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(), """
39
revision as _mod_revision,
45
from bzrlib.bundle import serializer
46
from bzrlib.revisiontree import RevisionTree
47
from bzrlib.store.versioned import VersionedFileStore
48
from bzrlib.testament import Testament
51
from bzrlib.decorators import needs_read_lock, needs_write_lock
52
from bzrlib.inter import InterObject
53
from bzrlib.inventory import (
59
from bzrlib import registry
60
from bzrlib.symbol_versioning import (
63
from bzrlib.trace import (
64
log_exception_quietly, note, mutter, mutter_callsite, warning)
67
# Old formats display a warning, but only once
68
_deprecation_warning_done = False
71
class CommitBuilder(object):
72
"""Provides an interface to build up a commit.
74
This allows describing a tree to be committed without needing to
75
know the internals of the format of the repository.
78
# all clients should supply tree roots.
79
record_root_entry = True
80
# the default CommitBuilder does not manage trees whose root is versioned.
81
_versioned_root = False
83
def __init__(self, repository, parents, config, timestamp=None,
84
timezone=None, committer=None, revprops=None,
86
"""Initiate a CommitBuilder.
88
:param repository: Repository to commit to.
89
:param parents: Revision ids of the parents of the new revision.
90
:param config: Configuration to use.
91
:param timestamp: Optional timestamp recorded for commit.
92
:param timezone: Optional timezone for timestamp.
93
:param committer: Optional committer to set for commit.
94
:param revprops: Optional dictionary of revision properties.
95
:param revision_id: Optional revision id.
100
self._committer = self._config.username()
102
self._committer = committer
104
self.new_inventory = Inventory(None)
105
self._new_revision_id = revision_id
106
self.parents = parents
107
self.repository = repository
110
if revprops is not None:
111
self._validate_revprops(revprops)
112
self._revprops.update(revprops)
114
if timestamp is None:
115
timestamp = time.time()
116
# Restrict resolution to 1ms
117
self._timestamp = round(timestamp, 3)
120
self._timezone = osutils.local_time_offset()
122
self._timezone = int(timezone)
124
self._generate_revision_if_needed()
125
self.__heads = graph.HeadsCache(repository.get_graph()).heads
126
self._basis_delta = []
127
# API compatibility, older code that used CommitBuilder did not call
128
# .record_delete(), which means the delta that is computed would not be
129
# valid. Callers that will call record_delete() should call
130
# .will_record_deletes() to indicate that.
131
self._recording_deletes = False
132
# memo'd check for no-op commits.
133
self._any_changes = False
135
def any_changes(self):
136
"""Return True if any entries were changed.
138
This includes merge-only changes. It is the core for the --unchanged
141
:return: True if any changes have occured.
143
return self._any_changes
145
def _validate_unicode_text(self, text, context):
146
"""Verify things like commit messages don't have bogus characters."""
148
raise ValueError('Invalid value for %s: %r' % (context, text))
150
def _validate_revprops(self, revprops):
151
for key, value in revprops.iteritems():
152
# We know that the XML serializers do not round trip '\r'
153
# correctly, so refuse to accept them
154
if not isinstance(value, basestring):
155
raise ValueError('revision property (%s) is not a valid'
156
' (unicode) string: %r' % (key, value))
157
self._validate_unicode_text(value,
158
'revision property (%s)' % (key,))
160
def commit(self, message):
161
"""Make the actual commit.
163
:return: The revision id of the recorded revision.
165
self._validate_unicode_text(message, 'commit message')
166
rev = _mod_revision.Revision(
167
timestamp=self._timestamp,
168
timezone=self._timezone,
169
committer=self._committer,
171
inventory_sha1=self.inv_sha1,
172
revision_id=self._new_revision_id,
173
properties=self._revprops)
174
rev.parent_ids = self.parents
175
self.repository.add_revision(self._new_revision_id, rev,
176
self.new_inventory, self._config)
177
self.repository.commit_write_group()
178
return self._new_revision_id
181
"""Abort the commit that is being built.
183
self.repository.abort_write_group()
185
def revision_tree(self):
186
"""Return the tree that was just committed.
188
After calling commit() this can be called to get a RevisionTree
189
representing the newly committed tree. This is preferred to
190
calling Repository.revision_tree() because that may require
191
deserializing the inventory, while we already have a copy in
194
if self.new_inventory is None:
195
self.new_inventory = self.repository.get_inventory(
196
self._new_revision_id)
197
return RevisionTree(self.repository, self.new_inventory,
198
self._new_revision_id)
200
def finish_inventory(self):
201
"""Tell the builder that the inventory is finished.
203
:return: The inventory id in the repository, which can be used with
204
repository.get_inventory.
206
if self.new_inventory is None:
207
# an inventory delta was accumulated without creating a new
209
basis_id = self.basis_delta_revision
210
self.inv_sha1 = self.repository.add_inventory_by_delta(
211
basis_id, self._basis_delta, self._new_revision_id,
214
if self.new_inventory.root is None:
215
raise AssertionError('Root entry should be supplied to'
216
' record_entry_contents, as of bzr 0.10.')
217
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
218
self.new_inventory.revision_id = self._new_revision_id
219
self.inv_sha1 = self.repository.add_inventory(
220
self._new_revision_id,
224
return self._new_revision_id
226
def _gen_revision_id(self):
227
"""Return new revision-id."""
228
return generate_ids.gen_revision_id(self._config.username(),
231
def _generate_revision_if_needed(self):
232
"""Create a revision id if None was supplied.
234
If the repository can not support user-specified revision ids
235
they should override this function and raise CannotSetRevisionId
236
if _new_revision_id is not None.
238
:raises: CannotSetRevisionId
240
if self._new_revision_id is None:
241
self._new_revision_id = self._gen_revision_id()
242
self.random_revid = True
244
self.random_revid = False
246
def _heads(self, file_id, revision_ids):
247
"""Calculate the graph heads for revision_ids in the graph of file_id.
249
This can use either a per-file graph or a global revision graph as we
250
have an identity relationship between the two graphs.
252
return self.__heads(revision_ids)
254
def _check_root(self, ie, parent_invs, tree):
255
"""Helper for record_entry_contents.
257
:param ie: An entry being added.
258
:param parent_invs: The inventories of the parent revisions of the
260
:param tree: The tree that is being committed.
262
# In this revision format, root entries have no knit or weave When
263
# serializing out to disk and back in root.revision is always
265
ie.revision = self._new_revision_id
267
def _require_root_change(self, tree):
268
"""Enforce an appropriate root object change.
270
This is called once when record_iter_changes is called, if and only if
271
the root was not in the delta calculated by record_iter_changes.
273
:param tree: The tree which is being committed.
275
# NB: if there are no parents then this method is not called, so no
276
# need to guard on parents having length.
277
entry = entry_factory['directory'](tree.path2id(''), '',
279
entry.revision = self._new_revision_id
280
self._basis_delta.append(('', '', entry.file_id, entry))
282
def _get_delta(self, ie, basis_inv, path):
283
"""Get a delta against the basis inventory for ie."""
284
if ie.file_id not in basis_inv:
286
result = (None, path, ie.file_id, ie)
287
self._basis_delta.append(result)
289
elif ie != basis_inv[ie.file_id]:
291
# TODO: avoid tis id2path call.
292
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
293
self._basis_delta.append(result)
299
def get_basis_delta(self):
300
"""Return the complete inventory delta versus the basis inventory.
302
This has been built up with the calls to record_delete and
303
record_entry_contents. The client must have already called
304
will_record_deletes() to indicate that they will be generating a
307
:return: An inventory delta, suitable for use with apply_delta, or
308
Repository.add_inventory_by_delta, etc.
310
if not self._recording_deletes:
311
raise AssertionError("recording deletes not activated.")
312
return self._basis_delta
314
def record_delete(self, path, file_id):
315
"""Record that a delete occured against a basis tree.
317
This is an optional API - when used it adds items to the basis_delta
318
being accumulated by the commit builder. It cannot be called unless the
319
method will_record_deletes() has been called to inform the builder that
320
a delta is being supplied.
322
:param path: The path of the thing deleted.
323
:param file_id: The file id that was deleted.
325
if not self._recording_deletes:
326
raise AssertionError("recording deletes not activated.")
327
delta = (path, None, file_id, None)
328
self._basis_delta.append(delta)
329
self._any_changes = True
332
def will_record_deletes(self):
333
"""Tell the commit builder that deletes are being notified.
335
This enables the accumulation of an inventory delta; for the resulting
336
commit to be valid, deletes against the basis MUST be recorded via
337
builder.record_delete().
339
self._recording_deletes = True
341
basis_id = self.parents[0]
343
basis_id = _mod_revision.NULL_REVISION
344
self.basis_delta_revision = basis_id
346
def record_entry_contents(self, ie, parent_invs, path, tree,
348
"""Record the content of ie from tree into the commit if needed.
350
Side effect: sets ie.revision when unchanged
352
:param ie: An inventory entry present in the commit.
353
:param parent_invs: The inventories of the parent revisions of the
355
:param path: The path the entry is at in the tree.
356
:param tree: The tree which contains this entry and should be used to
358
:param content_summary: Summary data from the tree about the paths
359
content - stat, length, exec, sha/link target. This is only
360
accessed when the entry has a revision of None - that is when it is
361
a candidate to commit.
362
:return: A tuple (change_delta, version_recorded, fs_hash).
363
change_delta is an inventory_delta change for this entry against
364
the basis tree of the commit, or None if no change occured against
366
version_recorded is True if a new version of the entry has been
367
recorded. For instance, committing a merge where a file was only
368
changed on the other side will return (delta, False).
369
fs_hash is either None, or the hash details for the path (currently
370
a tuple of the contents sha1 and the statvalue returned by
371
tree.get_file_with_stat()).
373
if self.new_inventory.root is None:
374
if ie.parent_id is not None:
375
raise errors.RootMissing()
376
self._check_root(ie, parent_invs, tree)
377
if ie.revision is None:
378
kind = content_summary[0]
380
# ie is carried over from a prior commit
382
# XXX: repository specific check for nested tree support goes here - if
383
# the repo doesn't want nested trees we skip it ?
384
if (kind == 'tree-reference' and
385
not self.repository._format.supports_tree_reference):
386
# mismatch between commit builder logic and repository:
387
# this needs the entry creation pushed down into the builder.
388
raise NotImplementedError('Missing repository subtree support.')
389
self.new_inventory.add(ie)
391
# TODO: slow, take it out of the inner loop.
393
basis_inv = parent_invs[0]
395
basis_inv = Inventory(root_id=None)
397
# ie.revision is always None if the InventoryEntry is considered
398
# for committing. We may record the previous parents revision if the
399
# content is actually unchanged against a sole head.
400
if ie.revision is not None:
401
if not self._versioned_root and path == '':
402
# repositories that do not version the root set the root's
403
# revision to the new commit even when no change occurs (more
404
# specifically, they do not record a revision on the root; and
405
# the rev id is assigned to the root during deserialisation -
406
# this masks when a change may have occurred against the basis.
407
# To match this we always issue a delta, because the revision
408
# of the root will always be changing.
409
if ie.file_id in basis_inv:
410
delta = (basis_inv.id2path(ie.file_id), path,
414
delta = (None, path, ie.file_id, ie)
415
self._basis_delta.append(delta)
416
return delta, False, None
418
# we don't need to commit this, because the caller already
419
# determined that an existing revision of this file is
420
# appropriate. If its not being considered for committing then
421
# it and all its parents to the root must be unaltered so
422
# no-change against the basis.
423
if ie.revision == self._new_revision_id:
424
raise AssertionError("Impossible situation, a skipped "
425
"inventory entry (%r) claims to be modified in this "
426
"commit (%r).", (ie, self._new_revision_id))
427
return None, False, None
428
# XXX: Friction: parent_candidates should return a list not a dict
429
# so that we don't have to walk the inventories again.
430
parent_candiate_entries = ie.parent_candidates(parent_invs)
431
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
433
for inv in parent_invs:
434
if ie.file_id in inv:
435
old_rev = inv[ie.file_id].revision
436
if old_rev in head_set:
437
heads.append(inv[ie.file_id].revision)
438
head_set.remove(inv[ie.file_id].revision)
441
# now we check to see if we need to write a new record to the
443
# We write a new entry unless there is one head to the ancestors, and
444
# the kind-derived content is unchanged.
446
# Cheapest check first: no ancestors, or more the one head in the
447
# ancestors, we write a new node.
451
# There is a single head, look it up for comparison
452
parent_entry = parent_candiate_entries[heads[0]]
453
# if the non-content specific data has changed, we'll be writing a
455
if (parent_entry.parent_id != ie.parent_id or
456
parent_entry.name != ie.name):
458
# now we need to do content specific checks:
460
# if the kind changed the content obviously has
461
if kind != parent_entry.kind:
463
# Stat cache fingerprint feedback for the caller - None as we usually
464
# don't generate one.
467
if content_summary[2] is None:
468
raise ValueError("Files must not have executable = None")
470
if (# if the file length changed we have to store:
471
parent_entry.text_size != content_summary[1] or
472
# if the exec bit has changed we have to store:
473
parent_entry.executable != content_summary[2]):
475
elif parent_entry.text_sha1 == content_summary[3]:
476
# all meta and content is unchanged (using a hash cache
477
# hit to check the sha)
478
ie.revision = parent_entry.revision
479
ie.text_size = parent_entry.text_size
480
ie.text_sha1 = parent_entry.text_sha1
481
ie.executable = parent_entry.executable
482
return self._get_delta(ie, basis_inv, path), False, None
484
# Either there is only a hash change(no hash cache entry,
485
# or same size content change), or there is no change on
487
# Provide the parent's hash to the store layer, so that the
488
# content is unchanged we will not store a new node.
489
nostore_sha = parent_entry.text_sha1
491
# We want to record a new node regardless of the presence or
492
# absence of a content change in the file.
494
ie.executable = content_summary[2]
495
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
497
text = file_obj.read()
501
ie.text_sha1, ie.text_size = self._add_text_to_weave(
502
ie.file_id, text, heads, nostore_sha)
503
# Let the caller know we generated a stat fingerprint.
504
fingerprint = (ie.text_sha1, stat_value)
505
except errors.ExistingContent:
506
# Turns out that the file content was unchanged, and we were
507
# only going to store a new node if it was changed. Carry over
509
ie.revision = parent_entry.revision
510
ie.text_size = parent_entry.text_size
511
ie.text_sha1 = parent_entry.text_sha1
512
ie.executable = parent_entry.executable
513
return self._get_delta(ie, basis_inv, path), False, None
514
elif kind == 'directory':
516
# all data is meta here, nothing specific to directory, so
518
ie.revision = parent_entry.revision
519
return self._get_delta(ie, basis_inv, path), False, None
520
self._add_text_to_weave(ie.file_id, '', heads, None)
521
elif kind == 'symlink':
522
current_link_target = content_summary[3]
524
# symlink target is not generic metadata, check if it has
526
if current_link_target != parent_entry.symlink_target:
529
# unchanged, carry over.
530
ie.revision = parent_entry.revision
531
ie.symlink_target = parent_entry.symlink_target
532
return self._get_delta(ie, basis_inv, path), False, None
533
ie.symlink_target = current_link_target
534
self._add_text_to_weave(ie.file_id, '', heads, None)
535
elif kind == 'tree-reference':
537
if content_summary[3] != parent_entry.reference_revision:
540
# unchanged, carry over.
541
ie.reference_revision = parent_entry.reference_revision
542
ie.revision = parent_entry.revision
543
return self._get_delta(ie, basis_inv, path), False, None
544
ie.reference_revision = content_summary[3]
545
self._add_text_to_weave(ie.file_id, '', heads, None)
547
raise NotImplementedError('unknown kind')
548
ie.revision = self._new_revision_id
549
self._any_changes = True
550
return self._get_delta(ie, basis_inv, path), True, fingerprint
552
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
553
_entry_factory=entry_factory):
554
"""Record a new tree via iter_changes.
556
:param tree: The tree to obtain text contents from for changed objects.
557
:param basis_revision_id: The revision id of the tree the iter_changes
558
has been generated against. Currently assumed to be the same
559
as self.parents[0] - if it is not, errors may occur.
560
:param iter_changes: An iter_changes iterator with the changes to apply
561
to basis_revision_id. The iterator must not include any items with
562
a current kind of None - missing items must be either filtered out
563
or errored-on beefore record_iter_changes sees the item.
