree
3176
RepositoryFormatKnit3, # rr, subtree
3177
RepositoryFormatKnitPack3, # rr, subtree
3178
RepositoryFormatKnitPack4, # rr, no subtree
3179
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
3180
RepositoryFormatKnitPack6RichRoot,# rr, no subtree
3181
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
3183
for format in norichroot:
3184
if format.rich_root_data:
3185
raise AssertionError('Format %s is a rich-root format'
3186
' but is included in the non-rich-root list'
3188
for format in richroot:
3189
if not format.rich_root_data:
3190
raise AssertionError('Format %s is not a rich-root format'
3191
' but is included in the rich-root list'
3193
# TODO: One alternative is to just check format.rich_root_data,
3194
# instead of keeping membership lists. However, the formats
3195
# *also* have to use the same 'Knit' style of storage
3196
# (line-deltas, fulltexts, etc.)
3197
return (isinstance(source._format, norichroot) and
3198
isinstance(target._format, richroot))
3199
except AttributeError:
3203
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3204
"""See InterRepository.fetch()."""
3205
from bzrlib.fetch import Knit1to2Fetcher
3206
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3207
self.source, self.source._format, self.target,
3208
self.target._format)
3209
f = Knit1to2Fetcher(to_repository=self.target,
3210
from_repository=self.source,
3211
last_revision=revision_id,
3212
pb=pb, find_ghosts=find_ghosts)
3213
return f.count_copied, f.failed_revisions
3216
class InterDifferingSerializer(InterKnitRepo):
3219
def _get_repo_format_to_test(self):
3223
def is_compatible(source, target):
3224
"""Be compatible with Knit2 source and Knit3 target"""
3225
if source.supports_rich_root() != target.supports_rich_root():
3227
# Ideally, we'd support fetching if the source had no tree references
3228
# even if it supported them...
3229
if (getattr(source, '_format.supports_tree_reference', False) and
3230
not getattr(target, '_format.supports_tree_reference', False)):
3234
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3235
"""Get the best delta and base for this revision.
3237
:return: (basis_id, delta)
3239
possible_trees = [(parent_id, cache[parent_id])
3240
for parent_id in parent_ids
3241
if parent_id in cache]
3242
if len(possible_trees) == 0:
3243
# There either aren't any parents, or the parents aren't in the
3244
# cache, so just use the last converted tree
3245
possible_trees.append((basis_id, cache[basis_id]))
3247
for basis_id, basis_tree in possible_trees:
3248
delta = tree.inventory._make_delta(basis_tree.inventory)
3249
deltas.append((len(delta), basis_id, delta))
3251
return deltas[0][1:]
3253
def _fetch_batch(self, revision_ids, basis_id, cache):
3254
"""Fetch across a few revisions.
3256
:param revision_ids: The revisions to copy
3257
:param basis_id: The revision_id of a tree that must be in cache, used
3258
as a basis for delta when no other base is available
3259
:param cache: A cache of RevisionTrees that we can use.
3260
:return: The revision_id of the last converted tree. The RevisionTree
3261
for it will be in cache
3263
# Walk though all revisions; get inventory deltas, copy referenced
3264
# texts that delta references, insert the delta, revision and
3268
pending_revisions = []
3269
parent_map = self.source.get_parent_map(revision_ids)
3270
for tree in self.source.revision_trees(revision_ids):
3271
current_revision_id = tree.get_revision_id()
3272
parent_ids = parent_map.get(current_revision_id, ())
3273
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3275
# Find text entries that need to be copied
3276
for old_path, new_path, file_id, entry in delta:
3277
if new_path is not None:
3278
if not (new_path or self.target.supports_rich_root()):
