13
13
# You should have received a copy of the GNU General Public License
14
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
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from cStringIO import StringIO
17
19
from bzrlib.lazy_import import lazy_import
18
20
lazy_import(globals(), """
21
25
from bzrlib import (
37
37
revision as _mod_revision,
42
from bzrlib.bundle import serializer
43
from bzrlib.recordcounter import RecordCounter
44
from bzrlib.revisiontree import InventoryRevisionTree
42
from bzrlib.revisiontree import RevisionTree
45
43
from bzrlib.store.versioned import VersionedFileStore
44
from bzrlib.store.text import TextStore
46
45
from bzrlib.testament import Testament
55
from bzrlib.decorators import needs_read_lock, needs_write_lock, only_raises
49
from bzrlib.decorators import needs_read_lock, needs_write_lock
56
50
from bzrlib.inter import InterObject
57
from bzrlib.inventory import (
63
from bzrlib.lock import _RelockDebugMixin, LogicalLockResult
64
from bzrlib.trace import (
65
log_exception_quietly, note, mutter, mutter_callsite, warning)
51
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
52
from bzrlib.symbol_versioning import (
56
from bzrlib.trace import mutter, note, warning
68
59
# Old formats display a warning, but only once
69
60
_deprecation_warning_done = False
72
class IsInWriteGroupError(errors.InternalBzrError):
74
_fmt = "May not refresh_data of repo %(repo)s while in a write group."
76
def __init__(self, repo):
77
errors.InternalBzrError.__init__(self, repo=repo)
80
class CommitBuilder(object):
81
"""Provides an interface to build up a commit.
83
This allows describing a tree to be committed without needing to
84
know the internals of the format of the repository.
87
# all clients should supply tree roots.
88
record_root_entry = True
89
# the default CommitBuilder does not manage trees whose root is versioned.
90
_versioned_root = False
91
# this commit builder supports the record_entry_contents interface
92
supports_record_entry_contents = True
94
def __init__(self, repository, parents, config, timestamp=None,
95
timezone=None, committer=None, revprops=None,
96
revision_id=None, lossy=False):
97
"""Initiate a CommitBuilder.
99
:param repository: Repository to commit to.
100
:param parents: Revision ids of the parents of the new revision.
101
:param timestamp: Optional timestamp recorded for commit.
102
:param timezone: Optional timezone for timestamp.
103
:param committer: Optional committer to set for commit.
104
:param revprops: Optional dictionary of revision properties.
105
:param revision_id: Optional revision id.
106
:param lossy: Whether to discard data that can not be natively
107
represented, when pushing to a foreign VCS
109
self._config = config
112
if committer is None:
113
self._committer = self._config.username()
114
elif not isinstance(committer, unicode):
115
self._committer = committer.decode() # throw if non-ascii
117
self._committer = committer
119
self.new_inventory = Inventory(None)
120
self._new_revision_id = revision_id
121
self.parents = parents
122
self.repository = repository
125
if revprops is not None:
126
self._validate_revprops(revprops)
127
self._revprops.update(revprops)
129
if timestamp is None:
130
timestamp = time.time()
131
# Restrict resolution to 1ms
132
self._timestamp = round(timestamp, 3)
135
self._timezone = osutils.local_time_offset()
137
self._timezone = int(timezone)
139
self._generate_revision_if_needed()
140
self.__heads = graph.HeadsCache(repository.get_graph()).heads
141
self._basis_delta = []
142
# API compatibility, older code that used CommitBuilder did not call
143
# .record_delete(), which means the delta that is computed would not be
144
# valid. Callers that will call record_delete() should call
145
# .will_record_deletes() to indicate that.
146
self._recording_deletes = False
147
# memo'd check for no-op commits.
148
self._any_changes = False
150
def any_changes(self):
151
"""Return True if any entries were changed.
153
This includes merge-only changes. It is the core for the --unchanged
156
:return: True if any changes have occured.
158
return self._any_changes
160
def _validate_unicode_text(self, text, context):
161
"""Verify things like commit messages don't have bogus characters."""
163
raise ValueError('Invalid value for %s: %r' % (context, text))
165
def _validate_revprops(self, revprops):
166
for key, value in revprops.iteritems():
167
# We know that the XML serializers do not round trip '\r'
168
# correctly, so refuse to accept them
169
if not isinstance(value, basestring):
170
raise ValueError('revision property (%s) is not a valid'
171
' (unicode) string: %r' % (key, value))
172
self._validate_unicode_text(value,
173
'revision property (%s)' % (key,))
175
def _ensure_fallback_inventories(self):
176
"""Ensure that appropriate inventories are available.
178
This only applies to repositories that are stacked, and is about
179
enusring the stacking invariants. Namely, that for any revision that is
180
present, we either have all of the file content, or we have the parent
181
inventory and the delta file content.
183
if not self.repository._fallback_repositories:
185
if not self.repository._format.supports_chks:
186
raise errors.BzrError("Cannot commit directly to a stacked branch"
187
" in pre-2a formats. See "
188
"https://bugs.launchpad.net/bzr/+bug/375013 for details.")
189
# This is a stacked repo, we need to make sure we have the parent
190
# inventories for the parents.
191
parent_keys = [(p,) for p in self.parents]
192
parent_map = self.repository.inventories._index.get_parent_map(parent_keys)
193
missing_parent_keys = set([pk for pk in parent_keys
194
if pk not in parent_map])
195
fallback_repos = list(reversed(self.repository._fallback_repositories))
196
missing_keys = [('inventories', pk[0])
197
for pk in missing_parent_keys]
199
while missing_keys and fallback_repos:
200
fallback_repo = fallback_repos.pop()
201
source = fallback_repo._get_source(self.repository._format)
202
sink = self.repository._get_sink()
203
stream = source.get_stream_for_missing_keys(missing_keys)
204
missing_keys = sink.insert_stream_without_locking(stream,
205
self.repository._format)
207
raise errors.BzrError('Unable to fill in parent inventories for a'
210
def commit(self, message):
211
"""Make the actual commit.
213
:return: The revision id of the recorded revision.
215
self._validate_unicode_text(message, 'commit message')
216
rev = _mod_revision.Revision(
217
timestamp=self._timestamp,
218
timezone=self._timezone,
219
committer=self._committer,
221
inventory_sha1=self.inv_sha1,
222
revision_id=self._new_revision_id,
223
properties=self._revprops)
224
rev.parent_ids = self.parents
225
self.repository.add_revision(self._new_revision_id, rev,
226
self.new_inventory, self._config)
227
self._ensure_fallback_inventories()
228
self.repository.commit_write_group()
229
return self._new_revision_id
232
"""Abort the commit that is being built.
234
self.repository.abort_write_group()
236
def revision_tree(self):
237
"""Return the tree that was just committed.
239
After calling commit() this can be called to get a
240
InventoryRevisionTree representing the newly committed tree. This is
241
preferred to calling Repository.revision_tree() because that may
242
require deserializing the inventory, while we already have a copy in
245
if self.new_inventory is None:
246
self.new_inventory = self.repository.get_inventory(
247
self._new_revision_id)
248
return InventoryRevisionTree(self.repository, self.new_inventory,
249
self._new_revision_id)
251
def finish_inventory(self):
252
"""Tell the builder that the inventory is finished.
254
:return: The inventory id in the repository, which can be used with
255
repository.get_inventory.
257
if self.new_inventory is None:
258
# an inventory delta was accumulated without creating a new
260
basis_id = self.basis_delta_revision
261
# We ignore the 'inventory' returned by add_inventory_by_delta
262
# because self.new_inventory is used to hint to the rest of the
263
# system what code path was taken
264
self.inv_sha1, _ = self.repository.add_inventory_by_delta(
265
basis_id, self._basis_delta, self._new_revision_id,
268
if self.new_inventory.root is None:
269
raise AssertionError('Root entry should be supplied to'
270
' record_entry_contents, as of bzr 0.10.')
271
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
272
self.new_inventory.revision_id = self._new_revision_id
273
self.inv_sha1 = self.repository.add_inventory(
274
self._new_revision_id,
278
return self._new_revision_id
280
def _gen_revision_id(self):
281
"""Return new revision-id."""
282
return generate_ids.gen_revision_id(self._committer, self._timestamp)
284
def _generate_revision_if_needed(self):
285
"""Create a revision id if None was supplied.
287
If the repository can not support user-specified revision ids
288
they should override this function and raise CannotSetRevisionId
289
if _new_revision_id is not None.
291
:raises: CannotSetRevisionId
293
if self._new_revision_id is None:
294
self._new_revision_id = self._gen_revision_id()
295
self.random_revid = True
297
self.random_revid = False
299
def _heads(self, file_id, revision_ids):
300
"""Calculate the graph heads for revision_ids in the graph of file_id.
302
This can use either a per-file graph or a global revision graph as we
303
have an identity relationship between the two graphs.
305
return self.__heads(revision_ids)
307
def _check_root(self, ie, parent_invs, tree):
308
"""Helper for record_entry_contents.
310
:param ie: An entry being added.
311
:param parent_invs: The inventories of the parent revisions of the
313
:param tree: The tree that is being committed.
315
# In this revision format, root entries have no knit or weave When
316
# serializing out to disk and back in root.revision is always
318
ie.revision = self._new_revision_id
320
def _require_root_change(self, tree):
321
"""Enforce an appropriate root object change.
323
This is called once when record_iter_changes is called, if and only if
324
the root was not in the delta calculated by record_iter_changes.
326
:param tree: The tree which is being committed.
328
if len(self.parents) == 0:
329
raise errors.RootMissing()
330
entry = entry_factory['directory'](tree.path2id(''), '',
332
entry.revision = self._new_revision_id
333
self._basis_delta.append(('', '', entry.file_id, entry))
335
def _get_delta(self, ie, basis_inv, path):
336
"""Get a delta against the basis inventory for ie."""
337
if ie.file_id not in basis_inv:
339
result = (None, path, ie.file_id, ie)
340
self._basis_delta.append(result)
342
elif ie != basis_inv[ie.file_id]:
344
# TODO: avoid tis id2path call.
345
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
346
self._basis_delta.append(result)
352
def get_basis_delta(self):
353
"""Return the complete inventory delta versus the basis inventory.
355
This has been built up with the calls to record_delete and
356
record_entry_contents. The client must have already called
357
will_record_deletes() to indicate that they will be generating a
360
:return: An inventory delta, suitable for use with apply_delta, or
361
Repository.add_inventory_by_delta, etc.
363
if not self._recording_deletes:
364
raise AssertionError("recording deletes not activated.")
365
return self._basis_delta
367
def record_delete(self, path, file_id):
368
"""Record that a delete occured against a basis tree.
370
This is an optional API - when used it adds items to the basis_delta
371
being accumulated by the commit builder. It cannot be called unless the
372
method will_record_deletes() has been called to inform the builder that
373
a delta is being supplied.
375
:param path: The path of the thing deleted.
376
:param file_id: The file id that was deleted.
378
if not self._recording_deletes:
379
raise AssertionError("recording deletes not activated.")
380
delta = (path, None, file_id, None)
381
self._basis_delta.append(delta)
382
self._any_changes = True
385
def will_record_deletes(self):
386
"""Tell the commit builder that deletes are being notified.
388
This enables the accumulation of an inventory delta; for the resulting
389
commit to be valid, deletes against the basis MUST be recorded via
390
builder.record_delete().
392
self._recording_deletes = True
394
basis_id = self.parents[0]
396
basis_id = _mod_revision.NULL_REVISION
397
self.basis_delta_revision = basis_id
399
def record_entry_contents(self, ie, parent_invs, path, tree,
401
"""Record the content of ie from tree into the commit if needed.
403
Side effect: sets ie.revision when unchanged
405
:param ie: An inventory entry present in the commit.
406
:param parent_invs: The inventories of the parent revisions of the
408
:param path: The path the entry is at in the tree.
409
:param tree: The tree which contains this entry and should be used to
411
:param content_summary: Summary data from the tree about the paths
412
content - stat, length, exec, sha/link target. This is only
413
accessed when the entry has a revision of None - that is when it is
414
a candidate to commit.
415
:return: A tuple (change_delta, version_recorded, fs_hash).
416
change_delta is an inventory_delta change for this entry against
417
the basis tree of the commit, or None if no change occured against
419
version_recorded is True if a new version of the entry has been
420
recorded. For instance, committing a merge where a file was only
421
changed on the other side will return (delta, False).
422
fs_hash is either None, or the hash details for the path (currently
423
a tuple of the contents sha1 and the statvalue returned by
424
tree.get_file_with_stat()).
426
if self.new_inventory.root is None:
427
if ie.parent_id is not None:
428
raise errors.RootMissing()
429
self._check_root(ie, parent_invs, tree)
430
if ie.revision is None:
431
kind = content_summary[0]
433
# ie is carried over from a prior commit
435
# XXX: repository specific check for nested tree support goes here - if
436
# the repo doesn't want nested trees we skip it ?
437
if (kind == 'tree-reference' and
438
not self.repository._format.supports_tree_reference):
439
# mismatch between commit builder logic and repository:
440
# this needs the entry creation pushed down into the builder.
441
raise NotImplementedError('Missing repository subtree support.')
442
self.new_inventory.add(ie)
444
# TODO: slow, take it out of the inner loop.
446
basis_inv = parent_invs[0]
448
basis_inv = Inventory(root_id=None)
450
# ie.revision is always None if the InventoryEntry is considered
451
# for committing. We may record the previous parents revision if the
452
# content is actually unchanged against a sole head.
453
if ie.revision is not None:
454
if not self._versioned_root and path == '':
455
# repositories that do not version the root set the root's
456
# revision to the new commit even when no change occurs (more
457
# specifically, they do not record a revision on the root; and
458
# the rev id is assigned to the root during deserialisation -
459
# this masks when a change may have occurred against the basis.
460
# To match this we always issue a delta, because the revision
461
# of the root will always be changing.
462
if ie.file_id in basis_inv:
463
delta = (basis_inv.id2path(ie.file_id), path,
467
delta = (None, path, ie.file_id, ie)
468
self._basis_delta.append(delta)
469
return delta, False, None
471
# we don't need to commit this, because the caller already
472
# determined that an existing revision of this file is
473
# appropriate. If it's not being considered for committing then
474
# it and all its parents to the root must be unaltered so
475
# no-change against the basis.
476
if ie.revision == self._new_revision_id:
477
raise AssertionError("Impossible situation, a skipped "
478
"inventory entry (%r) claims to be modified in this "
479
"commit (%r).", (ie, self._new_revision_id))
480
return None, False, None
481
# XXX: Friction: parent_candidates should return a list not a dict
482
# so that we don't have to walk the inventories again.
483
parent_candiate_entries = ie.parent_candidates(parent_invs)
484
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
486
for inv in parent_invs:
487
if ie.file_id in inv:
488
old_rev = inv[ie.file_id].revision
489
if old_rev in head_set:
490
heads.append(inv[ie.file_id].revision)
491
head_set.remove(inv[ie.file_id].revision)
494
# now we check to see if we need to write a new record to the
496
# We write a new entry unless there is one head to the ancestors, and
497
# the kind-derived content is unchanged.
499
# Cheapest check first: no ancestors, or more the one head in the
500
# ancestors, we write a new node.
504
# There is a single head, look it up for comparison
505
parent_entry = parent_candiate_entries[heads[0]]
506
# if the non-content specific data has changed, we'll be writing a
508
if (parent_entry.parent_id != ie.parent_id or
509
parent_entry.name != ie.name):
511
# now we need to do content specific checks:
513
# if the kind changed the content obviously has
514
if kind != parent_entry.kind:
516
# Stat cache fingerprint feedback for the caller - None as we usually
517
# don't generate one.
520
if content_summary[2] is None:
521
raise ValueError("Files must not have executable = None")
523
# We can't trust a check of the file length because of content
525
if (# if the exec bit has changed we have to store:
526
parent_entry.executable != content_summary[2]):
528
elif parent_entry.text_sha1 == content_summary[3]:
529
# all meta and content is unchanged (using a hash cache
530
# hit to check the sha)
531
ie.revision = parent_entry.revision
532
ie.text_size = parent_entry.text_size
533
ie.text_sha1 = parent_entry.text_sha1
534
ie.executable = parent_entry.executable
535
return self._get_delta(ie, basis_inv, path), False, None
537
# Either there is only a hash change(no hash cache entry,
538
# or same size content change), or there is no change on
540
# Provide the parent's hash to the store layer, so that the
541
# content is unchanged we will not store a new node.
542
nostore_sha = parent_entry.text_sha1
544
# We want to record a new node regardless of the presence or
545
# absence of a content change in the file.
547
ie.executable = content_summary[2]
548
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
550
text = file_obj.read()
554
ie.text_sha1, ie.text_size = self._add_text_to_weave(
555
ie.file_id, text, heads, nostore_sha)
556
# Let the caller know we generated a stat fingerprint.
557
fingerprint = (ie.text_sha1, stat_value)
558
except errors.ExistingContent:
559
# Turns out that the file content was unchanged, and we were
560
# only going to store a new node if it was changed. Carry over
562
ie.revision = parent_entry.revision
563
ie.text_size = parent_entry.text_size
564
ie.text_sha1 = parent_entry.text_sha1
565
ie.executable = parent_entry.executable
566
return self._get_delta(ie, basis_inv, path), False, None
567
elif kind == 'directory':
569
# all data is meta here, nothing specific to directory, so
571
ie.revision = parent_entry.revision
572
return self._get_delta(ie, basis_inv, path), False, None
573
self._add_text_to_weave(ie.file_id, '', heads, None)
574
elif kind == 'symlink':
575
current_link_target = content_summary[3]
577
# symlink target is not generic metadata, check if it has
579
if current_link_target != parent_entry.symlink_target:
582
# unchanged, carry over.
583
ie.revision = parent_entry.revision
584
ie.symlink_target = parent_entry.symlink_target
585
return self._get_delta(ie, basis_inv, path), False, None
586
ie.symlink_target = current_link_target
587
self._add_text_to_weave(ie.file_id, '', heads, None)
588
elif kind == 'tree-reference':
590
if content_summary[3] != parent_entry.reference_revision:
593
# unchanged, carry over.
594
ie.reference_revision = parent_entry.reference_revision
595
ie.revision = parent_entry.revision
596
return self._get_delta(ie, basis_inv, path), False, None
597
ie.reference_revision = content_summary[3]
598
if ie.reference_revision is None:
599
raise AssertionError("invalid content_summary for nested tree: %r"
600
% (content_summary,))
601
self._add_text_to_weave(ie.file_id, '', heads, None)
603
raise NotImplementedError('unknown kind')
604
ie.revision = self._new_revision_id
605
self._any_changes = True
606
return self._get_delta(ie, basis_inv, path), True, fingerprint
608
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
609
_entry_factory=entry_factory):
610
"""Record a new tree via iter_changes.
