1
# Copyright (C) 2005, 2006, 2007, 2008, 2009 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
from bzrlib.lazy_import import lazy_import
18
lazy_import(globals(), """
40
revision as _mod_revision,
46
from bzrlib.bundle import serializer
47
from bzrlib.revisiontree import RevisionTree
48
from bzrlib.store.versioned import VersionedFileStore
49
from bzrlib.testament import Testament
52
from bzrlib.decorators import needs_read_lock, needs_write_lock
53
from bzrlib.inter import InterObject
54
from bzrlib.inventory import (
60
from bzrlib import registry
61
from bzrlib.trace import (
62
log_exception_quietly, note, mutter, mutter_callsite, warning)
65
# Old formats display a warning, but only once
66
_deprecation_warning_done = False
69
class CommitBuilder(object):
70
"""Provides an interface to build up a commit.
72
This allows describing a tree to be committed without needing to
73
know the internals of the format of the repository.
76
# all clients should supply tree roots.
77
record_root_entry = True
78
# the default CommitBuilder does not manage trees whose root is versioned.
79
_versioned_root = False
81
def __init__(self, repository, parents, config, timestamp=None,
82
timezone=None, committer=None, revprops=None,
84
"""Initiate a CommitBuilder.
86
:param repository: Repository to commit to.
87
:param parents: Revision ids of the parents of the new revision.
88
:param config: Configuration to use.
89
:param timestamp: Optional timestamp recorded for commit.
90
:param timezone: Optional timezone for timestamp.
91
:param committer: Optional committer to set for commit.
92
:param revprops: Optional dictionary of revision properties.
93
:param revision_id: Optional revision id.
98
self._committer = self._config.username()
100
self._committer = committer
102
self.new_inventory = Inventory(None)
103
self._new_revision_id = revision_id
104
self.parents = parents
105
self.repository = repository
108
if revprops is not None:
109
self._validate_revprops(revprops)
110
self._revprops.update(revprops)
112
if timestamp is None:
113
timestamp = time.time()
114
# Restrict resolution to 1ms
115
self._timestamp = round(timestamp, 3)
118
self._timezone = osutils.local_time_offset()
120
self._timezone = int(timezone)
122
self._generate_revision_if_needed()
123
self.__heads = graph.HeadsCache(repository.get_graph()).heads
124
self._basis_delta = []
125
# API compatibility, older code that used CommitBuilder did not call
126
# .record_delete(), which means the delta that is computed would not be
127
# valid. Callers that will call record_delete() should call
128
# .will_record_deletes() to indicate that.
129
self._recording_deletes = False
130
# memo'd check for no-op commits.
131
self._any_changes = False
133
def any_changes(self):
134
"""Return True if any entries were changed.
136
This includes merge-only changes. It is the core for the --unchanged
139
:return: True if any changes have occured.
141
return self._any_changes
143
def _validate_unicode_text(self, text, context):
144
"""Verify things like commit messages don't have bogus characters."""
146
raise ValueError('Invalid value for %s: %r' % (context, text))
148
def _validate_revprops(self, revprops):
149
for key, value in revprops.iteritems():
150
# We know that the XML serializers do not round trip '\r'
151
# correctly, so refuse to accept them
152
if not isinstance(value, basestring):
153
raise ValueError('revision property (%s) is not a valid'
154
' (unicode) string: %r' % (key, value))
155
self._validate_unicode_text(value,
156
'revision property (%s)' % (key,))
158
def commit(self, message):
159
"""Make the actual commit.
161
:return: The revision id of the recorded revision.
163
self._validate_unicode_text(message, 'commit message')
164
rev = _mod_revision.Revision(
165
timestamp=self._timestamp,
166
timezone=self._timezone,
167
committer=self._committer,
169
inventory_sha1=self.inv_sha1,
170
revision_id=self._new_revision_id,
171
properties=self._revprops)
172
rev.parent_ids = self.parents
173
self.repository.add_revision(self._new_revision_id, rev,
174
self.new_inventory, self._config)
175
self.repository.commit_write_group()
176
return self._new_revision_id
179
"""Abort the commit that is being built.
181
self.repository.abort_write_group()
183
def revision_tree(self):
184
"""Return the tree that was just committed.
186
After calling commit() this can be called to get a RevisionTree
187
representing the newly committed tree. This is preferred to
188
calling Repository.revision_tree() because that may require
189
deserializing the inventory, while we already have a copy in
192
if self.new_inventory is None:
193
self.new_inventory = self.repository.get_inventory(
194
self._new_revision_id)
195
return RevisionTree(self.repository, self.new_inventory,
196
self._new_revision_id)
198
def finish_inventory(self):
199
"""Tell the builder that the inventory is finished.
201
:return: The inventory id in the repository, which can be used with
202
repository.get_inventory.
204
if self.new_inventory is None:
205
# an inventory delta was accumulated without creating a new
207
basis_id = self.basis_delta_revision
208
self.inv_sha1 = self.repository.add_inventory_by_delta(
209
basis_id, self._basis_delta, self._new_revision_id,
212
if self.new_inventory.root is None:
213
raise AssertionError('Root entry should be supplied to'
214
' record_entry_contents, as of bzr 0.10.')
215
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
216
self.new_inventory.revision_id = self._new_revision_id
217
self.inv_sha1 = self.repository.add_inventory(
218
self._new_revision_id,
222
return self._new_revision_id
224
def _gen_revision_id(self):
225
"""Return new revision-id."""
226
return generate_ids.gen_revision_id(self._config.username(),
229
def _generate_revision_if_needed(self):
230
"""Create a revision id if None was supplied.
232
If the repository can not support user-specified revision ids
233
they should override this function and raise CannotSetRevisionId
234
if _new_revision_id is not None.
236
:raises: CannotSetRevisionId
238
if self._new_revision_id is None:
239
self._new_revision_id = self._gen_revision_id()
240
self.random_revid = True
242
self.random_revid = False
244
def _heads(self, file_id, revision_ids):
245
"""Calculate the graph heads for revision_ids in the graph of file_id.
247
This can use either a per-file graph or a global revision graph as we
248
have an identity relationship between the two graphs.
250
return self.__heads(revision_ids)
252
def _check_root(self, ie, parent_invs, tree):
253
"""Helper for record_entry_contents.
255
:param ie: An entry being added.
256
:param parent_invs: The inventories of the parent revisions of the
258
:param tree: The tree that is being committed.
260
# In this revision format, root entries have no knit or weave When
261
# serializing out to disk and back in root.revision is always
263
ie.revision = self._new_revision_id
265
def _require_root_change(self, tree):
266
"""Enforce an appropriate root object change.
268
This is called once when record_iter_changes is called, if and only if
269
the root was not in the delta calculated by record_iter_changes.
271
:param tree: The tree which is being committed.
273
# NB: if there are no parents then this method is not called, so no
274
# need to guard on parents having length.
275
entry = entry_factory['directory'](tree.path2id(''), '',
277
entry.revision = self._new_revision_id
278
self._basis_delta.append(('', '', entry.file_id, entry))
280
def _get_delta(self, ie, basis_inv, path):
281
"""Get a delta against the basis inventory for ie."""
282
if ie.file_id not in basis_inv:
284
result = (None, path, ie.file_id, ie)
285
self._basis_delta.append(result)
287
elif ie != basis_inv[ie.file_id]:
289
# TODO: avoid tis id2path call.
290
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
291
self._basis_delta.append(result)
297
def get_basis_delta(self):
298
"""Return the complete inventory delta versus the basis inventory.
300
This has been built up with the calls to record_delete and
301
record_entry_contents. The client must have already called
302
will_record_deletes() to indicate that they will be generating a
305
:return: An inventory delta, suitable for use with apply_delta, or
306
Repository.add_inventory_by_delta, etc.
308
if not self._recording_deletes:
309
raise AssertionError("recording deletes not activated.")
310
return self._basis_delta
312
def record_delete(self, path, file_id):
313
"""Record that a delete occured against a basis tree.
315
This is an optional API - when used it adds items to the basis_delta
316
being accumulated by the commit builder. It cannot be called unless the
317
method will_record_deletes() has been called to inform the builder that
318
a delta is being supplied.
320
:param path: The path of the thing deleted.
321
:param file_id: The file id that was deleted.
323
if not self._recording_deletes:
324
raise AssertionError("recording deletes not activated.")
325
delta = (path, None, file_id, None)
326
self._basis_delta.append(delta)
327
self._any_changes = True
330
def will_record_deletes(self):
331
"""Tell the commit builder that deletes are being notified.
333
This enables the accumulation of an inventory delta; for the resulting
334
commit to be valid, deletes against the basis MUST be recorded via
335
builder.record_delete().
337
self._recording_deletes = True
339
basis_id = self.parents[0]
341
basis_id = _mod_revision.NULL_REVISION
342
self.basis_delta_revision = basis_id
344
def record_entry_contents(self, ie, parent_invs, path, tree,
346
"""Record the content of ie from tree into the commit if needed.
348
Side effect: sets ie.revision when unchanged
350
:param ie: An inventory entry present in the commit.
351
:param parent_invs: The inventories of the parent revisions of the
353
:param path: The path the entry is at in the tree.
354
:param tree: The tree which contains this entry and should be used to
356
:param content_summary: Summary data from the tree about the paths
357
content - stat, length, exec, sha/link target. This is only
358
accessed when the entry has a revision of None - that is when it is
359
a candidate to commit.
360
:return: A tuple (change_delta, version_recorded, fs_hash).
361
change_delta is an inventory_delta change for this entry against
362
the basis tree of the commit, or None if no change occured against
364
version_recorded is True if a new version of the entry has been
365
recorded. For instance, committing a merge where a file was only
366
changed on the other side will return (delta, False).
367
fs_hash is either None, or the hash details for the path (currently
368
a tuple of the contents sha1 and the statvalue returned by
369
tree.get_file_with_stat()).
371
if self.new_inventory.root is None:
372
if ie.parent_id is not None:
373
raise errors.RootMissing()
374
self._check_root(ie, parent_invs, tree)
375
if ie.revision is None:
376
kind = content_summary[0]
378
# ie is carried over from a prior commit
380
# XXX: repository specific check for nested tree support goes here - if
381
# the repo doesn't want nested trees we skip it ?
382
if (kind == 'tree-reference' and
383
not self.repository._format.supports_tree_reference):
384
# mismatch between commit builder logic and repository:
385
# this needs the entry creation pushed down into the builder.
386
raise NotImplementedError('Missing repository subtree support.')
387
self.new_inventory.add(ie)
389
# TODO: slow, take it out of the inner loop.
391
basis_inv = parent_invs[0]
393
basis_inv = Inventory(root_id=None)
395
# ie.revision is always None if the InventoryEntry is considered
396
# for committing. We may record the previous parents revision if the
397
# content is actually unchanged against a sole head.
398
if ie.revision is not None:
399
if not self._versioned_root and path == '':
400
# repositories that do not version the root set the root's
401
# revision to the new commit even when no change occurs (more
402
# specifically, they do not record a revision on the root; and
403
# the rev id is assigned to the root during deserialisation -
404
# this masks when a change may have occurred against the basis.
405
# To match this we always issue a delta, because the revision
406
# of the root will always be changing.
407
if ie.file_id in basis_inv:
408
delta = (basis_inv.id2path(ie.file_id), path,
412
delta = (None, path, ie.file_id, ie)
413
self._basis_delta.append(delta)
414
return delta, False, None
416
# we don't need to commit this, because the caller already
417
# determined that an existing revision of this file is
418
# appropriate. If its not being considered for committing then
419
# it and all its parents to the root must be unaltered so
420
# no-change against the basis.
421
if ie.revision == self._new_revision_id:
422
raise AssertionError("Impossible situation, a skipped "
423
"inventory entry (%r) claims to be modified in this "
424
"commit (%r).", (ie, self._new_revision_id))
425
return None, False, None
426
# XXX: Friction: parent_candidates should return a list not a dict
427
# so that we don't have to walk the inventories again.
428
parent_candiate_entries = ie.parent_candidates(parent_invs)
429
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
431
for inv in parent_invs:
432
if ie.file_id in inv:
433
old_rev = inv[ie.file_id].revision
434
if old_rev in head_set:
435
heads.append(inv[ie.file_id].revision)
436
head_set.remove(inv[ie.file_id].revision)
439
# now we check to see if we need to write a new record to the
441
# We write a new entry unless there is one head to the ancestors, and
442
# the kind-derived content is unchanged.
444
# Cheapest check first: no ancestors, or more the one head in the
445
# ancestors, we write a new node.
449
# There is a single head, look it up for comparison
450
parent_entry = parent_candiate_entries[heads[0]]
451
# if the non-content specific data has changed, we'll be writing a
453
if (parent_entry.parent_id != ie.parent_id or
454
parent_entry.name != ie.name):
456
# now we need to do content specific checks:
458
# if the kind changed the content obviously has
459
if kind != parent_entry.kind:
461
# Stat cache fingerprint feedback for the caller - None as we usually
462
# don't generate one.
465
if content_summary[2] is None:
466
raise ValueError("Files must not have executable = None")
468
if (# if the file length changed we have to store:
469
parent_entry.text_size != content_summary[1] or
470
# if the exec bit has changed we have to store:
471
parent_entry.executable != content_summary[2]):
473
elif parent_entry.text_sha1 == content_summary[3]:
474
# all meta and content is unchanged (using a hash cache
475
# hit to check the sha)
476
ie.revision = parent_entry.revision
477
ie.text_size = parent_entry.text_size
478
ie.text_sha1 = parent_entry.text_sha1
479
ie.executable = parent_entry.executable
480
return self._get_delta(ie, basis_inv, path), False, None
482
# Either there is only a hash change(no hash cache entry,
483
# or same size content change), or there is no change on
485
# Provide the parent's hash to the store layer, so that the
486
# content is unchanged we will not store a new node.
487
nostore_sha = parent_entry.text_sha1
489
# We want to record a new node regardless of the presence or
490
# absence of a content change in the file.
492
ie.executable = content_summary[2]
493
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
495
text = file_obj.read()
499
ie.text_sha1, ie.text_size = self._add_text_to_weave(
500
ie.file_id, text, heads, nostore_sha)
501
# Let the caller know we generated a stat fingerprint.
502
fingerprint = (ie.text_sha1, stat_value)
503
except errors.ExistingContent:
504
# Turns out that the file content was unchanged, and we were
505
# only going to store a new node if it was changed. Carry over
507
ie.revision = parent_entry.revision
508
ie.text_size = parent_entry.text_size
509
ie.text_sha1 = parent_entry.text_sha1
510
ie.executable = parent_entry.executable
511
return self._get_delta(ie, basis_inv, path), False, None
512
elif kind == 'directory':
514
# all data is meta here, nothing specific to directory, so
516
ie.revision = parent_entry.revision
517
return self._get_delta(ie, basis_inv, path), False, None
518
self._add_text_to_weave(ie.file_id, '', heads, None)
519
elif kind == 'symlink':
520
current_link_target = content_summary[3]
522
# symlink target is not generic metadata, check if it has
524
if current_link_target != parent_entry.symlink_target:
527
# unchanged, carry over.
528
ie.revision = parent_entry.revision
529
ie.symlink_target = parent_entry.symlink_target
530
return self._get_delta(ie, basis_inv, path), False, None
531
ie.symlink_target = current_link_target
532
self._add_text_to_weave(ie.file_id, '', heads, None)
533
elif kind == 'tree-reference':
535
if content_summary[3] != parent_entry.reference_revision:
538
# unchanged, carry over.
539
ie.reference_revision = parent_entry.reference_revision
540
ie.revision = parent_entry.revision
541
return self._get_delta(ie, basis_inv, path), False, None
542
ie.reference_revision = content_summary[3]
543
self._add_text_to_weave(ie.file_id, '', heads, None)
545
raise NotImplementedError('unknown kind')
546
ie.revision = self._new_revision_id
547
self._any_changes = True
548
return self._get_delta(ie, basis_inv, path), True, fingerprint
550
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
551
_entry_factory=entry_factory):
552
"""Record a new tree via iter_changes.
554
:param tree: The tree to obtain text contents from for changed objects.
555
:param basis_revision_id: The revision id of the tree the iter_changes
556
has been generated against. Currently assumed to be the same
557
as self.parents[0] - if it is not, errors may occur.
558
:param iter_changes: An iter_changes iterator with the changes to apply
559
to basis_revision_id. The iterator must not include any items with
560
a current kind of None - missing items must be either filtered out
561
or errored-on beefore record_iter_changes sees the item.
