1
# Copyright (C) 2005, 2006, 2007, 2008, 2009 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from bzrlib.lazy_import import lazy_import
18
lazy_import(globals(), """
38
revision as _mod_revision,
44
from bzrlib.bundle import serializer
45
from bzrlib.revisiontree import RevisionTree
46
from bzrlib.store.versioned import VersionedFileStore
47
from bzrlib.testament import Testament
50
from bzrlib.decorators import needs_read_lock, needs_write_lock
51
from bzrlib.inter import InterObject
52
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
53
from bzrlib import registry
54
from bzrlib.symbol_versioning import (
60
from bzrlib.trace import (
61
log_exception_quietly, note, mutter, mutter_callsite, warning)
64
# Old formats display a warning, but only once
65
_deprecation_warning_done = False
68
class CommitBuilder(object):
69
"""Provides an interface to build up a commit.
71
This allows describing a tree to be committed without needing to
72
know the internals of the format of the repository.
75
# all clients should supply tree roots.
76
record_root_entry = True
77
# the default CommitBuilder does not manage trees whose root is versioned.
78
_versioned_root = False
80
def __init__(self, repository, parents, config, timestamp=None,
81
timezone=None, committer=None, revprops=None,
83
"""Initiate a CommitBuilder.
85
:param repository: Repository to commit to.
86
:param parents: Revision ids of the parents of the new revision.
87
:param config: Configuration to use.
88
:param timestamp: Optional timestamp recorded for commit.
89
:param timezone: Optional timezone for timestamp.
90
:param committer: Optional committer to set for commit.
91
:param revprops: Optional dictionary of revision properties.
92
:param revision_id: Optional revision id.
97
self._committer = self._config.username()
99
self._committer = committer
101
self.new_inventory = Inventory(None)
102
self._new_revision_id = revision_id
103
self.parents = parents
104
self.repository = repository
107
if revprops is not None:
108
self._validate_revprops(revprops)
109
self._revprops.update(revprops)
111
if timestamp is None:
112
timestamp = time.time()
113
# Restrict resolution to 1ms
114
self._timestamp = round(timestamp, 3)
117
self._timezone = osutils.local_time_offset()
119
self._timezone = int(timezone)
121
self._generate_revision_if_needed()
122
self.__heads = graph.HeadsCache(repository.get_graph()).heads
123
self._basis_delta = []
124
# API compatibility, older code that used CommitBuilder did not call
125
# .record_delete(), which means the delta that is computed would not be
126
# valid. Callers that will call record_delete() should call
127
# .will_record_deletes() to indicate that.
128
self._recording_deletes = False
130
def _validate_unicode_text(self, text, context):
131
"""Verify things like commit messages don't have bogus characters."""
133
raise ValueError('Invalid value for %s: %r' % (context, text))
135
def _validate_revprops(self, revprops):
136
for key, value in revprops.iteritems():
137
# We know that the XML serializers do not round trip '\r'
138
# correctly, so refuse to accept them
139
if not isinstance(value, basestring):
140
raise ValueError('revision property (%s) is not a valid'
141
' (unicode) string: %r' % (key, value))
142
self._validate_unicode_text(value,
143
'revision property (%s)' % (key,))
145
def commit(self, message):
146
"""Make the actual commit.
148
:return: The revision id of the recorded revision.
150
self._validate_unicode_text(message, 'commit message')
151
rev = _mod_revision.Revision(
152
timestamp=self._timestamp,
153
timezone=self._timezone,
154
committer=self._committer,
156
inventory_sha1=self.inv_sha1,
157
revision_id=self._new_revision_id,
158
properties=self._revprops)
159
rev.parent_ids = self.parents
160
self.repository.add_revision(self._new_revision_id, rev,
161
self.new_inventory, self._config)
162
self.repository.commit_write_group()
163
return self._new_revision_id
166
"""Abort the commit that is being built.
168
self.repository.abort_write_group()
170
def revision_tree(self):
171
"""Return the tree that was just committed.
173
After calling commit() this can be called to get a RevisionTree
174
representing the newly committed tree. This is preferred to
175
calling Repository.revision_tree() because that may require
176
deserializing the inventory, while we already have a copy in
179
return RevisionTree(self.repository, self.new_inventory,
180
self._new_revision_id)
182
def finish_inventory(self):
183
"""Tell the builder that the inventory is finished."""
184
if self.new_inventory.root is None:
185
raise AssertionError('Root entry should be supplied to'
186
' record_entry_contents, as of bzr 0.10.')
187
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
188
self.new_inventory.revision_id = self._new_revision_id
189
self.inv_sha1 = self.repository.add_inventory(
190
self._new_revision_id,
195
def _gen_revision_id(self):
196
"""Return new revision-id."""
197
return generate_ids.gen_revision_id(self._config.username(),
200
def _generate_revision_if_needed(self):
201
"""Create a revision id if None was supplied.
203
If the repository can not support user-specified revision ids
204
they should override this function and raise CannotSetRevisionId
205
if _new_revision_id is not None.
207
:raises: CannotSetRevisionId
209
if self._new_revision_id is None:
210
self._new_revision_id = self._gen_revision_id()
211
self.random_revid = True
213
self.random_revid = False
215
def _heads(self, file_id, revision_ids):
216
"""Calculate the graph heads for revision_ids in the graph of file_id.
218
This can use either a per-file graph or a global revision graph as we
219
have an identity relationship between the two graphs.
221
return self.__heads(revision_ids)
223
def _check_root(self, ie, parent_invs, tree):
224
"""Helper for record_entry_contents.
226
:param ie: An entry being added.
227
:param parent_invs: The inventories of the parent revisions of the
229
:param tree: The tree that is being committed.
231
# In this revision format, root entries have no knit or weave When
232
# serializing out to disk and back in root.revision is always
234
ie.revision = self._new_revision_id
236
def _get_delta(self, ie, basis_inv, path):
237
"""Get a delta against the basis inventory for ie."""
238
if ie.file_id not in basis_inv:
240
result = (None, path, ie.file_id, ie)
241
self._basis_delta.append(result)
243
elif ie != basis_inv[ie.file_id]:
245
# TODO: avoid tis id2path call.
246
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
247
self._basis_delta.append(result)
253
def get_basis_delta(self):
254
"""Return the complete inventory delta versus the basis inventory.
256
This has been built up with the calls to record_delete and
257
record_entry_contents. The client must have already called
258
will_record_deletes() to indicate that they will be generating a
261
:return: An inventory delta, suitable for use with apply_delta, or
262
Repository.add_inventory_by_delta, etc.
264
if not self._recording_deletes:
265
raise AssertionError("recording deletes not activated.")
266
return self._basis_delta
268
def record_delete(self, path, file_id):
269
"""Record that a delete occured against a basis tree.
271
This is an optional API - when used it adds items to the basis_delta
272
being accumulated by the commit builder. It cannot be called unless the
273
method will_record_deletes() has been called to inform the builder that
274
a delta is being supplied.
276
:param path: The path of the thing deleted.
277
:param file_id: The file id that was deleted.
279
if not self._recording_deletes:
280
raise AssertionError("recording deletes not activated.")
281
delta = (path, None, file_id, None)
282
self._basis_delta.append(delta)
285
def will_record_deletes(self):
286
"""Tell the commit builder that deletes are being notified.
288
This enables the accumulation of an inventory delta; for the resulting
289
commit to be valid, deletes against the basis MUST be recorded via
290
builder.record_delete().
292
self._recording_deletes = True
294
def record_entry_contents(self, ie, parent_invs, path, tree,
296
"""Record the content of ie from tree into the commit if needed.
298
Side effect: sets ie.revision when unchanged
300
:param ie: An inventory entry present in the commit.
301
:param parent_invs: The inventories of the parent revisions of the
303
:param path: The path the entry is at in the tree.
304
:param tree: The tree which contains this entry and should be used to
306
:param content_summary: Summary data from the tree about the paths
307
content - stat, length, exec, sha/link target. This is only
308
accessed when the entry has a revision of None - that is when it is
309
a candidate to commit.
310
:return: A tuple (change_delta, version_recorded, fs_hash).
311
change_delta is an inventory_delta change for this entry against
312
the basis tree of the commit, or None if no change occured against
314
version_recorded is True if a new version of the entry has been
315
recorded. For instance, committing a merge where a file was only
316
changed on the other side will return (delta, False).
317
fs_hash is either None, or the hash details for the path (currently
318
a tuple of the contents sha1 and the statvalue returned by
319
tree.get_file_with_stat()).
321
if self.new_inventory.root is None:
322
if ie.parent_id is not None:
323
raise errors.RootMissing()
324
self._check_root(ie, parent_invs, tree)
325
if ie.revision is None:
326
kind = content_summary[0]
328
# ie is carried over from a prior commit
330
# XXX: repository specific check for nested tree support goes here - if
331
# the repo doesn't want nested trees we skip it ?
332
if (kind == 'tree-reference' and
333
not self.repository._format.supports_tree_reference):
334
# mismatch between commit builder logic and repository:
335
# this needs the entry creation pushed down into the builder.
336
raise NotImplementedError('Missing repository subtree support.')
337
self.new_inventory.add(ie)
339
# TODO: slow, take it out of the inner loop.
341
basis_inv = parent_invs[0]
343
basis_inv = Inventory(root_id=None)
345
# ie.revision is always None if the InventoryEntry is considered
346
# for committing. We may record the previous parents revision if the
347
# content is actually unchanged against a sole head.
348
if ie.revision is not None:
349
if not self._versioned_root and path == '':
350
# repositories that do not version the root set the root's
351
# revision to the new commit even when no change occurs (more
352
# specifically, they do not record a revision on the root; and
353
# the rev id is assigned to the root during deserialisation -
354
# this masks when a change may have occurred against the basis.
355
# To match this we always issue a delta, because the revision
356
# of the root will always be changing.
357
if ie.file_id in basis_inv:
358
delta = (basis_inv.id2path(ie.file_id), path,
362
delta = (None, path, ie.file_id, ie)
363
self._basis_delta.append(delta)
364
return delta, False, None
366
# we don't need to commit this, because the caller already
367
# determined that an existing revision of this file is
368
# appropriate. If its not being considered for committing then
369
# it and all its parents to the root must be unaltered so
370
# no-change against the basis.
371
if ie.revision == self._new_revision_id:
372
raise AssertionError("Impossible situation, a skipped "
373
"inventory entry (%r) claims to be modified in this "
374
"commit (%r).", (ie, self._new_revision_id))
375
return None, False, None
376
# XXX: Friction: parent_candidates should return a list not a dict
377
# so that we don't have to walk the inventories again.
378
parent_candiate_entries = ie.parent_candidates(parent_invs)
379
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
381
for inv in parent_invs:
382
if ie.file_id in inv:
383
old_rev = inv[ie.file_id].revision
384
if old_rev in head_set:
385
heads.append(inv[ie.file_id].revision)
386
head_set.remove(inv[ie.file_id].revision)
389
# now we check to see if we need to write a new record to the
391
# We write a new entry unless there is one head to the ancestors, and
392
# the kind-derived content is unchanged.
394
# Cheapest check first: no ancestors, or more the one head in the
395
# ancestors, we write a new node.
399
# There is a single head, look it up for comparison
400
parent_entry = parent_candiate_entries[heads[0]]
401
# if the non-content specific data has changed, we'll be writing a
403
if (parent_entry.parent_id != ie.parent_id or
404
parent_entry.name != ie.name):
406
# now we need to do content specific checks:
408
# if the kind changed the content obviously has
409
if kind != parent_entry.kind:
411
# Stat cache fingerprint feedback for the caller - None as we usually
412
# don't generate one.
415
if content_summary[2] is None:
416
raise ValueError("Files must not have executable = None")
418
if (# if the file length changed we have to store:
419
parent_entry.text_size != content_summary[1] or
420
# if the exec bit has changed we have to store:
421
parent_entry.executable != content_summary[2]):
423
elif parent_entry.text_sha1 == content_summary[3]:
424
# all meta and content is unchanged (using a hash cache
425
# hit to check the sha)
426
ie.revision = parent_entry.revision
427
ie.text_size = parent_entry.text_size
428
ie.text_sha1 = parent_entry.text_sha1
429
ie.executable = parent_entry.executable
430
return self._get_delta(ie, basis_inv, path), False, None
432
# Either there is only a hash change(no hash cache entry,
433
# or same size content change), or there is no change on
435
# Provide the parent's hash to the store layer, so that the
436
# content is unchanged we will not store a new node.
437
nostore_sha = parent_entry.text_sha1
439
# We want to record a new node regardless of the presence or
440
# absence of a content change in the file.
442
ie.executable = content_summary[2]
443
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
445
lines = file_obj.readlines()
449
ie.text_sha1, ie.text_size = self._add_text_to_weave(
450
ie.file_id, lines, heads, nostore_sha)
451
# Let the caller know we generated a stat fingerprint.
452
fingerprint = (ie.text_sha1, stat_value)
453
except errors.ExistingContent:
454
# Turns out that the file content was unchanged, and we were
455
# only going to store a new node if it was changed. Carry over
457
ie.revision = parent_entry.revision
458
ie.text_size = parent_entry.text_size
459
ie.text_sha1 = parent_entry.text_sha1
460
ie.executable = parent_entry.executable
461
return self._get_delta(ie, basis_inv, path), False, None
462
elif kind == 'directory':
464
# all data is meta here, nothing specific to directory, so
466
ie.revision = parent_entry.revision
467
return self._get_delta(ie, basis_inv, path), False, None
469
self._add_text_to_weave(ie.file_id, lines, heads, None)
470
elif kind == 'symlink':
471
current_link_target = content_summary[3]
473
# symlink target is not generic metadata, check if it has
475
if current_link_target != parent_entry.symlink_target:
478
# unchanged, carry over.
479
ie.revision = parent_entry.revision
480
ie.symlink_target = parent_entry.symlink_target
481
return self._get_delta(ie, basis_inv, path), False, None
482
ie.symlink_target = current_link_target
484
self._add_text_to_weave(ie.file_id, lines, heads, None)
485
elif kind == 'tree-reference':
487
if content_summary[3] != parent_entry.reference_revision:
490
# unchanged, carry over.
491
ie.reference_revision = parent_entry.reference_revision
492
ie.revision = parent_entry.revision
493
return self._get_delta(ie, basis_inv, path), False, None
494
ie.reference_revision = content_summary[3]
496
self._add_text_to_weave(ie.file_id, lines, heads, None)
498
raise NotImplementedError('unknown kind')
499
ie.revision = self._new_revision_id
500
return self._get_delta(ie, basis_inv, path), True, fingerprint
502
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
503
# Note: as we read the content directly from the tree, we know its not
504
# been turned into unicode or badly split - but a broken tree
505
# implementation could give us bad output from readlines() so this is
506
# not a guarantee of safety. What would be better is always checking
507
# the content during test suite execution. RBC 20070912
508
parent_keys = tuple((file_id, parent) for parent in parents)
509
return self.repository.texts.add_lines(
510
(file_id, self._new_revision_id), parent_keys, new_lines,
511
nostore_sha=nostore_sha, random_id=self.random_revid,
512
check_content=False)[0:2]
515
class RootCommitBuilder(CommitBuilder):
516
"""This commitbuilder actually records the root id"""
518
# the root entry gets versioned properly by this builder.
519
_versioned_root = True
521
def _check_root(self, ie, parent_invs, tree):
522
"""Helper for record_entry_contents.