612
:param tree: The tree to obtain text contents from for changed objects.
613
:param basis_revision_id: The revision id of the tree the iter_changes
614
has been generated against. Currently assumed to be the same
615
as self.parents[0] - if it is not, errors may occur.
616
:param iter_changes: An iter_changes iterator with the changes to apply
617
to basis_revision_id. The iterator must not include any items with
618
a current kind of None - missing items must be either filtered out
619
or errored-on beefore record_iter_changes sees the item.
620
:param _entry_factory: Private method to bind entry_factory locally for
622
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
625
# Create an inventory delta based on deltas between all the parents and
626
# deltas between all the parent inventories. We use inventory delta's
627
# between the inventory objects because iter_changes masks
628
# last-changed-field only changes.
630
# file_id -> change map, change is fileid, paths, changed, versioneds,
631
# parents, names, kinds, executables
633
# {file_id -> revision_id -> inventory entry, for entries in parent
634
# trees that are not parents[0]
638
revtrees = list(self.repository.revision_trees(self.parents))
639
except errors.NoSuchRevision:
640
# one or more ghosts, slow path.
642
for revision_id in self.parents:
644
revtrees.append(self.repository.revision_tree(revision_id))
645
except errors.NoSuchRevision:
647
basis_revision_id = _mod_revision.NULL_REVISION
649
revtrees.append(self.repository.revision_tree(
650
_mod_revision.NULL_REVISION))
651
# The basis inventory from a repository
653
basis_inv = revtrees[0].inventory
655
basis_inv = self.repository.revision_tree(
656
_mod_revision.NULL_REVISION).inventory
657
if len(self.parents) > 0:
658
if basis_revision_id != self.parents[0] and not ghost_basis:
660
"arbitrary basis parents not yet supported with merges")
661
for revtree in revtrees[1:]:
662
for change in revtree.inventory._make_delta(basis_inv):
663
if change[1] is None:
664
# Not present in this parent.
666
if change[2] not in merged_ids:
667
if change[0] is not None:
668
basis_entry = basis_inv[change[2]]
669
merged_ids[change[2]] = [
671
basis_entry.revision,
674
parent_entries[change[2]] = {
676
basis_entry.revision:basis_entry,
678
change[3].revision:change[3],
681
merged_ids[change[2]] = [change[3].revision]
682
parent_entries[change[2]] = {change[3].revision:change[3]}
684
merged_ids[change[2]].append(change[3].revision)
685
parent_entries[change[2]][change[3].revision] = change[3]
688
# Setup the changes from the tree:
689
# changes maps file_id -> (change, [parent revision_ids])
691
for change in iter_changes:
692
# This probably looks up in basis_inv way to much.
693
if change[1][0] is not None:
694
head_candidate = [basis_inv[change[0]].revision]
697
changes[change[0]] = change, merged_ids.get(change[0],
699
unchanged_merged = set(merged_ids) - set(changes)
700
# Extend the changes dict with synthetic changes to record merges of
702
for file_id in unchanged_merged:
703
# Record a merged version of these items that did not change vs the
704
# basis. This can be either identical parallel changes, or a revert
705
# of a specific file after a merge. The recorded content will be
706
# that of the current tree (which is the same as the basis), but
707
# the per-file graph will reflect a merge.
708
# NB:XXX: We are reconstructing path information we had, this
709
# should be preserved instead.
710
# inv delta change: (file_id, (path_in_source, path_in_target),
711
# changed_content, versioned, parent, name, kind,
714
basis_entry = basis_inv[file_id]
715
except errors.NoSuchId:
716
# a change from basis->some_parents but file_id isn't in basis
717
# so was new in the merge, which means it must have changed
718
# from basis -> current, and as it hasn't the add was reverted
719
# by the user. So we discard this change.
723
(basis_inv.id2path(file_id), tree.id2path(file_id)),
725
(basis_entry.parent_id, basis_entry.parent_id),
726
(basis_entry.name, basis_entry.name),
727
(basis_entry.kind, basis_entry.kind),
728
(basis_entry.executable, basis_entry.executable))
729
changes[file_id] = (change, merged_ids[file_id])
730
# changes contains tuples with the change and a set of inventory
731
# candidates for the file.
733
# old_path, new_path, file_id, new_inventory_entry
734
seen_root = False # Is the root in the basis delta?
735
inv_delta = self._basis_delta
736
modified_rev = self._new_revision_id
737
for change, head_candidates in changes.values():
738
if change[3][1]: # versioned in target.
739
# Several things may be happening here:
740
# We may have a fork in the per-file graph
741
# - record a change with the content from tree
742
# We may have a change against < all trees
743
# - carry over the tree that hasn't changed
744
# We may have a change against all trees
745
# - record the change with the content from tree
748
entry = _entry_factory[kind](file_id, change[5][1],
750
head_set = self._heads(change[0], set(head_candidates))
753
for head_candidate in head_candidates:
754
if head_candidate in head_set:
755
heads.append(head_candidate)
756
head_set.remove(head_candidate)
759
# Could be a carry-over situation:
760
parent_entry_revs = parent_entries.get(file_id, None)
761
if parent_entry_revs:
762
parent_entry = parent_entry_revs.get(heads[0], None)
765
if parent_entry is None:
766
# The parent iter_changes was called against is the one
767
# that is the per-file head, so any change is relevant
768
# iter_changes is valid.
769
carry_over_possible = False
771
# could be a carry over situation
772
# A change against the basis may just indicate a merge,
773
# we need to check the content against the source of the
774
# merge to determine if it was changed after the merge
776
if (parent_entry.kind != entry.kind or
777
parent_entry.parent_id != entry.parent_id or
778
parent_entry.name != entry.name):
779
# Metadata common to all entries has changed
780
# against per-file parent
781
carry_over_possible = False
783
carry_over_possible = True
784
# per-type checks for changes against the parent_entry
787
# Cannot be a carry-over situation
788
carry_over_possible = False
789
# Populate the entry in the delta
791
# XXX: There is still a small race here: If someone reverts the content of a file
792
# after iter_changes examines and decides it has changed,
793
# we will unconditionally record a new version even if some
794
# other process reverts it while commit is running (with
795
# the revert happening after iter_changes did its
798
entry.executable = True
800
entry.executable = False
801
if (carry_over_possible and
802
parent_entry.executable == entry.executable):
803
# Check the file length, content hash after reading
805
nostore_sha = parent_entry.text_sha1
808
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
810
text = file_obj.read()
814
entry.text_sha1, entry.text_size = self._add_text_to_weave(
815
file_id, text, heads, nostore_sha)
816
yield file_id, change[1][1], (entry.text_sha1, stat_value)
817
except errors.ExistingContent:
818
# No content change against a carry_over parent
819
# Perhaps this should also yield a fs hash update?
821
entry.text_size = parent_entry.text_size
822
entry.text_sha1 = parent_entry.text_sha1
823
elif kind == 'symlink':
825
entry.symlink_target = tree.get_symlink_target(file_id)
826
if (carry_over_possible and
827
parent_entry.symlink_target == entry.symlink_target):
830
self._add_text_to_weave(change[0], '', heads, None)
831
elif kind == 'directory':
832
if carry_over_possible:
835
# Nothing to set on the entry.
836
# XXX: split into the Root and nonRoot versions.
837
if change[1][1] != '' or self.repository.supports_rich_root():
838
self._add_text_to_weave(change[0], '', heads, None)
839
elif kind == 'tree-reference':
840
if not self.repository._format.supports_tree_reference:
841
# This isn't quite sane as an error, but we shouldn't
842
# ever see this code path in practice: tree's don't
843
# permit references when the repo doesn't support tree
845
raise errors.UnsupportedOperation(tree.add_reference,
847
reference_revision = tree.get_reference_revision(change[0])
848
entry.reference_revision = reference_revision
849
if (carry_over_possible and
850
parent_entry.reference_revision == reference_revision):
853
self._add_text_to_weave(change[0], '', heads, None)
855
raise AssertionError('unknown kind %r' % kind)
857
entry.revision = modified_rev
859
entry.revision = parent_entry.revision
862
new_path = change[1][1]
863
inv_delta.append((change[1][0], new_path, change[0], entry))
866
self.new_inventory = None
868
# This should perhaps be guarded by a check that the basis we
869
# commit against is the basis for the commit and if not do a delta
871
self._any_changes = True
873
# housekeeping root entry changes do not affect no-change commits.
874
self._require_root_change(tree)
875
self.basis_delta_revision = basis_revision_id
877
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
878
parent_keys = tuple([(file_id, parent) for parent in parents])
879
return self.repository.texts._add_text(
880
(file_id, self._new_revision_id), parent_keys, new_text,
881
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
884
class RootCommitBuilder(CommitBuilder):
885
"""This commitbuilder actually records the root id"""
887
# the root entry gets versioned properly by this builder.
888
_versioned_root = True
890
def _check_root(self, ie, parent_invs, tree):
891
"""Helper for record_entry_contents.
893
:param ie: An entry being added.
894
:param parent_invs: The inventories of the parent revisions of the
896
:param tree: The tree that is being committed.
899
def _require_root_change(self, tree):
900
"""Enforce an appropriate root object change.
902
This is called once when record_iter_changes is called, if and only if
903
the root was not in the delta calculated by record_iter_changes.
905
:param tree: The tree which is being committed.
907
# versioned roots do not change unless the tree found a change.
910
class RepositoryWriteLockResult(LogicalLockResult):
911
"""The result of write locking a repository.
913
:ivar repository_token: The token obtained from the underlying lock, or
915
:ivar unlock: A callable which will unlock the lock.
918
def __init__(self, unlock, repository_token):
919
LogicalLockResult.__init__(self, unlock)
920
self.repository_token = repository_token
923
return "RepositoryWriteLockResult(%s, %s)" % (self.repository_token,
927
63
######################################################################
931
class Repository(_RelockDebugMixin, controldir.ControlComponent):
66
class Repository(object):
932
67
"""Repository holding history for one or more branches.
934
69
The repository holds and retrieves historical information including
935
70
revisions and file history. It's normally accessed only by the Branch,
936
71
which views a particular line of development through that history.
938
The Repository builds on top of some byte storage facilies (the revisions,
939
signatures, inventories, texts and chk_bytes attributes) and a Transport,
940
which respectively provide byte storage and a means to access the (possibly
73
The Repository builds on top of Stores and a Transport, which respectively
74
describe the disk data format and the way of accessing the (possibly
943
The byte storage facilities are addressed via tuples, which we refer to
944
as 'keys' throughout the code base. Revision_keys, inventory_keys and
945
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
946
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
947
byte string made up of a hash identifier and a hash value.
948
We use this interface because it allows low friction with the underlying
949
code that implements disk indices, network encoding and other parts of
952
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
953
the serialised revisions for the repository. This can be used to obtain
954
revision graph information or to access raw serialised revisions.
955
The result of trying to insert data into the repository via this store
956
is undefined: it should be considered read-only except for implementors
958
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
959
the serialised signatures for the repository. This can be used to
960
obtain access to raw serialised signatures. The result of trying to
961
insert data into the repository via this store is undefined: it should
962
be considered read-only except for implementors of repositories.
963
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
964
the serialised inventories for the repository. This can be used to
965
obtain unserialised inventories. The result of trying to insert data
966
into the repository via this store is undefined: it should be
967
considered read-only except for implementors of repositories.
968
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
969
texts of files and directories for the repository. This can be used to
970
obtain file texts or file graphs. Note that Repository.iter_file_bytes
971
is usually a better interface for accessing file texts.
972
The result of trying to insert data into the repository via this store
973
is undefined: it should be considered read-only except for implementors
975
:ivar chk_bytes: A bzrlib.versionedfile.VersionedFiles instance containing
976
any data the repository chooses to store or have indexed by its hash.
977
The result of trying to insert data into the repository via this store
978
is undefined: it should be considered read-only except for implementors
980
:ivar _transport: Transport for file access to repository, typically
981
pointing to .bzr/repository.
984
# What class to use for a CommitBuilder. Often it's simpler to change this
985
# in a Repository class subclass rather than to override
986
# get_commit_builder.
987
_commit_builder_class = CommitBuilder
989
def abort_write_group(self, suppress_errors=False):
990
"""Commit the contents accrued within the current write group.
992
:param suppress_errors: if true, abort_write_group will catch and log
993
unexpected errors that happen during the abort, rather than
994
allowing them to propagate. Defaults to False.
996
:seealso: start_write_group.
998
if self._write_group is not self.get_transaction():
999
# has an unlock or relock occured ?
1002
'(suppressed) mismatched lock context and write group. %r, %r',
1003
self._write_group, self.get_transaction())
1005
raise errors.BzrError(
1006
'mismatched lock context and write group. %r, %r' %
1007
(self._write_group, self.get_transaction()))
1009
self._abort_write_group()
1010
except Exception, exc:
1011
self._write_group = None
1012
if not suppress_errors:
1014
mutter('abort_write_group failed')
1015
log_exception_quietly()
1016
note('bzr: ERROR (ignored): %s', exc)
1017
self._write_group = None
1019
def _abort_write_group(self):
1020
"""Template method for per-repository write group cleanup.
1022
This is called during abort before the write group is considered to be
1023
finished and should cleanup any internal state accrued during the write
1024
group. There is no requirement that data handed to the repository be
1025
*not* made available - this is not a rollback - but neither should any
1026
attempt be made to ensure that data added is fully commited. Abort is
1027
invoked when an error has occured so futher disk or network operations
1028
may not be possible or may error and if possible should not be
1032
def add_fallback_repository(self, repository):
1033
"""Add a repository to use for looking up data not held locally.
1035
:param repository: A repository.
1037
if not self._format.supports_external_lookups:
1038
raise errors.UnstackableRepositoryFormat(self._format, self.base)
1039
if self.is_locked():
1040
# This repository will call fallback.unlock() when we transition to
1041
# the unlocked state, so we make sure to increment the lock count
1042
repository.lock_read()
1043
self._check_fallback_repository(repository)
1044
self._fallback_repositories.append(repository)
1045
self.texts.add_fallback_versioned_files(repository.texts)
1046
self.inventories.add_fallback_versioned_files(repository.inventories)
1047
self.revisions.add_fallback_versioned_files(repository.revisions)
1048
self.signatures.add_fallback_versioned_files(repository.signatures)
1049
if self.chk_bytes is not None:
1050
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
1052
def _check_fallback_repository(self, repository):
1053
"""Check that this repository can fallback to repository safely.
1055
Raise an error if not.
1057
:param repository: A repository to fallback to.
1059
return InterRepository._assert_same_model(self, repository)
78
_file_ids_altered_regex = lazy_regex.lazy_compile(
79
r'file_id="(?P<file_id>[^"]+)"'
80
r'.*revision="(?P<revision_id>[^"]+)"'
1061
84
def add_inventory(self, revision_id, inv, parents):
1062
85
"""Add the inventory inv to the repository as revision_id.
1064
87
:param parents: The revision ids of the parents that revision_id
1065
88
is known to have and are in the repository already.
1067
:returns: The validator(which is a sha1 digest, though what is sha'd is
1068
repository format specific) of the serialized inventory.
90
returns the sha1 of the serialized inventory.
1070
if not self.is_in_write_group():
1071
raise AssertionError("%r not in write group" % (self,))
92
revision_id = osutils.safe_revision_id(revision_id)
1072
93
_mod_revision.check_not_reserved_id(revision_id)
1073
if not (inv.revision_id is None or inv.revision_id == revision_id):
1074
raise AssertionError(
1075
"Mismatch between inventory revision"
1076
" id and insertion revid (%r, %r)"
1077
% (inv.revision_id, revision_id))
1078
if inv.root is None:
1079
raise errors.RootMissing()
1080
return self._add_inventory_checked(revision_id, inv, parents)
1082
def _add_inventory_checked(self, revision_id, inv, parents):
1083
"""Add inv to the repository after checking the inputs.
1085
This function can be overridden to allow different inventory styles.
1087
:seealso: add_inventory, for the contract.
1089
inv_lines = self._serializer.write_inventory_to_lines(inv)
1090
return self._inventory_add_lines(revision_id, parents,
1091
inv_lines, check_content=False)
1093
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1094
parents, basis_inv=None, propagate_caches=False):
1095
"""Add a new inventory expressed as a delta against another revision.
1097
See the inventory developers documentation for the theory behind
1100
:param basis_revision_id: The inventory id the delta was created
1101
against. (This does not have to be a direct parent.)
1102
:param delta: The inventory delta (see Inventory.apply_delta for
1104
:param new_revision_id: The revision id that the inventory is being
1106
:param parents: The revision ids of the parents that revision_id is
1107
known to have and are in the repository already. These are supplied
1108
for repositories that depend on the inventory graph for revision
1109
graph access, as well as for those that pun ancestry with delta
1111
:param basis_inv: The basis inventory if it is already known,
1113
:param propagate_caches: If True, the caches for this inventory are
1114
copied to and updated for the result if possible.
1116
:returns: (validator, new_inv)
1117
The validator(which is a sha1 digest, though what is sha'd is
1118
repository format specific) of the serialized inventory, and the
1119
resulting inventory.
1121
if not self.is_in_write_group():
1122
raise AssertionError("%r not in write group" % (self,))
1123
_mod_revision.check_not_reserved_id(new_revision_id)
1124
basis_tree = self.revision_tree(basis_revision_id)
1125
basis_tree.lock_read()
1127
# Note that this mutates the inventory of basis_tree, which not all
1128
# inventory implementations may support: A better idiom would be to
1129
# return a new inventory, but as there is no revision tree cache in
1130
# repository this is safe for now - RBC 20081013
1131
if basis_inv is None:
1132
basis_inv = basis_tree.inventory
1133
basis_inv.apply_delta(delta)
1134
basis_inv.revision_id = new_revision_id
1135
return (self.add_inventory(new_revision_id, basis_inv, parents),
1140
def _inventory_add_lines(self, revision_id, parents, lines,
1141
check_content=True):
1142
"""Store lines in inv_vf and return the sha1 of the inventory."""
1143
parents = [(parent,) for parent in parents]
1144
result = self.inventories.add_lines((revision_id,), parents, lines,
1145
check_content=check_content)[0]
1146
self.inventories._access.flush()
94
assert inv.revision_id is None or inv.revision_id == revision_id, \
95
"Mismatch between inventory revision" \
96
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
97
assert inv.root is not None
98
inv_text = self.serialise_inventory(inv)
99
inv_sha1 = osutils.sha_string(inv_text)
100
inv_vf = self.control_weaves.get_weave('inventory',
101
self.get_transaction())
102
self._inventory_add_lines(inv_vf, revision_id, parents,
103
osutils.split_lines(inv_text))
106
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines):
108
for parent in parents:
110
final_parents.append(parent)
112
inv_vf.add_lines(revision_id, final_parents, lines)
1149
115
def add_revision(self, revision_id, rev, inv=None, config=None):
1150
116
"""Add rev to the revision store as revision_id.