564
:param _entry_factory: Private method to bind entry_factory locally for
566
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
569
# Create an inventory delta based on deltas between all the parents and
570
# deltas between all the parent inventories. We use inventory delta's
571
# between the inventory objects because iter_changes masks
572
# last-changed-field only changes.
574
# file_id -> change map, change is fileid, paths, changed, versioneds,
575
# parents, names, kinds, executables
577
# {file_id -> revision_id -> inventory entry, for entries in parent
578
# trees that are not parents[0]
582
revtrees = list(self.repository.revision_trees(self.parents))
583
except errors.NoSuchRevision:
584
# one or more ghosts, slow path.
586
for revision_id in self.parents:
588
revtrees.append(self.repository.revision_tree(revision_id))
589
except errors.NoSuchRevision:
591
basis_revision_id = _mod_revision.NULL_REVISION
593
revtrees.append(self.repository.revision_tree(
594
_mod_revision.NULL_REVISION))
595
# The basis inventory from a repository
597
basis_inv = revtrees[0].inventory
599
basis_inv = self.repository.revision_tree(
600
_mod_revision.NULL_REVISION).inventory
601
if len(self.parents) > 0:
602
if basis_revision_id != self.parents[0] and not ghost_basis:
604
"arbitrary basis parents not yet supported with merges")
605
for revtree in revtrees[1:]:
606
for change in revtree.inventory._make_delta(basis_inv):
607
if change[1] is None:
608
# Not present in this parent.
610
if change[2] not in merged_ids:
611
if change[0] is not None:
612
basis_entry = basis_inv[change[2]]
613
merged_ids[change[2]] = [
615
basis_entry.revision,
618
parent_entries[change[2]] = {
620
basis_entry.revision:basis_entry,
622
change[3].revision:change[3],
625
merged_ids[change[2]] = [change[3].revision]
626
parent_entries[change[2]] = {change[3].revision:change[3]}
628
merged_ids[change[2]].append(change[3].revision)
629
parent_entries[change[2]][change[3].revision] = change[3]
632
# Setup the changes from the tree:
633
# changes maps file_id -> (change, [parent revision_ids])
635
for change in iter_changes:
636
# This probably looks up in basis_inv way to much.
637
if change[1][0] is not None:
638
head_candidate = [basis_inv[change[0]].revision]
641
changes[change[0]] = change, merged_ids.get(change[0],
643
unchanged_merged = set(merged_ids) - set(changes)
644
# Extend the changes dict with synthetic changes to record merges of
646
for file_id in unchanged_merged:
647
# Record a merged version of these items that did not change vs the
648
# basis. This can be either identical parallel changes, or a revert
649
# of a specific file after a merge. The recorded content will be
650
# that of the current tree (which is the same as the basis), but
651
# the per-file graph will reflect a merge.
652
# NB:XXX: We are reconstructing path information we had, this
653
# should be preserved instead.
654
# inv delta change: (file_id, (path_in_source, path_in_target),
655
# changed_content, versioned, parent, name, kind,
658
basis_entry = basis_inv[file_id]
659
except errors.NoSuchId:
660
# a change from basis->some_parents but file_id isn't in basis
661
# so was new in the merge, which means it must have changed
662
# from basis -> current, and as it hasn't the add was reverted
663
# by the user. So we discard this change.
667
(basis_inv.id2path(file_id), tree.id2path(file_id)),
669
(basis_entry.parent_id, basis_entry.parent_id),
670
(basis_entry.name, basis_entry.name),
671
(basis_entry.kind, basis_entry.kind),
672
(basis_entry.executable, basis_entry.executable))
673
changes[file_id] = (change, merged_ids[file_id])
674
# changes contains tuples with the change and a set of inventory
675
# candidates for the file.
677
# old_path, new_path, file_id, new_inventory_entry
678
seen_root = False # Is the root in the basis delta?
679
inv_delta = self._basis_delta
680
modified_rev = self._new_revision_id
681
for change, head_candidates in changes.values():
682
if change[3][1]: # versioned in target.
683
# Several things may be happening here:
684
# We may have a fork in the per-file graph
685
# - record a change with the content from tree
686
# We may have a change against < all trees
687
# - carry over the tree that hasn't changed
688
# We may have a change against all trees
689
# - record the change with the content from tree
692
entry = _entry_factory[kind](file_id, change[5][1],
694
head_set = self._heads(change[0], set(head_candidates))
697
for head_candidate in head_candidates:
698
if head_candidate in head_set:
699
heads.append(head_candidate)
700
head_set.remove(head_candidate)
703
# Could be a carry-over situation:
704
parent_entry_revs = parent_entries.get(file_id, None)
705
if parent_entry_revs:
706
parent_entry = parent_entry_revs.get(heads[0], None)
709
if parent_entry is None:
710
# The parent iter_changes was called against is the one
711
# that is the per-file head, so any change is relevant
712
# iter_changes is valid.
713
carry_over_possible = False
715
# could be a carry over situation
716
# A change against the basis may just indicate a merge,
717
# we need to check the content against the source of the
718
# merge to determine if it was changed after the merge
720
if (parent_entry.kind != entry.kind or
721
parent_entry.parent_id != entry.parent_id or
722
parent_entry.name != entry.name):
723
# Metadata common to all entries has changed
724
# against per-file parent
725
carry_over_possible = False
727
carry_over_possible = True
728
# per-type checks for changes against the parent_entry
731
# Cannot be a carry-over situation
732
carry_over_possible = False
733
# Populate the entry in the delta
735
# XXX: There is still a small race here: If someone reverts the content of a file
736
# after iter_changes examines and decides it has changed,
737
# we will unconditionally record a new version even if some
738
# other process reverts it while commit is running (with
739
# the revert happening after iter_changes did it's
742
entry.executable = True
744
entry.executable = False
745
if (carry_over_possible and
746
parent_entry.executable == entry.executable):
747
# Check the file length, content hash after reading
749
nostore_sha = parent_entry.text_sha1
752
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
754
text = file_obj.read()
758
entry.text_sha1, entry.text_size = self._add_text_to_weave(
759
file_id, text, heads, nostore_sha)
760
yield file_id, change[1][1], (entry.text_sha1, stat_value)
761
except errors.ExistingContent:
762
# No content change against a carry_over parent
763
# Perhaps this should also yield a fs hash update?
765
entry.text_size = parent_entry.text_size
766
entry.text_sha1 = parent_entry.text_sha1
767
elif kind == 'symlink':
769
entry.symlink_target = tree.get_symlink_target(file_id)
770
if (carry_over_possible and
771
parent_entry.symlink_target == entry.symlink_target):
774
self._add_text_to_weave(change[0], '', heads, None)
775
elif kind == 'directory':
776
if carry_over_possible:
779
# Nothing to set on the entry.
780
# XXX: split into the Root and nonRoot versions.
781
if change[1][1] != '' or self.repository.supports_rich_root():
782
self._add_text_to_weave(change[0], '', heads, None)
783
elif kind == 'tree-reference':
784
if not self.repository._format.supports_tree_reference:
785
# This isn't quite sane as an error, but we shouldn't
786
# ever see this code path in practice: tree's don't
787
# permit references when the repo doesn't support tree
789
raise errors.UnsupportedOperation(tree.add_reference,
791
entry.reference_revision = \
792
tree.get_reference_revision(change[0])
793
if (carry_over_possible and
794
parent_entry.reference_revision == reference_revision):
797
self._add_text_to_weave(change[0], '', heads, None)
799
raise AssertionError('unknown kind %r' % kind)
801
entry.revision = modified_rev
803
entry.revision = parent_entry.revision
806
new_path = change[1][1]
807
inv_delta.append((change[1][0], new_path, change[0], entry))
810
self.new_inventory = None
812
self._any_changes = True
814
# housekeeping root entry changes do not affect no-change commits.
815
self._require_root_change(tree)
816
self.basis_delta_revision = basis_revision_id
818
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
819
parent_keys = tuple([(file_id, parent) for parent in parents])
820
return self.repository.texts._add_text(
821
(file_id, self._new_revision_id), parent_keys, new_text,
822
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
825
class RootCommitBuilder(CommitBuilder):
826
"""This commitbuilder actually records the root id"""
828
# the root entry gets versioned properly by this builder.
829
_versioned_root = True
831
def _check_root(self, ie, parent_invs, tree):
832
"""Helper for record_entry_contents.
834
:param ie: An entry being added.
835
:param parent_invs: The inventories of the parent revisions of the
837
:param tree: The tree that is being committed.
840
def _require_root_change(self, tree):
841
"""Enforce an appropriate root object change.
843
This is called once when record_iter_changes is called, if and only if
844
the root was not in the delta calculated by record_iter_changes.
846
:param tree: The tree which is being committed.
848
# versioned roots do not change unless the tree found a change.
851
######################################################################
854
class Repository(object):
855
"""Repository holding history for one or more branches.
857
The repository holds and retrieves historical information including
858
revisions and file history. It's normally accessed only by the Branch,
859
which views a particular line of development through that history.
861
The Repository builds on top of some byte storage facilies (the revisions,
862
signatures, inventories, texts and chk_bytes attributes) and a Transport,
863
which respectively provide byte storage and a means to access the (possibly
866
The byte storage facilities are addressed via tuples, which we refer to
867
as 'keys' throughout the code base. Revision_keys, inventory_keys and
868
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
869
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
870
byte string made up of a hash identifier and a hash value.
871
We use this interface because it allows low friction with the underlying
872
code that implements disk indices, network encoding and other parts of
875
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
876
the serialised revisions for the repository. This can be used to obtain
877
revision graph information or to access raw serialised revisions.
878
The result of trying to insert data into the repository via this store
879
is undefined: it should be considered read-only except for implementors
881
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
882
the serialised signatures for the repository. This can be used to
883
obtain access to raw serialised signatures. The result of trying to
884
insert data into the repository via this store is undefined: it should
885
be considered read-only except for implementors of repositories.
886
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
887
the serialised inventories for the repository. This can be used to
888
obtain unserialised inventories. The result of trying to insert data
889
into the repository via this store is undefined: it should be
890
considered read-only except for implementors of repositories.
891
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
892
texts of files and directories for the repository. This can be used to
893
obtain file texts or file graphs. Note that Repository.iter_file_bytes
894
is usually a better interface for accessing file texts.
895
The result of trying to insert data into the repository via this store
896
is undefined: it should be considered read-only except for implementors
898
:ivar chk_bytes: A bzrlib.versionedfile.VersionedFiles instance containing
899
any data the repository chooses to store or have indexed by its hash.
900
The result of trying to insert data into the repository via this store
901
is undefined: it should be considered read-only except for implementors
903
:ivar _transport: Transport for file access to repository, typically
904
pointing to .bzr/repository.
907
# What class to use for a CommitBuilder. Often its simpler to change this
908
# in a Repository class subclass rather than to override
909
# get_commit_builder.
910
_commit_builder_class = CommitBuilder
911
# The search regex used by xml based repositories to determine what things
912
# where changed in a single commit.
913
_file_ids_altered_regex = lazy_regex.lazy_compile(
914
r'file_id="(?P<file_id>[^"]+)"'
915
r'.* revision="(?P<revision_id>[^"]+)"'
918
def abort_write_group(self, suppress_errors=False):
919
"""Commit the contents accrued within the current write group.
921
:param suppress_errors: if true, abort_write_group will catch and log
922
unexpected errors that happen during the abort, rather than
923
allowing them to propagate. Defaults to False.
925
:seealso: start_write_group.
927
if self._write_group is not self.get_transaction():
928
# has an unlock or relock occured ?
929
raise errors.BzrError(
930
'mismatched lock context and write group. %r, %r' %
931
(self._write_group, self.get_transaction()))
933
self._abort_write_group()
934
except Exception, exc:
935
self._write_group = None
936
if not suppress_errors:
938
mutter('abort_write_group failed')
939
log_exception_quietly()
940
note('bzr: ERROR (ignored): %s', exc)
941
self._write_group = None
943
def _abort_write_group(self):
944
"""Template method for per-repository write group cleanup.
946
This is called during abort before the write group is considered to be
947
finished and should cleanup any internal state accrued during the write
948
group. There is no requirement that data handed to the repository be
949
*not* made available - this is not a rollback - but neither should any
950
attempt be made to ensure that data added is fully commited. Abort is
951
invoked when an error has occured so futher disk or network operations
952
may not be possible or may error and if possible should not be
956
def add_fallback_repository(self, repository):
957
"""Add a repository to use for looking up data not held locally.
959
:param repository: A repository.
961
if not self._format.supports_external_lookups:
962
raise errors.UnstackableRepositoryFormat(self._format, self.base)
964
# This repository will call fallback.unlock() when we transition to
965
# the unlocked state, so we make sure to increment the lock count
966
repository.lock_read()
967
self._check_fallback_repository(repository)
968
self._fallback_repositories.append(repository)
969
self.texts.add_fallback_versioned_files(repository.texts)
970
self.inventories.add_fallback_versioned_files(repository.inventories)
971
self.revisions.add_fallback_versioned_files(repository.revisions)
972
self.signatures.add_fallback_versioned_files(repository.signatures)
973
if self.chk_bytes is not None:
974
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
976
def _check_fallback_repository(self, repository):
977
"""Check that this repository can fallback to repository safely.
979
Raise an error if not.
981
:param repository: A repository to fallback to.
983
return InterRepository._assert_same_model(self, repository)
985
def add_inventory(self, revision_id, inv, parents):
986
"""Add the inventory inv to the repository as revision_id.
988
:param parents: The revision ids of the parents that revision_id
989
is known to have and are in the repository already.
991
:returns: The validator(which is a sha1 digest, though what is sha'd is
992
repository format specific) of the serialized inventory.
994
if not self.is_in_write_group():
995
raise AssertionError("%r not in write group" % (self,))
996
_mod_revision.check_not_reserved_id(revision_id)
997
if not (inv.revision_id is None or inv.revision_id == revision_id):
998
raise AssertionError(
999
"Mismatch between inventory revision"
1000
" id and insertion revid (%r, %r)"
1001
% (inv.revision_id, revision_id))
1002
if inv.root is None:
1003
raise AssertionError()
1004
return self._add_inventory_checked(revision_id, inv, parents)
1006
def _add_inventory_checked(self, revision_id, inv, parents):
1007
"""Add inv to the repository after checking the inputs.
1009
This function can be overridden to allow different inventory styles.
1011
:seealso: add_inventory, for the contract.
1013
inv_lines = self._serialise_inventory_to_lines(inv)
1014
return self._inventory_add_lines(revision_id, parents,
1015
inv_lines, check_content=False)
1017
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1018
parents, basis_inv=None, propagate_caches=False):
1019
"""Add a new inventory expressed as a delta against another revision.
1021
See the inventory developers documentation for the theory behind
1024
:param basis_revision_id: The inventory id the delta was created
1025
against. (This does not have to be a direct parent.)
1026
:param delta: The inventory delta (see Inventory.apply_delta for
1028
:param new_revision_id: The revision id that the inventory is being
1030
:param parents: The revision ids of the parents that revision_id is
1031
known to have and are in the repository already. These are supplied
1032
for repositories that depend on the inventory graph for revision
1033
graph access, as well as for those that pun ancestry with delta
1035
:param basis_inv: The basis inventory if it is already known,
1037
:param propagate_caches: If True, the caches for this inventory are
1038
copied to and updated for the result if possible.
1040
:returns: (validator, new_inv)
1041
The validator(which is a sha1 digest, though what is sha'd is
1042
repository format specific) of the serialized inventory, and the
1043
resulting inventory.
1045
if not self.is_in_write_group():
1046
raise AssertionError("%r not in write group" % (self,))
1047
_mod_revision.check_not_reserved_id(new_revision_id)
1048
basis_tree = self.revision_tree(basis_revision_id)
1049
basis_tree.lock_read()
1051
# Note that this mutates the inventory of basis_tree, which not all
1052
# inventory implementations may support: A better idiom would be to
1053
# return a new inventory, but as there is no revision tree cache in
1054
# repository this is safe for now - RBC 20081013
1055
if basis_inv is None:
1056
basis_inv = basis_tree.inventory
1057
basis_inv.apply_delta(delta)
1058
basis_inv.revision_id = new_revision_id
1059
return (self.add_inventory(new_revision_id, basis_inv, parents),
1064
def _inventory_add_lines(self, revision_id, parents, lines,
1065
check_content=True):
1066
"""Store lines in inv_vf and return the sha1 of the inventory."""
1067
parents = [(parent,) for parent in parents]
1068
return self.inventories.add_lines((revision_id,), parents, lines,
1069
check_content=check_content)[0]
1071
def add_revision(self, revision_id, rev, inv=None, config=None):
1072
"""Add rev to the revision store as revision_id.
1074
:param revision_id: the revision id to use.
1075
:param rev: The revision object.
1076
:param inv: The inventory for the revision. if None, it will be looked
1077
up in the inventory storer
1078
:param config: If None no digital signature will be created.
1079
If supplied its signature_needed method will be used
1080
to determine if a signature should be made.
1082
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
1084
_mod_revision.check_not_reserved_id(revision_id)
1085
if config is not None and config.signature_needed():
1087
inv = self.get_inventory(revision_id)
1088
plaintext = Testament(rev, inv).as_short_text()
1089
self.store_revision_signature(
1090
gpg.GPGStrategy(config), plaintext, revision_id)
1091
# check inventory present
1092
if not self.inventories.get_parent_map([(revision_id,)]):
1094
raise errors.WeaveRevisionNotPresent(revision_id,
1097
# yes, this is not suitable for adding with ghosts.
1098
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
1101
key = (revision_id,)
1102
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
1103
self._add_revision(rev)
1105
def _add_revision(self, revision):
1106
text = self._serializer.write_revision_to_string(revision)
1107
key = (revision.revision_id,)
1108
parents = tuple((parent,) for parent in revision.parent_ids)
1109
self.revisions.add_lines(key, parents, osutils.split_lines(text))
1111
def all_revision_ids(self):
1112
"""Returns a list of all the revision ids in the repository.
1114
This is conceptually deprecated because code should generally work on
1115
the graph reachable from a particular revision, and ignore any other
1116
revisions that might be present. There is no direct replacement
1119
if 'evil' in debug.debug_flags:
1120
mutter_callsite(2, "all_revision_ids is linear with history.")
1121
return self._all_revision_ids()
1123
def _all_revision_ids(self):
1124
"""Returns a list of all the revision ids in the repository.
1126
These are in as much topological order as the underlying store can
1129
raise NotImplementedError(self._all_revision_ids)
1131
def break_lock(self):
1132
"""Break a lock if one is present from another instance.
1134
Uses the ui factory to ask for confirmation if the lock may be from
1137
self.control_files.break_lock()
1140
def _eliminate_revisions_not_present(self, revision_ids):
1141
"""Check every revision id in revision_ids to see if we have it.
1143
Returns a set of the present revisions.
1146
graph = self.get_graph()
1147
parent_map = graph.get_parent_map(revision_ids)
1148
# The old API returned a list, should this actually be a set?
1149
return parent_map.keys()
1152
def create(a_bzrdir):
1153
"""Construct the current default format repository in a_bzrdir."""
1154
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
1156
def __init__(self, _format, a_bzrdir, control_files):
1157
"""instantiate a Repository.
1159
:param _format: The format of the repository on disk.
1160
:param a_bzrdir: The BzrDir of the repository.
