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.trace import (
61
log_exception_quietly, note, mutter, mutter_callsite, warning)
64
# Old formats display a warning, but only once
65
_deprecation_warning_done = False
68
class CommitBuilder(object):
69
"""Provides an interface to build up a commit.
71
This allows describing a tree to be committed without needing to
72
know the internals of the format of the repository.
75
# all clients should supply tree roots.
76
record_root_entry = True
77
# the default CommitBuilder does not manage trees whose root is versioned.
78
_versioned_root = False
80
def __init__(self, repository, parents, config, timestamp=None,
81
timezone=None, committer=None, revprops=None,
83
"""Initiate a CommitBuilder.
85
:param repository: Repository to commit to.
86
:param parents: Revision ids of the parents of the new revision.
87
:param config: Configuration to use.
88
:param timestamp: Optional timestamp recorded for commit.
89
:param timezone: Optional timezone for timestamp.
90
:param committer: Optional committer to set for commit.
91
:param revprops: Optional dictionary of revision properties.
92
:param revision_id: Optional revision id.
97
self._committer = self._config.username()
99
self._committer = committer
101
self.new_inventory = Inventory(None)
102
self._new_revision_id = revision_id
103
self.parents = parents
104
self.repository = repository
107
if revprops is not None:
108
self._validate_revprops(revprops)
109
self._revprops.update(revprops)
111
if timestamp is None:
112
timestamp = time.time()
113
# Restrict resolution to 1ms
114
self._timestamp = round(timestamp, 3)
117
self._timezone = osutils.local_time_offset()
119
self._timezone = int(timezone)
121
self._generate_revision_if_needed()
122
self.__heads = graph.HeadsCache(repository.get_graph()).heads
123
self._basis_delta = []
124
# API compatibility, older code that used CommitBuilder did not call
125
# .record_delete(), which means the delta that is computed would not be
126
# valid. Callers that will call record_delete() should call
127
# .will_record_deletes() to indicate that.
128
self._recording_deletes = False
129
# memo'd check for no-op commits.
130
self._any_changes = False
132
def any_changes(self):
133
"""Return True if any entries were changed.
135
This includes merge-only changes. It is the core for the --unchanged
138
:return: True if any changes have occured.
140
return self._any_changes
142
def _validate_unicode_text(self, text, context):
143
"""Verify things like commit messages don't have bogus characters."""
145
raise ValueError('Invalid value for %s: %r' % (context, text))
147
def _validate_revprops(self, revprops):
148
for key, value in revprops.iteritems():
149
# We know that the XML serializers do not round trip '\r'
150
# correctly, so refuse to accept them
151
if not isinstance(value, basestring):
152
raise ValueError('revision property (%s) is not a valid'
153
' (unicode) string: %r' % (key, value))
154
self._validate_unicode_text(value,
155
'revision property (%s)' % (key,))
157
def commit(self, message):
158
"""Make the actual commit.
160
:return: The revision id of the recorded revision.
162
self._validate_unicode_text(message, 'commit message')
163
rev = _mod_revision.Revision(
164
timestamp=self._timestamp,
165
timezone=self._timezone,
166
committer=self._committer,
168
inventory_sha1=self.inv_sha1,
169
revision_id=self._new_revision_id,
170
properties=self._revprops)
171
rev.parent_ids = self.parents
172
self.repository.add_revision(self._new_revision_id, rev,
173
self.new_inventory, self._config)
174
self.repository.commit_write_group()
175
return self._new_revision_id
178
"""Abort the commit that is being built.
180
self.repository.abort_write_group()
182
def revision_tree(self):
183
"""Return the tree that was just committed.
185
After calling commit() this can be called to get a RevisionTree
186
representing the newly committed tree. This is preferred to
187
calling Repository.revision_tree() because that may require
188
deserializing the inventory, while we already have a copy in
191
if self.new_inventory is None:
192
self.new_inventory = self.repository.get_inventory(
193
self._new_revision_id)
194
return RevisionTree(self.repository, self.new_inventory,
195
self._new_revision_id)
197
def finish_inventory(self):
198
"""Tell the builder that the inventory is finished.
200
:return: The inventory id in the repository, which can be used with
201
repository.get_inventory.
203
if self.new_inventory is None:
204
# an inventory delta was accumulated without creating a new
206
basis_id = self.basis_delta_revision
207
self.inv_sha1 = self.repository.add_inventory_by_delta(
208
basis_id, self._basis_delta, self._new_revision_id,
211
if self.new_inventory.root is None:
212
raise AssertionError('Root entry should be supplied to'
213
' record_entry_contents, as of bzr 0.10.')
214
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
215
self.new_inventory.revision_id = self._new_revision_id
216
self.inv_sha1 = self.repository.add_inventory(
217
self._new_revision_id,
221
return self._new_revision_id
223
def _gen_revision_id(self):
224
"""Return new revision-id."""
225
return generate_ids.gen_revision_id(self._config.username(),
228
def _generate_revision_if_needed(self):
229
"""Create a revision id if None was supplied.
231
If the repository can not support user-specified revision ids
232
they should override this function and raise CannotSetRevisionId
233
if _new_revision_id is not None.
235
:raises: CannotSetRevisionId
237
if self._new_revision_id is None:
238
self._new_revision_id = self._gen_revision_id()
239
self.random_revid = True
241
self.random_revid = False
243
def _heads(self, file_id, revision_ids):
244
"""Calculate the graph heads for revision_ids in the graph of file_id.
246
This can use either a per-file graph or a global revision graph as we
247
have an identity relationship between the two graphs.
249
return self.__heads(revision_ids)
251
def _check_root(self, ie, parent_invs, tree):
252
"""Helper for record_entry_contents.
254
:param ie: An entry being added.
255
:param parent_invs: The inventories of the parent revisions of the
257
:param tree: The tree that is being committed.
259
# In this revision format, root entries have no knit or weave When
260
# serializing out to disk and back in root.revision is always
262
ie.revision = self._new_revision_id
264
def _require_root_change(self, tree):
265
"""Enforce an appropriate root object change.
267
This is called once when record_iter_changes is called, if and only if
268
the root was not in the delta calculated by record_iter_changes.
270
:param tree: The tree which is being committed.
272
# NB: if there are no parents then this method is not called, so no
273
# need to guard on parents having length.
274
entry = entry_factory['directory'](tree.path2id(''), '',
276
entry.revision = self._new_revision_id
277
self._basis_delta.append(('', '', entry.file_id, entry))
279
def _get_delta(self, ie, basis_inv, path):
280
"""Get a delta against the basis inventory for ie."""
281
if ie.file_id not in basis_inv:
283
result = (None, path, ie.file_id, ie)
284
self._basis_delta.append(result)
286
elif ie != basis_inv[ie.file_id]:
288
# TODO: avoid tis id2path call.
289
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
290
self._basis_delta.append(result)
296
def get_basis_delta(self):
297
"""Return the complete inventory delta versus the basis inventory.
299
This has been built up with the calls to record_delete and
300
record_entry_contents. The client must have already called
301
will_record_deletes() to indicate that they will be generating a
304
:return: An inventory delta, suitable for use with apply_delta, or
305
Repository.add_inventory_by_delta, etc.
307
if not self._recording_deletes:
308
raise AssertionError("recording deletes not activated.")
309
return self._basis_delta
311
def record_delete(self, path, file_id):
312
"""Record that a delete occured against a basis tree.
314
This is an optional API - when used it adds items to the basis_delta
315
being accumulated by the commit builder. It cannot be called unless the
316
method will_record_deletes() has been called to inform the builder that
317
a delta is being supplied.
319
:param path: The path of the thing deleted.
320
:param file_id: The file id that was deleted.
322
if not self._recording_deletes:
323
raise AssertionError("recording deletes not activated.")
324
delta = (path, None, file_id, None)
325
self._basis_delta.append(delta)
326
self._any_changes = True
329
def will_record_deletes(self):
330
"""Tell the commit builder that deletes are being notified.
332
This enables the accumulation of an inventory delta; for the resulting
333
commit to be valid, deletes against the basis MUST be recorded via
334
builder.record_delete().
336
self._recording_deletes = True
338
basis_id = self.parents[0]
340
basis_id = _mod_revision.NULL_REVISION
341
self.basis_delta_revision = basis_id
343
def record_entry_contents(self, ie, parent_invs, path, tree,
345
"""Record the content of ie from tree into the commit if needed.
347
Side effect: sets ie.revision when unchanged
349
:param ie: An inventory entry present in the commit.
350
:param parent_invs: The inventories of the parent revisions of the
352
:param path: The path the entry is at in the tree.
353
:param tree: The tree which contains this entry and should be used to
355
:param content_summary: Summary data from the tree about the paths
356
content - stat, length, exec, sha/link target. This is only
357
accessed when the entry has a revision of None - that is when it is
358
a candidate to commit.
359
:return: A tuple (change_delta, version_recorded, fs_hash).
360
change_delta is an inventory_delta change for this entry against
361
the basis tree of the commit, or None if no change occured against
363
version_recorded is True if a new version of the entry has been
364
recorded. For instance, committing a merge where a file was only
365
changed on the other side will return (delta, False).
366
fs_hash is either None, or the hash details for the path (currently
367
a tuple of the contents sha1 and the statvalue returned by
368
tree.get_file_with_stat()).
370
if self.new_inventory.root is None:
371
if ie.parent_id is not None:
372
raise errors.RootMissing()
373
self._check_root(ie, parent_invs, tree)
374
if ie.revision is None:
375
kind = content_summary[0]
377
# ie is carried over from a prior commit
379
# XXX: repository specific check for nested tree support goes here - if
380
# the repo doesn't want nested trees we skip it ?
381
if (kind == 'tree-reference' and
382
not self.repository._format.supports_tree_reference):
383
# mismatch between commit builder logic and repository:
384
# this needs the entry creation pushed down into the builder.
385
raise NotImplementedError('Missing repository subtree support.')
386
self.new_inventory.add(ie)
388
# TODO: slow, take it out of the inner loop.
390
basis_inv = parent_invs[0]
392
basis_inv = Inventory(root_id=None)
394
# ie.revision is always None if the InventoryEntry is considered
395
# for committing. We may record the previous parents revision if the
396
# content is actually unchanged against a sole head.
397
if ie.revision is not None:
398
if not self._versioned_root and path == '':
399
# repositories that do not version the root set the root's
400
# revision to the new commit even when no change occurs (more
401
# specifically, they do not record a revision on the root; and
402
# the rev id is assigned to the root during deserialisation -
403
# this masks when a change may have occurred against the basis.
404
# To match this we always issue a delta, because the revision
405
# of the root will always be changing.
406
if ie.file_id in basis_inv:
407
delta = (basis_inv.id2path(ie.file_id), path,
411
delta = (None, path, ie.file_id, ie)
412
self._basis_delta.append(delta)
413
return delta, False, None
415
# we don't need to commit this, because the caller already
416
# determined that an existing revision of this file is
417
# appropriate. If its not being considered for committing then
418
# it and all its parents to the root must be unaltered so
419
# no-change against the basis.
420
if ie.revision == self._new_revision_id:
421
raise AssertionError("Impossible situation, a skipped "
422
"inventory entry (%r) claims to be modified in this "
423
"commit (%r).", (ie, self._new_revision_id))
424
return None, False, None
425
# XXX: Friction: parent_candidates should return a list not a dict
426
# so that we don't have to walk the inventories again.
427
parent_candiate_entries = ie.parent_candidates(parent_invs)
428
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
430
for inv in parent_invs:
431
if ie.file_id in inv:
432
old_rev = inv[ie.file_id].revision
433
if old_rev in head_set:
434
heads.append(inv[ie.file_id].revision)
435
head_set.remove(inv[ie.file_id].revision)
438
# now we check to see if we need to write a new record to the
440
# We write a new entry unless there is one head to the ancestors, and
441
# the kind-derived content is unchanged.
443
# Cheapest check first: no ancestors, or more the one head in the
444
# ancestors, we write a new node.
448
# There is a single head, look it up for comparison
449
parent_entry = parent_candiate_entries[heads[0]]
450
# if the non-content specific data has changed, we'll be writing a
452
if (parent_entry.parent_id != ie.parent_id or
453
parent_entry.name != ie.name):
455
# now we need to do content specific checks:
457
# if the kind changed the content obviously has
458
if kind != parent_entry.kind:
460
# Stat cache fingerprint feedback for the caller - None as we usually
461
# don't generate one.
464
if content_summary[2] is None:
465
raise ValueError("Files must not have executable = None")
467
if (# if the file length changed we have to store:
468
parent_entry.text_size != content_summary[1] or
469
# if the exec bit has changed we have to store:
470
parent_entry.executable != content_summary[2]):
472
elif parent_entry.text_sha1 == content_summary[3]:
473
# all meta and content is unchanged (using a hash cache
474
# hit to check the sha)
475
ie.revision = parent_entry.revision
476
ie.text_size = parent_entry.text_size
477
ie.text_sha1 = parent_entry.text_sha1
478
ie.executable = parent_entry.executable
479
return self._get_delta(ie, basis_inv, path), False, None
481
# Either there is only a hash change(no hash cache entry,
482
# or same size content change), or there is no change on
484
# Provide the parent's hash to the store layer, so that the
485
# content is unchanged we will not store a new node.
486
nostore_sha = parent_entry.text_sha1
488
# We want to record a new node regardless of the presence or
489
# absence of a content change in the file.
491
ie.executable = content_summary[2]
492
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
494
text = file_obj.read()
498
ie.text_sha1, ie.text_size = self._add_text_to_weave(
499
ie.file_id, text, heads, nostore_sha)
500
# Let the caller know we generated a stat fingerprint.
501
fingerprint = (ie.text_sha1, stat_value)
502
except errors.ExistingContent:
503
# Turns out that the file content was unchanged, and we were
504
# only going to store a new node if it was changed. Carry over
506
ie.revision = parent_entry.revision
507
ie.text_size = parent_entry.text_size
508
ie.text_sha1 = parent_entry.text_sha1
509
ie.executable = parent_entry.executable
510
return self._get_delta(ie, basis_inv, path), False, None
511
elif kind == 'directory':
513
# all data is meta here, nothing specific to directory, so
515
ie.revision = parent_entry.revision
516
return self._get_delta(ie, basis_inv, path), False, None
517
self._add_text_to_weave(ie.file_id, '', heads, None)
518
elif kind == 'symlink':
519
current_link_target = content_summary[3]
521
# symlink target is not generic metadata, check if it has
523
if current_link_target != parent_entry.symlink_target:
526
# unchanged, carry over.
527
ie.revision = parent_entry.revision
528
ie.symlink_target = parent_entry.symlink_target
529
return self._get_delta(ie, basis_inv, path), False, None
530
ie.symlink_target = current_link_target
531
self._add_text_to_weave(ie.file_id, '', heads, None)
532
elif kind == 'tree-reference':
534
if content_summary[3] != parent_entry.reference_revision:
537
# unchanged, carry over.
538
ie.reference_revision = parent_entry.reference_revision
539
ie.revision = parent_entry.revision
540
return self._get_delta(ie, basis_inv, path), False, None
541
ie.reference_revision = content_summary[3]
542
self._add_text_to_weave(ie.file_id, '', heads, None)
544
raise NotImplementedError('unknown kind')
545
ie.revision = self._new_revision_id
546
self._any_changes = True
547
return self._get_delta(ie, basis_inv, path), True, fingerprint
549
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
550
_entry_factory=entry_factory):
551
"""Record a new tree via iter_changes.
553
:param tree: The tree to obtain text contents from for changed objects.
554
:param basis_revision_id: The revision id of the tree the iter_changes
555
has been generated against. Currently assumed to be the same
556
as self.parents[0] - if it is not, errors may occur.
557
:param iter_changes: An iter_changes iterator with the changes to apply
558
to basis_revision_id. The iterator must not include any items with
559
a current kind of None - missing items must be either filtered out
560
or errored-on beefore record_iter_changes sees the item.
561
:param _entry_factory: Private method to bind entry_factory locally for
563
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
566
# Create an inventory delta based on deltas between all the parents and
567
# deltas between all the parent inventories. We use inventory delta's
568
# between the inventory objects because iter_changes masks
569
# last-changed-field only changes.
571
# file_id -> change map, change is fileid, paths, changed, versioneds,
572
# parents, names, kinds, executables
574
# {file_id -> revision_id -> inventory entry, for entries in parent
575
# trees that are not parents[0]
579
revtrees = list(self.repository.revision_trees(self.parents))
580
except errors.NoSuchRevision:
581
# one or more ghosts, slow path.
583
for revision_id in self.parents:
585
revtrees.append(self.repository.revision_tree(revision_id))
586
except errors.NoSuchRevision:
588
basis_revision_id = _mod_revision.NULL_REVISION
590
revtrees.append(self.repository.revision_tree(
591
_mod_revision.NULL_REVISION))
592
# The basis inventory from a repository
594
basis_inv = revtrees[0].inventory
596
basis_inv = self.repository.revision_tree(
597
_mod_revision.NULL_REVISION).inventory
598
if len(self.parents) > 0:
599
if basis_revision_id != self.parents[0] and not ghost_basis:
601
"arbitrary basis parents not yet supported with merges")
602
for revtree in revtrees[1:]:
603
for change in revtree.inventory._make_delta(basis_inv):
604
if change[1] is None:
605
# Not present in this parent.
607
if change[2] not in merged_ids:
608
if change[0] is not None:
609
basis_entry = basis_inv[change[2]]
610
merged_ids[change[2]] = [
612
basis_entry.revision,
615
parent_entries[change[2]] = {
617
basis_entry.revision:basis_entry,
619
change[3].revision:change[3],
622
merged_ids[change[2]] = [change[3].revision]
623
parent_entries[change[2]] = {change[3].revision:change[3]}
625
merged_ids[change[2]].append(change[3].revision)
626
parent_entries[change[2]][change[3].revision] = change[3]
629
# Setup the changes from the tree:
630
# changes maps file_id -> (change, [parent revision_ids])
632
for change in iter_changes:
633
# This probably looks up in basis_inv way to much.
634
if change[1][0] is not None:
635
head_candidate = [basis_inv[change[0]].revision]
638
changes[change[0]] = change, merged_ids.get(change[0],
640
unchanged_merged = set(merged_ids) - set(changes)
641
# Extend the changes dict with synthetic changes to record merges of
643
for file_id in unchanged_merged:
644
# Record a merged version of these items that did not change vs the
645
# basis. This can be either identical parallel changes, or a revert
646
# of a specific file after a merge. The recorded content will be
647
# that of the current tree (which is the same as the basis), but
648
# the per-file graph will reflect a merge.
649
# NB:XXX: We are reconstructing path information we had, this
650
# should be preserved instead.
651
# inv delta change: (file_id, (path_in_source, path_in_target),
652
# changed_content, versioned, parent, name, kind,
655
basis_entry = basis_inv[file_id]
656
except errors.NoSuchId:
657
# a change from basis->some_parents but file_id isn't in basis
658
# so was new in the merge, which means it must have changed
659
# from basis -> current, and as it hasn't the add was reverted
660
# by the user. So we discard this change.
664
(basis_inv.id2path(file_id), tree.id2path(file_id)),
666
(basis_entry.parent_id, basis_entry.parent_id),
667
(basis_entry.name, basis_entry.name),
668
(basis_entry.kind, basis_entry.kind),
669
(basis_entry.executable, basis_entry.executable))
670
changes[file_id] = (change, merged_ids[file_id])
671
# changes contains tuples with the change and a set of inventory
672
# candidates for the file.
674
# old_path, new_path, file_id, new_inventory_entry
675
seen_root = False # Is the root in the basis delta?