1223
187
Returns a set of the present revisions.
1226
graph = self.get_graph()
1227
parent_map = graph.get_parent_map(revision_ids)
1228
# The old API returned a list, should this actually be a set?
1229
return parent_map.keys()
1231
def _check_inventories(self, checker):
1232
"""Check the inventories found from the revision scan.
1234
This is responsible for verifying the sha1 of inventories and
1235
creating a pending_keys set that covers data referenced by inventories.
1237
bar = ui.ui_factory.nested_progress_bar()
1239
self._do_check_inventories(checker, bar)
1243
def _do_check_inventories(self, checker, bar):
1244
"""Helper for _check_inventories."""
1246
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1247
kinds = ['chk_bytes', 'texts']
1248
count = len(checker.pending_keys)
1249
bar.update("inventories", 0, 2)
1250
current_keys = checker.pending_keys
1251
checker.pending_keys = {}
1252
# Accumulate current checks.
1253
for key in current_keys:
1254
if key[0] != 'inventories' and key[0] not in kinds:
1255
checker._report_items.append('unknown key type %r' % (key,))
1256
keys[key[0]].add(key[1:])
1257
if keys['inventories']:
1258
# NB: output order *should* be roughly sorted - topo or
1259
# inverse topo depending on repository - either way decent
1260
# to just delta against. However, pre-CHK formats didn't
1261
# try to optimise inventory layout on disk. As such the
1262
# pre-CHK code path does not use inventory deltas.
1264
for record in self.inventories.check(keys=keys['inventories']):
1265
if record.storage_kind == 'absent':
1266
checker._report_items.append(
1267
'Missing inventory {%s}' % (record.key,))
1269
last_object = self._check_record('inventories', record,
1270
checker, last_object,
1271
current_keys[('inventories',) + record.key])
1272
del keys['inventories']
1275
bar.update("texts", 1)
1276
while (checker.pending_keys or keys['chk_bytes']
1278
# Something to check.
1279
current_keys = checker.pending_keys
1280
checker.pending_keys = {}
1281
# Accumulate current checks.
1282
for key in current_keys:
1283
if key[0] not in kinds:
1284
checker._report_items.append('unknown key type %r' % (key,))
1285
keys[key[0]].add(key[1:])
1286
# Check the outermost kind only - inventories || chk_bytes || texts
1290
for record in getattr(self, kind).check(keys=keys[kind]):
1291
if record.storage_kind == 'absent':
1292
checker._report_items.append(
1293
'Missing %s {%s}' % (kind, record.key,))
1295
last_object = self._check_record(kind, record,
1296
checker, last_object, current_keys[(kind,) + record.key])
1300
def _check_record(self, kind, record, checker, last_object, item_data):
1301
"""Check a single text from this repository."""
1302
if kind == 'inventories':
1303
rev_id = record.key[0]
1304
inv = self._deserialise_inventory(rev_id,
1305
record.get_bytes_as('fulltext'))
1306
if last_object is not None:
1307
delta = inv._make_delta(last_object)
1308
for old_path, path, file_id, ie in delta:
1311
ie.check(checker, rev_id, inv)
1313
for path, ie in inv.iter_entries():
1314
ie.check(checker, rev_id, inv)
1315
if self._format.fast_deltas:
1317
elif kind == 'chk_bytes':
1318
# No code written to check chk_bytes for this repo format.
1319
checker._report_items.append(
1320
'unsupported key type chk_bytes for %s' % (record.key,))
1321
elif kind == 'texts':
1322
self._check_text(record, checker, item_data)
1324
checker._report_items.append(
1325
'unknown key type %s for %s' % (kind, record.key))
1327
def _check_text(self, record, checker, item_data):
1328
"""Check a single text."""
1329
# Check it is extractable.
1330
# TODO: check length.
1331
if record.storage_kind == 'chunked':
1332
chunks = record.get_bytes_as(record.storage_kind)
1333
sha1 = osutils.sha_strings(chunks)
1334
length = sum(map(len, chunks))
1336
content = record.get_bytes_as('fulltext')
1337
sha1 = osutils.sha_string(content)
1338
length = len(content)
1339
if item_data and sha1 != item_data[1]:
1340
checker._report_items.append(
1341
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1342
(record.key, sha1, item_data[1], item_data[2]))
190
for id in revision_ids:
191
if self.has_revision(id):
1345
196
def create(a_bzrdir):
1346
197
"""Construct the current default format repository in a_bzrdir."""
1347
198
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
1349
def __init__(self, _format, a_bzrdir, control_files):
200
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1350
201
"""instantiate a Repository.
1352
203
:param _format: The format of the repository on disk.
1353
204
:param a_bzrdir: The BzrDir of the repository.
206
In the future we will have a single api for all stores for
207
getting file texts, inventories and revisions, then
208
this construct will accept instances of those things.
1355
# In the future we will have a single api for all stores for
1356
# getting file texts, inventories and revisions, then
1357
# this construct will accept instances of those things.
1358
210
super(Repository, self).__init__()
1359
211
self._format = _format
1360
212
# the following are part of the public API for Repository:
1361
213
self.bzrdir = a_bzrdir
1362
214
self.control_files = control_files
1363
self._transport = control_files._transport
1364
self.base = self._transport.base
1366
self._reconcile_does_inventory_gc = True
1367
self._reconcile_fixes_text_parents = False
1368
self._reconcile_backsup_inventory = True
1369
self._write_group = None
1370
# Additional places to query for data.
1371
self._fallback_repositories = []
1372
# An InventoryEntry cache, used during deserialization
1373
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1374
# Is it safe to return inventory entries directly from the entry cache,
1375
# rather copying them?
1376
self._safe_to_return_from_cache = False
1379
def user_transport(self):
1380
return self.bzrdir.user_transport
1383
def control_transport(self):
1384
return self._transport
215
self._revision_store = _revision_store
216
self.text_store = text_store
217
# backwards compatibility
218
self.weave_store = text_store
219
# not right yet - should be more semantically clear ?
221
self.control_store = control_store
222
self.control_weaves = control_store
223
# TODO: make sure to construct the right store classes, etc, depending
224
# on whether escaping is required.
225
self._warn_if_deprecated()
1386
227
def __repr__(self):
1387
if self._fallback_repositories:
1388
return '%s(%r, fallback_repositories=%r)' % (
1389
self.__class__.__name__,
1391
self._fallback_repositories)
1393
return '%s(%r)' % (self.__class__.__name__,
1396
def _has_same_fallbacks(self, other_repo):
1397
"""Returns true if the repositories have the same fallbacks."""
1398
my_fb = self._fallback_repositories
1399
other_fb = other_repo._fallback_repositories
1400
if len(my_fb) != len(other_fb):
1402
for f, g in zip(my_fb, other_fb):
1403
if not f.has_same_location(g):
1407
def has_same_location(self, other):
1408
"""Returns a boolean indicating if this repository is at the same
1409
location as another repository.
1411
This might return False even when two repository objects are accessing
1412
the same physical repository via different URLs.
1414
if self.__class__ is not other.__class__:
1416
return (self._transport.base == other._transport.base)
1418
def is_in_write_group(self):
1419
"""Return True if there is an open write group.
1421
:seealso: start_write_group.
1423
return self._write_group is not None
228
return '%s(%r)' % (self.__class__.__name__,
229
self.bzrdir.transport.base)
1425
231
def is_locked(self):
1426
232
return self.control_files.is_locked()
1428
def is_write_locked(self):
1429
"""Return True if this object is write locked."""
1430
return self.is_locked() and self.control_files._lock_mode == 'w'
1432
234
def lock_write(self, token=None):
1433
235
"""Lock this repository for writing.
1435
This causes caching within the repository obejct to start accumlating
1436
data during reads, and allows a 'write_group' to be obtained. Write
1437
groups must be used for actual data insertion.
1439
A token should be passed in if you know that you have locked the object
1440
some other way, and need to synchronise this object's state with that
1443
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1445
237
:param token: if this is already locked, then lock_write will fail
1446
238
unless the token matches the existing lock.
1447
239
:returns: a token if this instance supports tokens, otherwise None.
1888
401
# Most control formats need the repository to be specifically
1889
402
# created, but on some old all-in-one formats it's not needed
1891
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
404
dest_repo = self._format.initialize(a_bzrdir, shared=self.is_shared())
1892
405
except errors.UninitializableFormat:
1893
406
dest_repo = a_bzrdir.open_repository()
407
self.copy_content_into(dest_repo, revision_id)
1894
408
return dest_repo
1896
def _get_sink(self):
1897
"""Return a sink for streaming into this repository."""
1898
return StreamSink(self)
1900
def _get_source(self, to_format):
1901
"""Return a source for streaming from this repository."""
1902
return StreamSource(self, to_format)
1904
410
@needs_read_lock
1905
411
def has_revision(self, revision_id):
1906
412
"""True if this repository has a copy of the revision."""
1907
return revision_id in self.has_revisions((revision_id,))
1910
def has_revisions(self, revision_ids):
1911
"""Probe to find out the presence of multiple revisions.
1913
:param revision_ids: An iterable of revision_ids.
1914
:return: A set of the revision_ids that were present.
1916
parent_map = self.revisions.get_parent_map(
1917
[(rev_id,) for rev_id in revision_ids])
1919
if _mod_revision.NULL_REVISION in revision_ids:
1920
result.add(_mod_revision.NULL_REVISION)
1921
result.update([key[0] for key in parent_map])
1925
def get_revision(self, revision_id):
1926
"""Return the Revision object for a named revision."""
1927
return self.get_revisions([revision_id])[0]
413
revision_id = osutils.safe_revision_id(revision_id)
414
return self._revision_store.has_revision_id(revision_id,
415
self.get_transaction())
1929
417
@needs_read_lock
1930
418
def get_revision_reconcile(self, revision_id):
1931
419
"""'reconcile' helper routine that allows access to a revision always.
1933
421
This variant of get_revision does not cross check the weave graph
1934
422
against the revision one as get_revision does: but it should only
1935
423
be used by reconcile, or reconcile-alike commands that are correcting
1936
424
or testing the revision graph.
1938
return self._get_revisions([revision_id])[0]
426
if not revision_id or not isinstance(revision_id, basestring):
427
raise errors.InvalidRevisionId(revision_id=revision_id,
429
return self.get_revisions([revision_id])[0]
1940
431
@needs_read_lock
1941
432
def get_revisions(self, revision_ids):
1942
"""Get many revisions at once.
1944
Repositories that need to check data on every revision read should
1945
subclass this method.
1947
return self._get_revisions(revision_ids)
1950
def _get_revisions(self, revision_ids):
1951
"""Core work logic to get many revisions without sanity checks."""
1953
for revid, rev in self._iter_revisions(revision_ids):
1955
raise errors.NoSuchRevision(self, revid)
1957
return [revs[revid] for revid in revision_ids]
1959
def _iter_revisions(self, revision_ids):
1960
"""Iterate over revision objects.
1962
:param revision_ids: An iterable of revisions to examine. None may be
1963
passed to request all revisions known to the repository. Note that
1964
not all repositories can find unreferenced revisions; for those
1965
repositories only referenced ones will be returned.
1966
:return: An iterator of (revid, revision) tuples. Absent revisions (
1967
those asked for but not available) are returned as (revid, None).
1969
if revision_ids is None:
1970
revision_ids = self.all_revision_ids()
1972
for rev_id in revision_ids:
1973
if not rev_id or not isinstance(rev_id, basestring):
1974
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1975
keys = [(key,) for key in revision_ids]
1976
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1977
for record in stream:
1978
revid = record.key[0]
1979
if record.storage_kind == 'absent':
1982
text = record.get_bytes_as('fulltext')
1983
rev = self._serializer.read_revision_from_string(text)
1986
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
433
revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
434
revs = self._revision_store.get_revisions(revision_ids,
435
self.get_transaction())
437
assert not isinstance(rev.revision_id, unicode)
438
for parent_id in rev.parent_ids:
439
assert not isinstance(parent_id, unicode)
443
def get_revision_xml(self, revision_id):
444
# TODO: jam 20070210 This shouldn't be necessary since get_revision
445
# would have already do it.
446
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
447
revision_id = osutils.safe_revision_id(revision_id)
448
rev = self.get_revision(revision_id)
450
# the current serializer..
451
self._revision_store._serializer.write_revision(rev, rev_tmp)
453
return rev_tmp.getvalue()
456
def get_revision(self, revision_id):
457
"""Return the Revision object for a named revision"""
458
# TODO: jam 20070210 get_revision_reconcile should do this for us
459
revision_id = osutils.safe_revision_id(revision_id)
460
r = self.get_revision_reconcile(revision_id)
461
# weave corruption can lead to absent revision markers that should be
463
# the following test is reasonably cheap (it needs a single weave read)
464
# and the weave is cached in read transactions. In write transactions
465
# it is not cached but typically we only read a small number of
466
# revisions. For knits when they are introduced we will probably want
467
# to ensure that caching write transactions are in use.
468
inv = self.get_inventory_weave()
469
self._check_revision_parents(r, inv)
473
def get_deltas_for_revisions(self, revisions):
1987
474
"""Produce a generator of revision deltas.
1989
476
Note that the input is a sequence of REVISIONS, not revision_ids.
1990
477
Trees will be held in memory until the generator exits.
1991
478
Each delta is relative to the revision's lefthand predecessor.
1993
:param specific_fileids: if not None, the result is filtered
1994
so that only those file-ids, their parents and their
1995
children are included.
1997
# Get the revision-ids of interest
1998
480
required_trees = set()
1999
481
for revision in revisions:
2000
482
required_trees.add(revision.revision_id)
2001
483
required_trees.update(revision.parent_ids[:1])
2003
# Get the matching filtered trees. Note that it's more
2004
# efficient to pass filtered trees to changes_from() rather
2005
# than doing the filtering afterwards. changes_from() could
2006
# arguably do the filtering itself but it's path-based, not
2007
# file-id based, so filtering before or afterwards is
2009
if specific_fileids is None:
2010
trees = dict((t.get_revision_id(), t) for
2011
t in self.revision_trees(required_trees))
2013
trees = dict((t.get_revision_id(), t) for
2014
t in self._filtered_revision_trees(required_trees,
2017
# Calculate the deltas
484
trees = dict((t.get_revision_id(), t) for
485
t in self.revision_trees(required_trees))
2018
486
for revision in revisions:
2019
487
if not revision.parent_ids:
2020
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
488
old_tree = self.revision_tree(None)
2022
490
old_tree = trees[revision.parent_ids[0]]
2023
491
yield trees[revision.revision_id].changes_from(old_tree)
2025
493
@needs_read_lock
2026
def get_revision_delta(self, revision_id, specific_fileids=None):
494
def get_revision_delta(self, revision_id):
2027
495
"""Return the delta for one revision.
2029
497
The delta is relative to the left-hand predecessor of the
2032
:param specific_fileids: if not None, the result is filtered
2033
so that only those file-ids, their parents and their
2034
children are included.
2036
500
r = self.get_revision(revision_id)
2037
return list(self.get_deltas_for_revisions([r],
2038
specific_fileids=specific_fileids))[0]
501
return list(self.get_deltas_for_revisions([r]))[0]
503
def _check_revision_parents(self, revision, inventory):
504
"""Private to Repository and Fetch.
506
This checks the parentage of revision in an inventory weave for
507
consistency and is only applicable to inventory-weave-for-ancestry
508
using repository formats & fetchers.
510
weave_parents = inventory.get_parents(revision.revision_id)
511
weave_names = inventory.versions()
512
for parent_id in revision.parent_ids:
513
if parent_id in weave_names:
514
# this parent must not be a ghost.
515
if not parent_id in weave_parents:
517
raise errors.CorruptRepository(self)
2040
519
@needs_write_lock
2041
520
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
521
revision_id = osutils.safe_revision_id(revision_id)
2042
522
signature = gpg_strategy.sign(plaintext)
2043
self.add_signature_text(revision_id, signature)
2046
def add_signature_text(self, revision_id, signature):
2047
self.signatures.add_lines((revision_id,), (),
2048
osutils.split_lines(signature))
2050
def find_text_key_references(self):
2051
"""Find the text key references within the repository.
2053
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2054
to whether they were referred to by the inventory of the
2055
revision_id that they contain. The inventory texts from all present
2056
revision ids are assessed to generate this report.
2058
revision_keys = self.revisions.keys()
2059
w = self.inventories
2060
pb = ui.ui_factory.nested_progress_bar()
2062
return self._serializer._find_text_key_references(
2063
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
2067
def _inventory_xml_lines_for_keys(self, keys):
2068
"""Get a line iterator of the sort needed for findind references.
2070
Not relevant for non-xml inventory repositories.
2072
Ghosts in revision_keys are ignored.
2074
:param revision_keys: The revision keys for the inventories to inspect.
2075
:return: An iterator over (inventory line, revid) for the fulltexts of
2076
all of the xml inventories specified by revision_keys.
2078
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2079
for record in stream:
2080
if record.storage_kind != 'absent':
2081
chunks = record.get_bytes_as('chunked')
2082
revid = record.key[-1]
2083
lines = osutils.chunks_to_lines(chunks)
2087
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2089
"""Helper routine for fileids_altered_by_revision_ids.
2091
This performs the translation of xml lines to revision ids.
2093
:param line_iterator: An iterator of lines, origin_version_id
2094
:param revision_keys: The revision ids to filter for. This should be a
2095
set or other type which supports efficient __contains__ lookups, as
2096
the revision key from each parsed line will be looked up in the
2097
revision_keys filter.
2098
:return: a dictionary mapping altered file-ids to an iterable of
2099
revision_ids. Each altered file-ids has the exact revision_ids that
2100
altered it listed explicitly.
2102
seen = set(self._serializer._find_text_key_references(
2103
line_iterator).iterkeys())
2104
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2105
parent_seen = set(self._serializer._find_text_key_references(
2106
self._inventory_xml_lines_for_keys(parent_keys)))
2107
new_keys = seen - parent_seen
2109
setdefault = result.setdefault
2110
for key in new_keys:
2111
setdefault(key[0], set()).add(key[-1])
2114
def _find_parent_ids_of_revisions(self, revision_ids):
2115
"""Find all parent ids that are mentioned in the revision graph.
2117
:return: set of revisions that are parents of revision_ids which are
2118
not part of revision_ids themselves
2120
parent_map = self.get_parent_map(revision_ids)
2122
map(parent_ids.update, parent_map.itervalues())
2123
parent_ids.difference_update(revision_ids)
2124
parent_ids.discard(_mod_revision.NULL_REVISION)
2127
def _find_parent_keys_of_revisions(self, revision_keys):
2128
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2130
:param revision_keys: An iterable of revision_keys.