3279
# We don't copy the text for the root node unless the
3280
# target supports_rich_root.
3283
# "if entry.revision == current_revision_id" ?
3284
if entry.revision == current_revision_id:
3285
text_keys.add((file_id, entry.revision))
3286
revision = self.source.get_revision(current_revision_id)
3287
pending_deltas.append((basis_id, delta,
3288
current_revision_id, revision.parent_ids))
3289
pending_revisions.append(revision)
3290
cache[current_revision_id] = tree
3291
basis_id = current_revision_id
3293
from_texts = self.source.texts
3294
to_texts = self.target.texts
3295
to_texts.insert_record_stream(from_texts.get_record_stream(
3296
text_keys, self.target._fetch_order,
3297
not self.target._fetch_uses_deltas))
3299
for delta in pending_deltas:
3300
self.target.add_inventory_by_delta(*delta)
3301
# insert signatures and revisions
3302
for revision in pending_revisions:
3304
signature = self.source.get_signature_text(
3305
revision.revision_id)
3306
self.target.add_signature_text(revision.revision_id,
3308
except errors.NoSuchRevision:
3310
self.target.add_revision(revision.revision_id, revision)
3313
def _fetch_all_revisions(self, revision_ids, pb):
3314
"""Fetch everything for the list of revisions.
3316
:param revision_ids: The list of revisions to fetch. Must be in
3318
:param pb: A ProgressBar
3321
basis_id, basis_tree = self._get_basis(revision_ids[0])
3323
cache = lru_cache.LRUCache(100)
3324
cache[basis_id] = basis_tree
3325
del basis_tree # We don't want to hang on to it here
3326
for offset in range(0, len(revision_ids), batch_size):
3327
self.target.start_write_group()
3329
pb.update('Transferring revisions', offset,
3331
batch = revision_ids[offset:offset+batch_size]
3332
basis_id = self._fetch_batch(batch, basis_id, cache)
3334
self.target.abort_write_group()
3337
self.target.commit_write_group()
3338
pb.update('Transferring revisions', len(revision_ids),
3342
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3343
"""See InterRepository.fetch()."""
3344
revision_ids = self.target.search_missing_revision_ids(self.source,
3345
revision_id, find_ghosts=find_ghosts).get_keys()
3346
if not revision_ids:
3348
revision_ids = tsort.topo_sort(
3349
self.source.get_graph().get_parent_map(revision_ids))
3351
my_pb = ui.ui_factory.nested_progress_bar()
3356
self._fetch_all_revisions(revision_ids, pb)
3358
if my_pb is not None:
3360
return len(revision_ids), 0
3362
def _get_basis(self, first_revision_id):
3363
"""Get a revision and tree which exists in the target.
3365
This assumes that first_revision_id is selected for transmission
3366
because all other ancestors are already present. If we can't find an
3367
ancestor we fall back to NULL_REVISION since we know that is safe.
3369
:return: (basis_id, basis_tree)
3371
first_rev = self.source.get_revision(first_revision_id)
3373
basis_id = first_rev.parent_ids[0]
3374
# only valid as a basis if the target has it
3375
self.target.get_revision(basis_id)
3376
# Try to get a basis tree - if its a ghost it will hit the
3377
# NoSuchRevision case.
3378
basis_tree = self.source.revision_tree(basis_id)
3379
except (IndexError, errors.NoSuchRevision):
3380
basis_id = _mod_revision.NULL_REVISION
3381
basis_tree = self.source.revision_tree(basis_id)
3382
return basis_id, basis_tree
3385
class InterOtherToRemote(InterRepository):
3386
"""An InterRepository that simply delegates to the 'real' InterRepository
3387
calculated for (source, target._real_repository).
3390
_walk_to_common_revisions_batch_size = 50
3392
def __init__(self, source, target):
3393
InterRepository.__init__(self, source, target)
3394
self._real_inter = None
3397
def is_compatible(source, target):
3398
if isinstance(target, remote.RemoteRepository):
3402
def _ensure_real_inter(self):
3403
if self._real_inter is None:
3404
self.target._ensure_real()
3405
real_target = self.target._real_repository
3406
self._real_inter = InterRepository.get(self.source, real_target)
3407
# Make _real_inter use the RemoteRepository for get_parent_map
3408
self._real_inter.target_get_graph = self.target.get_graph
3409
self._real_inter.target_get_parent_map = self.target.get_parent_map
3411
def copy_content(self, revision_id=None):
3412
self._ensure_real_inter()
3413
self._real_inter.copy_content(revision_id=revision_id)
3415
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3416
self._ensure_real_inter()
3417
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3418
find_ghosts=find_ghosts)
3421
def _get_repo_format_to_test(self):
3425
class InterRemoteToOther(InterRepository):
3427
def __init__(self, source, target):
3428
InterRepository.__init__(self, source, target)
3429
self._real_inter = None
3432
def is_compatible(source, target):
3433
if not isinstance(source, remote.RemoteRepository):