524
:param ie: An entry being added.
525
:param parent_invs: The inventories of the parent revisions of the
527
:param tree: The tree that is being committed.
531
######################################################################
534
class Repository(object):
535
"""Repository holding history for one or more branches.
537
The repository holds and retrieves historical information including
538
revisions and file history. It's normally accessed only by the Branch,
539
which views a particular line of development through that history.
541
The Repository builds on top of some byte storage facilies (the revisions,
542
signatures, inventories and texts attributes) and a Transport, which
543
respectively provide byte storage and a means to access the (possibly
546
The byte storage facilities are addressed via tuples, which we refer to
547
as 'keys' throughout the code base. Revision_keys, inventory_keys and
548
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
549
(file_id, revision_id). We use this interface because it allows low
550
friction with the underlying code that implements disk indices, network
551
encoding and other parts of bzrlib.
553
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
554
the serialised revisions for the repository. This can be used to obtain
555
revision graph information or to access raw serialised revisions.
556
The result of trying to insert data into the repository via this store
557
is undefined: it should be considered read-only except for implementors
559
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
560
the serialised signatures for the repository. This can be used to
561
obtain access to raw serialised signatures. The result of trying to
562
insert data into the repository via this store is undefined: it should
563
be considered read-only except for implementors of repositories.
564
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
565
the serialised inventories for the repository. This can be used to
566
obtain unserialised inventories. The result of trying to insert data
567
into the repository via this store is undefined: it should be
568
considered read-only except for implementors of repositories.
569
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
570
texts of files and directories for the repository. This can be used to
571
obtain file texts or file graphs. Note that Repository.iter_file_bytes
572
is usually a better interface for accessing file texts.
573
The result of trying to insert data into the repository via this store
574
is undefined: it should be considered read-only except for implementors
576
:ivar _transport: Transport for file access to repository, typically
577
pointing to .bzr/repository.
580
# What class to use for a CommitBuilder. Often its simpler to change this
581
# in a Repository class subclass rather than to override
582
# get_commit_builder.
583
_commit_builder_class = CommitBuilder
584
# The search regex used by xml based repositories to determine what things
585
# where changed in a single commit.
586
_file_ids_altered_regex = lazy_regex.lazy_compile(
587
r'file_id="(?P<file_id>[^"]+)"'
588
r'.* revision="(?P<revision_id>[^"]+)"'
591
def abort_write_group(self, suppress_errors=False):
592
"""Commit the contents accrued within the current write group.
594
:param suppress_errors: if true, abort_write_group will catch and log
595
unexpected errors that happen during the abort, rather than
596
allowing them to propagate. Defaults to False.
598
:seealso: start_write_group.
600
if self._write_group is not self.get_transaction():
601
# has an unlock or relock occured ?
602
raise errors.BzrError('mismatched lock context and write group.')
604
self._abort_write_group()
605
except Exception, exc:
606
self._write_group = None
607
if not suppress_errors:
609
mutter('abort_write_group failed')
610
log_exception_quietly()
611
note('bzr: ERROR (ignored): %s', exc)
612
self._write_group = None
614
def _abort_write_group(self):
615
"""Template method for per-repository write group cleanup.
617
This is called during abort before the write group is considered to be
618
finished and should cleanup any internal state accrued during the write
619
group. There is no requirement that data handed to the repository be
620
*not* made available - this is not a rollback - but neither should any
621
attempt be made to ensure that data added is fully commited. Abort is
622
invoked when an error has occured so futher disk or network operations
623
may not be possible or may error and if possible should not be
627
def add_fallback_repository(self, repository):
628
"""Add a repository to use for looking up data not held locally.
630
:param repository: A repository.
632
if not self._format.supports_external_lookups:
633
raise errors.UnstackableRepositoryFormat(self._format, self.base)
634
self._check_fallback_repository(repository)
635
self._fallback_repositories.append(repository)
636
self.texts.add_fallback_versioned_files(repository.texts)
637
self.inventories.add_fallback_versioned_files(repository.inventories)
638
self.revisions.add_fallback_versioned_files(repository.revisions)
639
self.signatures.add_fallback_versioned_files(repository.signatures)
641
def _check_fallback_repository(self, repository):
642
"""Check that this repository can fallback to repository safely.
644
Raise an error if not.
646
:param repository: A repository to fallback to.
648
return InterRepository._assert_same_model(self, repository)
650
def add_inventory(self, revision_id, inv, parents):
651
"""Add the inventory inv to the repository as revision_id.
653
:param parents: The revision ids of the parents that revision_id
654
is known to have and are in the repository already.
656
:returns: The validator(which is a sha1 digest, though what is sha'd is
657
repository format specific) of the serialized inventory.
659
if not self.is_in_write_group():
660
raise AssertionError("%r not in write group" % (self,))
661
_mod_revision.check_not_reserved_id(revision_id)
662
if not (inv.revision_id is None or inv.revision_id == revision_id):
663
raise AssertionError(
664
"Mismatch between inventory revision"
665
" id and insertion revid (%r, %r)"
666
% (inv.revision_id, revision_id))
668
raise AssertionError()
669
inv_lines = self._serialise_inventory_to_lines(inv)
670
return self._inventory_add_lines(revision_id, parents,
671
inv_lines, check_content=False)
673
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
675
"""Add a new inventory expressed as a delta against another revision.
677
:param basis_revision_id: The inventory id the delta was created
678
against. (This does not have to be a direct parent.)
679
:param delta: The inventory delta (see Inventory.apply_delta for
681
:param new_revision_id: The revision id that the inventory is being
683
:param parents: The revision ids of the parents that revision_id is
684
known to have and are in the repository already. These are supplied
685
for repositories that depend on the inventory graph for revision
686
graph access, as well as for those that pun ancestry with delta
689
:returns: (validator, new_inv)
690
The validator(which is a sha1 digest, though what is sha'd is
691
repository format specific) of the serialized inventory, and the
694
if not self.is_in_write_group():
695
raise AssertionError("%r not in write group" % (self,))
696
_mod_revision.check_not_reserved_id(new_revision_id)
697
basis_tree = self.revision_tree(basis_revision_id)
698
basis_tree.lock_read()
700
# Note that this mutates the inventory of basis_tree, which not all
701
# inventory implementations may support: A better idiom would be to
702
# return a new inventory, but as there is no revision tree cache in
703
# repository this is safe for now - RBC 20081013
704
basis_inv = basis_tree.inventory
705
basis_inv.apply_delta(delta)
706
basis_inv.revision_id = new_revision_id
707
return (self.add_inventory(new_revision_id, basis_inv, parents),
712
def _inventory_add_lines(self, revision_id, parents, lines,
714
"""Store lines in inv_vf and return the sha1 of the inventory."""
715
parents = [(parent,) for parent in parents]
716
return self.inventories.add_lines((revision_id,), parents, lines,
717
check_content=check_content)[0]
719
def add_revision(self, revision_id, rev, inv=None, config=None):
720
"""Add rev to the revision store as revision_id.
722
:param revision_id: the revision id to use.
723
:param rev: The revision object.
724
:param inv: The inventory for the revision. if None, it will be looked
725
up in the inventory storer
726
:param config: If None no digital signature will be created.
727
If supplied its signature_needed method will be used
728
to determine if a signature should be made.
730
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
732
_mod_revision.check_not_reserved_id(revision_id)
733
if config is not None and config.signature_needed():
735
inv = self.get_inventory(revision_id)
736
plaintext = Testament(rev, inv).as_short_text()
737
self.store_revision_signature(
738
gpg.GPGStrategy(config), plaintext, revision_id)
739
# check inventory present
740
if not self.inventories.get_parent_map([(revision_id,)]):
742
raise errors.WeaveRevisionNotPresent(revision_id,
745
# yes, this is not suitable for adding with ghosts.
746
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
750
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
751
self._add_revision(rev)
753
def _add_revision(self, revision):
754
text = self._serializer.write_revision_to_string(revision)
755
key = (revision.revision_id,)
756
parents = tuple((parent,) for parent in revision.parent_ids)
757
self.revisions.add_lines(key, parents, osutils.split_lines(text))
759
def all_revision_ids(self):
760
"""Returns a list of all the revision ids in the repository.
762
This is conceptually deprecated because code should generally work on
763
the graph reachable from a particular revision, and ignore any other
764
revisions that might be present. There is no direct replacement
767
if 'evil' in debug.debug_flags:
768
mutter_callsite(2, "all_revision_ids is linear with history.")
769
return self._all_revision_ids()
771
def _all_revision_ids(self):
772
"""Returns a list of all the revision ids in the repository.
774
These are in as much topological order as the underlying store can
777
raise NotImplementedError(self._all_revision_ids)
779
def break_lock(self):
780
"""Break a lock if one is present from another instance.
782
Uses the ui factory to ask for confirmation if the lock may be from
785
self.control_files.break_lock()
788
def _eliminate_revisions_not_present(self, revision_ids):
789
"""Check every revision id in revision_ids to see if we have it.
791
Returns a set of the present revisions.
794
graph = self.get_graph()
795
parent_map = graph.get_parent_map(revision_ids)
796
# The old API returned a list, should this actually be a set?
797
return parent_map.keys()
800
def create(a_bzrdir):
801
"""Construct the current default format repository in a_bzrdir."""
802
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
804
def __init__(self, _format, a_bzrdir, control_files):
805
"""instantiate a Repository.
807
:param _format: The format of the repository on disk.
808
:param a_bzrdir: The BzrDir of the repository.
810
In the future we will have a single api for all stores for
811
getting file texts, inventories and revisions, then
812
this construct will accept instances of those things.
814
super(Repository, self).__init__()
815
self._format = _format
816
# the following are part of the public API for Repository:
817
self.bzrdir = a_bzrdir
818
self.control_files = control_files
819
self._transport = control_files._transport
820
self.base = self._transport.base
822
self._reconcile_does_inventory_gc = True
823
self._reconcile_fixes_text_parents = False
824
self._reconcile_backsup_inventory = True
825
# not right yet - should be more semantically clear ?
827
# TODO: make sure to construct the right store classes, etc, depending
828
# on whether escaping is required.
829
self._warn_if_deprecated()
830
self._write_group = None
831
# Additional places to query for data.
832
self._fallback_repositories = []
833
# An InventoryEntry cache, used during deserialization
834
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
837
return '%s(%r)' % (self.__class__.__name__,
840
def has_same_location(self, other):
841
"""Returns a boolean indicating if this repository is at the same
842
location as another repository.
844
This might return False even when two repository objects are accessing
845
the same physical repository via different URLs.
847
if self.__class__ is not other.__class__:
849
return (self._transport.base == other._transport.base)
851
def is_in_write_group(self):
852
"""Return True if there is an open write group.
854
:seealso: start_write_group.
856
return self._write_group is not None
859
return self.control_files.is_locked()
861
def is_write_locked(self):
862
"""Return True if this object is write locked."""
863
return self.is_locked() and self.control_files._lock_mode == 'w'
865
def lock_write(self, token=None):
866
"""Lock this repository for writing.
868
This causes caching within the repository obejct to start accumlating
869
data during reads, and allows a 'write_group' to be obtained. Write
870
groups must be used for actual data insertion.
872
:param token: if this is already locked, then lock_write will fail
873
unless the token matches the existing lock.
874
:returns: a token if this instance supports tokens, otherwise None.
875
:raises TokenLockingNotSupported: when a token is given but this
876
instance doesn't support using token locks.
877
:raises MismatchedToken: if the specified token doesn't match the token
878
of the existing lock.
879
:seealso: start_write_group.
881
A token should be passed in if you know that you have locked the object
882
some other way, and need to synchronise this object's state with that
885
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
887
result = self.control_files.lock_write(token=token)
888
for repo in self._fallback_repositories:
889
# Writes don't affect fallback repos
895
self.control_files.lock_read()
896
for repo in self._fallback_repositories:
900
def get_physical_lock_status(self):
901
return self.control_files.get_physical_lock_status()
903
def leave_lock_in_place(self):
904
"""Tell this repository not to release the physical lock when this
907
If lock_write doesn't return a token, then this method is not supported.
909
self.control_files.leave_in_place()
911
def dont_leave_lock_in_place(self):
912
"""Tell this repository to release the physical lock when this
913
object is unlocked, even if it didn't originally acquire it.
915
If lock_write doesn't return a token, then this method is not supported.
917
self.control_files.dont_leave_in_place()
920
def gather_stats(self, revid=None, committers=None):
921
"""Gather statistics from a revision id.
923
:param revid: The revision id to gather statistics from, if None, then
924
no revision specific statistics are gathered.
925
:param committers: Optional parameter controlling whether to grab
926
a count of committers from the revision specific statistics.
927
:return: A dictionary of statistics. Currently this contains:
928
committers: The number of committers if requested.
929
firstrev: A tuple with timestamp, timezone for the penultimate left
930
most ancestor of revid, if revid is not the NULL_REVISION.
931
latestrev: A tuple with timestamp, timezone for revid, if revid is
932
not the NULL_REVISION.
933
revisions: The total revision count in the repository.
934
size: An estimate disk size of the repository in bytes.
937
if revid and committers:
938
result['committers'] = 0
939
if revid and revid != _mod_revision.NULL_REVISION:
941
all_committers = set()
942
revisions = self.get_ancestry(revid)
943
# pop the leading None
945
first_revision = None
947
# ignore the revisions in the middle - just grab first and last
948
revisions = revisions[0], revisions[-1]
949
for revision in self.get_revisions(revisions):
950
if not first_revision:
951
first_revision = revision
953
all_committers.add(revision.committer)
954
last_revision = revision
956
result['committers'] = len(all_committers)
957
result['firstrev'] = (first_revision.timestamp,
958
first_revision.timezone)
959
result['latestrev'] = (last_revision.timestamp,
960
last_revision.timezone)
962
# now gather global repository information
963
# XXX: This is available for many repos regardless of listability.
964
if self.bzrdir.root_transport.listable():
965
# XXX: do we want to __define len__() ?
966
# Maybe the versionedfiles object should provide a different
967
# method to get the number of keys.
968
result['revisions'] = len(self.revisions.keys())
972
def find_branches(self, using=False):
973
"""Find branches underneath this repository.
975
This will include branches inside other branches.
977
:param using: If True, list only branches using this repository.
979
if using and not self.is_shared():
981
return [self.bzrdir.open_branch()]
982
except errors.NotBranchError:
984
class Evaluator(object):
987
self.first_call = True
989
def __call__(self, bzrdir):
990
# On the first call, the parameter is always the bzrdir
991
# containing the current repo.
992
if not self.first_call:
994
repository = bzrdir.open_repository()
995
except errors.NoRepositoryPresent:
998
return False, (None, repository)
999
self.first_call = False
1001
value = (bzrdir.open_branch(), None)
1002
except errors.NotBranchError:
1003
value = (None, None)
1007
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1008
self.bzrdir.root_transport, evaluate=Evaluator()):
1009
if branch is not None:
1010
branches.append(branch)
1011
if not using and repository is not None:
1012
branches.extend(repository.find_branches())
1016
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1017
"""Return the revision ids that other has that this does not.