676
inv_delta = self._basis_delta
677
modified_rev = self._new_revision_id
678
for change, head_candidates in changes.values():
679
if change[3][1]: # versioned in target.
680
# Several things may be happening here:
681
# We may have a fork in the per-file graph
682
# - record a change with the content from tree
683
# We may have a change against < all trees
684
# - carry over the tree that hasn't changed
685
# We may have a change against all trees
686
# - record the change with the content from tree
689
entry = _entry_factory[kind](file_id, change[5][1],
691
head_set = self._heads(change[0], set(head_candidates))
694
for head_candidate in head_candidates:
695
if head_candidate in head_set:
696
heads.append(head_candidate)
697
head_set.remove(head_candidate)
700
# Could be a carry-over situation:
701
parent_entry_revs = parent_entries.get(file_id, None)
702
if parent_entry_revs:
703
parent_entry = parent_entry_revs.get(heads[0], None)
706
if parent_entry is None:
707
# The parent iter_changes was called against is the one
708
# that is the per-file head, so any change is relevant
709
# iter_changes is valid.
710
carry_over_possible = False
712
# could be a carry over situation
713
# A change against the basis may just indicate a merge,
714
# we need to check the content against the source of the
715
# merge to determine if it was changed after the merge
717
if (parent_entry.kind != entry.kind or
718
parent_entry.parent_id != entry.parent_id or
719
parent_entry.name != entry.name):
720
# Metadata common to all entries has changed
721
# against per-file parent
722
carry_over_possible = False
724
carry_over_possible = True
725
# per-type checks for changes against the parent_entry
728
# Cannot be a carry-over situation
729
carry_over_possible = False
730
# Populate the entry in the delta
732
# XXX: There is still a small race here: If someone reverts the content of a file
733
# after iter_changes examines and decides it has changed,
734
# we will unconditionally record a new version even if some
735
# other process reverts it while commit is running (with
736
# the revert happening after iter_changes did it's
739
entry.executable = True
741
entry.executable = False
742
if (carry_over_possible and
743
parent_entry.executable == entry.executable):
744
# Check the file length, content hash after reading
746
nostore_sha = parent_entry.text_sha1
749
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
751
text = file_obj.read()
755
entry.text_sha1, entry.text_size = self._add_text_to_weave(
756
file_id, text, heads, nostore_sha)
757
yield file_id, change[1][1], (entry.text_sha1, stat_value)
758
except errors.ExistingContent:
759
# No content change against a carry_over parent
760
# Perhaps this should also yield a fs hash update?
762
entry.text_size = parent_entry.text_size
763
entry.text_sha1 = parent_entry.text_sha1
764
elif kind == 'symlink':
766
entry.symlink_target = tree.get_symlink_target(file_id)
767
if (carry_over_possible and
768
parent_entry.symlink_target == entry.symlink_target):
771
self._add_text_to_weave(change[0], '', heads, None)
772
elif kind == 'directory':
773
if carry_over_possible:
776
# Nothing to set on the entry.
777
# XXX: split into the Root and nonRoot versions.
778
if change[1][1] != '' or self.repository.supports_rich_root():
779
self._add_text_to_weave(change[0], '', heads, None)
780
elif kind == 'tree-reference':
781
if not self.repository._format.supports_tree_reference:
782
# This isn't quite sane as an error, but we shouldn't
783
# ever see this code path in practice: tree's don't
784
# permit references when the repo doesn't support tree
786
raise errors.UnsupportedOperation(tree.add_reference,
788
reference_revision = tree.get_reference_revision(change[0])
789
entry.reference_revision = reference_revision
790
if (carry_over_possible and
791
parent_entry.reference_revision == reference_revision):
794
self._add_text_to_weave(change[0], '', heads, None)
796
raise AssertionError('unknown kind %r' % kind)
798
entry.revision = modified_rev
800
entry.revision = parent_entry.revision
803
new_path = change[1][1]
804
inv_delta.append((change[1][0], new_path, change[0], entry))
807
self.new_inventory = None
809
self._any_changes = True
811
# housekeeping root entry changes do not affect no-change commits.
812
self._require_root_change(tree)
813
self.basis_delta_revision = basis_revision_id
815
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
816
parent_keys = tuple([(file_id, parent) for parent in parents])
817
return self.repository.texts._add_text(
818
(file_id, self._new_revision_id), parent_keys, new_text,
819
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
822
class RootCommitBuilder(CommitBuilder):
823
"""This commitbuilder actually records the root id"""
825
# the root entry gets versioned properly by this builder.
826
_versioned_root = True
828
def _check_root(self, ie, parent_invs, tree):
829
"""Helper for record_entry_contents.
831
:param ie: An entry being added.
832
:param parent_invs: The inventories of the parent revisions of the
834
:param tree: The tree that is being committed.
837
def _require_root_change(self, tree):
838
"""Enforce an appropriate root object change.
840
This is called once when record_iter_changes is called, if and only if
841
the root was not in the delta calculated by record_iter_changes.
843
:param tree: The tree which is being committed.
845
# versioned roots do not change unless the tree found a change.
848
######################################################################
851
class Repository(object):
852
"""Repository holding history for one or more branches.
854
The repository holds and retrieves historical information including
855
revisions and file history. It's normally accessed only by the Branch,
856
which views a particular line of development through that history.
858
The Repository builds on top of some byte storage facilies (the revisions,
859
signatures, inventories, texts and chk_bytes attributes) and a Transport,
860
which respectively provide byte storage and a means to access the (possibly
863
The byte storage facilities are addressed via tuples, which we refer to
864
as 'keys' throughout the code base. Revision_keys, inventory_keys and
865
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
866
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
867
byte string made up of a hash identifier and a hash value.
868
We use this interface because it allows low friction with the underlying
869
code that implements disk indices, network encoding and other parts of
872
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
873
the serialised revisions for the repository. This can be used to obtain
874
revision graph information or to access raw serialised revisions.
875
The result of trying to insert data into the repository via this store
876
is undefined: it should be considered read-only except for implementors
878
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
879
the serialised signatures for the repository. This can be used to
880
obtain access to raw serialised signatures. The result of trying to
881
insert data into the repository via this store is undefined: it should
882
be considered read-only except for implementors of repositories.
883
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
884
the serialised inventories for the repository. This can be used to
885
obtain unserialised inventories. The result of trying to insert data
886
into the repository via this store is undefined: it should be
887
considered read-only except for implementors of repositories.
888
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
889
texts of files and directories for the repository. This can be used to
890
obtain file texts or file graphs. Note that Repository.iter_file_bytes
891
is usually a better interface for accessing file texts.
892
The result of trying to insert data into the repository via this store
893
is undefined: it should be considered read-only except for implementors
895
:ivar chk_bytes: A bzrlib.versionedfile.VersionedFiles instance containing
896
any data the repository chooses to store or have indexed by its hash.
897
The result of trying to insert data into the repository via this store
898
is undefined: it should be considered read-only except for implementors
900
:ivar _transport: Transport for file access to repository, typically
901
pointing to .bzr/repository.
904
# What class to use for a CommitBuilder. Often its simpler to change this
905
# in a Repository class subclass rather than to override
906
# get_commit_builder.
907
_commit_builder_class = CommitBuilder
908
# The search regex used by xml based repositories to determine what things
909
# where changed in a single commit.
910
_file_ids_altered_regex = lazy_regex.lazy_compile(
911
r'file_id="(?P<file_id>[^"]+)"'
912
r'.* revision="(?P<revision_id>[^"]+)"'
915
def abort_write_group(self, suppress_errors=False):
916
"""Commit the contents accrued within the current write group.
918
:param suppress_errors: if true, abort_write_group will catch and log
919
unexpected errors that happen during the abort, rather than
920
allowing them to propagate. Defaults to False.
922
:seealso: start_write_group.
924
if self._write_group is not self.get_transaction():
925
# has an unlock or relock occured ?
926
raise errors.BzrError(
927
'mismatched lock context and write group. %r, %r' %
928
(self._write_group, self.get_transaction()))
930
self._abort_write_group()
931
except Exception, exc:
932
self._write_group = None
933
if not suppress_errors:
935
mutter('abort_write_group failed')
936
log_exception_quietly()
937
note('bzr: ERROR (ignored): %s', exc)
938
self._write_group = None
940
def _abort_write_group(self):
941
"""Template method for per-repository write group cleanup.
943
This is called during abort before the write group is considered to be
944
finished and should cleanup any internal state accrued during the write
945
group. There is no requirement that data handed to the repository be
946
*not* made available - this is not a rollback - but neither should any
947
attempt be made to ensure that data added is fully commited. Abort is
948
invoked when an error has occured so futher disk or network operations
949
may not be possible or may error and if possible should not be
953
def add_fallback_repository(self, repository):
954
"""Add a repository to use for looking up data not held locally.
956
:param repository: A repository.
958
if not self._format.supports_external_lookups:
959
raise errors.UnstackableRepositoryFormat(self._format, self.base)
961
# This repository will call fallback.unlock() when we transition to
962
# the unlocked state, so we make sure to increment the lock count
963
repository.lock_read()
964
self._check_fallback_repository(repository)
965
self._fallback_repositories.append(repository)
966
self.texts.add_fallback_versioned_files(repository.texts)
967
self.inventories.add_fallback_versioned_files(repository.inventories)
968
self.revisions.add_fallback_versioned_files(repository.revisions)
969
self.signatures.add_fallback_versioned_files(repository.signatures)
970
if self.chk_bytes is not None:
971
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
973
def _check_fallback_repository(self, repository):
974
"""Check that this repository can fallback to repository safely.
976
Raise an error if not.
978
:param repository: A repository to fallback to.
980
return InterRepository._assert_same_model(self, repository)
982
def add_inventory(self, revision_id, inv, parents):
983
"""Add the inventory inv to the repository as revision_id.
985
:param parents: The revision ids of the parents that revision_id
986
is known to have and are in the repository already.
988
:returns: The validator(which is a sha1 digest, though what is sha'd is
989
repository format specific) of the serialized inventory.
991
if not self.is_in_write_group():
992
raise AssertionError("%r not in write group" % (self,))
993
_mod_revision.check_not_reserved_id(revision_id)
994
if not (inv.revision_id is None or inv.revision_id == revision_id):
995
raise AssertionError(
996
"Mismatch between inventory revision"
997
" id and insertion revid (%r, %r)"
998
% (inv.revision_id, revision_id))
1000
raise AssertionError()
1001
return self._add_inventory_checked(revision_id, inv, parents)
1003
def _add_inventory_checked(self, revision_id, inv, parents):
1004
"""Add inv to the repository after checking the inputs.
1006
This function can be overridden to allow different inventory styles.
1008
:seealso: add_inventory, for the contract.
1010
inv_lines = self._serialise_inventory_to_lines(inv)
1011
return self._inventory_add_lines(revision_id, parents,
1012
inv_lines, check_content=False)
1014
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1015
parents, basis_inv=None, propagate_caches=False):
1016
"""Add a new inventory expressed as a delta against another revision.
1018
See the inventory developers documentation for the theory behind
1021
:param basis_revision_id: The inventory id the delta was created
1022
against. (This does not have to be a direct parent.)
1023
:param delta: The inventory delta (see Inventory.apply_delta for
1025
:param new_revision_id: The revision id that the inventory is being
1027
:param parents: The revision ids of the parents that revision_id is
1028
known to have and are in the repository already. These are supplied
1029
for repositories that depend on the inventory graph for revision
1030
graph access, as well as for those that pun ancestry with delta
1032
:param basis_inv: The basis inventory if it is already known,
1034
:param propagate_caches: If True, the caches for this inventory are
1035
copied to and updated for the result if possible.
1037
:returns: (validator, new_inv)
1038
The validator(which is a sha1 digest, though what is sha'd is
1039
repository format specific) of the serialized inventory, and the
1040
resulting inventory.
1042
if not self.is_in_write_group():
1043
raise AssertionError("%r not in write group" % (self,))
1044
_mod_revision.check_not_reserved_id(new_revision_id)
1045
basis_tree = self.revision_tree(basis_revision_id)
1046
basis_tree.lock_read()
1048
# Note that this mutates the inventory of basis_tree, which not all
1049
# inventory implementations may support: A better idiom would be to
1050
# return a new inventory, but as there is no revision tree cache in
1051
# repository this is safe for now - RBC 20081013
1052
if basis_inv is None:
1053
basis_inv = basis_tree.inventory
1054
basis_inv.apply_delta(delta)
1055
basis_inv.revision_id = new_revision_id
1056
return (self.add_inventory(new_revision_id, basis_inv, parents),
1061
def _inventory_add_lines(self, revision_id, parents, lines,
1062
check_content=True):
1063
"""Store lines in inv_vf and return the sha1 of the inventory."""
1064
parents = [(parent,) for parent in parents]
1065
return self.inventories.add_lines((revision_id,), parents, lines,
1066
check_content=check_content)[0]
1068
def add_revision(self, revision_id, rev, inv=None, config=None):
1069
"""Add rev to the revision store as revision_id.
1071
:param revision_id: the revision id to use.
1072
:param rev: The revision object.
1073
:param inv: The inventory for the revision. if None, it will be looked
1074
up in the inventory storer
1075
:param config: If None no digital signature will be created.
1076
If supplied its signature_needed method will be used
1077
to determine if a signature should be made.
1079
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
1081
_mod_revision.check_not_reserved_id(revision_id)
1082
if config is not None and config.signature_needed():
1084
inv = self.get_inventory(revision_id)
1085
plaintext = Testament(rev, inv).as_short_text()
1086
self.store_revision_signature(
1087
gpg.GPGStrategy(config), plaintext, revision_id)
1088
# check inventory present
1089
if not self.inventories.get_parent_map([(revision_id,)]):
1091
raise errors.WeaveRevisionNotPresent(revision_id,
1094
# yes, this is not suitable for adding with ghosts.
1095
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
1098
key = (revision_id,)
1099
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
1100
self._add_revision(rev)
1102
def _add_revision(self, revision):
1103
text = self._serializer.write_revision_to_string(revision)
1104
key = (revision.revision_id,)
1105
parents = tuple((parent,) for parent in revision.parent_ids)
1106
self.revisions.add_lines(key, parents, osutils.split_lines(text))
1108
def all_revision_ids(self):
1109
"""Returns a list of all the revision ids in the repository.
1111
This is conceptually deprecated because code should generally work on
1112
the graph reachable from a particular revision, and ignore any other
1113
revisions that might be present. There is no direct replacement
1116
if 'evil' in debug.debug_flags:
1117
mutter_callsite(2, "all_revision_ids is linear with history.")
1118
return self._all_revision_ids()
1120
def _all_revision_ids(self):
1121
"""Returns a list of all the revision ids in the repository.
1123
These are in as much topological order as the underlying store can
1126
raise NotImplementedError(self._all_revision_ids)
1128
def break_lock(self):
1129
"""Break a lock if one is present from another instance.
1131
Uses the ui factory to ask for confirmation if the lock may be from
1134
self.control_files.break_lock()
1137
def _eliminate_revisions_not_present(self, revision_ids):
1138
"""Check every revision id in revision_ids to see if we have it.
1140
Returns a set of the present revisions.
1143
graph = self.get_graph()
1144
parent_map = graph.get_parent_map(revision_ids)
1145
# The old API returned a list, should this actually be a set?
1146
return parent_map.keys()
1149
def create(a_bzrdir):
1150
"""Construct the current default format repository in a_bzrdir."""
1151
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
1153
def __init__(self, _format, a_bzrdir, control_files):
1154
"""instantiate a Repository.
1156
:param _format: The format of the repository on disk.
1157
:param a_bzrdir: The BzrDir of the repository.
1159
In the future we will have a single api for all stores for
1160
getting file texts, inventories and revisions, then
1161
this construct will accept instances of those things.