3435
# Is source's model compatible with target's model?
3436
source._ensure_real()
3437
real_source = source._real_repository
3438
if isinstance(real_source, remote.RemoteRepository):
3439
raise NotImplementedError(
3440
"We don't support remote repos backed by remote repos yet.")
3441
return InterRepository._same_model(real_source, target)
3443
def _ensure_real_inter(self):
3444
if self._real_inter is None:
3445
self.source._ensure_real()
3446
real_source = self.source._real_repository
3447
self._real_inter = InterRepository.get(real_source, self.target)
3449
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3450
self._ensure_real_inter()
3451
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3452
find_ghosts=find_ghosts)
3454
def copy_content(self, revision_id=None):
3455
self._ensure_real_inter()
3456
self._real_inter.copy_content(revision_id=revision_id)
3459
def _get_repo_format_to_test(self):
3464
class InterPackToRemotePack(InterPackRepo):
3465
"""A specialisation of InterPackRepo for a target that is a
3468
This will use the get_parent_map RPC rather than plain readvs, and also
3469
uses an RPC for autopacking.
3472
_walk_to_common_revisions_batch_size = 50
3475
def is_compatible(source, target):
3476
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3477
if isinstance(source._format, RepositoryFormatPack):
3478
if isinstance(target, remote.RemoteRepository):
3479
target._ensure_real()
3480
if isinstance(target._real_repository._format,
3481
RepositoryFormatPack):
3482
if InterRepository._same_model(source, target):
3486
def _autopack(self):
3487
self.target.autopack()
3490
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3491
"""See InterRepository.fetch()."""
3492
# Always fetch using the generic streaming fetch code, to allow
3493
# streaming fetching into remote servers.
3494
from bzrlib.fetch import RepoFetcher
3495
fetcher = RepoFetcher(self.target, self.source, revision_id,
3497
return fetcher.count_copied, fetcher.failed_revisions
3499
def _get_target_pack_collection(self):
3500
return self.target._real_repository._pack_collection
3503
def _get_repo_format_to_test(self):
3507
InterRepository.register_optimiser(InterDifferingSerializer)
3508
InterRepository.register_optimiser(InterSameDataRepository)
3509
InterRepository.register_optimiser(InterWeaveRepo)
3510
InterRepository.register_optimiser(InterKnitRepo)
3511
InterRepository.register_optimiser(InterModel1and2)
3512
InterRepository.register_optimiser(InterKnit1and2)
3513
InterRepository.register_optimiser(InterPackRepo)
3514
InterRepository.register_optimiser(InterOtherToRemote)
3515
InterRepository.register_optimiser(InterRemoteToOther)
3516
InterRepository.register_optimiser(InterPackToRemotePack)
3519
class CopyConverter(object):
3520
"""A repository conversion tool which just performs a copy of the content.
3522
This is slow but quite reliable.
3525
def __init__(self, target_format):
3526
"""Create a CopyConverter.
3528
:param target_format: The format the resulting repository should be.
3530
self.target_format = target_format
3532
def convert(self, repo, pb):
3533
"""Perform the conversion of to_convert, giving feedback via pb.
3535
:param to_convert: The disk object to convert.
3536
:param pb: a progress bar to use for progress information.