1019
These are returned in topological order.
1021
revision_id: only return revision ids included by revision_id.
1023
return InterRepository.get(other, self).search_missing_revision_ids(
1024
revision_id, find_ghosts)
1026
@deprecated_method(one_two)
1028
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1029
"""Return the revision ids that other has that this does not.
1031
These are returned in topological order.
1033
revision_id: only return revision ids included by revision_id.
1035
keys = self.search_missing_revision_ids(
1036
other, revision_id, find_ghosts).get_keys()
1039
parents = other.get_graph().get_parent_map(keys)
1042
return tsort.topo_sort(parents)
1046
"""Open the repository rooted at base.
1048
For instance, if the repository is at URL/.bzr/repository,
1049
Repository.open(URL) -> a Repository instance.
1051
control = bzrdir.BzrDir.open(base)
1052
return control.open_repository()
1054
def copy_content_into(self, destination, revision_id=None):
1055
"""Make a complete copy of the content in self into destination.
1057
This is a destructive operation! Do not use it on existing
1060
return InterRepository.get(self, destination).copy_content(revision_id)
1062
def commit_write_group(self):
1063
"""Commit the contents accrued within the current write group.
1065
:seealso: start_write_group.
1067
if self._write_group is not self.get_transaction():
1068
# has an unlock or relock occured ?
1069
raise errors.BzrError('mismatched lock context %r and '
1071
(self.get_transaction(), self._write_group))
1072
self._commit_write_group()
1073
self._write_group = None
1075
def _commit_write_group(self):
1076
"""Template method for per-repository write group cleanup.
1078
This is called before the write group is considered to be
1079
finished and should ensure that all data handed to the repository
1080
for writing during the write group is safely committed (to the
1081
extent possible considering file system caching etc).
1084
def suspend_write_group(self):
1085
raise errors.UnsuspendableWriteGroup(self)
1087
def resume_write_group(self, tokens):
1088
if not self.is_write_locked():
1089
raise errors.NotWriteLocked(self)
1090
if self._write_group:
1091
raise errors.BzrError('already in a write group')
1092
self._resume_write_group(tokens)
1093
# so we can detect unlock/relock - the write group is now entered.
1094
self._write_group = self.get_transaction()
1096
def _resume_write_group(self, tokens):
1097
raise errors.UnsuspendableWriteGroup(self)
1099
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
1101
"""Fetch the content required to construct revision_id from source.
1103
If revision_id is None and fetch_spec is None, then all content is
1106
fetch() may not be used when the repository is in a write group -
1107
either finish the current write group before using fetch, or use
1108
fetch before starting the write group.
1110
:param find_ghosts: Find and copy revisions in the source that are
1111
ghosts in the target (and not reachable directly by walking out to
1112
the first-present revision in target from revision_id).
1113
:param revision_id: If specified, all the content needed for this
1114
revision ID will be copied to the target. Fetch will determine for
1115
itself which content needs to be copied.
1116
:param fetch_spec: If specified, a SearchResult or
1117
PendingAncestryResult that describes which revisions to copy. This
1118
allows copying multiple heads at once. Mutually exclusive with
1121
if fetch_spec is not None and revision_id is not None:
1122
raise AssertionError(
1123
"fetch_spec and revision_id are mutually exclusive.")
1124
if self.is_in_write_group():
1125
raise errors.BzrError("May not fetch while in a write group.")
1126
# fast path same-url fetch operations
1127
if self.has_same_location(source) and fetch_spec is None:
1128
# check that last_revision is in 'from' and then return a
1130
if (revision_id is not None and
1131
not _mod_revision.is_null(revision_id)):
1132
self.get_revision(revision_id)
1134
# if there is no specific appropriate InterRepository, this will get
1135
# the InterRepository base class, which raises an
1136
# IncompatibleRepositories when asked to fetch.
1137
inter = InterRepository.get(source, self)
1138
return inter.fetch(revision_id=revision_id, pb=pb,
1139
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
1141
def create_bundle(self, target, base, fileobj, format=None):
1142
return serializer.write_bundle(self, target, base, fileobj, format)
1144
def get_commit_builder(self, branch, parents, config, timestamp=None,
1145
timezone=None, committer=None, revprops=None,
1147
"""Obtain a CommitBuilder for this repository.
1149
:param branch: Branch to commit to.
1150
:param parents: Revision ids of the parents of the new revision.
1151
:param config: Configuration to use.
1152
:param timestamp: Optional timestamp recorded for commit.
1153
:param timezone: Optional timezone for timestamp.
1154
:param committer: Optional committer to set for commit.
1155
:param revprops: Optional dictionary of revision properties.
1156
:param revision_id: Optional revision id.
1158
result = self._commit_builder_class(self, parents, config,
1159
timestamp, timezone, committer, revprops, revision_id)
1160
self.start_write_group()
1164
if (self.control_files._lock_count == 1 and
1165
self.control_files._lock_mode == 'w'):
1166
if self._write_group is not None:
1167
self.abort_write_group()
1168
self.control_files.unlock()
1169
raise errors.BzrError(
1170
'Must end write groups before releasing write locks.')
1171
self.control_files.unlock()
1172
if self.control_files._lock_count == 0:
1173
self._inventory_entry_cache.clear()
1174
for repo in self._fallback_repositories:
1178
def clone(self, a_bzrdir, revision_id=None):
1179
"""Clone this repository into a_bzrdir using the current format.
1181
Currently no check is made that the format of this repository and
1182
the bzrdir format are compatible. FIXME RBC 20060201.
1184
:return: The newly created destination repository.
1186
# TODO: deprecate after 0.16; cloning this with all its settings is
1187
# probably not very useful -- mbp 20070423
1188
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1189
self.copy_content_into(dest_repo, revision_id)
1192
def start_write_group(self):
1193
"""Start a write group in the repository.
1195
Write groups are used by repositories which do not have a 1:1 mapping
1196
between file ids and backend store to manage the insertion of data from
1197
both fetch and commit operations.
1199
A write lock is required around the start_write_group/commit_write_group
1200
for the support of lock-requiring repository formats.
1202
One can only insert data into a repository inside a write group.
1206
if not self.is_write_locked():
1207
raise errors.NotWriteLocked(self)
1208
if self._write_group:
1209
raise errors.BzrError('already in a write group')
1210
self._start_write_group()
1211
# so we can detect unlock/relock - the write group is now entered.
1212
self._write_group = self.get_transaction()
1214
def _start_write_group(self):
1215
"""Template method for per-repository write group startup.
1217
This is called before the write group is considered to be
1222
def sprout(self, to_bzrdir, revision_id=None):
1223
"""Create a descendent repository for new development.
1225
Unlike clone, this does not copy the settings of the repository.
1227
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1228
dest_repo.fetch(self, revision_id=revision_id)
1231
def _create_sprouting_repo(self, a_bzrdir, shared):
1232
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1233
# use target default format.
1234
dest_repo = a_bzrdir.create_repository()
1236
# Most control formats need the repository to be specifically
1237
# created, but on some old all-in-one formats it's not needed
1239
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1240
except errors.UninitializableFormat:
1241
dest_repo = a_bzrdir.open_repository()
1244
def _get_sink(self):
1245
"""Return a sink for streaming into this repository."""
1246
return StreamSink(self)
1248
def _get_source(self, to_format):
1249
"""Return a source for streaming from this repository."""
1250
return StreamSource(self, to_format)
1253
def has_revision(self, revision_id):
1254
"""True if this repository has a copy of the revision."""
1255
return revision_id in self.has_revisions((revision_id,))
1258
def has_revisions(self, revision_ids):
1259
"""Probe to find out the presence of multiple revisions.
1261
:param revision_ids: An iterable of revision_ids.
1262
:return: A set of the revision_ids that were present.
1264
parent_map = self.revisions.get_parent_map(
1265
[(rev_id,) for rev_id in revision_ids])
1267
if _mod_revision.NULL_REVISION in revision_ids:
1268
result.add(_mod_revision.NULL_REVISION)
1269
result.update([key[0] for key in parent_map])
1273
def get_revision(self, revision_id):
1274
"""Return the Revision object for a named revision."""
1275
return self.get_revisions([revision_id])[0]
1278
def get_revision_reconcile(self, revision_id):
1279
"""'reconcile' helper routine that allows access to a revision always.
1281
This variant of get_revision does not cross check the weave graph
1282
against the revision one as get_revision does: but it should only
1283
be used by reconcile, or reconcile-alike commands that are correcting
1284
or testing the revision graph.
1286
return self._get_revisions([revision_id])[0]
1289
def get_revisions(self, revision_ids):
1290
"""Get many revisions at once."""
1291
return self._get_revisions(revision_ids)
1294
def _get_revisions(self, revision_ids):
1295
"""Core work logic to get many revisions without sanity checks."""
1296
for rev_id in revision_ids:
1297
if not rev_id or not isinstance(rev_id, basestring):
1298
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1299
keys = [(key,) for key in revision_ids]
1300
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1302
for record in stream:
1303
if record.storage_kind == 'absent':
1304
raise errors.NoSuchRevision(self, record.key[0])
1305
text = record.get_bytes_as('fulltext')
1306
rev = self._serializer.read_revision_from_string(text)
1307
revs[record.key[0]] = rev
1308
return [revs[revid] for revid in revision_ids]
1311
def get_revision_xml(self, revision_id):
1312
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1313
# would have already do it.
1314
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1315
rev = self.get_revision(revision_id)
1316
rev_tmp = cStringIO.StringIO()
1317
# the current serializer..
1318
self._serializer.write_revision(rev, rev_tmp)
1320
return rev_tmp.getvalue()
1322
def get_deltas_for_revisions(self, revisions):
1323
"""Produce a generator of revision deltas.
1325
Note that the input is a sequence of REVISIONS, not revision_ids.
1326
Trees will be held in memory until the generator exits.
1327
Each delta is relative to the revision's lefthand predecessor.
1329
required_trees = set()
1330
for revision in revisions:
1331
required_trees.add(revision.revision_id)
1332
required_trees.update(revision.parent_ids[:1])
1333
trees = dict((t.get_revision_id(), t) for
1334
t in self.revision_trees(required_trees))
1335
for revision in revisions:
1336
if not revision.parent_ids:
1337
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1339
old_tree = trees[revision.parent_ids[0]]
1340
yield trees[revision.revision_id].changes_from(old_tree)
1343
def get_revision_delta(self, revision_id):
1344
"""Return the delta for one revision.
1346
The delta is relative to the left-hand predecessor of the
1349
r = self.get_revision(revision_id)
1350
return list(self.get_deltas_for_revisions([r]))[0]
1353
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1354
signature = gpg_strategy.sign(plaintext)
1355
self.add_signature_text(revision_id, signature)
1358
def add_signature_text(self, revision_id, signature):
1359
self.signatures.add_lines((revision_id,), (),
1360
osutils.split_lines(signature))
1362
def find_text_key_references(self):
1363
"""Find the text key references within the repository.
1365
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1366
revision_ids. Each altered file-ids has the exact revision_ids that
1367
altered it listed explicitly.
1368
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1369
to whether they were referred to by the inventory of the
1370
revision_id that they contain. The inventory texts from all present
1371
revision ids are assessed to generate this report.
1373
revision_keys = self.revisions.keys()
1374
w = self.inventories
1375
pb = ui.ui_factory.nested_progress_bar()
1377
return self._find_text_key_references_from_xml_inventory_lines(
1378
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1382
def _find_text_key_references_from_xml_inventory_lines(self,
1384
"""Core routine for extracting references to texts from inventories.
1386
This performs the translation of xml lines to revision ids.
1388
:param line_iterator: An iterator of lines, origin_version_id
1389
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1390
to whether they were referred to by the inventory of the
1391
revision_id that they contain. Note that if that revision_id was
1392
not part of the line_iterator's output then False will be given -
1393
even though it may actually refer to that key.
1395
if not self._serializer.support_altered_by_hack:
1396
raise AssertionError(
1397
"_find_text_key_references_from_xml_inventory_lines only "
1398
"supported for branches which store inventory as unnested xml"
1399
", not on %r" % self)
1402
# this code needs to read every new line in every inventory for the
1403
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1404
# not present in one of those inventories is unnecessary but not
1405
# harmful because we are filtering by the revision id marker in the
1406
# inventory lines : we only select file ids altered in one of those
1407
# revisions. We don't need to see all lines in the inventory because
1408
# only those added in an inventory in rev X can contain a revision=X
1410
unescape_revid_cache = {}
1411
unescape_fileid_cache = {}
1413
# jam 20061218 In a big fetch, this handles hundreds of thousands
1414
# of lines, so it has had a lot of inlining and optimizing done.
1415
# Sorry that it is a little bit messy.
1416
# Move several functions to be local variables, since this is a long
1418
search = self._file_ids_altered_regex.search
1419
unescape = _unescape_xml
1420
setdefault = result.setdefault
1421
for line, line_key in line_iterator:
1422
match = search(line)
1425
# One call to match.group() returning multiple items is quite a
1426
# bit faster than 2 calls to match.group() each returning 1
1427
file_id, revision_id = match.group('file_id', 'revision_id')
1429
# Inlining the cache lookups helps a lot when you make 170,000
1430
# lines and 350k ids, versus 8.4 unique ids.
1431
# Using a cache helps in 2 ways:
1432
# 1) Avoids unnecessary decoding calls
1433
# 2) Re-uses cached strings, which helps in future set and
1435
# (2) is enough that removing encoding entirely along with
1436
# the cache (so we are using plain strings) results in no
1437
# performance improvement.
1439
revision_id = unescape_revid_cache[revision_id]
1441
unescaped = unescape(revision_id)
1442
unescape_revid_cache[revision_id] = unescaped
1443
revision_id = unescaped
1445
# Note that unconditionally unescaping means that we deserialise
1446
# every fileid, which for general 'pull' is not great, but we don't
1447
# really want to have some many fulltexts that this matters anyway.
1450
file_id = unescape_fileid_cache[file_id]
1452
unescaped = unescape(file_id)
1453
unescape_fileid_cache[file_id] = unescaped
1456
key = (file_id, revision_id)
1457
setdefault(key, False)
1458
if revision_id == line_key[-1]:
1462
def _inventory_xml_lines_for_keys(self, keys):
1463
"""Get a line iterator of the sort needed for findind references.
1465
Not relevant for non-xml inventory repositories.
1467
Ghosts in revision_keys are ignored.
1469
:param revision_keys: The revision keys for the inventories to inspect.
1470
:return: An iterator over (inventory line, revid) for the fulltexts of
1471
all of the xml inventories specified by revision_keys.
1473
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1474
for record in stream:
1475
if record.storage_kind != 'absent':
1476
chunks = record.get_bytes_as('chunked')
1477
revid = record.key[-1]
1478
lines = osutils.chunks_to_lines(chunks)
1482
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1484
"""Helper routine for fileids_altered_by_revision_ids.