1162
In the future we will have a single api for all stores for
1163
getting file texts, inventories and revisions, then
1164
this construct will accept instances of those things.
1166
super(Repository, self).__init__()
1167
self._format = _format
1168
# the following are part of the public API for Repository:
1169
self.bzrdir = a_bzrdir
1170
self.control_files = control_files
1171
self._transport = control_files._transport
1172
self.base = self._transport.base
1174
self._reconcile_does_inventory_gc = True
1175
self._reconcile_fixes_text_parents = False
1176
self._reconcile_backsup_inventory = True
1177
# not right yet - should be more semantically clear ?
1179
# TODO: make sure to construct the right store classes, etc, depending
1180
# on whether escaping is required.
1181
self._warn_if_deprecated()
1182
self._write_group = None
1183
# Additional places to query for data.
1184
self._fallback_repositories = []
1185
# An InventoryEntry cache, used during deserialization
1186
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1189
return '%s(%r)' % (self.__class__.__name__,
1192
def has_same_location(self, other):
1193
"""Returns a boolean indicating if this repository is at the same
1194
location as another repository.
1196
This might return False even when two repository objects are accessing
1197
the same physical repository via different URLs.
1199
if self.__class__ is not other.__class__:
1201
return (self._transport.base == other._transport.base)
1203
def is_in_write_group(self):
1204
"""Return True if there is an open write group.
1206
:seealso: start_write_group.
1208
return self._write_group is not None
1210
def is_locked(self):
1211
return self.control_files.is_locked()
1213
def is_write_locked(self):
1214
"""Return True if this object is write locked."""
1215
return self.is_locked() and self.control_files._lock_mode == 'w'
1217
def lock_write(self, token=None):
1218
"""Lock this repository for writing.
1220
This causes caching within the repository obejct to start accumlating
1221
data during reads, and allows a 'write_group' to be obtained. Write
1222
groups must be used for actual data insertion.
1224
:param token: if this is already locked, then lock_write will fail
1225
unless the token matches the existing lock.
1226
:returns: a token if this instance supports tokens, otherwise None.
1227
:raises TokenLockingNotSupported: when a token is given but this
1228
instance doesn't support using token locks.
1229
:raises MismatchedToken: if the specified token doesn't match the token
1230
of the existing lock.
1231
:seealso: start_write_group.
1233
A token should be passed in if you know that you have locked the object
1234
some other way, and need to synchronise this object's state with that
1237
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1239
locked = self.is_locked()
1240
result = self.control_files.lock_write(token=token)
1242
for repo in self._fallback_repositories:
1243
# Writes don't affect fallback repos
1245
self._refresh_data()
1248
def lock_read(self):
1249
locked = self.is_locked()
1250
self.control_files.lock_read()
1252
for repo in self._fallback_repositories:
1254
self._refresh_data()
1256
def get_physical_lock_status(self):
1257
return self.control_files.get_physical_lock_status()
1259
def leave_lock_in_place(self):
1260
"""Tell this repository not to release the physical lock when this
1263
If lock_write doesn't return a token, then this method is not supported.
1265
self.control_files.leave_in_place()
1267
def dont_leave_lock_in_place(self):
1268
"""Tell this repository to release the physical lock when this
1269
object is unlocked, even if it didn't originally acquire it.
1271
If lock_write doesn't return a token, then this method is not supported.
1273
self.control_files.dont_leave_in_place()
1276
def gather_stats(self, revid=None, committers=None):
1277
"""Gather statistics from a revision id.
1279
:param revid: The revision id to gather statistics from, if None, then
1280
no revision specific statistics are gathered.
1281
:param committers: Optional parameter controlling whether to grab
1282
a count of committers from the revision specific statistics.
1283
:return: A dictionary of statistics. Currently this contains:
1284
committers: The number of committers if requested.
1285
firstrev: A tuple with timestamp, timezone for the penultimate left
1286
most ancestor of revid, if revid is not the NULL_REVISION.
1287
latestrev: A tuple with timestamp, timezone for revid, if revid is
1288
not the NULL_REVISION.
1289
revisions: The total revision count in the repository.
1290
size: An estimate disk size of the repository in bytes.
1293
if revid and committers:
1294
result['committers'] = 0
1295
if revid and revid != _mod_revision.NULL_REVISION:
1297
all_committers = set()
1298
revisions = self.get_ancestry(revid)
1299
# pop the leading None
1301
first_revision = None
1303
# ignore the revisions in the middle - just grab first and last
1304
revisions = revisions[0], revisions[-1]
1305
for revision in self.get_revisions(revisions):
1306
if not first_revision:
1307
first_revision = revision
1309
all_committers.add(revision.committer)
1310
last_revision = revision
1312
result['committers'] = len(all_committers)
1313
result['firstrev'] = (first_revision.timestamp,
1314
first_revision.timezone)
1315
result['latestrev'] = (last_revision.timestamp,
1316
last_revision.timezone)
1318
# now gather global repository information
1319
# XXX: This is available for many repos regardless of listability.
1320
if self.bzrdir.root_transport.listable():
1321
# XXX: do we want to __define len__() ?
1322
# Maybe the versionedfiles object should provide a different
1323
# method to get the number of keys.
1324
result['revisions'] = len(self.revisions.keys())
1325
# result['size'] = t
1328
def find_branches(self, using=False):
1329
"""Find branches underneath this repository.
1331
This will include branches inside other branches.
1333
:param using: If True, list only branches using this repository.
1335
if using and not self.is_shared():
1337
return [self.bzrdir.open_branch()]
1338
except errors.NotBranchError:
1340
class Evaluator(object):
1343
self.first_call = True
1345
def __call__(self, bzrdir):
1346
# On the first call, the parameter is always the bzrdir
1347
# containing the current repo.
1348
if not self.first_call:
1350
repository = bzrdir.open_repository()
1351
except errors.NoRepositoryPresent:
1354
return False, (None, repository)
1355
self.first_call = False
1357
value = (bzrdir.open_branch(), None)
1358
except errors.NotBranchError:
1359
value = (None, None)
1363
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1364
self.bzrdir.root_transport, evaluate=Evaluator()):
1365
if branch is not None:
1366
branches.append(branch)
1367
if not using and repository is not None:
1368
branches.extend(repository.find_branches())
1372
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1373
"""Return the revision ids that other has that this does not.
1375
These are returned in topological order.
1377
revision_id: only return revision ids included by revision_id.
1379
return InterRepository.get(other, self).search_missing_revision_ids(
1380
revision_id, find_ghosts)
1384
"""Open the repository rooted at base.
1386
For instance, if the repository is at URL/.bzr/repository,
1387
Repository.open(URL) -> a Repository instance.
1389
control = bzrdir.BzrDir.open(base)
1390
return control.open_repository()
1392
def copy_content_into(self, destination, revision_id=None):
1393
"""Make a complete copy of the content in self into destination.
1395
This is a destructive operation! Do not use it on existing
1398
return InterRepository.get(self, destination).copy_content(revision_id)
1400
def commit_write_group(self):
1401
"""Commit the contents accrued within the current write group.
1403
:seealso: start_write_group.
1405
if self._write_group is not self.get_transaction():
1406
# has an unlock or relock occured ?
1407
raise errors.BzrError('mismatched lock context %r and '
1409
(self.get_transaction(), self._write_group))
1410
result = self._commit_write_group()
1411
self._write_group = None
1414
def _commit_write_group(self):
1415
"""Template method for per-repository write group cleanup.
1417
This is called before the write group is considered to be
1418
finished and should ensure that all data handed to the repository
1419
for writing during the write group is safely committed (to the
1420
extent possible considering file system caching etc).
1423
def suspend_write_group(self):
1424
raise errors.UnsuspendableWriteGroup(self)
1426
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1427
"""Return the keys of missing inventory parents for revisions added in
1430
A revision is not complete if the inventory delta for that revision
1431
cannot be calculated. Therefore if the parent inventories of a
1432
revision are not present, the revision is incomplete, and e.g. cannot
1433
be streamed by a smart server. This method finds missing inventory
1434
parents for revisions added in this write group.
1436
if not self._format.supports_external_lookups:
1437
# This is only an issue for stacked repositories
1439
if not self.is_in_write_group():
1440
raise AssertionError('not in a write group')
1442
# XXX: We assume that every added revision already has its
1443
# corresponding inventory, so we only check for parent inventories that
1444
# might be missing, rather than all inventories.
1445
parents = set(self.revisions._index.get_missing_parents())
1446
parents.discard(_mod_revision.NULL_REVISION)
1447
unstacked_inventories = self.inventories._index
1448
present_inventories = unstacked_inventories.get_parent_map(
1449
key[-1:] for key in parents)
1450
parents.difference_update(present_inventories)
1451
if len(parents) == 0:
1452
# No missing parent inventories.
1454
if not check_for_missing_texts:
1455
return set(('inventories', rev_id) for (rev_id,) in parents)
1456
# Ok, now we have a list of missing inventories. But these only matter
1457
# if the inventories that reference them are missing some texts they
1458
# appear to introduce.
1459
# XXX: Texts referenced by all added inventories need to be present,
1460
# but at the moment we're only checking for texts referenced by
1461
# inventories at the graph's edge.
1462
key_deps = self.revisions._index._key_dependencies
1463
key_deps.add_keys(present_inventories)
1464
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1465
file_ids = self.fileids_altered_by_revision_ids(referrers)
1466
missing_texts = set()
1467
for file_id, version_ids in file_ids.iteritems():
1468
missing_texts.update(
1469
(file_id, version_id) for version_id in version_ids)
1470
present_texts = self.texts.get_parent_map(missing_texts)
1471
missing_texts.difference_update(present_texts)
1472
if not missing_texts:
1473
# No texts are missing, so all revisions and their deltas are
1476
# Alternatively the text versions could be returned as the missing
1477
# keys, but this is likely to be less data.
1478
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1481
def refresh_data(self):
1482
"""Re-read any data needed to to synchronise with disk.
1484
This method is intended to be called after another repository instance
1485
(such as one used by a smart server) has inserted data into the
1486
repository. It may not be called during a write group, but may be
1487
called at any other time.
1489
if self.is_in_write_group():
1490
raise errors.InternalBzrError(
1491
"May not refresh_data while in a write group.")
1492
self._refresh_data()
1494
def resume_write_group(self, tokens):
1495
if not self.is_write_locked():
1496
raise errors.NotWriteLocked(self)
1497
if self._write_group:
1498
raise errors.BzrError('already in a write group')
1499
self._resume_write_group(tokens)
1500
# so we can detect unlock/relock - the write group is now entered.
1501
self._write_group = self.get_transaction()
1503
def _resume_write_group(self, tokens):
1504
raise errors.UnsuspendableWriteGroup(self)
1506
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
1508
"""Fetch the content required to construct revision_id from source.
1510
If revision_id is None and fetch_spec is None, then all content is
1513
fetch() may not be used when the repository is in a write group -
1514
either finish the current write group before using fetch, or use
1515
fetch before starting the write group.
1517
:param find_ghosts: Find and copy revisions in the source that are
1518
ghosts in the target (and not reachable directly by walking out to
1519
the first-present revision in target from revision_id).
1520
:param revision_id: If specified, all the content needed for this
1521
revision ID will be copied to the target. Fetch will determine for
1522
itself which content needs to be copied.
1523
:param fetch_spec: If specified, a SearchResult or
1524
PendingAncestryResult that describes which revisions to copy. This
1525
allows copying multiple heads at once. Mutually exclusive with
1528
if fetch_spec is not None and revision_id is not None:
1529
raise AssertionError(
1530
"fetch_spec and revision_id are mutually exclusive.")
1531
if self.is_in_write_group():
1532
raise errors.InternalBzrError(
1533
"May not fetch while in a write group.")
1534
# fast path same-url fetch operations
1535
if self.has_same_location(source) and fetch_spec is None:
1536
# check that last_revision is in 'from' and then return a
1538
if (revision_id is not None and
1539
not _mod_revision.is_null(revision_id)):
1540
self.get_revision(revision_id)
1542
# if there is no specific appropriate InterRepository, this will get
1543
# the InterRepository base class, which raises an
1544
# IncompatibleRepositories when asked to fetch.
1545
inter = InterRepository.get(source, self)
1546
return inter.fetch(revision_id=revision_id, pb=pb,
1547
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
1549
def create_bundle(self, target, base, fileobj, format=None):
1550
return serializer.write_bundle(self, target, base, fileobj, format)
1552
def get_commit_builder(self, branch, parents, config, timestamp=None,
1553
timezone=None, committer=None, revprops=None,
1555
"""Obtain a CommitBuilder for this repository.
1557
:param branch: Branch to commit to.
1558
:param parents: Revision ids of the parents of the new revision.
1559
:param config: Configuration to use.
1560
:param timestamp: Optional timestamp recorded for commit.
1561
:param timezone: Optional timezone for timestamp.
1562
:param committer: Optional committer to set for commit.
1563
:param revprops: Optional dictionary of revision properties.
1564
:param revision_id: Optional revision id.
1566
result = self._commit_builder_class(self, parents, config,
1567
timestamp, timezone, committer, revprops, revision_id)
1568
self.start_write_group()
1572
if (self.control_files._lock_count == 1 and
1573
self.control_files._lock_mode == 'w'):
1574
if self._write_group is not None:
1575
self.abort_write_group()
1576
self.control_files.unlock()
1577
raise errors.BzrError(
1578
'Must end write groups before releasing write locks.')
1579
self.control_files.unlock()
1580
if self.control_files._lock_count == 0:
1581
self._inventory_entry_cache.clear()
1582
for repo in self._fallback_repositories:
1586
def clone(self, a_bzrdir, revision_id=None):
1587
"""Clone this repository into a_bzrdir using the current format.
1589
Currently no check is made that the format of this repository and
1590
the bzrdir format are compatible. FIXME RBC 20060201.
1592
:return: The newly created destination repository.
1594
# TODO: deprecate after 0.16; cloning this with all its settings is
1595
# probably not very useful -- mbp 20070423
1596
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1597
self.copy_content_into(dest_repo, revision_id)
1600
def start_write_group(self):
1601
"""Start a write group in the repository.
1603
Write groups are used by repositories which do not have a 1:1 mapping
1604
between file ids and backend store to manage the insertion of data from
1605
both fetch and commit operations.
1607
A write lock is required around the start_write_group/commit_write_group
1608
for the support of lock-requiring repository formats.
1610
One can only insert data into a repository inside a write group.
1614
if not self.is_write_locked():
1615
raise errors.NotWriteLocked(self)
1616
if self._write_group:
1617
raise errors.BzrError('already in a write group')
1618
self._start_write_group()
1619
# so we can detect unlock/relock - the write group is now entered.
1620
self._write_group = self.get_transaction()
1622
def _start_write_group(self):
1623
"""Template method for per-repository write group startup.
1625
This is called before the write group is considered to be
1630
def sprout(self, to_bzrdir, revision_id=None):
1631
"""Create a descendent repository for new development.
1633
Unlike clone, this does not copy the settings of the repository.
1635
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1636
dest_repo.fetch(self, revision_id=revision_id)
1639
def _create_sprouting_repo(self, a_bzrdir, shared):
1640
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1641
# use target default format.
1642
dest_repo = a_bzrdir.create_repository()
1644
# Most control formats need the repository to be specifically
1645
# created, but on some old all-in-one formats it's not needed
1647
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1648
except errors.UninitializableFormat:
1649
dest_repo = a_bzrdir.open_repository()
1652
def _get_sink(self):
1653
"""Return a sink for streaming into this repository."""
1654
return StreamSink(self)
1656
def _get_source(self, to_format):
1657
"""Return a source for streaming from this repository."""
1658
return StreamSource(self, to_format)
1661
def has_revision(self, revision_id):
1662
"""True if this repository has a copy of the revision."""
1663
return revision_id in self.has_revisions((revision_id,))
1666
def has_revisions(self, revision_ids):
1667
"""Probe to find out the presence of multiple revisions.
1669
:param revision_ids: An iterable of revision_ids.
1670
:return: A set of the revision_ids that were present.
1672
parent_map = self.revisions.get_parent_map(
1673
[(rev_id,) for rev_id in revision_ids])
1675
if _mod_revision.NULL_REVISION in revision_ids:
1676
result.add(_mod_revision.NULL_REVISION)
1677
result.update([key[0] for key in parent_map])
1681
def get_revision(self, revision_id):
1682
"""Return the Revision object for a named revision."""
1683
return self.get_revisions([revision_id])[0]
1686
def get_revision_reconcile(self, revision_id):
1687
"""'reconcile' helper routine that allows access to a revision always.
1689
This variant of get_revision does not cross check the weave graph
1690
against the revision one as get_revision does: but it should only
1691
be used by reconcile, or reconcile-alike commands that are correcting
1692
or testing the revision graph.
1694
return self._get_revisions([revision_id])[0]
1697
def get_revisions(self, revision_ids):
1698
"""Get many revisions at once."""
1699
return self._get_revisions(revision_ids)
1702
def _get_revisions(self, revision_ids):
1703
"""Core work logic to get many revisions without sanity checks."""
1704
for rev_id in revision_ids:
1705
if not rev_id or not isinstance(rev_id, basestring):
1706
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1707
keys = [(key,) for key in revision_ids]
1708
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1710
for record in stream:
1711
if record.storage_kind == 'absent':
1712
raise errors.NoSuchRevision(self, record.key[0])
1713
text = record.get_bytes_as('fulltext')
1714
rev = self._serializer.read_revision_from_string(text)
1715
revs[record.key[0]] = rev
1716
return [revs[revid] for revid in revision_ids]
1719
def get_revision_xml(self, revision_id):
1720
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1721
# would have already do it.
1722
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1723
# TODO: this can't just be replaced by:
1724
# return self._serializer.write_revision_to_string(
1725
# self.get_revision(revision_id))
1726
# as cStringIO preservers the encoding unlike write_revision_to_string
1727
# or some other call down the path.
1728
rev = self.get_revision(revision_id)
1729
rev_tmp = cStringIO.StringIO()
1730
# the current serializer..
1731
self._serializer.write_revision(rev, rev_tmp)
1733
return rev_tmp.getvalue()
1735
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1736
"""Produce a generator of revision deltas.