2131
:return: The parents of all revision_keys that are not already in
2134
parent_map = self.revisions.get_parent_map(revision_keys)
2136
map(parent_keys.update, parent_map.itervalues())
2137
parent_keys.difference_update(revision_keys)
2138
parent_keys.discard(_mod_revision.NULL_REVISION)
2141
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
523
self._revision_store.add_revision_signature_text(revision_id,
525
self.get_transaction())
527
def fileids_altered_by_revision_ids(self, revision_ids):
2142
528
"""Find the file ids and versions affected by revisions.
2144
530
:param revisions: an iterable containing revision ids.
2145
:param _inv_weave: The inventory weave from this repository or None.
2146
If None, the inventory weave will be opened automatically.
2147
531
:return: a dictionary mapping altered file-ids to an iterable of
2148
532
revision_ids. Each altered file-ids has the exact revision_ids that
2149
533
altered it listed explicitly.
2151
selected_keys = set((revid,) for revid in revision_ids)
2152
w = _inv_weave or self.inventories
2153
return self._find_file_ids_from_xml_inventory_lines(
2154
w.iter_lines_added_or_present_in_keys(
2155
selected_keys, pb=None),
2158
def iter_files_bytes(self, desired_files):
2159
"""Iterate through file versions.
2161
Files will not necessarily be returned in the order they occur in
2162
desired_files. No specific order is guaranteed.
2164
Yields pairs of identifier, bytes_iterator. identifier is an opaque
2165
value supplied by the caller as part of desired_files. It should
2166
uniquely identify the file version in the caller's context. (Examples:
2167
an index number or a TreeTransform trans_id.)
2169
bytes_iterator is an iterable of bytestrings for the file. The
2170
kind of iterable and length of the bytestrings are unspecified, but for
2171
this implementation, it is a list of bytes produced by
2172
VersionedFile.get_record_stream().
2174
:param desired_files: a list of (file_id, revision_id, identifier)
2178
for file_id, revision_id, callable_data in desired_files:
2179
text_keys[(file_id, revision_id)] = callable_data
2180
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2181
if record.storage_kind == 'absent':
2182
raise errors.RevisionNotPresent(record.key, self)
2183
yield text_keys[record.key], record.get_bytes_as('chunked')
2185
def _generate_text_key_index(self, text_key_references=None,
2187
"""Generate a new text key index for the repository.
2189
This is an expensive function that will take considerable time to run.
2191
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2192
list of parents, also text keys. When a given key has no parents,
2193
the parents list will be [NULL_REVISION].
2195
# All revisions, to find inventory parents.
2196
if ancestors is None:
2197
graph = self.get_graph()
2198
ancestors = graph.get_parent_map(self.all_revision_ids())
2199
if text_key_references is None:
2200
text_key_references = self.find_text_key_references()
535
assert self._serializer.support_altered_by_hack, \
536
("fileids_altered_by_revision_ids only supported for branches "
537
"which store inventory as unnested xml, not on %r" % self)
538
selected_revision_ids = set(osutils.safe_revision_id(r)
539
for r in revision_ids)
540
w = self.get_inventory_weave()
543
# this code needs to read every new line in every inventory for the
544
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
545
# not present in one of those inventories is unnecessary but not
546
# harmful because we are filtering by the revision id marker in the
547
# inventory lines : we only select file ids altered in one of those
548
# revisions. We don't need to see all lines in the inventory because
549
# only those added in an inventory in rev X can contain a revision=X
551
unescape_revid_cache = {}
552
unescape_fileid_cache = {}
554
# jam 20061218 In a big fetch, this handles hundreds of thousands
555
# of lines, so it has had a lot of inlining and optimizing done.
556
# Sorry that it is a little bit messy.
557
# Move several functions to be local variables, since this is a long
559
search = self._file_ids_altered_regex.search
560
unescape = _unescape_xml
561
setdefault = result.setdefault
2201
562
pb = ui.ui_factory.nested_progress_bar()
2203
return self._do_generate_text_key_index(ancestors,
2204
text_key_references, pb)
564
for line in w.iter_lines_added_or_present_in_versions(
565
selected_revision_ids, pb=pb):
569
# One call to match.group() returning multiple items is quite a
570
# bit faster than 2 calls to match.group() each returning 1
571
file_id, revision_id = match.group('file_id', 'revision_id')
573
# Inlining the cache lookups helps a lot when you make 170,000
574
# lines and 350k ids, versus 8.4 unique ids.
575
# Using a cache helps in 2 ways:
576
# 1) Avoids unnecessary decoding calls
577
# 2) Re-uses cached strings, which helps in future set and
579
# (2) is enough that removing encoding entirely along with
580
# the cache (so we are using plain strings) results in no
581
# performance improvement.
583
revision_id = unescape_revid_cache[revision_id]
585
unescaped = unescape(revision_id)
586
unescape_revid_cache[revision_id] = unescaped
587
revision_id = unescaped
589
if revision_id in selected_revision_ids:
591
file_id = unescape_fileid_cache[file_id]
593
unescaped = unescape(file_id)
594
unescape_fileid_cache[file_id] = unescaped
596
setdefault(file_id, set()).add(revision_id)
2208
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2209
"""Helper for _generate_text_key_index to avoid deep nesting."""
2210
revision_order = tsort.topo_sort(ancestors)
2211
invalid_keys = set()
2213
for revision_id in revision_order:
2214
revision_keys[revision_id] = set()
2215
text_count = len(text_key_references)
2216
# a cache of the text keys to allow reuse; costs a dict of all the
2217
# keys, but saves a 2-tuple for every child of a given key.
2219
for text_key, valid in text_key_references.iteritems():
2221
invalid_keys.add(text_key)
2223
revision_keys[text_key[1]].add(text_key)
2224
text_key_cache[text_key] = text_key
2225
del text_key_references
2227
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2228
NULL_REVISION = _mod_revision.NULL_REVISION
2229
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2230
# too small for large or very branchy trees. However, for 55K path
2231
# trees, it would be easy to use too much memory trivially. Ideally we
2232
# could gauge this by looking at available real memory etc, but this is
2233
# always a tricky proposition.
2234
inventory_cache = lru_cache.LRUCache(10)
2235
batch_size = 10 # should be ~150MB on a 55K path tree
2236
batch_count = len(revision_order) / batch_size + 1
2238
pb.update("Calculating text parents", processed_texts, text_count)
2239
for offset in xrange(batch_count):
2240
to_query = revision_order[offset * batch_size:(offset + 1) *
2244
for revision_id in to_query:
2245
parent_ids = ancestors[revision_id]
2246
for text_key in revision_keys[revision_id]:
2247
pb.update("Calculating text parents", processed_texts)
2248
processed_texts += 1
2249
candidate_parents = []
2250
for parent_id in parent_ids:
2251
parent_text_key = (text_key[0], parent_id)
2253
check_parent = parent_text_key not in \
2254
revision_keys[parent_id]
2256
# the parent parent_id is a ghost:
2257
check_parent = False
2258
# truncate the derived graph against this ghost.
2259
parent_text_key = None
2261
# look at the parent commit details inventories to
2262
# determine possible candidates in the per file graph.
2265
inv = inventory_cache[parent_id]
2267
inv = self.revision_tree(parent_id).inventory
2268
inventory_cache[parent_id] = inv
2270
parent_entry = inv[text_key[0]]
2271
except (KeyError, errors.NoSuchId):
2273
if parent_entry is not None:
2275
text_key[0], parent_entry.revision)
2277
parent_text_key = None
2278
if parent_text_key is not None:
2279
candidate_parents.append(
2280
text_key_cache[parent_text_key])
2281
parent_heads = text_graph.heads(candidate_parents)
2282
new_parents = list(parent_heads)
2283
new_parents.sort(key=lambda x:candidate_parents.index(x))
2284
if new_parents == []:
2285
new_parents = [NULL_REVISION]
2286
text_index[text_key] = new_parents
2288
for text_key in invalid_keys:
2289
text_index[text_key] = [NULL_REVISION]
2292
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2293
"""Get an iterable listing the keys of all the data introduced by a set
2296
The keys will be ordered so that the corresponding items can be safely
2297
fetched and inserted in that order.
2299
:returns: An iterable producing tuples of (knit-kind, file-id,
2300
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2301
'revisions'. file-id is None unless knit-kind is 'file'.
2303
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2306
for result in self._find_non_file_keys_to_fetch(revision_ids):
2309
def _find_file_keys_to_fetch(self, revision_ids, pb):
2310
# XXX: it's a bit weird to control the inventory weave caching in this
2311
# generator. Ideally the caching would be done in fetch.py I think. Or
2312
# maybe this generator should explicitly have the contract that it
2313
# should not be iterated until the previously yielded item has been
2315
inv_w = self.inventories
2317
# file ids that changed
2318
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2320
num_file_ids = len(file_ids)
2321
for file_id, altered_versions in file_ids.iteritems():
2323
pb.update("Fetch texts", count, num_file_ids)
2325
yield ("file", file_id, altered_versions)
2327
def _find_non_file_keys_to_fetch(self, revision_ids):
2329
yield ("inventory", None, revision_ids)
2332
# XXX: Note ATM no callers actually pay attention to this return
2333
# instead they just use the list of revision ids and ignore
2334
# missing sigs. Consider removing this work entirely
2335
revisions_with_signatures = set(self.signatures.get_parent_map(
2336
[(r,) for r in revision_ids]))
2337
revisions_with_signatures = set(
2338
[r for (r,) in revisions_with_signatures])
2339
revisions_with_signatures.intersection_update(revision_ids)
2340
yield ("signatures", None, revisions_with_signatures)
2343
yield ("revisions", None, revision_ids)
602
def get_inventory_weave(self):
603
return self.control_weaves.get_weave('inventory',
604
self.get_transaction())
2345
606
@needs_read_lock
2346
607
def get_inventory(self, revision_id):
2347
"""Get Inventory object by revision id."""
2348
return self.iter_inventories([revision_id]).next()
2350
def iter_inventories(self, revision_ids, ordering=None):
2351
"""Get many inventories by revision_ids.
2353
This will buffer some or all of the texts used in constructing the
2354
inventories in memory, but will only parse a single inventory at a
2357
:param revision_ids: The expected revision ids of the inventories.
2358
:param ordering: optional ordering, e.g. 'topological'. If not
2359
specified, the order of revision_ids will be preserved (by
2360
buffering if necessary).
2361
:return: An iterator of inventories.
2363
if ((None in revision_ids)
2364
or (_mod_revision.NULL_REVISION in revision_ids)):
2365
raise ValueError('cannot get null revision inventory')
2366
return self._iter_inventories(revision_ids, ordering)
2368
def _iter_inventories(self, revision_ids, ordering):
2369
"""single-document based inventory iteration."""
2370
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2371
for text, revision_id in inv_xmls:
2372
yield self._deserialise_inventory(revision_id, text)
2374
def _iter_inventory_xmls(self, revision_ids, ordering):
2375
if ordering is None:
2376
order_as_requested = True
2377
ordering = 'unordered'
2379
order_as_requested = False
2380
keys = [(revision_id,) for revision_id in revision_ids]
2383
if order_as_requested:
2384
key_iter = iter(keys)
2385
next_key = key_iter.next()
2386
stream = self.inventories.get_record_stream(keys, ordering, True)
2388
for record in stream:
2389
if record.storage_kind != 'absent':
2390
chunks = record.get_bytes_as('chunked')
2391
if order_as_requested:
2392
text_chunks[record.key] = chunks
2394
yield ''.join(chunks), record.key[-1]
2396
raise errors.NoSuchRevision(self, record.key)
2397
if order_as_requested:
2398
# Yield as many results as we can while preserving order.
2399
while next_key in text_chunks:
2400
chunks = text_chunks.pop(next_key)
2401
yield ''.join(chunks), next_key[-1]
2403
next_key = key_iter.next()
2404
except StopIteration:
2405
# We still want to fully consume the get_record_stream,
2406
# just in case it is not actually finished at this point
2410
def _deserialise_inventory(self, revision_id, xml):
2411
"""Transform the xml into an inventory object.
608
"""Get Inventory object by hash."""
609
# TODO: jam 20070210 Technically we don't need to sanitize, since all
610
# called functions must sanitize.
611
revision_id = osutils.safe_revision_id(revision_id)
612
return self.deserialise_inventory(
613
revision_id, self.get_inventory_xml(revision_id))
615
def deserialise_inventory(self, revision_id, xml):
616
"""Transform the xml into an inventory object.
2413
618
:param revision_id: The expected revision id of the inventory.
2414
619
:param xml: A serialised inventory.
2416
result = self._serializer.read_inventory_from_string(xml, revision_id,
2417
entry_cache=self._inventory_entry_cache,
2418
return_from_cache=self._safe_to_return_from_cache)
2419
if result.revision_id != revision_id:
2420
raise AssertionError('revision id mismatch %s != %s' % (
2421
result.revision_id, revision_id))
621
revision_id = osutils.safe_revision_id(revision_id)
622
result = self._serializer.read_inventory_from_string(xml)
623
result.root.revision = revision_id
2424
def get_serializer_format(self):
2425
return self._serializer.format_num
626
def serialise_inventory(self, inv):
627
return self._serializer.write_inventory_to_string(inv)
2427
629
@needs_read_lock
2428
def _get_inventory_xml(self, revision_id):
2429
"""Get serialized inventory as a string."""
2430
texts = self._iter_inventory_xmls([revision_id], 'unordered')
630
def get_inventory_xml(self, revision_id):
631
"""Get inventory XML as a file object."""
632
revision_id = osutils.safe_revision_id(revision_id)
2432
text, revision_id = texts.next()
2433
except StopIteration:
634
assert isinstance(revision_id, str), type(revision_id)
635
iw = self.get_inventory_weave()
636
return iw.get_text(revision_id)
2434
638
raise errors.HistoryMissing(self, 'inventory', revision_id)
2437
def get_rev_id_for_revno(self, revno, known_pair):
2438
"""Return the revision id of a revno, given a later (revno, revid)
2439
pair in the same history.
2441
:return: if found (True, revid). If the available history ran out
2442
before reaching the revno, then this returns
2443
(False, (closest_revno, closest_revid)).
2445
known_revno, known_revid = known_pair
2446
partial_history = [known_revid]
2447
distance_from_known = known_revno - revno
2448
if distance_from_known < 0:
2450
'requested revno (%d) is later than given known revno (%d)'
2451
% (revno, known_revno))
2454
self, partial_history, stop_index=distance_from_known)
2455
except errors.RevisionNotPresent, err:
2456
if err.revision_id == known_revid:
2457
# The start revision (known_revid) wasn't found.
2459
# This is a stacked repository with no fallbacks, or a there's a
2460
# left-hand ghost. Either way, even though the revision named in
2461
# the error isn't in this repo, we know it's the next step in this
2462
# left-hand history.
2463
partial_history.append(err.revision_id)
2464
if len(partial_history) <= distance_from_known:
2465
# Didn't find enough history to get a revid for the revno.
2466
earliest_revno = known_revno - len(partial_history) + 1
2467
return (False, (earliest_revno, partial_history[-1]))
2468
if len(partial_history) - 1 > distance_from_known:
2469
raise AssertionError('_iter_for_revno returned too much history')
2470
return (True, partial_history[-1])
641
def get_inventory_sha1(self, revision_id):
642
"""Return the sha1 hash of the inventory entry
644
# TODO: jam 20070210 Shouldn't this be deprecated / removed?
645
revision_id = osutils.safe_revision_id(revision_id)
646
return self.get_revision(revision_id).inventory_sha1
649
def get_revision_graph(self, revision_id=None):
650
"""Return a dictionary containing the revision graph.
652
:param revision_id: The revision_id to get a graph from. If None, then
653
the entire revision graph is returned. This is a deprecated mode of
654
operation and will be removed in the future.
655
:return: a dictionary of revision_id->revision_parents_list.
657
# special case NULL_REVISION
658
if revision_id == _mod_revision.NULL_REVISION:
660
revision_id = osutils.safe_revision_id(revision_id)
661
a_weave = self.get_inventory_weave()
662
all_revisions = self._eliminate_revisions_not_present(
664
entire_graph = dict([(node, a_weave.get_parents(node)) for
665
node in all_revisions])
666
if revision_id is None:
668
elif revision_id not in entire_graph:
669
raise errors.NoSuchRevision(self, revision_id)
671
# add what can be reached from revision_id
673
pending = set([revision_id])
674
while len(pending) > 0:
676
result[node] = entire_graph[node]
677
for revision_id in result[node]:
678
if revision_id not in result:
679
pending.add(revision_id)
683
def get_revision_graph_with_ghosts(self, revision_ids=None):
684
"""Return a graph of the revisions with ghosts marked as applicable.
686
:param revision_ids: an iterable of revisions to graph or None for all.
687
:return: a Graph object with the graph reachable from revision_ids.
689
result = graph.Graph()
691
pending = set(self.all_revision_ids())
694
pending = set(osutils.safe_revision_id(r) for r in revision_ids)
695
# special case NULL_REVISION
696
if _mod_revision.NULL_REVISION in pending:
697
pending.remove(_mod_revision.NULL_REVISION)
698
required = set(pending)
701
revision_id = pending.pop()
703
rev = self.get_revision(revision_id)
704
except errors.NoSuchRevision:
705
if revision_id in required:
708
result.add_ghost(revision_id)
710
for parent_id in rev.parent_ids:
711
# is this queued or done ?
712
if (parent_id not in pending and
713
parent_id not in done):
715
pending.add(parent_id)
716
result.add_node(revision_id, rev.parent_ids)
717
done.add(revision_id)
720
def _get_history_vf(self):
721
"""Get a versionedfile whose history graph reflects all revisions.
723
For weave repositories, this is the inventory weave.
725
return self.get_inventory_weave()
2472
727
def iter_reverse_revision_history(self, revision_id):
2473
728
"""Iterate backwards through revision ids in the lefthand history
2490
771
reconciler = RepoReconciler(self, thorough=thorough)
2491
772
reconciler.reconcile()
2492
773
return reconciler
2494
def _refresh_data(self):
2495
"""Helper called from lock_* to ensure coherency with disk.
2497
The default implementation does nothing; it is however possible
2498
for repositories to maintain loaded indices across multiple locks
2499
by checking inside their implementation of this method to see
2500
whether their indices are still valid. This depends of course on
2501
the disk format being validatable in this manner. This method is
2502
also called by the refresh_data() public interface to cause a refresh
2503
to occur while in a write lock so that data inserted by a smart server
2504
push operation is visible on the client's instance of the physical
2508
775
@needs_read_lock
2509
776
def revision_tree(self, revision_id):
2510
777
"""Return Tree for a revision on this branch.