1486
This performs the translation of xml lines to revision ids.
1488
:param line_iterator: An iterator of lines, origin_version_id
1489
:param revision_ids: The revision ids to filter for. This should be a
1490
set or other type which supports efficient __contains__ lookups, as
1491
the revision id from each parsed line will be looked up in the
1492
revision_ids filter.
1493
:return: a dictionary mapping altered file-ids to an iterable of
1494
revision_ids. Each altered file-ids has the exact revision_ids that
1495
altered it listed explicitly.
1497
seen = set(self._find_text_key_references_from_xml_inventory_lines(
1498
line_iterator).iterkeys())
1499
# Note that revision_ids are revision keys.
1500
parent_maps = self.revisions.get_parent_map(revision_ids)
1502
map(parents.update, parent_maps.itervalues())
1503
parents.difference_update(revision_ids)
1504
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
1505
self._inventory_xml_lines_for_keys(parents)))
1506
new_keys = seen - parent_seen
1508
setdefault = result.setdefault
1509
for key in new_keys:
1510
setdefault(key[0], set()).add(key[-1])
1513
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1514
"""Find the file ids and versions affected by revisions.
1516
:param revisions: an iterable containing revision ids.
1517
:param _inv_weave: The inventory weave from this repository or None.
1518
If None, the inventory weave will be opened automatically.
1519
:return: a dictionary mapping altered file-ids to an iterable of
1520
revision_ids. Each altered file-ids has the exact revision_ids that
1521
altered it listed explicitly.
1523
selected_keys = set((revid,) for revid in revision_ids)
1524
w = _inv_weave or self.inventories
1525
pb = ui.ui_factory.nested_progress_bar()
1527
return self._find_file_ids_from_xml_inventory_lines(
1528
w.iter_lines_added_or_present_in_keys(
1529
selected_keys, pb=pb),
1534
def iter_files_bytes(self, desired_files):
1535
"""Iterate through file versions.
1537
Files will not necessarily be returned in the order they occur in
1538
desired_files. No specific order is guaranteed.
1540
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1541
value supplied by the caller as part of desired_files. It should
1542
uniquely identify the file version in the caller's context. (Examples:
1543
an index number or a TreeTransform trans_id.)
1545
bytes_iterator is an iterable of bytestrings for the file. The
1546
kind of iterable and length of the bytestrings are unspecified, but for
1547
this implementation, it is a list of bytes produced by
1548
VersionedFile.get_record_stream().
1550
:param desired_files: a list of (file_id, revision_id, identifier)
1554
for file_id, revision_id, callable_data in desired_files:
1555
text_keys[(file_id, revision_id)] = callable_data
1556
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1557
if record.storage_kind == 'absent':
1558
raise errors.RevisionNotPresent(record.key, self)
1559
yield text_keys[record.key], record.get_bytes_as('fulltext')
1561
def _generate_text_key_index(self, text_key_references=None,
1563
"""Generate a new text key index for the repository.
1565
This is an expensive function that will take considerable time to run.
1567
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1568
list of parents, also text keys. When a given key has no parents,
1569
the parents list will be [NULL_REVISION].
1571
# All revisions, to find inventory parents.
1572
if ancestors is None:
1573
graph = self.get_graph()
1574
ancestors = graph.get_parent_map(self.all_revision_ids())
1575
if text_key_references is None:
1576
text_key_references = self.find_text_key_references()
1577
pb = ui.ui_factory.nested_progress_bar()
1579
return self._do_generate_text_key_index(ancestors,
1580
text_key_references, pb)
1584
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1585
"""Helper for _generate_text_key_index to avoid deep nesting."""
1586
revision_order = tsort.topo_sort(ancestors)
1587
invalid_keys = set()
1589
for revision_id in revision_order:
1590
revision_keys[revision_id] = set()
1591
text_count = len(text_key_references)
1592
# a cache of the text keys to allow reuse; costs a dict of all the
1593
# keys, but saves a 2-tuple for every child of a given key.
1595
for text_key, valid in text_key_references.iteritems():
1597
invalid_keys.add(text_key)
1599
revision_keys[text_key[1]].add(text_key)
1600
text_key_cache[text_key] = text_key
1601
del text_key_references
1603
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1604
NULL_REVISION = _mod_revision.NULL_REVISION
1605
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1606
# too small for large or very branchy trees. However, for 55K path
1607
# trees, it would be easy to use too much memory trivially. Ideally we
1608
# could gauge this by looking at available real memory etc, but this is
1609
# always a tricky proposition.
1610
inventory_cache = lru_cache.LRUCache(10)
1611
batch_size = 10 # should be ~150MB on a 55K path tree
1612
batch_count = len(revision_order) / batch_size + 1
1614
pb.update("Calculating text parents", processed_texts, text_count)
1615
for offset in xrange(batch_count):
1616
to_query = revision_order[offset * batch_size:(offset + 1) *
1620
for rev_tree in self.revision_trees(to_query):
1621
revision_id = rev_tree.get_revision_id()
1622
parent_ids = ancestors[revision_id]
1623
for text_key in revision_keys[revision_id]:
1624
pb.update("Calculating text parents", processed_texts)
1625
processed_texts += 1
1626
candidate_parents = []
1627
for parent_id in parent_ids:
1628
parent_text_key = (text_key[0], parent_id)
1630
check_parent = parent_text_key not in \
1631
revision_keys[parent_id]
1633
# the parent parent_id is a ghost:
1634
check_parent = False
1635
# truncate the derived graph against this ghost.
1636
parent_text_key = None
1638
# look at the parent commit details inventories to
1639
# determine possible candidates in the per file graph.
1642
inv = inventory_cache[parent_id]
1644
inv = self.revision_tree(parent_id).inventory
1645
inventory_cache[parent_id] = inv
1646
parent_entry = inv._byid.get(text_key[0], None)
1647
if parent_entry is not None:
1649
text_key[0], parent_entry.revision)
1651
parent_text_key = None
1652
if parent_text_key is not None:
1653
candidate_parents.append(
1654
text_key_cache[parent_text_key])
1655
parent_heads = text_graph.heads(candidate_parents)
1656
new_parents = list(parent_heads)
1657
new_parents.sort(key=lambda x:candidate_parents.index(x))
1658
if new_parents == []:
1659
new_parents = [NULL_REVISION]
1660
text_index[text_key] = new_parents
1662
for text_key in invalid_keys:
1663
text_index[text_key] = [NULL_REVISION]
1666
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1667
"""Get an iterable listing the keys of all the data introduced by a set
1670
The keys will be ordered so that the corresponding items can be safely
1671
fetched and inserted in that order.
1673
:returns: An iterable producing tuples of (knit-kind, file-id,
1674
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1675
'revisions'. file-id is None unless knit-kind is 'file'.
1677
# XXX: it's a bit weird to control the inventory weave caching in this
1678
# generator. Ideally the caching would be done in fetch.py I think. Or
1679
# maybe this generator should explicitly have the contract that it
1680
# should not be iterated until the previously yielded item has been
1682
inv_w = self.inventories
1684
# file ids that changed
1685
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1687
num_file_ids = len(file_ids)
1688
for file_id, altered_versions in file_ids.iteritems():
1689
if _files_pb is not None:
1690
_files_pb.update("fetch texts", count, num_file_ids)
1692
yield ("file", file_id, altered_versions)
1693
# We're done with the files_pb. Note that it finished by the caller,
1694
# just as it was created by the caller.
1698
yield ("inventory", None, revision_ids)
1701
# XXX: Note ATM no callers actually pay attention to this return
1702
# instead they just use the list of revision ids and ignore
1703
# missing sigs. Consider removing this work entirely
1704
revisions_with_signatures = set(self.signatures.get_parent_map(
1705
[(r,) for r in revision_ids]))
1706
revisions_with_signatures = set(
1707
[r for (r,) in revisions_with_signatures])
1708
revisions_with_signatures.intersection_update(revision_ids)
1709
yield ("signatures", None, revisions_with_signatures)
1712
yield ("revisions", None, revision_ids)
1715
def get_inventory(self, revision_id):
1716
"""Get Inventory object by revision id."""
1717
return self.iter_inventories([revision_id]).next()
1719
def iter_inventories(self, revision_ids):
1720
"""Get many inventories by revision_ids.
1722
This will buffer some or all of the texts used in constructing the
1723
inventories in memory, but will only parse a single inventory at a
1726
:return: An iterator of inventories.
1728
if ((None in revision_ids)
1729
or (_mod_revision.NULL_REVISION in revision_ids)):
1730
raise ValueError('cannot get null revision inventory')
1731
return self._iter_inventories(revision_ids)
1733
def _iter_inventories(self, revision_ids):
1734
"""single-document based inventory iteration."""
1735
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1736
yield self.deserialise_inventory(revision_id, text)
1738
def _iter_inventory_xmls(self, revision_ids):
1739
keys = [(revision_id,) for revision_id in revision_ids]
1740
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1742
for record in stream:
1743
if record.storage_kind != 'absent':
1744
text_chunks[record.key] = record.get_bytes_as('chunked')
1746
raise errors.NoSuchRevision(self, record.key)
1748
chunks = text_chunks.pop(key)
1749
yield ''.join(chunks), key[-1]
1751
def deserialise_inventory(self, revision_id, xml):
1752
"""Transform the xml into an inventory object.
1754
:param revision_id: The expected revision id of the inventory.
1755
:param xml: A serialised inventory.
1757
result = self._serializer.read_inventory_from_string(xml, revision_id,
1758
entry_cache=self._inventory_entry_cache)
1759
if result.revision_id != revision_id:
1760
raise AssertionError('revision id mismatch %s != %s' % (
1761
result.revision_id, revision_id))
1764
def serialise_inventory(self, inv):
1765
return self._serializer.write_inventory_to_string(inv)
1767
def _serialise_inventory_to_lines(self, inv):
1768
return self._serializer.write_inventory_to_lines(inv)
1770
def get_serializer_format(self):
1771
return self._serializer.format_num
1774
def get_inventory_xml(self, revision_id):
1775
"""Get inventory XML as a file object."""
1776
texts = self._iter_inventory_xmls([revision_id])
1778
text, revision_id = texts.next()
1779
except StopIteration:
1780
raise errors.HistoryMissing(self, 'inventory', revision_id)
1784
def get_inventory_sha1(self, revision_id):
1785
"""Return the sha1 hash of the inventory entry
1787
return self.get_revision(revision_id).inventory_sha1
1789
def iter_reverse_revision_history(self, revision_id):
1790
"""Iterate backwards through revision ids in the lefthand history
1792
:param revision_id: The revision id to start with. All its lefthand
1793
ancestors will be traversed.
1795
graph = self.get_graph()
1796
next_id = revision_id
1798
if next_id in (None, _mod_revision.NULL_REVISION):
1801
# Note: The following line may raise KeyError in the event of
1802
# truncated history. We decided not to have a try:except:raise
1803
# RevisionNotPresent here until we see a use for it, because of the
1804
# cost in an inner loop that is by its very nature O(history).
1805
# Robert Collins 20080326
1806
parents = graph.get_parent_map([next_id])[next_id]
1807
if len(parents) == 0:
1810
next_id = parents[0]
1813
def get_revision_inventory(self, revision_id):
1814
"""Return inventory of a past revision."""
1815
# TODO: Unify this with get_inventory()
1816
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1817
# must be the same as its revision, so this is trivial.
1818
if revision_id is None:
1819
# This does not make sense: if there is no revision,
1820
# then it is the current tree inventory surely ?!
1821
# and thus get_root_id() is something that looks at the last
1822
# commit on the branch, and the get_root_id is an inventory check.
1823
raise NotImplementedError
1824
# return Inventory(self.get_root_id())
1826
return self.get_inventory(revision_id)
1828
def is_shared(self):
1829
"""Return True if this repository is flagged as a shared repository."""
1830
raise NotImplementedError(self.is_shared)
1833
def reconcile(self, other=None, thorough=False):
1834
"""Reconcile this repository."""
1835
from bzrlib.reconcile import RepoReconciler
1836
reconciler = RepoReconciler(self, thorough=thorough)
1837
reconciler.reconcile()
1840
def _refresh_data(self):
1841
"""Helper called from lock_* to ensure coherency with disk.
1843
The default implementation does nothing; it is however possible
1844
for repositories to maintain loaded indices across multiple locks
1845
by checking inside their implementation of this method to see
1846
whether their indices are still valid. This depends of course on
1847
the disk format being validatable in this manner.
1851
def revision_tree(self, revision_id):
1852
"""Return Tree for a revision on this branch.
1854
`revision_id` may be NULL_REVISION for the empty tree revision.
1856
revision_id = _mod_revision.ensure_null(revision_id)
1857
# TODO: refactor this to use an existing revision object
1858
# so we don't need to read it in twice.
1859
if revision_id == _mod_revision.NULL_REVISION:
1860
return RevisionTree(self, Inventory(root_id=None),
1861
_mod_revision.NULL_REVISION)
1863
inv = self.get_revision_inventory(revision_id)
1864
return RevisionTree(self, inv, revision_id)
1866
def revision_trees(self, revision_ids):
1867
"""Return Tree for a revision on this branch.
1869
`revision_id` may not be None or 'null:'"""
1870
inventories = self.iter_inventories(revision_ids)
1871
for inv in inventories:
1872
yield RevisionTree(self, inv, inv.revision_id)
1875
def get_ancestry(self, revision_id, topo_sorted=True):
1876
"""Return a list of revision-ids integrated by a revision.
1878
The first element of the list is always None, indicating the origin
1879
revision. This might change when we have history horizons, or
1880
perhaps we should have a new API.
1882
This is topologically sorted.
1884
if _mod_revision.is_null(revision_id):
1886
if not self.has_revision(revision_id):
1887
raise errors.NoSuchRevision(self, revision_id)
1888
graph = self.get_graph()
1890
search = graph._make_breadth_first_searcher([revision_id])
1893
found, ghosts = search.next_with_ghosts()
1894
except StopIteration:
1897
if _mod_revision.NULL_REVISION in keys:
1898
keys.remove(_mod_revision.NULL_REVISION)
1900
parent_map = graph.get_parent_map(keys)
1901
keys = tsort.topo_sort(parent_map)
1902
return [None] + list(keys)
1905
"""Compress the data within the repository.
1907
This operation only makes sense for some repository types. For other
1908
types it should be a no-op that just returns.
1910
This stub method does not require a lock, but subclasses should use
1911
@needs_write_lock as this is a long running call its reasonable to
1912
implicitly lock for the user.
1916
@deprecated_method(one_six)
1917
def print_file(self, file, revision_id):
1918
"""Print `file` to stdout.
1920
FIXME RBC 20060125 as John Meinel points out this is a bad api
1921
- it writes to stdout, it assumes that that is valid etc. Fix
1922
by creating a new more flexible convenience function.