1738
Note that the input is a sequence of REVISIONS, not revision_ids.
1739
Trees will be held in memory until the generator exits.
1740
Each delta is relative to the revision's lefthand predecessor.
1742
:param specific_fileids: if not None, the result is filtered
1743
so that only those file-ids, their parents and their
1744
children are included.
1746
# Get the revision-ids of interest
1747
required_trees = set()
1748
for revision in revisions:
1749
required_trees.add(revision.revision_id)
1750
required_trees.update(revision.parent_ids[:1])
1752
# Get the matching filtered trees. Note that it's more
1753
# efficient to pass filtered trees to changes_from() rather
1754
# than doing the filtering afterwards. changes_from() could
1755
# arguably do the filtering itself but it's path-based, not
1756
# file-id based, so filtering before or afterwards is
1758
if specific_fileids is None:
1759
trees = dict((t.get_revision_id(), t) for
1760
t in self.revision_trees(required_trees))
1762
trees = dict((t.get_revision_id(), t) for
1763
t in self._filtered_revision_trees(required_trees,
1766
# Calculate the deltas
1767
for revision in revisions:
1768
if not revision.parent_ids:
1769
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1771
old_tree = trees[revision.parent_ids[0]]
1772
yield trees[revision.revision_id].changes_from(old_tree)
1775
def get_revision_delta(self, revision_id, specific_fileids=None):
1776
"""Return the delta for one revision.
1778
The delta is relative to the left-hand predecessor of the
1781
:param specific_fileids: if not None, the result is filtered
1782
so that only those file-ids, their parents and their
1783
children are included.
1785
r = self.get_revision(revision_id)
1786
return list(self.get_deltas_for_revisions([r],
1787
specific_fileids=specific_fileids))[0]
1790
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1791
signature = gpg_strategy.sign(plaintext)
1792
self.add_signature_text(revision_id, signature)
1795
def add_signature_text(self, revision_id, signature):
1796
self.signatures.add_lines((revision_id,), (),
1797
osutils.split_lines(signature))
1799
def find_text_key_references(self):
1800
"""Find the text key references within the repository.
1802
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1803
to whether they were referred to by the inventory of the
1804
revision_id that they contain. The inventory texts from all present
1805
revision ids are assessed to generate this report.
1807
revision_keys = self.revisions.keys()
1808
w = self.inventories
1809
pb = ui.ui_factory.nested_progress_bar()
1811
return self._find_text_key_references_from_xml_inventory_lines(
1812
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1816
def _find_text_key_references_from_xml_inventory_lines(self,
1818
"""Core routine for extracting references to texts from inventories.
1820
This performs the translation of xml lines to revision ids.
1822
:param line_iterator: An iterator of lines, origin_version_id
1823
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1824
to whether they were referred to by the inventory of the
1825
revision_id that they contain. Note that if that revision_id was
1826
not part of the line_iterator's output then False will be given -
1827
even though it may actually refer to that key.
1829
if not self._serializer.support_altered_by_hack:
1830
raise AssertionError(
1831
"_find_text_key_references_from_xml_inventory_lines only "
1832
"supported for branches which store inventory as unnested xml"
1833
", not on %r" % self)
1836
# this code needs to read every new line in every inventory for the
1837
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1838
# not present in one of those inventories is unnecessary but not
1839
# harmful because we are filtering by the revision id marker in the
1840
# inventory lines : we only select file ids altered in one of those
1841
# revisions. We don't need to see all lines in the inventory because
1842
# only those added in an inventory in rev X can contain a revision=X
1844
unescape_revid_cache = {}
1845
unescape_fileid_cache = {}
1847
# jam 20061218 In a big fetch, this handles hundreds of thousands
1848
# of lines, so it has had a lot of inlining and optimizing done.
1849
# Sorry that it is a little bit messy.
1850
# Move several functions to be local variables, since this is a long
1852
search = self._file_ids_altered_regex.search
1853
unescape = _unescape_xml
1854
setdefault = result.setdefault
1855
for line, line_key in line_iterator:
1856
match = search(line)
1859
# One call to match.group() returning multiple items is quite a
1860
# bit faster than 2 calls to match.group() each returning 1
1861
file_id, revision_id = match.group('file_id', 'revision_id')
1863
# Inlining the cache lookups helps a lot when you make 170,000
1864
# lines and 350k ids, versus 8.4 unique ids.
1865
# Using a cache helps in 2 ways:
1866
# 1) Avoids unnecessary decoding calls
1867
# 2) Re-uses cached strings, which helps in future set and
1869
# (2) is enough that removing encoding entirely along with
1870
# the cache (so we are using plain strings) results in no
1871
# performance improvement.
1873
revision_id = unescape_revid_cache[revision_id]
1875
unescaped = unescape(revision_id)
1876
unescape_revid_cache[revision_id] = unescaped
1877
revision_id = unescaped
1879
# Note that unconditionally unescaping means that we deserialise
1880
# every fileid, which for general 'pull' is not great, but we don't
1881
# really want to have some many fulltexts that this matters anyway.
1884
file_id = unescape_fileid_cache[file_id]
1886
unescaped = unescape(file_id)
1887
unescape_fileid_cache[file_id] = unescaped
1890
key = (file_id, revision_id)
1891
setdefault(key, False)
1892
if revision_id == line_key[-1]:
1896
def _inventory_xml_lines_for_keys(self, keys):
1897
"""Get a line iterator of the sort needed for findind references.
1899
Not relevant for non-xml inventory repositories.
1901
Ghosts in revision_keys are ignored.
1903
:param revision_keys: The revision keys for the inventories to inspect.
1904
:return: An iterator over (inventory line, revid) for the fulltexts of
1905
all of the xml inventories specified by revision_keys.
1907
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1908
for record in stream:
1909
if record.storage_kind != 'absent':
1910
chunks = record.get_bytes_as('chunked')
1911
revid = record.key[-1]
1912
lines = osutils.chunks_to_lines(chunks)
1916
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1918
"""Helper routine for fileids_altered_by_revision_ids.
1920
This performs the translation of xml lines to revision ids.
1922
:param line_iterator: An iterator of lines, origin_version_id
1923
:param revision_keys: The revision ids to filter for. This should be a
1924
set or other type which supports efficient __contains__ lookups, as
1925
the revision key from each parsed line will be looked up in the
1926
revision_keys filter.
1927
:return: a dictionary mapping altered file-ids to an iterable of
1928
revision_ids. Each altered file-ids has the exact revision_ids that
1929
altered it listed explicitly.
1931
seen = set(self._find_text_key_references_from_xml_inventory_lines(
1932
line_iterator).iterkeys())
1933
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
1934
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
1935
self._inventory_xml_lines_for_keys(parent_keys)))
1936
new_keys = seen - parent_seen
1938
setdefault = result.setdefault
1939
for key in new_keys:
1940
setdefault(key[0], set()).add(key[-1])
1943
def _find_parent_ids_of_revisions(self, revision_ids):
1944
"""Find all parent ids that are mentioned in the revision graph.
1946
:return: set of revisions that are parents of revision_ids which are
1947
not part of revision_ids themselves
1949
parent_map = self.get_parent_map(revision_ids)
1951
map(parent_ids.update, parent_map.itervalues())
1952
parent_ids.difference_update(revision_ids)
1953
parent_ids.discard(_mod_revision.NULL_REVISION)
1956
def _find_parent_keys_of_revisions(self, revision_keys):
1957
"""Similar to _find_parent_ids_of_revisions, but used with keys.
1959
:param revision_keys: An iterable of revision_keys.
1960
:return: The parents of all revision_keys that are not already in
1963
parent_map = self.revisions.get_parent_map(revision_keys)
1965
map(parent_keys.update, parent_map.itervalues())
1966
parent_keys.difference_update(revision_keys)
1967
parent_keys.discard(_mod_revision.NULL_REVISION)
1970
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1971
"""Find the file ids and versions affected by revisions.
1973
:param revisions: an iterable containing revision ids.
1974
:param _inv_weave: The inventory weave from this repository or None.
1975
If None, the inventory weave will be opened automatically.
1976
:return: a dictionary mapping altered file-ids to an iterable of
1977
revision_ids. Each altered file-ids has the exact revision_ids that
1978
altered it listed explicitly.
1980
selected_keys = set((revid,) for revid in revision_ids)
1981
w = _inv_weave or self.inventories
1982
pb = ui.ui_factory.nested_progress_bar()
1984
return self._find_file_ids_from_xml_inventory_lines(
1985
w.iter_lines_added_or_present_in_keys(
1986
selected_keys, pb=pb),
1991
def iter_files_bytes(self, desired_files):
1992
"""Iterate through file versions.
1994
Files will not necessarily be returned in the order they occur in
1995
desired_files. No specific order is guaranteed.
1997
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1998
value supplied by the caller as part of desired_files. It should
1999
uniquely identify the file version in the caller's context. (Examples:
2000
an index number or a TreeTransform trans_id.)
2002
bytes_iterator is an iterable of bytestrings for the file. The
2003
kind of iterable and length of the bytestrings are unspecified, but for
2004
this implementation, it is a list of bytes produced by
2005
VersionedFile.get_record_stream().
2007
:param desired_files: a list of (file_id, revision_id, identifier)
2011
for file_id, revision_id, callable_data in desired_files:
2012
text_keys[(file_id, revision_id)] = callable_data
2013
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2014
if record.storage_kind == 'absent':
2015
raise errors.RevisionNotPresent(record.key, self)
2016
yield text_keys[record.key], record.get_bytes_as('chunked')
2018
def _generate_text_key_index(self, text_key_references=None,
2020
"""Generate a new text key index for the repository.
2022
This is an expensive function that will take considerable time to run.
2024
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2025
list of parents, also text keys. When a given key has no parents,
2026
the parents list will be [NULL_REVISION].
2028
# All revisions, to find inventory parents.
2029
if ancestors is None:
2030
graph = self.get_graph()
2031
ancestors = graph.get_parent_map(self.all_revision_ids())
2032
if text_key_references is None:
2033
text_key_references = self.find_text_key_references()
2034
pb = ui.ui_factory.nested_progress_bar()
2036
return self._do_generate_text_key_index(ancestors,
2037
text_key_references, pb)
2041
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2042
"""Helper for _generate_text_key_index to avoid deep nesting."""
2043
revision_order = tsort.topo_sort(ancestors)
2044
invalid_keys = set()
2046
for revision_id in revision_order:
2047
revision_keys[revision_id] = set()
2048
text_count = len(text_key_references)
2049
# a cache of the text keys to allow reuse; costs a dict of all the
2050
# keys, but saves a 2-tuple for every child of a given key.
2052
for text_key, valid in text_key_references.iteritems():
2054
invalid_keys.add(text_key)
2056
revision_keys[text_key[1]].add(text_key)
2057
text_key_cache[text_key] = text_key
2058
del text_key_references
2060
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2061
NULL_REVISION = _mod_revision.NULL_REVISION
2062
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2063
# too small for large or very branchy trees. However, for 55K path
2064
# trees, it would be easy to use too much memory trivially. Ideally we
2065
# could gauge this by looking at available real memory etc, but this is
2066
# always a tricky proposition.
2067
inventory_cache = lru_cache.LRUCache(10)
2068
batch_size = 10 # should be ~150MB on a 55K path tree
2069
batch_count = len(revision_order) / batch_size + 1
2071
pb.update("Calculating text parents", processed_texts, text_count)
2072
for offset in xrange(batch_count):
2073
to_query = revision_order[offset * batch_size:(offset + 1) *
2077
for rev_tree in self.revision_trees(to_query):
2078
revision_id = rev_tree.get_revision_id()
2079
parent_ids = ancestors[revision_id]
2080
for text_key in revision_keys[revision_id]:
2081
pb.update("Calculating text parents", processed_texts)
2082
processed_texts += 1
2083
candidate_parents = []
2084
for parent_id in parent_ids:
2085
parent_text_key = (text_key[0], parent_id)
2087
check_parent = parent_text_key not in \
2088
revision_keys[parent_id]
2090
# the parent parent_id is a ghost:
2091
check_parent = False
2092
# truncate the derived graph against this ghost.
2093
parent_text_key = None
2095
# look at the parent commit details inventories to
2096
# determine possible candidates in the per file graph.
2099
inv = inventory_cache[parent_id]
2101
inv = self.revision_tree(parent_id).inventory
2102
inventory_cache[parent_id] = inv
2104
parent_entry = inv[text_key[0]]
2105
except (KeyError, errors.NoSuchId):
2107
if parent_entry is not None:
2109
text_key[0], parent_entry.revision)
2111
parent_text_key = None
2112
if parent_text_key is not None:
2113
candidate_parents.append(
2114
text_key_cache[parent_text_key])
2115
parent_heads = text_graph.heads(candidate_parents)
2116
new_parents = list(parent_heads)
2117
new_parents.sort(key=lambda x:candidate_parents.index(x))
2118
if new_parents == []:
2119
new_parents = [NULL_REVISION]
2120
text_index[text_key] = new_parents
2122
for text_key in invalid_keys:
2123
text_index[text_key] = [NULL_REVISION]
2126
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2127
"""Get an iterable listing the keys of all the data introduced by a set
2130
The keys will be ordered so that the corresponding items can be safely
2131
fetched and inserted in that order.
2133
:returns: An iterable producing tuples of (knit-kind, file-id,
2134
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2135
'revisions'. file-id is None unless knit-kind is 'file'.
2137
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2140
for result in self._find_non_file_keys_to_fetch(revision_ids):
2143
def _find_file_keys_to_fetch(self, revision_ids, pb):
2144
# XXX: it's a bit weird to control the inventory weave caching in this
2145
# generator. Ideally the caching would be done in fetch.py I think. Or
2146
# maybe this generator should explicitly have the contract that it
2147
# should not be iterated until the previously yielded item has been
2149
inv_w = self.inventories
2151
# file ids that changed
2152
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2154
num_file_ids = len(file_ids)
2155
for file_id, altered_versions in file_ids.iteritems():
2157
pb.update("fetch texts", count, num_file_ids)
2159
yield ("file", file_id, altered_versions)
2161
def _find_non_file_keys_to_fetch(self, revision_ids):
2163
yield ("inventory", None, revision_ids)
2166
# XXX: Note ATM no callers actually pay attention to this return
2167
# instead they just use the list of revision ids and ignore
2168
# missing sigs. Consider removing this work entirely
2169
revisions_with_signatures = set(self.signatures.get_parent_map(
2170
[(r,) for r in revision_ids]))
2171
revisions_with_signatures = set(
2172
[r for (r,) in revisions_with_signatures])
2173
revisions_with_signatures.intersection_update(revision_ids)
2174
yield ("signatures", None, revisions_with_signatures)
2177
yield ("revisions", None, revision_ids)
2180
def get_inventory(self, revision_id):
2181
"""Get Inventory object by revision id."""
2182
return self.iter_inventories([revision_id]).next()
2184
def iter_inventories(self, revision_ids):
2185
"""Get many inventories by revision_ids.
2187
This will buffer some or all of the texts used in constructing the
2188
inventories in memory, but will only parse a single inventory at a
2191
:param revision_ids: The expected revision ids of the inventories.
2192
:return: An iterator of inventories.
2194
if ((None in revision_ids)
2195
or (_mod_revision.NULL_REVISION in revision_ids)):
2196
raise ValueError('cannot get null revision inventory')
2197
return self._iter_inventories(revision_ids)
2199
def _iter_inventories(self, revision_ids):
2200
"""single-document based inventory iteration."""
2201
for text, revision_id in self._iter_inventory_xmls(revision_ids):
2202
yield self.deserialise_inventory(revision_id, text)
2204
def _iter_inventory_xmls(self, revision_ids):
2205
keys = [(revision_id,) for revision_id in revision_ids]
2206
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2208
for record in stream:
2209
if record.storage_kind != 'absent':
2210
text_chunks[record.key] = record.get_bytes_as('chunked')
2212
raise errors.NoSuchRevision(self, record.key)
2214
chunks = text_chunks.pop(key)
2215
yield ''.join(chunks), key[-1]
2217
def deserialise_inventory(self, revision_id, xml):
2218
"""Transform the xml into an inventory object.
2220
:param revision_id: The expected revision id of the inventory.
2221
:param xml: A serialised inventory.
2223
result = self._serializer.read_inventory_from_string(xml, revision_id,
2224
entry_cache=self._inventory_entry_cache)
2225
if result.revision_id != revision_id:
2226
raise AssertionError('revision id mismatch %s != %s' % (
2227
result.revision_id, revision_id))
2230
def serialise_inventory(self, inv):
2231
return self._serializer.write_inventory_to_string(inv)
2233
def _serialise_inventory_to_lines(self, inv):
2234
return self._serializer.write_inventory_to_lines(inv)
2236
def get_serializer_format(self):
2237
return self._serializer.format_num
2240
def get_inventory_xml(self, revision_id):
2241
"""Get inventory XML as a file object."""
2242
texts = self._iter_inventory_xmls([revision_id])
2244
text, revision_id = texts.next()
2245
except StopIteration:
2246
raise errors.HistoryMissing(self, 'inventory', revision_id)
2250
def get_inventory_sha1(self, revision_id):
2251
"""Return the sha1 hash of the inventory entry
2253
return self.get_revision(revision_id).inventory_sha1
2255
def get_rev_id_for_revno(self, revno, known_pair):
2256
"""Return the revision id of a revno, given a later (revno, revid)
2257
pair in the same history.
2259
:return: if found (True, revid). If the available history ran out
2260
before reaching the revno, then this returns
2261
(False, (closest_revno, closest_revid)).