1163
super(Repository, self).__init__()
1164
self._format = _format
1165
# the following are part of the public API for Repository:
1166
self.bzrdir = a_bzrdir
1167
self.control_files = control_files
1168
self._transport = control_files._transport
1169
self.base = self._transport.base
1171
self._reconcile_does_inventory_gc = True
1172
self._reconcile_fixes_text_parents = False
1173
self._reconcile_backsup_inventory = True
1174
# not right yet - should be more semantically clear ?
1176
# TODO: make sure to construct the right store classes, etc, depending
1177
# on whether escaping is required.
1178
self._warn_if_deprecated()
1179
self._write_group = None
1180
# Additional places to query for data.
1181
self._fallback_repositories = []
1182
# An InventoryEntry cache, used during deserialization
1183
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1186
return '%s(%r)' % (self.__class__.__name__,
1189
def has_same_location(self, other):
1190
"""Returns a boolean indicating if this repository is at the same
1191
location as another repository.
1193
This might return False even when two repository objects are accessing
1194
the same physical repository via different URLs.
1196
if self.__class__ is not other.__class__:
1198
return (self._transport.base == other._transport.base)
1200
def is_in_write_group(self):
1201
"""Return True if there is an open write group.
1203
:seealso: start_write_group.
1205
return self._write_group is not None
1207
def is_locked(self):
1208
return self.control_files.is_locked()
1210
def is_write_locked(self):
1211
"""Return True if this object is write locked."""
1212
return self.is_locked() and self.control_files._lock_mode == 'w'
1214
def lock_write(self, token=None):
1215
"""Lock this repository for writing.
1217
This causes caching within the repository obejct to start accumlating
1218
data during reads, and allows a 'write_group' to be obtained. Write
1219
groups must be used for actual data insertion.
1221
:param token: if this is already locked, then lock_write will fail
1222
unless the token matches the existing lock.
1223
:returns: a token if this instance supports tokens, otherwise None.
1224
:raises TokenLockingNotSupported: when a token is given but this
1225
instance doesn't support using token locks.
1226
:raises MismatchedToken: if the specified token doesn't match the token
1227
of the existing lock.
1228
:seealso: start_write_group.
1230
A token should be passed in if you know that you have locked the object
1231
some other way, and need to synchronise this object's state with that
1234
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1236
locked = self.is_locked()
1237
result = self.control_files.lock_write(token=token)
1239
for repo in self._fallback_repositories:
1240
# Writes don't affect fallback repos
1242
self._refresh_data()
1245
def lock_read(self):
1246
locked = self.is_locked()
1247
self.control_files.lock_read()
1249
for repo in self._fallback_repositories:
1251
self._refresh_data()
1253
def get_physical_lock_status(self):
1254
return self.control_files.get_physical_lock_status()
1256
def leave_lock_in_place(self):
1257
"""Tell this repository not to release the physical lock when this
1260
If lock_write doesn't return a token, then this method is not supported.
1262
self.control_files.leave_in_place()
1264
def dont_leave_lock_in_place(self):
1265
"""Tell this repository to release the physical lock when this
1266
object is unlocked, even if it didn't originally acquire it.
1268
If lock_write doesn't return a token, then this method is not supported.
1270
self.control_files.dont_leave_in_place()
1273
def gather_stats(self, revid=None, committers=None):
1274
"""Gather statistics from a revision id.
1276
:param revid: The revision id to gather statistics from, if None, then
1277
no revision specific statistics are gathered.
1278
:param committers: Optional parameter controlling whether to grab
1279
a count of committers from the revision specific statistics.
1280
:return: A dictionary of statistics. Currently this contains:
1281
committers: The number of committers if requested.
1282
firstrev: A tuple with timestamp, timezone for the penultimate left
1283
most ancestor of revid, if revid is not the NULL_REVISION.
1284
latestrev: A tuple with timestamp, timezone for revid, if revid is
1285
not the NULL_REVISION.
1286
revisions: The total revision count in the repository.
1287
size: An estimate disk size of the repository in bytes.
1290
if revid and committers:
1291
result['committers'] = 0
1292
if revid and revid != _mod_revision.NULL_REVISION:
1294
all_committers = set()
1295
revisions = self.get_ancestry(revid)
1296
# pop the leading None
1298
first_revision = None
1300
# ignore the revisions in the middle - just grab first and last
1301
revisions = revisions[0], revisions[-1]
1302
for revision in self.get_revisions(revisions):
1303
if not first_revision:
1304
first_revision = revision
1306
all_committers.add(revision.committer)
1307
last_revision = revision
1309
result['committers'] = len(all_committers)
1310
result['firstrev'] = (first_revision.timestamp,
1311
first_revision.timezone)
1312
result['latestrev'] = (last_revision.timestamp,
1313
last_revision.timezone)
1315
# now gather global repository information
1316
# XXX: This is available for many repos regardless of listability.
1317
if self.bzrdir.root_transport.listable():
1318
# XXX: do we want to __define len__() ?
1319
# Maybe the versionedfiles object should provide a different
1320
# method to get the number of keys.
1321
result['revisions'] = len(self.revisions.keys())
1322
# result['size'] = t
1325
def find_branches(self, using=False):
1326
"""Find branches underneath this repository.
1328
This will include branches inside other branches.
1330
:param using: If True, list only branches using this repository.
1332
if using and not self.is_shared():
1334
return [self.bzrdir.open_branch()]
1335
except errors.NotBranchError:
1337
class Evaluator(object):
1340
self.first_call = True
1342
def __call__(self, bzrdir):
1343
# On the first call, the parameter is always the bzrdir
1344
# containing the current repo.
1345
if not self.first_call:
1347
repository = bzrdir.open_repository()
1348
except errors.NoRepositoryPresent:
1351
return False, (None, repository)
1352
self.first_call = False
1354
value = (bzrdir.open_branch(), None)
1355
except errors.NotBranchError:
1356
value = (None, None)
1360
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1361
self.bzrdir.root_transport, evaluate=Evaluator()):
1362
if branch is not None:
1363
branches.append(branch)
1364
if not using and repository is not None:
1365
branches.extend(repository.find_branches())
1369
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1370
"""Return the revision ids that other has that this does not.
1372
These are returned in topological order.
1374
revision_id: only return revision ids included by revision_id.
1376
return InterRepository.get(other, self).search_missing_revision_ids(
1377
revision_id, find_ghosts)
1381
"""Open the repository rooted at base.
1383
For instance, if the repository is at URL/.bzr/repository,
1384
Repository.open(URL) -> a Repository instance.
1386
control = bzrdir.BzrDir.open(base)
1387
return control.open_repository()
1389
def copy_content_into(self, destination, revision_id=None):
1390
"""Make a complete copy of the content in self into destination.
1392
This is a destructive operation! Do not use it on existing
1395
return InterRepository.get(self, destination).copy_content(revision_id)
1397
def commit_write_group(self):
1398
"""Commit the contents accrued within the current write group.
1400
:seealso: start_write_group.
1402
if self._write_group is not self.get_transaction():
1403
# has an unlock or relock occured ?
1404
raise errors.BzrError('mismatched lock context %r and '
1406
(self.get_transaction(), self._write_group))
1407
result = self._commit_write_group()
1408
self._write_group = None
1411
def _commit_write_group(self):
1412
"""Template method for per-repository write group cleanup.
1414
This is called before the write group is considered to be
1415
finished and should ensure that all data handed to the repository
1416
for writing during the write group is safely committed (to the
1417
extent possible considering file system caching etc).
1420
def suspend_write_group(self):
1421
raise errors.UnsuspendableWriteGroup(self)
1423
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1424
"""Return the keys of missing inventory parents for revisions added in
1427
A revision is not complete if the inventory delta for that revision
1428
cannot be calculated. Therefore if the parent inventories of a
1429
revision are not present, the revision is incomplete, and e.g. cannot
1430
be streamed by a smart server. This method finds missing inventory
1431
parents for revisions added in this write group.
1433
if not self._format.supports_external_lookups:
1434
# This is only an issue for stacked repositories
1436
if not self.is_in_write_group():
1437
raise AssertionError('not in a write group')
1439
# XXX: We assume that every added revision already has its
1440
# corresponding inventory, so we only check for parent inventories that
1441
# might be missing, rather than all inventories.
1442
parents = set(self.revisions._index.get_missing_parents())
1443
parents.discard(_mod_revision.NULL_REVISION)
1444
unstacked_inventories = self.inventories._index
1445
present_inventories = unstacked_inventories.get_parent_map(
1446
key[-1:] for key in parents)
1447
parents.difference_update(present_inventories)
1448
if len(parents) == 0:
1449
# No missing parent inventories.
1451
if not check_for_missing_texts:
1452
return set(('inventories', rev_id) for (rev_id,) in parents)
1453
# Ok, now we have a list of missing inventories. But these only matter
1454
# if the inventories that reference them are missing some texts they
1455
# appear to introduce.
1456
# XXX: Texts referenced by all added inventories need to be present,
1457
# but at the moment we're only checking for texts referenced by
1458
# inventories at the graph's edge.
1459
key_deps = self.revisions._index._key_dependencies
1460
key_deps.add_keys(present_inventories)
1461
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1462
file_ids = self.fileids_altered_by_revision_ids(referrers)
1463
missing_texts = set()
1464
for file_id, version_ids in file_ids.iteritems():
1465
missing_texts.update(
1466
(file_id, version_id) for version_id in version_ids)
1467
present_texts = self.texts.get_parent_map(missing_texts)
1468
missing_texts.difference_update(present_texts)
1469
if not missing_texts:
1470
# No texts are missing, so all revisions and their deltas are
1473
# Alternatively the text versions could be returned as the missing
1474
# keys, but this is likely to be less data.
1475
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1478
def refresh_data(self):
1479
"""Re-read any data needed to to synchronise with disk.
1481
This method is intended to be called after another repository instance
1482
(such as one used by a smart server) has inserted data into the
1483
repository. It may not be called during a write group, but may be
1484
called at any other time.
1486
if self.is_in_write_group():
1487
raise errors.InternalBzrError(
1488
"May not refresh_data while in a write group.")
1489
self._refresh_data()
1491
def resume_write_group(self, tokens):
1492
if not self.is_write_locked():
1493
raise errors.NotWriteLocked(self)
1494
if self._write_group:
1495
raise errors.BzrError('already in a write group')
1496
self._resume_write_group(tokens)
1497
# so we can detect unlock/relock - the write group is now entered.
1498
self._write_group = self.get_transaction()
1500
def _resume_write_group(self, tokens):
1501
raise errors.UnsuspendableWriteGroup(self)
1503
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
1505
"""Fetch the content required to construct revision_id from source.
1507
If revision_id is None and fetch_spec is None, then all content is
1510
fetch() may not be used when the repository is in a write group -
1511
either finish the current write group before using fetch, or use
1512
fetch before starting the write group.
1514
:param find_ghosts: Find and copy revisions in the source that are
1515
ghosts in the target (and not reachable directly by walking out to
1516
the first-present revision in target from revision_id).
1517
:param revision_id: If specified, all the content needed for this
1518
revision ID will be copied to the target. Fetch will determine for
1519
itself which content needs to be copied.
1520
:param fetch_spec: If specified, a SearchResult or
1521
PendingAncestryResult that describes which revisions to copy. This
1522
allows copying multiple heads at once. Mutually exclusive with
1525
if fetch_spec is not None and revision_id is not None:
1526
raise AssertionError(
1527
"fetch_spec and revision_id are mutually exclusive.")
1528
if self.is_in_write_group():
1529
raise errors.InternalBzrError(
1530
"May not fetch while in a write group.")
1531
# fast path same-url fetch operations
1532
if self.has_same_location(source) and fetch_spec is None:
1533
# check that last_revision is in 'from' and then return a
1535
if (revision_id is not None and
1536
not _mod_revision.is_null(revision_id)):
1537
self.get_revision(revision_id)
1539
# if there is no specific appropriate InterRepository, this will get
1540
# the InterRepository base class, which raises an
1541
# IncompatibleRepositories when asked to fetch.
1542
inter = InterRepository.get(source, self)
1543
return inter.fetch(revision_id=revision_id, pb=pb,
1544
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
1546
def create_bundle(self, target, base, fileobj, format=None):
1547
return serializer.write_bundle(self, target, base, fileobj, format)
1549
def get_commit_builder(self, branch, parents, config, timestamp=None,
1550
timezone=None, committer=None, revprops=None,
1552
"""Obtain a CommitBuilder for this repository.
1554
:param branch: Branch to commit to.
1555
:param parents: Revision ids of the parents of the new revision.
1556
:param config: Configuration to use.
1557
:param timestamp: Optional timestamp recorded for commit.
1558
:param timezone: Optional timezone for timestamp.
1559
:param committer: Optional committer to set for commit.
1560
:param revprops: Optional dictionary of revision properties.
1561
:param revision_id: Optional revision id.
1563
result = self._commit_builder_class(self, parents, config,
1564
timestamp, timezone, committer, revprops, revision_id)
1565
self.start_write_group()
1569
if (self.control_files._lock_count == 1 and
1570
self.control_files._lock_mode == 'w'):
1571
if self._write_group is not None:
1572
self.abort_write_group()
1573
self.control_files.unlock()
1574
raise errors.BzrError(
1575
'Must end write groups before releasing write locks.')
1576
self.control_files.unlock()
1577
if self.control_files._lock_count == 0:
1578
self._inventory_entry_cache.clear()
1579
for repo in self._fallback_repositories:
1583
def clone(self, a_bzrdir, revision_id=None):
1584
"""Clone this repository into a_bzrdir using the current format.
1586
Currently no check is made that the format of this repository and
1587
the bzrdir format are compatible. FIXME RBC 20060201.
1589
:return: The newly created destination repository.
1591
# TODO: deprecate after 0.16; cloning this with all its settings is
1592
# probably not very useful -- mbp 20070423
1593
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1594
self.copy_content_into(dest_repo, revision_id)
1597
def start_write_group(self):
1598
"""Start a write group in the repository.
1600
Write groups are used by repositories which do not have a 1:1 mapping
1601
between file ids and backend store to manage the insertion of data from
1602
both fetch and commit operations.
1604
A write lock is required around the start_write_group/commit_write_group
1605
for the support of lock-requiring repository formats.
1607
One can only insert data into a repository inside a write group.
1611
if not self.is_write_locked():
1612
raise errors.NotWriteLocked(self)
1613
if self._write_group:
1614
raise errors.BzrError('already in a write group')
1615
self._start_write_group()
1616
# so we can detect unlock/relock - the write group is now entered.
1617
self._write_group = self.get_transaction()
1619
def _start_write_group(self):
1620
"""Template method for per-repository write group startup.
1622
This is called before the write group is considered to be
1627
def sprout(self, to_bzrdir, revision_id=None):
1628
"""Create a descendent repository for new development.
1630
Unlike clone, this does not copy the settings of the repository.
1632
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1633
dest_repo.fetch(self, revision_id=revision_id)
1636
def _create_sprouting_repo(self, a_bzrdir, shared):
1637
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1638
# use target default format.
1639
dest_repo = a_bzrdir.create_repository()
1641
# Most control formats need the repository to be specifically
1642
# created, but on some old all-in-one formats it's not needed
1644
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1645
except errors.UninitializableFormat:
1646
dest_repo = a_bzrdir.open_repository()
1649
def _get_sink(self):
1650
"""Return a sink for streaming into this repository."""
1651
return StreamSink(self)
1653
def _get_source(self, to_format):
1654
"""Return a source for streaming from this repository."""
1655
return StreamSource(self, to_format)
1658
def has_revision(self, revision_id):
1659
"""True if this repository has a copy of the revision."""
1660
return revision_id in self.has_revisions((revision_id,))
1663
def has_revisions(self, revision_ids):
1664
"""Probe to find out the presence of multiple revisions.
1666
:param revision_ids: An iterable of revision_ids.
1667
:return: A set of the revision_ids that were present.
1669
parent_map = self.revisions.get_parent_map(
1670
[(rev_id,) for rev_id in revision_ids])
1672
if _mod_revision.NULL_REVISION in revision_ids:
1673
result.add(_mod_revision.NULL_REVISION)
1674
result.update([key[0] for key in parent_map])
1678
def get_revision(self, revision_id):
1679
"""Return the Revision object for a named revision."""
1680
return self.get_revisions([revision_id])[0]
1683
def get_revision_reconcile(self, revision_id):
1684
"""'reconcile' helper routine that allows access to a revision always.
1686
This variant of get_revision does not cross check the weave graph
1687
against the revision one as get_revision does: but it should only
1688
be used by reconcile, or reconcile-alike commands that are correcting
1689
or testing the revision graph.
1691
return self._get_revisions([revision_id])[0]
1694
def get_revisions(self, revision_ids):
1695
"""Get many revisions at once."""
1696
return self._get_revisions(revision_ids)
1699
def _get_revisions(self, revision_ids):
1700
"""Core work logic to get many revisions without sanity checks."""
1701
for rev_id in revision_ids:
1702
if not rev_id or not isinstance(rev_id, basestring):
1703
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1704
keys = [(key,) for key in revision_ids]
1705
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1707
for record in stream:
1708
if record.storage_kind == 'absent':
1709
raise errors.NoSuchRevision(self, record.key[0])
1710
text = record.get_bytes_as('fulltext')
1711
rev = self._serializer.read_revision_from_string(text)
1712
revs[record.key[0]] = rev
1713
return [revs[revid] for revid in revision_ids]
1716
def get_revision_xml(self, revision_id):
1717
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1718
# would have already do it.
1719
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1720
# TODO: this can't just be replaced by:
1721
# return self._serializer.write_revision_to_string(
1722
# self.get_revision(revision_id))
1723
# as cStringIO preservers the encoding unlike write_revision_to_string
1724
# or some other call down the path.
1725
rev = self.get_revision(revision_id)
1726
rev_tmp = cStringIO.StringIO()
1727
# the current serializer..
1728
self._serializer.write_revision(rev, rev_tmp)
1730
return rev_tmp.getvalue()
1732
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1733
"""Produce a generator of revision deltas.