562
:param _entry_factory: Private method to bind entry_factory locally for
564
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
567
# Create an inventory delta based on deltas between all the parents and
568
# deltas between all the parent inventories. We use inventory delta's
569
# between the inventory objects because iter_changes masks
570
# last-changed-field only changes.
572
# file_id -> change map, change is fileid, paths, changed, versioneds,
573
# parents, names, kinds, executables
575
# {file_id -> revision_id -> inventory entry, for entries in parent
576
# trees that are not parents[0]
580
revtrees = list(self.repository.revision_trees(self.parents))
581
except errors.NoSuchRevision:
582
# one or more ghosts, slow path.
584
for revision_id in self.parents:
586
revtrees.append(self.repository.revision_tree(revision_id))
587
except errors.NoSuchRevision:
589
basis_revision_id = _mod_revision.NULL_REVISION
591
revtrees.append(self.repository.revision_tree(
592
_mod_revision.NULL_REVISION))
593
# The basis inventory from a repository
595
basis_inv = revtrees[0].inventory
597
basis_inv = self.repository.revision_tree(
598
_mod_revision.NULL_REVISION).inventory
599
if len(self.parents) > 0:
600
if basis_revision_id != self.parents[0] and not ghost_basis:
602
"arbitrary basis parents not yet supported with merges")
603
for revtree in revtrees[1:]:
604
for change in revtree.inventory._make_delta(basis_inv):
605
if change[1] is None:
606
# Not present in this parent.
608
if change[2] not in merged_ids:
609
if change[0] is not None:
610
basis_entry = basis_inv[change[2]]
611
merged_ids[change[2]] = [
613
basis_entry.revision,
616
parent_entries[change[2]] = {
618
basis_entry.revision:basis_entry,
620
change[3].revision:change[3],
623
merged_ids[change[2]] = [change[3].revision]
624
parent_entries[change[2]] = {change[3].revision:change[3]}
626
merged_ids[change[2]].append(change[3].revision)
627
parent_entries[change[2]][change[3].revision] = change[3]
630
# Setup the changes from the tree:
631
# changes maps file_id -> (change, [parent revision_ids])
633
for change in iter_changes:
634
# This probably looks up in basis_inv way to much.
635
if change[1][0] is not None:
636
head_candidate = [basis_inv[change[0]].revision]
639
changes[change[0]] = change, merged_ids.get(change[0],
641
unchanged_merged = set(merged_ids) - set(changes)
642
# Extend the changes dict with synthetic changes to record merges of
644
for file_id in unchanged_merged:
645
# Record a merged version of these items that did not change vs the
646
# basis. This can be either identical parallel changes, or a revert
647
# of a specific file after a merge. The recorded content will be
648
# that of the current tree (which is the same as the basis), but
649
# the per-file graph will reflect a merge.
650
# NB:XXX: We are reconstructing path information we had, this
651
# should be preserved instead.
652
# inv delta change: (file_id, (path_in_source, path_in_target),
653
# changed_content, versioned, parent, name, kind,
656
basis_entry = basis_inv[file_id]
657
except errors.NoSuchId:
658
# a change from basis->some_parents but file_id isn't in basis
659
# so was new in the merge, which means it must have changed
660
# from basis -> current, and as it hasn't the add was reverted
661
# by the user. So we discard this change.
665
(basis_inv.id2path(file_id), tree.id2path(file_id)),
667
(basis_entry.parent_id, basis_entry.parent_id),
668
(basis_entry.name, basis_entry.name),
669
(basis_entry.kind, basis_entry.kind),
670
(basis_entry.executable, basis_entry.executable))
671
changes[file_id] = (change, merged_ids[file_id])
672
# changes contains tuples with the change and a set of inventory
673
# candidates for the file.
675
# old_path, new_path, file_id, new_inventory_entry
676
seen_root = False # Is the root in the basis delta?
677
inv_delta = self._basis_delta
678
modified_rev = self._new_revision_id
679
for change, head_candidates in changes.values():
680
if change[3][1]: # versioned in target.
681
# Several things may be happening here:
682
# We may have a fork in the per-file graph
683
# - record a change with the content from tree
684
# We may have a change against < all trees
685
# - carry over the tree that hasn't changed
686
# We may have a change against all trees
687
# - record the change with the content from tree
690
entry = _entry_factory[kind](file_id, change[5][1],
692
head_set = self._heads(change[0], set(head_candidates))
695
for head_candidate in head_candidates:
696
if head_candidate in head_set:
697
heads.append(head_candidate)
698
head_set.remove(head_candidate)
701
# Could be a carry-over situation:
702
parent_entry_revs = parent_entries.get(file_id, None)
703
if parent_entry_revs:
704
parent_entry = parent_entry_revs.get(heads[0], None)
707
if parent_entry is None:
708
# The parent iter_changes was called against is the one
709
# that is the per-file head, so any change is relevant
710
# iter_changes is valid.
711
carry_over_possible = False
713
# could be a carry over situation
714
# A change against the basis may just indicate a merge,
715
# we need to check the content against the source of the
716
# merge to determine if it was changed after the merge
718
if (parent_entry.kind != entry.kind or
719
parent_entry.parent_id != entry.parent_id or
720
parent_entry.name != entry.name):
721
# Metadata common to all entries has changed
722
# against per-file parent
723
carry_over_possible = False
725
carry_over_possible = True
726
# per-type checks for changes against the parent_entry
729
# Cannot be a carry-over situation
730
carry_over_possible = False
731
# Populate the entry in the delta
733
# XXX: There is still a small race here: If someone reverts the content of a file
734
# after iter_changes examines and decides it has changed,
735
# we will unconditionally record a new version even if some
736
# other process reverts it while commit is running (with
737
# the revert happening after iter_changes did it's
740
entry.executable = True
742
entry.executable = False
743
if (carry_over_possible and
744
parent_entry.executable == entry.executable):
745
# Check the file length, content hash after reading
747
nostore_sha = parent_entry.text_sha1
750
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
752
text = file_obj.read()
756
entry.text_sha1, entry.text_size = self._add_text_to_weave(
757
file_id, text, heads, nostore_sha)
758
yield file_id, change[1][1], (entry.text_sha1, stat_value)
759
except errors.ExistingContent:
760
# No content change against a carry_over parent
761
# Perhaps this should also yield a fs hash update?
763
entry.text_size = parent_entry.text_size
764
entry.text_sha1 = parent_entry.text_sha1
765
elif kind == 'symlink':
767
entry.symlink_target = tree.get_symlink_target(file_id)
768
if (carry_over_possible and
769
parent_entry.symlink_target == entry.symlink_target):
772
self._add_text_to_weave(change[0], '', heads, None)
773
elif kind == 'directory':
774
if carry_over_possible:
777
# Nothing to set on the entry.
778
# XXX: split into the Root and nonRoot versions.
779
if change[1][1] != '' or self.repository.supports_rich_root():
780
self._add_text_to_weave(change[0], '', heads, None)
781
elif kind == 'tree-reference':
782
if not self.repository._format.supports_tree_reference:
783
# This isn't quite sane as an error, but we shouldn't
784
# ever see this code path in practice: tree's don't
785
# permit references when the repo doesn't support tree
787
raise errors.UnsupportedOperation(tree.add_reference,
789
reference_revision = tree.get_reference_revision(change[0])
790
entry.reference_revision = reference_revision
791
if (carry_over_possible and
792
parent_entry.reference_revision == reference_revision):
795
self._add_text_to_weave(change[0], '', heads, None)
797
raise AssertionError('unknown kind %r' % kind)
799
entry.revision = modified_rev
801
entry.revision = parent_entry.revision
804
new_path = change[1][1]
805
inv_delta.append((change[1][0], new_path, change[0], entry))
808
self.new_inventory = None
810
self._any_changes = True
812
# housekeeping root entry changes do not affect no-change commits.
813
self._require_root_change(tree)
814
self.basis_delta_revision = basis_revision_id
816
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
817
parent_keys = tuple([(file_id, parent) for parent in parents])
818
return self.repository.texts._add_text(
819
(file_id, self._new_revision_id), parent_keys, new_text,
820
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
823
class RootCommitBuilder(CommitBuilder):
824
"""This commitbuilder actually records the root id"""
826
# the root entry gets versioned properly by this builder.
827
_versioned_root = True
829
def _check_root(self, ie, parent_invs, tree):
830
"""Helper for record_entry_contents.
832
:param ie: An entry being added.
833
:param parent_invs: The inventories of the parent revisions of the
835
:param tree: The tree that is being committed.
838
def _require_root_change(self, tree):
839
"""Enforce an appropriate root object change.
841
This is called once when record_iter_changes is called, if and only if
842
the root was not in the delta calculated by record_iter_changes.
844
:param tree: The tree which is being committed.
846
# versioned roots do not change unless the tree found a change.
849
######################################################################
853
class Repository(object):
854
"""Repository holding history for one or more branches.
856
The repository holds and retrieves historical information including
857
revisions and file history. It's normally accessed only by the Branch,
858
which views a particular line of development through that history.
860
The Repository builds on top of some byte storage facilies (the revisions,
861
signatures, inventories, texts and chk_bytes attributes) and a Transport,
862
which respectively provide byte storage and a means to access the (possibly
865
The byte storage facilities are addressed via tuples, which we refer to
866
as 'keys' throughout the code base. Revision_keys, inventory_keys and
867
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
868
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
869
byte string made up of a hash identifier and a hash value.
870
We use this interface because it allows low friction with the underlying
871
code that implements disk indices, network encoding and other parts of
874
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
875
the serialised revisions for the repository. This can be used to obtain
876
revision graph information or to access raw serialised revisions.
877
The result of trying to insert data into the repository via this store
878
is undefined: it should be considered read-only except for implementors
880
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
881
the serialised signatures for the repository. This can be used to
882
obtain access to raw serialised signatures. The result of trying to
883
insert data into the repository via this store is undefined: it should
884
be considered read-only except for implementors of repositories.
885
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
886
the serialised inventories for the repository. This can be used to
887
obtain unserialised inventories. The result of trying to insert data
888
into the repository via this store is undefined: it should be
889
considered read-only except for implementors of repositories.
890
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
891
texts of files and directories for the repository. This can be used to
892
obtain file texts or file graphs. Note that Repository.iter_file_bytes
893
is usually a better interface for accessing file texts.
894
The result of trying to insert data into the repository via this store
895
is undefined: it should be considered read-only except for implementors
897
:ivar chk_bytes: A bzrlib.versionedfile.VersionedFiles instance containing
898
any data the repository chooses to store or have indexed by its hash.
899
The result of trying to insert data into the repository via this store
900
is undefined: it should be considered read-only except for implementors
902
:ivar _transport: Transport for file access to repository, typically
903
pointing to .bzr/repository.
906
# What class to use for a CommitBuilder. Often its simpler to change this
907
# in a Repository class subclass rather than to override
908
# get_commit_builder.
909
_commit_builder_class = CommitBuilder
910
# The search regex used by xml based repositories to determine what things
911
# where changed in a single commit.
912
_file_ids_altered_regex = lazy_regex.lazy_compile(
913
r'file_id="(?P<file_id>[^"]+)"'
914
r'.* revision="(?P<revision_id>[^"]+)"'
917
def abort_write_group(self, suppress_errors=False):
918
"""Commit the contents accrued within the current write group.
920
:param suppress_errors: if true, abort_write_group will catch and log
921
unexpected errors that happen during the abort, rather than
922
allowing them to propagate. Defaults to False.
924
:seealso: start_write_group.
926
if self._write_group is not self.get_transaction():
927
# has an unlock or relock occured ?
930
'(suppressed) mismatched lock context and write group. %r, %r',
931
self._write_group, self.get_transaction())
933
raise errors.BzrError(
934
'mismatched lock context and write group. %r, %r' %
935
(self._write_group, self.get_transaction()))
937
self._abort_write_group()
938
except Exception, exc:
939
self._write_group = None
940
if not suppress_errors:
942
mutter('abort_write_group failed')
943
log_exception_quietly()
944
note('bzr: ERROR (ignored): %s', exc)
945
self._write_group = None
947
def _abort_write_group(self):
948
"""Template method for per-repository write group cleanup.
950
This is called during abort before the write group is considered to be
951
finished and should cleanup any internal state accrued during the write
952
group. There is no requirement that data handed to the repository be
953
*not* made available - this is not a rollback - but neither should any
954
attempt be made to ensure that data added is fully commited. Abort is
955
invoked when an error has occured so futher disk or network operations
956
may not be possible or may error and if possible should not be
960
def add_fallback_repository(self, repository):
961
"""Add a repository to use for looking up data not held locally.
963
:param repository: A repository.
965
if not self._format.supports_external_lookups:
966
raise errors.UnstackableRepositoryFormat(self._format, self.base)
968
# This repository will call fallback.unlock() when we transition to
969
# the unlocked state, so we make sure to increment the lock count
970
repository.lock_read()
971
self._check_fallback_repository(repository)
972
self._fallback_repositories.append(repository)
973
self.texts.add_fallback_versioned_files(repository.texts)
974
self.inventories.add_fallback_versioned_files(repository.inventories)
975
self.revisions.add_fallback_versioned_files(repository.revisions)
976
self.signatures.add_fallback_versioned_files(repository.signatures)
977
if self.chk_bytes is not None:
978
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
980
def _check_fallback_repository(self, repository):
981
"""Check that this repository can fallback to repository safely.
983
Raise an error if not.
985
:param repository: A repository to fallback to.
987
return InterRepository._assert_same_model(self, repository)
989
def add_inventory(self, revision_id, inv, parents):
990
"""Add the inventory inv to the repository as revision_id.
992
:param parents: The revision ids of the parents that revision_id
993
is known to have and are in the repository already.
995
:returns: The validator(which is a sha1 digest, though what is sha'd is
996
repository format specific) of the serialized inventory.
998
if not self.is_in_write_group():
999
raise AssertionError("%r not in write group" % (self,))
1000
_mod_revision.check_not_reserved_id(revision_id)
1001
if not (inv.revision_id is None or inv.revision_id == revision_id):
1002
raise AssertionError(
1003
"Mismatch between inventory revision"
1004
" id and insertion revid (%r, %r)"
1005
% (inv.revision_id, revision_id))
1006
if inv.root is None:
1007
raise AssertionError()
1008
return self._add_inventory_checked(revision_id, inv, parents)
1010
def _add_inventory_checked(self, revision_id, inv, parents):
1011
"""Add inv to the repository after checking the inputs.
1013
This function can be overridden to allow different inventory styles.
1015
:seealso: add_inventory, for the contract.
1017
inv_lines = self._serialise_inventory_to_lines(inv)
1018
return self._inventory_add_lines(revision_id, parents,
1019
inv_lines, check_content=False)
1021
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1022
parents, basis_inv=None, propagate_caches=False):
1023
"""Add a new inventory expressed as a delta against another revision.
1025
See the inventory developers documentation for the theory behind
1028
:param basis_revision_id: The inventory id the delta was created
1029
against. (This does not have to be a direct parent.)
1030
:param delta: The inventory delta (see Inventory.apply_delta for
1032
:param new_revision_id: The revision id that the inventory is being
1034
:param parents: The revision ids of the parents that revision_id is
1035
known to have and are in the repository already. These are supplied
1036
for repositories that depend on the inventory graph for revision
1037
graph access, as well as for those that pun ancestry with delta
1039
:param basis_inv: The basis inventory if it is already known,
1041
:param propagate_caches: If True, the caches for this inventory are
1042
copied to and updated for the result if possible.
1044
:returns: (validator, new_inv)
1045
The validator(which is a sha1 digest, though what is sha'd is
1046
repository format specific) of the serialized inventory, and the
1047
resulting inventory.
1049
if not self.is_in_write_group():
1050
raise AssertionError("%r not in write group" % (self,))
1051
_mod_revision.check_not_reserved_id(new_revision_id)
1052
basis_tree = self.revision_tree(basis_revision_id)
1053
basis_tree.lock_read()
1055
# Note that this mutates the inventory of basis_tree, which not all
1056
# inventory implementations may support: A better idiom would be to
1057
# return a new inventory, but as there is no revision tree cache in
1058
# repository this is safe for now - RBC 20081013
1059
if basis_inv is None:
1060
basis_inv = basis_tree.inventory
1061
basis_inv.apply_delta(delta)
1062
basis_inv.revision_id = new_revision_id
1063
return (self.add_inventory(new_revision_id, basis_inv, parents),
1068
def _inventory_add_lines(self, revision_id, parents, lines,
1069
check_content=True):
1070
"""Store lines in inv_vf and return the sha1 of the inventory."""
1071
parents = [(parent,) for parent in parents]
1072
result = self.inventories.add_lines((revision_id,), parents, lines,
1073
check_content=check_content)[0]
1074
self.inventories._access.flush()
1077
def add_revision(self, revision_id, rev, inv=None, config=None):
1078
"""Add rev to the revision store as revision_id.
1080
:param revision_id: the revision id to use.
1081
:param rev: The revision object.
1082
:param inv: The inventory for the revision. if None, it will be looked
1083
up in the inventory storer
1084
:param config: If None no digital signature will be created.
1085
If supplied its signature_needed method will be used
1086
to determine if a signature should be made.
1088
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
1090
_mod_revision.check_not_reserved_id(revision_id)
1091
if config is not None and config.signature_needed():
1093
inv = self.get_inventory(revision_id)
1094
plaintext = Testament(rev, inv).as_short_text()
1095
self.store_revision_signature(
1096
gpg.GPGStrategy(config), plaintext, revision_id)
1097
# check inventory present
1098
if not self.inventories.get_parent_map([(revision_id,)]):
1100
raise errors.WeaveRevisionNotPresent(revision_id,
1103
# yes, this is not suitable for adding with ghosts.
1104
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
1107
key = (revision_id,)
1108
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
1109
self._add_revision(rev)
1111
def _add_revision(self, revision):
1112
text = self._serializer.write_revision_to_string(revision)
1113
key = (revision.revision_id,)
1114
parents = tuple((parent,) for parent in revision.parent_ids)
1115
self.revisions.add_lines(key, parents, osutils.split_lines(text))
1117
def all_revision_ids(self):
1118
"""Returns a list of all the revision ids in the repository.
1120
This is conceptually deprecated because code should generally work on
1121
the graph reachable from a particular revision, and ignore any other
1122
revisions that might be present. There is no direct replacement
1125
if 'evil' in debug.debug_flags:
1126
mutter_callsite(2, "all_revision_ids is linear with history.")
1127
return self._all_revision_ids()
1129
def _all_revision_ids(self):
1130
"""Returns a list of all the revision ids in the repository.
1132
These are in as much topological order as the underlying store can
1135
raise NotImplementedError(self._all_revision_ids)
1137
def break_lock(self):
1138
"""Break a lock if one is present from another instance.
1140
Uses the ui factory to ask for confirmation if the lock may be from
1143
self.control_files.break_lock()
1146
def _eliminate_revisions_not_present(self, revision_ids):
1147
"""Check every revision id in revision_ids to see if we have it.
1149
Returns a set of the present revisions.
1152
graph = self.get_graph()
1153
parent_map = graph.get_parent_map(revision_ids)
1154
# The old API returned a list, should this actually be a set?
1155
return parent_map.keys()
1158
def create(a_bzrdir):
1159
"""Construct the current default format repository in a_bzrdir."""
1160
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
1162
def __init__(self, _format, a_bzrdir, control_files):
1163
"""instantiate a Repository.
1165
:param _format: The format of the repository on disk.