2263
known_revno, known_revid = known_pair
2264
partial_history = [known_revid]
2265
distance_from_known = known_revno - revno
2266
if distance_from_known < 0:
2268
'requested revno (%d) is later than given known revno (%d)'
2269
% (revno, known_revno))
2272
self, partial_history, stop_index=distance_from_known)
2273
except errors.RevisionNotPresent, err:
2274
if err.revision_id == known_revid:
2275
# The start revision (known_revid) wasn't found.
2277
# This is a stacked repository with no fallbacks, or a there's a
2278
# left-hand ghost. Either way, even though the revision named in
2279
# the error isn't in this repo, we know it's the next step in this
2280
# left-hand history.
2281
partial_history.append(err.revision_id)
2282
if len(partial_history) <= distance_from_known:
2283
# Didn't find enough history to get a revid for the revno.
2284
earliest_revno = known_revno - len(partial_history) + 1
2285
return (False, (earliest_revno, partial_history[-1]))
2286
if len(partial_history) - 1 > distance_from_known:
2287
raise AssertionError('_iter_for_revno returned too much history')
2288
return (True, partial_history[-1])
2290
def iter_reverse_revision_history(self, revision_id):
2291
"""Iterate backwards through revision ids in the lefthand history
2293
:param revision_id: The revision id to start with. All its lefthand
2294
ancestors will be traversed.
2296
graph = self.get_graph()
2297
next_id = revision_id
2299
if next_id in (None, _mod_revision.NULL_REVISION):
2302
parents = graph.get_parent_map([next_id])[next_id]
2304
raise errors.RevisionNotPresent(next_id, self)
2306
if len(parents) == 0:
2309
next_id = parents[0]
2312
def get_revision_inventory(self, revision_id):
2313
"""Return inventory of a past revision."""
2314
# TODO: Unify this with get_inventory()
2315
# bzr 0.0.6 and later imposes the constraint that the inventory_id
2316
# must be the same as its revision, so this is trivial.
2317
if revision_id is None:
2318
# This does not make sense: if there is no revision,
2319
# then it is the current tree inventory surely ?!
2320
# and thus get_root_id() is something that looks at the last
2321
# commit on the branch, and the get_root_id is an inventory check.
2322
raise NotImplementedError
2323
# return Inventory(self.get_root_id())
2325
return self.get_inventory(revision_id)
2327
def is_shared(self):
2328
"""Return True if this repository is flagged as a shared repository."""
2329
raise NotImplementedError(self.is_shared)
2332
def reconcile(self, other=None, thorough=False):
2333
"""Reconcile this repository."""
2334
from bzrlib.reconcile import RepoReconciler
2335
reconciler = RepoReconciler(self, thorough=thorough)
2336
reconciler.reconcile()
2339
def _refresh_data(self):
2340
"""Helper called from lock_* to ensure coherency with disk.
2342
The default implementation does nothing; it is however possible
2343
for repositories to maintain loaded indices across multiple locks
2344
by checking inside their implementation of this method to see
2345
whether their indices are still valid. This depends of course on
2346
the disk format being validatable in this manner. This method is
2347
also called by the refresh_data() public interface to cause a refresh
2348
to occur while in a write lock so that data inserted by a smart server
2349
push operation is visible on the client's instance of the physical
2354
def revision_tree(self, revision_id):
2355
"""Return Tree for a revision on this branch.
2357
`revision_id` may be NULL_REVISION for the empty tree revision.
2359
revision_id = _mod_revision.ensure_null(revision_id)
2360
# TODO: refactor this to use an existing revision object
2361
# so we don't need to read it in twice.
2362
if revision_id == _mod_revision.NULL_REVISION:
2363
return RevisionTree(self, Inventory(root_id=None),
2364
_mod_revision.NULL_REVISION)
2366
inv = self.get_revision_inventory(revision_id)
2367
return RevisionTree(self, inv, revision_id)
2369
def revision_trees(self, revision_ids):
2370
"""Return Trees for revisions in this repository.
2372
:param revision_ids: a sequence of revision-ids;
2373
a revision-id may not be None or 'null:'
2375
inventories = self.iter_inventories(revision_ids)
2376
for inv in inventories:
2377
yield RevisionTree(self, inv, inv.revision_id)
2379
def _filtered_revision_trees(self, revision_ids, file_ids):
2380
"""Return Tree for a revision on this branch with only some files.
2382
:param revision_ids: a sequence of revision-ids;
2383
a revision-id may not be None or 'null:'
2384
:param file_ids: if not None, the result is filtered
2385
so that only those file-ids, their parents and their
2386
children are included.
2388
inventories = self.iter_inventories(revision_ids)
2389
for inv in inventories:
2390
# Should we introduce a FilteredRevisionTree class rather
2391
# than pre-filter the inventory here?
2392
filtered_inv = inv.filter(file_ids)
2393
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
2396
def get_ancestry(self, revision_id, topo_sorted=True):
2397
"""Return a list of revision-ids integrated by a revision.
2399
The first element of the list is always None, indicating the origin
2400
revision. This might change when we have history horizons, or
2401
perhaps we should have a new API.
2403
This is topologically sorted.
2405
if _mod_revision.is_null(revision_id):
2407
if not self.has_revision(revision_id):
2408
raise errors.NoSuchRevision(self, revision_id)
2409
graph = self.get_graph()
2411
search = graph._make_breadth_first_searcher([revision_id])
2414
found, ghosts = search.next_with_ghosts()
2415
except StopIteration:
2418
if _mod_revision.NULL_REVISION in keys:
2419
keys.remove(_mod_revision.NULL_REVISION)
2421
parent_map = graph.get_parent_map(keys)
2422
keys = tsort.topo_sort(parent_map)
2423
return [None] + list(keys)
2425
def pack(self, hint=None):
2426
"""Compress the data within the repository.
2428
This operation only makes sense for some repository types. For other
2429
types it should be a no-op that just returns.
2431
This stub method does not require a lock, but subclasses should use
2432
@needs_write_lock as this is a long running call its reasonable to
2433
implicitly lock for the user.
2435
:param hint: If not supplied, the whole repository is packed.
2436
If supplied, the repository may use the hint parameter as a
2437
hint for the parts of the repository to pack. A hint can be
2438
obtained from the result of commit_write_group(). Out of
2439
date hints are simply ignored, because concurrent operations
2440
can obsolete them rapidly.
2443
def get_transaction(self):
2444
return self.control_files.get_transaction()
2446
def get_parent_map(self, revision_ids):
2447
"""See graph.StackedParentsProvider.get_parent_map"""
2448
# revisions index works in keys; this just works in revisions
2449
# therefore wrap and unwrap
2452
for revision_id in revision_ids:
2453
if revision_id == _mod_revision.NULL_REVISION:
2454
result[revision_id] = ()
2455
elif revision_id is None:
2456
raise ValueError('get_parent_map(None) is not valid')
2458
query_keys.append((revision_id ,))
2459
for ((revision_id,), parent_keys) in \
2460
self.revisions.get_parent_map(query_keys).iteritems():
2462
result[revision_id] = tuple(parent_revid
2463
for (parent_revid,) in parent_keys)
2465
result[revision_id] = (_mod_revision.NULL_REVISION,)
2468
def _make_parents_provider(self):
2471
def get_graph(self, other_repository=None):
2472
"""Return the graph walker for this repository format"""
2473
parents_provider = self._make_parents_provider()
2474
if (other_repository is not None and
2475
not self.has_same_location(other_repository)):
2476
parents_provider = graph.StackedParentsProvider(
2477
[parents_provider, other_repository._make_parents_provider()])
2478
return graph.Graph(parents_provider)
2480
def _get_versioned_file_checker(self, text_key_references=None):
2481
"""Return an object suitable for checking versioned files.
2483
:param text_key_references: if non-None, an already built
2484
dictionary mapping text keys ((fileid, revision_id) tuples)
2485
to whether they were referred to by the inventory of the
2486
revision_id that they contain. If None, this will be
2489
return _VersionedFileChecker(self,
2490
text_key_references=text_key_references)
2492
def revision_ids_to_search_result(self, result_set):
2493
"""Convert a set of revision ids to a graph SearchResult."""
2494
result_parents = set()
2495
for parents in self.get_graph().get_parent_map(
2496
result_set).itervalues():
2497
result_parents.update(parents)
2498
included_keys = result_set.intersection(result_parents)
2499
start_keys = result_set.difference(included_keys)
2500
exclude_keys = result_parents.difference(result_set)
2501
result = graph.SearchResult(start_keys, exclude_keys,
2502
len(result_set), result_set)
2506
def set_make_working_trees(self, new_value):
2507
"""Set the policy flag for making working trees when creating branches.
2509
This only applies to branches that use this repository.
2511
The default is 'True'.
2512
:param new_value: True to restore the default, False to disable making
2515
raise NotImplementedError(self.set_make_working_trees)
2517
def make_working_trees(self):
2518
"""Returns the policy for making working trees on new branches."""
2519
raise NotImplementedError(self.make_working_trees)
2522
def sign_revision(self, revision_id, gpg_strategy):
2523
plaintext = Testament.from_revision(self, revision_id).as_short_text()
2524
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2527
def has_signature_for_revision_id(self, revision_id):
2528
"""Query for a revision signature for revision_id in the repository."""
2529
if not self.has_revision(revision_id):
2530
raise errors.NoSuchRevision(self, revision_id)
2531
sig_present = (1 == len(
2532
self.signatures.get_parent_map([(revision_id,)])))
2536
def get_signature_text(self, revision_id):
2537
"""Return the text for a signature."""
2538
stream = self.signatures.get_record_stream([(revision_id,)],
2540
record = stream.next()
2541
if record.storage_kind == 'absent':
2542
raise errors.NoSuchRevision(self, revision_id)
2543
return record.get_bytes_as('fulltext')
2546
def check(self, revision_ids=None):
2547
"""Check consistency of all history of given revision_ids.
2549
Different repository implementations should override _check().
2551
:param revision_ids: A non-empty list of revision_ids whose ancestry
2552
will be checked. Typically the last revision_id of a branch.
2554
return self._check(revision_ids)
2556
def _check(self, revision_ids):
2557
result = check.Check(self)
2561
def _warn_if_deprecated(self):
2562
global _deprecation_warning_done
2563
if _deprecation_warning_done:
2565
_deprecation_warning_done = True
2566
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2567
% (self._format, self.bzrdir.transport.base))
2569
def supports_rich_root(self):
2570
return self._format.rich_root_data
2572
def _check_ascii_revisionid(self, revision_id, method):
2573
"""Private helper for ascii-only repositories."""
2574
# weave repositories refuse to store revisionids that are non-ascii.
2575
if revision_id is not None:
2576
# weaves require ascii revision ids.
2577
if isinstance(revision_id, unicode):
2579
revision_id.encode('ascii')
2580
except UnicodeEncodeError:
2581
raise errors.NonAsciiRevisionId(method, self)
2584
revision_id.decode('ascii')
2585
except UnicodeDecodeError:
2586
raise errors.NonAsciiRevisionId(method, self)
2588
def revision_graph_can_have_wrong_parents(self):
2589
"""Is it possible for this repository to have a revision graph with
2592
If True, then this repository must also implement
2593
_find_inconsistent_revision_parents so that check and reconcile can
2594
check for inconsistencies before proceeding with other checks that may
2595
depend on the revision index being consistent.
2597
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2600
# remove these delegates a while after bzr 0.15
2601
def __make_delegated(name, from_module):
2602
def _deprecated_repository_forwarder():
2603
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2604
% (name, from_module),
2607
m = __import__(from_module, globals(), locals(), [name])
2609
return getattr(m, name)
2610
except AttributeError:
2611
raise AttributeError('module %s has no name %s'
2613
globals()[name] = _deprecated_repository_forwarder
2616
'AllInOneRepository',
2617
'WeaveMetaDirRepository',
2618
'PreSplitOutRepositoryFormat',
2619
'RepositoryFormat4',
2620
'RepositoryFormat5',
2621
'RepositoryFormat6',
2622
'RepositoryFormat7',
2624
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2628
'RepositoryFormatKnit',
2629
'RepositoryFormatKnit1',
2631
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2634
def install_revision(repository, rev, revision_tree):
2635
"""Install all revision data into a repository."""
2636
install_revisions(repository, [(rev, revision_tree, None)])
2639
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2640
"""Install all revision data into a repository.
2642
Accepts an iterable of revision, tree, signature tuples. The signature
2645
repository.start_write_group()
2647
inventory_cache = lru_cache.LRUCache(10)
2648
for n, (revision, revision_tree, signature) in enumerate(iterable):
2649
_install_revision(repository, revision, revision_tree, signature,
2652
pb.update('Transferring revisions', n + 1, num_revisions)
2654
repository.abort_write_group()
2657
repository.commit_write_group()
2660
def _install_revision(repository, rev, revision_tree, signature,
2662
"""Install all revision data into a repository."""
2663
present_parents = []
2665
for p_id in rev.parent_ids:
2666
if repository.has_revision(p_id):
2667
present_parents.append(p_id)
2668
parent_trees[p_id] = repository.revision_tree(p_id)
2670
parent_trees[p_id] = repository.revision_tree(
2671
_mod_revision.NULL_REVISION)
2673
inv = revision_tree.inventory
2674
entries = inv.iter_entries()
2675
# backwards compatibility hack: skip the root id.
2676
if not repository.supports_rich_root():
2677
path, root = entries.next()
2678
if root.revision != rev.revision_id:
2679
raise errors.IncompatibleRevision(repr(repository))
2681
for path, ie in entries:
2682
text_keys[(ie.file_id, ie.revision)] = ie
2683
text_parent_map = repository.texts.get_parent_map(text_keys)
2684
missing_texts = set(text_keys) - set(text_parent_map)
2685
# Add the texts that are not already present
2686
for text_key in missing_texts:
2687
ie = text_keys[text_key]
2689
# FIXME: TODO: The following loop overlaps/duplicates that done by
2690
# commit to determine parents. There is a latent/real bug here where
2691
# the parents inserted are not those commit would do - in particular
2692
# they are not filtered by heads(). RBC, AB
2693
for revision, tree in parent_trees.iteritems():
2694
if ie.file_id not in tree:
2696
parent_id = tree.inventory[ie.file_id].revision
2697
if parent_id in text_parents:
2699
text_parents.append((ie.file_id, parent_id))
2700
lines = revision_tree.get_file(ie.file_id).readlines()
2701
repository.texts.add_lines(text_key, text_parents, lines)
2703
# install the inventory
2704
if repository._format._commit_inv_deltas and len(rev.parent_ids):
2705
# Cache this inventory
2706
inventory_cache[rev.revision_id] = inv
2708
basis_inv = inventory_cache[rev.parent_ids[0]]
2710
repository.add_inventory(rev.revision_id, inv, present_parents)
2712
delta = inv._make_delta(basis_inv)
2713
repository.add_inventory_by_delta(rev.parent_ids[0], delta,
2714
rev.revision_id, present_parents)
2716
repository.add_inventory(rev.revision_id, inv, present_parents)
2717
except errors.RevisionAlreadyPresent:
2719
if signature is not None:
2720
repository.add_signature_text(rev.revision_id, signature)
2721
repository.add_revision(rev.revision_id, rev, inv)
2724
class MetaDirRepository(Repository):
2725
"""Repositories in the new meta-dir layout.
2727
:ivar _transport: Transport for access to repository control files,
2728
typically pointing to .bzr/repository.
2731
def __init__(self, _format, a_bzrdir, control_files):
2732
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2733
self._transport = control_files._transport
2735
def is_shared(self):
2736
"""Return True if this repository is flagged as a shared repository."""
2737
return self._transport.has('shared-storage')
2740
def set_make_working_trees(self, new_value):
2741
"""Set the policy flag for making working trees when creating branches.
2743
This only applies to branches that use this repository.
2745
The default is 'True'.
2746
:param new_value: True to restore the default, False to disable making
2751
self._transport.delete('no-working-trees')
2752
except errors.NoSuchFile:
2755
self._transport.put_bytes('no-working-trees', '',
2756
mode=self.bzrdir._get_file_mode())
2758
def make_working_trees(self):
2759
"""Returns the policy for making working trees on new branches."""
2760
return not self._transport.has('no-working-trees')
2763
class MetaDirVersionedFileRepository(MetaDirRepository):
2764
"""Repositories in a meta-dir, that work via versioned file objects."""
2766
def __init__(self, _format, a_bzrdir, control_files):
2767
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2771
network_format_registry = registry.FormatRegistry()
2772
"""Registry of formats indexed by their network name.
2774
The network name for a repository format is an identifier that can be used when
2775
referring to formats with smart server operations. See
2776
RepositoryFormat.network_name() for more detail.
2780
format_registry = registry.FormatRegistry(network_format_registry)
2781
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2783
This can contain either format instances themselves, or classes/factories that
2784
can be called to obtain one.
2788
#####################################################################
2789
# Repository Formats
2791
class RepositoryFormat(object):
2792
"""A repository format.
2794
Formats provide four things:
2795
* An initialization routine to construct repository data on disk.
2796
* a optional format string which is used when the BzrDir supports
2798
* an open routine which returns a Repository instance.
2799
* A network name for referring to the format in smart server RPC
2802
There is one and only one Format subclass for each on-disk format. But
2803
there can be one Repository subclass that is used for several different
2804
formats. The _format attribute on a Repository instance can be used to
2805
determine the disk format.