2512
`revision_id` may be NULL_REVISION for the empty tree revision.
779
`revision_id` may be None for the empty tree revision.
2514
revision_id = _mod_revision.ensure_null(revision_id)
2515
781
# TODO: refactor this to use an existing revision object
2516
782
# so we don't need to read it in twice.
2517
if revision_id == _mod_revision.NULL_REVISION:
2518
return InventoryRevisionTree(self,
2519
Inventory(root_id=None), _mod_revision.NULL_REVISION)
783
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
784
return RevisionTree(self, Inventory(root_id=None),
785
_mod_revision.NULL_REVISION)
2521
inv = self.get_inventory(revision_id)
2522
return InventoryRevisionTree(self, inv, revision_id)
787
revision_id = osutils.safe_revision_id(revision_id)
788
inv = self.get_revision_inventory(revision_id)
789
return RevisionTree(self, inv, revision_id)
2524
792
def revision_trees(self, revision_ids):
2525
"""Return Trees for revisions in this repository.
2527
:param revision_ids: a sequence of revision-ids;
2528
a revision-id may not be None or 'null:'
2530
inventories = self.iter_inventories(revision_ids)
2531
for inv in inventories:
2532
yield InventoryRevisionTree(self, inv, inv.revision_id)
2534
def _filtered_revision_trees(self, revision_ids, file_ids):
2535
"""Return Tree for a revision on this branch with only some files.
2537
:param revision_ids: a sequence of revision-ids;
2538
a revision-id may not be None or 'null:'
2539
:param file_ids: if not None, the result is filtered
2540
so that only those file-ids, their parents and their
2541
children are included.
2543
inventories = self.iter_inventories(revision_ids)
2544
for inv in inventories:
2545
# Should we introduce a FilteredRevisionTree class rather
2546
# than pre-filter the inventory here?
2547
filtered_inv = inv.filter(file_ids)
2548
yield InventoryRevisionTree(self, filtered_inv, filtered_inv.revision_id)
793
"""Return Tree for a revision on this branch.
795
`revision_id` may not be None or 'null:'"""
796
assert None not in revision_ids
797
assert _mod_revision.NULL_REVISION not in revision_ids
798
texts = self.get_inventory_weave().get_texts(revision_ids)
799
for text, revision_id in zip(texts, revision_ids):
800
inv = self.deserialise_inventory(revision_id, text)
801
yield RevisionTree(self, inv, revision_id)
2550
803
@needs_read_lock
2551
def get_ancestry(self, revision_id, topo_sorted=True):
804
def get_ancestry(self, revision_id):
2552
805
"""Return a list of revision-ids integrated by a revision.
2554
The first element of the list is always None, indicating the origin
2555
revision. This might change when we have history horizons, or
807
The first element of the list is always None, indicating the origin
808
revision. This might change when we have history horizons, or
2556
809
perhaps we should have a new API.
2558
811
This is topologically sorted.
2560
if _mod_revision.is_null(revision_id):
813
if revision_id is None:
815
revision_id = osutils.safe_revision_id(revision_id)
2562
816
if not self.has_revision(revision_id):
2563
817
raise errors.NoSuchRevision(self, revision_id)
2564
graph = self.get_graph()
2566
search = graph._make_breadth_first_searcher([revision_id])
2569
found, ghosts = search.next_with_ghosts()
2570
except StopIteration:
2573
if _mod_revision.NULL_REVISION in keys:
2574
keys.remove(_mod_revision.NULL_REVISION)
2576
parent_map = graph.get_parent_map(keys)
2577
keys = tsort.topo_sort(parent_map)
2578
return [None] + list(keys)
2580
def pack(self, hint=None, clean_obsolete_packs=False):
2581
"""Compress the data within the repository.
2583
This operation only makes sense for some repository types. For other
2584
types it should be a no-op that just returns.
2586
This stub method does not require a lock, but subclasses should use
2587
@needs_write_lock as this is a long running call it's reasonable to
2588
implicitly lock for the user.
2590
:param hint: If not supplied, the whole repository is packed.
2591
If supplied, the repository may use the hint parameter as a
2592
hint for the parts of the repository to pack. A hint can be
2593
obtained from the result of commit_write_group(). Out of
2594
date hints are simply ignored, because concurrent operations
2595
can obsolete them rapidly.
2597
:param clean_obsolete_packs: Clean obsolete packs immediately after
818
w = self.get_inventory_weave()
819
candidates = w.get_ancestry(revision_id)
820
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
823
def print_file(self, file, revision_id):
824
"""Print `file` to stdout.
826
FIXME RBC 20060125 as John Meinel points out this is a bad api
827
- it writes to stdout, it assumes that that is valid etc. Fix
828
by creating a new more flexible convenience function.
830
revision_id = osutils.safe_revision_id(revision_id)
831
tree = self.revision_tree(revision_id)
832
# use inventory as it was in that revision
833
file_id = tree.inventory.path2id(file)
835
# TODO: jam 20060427 Write a test for this code path
836
# it had a bug in it, and was raising the wrong
838
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
839
tree.print_file(file_id)
2601
841
def get_transaction(self):
2602
842
return self.control_files.get_transaction()
2604
def get_parent_map(self, revision_ids):
2605
"""See graph.StackedParentsProvider.get_parent_map"""
2606
# revisions index works in keys; this just works in revisions
2607
# therefore wrap and unwrap
2610
for revision_id in revision_ids:
2611
if revision_id == _mod_revision.NULL_REVISION:
2612
result[revision_id] = ()
2613
elif revision_id is None:
2614
raise ValueError('get_parent_map(None) is not valid')
2616
query_keys.append((revision_id ,))
2617
for ((revision_id,), parent_keys) in \
2618
self.revisions.get_parent_map(query_keys).iteritems():
2620
result[revision_id] = tuple([parent_revid
2621
for (parent_revid,) in parent_keys])
2623
result[revision_id] = (_mod_revision.NULL_REVISION,)
2626
def _make_parents_provider(self):
2630
def get_known_graph_ancestry(self, revision_ids):
2631
"""Return the known graph for a set of revision ids and their ancestors.
2633
st = static_tuple.StaticTuple
2634
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2635
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2636
return graph.GraphThunkIdsToKeys(known_graph)
2638
def get_graph(self, other_repository=None):
2639
"""Return the graph walker for this repository format"""
2640
parents_provider = self._make_parents_provider()
2641
if (other_repository is not None and
2642
not self.has_same_location(other_repository)):
2643
parents_provider = graph.StackedParentsProvider(
2644
[parents_provider, other_repository._make_parents_provider()])
2645
return graph.Graph(parents_provider)
2647
def _get_versioned_file_checker(self, text_key_references=None,
2649
"""Return an object suitable for checking versioned files.
2651
:param text_key_references: if non-None, an already built
2652
dictionary mapping text keys ((fileid, revision_id) tuples)
2653
to whether they were referred to by the inventory of the
2654
revision_id that they contain. If None, this will be
2656
:param ancestors: Optional result from
2657
self.get_graph().get_parent_map(self.all_revision_ids()) if already
2660
return _VersionedFileChecker(self,
2661
text_key_references=text_key_references, ancestors=ancestors)
2663
def revision_ids_to_search_result(self, result_set):
2664
"""Convert a set of revision ids to a graph SearchResult."""
2665
result_parents = set()
2666
for parents in self.get_graph().get_parent_map(
2667
result_set).itervalues():
2668
result_parents.update(parents)
2669
included_keys = result_set.intersection(result_parents)
2670
start_keys = result_set.difference(included_keys)
2671
exclude_keys = result_parents.difference(result_set)
2672
result = graph.SearchResult(start_keys, exclude_keys,
2673
len(result_set), result_set)
844
def revision_parents(self, revision_id):
845
revision_id = osutils.safe_revision_id(revision_id)
846
return self.get_inventory_weave().parent_names(revision_id)
2676
848
@needs_write_lock
2677
849
def set_make_working_trees(self, new_value):
3323
1312
InterRepository.get(other).method_name(parameters).
3326
_walk_to_common_revisions_batch_size = 50
3327
1315
_optimisers = []
3328
1316
"""The available optimised InterRepository types."""
3331
1318
def copy_content(self, revision_id=None):
3332
"""Make a complete copy of the content in self into destination.
3334
This is a destructive operation! Do not use it on existing
3337
:param revision_id: Only copy the content needed to construct
3338
revision_id and its parents.
3341
self.target.set_make_working_trees(self.source.make_working_trees())
3342
except NotImplementedError:
3344
self.target.fetch(self.source, revision_id=revision_id)
3347
def fetch(self, revision_id=None, find_ghosts=False,
1319
raise NotImplementedError(self.copy_content)
1321
def fetch(self, revision_id=None, pb=None):
3349
1322
"""Fetch the content required to construct revision_id.
3351
1324
The content is copied from self.source to self.target.
3353
1326
:param revision_id: if None all content is copied, if NULL_REVISION no
3354
1327
content is copied.
3357
ui.ui_factory.warn_experimental_format_fetch(self)
3358
from bzrlib.fetch import RepoFetcher
3359
# See <https://launchpad.net/bugs/456077> asking for a warning here
3360
if self.source._format.network_name() != self.target._format.network_name():
3361
ui.ui_factory.show_user_warning('cross_format_fetch',
3362
from_format=self.source._format,
3363
to_format=self.target._format)
3364
f = RepoFetcher(to_repository=self.target,
3365
from_repository=self.source,
3366
last_revision=revision_id,
3367
fetch_spec=fetch_spec,
3368
find_ghosts=find_ghosts)
3370
def _walk_to_common_revisions(self, revision_ids, if_present_ids=None):
3371
"""Walk out from revision_ids in source to revisions target has.
3373
:param revision_ids: The start point for the search.
3374
:return: A set of revision ids.
3376
target_graph = self.target.get_graph()
3377
revision_ids = frozenset(revision_ids)
3379
all_wanted_revs = revision_ids.union(if_present_ids)
3381
all_wanted_revs = revision_ids
3382
missing_revs = set()
3383
source_graph = self.source.get_graph()
3384
# ensure we don't pay silly lookup costs.
3385
searcher = source_graph._make_breadth_first_searcher(all_wanted_revs)
3386
null_set = frozenset([_mod_revision.NULL_REVISION])
3387
searcher_exhausted = False
3391
# Iterate the searcher until we have enough next_revs
3392
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3394
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3395
next_revs.update(next_revs_part)
3396
ghosts.update(ghosts_part)
3397
except StopIteration:
3398
searcher_exhausted = True
3400
# If there are ghosts in the source graph, and the caller asked for
3401
# them, make sure that they are present in the target.
3402
# We don't care about other ghosts as we can't fetch them and
3403
# haven't been asked to.
3404
ghosts_to_check = set(revision_ids.intersection(ghosts))
3405
revs_to_get = set(next_revs).union(ghosts_to_check)
3407
have_revs = set(target_graph.get_parent_map(revs_to_get))
3408
# we always have NULL_REVISION present.
3409
have_revs = have_revs.union(null_set)
3410
# Check if the target is missing any ghosts we need.
3411
ghosts_to_check.difference_update(have_revs)
3413
# One of the caller's revision_ids is a ghost in both the
3414
# source and the target.
3415
raise errors.NoSuchRevision(
3416
self.source, ghosts_to_check.pop())
3417
missing_revs.update(next_revs - have_revs)
3418
# Because we may have walked past the original stop point, make
3419
# sure everything is stopped
3420
stop_revs = searcher.find_seen_ancestors(have_revs)
3421
searcher.stop_searching_any(stop_revs)
3422
if searcher_exhausted:
3424
return searcher.get_result()
1328
:param pb: optional progress bar to use for progress reports. If not
1329
provided a default one will be created.
1331
Returns the copied revision count and the failed revisions in a tuple:
1334
raise NotImplementedError(self.fetch)
3426
1336
@needs_read_lock
3427
def search_missing_revision_ids(self,
3428
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
3429
find_ghosts=True, revision_ids=None, if_present_ids=None):
1337
def missing_revision_ids(self, revision_id=None):
3430
1338
"""Return the revision ids that source has that target does not.
1340
These are returned in topological order.
3432
1342
:param revision_id: only return revision ids included by this
3434
:param revision_ids: return revision ids included by these
3435
revision_ids. NoSuchRevision will be raised if any of these
3436
revisions are not present.
3437
:param if_present_ids: like revision_ids, but will not cause
3438
NoSuchRevision if any of these are absent, instead they will simply
3439
not be in the result. This is useful for e.g. finding revisions
3440
to fetch for tags, which may reference absent revisions.
3441
:param find_ghosts: If True find missing revisions in deep history
3442
rather than just finding the surface difference.
3443
:return: A bzrlib.graph.SearchResult.
3445
if symbol_versioning.deprecated_passed(revision_id):
3446
symbol_versioning.warn(
3447
'search_missing_revision_ids(revision_id=...) was '
3448
'deprecated in 2.4. Use revision_ids=[...] instead.',
3449
DeprecationWarning, stacklevel=2)
3450
if revision_ids is not None:
3451
raise AssertionError(
3452
'revision_ids is mutually exclusive with revision_id')
3453
if revision_id is not None:
3454
revision_ids = [revision_id]
3456
# stop searching at found target revisions.
3457
if not find_ghosts and (revision_ids is not None or if_present_ids is
3459
return self._walk_to_common_revisions(revision_ids,
3460
if_present_ids=if_present_ids)
3461
1345
# generic, possibly worst case, slow code path.
3462
1346
target_ids = set(self.target.all_revision_ids())
3463
source_ids = self._present_source_revisions_for(
3464
revision_ids, if_present_ids)
1347
if revision_id is not None:
1348
# TODO: jam 20070210 InterRepository is internal enough that it
1349
# should assume revision_ids are already utf-8
1350
revision_id = osutils.safe_revision_id(revision_id)
1351
source_ids = self.source.get_ancestry(revision_id)
1352
assert source_ids[0] is None
1355
source_ids = self.source.all_revision_ids()
3465
1356
result_set = set(source_ids).difference(target_ids)
3466
return self.source.revision_ids_to_search_result(result_set)
3468
def _present_source_revisions_for(self, revision_ids, if_present_ids=None):
3469
"""Returns set of all revisions in ancestry of revision_ids present in
3472
:param revision_ids: if None, all revisions in source are returned.
3473
:param if_present_ids: like revision_ids, but if any/all of these are
3474
absent no error is raised.
3476
if revision_ids is not None or if_present_ids is not None:
3477
# First, ensure all specified revisions exist. Callers expect
3478
# NoSuchRevision when they pass absent revision_ids here.
3479
if revision_ids is None:
3480
revision_ids = set()
3481
if if_present_ids is None:
3482
if_present_ids = set()
3483
revision_ids = set(revision_ids)
3484
if_present_ids = set(if_present_ids)
3485
all_wanted_ids = revision_ids.union(if_present_ids)
3486
graph = self.source.get_graph()
3487
present_revs = set(graph.get_parent_map(all_wanted_ids))
3488
missing = revision_ids.difference(present_revs)
3490
raise errors.NoSuchRevision(self.source, missing.pop())
3491
found_ids = all_wanted_ids.intersection(present_revs)
3492
source_ids = [rev_id for (rev_id, parents) in
3493
graph.iter_ancestry(found_ids)
3494
if rev_id != _mod_revision.NULL_REVISION
3495
and parents is not None]
3497
source_ids = self.source.all_revision_ids()
3498
return set(source_ids)
3501
def _same_model(source, target):
3502
"""True if source and target have the same data representation.
3504
Note: this is always called on the base class; overriding it in a
3505
subclass will have no effect.
3508
InterRepository._assert_same_model(source, target)
3510
except errors.IncompatibleRepositories, e:
3514
def _assert_same_model(source, target):
3515
"""Raise an exception if two repositories do not use the same model.
3517
if source.supports_rich_root() != target.supports_rich_root():
3518
raise errors.IncompatibleRepositories(source, target,
3519
"different rich-root support")
3520
if source._serializer != target._serializer:
3521
raise errors.IncompatibleRepositories(source, target,
3522
"different serializers")
1357
# this may look like a no-op: its not. It preserves the ordering
1358
# other_ids had while only returning the members from other_ids
1359
# that we've decided we need.
1360
return [rev_id for rev_id in source_ids if rev_id in result_set]
3525
1363
class InterSameDataRepository(InterRepository):
3526
1364
"""Code for converting between repositories that represent the same data.
3528
1366
Data format and model must match for this to work.
3532
1370
def _get_repo_format_to_test(self):
3533
"""Repository format for testing with.
3535
InterSameData can pull from subtree to subtree and from non-subtree to
3536
non-subtree, so we test this with the richest repository format.
1371
"""Repository format for testing with."""
1372
return RepositoryFormat.get_default_format()
1375
def is_compatible(source, target):
1376
if source.supports_rich_root() != target.supports_rich_root():
1378
if source._serializer != target._serializer:
1383
def copy_content(self, revision_id=None):
1384
"""Make a complete copy of the content in self into destination.
1386
This is a destructive operation! Do not use it on existing
1389
:param revision_id: Only copy the content needed to construct
1390
revision_id and its parents.
1393
self.target.set_make_working_trees(self.source.make_working_trees())
1394
except NotImplementedError:
1396
# TODO: jam 20070210 This is fairly internal, so we should probably
1397
# just assert that revision_id is not unicode.
1398
revision_id = osutils.safe_revision_id(revision_id)
1399
# but don't bother fetching if we have the needed data now.
1400
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
1401
self.target.has_revision(revision_id)):
1403
self.target.fetch(self.source, revision_id=revision_id)
1406
def fetch(self, revision_id=None, pb=None):
1407
"""See InterRepository.fetch()."""
1408
from bzrlib.fetch import GenericRepoFetcher
1409
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1410
self.source, self.source._format, self.target,
1411
self.target._format)
1412
# TODO: jam 20070210 This should be an assert, not a translate
1413
revision_id = osutils.safe_revision_id(revision_id)
1414
f = GenericRepoFetcher(to_repository=self.target,
1415
from_repository=self.source,
1416
last_revision=revision_id,
1418
return f.count_copied, f.failed_revisions
1421
class InterWeaveRepo(InterSameDataRepository):
1422
"""Optimised code paths between Weave based repositories."""
1425
def _get_repo_format_to_test(self):
1426
from bzrlib.repofmt import weaverepo
1427
return weaverepo.RepositoryFormat7()
1430
def is_compatible(source, target):
1431
"""Be compatible with known Weave formats.
1433
We don't test for the stores being of specific types because that
1434
could lead to confusing results, and there is no need to be
1437
from bzrlib.repofmt.weaverepo import (
1443
return (isinstance(source._format, (RepositoryFormat5,
1445
RepositoryFormat7)) and
1446
isinstance(target._format, (RepositoryFormat5,
1448
RepositoryFormat7)))
1449
except AttributeError:
1453
def copy_content(self, revision_id=None):
1454
"""See InterRepository.copy_content()."""