1166
:param a_bzrdir: The BzrDir of the repository.
1168
In the future we will have a single api for all stores for
1169
getting file texts, inventories and revisions, then
1170
this construct will accept instances of those things.
1172
super(Repository, self).__init__()
1173
self._format = _format
1174
# the following are part of the public API for Repository:
1175
self.bzrdir = a_bzrdir
1176
self.control_files = control_files
1177
self._transport = control_files._transport
1178
self.base = self._transport.base
1180
self._reconcile_does_inventory_gc = True
1181
self._reconcile_fixes_text_parents = False
1182
self._reconcile_backsup_inventory = True
1183
# not right yet - should be more semantically clear ?
1185
# TODO: make sure to construct the right store classes, etc, depending
1186
# on whether escaping is required.
1187
self._warn_if_deprecated()
1188
self._write_group = None
1189
# Additional places to query for data.
1190
self._fallback_repositories = []
1191
# An InventoryEntry cache, used during deserialization
1192
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1195
if self._fallback_repositories:
1196
return '%s(%r, fallback_repositories=%r)' % (
1197
self.__class__.__name__,
1199
self._fallback_repositories)
1201
return '%s(%r)' % (self.__class__.__name__,
1204
def _has_same_fallbacks(self, other_repo):
1205
"""Returns true if the repositories have the same fallbacks."""
1206
my_fb = self._fallback_repositories
1207
other_fb = other_repo._fallback_repositories
1208
if len(my_fb) != len(other_fb):
1210
for f, g in zip(my_fb, other_fb):
1211
if not f.has_same_location(g):
1215
def has_same_location(self, other):
1216
"""Returns a boolean indicating if this repository is at the same
1217
location as another repository.
1219
This might return False even when two repository objects are accessing
1220
the same physical repository via different URLs.
1222
if self.__class__ is not other.__class__:
1224
return (self._transport.base == other._transport.base)
1226
def is_in_write_group(self):
1227
"""Return True if there is an open write group.
1229
:seealso: start_write_group.
1231
return self._write_group is not None
1233
def is_locked(self):
1234
return self.control_files.is_locked()
1236
def is_write_locked(self):
1237
"""Return True if this object is write locked."""
1238
return self.is_locked() and self.control_files._lock_mode == 'w'
1240
def lock_write(self, token=None):
1241
"""Lock this repository for writing.
1243
This causes caching within the repository obejct to start accumlating
1244
data during reads, and allows a 'write_group' to be obtained. Write
1245
groups must be used for actual data insertion.
1247
:param token: if this is already locked, then lock_write will fail
1248
unless the token matches the existing lock.
1249
:returns: a token if this instance supports tokens, otherwise None.
1250
:raises TokenLockingNotSupported: when a token is given but this
1251
instance doesn't support using token locks.
1252
:raises MismatchedToken: if the specified token doesn't match the token
1253
of the existing lock.
1254
:seealso: start_write_group.
1256
A token should be passed in if you know that you have locked the object
1257
some other way, and need to synchronise this object's state with that
1260
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1262
locked = self.is_locked()
1263
result = self.control_files.lock_write(token=token)
1265
for repo in self._fallback_repositories:
1266
# Writes don't affect fallback repos
1268
self._refresh_data()
1271
def lock_read(self):
1272
locked = self.is_locked()
1273
self.control_files.lock_read()
1275
for repo in self._fallback_repositories:
1277
self._refresh_data()
1279
def get_physical_lock_status(self):
1280
return self.control_files.get_physical_lock_status()
1282
def leave_lock_in_place(self):
1283
"""Tell this repository not to release the physical lock when this
1286
If lock_write doesn't return a token, then this method is not supported.
1288
self.control_files.leave_in_place()
1290
def dont_leave_lock_in_place(self):
1291
"""Tell this repository to release the physical lock when this
1292
object is unlocked, even if it didn't originally acquire it.
1294
If lock_write doesn't return a token, then this method is not supported.
1296
self.control_files.dont_leave_in_place()
1299
def gather_stats(self, revid=None, committers=None):
1300
"""Gather statistics from a revision id.
1302
:param revid: The revision id to gather statistics from, if None, then
1303
no revision specific statistics are gathered.
1304
:param committers: Optional parameter controlling whether to grab
1305
a count of committers from the revision specific statistics.
1306
:return: A dictionary of statistics. Currently this contains:
1307
committers: The number of committers if requested.
1308
firstrev: A tuple with timestamp, timezone for the penultimate left
1309
most ancestor of revid, if revid is not the NULL_REVISION.
1310
latestrev: A tuple with timestamp, timezone for revid, if revid is
1311
not the NULL_REVISION.
1312
revisions: The total revision count in the repository.
1313
size: An estimate disk size of the repository in bytes.
1316
if revid and committers:
1317
result['committers'] = 0
1318
if revid and revid != _mod_revision.NULL_REVISION:
1320
all_committers = set()
1321
revisions = self.get_ancestry(revid)
1322
# pop the leading None
1324
first_revision = None
1326
# ignore the revisions in the middle - just grab first and last
1327
revisions = revisions[0], revisions[-1]
1328
for revision in self.get_revisions(revisions):
1329
if not first_revision:
1330
first_revision = revision
1332
all_committers.add(revision.committer)
1333
last_revision = revision
1335
result['committers'] = len(all_committers)
1336
result['firstrev'] = (first_revision.timestamp,
1337
first_revision.timezone)
1338
result['latestrev'] = (last_revision.timestamp,
1339
last_revision.timezone)
1341
# now gather global repository information
1342
# XXX: This is available for many repos regardless of listability.
1343
if self.bzrdir.root_transport.listable():
1344
# XXX: do we want to __define len__() ?
1345
# Maybe the versionedfiles object should provide a different
1346
# method to get the number of keys.
1347
result['revisions'] = len(self.revisions.keys())
1348
# result['size'] = t
1351
def find_branches(self, using=False):
1352
"""Find branches underneath this repository.
1354
This will include branches inside other branches.
1356
:param using: If True, list only branches using this repository.
1358
if using and not self.is_shared():
1360
return [self.bzrdir.open_branch()]
1361
except errors.NotBranchError:
1363
class Evaluator(object):
1366
self.first_call = True
1368
def __call__(self, bzrdir):
1369
# On the first call, the parameter is always the bzrdir
1370
# containing the current repo.
1371
if not self.first_call:
1373
repository = bzrdir.open_repository()
1374
except errors.NoRepositoryPresent:
1377
return False, (None, repository)
1378
self.first_call = False
1380
value = (bzrdir.open_branch(), None)
1381
except errors.NotBranchError:
1382
value = (None, None)
1386
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1387
self.bzrdir.root_transport, evaluate=Evaluator()):
1388
if branch is not None:
1389
branches.append(branch)
1390
if not using and repository is not None:
1391
branches.extend(repository.find_branches())
1395
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1396
"""Return the revision ids that other has that this does not.
1398
These are returned in topological order.
1400
revision_id: only return revision ids included by revision_id.
1402
return InterRepository.get(other, self).search_missing_revision_ids(
1403
revision_id, find_ghosts)
1407
"""Open the repository rooted at base.
1409
For instance, if the repository is at URL/.bzr/repository,
1410
Repository.open(URL) -> a Repository instance.
1412
control = bzrdir.BzrDir.open(base)
1413
return control.open_repository()
1415
def copy_content_into(self, destination, revision_id=None):
1416
"""Make a complete copy of the content in self into destination.
1418
This is a destructive operation! Do not use it on existing
1421
return InterRepository.get(self, destination).copy_content(revision_id)
1423
def commit_write_group(self):
1424
"""Commit the contents accrued within the current write group.
1426
:seealso: start_write_group.
1428
if self._write_group is not self.get_transaction():
1429
# has an unlock or relock occured ?
1430
raise errors.BzrError('mismatched lock context %r and '
1432
(self.get_transaction(), self._write_group))
1433
result = self._commit_write_group()
1434
self._write_group = None
1437
def _commit_write_group(self):
1438
"""Template method for per-repository write group cleanup.
1440
This is called before the write group is considered to be
1441
finished and should ensure that all data handed to the repository
1442
for writing during the write group is safely committed (to the
1443
extent possible considering file system caching etc).
1446
def suspend_write_group(self):
1447
raise errors.UnsuspendableWriteGroup(self)
1449
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1450
"""Return the keys of missing inventory parents for revisions added in
1453
A revision is not complete if the inventory delta for that revision
1454
cannot be calculated. Therefore if the parent inventories of a
1455
revision are not present, the revision is incomplete, and e.g. cannot
1456
be streamed by a smart server. This method finds missing inventory
1457
parents for revisions added in this write group.
1459
if not self._format.supports_external_lookups:
1460
# This is only an issue for stacked repositories
1462
if not self.is_in_write_group():
1463
raise AssertionError('not in a write group')
1465
# XXX: We assume that every added revision already has its
1466
# corresponding inventory, so we only check for parent inventories that
1467
# might be missing, rather than all inventories.
1468
parents = set(self.revisions._index.get_missing_parents())
1469
parents.discard(_mod_revision.NULL_REVISION)
1470
unstacked_inventories = self.inventories._index
1471
present_inventories = unstacked_inventories.get_parent_map(
1472
key[-1:] for key in parents)
1473
parents.difference_update(present_inventories)
1474
if len(parents) == 0:
1475
# No missing parent inventories.
1477
if not check_for_missing_texts:
1478
return set(('inventories', rev_id) for (rev_id,) in parents)
1479
# Ok, now we have a list of missing inventories. But these only matter
1480
# if the inventories that reference them are missing some texts they
1481
# appear to introduce.
1482
# XXX: Texts referenced by all added inventories need to be present,
1483
# but at the moment we're only checking for texts referenced by
1484
# inventories at the graph's edge.
1485
key_deps = self.revisions._index._key_dependencies
1486
key_deps.add_keys(present_inventories)
1487
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1488
file_ids = self.fileids_altered_by_revision_ids(referrers)
1489
missing_texts = set()
1490
for file_id, version_ids in file_ids.iteritems():
1491
missing_texts.update(
1492
(file_id, version_id) for version_id in version_ids)
1493
present_texts = self.texts.get_parent_map(missing_texts)
1494
missing_texts.difference_update(present_texts)
1495
if not missing_texts:
1496
# No texts are missing, so all revisions and their deltas are
1499
# Alternatively the text versions could be returned as the missing
1500
# keys, but this is likely to be less data.
1501
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1504
def refresh_data(self):
1505
"""Re-read any data needed to to synchronise with disk.
1507
This method is intended to be called after another repository instance
1508
(such as one used by a smart server) has inserted data into the
1509
repository. It may not be called during a write group, but may be
1510
called at any other time.
1512
if self.is_in_write_group():
1513
raise errors.InternalBzrError(
1514
"May not refresh_data while in a write group.")
1515
self._refresh_data()
1517
def resume_write_group(self, tokens):
1518
if not self.is_write_locked():
1519
raise errors.NotWriteLocked(self)
1520
if self._write_group:
1521
raise errors.BzrError('already in a write group')
1522
self._resume_write_group(tokens)
1523
# so we can detect unlock/relock - the write group is now entered.
1524
self._write_group = self.get_transaction()
1526
def _resume_write_group(self, tokens):
1527
raise errors.UnsuspendableWriteGroup(self)
1529
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
1531
"""Fetch the content required to construct revision_id from source.
1533
If revision_id is None and fetch_spec is None, then all content is
1536
fetch() may not be used when the repository is in a write group -
1537
either finish the current write group before using fetch, or use
1538
fetch before starting the write group.
1540
:param find_ghosts: Find and copy revisions in the source that are
1541
ghosts in the target (and not reachable directly by walking out to
1542
the first-present revision in target from revision_id).
1543
:param revision_id: If specified, all the content needed for this
1544
revision ID will be copied to the target. Fetch will determine for
1545
itself which content needs to be copied.
1546
:param fetch_spec: If specified, a SearchResult or
1547
PendingAncestryResult that describes which revisions to copy. This
1548
allows copying multiple heads at once. Mutually exclusive with
1551
if fetch_spec is not None and revision_id is not None:
1552
raise AssertionError(
1553
"fetch_spec and revision_id are mutually exclusive.")
1554
if self.is_in_write_group():
1555
raise errors.InternalBzrError(
1556
"May not fetch while in a write group.")
1557
# fast path same-url fetch operations
1558
# TODO: lift out to somewhere common with RemoteRepository
1559
# <https://bugs.edge.launchpad.net/bzr/+bug/401646>
1560
if (self.has_same_location(source)
1561
and fetch_spec is None
1562
and self._has_same_fallbacks(source)):
1563
# check that last_revision is in 'from' and then return a
1565
if (revision_id is not None and
1566
not _mod_revision.is_null(revision_id)):
1567
self.get_revision(revision_id)
1569
# if there is no specific appropriate InterRepository, this will get
1570
# the InterRepository base class, which raises an
1571
# IncompatibleRepositories when asked to fetch.
1572
inter = InterRepository.get(source, self)
1573
return inter.fetch(revision_id=revision_id, pb=pb,
1574
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
1576
def create_bundle(self, target, base, fileobj, format=None):
1577
return serializer.write_bundle(self, target, base, fileobj, format)
1579
def get_commit_builder(self, branch, parents, config, timestamp=None,
1580
timezone=None, committer=None, revprops=None,
1582
"""Obtain a CommitBuilder for this repository.
1584
:param branch: Branch to commit to.
1585
:param parents: Revision ids of the parents of the new revision.
1586
:param config: Configuration to use.
1587
:param timestamp: Optional timestamp recorded for commit.
1588
:param timezone: Optional timezone for timestamp.
1589
:param committer: Optional committer to set for commit.
1590
:param revprops: Optional dictionary of revision properties.
1591
:param revision_id: Optional revision id.
1593
result = self._commit_builder_class(self, parents, config,
1594
timestamp, timezone, committer, revprops, revision_id)
1595
self.start_write_group()
1599
if (self.control_files._lock_count == 1 and
1600
self.control_files._lock_mode == 'w'):
1601
if self._write_group is not None:
1602
self.abort_write_group()
1603
self.control_files.unlock()
1604
raise errors.BzrError(
1605
'Must end write groups before releasing write locks.')
1606
self.control_files.unlock()
1607
if self.control_files._lock_count == 0:
1608
self._inventory_entry_cache.clear()
1609
for repo in self._fallback_repositories:
1613
def clone(self, a_bzrdir, revision_id=None):
1614
"""Clone this repository into a_bzrdir using the current format.
1616
Currently no check is made that the format of this repository and
1617
the bzrdir format are compatible. FIXME RBC 20060201.
1619
:return: The newly created destination repository.
1621
# TODO: deprecate after 0.16; cloning this with all its settings is
1622
# probably not very useful -- mbp 20070423
1623
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1624
self.copy_content_into(dest_repo, revision_id)
1627
def start_write_group(self):
1628
"""Start a write group in the repository.
1630
Write groups are used by repositories which do not have a 1:1 mapping
1631
between file ids and backend store to manage the insertion of data from
1632
both fetch and commit operations.
1634
A write lock is required around the start_write_group/commit_write_group
1635
for the support of lock-requiring repository formats.
1637
One can only insert data into a repository inside a write group.
1641
if not self.is_write_locked():
1642
raise errors.NotWriteLocked(self)
1643
if self._write_group:
1644
raise errors.BzrError('already in a write group')
1645
self._start_write_group()
1646
# so we can detect unlock/relock - the write group is now entered.
1647
self._write_group = self.get_transaction()
1649
def _start_write_group(self):
1650
"""Template method for per-repository write group startup.
1652
This is called before the write group is considered to be
1657
def sprout(self, to_bzrdir, revision_id=None):
1658
"""Create a descendent repository for new development.
1660
Unlike clone, this does not copy the settings of the repository.
1662
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1663
dest_repo.fetch(self, revision_id=revision_id)
1666
def _create_sprouting_repo(self, a_bzrdir, shared):
1667
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1668
# use target default format.
1669
dest_repo = a_bzrdir.create_repository()
1671
# Most control formats need the repository to be specifically
1672
# created, but on some old all-in-one formats it's not needed
1674
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1675
except errors.UninitializableFormat:
1676
dest_repo = a_bzrdir.open_repository()
1679
def _get_sink(self):
1680
"""Return a sink for streaming into this repository."""
1681
return StreamSink(self)
1683
def _get_source(self, to_format):
1684
"""Return a source for streaming from this repository."""
1685
return StreamSource(self, to_format)
1688
def has_revision(self, revision_id):
1689
"""True if this repository has a copy of the revision."""
1690
return revision_id in self.has_revisions((revision_id,))
1693
def has_revisions(self, revision_ids):
1694
"""Probe to find out the presence of multiple revisions.
1696
:param revision_ids: An iterable of revision_ids.
1697
:return: A set of the revision_ids that were present.
1699
parent_map = self.revisions.get_parent_map(
1700
[(rev_id,) for rev_id in revision_ids])
1702
if _mod_revision.NULL_REVISION in revision_ids:
1703
result.add(_mod_revision.NULL_REVISION)
1704
result.update([key[0] for key in parent_map])
1708
def get_revision(self, revision_id):
1709
"""Return the Revision object for a named revision."""
1710
return self.get_revisions([revision_id])[0]
1713
def get_revision_reconcile(self, revision_id):
1714
"""'reconcile' helper routine that allows access to a revision always.
1716
This variant of get_revision does not cross check the weave graph
1717
against the revision one as get_revision does: but it should only
1718
be used by reconcile, or reconcile-alike commands that are correcting
1719
or testing the revision graph.
1721
return self._get_revisions([revision_id])[0]
1724
def get_revisions(self, revision_ids):
1725
"""Get many revisions at once."""
1726
return self._get_revisions(revision_ids)
1729
def _get_revisions(self, revision_ids):
1730
"""Core work logic to get many revisions without sanity checks."""