1735
Note that the input is a sequence of REVISIONS, not revision_ids.
1736
Trees will be held in memory until the generator exits.
1737
Each delta is relative to the revision's lefthand predecessor.
1739
:param specific_fileids: if not None, the result is filtered
1740
so that only those file-ids, their parents and their
1741
children are included.
1743
# Get the revision-ids of interest
1744
required_trees = set()
1745
for revision in revisions:
1746
required_trees.add(revision.revision_id)
1747
required_trees.update(revision.parent_ids[:1])
1749
# Get the matching filtered trees. Note that it's more
1750
# efficient to pass filtered trees to changes_from() rather
1751
# than doing the filtering afterwards. changes_from() could
1752
# arguably do the filtering itself but it's path-based, not
1753
# file-id based, so filtering before or afterwards is
1755
if specific_fileids is None:
1756
trees = dict((t.get_revision_id(), t) for
1757
t in self.revision_trees(required_trees))
1759
trees = dict((t.get_revision_id(), t) for
1760
t in self._filtered_revision_trees(required_trees,
1763
# Calculate the deltas
1764
for revision in revisions:
1765
if not revision.parent_ids:
1766
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1768
old_tree = trees[revision.parent_ids[0]]
1769
yield trees[revision.revision_id].changes_from(old_tree)
1772
def get_revision_delta(self, revision_id, specific_fileids=None):
1773
"""Return the delta for one revision.
1775
The delta is relative to the left-hand predecessor of the
1778
:param specific_fileids: if not None, the result is filtered
1779
so that only those file-ids, their parents and their
1780
children are included.
1782
r = self.get_revision(revision_id)
1783
return list(self.get_deltas_for_revisions([r],
1784
specific_fileids=specific_fileids))[0]
1787
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1788
signature = gpg_strategy.sign(plaintext)
1789
self.add_signature_text(revision_id, signature)
1792
def add_signature_text(self, revision_id, signature):
1793
self.signatures.add_lines((revision_id,), (),
1794
osutils.split_lines(signature))
1796
def find_text_key_references(self):
1797
"""Find the text key references within the repository.
1799
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1800
to whether they were referred to by the inventory of the
1801
revision_id that they contain. The inventory texts from all present
1802
revision ids are assessed to generate this report.
1804
revision_keys = self.revisions.keys()
1805
w = self.inventories
1806
pb = ui.ui_factory.nested_progress_bar()
1808
return self._find_text_key_references_from_xml_inventory_lines(
1809
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1813
def _find_text_key_references_from_xml_inventory_lines(self,
1815
"""Core routine for extracting references to texts from inventories.
1817
This performs the translation of xml lines to revision ids.
1819
:param line_iterator: An iterator of lines, origin_version_id
1820
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1821
to whether they were referred to by the inventory of the
1822
revision_id that they contain. Note that if that revision_id was
1823
not part of the line_iterator's output then False will be given -
1824
even though it may actually refer to that key.
1826
if not self._serializer.support_altered_by_hack:
1827
raise AssertionError(
1828
"_find_text_key_references_from_xml_inventory_lines only "
1829
"supported for branches which store inventory as unnested xml"
1830
", not on %r" % self)
1833
# this code needs to read every new line in every inventory for the
1834
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1835
# not present in one of those inventories is unnecessary but not
1836
# harmful because we are filtering by the revision id marker in the
1837
# inventory lines : we only select file ids altered in one of those
1838
# revisions. We don't need to see all lines in the inventory because
1839
# only those added in an inventory in rev X can contain a revision=X
1841
unescape_revid_cache = {}
1842
unescape_fileid_cache = {}
1844
# jam 20061218 In a big fetch, this handles hundreds of thousands
1845
# of lines, so it has had a lot of inlining and optimizing done.
1846
# Sorry that it is a little bit messy.
1847
# Move several functions to be local variables, since this is a long
1849
search = self._file_ids_altered_regex.search
1850
unescape = _unescape_xml
1851
setdefault = result.setdefault
1852
for line, line_key in line_iterator:
1853
match = search(line)
1856
# One call to match.group() returning multiple items is quite a
1857
# bit faster than 2 calls to match.group() each returning 1
1858
file_id, revision_id = match.group('file_id', 'revision_id')
1860
# Inlining the cache lookups helps a lot when you make 170,000
1861
# lines and 350k ids, versus 8.4 unique ids.
1862
# Using a cache helps in 2 ways:
1863
# 1) Avoids unnecessary decoding calls
1864
# 2) Re-uses cached strings, which helps in future set and
1866
# (2) is enough that removing encoding entirely along with
1867
# the cache (so we are using plain strings) results in no
1868
# performance improvement.
1870
revision_id = unescape_revid_cache[revision_id]
1872
unescaped = unescape(revision_id)
1873
unescape_revid_cache[revision_id] = unescaped
1874
revision_id = unescaped
1876
# Note that unconditionally unescaping means that we deserialise
1877
# every fileid, which for general 'pull' is not great, but we don't
1878
# really want to have some many fulltexts that this matters anyway.
1881
file_id = unescape_fileid_cache[file_id]
1883
unescaped = unescape(file_id)
1884
unescape_fileid_cache[file_id] = unescaped
1887
key = (file_id, revision_id)
1888
setdefault(key, False)
1889
if revision_id == line_key[-1]:
1893
def _inventory_xml_lines_for_keys(self, keys):
1894
"""Get a line iterator of the sort needed for findind references.
1896
Not relevant for non-xml inventory repositories.
1898
Ghosts in revision_keys are ignored.
1900
:param revision_keys: The revision keys for the inventories to inspect.
1901
:return: An iterator over (inventory line, revid) for the fulltexts of
1902
all of the xml inventories specified by revision_keys.
1904
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1905
for record in stream:
1906
if record.storage_kind != 'absent':
1907
chunks = record.get_bytes_as('chunked')
1908
revid = record.key[-1]
1909
lines = osutils.chunks_to_lines(chunks)
1913
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1915
"""Helper routine for fileids_altered_by_revision_ids.
1917
This performs the translation of xml lines to revision ids.
1919
:param line_iterator: An iterator of lines, origin_version_id
1920
:param revision_keys: The revision ids to filter for. This should be a
1921
set or other type which supports efficient __contains__ lookups, as
1922
the revision key from each parsed line will be looked up in the
1923
revision_keys filter.
1924
:return: a dictionary mapping altered file-ids to an iterable of
1925
revision_ids. Each altered file-ids has the exact revision_ids that
1926
altered it listed explicitly.
1928
seen = set(self._find_text_key_references_from_xml_inventory_lines(
1929
line_iterator).iterkeys())
1930
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
1931
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
1932
self._inventory_xml_lines_for_keys(parent_keys)))
1933
new_keys = seen - parent_seen
1935
setdefault = result.setdefault
1936
for key in new_keys:
1937
setdefault(key[0], set()).add(key[-1])
1940
def _find_parent_ids_of_revisions(self, revision_ids):
1941
"""Find all parent ids that are mentioned in the revision graph.
1943
:return: set of revisions that are parents of revision_ids which are
1944
not part of revision_ids themselves
1946
parent_map = self.get_parent_map(revision_ids)
1948
map(parent_ids.update, parent_map.itervalues())
1949
parent_ids.difference_update(revision_ids)
1950
parent_ids.discard(_mod_revision.NULL_REVISION)
1953
def _find_parent_keys_of_revisions(self, revision_keys):
1954
"""Similar to _find_parent_ids_of_revisions, but used with keys.
1956
:param revision_keys: An iterable of revision_keys.
1957
:return: The parents of all revision_keys that are not already in
1960
parent_map = self.revisions.get_parent_map(revision_keys)
1962
map(parent_keys.update, parent_map.itervalues())
1963
parent_keys.difference_update(revision_keys)
1964
parent_keys.discard(_mod_revision.NULL_REVISION)
1967
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1968
"""Find the file ids and versions affected by revisions.
1970
:param revisions: an iterable containing revision ids.
1971
:param _inv_weave: The inventory weave from this repository or None.
1972
If None, the inventory weave will be opened automatically.
1973
:return: a dictionary mapping altered file-ids to an iterable of
1974
revision_ids. Each altered file-ids has the exact revision_ids that
1975
altered it listed explicitly.
1977
selected_keys = set((revid,) for revid in revision_ids)
1978
w = _inv_weave or self.inventories
1979
pb = ui.ui_factory.nested_progress_bar()
1981
return self._find_file_ids_from_xml_inventory_lines(
1982
w.iter_lines_added_or_present_in_keys(
1983
selected_keys, pb=pb),
1988
def iter_files_bytes(self, desired_files):
1989
"""Iterate through file versions.
1991
Files will not necessarily be returned in the order they occur in
1992
desired_files. No specific order is guaranteed.
1994
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1995
value supplied by the caller as part of desired_files. It should
1996
uniquely identify the file version in the caller's context. (Examples:
1997
an index number or a TreeTransform trans_id.)
1999
bytes_iterator is an iterable of bytestrings for the file. The
2000
kind of iterable and length of the bytestrings are unspecified, but for
2001
this implementation, it is a list of bytes produced by
2002
VersionedFile.get_record_stream().
2004
:param desired_files: a list of (file_id, revision_id, identifier)
2008
for file_id, revision_id, callable_data in desired_files:
2009
text_keys[(file_id, revision_id)] = callable_data
2010
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2011
if record.storage_kind == 'absent':
2012
raise errors.RevisionNotPresent(record.key, self)
2013
yield text_keys[record.key], record.get_bytes_as('chunked')
2015
def _generate_text_key_index(self, text_key_references=None,
2017
"""Generate a new text key index for the repository.
2019
This is an expensive function that will take considerable time to run.
2021
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2022
list of parents, also text keys. When a given key has no parents,
2023
the parents list will be [NULL_REVISION].
2025
# All revisions, to find inventory parents.
2026
if ancestors is None:
2027
graph = self.get_graph()
2028
ancestors = graph.get_parent_map(self.all_revision_ids())
2029
if text_key_references is None:
2030
text_key_references = self.find_text_key_references()
2031
pb = ui.ui_factory.nested_progress_bar()
2033
return self._do_generate_text_key_index(ancestors,
2034
text_key_references, pb)
2038
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2039
"""Helper for _generate_text_key_index to avoid deep nesting."""
2040
revision_order = tsort.topo_sort(ancestors)
2041
invalid_keys = set()
2043
for revision_id in revision_order:
2044
revision_keys[revision_id] = set()
2045
text_count = len(text_key_references)
2046
# a cache of the text keys to allow reuse; costs a dict of all the
2047
# keys, but saves a 2-tuple for every child of a given key.
2049
for text_key, valid in text_key_references.iteritems():
2051
invalid_keys.add(text_key)
2053
revision_keys[text_key[1]].add(text_key)
2054
text_key_cache[text_key] = text_key
2055
del text_key_references
2057
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2058
NULL_REVISION = _mod_revision.NULL_REVISION
2059
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2060
# too small for large or very branchy trees. However, for 55K path
2061
# trees, it would be easy to use too much memory trivially. Ideally we
2062
# could gauge this by looking at available real memory etc, but this is
2063
# always a tricky proposition.
2064
inventory_cache = lru_cache.LRUCache(10)
2065
batch_size = 10 # should be ~150MB on a 55K path tree
2066
batch_count = len(revision_order) / batch_size + 1
2068
pb.update("Calculating text parents", processed_texts, text_count)
2069
for offset in xrange(batch_count):
2070
to_query = revision_order[offset * batch_size:(offset + 1) *
2074
for rev_tree in self.revision_trees(to_query):
2075
revision_id = rev_tree.get_revision_id()
2076
parent_ids = ancestors[revision_id]
2077
for text_key in revision_keys[revision_id]:
2078
pb.update("Calculating text parents", processed_texts)
2079
processed_texts += 1
2080
candidate_parents = []
2081
for parent_id in parent_ids:
2082
parent_text_key = (text_key[0], parent_id)
2084
check_parent = parent_text_key not in \
2085
revision_keys[parent_id]
2087
# the parent parent_id is a ghost:
2088
check_parent = False
2089
# truncate the derived graph against this ghost.
2090
parent_text_key = None
2092
# look at the parent commit details inventories to
2093
# determine possible candidates in the per file graph.
2096
inv = inventory_cache[parent_id]
2098
inv = self.revision_tree(parent_id).inventory
2099
inventory_cache[parent_id] = inv
2101
parent_entry = inv[text_key[0]]
2102
except (KeyError, errors.NoSuchId):
2104
if parent_entry is not None:
2106
text_key[0], parent_entry.revision)
2108
parent_text_key = None
2109
if parent_text_key is not None:
2110
candidate_parents.append(
2111
text_key_cache[parent_text_key])
2112
parent_heads = text_graph.heads(candidate_parents)
2113
new_parents = list(parent_heads)
2114
new_parents.sort(key=lambda x:candidate_parents.index(x))
2115
if new_parents == []:
2116
new_parents = [NULL_REVISION]
2117
text_index[text_key] = new_parents
2119
for text_key in invalid_keys:
2120
text_index[text_key] = [NULL_REVISION]
2123
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2124
"""Get an iterable listing the keys of all the data introduced by a set
2127
The keys will be ordered so that the corresponding items can be safely
2128
fetched and inserted in that order.
2130
:returns: An iterable producing tuples of (knit-kind, file-id,
2131
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2132
'revisions'. file-id is None unless knit-kind is 'file'.
2134
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2137
for result in self._find_non_file_keys_to_fetch(revision_ids):
2140
def _find_file_keys_to_fetch(self, revision_ids, pb):
2141
# XXX: it's a bit weird to control the inventory weave caching in this
2142
# generator. Ideally the caching would be done in fetch.py I think. Or
2143
# maybe this generator should explicitly have the contract that it
2144
# should not be iterated until the previously yielded item has been
2146
inv_w = self.inventories
2148
# file ids that changed
2149
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2151
num_file_ids = len(file_ids)
2152
for file_id, altered_versions in file_ids.iteritems():
2154
pb.update("fetch texts", count, num_file_ids)
2156
yield ("file", file_id, altered_versions)
2158
def _find_non_file_keys_to_fetch(self, revision_ids):
2160
yield ("inventory", None, revision_ids)
2163
# XXX: Note ATM no callers actually pay attention to this return
2164
# instead they just use the list of revision ids and ignore
2165
# missing sigs. Consider removing this work entirely
2166
revisions_with_signatures = set(self.signatures.get_parent_map(
2167
[(r,) for r in revision_ids]))
2168
revisions_with_signatures = set(
2169
[r for (r,) in revisions_with_signatures])
2170
revisions_with_signatures.intersection_update(revision_ids)
2171
yield ("signatures", None, revisions_with_signatures)
2174
yield ("revisions", None, revision_ids)
2177
def get_inventory(self, revision_id):
2178
"""Get Inventory object by revision id."""
2179
return self.iter_inventories([revision_id]).next()
2181
def iter_inventories(self, revision_ids):
2182
"""Get many inventories by revision_ids.
2184
This will buffer some or all of the texts used in constructing the
2185
inventories in memory, but will only parse a single inventory at a
2188
:param revision_ids: The expected revision ids of the inventories.
2189
:return: An iterator of inventories.
2191
if ((None in revision_ids)
2192
or (_mod_revision.NULL_REVISION in revision_ids)):
2193
import pdb; pdb.set_trace()
2194
raise ValueError('cannot get null revision inventory')
2195
return self._iter_inventories(revision_ids)
2197
def _iter_inventories(self, revision_ids):
2198
"""single-document based inventory iteration."""
2199
for text, revision_id in self._iter_inventory_xmls(revision_ids):
2200
yield self.deserialise_inventory(revision_id, text)
2202
def _iter_inventory_xmls(self, revision_ids):
2203
keys = [(revision_id,) for revision_id in revision_ids]
2204
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2206
for record in stream:
2207
if record.storage_kind != 'absent':
2208
text_chunks[record.key] = record.get_bytes_as('chunked')
2210
raise errors.NoSuchRevision(self, record.key)
2212
chunks = text_chunks.pop(key)
2213
yield ''.join(chunks), key[-1]
2215
def deserialise_inventory(self, revision_id, xml):
2216
"""Transform the xml into an inventory object.
2218
:param revision_id: The expected revision id of the inventory.
2219
:param xml: A serialised inventory.
2221
result = self._serializer.read_inventory_from_string(xml, revision_id,
2222
entry_cache=self._inventory_entry_cache)
2223
if result.revision_id != revision_id:
2224
raise AssertionError('revision id mismatch %s != %s' % (
2225
result.revision_id, revision_id))
2228
def serialise_inventory(self, inv):
2229
return self._serializer.write_inventory_to_string(inv)
2231
def _serialise_inventory_to_lines(self, inv):
2232
return self._serializer.write_inventory_to_lines(inv)
2234
def get_serializer_format(self):
2235
return self._serializer.format_num
2238
def get_inventory_xml(self, revision_id):
2239
"""Get inventory XML as a file object."""
2240
texts = self._iter_inventory_xmls([revision_id])
2242
text, revision_id = texts.next()
2243
except StopIteration:
2244
raise errors.HistoryMissing(self, 'inventory', revision_id)
2248
def get_inventory_sha1(self, revision_id):
2249
"""Return the sha1 hash of the inventory entry
2251
return self.get_revision(revision_id).inventory_sha1
2253
def get_rev_id_for_revno(self, revno, known_pair):
2254
"""Return the revision id of a revno, given a later (revno, revid)
2255
pair in the same history.
2257
:return: if found (True, revid). If the available history ran out
2258
before reaching the revno, then this returns
2259
(False, (closest_revno, closest_revid)).