1455
# weave specific optimised path:
1456
# TODO: jam 20070210 Internal, should be an assert, not translate
1457
revision_id = osutils.safe_revision_id(revision_id)
1459
self.target.set_make_working_trees(self.source.make_working_trees())
1460
except NotImplementedError:
1462
# FIXME do not peek!
1463
if self.source.control_files._transport.listable():
1464
pb = ui.ui_factory.nested_progress_bar()
1466
self.target.weave_store.copy_all_ids(
1467
self.source.weave_store,
1469
from_transaction=self.source.get_transaction(),
1470
to_transaction=self.target.get_transaction())
1471
pb.update('copying inventory', 0, 1)
1472
self.target.control_weaves.copy_multi(
1473
self.source.control_weaves, ['inventory'],
1474
from_transaction=self.source.get_transaction(),
1475
to_transaction=self.target.get_transaction())
1476
self.target._revision_store.text_store.copy_all_ids(
1477
self.source._revision_store.text_store,
1482
self.target.fetch(self.source, revision_id=revision_id)
1485
def fetch(self, revision_id=None, pb=None):
1486
"""See InterRepository.fetch()."""
1487
from bzrlib.fetch import GenericRepoFetcher
1488
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1489
self.source, self.source._format, self.target, self.target._format)
1490
# TODO: jam 20070210 This should be an assert, not a translate
1491
revision_id = osutils.safe_revision_id(revision_id)
1492
f = GenericRepoFetcher(to_repository=self.target,
1493
from_repository=self.source,
1494
last_revision=revision_id,
1496
return f.count_copied, f.failed_revisions
1499
def missing_revision_ids(self, revision_id=None):
1500
"""See InterRepository.missing_revision_ids()."""
1501
# we want all revisions to satisfy revision_id in source.
1502
# but we don't want to stat every file here and there.
1503
# we want then, all revisions other needs to satisfy revision_id
1504
# checked, but not those that we have locally.
1505
# so the first thing is to get a subset of the revisions to
1506
# satisfy revision_id in source, and then eliminate those that
1507
# we do already have.
1508
# this is slow on high latency connection to self, but as as this
1509
# disk format scales terribly for push anyway due to rewriting
1510
# inventory.weave, this is considered acceptable.
1512
if revision_id is not None:
1513
source_ids = self.source.get_ancestry(revision_id)
1514
assert source_ids[0] is None
1517
source_ids = self.source._all_possible_ids()
1518
source_ids_set = set(source_ids)
1519
# source_ids is the worst possible case we may need to pull.
1520
# now we want to filter source_ids against what we actually
1521
# have in target, but don't try to check for existence where we know
1522
# we do not have a revision as that would be pointless.
1523
target_ids = set(self.target._all_possible_ids())
1524
possibly_present_revisions = target_ids.intersection(source_ids_set)
1525
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
1526
required_revisions = source_ids_set.difference(actually_present_revisions)
1527
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
1528
if revision_id is not None:
1529
# we used get_ancestry to determine source_ids then we are assured all
1530
# revisions referenced are present as they are installed in topological order.
1531
# and the tip revision was validated by get_ancestry.
1532
return required_topo_revisions
1534
# if we just grabbed the possibly available ids, then
1535
# we only have an estimate of whats available and need to validate
1536
# that against the revision records.
1537
return self.source._eliminate_revisions_not_present(required_topo_revisions)
1540
class InterKnitRepo(InterSameDataRepository):
1541
"""Optimised code paths between Knit based repositories."""
1544
def _get_repo_format_to_test(self):
3538
1545
from bzrlib.repofmt import knitrepo
3539
return knitrepo.RepositoryFormatKnit3()
1546
return knitrepo.RepositoryFormatKnit1()
3542
1549
def is_compatible(source, target):
3543
return InterRepository._same_model(source, target)
3546
class InterDifferingSerializer(InterRepository):
1550
"""Be compatible with known Knit formats.
1552
We don't test for the stores being of specific types because that
1553
could lead to confusing results, and there is no need to be
1556
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit1
1558
return (isinstance(source._format, (RepositoryFormatKnit1)) and
1559
isinstance(target._format, (RepositoryFormatKnit1)))
1560
except AttributeError:
1564
def fetch(self, revision_id=None, pb=None):
1565
"""See InterRepository.fetch()."""
1566
from bzrlib.fetch import KnitRepoFetcher
1567
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1568
self.source, self.source._format, self.target, self.target._format)
1569
# TODO: jam 20070210 This should be an assert, not a translate
1570
revision_id = osutils.safe_revision_id(revision_id)
1571
f = KnitRepoFetcher(to_repository=self.target,
1572
from_repository=self.source,
1573
last_revision=revision_id,
1575
return f.count_copied, f.failed_revisions
1578
def missing_revision_ids(self, revision_id=None):
1579
"""See InterRepository.missing_revision_ids()."""
1580
if revision_id is not None:
1581
source_ids = self.source.get_ancestry(revision_id)
1582
assert source_ids[0] is None
1585
source_ids = self.source._all_possible_ids()
1586
source_ids_set = set(source_ids)
1587
# source_ids is the worst possible case we may need to pull.
1588
# now we want to filter source_ids against what we actually
1589
# have in target, but don't try to check for existence where we know
1590
# we do not have a revision as that would be pointless.
1591
target_ids = set(self.target._all_possible_ids())
1592
possibly_present_revisions = target_ids.intersection(source_ids_set)
1593
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
1594
required_revisions = source_ids_set.difference(actually_present_revisions)
1595
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
1596
if revision_id is not None:
1597
# we used get_ancestry to determine source_ids then we are assured all
1598
# revisions referenced are present as they are installed in topological order.
1599
# and the tip revision was validated by get_ancestry.
1600
return required_topo_revisions
1602
# if we just grabbed the possibly available ids, then
1603
# we only have an estimate of whats available and need to validate
1604
# that against the revision records.
1605
return self.source._eliminate_revisions_not_present(required_topo_revisions)
1608
class InterModel1and2(InterRepository):
3549
1611
def _get_repo_format_to_test(self):
3553
1615
def is_compatible(source, target):
3554
"""Be compatible with Knit2 source and Knit3 target"""
3555
# This is redundant with format.check_conversion_target(), however that
3556
# raises an exception, and we just want to say "False" as in we won't
3557
# support converting between these formats.
3558
if 'IDS_never' in debug.debug_flags:
3560
if source.supports_rich_root() and not target.supports_rich_root():
3562
if (source._format.supports_tree_reference
3563
and not target._format.supports_tree_reference):
3565
if target._fallback_repositories and target._format.supports_chks:
3566
# IDS doesn't know how to copy CHKs for the parent inventories it
3567
# adds to stacked repos.
3569
if 'IDS_always' in debug.debug_flags:
1616
if not source.supports_rich_root() and target.supports_rich_root():
3571
# Only use this code path for local source and target. IDS does far
3572
# too much IO (both bandwidth and roundtrips) over a network.
3573
if not source.bzrdir.transport.base.startswith('file:///'):
3575
if not target.bzrdir.transport.base.startswith('file:///'):
3579
def _get_trees(self, revision_ids, cache):
3581
for rev_id in revision_ids:
3583
possible_trees.append((rev_id, cache[rev_id]))
3585
# Not cached, but inventory might be present anyway.
3587
tree = self.source.revision_tree(rev_id)
3588
except errors.NoSuchRevision:
3589
# Nope, parent is ghost.
3592
cache[rev_id] = tree
3593
possible_trees.append((rev_id, tree))
3594
return possible_trees
3596
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3597
"""Get the best delta and base for this revision.
3599
:return: (basis_id, delta)
3602
# Generate deltas against each tree, to find the shortest.
3603
texts_possibly_new_in_tree = set()
3604
for basis_id, basis_tree in possible_trees:
3605
delta = tree.inventory._make_delta(basis_tree.inventory)
3606
for old_path, new_path, file_id, new_entry in delta:
3607
if new_path is None:
3608
# This file_id isn't present in the new rev, so we don't
3612
# Rich roots are handled elsewhere...
3614
kind = new_entry.kind
3615
if kind != 'directory' and kind != 'file':
3616
# No text record associated with this inventory entry.
3618
# This is a directory or file that has changed somehow.
3619
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3620
deltas.append((len(delta), basis_id, delta))
3622
return deltas[0][1:]
3624
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3625
"""Find all parent revisions that are absent, but for which the
3626
inventory is present, and copy those inventories.
3628
This is necessary to preserve correctness when the source is stacked
3629
without fallbacks configured. (Note that in cases like upgrade the
3630
source may be not have _fallback_repositories even though it is
3634
for parents in parent_map.values():
3635
parent_revs.update(parents)
3636
present_parents = self.source.get_parent_map(parent_revs)
3637
absent_parents = set(parent_revs).difference(present_parents)
3638
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3639
(rev_id,) for rev_id in absent_parents)
3640
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3641
for parent_tree in self.source.revision_trees(parent_inv_ids):
3642
current_revision_id = parent_tree.get_revision_id()
3643
parents_parents_keys = parent_invs_keys_for_stacking[
3644
(current_revision_id,)]
3645
parents_parents = [key[-1] for key in parents_parents_keys]
3646
basis_id = _mod_revision.NULL_REVISION
3647
basis_tree = self.source.revision_tree(basis_id)
3648
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3649
self.target.add_inventory_by_delta(
3650
basis_id, delta, current_revision_id, parents_parents)
3651
cache[current_revision_id] = parent_tree
3653
def _fetch_batch(self, revision_ids, basis_id, cache):
3654
"""Fetch across a few revisions.
3656
:param revision_ids: The revisions to copy
3657
:param basis_id: The revision_id of a tree that must be in cache, used
3658
as a basis for delta when no other base is available
3659
:param cache: A cache of RevisionTrees that we can use.
3660
:return: The revision_id of the last converted tree. The RevisionTree
3661
for it will be in cache
3663
# Walk though all revisions; get inventory deltas, copy referenced
3664
# texts that delta references, insert the delta, revision and
3666
root_keys_to_create = set()
3669
pending_revisions = []
3670
parent_map = self.source.get_parent_map(revision_ids)
3671
self._fetch_parent_invs_for_stacking(parent_map, cache)
3672
self.source._safe_to_return_from_cache = True
3673
for tree in self.source.revision_trees(revision_ids):
3674
# Find a inventory delta for this revision.
3675
# Find text entries that need to be copied, too.
3676
current_revision_id = tree.get_revision_id()
3677
parent_ids = parent_map.get(current_revision_id, ())
3678
parent_trees = self._get_trees(parent_ids, cache)
3679
possible_trees = list(parent_trees)
3680
if len(possible_trees) == 0:
3681
# There either aren't any parents, or the parents are ghosts,
3682
# so just use the last converted tree.
3683
possible_trees.append((basis_id, cache[basis_id]))
3684
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3686
revision = self.source.get_revision(current_revision_id)
3687
pending_deltas.append((basis_id, delta,
3688
current_revision_id, revision.parent_ids))
3689
if self._converting_to_rich_root:
3690
self._revision_id_to_root_id[current_revision_id] = \
3692
# Determine which texts are in present in this revision but not in
3693
# any of the available parents.
3694
texts_possibly_new_in_tree = set()
3695
for old_path, new_path, file_id, entry in delta:
3696
if new_path is None:
3697
# This file_id isn't present in the new rev
3701
if not self.target.supports_rich_root():
3702
# The target doesn't support rich root, so we don't
3705
if self._converting_to_rich_root:
3706
# This can't be copied normally, we have to insert
3708
root_keys_to_create.add((file_id, entry.revision))
3711
texts_possibly_new_in_tree.add((file_id, entry.revision))
3712
for basis_id, basis_tree in possible_trees:
3713
basis_inv = basis_tree.inventory
3714
for file_key in list(texts_possibly_new_in_tree):
3715
file_id, file_revision = file_key
3717
entry = basis_inv[file_id]
3718
except errors.NoSuchId:
3720
if entry.revision == file_revision:
3721
texts_possibly_new_in_tree.remove(file_key)
3722
text_keys.update(texts_possibly_new_in_tree)
3723
pending_revisions.append(revision)
3724
cache[current_revision_id] = tree
3725
basis_id = current_revision_id
3726
self.source._safe_to_return_from_cache = False
3728
from_texts = self.source.texts
3729
to_texts = self.target.texts
3730
if root_keys_to_create:
3731
root_stream = _mod_fetch._new_root_data_stream(
3732
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3734
to_texts.insert_record_stream(root_stream)
3735
to_texts.insert_record_stream(from_texts.get_record_stream(
3736
text_keys, self.target._format._fetch_order,
3737
not self.target._format._fetch_uses_deltas))
3738
# insert inventory deltas
3739
for delta in pending_deltas:
3740
self.target.add_inventory_by_delta(*delta)
3741
if self.target._fallback_repositories:
3742
# Make sure this stacked repository has all the parent inventories
3743
# for the new revisions that we are about to insert. We do this
3744
# before adding the revisions so that no revision is added until
3745
# all the inventories it may depend on are added.
3746
# Note that this is overzealous, as we may have fetched these in an
3749
revision_ids = set()
3750
for revision in pending_revisions:
3751
revision_ids.add(revision.revision_id)
3752
parent_ids.update(revision.parent_ids)
3753
parent_ids.difference_update(revision_ids)
3754
parent_ids.discard(_mod_revision.NULL_REVISION)
3755
parent_map = self.source.get_parent_map(parent_ids)
3756
# we iterate over parent_map and not parent_ids because we don't
3757
# want to try copying any revision which is a ghost
3758
for parent_tree in self.source.revision_trees(parent_map):
3759
current_revision_id = parent_tree.get_revision_id()
3760
parents_parents = parent_map[current_revision_id]
3761
possible_trees = self._get_trees(parents_parents, cache)
3762
if len(possible_trees) == 0:
3763
# There either aren't any parents, or the parents are
3764
# ghosts, so just use the last converted tree.
3765
possible_trees.append((basis_id, cache[basis_id]))
3766
basis_id, delta = self._get_delta_for_revision(parent_tree,
3767
parents_parents, possible_trees)
3768
self.target.add_inventory_by_delta(
3769
basis_id, delta, current_revision_id, parents_parents)
3770
# insert signatures and revisions
3771
for revision in pending_revisions:
3773
signature = self.source.get_signature_text(
3774
revision.revision_id)
3775
self.target.add_signature_text(revision.revision_id,
3777
except errors.NoSuchRevision:
3779
self.target.add_revision(revision.revision_id, revision)
3782
def _fetch_all_revisions(self, revision_ids, pb):
3783
"""Fetch everything for the list of revisions.
3785
:param revision_ids: The list of revisions to fetch. Must be in
3787
:param pb: A ProgressTask
3790
basis_id, basis_tree = self._get_basis(revision_ids[0])
3792
cache = lru_cache.LRUCache(100)
3793
cache[basis_id] = basis_tree
3794
del basis_tree # We don't want to hang on to it here
3798
for offset in range(0, len(revision_ids), batch_size):
3799
self.target.start_write_group()
3801
pb.update('Transferring revisions', offset,
3803
batch = revision_ids[offset:offset+batch_size]
3804
basis_id = self._fetch_batch(batch, basis_id, cache)
3806
self.source._safe_to_return_from_cache = False
3807
self.target.abort_write_group()
3810
hint = self.target.commit_write_group()
3813
if hints and self.target._format.pack_compresses:
3814
self.target.pack(hint=hints)
3815
pb.update('Transferring revisions', len(revision_ids),
3819
def fetch(self, revision_id=None, find_ghosts=False,
3821
"""See InterRepository.fetch()."""
3822
if fetch_spec is not None:
3823
revision_ids = fetch_spec.get_keys()
3826
ui.ui_factory.warn_experimental_format_fetch(self)
3827
if (not self.source.supports_rich_root()
3828
and self.target.supports_rich_root()):
3829
self._converting_to_rich_root = True
3830
self._revision_id_to_root_id = {}
3832
self._converting_to_rich_root = False
3833
# See <https://launchpad.net/bugs/456077> asking for a warning here
3834
if self.source._format.network_name() != self.target._format.network_name():
3835
ui.ui_factory.show_user_warning('cross_format_fetch',
3836
from_format=self.source._format,
3837
to_format=self.target._format)
3838
if revision_ids is None:
3840
search_revision_ids = [revision_id]
3842
search_revision_ids = None
3843
revision_ids = self.target.search_missing_revision_ids(self.source,
3844
revision_ids=search_revision_ids,
3845
find_ghosts=find_ghosts).get_keys()
3846
if not revision_ids:
3848
revision_ids = tsort.topo_sort(
3849
self.source.get_graph().get_parent_map(revision_ids))
3850
if not revision_ids:
3852
# Walk though all revisions; get inventory deltas, copy referenced
3853
# texts that delta references, insert the delta, revision and
3855
pb = ui.ui_factory.nested_progress_bar()
3857
self._fetch_all_revisions(revision_ids, pb)
3860
return len(revision_ids), 0
3862
def _get_basis(self, first_revision_id):
3863
"""Get a revision and tree which exists in the target.
3865
This assumes that first_revision_id is selected for transmission
3866
because all other ancestors are already present. If we can't find an
3867
ancestor we fall back to NULL_REVISION since we know that is safe.
3869
:return: (basis_id, basis_tree)
3871
first_rev = self.source.get_revision(first_revision_id)
3873
basis_id = first_rev.parent_ids[0]
3874
# only valid as a basis if the target has it
3875
self.target.get_revision(basis_id)
3876
# Try to get a basis tree - if it's a ghost it will hit the
3877
# NoSuchRevision case.
3878
basis_tree = self.source.revision_tree(basis_id)
3879
except (IndexError, errors.NoSuchRevision):
3880
basis_id = _mod_revision.NULL_REVISION
3881
basis_tree = self.source.revision_tree(basis_id)
3882
return basis_id, basis_tree
3885
InterRepository.register_optimiser(InterDifferingSerializer)
1622
def fetch(self, revision_id=None, pb=None):
1623
"""See InterRepository.fetch()."""
1624
from bzrlib.fetch import Model1toKnit2Fetcher
1625
# TODO: jam 20070210 This should be an assert, not a translate
1626
revision_id = osutils.safe_revision_id(revision_id)
1627
f = Model1toKnit2Fetcher(to_repository=self.target,
1628
from_repository=self.source,
1629
last_revision=revision_id,
1631
return f.count_copied, f.failed_revisions
1634
def copy_content(self, revision_id=None):
1635
"""Make a complete copy of the content in self into destination.
1637
This is a destructive operation! Do not use it on existing
1640
:param revision_id: Only copy the content needed to construct
1641
revision_id and its parents.