1731
for rev_id in revision_ids:
1732
if not rev_id or not isinstance(rev_id, basestring):
1733
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1734
keys = [(key,) for key in revision_ids]
1735
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1737
for record in stream:
1738
if record.storage_kind == 'absent':
1739
raise errors.NoSuchRevision(self, record.key[0])
1740
text = record.get_bytes_as('fulltext')
1741
rev = self._serializer.read_revision_from_string(text)
1742
revs[record.key[0]] = rev
1743
return [revs[revid] for revid in revision_ids]
1746
def get_revision_xml(self, revision_id):
1747
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1748
# would have already do it.
1749
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1750
# TODO: this can't just be replaced by:
1751
# return self._serializer.write_revision_to_string(
1752
# self.get_revision(revision_id))
1753
# as cStringIO preservers the encoding unlike write_revision_to_string
1754
# or some other call down the path.
1755
rev = self.get_revision(revision_id)
1756
rev_tmp = cStringIO.StringIO()
1757
# the current serializer..
1758
self._serializer.write_revision(rev, rev_tmp)
1760
return rev_tmp.getvalue()
1762
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1763
"""Produce a generator of revision deltas.
1765
Note that the input is a sequence of REVISIONS, not revision_ids.
1766
Trees will be held in memory until the generator exits.
1767
Each delta is relative to the revision's lefthand predecessor.
1769
:param specific_fileids: if not None, the result is filtered
1770
so that only those file-ids, their parents and their
1771
children are included.
1773
# Get the revision-ids of interest
1774
required_trees = set()
1775
for revision in revisions:
1776
required_trees.add(revision.revision_id)
1777
required_trees.update(revision.parent_ids[:1])
1779
# Get the matching filtered trees. Note that it's more
1780
# efficient to pass filtered trees to changes_from() rather
1781
# than doing the filtering afterwards. changes_from() could
1782
# arguably do the filtering itself but it's path-based, not
1783
# file-id based, so filtering before or afterwards is
1785
if specific_fileids is None:
1786
trees = dict((t.get_revision_id(), t) for
1787
t in self.revision_trees(required_trees))
1789
trees = dict((t.get_revision_id(), t) for
1790
t in self._filtered_revision_trees(required_trees,
1793
# Calculate the deltas
1794
for revision in revisions:
1795
if not revision.parent_ids:
1796
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1798
old_tree = trees[revision.parent_ids[0]]
1799
yield trees[revision.revision_id].changes_from(old_tree)
1802
def get_revision_delta(self, revision_id, specific_fileids=None):
1803
"""Return the delta for one revision.
1805
The delta is relative to the left-hand predecessor of the
1808
:param specific_fileids: if not None, the result is filtered
1809
so that only those file-ids, their parents and their
1810
children are included.
1812
r = self.get_revision(revision_id)
1813
return list(self.get_deltas_for_revisions([r],
1814
specific_fileids=specific_fileids))[0]
1817
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1818
signature = gpg_strategy.sign(plaintext)
1819
self.add_signature_text(revision_id, signature)
1822
def add_signature_text(self, revision_id, signature):
1823
self.signatures.add_lines((revision_id,), (),
1824
osutils.split_lines(signature))
1826
def find_text_key_references(self):
1827
"""Find the text key references within the repository.
1829
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1830
to whether they were referred to by the inventory of the
1831
revision_id that they contain. The inventory texts from all present
1832
revision ids are assessed to generate this report.
1834
revision_keys = self.revisions.keys()
1835
w = self.inventories
1836
pb = ui.ui_factory.nested_progress_bar()
1838
return self._find_text_key_references_from_xml_inventory_lines(
1839
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1843
def _find_text_key_references_from_xml_inventory_lines(self,
1845
"""Core routine for extracting references to texts from inventories.
1847
This performs the translation of xml lines to revision ids.
1849
:param line_iterator: An iterator of lines, origin_version_id
1850
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1851
to whether they were referred to by the inventory of the
1852
revision_id that they contain. Note that if that revision_id was
1853
not part of the line_iterator's output then False will be given -
1854
even though it may actually refer to that key.
1856
if not self._serializer.support_altered_by_hack:
1857
raise AssertionError(
1858
"_find_text_key_references_from_xml_inventory_lines only "
1859
"supported for branches which store inventory as unnested xml"
1860
", not on %r" % self)
1863
# this code needs to read every new line in every inventory for the
1864
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1865
# not present in one of those inventories is unnecessary but not
1866
# harmful because we are filtering by the revision id marker in the
1867
# inventory lines : we only select file ids altered in one of those
1868
# revisions. We don't need to see all lines in the inventory because
1869
# only those added in an inventory in rev X can contain a revision=X
1871
unescape_revid_cache = {}
1872
unescape_fileid_cache = {}
1874
# jam 20061218 In a big fetch, this handles hundreds of thousands
1875
# of lines, so it has had a lot of inlining and optimizing done.
1876
# Sorry that it is a little bit messy.
1877
# Move several functions to be local variables, since this is a long
1879
search = self._file_ids_altered_regex.search
1880
unescape = _unescape_xml
1881
setdefault = result.setdefault
1882
for line, line_key in line_iterator:
1883
match = search(line)
1886
# One call to match.group() returning multiple items is quite a
1887
# bit faster than 2 calls to match.group() each returning 1
1888
file_id, revision_id = match.group('file_id', 'revision_id')
1890
# Inlining the cache lookups helps a lot when you make 170,000
1891
# lines and 350k ids, versus 8.4 unique ids.
1892
# Using a cache helps in 2 ways:
1893
# 1) Avoids unnecessary decoding calls
1894
# 2) Re-uses cached strings, which helps in future set and
1896
# (2) is enough that removing encoding entirely along with
1897
# the cache (so we are using plain strings) results in no
1898
# performance improvement.
1900
revision_id = unescape_revid_cache[revision_id]
1902
unescaped = unescape(revision_id)
1903
unescape_revid_cache[revision_id] = unescaped
1904
revision_id = unescaped
1906
# Note that unconditionally unescaping means that we deserialise
1907
# every fileid, which for general 'pull' is not great, but we don't
1908
# really want to have some many fulltexts that this matters anyway.
1911
file_id = unescape_fileid_cache[file_id]
1913
unescaped = unescape(file_id)
1914
unescape_fileid_cache[file_id] = unescaped
1917
key = (file_id, revision_id)
1918
setdefault(key, False)
1919
if revision_id == line_key[-1]:
1923
def _inventory_xml_lines_for_keys(self, keys):
1924
"""Get a line iterator of the sort needed for findind references.
1926
Not relevant for non-xml inventory repositories.
1928
Ghosts in revision_keys are ignored.
1930
:param revision_keys: The revision keys for the inventories to inspect.
1931
:return: An iterator over (inventory line, revid) for the fulltexts of
1932
all of the xml inventories specified by revision_keys.
1934
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1935
for record in stream:
1936
if record.storage_kind != 'absent':
1937
chunks = record.get_bytes_as('chunked')
1938
revid = record.key[-1]
1939
lines = osutils.chunks_to_lines(chunks)
1943
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1945
"""Helper routine for fileids_altered_by_revision_ids.
1947
This performs the translation of xml lines to revision ids.
1949
:param line_iterator: An iterator of lines, origin_version_id
1950
:param revision_keys: The revision ids to filter for. This should be a
1951
set or other type which supports efficient __contains__ lookups, as
1952
the revision key from each parsed line will be looked up in the
1953
revision_keys filter.
1954
:return: a dictionary mapping altered file-ids to an iterable of
1955
revision_ids. Each altered file-ids has the exact revision_ids that
1956
altered it listed explicitly.
1958
seen = set(self._find_text_key_references_from_xml_inventory_lines(
1959
line_iterator).iterkeys())
1960
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
1961
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
1962
self._inventory_xml_lines_for_keys(parent_keys)))
1963
new_keys = seen - parent_seen
1965
setdefault = result.setdefault
1966
for key in new_keys:
1967
setdefault(key[0], set()).add(key[-1])
1970
def _find_parent_ids_of_revisions(self, revision_ids):
1971
"""Find all parent ids that are mentioned in the revision graph.
1973
:return: set of revisions that are parents of revision_ids which are
1974
not part of revision_ids themselves
1976
parent_map = self.get_parent_map(revision_ids)
1978
map(parent_ids.update, parent_map.itervalues())
1979
parent_ids.difference_update(revision_ids)
1980
parent_ids.discard(_mod_revision.NULL_REVISION)
1983
def _find_parent_keys_of_revisions(self, revision_keys):
1984
"""Similar to _find_parent_ids_of_revisions, but used with keys.
1986
:param revision_keys: An iterable of revision_keys.
1987
:return: The parents of all revision_keys that are not already in
1990
parent_map = self.revisions.get_parent_map(revision_keys)
1992
map(parent_keys.update, parent_map.itervalues())
1993
parent_keys.difference_update(revision_keys)
1994
parent_keys.discard(_mod_revision.NULL_REVISION)
1997
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1998
"""Find the file ids and versions affected by revisions.
2000
:param revisions: an iterable containing revision ids.
2001
:param _inv_weave: The inventory weave from this repository or None.
2002
If None, the inventory weave will be opened automatically.
2003
:return: a dictionary mapping altered file-ids to an iterable of
2004
revision_ids. Each altered file-ids has the exact revision_ids that
2005
altered it listed explicitly.
2007
selected_keys = set((revid,) for revid in revision_ids)
2008
w = _inv_weave or self.inventories
2009
pb = ui.ui_factory.nested_progress_bar()
2011
return self._find_file_ids_from_xml_inventory_lines(
2012
w.iter_lines_added_or_present_in_keys(
2013
selected_keys, pb=pb),
2018
def iter_files_bytes(self, desired_files):
2019
"""Iterate through file versions.
2021
Files will not necessarily be returned in the order they occur in
2022
desired_files. No specific order is guaranteed.
2024
Yields pairs of identifier, bytes_iterator. identifier is an opaque
2025
value supplied by the caller as part of desired_files. It should
2026
uniquely identify the file version in the caller's context. (Examples:
2027
an index number or a TreeTransform trans_id.)
2029
bytes_iterator is an iterable of bytestrings for the file. The
2030
kind of iterable and length of the bytestrings are unspecified, but for
2031
this implementation, it is a list of bytes produced by
2032
VersionedFile.get_record_stream().
2034
:param desired_files: a list of (file_id, revision_id, identifier)
2038
for file_id, revision_id, callable_data in desired_files:
2039
text_keys[(file_id, revision_id)] = callable_data
2040
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2041
if record.storage_kind == 'absent':
2042
raise errors.RevisionNotPresent(record.key, self)
2043
yield text_keys[record.key], record.get_bytes_as('chunked')
2045
def _generate_text_key_index(self, text_key_references=None,
2047
"""Generate a new text key index for the repository.
2049
This is an expensive function that will take considerable time to run.
2051
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2052
list of parents, also text keys. When a given key has no parents,
2053
the parents list will be [NULL_REVISION].
2055
# All revisions, to find inventory parents.
2056
if ancestors is None:
2057
graph = self.get_graph()
2058
ancestors = graph.get_parent_map(self.all_revision_ids())
2059
if text_key_references is None:
2060
text_key_references = self.find_text_key_references()
2061
pb = ui.ui_factory.nested_progress_bar()
2063
return self._do_generate_text_key_index(ancestors,
2064
text_key_references, pb)
2068
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2069
"""Helper for _generate_text_key_index to avoid deep nesting."""
2070
revision_order = tsort.topo_sort(ancestors)
2071
invalid_keys = set()
2073
for revision_id in revision_order:
2074
revision_keys[revision_id] = set()
2075
text_count = len(text_key_references)
2076
# a cache of the text keys to allow reuse; costs a dict of all the
2077
# keys, but saves a 2-tuple for every child of a given key.
2079
for text_key, valid in text_key_references.iteritems():
2081
invalid_keys.add(text_key)
2083
revision_keys[text_key[1]].add(text_key)
2084
text_key_cache[text_key] = text_key
2085
del text_key_references
2087
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2088
NULL_REVISION = _mod_revision.NULL_REVISION
2089
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2090
# too small for large or very branchy trees. However, for 55K path
2091
# trees, it would be easy to use too much memory trivially. Ideally we
2092
# could gauge this by looking at available real memory etc, but this is
2093
# always a tricky proposition.
2094
inventory_cache = lru_cache.LRUCache(10)
2095
batch_size = 10 # should be ~150MB on a 55K path tree
2096
batch_count = len(revision_order) / batch_size + 1
2098
pb.update("Calculating text parents", processed_texts, text_count)
2099
for offset in xrange(batch_count):
2100
to_query = revision_order[offset * batch_size:(offset + 1) *
2104
for rev_tree in self.revision_trees(to_query):
2105
revision_id = rev_tree.get_revision_id()
2106
parent_ids = ancestors[revision_id]
2107
for text_key in revision_keys[revision_id]:
2108
pb.update("Calculating text parents", processed_texts)
2109
processed_texts += 1
2110
candidate_parents = []
2111
for parent_id in parent_ids:
2112
parent_text_key = (text_key[0], parent_id)
2114
check_parent = parent_text_key not in \
2115
revision_keys[parent_id]
2117
# the parent parent_id is a ghost:
2118
check_parent = False
2119
# truncate the derived graph against this ghost.
2120
parent_text_key = None
2122
# look at the parent commit details inventories to
2123
# determine possible candidates in the per file graph.
2126
inv = inventory_cache[parent_id]
2128
inv = self.revision_tree(parent_id).inventory
2129
inventory_cache[parent_id] = inv
2131
parent_entry = inv[text_key[0]]
2132
except (KeyError, errors.NoSuchId):
2134
if parent_entry is not None:
2136
text_key[0], parent_entry.revision)
2138
parent_text_key = None
2139
if parent_text_key is not None:
2140
candidate_parents.append(
2141
text_key_cache[parent_text_key])
2142
parent_heads = text_graph.heads(candidate_parents)
2143
new_parents = list(parent_heads)
2144
new_parents.sort(key=lambda x:candidate_parents.index(x))
2145
if new_parents == []:
2146
new_parents = [NULL_REVISION]
2147
text_index[text_key] = new_parents
2149
for text_key in invalid_keys:
2150
text_index[text_key] = [NULL_REVISION]
2153
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2154
"""Get an iterable listing the keys of all the data introduced by a set
2157
The keys will be ordered so that the corresponding items can be safely
2158
fetched and inserted in that order.
2160
:returns: An iterable producing tuples of (knit-kind, file-id,
2161
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2162
'revisions'. file-id is None unless knit-kind is 'file'.
2164
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2167
for result in self._find_non_file_keys_to_fetch(revision_ids):
2170
def _find_file_keys_to_fetch(self, revision_ids, pb):
2171
# XXX: it's a bit weird to control the inventory weave caching in this
2172
# generator. Ideally the caching would be done in fetch.py I think. Or
2173
# maybe this generator should explicitly have the contract that it
2174
# should not be iterated until the previously yielded item has been
2176
inv_w = self.inventories
2178
# file ids that changed
2179
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2181
num_file_ids = len(file_ids)
2182
for file_id, altered_versions in file_ids.iteritems():
2184
pb.update("fetch texts", count, num_file_ids)
2186
yield ("file", file_id, altered_versions)
2188
def _find_non_file_keys_to_fetch(self, revision_ids):
2190
yield ("inventory", None, revision_ids)
2193
# XXX: Note ATM no callers actually pay attention to this return
2194
# instead they just use the list of revision ids and ignore
2195
# missing sigs. Consider removing this work entirely
2196
revisions_with_signatures = set(self.signatures.get_parent_map(
2197
[(r,) for r in revision_ids]))
2198
revisions_with_signatures = set(
2199
[r for (r,) in revisions_with_signatures])
2200
revisions_with_signatures.intersection_update(revision_ids)
2201
yield ("signatures", None, revisions_with_signatures)
2204
yield ("revisions", None, revision_ids)
2207
def get_inventory(self, revision_id):
2208
"""Get Inventory object by revision id."""
2209
return self.iter_inventories([revision_id]).next()
2211
def iter_inventories(self, revision_ids, ordering=None):
2212
"""Get many inventories by revision_ids.
2214
This will buffer some or all of the texts used in constructing the
2215
inventories in memory, but will only parse a single inventory at a
2218
:param revision_ids: The expected revision ids of the inventories.
2219
:param ordering: optional ordering, e.g. 'topological'. If not
2220
specified, the order of revision_ids will be preserved (by
2221
buffering if necessary).
2222
:return: An iterator of inventories.
2224
if ((None in revision_ids)
2225
or (_mod_revision.NULL_REVISION in revision_ids)):
2226
raise ValueError('cannot get null revision inventory')
2227
return self._iter_inventories(revision_ids, ordering)
2229
def _iter_inventories(self, revision_ids, ordering):
2230
"""single-document based inventory iteration."""
2231
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2232
for text, revision_id in inv_xmls:
2233
yield self.deserialise_inventory(revision_id, text)
2235
def _iter_inventory_xmls(self, revision_ids, ordering):
2236
if ordering is None:
2237
order_as_requested = True
2238
ordering = 'unordered'
2240
order_as_requested = False
2241
keys = [(revision_id,) for revision_id in revision_ids]
2244
if order_as_requested:
2245
key_iter = iter(keys)
2246
next_key = key_iter.next()
2247
stream = self.inventories.get_record_stream(keys, ordering, True)
2249
for record in stream:
2250
if record.storage_kind != 'absent':
2251
chunks = record.get_bytes_as('chunked')
2252
if order_as_requested:
2253
text_chunks[record.key] = chunks
2255
yield ''.join(chunks), record.key[-1]
2257
raise errors.NoSuchRevision(self, record.key)
2258
if order_as_requested:
2259
# Yield as many results as we can while preserving order.
2260
while next_key in text_chunks:
2261
chunks = text_chunks.pop(next_key)
2262
yield ''.join(chunks), next_key[-1]
2264
next_key = key_iter.next()
2265
except StopIteration:
2266
# We still want to fully consume the get_record_stream,
2267
# just in case it is not actually finished at this point
2271
def deserialise_inventory(self, revision_id, xml):
2272
"""Transform the xml into an inventory object.