2261
known_revno, known_revid = known_pair
2262
partial_history = [known_revid]
2263
distance_from_known = known_revno - revno
2264
if distance_from_known < 0:
2266
'requested revno (%d) is later than given known revno (%d)'
2267
% (revno, known_revno))
2270
self, partial_history, stop_index=distance_from_known)
2271
except errors.RevisionNotPresent, err:
2272
if err.revision_id == known_revid:
2273
# The start revision (known_revid) wasn't found.
2275
# This is a stacked repository with no fallbacks, or a there's a
2276
# left-hand ghost. Either way, even though the revision named in
2277
# the error isn't in this repo, we know it's the next step in this
2278
# left-hand history.
2279
partial_history.append(err.revision_id)
2280
if len(partial_history) <= distance_from_known:
2281
# Didn't find enough history to get a revid for the revno.
2282
earliest_revno = known_revno - len(partial_history) + 1
2283
return (False, (earliest_revno, partial_history[-1]))
2284
if len(partial_history) - 1 > distance_from_known:
2285
raise AssertionError('_iter_for_revno returned too much history')
2286
return (True, partial_history[-1])
2288
def iter_reverse_revision_history(self, revision_id):
2289
"""Iterate backwards through revision ids in the lefthand history
2291
:param revision_id: The revision id to start with. All its lefthand
2292
ancestors will be traversed.
2294
graph = self.get_graph()
2295
next_id = revision_id
2297
if next_id in (None, _mod_revision.NULL_REVISION):
2300
parents = graph.get_parent_map([next_id])[next_id]
2302
raise errors.RevisionNotPresent(next_id, self)
2304
if len(parents) == 0:
2307
next_id = parents[0]
2310
def get_revision_inventory(self, revision_id):
2311
"""Return inventory of a past revision."""
2312
# TODO: Unify this with get_inventory()
2313
# bzr 0.0.6 and later imposes the constraint that the inventory_id
2314
# must be the same as its revision, so this is trivial.
2315
if revision_id is None:
2316
# This does not make sense: if there is no revision,
2317
# then it is the current tree inventory surely ?!
2318
# and thus get_root_id() is something that looks at the last
2319
# commit on the branch, and the get_root_id is an inventory check.
2320
raise NotImplementedError
2321
# return Inventory(self.get_root_id())
2323
return self.get_inventory(revision_id)
2325
def is_shared(self):
2326
"""Return True if this repository is flagged as a shared repository."""
2327
raise NotImplementedError(self.is_shared)
2330
def reconcile(self, other=None, thorough=False):
2331
"""Reconcile this repository."""
2332
from bzrlib.reconcile import RepoReconciler
2333
reconciler = RepoReconciler(self, thorough=thorough)
2334
reconciler.reconcile()
2337
def _refresh_data(self):
2338
"""Helper called from lock_* to ensure coherency with disk.
2340
The default implementation does nothing; it is however possible
2341
for repositories to maintain loaded indices across multiple locks
2342
by checking inside their implementation of this method to see
2343
whether their indices are still valid. This depends of course on
2344
the disk format being validatable in this manner. This method is
2345
also called by the refresh_data() public interface to cause a refresh
2346
to occur while in a write lock so that data inserted by a smart server
2347
push operation is visible on the client's instance of the physical
2352
def revision_tree(self, revision_id):
2353
"""Return Tree for a revision on this branch.
2355
`revision_id` may be NULL_REVISION for the empty tree revision.
2357
revision_id = _mod_revision.ensure_null(revision_id)
2358
# TODO: refactor this to use an existing revision object
2359
# so we don't need to read it in twice.
2360
if revision_id == _mod_revision.NULL_REVISION:
2361
return RevisionTree(self, Inventory(root_id=None),
2362
_mod_revision.NULL_REVISION)
2364
inv = self.get_revision_inventory(revision_id)
2365
return RevisionTree(self, inv, revision_id)
2367
def revision_trees(self, revision_ids):
2368
"""Return Trees for revisions in this repository.
2370
:param revision_ids: a sequence of revision-ids;
2371
a revision-id may not be None or 'null:'
2373
inventories = self.iter_inventories(revision_ids)
2374
for inv in inventories:
2375
yield RevisionTree(self, inv, inv.revision_id)
2377
def _filtered_revision_trees(self, revision_ids, file_ids):
2378
"""Return Tree for a revision on this branch with only some files.
2380
:param revision_ids: a sequence of revision-ids;
2381
a revision-id may not be None or 'null:'
2382
:param file_ids: if not None, the result is filtered
2383
so that only those file-ids, their parents and their
2384
children are included.
2386
inventories = self.iter_inventories(revision_ids)
2387
for inv in inventories:
2388
# Should we introduce a FilteredRevisionTree class rather
2389
# than pre-filter the inventory here?
2390
filtered_inv = inv.filter(file_ids)
2391
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
2394
def get_ancestry(self, revision_id, topo_sorted=True):
2395
"""Return a list of revision-ids integrated by a revision.
2397
The first element of the list is always None, indicating the origin
2398
revision. This might change when we have history horizons, or
2399
perhaps we should have a new API.
2401
This is topologically sorted.
2403
if _mod_revision.is_null(revision_id):
2405
if not self.has_revision(revision_id):
2406
raise errors.NoSuchRevision(self, revision_id)
2407
graph = self.get_graph()
2409
search = graph._make_breadth_first_searcher([revision_id])
2412
found, ghosts = search.next_with_ghosts()
2413
except StopIteration:
2416
if _mod_revision.NULL_REVISION in keys:
2417
keys.remove(_mod_revision.NULL_REVISION)
2419
parent_map = graph.get_parent_map(keys)
2420
keys = tsort.topo_sort(parent_map)
2421
return [None] + list(keys)
2423
def pack(self, hint=None):
2424
"""Compress the data within the repository.
2426
This operation only makes sense for some repository types. For other
2427
types it should be a no-op that just returns.
2429
This stub method does not require a lock, but subclasses should use
2430
@needs_write_lock as this is a long running call its reasonable to
2431
implicitly lock for the user.
2433
:param hint: If not supplied, the whole repository is packed.
2434
If supplied, the repository may use the hint parameter as a
2435
hint for the parts of the repository to pack. A hint can be
2436
obtained from the result of commit_write_group(). Out of
2437
date hints are simply ignored, because concurrent operations
2438
can obsolete them rapidly.
2441
def get_transaction(self):
2442
return self.control_files.get_transaction()
2444
def get_parent_map(self, revision_ids):
2445
"""See graph.StackedParentsProvider.get_parent_map"""
2446
# revisions index works in keys; this just works in revisions
2447
# therefore wrap and unwrap
2450
for revision_id in revision_ids:
2451
if revision_id == _mod_revision.NULL_REVISION:
2452
result[revision_id] = ()
2453
elif revision_id is None:
2454
raise ValueError('get_parent_map(None) is not valid')
2456
query_keys.append((revision_id ,))
2457
for ((revision_id,), parent_keys) in \
2458
self.revisions.get_parent_map(query_keys).iteritems():
2460
result[revision_id] = tuple(parent_revid
2461
for (parent_revid,) in parent_keys)
2463
result[revision_id] = (_mod_revision.NULL_REVISION,)
2466
def _make_parents_provider(self):
2469
def get_graph(self, other_repository=None):
2470
"""Return the graph walker for this repository format"""
2471
parents_provider = self._make_parents_provider()
2472
if (other_repository is not None and
2473
not self.has_same_location(other_repository)):
2474
parents_provider = graph.StackedParentsProvider(
2475
[parents_provider, other_repository._make_parents_provider()])
2476
return graph.Graph(parents_provider)
2478
def _get_versioned_file_checker(self, text_key_references=None):
2479
"""Return an object suitable for checking versioned files.
2481
:param text_key_references: if non-None, an already built
2482
dictionary mapping text keys ((fileid, revision_id) tuples)
2483
to whether they were referred to by the inventory of the
2484
revision_id that they contain. If None, this will be
2487
return _VersionedFileChecker(self,
2488
text_key_references=text_key_references)
2490
def revision_ids_to_search_result(self, result_set):
2491
"""Convert a set of revision ids to a graph SearchResult."""
2492
result_parents = set()
2493
for parents in self.get_graph().get_parent_map(
2494
result_set).itervalues():
2495
result_parents.update(parents)
2496
included_keys = result_set.intersection(result_parents)
2497
start_keys = result_set.difference(included_keys)
2498
exclude_keys = result_parents.difference(result_set)
2499
result = graph.SearchResult(start_keys, exclude_keys,
2500
len(result_set), result_set)
2504
def set_make_working_trees(self, new_value):
2505
"""Set the policy flag for making working trees when creating branches.
2507
This only applies to branches that use this repository.
2509
The default is 'True'.
2510
:param new_value: True to restore the default, False to disable making
2513
raise NotImplementedError(self.set_make_working_trees)
2515
def make_working_trees(self):
2516
"""Returns the policy for making working trees on new branches."""
2517
raise NotImplementedError(self.make_working_trees)
2520
def sign_revision(self, revision_id, gpg_strategy):
2521
plaintext = Testament.from_revision(self, revision_id).as_short_text()
2522
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2525
def has_signature_for_revision_id(self, revision_id):
2526
"""Query for a revision signature for revision_id in the repository."""
2527
if not self.has_revision(revision_id):
2528
raise errors.NoSuchRevision(self, revision_id)
2529
sig_present = (1 == len(
2530
self.signatures.get_parent_map([(revision_id,)])))
2534
def get_signature_text(self, revision_id):
2535
"""Return the text for a signature."""
2536
stream = self.signatures.get_record_stream([(revision_id,)],
2538
record = stream.next()
2539
if record.storage_kind == 'absent':
2540
raise errors.NoSuchRevision(self, revision_id)
2541
return record.get_bytes_as('fulltext')
2544
def check(self, revision_ids=None):
2545
"""Check consistency of all history of given revision_ids.
2547
Different repository implementations should override _check().
2549
:param revision_ids: A non-empty list of revision_ids whose ancestry
2550
will be checked. Typically the last revision_id of a branch.
2552
return self._check(revision_ids)
2554
def _check(self, revision_ids):
2555
result = check.Check(self)
2559
def _warn_if_deprecated(self):
2560
global _deprecation_warning_done
2561
if _deprecation_warning_done:
2563
_deprecation_warning_done = True
2564
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2565
% (self._format, self.bzrdir.transport.base))
2567
def supports_rich_root(self):
2568
return self._format.rich_root_data
2570
def _check_ascii_revisionid(self, revision_id, method):
2571
"""Private helper for ascii-only repositories."""
2572
# weave repositories refuse to store revisionids that are non-ascii.
2573
if revision_id is not None:
2574
# weaves require ascii revision ids.
2575
if isinstance(revision_id, unicode):
2577
revision_id.encode('ascii')
2578
except UnicodeEncodeError:
2579
raise errors.NonAsciiRevisionId(method, self)
2582
revision_id.decode('ascii')
2583
except UnicodeDecodeError:
2584
raise errors.NonAsciiRevisionId(method, self)
2586
def revision_graph_can_have_wrong_parents(self):
2587
"""Is it possible for this repository to have a revision graph with
2590
If True, then this repository must also implement
2591
_find_inconsistent_revision_parents so that check and reconcile can
2592
check for inconsistencies before proceeding with other checks that may
2593
depend on the revision index being consistent.
2595
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2598
# remove these delegates a while after bzr 0.15
2599
def __make_delegated(name, from_module):
2600
def _deprecated_repository_forwarder():
2601
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2602
% (name, from_module),
2605
m = __import__(from_module, globals(), locals(), [name])
2607
return getattr(m, name)
2608
except AttributeError:
2609
raise AttributeError('module %s has no name %s'
2611
globals()[name] = _deprecated_repository_forwarder
2614
'AllInOneRepository',
2615
'WeaveMetaDirRepository',
2616
'PreSplitOutRepositoryFormat',
2617
'RepositoryFormat4',
2618
'RepositoryFormat5',
2619
'RepositoryFormat6',
2620
'RepositoryFormat7',
2622
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2626
'RepositoryFormatKnit',
2627
'RepositoryFormatKnit1',
2629
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2632
def install_revision(repository, rev, revision_tree):
2633
"""Install all revision data into a repository."""
2634
install_revisions(repository, [(rev, revision_tree, None)])
2637
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2638
"""Install all revision data into a repository.
2640
Accepts an iterable of revision, tree, signature tuples. The signature
2643
repository.start_write_group()
2645
inventory_cache = lru_cache.LRUCache(10)
2646
for n, (revision, revision_tree, signature) in enumerate(iterable):
2647
_install_revision(repository, revision, revision_tree, signature,
2650
pb.update('Transferring revisions', n + 1, num_revisions)
2652
repository.abort_write_group()
2655
repository.commit_write_group()
2658
def _install_revision(repository, rev, revision_tree, signature,
2660
"""Install all revision data into a repository."""
2661
present_parents = []
2663
for p_id in rev.parent_ids:
2664
if repository.has_revision(p_id):
2665
present_parents.append(p_id)
2666
parent_trees[p_id] = repository.revision_tree(p_id)
2668
parent_trees[p_id] = repository.revision_tree(
2669
_mod_revision.NULL_REVISION)
2671
inv = revision_tree.inventory
2672
entries = inv.iter_entries()
2673
# backwards compatibility hack: skip the root id.
2674
if not repository.supports_rich_root():
2675
path, root = entries.next()
2676
if root.revision != rev.revision_id:
2677
raise errors.IncompatibleRevision(repr(repository))
2679
for path, ie in entries:
2680
text_keys[(ie.file_id, ie.revision)] = ie
2681
text_parent_map = repository.texts.get_parent_map(text_keys)
2682
missing_texts = set(text_keys) - set(text_parent_map)
2683
# Add the texts that are not already present
2684
for text_key in missing_texts:
2685
ie = text_keys[text_key]
2687
# FIXME: TODO: The following loop overlaps/duplicates that done by
2688
# commit to determine parents. There is a latent/real bug here where
2689
# the parents inserted are not those commit would do - in particular
2690
# they are not filtered by heads(). RBC, AB
2691
for revision, tree in parent_trees.iteritems():
2692
if ie.file_id not in tree:
2694
parent_id = tree.inventory[ie.file_id].revision
2695
if parent_id in text_parents:
2697
text_parents.append((ie.file_id, parent_id))
2698
lines = revision_tree.get_file(ie.file_id).readlines()
2699
repository.texts.add_lines(text_key, text_parents, lines)
2701
# install the inventory
2702
if repository._format._commit_inv_deltas and len(rev.parent_ids):
2703
# Cache this inventory
2704
inventory_cache[rev.revision_id] = inv
2706
basis_inv = inventory_cache[rev.parent_ids[0]]
2708
repository.add_inventory(rev.revision_id, inv, present_parents)
2710
delta = inv._make_delta(basis_inv)
2711
repository.add_inventory_by_delta(rev.parent_ids[0], delta,
2712
rev.revision_id, present_parents)
2714
repository.add_inventory(rev.revision_id, inv, present_parents)
2715
except errors.RevisionAlreadyPresent:
2717
if signature is not None:
2718
repository.add_signature_text(rev.revision_id, signature)
2719
repository.add_revision(rev.revision_id, rev, inv)
2722
class MetaDirRepository(Repository):
2723
"""Repositories in the new meta-dir layout.
2725
:ivar _transport: Transport for access to repository control files,
2726
typically pointing to .bzr/repository.
2729
def __init__(self, _format, a_bzrdir, control_files):
2730
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2731
self._transport = control_files._transport
2733
def is_shared(self):
2734
"""Return True if this repository is flagged as a shared repository."""
2735
return self._transport.has('shared-storage')
2738
def set_make_working_trees(self, new_value):
2739
"""Set the policy flag for making working trees when creating branches.
2741
This only applies to branches that use this repository.
2743
The default is 'True'.
2744
:param new_value: True to restore the default, False to disable making
2749
self._transport.delete('no-working-trees')
2750
except errors.NoSuchFile:
2753
self._transport.put_bytes('no-working-trees', '',
2754
mode=self.bzrdir._get_file_mode())
2756
def make_working_trees(self):
2757
"""Returns the policy for making working trees on new branches."""
2758
return not self._transport.has('no-working-trees')
2761
class MetaDirVersionedFileRepository(MetaDirRepository):
2762
"""Repositories in a meta-dir, that work via versioned file objects."""
2764
def __init__(self, _format, a_bzrdir, control_files):
2765
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2769
network_format_registry = registry.FormatRegistry()
2770
"""Registry of formats indexed by their network name.
2772
The network name for a repository format is an identifier that can be used when
2773
referring to formats with smart server operations. See
2774
RepositoryFormat.network_name() for more detail.
2778
format_registry = registry.FormatRegistry(network_format_registry)
2779
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2781
This can contain either format instances themselves, or classes/factories that
2782
can be called to obtain one.
2786
#####################################################################
2787
# Repository Formats
2789
class RepositoryFormat(object):
2790
"""A repository format.
2792
Formats provide four things:
2793
* An initialization routine to construct repository data on disk.
2794
* a optional format string which is used when the BzrDir supports
2796
* an open routine which returns a Repository instance.
2797
* A network name for referring to the format in smart server RPC
2800
There is one and only one Format subclass for each on-disk format. But
2801
there can be one Repository subclass that is used for several different
2802
formats. The _format attribute on a Repository instance can be used to
2803
determine the disk format.