2807
Formats are placed in a registry by their format string for reference
2808
during opening. These should be subclasses of RepositoryFormat for
2811
Once a format is deprecated, just deprecate the initialize and open
2812
methods on the format class. Do not deprecate the object, as the
2813
object may be created even when a repository instance hasn't been
2816
Common instance attributes:
2817
_matchingbzrdir - the bzrdir format that the repository format was
2818
originally written to work with. This can be used if manually
2819
constructing a bzrdir and repository, or more commonly for test suite
2823
# Set to True or False in derived classes. True indicates that the format
2824
# supports ghosts gracefully.
2825
supports_ghosts = None
2826
# Can this repository be given external locations to lookup additional
2827
# data. Set to True or False in derived classes.
2828
supports_external_lookups = None
2829
# Does this format support CHK bytestring lookups. Set to True or False in
2831
supports_chks = None
2832
# Should commit add an inventory, or an inventory delta to the repository.
2833
_commit_inv_deltas = True
2834
# What order should fetch operations request streams in?
2835
# The default is unordered as that is the cheapest for an origin to
2837
_fetch_order = 'unordered'
2838
# Does this repository format use deltas that can be fetched as-deltas ?
2839
# (E.g. knits, where the knit deltas can be transplanted intact.
2840
# We default to False, which will ensure that enough data to get
2841
# a full text out of any fetch stream will be grabbed.
2842
_fetch_uses_deltas = False
2843
# Should fetch trigger a reconcile after the fetch? Only needed for
2844
# some repository formats that can suffer internal inconsistencies.
2845
_fetch_reconcile = False
2846
# Does this format have < O(tree_size) delta generation. Used to hint what
2847
# code path for commit, amongst other things.
2849
# Does doing a pack operation compress data? Useful for the pack UI command
2850
# (so if there is one pack, the operation can still proceed because it may
2851
# help), and for fetching when data won't have come from the same
2853
pack_compresses = False
2856
return "<%s>" % self.__class__.__name__
2858
def __eq__(self, other):
2859
# format objects are generally stateless
2860
return isinstance(other, self.__class__)
2862
def __ne__(self, other):
2863
return not self == other
2866
def find_format(klass, a_bzrdir):
2867
"""Return the format for the repository object in a_bzrdir.
2869
This is used by bzr native formats that have a "format" file in
2870
the repository. Other methods may be used by different types of
2874
transport = a_bzrdir.get_repository_transport(None)
2875
format_string = transport.get("format").read()
2876
return format_registry.get(format_string)
2877
except errors.NoSuchFile:
2878
raise errors.NoRepositoryPresent(a_bzrdir)
2880
raise errors.UnknownFormatError(format=format_string,
2884
def register_format(klass, format):
2885
format_registry.register(format.get_format_string(), format)
2888
def unregister_format(klass, format):
2889
format_registry.remove(format.get_format_string())
2892
def get_default_format(klass):
2893
"""Return the current default format."""
2894
from bzrlib import bzrdir
2895
return bzrdir.format_registry.make_bzrdir('default').repository_format
2897
def get_format_string(self):
2898
"""Return the ASCII format string that identifies this format.
2900
Note that in pre format ?? repositories the format string is
2901
not permitted nor written to disk.
2903
raise NotImplementedError(self.get_format_string)
2905
def get_format_description(self):
2906
"""Return the short description for this format."""
2907
raise NotImplementedError(self.get_format_description)
2909
# TODO: this shouldn't be in the base class, it's specific to things that
2910
# use weaves or knits -- mbp 20070207
2911
def _get_versioned_file_store(self,
2916
versionedfile_class=None,
2917
versionedfile_kwargs={},
2919
if versionedfile_class is None:
2920
versionedfile_class = self._versionedfile_class
2921
weave_transport = control_files._transport.clone(name)
2922
dir_mode = control_files._dir_mode
2923
file_mode = control_files._file_mode
2924
return VersionedFileStore(weave_transport, prefixed=prefixed,
2926
file_mode=file_mode,
2927
versionedfile_class=versionedfile_class,
2928
versionedfile_kwargs=versionedfile_kwargs,
2931
def initialize(self, a_bzrdir, shared=False):
2932
"""Initialize a repository of this format in a_bzrdir.
2934
:param a_bzrdir: The bzrdir to put the new repository in it.
2935
:param shared: The repository should be initialized as a sharable one.
2936
:returns: The new repository object.
2938
This may raise UninitializableFormat if shared repository are not
2939
compatible the a_bzrdir.
2941
raise NotImplementedError(self.initialize)
2943
def is_supported(self):
2944
"""Is this format supported?
2946
Supported formats must be initializable and openable.
2947
Unsupported formats may not support initialization or committing or
2948
some other features depending on the reason for not being supported.
2952
def network_name(self):
2953
"""A simple byte string uniquely identifying this format for RPC calls.
2955
MetaDir repository formats use their disk format string to identify the
2956
repository over the wire. All in one formats such as bzr < 0.8, and
2957
foreign formats like svn/git and hg should use some marker which is
2958
unique and immutable.
2960
raise NotImplementedError(self.network_name)
2962
def check_conversion_target(self, target_format):
2963
raise NotImplementedError(self.check_conversion_target)
2965
def open(self, a_bzrdir, _found=False):
2966
"""Return an instance of this format for the bzrdir a_bzrdir.
2968
_found is a private parameter, do not use it.
2970
raise NotImplementedError(self.open)
2973
class MetaDirRepositoryFormat(RepositoryFormat):
2974
"""Common base class for the new repositories using the metadir layout."""
2976
rich_root_data = False
2977
supports_tree_reference = False
2978
supports_external_lookups = False
2981
def _matchingbzrdir(self):
2982
matching = bzrdir.BzrDirMetaFormat1()
2983
matching.repository_format = self
2987
super(MetaDirRepositoryFormat, self).__init__()
2989
def _create_control_files(self, a_bzrdir):
2990
"""Create the required files and the initial control_files object."""
2991
# FIXME: RBC 20060125 don't peek under the covers
2992
# NB: no need to escape relative paths that are url safe.
2993
repository_transport = a_bzrdir.get_repository_transport(self)
2994
control_files = lockable_files.LockableFiles(repository_transport,
2995
'lock', lockdir.LockDir)
2996
control_files.create_lock()
2997
return control_files
2999
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
3000
"""Upload the initial blank content."""
3001
control_files = self._create_control_files(a_bzrdir)
3002
control_files.lock_write()
3003
transport = control_files._transport
3005
utf8_files += [('shared-storage', '')]
3007
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
3008
for (filename, content_stream) in files:
3009
transport.put_file(filename, content_stream,
3010
mode=a_bzrdir._get_file_mode())
3011
for (filename, content_bytes) in utf8_files:
3012
transport.put_bytes_non_atomic(filename, content_bytes,
3013
mode=a_bzrdir._get_file_mode())
3015
control_files.unlock()
3017
def network_name(self):
3018
"""Metadir formats have matching disk and network format strings."""
3019
return self.get_format_string()
3022
# Pre-0.8 formats that don't have a disk format string (because they are
3023
# versioned by the matching control directory). We use the control directories
3024
# disk format string as a key for the network_name because they meet the
3025
# constraints (simple string, unique, immutable).
3026
network_format_registry.register_lazy(
3027
"Bazaar-NG branch, format 5\n",
3028
'bzrlib.repofmt.weaverepo',
3029
'RepositoryFormat5',
3031
network_format_registry.register_lazy(
3032
"Bazaar-NG branch, format 6\n",
3033
'bzrlib.repofmt.weaverepo',
3034
'RepositoryFormat6',
3037
# formats which have no format string are not discoverable or independently
3038
# creatable on disk, so are not registered in format_registry. They're
3039
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
3040
# needed, it's constructed directly by the BzrDir. Non-native formats where
3041
# the repository is not separately opened are similar.
3043
format_registry.register_lazy(
3044
'Bazaar-NG Repository format 7',
3045
'bzrlib.repofmt.weaverepo',
3049
format_registry.register_lazy(
3050
'Bazaar-NG Knit Repository Format 1',
3051
'bzrlib.repofmt.knitrepo',
3052
'RepositoryFormatKnit1',
3055
format_registry.register_lazy(
3056
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
3057
'bzrlib.repofmt.knitrepo',
3058
'RepositoryFormatKnit3',
3061
format_registry.register_lazy(
3062
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
3063
'bzrlib.repofmt.knitrepo',
3064
'RepositoryFormatKnit4',
3067
# Pack-based formats. There is one format for pre-subtrees, and one for
3068
# post-subtrees to allow ease of testing.
3069
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
3070
format_registry.register_lazy(
3071
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
3072
'bzrlib.repofmt.pack_repo',
3073
'RepositoryFormatKnitPack1',
3075
format_registry.register_lazy(
3076
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
3077
'bzrlib.repofmt.pack_repo',
3078
'RepositoryFormatKnitPack3',
3080
format_registry.register_lazy(
3081
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
3082
'bzrlib.repofmt.pack_repo',
3083
'RepositoryFormatKnitPack4',
3085
format_registry.register_lazy(
3086
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
3087
'bzrlib.repofmt.pack_repo',
3088
'RepositoryFormatKnitPack5',
3090
format_registry.register_lazy(
3091
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
3092
'bzrlib.repofmt.pack_repo',
3093
'RepositoryFormatKnitPack5RichRoot',
3095
format_registry.register_lazy(
3096
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
3097
'bzrlib.repofmt.pack_repo',
3098
'RepositoryFormatKnitPack5RichRootBroken',
3100
format_registry.register_lazy(
3101
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
3102
'bzrlib.repofmt.pack_repo',
3103
'RepositoryFormatKnitPack6',
3105
format_registry.register_lazy(
3106
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
3107
'bzrlib.repofmt.pack_repo',
3108
'RepositoryFormatKnitPack6RichRoot',
3111
# Development formats.
3112
# Obsolete but kept pending a CHK based subtree format.
3113
format_registry.register_lazy(
3114
("Bazaar development format 2 with subtree support "
3115
"(needs bzr.dev from before 1.8)\n"),
3116
'bzrlib.repofmt.pack_repo',
3117
'RepositoryFormatPackDevelopment2Subtree',
3120
# 1.14->1.16 go below here
3121
format_registry.register_lazy(
3122
'Bazaar development format - group compression and chk inventory'
3123
' (needs bzr.dev from 1.14)\n',
3124
'bzrlib.repofmt.groupcompress_repo',
3125
'RepositoryFormatCHK1',
3128
format_registry.register_lazy(
3129
'Bazaar development format - chk repository with bencode revision '
3130
'serialization (needs bzr.dev from 1.16)\n',
3131
'bzrlib.repofmt.groupcompress_repo',
3132
'RepositoryFormatCHK2',
3134
format_registry.register_lazy(
3135
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3136
'bzrlib.repofmt.groupcompress_repo',
3137
'RepositoryFormat2a',
3141
class InterRepository(InterObject):
3142
"""This class represents operations taking place between two repositories.
3144
Its instances have methods like copy_content and fetch, and contain
3145
references to the source and target repositories these operations can be
3148
Often we will provide convenience methods on 'repository' which carry out
3149
operations with another repository - they will always forward to
3150
InterRepository.get(other).method_name(parameters).
3153
_walk_to_common_revisions_batch_size = 50
3155
"""The available optimised InterRepository types."""
3158
def copy_content(self, revision_id=None):
3159
"""Make a complete copy of the content in self into destination.
3161
This is a destructive operation! Do not use it on existing
3164
:param revision_id: Only copy the content needed to construct
3165
revision_id and its parents.
3168
self.target.set_make_working_trees(self.source.make_working_trees())
3169
except NotImplementedError:
3171
self.target.fetch(self.source, revision_id=revision_id)
3174
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3176
"""Fetch the content required to construct revision_id.
3178
The content is copied from self.source to self.target.
3180
:param revision_id: if None all content is copied, if NULL_REVISION no
3182
:param pb: optional progress bar to use for progress reports. If not
3183
provided a default one will be created.
3186
from bzrlib.fetch import RepoFetcher
3187
f = RepoFetcher(to_repository=self.target,
3188
from_repository=self.source,
3189
last_revision=revision_id,
3190
fetch_spec=fetch_spec,
3191
pb=pb, find_ghosts=find_ghosts)
3193
def _walk_to_common_revisions(self, revision_ids):
3194
"""Walk out from revision_ids in source to revisions target has.
3196
:param revision_ids: The start point for the search.
3197
:return: A set of revision ids.
3199
target_graph = self.target.get_graph()
3200
revision_ids = frozenset(revision_ids)
3201
missing_revs = set()
3202
source_graph = self.source.get_graph()
3203
# ensure we don't pay silly lookup costs.
3204
searcher = source_graph._make_breadth_first_searcher(revision_ids)
3205
null_set = frozenset([_mod_revision.NULL_REVISION])
3206
searcher_exhausted = False
3210
# Iterate the searcher until we have enough next_revs
3211
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3213
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3214
next_revs.update(next_revs_part)
3215
ghosts.update(ghosts_part)
3216
except StopIteration:
3217
searcher_exhausted = True
3219
# If there are ghosts in the source graph, and the caller asked for
3220
# them, make sure that they are present in the target.
3221
# We don't care about other ghosts as we can't fetch them and
3222
# haven't been asked to.
3223
ghosts_to_check = set(revision_ids.intersection(ghosts))
3224
revs_to_get = set(next_revs).union(ghosts_to_check)
3226
have_revs = set(target_graph.get_parent_map(revs_to_get))
3227
# we always have NULL_REVISION present.
3228
have_revs = have_revs.union(null_set)
3229
# Check if the target is missing any ghosts we need.
3230
ghosts_to_check.difference_update(have_revs)
3232
# One of the caller's revision_ids is a ghost in both the
3233
# source and the target.
3234
raise errors.NoSuchRevision(
3235
self.source, ghosts_to_check.pop())
3236
missing_revs.update(next_revs - have_revs)
3237
# Because we may have walked past the original stop point, make
3238
# sure everything is stopped
3239
stop_revs = searcher.find_seen_ancestors(have_revs)
3240
searcher.stop_searching_any(stop_revs)
3241
if searcher_exhausted:
3243
return searcher.get_result()
3246
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3247
"""Return the revision ids that source has that target does not.
3249
:param revision_id: only return revision ids included by this
3251
:param find_ghosts: If True find missing revisions in deep history
3252
rather than just finding the surface difference.
3253
:return: A bzrlib.graph.SearchResult.
3255
# stop searching at found target revisions.
3256
if not find_ghosts and revision_id is not None:
3257
return self._walk_to_common_revisions([revision_id])
3258
# generic, possibly worst case, slow code path.
3259
target_ids = set(self.target.all_revision_ids())
3260
if revision_id is not None:
3261
source_ids = self.source.get_ancestry(revision_id)
3262
if source_ids[0] is not None:
3263
raise AssertionError()
3266
source_ids = self.source.all_revision_ids()
3267
result_set = set(source_ids).difference(target_ids)
3268
return self.source.revision_ids_to_search_result(result_set)
3271
def _same_model(source, target):
3272
"""True if source and target have the same data representation.
3274
Note: this is always called on the base class; overriding it in a
3275
subclass will have no effect.
3278
InterRepository._assert_same_model(source, target)
3280
except errors.IncompatibleRepositories, e:
3284
def _assert_same_model(source, target):
3285
"""Raise an exception if two repositories do not use the same model.
3287
if source.supports_rich_root() != target.supports_rich_root():
3288
raise errors.IncompatibleRepositories(source, target,
3289
"different rich-root support")
3290
if source._serializer != target._serializer:
3291
raise errors.IncompatibleRepositories(source, target,
3292
"different serializers")
3295
class InterSameDataRepository(InterRepository):
3296
"""Code for converting between repositories that represent the same data.
3298
Data format and model must match for this to work.
3302
def _get_repo_format_to_test(self):
3303
"""Repository format for testing with.
3305
InterSameData can pull from subtree to subtree and from non-subtree to
3306
non-subtree, so we test this with the richest repository format.
3308
from bzrlib.repofmt import knitrepo
3309
return knitrepo.RepositoryFormatKnit3()
3312
def is_compatible(source, target):
3313
return InterRepository._same_model(source, target)
3316
class InterWeaveRepo(InterSameDataRepository):
3317
"""Optimised code paths between Weave based repositories.
3319
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
3320
implemented lazy inter-object optimisation.
3324
def _get_repo_format_to_test(self):
3325
from bzrlib.repofmt import weaverepo
3326
return weaverepo.RepositoryFormat7()
3329
def is_compatible(source, target):
3330
"""Be compatible with known Weave formats.
3332
We don't test for the stores being of specific types because that
3333
could lead to confusing results, and there is no need to be
3336
from bzrlib.repofmt.weaverepo import (
3342
return (isinstance(source._format, (RepositoryFormat5,
3344
RepositoryFormat7)) and
3345
isinstance(target._format, (RepositoryFormat5,
3347
RepositoryFormat7)))
3348
except AttributeError:
3352
def copy_content(self, revision_id=None):
3353
"""See InterRepository.copy_content()."""
3354
# weave specific optimised path:
3356
self.target.set_make_working_trees(self.source.make_working_trees())
3357
except (errors.RepositoryUpgradeRequired, NotImplemented):
3359
# FIXME do not peek!
3360
if self.source._transport.listable():
3361
pb = ui.ui_factory.nested_progress_bar()
3363
self.target.texts.insert_record_stream(
3364
self.source.texts.get_record_stream(
3365
self.source.texts.keys(), 'topological', False))
3366
pb.update('copying inventory', 0, 1)
3367
self.target.inventories.insert_record_stream(
3368
self.source.inventories.get_record_stream(
3369
self.source.inventories.keys(), 'topological', False))
3370
self.target.signatures.insert_record_stream(
3371
self.source.signatures.get_record_stream(
3372
self.source.signatures.keys(),
3374
self.target.revisions.insert_record_stream(
3375
self.source.revisions.get_record_stream(
3376
self.source.revisions.keys(),
3377
'topological', True))
3381
self.target.fetch(self.source, revision_id=revision_id)
3384
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3385
"""See InterRepository.missing_revision_ids()."""