2274
:param revision_id: The expected revision id of the inventory.
2275
:param xml: A serialised inventory.
2277
result = self._serializer.read_inventory_from_string(xml, revision_id,
2278
entry_cache=self._inventory_entry_cache)
2279
if result.revision_id != revision_id:
2280
raise AssertionError('revision id mismatch %s != %s' % (
2281
result.revision_id, revision_id))
2284
def serialise_inventory(self, inv):
2285
return self._serializer.write_inventory_to_string(inv)
2287
def _serialise_inventory_to_lines(self, inv):
2288
return self._serializer.write_inventory_to_lines(inv)
2290
def get_serializer_format(self):
2291
return self._serializer.format_num
2294
def get_inventory_xml(self, revision_id):
2295
"""Get inventory XML as a file object."""
2296
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2298
text, revision_id = texts.next()
2299
except StopIteration:
2300
raise errors.HistoryMissing(self, 'inventory', revision_id)
2304
def get_inventory_sha1(self, revision_id):
2305
"""Return the sha1 hash of the inventory entry
2307
return self.get_revision(revision_id).inventory_sha1
2309
def get_rev_id_for_revno(self, revno, known_pair):
2310
"""Return the revision id of a revno, given a later (revno, revid)
2311
pair in the same history.
2313
:return: if found (True, revid). If the available history ran out
2314
before reaching the revno, then this returns
2315
(False, (closest_revno, closest_revid)).
2317
known_revno, known_revid = known_pair
2318
partial_history = [known_revid]
2319
distance_from_known = known_revno - revno
2320
if distance_from_known < 0:
2322
'requested revno (%d) is later than given known revno (%d)'
2323
% (revno, known_revno))
2326
self, partial_history, stop_index=distance_from_known)
2327
except errors.RevisionNotPresent, err:
2328
if err.revision_id == known_revid:
2329
# The start revision (known_revid) wasn't found.
2331
# This is a stacked repository with no fallbacks, or a there's a
2332
# left-hand ghost. Either way, even though the revision named in
2333
# the error isn't in this repo, we know it's the next step in this
2334
# left-hand history.
2335
partial_history.append(err.revision_id)
2336
if len(partial_history) <= distance_from_known:
2337
# Didn't find enough history to get a revid for the revno.
2338
earliest_revno = known_revno - len(partial_history) + 1
2339
return (False, (earliest_revno, partial_history[-1]))
2340
if len(partial_history) - 1 > distance_from_known:
2341
raise AssertionError('_iter_for_revno returned too much history')
2342
return (True, partial_history[-1])
2344
def iter_reverse_revision_history(self, revision_id):
2345
"""Iterate backwards through revision ids in the lefthand history
2347
:param revision_id: The revision id to start with. All its lefthand
2348
ancestors will be traversed.
2350
graph = self.get_graph()
2351
next_id = revision_id
2353
if next_id in (None, _mod_revision.NULL_REVISION):
2356
parents = graph.get_parent_map([next_id])[next_id]
2358
raise errors.RevisionNotPresent(next_id, self)
2360
if len(parents) == 0:
2363
next_id = parents[0]
2366
def get_revision_inventory(self, revision_id):
2367
"""Return inventory of a past revision."""
2368
# TODO: Unify this with get_inventory()
2369
# bzr 0.0.6 and later imposes the constraint that the inventory_id
2370
# must be the same as its revision, so this is trivial.
2371
if revision_id is None:
2372
# This does not make sense: if there is no revision,
2373
# then it is the current tree inventory surely ?!
2374
# and thus get_root_id() is something that looks at the last
2375
# commit on the branch, and the get_root_id is an inventory check.
2376
raise NotImplementedError
2377
# return Inventory(self.get_root_id())
2379
return self.get_inventory(revision_id)
2381
def is_shared(self):
2382
"""Return True if this repository is flagged as a shared repository."""
2383
raise NotImplementedError(self.is_shared)
2386
def reconcile(self, other=None, thorough=False):
2387
"""Reconcile this repository."""
2388
from bzrlib.reconcile import RepoReconciler
2389
reconciler = RepoReconciler(self, thorough=thorough)
2390
reconciler.reconcile()
2393
def _refresh_data(self):
2394
"""Helper called from lock_* to ensure coherency with disk.
2396
The default implementation does nothing; it is however possible
2397
for repositories to maintain loaded indices across multiple locks
2398
by checking inside their implementation of this method to see
2399
whether their indices are still valid. This depends of course on
2400
the disk format being validatable in this manner. This method is
2401
also called by the refresh_data() public interface to cause a refresh
2402
to occur while in a write lock so that data inserted by a smart server
2403
push operation is visible on the client's instance of the physical
2408
def revision_tree(self, revision_id):
2409
"""Return Tree for a revision on this branch.
2411
`revision_id` may be NULL_REVISION for the empty tree revision.
2413
revision_id = _mod_revision.ensure_null(revision_id)
2414
# TODO: refactor this to use an existing revision object
2415
# so we don't need to read it in twice.
2416
if revision_id == _mod_revision.NULL_REVISION:
2417
return RevisionTree(self, Inventory(root_id=None),
2418
_mod_revision.NULL_REVISION)
2420
inv = self.get_revision_inventory(revision_id)
2421
return RevisionTree(self, inv, revision_id)
2423
def revision_trees(self, revision_ids):
2424
"""Return Trees for revisions in this repository.
2426
:param revision_ids: a sequence of revision-ids;
2427
a revision-id may not be None or 'null:'
2429
inventories = self.iter_inventories(revision_ids)
2430
for inv in inventories:
2431
yield RevisionTree(self, inv, inv.revision_id)
2433
def _filtered_revision_trees(self, revision_ids, file_ids):
2434
"""Return Tree for a revision on this branch with only some files.
2436
:param revision_ids: a sequence of revision-ids;
2437
a revision-id may not be None or 'null:'
2438
:param file_ids: if not None, the result is filtered
2439
so that only those file-ids, their parents and their
2440
children are included.
2442
inventories = self.iter_inventories(revision_ids)
2443
for inv in inventories:
2444
# Should we introduce a FilteredRevisionTree class rather
2445
# than pre-filter the inventory here?
2446
filtered_inv = inv.filter(file_ids)
2447
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
2450
def get_ancestry(self, revision_id, topo_sorted=True):
2451
"""Return a list of revision-ids integrated by a revision.
2453
The first element of the list is always None, indicating the origin
2454
revision. This might change when we have history horizons, or
2455
perhaps we should have a new API.
2457
This is topologically sorted.
2459
if _mod_revision.is_null(revision_id):
2461
if not self.has_revision(revision_id):
2462
raise errors.NoSuchRevision(self, revision_id)
2463
graph = self.get_graph()
2465
search = graph._make_breadth_first_searcher([revision_id])
2468
found, ghosts = search.next_with_ghosts()
2469
except StopIteration:
2472
if _mod_revision.NULL_REVISION in keys:
2473
keys.remove(_mod_revision.NULL_REVISION)
2475
parent_map = graph.get_parent_map(keys)
2476
keys = tsort.topo_sort(parent_map)
2477
return [None] + list(keys)
2479
def pack(self, hint=None):
2480
"""Compress the data within the repository.
2482
This operation only makes sense for some repository types. For other
2483
types it should be a no-op that just returns.
2485
This stub method does not require a lock, but subclasses should use
2486
@needs_write_lock as this is a long running call its reasonable to
2487
implicitly lock for the user.
2489
:param hint: If not supplied, the whole repository is packed.
2490
If supplied, the repository may use the hint parameter as a
2491
hint for the parts of the repository to pack. A hint can be
2492
obtained from the result of commit_write_group(). Out of
2493
date hints are simply ignored, because concurrent operations
2494
can obsolete them rapidly.
2497
def get_transaction(self):
2498
return self.control_files.get_transaction()
2500
def get_parent_map(self, revision_ids):
2501
"""See graph.StackedParentsProvider.get_parent_map"""
2502
# revisions index works in keys; this just works in revisions
2503
# therefore wrap and unwrap
2506
for revision_id in revision_ids:
2507
if revision_id == _mod_revision.NULL_REVISION:
2508
result[revision_id] = ()
2509
elif revision_id is None:
2510
raise ValueError('get_parent_map(None) is not valid')
2512
query_keys.append((revision_id ,))
2513
for ((revision_id,), parent_keys) in \
2514
self.revisions.get_parent_map(query_keys).iteritems():
2516
result[revision_id] = tuple(parent_revid
2517
for (parent_revid,) in parent_keys)
2519
result[revision_id] = (_mod_revision.NULL_REVISION,)
2522
def _make_parents_provider(self):
2525
def get_graph(self, other_repository=None):
2526
"""Return the graph walker for this repository format"""
2527
parents_provider = self._make_parents_provider()
2528
if (other_repository is not None and
2529
not self.has_same_location(other_repository)):
2530
parents_provider = graph.StackedParentsProvider(
2531
[parents_provider, other_repository._make_parents_provider()])
2532
return graph.Graph(parents_provider)
2534
def _get_versioned_file_checker(self, text_key_references=None):
2535
"""Return an object suitable for checking versioned files.
2537
:param text_key_references: if non-None, an already built
2538
dictionary mapping text keys ((fileid, revision_id) tuples)
2539
to whether they were referred to by the inventory of the
2540
revision_id that they contain. If None, this will be
2543
return _VersionedFileChecker(self,
2544
text_key_references=text_key_references)
2546
def revision_ids_to_search_result(self, result_set):
2547
"""Convert a set of revision ids to a graph SearchResult."""
2548
result_parents = set()
2549
for parents in self.get_graph().get_parent_map(
2550
result_set).itervalues():
2551
result_parents.update(parents)
2552
included_keys = result_set.intersection(result_parents)
2553
start_keys = result_set.difference(included_keys)
2554
exclude_keys = result_parents.difference(result_set)
2555
result = graph.SearchResult(start_keys, exclude_keys,
2556
len(result_set), result_set)
2560
def set_make_working_trees(self, new_value):
2561
"""Set the policy flag for making working trees when creating branches.
2563
This only applies to branches that use this repository.
2565
The default is 'True'.
2566
:param new_value: True to restore the default, False to disable making
2569
raise NotImplementedError(self.set_make_working_trees)
2571
def make_working_trees(self):
2572
"""Returns the policy for making working trees on new branches."""
2573
raise NotImplementedError(self.make_working_trees)
2576
def sign_revision(self, revision_id, gpg_strategy):
2577
plaintext = Testament.from_revision(self, revision_id).as_short_text()
2578
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2581
def has_signature_for_revision_id(self, revision_id):
2582
"""Query for a revision signature for revision_id in the repository."""
2583
if not self.has_revision(revision_id):
2584
raise errors.NoSuchRevision(self, revision_id)
2585
sig_present = (1 == len(
2586
self.signatures.get_parent_map([(revision_id,)])))
2590
def get_signature_text(self, revision_id):
2591
"""Return the text for a signature."""
2592
stream = self.signatures.get_record_stream([(revision_id,)],
2594
record = stream.next()
2595
if record.storage_kind == 'absent':
2596
raise errors.NoSuchRevision(self, revision_id)
2597
return record.get_bytes_as('fulltext')
2600
def check(self, revision_ids=None):
2601
"""Check consistency of all history of given revision_ids.
2603
Different repository implementations should override _check().
2605
:param revision_ids: A non-empty list of revision_ids whose ancestry
2606
will be checked. Typically the last revision_id of a branch.
2608
return self._check(revision_ids)
2610
def _check(self, revision_ids):
2611
result = check.Check(self)
2615
def _warn_if_deprecated(self):
2616
global _deprecation_warning_done
2617
if _deprecation_warning_done:
2619
_deprecation_warning_done = True
2620
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2621
% (self._format, self.bzrdir.transport.base))
2623
def supports_rich_root(self):
2624
return self._format.rich_root_data
2626
def _check_ascii_revisionid(self, revision_id, method):
2627
"""Private helper for ascii-only repositories."""
2628
# weave repositories refuse to store revisionids that are non-ascii.
2629
if revision_id is not None:
2630
# weaves require ascii revision ids.
2631
if isinstance(revision_id, unicode):
2633
revision_id.encode('ascii')
2634
except UnicodeEncodeError:
2635
raise errors.NonAsciiRevisionId(method, self)
2638
revision_id.decode('ascii')
2639
except UnicodeDecodeError:
2640
raise errors.NonAsciiRevisionId(method, self)
2642
def revision_graph_can_have_wrong_parents(self):
2643
"""Is it possible for this repository to have a revision graph with
2646
If True, then this repository must also implement
2647
_find_inconsistent_revision_parents so that check and reconcile can
2648
check for inconsistencies before proceeding with other checks that may
2649
depend on the revision index being consistent.
2651
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2654
# remove these delegates a while after bzr 0.15
2655
def __make_delegated(name, from_module):
2656
def _deprecated_repository_forwarder():
2657
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2658
% (name, from_module),
2661
m = __import__(from_module, globals(), locals(), [name])
2663
return getattr(m, name)
2664
except AttributeError:
2665
raise AttributeError('module %s has no name %s'
2667
globals()[name] = _deprecated_repository_forwarder
2670
'AllInOneRepository',
2671
'WeaveMetaDirRepository',
2672
'PreSplitOutRepositoryFormat',
2673
'RepositoryFormat4',
2674
'RepositoryFormat5',
2675
'RepositoryFormat6',
2676
'RepositoryFormat7',
2678
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2682
'RepositoryFormatKnit',
2683
'RepositoryFormatKnit1',
2685
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2688
def install_revision(repository, rev, revision_tree):
2689
"""Install all revision data into a repository."""
2690
install_revisions(repository, [(rev, revision_tree, None)])
2693
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2694
"""Install all revision data into a repository.
2696
Accepts an iterable of revision, tree, signature tuples. The signature
2699
repository.start_write_group()
2701
inventory_cache = lru_cache.LRUCache(10)
2702
for n, (revision, revision_tree, signature) in enumerate(iterable):
2703
_install_revision(repository, revision, revision_tree, signature,
2706
pb.update('Transferring revisions', n + 1, num_revisions)
2708
repository.abort_write_group()
2711
repository.commit_write_group()
2714
def _install_revision(repository, rev, revision_tree, signature,
2716
"""Install all revision data into a repository."""
2717
present_parents = []
2719
for p_id in rev.parent_ids:
2720
if repository.has_revision(p_id):
2721
present_parents.append(p_id)
2722
parent_trees[p_id] = repository.revision_tree(p_id)
2724
parent_trees[p_id] = repository.revision_tree(
2725
_mod_revision.NULL_REVISION)
2727
inv = revision_tree.inventory
2728
entries = inv.iter_entries()
2729
# backwards compatibility hack: skip the root id.
2730
if not repository.supports_rich_root():
2731
path, root = entries.next()
2732
if root.revision != rev.revision_id:
2733
raise errors.IncompatibleRevision(repr(repository))
2735
for path, ie in entries:
2736
text_keys[(ie.file_id, ie.revision)] = ie
2737
text_parent_map = repository.texts.get_parent_map(text_keys)
2738
missing_texts = set(text_keys) - set(text_parent_map)
2739
# Add the texts that are not already present
2740
for text_key in missing_texts:
2741
ie = text_keys[text_key]
2743
# FIXME: TODO: The following loop overlaps/duplicates that done by
2744
# commit to determine parents. There is a latent/real bug here where
2745
# the parents inserted are not those commit would do - in particular
2746
# they are not filtered by heads(). RBC, AB
2747
for revision, tree in parent_trees.iteritems():
2748
if ie.file_id not in tree:
2750
parent_id = tree.inventory[ie.file_id].revision
2751
if parent_id in text_parents:
2753
text_parents.append((ie.file_id, parent_id))
2754
lines = revision_tree.get_file(ie.file_id).readlines()
2755
repository.texts.add_lines(text_key, text_parents, lines)
2757
# install the inventory
2758
if repository._format._commit_inv_deltas and len(rev.parent_ids):
2759
# Cache this inventory
2760
inventory_cache[rev.revision_id] = inv
2762
basis_inv = inventory_cache[rev.parent_ids[0]]
2764
repository.add_inventory(rev.revision_id, inv, present_parents)
2766
delta = inv._make_delta(basis_inv)
2767
repository.add_inventory_by_delta(rev.parent_ids[0], delta,
2768
rev.revision_id, present_parents)
2770
repository.add_inventory(rev.revision_id, inv, present_parents)
2771
except errors.RevisionAlreadyPresent:
2773
if signature is not None:
2774
repository.add_signature_text(rev.revision_id, signature)
2775
repository.add_revision(rev.revision_id, rev, inv)
2778
class MetaDirRepository(Repository):
2779
"""Repositories in the new meta-dir layout.
2781
:ivar _transport: Transport for access to repository control files,
2782
typically pointing to .bzr/repository.
2785
def __init__(self, _format, a_bzrdir, control_files):
2786
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2787
self._transport = control_files._transport
2789
def is_shared(self):
2790
"""Return True if this repository is flagged as a shared repository."""
2791
return self._transport.has('shared-storage')
2794
def set_make_working_trees(self, new_value):
2795
"""Set the policy flag for making working trees when creating branches.
2797
This only applies to branches that use this repository.
2799
The default is 'True'.
2800
:param new_value: True to restore the default, False to disable making
2805
self._transport.delete('no-working-trees')
2806
except errors.NoSuchFile:
2809
self._transport.put_bytes('no-working-trees', '',
2810
mode=self.bzrdir._get_file_mode())
2812
def make_working_trees(self):
2813
"""Returns the policy for making working trees on new branches."""
2814
return not self._transport.has('no-working-trees')
2817
class MetaDirVersionedFileRepository(MetaDirRepository):
2818
"""Repositories in a meta-dir, that work via versioned file objects."""
2820
def __init__(self, _format, a_bzrdir, control_files):
2821
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2825
network_format_registry = registry.FormatRegistry()
2826
"""Registry of formats indexed by their network name.