1644
self.target.set_make_working_trees(self.source.make_working_trees())
1645
except NotImplementedError:
1647
# TODO: jam 20070210 Internal, assert, don't translate
1648
revision_id = osutils.safe_revision_id(revision_id)
1649
# but don't bother fetching if we have the needed data now.
1650
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
1651
self.target.has_revision(revision_id)):
1653
self.target.fetch(self.source, revision_id=revision_id)
1656
class InterKnit1and2(InterKnitRepo):
1659
def _get_repo_format_to_test(self):
1663
def is_compatible(source, target):
1664
"""Be compatible with Knit1 source and Knit3 target"""
1665
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
1667
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit1, \
1668
RepositoryFormatKnit3
1669
return (isinstance(source._format, (RepositoryFormatKnit1)) and
1670
isinstance(target._format, (RepositoryFormatKnit3)))
1671
except AttributeError:
1675
def fetch(self, revision_id=None, pb=None):
1676
"""See InterRepository.fetch()."""
1677
from bzrlib.fetch import Knit1to2Fetcher
1678
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1679
self.source, self.source._format, self.target,
1680
self.target._format)
1681
# TODO: jam 20070210 This should be an assert, not a translate
1682
revision_id = osutils.safe_revision_id(revision_id)
1683
f = Knit1to2Fetcher(to_repository=self.target,
1684
from_repository=self.source,
1685
last_revision=revision_id,
1687
return f.count_copied, f.failed_revisions
3886
1690
InterRepository.register_optimiser(InterSameDataRepository)
1691
InterRepository.register_optimiser(InterWeaveRepo)
1692
InterRepository.register_optimiser(InterKnitRepo)
1693
InterRepository.register_optimiser(InterModel1and2)
1694
InterRepository.register_optimiser(InterKnit1and2)
1697
class RepositoryTestProviderAdapter(object):
1698
"""A tool to generate a suite testing multiple repository formats at once.
1700
This is done by copying the test once for each transport and injecting
1701
the transport_server, transport_readonly_server, and bzrdir_format and
1702
repository_format classes into each copy. Each copy is also given a new id()
1703
to make it easy to identify.
1706
def __init__(self, transport_server, transport_readonly_server, formats):
1707
self._transport_server = transport_server
1708
self._transport_readonly_server = transport_readonly_server
1709
self._formats = formats
1711
def adapt(self, test):
1712
result = unittest.TestSuite()
1713
for repository_format, bzrdir_format in self._formats:
1714
from copy import deepcopy
1715
new_test = deepcopy(test)
1716
new_test.transport_server = self._transport_server
1717
new_test.transport_readonly_server = self._transport_readonly_server
1718
new_test.bzrdir_format = bzrdir_format
1719
new_test.repository_format = repository_format
1720
def make_new_test_id():
1721
new_id = "%s(%s)" % (new_test.id(), repository_format.__class__.__name__)
1722
return lambda: new_id
1723
new_test.id = make_new_test_id()
1724
result.addTest(new_test)
1728
class InterRepositoryTestProviderAdapter(object):
1729
"""A tool to generate a suite testing multiple inter repository formats.
1731
This is done by copying the test once for each interrepo provider and injecting
1732
the transport_server, transport_readonly_server, repository_format and
1733
repository_to_format classes into each copy.
1734
Each copy is also given a new id() to make it easy to identify.
1737
def __init__(self, transport_server, transport_readonly_server, formats):
1738
self._transport_server = transport_server
1739
self._transport_readonly_server = transport_readonly_server
1740
self._formats = formats
1742
def adapt(self, test):
1743
result = unittest.TestSuite()
1744
for interrepo_class, repository_format, repository_format_to in self._formats:
1745
from copy import deepcopy
1746
new_test = deepcopy(test)
1747
new_test.transport_server = self._transport_server
1748
new_test.transport_readonly_server = self._transport_readonly_server
1749
new_test.interrepo_class = interrepo_class
1750
new_test.repository_format = repository_format
1751
new_test.repository_format_to = repository_format_to
1752
def make_new_test_id():
1753
new_id = "%s(%s)" % (new_test.id(), interrepo_class.__name__)
1754
return lambda: new_id
1755
new_test.id = make_new_test_id()
1756
result.addTest(new_test)
1760
def default_test_list():
1761
"""Generate the default list of interrepo permutations to test."""
1762
from bzrlib.repofmt import knitrepo, weaverepo
1764
# test the default InterRepository between format 6 and the current
1766
# XXX: robertc 20060220 reinstate this when there are two supported
1767
# formats which do not have an optimal code path between them.
1768
#result.append((InterRepository,
1769
# RepositoryFormat6(),
1770
# RepositoryFormatKnit1()))
1771
for optimiser_class in InterRepository._optimisers:
1772
format_to_test = optimiser_class._get_repo_format_to_test()
1773
if format_to_test is not None:
1774
result.append((optimiser_class,
1775
format_to_test, format_to_test))
1776
# if there are specific combinations we want to use, we can add them
1778
result.append((InterModel1and2,
1779
weaverepo.RepositoryFormat5(),
1780
knitrepo.RepositoryFormatKnit3()))
1781
result.append((InterKnit1and2,
1782
knitrepo.RepositoryFormatKnit1(),
1783
knitrepo.RepositoryFormatKnit3()))
3889
1787
class CopyConverter(object):
3890
1788
"""A repository conversion tool which just performs a copy of the content.
3892
1790
This is slow but quite reliable.
3898
1796
:param target_format: The format the resulting repository should be.
3900
1798
self.target_format = target_format
3902
1800
def convert(self, repo, pb):
3903
1801
"""Perform the conversion of to_convert, giving feedback via pb.
3905
1803
:param to_convert: The disk object to convert.
3906
1804
:param pb: a progress bar to use for progress information.
3908
pb = ui.ui_factory.nested_progress_bar()
3911
1809
# this is only useful with metadir layouts - separated repo content.
3912
1810
# trigger an assertion if not such
3913
1811
repo._format.get_format_string()
3914
1812
self.repo_dir = repo.bzrdir
3915
pb.update('Moving repository to repository.backup')
1813
self.step('Moving repository to repository.backup')
3916
1814
self.repo_dir.transport.move('repository', 'repository.backup')
3917
1815
backup_transport = self.repo_dir.transport.clone('repository.backup')
3918
1816
repo._format.check_conversion_target(self.target_format)
3919
1817
self.source_repo = repo._format.open(self.repo_dir,
3921
1819
_override_transport=backup_transport)
3922
pb.update('Creating new repository')
1820
self.step('Creating new repository')
3923
1821
converted = self.target_format.initialize(self.repo_dir,
3924
1822
self.source_repo.is_shared())
3925
1823
converted.lock_write()
3927
pb.update('Copying content')
1825
self.step('Copying content into repository.')
3928
1826
self.source_repo.copy_content_into(converted)
3930
1828
converted.unlock()
3931
pb.update('Deleting old repository content')
1829
self.step('Deleting old repository content.')
3932
1830
self.repo_dir.transport.delete_tree('repository.backup')
3933
ui.ui_factory.note('repository converted')
3937
class _VersionedFileChecker(object):
3939
def __init__(self, repository, text_key_references=None, ancestors=None):
1831
self.pb.note('repository converted')
1833
def step(self, message):
1834
"""Update the pb by a step."""
1836
self.pb.update(message, self.count, self.total)
1839
class CommitBuilder(object):
1840
"""Provides an interface to build up a commit.
1842
This allows describing a tree to be committed without needing to
1843
know the internals of the format of the repository.
1846
record_root_entry = False
1847
def __init__(self, repository, parents, config, timestamp=None,
1848
timezone=None, committer=None, revprops=None,
1850
"""Initiate a CommitBuilder.
1852
:param repository: Repository to commit to.
1853
:param parents: Revision ids of the parents of the new revision.
1854
:param config: Configuration to use.
1855
:param timestamp: Optional timestamp recorded for commit.
1856
:param timezone: Optional timezone for timestamp.
1857
:param committer: Optional committer to set for commit.
1858
:param revprops: Optional dictionary of revision properties.
1859
:param revision_id: Optional revision id.
1861
self._config = config
1863
if committer is None:
1864
self._committer = self._config.username()
1866
assert isinstance(committer, basestring), type(committer)
1867
self._committer = committer
1869
self.new_inventory = Inventory(None)
1870
self._new_revision_id = osutils.safe_revision_id(revision_id)
1871
self.parents = parents
3940
1872
self.repository = repository
3941
self.text_index = self.repository._generate_text_key_index(
3942
text_key_references=text_key_references, ancestors=ancestors)
3944
def calculate_file_version_parents(self, text_key):
3945
"""Calculate the correct parents for a file version according to
3948
parent_keys = self.text_index[text_key]
3949
if parent_keys == [_mod_revision.NULL_REVISION]:
3951
return tuple(parent_keys)
3953
def check_file_version_parents(self, texts, progress_bar=None):
3954
"""Check the parents stored in a versioned file are correct.
3956
It also detects file versions that are not referenced by their
3957
corresponding revision's inventory.
3959
:returns: A tuple of (wrong_parents, dangling_file_versions).
3960
wrong_parents is a dict mapping {revision_id: (stored_parents,
3961
correct_parents)} for each revision_id where the stored parents
3962
are not correct. dangling_file_versions is a set of (file_id,
3963
revision_id) tuples for versions that are present in this versioned
3964
file, but not used by the corresponding inventory.
3966
local_progress = None
3967
if progress_bar is None:
3968
local_progress = ui.ui_factory.nested_progress_bar()
3969
progress_bar = local_progress
3971
return self._check_file_version_parents(texts, progress_bar)
3974
local_progress.finished()
3976
def _check_file_version_parents(self, texts, progress_bar):
3977
"""See check_file_version_parents."""
3979
self.file_ids = set([file_id for file_id, _ in
3980
self.text_index.iterkeys()])
3981
# text keys is now grouped by file_id
3982
n_versions = len(self.text_index)
3983
progress_bar.update('loading text store', 0, n_versions)
3984
parent_map = self.repository.texts.get_parent_map(self.text_index)
3985
# On unlistable transports this could well be empty/error...
3986
text_keys = self.repository.texts.keys()
3987
unused_keys = frozenset(text_keys) - set(self.text_index)
3988
for num, key in enumerate(self.text_index.iterkeys()):
3989
progress_bar.update('checking text graph', num, n_versions)
3990
correct_parents = self.calculate_file_version_parents(key)
3992
knit_parents = parent_map[key]
3993
except errors.RevisionNotPresent:
3996
if correct_parents != knit_parents:
3997
wrong_parents[key] = (knit_parents, correct_parents)
3998
return wrong_parents, unused_keys
4001
def _strip_NULL_ghosts(revision_graph):
4002
"""Also don't use this. more compatibility code for unmigrated clients."""
4003
# Filter ghosts, and null:
4004
if _mod_revision.NULL_REVISION in revision_graph:
4005
del revision_graph[_mod_revision.NULL_REVISION]
4006
for key, parents in revision_graph.items():
4007
revision_graph[key] = tuple(parent for parent in parents if parent
4009
return revision_graph
4012
class StreamSink(object):
4013
"""An object that can insert a stream into a repository.
4015
This interface handles the complexity of reserialising inventories and
4016
revisions from different formats, and allows unidirectional insertion into
4017
stacked repositories without looking for the missing basis parents
4021
def __init__(self, target_repo):
4022
self.target_repo = target_repo
4024
def insert_stream(self, stream, src_format, resume_tokens):
4025
"""Insert a stream's content into the target repository.
4027
:param src_format: a bzr repository format.
4029
:return: a list of resume tokens and an iterable of keys additional
4030
items required before the insertion can be completed.
4032
self.target_repo.lock_write()
4035
self.target_repo.resume_write_group(resume_tokens)
4038
self.target_repo.start_write_group()
4041
# locked_insert_stream performs a commit|suspend.
4042
missing_keys = self.insert_stream_without_locking(stream,
4043
src_format, is_resume)
4045
# suspend the write group and tell the caller what we is
4046
# missing. We know we can suspend or else we would not have
4047
# entered this code path. (All repositories that can handle
4048
# missing keys can handle suspending a write group).
4049
write_group_tokens = self.target_repo.suspend_write_group()
4050
return write_group_tokens, missing_keys
4051
hint = self.target_repo.commit_write_group()
4052
to_serializer = self.target_repo._format._serializer
4053
src_serializer = src_format._serializer
4054
if (to_serializer != src_serializer and
4055
self.target_repo._format.pack_compresses):
4056
self.target_repo.pack(hint=hint)
4059
self.target_repo.abort_write_group(suppress_errors=True)
4062
self.target_repo.unlock()
4064
def insert_stream_without_locking(self, stream, src_format,
4066
"""Insert a stream's content into the target repository.
4068
This assumes that you already have a locked repository and an active
4071
:param src_format: a bzr repository format.
4072
:param is_resume: Passed down to get_missing_parent_inventories to
4073
indicate if we should be checking for missing texts at the same
4076
:return: A set of keys that are missing.
4078
if not self.target_repo.is_write_locked():
4079
raise errors.ObjectNotLocked(self)
4080
if not self.target_repo.is_in_write_group():
4081
raise errors.BzrError('you must already be in a write group')
4082
to_serializer = self.target_repo._format._serializer
4083
src_serializer = src_format._serializer
4085
if to_serializer == src_serializer:
4086
# If serializers match and the target is a pack repository, set the
4087
# write cache size on the new pack. This avoids poor performance
4088
# on transports where append is unbuffered (such as
4089
# RemoteTransport). This is safe to do because nothing should read
4090
# back from the target repository while a stream with matching
4091
# serialization is being inserted.
4092
# The exception is that a delta record from the source that should
4093
# be a fulltext may need to be expanded by the target (see
4094
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4095
# explicitly flush any buffered writes first in that rare case.
4097
new_pack = self.target_repo._pack_collection._new_pack
4098
except AttributeError:
4099
# Not a pack repository
4102
new_pack.set_write_cache_size(1024*1024)
4103
for substream_type, substream in stream:
4104
if 'stream' in debug.debug_flags:
4105
mutter('inserting substream: %s', substream_type)
4106
if substream_type == 'texts':
4107
self.target_repo.texts.insert_record_stream(substream)
4108
elif substream_type == 'inventories':
4109
if src_serializer == to_serializer:
4110
self.target_repo.inventories.insert_record_stream(
4113
self._extract_and_insert_inventories(
4114
substream, src_serializer)
4115
elif substream_type == 'inventory-deltas':
4116
self._extract_and_insert_inventory_deltas(
4117
substream, src_serializer)
4118
elif substream_type == 'chk_bytes':
4119
# XXX: This doesn't support conversions, as it assumes the
4120
# conversion was done in the fetch code.
4121
self.target_repo.chk_bytes.insert_record_stream(substream)
4122
elif substream_type == 'revisions':
4123
# This may fallback to extract-and-insert more often than
4124
# required if the serializers are different only in terms of
4126
if src_serializer == to_serializer:
4127
self.target_repo.revisions.insert_record_stream(substream)
4129
self._extract_and_insert_revisions(substream,
4131
elif substream_type == 'signatures':
4132
self.target_repo.signatures.insert_record_stream(substream)
4134
raise AssertionError('kaboom! %s' % (substream_type,))
4135
# Done inserting data, and the missing_keys calculations will try to
4136
# read back from the inserted data, so flush the writes to the new pack
4137
# (if this is pack format).
4138
if new_pack is not None:
4139
new_pack._write_data('', flush=True)
4140
# Find all the new revisions (including ones from resume_tokens)
4141
missing_keys = self.target_repo.get_missing_parent_inventories(
4142
check_for_missing_texts=is_resume)
4144
for prefix, versioned_file in (
4145
('texts', self.target_repo.texts),
4146
('inventories', self.target_repo.inventories),
4147
('revisions', self.target_repo.revisions),
4148
('signatures', self.target_repo.signatures),
4149
('chk_bytes', self.target_repo.chk_bytes),
4151
if versioned_file is None:
4153
# TODO: key is often going to be a StaticTuple object
4154
# I don't believe we can define a method by which
4155
# (prefix,) + StaticTuple will work, though we could
4156
# define a StaticTuple.sq_concat that would allow you to
4157
# pass in either a tuple or a StaticTuple as the second
4158
# object, so instead we could have:
4159
# StaticTuple(prefix) + key here...
4160
missing_keys.update((prefix,) + key for key in
4161
versioned_file.get_missing_compression_parent_keys())
4162
except NotImplementedError:
4163
# cannot even attempt suspending, and missing would have failed
4164
# during stream insertion.
4165
missing_keys = set()
4168
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4169
target_rich_root = self.target_repo._format.rich_root_data
4170
target_tree_refs = self.target_repo._format.supports_tree_reference
4171
for record in substream:
4172
# Insert the delta directly
4173
inventory_delta_bytes = record.get_bytes_as('fulltext')
4174
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4176
parse_result = deserialiser.parse_text_bytes(
4177
inventory_delta_bytes)
4178
except inventory_delta.IncompatibleInventoryDelta, err:
4179
mutter("Incompatible delta: %s", err.msg)
4180
raise errors.IncompatibleRevision(self.target_repo._format)
4181
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4182
revision_id = new_id
4183
parents = [key[0] for key in record.parents]
4184
self.target_repo.add_inventory_by_delta(
4185
basis_id, inv_delta, revision_id, parents)
4187
def _extract_and_insert_inventories(self, substream, serializer,
4189
"""Generate a new inventory versionedfile in target, converting data.
4191
The inventory is retrieved from the source, (deserializing it), and
4192
stored in the target (reserializing it in a different format).
4194
target_rich_root = self.target_repo._format.rich_root_data
4195
target_tree_refs = self.target_repo._format.supports_tree_reference
4196
for record in substream:
4197
# It's not a delta, so it must be a fulltext in the source
4198
# serializer's format.
4199
bytes = record.get_bytes_as('fulltext')
4200
revision_id = record.key[0]
4201
inv = serializer.read_inventory_from_string(bytes, revision_id)
4202
parents = [key[0] for key in record.parents]
4203
self.target_repo.add_inventory(revision_id, inv, parents)
4204
# No need to keep holding this full inv in memory when the rest of
4205
# the substream is likely to be all deltas.
4208
def _extract_and_insert_revisions(self, substream, serializer):
4209
for record in substream:
4210
bytes = record.get_bytes_as('fulltext')
4211
revision_id = record.key[0]
4212
rev = serializer.read_revision_from_string(bytes)
4213
if rev.revision_id != revision_id:
4214
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4215
self.target_repo.add_revision(revision_id, rev)
4218
if self.target_repo._format._fetch_reconcile:
4219
self.target_repo.reconcile()
4222
class StreamSource(object):
4223
"""A source of a stream for fetching between repositories."""
4225
def __init__(self, from_repository, to_format):
4226
"""Create a StreamSource streaming from from_repository."""