3386
# we want all revisions to satisfy revision_id in source.
3387
# but we don't want to stat every file here and there.
3388
# we want then, all revisions other needs to satisfy revision_id
3389
# checked, but not those that we have locally.
3390
# so the first thing is to get a subset of the revisions to
3391
# satisfy revision_id in source, and then eliminate those that
3392
# we do already have.
3393
# this is slow on high latency connection to self, but as this
3394
# disk format scales terribly for push anyway due to rewriting
3395
# inventory.weave, this is considered acceptable.
3397
if revision_id is not None:
3398
source_ids = self.source.get_ancestry(revision_id)
3399
if source_ids[0] is not None:
3400
raise AssertionError()
3403
source_ids = self.source._all_possible_ids()
3404
source_ids_set = set(source_ids)
3405
# source_ids is the worst possible case we may need to pull.
3406
# now we want to filter source_ids against what we actually
3407
# have in target, but don't try to check for existence where we know
3408
# we do not have a revision as that would be pointless.
3409
target_ids = set(self.target._all_possible_ids())
3410
possibly_present_revisions = target_ids.intersection(source_ids_set)
3411
actually_present_revisions = set(
3412
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3413
required_revisions = source_ids_set.difference(actually_present_revisions)
3414
if revision_id is not None:
3415
# we used get_ancestry to determine source_ids then we are assured all
3416
# revisions referenced are present as they are installed in topological order.
3417
# and the tip revision was validated by get_ancestry.
3418
result_set = required_revisions
3420
# if we just grabbed the possibly available ids, then
3421
# we only have an estimate of whats available and need to validate
3422
# that against the revision records.
3424
self.source._eliminate_revisions_not_present(required_revisions))
3425
return self.source.revision_ids_to_search_result(result_set)
3428
class InterKnitRepo(InterSameDataRepository):
3429
"""Optimised code paths between Knit based repositories."""
3432
def _get_repo_format_to_test(self):
3433
from bzrlib.repofmt import knitrepo
3434
return knitrepo.RepositoryFormatKnit1()
3437
def is_compatible(source, target):
3438
"""Be compatible with known Knit formats.
3440
We don't test for the stores being of specific types because that
3441
could lead to confusing results, and there is no need to be
3444
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
3446
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
3447
isinstance(target._format, RepositoryFormatKnit))
3448
except AttributeError:
3450
return are_knits and InterRepository._same_model(source, target)
3453
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3454
"""See InterRepository.missing_revision_ids()."""
3455
if revision_id is not None:
3456
source_ids = self.source.get_ancestry(revision_id)
3457
if source_ids[0] is not None:
3458
raise AssertionError()
3461
source_ids = self.source.all_revision_ids()
3462
source_ids_set = set(source_ids)
3463
# source_ids is the worst possible case we may need to pull.
3464
# now we want to filter source_ids against what we actually
3465
# have in target, but don't try to check for existence where we know
3466
# we do not have a revision as that would be pointless.
3467
target_ids = set(self.target.all_revision_ids())
3468
possibly_present_revisions = target_ids.intersection(source_ids_set)
3469
actually_present_revisions = set(
3470
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3471
required_revisions = source_ids_set.difference(actually_present_revisions)
3472
if revision_id is not None:
3473
# we used get_ancestry to determine source_ids then we are assured all
3474
# revisions referenced are present as they are installed in topological order.
3475
# and the tip revision was validated by get_ancestry.
3476
result_set = required_revisions
3478
# if we just grabbed the possibly available ids, then
3479
# we only have an estimate of whats available and need to validate
3480
# that against the revision records.
3482
self.source._eliminate_revisions_not_present(required_revisions))
3483
return self.source.revision_ids_to_search_result(result_set)
3486
class InterDifferingSerializer(InterRepository):
3489
def _get_repo_format_to_test(self):
3493
def is_compatible(source, target):
3494
"""Be compatible with Knit2 source and Knit3 target"""
3495
# This is redundant with format.check_conversion_target(), however that
3496
# raises an exception, and we just want to say "False" as in we won't
3497
# support converting between these formats.
3498
if source.supports_rich_root() and not target.supports_rich_root():
3500
if (source._format.supports_tree_reference
3501
and not target._format.supports_tree_reference):
3505
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3506
"""Get the best delta and base for this revision.
3508
:return: (basis_id, delta)
3510
possible_trees = [(parent_id, cache[parent_id])
3511
for parent_id in parent_ids
3512
if parent_id in cache]
3513
if len(possible_trees) == 0:
3514
# There either aren't any parents, or the parents aren't in the
3515
# cache, so just use the last converted tree
3516
possible_trees.append((basis_id, cache[basis_id]))
3518
for basis_id, basis_tree in possible_trees:
3519
delta = tree.inventory._make_delta(basis_tree.inventory)
3520
deltas.append((len(delta), basis_id, delta))
3522
return deltas[0][1:]
3524
def _get_parent_keys(self, root_key, parent_map):
3525
"""Get the parent keys for a given root id."""
3526
root_id, rev_id = root_key
3527
# Include direct parents of the revision, but only if they used
3528
# the same root_id and are heads.
3530
for parent_id in parent_map[rev_id]:
3531
if parent_id == _mod_revision.NULL_REVISION:
3533
if parent_id not in self._revision_id_to_root_id:
3534
# We probably didn't read this revision, go spend the
3535
# extra effort to actually check
3537
tree = self.source.revision_tree(parent_id)
3538
except errors.NoSuchRevision:
3539
# Ghost, fill out _revision_id_to_root_id in case we
3540
# encounter this again.
3541
# But set parent_root_id to None since we don't really know
3542
parent_root_id = None
3544
parent_root_id = tree.get_root_id()
3545
self._revision_id_to_root_id[parent_id] = None
3547
parent_root_id = self._revision_id_to_root_id[parent_id]
3548
if root_id == parent_root_id:
3549
# With stacking we _might_ want to refer to a non-local
3550
# revision, but this code path only applies when we have the
3551
# full content available, so ghosts really are ghosts, not just
3552
# the edge of local data.
3553
parent_keys.append((parent_id,))
3555
# root_id may be in the parent anyway.
3557
tree = self.source.revision_tree(parent_id)
3558
except errors.NoSuchRevision:
3559
# ghost, can't refer to it.
3563
parent_keys.append((tree.inventory[root_id].revision,))
3564
except errors.NoSuchId:
3567
g = graph.Graph(self.source.revisions)
3568
heads = g.heads(parent_keys)
3570
for key in parent_keys:
3571
if key in heads and key not in selected_keys:
3572
selected_keys.append(key)
3573
return tuple([(root_id,)+ key for key in selected_keys])
3575
def _new_root_data_stream(self, root_keys_to_create, parent_map):
3576
for root_key in root_keys_to_create:
3577
parent_keys = self._get_parent_keys(root_key, parent_map)
3578
yield versionedfile.FulltextContentFactory(root_key,
3579
parent_keys, None, '')
3581
def _fetch_batch(self, revision_ids, basis_id, cache):
3582
"""Fetch across a few revisions.
3584
:param revision_ids: The revisions to copy
3585
:param basis_id: The revision_id of a tree that must be in cache, used
3586
as a basis for delta when no other base is available
3587
:param cache: A cache of RevisionTrees that we can use.
3588
:return: The revision_id of the last converted tree. The RevisionTree
3589
for it will be in cache
3591
# Walk though all revisions; get inventory deltas, copy referenced
3592
# texts that delta references, insert the delta, revision and
3594
root_keys_to_create = set()
3597
pending_revisions = []
3598
parent_map = self.source.get_parent_map(revision_ids)
3599
for tree in self.source.revision_trees(revision_ids):
3600
current_revision_id = tree.get_revision_id()
3601
parent_ids = parent_map.get(current_revision_id, ())
3602
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3604
if self._converting_to_rich_root:
3605
self._revision_id_to_root_id[current_revision_id] = \
3607
# Find text entries that need to be copied
3608
for old_path, new_path, file_id, entry in delta:
3609
if new_path is not None:
3612
if not self.target.supports_rich_root():
3613
# The target doesn't support rich root, so we don't
3616
if self._converting_to_rich_root:
3617
# This can't be copied normally, we have to insert
3619
root_keys_to_create.add((file_id, entry.revision))
3621
text_keys.add((file_id, entry.revision))
3622
revision = self.source.get_revision(current_revision_id)
3623
pending_deltas.append((basis_id, delta,
3624
current_revision_id, revision.parent_ids))
3625
pending_revisions.append(revision)
3626
cache[current_revision_id] = tree
3627
basis_id = current_revision_id
3629
from_texts = self.source.texts
3630
to_texts = self.target.texts
3631
if root_keys_to_create:
3632
root_stream = self._new_root_data_stream(root_keys_to_create,
3634
to_texts.insert_record_stream(root_stream)
3635
to_texts.insert_record_stream(from_texts.get_record_stream(
3636
text_keys, self.target._format._fetch_order,
3637
not self.target._format._fetch_uses_deltas))
3638
# insert inventory deltas
3639
for delta in pending_deltas:
3640
self.target.add_inventory_by_delta(*delta)
3641
if self.target._fallback_repositories:
3642
# Make sure this stacked repository has all the parent inventories
3643
# for the new revisions that we are about to insert. We do this
3644
# before adding the revisions so that no revision is added until
3645
# all the inventories it may depend on are added.
3647
revision_ids = set()
3648
for revision in pending_revisions:
3649
revision_ids.add(revision.revision_id)
3650
parent_ids.update(revision.parent_ids)
3651
parent_ids.difference_update(revision_ids)
3652
parent_ids.discard(_mod_revision.NULL_REVISION)
3653
parent_map = self.source.get_parent_map(parent_ids)
3654
for parent_tree in self.source.revision_trees(parent_ids):
3655
basis_id, delta = self._get_delta_for_revision(tree, parent_ids, basis_id, cache)
3656
current_revision_id = parent_tree.get_revision_id()
3657
parents_parents = parent_map[current_revision_id]
3658
self.target.add_inventory_by_delta(
3659
basis_id, delta, current_revision_id, parents_parents)
3660
# insert signatures and revisions
3661
for revision in pending_revisions:
3663
signature = self.source.get_signature_text(
3664
revision.revision_id)
3665
self.target.add_signature_text(revision.revision_id,
3667
except errors.NoSuchRevision:
3669
self.target.add_revision(revision.revision_id, revision)
3672
def _fetch_all_revisions(self, revision_ids, pb):
3673
"""Fetch everything for the list of revisions.
3675
:param revision_ids: The list of revisions to fetch. Must be in
3677
:param pb: A ProgressBar
3680
basis_id, basis_tree = self._get_basis(revision_ids[0])
3682
cache = lru_cache.LRUCache(100)
3683
cache[basis_id] = basis_tree
3684
del basis_tree # We don't want to hang on to it here
3686
for offset in range(0, len(revision_ids), batch_size):
3687
self.target.start_write_group()
3689
pb.update('Transferring revisions', offset,
3691
batch = revision_ids[offset:offset+batch_size]
3692
basis_id = self._fetch_batch(batch, basis_id, cache)
3694
self.target.abort_write_group()
3697
hint = self.target.commit_write_group()
3700
if hints and self.target._format.pack_compresses:
3701
self.target.pack(hint=hints)
3702
pb.update('Transferring revisions', len(revision_ids),
3706
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3708
"""See InterRepository.fetch()."""
3709
if fetch_spec is not None:
3710
raise AssertionError("Not implemented yet...")
3711
if (not self.source.supports_rich_root()
3712
and self.target.supports_rich_root()):
3713
self._converting_to_rich_root = True
3714
self._revision_id_to_root_id = {}
3716
self._converting_to_rich_root = False
3717
revision_ids = self.target.search_missing_revision_ids(self.source,
3718
revision_id, find_ghosts=find_ghosts).get_keys()
3719
if not revision_ids:
3721
revision_ids = tsort.topo_sort(
3722
self.source.get_graph().get_parent_map(revision_ids))
3723
if not revision_ids:
3725
# Walk though all revisions; get inventory deltas, copy referenced
3726
# texts that delta references, insert the delta, revision and
3728
first_rev = self.source.get_revision(revision_ids[0])
3730
my_pb = ui.ui_factory.nested_progress_bar()
3733
symbol_versioning.warn(
3734
symbol_versioning.deprecated_in((1, 14, 0))
3735
% "pb parameter to fetch()")
3738
self._fetch_all_revisions(revision_ids, pb)
3740
if my_pb is not None:
3742
return len(revision_ids), 0
3744
def _get_basis(self, first_revision_id):
3745
"""Get a revision and tree which exists in the target.
3747
This assumes that first_revision_id is selected for transmission
3748
because all other ancestors are already present. If we can't find an
3749
ancestor we fall back to NULL_REVISION since we know that is safe.
3751
:return: (basis_id, basis_tree)
3753
first_rev = self.source.get_revision(first_revision_id)
3755
basis_id = first_rev.parent_ids[0]
3756
# only valid as a basis if the target has it
3757
self.target.get_revision(basis_id)
3758
# Try to get a basis tree - if its a ghost it will hit the
3759
# NoSuchRevision case.
3760
basis_tree = self.source.revision_tree(basis_id)
3761
except (IndexError, errors.NoSuchRevision):
3762
basis_id = _mod_revision.NULL_REVISION
3763
basis_tree = self.source.revision_tree(basis_id)
3764
return basis_id, basis_tree
3767
InterRepository.register_optimiser(InterDifferingSerializer)
3768
InterRepository.register_optimiser(InterSameDataRepository)
3769
InterRepository.register_optimiser(InterWeaveRepo)
3770
InterRepository.register_optimiser(InterKnitRepo)
3773
class CopyConverter(object):
3774
"""A repository conversion tool which just performs a copy of the content.
3776
This is slow but quite reliable.
3779
def __init__(self, target_format):
3780
"""Create a CopyConverter.
3782
:param target_format: The format the resulting repository should be.
3784
self.target_format = target_format
3786
def convert(self, repo, pb):
3787
"""Perform the conversion of to_convert, giving feedback via pb.
3789
:param to_convert: The disk object to convert.
3790
:param pb: a progress bar to use for progress information.
3795
# this is only useful with metadir layouts - separated repo content.
3796
# trigger an assertion if not such
3797
repo._format.get_format_string()
3798
self.repo_dir = repo.bzrdir
3799
self.step('Moving repository to repository.backup')
3800
self.repo_dir.transport.move('repository', 'repository.backup')
3801
backup_transport = self.repo_dir.transport.clone('repository.backup')
3802
repo._format.check_conversion_target(self.target_format)
3803
self.source_repo = repo._format.open(self.repo_dir,
3805
_override_transport=backup_transport)
3806
self.step('Creating new repository')
3807
converted = self.target_format.initialize(self.repo_dir,
3808
self.source_repo.is_shared())
3809
converted.lock_write()
3811
self.step('Copying content into repository.')
3812
self.source_repo.copy_content_into(converted)
3815
self.step('Deleting old repository content.')
3816
self.repo_dir.transport.delete_tree('repository.backup')
3817
self.pb.note('repository converted')
3819
def step(self, message):
3820
"""Update the pb by a step."""
3822
self.pb.update(message, self.count, self.total)
3834
def _unescaper(match, _map=_unescape_map):
3835
code = match.group(1)
3839
if not code.startswith('#'):
3841
return unichr(int(code[1:])).encode('utf8')
3847
def _unescape_xml(data):
3848
"""Unescape predefined XML entities in a string of data."""
3850
if _unescape_re is None:
3851
_unescape_re = re.compile('\&([^;]*);')
3852
return _unescape_re.sub(_unescaper, data)
3855
class _VersionedFileChecker(object):
3857
def __init__(self, repository, text_key_references=None):
3858
self.repository = repository
3859
self.text_index = self.repository._generate_text_key_index(
3860
text_key_references=text_key_references)
3862
def calculate_file_version_parents(self, text_key):
3863
"""Calculate the correct parents for a file version according to
3866
parent_keys = self.text_index[text_key]
3867
if parent_keys == [_mod_revision.NULL_REVISION]:
3869
return tuple(parent_keys)
3871
def check_file_version_parents(self, texts, progress_bar=None):
3872
"""Check the parents stored in a versioned file are correct.
3874
It also detects file versions that are not referenced by their
3875
corresponding revision's inventory.
3877
:returns: A tuple of (wrong_parents, dangling_file_versions).
3878
wrong_parents is a dict mapping {revision_id: (stored_parents,
3879
correct_parents)} for each revision_id where the stored parents
3880
are not correct. dangling_file_versions is a set of (file_id,
3881
revision_id) tuples for versions that are present in this versioned
3882
file, but not used by the corresponding inventory.
3885
self.file_ids = set([file_id for file_id, _ in
3886
self.text_index.iterkeys()])
3887
# text keys is now grouped by file_id
3888
n_weaves = len(self.file_ids)
3889
files_in_revisions = {}
3890
revisions_of_files = {}
3891
n_versions = len(self.text_index)
3892
progress_bar.update('loading text store', 0, n_versions)
3893
parent_map = self.repository.texts.get_parent_map(self.text_index)
3894
# On unlistable transports this could well be empty/error...
3895
text_keys = self.repository.texts.keys()
3896
unused_keys = frozenset(text_keys) - set(self.text_index)
3897
for num, key in enumerate(self.text_index.iterkeys()):
3898
if progress_bar is not None:
3899
progress_bar.update('checking text graph', num, n_versions)
3900
correct_parents = self.calculate_file_version_parents(key)
3902
knit_parents = parent_map[key]
3903
except errors.RevisionNotPresent:
3906
if correct_parents != knit_parents:
3907
wrong_parents[key] = (knit_parents, correct_parents)
3908
return wrong_parents, unused_keys
3911
def _old_get_graph(repository, revision_id):
3912
"""DO NOT USE. That is all. I'm serious."""
3913
graph = repository.get_graph()
3914
revision_graph = dict(((key, value) for key, value in
3915
graph.iter_ancestry([revision_id]) if value is not None))
3916
return _strip_NULL_ghosts(revision_graph)
3919
def _strip_NULL_ghosts(revision_graph):
3920
"""Also don't use this. more compatibility code for unmigrated clients."""