2828
The network name for a repository format is an identifier that can be used when
2829
referring to formats with smart server operations. See
2830
RepositoryFormat.network_name() for more detail.
2834
format_registry = registry.FormatRegistry(network_format_registry)
2835
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2837
This can contain either format instances themselves, or classes/factories that
2838
can be called to obtain one.
2842
#####################################################################
2843
# Repository Formats
2845
class RepositoryFormat(object):
2846
"""A repository format.
2848
Formats provide four things:
2849
* An initialization routine to construct repository data on disk.
2850
* a optional format string which is used when the BzrDir supports
2852
* an open routine which returns a Repository instance.
2853
* A network name for referring to the format in smart server RPC
2856
There is one and only one Format subclass for each on-disk format. But
2857
there can be one Repository subclass that is used for several different
2858
formats. The _format attribute on a Repository instance can be used to
2859
determine the disk format.
2861
Formats are placed in a registry by their format string for reference
2862
during opening. These should be subclasses of RepositoryFormat for
2865
Once a format is deprecated, just deprecate the initialize and open
2866
methods on the format class. Do not deprecate the object, as the
2867
object may be created even when a repository instance hasn't been
2870
Common instance attributes:
2871
_matchingbzrdir - the bzrdir format that the repository format was
2872
originally written to work with. This can be used if manually
2873
constructing a bzrdir and repository, or more commonly for test suite
2877
# Set to True or False in derived classes. True indicates that the format
2878
# supports ghosts gracefully.
2879
supports_ghosts = None
2880
# Can this repository be given external locations to lookup additional
2881
# data. Set to True or False in derived classes.
2882
supports_external_lookups = None
2883
# Does this format support CHK bytestring lookups. Set to True or False in
2885
supports_chks = None
2886
# Should commit add an inventory, or an inventory delta to the repository.
2887
_commit_inv_deltas = True
2888
# What order should fetch operations request streams in?
2889
# The default is unordered as that is the cheapest for an origin to
2891
_fetch_order = 'unordered'
2892
# Does this repository format use deltas that can be fetched as-deltas ?
2893
# (E.g. knits, where the knit deltas can be transplanted intact.
2894
# We default to False, which will ensure that enough data to get
2895
# a full text out of any fetch stream will be grabbed.
2896
_fetch_uses_deltas = False
2897
# Should fetch trigger a reconcile after the fetch? Only needed for
2898
# some repository formats that can suffer internal inconsistencies.
2899
_fetch_reconcile = False
2900
# Does this format have < O(tree_size) delta generation. Used to hint what
2901
# code path for commit, amongst other things.
2903
# Does doing a pack operation compress data? Useful for the pack UI command
2904
# (so if there is one pack, the operation can still proceed because it may
2905
# help), and for fetching when data won't have come from the same
2907
pack_compresses = False
2910
return "<%s>" % self.__class__.__name__
2912
def __eq__(self, other):
2913
# format objects are generally stateless
2914
return isinstance(other, self.__class__)
2916
def __ne__(self, other):
2917
return not self == other
2920
def find_format(klass, a_bzrdir):
2921
"""Return the format for the repository object in a_bzrdir.
2923
This is used by bzr native formats that have a "format" file in
2924
the repository. Other methods may be used by different types of
2928
transport = a_bzrdir.get_repository_transport(None)
2929
format_string = transport.get("format").read()
2930
return format_registry.get(format_string)
2931
except errors.NoSuchFile:
2932
raise errors.NoRepositoryPresent(a_bzrdir)
2934
raise errors.UnknownFormatError(format=format_string,
2938
def register_format(klass, format):
2939
format_registry.register(format.get_format_string(), format)
2942
def unregister_format(klass, format):
2943
format_registry.remove(format.get_format_string())
2946
def get_default_format(klass):
2947
"""Return the current default format."""
2948
from bzrlib import bzrdir
2949
return bzrdir.format_registry.make_bzrdir('default').repository_format
2951
def get_format_string(self):
2952
"""Return the ASCII format string that identifies this format.
2954
Note that in pre format ?? repositories the format string is
2955
not permitted nor written to disk.
2957
raise NotImplementedError(self.get_format_string)
2959
def get_format_description(self):
2960
"""Return the short description for this format."""
2961
raise NotImplementedError(self.get_format_description)
2963
# TODO: this shouldn't be in the base class, it's specific to things that
2964
# use weaves or knits -- mbp 20070207
2965
def _get_versioned_file_store(self,
2970
versionedfile_class=None,
2971
versionedfile_kwargs={},
2973
if versionedfile_class is None:
2974
versionedfile_class = self._versionedfile_class
2975
weave_transport = control_files._transport.clone(name)
2976
dir_mode = control_files._dir_mode
2977
file_mode = control_files._file_mode
2978
return VersionedFileStore(weave_transport, prefixed=prefixed,
2980
file_mode=file_mode,
2981
versionedfile_class=versionedfile_class,
2982
versionedfile_kwargs=versionedfile_kwargs,
2985
def initialize(self, a_bzrdir, shared=False):
2986
"""Initialize a repository of this format in a_bzrdir.
2988
:param a_bzrdir: The bzrdir to put the new repository in it.
2989
:param shared: The repository should be initialized as a sharable one.
2990
:returns: The new repository object.
2992
This may raise UninitializableFormat if shared repository are not
2993
compatible the a_bzrdir.
2995
raise NotImplementedError(self.initialize)
2997
def is_supported(self):
2998
"""Is this format supported?
3000
Supported formats must be initializable and openable.
3001
Unsupported formats may not support initialization or committing or
3002
some other features depending on the reason for not being supported.
3006
def network_name(self):
3007
"""A simple byte string uniquely identifying this format for RPC calls.
3009
MetaDir repository formats use their disk format string to identify the
3010
repository over the wire. All in one formats such as bzr < 0.8, and
3011
foreign formats like svn/git and hg should use some marker which is
3012
unique and immutable.
3014
raise NotImplementedError(self.network_name)
3016
def check_conversion_target(self, target_format):
3017
raise NotImplementedError(self.check_conversion_target)
3019
def open(self, a_bzrdir, _found=False):
3020
"""Return an instance of this format for the bzrdir a_bzrdir.
3022
_found is a private parameter, do not use it.
3024
raise NotImplementedError(self.open)
3027
class MetaDirRepositoryFormat(RepositoryFormat):
3028
"""Common base class for the new repositories using the metadir layout."""
3030
rich_root_data = False
3031
supports_tree_reference = False
3032
supports_external_lookups = False
3035
def _matchingbzrdir(self):
3036
matching = bzrdir.BzrDirMetaFormat1()
3037
matching.repository_format = self
3041
super(MetaDirRepositoryFormat, self).__init__()
3043
def _create_control_files(self, a_bzrdir):
3044
"""Create the required files and the initial control_files object."""
3045
# FIXME: RBC 20060125 don't peek under the covers
3046
# NB: no need to escape relative paths that are url safe.
3047
repository_transport = a_bzrdir.get_repository_transport(self)
3048
control_files = lockable_files.LockableFiles(repository_transport,
3049
'lock', lockdir.LockDir)
3050
control_files.create_lock()
3051
return control_files
3053
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
3054
"""Upload the initial blank content."""
3055
control_files = self._create_control_files(a_bzrdir)
3056
control_files.lock_write()
3057
transport = control_files._transport
3059
utf8_files += [('shared-storage', '')]
3061
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
3062
for (filename, content_stream) in files:
3063
transport.put_file(filename, content_stream,
3064
mode=a_bzrdir._get_file_mode())
3065
for (filename, content_bytes) in utf8_files:
3066
transport.put_bytes_non_atomic(filename, content_bytes,
3067
mode=a_bzrdir._get_file_mode())
3069
control_files.unlock()
3071
def network_name(self):
3072
"""Metadir formats have matching disk and network format strings."""
3073
return self.get_format_string()
3076
# Pre-0.8 formats that don't have a disk format string (because they are
3077
# versioned by the matching control directory). We use the control directories
3078
# disk format string as a key for the network_name because they meet the
3079
# constraints (simple string, unique, immutable).
3080
network_format_registry.register_lazy(
3081
"Bazaar-NG branch, format 5\n",
3082
'bzrlib.repofmt.weaverepo',
3083
'RepositoryFormat5',
3085
network_format_registry.register_lazy(
3086
"Bazaar-NG branch, format 6\n",
3087
'bzrlib.repofmt.weaverepo',
3088
'RepositoryFormat6',
3091
# formats which have no format string are not discoverable or independently
3092
# creatable on disk, so are not registered in format_registry. They're
3093
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
3094
# needed, it's constructed directly by the BzrDir. Non-native formats where
3095
# the repository is not separately opened are similar.
3097
format_registry.register_lazy(
3098
'Bazaar-NG Repository format 7',
3099
'bzrlib.repofmt.weaverepo',
3103
format_registry.register_lazy(
3104
'Bazaar-NG Knit Repository Format 1',
3105
'bzrlib.repofmt.knitrepo',
3106
'RepositoryFormatKnit1',
3109
format_registry.register_lazy(
3110
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
3111
'bzrlib.repofmt.knitrepo',
3112
'RepositoryFormatKnit3',
3115
format_registry.register_lazy(
3116
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
3117
'bzrlib.repofmt.knitrepo',
3118
'RepositoryFormatKnit4',
3121
# Pack-based formats. There is one format for pre-subtrees, and one for
3122
# post-subtrees to allow ease of testing.
3123
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
3124
format_registry.register_lazy(
3125
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
3126
'bzrlib.repofmt.pack_repo',
3127
'RepositoryFormatKnitPack1',
3129
format_registry.register_lazy(
3130
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
3131
'bzrlib.repofmt.pack_repo',
3132
'RepositoryFormatKnitPack3',
3134
format_registry.register_lazy(
3135
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
3136
'bzrlib.repofmt.pack_repo',
3137
'RepositoryFormatKnitPack4',
3139
format_registry.register_lazy(
3140
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
3141
'bzrlib.repofmt.pack_repo',
3142
'RepositoryFormatKnitPack5',
3144
format_registry.register_lazy(
3145
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
3146
'bzrlib.repofmt.pack_repo',
3147
'RepositoryFormatKnitPack5RichRoot',
3149
format_registry.register_lazy(
3150
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
3151
'bzrlib.repofmt.pack_repo',
3152
'RepositoryFormatKnitPack5RichRootBroken',
3154
format_registry.register_lazy(
3155
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
3156
'bzrlib.repofmt.pack_repo',
3157
'RepositoryFormatKnitPack6',
3159
format_registry.register_lazy(
3160
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
3161
'bzrlib.repofmt.pack_repo',
3162
'RepositoryFormatKnitPack6RichRoot',
3165
# Development formats.
3166
# Obsolete but kept pending a CHK based subtree format.
3167
format_registry.register_lazy(
3168
("Bazaar development format 2 with subtree support "
3169
"(needs bzr.dev from before 1.8)\n"),
3170
'bzrlib.repofmt.pack_repo',
3171
'RepositoryFormatPackDevelopment2Subtree',
3174
# 1.14->1.16 go below here
3175
format_registry.register_lazy(
3176
'Bazaar development format - group compression and chk inventory'
3177
' (needs bzr.dev from 1.14)\n',
3178
'bzrlib.repofmt.groupcompress_repo',
3179
'RepositoryFormatCHK1',
3182
format_registry.register_lazy(
3183
'Bazaar development format - chk repository with bencode revision '
3184
'serialization (needs bzr.dev from 1.16)\n',
3185
'bzrlib.repofmt.groupcompress_repo',
3186
'RepositoryFormatCHK2',
3188
format_registry.register_lazy(
3189
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3190
'bzrlib.repofmt.groupcompress_repo',
3191
'RepositoryFormat2a',
3195
class InterRepository(InterObject):
3196
"""This class represents operations taking place between two repositories.
3198
Its instances have methods like copy_content and fetch, and contain
3199
references to the source and target repositories these operations can be
3202
Often we will provide convenience methods on 'repository' which carry out
3203
operations with another repository - they will always forward to
3204
InterRepository.get(other).method_name(parameters).
3207
_walk_to_common_revisions_batch_size = 50
3209
"""The available optimised InterRepository types."""
3212
def copy_content(self, revision_id=None):
3213
"""Make a complete copy of the content in self into destination.
3215
This is a destructive operation! Do not use it on existing
3218
:param revision_id: Only copy the content needed to construct
3219
revision_id and its parents.
3222
self.target.set_make_working_trees(self.source.make_working_trees())
3223
except NotImplementedError:
3225
self.target.fetch(self.source, revision_id=revision_id)
3228
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3230
"""Fetch the content required to construct revision_id.
3232
The content is copied from self.source to self.target.
3234
:param revision_id: if None all content is copied, if NULL_REVISION no
3236
:param pb: optional progress bar to use for progress reports. If not
3237
provided a default one will be created.
3240
from bzrlib.fetch import RepoFetcher
3241
f = RepoFetcher(to_repository=self.target,
3242
from_repository=self.source,
3243
last_revision=revision_id,
3244
fetch_spec=fetch_spec,
3245
pb=pb, find_ghosts=find_ghosts)
3247
def _walk_to_common_revisions(self, revision_ids):
3248
"""Walk out from revision_ids in source to revisions target has.
3250
:param revision_ids: The start point for the search.
3251
:return: A set of revision ids.
3253
target_graph = self.target.get_graph()
3254
revision_ids = frozenset(revision_ids)
3255
missing_revs = set()
3256
source_graph = self.source.get_graph()
3257
# ensure we don't pay silly lookup costs.
3258
searcher = source_graph._make_breadth_first_searcher(revision_ids)
3259
null_set = frozenset([_mod_revision.NULL_REVISION])
3260
searcher_exhausted = False
3264
# Iterate the searcher until we have enough next_revs
3265
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3267
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3268
next_revs.update(next_revs_part)
3269
ghosts.update(ghosts_part)
3270
except StopIteration:
3271
searcher_exhausted = True
3273
# If there are ghosts in the source graph, and the caller asked for
3274
# them, make sure that they are present in the target.
3275
# We don't care about other ghosts as we can't fetch them and
3276
# haven't been asked to.
3277
ghosts_to_check = set(revision_ids.intersection(ghosts))
3278
revs_to_get = set(next_revs).union(ghosts_to_check)
3280
have_revs = set(target_graph.get_parent_map(revs_to_get))
3281
# we always have NULL_REVISION present.
3282
have_revs = have_revs.union(null_set)
3283
# Check if the target is missing any ghosts we need.
3284
ghosts_to_check.difference_update(have_revs)
3286
# One of the caller's revision_ids is a ghost in both the
3287
# source and the target.
3288
raise errors.NoSuchRevision(
3289
self.source, ghosts_to_check.pop())
3290
missing_revs.update(next_revs - have_revs)
3291
# Because we may have walked past the original stop point, make
3292
# sure everything is stopped
3293
stop_revs = searcher.find_seen_ancestors(have_revs)
3294
searcher.stop_searching_any(stop_revs)
3295
if searcher_exhausted:
3297
return searcher.get_result()
3300
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3301
"""Return the revision ids that source has that target does not.
3303
:param revision_id: only return revision ids included by this
3305
:param find_ghosts: If True find missing revisions in deep history
3306
rather than just finding the surface difference.
3307
:return: A bzrlib.graph.SearchResult.
3309
# stop searching at found target revisions.
3310
if not find_ghosts and revision_id is not None:
3311
return self._walk_to_common_revisions([revision_id])
3312
# generic, possibly worst case, slow code path.
3313
target_ids = set(self.target.all_revision_ids())
3314
if revision_id is not None:
3315
source_ids = self.source.get_ancestry(revision_id)
3316
if source_ids[0] is not None:
3317
raise AssertionError()
3320
source_ids = self.source.all_revision_ids()
3321
result_set = set(source_ids).difference(target_ids)
3322
return self.source.revision_ids_to_search_result(result_set)
3325
def _same_model(source, target):
3326
"""True if source and target have the same data representation.
3328
Note: this is always called on the base class; overriding it in a
3329
subclass will have no effect.
3332
InterRepository._assert_same_model(source, target)
3334
except errors.IncompatibleRepositories, e:
3338
def _assert_same_model(source, target):
3339
"""Raise an exception if two repositories do not use the same model.
3341
if source.supports_rich_root() != target.supports_rich_root():
3342
raise errors.IncompatibleRepositories(source, target,
3343
"different rich-root support")
3344
if source._serializer != target._serializer:
3345
raise errors.IncompatibleRepositories(source, target,
3346
"different serializers")
3349
class InterSameDataRepository(InterRepository):
3350
"""Code for converting between repositories that represent the same data.
3352
Data format and model must match for this to work.
3356
def _get_repo_format_to_test(self):
3357
"""Repository format for testing with.
3359
InterSameData can pull from subtree to subtree and from non-subtree to
3360
non-subtree, so we test this with the richest repository format.
3362
from bzrlib.repofmt import knitrepo
3363
return knitrepo.RepositoryFormatKnit3()
3366
def is_compatible(source, target):
3367
return InterRepository._same_model(source, target)
3370
class InterWeaveRepo(InterSameDataRepository):
3371
"""Optimised code paths between Weave based repositories.
3373
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
3374
implemented lazy inter-object optimisation.
3378
def _get_repo_format_to_test(self):
3379
from bzrlib.repofmt import weaverepo
3380
return weaverepo.RepositoryFormat7()
3383
def is_compatible(source, target):
3384
"""Be compatible with known Weave formats.
3386
We don't test for the stores being of specific types because that
3387
could lead to confusing results, and there is no need to be
3390
from bzrlib.repofmt.weaverepo import (
3396
return (isinstance(source._format, (RepositoryFormat5,
3398
RepositoryFormat7)) and
3399
isinstance(target._format, (RepositoryFormat5,
3401
RepositoryFormat7)))
3402
except AttributeError:
3406
def copy_content(self, revision_id=None):
3407
"""See InterRepository.copy_content()."""