3541
# this is only useful with metadir layouts - separated repo content.
3542
# trigger an assertion if not such
3543
repo._format.get_format_string()
3544
self.repo_dir = repo.bzrdir
3545
self.step('Moving repository to repository.backup')
3546
self.repo_dir.transport.move('repository', 'repository.backup')
3547
backup_transport = self.repo_dir.transport.clone('repository.backup')
3548
repo._format.check_conversion_target(self.target_format)
3549
self.source_repo = repo._format.open(self.repo_dir,
3551
_override_transport=backup_transport)
3552
self.step('Creating new repository')
3553
converted = self.target_format.initialize(self.repo_dir,
3554
self.source_repo.is_shared())
3555
converted.lock_write()
3557
self.step('Copying content into repository.')
3558
self.source_repo.copy_content_into(converted)
3561
self.step('Deleting old repository content.')
3562
self.repo_dir.transport.delete_tree('repository.backup')
3563
self.pb.note('repository converted')
3565
def step(self, message):
3566
"""Update the pb by a step."""
3568
self.pb.update(message, self.count, self.total)
3580
def _unescaper(match, _map=_unescape_map):
3581
code = match.group(1)
3585
if not code.startswith('#'):
3587
return unichr(int(code[1:])).encode('utf8')
3593
def _unescape_xml(data):
3594
"""Unescape predefined XML entities in a string of data."""
3596
if _unescape_re is None:
3597
_unescape_re = re.compile('\&([^;]*);')
3598
return _unescape_re.sub(_unescaper, data)
3601
class _VersionedFileChecker(object):
3603
def __init__(self, repository):
3604
self.repository = repository
3605
self.text_index = self.repository._generate_text_key_index()
3607
def calculate_file_version_parents(self, text_key):
3608
"""Calculate the correct parents for a file version according to
3611
parent_keys = self.text_index[text_key]
3612
if parent_keys == [_mod_revision.NULL_REVISION]:
3614
return tuple(parent_keys)
3616
def check_file_version_parents(self, texts, progress_bar=None):
3617
"""Check the parents stored in a versioned file are correct.
3619
It also detects file versions that are not referenced by their
3620
corresponding revision's inventory.
3622
:returns: A tuple of (wrong_parents, dangling_file_versions).
3623
wrong_parents is a dict mapping {revision_id: (stored_parents,
3624
correct_parents)} for each revision_id where the stored parents
3625
are not correct. dangling_file_versions is a set of (file_id,
3626
revision_id) tuples for versions that are present in this versioned
3627
file, but not used by the corresponding inventory.
3630
self.file_ids = set([file_id for file_id, _ in
3631
self.text_index.iterkeys()])
3632
# text keys is now grouped by file_id
3633
n_weaves = len(self.file_ids)
3634
files_in_revisions = {}
3635
revisions_of_files = {}
3636
n_versions = len(self.text_index)
3637
progress_bar.update('loading text store', 0, n_versions)
3638
parent_map = self.repository.texts.get_parent_map(self.text_index)
3639
# On unlistable transports this could well be empty/error...
3640
text_keys = self.repository.texts.keys()
3641
unused_keys = frozenset(text_keys) - set(self.text_index)
3642
for num, key in enumerate(self.text_index.iterkeys()):
3643
if progress_bar is not None:
3644
progress_bar.update('checking text graph', num, n_versions)
3645
correct_parents = self.calculate_file_version_parents(key)
3647
knit_parents = parent_map[key]
3648
except errors.RevisionNotPresent:
3651
if correct_parents != knit_parents:
3652
wrong_parents[key] = (knit_parents, correct_parents)
3653
return wrong_parents, unused_keys
3656
def _old_get_graph(repository, revision_id):
3657
"""DO NOT USE. That is all. I'm serious."""
3658
graph = repository.get_graph()
3659
revision_graph = dict(((key, value) for key, value in
3660
graph.iter_ancestry([revision_id]) if value is not None))
3661
return _strip_NULL_ghosts(revision_graph)
3664
def _strip_NULL_ghosts(revision_graph):
3665
"""Also don't use this. more compatibility code for unmigrated clients."""