1924
tree = self.revision_tree(revision_id)
1925
# use inventory as it was in that revision
1926
file_id = tree.inventory.path2id(file)
1928
# TODO: jam 20060427 Write a test for this code path
1929
# it had a bug in it, and was raising the wrong
1931
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1932
tree.print_file(file_id)
1934
def get_transaction(self):
1935
return self.control_files.get_transaction()
1937
@deprecated_method(one_one)
1938
def get_parents(self, revision_ids):
1939
"""See StackedParentsProvider.get_parents"""
1940
parent_map = self.get_parent_map(revision_ids)
1941
return [parent_map.get(r, None) for r in revision_ids]
1943
def get_parent_map(self, revision_ids):
1944
"""See graph._StackedParentsProvider.get_parent_map"""
1945
# revisions index works in keys; this just works in revisions
1946
# therefore wrap and unwrap
1949
for revision_id in revision_ids:
1950
if revision_id == _mod_revision.NULL_REVISION:
1951
result[revision_id] = ()
1952
elif revision_id is None:
1953
raise ValueError('get_parent_map(None) is not valid')
1955
query_keys.append((revision_id ,))
1956
for ((revision_id,), parent_keys) in \
1957
self.revisions.get_parent_map(query_keys).iteritems():
1959
result[revision_id] = tuple(parent_revid
1960
for (parent_revid,) in parent_keys)
1962
result[revision_id] = (_mod_revision.NULL_REVISION,)
1965
def _make_parents_provider(self):
1968
def get_graph(self, other_repository=None):
1969
"""Return the graph walker for this repository format"""
1970
parents_provider = self._make_parents_provider()
1971
if (other_repository is not None and
1972
not self.has_same_location(other_repository)):
1973
parents_provider = graph._StackedParentsProvider(
1974
[parents_provider, other_repository._make_parents_provider()])
1975
return graph.Graph(parents_provider)
1977
def _get_versioned_file_checker(self):
1978
"""Return an object suitable for checking versioned files."""
1979
return _VersionedFileChecker(self)
1981
def revision_ids_to_search_result(self, result_set):
1982
"""Convert a set of revision ids to a graph SearchResult."""
1983
result_parents = set()
1984
for parents in self.get_graph().get_parent_map(
1985
result_set).itervalues():
1986
result_parents.update(parents)
1987
included_keys = result_set.intersection(result_parents)
1988
start_keys = result_set.difference(included_keys)
1989
exclude_keys = result_parents.difference(result_set)
1990
result = graph.SearchResult(start_keys, exclude_keys,
1991
len(result_set), result_set)
1995
def set_make_working_trees(self, new_value):
1996
"""Set the policy flag for making working trees when creating branches.
1998
This only applies to branches that use this repository.
2000
The default is 'True'.
2001
:param new_value: True to restore the default, False to disable making
2004
raise NotImplementedError(self.set_make_working_trees)
2006
def make_working_trees(self):
2007
"""Returns the policy for making working trees on new branches."""
2008
raise NotImplementedError(self.make_working_trees)
2011
def sign_revision(self, revision_id, gpg_strategy):
2012
plaintext = Testament.from_revision(self, revision_id).as_short_text()
2013
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2016
def has_signature_for_revision_id(self, revision_id):
2017
"""Query for a revision signature for revision_id in the repository."""
2018
if not self.has_revision(revision_id):
2019
raise errors.NoSuchRevision(self, revision_id)
2020
sig_present = (1 == len(
2021
self.signatures.get_parent_map([(revision_id,)])))
2025
def get_signature_text(self, revision_id):
2026
"""Return the text for a signature."""
2027
stream = self.signatures.get_record_stream([(revision_id,)],
2029
record = stream.next()
2030
if record.storage_kind == 'absent':
2031
raise errors.NoSuchRevision(self, revision_id)
2032
return record.get_bytes_as('fulltext')
2035
def check(self, revision_ids=None):
2036
"""Check consistency of all history of given revision_ids.
2038
Different repository implementations should override _check().
2040
:param revision_ids: A non-empty list of revision_ids whose ancestry
2041
will be checked. Typically the last revision_id of a branch.
2043
return self._check(revision_ids)
2045
def _check(self, revision_ids):
2046
result = check.Check(self)
2050
def _warn_if_deprecated(self):
2051
global _deprecation_warning_done
2052
if _deprecation_warning_done:
2054
_deprecation_warning_done = True
2055
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2056
% (self._format, self.bzrdir.transport.base))
2058
def supports_rich_root(self):
2059
return self._format.rich_root_data
2061
def _check_ascii_revisionid(self, revision_id, method):
2062
"""Private helper for ascii-only repositories."""
2063
# weave repositories refuse to store revisionids that are non-ascii.
2064
if revision_id is not None:
2065
# weaves require ascii revision ids.
2066
if isinstance(revision_id, unicode):
2068
revision_id.encode('ascii')
2069
except UnicodeEncodeError:
2070
raise errors.NonAsciiRevisionId(method, self)
2073
revision_id.decode('ascii')
2074
except UnicodeDecodeError:
2075
raise errors.NonAsciiRevisionId(method, self)
2077
def revision_graph_can_have_wrong_parents(self):
2078
"""Is it possible for this repository to have a revision graph with
2081
If True, then this repository must also implement
2082
_find_inconsistent_revision_parents so that check and reconcile can
2083
check for inconsistencies before proceeding with other checks that may
2084
depend on the revision index being consistent.
2086
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2089
# remove these delegates a while after bzr 0.15
2090
def __make_delegated(name, from_module):
2091
def _deprecated_repository_forwarder():
2092
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2093
% (name, from_module),
2096
m = __import__(from_module, globals(), locals(), [name])
2098
return getattr(m, name)
2099
except AttributeError:
2100
raise AttributeError('module %s has no name %s'
2102
globals()[name] = _deprecated_repository_forwarder
2105
'AllInOneRepository',
2106
'WeaveMetaDirRepository',
2107
'PreSplitOutRepositoryFormat',
2108
'RepositoryFormat4',
2109
'RepositoryFormat5',
2110
'RepositoryFormat6',
2111
'RepositoryFormat7',
2113
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2117
'RepositoryFormatKnit',
2118
'RepositoryFormatKnit1',
2120
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2123
def install_revision(repository, rev, revision_tree):
2124
"""Install all revision data into a repository."""
2125
install_revisions(repository, [(rev, revision_tree, None)])
2128
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2129
"""Install all revision data into a repository.
2131
Accepts an iterable of revision, tree, signature tuples. The signature
2134
repository.start_write_group()
2136
for n, (revision, revision_tree, signature) in enumerate(iterable):
2137
_install_revision(repository, revision, revision_tree, signature)
2139
pb.update('Transferring revisions', n + 1, num_revisions)
2141
repository.abort_write_group()
2144
repository.commit_write_group()
2147
def _install_revision(repository, rev, revision_tree, signature):
2148
"""Install all revision data into a repository."""
2149
present_parents = []
2151
for p_id in rev.parent_ids:
2152
if repository.has_revision(p_id):
2153
present_parents.append(p_id)
2154
parent_trees[p_id] = repository.revision_tree(p_id)
2156
parent_trees[p_id] = repository.revision_tree(
2157
_mod_revision.NULL_REVISION)
2159
inv = revision_tree.inventory
2160
entries = inv.iter_entries()
2161
# backwards compatibility hack: skip the root id.
2162
if not repository.supports_rich_root():
2163
path, root = entries.next()
2164
if root.revision != rev.revision_id:
2165
raise errors.IncompatibleRevision(repr(repository))
2167
for path, ie in entries:
2168
text_keys[(ie.file_id, ie.revision)] = ie
2169
text_parent_map = repository.texts.get_parent_map(text_keys)
2170
missing_texts = set(text_keys) - set(text_parent_map)
2171
# Add the texts that are not already present
2172
for text_key in missing_texts:
2173
ie = text_keys[text_key]
2175
# FIXME: TODO: The following loop overlaps/duplicates that done by
2176
# commit to determine parents. There is a latent/real bug here where
2177
# the parents inserted are not those commit would do - in particular
2178
# they are not filtered by heads(). RBC, AB
2179
for revision, tree in parent_trees.iteritems():
2180
if ie.file_id not in tree:
2182
parent_id = tree.inventory[ie.file_id].revision
2183
if parent_id in text_parents:
2185
text_parents.append((ie.file_id, parent_id))
2186
lines = revision_tree.get_file(ie.file_id).readlines()
2187
repository.texts.add_lines(text_key, text_parents, lines)
2189
# install the inventory
2190
repository.add_inventory(rev.revision_id, inv, present_parents)
2191
except errors.RevisionAlreadyPresent:
2193
if signature is not None:
2194
repository.add_signature_text(rev.revision_id, signature)
2195
repository.add_revision(rev.revision_id, rev, inv)
2198
class MetaDirRepository(Repository):
2199
"""Repositories in the new meta-dir layout.
2201
:ivar _transport: Transport for access to repository control files,
2202
typically pointing to .bzr/repository.
2205
def __init__(self, _format, a_bzrdir, control_files):
2206
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2207
self._transport = control_files._transport
2209
def is_shared(self):
2210
"""Return True if this repository is flagged as a shared repository."""
2211
return self._transport.has('shared-storage')
2214
def set_make_working_trees(self, new_value):
2215
"""Set the policy flag for making working trees when creating branches.
2217
This only applies to branches that use this repository.
2219
The default is 'True'.
2220
:param new_value: True to restore the default, False to disable making
2225
self._transport.delete('no-working-trees')
2226
except errors.NoSuchFile:
2229
self._transport.put_bytes('no-working-trees', '',
2230
mode=self.bzrdir._get_file_mode())
2232
def make_working_trees(self):
2233
"""Returns the policy for making working trees on new branches."""
2234
return not self._transport.has('no-working-trees')
2237
class MetaDirVersionedFileRepository(MetaDirRepository):
2238
"""Repositories in a meta-dir, that work via versioned file objects."""
2240
def __init__(self, _format, a_bzrdir, control_files):
2241
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2245
network_format_registry = registry.FormatRegistry()
2246
"""Registry of formats indexed by their network name.
2248
The network name for a repository format is an identifier that can be used when
2249
referring to formats with smart server operations. See
2250
RepositoryFormat.network_name() for more detail.
2254
format_registry = registry.FormatRegistry(network_format_registry)
2255
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2257
This can contain either format instances themselves, or classes/factories that
2258
can be called to obtain one.
2262
#####################################################################
2263
# Repository Formats
2265
class RepositoryFormat(object):
2266
"""A repository format.
2268
Formats provide four things:
2269
* An initialization routine to construct repository data on disk.
2270
* a optional format string which is used when the BzrDir supports
2272
* an open routine which returns a Repository instance.
2273
* A network name for referring to the format in smart server RPC
2276
There is one and only one Format subclass for each on-disk format. But
2277
there can be one Repository subclass that is used for several different
2278
formats. The _format attribute on a Repository instance can be used to
2279
determine the disk format.
2281
Formats are placed in a registry by their format string for reference
2282
during opening. These should be subclasses of RepositoryFormat for
2285
Once a format is deprecated, just deprecate the initialize and open
2286
methods on the format class. Do not deprecate the object, as the
2287
object may be created even when a repository instnace hasn't been
2290
Common instance attributes:
2291
_matchingbzrdir - the bzrdir format that the repository format was
2292
originally written to work with. This can be used if manually
2293
constructing a bzrdir and repository, or more commonly for test suite
2297
# Set to True or False in derived classes. True indicates that the format
2298
# supports ghosts gracefully.
2299
supports_ghosts = None
2300
# Can this repository be given external locations to lookup additional
2301
# data. Set to True or False in derived classes.
2302
supports_external_lookups = None
2303
# What order should fetch operations request streams in?
2304
# The default is unordered as that is the cheapest for an origin to
2306
_fetch_order = 'unordered'
2307
# Does this repository format use deltas that can be fetched as-deltas ?
2308
# (E.g. knits, where the knit deltas can be transplanted intact.
2309
# We default to False, which will ensure that enough data to get
2310
# a full text out of any fetch stream will be grabbed.
2311
_fetch_uses_deltas = False
2312
# Should fetch trigger a reconcile after the fetch? Only needed for
2313
# some repository formats that can suffer internal inconsistencies.
2314
_fetch_reconcile = False
2317
return "<%s>" % self.__class__.__name__
2319
def __eq__(self, other):
2320
# format objects are generally stateless
2321
return isinstance(other, self.__class__)
2323
def __ne__(self, other):
2324
return not self == other
2327
def find_format(klass, a_bzrdir):
2328
"""Return the format for the repository object in a_bzrdir.
2330
This is used by bzr native formats that have a "format" file in
2331
the repository. Other methods may be used by different types of
2335
transport = a_bzrdir.get_repository_transport(None)
2336
format_string = transport.get("format").read()
2337
return format_registry.get(format_string)
2338
except errors.NoSuchFile:
2339
raise errors.NoRepositoryPresent(a_bzrdir)
2341
raise errors.UnknownFormatError(format=format_string,
2345
def register_format(klass, format):
2346
format_registry.register(format.get_format_string(), format)
2349
def unregister_format(klass, format):
2350
format_registry.remove(format.get_format_string())
2353
def get_default_format(klass):
2354
"""Return the current default format."""
2355
from bzrlib import bzrdir
2356
return bzrdir.format_registry.make_bzrdir('default').repository_format
2358
def get_format_string(self):
2359
"""Return the ASCII format string that identifies this format.
2361
Note that in pre format ?? repositories the format string is
2362
not permitted nor written to disk.
2364
raise NotImplementedError(self.get_format_string)
2366
def get_format_description(self):
2367
"""Return the short description for this format."""
2368
raise NotImplementedError(self.get_format_description)
2370
# TODO: this shouldn't be in the base class, it's specific to things that
2371
# use weaves or knits -- mbp 20070207
2372
def _get_versioned_file_store(self,
2377
versionedfile_class=None,
2378
versionedfile_kwargs={},
2380
if versionedfile_class is None:
2381
versionedfile_class = self._versionedfile_class
2382
weave_transport = control_files._transport.clone(name)
2383
dir_mode = control_files._dir_mode
2384
file_mode = control_files._file_mode
2385
return VersionedFileStore(weave_transport, prefixed=prefixed,
2387
file_mode=file_mode,
2388
versionedfile_class=versionedfile_class,
2389
versionedfile_kwargs=versionedfile_kwargs,
2392
def initialize(self, a_bzrdir, shared=False):
2393
"""Initialize a repository of this format in a_bzrdir.
2395
:param a_bzrdir: The bzrdir to put the new repository in it.
2396
:param shared: The repository should be initialized as a sharable one.
2397
:returns: The new repository object.
2399
This may raise UninitializableFormat if shared repository are not
2400
compatible the a_bzrdir.
2402
raise NotImplementedError(self.initialize)
2404
def is_supported(self):
2405
"""Is this format supported?
2407
Supported formats must be initializable and openable.
2408
Unsupported formats may not support initialization or committing or
2409
some other features depending on the reason for not being supported.
2413
def network_name(self):
2414
"""A simple byte string uniquely identifying this format for RPC calls.
2416
MetaDir repository formats use their disk format string to identify the
2417
repository over the wire. All in one formats such as bzr < 0.8, and
2418
foreign formats like svn/git and hg should use some marker which is
2419
unique and immutable.