2805
Formats are placed in a registry by their format string for reference
2806
during opening. These should be subclasses of RepositoryFormat for
2809
Once a format is deprecated, just deprecate the initialize and open
2810
methods on the format class. Do not deprecate the object, as the
2811
object may be created even when a repository instance hasn't been
2814
Common instance attributes:
2815
_matchingbzrdir - the bzrdir format that the repository format was
2816
originally written to work with. This can be used if manually
2817
constructing a bzrdir and repository, or more commonly for test suite
2821
# Set to True or False in derived classes. True indicates that the format
2822
# supports ghosts gracefully.
2823
supports_ghosts = None
2824
# Can this repository be given external locations to lookup additional
2825
# data. Set to True or False in derived classes.
2826
supports_external_lookups = None
2827
# Does this format support CHK bytestring lookups. Set to True or False in
2829
supports_chks = None
2830
# Should commit add an inventory, or an inventory delta to the repository.
2831
_commit_inv_deltas = True
2832
# What order should fetch operations request streams in?
2833
# The default is unordered as that is the cheapest for an origin to
2835
_fetch_order = 'unordered'
2836
# Does this repository format use deltas that can be fetched as-deltas ?
2837
# (E.g. knits, where the knit deltas can be transplanted intact.
2838
# We default to False, which will ensure that enough data to get
2839
# a full text out of any fetch stream will be grabbed.
2840
_fetch_uses_deltas = False
2841
# Should fetch trigger a reconcile after the fetch? Only needed for
2842
# some repository formats that can suffer internal inconsistencies.
2843
_fetch_reconcile = False
2844
# Does this format have < O(tree_size) delta generation. Used to hint what
2845
# code path for commit, amongst other things.
2847
# Does doing a pack operation compress data? Useful for the pack UI command
2848
# (so if there is one pack, the operation can still proceed because it may
2849
# help), and for fetching when data won't have come from the same
2851
pack_compresses = False
2854
return "<%s>" % self.__class__.__name__
2856
def __eq__(self, other):
2857
# format objects are generally stateless
2858
return isinstance(other, self.__class__)
2860
def __ne__(self, other):
2861
return not self == other
2864
def find_format(klass, a_bzrdir):
2865
"""Return the format for the repository object in a_bzrdir.
2867
This is used by bzr native formats that have a "format" file in
2868
the repository. Other methods may be used by different types of
2872
transport = a_bzrdir.get_repository_transport(None)
2873
format_string = transport.get("format").read()
2874
return format_registry.get(format_string)
2875
except errors.NoSuchFile:
2876
raise errors.NoRepositoryPresent(a_bzrdir)
2878
raise errors.UnknownFormatError(format=format_string,
2882
def register_format(klass, format):
2883
format_registry.register(format.get_format_string(), format)
2886
def unregister_format(klass, format):
2887
format_registry.remove(format.get_format_string())
2890
def get_default_format(klass):
2891
"""Return the current default format."""
2892
from bzrlib import bzrdir
2893
return bzrdir.format_registry.make_bzrdir('default').repository_format
2895
def get_format_string(self):
2896
"""Return the ASCII format string that identifies this format.
2898
Note that in pre format ?? repositories the format string is
2899
not permitted nor written to disk.
2901
raise NotImplementedError(self.get_format_string)
2903
def get_format_description(self):
2904
"""Return the short description for this format."""
2905
raise NotImplementedError(self.get_format_description)
2907
# TODO: this shouldn't be in the base class, it's specific to things that
2908
# use weaves or knits -- mbp 20070207
2909
def _get_versioned_file_store(self,
2914
versionedfile_class=None,
2915
versionedfile_kwargs={},
2917
if versionedfile_class is None:
2918
versionedfile_class = self._versionedfile_class
2919
weave_transport = control_files._transport.clone(name)
2920
dir_mode = control_files._dir_mode
2921
file_mode = control_files._file_mode
2922
return VersionedFileStore(weave_transport, prefixed=prefixed,
2924
file_mode=file_mode,
2925
versionedfile_class=versionedfile_class,
2926
versionedfile_kwargs=versionedfile_kwargs,
2929
def initialize(self, a_bzrdir, shared=False):
2930
"""Initialize a repository of this format in a_bzrdir.
2932
:param a_bzrdir: The bzrdir to put the new repository in it.
2933
:param shared: The repository should be initialized as a sharable one.
2934
:returns: The new repository object.
2936
This may raise UninitializableFormat if shared repository are not
2937
compatible the a_bzrdir.
2939
raise NotImplementedError(self.initialize)
2941
def is_supported(self):
2942
"""Is this format supported?
2944
Supported formats must be initializable and openable.
2945
Unsupported formats may not support initialization or committing or
2946
some other features depending on the reason for not being supported.
2950
def network_name(self):
2951
"""A simple byte string uniquely identifying this format for RPC calls.
2953
MetaDir repository formats use their disk format string to identify the
2954
repository over the wire. All in one formats such as bzr < 0.8, and
2955
foreign formats like svn/git and hg should use some marker which is
2956
unique and immutable.
2958
raise NotImplementedError(self.network_name)
2960
def check_conversion_target(self, target_format):
2961
raise NotImplementedError(self.check_conversion_target)
2963
def open(self, a_bzrdir, _found=False):
2964
"""Return an instance of this format for the bzrdir a_bzrdir.
2966
_found is a private parameter, do not use it.
2968
raise NotImplementedError(self.open)
2971
class MetaDirRepositoryFormat(RepositoryFormat):
2972
"""Common base class for the new repositories using the metadir layout."""
2974
rich_root_data = False
2975
supports_tree_reference = False
2976
supports_external_lookups = False
2979
def _matchingbzrdir(self):
2980
matching = bzrdir.BzrDirMetaFormat1()
2981
matching.repository_format = self
2985
super(MetaDirRepositoryFormat, self).__init__()
2987
def _create_control_files(self, a_bzrdir):
2988
"""Create the required files and the initial control_files object."""
2989
# FIXME: RBC 20060125 don't peek under the covers
2990
# NB: no need to escape relative paths that are url safe.
2991
repository_transport = a_bzrdir.get_repository_transport(self)
2992
control_files = lockable_files.LockableFiles(repository_transport,
2993
'lock', lockdir.LockDir)
2994
control_files.create_lock()
2995
return control_files
2997
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2998
"""Upload the initial blank content."""
2999
control_files = self._create_control_files(a_bzrdir)
3000
control_files.lock_write()
3001
transport = control_files._transport
3003
utf8_files += [('shared-storage', '')]
3005
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
3006
for (filename, content_stream) in files:
3007
transport.put_file(filename, content_stream,
3008
mode=a_bzrdir._get_file_mode())
3009
for (filename, content_bytes) in utf8_files:
3010
transport.put_bytes_non_atomic(filename, content_bytes,
3011
mode=a_bzrdir._get_file_mode())
3013
control_files.unlock()
3015
def network_name(self):
3016
"""Metadir formats have matching disk and network format strings."""
3017
return self.get_format_string()
3020
# Pre-0.8 formats that don't have a disk format string (because they are
3021
# versioned by the matching control directory). We use the control directories
3022
# disk format string as a key for the network_name because they meet the
3023
# constraints (simple string, unique, immutable).
3024
network_format_registry.register_lazy(
3025
"Bazaar-NG branch, format 5\n",
3026
'bzrlib.repofmt.weaverepo',
3027
'RepositoryFormat5',
3029
network_format_registry.register_lazy(
3030
"Bazaar-NG branch, format 6\n",
3031
'bzrlib.repofmt.weaverepo',
3032
'RepositoryFormat6',
3035
# formats which have no format string are not discoverable or independently
3036
# creatable on disk, so are not registered in format_registry. They're
3037
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
3038
# needed, it's constructed directly by the BzrDir. Non-native formats where
3039
# the repository is not separately opened are similar.
3041
format_registry.register_lazy(
3042
'Bazaar-NG Repository format 7',
3043
'bzrlib.repofmt.weaverepo',
3047
format_registry.register_lazy(
3048
'Bazaar-NG Knit Repository Format 1',
3049
'bzrlib.repofmt.knitrepo',
3050
'RepositoryFormatKnit1',
3053
format_registry.register_lazy(
3054
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
3055
'bzrlib.repofmt.knitrepo',
3056
'RepositoryFormatKnit3',
3059
format_registry.register_lazy(
3060
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
3061
'bzrlib.repofmt.knitrepo',
3062
'RepositoryFormatKnit4',
3065
# Pack-based formats. There is one format for pre-subtrees, and one for
3066
# post-subtrees to allow ease of testing.
3067
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
3068
format_registry.register_lazy(
3069
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
3070
'bzrlib.repofmt.pack_repo',
3071
'RepositoryFormatKnitPack1',
3073
format_registry.register_lazy(
3074
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
3075
'bzrlib.repofmt.pack_repo',
3076
'RepositoryFormatKnitPack3',
3078
format_registry.register_lazy(
3079
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
3080
'bzrlib.repofmt.pack_repo',
3081
'RepositoryFormatKnitPack4',
3083
format_registry.register_lazy(
3084
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
3085
'bzrlib.repofmt.pack_repo',
3086
'RepositoryFormatKnitPack5',
3088
format_registry.register_lazy(
3089
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
3090
'bzrlib.repofmt.pack_repo',
3091
'RepositoryFormatKnitPack5RichRoot',
3093
format_registry.register_lazy(
3094
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
3095
'bzrlib.repofmt.pack_repo',
3096
'RepositoryFormatKnitPack5RichRootBroken',
3098
format_registry.register_lazy(
3099
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
3100
'bzrlib.repofmt.pack_repo',
3101
'RepositoryFormatKnitPack6',
3103
format_registry.register_lazy(
3104
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
3105
'bzrlib.repofmt.pack_repo',
3106
'RepositoryFormatKnitPack6RichRoot',
3109
# Development formats.
3110
# Obsolete but kept pending a CHK based subtree format.
3111
format_registry.register_lazy(
3112
("Bazaar development format 2 with subtree support "
3113
"(needs bzr.dev from before 1.8)\n"),
3114
'bzrlib.repofmt.pack_repo',
3115
'RepositoryFormatPackDevelopment2Subtree',
3118
# 1.14->1.16 go below here
3119
format_registry.register_lazy(
3120
'Bazaar development format - group compression and chk inventory'
3121
' (needs bzr.dev from 1.14)\n',
3122
'bzrlib.repofmt.groupcompress_repo',
3123
'RepositoryFormatCHK1',
3126
format_registry.register_lazy(
3127
'Bazaar development format - chk repository with bencode revision '
3128
'serialization (needs bzr.dev from 1.16)\n',
3129
'bzrlib.repofmt.groupcompress_repo',
3130
'RepositoryFormatCHK2',
3132
format_registry.register_lazy(
3133
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3134
'bzrlib.repofmt.groupcompress_repo',
3135
'RepositoryFormat2a',
3139
class InterRepository(InterObject):
3140
"""This class represents operations taking place between two repositories.
3142
Its instances have methods like copy_content and fetch, and contain
3143
references to the source and target repositories these operations can be
3146
Often we will provide convenience methods on 'repository' which carry out
3147
operations with another repository - they will always forward to
3148
InterRepository.get(other).method_name(parameters).
3151
_walk_to_common_revisions_batch_size = 50
3153
"""The available optimised InterRepository types."""
3156
def copy_content(self, revision_id=None):
3157
"""Make a complete copy of the content in self into destination.
3159
This is a destructive operation! Do not use it on existing
3162
:param revision_id: Only copy the content needed to construct
3163
revision_id and its parents.
3166
self.target.set_make_working_trees(self.source.make_working_trees())
3167
except NotImplementedError:
3169
self.target.fetch(self.source, revision_id=revision_id)
3172
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3174
"""Fetch the content required to construct revision_id.
3176
The content is copied from self.source to self.target.
3178
:param revision_id: if None all content is copied, if NULL_REVISION no
3180
:param pb: optional progress bar to use for progress reports. If not
3181
provided a default one will be created.
3184
from bzrlib.fetch import RepoFetcher
3185
f = RepoFetcher(to_repository=self.target,
3186
from_repository=self.source,
3187
last_revision=revision_id,
3188
fetch_spec=fetch_spec,
3189
pb=pb, find_ghosts=find_ghosts)
3191
def _walk_to_common_revisions(self, revision_ids):
3192
"""Walk out from revision_ids in source to revisions target has.
3194
:param revision_ids: The start point for the search.
3195
:return: A set of revision ids.
3197
target_graph = self.target.get_graph()
3198
revision_ids = frozenset(revision_ids)
3199
missing_revs = set()
3200
source_graph = self.source.get_graph()
3201
# ensure we don't pay silly lookup costs.
3202
searcher = source_graph._make_breadth_first_searcher(revision_ids)
3203
null_set = frozenset([_mod_revision.NULL_REVISION])
3204
searcher_exhausted = False
3208
# Iterate the searcher until we have enough next_revs
3209
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3211
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3212
next_revs.update(next_revs_part)
3213
ghosts.update(ghosts_part)
3214
except StopIteration:
3215
searcher_exhausted = True
3217
# If there are ghosts in the source graph, and the caller asked for
3218
# them, make sure that they are present in the target.
3219
# We don't care about other ghosts as we can't fetch them and
3220
# haven't been asked to.
3221
ghosts_to_check = set(revision_ids.intersection(ghosts))
3222
revs_to_get = set(next_revs).union(ghosts_to_check)
3224
have_revs = set(target_graph.get_parent_map(revs_to_get))
3225
# we always have NULL_REVISION present.
3226
have_revs = have_revs.union(null_set)
3227
# Check if the target is missing any ghosts we need.
3228
ghosts_to_check.difference_update(have_revs)
3230
# One of the caller's revision_ids is a ghost in both the
3231
# source and the target.
3232
raise errors.NoSuchRevision(
3233
self.source, ghosts_to_check.pop())
3234
missing_revs.update(next_revs - have_revs)
3235
# Because we may have walked past the original stop point, make
3236
# sure everything is stopped
3237
stop_revs = searcher.find_seen_ancestors(have_revs)
3238
searcher.stop_searching_any(stop_revs)
3239
if searcher_exhausted:
3241
return searcher.get_result()
3244
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3245
"""Return the revision ids that source has that target does not.
3247
:param revision_id: only return revision ids included by this
3249
:param find_ghosts: If True find missing revisions in deep history
3250
rather than just finding the surface difference.
3251
:return: A bzrlib.graph.SearchResult.
3253
# stop searching at found target revisions.
3254
if not find_ghosts and revision_id is not None:
3255
return self._walk_to_common_revisions([revision_id])
3256
# generic, possibly worst case, slow code path.
3257
target_ids = set(self.target.all_revision_ids())
3258
if revision_id is not None:
3259
source_ids = self.source.get_ancestry(revision_id)
3260
if source_ids[0] is not None:
3261
raise AssertionError()
3264
source_ids = self.source.all_revision_ids()
3265
result_set = set(source_ids).difference(target_ids)
3266
return self.source.revision_ids_to_search_result(result_set)
3269
def _same_model(source, target):
3270
"""True if source and target have the same data representation.
3272
Note: this is always called on the base class; overriding it in a
3273
subclass will have no effect.
3276
InterRepository._assert_same_model(source, target)
3278
except errors.IncompatibleRepositories, e:
3282
def _assert_same_model(source, target):
3283
"""Raise an exception if two repositories do not use the same model.
3285
if source.supports_rich_root() != target.supports_rich_root():
3286
raise errors.IncompatibleRepositories(source, target,
3287
"different rich-root support")
3288
if source._serializer != target._serializer:
3289
raise errors.IncompatibleRepositories(source, target,
3290
"different serializers")
3293
class InterSameDataRepository(InterRepository):
3294
"""Code for converting between repositories that represent the same data.
3296
Data format and model must match for this to work.
3300
def _get_repo_format_to_test(self):
3301
"""Repository format for testing with.
3303
InterSameData can pull from subtree to subtree and from non-subtree to
3304
non-subtree, so we test this with the richest repository format.
3306
from bzrlib.repofmt import knitrepo
3307
return knitrepo.RepositoryFormatKnit3()
3310
def is_compatible(source, target):
3311
return InterRepository._same_model(source, target)
3314
class InterWeaveRepo(InterSameDataRepository):
3315
"""Optimised code paths between Weave based repositories.
3317
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
3318
implemented lazy inter-object optimisation.
3322
def _get_repo_format_to_test(self):
3323
from bzrlib.repofmt import weaverepo
3324
return weaverepo.RepositoryFormat7()
3327
def is_compatible(source, target):
3328
"""Be compatible with known Weave formats.
3330
We don't test for the stores being of specific types because that
3331
could lead to confusing results, and there is no need to be
3334
from bzrlib.repofmt.weaverepo import (
3340
return (isinstance(source._format, (RepositoryFormat5,
3342
RepositoryFormat7)) and
3343
isinstance(target._format, (RepositoryFormat5,
3345
RepositoryFormat7)))
3346
except AttributeError:
3350
def copy_content(self, revision_id=None):
3351
"""See InterRepository.copy_content()."""
3352
# weave specific optimised path:
3354
self.target.set_make_working_trees(self.source.make_working_trees())
3355
except (errors.RepositoryUpgradeRequired, NotImplemented):
3357
# FIXME do not peek!
3358
if self.source._transport.listable():
3359
pb = ui.ui_factory.nested_progress_bar()
3361
self.target.texts.insert_record_stream(
3362
self.source.texts.get_record_stream(
3363
self.source.texts.keys(), 'topological', False))
3364
pb.update('copying inventory', 0, 1)
3365
self.target.inventories.insert_record_stream(
3366
self.source.inventories.get_record_stream(
3367
self.source.inventories.keys(), 'topological', False))
3368
self.target.signatures.insert_record_stream(
3369
self.source.signatures.get_record_stream(
3370
self.source.signatures.keys(),
3372
self.target.revisions.insert_record_stream(
3373
self.source.revisions.get_record_stream(
3374
self.source.revisions.keys(),
3375
'topological', True))
3379
self.target.fetch(self.source, revision_id=revision_id)
3382
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3383
"""See InterRepository.missing_revision_ids()."""