3408
# weave specific optimised path:
3410
self.target.set_make_working_trees(self.source.make_working_trees())
3411
except (errors.RepositoryUpgradeRequired, NotImplemented):
3413
# FIXME do not peek!
3414
if self.source._transport.listable():
3415
pb = ui.ui_factory.nested_progress_bar()
3417
self.target.texts.insert_record_stream(
3418
self.source.texts.get_record_stream(
3419
self.source.texts.keys(), 'topological', False))
3420
pb.update('copying inventory', 0, 1)
3421
self.target.inventories.insert_record_stream(
3422
self.source.inventories.get_record_stream(
3423
self.source.inventories.keys(), 'topological', False))
3424
self.target.signatures.insert_record_stream(
3425
self.source.signatures.get_record_stream(
3426
self.source.signatures.keys(),
3428
self.target.revisions.insert_record_stream(
3429
self.source.revisions.get_record_stream(
3430
self.source.revisions.keys(),
3431
'topological', True))
3435
self.target.fetch(self.source, revision_id=revision_id)
3438
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3439
"""See InterRepository.missing_revision_ids()."""
3440
# we want all revisions to satisfy revision_id in source.
3441
# but we don't want to stat every file here and there.
3442
# we want then, all revisions other needs to satisfy revision_id
3443
# checked, but not those that we have locally.
3444
# so the first thing is to get a subset of the revisions to
3445
# satisfy revision_id in source, and then eliminate those that
3446
# we do already have.
3447
# this is slow on high latency connection to self, but as this
3448
# disk format scales terribly for push anyway due to rewriting
3449
# inventory.weave, this is considered acceptable.
3451
if revision_id is not None:
3452
source_ids = self.source.get_ancestry(revision_id)
3453
if source_ids[0] is not None:
3454
raise AssertionError()
3457
source_ids = self.source._all_possible_ids()
3458
source_ids_set = set(source_ids)
3459
# source_ids is the worst possible case we may need to pull.
3460
# now we want to filter source_ids against what we actually
3461
# have in target, but don't try to check for existence where we know
3462
# we do not have a revision as that would be pointless.
3463
target_ids = set(self.target._all_possible_ids())
3464
possibly_present_revisions = target_ids.intersection(source_ids_set)
3465
actually_present_revisions = set(
3466
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3467
required_revisions = source_ids_set.difference(actually_present_revisions)
3468
if revision_id is not None:
3469
# we used get_ancestry to determine source_ids then we are assured all
3470
# revisions referenced are present as they are installed in topological order.
3471
# and the tip revision was validated by get_ancestry.
3472
result_set = required_revisions
3474
# if we just grabbed the possibly available ids, then
3475
# we only have an estimate of whats available and need to validate
3476
# that against the revision records.
3478
self.source._eliminate_revisions_not_present(required_revisions))
3479
return self.source.revision_ids_to_search_result(result_set)
3482
class InterKnitRepo(InterSameDataRepository):
3483
"""Optimised code paths between Knit based repositories."""
3486
def _get_repo_format_to_test(self):
3487
from bzrlib.repofmt import knitrepo
3488
return knitrepo.RepositoryFormatKnit1()
3491
def is_compatible(source, target):
3492
"""Be compatible with known Knit formats.
3494
We don't test for the stores being of specific types because that
3495
could lead to confusing results, and there is no need to be
3498
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
3500
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
3501
isinstance(target._format, RepositoryFormatKnit))
3502
except AttributeError:
3504
return are_knits and InterRepository._same_model(source, target)
3507
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3508
"""See InterRepository.missing_revision_ids()."""
3509
if revision_id is not None:
3510
source_ids = self.source.get_ancestry(revision_id)
3511
if source_ids[0] is not None:
3512
raise AssertionError()
3515
source_ids = self.source.all_revision_ids()
3516
source_ids_set = set(source_ids)
3517
# source_ids is the worst possible case we may need to pull.
3518
# now we want to filter source_ids against what we actually
3519
# have in target, but don't try to check for existence where we know
3520
# we do not have a revision as that would be pointless.
3521
target_ids = set(self.target.all_revision_ids())
3522
possibly_present_revisions = target_ids.intersection(source_ids_set)
3523
actually_present_revisions = set(
3524
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3525
required_revisions = source_ids_set.difference(actually_present_revisions)
3526
if revision_id is not None:
3527
# we used get_ancestry to determine source_ids then we are assured all
3528
# revisions referenced are present as they are installed in topological order.
3529
# and the tip revision was validated by get_ancestry.
3530
result_set = required_revisions
3532
# if we just grabbed the possibly available ids, then
3533
# we only have an estimate of whats available and need to validate
3534
# that against the revision records.
3536
self.source._eliminate_revisions_not_present(required_revisions))
3537
return self.source.revision_ids_to_search_result(result_set)
3540
class InterDifferingSerializer(InterRepository):
3543
def _get_repo_format_to_test(self):
3547
def is_compatible(source, target):
3548
"""Be compatible with Knit2 source and Knit3 target"""
3549
# This is redundant with format.check_conversion_target(), however that
3550
# raises an exception, and we just want to say "False" as in we won't
3551
# support converting between these formats.
3552
if 'IDS:never' in debug.debug_flags:
3554
if source.supports_rich_root() and not target.supports_rich_root():
3556
if (source._format.supports_tree_reference
3557
and not target._format.supports_tree_reference):
3559
if target._fallback_repositories and target._format.supports_chks:
3560
# IDS doesn't know how to copy CHKs for the parent inventories it
3561
# adds to stacked repos.
3563
if 'IDS:always' in debug.debug_flags:
3565
# Only use this code path for local source and target. IDS does far
3566
# too much IO (both bandwidth and roundtrips) over a network.
3567
if not source.bzrdir.transport.base.startswith('file:///'):
3569
if not target.bzrdir.transport.base.startswith('file:///'):
3573
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3574
"""Get the best delta and base for this revision.
3576
:return: (basis_id, delta)
3578
possible_trees = [(parent_id, cache[parent_id])
3579
for parent_id in parent_ids
3580
if parent_id in cache]
3581
if len(possible_trees) == 0:
3582
# There either aren't any parents, or the parents aren't in the
3583
# cache, so just use the last converted tree
3584
possible_trees.append((basis_id, cache[basis_id]))
3586
for basis_id, basis_tree in possible_trees:
3587
delta = tree.inventory._make_delta(basis_tree.inventory)
3588
deltas.append((len(delta), basis_id, delta))
3590
return deltas[0][1:]
3592
def _fetch_batch(self, revision_ids, basis_id, cache):
3593
"""Fetch across a few revisions.
3595
:param revision_ids: The revisions to copy
3596
:param basis_id: The revision_id of a tree that must be in cache, used
3597
as a basis for delta when no other base is available
3598
:param cache: A cache of RevisionTrees that we can use.
3599
:return: The revision_id of the last converted tree. The RevisionTree
3600
for it will be in cache
3602
# Walk though all revisions; get inventory deltas, copy referenced
3603
# texts that delta references, insert the delta, revision and
3605
root_keys_to_create = set()
3608
pending_revisions = []
3609
parent_map = self.source.get_parent_map(revision_ids)
3610
for tree in self.source.revision_trees(revision_ids):
3611
current_revision_id = tree.get_revision_id()
3612
parent_ids = parent_map.get(current_revision_id, ())
3613
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3615
if self._converting_to_rich_root:
3616
self._revision_id_to_root_id[current_revision_id] = \
3618
# Find text entries that need to be copied
3619
for old_path, new_path, file_id, entry in delta:
3620
if new_path is not None:
3623
if not self.target.supports_rich_root():
3624
# The target doesn't support rich root, so we don't
3627
if self._converting_to_rich_root:
3628
# This can't be copied normally, we have to insert
3630
root_keys_to_create.add((file_id, entry.revision))
3632
text_keys.add((file_id, entry.revision))
3633
revision = self.source.get_revision(current_revision_id)
3634
pending_deltas.append((basis_id, delta,
3635
current_revision_id, revision.parent_ids))
3636
pending_revisions.append(revision)
3637
cache[current_revision_id] = tree
3638
basis_id = current_revision_id
3640
from_texts = self.source.texts
3641
to_texts = self.target.texts
3642
if root_keys_to_create:
3643
from bzrlib.fetch import _new_root_data_stream
3644
root_stream = _new_root_data_stream(
3645
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3647
to_texts.insert_record_stream(root_stream)
3648
to_texts.insert_record_stream(from_texts.get_record_stream(
3649
text_keys, self.target._format._fetch_order,
3650
not self.target._format._fetch_uses_deltas))
3651
# insert inventory deltas
3652
for delta in pending_deltas:
3653
self.target.add_inventory_by_delta(*delta)
3654
if self.target._fallback_repositories:
3655
# Make sure this stacked repository has all the parent inventories
3656
# for the new revisions that we are about to insert. We do this
3657
# before adding the revisions so that no revision is added until
3658
# all the inventories it may depend on are added.
3660
revision_ids = set()
3661
for revision in pending_revisions:
3662
revision_ids.add(revision.revision_id)
3663
parent_ids.update(revision.parent_ids)
3664
parent_ids.difference_update(revision_ids)
3665
parent_ids.discard(_mod_revision.NULL_REVISION)
3666
parent_map = self.source.get_parent_map(parent_ids)
3667
for parent_tree in self.source.revision_trees(parent_ids):
3668
basis_id, delta = self._get_delta_for_revision(tree, parent_ids, basis_id, cache)
3669
current_revision_id = parent_tree.get_revision_id()
3670
parents_parents = parent_map[current_revision_id]
3671
self.target.add_inventory_by_delta(
3672
basis_id, delta, current_revision_id, parents_parents)
3673
# insert signatures and revisions
3674
for revision in pending_revisions:
3676
signature = self.source.get_signature_text(
3677
revision.revision_id)
3678
self.target.add_signature_text(revision.revision_id,
3680
except errors.NoSuchRevision:
3682
self.target.add_revision(revision.revision_id, revision)
3685
def _fetch_all_revisions(self, revision_ids, pb):
3686
"""Fetch everything for the list of revisions.
3688
:param revision_ids: The list of revisions to fetch. Must be in
3690
:param pb: A ProgressTask
3693
basis_id, basis_tree = self._get_basis(revision_ids[0])
3695
cache = lru_cache.LRUCache(100)
3696
cache[basis_id] = basis_tree
3697
del basis_tree # We don't want to hang on to it here
3699
for offset in range(0, len(revision_ids), batch_size):
3700
self.target.start_write_group()
3702
pb.update('Transferring revisions', offset,
3704
batch = revision_ids[offset:offset+batch_size]
3705
basis_id = self._fetch_batch(batch, basis_id, cache)
3707
self.target.abort_write_group()
3710
hint = self.target.commit_write_group()
3713
if hints and self.target._format.pack_compresses:
3714
self.target.pack(hint=hints)
3715
pb.update('Transferring revisions', len(revision_ids),
3719
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3721
"""See InterRepository.fetch()."""
3722
if fetch_spec is not None:
3723
raise AssertionError("Not implemented yet...")
3724
if (not self.source.supports_rich_root()
3725
and self.target.supports_rich_root()):
3726
self._converting_to_rich_root = True
3727
self._revision_id_to_root_id = {}
3729
self._converting_to_rich_root = False
3730
revision_ids = self.target.search_missing_revision_ids(self.source,
3731
revision_id, find_ghosts=find_ghosts).get_keys()
3732
if not revision_ids:
3734
revision_ids = tsort.topo_sort(
3735
self.source.get_graph().get_parent_map(revision_ids))
3736
if not revision_ids:
3738
# Walk though all revisions; get inventory deltas, copy referenced
3739
# texts that delta references, insert the delta, revision and
3742
my_pb = ui.ui_factory.nested_progress_bar()
3745
symbol_versioning.warn(
3746
symbol_versioning.deprecated_in((1, 14, 0))
3747
% "pb parameter to fetch()")
3750
self._fetch_all_revisions(revision_ids, pb)
3752
if my_pb is not None:
3754
return len(revision_ids), 0
3756
def _get_basis(self, first_revision_id):
3757
"""Get a revision and tree which exists in the target.
3759
This assumes that first_revision_id is selected for transmission
3760
because all other ancestors are already present. If we can't find an
3761
ancestor we fall back to NULL_REVISION since we know that is safe.
3763
:return: (basis_id, basis_tree)
3765
first_rev = self.source.get_revision(first_revision_id)
3767
basis_id = first_rev.parent_ids[0]
3768
# only valid as a basis if the target has it
3769
self.target.get_revision(basis_id)
3770
# Try to get a basis tree - if its a ghost it will hit the
3771
# NoSuchRevision case.
3772
basis_tree = self.source.revision_tree(basis_id)
3773
except (IndexError, errors.NoSuchRevision):
3774
basis_id = _mod_revision.NULL_REVISION
3775
basis_tree = self.source.revision_tree(basis_id)
3776
return basis_id, basis_tree
3779
InterRepository.register_optimiser(InterDifferingSerializer)
3780
InterRepository.register_optimiser(InterSameDataRepository)
3781
InterRepository.register_optimiser(InterWeaveRepo)
3782
InterRepository.register_optimiser(InterKnitRepo)
3785
class CopyConverter(object):
3786
"""A repository conversion tool which just performs a copy of the content.
3788
This is slow but quite reliable.
3791
def __init__(self, target_format):
3792
"""Create a CopyConverter.
3794
:param target_format: The format the resulting repository should be.
3796
self.target_format = target_format
3798
def convert(self, repo, pb):
3799
"""Perform the conversion of to_convert, giving feedback via pb.
3801
:param to_convert: The disk object to convert.
3802
:param pb: a progress bar to use for progress information.
3807
# this is only useful with metadir layouts - separated repo content.
3808
# trigger an assertion if not such
3809
repo._format.get_format_string()
3810
self.repo_dir = repo.bzrdir
3811
self.step('Moving repository to repository.backup')
3812
self.repo_dir.transport.move('repository', 'repository.backup')
3813
backup_transport = self.repo_dir.transport.clone('repository.backup')
3814
repo._format.check_conversion_target(self.target_format)
3815
self.source_repo = repo._format.open(self.repo_dir,
3817
_override_transport=backup_transport)
3818
self.step('Creating new repository')
3819
converted = self.target_format.initialize(self.repo_dir,
3820
self.source_repo.is_shared())
3821
converted.lock_write()
3823
self.step('Copying content into repository.')
3824
self.source_repo.copy_content_into(converted)
3827
self.step('Deleting old repository content.')
3828
self.repo_dir.transport.delete_tree('repository.backup')
3829
self.pb.note('repository converted')
3831
def step(self, message):
3832
"""Update the pb by a step."""
3834
self.pb.update(message, self.count, self.total)
3846
def _unescaper(match, _map=_unescape_map):
3847
code = match.group(1)
3851
if not code.startswith('#'):
3853
return unichr(int(code[1:])).encode('utf8')
3859
def _unescape_xml(data):
3860
"""Unescape predefined XML entities in a string of data."""
3862
if _unescape_re is None:
3863
_unescape_re = re.compile('\&([^;]*);')
3864
return _unescape_re.sub(_unescaper, data)
3867
class _VersionedFileChecker(object):
3869
def __init__(self, repository, text_key_references=None):
3870
self.repository = repository
3871
self.text_index = self.repository._generate_text_key_index(
3872
text_key_references=text_key_references)
3874
def calculate_file_version_parents(self, text_key):
3875
"""Calculate the correct parents for a file version according to
3878
parent_keys = self.text_index[text_key]
3879
if parent_keys == [_mod_revision.NULL_REVISION]:
3881
return tuple(parent_keys)
3883
def check_file_version_parents(self, texts, progress_bar=None):
3884
"""Check the parents stored in a versioned file are correct.
3886
It also detects file versions that are not referenced by their
3887
corresponding revision's inventory.
3889
:returns: A tuple of (wrong_parents, dangling_file_versions).
3890
wrong_parents is a dict mapping {revision_id: (stored_parents,
3891
correct_parents)} for each revision_id where the stored parents
3892
are not correct. dangling_file_versions is a set of (file_id,
3893
revision_id) tuples for versions that are present in this versioned
3894
file, but not used by the corresponding inventory.
3897
self.file_ids = set([file_id for file_id, _ in
3898
self.text_index.iterkeys()])
3899
# text keys is now grouped by file_id
3900
n_versions = len(self.text_index)
3901
progress_bar.update('loading text store', 0, n_versions)
3902
parent_map = self.repository.texts.get_parent_map(self.text_index)
3903
# On unlistable transports this could well be empty/error...
3904
text_keys = self.repository.texts.keys()
3905
unused_keys = frozenset(text_keys) - set(self.text_index)
3906
for num, key in enumerate(self.text_index.iterkeys()):
3907
if progress_bar is not None:
3908
progress_bar.update('checking text graph', num, n_versions)
3909
correct_parents = self.calculate_file_version_parents(key)
3911
knit_parents = parent_map[key]
3912
except errors.RevisionNotPresent:
3915
if correct_parents != knit_parents:
3916
wrong_parents[key] = (knit_parents, correct_parents)
3917
return wrong_parents, unused_keys
3920
def _old_get_graph(repository, revision_id):
3921
"""DO NOT USE. That is all. I'm serious."""
3922
graph = repository.get_graph()
3923
revision_graph = dict(((key, value) for key, value in
3924
graph.iter_ancestry([revision_id]) if value is not None))
3925
return _strip_NULL_ghosts(revision_graph)
3928
def _strip_NULL_ghosts(revision_graph):
3929
"""Also don't use this. more compatibility code for unmigrated clients."""
3930
# Filter ghosts, and null:
3931
if _mod_revision.NULL_REVISION in revision_graph:
3932
del revision_graph[_mod_revision.NULL_REVISION]
3933
for key, parents in revision_graph.items():
3934
revision_graph[key] = tuple(parent for parent in parents if parent
3936
return revision_graph
3939
class StreamSink(object):
3940
"""An object that can insert a stream into a repository.