4227
self.from_repository = from_repository
4228
self.to_format = to_format
4229
self._record_counter = RecordCounter()
4231
def delta_on_metadata(self):
4232
"""Return True if delta's are permitted on metadata streams.
4234
That is on revisions and signatures.
4236
src_serializer = self.from_repository._format._serializer
4237
target_serializer = self.to_format._serializer
4238
return (self.to_format._fetch_uses_deltas and
4239
src_serializer == target_serializer)
4241
def _fetch_revision_texts(self, revs):
4242
# fetch signatures first and then the revision texts
4243
# may need to be a InterRevisionStore call here.
4244
from_sf = self.from_repository.signatures
4245
# A missing signature is just skipped.
4246
keys = [(rev_id,) for rev_id in revs]
4247
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4249
self.to_format._fetch_order,
4250
not self.to_format._fetch_uses_deltas))
4251
# If a revision has a delta, this is actually expanded inside the
4252
# insert_record_stream code now, which is an alternate fix for
4254
from_rf = self.from_repository.revisions
4255
revisions = from_rf.get_record_stream(
4257
self.to_format._fetch_order,
4258
not self.delta_on_metadata())
4259
return [('signatures', signatures), ('revisions', revisions)]
4261
def _generate_root_texts(self, revs):
4262
"""This will be called by get_stream between fetching weave texts and
4263
fetching the inventory weave.
4265
if self._rich_root_upgrade():
4266
return _mod_fetch.Inter1and2Helper(
4267
self.from_repository).generate_root_texts(revs)
4271
def get_stream(self, search):
4273
revs = search.get_keys()
4274
graph = self.from_repository.get_graph()
4275
revs = tsort.topo_sort(graph.get_parent_map(revs))
4276
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4278
for knit_kind, file_id, revisions in data_to_fetch:
4279
if knit_kind != phase:
4281
# Make a new progress bar for this phase
4282
if knit_kind == "file":
4283
# Accumulate file texts
4284
text_keys.extend([(file_id, revision) for revision in
4286
elif knit_kind == "inventory":
4287
# Now copy the file texts.
4288
from_texts = self.from_repository.texts
4289
yield ('texts', from_texts.get_record_stream(
4290
text_keys, self.to_format._fetch_order,
4291
not self.to_format._fetch_uses_deltas))
4292
# Cause an error if a text occurs after we have done the
4295
# Before we process the inventory we generate the root
4296
# texts (if necessary) so that the inventories references
4298
for _ in self._generate_root_texts(revs):
4300
# we fetch only the referenced inventories because we do not
4301
# know for unselected inventories whether all their required
4302
# texts are present in the other repository - it could be
4304
for info in self._get_inventory_stream(revs):
4306
elif knit_kind == "signatures":
4307
# Nothing to do here; this will be taken care of when
4308
# _fetch_revision_texts happens.
4310
elif knit_kind == "revisions":
4311
for record in self._fetch_revision_texts(revs):
4314
raise AssertionError("Unknown knit kind %r" % knit_kind)
4316
def get_stream_for_missing_keys(self, missing_keys):
4317
# missing keys can only occur when we are byte copying and not
4318
# translating (because translation means we don't send
4319
# unreconstructable deltas ever).
4321
keys['texts'] = set()
4322
keys['revisions'] = set()
4323
keys['inventories'] = set()
4324
keys['chk_bytes'] = set()
4325
keys['signatures'] = set()
4326
for key in missing_keys:
4327
keys[key[0]].add(key[1:])
4328
if len(keys['revisions']):
4329
# If we allowed copying revisions at this point, we could end up
4330
# copying a revision without copying its required texts: a
4331
# violation of the requirements for repository integrity.
4332
raise AssertionError(
4333
'cannot copy revisions to fill in missing deltas %s' % (
4334
keys['revisions'],))
4335
for substream_kind, keys in keys.iteritems():
4336
vf = getattr(self.from_repository, substream_kind)
4337
if vf is None and keys:
4338
raise AssertionError(
4339
"cannot fill in keys for a versioned file we don't"
4340
" have: %s needs %s" % (substream_kind, keys))
4342
# No need to stream something we don't have
4344
if substream_kind == 'inventories':
4345
# Some missing keys are genuinely ghosts, filter those out.
4346
present = self.from_repository.inventories.get_parent_map(keys)
4347
revs = [key[0] for key in present]
4348
# Get the inventory stream more-or-less as we do for the
4349
# original stream; there's no reason to assume that records
4350
# direct from the source will be suitable for the sink. (Think
4351
# e.g. 2a -> 1.9-rich-root).
4352
for info in self._get_inventory_stream(revs, missing=True):
4356
# Ask for full texts always so that we don't need more round trips
4357
# after this stream.
4358
# Some of the missing keys are genuinely ghosts, so filter absent
4359
# records. The Sink is responsible for doing another check to
4360
# ensure that ghosts don't introduce missing data for future
4362
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4363
self.to_format._fetch_order, True))
4364
yield substream_kind, stream
4366
def inventory_fetch_order(self):
4367
if self._rich_root_upgrade():
4368
return 'topological'
4370
return self.to_format._fetch_order
4372
def _rich_root_upgrade(self):
4373
return (not self.from_repository._format.rich_root_data and
4374
self.to_format.rich_root_data)
4376
def _get_inventory_stream(self, revision_ids, missing=False):
4377
from_format = self.from_repository._format
4378
if (from_format.supports_chks and self.to_format.supports_chks and
4379
from_format.network_name() == self.to_format.network_name()):
4380
raise AssertionError(
4381
"this case should be handled by GroupCHKStreamSource")
4382
elif 'forceinvdeltas' in debug.debug_flags:
4383
return self._get_convertable_inventory_stream(revision_ids,
4384
delta_versus_null=missing)
4385
elif from_format.network_name() == self.to_format.network_name():
4387
return self._get_simple_inventory_stream(revision_ids,
4389
elif (not from_format.supports_chks and not self.to_format.supports_chks
4390
and from_format._serializer == self.to_format._serializer):
4391
# Essentially the same format.
4392
return self._get_simple_inventory_stream(revision_ids,
4395
# Any time we switch serializations, we want to use an
4396
# inventory-delta based approach.
4397
return self._get_convertable_inventory_stream(revision_ids,
4398
delta_versus_null=missing)
4400
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4401
# NB: This currently reopens the inventory weave in source;
4402
# using a single stream interface instead would avoid this.
4403
from_weave = self.from_repository.inventories
4405
delta_closure = True
4407
delta_closure = not self.delta_on_metadata()
4408
yield ('inventories', from_weave.get_record_stream(
4409
[(rev_id,) for rev_id in revision_ids],
4410
self.inventory_fetch_order(), delta_closure))
4412
def _get_convertable_inventory_stream(self, revision_ids,
4413
delta_versus_null=False):
4414
# The two formats are sufficiently different that there is no fast
4415
# path, so we need to send just inventorydeltas, which any
4416
# sufficiently modern client can insert into any repository.
4417
# The StreamSink code expects to be able to
4418
# convert on the target, so we need to put bytes-on-the-wire that can
4419
# be converted. That means inventory deltas (if the remote is <1.19,
4420
# RemoteStreamSink will fallback to VFS to insert the deltas).
4421
yield ('inventory-deltas',
4422
self._stream_invs_as_deltas(revision_ids,
4423
delta_versus_null=delta_versus_null))
4425
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4426
"""Return a stream of inventory-deltas for the given rev ids.
4428
:param revision_ids: The list of inventories to transmit
4429
:param delta_versus_null: Don't try to find a minimal delta for this
4430
entry, instead compute the delta versus the NULL_REVISION. This
4431
effectively streams a complete inventory. Used for stuff like
4432
filling in missing parents, etc.
4434
from_repo = self.from_repository
4435
revision_keys = [(rev_id,) for rev_id in revision_ids]
4436
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4437
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4439
inventories = self.from_repository.iter_inventories(
4440
revision_ids, 'topological')
4441
format = from_repo._format
4442
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4443
inventory_cache = lru_cache.LRUCache(50)
4444
null_inventory = from_repo.revision_tree(
4445
_mod_revision.NULL_REVISION).inventory
4446
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4447
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4448
# repo back into a non-rich-root repo ought to be allowed)
4449
serializer = inventory_delta.InventoryDeltaSerializer(
4450
versioned_root=format.rich_root_data,
4451
tree_references=format.supports_tree_reference)
4452
for inv in inventories:
4453
key = (inv.revision_id,)
4454
parent_keys = parent_map.get(key, ())
4456
if not delta_versus_null and parent_keys:
4457
# The caller did not ask for complete inventories and we have
4458
# some parents that we can delta against. Make a delta against
4459
# each parent so that we can find the smallest.
4460
parent_ids = [parent_key[0] for parent_key in parent_keys]
4461
for parent_id in parent_ids:
4462
if parent_id not in invs_sent_so_far:
4463
# We don't know that the remote side has this basis, so
4466
if parent_id == _mod_revision.NULL_REVISION:
4467
parent_inv = null_inventory
4469
parent_inv = inventory_cache.get(parent_id, None)
4470
if parent_inv is None:
4471
parent_inv = from_repo.get_inventory(parent_id)
4472
candidate_delta = inv._make_delta(parent_inv)
4473
if (delta is None or
4474
len(delta) > len(candidate_delta)):
4475
delta = candidate_delta
4476
basis_id = parent_id
4478
# Either none of the parents ended up being suitable, or we
4479
# were asked to delta against NULL
4480
basis_id = _mod_revision.NULL_REVISION
4481
delta = inv._make_delta(null_inventory)
4482
invs_sent_so_far.add(inv.revision_id)
4483
inventory_cache[inv.revision_id] = inv
4484
delta_serialized = ''.join(
4485
serializer.delta_to_lines(basis_id, key[-1], delta))
4486
yield versionedfile.FulltextContentFactory(
4487
key, parent_keys, None, delta_serialized)
4490
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4491
stop_revision=None):
4492
"""Extend the partial history to include a given index
4494
If a stop_index is supplied, stop when that index has been reached.
4495
If a stop_revision is supplied, stop when that revision is
4496
encountered. Otherwise, stop when the beginning of history is
4499
:param stop_index: The index which should be present. When it is
4500
present, history extension will stop.
4501
:param stop_revision: The revision id which should be present. When
4502
it is encountered, history extension will stop.
4504
start_revision = partial_history_cache[-1]
4505
iterator = repo.iter_reverse_revision_history(start_revision)
1875
if revprops is not None:
1876
self._revprops.update(revprops)
1878
if timestamp is None:
1879
timestamp = time.time()
1880
# Restrict resolution to 1ms
1881
self._timestamp = round(timestamp, 3)
1883
if timezone is None:
1884
self._timezone = osutils.local_time_offset()
1886
self._timezone = int(timezone)
1888
self._generate_revision_if_needed()
1890
def commit(self, message):
1891
"""Make the actual commit.
1893
:return: The revision id of the recorded revision.
1895
rev = _mod_revision.Revision(
1896
timestamp=self._timestamp,
1897
timezone=self._timezone,
1898
committer=self._committer,
1900
inventory_sha1=self.inv_sha1,
1901
revision_id=self._new_revision_id,
1902
properties=self._revprops)
1903
rev.parent_ids = self.parents
1904
self.repository.add_revision(self._new_revision_id, rev,
1905
self.new_inventory, self._config)
1906
return self._new_revision_id
1908
def revision_tree(self):
1909
"""Return the tree that was just committed.
1911
After calling commit() this can be called to get a RevisionTree
1912
representing the newly committed tree. This is preferred to
1913
calling Repository.revision_tree() because that may require
1914
deserializing the inventory, while we already have a copy in
1917
return RevisionTree(self.repository, self.new_inventory,
1918
self._new_revision_id)
1920
def finish_inventory(self):
1921
"""Tell the builder that the inventory is finished."""
1922
if self.new_inventory.root is None:
1923
symbol_versioning.warn('Root entry should be supplied to'
1924
' record_entry_contents, as of bzr 0.10.',
1925
DeprecationWarning, stacklevel=2)
1926
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
1927
self.new_inventory.revision_id = self._new_revision_id
1928
self.inv_sha1 = self.repository.add_inventory(
1929
self._new_revision_id,
1934
def _gen_revision_id(self):
1935
"""Return new revision-id."""
1936
return generate_ids.gen_revision_id(self._config.username(),
1939
def _generate_revision_if_needed(self):
1940
"""Create a revision id if None was supplied.
1942
If the repository can not support user-specified revision ids
1943
they should override this function and raise CannotSetRevisionId
1944
if _new_revision_id is not None.
1946
:raises: CannotSetRevisionId
1948
if self._new_revision_id is None:
1949
self._new_revision_id = self._gen_revision_id()
1951
def record_entry_contents(self, ie, parent_invs, path, tree):
1952
"""Record the content of ie from tree into the commit if needed.
1954
Side effect: sets ie.revision when unchanged
1956
:param ie: An inventory entry present in the commit.
1957
:param parent_invs: The inventories of the parent revisions of the
1959
:param path: The path the entry is at in the tree.
1960
:param tree: The tree which contains this entry and should be used to
1963
if self.new_inventory.root is None and ie.parent_id is not None:
1964
symbol_versioning.warn('Root entry should be supplied to'
1965
' record_entry_contents, as of bzr 0.10.',
1966
DeprecationWarning, stacklevel=2)
1967
self.record_entry_contents(tree.inventory.root.copy(), parent_invs,
1969
self.new_inventory.add(ie)
1971
# ie.revision is always None if the InventoryEntry is considered
1972
# for committing. ie.snapshot will record the correct revision
1973
# which may be the sole parent if it is untouched.
1974
if ie.revision is not None:
1977
# In this revision format, root entries have no knit or weave
1978
if ie is self.new_inventory.root:
1979
# When serializing out to disk and back in
1980
# root.revision is always _new_revision_id
1981
ie.revision = self._new_revision_id
1983
previous_entries = ie.find_previous_heads(
1985
self.repository.weave_store,
1986
self.repository.get_transaction())
1987
# we are creating a new revision for ie in the history store
1989
ie.snapshot(self._new_revision_id, path, previous_entries, tree, self)
1991
def modified_directory(self, file_id, file_parents):
1992
"""Record the presence of a symbolic link.
1994
:param file_id: The file_id of the link to record.
1995
:param file_parents: The per-file parent revision ids.
1997
self._add_text_to_weave(file_id, [], file_parents.keys())
1999
def modified_reference(self, file_id, file_parents):
2000
"""Record the modification of a reference.
2002
:param file_id: The file_id of the link to record.
2003
:param file_parents: The per-file parent revision ids.
2005
self._add_text_to_weave(file_id, [], file_parents.keys())
2007
def modified_file_text(self, file_id, file_parents,
2008
get_content_byte_lines, text_sha1=None,
2010
"""Record the text of file file_id
2012
:param file_id: The file_id of the file to record the text of.
2013
:param file_parents: The per-file parent revision ids.
2014
:param get_content_byte_lines: A callable which will return the byte
2016
:param text_sha1: Optional SHA1 of the file contents.
2017
:param text_size: Optional size of the file contents.
2019
# mutter('storing text of file {%s} in revision {%s} into %r',
2020
# file_id, self._new_revision_id, self.repository.weave_store)
2021
# special case to avoid diffing on renames or
2023
if (len(file_parents) == 1
2024
and text_sha1 == file_parents.values()[0].text_sha1
2025
and text_size == file_parents.values()[0].text_size):
2026
previous_ie = file_parents.values()[0]
2027
versionedfile = self.repository.weave_store.get_weave(file_id,
2028
self.repository.get_transaction())
2029
versionedfile.clone_text(self._new_revision_id,
2030
previous_ie.revision, file_parents.keys())
2031
return text_sha1, text_size
2033
new_lines = get_content_byte_lines()
2034
# TODO: Rather than invoking sha_strings here, _add_text_to_weave
2035
# should return the SHA1 and size
2036
self._add_text_to_weave(file_id, new_lines, file_parents.keys())
2037
return osutils.sha_strings(new_lines), \
2038
sum(map(len, new_lines))
2040
def modified_link(self, file_id, file_parents, link_target):
2041
"""Record the presence of a symbolic link.
2043
:param file_id: The file_id of the link to record.
2044
:param file_parents: The per-file parent revision ids.
2045
:param link_target: Target location of this link.
2047
self._add_text_to_weave(file_id, [], file_parents.keys())
2049
def _add_text_to_weave(self, file_id, new_lines, parents):
2050
versionedfile = self.repository.weave_store.get_weave_or_empty(
2051
file_id, self.repository.get_transaction())
2052
versionedfile.add_lines(self._new_revision_id, parents, new_lines)
2053
versionedfile.clear_cache()
2056
class _CommitBuilder(CommitBuilder):
2057
"""Temporary class so old CommitBuilders are detected properly
2059
Note: CommitBuilder works whether or not root entry is recorded.
2062
record_root_entry = True
2065
class RootCommitBuilder(CommitBuilder):
2066
"""This commitbuilder actually records the root id"""
2068
record_root_entry = True
2070
def record_entry_contents(self, ie, parent_invs, path, tree):
2071
"""Record the content of ie from tree into the commit if needed.
2073
Side effect: sets ie.revision when unchanged
2075
:param ie: An inventory entry present in the commit.
2076
:param parent_invs: The inventories of the parent revisions of the
2078
:param path: The path the entry is at in the tree.
2079
:param tree: The tree which contains this entry and should be used to
2082
assert self.new_inventory.root is not None or ie.parent_id is None
2083
self.new_inventory.add(ie)
2085
# ie.revision is always None if the InventoryEntry is considered
2086
# for committing. ie.snapshot will record the correct revision
2087
# which may be the sole parent if it is untouched.
2088
if ie.revision is not None:
2091
previous_entries = ie.find_previous_heads(
2093
self.repository.weave_store,
2094
self.repository.get_transaction())
2095
# we are creating a new revision for ie in the history store
2097
ie.snapshot(self._new_revision_id, path, previous_entries, tree, self)
2109
def _unescaper(match, _map=_unescape_map):
2110
code = match.group(1)
4507
#skip the last revision in the list
4510
if (stop_index is not None and
4511
len(partial_history_cache) > stop_index):
4513
if partial_history_cache[-1] == stop_revision:
4515
revision_id = iterator.next()
4516
partial_history_cache.append(revision_id)
4517
except StopIteration:
2114
if not code.startswith('#'):
2116
return unichr(int(code[1:])).encode('utf8')
2122
def _unescape_xml(data):
2123
"""Unescape predefined XML entities in a string of data."""
2125
if _unescape_re is None:
2126
_unescape_re = re.compile('\&([^;]*);')
2127
return _unescape_re.sub(_unescaper, data)
added
removed
bzrlib/repository.py