3384
# we want all revisions to satisfy revision_id in source.
3385
# but we don't want to stat every file here and there.
3386
# we want then, all revisions other needs to satisfy revision_id
3387
# checked, but not those that we have locally.
3388
# so the first thing is to get a subset of the revisions to
3389
# satisfy revision_id in source, and then eliminate those that
3390
# we do already have.
3391
# this is slow on high latency connection to self, but as this
3392
# disk format scales terribly for push anyway due to rewriting
3393
# inventory.weave, this is considered acceptable.
3395
if revision_id is not None:
3396
source_ids = self.source.get_ancestry(revision_id)
3397
if source_ids[0] is not None:
3398
raise AssertionError()
3401
source_ids = self.source._all_possible_ids()
3402
source_ids_set = set(source_ids)
3403
# source_ids is the worst possible case we may need to pull.
3404
# now we want to filter source_ids against what we actually
3405
# have in target, but don't try to check for existence where we know
3406
# we do not have a revision as that would be pointless.
3407
target_ids = set(self.target._all_possible_ids())
3408
possibly_present_revisions = target_ids.intersection(source_ids_set)
3409
actually_present_revisions = set(
3410
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3411
required_revisions = source_ids_set.difference(actually_present_revisions)
3412
if revision_id is not None:
3413
# we used get_ancestry to determine source_ids then we are assured all
3414
# revisions referenced are present as they are installed in topological order.
3415
# and the tip revision was validated by get_ancestry.
3416
result_set = required_revisions
3418
# if we just grabbed the possibly available ids, then
3419
# we only have an estimate of whats available and need to validate
3420
# that against the revision records.
3422
self.source._eliminate_revisions_not_present(required_revisions))
3423
return self.source.revision_ids_to_search_result(result_set)
3426
class InterKnitRepo(InterSameDataRepository):
3427
"""Optimised code paths between Knit based repositories."""
3430
def _get_repo_format_to_test(self):
3431
from bzrlib.repofmt import knitrepo
3432
return knitrepo.RepositoryFormatKnit1()
3435
def is_compatible(source, target):
3436
"""Be compatible with known Knit formats.
3438
We don't test for the stores being of specific types because that
3439
could lead to confusing results, and there is no need to be
3442
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
3444
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
3445
isinstance(target._format, RepositoryFormatKnit))
3446
except AttributeError:
3448
return are_knits and InterRepository._same_model(source, target)
3451
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3452
"""See InterRepository.missing_revision_ids()."""
3453
if revision_id is not None:
3454
source_ids = self.source.get_ancestry(revision_id)
3455
if source_ids[0] is not None:
3456
raise AssertionError()
3459
source_ids = self.source.all_revision_ids()
3460
source_ids_set = set(source_ids)
3461
# source_ids is the worst possible case we may need to pull.
3462
# now we want to filter source_ids against what we actually
3463
# have in target, but don't try to check for existence where we know
3464
# we do not have a revision as that would be pointless.
3465
target_ids = set(self.target.all_revision_ids())
3466
possibly_present_revisions = target_ids.intersection(source_ids_set)
3467
actually_present_revisions = set(
3468
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3469
required_revisions = source_ids_set.difference(actually_present_revisions)
3470
if revision_id is not None:
3471
# we used get_ancestry to determine source_ids then we are assured all
3472
# revisions referenced are present as they are installed in topological order.
3473
# and the tip revision was validated by get_ancestry.
3474
result_set = required_revisions
3476
# if we just grabbed the possibly available ids, then
3477
# we only have an estimate of whats available and need to validate
3478
# that against the revision records.
3480
self.source._eliminate_revisions_not_present(required_revisions))
3481
return self.source.revision_ids_to_search_result(result_set)
3484
class InterDifferingSerializer(InterRepository):
3487
def _get_repo_format_to_test(self):
3491
def is_compatible(source, target):
3492
"""Be compatible with Knit2 source and Knit3 target"""
3493
# This is redundant with format.check_conversion_target(), however that
3494
# raises an exception, and we just want to say "False" as in we won't
3495
# support converting between these formats.
3496
if source.supports_rich_root() and not target.supports_rich_root():
3498
if (source._format.supports_tree_reference
3499
and not target._format.supports_tree_reference):
3503
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3504
"""Get the best delta and base for this revision.
3506
:return: (basis_id, delta)
3508
possible_trees = [(parent_id, cache[parent_id])
3509
for parent_id in parent_ids
3510
if parent_id in cache]
3511
if len(possible_trees) == 0:
3512
# There either aren't any parents, or the parents aren't in the
3513
# cache, so just use the last converted tree
3514
possible_trees.append((basis_id, cache[basis_id]))
3516
for basis_id, basis_tree in possible_trees:
3517
delta = tree.inventory._make_delta(basis_tree.inventory)
3518
deltas.append((len(delta), basis_id, delta))
3520
return deltas[0][1:]
3522
def _get_parent_keys(self, root_key, parent_map):
3523
"""Get the parent keys for a given root id."""
3524
root_id, rev_id = root_key
3525
# Include direct parents of the revision, but only if they used
3526
# the same root_id and are heads.
3528
for parent_id in parent_map[rev_id]:
3529
if parent_id == _mod_revision.NULL_REVISION:
3531
if parent_id not in self._revision_id_to_root_id:
3532
# We probably didn't read this revision, go spend the
3533
# extra effort to actually check
3535
tree = self.source.revision_tree(parent_id)
3536
except errors.NoSuchRevision:
3537
# Ghost, fill out _revision_id_to_root_id in case we
3538
# encounter this again.
3539
# But set parent_root_id to None since we don't really know
3540
parent_root_id = None
3542
parent_root_id = tree.get_root_id()
3543
self._revision_id_to_root_id[parent_id] = None
3545
parent_root_id = self._revision_id_to_root_id[parent_id]
3546
if root_id == parent_root_id:
3547
# With stacking we _might_ want to refer to a non-local
3548
# revision, but this code path only applies when we have the
3549
# full content available, so ghosts really are ghosts, not just
3550
# the edge of local data.
3551
parent_keys.append((parent_id,))
3553
# root_id may be in the parent anyway.
3555
tree = self.source.revision_tree(parent_id)
3556
except errors.NoSuchRevision:
3557
# ghost, can't refer to it.
3561
parent_keys.append((tree.inventory[root_id].revision,))
3562
except errors.NoSuchId:
3565
g = graph.Graph(self.source.revisions)
3566
heads = g.heads(parent_keys)
3568
for key in parent_keys:
3569
if key in heads and key not in selected_keys:
3570
selected_keys.append(key)
3571
return tuple([(root_id,)+ key for key in selected_keys])
3573
def _new_root_data_stream(self, root_keys_to_create, parent_map):
3574
for root_key in root_keys_to_create:
3575
parent_keys = self._get_parent_keys(root_key, parent_map)
3576
yield versionedfile.FulltextContentFactory(root_key,
3577
parent_keys, None, '')
3579
def _fetch_batch(self, revision_ids, basis_id, cache):
3580
"""Fetch across a few revisions.
3582
:param revision_ids: The revisions to copy
3583
:param basis_id: The revision_id of a tree that must be in cache, used
3584
as a basis for delta when no other base is available
3585
:param cache: A cache of RevisionTrees that we can use.
3586
:return: The revision_id of the last converted tree. The RevisionTree
3587
for it will be in cache
3589
# Walk though all revisions; get inventory deltas, copy referenced
3590
# texts that delta references, insert the delta, revision and
3592
root_keys_to_create = set()
3595
pending_revisions = []
3596
parent_map = self.source.get_parent_map(revision_ids)
3597
for tree in self.source.revision_trees(revision_ids):
3598
current_revision_id = tree.get_revision_id()
3599
parent_ids = parent_map.get(current_revision_id, ())
3600
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3602
if self._converting_to_rich_root:
3603
self._revision_id_to_root_id[current_revision_id] = \
3605
# Find text entries that need to be copied
3606
for old_path, new_path, file_id, entry in delta:
3607
if new_path is not None:
3610
if not self.target.supports_rich_root():
3611
# The target doesn't support rich root, so we don't
3614
if self._converting_to_rich_root:
3615
# This can't be copied normally, we have to insert
3617
root_keys_to_create.add((file_id, entry.revision))
3619
text_keys.add((file_id, entry.revision))
3620
revision = self.source.get_revision(current_revision_id)
3621
pending_deltas.append((basis_id, delta,
3622
current_revision_id, revision.parent_ids))
3623
pending_revisions.append(revision)
3624
cache[current_revision_id] = tree
3625
basis_id = current_revision_id
3627
from_texts = self.source.texts
3628
to_texts = self.target.texts
3629
if root_keys_to_create:
3630
root_stream = self._new_root_data_stream(root_keys_to_create,
3632
to_texts.insert_record_stream(root_stream)
3633
to_texts.insert_record_stream(from_texts.get_record_stream(
3634
text_keys, self.target._format._fetch_order,
3635
not self.target._format._fetch_uses_deltas))
3636
# insert inventory deltas
3637
for delta in pending_deltas:
3638
self.target.add_inventory_by_delta(*delta)
3639
if self.target._fallback_repositories:
3640
# Make sure this stacked repository has all the parent inventories
3641
# for the new revisions that we are about to insert. We do this
3642
# before adding the revisions so that no revision is added until
3643
# all the inventories it may depend on are added.
3645
revision_ids = set()
3646
for revision in pending_revisions:
3647
revision_ids.add(revision.revision_id)
3648
parent_ids.update(revision.parent_ids)
3649
parent_ids.difference_update(revision_ids)
3650
parent_ids.discard(_mod_revision.NULL_REVISION)
3651
parent_map = self.source.get_parent_map(parent_ids)
3652
for parent_tree in self.source.revision_trees(parent_ids):
3653
basis_id, delta = self._get_delta_for_revision(tree, parent_ids, basis_id, cache)
3654
current_revision_id = parent_tree.get_revision_id()
3655
parents_parents = parent_map[current_revision_id]
3656
self.target.add_inventory_by_delta(
3657
basis_id, delta, current_revision_id, parents_parents)
3658
# insert signatures and revisions
3659
for revision in pending_revisions:
3661
signature = self.source.get_signature_text(
3662
revision.revision_id)
3663
self.target.add_signature_text(revision.revision_id,
3665
except errors.NoSuchRevision:
3667
self.target.add_revision(revision.revision_id, revision)
3670
def _fetch_all_revisions(self, revision_ids, pb):
3671
"""Fetch everything for the list of revisions.
3673
:param revision_ids: The list of revisions to fetch. Must be in
3675
:param pb: A ProgressBar
3678
basis_id, basis_tree = self._get_basis(revision_ids[0])
3680
cache = lru_cache.LRUCache(100)
3681
cache[basis_id] = basis_tree
3682
del basis_tree # We don't want to hang on to it here
3684
for offset in range(0, len(revision_ids), batch_size):
3685
self.target.start_write_group()
3687
pb.update('Transferring revisions', offset,
3689
batch = revision_ids[offset:offset+batch_size]
3690
basis_id = self._fetch_batch(batch, basis_id, cache)
3692
self.target.abort_write_group()
3695
hint = self.target.commit_write_group()
3698
if hints and self.target._format.pack_compresses:
3699
self.target.pack(hint=hints)
3700
pb.update('Transferring revisions', len(revision_ids),
3704
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3706
"""See InterRepository.fetch()."""
3707
if fetch_spec is not None:
3708
raise AssertionError("Not implemented yet...")
3709
if (not self.source.supports_rich_root()
3710
and self.target.supports_rich_root()):
3711
self._converting_to_rich_root = True
3712
self._revision_id_to_root_id = {}
3714
self._converting_to_rich_root = False
3715
revision_ids = self.target.search_missing_revision_ids(self.source,
3716
revision_id, find_ghosts=find_ghosts).get_keys()
3717
if not revision_ids:
3719
revision_ids = tsort.topo_sort(
3720
self.source.get_graph().get_parent_map(revision_ids))
3721
if not revision_ids:
3723
# Walk though all revisions; get inventory deltas, copy referenced
3724
# texts that delta references, insert the delta, revision and
3726
first_rev = self.source.get_revision(revision_ids[0])
3728
my_pb = ui.ui_factory.nested_progress_bar()
3731
symbol_versioning.warn(
3732
symbol_versioning.deprecated_in((1, 14, 0))
3733
% "pb parameter to fetch()")
3736
self._fetch_all_revisions(revision_ids, pb)
3738
if my_pb is not None:
3740
return len(revision_ids), 0
3742
def _get_basis(self, first_revision_id):
3743
"""Get a revision and tree which exists in the target.
3745
This assumes that first_revision_id is selected for transmission
3746
because all other ancestors are already present. If we can't find an
3747
ancestor we fall back to NULL_REVISION since we know that is safe.
3749
:return: (basis_id, basis_tree)
3751
first_rev = self.source.get_revision(first_revision_id)
3753
basis_id = first_rev.parent_ids[0]
3754
# only valid as a basis if the target has it
3755
self.target.get_revision(basis_id)
3756
# Try to get a basis tree - if its a ghost it will hit the
3757
# NoSuchRevision case.
3758
basis_tree = self.source.revision_tree(basis_id)
3759
except (IndexError, errors.NoSuchRevision):
3760
basis_id = _mod_revision.NULL_REVISION
3761
basis_tree = self.source.revision_tree(basis_id)
3762
return basis_id, basis_tree
3765
InterRepository.register_optimiser(InterDifferingSerializer)
3766
InterRepository.register_optimiser(InterSameDataRepository)
3767
InterRepository.register_optimiser(InterWeaveRepo)
3768
InterRepository.register_optimiser(InterKnitRepo)
3771
class CopyConverter(object):
3772
"""A repository conversion tool which just performs a copy of the content.
3774
This is slow but quite reliable.
3777
def __init__(self, target_format):
3778
"""Create a CopyConverter.
3780
:param target_format: The format the resulting repository should be.
3782
self.target_format = target_format
3784
def convert(self, repo, pb):
3785
"""Perform the conversion of to_convert, giving feedback via pb.
3787
:param to_convert: The disk object to convert.
3788
:param pb: a progress bar to use for progress information.
3793
# this is only useful with metadir layouts - separated repo content.
3794
# trigger an assertion if not such
3795
repo._format.get_format_string()
3796
self.repo_dir = repo.bzrdir
3797
self.step('Moving repository to repository.backup')
3798
self.repo_dir.transport.move('repository', 'repository.backup')
3799
backup_transport = self.repo_dir.transport.clone('repository.backup')
3800
repo._format.check_conversion_target(self.target_format)
3801
self.source_repo = repo._format.open(self.repo_dir,
3803
_override_transport=backup_transport)
3804
self.step('Creating new repository')
3805
converted = self.target_format.initialize(self.repo_dir,
3806
self.source_repo.is_shared())
3807
converted.lock_write()
3809
self.step('Copying content into repository.')
3810
self.source_repo.copy_content_into(converted)
3813
self.step('Deleting old repository content.')
3814
self.repo_dir.transport.delete_tree('repository.backup')
3815
self.pb.note('repository converted')
3817
def step(self, message):
3818
"""Update the pb by a step."""
3820
self.pb.update(message, self.count, self.total)
3832
def _unescaper(match, _map=_unescape_map):
3833
code = match.group(1)
3837
if not code.startswith('#'):
3839
return unichr(int(code[1:])).encode('utf8')
3845
def _unescape_xml(data):
3846
"""Unescape predefined XML entities in a string of data."""
3848
if _unescape_re is None:
3849
_unescape_re = re.compile('\&([^;]*);')
3850
return _unescape_re.sub(_unescaper, data)
3853
class _VersionedFileChecker(object):
3855
def __init__(self, repository, text_key_references=None):
3856
self.repository = repository
3857
self.text_index = self.repository._generate_text_key_index(
3858
text_key_references=text_key_references)
3860
def calculate_file_version_parents(self, text_key):
3861
"""Calculate the correct parents for a file version according to
3864
parent_keys = self.text_index[text_key]
3865
if parent_keys == [_mod_revision.NULL_REVISION]:
3867
return tuple(parent_keys)
3869
def check_file_version_parents(self, texts, progress_bar=None):
3870
"""Check the parents stored in a versioned file are correct.
3872
It also detects file versions that are not referenced by their
3873
corresponding revision's inventory.
3875
:returns: A tuple of (wrong_parents, dangling_file_versions).
3876
wrong_parents is a dict mapping {revision_id: (stored_parents,
3877
correct_parents)} for each revision_id where the stored parents
3878
are not correct. dangling_file_versions is a set of (file_id,
3879
revision_id) tuples for versions that are present in this versioned
3880
file, but not used by the corresponding inventory.
3883
self.file_ids = set([file_id for file_id, _ in
3884
self.text_index.iterkeys()])
3885
# text keys is now grouped by file_id
3886
n_weaves = len(self.file_ids)
3887
files_in_revisions = {}
3888
revisions_of_files = {}
3889
n_versions = len(self.text_index)
3890
progress_bar.update('loading text store', 0, n_versions)
3891
parent_map = self.repository.texts.get_parent_map(self.text_index)
3892
# On unlistable transports this could well be empty/error...
3893
text_keys = self.repository.texts.keys()
3894
unused_keys = frozenset(text_keys) - set(self.text_index)
3895
for num, key in enumerate(self.text_index.iterkeys()):
3896
if progress_bar is not None:
3897
progress_bar.update('checking text graph', num, n_versions)
3898
correct_parents = self.calculate_file_version_parents(key)
3900
knit_parents = parent_map[key]
3901
except errors.RevisionNotPresent:
3904
if correct_parents != knit_parents:
3905
wrong_parents[key] = (knit_parents, correct_parents)
3906
return wrong_parents, unused_keys
3909
def _old_get_graph(repository, revision_id):
3910
"""DO NOT USE. That is all. I'm serious."""
3911
graph = repository.get_graph()
3912
revision_graph = dict(((key, value) for key, value in
3913
graph.iter_ancestry([revision_id]) if value is not None))
3914
return _strip_NULL_ghosts(revision_graph)
3917
def _strip_NULL_ghosts(revision_graph):
3918
"""Also don't use this. more compatibility code for unmigrated clients."""