3666
# Filter ghosts, and null:
3667
if _mod_revision.NULL_REVISION in revision_graph:
3668
del revision_graph[_mod_revision.NULL_REVISION]
3669
for key, parents in revision_graph.items():
3670
revision_graph[key] = tuple(parent for parent in parents if parent
3672
return revision_graph
3675
class StreamSink(object):
3676
"""An object that can insert a stream into a repository.
3678
This interface handles the complexity of reserialising inventories and
3679
revisions from different formats, and allows unidirectional insertion into
3680
stacked repositories without looking for the missing basis parents
3684
def __init__(self, target_repo):
3685
self.target_repo = target_repo
3687
def insert_stream(self, stream, src_format, resume_tokens):
3688
"""Insert a stream's content into the target repository.
3690
:param src_format: a bzr repository format.
3692
:return: a list of resume tokens and an iterable of keys additional
3693
items required before the insertion can be completed.
3695
self.target_repo.lock_write()
3698
self.target_repo.resume_write_group(resume_tokens)
3700
self.target_repo.start_write_group()
3702
# locked_insert_stream performs a commit|suspend.
3703
return self._locked_insert_stream(stream, src_format)
3705
self.target_repo.abort_write_group(suppress_errors=True)
3708
self.target_repo.unlock()
3710
def _locked_insert_stream(self, stream, src_format):
3711
to_serializer = self.target_repo._format._serializer
3712
src_serializer = src_format._serializer
3713
for substream_type, substream in stream:
3714
if substream_type == 'texts':
3715
self.target_repo.texts.insert_record_stream(substream)
3716
elif substream_type == 'inventories':
3717
if src_serializer == to_serializer:
3718
self.target_repo.inventories.insert_record_stream(
3721
self._extract_and_insert_inventories(
3722
substream, src_serializer)
3723
elif substream_type == 'revisions':
3724
# This may fallback to extract-and-insert more often than
3725
# required if the serializers are different only in terms of
3727
if src_serializer == to_serializer:
3728
self.target_repo.revisions.insert_record_stream(
3731
self._extract_and_insert_revisions(substream,
3733
elif substream_type == 'signatures':
3734
self.target_repo.signatures.insert_record_stream(substream)
3736
raise AssertionError('kaboom! %s' % (substream_type,))
3738
missing_keys = set()
3739
for prefix, versioned_file in (
3740
('texts', self.target_repo.texts),
3741
('inventories', self.target_repo.inventories),
3742
('revisions', self.target_repo.revisions),
3743
('signatures', self.target_repo.signatures),
3745
missing_keys.update((prefix,) + key for key in
3746
versioned_file.get_missing_compression_parent_keys())
3747
except NotImplementedError:
3748
# cannot even attempt suspending, and missing would have failed
3749
# during stream insertion.
3750
missing_keys = set()
3753
# suspend the write group and tell the caller what we is
3754
# missing. We know we can suspend or else we would not have
3755
# entered this code path. (All repositories that can handle
3756
# missing keys can handle suspending a write group).
3757
write_group_tokens = self.target_repo.suspend_write_group()
3758
return write_group_tokens, missing_keys
3759
self.target_repo.commit_write_group()
3762
def _extract_and_insert_inventories(self, substream, serializer):
3763
"""Generate a new inventory versionedfile in target, converting data.
3765
The inventory is retrieved from the source, (deserializing it), and
3766
stored in the target (reserializing it in a different format).
3768
for record in substream:
3769
bytes = record.get_bytes_as('fulltext')
3770
revision_id = record.key[0]
3771
inv = serializer.read_inventory_from_string(bytes, revision_id)
3772
parents = [key[0] for key in record.parents]
3773
self.target_repo.add_inventory(revision_id, inv, parents)
3775
def _extract_and_insert_revisions(self, substream, serializer):
3776
for record in substream:
3777
bytes = record.get_bytes_as('fulltext')
3778
revision_id = record.key[0]
3779
rev = serializer.read_revision_from_string(bytes)
3780
if rev.revision_id != revision_id:
3781
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
3782
self.target_repo.add_revision(revision_id, rev)
3785
if self.target_repo._fetch_reconcile:
3786
self.target_repo.reconcile()
added
removed
bzrlib/repository.py