2421
raise NotImplementedError(self.network_name)
2423
def check_conversion_target(self, target_format):
2424
raise NotImplementedError(self.check_conversion_target)
2426
def open(self, a_bzrdir, _found=False):
2427
"""Return an instance of this format for the bzrdir a_bzrdir.
2429
_found is a private parameter, do not use it.
2431
raise NotImplementedError(self.open)
2434
class MetaDirRepositoryFormat(RepositoryFormat):
2435
"""Common base class for the new repositories using the metadir layout."""
2437
rich_root_data = False
2438
supports_tree_reference = False
2439
supports_external_lookups = False
2442
def _matchingbzrdir(self):
2443
matching = bzrdir.BzrDirMetaFormat1()
2444
matching.repository_format = self
2448
super(MetaDirRepositoryFormat, self).__init__()
2450
def _create_control_files(self, a_bzrdir):
2451
"""Create the required files and the initial control_files object."""
2452
# FIXME: RBC 20060125 don't peek under the covers
2453
# NB: no need to escape relative paths that are url safe.
2454
repository_transport = a_bzrdir.get_repository_transport(self)
2455
control_files = lockable_files.LockableFiles(repository_transport,
2456
'lock', lockdir.LockDir)
2457
control_files.create_lock()
2458
return control_files
2460
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2461
"""Upload the initial blank content."""
2462
control_files = self._create_control_files(a_bzrdir)
2463
control_files.lock_write()
2464
transport = control_files._transport
2466
utf8_files += [('shared-storage', '')]
2468
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2469
for (filename, content_stream) in files:
2470
transport.put_file(filename, content_stream,
2471
mode=a_bzrdir._get_file_mode())
2472
for (filename, content_bytes) in utf8_files:
2473
transport.put_bytes_non_atomic(filename, content_bytes,
2474
mode=a_bzrdir._get_file_mode())
2476
control_files.unlock()
2478
def network_name(self):
2479
"""Metadir formats have matching disk and network format strings."""
2480
return self.get_format_string()
2483
# Pre-0.8 formats that don't have a disk format string (because they are
2484
# versioned by the matching control directory). We use the control directories
2485
# disk format string as a key for the network_name because they meet the
2486
# constraints (simple string, unique, immmutable).
2487
network_format_registry.register_lazy(
2488
"Bazaar-NG branch, format 5\n",
2489
'bzrlib.repofmt.weaverepo',
2490
'RepositoryFormat5',
2492
network_format_registry.register_lazy(
2493
"Bazaar-NG branch, format 6\n",
2494
'bzrlib.repofmt.weaverepo',
2495
'RepositoryFormat6',
2498
# formats which have no format string are not discoverable or independently
2499
# creatable on disk, so are not registered in format_registry. They're
2500
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2501
# needed, it's constructed directly by the BzrDir. Non-native formats where
2502
# the repository is not separately opened are similar.
2504
format_registry.register_lazy(
2505
'Bazaar-NG Repository format 7',
2506
'bzrlib.repofmt.weaverepo',
2510
format_registry.register_lazy(
2511
'Bazaar-NG Knit Repository Format 1',
2512
'bzrlib.repofmt.knitrepo',
2513
'RepositoryFormatKnit1',
2516
format_registry.register_lazy(
2517
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2518
'bzrlib.repofmt.knitrepo',
2519
'RepositoryFormatKnit3',
2522
format_registry.register_lazy(
2523
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2524
'bzrlib.repofmt.knitrepo',
2525
'RepositoryFormatKnit4',
2528
# Pack-based formats. There is one format for pre-subtrees, and one for
2529
# post-subtrees to allow ease of testing.
2530
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2531
format_registry.register_lazy(
2532
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2533
'bzrlib.repofmt.pack_repo',
2534
'RepositoryFormatKnitPack1',
2536
format_registry.register_lazy(
2537
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2538
'bzrlib.repofmt.pack_repo',
2539
'RepositoryFormatKnitPack3',
2541
format_registry.register_lazy(
2542
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2543
'bzrlib.repofmt.pack_repo',
2544
'RepositoryFormatKnitPack4',
2546
format_registry.register_lazy(
2547
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2548
'bzrlib.repofmt.pack_repo',
2549
'RepositoryFormatKnitPack5',
2551
format_registry.register_lazy(
2552
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2553
'bzrlib.repofmt.pack_repo',
2554
'RepositoryFormatKnitPack5RichRoot',
2556
format_registry.register_lazy(
2557
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2558
'bzrlib.repofmt.pack_repo',
2559
'RepositoryFormatKnitPack5RichRootBroken',
2561
format_registry.register_lazy(
2562
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2563
'bzrlib.repofmt.pack_repo',
2564
'RepositoryFormatKnitPack6',
2566
format_registry.register_lazy(
2567
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2568
'bzrlib.repofmt.pack_repo',
2569
'RepositoryFormatKnitPack6RichRoot',
2572
# Development formats.
2573
# 1.7->1.8 go below here
2574
format_registry.register_lazy(
2575
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2576
'bzrlib.repofmt.pack_repo',
2577
'RepositoryFormatPackDevelopment2',
2579
format_registry.register_lazy(
2580
("Bazaar development format 2 with subtree support "
2581
"(needs bzr.dev from before 1.8)\n"),
2582
'bzrlib.repofmt.pack_repo',
2583
'RepositoryFormatPackDevelopment2Subtree',
2587
class InterRepository(InterObject):
2588
"""This class represents operations taking place between two repositories.
2590
Its instances have methods like copy_content and fetch, and contain
2591
references to the source and target repositories these operations can be
2594
Often we will provide convenience methods on 'repository' which carry out
2595
operations with another repository - they will always forward to
2596
InterRepository.get(other).method_name(parameters).
2599
_walk_to_common_revisions_batch_size = 1
2601
"""The available optimised InterRepository types."""
2603
def __init__(self, source, target):
2604
InterObject.__init__(self, source, target)
2605
# These two attributes may be overridden by e.g. InterOtherToRemote to
2606
# provide a faster implementation.
2607
self.target_get_graph = self.target.get_graph
2608
self.target_get_parent_map = self.target.get_parent_map
2611
def copy_content(self, revision_id=None):
2612
"""Make a complete copy of the content in self into destination.
2614
This is a destructive operation! Do not use it on existing
2617
:param revision_id: Only copy the content needed to construct
2618
revision_id and its parents.
2621
self.target.set_make_working_trees(self.source.make_working_trees())
2622
except NotImplementedError:
2624
self.target.fetch(self.source, revision_id=revision_id)
2627
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2629
"""Fetch the content required to construct revision_id.
2631
The content is copied from self.source to self.target.
2633
:param revision_id: if None all content is copied, if NULL_REVISION no
2635
:param pb: optional progress bar to use for progress reports. If not
2636
provided a default one will be created.
2639
from bzrlib.fetch import RepoFetcher
2640
f = RepoFetcher(to_repository=self.target,
2641
from_repository=self.source,
2642
last_revision=revision_id,
2643
fetch_spec=fetch_spec,
2644
pb=pb, find_ghosts=find_ghosts)
2646
def _walk_to_common_revisions(self, revision_ids):
2647
"""Walk out from revision_ids in source to revisions target has.
2649
:param revision_ids: The start point for the search.
2650
:return: A set of revision ids.
2652
target_graph = self.target_get_graph()
2653
revision_ids = frozenset(revision_ids)
2654
# Fast path for the case where all the revisions are already in the
2656
# (Although this does incur an extra round trip for the
2657
# fairly common case where the target doesn't already have the revision
2659
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2660
return graph.SearchResult(revision_ids, set(), 0, set())
2661
missing_revs = set()
2662
source_graph = self.source.get_graph()
2663
# ensure we don't pay silly lookup costs.
2664
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2665
null_set = frozenset([_mod_revision.NULL_REVISION])
2666
searcher_exhausted = False
2670
# Iterate the searcher until we have enough next_revs
2671
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2673
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2674
next_revs.update(next_revs_part)
2675
ghosts.update(ghosts_part)
2676
except StopIteration:
2677
searcher_exhausted = True
2679
# If there are ghosts in the source graph, and the caller asked for
2680
# them, make sure that they are present in the target.
2681
# We don't care about other ghosts as we can't fetch them and
2682
# haven't been asked to.
2683
ghosts_to_check = set(revision_ids.intersection(ghosts))
2684
revs_to_get = set(next_revs).union(ghosts_to_check)
2686
have_revs = set(target_graph.get_parent_map(revs_to_get))
2687
# we always have NULL_REVISION present.
2688
have_revs = have_revs.union(null_set)
2689
# Check if the target is missing any ghosts we need.
2690
ghosts_to_check.difference_update(have_revs)
2692
# One of the caller's revision_ids is a ghost in both the
2693
# source and the target.
2694
raise errors.NoSuchRevision(
2695
self.source, ghosts_to_check.pop())
2696
missing_revs.update(next_revs - have_revs)
2697
# Because we may have walked past the original stop point, make
2698
# sure everything is stopped
2699
stop_revs = searcher.find_seen_ancestors(have_revs)
2700
searcher.stop_searching_any(stop_revs)
2701
if searcher_exhausted:
2703
return searcher.get_result()
2705
@deprecated_method(one_two)
2707
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2708
"""Return the revision ids that source has that target does not.
2710
These are returned in topological order.
2712
:param revision_id: only return revision ids included by this
2714
:param find_ghosts: If True find missing revisions in deep history
2715
rather than just finding the surface difference.
2717
return list(self.search_missing_revision_ids(
2718
revision_id, find_ghosts).get_keys())
2721
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2722
"""Return the revision ids that source has that target does not.
2724
:param revision_id: only return revision ids included by this
2726
:param find_ghosts: If True find missing revisions in deep history
2727
rather than just finding the surface difference.
2728
:return: A bzrlib.graph.SearchResult.
2730
# stop searching at found target revisions.
2731
if not find_ghosts and revision_id is not None:
2732
return self._walk_to_common_revisions([revision_id])
2733
# generic, possibly worst case, slow code path.
2734
target_ids = set(self.target.all_revision_ids())
2735
if revision_id is not None:
2736
source_ids = self.source.get_ancestry(revision_id)
2737
if source_ids[0] is not None:
2738
raise AssertionError()
2741
source_ids = self.source.all_revision_ids()
2742
result_set = set(source_ids).difference(target_ids)
2743
return self.source.revision_ids_to_search_result(result_set)
2746
def _same_model(source, target):
2747
"""True if source and target have the same data representation.
2749
Note: this is always called on the base class; overriding it in a
2750
subclass will have no effect.
2753
InterRepository._assert_same_model(source, target)
2755
except errors.IncompatibleRepositories, e:
2759
def _assert_same_model(source, target):
2760
"""Raise an exception if two repositories do not use the same model.
2762
if source.supports_rich_root() != target.supports_rich_root():
2763
raise errors.IncompatibleRepositories(source, target,
2764
"different rich-root support")
2765
if source._serializer != target._serializer:
2766
raise errors.IncompatibleRepositories(source, target,
2767
"different serializers")
2770
class InterSameDataRepository(InterRepository):
2771
"""Code for converting between repositories that represent the same data.
2773
Data format and model must match for this to work.
2777
def _get_repo_format_to_test(self):
2778
"""Repository format for testing with.
2780
InterSameData can pull from subtree to subtree and from non-subtree to
2781
non-subtree, so we test this with the richest repository format.
2783
from bzrlib.repofmt import knitrepo
2784
return knitrepo.RepositoryFormatKnit3()
2787
def is_compatible(source, target):
2788
return InterRepository._same_model(source, target)
2791
class InterWeaveRepo(InterSameDataRepository):
2792
"""Optimised code paths between Weave based repositories.
2794
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2795
implemented lazy inter-object optimisation.
2799
def _get_repo_format_to_test(self):
2800
from bzrlib.repofmt import weaverepo
2801
return weaverepo.RepositoryFormat7()
2804
def is_compatible(source, target):
2805
"""Be compatible with known Weave formats.
2807
We don't test for the stores being of specific types because that
2808
could lead to confusing results, and there is no need to be
2811
from bzrlib.repofmt.weaverepo import (
2817
return (isinstance(source._format, (RepositoryFormat5,
2819
RepositoryFormat7)) and
2820
isinstance(target._format, (RepositoryFormat5,
2822
RepositoryFormat7)))
2823
except AttributeError:
2827
def copy_content(self, revision_id=None):
2828
"""See InterRepository.copy_content()."""
2829
# weave specific optimised path:
2831
self.target.set_make_working_trees(self.source.make_working_trees())
2832
except (errors.RepositoryUpgradeRequired, NotImplemented):
2834
# FIXME do not peek!
2835
if self.source._transport.listable():
2836
pb = ui.ui_factory.nested_progress_bar()
2838
self.target.texts.insert_record_stream(
2839
self.source.texts.get_record_stream(
2840
self.source.texts.keys(), 'topological', False))
2841
pb.update('copying inventory', 0, 1)
2842
self.target.inventories.insert_record_stream(
2843
self.source.inventories.get_record_stream(
2844
self.source.inventories.keys(), 'topological', False))
2845
self.target.signatures.insert_record_stream(
2846
self.source.signatures.get_record_stream(
2847
self.source.signatures.keys(),
2849
self.target.revisions.insert_record_stream(
2850
self.source.revisions.get_record_stream(
2851
self.source.revisions.keys(),
2852
'topological', True))
2856
self.target.fetch(self.source, revision_id=revision_id)
2859
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2860
"""See InterRepository.missing_revision_ids()."""
2861
# we want all revisions to satisfy revision_id in source.
2862
# but we don't want to stat every file here and there.
2863
# we want then, all revisions other needs to satisfy revision_id
2864
# checked, but not those that we have locally.
2865
# so the first thing is to get a subset of the revisions to
2866
# satisfy revision_id in source, and then eliminate those that
2867
# we do already have.
2868
# this is slow on high latency connection to self, but as as this
2869
# disk format scales terribly for push anyway due to rewriting
2870
# inventory.weave, this is considered acceptable.
2872
if revision_id is not None:
2873
source_ids = self.source.get_ancestry(revision_id)
2874
if source_ids[0] is not None:
2875
raise AssertionError()
2878
source_ids = self.source._all_possible_ids()
2879
source_ids_set = set(source_ids)
2880
# source_ids is the worst possible case we may need to pull.
2881
# now we want to filter source_ids against what we actually
2882
# have in target, but don't try to check for existence where we know
2883
# we do not have a revision as that would be pointless.
2884
target_ids = set(self.target._all_possible_ids())
2885
possibly_present_revisions = target_ids.intersection(source_ids_set)
2886
actually_present_revisions = set(
2887
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2888
required_revisions = source_ids_set.difference(actually_present_revisions)
2889
if revision_id is not None:
2890
# we used get_ancestry to determine source_ids then we are assured all
2891
# revisions referenced are present as they are installed in topological order.
2892
# and the tip revision was validated by get_ancestry.
2893
result_set = required_revisions
2895
# if we just grabbed the possibly available ids, then
2896
# we only have an estimate of whats available and need to validate
2897
# that against the revision records.
2899
self.source._eliminate_revisions_not_present(required_revisions))
2900
return self.source.revision_ids_to_search_result(result_set)
2903
class InterKnitRepo(InterSameDataRepository):
2904
"""Optimised code paths between Knit based repositories."""