3919
# Filter ghosts, and null:
3920
if _mod_revision.NULL_REVISION in revision_graph:
3921
del revision_graph[_mod_revision.NULL_REVISION]
3922
for key, parents in revision_graph.items():
3923
revision_graph[key] = tuple(parent for parent in parents if parent
3925
return revision_graph
3928
class StreamSink(object):
3929
"""An object that can insert a stream into a repository.
3931
This interface handles the complexity of reserialising inventories and
3932
revisions from different formats, and allows unidirectional insertion into
3933
stacked repositories without looking for the missing basis parents
3937
def __init__(self, target_repo):
3938
self.target_repo = target_repo
3940
def insert_stream(self, stream, src_format, resume_tokens):
3941
"""Insert a stream's content into the target repository.
3943
:param src_format: a bzr repository format.
3945
:return: a list of resume tokens and an iterable of keys additional
3946
items required before the insertion can be completed.
3948
self.target_repo.lock_write()
3951
self.target_repo.resume_write_group(resume_tokens)
3954
self.target_repo.start_write_group()
3957
# locked_insert_stream performs a commit|suspend.
3958
return self._locked_insert_stream(stream, src_format, is_resume)
3960
self.target_repo.abort_write_group(suppress_errors=True)
3963
self.target_repo.unlock()
3965
def _locked_insert_stream(self, stream, src_format, is_resume):
3966
to_serializer = self.target_repo._format._serializer
3967
src_serializer = src_format._serializer
3969
if to_serializer == src_serializer:
3970
# If serializers match and the target is a pack repository, set the
3971
# write cache size on the new pack. This avoids poor performance
3972
# on transports where append is unbuffered (such as
3973
# RemoteTransport). This is safe to do because nothing should read
3974
# back from the target repository while a stream with matching
3975
# serialization is being inserted.
3976
# The exception is that a delta record from the source that should
3977
# be a fulltext may need to be expanded by the target (see
3978
# test_fetch_revisions_with_deltas_into_pack); but we take care to
3979
# explicitly flush any buffered writes first in that rare case.
3981
new_pack = self.target_repo._pack_collection._new_pack
3982
except AttributeError:
3983
# Not a pack repository
3986
new_pack.set_write_cache_size(1024*1024)
3987
for substream_type, substream in stream:
3988
if substream_type == 'texts':
3989
self.target_repo.texts.insert_record_stream(substream)
3990
elif substream_type == 'inventories':
3991
if src_serializer == to_serializer:
3992
self.target_repo.inventories.insert_record_stream(
3995
self._extract_and_insert_inventories(
3996
substream, src_serializer)
3997
elif substream_type == 'chk_bytes':
3998
# XXX: This doesn't support conversions, as it assumes the
3999
# conversion was done in the fetch code.
4000
self.target_repo.chk_bytes.insert_record_stream(substream)
4001
elif substream_type == 'revisions':
4002
# This may fallback to extract-and-insert more often than
4003
# required if the serializers are different only in terms of
4005
if src_serializer == to_serializer:
4006
self.target_repo.revisions.insert_record_stream(
4009
self._extract_and_insert_revisions(substream,
4011
elif substream_type == 'signatures':
4012
self.target_repo.signatures.insert_record_stream(substream)
4014
raise AssertionError('kaboom! %s' % (substream_type,))
4015
# Done inserting data, and the missing_keys calculations will try to
4016
# read back from the inserted data, so flush the writes to the new pack
4017
# (if this is pack format).
4018
if new_pack is not None:
4019
new_pack._write_data('', flush=True)
4020
# Find all the new revisions (including ones from resume_tokens)
4021
missing_keys = self.target_repo.get_missing_parent_inventories(
4022
check_for_missing_texts=is_resume)
4024
for prefix, versioned_file in (
4025
('texts', self.target_repo.texts),
4026
('inventories', self.target_repo.inventories),
4027
('revisions', self.target_repo.revisions),
4028
('signatures', self.target_repo.signatures),
4029
('chk_bytes', self.target_repo.chk_bytes),
4031
if versioned_file is None:
4033
missing_keys.update((prefix,) + key for key in
4034
versioned_file.get_missing_compression_parent_keys())
4035
except NotImplementedError:
4036
# cannot even attempt suspending, and missing would have failed
4037
# during stream insertion.
4038
missing_keys = set()
4041
# suspend the write group and tell the caller what we is
4042
# missing. We know we can suspend or else we would not have
4043
# entered this code path. (All repositories that can handle
4044
# missing keys can handle suspending a write group).
4045
write_group_tokens = self.target_repo.suspend_write_group()
4046
return write_group_tokens, missing_keys
4047
hint = self.target_repo.commit_write_group()
4048
if (to_serializer != src_serializer and
4049
self.target_repo._format.pack_compresses):
4050
self.target_repo.pack(hint=hint)
4053
def _extract_and_insert_inventories(self, substream, serializer):
4054
"""Generate a new inventory versionedfile in target, converting data.
4056
The inventory is retrieved from the source, (deserializing it), and
4057
stored in the target (reserializing it in a different format).
4059
for record in substream:
4060
bytes = record.get_bytes_as('fulltext')
4061
revision_id = record.key[0]
4062
inv = serializer.read_inventory_from_string(bytes, revision_id)
4063
parents = [key[0] for key in record.parents]
4064
self.target_repo.add_inventory(revision_id, inv, parents)
4066
def _extract_and_insert_revisions(self, substream, serializer):
4067
for record in substream:
4068
bytes = record.get_bytes_as('fulltext')
4069
revision_id = record.key[0]
4070
rev = serializer.read_revision_from_string(bytes)
4071
if rev.revision_id != revision_id:
4072
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4073
self.target_repo.add_revision(revision_id, rev)
4076
if self.target_repo._format._fetch_reconcile:
4077
self.target_repo.reconcile()
4080
class StreamSource(object):
4081
"""A source of a stream for fetching between repositories."""
4083
def __init__(self, from_repository, to_format):
4084
"""Create a StreamSource streaming from from_repository."""
4085
self.from_repository = from_repository
4086
self.to_format = to_format
4088
def delta_on_metadata(self):
4089
"""Return True if delta's are permitted on metadata streams.
4091
That is on revisions and signatures.
4093
src_serializer = self.from_repository._format._serializer
4094
target_serializer = self.to_format._serializer
4095
return (self.to_format._fetch_uses_deltas and
4096
src_serializer == target_serializer)
4098
def _fetch_revision_texts(self, revs):
4099
# fetch signatures first and then the revision texts
4100
# may need to be a InterRevisionStore call here.
4101
from_sf = self.from_repository.signatures
4102
# A missing signature is just skipped.
4103
keys = [(rev_id,) for rev_id in revs]
4104
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4106
self.to_format._fetch_order,
4107
not self.to_format._fetch_uses_deltas))
4108
# If a revision has a delta, this is actually expanded inside the
4109
# insert_record_stream code now, which is an alternate fix for
4111
from_rf = self.from_repository.revisions
4112
revisions = from_rf.get_record_stream(
4114
self.to_format._fetch_order,
4115
not self.delta_on_metadata())
4116
return [('signatures', signatures), ('revisions', revisions)]
4118
def _generate_root_texts(self, revs):
4119
"""This will be called by __fetch between fetching weave texts and
4120
fetching the inventory weave.
4122
Subclasses should override this if they need to generate root texts
4123
after fetching weave texts.
4125
if self._rich_root_upgrade():
4127
return bzrlib.fetch.Inter1and2Helper(
4128
self.from_repository).generate_root_texts(revs)
4132
def get_stream(self, search):
4134
revs = search.get_keys()
4135
graph = self.from_repository.get_graph()
4136
revs = list(graph.iter_topo_order(revs))
4137
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4139
for knit_kind, file_id, revisions in data_to_fetch:
4140
if knit_kind != phase:
4142
# Make a new progress bar for this phase
4143
if knit_kind == "file":
4144
# Accumulate file texts
4145
text_keys.extend([(file_id, revision) for revision in
4147
elif knit_kind == "inventory":
4148
# Now copy the file texts.
4149
from_texts = self.from_repository.texts
4150
yield ('texts', from_texts.get_record_stream(
4151
text_keys, self.to_format._fetch_order,
4152
not self.to_format._fetch_uses_deltas))
4153
# Cause an error if a text occurs after we have done the
4156
# Before we process the inventory we generate the root
4157
# texts (if necessary) so that the inventories references
4159
for _ in self._generate_root_texts(revs):
4161
# NB: This currently reopens the inventory weave in source;
4162
# using a single stream interface instead would avoid this.
4163
from_weave = self.from_repository.inventories
4164
# we fetch only the referenced inventories because we do not
4165
# know for unselected inventories whether all their required
4166
# texts are present in the other repository - it could be
4168
for info in self._get_inventory_stream(revs):
4170
elif knit_kind == "signatures":
4171
# Nothing to do here; this will be taken care of when
4172
# _fetch_revision_texts happens.
4174
elif knit_kind == "revisions":
4175
for record in self._fetch_revision_texts(revs):
4178
raise AssertionError("Unknown knit kind %r" % knit_kind)
4180
def get_stream_for_missing_keys(self, missing_keys):
4181
# missing keys can only occur when we are byte copying and not
4182
# translating (because translation means we don't send
4183
# unreconstructable deltas ever).
4185
keys['texts'] = set()
4186
keys['revisions'] = set()
4187
keys['inventories'] = set()
4188
keys['chk_bytes'] = set()
4189
keys['signatures'] = set()
4190
for key in missing_keys:
4191
keys[key[0]].add(key[1:])
4192
if len(keys['revisions']):
4193
# If we allowed copying revisions at this point, we could end up
4194
# copying a revision without copying its required texts: a
4195
# violation of the requirements for repository integrity.
4196
raise AssertionError(
4197
'cannot copy revisions to fill in missing deltas %s' % (
4198
keys['revisions'],))
4199
for substream_kind, keys in keys.iteritems():
4200
vf = getattr(self.from_repository, substream_kind)
4201
if vf is None and keys:
4202
raise AssertionError(
4203
"cannot fill in keys for a versioned file we don't"
4204
" have: %s needs %s" % (substream_kind, keys))
4206
# No need to stream something we don't have
4208
# Ask for full texts always so that we don't need more round trips
4209
# after this stream.
4210
# Some of the missing keys are genuinely ghosts, so filter absent
4211
# records. The Sink is responsible for doing another check to
4212
# ensure that ghosts don't introduce missing data for future
4214
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4215
self.to_format._fetch_order, True))
4216
yield substream_kind, stream
4218
def inventory_fetch_order(self):
4219
if self._rich_root_upgrade():
4220
return 'topological'
4222
return self.to_format._fetch_order
4224
def _rich_root_upgrade(self):
4225
return (not self.from_repository._format.rich_root_data and
4226
self.to_format.rich_root_data)
4228
def _get_inventory_stream(self, revision_ids):
4229
from_format = self.from_repository._format
4230
if (from_format.supports_chks and self.to_format.supports_chks
4231
and (from_format._serializer == self.to_format._serializer)):
4232
# Both sides support chks, and they use the same serializer, so it
4233
# is safe to transmit the chk pages and inventory pages across
4235
return self._get_chk_inventory_stream(revision_ids)
4236
elif (not from_format.supports_chks):
4237
# Source repository doesn't support chks. So we can transmit the
4238
# inventories 'as-is' and either they are just accepted on the
4239
# target, or the Sink will properly convert it.
4240
return self._get_simple_inventory_stream(revision_ids)
4242
# XXX: Hack to make not-chk->chk fetch: copy the inventories as
4243
# inventories. Note that this should probably be done somehow
4244
# as part of bzrlib.repository.StreamSink. Except JAM couldn't
4245
# figure out how a non-chk repository could possibly handle
4246
# deserializing an inventory stream from a chk repo, as it
4247
# doesn't have a way to understand individual pages.
4248
return self._get_convertable_inventory_stream(revision_ids)
4250
def _get_simple_inventory_stream(self, revision_ids):
4251
from_weave = self.from_repository.inventories
4252
yield ('inventories', from_weave.get_record_stream(
4253
[(rev_id,) for rev_id in revision_ids],
4254
self.inventory_fetch_order(),
4255
not self.delta_on_metadata()))
4257
def _get_chk_inventory_stream(self, revision_ids):
4258
"""Fetch the inventory texts, along with the associated chk maps."""
4259
# We want an inventory outside of the search set, so that we can filter
4260
# out uninteresting chk pages. For now we use
4261
# _find_revision_outside_set, but if we had a Search with cut_revs, we
4262
# could use that instead.
4263
start_rev_id = self.from_repository._find_revision_outside_set(
4265
start_rev_key = (start_rev_id,)
4266
inv_keys_to_fetch = [(rev_id,) for rev_id in revision_ids]
4267
if start_rev_id != _mod_revision.NULL_REVISION:
4268
inv_keys_to_fetch.append((start_rev_id,))
4269
# Any repo that supports chk_bytes must also support out-of-order
4270
# insertion. At least, that is how we expect it to work
4271
# We use get_record_stream instead of iter_inventories because we want
4272
# to be able to insert the stream as well. We could instead fetch
4273
# allowing deltas, and then iter_inventories, but we don't know whether
4274
# source or target is more 'local' anway.
4275
inv_stream = self.from_repository.inventories.get_record_stream(
4276
inv_keys_to_fetch, 'unordered',
4277
True) # We need them as full-texts so we can find their references
4278
uninteresting_chk_roots = set()
4279
interesting_chk_roots = set()
4280
def filter_inv_stream(inv_stream):
4281
for idx, record in enumerate(inv_stream):
4282
### child_pb.update('fetch inv', idx, len(inv_keys_to_fetch))
4283
bytes = record.get_bytes_as('fulltext')
4284
chk_inv = inventory.CHKInventory.deserialise(
4285
self.from_repository.chk_bytes, bytes, record.key)
4286
if record.key == start_rev_key:
4287
uninteresting_chk_roots.add(chk_inv.id_to_entry.key())
4288
p_id_map = chk_inv.parent_id_basename_to_file_id
4289
if p_id_map is not None:
4290
uninteresting_chk_roots.add(p_id_map.key())
4293
interesting_chk_roots.add(chk_inv.id_to_entry.key())
4294
p_id_map = chk_inv.parent_id_basename_to_file_id
4295
if p_id_map is not None:
4296
interesting_chk_roots.add(p_id_map.key())
4297
### pb.update('fetch inventory', 0, 2)
4298
yield ('inventories', filter_inv_stream(inv_stream))
4299
# Now that we have worked out all of the interesting root nodes, grab
4300
# all of the interesting pages and insert them
4301
### pb.update('fetch inventory', 1, 2)
4302
interesting = chk_map.iter_interesting_nodes(
4303
self.from_repository.chk_bytes, interesting_chk_roots,
4304
uninteresting_chk_roots)
4305
def to_stream_adapter():
4306
"""Adapt the iter_interesting_nodes result to a single stream.
4308
iter_interesting_nodes returns records as it processes them, along
4309
with keys. However, we only want to return the records themselves.
4311
for record, items in interesting:
4312
if record is not None:
4314
# XXX: We could instead call get_record_stream(records.keys())
4315
# ATM, this will always insert the records as fulltexts, and
4316
# requires that you can hang on to records once you have gone
4317
# on to the next one. Further, it causes the target to
4318
# recompress the data. Testing shows it to be faster than
4319
# requesting the records again, though.
4320
yield ('chk_bytes', to_stream_adapter())
4321
### pb.update('fetch inventory', 2, 2)
4323
def _get_convertable_inventory_stream(self, revision_ids):
4324
# XXX: One of source or target is using chks, and they don't have
4325
# compatible serializations. The StreamSink code expects to be
4326
# able to convert on the target, so we need to put
4327
# bytes-on-the-wire that can be converted
4328
yield ('inventories', self._stream_invs_as_fulltexts(revision_ids))
4330
def _stream_invs_as_fulltexts(self, revision_ids):
4331
from_repo = self.from_repository
4332
from_serializer = from_repo._format._serializer
4333
revision_keys = [(rev_id,) for rev_id in revision_ids]
4334
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4335
for inv in self.from_repository.iter_inventories(revision_ids):
4336
# XXX: This is a bit hackish, but it works. Basically,
4337
# CHKSerializer 'accidentally' supports
4338
# read/write_inventory_to_string, even though that is never
4339
# the format that is stored on disk. It *does* give us a
4340
# single string representation for an inventory, so live with
4342
# This would be far better if we had a 'serialized inventory
4343
# delta' form. Then we could use 'inventory._make_delta', and
4344
# transmit that. This would both be faster to generate, and
4345
# result in fewer bytes-on-the-wire.
4346
as_bytes = from_serializer.write_inventory_to_string(inv)
4347
key = (inv.revision_id,)
4348
parent_keys = parent_map.get(key, ())
4349
yield versionedfile.FulltextContentFactory(
4350
key, parent_keys, None, as_bytes)
4353
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4354
stop_revision=None):
4355
"""Extend the partial history to include a given index
4357
If a stop_index is supplied, stop when that index has been reached.
4358
If a stop_revision is supplied, stop when that revision is
4359
encountered. Otherwise, stop when the beginning of history is
4362
:param stop_index: The index which should be present. When it is
4363
present, history extension will stop.
4364
:param stop_revision: The revision id which should be present. When
4365
it is encountered, history extension will stop.
4367
start_revision = partial_history_cache[-1]
4368
iterator = repo.iter_reverse_revision_history(start_revision)
4370
#skip the last revision in the list
4373
if (stop_index is not None and
4374
len(partial_history_cache) > stop_index):
4376
if partial_history_cache[-1] == stop_revision:
4378
revision_id = iterator.next()
4379
partial_history_cache.append(revision_id)
4380
except StopIteration:
added
removed
bzrlib/repository.py