3921
# Filter ghosts, and null:
3922
if _mod_revision.NULL_REVISION in revision_graph:
3923
del revision_graph[_mod_revision.NULL_REVISION]
3924
for key, parents in revision_graph.items():
3925
revision_graph[key] = tuple(parent for parent in parents if parent
3927
return revision_graph
3930
class StreamSink(object):
3931
"""An object that can insert a stream into a repository.
3933
This interface handles the complexity of reserialising inventories and
3934
revisions from different formats, and allows unidirectional insertion into
3935
stacked repositories without looking for the missing basis parents
3939
def __init__(self, target_repo):
3940
self.target_repo = target_repo
3942
def insert_stream(self, stream, src_format, resume_tokens):
3943
"""Insert a stream's content into the target repository.
3945
:param src_format: a bzr repository format.
3947
:return: a list of resume tokens and an iterable of keys additional
3948
items required before the insertion can be completed.
3950
self.target_repo.lock_write()
3953
self.target_repo.resume_write_group(resume_tokens)
3956
self.target_repo.start_write_group()
3959
# locked_insert_stream performs a commit|suspend.
3960
return self._locked_insert_stream(stream, src_format, is_resume)
3962
self.target_repo.abort_write_group(suppress_errors=True)
3965
self.target_repo.unlock()
3967
def _locked_insert_stream(self, stream, src_format, is_resume):
3968
to_serializer = self.target_repo._format._serializer
3969
src_serializer = src_format._serializer
3971
if to_serializer == src_serializer:
3972
# If serializers match and the target is a pack repository, set the
3973
# write cache size on the new pack. This avoids poor performance
3974
# on transports where append is unbuffered (such as
3975
# RemoteTransport). This is safe to do because nothing should read
3976
# back from the target repository while a stream with matching
3977
# serialization is being inserted.
3978
# The exception is that a delta record from the source that should
3979
# be a fulltext may need to be expanded by the target (see
3980
# test_fetch_revisions_with_deltas_into_pack); but we take care to
3981
# explicitly flush any buffered writes first in that rare case.
3983
new_pack = self.target_repo._pack_collection._new_pack
3984
except AttributeError:
3985
# Not a pack repository
3988
new_pack.set_write_cache_size(1024*1024)
3989
for substream_type, substream in stream:
3990
if substream_type == 'texts':
3991
self.target_repo.texts.insert_record_stream(substream)
3992
elif substream_type == 'inventories':
3993
if src_serializer == to_serializer:
3994
self.target_repo.inventories.insert_record_stream(
3997
self._extract_and_insert_inventories(
3998
substream, src_serializer)
3999
elif substream_type == 'chk_bytes':
4000
# XXX: This doesn't support conversions, as it assumes the
4001
# conversion was done in the fetch code.
4002
self.target_repo.chk_bytes.insert_record_stream(substream)
4003
elif substream_type == 'revisions':
4004
# This may fallback to extract-and-insert more often than
4005
# required if the serializers are different only in terms of
4007
if src_serializer == to_serializer:
4008
self.target_repo.revisions.insert_record_stream(
4011
self._extract_and_insert_revisions(substream,
4013
elif substream_type == 'signatures':
4014
self.target_repo.signatures.insert_record_stream(substream)
4016
raise AssertionError('kaboom! %s' % (substream_type,))
4017
# Done inserting data, and the missing_keys calculations will try to
4018
# read back from the inserted data, so flush the writes to the new pack
4019
# (if this is pack format).
4020
if new_pack is not None:
4021
new_pack._write_data('', flush=True)
4022
# Find all the new revisions (including ones from resume_tokens)
4023
missing_keys = self.target_repo.get_missing_parent_inventories(
4024
check_for_missing_texts=is_resume)
4026
for prefix, versioned_file in (
4027
('texts', self.target_repo.texts),
4028
('inventories', self.target_repo.inventories),
4029
('revisions', self.target_repo.revisions),
4030
('signatures', self.target_repo.signatures),
4031
('chk_bytes', self.target_repo.chk_bytes),
4033
if versioned_file is None:
4035
missing_keys.update((prefix,) + key for key in
4036
versioned_file.get_missing_compression_parent_keys())
4037
except NotImplementedError:
4038
# cannot even attempt suspending, and missing would have failed
4039
# during stream insertion.
4040
missing_keys = set()
4043
# suspend the write group and tell the caller what we is
4044
# missing. We know we can suspend or else we would not have
4045
# entered this code path. (All repositories that can handle
4046
# missing keys can handle suspending a write group).
4047
write_group_tokens = self.target_repo.suspend_write_group()
4048
return write_group_tokens, missing_keys
4049
hint = self.target_repo.commit_write_group()
4050
if (to_serializer != src_serializer and
4051
self.target_repo._format.pack_compresses):
4052
self.target_repo.pack(hint=hint)
4055
def _extract_and_insert_inventories(self, substream, serializer):
4056
"""Generate a new inventory versionedfile in target, converting data.
4058
The inventory is retrieved from the source, (deserializing it), and
4059
stored in the target (reserializing it in a different format).
4061
for record in substream:
4062
bytes = record.get_bytes_as('fulltext')
4063
revision_id = record.key[0]
4064
inv = serializer.read_inventory_from_string(bytes, revision_id)
4065
parents = [key[0] for key in record.parents]
4066
self.target_repo.add_inventory(revision_id, inv, parents)
4068
def _extract_and_insert_revisions(self, substream, serializer):
4069
for record in substream:
4070
bytes = record.get_bytes_as('fulltext')
4071
revision_id = record.key[0]
4072
rev = serializer.read_revision_from_string(bytes)
4073
if rev.revision_id != revision_id:
4074
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4075
self.target_repo.add_revision(revision_id, rev)
4078
if self.target_repo._format._fetch_reconcile:
4079
self.target_repo.reconcile()
4082
class StreamSource(object):
4083
"""A source of a stream for fetching between repositories."""
4085
def __init__(self, from_repository, to_format):
4086
"""Create a StreamSource streaming from from_repository."""
4087
self.from_repository = from_repository
4088
self.to_format = to_format
4090
def delta_on_metadata(self):
4091
"""Return True if delta's are permitted on metadata streams.
4093
That is on revisions and signatures.
4095
src_serializer = self.from_repository._format._serializer
4096
target_serializer = self.to_format._serializer
4097
return (self.to_format._fetch_uses_deltas and
4098
src_serializer == target_serializer)
4100
def _fetch_revision_texts(self, revs):
4101
# fetch signatures first and then the revision texts
4102
# may need to be a InterRevisionStore call here.
4103
from_sf = self.from_repository.signatures
4104
# A missing signature is just skipped.
4105
keys = [(rev_id,) for rev_id in revs]
4106
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4108
self.to_format._fetch_order,
4109
not self.to_format._fetch_uses_deltas))
4110
# If a revision has a delta, this is actually expanded inside the
4111
# insert_record_stream code now, which is an alternate fix for
4113
from_rf = self.from_repository.revisions
4114
revisions = from_rf.get_record_stream(
4116
self.to_format._fetch_order,
4117
not self.delta_on_metadata())
4118
return [('signatures', signatures), ('revisions', revisions)]
4120
def _generate_root_texts(self, revs):
4121
"""This will be called by __fetch between fetching weave texts and
4122
fetching the inventory weave.
4124
Subclasses should override this if they need to generate root texts
4125
after fetching weave texts.
4127
if self._rich_root_upgrade():
4129
return bzrlib.fetch.Inter1and2Helper(
4130
self.from_repository).generate_root_texts(revs)
4134
def get_stream(self, search):
4136
revs = search.get_keys()
4137
graph = self.from_repository.get_graph()
4138
revs = list(graph.iter_topo_order(revs))
4139
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4141
for knit_kind, file_id, revisions in data_to_fetch:
4142
if knit_kind != phase:
4144
# Make a new progress bar for this phase
4145
if knit_kind == "file":
4146
# Accumulate file texts
4147
text_keys.extend([(file_id, revision) for revision in
4149
elif knit_kind == "inventory":
4150
# Now copy the file texts.
4151
from_texts = self.from_repository.texts
4152
yield ('texts', from_texts.get_record_stream(
4153
text_keys, self.to_format._fetch_order,
4154
not self.to_format._fetch_uses_deltas))
4155
# Cause an error if a text occurs after we have done the
4158
# Before we process the inventory we generate the root
4159
# texts (if necessary) so that the inventories references
4161
for _ in self._generate_root_texts(revs):
4163
# NB: This currently reopens the inventory weave in source;
4164
# using a single stream interface instead would avoid this.
4165
from_weave = self.from_repository.inventories
4166
# we fetch only the referenced inventories because we do not
4167
# know for unselected inventories whether all their required
4168
# texts are present in the other repository - it could be
4170
for info in self._get_inventory_stream(revs):
4172
elif knit_kind == "signatures":
4173
# Nothing to do here; this will be taken care of when
4174
# _fetch_revision_texts happens.
4176
elif knit_kind == "revisions":
4177
for record in self._fetch_revision_texts(revs):
4180
raise AssertionError("Unknown knit kind %r" % knit_kind)
4182
def get_stream_for_missing_keys(self, missing_keys):
4183
# missing keys can only occur when we are byte copying and not
4184
# translating (because translation means we don't send
4185
# unreconstructable deltas ever).
4187
keys['texts'] = set()
4188
keys['revisions'] = set()
4189
keys['inventories'] = set()
4190
keys['chk_bytes'] = set()
4191
keys['signatures'] = set()
4192
for key in missing_keys:
4193
keys[key[0]].add(key[1:])
4194
if len(keys['revisions']):
4195
# If we allowed copying revisions at this point, we could end up
4196
# copying a revision without copying its required texts: a
4197
# violation of the requirements for repository integrity.
4198
raise AssertionError(
4199
'cannot copy revisions to fill in missing deltas %s' % (
4200
keys['revisions'],))
4201
for substream_kind, keys in keys.iteritems():
4202
vf = getattr(self.from_repository, substream_kind)
4203
if vf is None and keys:
4204
raise AssertionError(
4205
"cannot fill in keys for a versioned file we don't"
4206
" have: %s needs %s" % (substream_kind, keys))
4208
# No need to stream something we don't have
4210
# Ask for full texts always so that we don't need more round trips
4211
# after this stream.
4212
# Some of the missing keys are genuinely ghosts, so filter absent
4213
# records. The Sink is responsible for doing another check to
4214
# ensure that ghosts don't introduce missing data for future
4216
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4217
self.to_format._fetch_order, True))
4218
yield substream_kind, stream
4220
def inventory_fetch_order(self):
4221
if self._rich_root_upgrade():
4222
return 'topological'
4224
return self.to_format._fetch_order
4226
def _rich_root_upgrade(self):
4227
return (not self.from_repository._format.rich_root_data and
4228
self.to_format.rich_root_data)
4230
def _get_inventory_stream(self, revision_ids):
4231
from_format = self.from_repository._format
4232
if (from_format.supports_chks and self.to_format.supports_chks
4233
and (from_format._serializer == self.to_format._serializer)):
4234
# Both sides support chks, and they use the same serializer, so it
4235
# is safe to transmit the chk pages and inventory pages across
4237
return self._get_chk_inventory_stream(revision_ids)
4238
elif (not from_format.supports_chks):
4239
# Source repository doesn't support chks. So we can transmit the
4240
# inventories 'as-is' and either they are just accepted on the
4241
# target, or the Sink will properly convert it.
4242
return self._get_simple_inventory_stream(revision_ids)
4244
# XXX: Hack to make not-chk->chk fetch: copy the inventories as
4245
# inventories. Note that this should probably be done somehow
4246
# as part of bzrlib.repository.StreamSink. Except JAM couldn't
4247
# figure out how a non-chk repository could possibly handle
4248
# deserializing an inventory stream from a chk repo, as it
4249
# doesn't have a way to understand individual pages.
4250
return self._get_convertable_inventory_stream(revision_ids)
4252
def _get_simple_inventory_stream(self, revision_ids):
4253
from_weave = self.from_repository.inventories
4254
yield ('inventories', from_weave.get_record_stream(
4255
[(rev_id,) for rev_id in revision_ids],
4256
self.inventory_fetch_order(),
4257
not self.delta_on_metadata()))
4259
def _get_chk_inventory_stream(self, revision_ids):
4260
"""Fetch the inventory texts, along with the associated chk maps."""
4261
# We want an inventory outside of the search set, so that we can filter
4262
# out uninteresting chk pages. For now we use
4263
# _find_revision_outside_set, but if we had a Search with cut_revs, we
4264
# could use that instead.
4265
start_rev_id = self.from_repository._find_revision_outside_set(
4267
start_rev_key = (start_rev_id,)
4268
inv_keys_to_fetch = [(rev_id,) for rev_id in revision_ids]
4269
if start_rev_id != _mod_revision.NULL_REVISION:
4270
inv_keys_to_fetch.append((start_rev_id,))
4271
# Any repo that supports chk_bytes must also support out-of-order
4272
# insertion. At least, that is how we expect it to work
4273
# We use get_record_stream instead of iter_inventories because we want
4274
# to be able to insert the stream as well. We could instead fetch
4275
# allowing deltas, and then iter_inventories, but we don't know whether
4276
# source or target is more 'local' anway.
4277
inv_stream = self.from_repository.inventories.get_record_stream(
4278
inv_keys_to_fetch, 'unordered',
4279
True) # We need them as full-texts so we can find their references
4280
uninteresting_chk_roots = set()
4281
interesting_chk_roots = set()
4282
def filter_inv_stream(inv_stream):
4283
for idx, record in enumerate(inv_stream):
4284
### child_pb.update('fetch inv', idx, len(inv_keys_to_fetch))
4285
bytes = record.get_bytes_as('fulltext')
4286
chk_inv = inventory.CHKInventory.deserialise(
4287
self.from_repository.chk_bytes, bytes, record.key)
4288
if record.key == start_rev_key:
4289
uninteresting_chk_roots.add(chk_inv.id_to_entry.key())
4290
p_id_map = chk_inv.parent_id_basename_to_file_id
4291
if p_id_map is not None:
4292
uninteresting_chk_roots.add(p_id_map.key())
4295
interesting_chk_roots.add(chk_inv.id_to_entry.key())
4296
p_id_map = chk_inv.parent_id_basename_to_file_id
4297
if p_id_map is not None:
4298
interesting_chk_roots.add(p_id_map.key())
4299
### pb.update('fetch inventory', 0, 2)
4300
yield ('inventories', filter_inv_stream(inv_stream))
4301
# Now that we have worked out all of the interesting root nodes, grab
4302
# all of the interesting pages and insert them
4303
### pb.update('fetch inventory', 1, 2)
4304
interesting = chk_map.iter_interesting_nodes(
4305
self.from_repository.chk_bytes, interesting_chk_roots,
4306
uninteresting_chk_roots)
4307
def to_stream_adapter():
4308
"""Adapt the iter_interesting_nodes result to a single stream.
4310
iter_interesting_nodes returns records as it processes them, along
4311
with keys. However, we only want to return the records themselves.
4313
for record, items in interesting:
4314
if record is not None:
4316
# XXX: We could instead call get_record_stream(records.keys())
4317
# ATM, this will always insert the records as fulltexts, and
4318
# requires that you can hang on to records once you have gone
4319
# on to the next one. Further, it causes the target to
4320
# recompress the data. Testing shows it to be faster than
4321
# requesting the records again, though.
4322
yield ('chk_bytes', to_stream_adapter())
4323
### pb.update('fetch inventory', 2, 2)
4325
def _get_convertable_inventory_stream(self, revision_ids):
4326
# XXX: One of source or target is using chks, and they don't have
4327
# compatible serializations. The StreamSink code expects to be
4328
# able to convert on the target, so we need to put
4329
# bytes-on-the-wire that can be converted
4330
yield ('inventories', self._stream_invs_as_fulltexts(revision_ids))
4332
def _stream_invs_as_fulltexts(self, revision_ids):
4333
from_repo = self.from_repository
4334
from_serializer = from_repo._format._serializer
4335
revision_keys = [(rev_id,) for rev_id in revision_ids]
4336
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4337
for inv in self.from_repository.iter_inventories(revision_ids):
4338
# XXX: This is a bit hackish, but it works. Basically,
4339
# CHKSerializer 'accidentally' supports
4340
# read/write_inventory_to_string, even though that is never
4341
# the format that is stored on disk. It *does* give us a
4342
# single string representation for an inventory, so live with
4344
# This would be far better if we had a 'serialized inventory
4345
# delta' form. Then we could use 'inventory._make_delta', and
4346
# transmit that. This would both be faster to generate, and
4347
# result in fewer bytes-on-the-wire.
4348
as_bytes = from_serializer.write_inventory_to_string(inv)
4349
key = (inv.revision_id,)
4350
parent_keys = parent_map.get(key, ())
4351
yield versionedfile.FulltextContentFactory(
4352
key, parent_keys, None, as_bytes)
4355
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4356
stop_revision=None):
4357
"""Extend the partial history to include a given index
4359
If a stop_index is supplied, stop when that index has been reached.
4360
If a stop_revision is supplied, stop when that revision is
4361
encountered. Otherwise, stop when the beginning of history is
4364
:param stop_index: The index which should be present. When it is
4365
present, history extension will stop.
4366
:param stop_revision: The revision id which should be present. When
4367
it is encountered, history extension will stop.
4369
start_revision = partial_history_cache[-1]
4370
iterator = repo.iter_reverse_revision_history(start_revision)
4372
#skip the last revision in the list
4375
if (stop_index is not None and
4376
len(partial_history_cache) > stop_index):
4378
if partial_history_cache[-1] == stop_revision:
4380
revision_id = iterator.next()
4381
partial_history_cache.append(revision_id)
4382
except StopIteration:
added
removed
bzrlib/repository.py