2907
def _get_repo_format_to_test(self):
2908
from bzrlib.repofmt import knitrepo
2909
return knitrepo.RepositoryFormatKnit1()
2912
def is_compatible(source, target):
2913
"""Be compatible with known Knit formats.
2915
We don't test for the stores being of specific types because that
2916
could lead to confusing results, and there is no need to be
2919
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2921
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2922
isinstance(target._format, RepositoryFormatKnit))
2923
except AttributeError:
2925
return are_knits and InterRepository._same_model(source, target)
2928
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2929
"""See InterRepository.missing_revision_ids()."""
2930
if revision_id is not None:
2931
source_ids = self.source.get_ancestry(revision_id)
2932
if source_ids[0] is not None:
2933
raise AssertionError()
2936
source_ids = self.source.all_revision_ids()
2937
source_ids_set = set(source_ids)
2938
# source_ids is the worst possible case we may need to pull.
2939
# now we want to filter source_ids against what we actually
2940
# have in target, but don't try to check for existence where we know
2941
# we do not have a revision as that would be pointless.
2942
target_ids = set(self.target.all_revision_ids())
2943
possibly_present_revisions = target_ids.intersection(source_ids_set)
2944
actually_present_revisions = set(
2945
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2946
required_revisions = source_ids_set.difference(actually_present_revisions)
2947
if revision_id is not None:
2948
# we used get_ancestry to determine source_ids then we are assured all
2949
# revisions referenced are present as they are installed in topological order.
2950
# and the tip revision was validated by get_ancestry.
2951
result_set = required_revisions
2953
# if we just grabbed the possibly available ids, then
2954
# we only have an estimate of whats available and need to validate
2955
# that against the revision records.
2957
self.source._eliminate_revisions_not_present(required_revisions))
2958
return self.source.revision_ids_to_search_result(result_set)
2961
class InterPackRepo(InterSameDataRepository):
2962
"""Optimised code paths between Pack based repositories."""
2965
def _get_repo_format_to_test(self):
2966
from bzrlib.repofmt import pack_repo
2967
return pack_repo.RepositoryFormatKnitPack1()
2970
def is_compatible(source, target):
2971
"""Be compatible with known Pack formats.
2973
We don't test for the stores being of specific types because that
2974
could lead to confusing results, and there is no need to be
2977
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2979
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2980
isinstance(target._format, RepositoryFormatPack))
2981
except AttributeError:
2983
return are_packs and InterRepository._same_model(source, target)
2986
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2988
"""See InterRepository.fetch()."""
2989
if (len(self.source._fallback_repositories) > 0 or
2990
len(self.target._fallback_repositories) > 0):
2991
# The pack layer is not aware of fallback repositories, so when
2992
# fetching from a stacked repository or into a stacked repository
2993
# we use the generic fetch logic which uses the VersionedFiles
2994
# attributes on repository.
2995
from bzrlib.fetch import RepoFetcher
2996
fetcher = RepoFetcher(self.target, self.source, revision_id,
2997
pb, find_ghosts, fetch_spec=fetch_spec)
2998
if fetch_spec is not None:
2999
if len(list(fetch_spec.heads)) != 1:
3000
raise AssertionError(
3001
"InterPackRepo.fetch doesn't support "
3002
"fetching multiple heads yet.")
3003
revision_id = fetch_spec.heads[0]
3005
if revision_id is None:
3007
# everything to do - use pack logic
3008
# to fetch from all packs to one without
3009
# inventory parsing etc, IFF nothing to be copied is in the target.
3011
source_revision_ids = frozenset(self.source.all_revision_ids())
3012
revision_ids = source_revision_ids - \
3013
frozenset(self.target_get_parent_map(source_revision_ids))
3014
revision_keys = [(revid,) for revid in revision_ids]
3015
target_pack_collection = self._get_target_pack_collection()
3016
index = target_pack_collection.revision_index.combined_index
3017
present_revision_ids = set(item[1][0] for item in
3018
index.iter_entries(revision_keys))
3019
revision_ids = set(revision_ids) - present_revision_ids
3020
# implementing the TODO will involve:
3021
# - detecting when all of a pack is selected
3022
# - avoiding as much as possible pre-selection, so the
3023
# more-core routines such as create_pack_from_packs can filter in
3024
# a just-in-time fashion. (though having a HEADS list on a
3025
# repository might make this a lot easier, because we could
3026
# sensibly detect 'new revisions' without doing a full index scan.
3027
elif _mod_revision.is_null(revision_id):
3032
revision_ids = self.search_missing_revision_ids(revision_id,
3033
find_ghosts=find_ghosts).get_keys()
3034
except errors.NoSuchRevision:
3035
raise errors.InstallFailed([revision_id])
3036
if len(revision_ids) == 0:
3038
return self._pack(self.source, self.target, revision_ids)
3040
def _pack(self, source, target, revision_ids):
3041
from bzrlib.repofmt.pack_repo import Packer
3042
target_pack_collection = self._get_target_pack_collection()
3043
packs = source._pack_collection.all_packs()
3044
pack = Packer(target_pack_collection, packs, '.fetch',
3045
revision_ids).pack()
3046
if pack is not None:
3047
target_pack_collection._save_pack_names()
3048
copied_revs = pack.get_revision_count()
3049
# Trigger an autopack. This may duplicate effort as we've just done
3050
# a pack creation, but for now it is simpler to think about as
3051
# 'upload data, then repack if needed'.
3053
return (copied_revs, [])
3057
def _autopack(self):
3058
self.target._pack_collection.autopack()
3060
def _get_target_pack_collection(self):
3061
return self.target._pack_collection
3064
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3065
"""See InterRepository.missing_revision_ids().
3067
:param find_ghosts: Find ghosts throughout the ancestry of
3070
if not find_ghosts and revision_id is not None:
3071
return self._walk_to_common_revisions([revision_id])
3072
elif revision_id is not None:
3073
# Find ghosts: search for revisions pointing from one repository to
3074
# the other, and vice versa, anywhere in the history of revision_id.
3075
graph = self.target_get_graph(other_repository=self.source)
3076
searcher = graph._make_breadth_first_searcher([revision_id])
3080
next_revs, ghosts = searcher.next_with_ghosts()
3081
except StopIteration:
3083
if revision_id in ghosts:
3084
raise errors.NoSuchRevision(self.source, revision_id)
3085
found_ids.update(next_revs)
3086
found_ids.update(ghosts)
3087
found_ids = frozenset(found_ids)
3088
# Double query here: should be able to avoid this by changing the
3089
# graph api further.
3090
result_set = found_ids - frozenset(
3091
self.target_get_parent_map(found_ids))
3093
source_ids = self.source.all_revision_ids()
3094
# source_ids is the worst possible case we may need to pull.
3095
# now we want to filter source_ids against what we actually
3096
# have in target, but don't try to check for existence where we know
3097
# we do not have a revision as that would be pointless.
3098
target_ids = set(self.target.all_revision_ids())
3099
result_set = set(source_ids).difference(target_ids)
3100
return self.source.revision_ids_to_search_result(result_set)
3103
class InterDifferingSerializer(InterKnitRepo):
3106
def _get_repo_format_to_test(self):
3110
def is_compatible(source, target):
3111
"""Be compatible with Knit2 source and Knit3 target"""
3112
if source.supports_rich_root() != target.supports_rich_root():
3114
# Ideally, we'd support fetching if the source had no tree references
3115
# even if it supported them...
3116
if (getattr(source, '_format.supports_tree_reference', False) and
3117
not getattr(target, '_format.supports_tree_reference', False)):
3121
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3122
"""Get the best delta and base for this revision.
3124
:return: (basis_id, delta)
3126
possible_trees = [(parent_id, cache[parent_id])
3127
for parent_id in parent_ids
3128
if parent_id in cache]
3129
if len(possible_trees) == 0:
3130
# There either aren't any parents, or the parents aren't in the
3131
# cache, so just use the last converted tree
3132
possible_trees.append((basis_id, cache[basis_id]))
3134
for basis_id, basis_tree in possible_trees:
3135
delta = tree.inventory._make_delta(basis_tree.inventory)
3136
deltas.append((len(delta), basis_id, delta))
3138
return deltas[0][1:]
3140
def _fetch_batch(self, revision_ids, basis_id, cache):
3141
"""Fetch across a few revisions.
3143
:param revision_ids: The revisions to copy
3144
:param basis_id: The revision_id of a tree that must be in cache, used
3145
as a basis for delta when no other base is available
3146
:param cache: A cache of RevisionTrees that we can use.
3147
:return: The revision_id of the last converted tree. The RevisionTree
3148
for it will be in cache
3150
# Walk though all revisions; get inventory deltas, copy referenced
3151
# texts that delta references, insert the delta, revision and
3155
pending_revisions = []
3156
parent_map = self.source.get_parent_map(revision_ids)
3157
for tree in self.source.revision_trees(revision_ids):
3158
current_revision_id = tree.get_revision_id()
3159
parent_ids = parent_map.get(current_revision_id, ())
3160
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3162
# Find text entries that need to be copied
3163
for old_path, new_path, file_id, entry in delta:
3164
if new_path is not None:
3165
if not (new_path or self.target.supports_rich_root()):
3166
# We don't copy the text for the root node unless the
3167
# target supports_rich_root.
3169
text_keys.add((file_id, entry.revision))
3170
revision = self.source.get_revision(current_revision_id)
3171
pending_deltas.append((basis_id, delta,
3172
current_revision_id, revision.parent_ids))
3173
pending_revisions.append(revision)
3174
cache[current_revision_id] = tree
3175
basis_id = current_revision_id
3177
from_texts = self.source.texts
3178
to_texts = self.target.texts
3179
to_texts.insert_record_stream(from_texts.get_record_stream(
3180
text_keys, self.target._format._fetch_order,
3181
not self.target._format._fetch_uses_deltas))
3183
for delta in pending_deltas:
3184
self.target.add_inventory_by_delta(*delta)
3185
# insert signatures and revisions
3186
for revision in pending_revisions:
3188
signature = self.source.get_signature_text(
3189
revision.revision_id)
3190
self.target.add_signature_text(revision.revision_id,
3192
except errors.NoSuchRevision:
3194
self.target.add_revision(revision.revision_id, revision)
3197
def _fetch_all_revisions(self, revision_ids, pb):
3198
"""Fetch everything for the list of revisions.
3200
:param revision_ids: The list of revisions to fetch. Must be in
3202
:param pb: A ProgressBar
3205
basis_id, basis_tree = self._get_basis(revision_ids[0])
3207
cache = lru_cache.LRUCache(100)
3208
cache[basis_id] = basis_tree
3209
del basis_tree # We don't want to hang on to it here
3210
for offset in range(0, len(revision_ids), batch_size):
3211
self.target.start_write_group()
3213
pb.update('Transferring revisions', offset,
3215
batch = revision_ids[offset:offset+batch_size]
3216
basis_id = self._fetch_batch(batch, basis_id, cache)
3218
self.target.abort_write_group()
3221
self.target.commit_write_group()
3222
pb.update('Transferring revisions', len(revision_ids),
3226
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3228
"""See InterRepository.fetch()."""
3229
if fetch_spec is not None:
3230
raise AssertionError("Not implemented yet...")
3231
revision_ids = self.target.search_missing_revision_ids(self.source,
3232
revision_id, find_ghosts=find_ghosts).get_keys()
3233
if not revision_ids:
3235
revision_ids = tsort.topo_sort(
3236
self.source.get_graph().get_parent_map(revision_ids))
3238
my_pb = ui.ui_factory.nested_progress_bar()
3241
symbol_versioning.warn(
3242
symbol_versioning.deprecated_in((1, 14, 0))
3243
% "pb parameter to fetch()")
3246
self._fetch_all_revisions(revision_ids, pb)
3248
if my_pb is not None:
3250
return len(revision_ids), 0
3252
def _get_basis(self, first_revision_id):
3253
"""Get a revision and tree which exists in the target.
3255
This assumes that first_revision_id is selected for transmission
3256
because all other ancestors are already present. If we can't find an
3257
ancestor we fall back to NULL_REVISION since we know that is safe.
3259
:return: (basis_id, basis_tree)
3261
first_rev = self.source.get_revision(first_revision_id)
3263
basis_id = first_rev.parent_ids[0]
3264
# only valid as a basis if the target has it
3265
self.target.get_revision(basis_id)
3266
# Try to get a basis tree - if its a ghost it will hit the
3267
# NoSuchRevision case.
3268
basis_tree = self.source.revision_tree(basis_id)
3269
except (IndexError, errors.NoSuchRevision):
3270
basis_id = _mod_revision.NULL_REVISION
3271
basis_tree = self.source.revision_tree(basis_id)
3272
return basis_id, basis_tree
3275
class InterOtherToRemote(InterRepository):
3276
"""An InterRepository that simply delegates to the 'real' InterRepository
3277
calculated for (source, target._real_repository).
3280
_walk_to_common_revisions_batch_size = 50
3282
def __init__(self, source, target):
3283
InterRepository.__init__(self, source, target)
3284
self._real_inter = None
3287
def is_compatible(source, target):
3288
if isinstance(target, remote.RemoteRepository):
3292
def _ensure_real_inter(self):
3293
if self._real_inter is None:
3294
self.target._ensure_real()
3295
real_target = self.target._real_repository
3296
self._real_inter = InterRepository.get(self.source, real_target)
3297
# Make _real_inter use the RemoteRepository for get_parent_map
3298
self._real_inter.target_get_graph = self.target.get_graph
3299
self._real_inter.target_get_parent_map = self.target.get_parent_map
3301
def copy_content(self, revision_id=None):
3302
self._ensure_real_inter()
3303
self._real_inter.copy_content(revision_id=revision_id)
3305
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3307
self._ensure_real_inter()
3308
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3309
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
3312
def _get_repo_format_to_test(self):
3316
class InterRemoteToOther(InterRepository):
3318
def __init__(self, source, target):
3319
InterRepository.__init__(self, source, target)
3320
self._real_inter = None
3323
def is_compatible(source, target):
3324
if not isinstance(source, remote.RemoteRepository):
3326
return InterRepository._same_model(source, target)
3328
def _ensure_real_inter(self):
3329
if self._real_inter is None:
3330
self.source._ensure_real()
3331
real_source = self.source._real_repository
3332
self._real_inter = InterRepository.get(real_source, self.target)
3334
def copy_content(self, revision_id=None):
3335
self._ensure_real_inter()
3336
self._real_inter.copy_content(revision_id=revision_id)
3339
def _get_repo_format_to_test(self):
3344
class InterPackToRemotePack(InterPackRepo):
3345
"""A specialisation of InterPackRepo for a target that is a
3348
This will use the get_parent_map RPC rather than plain readvs, and also
3349
uses an RPC for autopacking.