3942
This interface handles the complexity of reserialising inventories and
3943
revisions from different formats, and allows unidirectional insertion into
3944
stacked repositories without looking for the missing basis parents
3948
def __init__(self, target_repo):
3949
self.target_repo = target_repo
3951
def insert_stream(self, stream, src_format, resume_tokens):
3952
"""Insert a stream's content into the target repository.
3954
:param src_format: a bzr repository format.
3956
:return: a list of resume tokens and an iterable of keys additional
3957
items required before the insertion can be completed.
3959
self.target_repo.lock_write()
3962
self.target_repo.resume_write_group(resume_tokens)
3965
self.target_repo.start_write_group()
3968
# locked_insert_stream performs a commit|suspend.
3969
return self._locked_insert_stream(stream, src_format, is_resume)
3971
self.target_repo.abort_write_group(suppress_errors=True)
3974
self.target_repo.unlock()
3976
def _locked_insert_stream(self, stream, src_format, is_resume):
3977
to_serializer = self.target_repo._format._serializer
3978
src_serializer = src_format._serializer
3980
if to_serializer == src_serializer:
3981
# If serializers match and the target is a pack repository, set the
3982
# write cache size on the new pack. This avoids poor performance
3983
# on transports where append is unbuffered (such as
3984
# RemoteTransport). This is safe to do because nothing should read
3985
# back from the target repository while a stream with matching
3986
# serialization is being inserted.
3987
# The exception is that a delta record from the source that should
3988
# be a fulltext may need to be expanded by the target (see
3989
# test_fetch_revisions_with_deltas_into_pack); but we take care to
3990
# explicitly flush any buffered writes first in that rare case.
3992
new_pack = self.target_repo._pack_collection._new_pack
3993
except AttributeError:
3994
# Not a pack repository
3997
new_pack.set_write_cache_size(1024*1024)
3998
for substream_type, substream in stream:
3999
if 'stream' in debug.debug_flags:
4000
mutter('inserting substream: %s', substream_type)
4001
if substream_type == 'texts':
4002
self.target_repo.texts.insert_record_stream(substream)
4003
elif substream_type == 'inventories':
4004
if src_serializer == to_serializer:
4005
self.target_repo.inventories.insert_record_stream(
4008
self._extract_and_insert_inventories(
4009
substream, src_serializer)
4010
elif substream_type == 'inventory-deltas':
4011
self._extract_and_insert_inventory_deltas(
4012
substream, src_serializer)
4013
elif substream_type == 'chk_bytes':
4014
# XXX: This doesn't support conversions, as it assumes the
4015
# conversion was done in the fetch code.
4016
self.target_repo.chk_bytes.insert_record_stream(substream)
4017
elif substream_type == 'revisions':
4018
# This may fallback to extract-and-insert more often than
4019
# required if the serializers are different only in terms of
4021
if src_serializer == to_serializer:
4022
self.target_repo.revisions.insert_record_stream(
4025
self._extract_and_insert_revisions(substream,
4027
elif substream_type == 'signatures':
4028
self.target_repo.signatures.insert_record_stream(substream)
4030
raise AssertionError('kaboom! %s' % (substream_type,))
4031
# Done inserting data, and the missing_keys calculations will try to
4032
# read back from the inserted data, so flush the writes to the new pack
4033
# (if this is pack format).
4034
if new_pack is not None:
4035
new_pack._write_data('', flush=True)
4036
# Find all the new revisions (including ones from resume_tokens)
4037
missing_keys = self.target_repo.get_missing_parent_inventories(
4038
check_for_missing_texts=is_resume)
4040
for prefix, versioned_file in (
4041
('texts', self.target_repo.texts),
4042
('inventories', self.target_repo.inventories),
4043
('revisions', self.target_repo.revisions),
4044
('signatures', self.target_repo.signatures),
4045
('chk_bytes', self.target_repo.chk_bytes),
4047
if versioned_file is None:
4049
missing_keys.update((prefix,) + key for key in
4050
versioned_file.get_missing_compression_parent_keys())
4051
except NotImplementedError:
4052
# cannot even attempt suspending, and missing would have failed
4053
# during stream insertion.
4054
missing_keys = set()
4057
# suspend the write group and tell the caller what we is
4058
# missing. We know we can suspend or else we would not have
4059
# entered this code path. (All repositories that can handle
4060
# missing keys can handle suspending a write group).
4061
write_group_tokens = self.target_repo.suspend_write_group()
4062
return write_group_tokens, missing_keys
4063
hint = self.target_repo.commit_write_group()
4064
if (to_serializer != src_serializer and
4065
self.target_repo._format.pack_compresses):
4066
self.target_repo.pack(hint=hint)
4069
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4070
target_rich_root = self.target_repo._format.rich_root_data
4071
target_tree_refs = self.target_repo._format.supports_tree_reference
4072
for record in substream:
4073
# Insert the delta directly
4074
inventory_delta_bytes = record.get_bytes_as('fulltext')
4075
deserialiser = inventory_delta.InventoryDeltaSerializer()
4076
parse_result = deserialiser.parse_text_bytes(inventory_delta_bytes)
4077
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4078
# Make sure the delta is compatible with the target
4079
if rich_root and not target_rich_root:
4080
raise errors.IncompatibleRevision(self.target_repo._format)
4081
if tree_refs and not target_tree_refs:
4082
# The source supports tree refs and the target doesn't. Check
4083
# the delta for tree refs; if it has any we can't insert it.
4084
for delta_item in inv_delta:
4085
entry = delta_item[3]
4086
if entry.kind == 'tree-reference':
4087
raise errors.IncompatibleRevision(
4088
self.target_repo._format)
4089
revision_id = new_id
4090
parents = [key[0] for key in record.parents]
4091
self.target_repo.add_inventory_by_delta(
4092
basis_id, inv_delta, revision_id, parents)
4094
def _extract_and_insert_inventories(self, substream, serializer,
4096
"""Generate a new inventory versionedfile in target, converting data.
4098
The inventory is retrieved from the source, (deserializing it), and
4099
stored in the target (reserializing it in a different format).
4101
target_rich_root = self.target_repo._format.rich_root_data
4102
target_tree_refs = self.target_repo._format.supports_tree_reference
4103
for record in substream:
4104
# It's not a delta, so it must be a fulltext in the source
4105
# serializer's format.
4106
bytes = record.get_bytes_as('fulltext')
4107
revision_id = record.key[0]
4108
inv = serializer.read_inventory_from_string(bytes, revision_id)
4109
parents = [key[0] for key in record.parents]
4110
self.target_repo.add_inventory(revision_id, inv, parents)
4111
# No need to keep holding this full inv in memory when the rest of
4112
# the substream is likely to be all deltas.
4115
def _extract_and_insert_revisions(self, substream, serializer):
4116
for record in substream:
4117
bytes = record.get_bytes_as('fulltext')
4118
revision_id = record.key[0]
4119
rev = serializer.read_revision_from_string(bytes)
4120
if rev.revision_id != revision_id:
4121
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4122
self.target_repo.add_revision(revision_id, rev)
4125
if self.target_repo._format._fetch_reconcile:
4126
self.target_repo.reconcile()
4129
class StreamSource(object):
4130
"""A source of a stream for fetching between repositories."""
4132
def __init__(self, from_repository, to_format):
4133
"""Create a StreamSource streaming from from_repository."""
4134
self.from_repository = from_repository
4135
self.to_format = to_format
4137
def delta_on_metadata(self):
4138
"""Return True if delta's are permitted on metadata streams.
4140
That is on revisions and signatures.
4142
src_serializer = self.from_repository._format._serializer
4143
target_serializer = self.to_format._serializer
4144
return (self.to_format._fetch_uses_deltas and
4145
src_serializer == target_serializer)
4147
def _fetch_revision_texts(self, revs):
4148
# fetch signatures first and then the revision texts
4149
# may need to be a InterRevisionStore call here.
4150
from_sf = self.from_repository.signatures
4151
# A missing signature is just skipped.
4152
keys = [(rev_id,) for rev_id in revs]
4153
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4155
self.to_format._fetch_order,
4156
not self.to_format._fetch_uses_deltas))
4157
# If a revision has a delta, this is actually expanded inside the
4158
# insert_record_stream code now, which is an alternate fix for
4160
from_rf = self.from_repository.revisions
4161
revisions = from_rf.get_record_stream(
4163
self.to_format._fetch_order,
4164
not self.delta_on_metadata())
4165
return [('signatures', signatures), ('revisions', revisions)]
4167
def _generate_root_texts(self, revs):
4168
"""This will be called by get_stream between fetching weave texts and
4169
fetching the inventory weave.
4171
if self._rich_root_upgrade():
4173
return bzrlib.fetch.Inter1and2Helper(
4174
self.from_repository).generate_root_texts(revs)
4178
def get_stream(self, search):
4180
revs = search.get_keys()
4181
graph = self.from_repository.get_graph()
4182
revs = list(graph.iter_topo_order(revs))
4183
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4185
for knit_kind, file_id, revisions in data_to_fetch:
4186
if knit_kind != phase:
4188
# Make a new progress bar for this phase
4189
if knit_kind == "file":
4190
# Accumulate file texts
4191
text_keys.extend([(file_id, revision) for revision in
4193
elif knit_kind == "inventory":
4194
# Now copy the file texts.
4195
from_texts = self.from_repository.texts
4196
yield ('texts', from_texts.get_record_stream(
4197
text_keys, self.to_format._fetch_order,
4198
not self.to_format._fetch_uses_deltas))
4199
# Cause an error if a text occurs after we have done the
4202
# Before we process the inventory we generate the root
4203
# texts (if necessary) so that the inventories references
4205
for _ in self._generate_root_texts(revs):
4207
# we fetch only the referenced inventories because we do not
4208
# know for unselected inventories whether all their required
4209
# texts are present in the other repository - it could be
4211
for info in self._get_inventory_stream(revs):
4213
elif knit_kind == "signatures":
4214
# Nothing to do here; this will be taken care of when
4215
# _fetch_revision_texts happens.
4217
elif knit_kind == "revisions":
4218
for record in self._fetch_revision_texts(revs):
4221
raise AssertionError("Unknown knit kind %r" % knit_kind)
4223
def get_stream_for_missing_keys(self, missing_keys):
4224
# missing keys can only occur when we are byte copying and not
4225
# translating (because translation means we don't send
4226
# unreconstructable deltas ever).
4228
keys['texts'] = set()
4229
keys['revisions'] = set()
4230
keys['inventories'] = set()
4231
keys['chk_bytes'] = set()
4232
keys['signatures'] = set()
4233
for key in missing_keys:
4234
keys[key[0]].add(key[1:])
4235
if len(keys['revisions']):
4236
# If we allowed copying revisions at this point, we could end up
4237
# copying a revision without copying its required texts: a
4238
# violation of the requirements for repository integrity.
4239
raise AssertionError(
4240
'cannot copy revisions to fill in missing deltas %s' % (
4241
keys['revisions'],))
4242
for substream_kind, keys in keys.iteritems():
4243
vf = getattr(self.from_repository, substream_kind)
4244
if vf is None and keys:
4245
raise AssertionError(
4246
"cannot fill in keys for a versioned file we don't"
4247
" have: %s needs %s" % (substream_kind, keys))
4249
# No need to stream something we don't have
4251
if substream_kind == 'inventories':
4252
# Some missing keys are genuinely ghosts, filter those out.
4253
present = self.from_repository.inventories.get_parent_map(keys)
4254
revs = [key[0] for key in present]
4255
# As with the original stream, we may need to generate root
4256
# texts for the inventories we're about to stream.
4257
for _ in self._generate_root_texts(revs):
4259
# Get the inventory stream more-or-less as we do for the
4260
# original stream; there's no reason to assume that records
4261
# direct from the source will be suitable for the sink. (Think
4262
# e.g. 2a -> 1.9-rich-root).
4263
for info in self._get_inventory_stream(revs, missing=True):
4267
# Ask for full texts always so that we don't need more round trips
4268
# after this stream.
4269
# Some of the missing keys are genuinely ghosts, so filter absent
4270
# records. The Sink is responsible for doing another check to
4271
# ensure that ghosts don't introduce missing data for future
4273
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4274
self.to_format._fetch_order, True))
4275
yield substream_kind, stream
4277
def inventory_fetch_order(self):
4278
if self._rich_root_upgrade():
4279
return 'topological'
4281
return self.to_format._fetch_order
4283
def _rich_root_upgrade(self):
4284
return (not self.from_repository._format.rich_root_data and
4285
self.to_format.rich_root_data)
4287
def _get_inventory_stream(self, revision_ids, missing=False):
4288
from_format = self.from_repository._format
4289
if (from_format.supports_chks and self.to_format.supports_chks and
4290
from_format.network_name() == self.to_format.network_name()):
4291
raise AssertionError(
4292
"this case should be handled by GroupCHKStreamSource")
4293
elif 'forceinvdeltas' in debug.debug_flags:
4294
return self._get_convertable_inventory_stream(revision_ids,
4295
delta_versus_null=missing)
4296
elif from_format.network_name() == self.to_format.network_name():
4298
return self._get_simple_inventory_stream(revision_ids,
4300
elif (not from_format.supports_chks and not self.to_format.supports_chks
4301
and from_format._serializer == self.to_format._serializer):
4302
# Essentially the same format.
4303
return self._get_simple_inventory_stream(revision_ids,
4306
# Any time we switch serializations, we want to use an
4307
# inventory-delta based approach.
4308
return self._get_convertable_inventory_stream(revision_ids,
4309
delta_versus_null=missing)
4311
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4312
# NB: This currently reopens the inventory weave in source;
4313
# using a single stream interface instead would avoid this.
4314
from_weave = self.from_repository.inventories
4316
delta_closure = True
4318
delta_closure = not self.delta_on_metadata()
4319
yield ('inventories', from_weave.get_record_stream(
4320
[(rev_id,) for rev_id in revision_ids],
4321
self.inventory_fetch_order(), delta_closure))
4323
def _get_convertable_inventory_stream(self, revision_ids,
4324
delta_versus_null=False):
4325
# The source is using CHKs, but the target either doesn't or is has a
4326
# different serializer. The StreamSink code expects to be able to
4327
# convert on the target, so we need to put bytes-on-the-wire that can
4328
# be converted. That means inventory deltas (if the remote is <1.18,
4329
# RemoteStreamSink will fallback to VFS to insert the deltas).
4330
yield ('inventory-deltas',
4331
self._stream_invs_as_deltas(revision_ids,
4332
delta_versus_null=delta_versus_null))
4334
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4335
"""Return a stream of inventory-deltas for the given rev ids.
4337
:param revision_ids: The list of inventories to transmit
4338
:param delta_versus_null: Don't try to find a minimal delta for this
4339
entry, instead compute the delta versus the NULL_REVISION. This
4340
effectively streams a complete inventory. Used for stuff like
4341
filling in missing parents, etc.
4343
from_repo = self.from_repository
4344
revision_keys = [(rev_id,) for rev_id in revision_ids]
4345
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4346
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4348
inventories = self.from_repository.iter_inventories(
4349
revision_ids, 'topological')
4350
# XXX: ideally these flags would be per-revision, not per-repo (e.g.
4351
# streaming a non-rich-root revision out of a rich-root repo back into
4352
# a non-rich-root repo ought to be allowed)
4353
format = from_repo._format
4354
flags = (format.rich_root_data, format.supports_tree_reference)
4355
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4356
inventory_cache = lru_cache.LRUCache(50)
4357
null_inventory = from_repo.revision_tree(
4358
_mod_revision.NULL_REVISION).inventory
4359
serializer = inventory_delta.InventoryDeltaSerializer()
4360
serializer.require_flags(*flags)
4361
for inv in inventories:
4362
key = (inv.revision_id,)
4363
parent_keys = parent_map.get(key, ())
4365
if not delta_versus_null and parent_keys:
4366
# The caller did not ask for complete inventories and we have
4367
# some parents that we can delta against. Make a delta against
4368
# each parent so that we can find the smallest.
4369
parent_ids = [parent_key[0] for parent_key in parent_keys]
4370
for parent_id in parent_ids:
4371
if parent_id not in invs_sent_so_far:
4372
# We don't know that the remote side has this basis, so
4375
if parent_id == _mod_revision.NULL_REVISION:
4376
parent_inv = null_inventory
4378
parent_inv = inventory_cache.get(parent_id, None)
4379
if parent_inv is None:
4380
parent_inv = from_repo.get_inventory(parent_id)
4381
candidate_delta = inv._make_delta(parent_inv)
4382
if (delta is None or
4383
len(delta) > len(candidate_delta)):
4384
delta = candidate_delta
4385
basis_id = parent_id
4387
# Either none of the parents ended up being suitable, or we
4388
# were asked to delta against NULL
4389
basis_id = _mod_revision.NULL_REVISION
4390
delta = inv._make_delta(null_inventory)
4391
invs_sent_so_far.add(inv.revision_id)
4392
inventory_cache[inv.revision_id] = inv
4393
delta_serialized = ''.join(
4394
serializer.delta_to_lines(basis_id, key[-1], delta))
4395
yield versionedfile.FulltextContentFactory(
4396
key, parent_keys, None, delta_serialized)
4399
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4400
stop_revision=None):
4401
"""Extend the partial history to include a given index
4403
If a stop_index is supplied, stop when that index has been reached.
4404
If a stop_revision is supplied, stop when that revision is
4405
encountered. Otherwise, stop when the beginning of history is
4408
:param stop_index: The index which should be present. When it is
4409
present, history extension will stop.
4410
:param stop_revision: The revision id which should be present. When
4411
it is encountered, history extension will stop.
4413
start_revision = partial_history_cache[-1]
4414
iterator = repo.iter_reverse_revision_history(start_revision)
4416
#skip the last revision in the list
4419
if (stop_index is not None and
4420
len(partial_history_cache) > stop_index):
4422
if partial_history_cache[-1] == stop_revision:
4424
revision_id = iterator.next()
4425
partial_history_cache.append(revision_id)
4426
except StopIteration:
added
removed
bzrlib/repository.py