3352
_walk_to_common_revisions_batch_size = 50
3355
def is_compatible(source, target):
3356
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3357
if isinstance(source._format, RepositoryFormatPack):
3358
if isinstance(target, remote.RemoteRepository):
3359
target._format._ensure_real()
3360
if isinstance(target._format._custom_format,
3361
RepositoryFormatPack):
3362
if InterRepository._same_model(source, target):
3366
def _autopack(self):
3367
self.target.autopack()
3370
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3372
"""See InterRepository.fetch()."""
3373
if self.target._client._medium._is_remote_before((1, 13)):
3374
# The server won't support the insert_stream RPC, so just use
3375
# regular InterPackRepo logic. This avoids a bug that causes many
3376
# round-trips for small append calls.
3377
return InterPackRepo.fetch(self, revision_id=revision_id, pb=pb,
3378
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
3379
# Always fetch using the generic streaming fetch code, to allow
3380
# streaming fetching into remote servers.
3381
from bzrlib.fetch import RepoFetcher
3382
fetcher = RepoFetcher(self.target, self.source, revision_id,
3383
pb, find_ghosts, fetch_spec=fetch_spec)
3385
def _get_target_pack_collection(self):
3386
return self.target._real_repository._pack_collection
3389
def _get_repo_format_to_test(self):
3393
InterRepository.register_optimiser(InterDifferingSerializer)
3394
InterRepository.register_optimiser(InterSameDataRepository)
3395
InterRepository.register_optimiser(InterWeaveRepo)
3396
InterRepository.register_optimiser(InterKnitRepo)
3397
InterRepository.register_optimiser(InterPackRepo)
3398
InterRepository.register_optimiser(InterOtherToRemote)
3399
InterRepository.register_optimiser(InterRemoteToOther)
3400
InterRepository.register_optimiser(InterPackToRemotePack)
3403
class CopyConverter(object):
3404
"""A repository conversion tool which just performs a copy of the content.
3406
This is slow but quite reliable.
3409
def __init__(self, target_format):
3410
"""Create a CopyConverter.
3412
:param target_format: The format the resulting repository should be.
3414
self.target_format = target_format
3416
def convert(self, repo, pb):
3417
"""Perform the conversion of to_convert, giving feedback via pb.
3419
:param to_convert: The disk object to convert.
3420
:param pb: a progress bar to use for progress information.
3425
# this is only useful with metadir layouts - separated repo content.
3426
# trigger an assertion if not such
3427
repo._format.get_format_string()
3428
self.repo_dir = repo.bzrdir
3429
self.step('Moving repository to repository.backup')
3430
self.repo_dir.transport.move('repository', 'repository.backup')
3431
backup_transport = self.repo_dir.transport.clone('repository.backup')
3432
repo._format.check_conversion_target(self.target_format)
3433
self.source_repo = repo._format.open(self.repo_dir,
3435
_override_transport=backup_transport)
3436
self.step('Creating new repository')
3437
converted = self.target_format.initialize(self.repo_dir,
3438
self.source_repo.is_shared())
3439
converted.lock_write()
3441
self.step('Copying content into repository.')
3442
self.source_repo.copy_content_into(converted)
3445
self.step('Deleting old repository content.')
3446
self.repo_dir.transport.delete_tree('repository.backup')
3447
self.pb.note('repository converted')
3449
def step(self, message):
3450
"""Update the pb by a step."""
3452
self.pb.update(message, self.count, self.total)
3464
def _unescaper(match, _map=_unescape_map):
3465
code = match.group(1)
3469
if not code.startswith('#'):
3471
return unichr(int(code[1:])).encode('utf8')
3477
def _unescape_xml(data):
3478
"""Unescape predefined XML entities in a string of data."""
3480
if _unescape_re is None:
3481
_unescape_re = re.compile('\&([^;]*);')
3482
return _unescape_re.sub(_unescaper, data)
3485
class _VersionedFileChecker(object):
3487
def __init__(self, repository):
3488
self.repository = repository
3489
self.text_index = self.repository._generate_text_key_index()
3491
def calculate_file_version_parents(self, text_key):
3492
"""Calculate the correct parents for a file version according to
3495
parent_keys = self.text_index[text_key]
3496
if parent_keys == [_mod_revision.NULL_REVISION]:
3498
return tuple(parent_keys)
3500
def check_file_version_parents(self, texts, progress_bar=None):
3501
"""Check the parents stored in a versioned file are correct.
3503
It also detects file versions that are not referenced by their
3504
corresponding revision's inventory.
3506
:returns: A tuple of (wrong_parents, dangling_file_versions).
3507
wrong_parents is a dict mapping {revision_id: (stored_parents,
3508
correct_parents)} for each revision_id where the stored parents
3509
are not correct. dangling_file_versions is a set of (file_id,
3510
revision_id) tuples for versions that are present in this versioned
3511
file, but not used by the corresponding inventory.
3514
self.file_ids = set([file_id for file_id, _ in
3515
self.text_index.iterkeys()])
3516
# text keys is now grouped by file_id
3517
n_weaves = len(self.file_ids)
3518
files_in_revisions = {}
3519
revisions_of_files = {}
3520
n_versions = len(self.text_index)
3521
progress_bar.update('loading text store', 0, n_versions)
3522
parent_map = self.repository.texts.get_parent_map(self.text_index)
3523
# On unlistable transports this could well be empty/error...
3524
text_keys = self.repository.texts.keys()
3525
unused_keys = frozenset(text_keys) - set(self.text_index)
3526
for num, key in enumerate(self.text_index.iterkeys()):
3527
if progress_bar is not None:
3528
progress_bar.update('checking text graph', num, n_versions)
3529
correct_parents = self.calculate_file_version_parents(key)
3531
knit_parents = parent_map[key]
3532
except errors.RevisionNotPresent:
3535
if correct_parents != knit_parents:
3536
wrong_parents[key] = (knit_parents, correct_parents)
3537
return wrong_parents, unused_keys
3540
def _old_get_graph(repository, revision_id):
3541
"""DO NOT USE. That is all. I'm serious."""
3542
graph = repository.get_graph()
3543
revision_graph = dict(((key, value) for key, value in
3544
graph.iter_ancestry([revision_id]) if value is not None))
3545
return _strip_NULL_ghosts(revision_graph)
3548
def _strip_NULL_ghosts(revision_graph):
3549
"""Also don't use this. more compatibility code for unmigrated clients."""
3550
# Filter ghosts, and null:
3551
if _mod_revision.NULL_REVISION in revision_graph:
3552
del revision_graph[_mod_revision.NULL_REVISION]
3553
for key, parents in revision_graph.items():
3554
revision_graph[key] = tuple(parent for parent in parents if parent
3556
return revision_graph
3559
class StreamSink(object):
3560
"""An object that can insert a stream into a repository.
3562
This interface handles the complexity of reserialising inventories and
3563
revisions from different formats, and allows unidirectional insertion into
3564
stacked repositories without looking for the missing basis parents
3568
def __init__(self, target_repo):
3569
self.target_repo = target_repo
3571
def insert_stream(self, stream, src_format, resume_tokens):
3572
"""Insert a stream's content into the target repository.
3574
:param src_format: a bzr repository format.
3576
:return: a list of resume tokens and an iterable of keys additional
3577
items required before the insertion can be completed.
3579
self.target_repo.lock_write()
3582
self.target_repo.resume_write_group(resume_tokens)
3584
self.target_repo.start_write_group()
3586
# locked_insert_stream performs a commit|suspend.
3587
return self._locked_insert_stream(stream, src_format)
3589
self.target_repo.abort_write_group(suppress_errors=True)
3592
self.target_repo.unlock()
3594
def _locked_insert_stream(self, stream, src_format):
3595
to_serializer = self.target_repo._format._serializer
3596
src_serializer = src_format._serializer
3597
for substream_type, substream in stream:
3598
if substream_type == 'texts':
3599
self.target_repo.texts.insert_record_stream(substream)
3600
elif substream_type == 'inventories':
3601
if src_serializer == to_serializer:
3602
self.target_repo.inventories.insert_record_stream(
3605
self._extract_and_insert_inventories(
3606
substream, src_serializer)
3607
elif substream_type == 'revisions':
3608
# This may fallback to extract-and-insert more often than
3609
# required if the serializers are different only in terms of
3611
if src_serializer == to_serializer:
3612
self.target_repo.revisions.insert_record_stream(
3615
self._extract_and_insert_revisions(substream,
3617
elif substream_type == 'signatures':
3618
self.target_repo.signatures.insert_record_stream(substream)
3620
raise AssertionError('kaboom! %s' % (substream_type,))
3622
missing_keys = set()
3623
for prefix, versioned_file in (
3624
('texts', self.target_repo.texts),
3625
('inventories', self.target_repo.inventories),
3626
('revisions', self.target_repo.revisions),
3627
('signatures', self.target_repo.signatures),
3629
missing_keys.update((prefix,) + key for key in
3630
versioned_file.get_missing_compression_parent_keys())
3631
except NotImplementedError:
3632
# cannot even attempt suspending, and missing would have failed
3633
# during stream insertion.
3634
missing_keys = set()
3637
# suspend the write group and tell the caller what we is
3638
# missing. We know we can suspend or else we would not have
3639
# entered this code path. (All repositories that can handle
3640
# missing keys can handle suspending a write group).
3641
write_group_tokens = self.target_repo.suspend_write_group()
3642
return write_group_tokens, missing_keys
3643
self.target_repo.commit_write_group()
3646
def _extract_and_insert_inventories(self, substream, serializer):
3647
"""Generate a new inventory versionedfile in target, converting data.
3649
The inventory is retrieved from the source, (deserializing it), and
3650
stored in the target (reserializing it in a different format).
3652
for record in substream:
3653
bytes = record.get_bytes_as('fulltext')
3654
revision_id = record.key[0]
3655
inv = serializer.read_inventory_from_string(bytes, revision_id)
3656
parents = [key[0] for key in record.parents]
3657
self.target_repo.add_inventory(revision_id, inv, parents)
3659
def _extract_and_insert_revisions(self, substream, serializer):
3660
for record in substream:
3661
bytes = record.get_bytes_as('fulltext')
3662
revision_id = record.key[0]
3663
rev = serializer.read_revision_from_string(bytes)
3664
if rev.revision_id != revision_id:
3665
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
3666
self.target_repo.add_revision(revision_id, rev)
3669
if self.target_repo._format._fetch_reconcile:
3670
self.target_repo.reconcile()
3673
class StreamSource(object):
3674
"""A source of a stream for fetching between repositories."""
3676
def __init__(self, from_repository, to_format):
3677
"""Create a StreamSource streaming from from_repository."""
3678
self.from_repository = from_repository
3679
self.to_format = to_format
3681
def delta_on_metadata(self):
3682
"""Return True if delta's are permitted on metadata streams.
3684
That is on revisions and signatures.
3686
src_serializer = self.from_repository._format._serializer
3687
target_serializer = self.to_format._serializer
3688
return (self.to_format._fetch_uses_deltas and
3689
src_serializer == target_serializer)
3691
def _fetch_revision_texts(self, revs):
3692
# fetch signatures first and then the revision texts
3693
# may need to be a InterRevisionStore call here.
3694
from_sf = self.from_repository.signatures
3695
# A missing signature is just skipped.
3696
keys = [(rev_id,) for rev_id in revs]
3697
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
3699
self.to_format._fetch_order,
3700
not self.to_format._fetch_uses_deltas))
3701
# If a revision has a delta, this is actually expanded inside the
3702
# insert_record_stream code now, which is an alternate fix for
3704
from_rf = self.from_repository.revisions
3705
revisions = from_rf.get_record_stream(
3707
self.to_format._fetch_order,
3708
not self.delta_on_metadata())
3709
return [('signatures', signatures), ('revisions', revisions)]
3711
def _generate_root_texts(self, revs):
3712
"""This will be called by __fetch between fetching weave texts and
3713
fetching the inventory weave.
3715
Subclasses should override this if they need to generate root texts
3716
after fetching weave texts.
3718
if self._rich_root_upgrade():
3720
return bzrlib.fetch.Inter1and2Helper(
3721
self.from_repository).generate_root_texts(revs)
3725
def get_stream(self, search):
3727
revs = search.get_keys()
3728
graph = self.from_repository.get_graph()
3729
revs = list(graph.iter_topo_order(revs))
3730
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
3732
for knit_kind, file_id, revisions in data_to_fetch:
3733
if knit_kind != phase:
3735
# Make a new progress bar for this phase
3736
if knit_kind == "file":
3737
# Accumulate file texts
3738
text_keys.extend([(file_id, revision) for revision in
3740
elif knit_kind == "inventory":
3741
# Now copy the file texts.
3742
from_texts = self.from_repository.texts
3743
yield ('texts', from_texts.get_record_stream(
3744
text_keys, self.to_format._fetch_order,
3745
not self.to_format._fetch_uses_deltas))
3746
# Cause an error if a text occurs after we have done the
3749
# Before we process the inventory we generate the root
3750
# texts (if necessary) so that the inventories references
3752
for _ in self._generate_root_texts(revs):
3754
# NB: This currently reopens the inventory weave in source;
3755
# using a single stream interface instead would avoid this.
3756
from_weave = self.from_repository.inventories
3757
# we fetch only the referenced inventories because we do not
3758
# know for unselected inventories whether all their required
3759
# texts are present in the other repository - it could be
3761
yield ('inventories', from_weave.get_record_stream(
3762
[(rev_id,) for rev_id in revs],
3763
self.inventory_fetch_order(),
3764
not self.delta_on_metadata()))
3765
elif knit_kind == "signatures":
3766
# Nothing to do here; this will be taken care of when
3767
# _fetch_revision_texts happens.
3769
elif knit_kind == "revisions":
3770
for record in self._fetch_revision_texts(revs):
3773
raise AssertionError("Unknown knit kind %r" % knit_kind)
3775
def get_stream_for_missing_keys(self, missing_keys):
3776
# missing keys can only occur when we are byte copying and not
3777
# translating (because translation means we don't send
3778
# unreconstructable deltas ever).
3780
keys['texts'] = set()
3781
keys['revisions'] = set()
3782
keys['inventories'] = set()
3783
keys['signatures'] = set()
3784
for key in missing_keys:
3785
keys[key[0]].add(key[1:])
3786
if len(keys['revisions']):
3787
# If we allowed copying revisions at this point, we could end up
3788
# copying a revision without copying its required texts: a
3789
# violation of the requirements for repository integrity.
3790
raise AssertionError(
3791
'cannot copy revisions to fill in missing deltas %s' % (
3792
keys['revisions'],))
3793
for substream_kind, keys in keys.iteritems():
3794
vf = getattr(self.from_repository, substream_kind)
3795
# Ask for full texts always so that we don't need more round trips
3796
# after this stream.
3797
stream = vf.get_record_stream(keys,
3798
self.to_format._fetch_order, True)
3799
yield substream_kind, stream
3801
def inventory_fetch_order(self):
3802
if self._rich_root_upgrade():
3803
return 'topological'
3805
return self.to_format._fetch_order
3807
def _rich_root_upgrade(self):
3808
return (not self.from_repository._format.rich_root_data and
3809
self.to_format.rich_root_data)
added
removed
bzrlib/repository.py
