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 (
57
from bzrlib.trace import (
58
log_exception_quietly, note, mutter, mutter_callsite, warning)
61
# Old formats display a warning, but only once
62
_deprecation_warning_done = False
65
class CommitBuilder(object):
66
"""Provides an interface to build up a commit.
68
This allows describing a tree to be committed without needing to
69
know the internals of the format of the repository.
72
# all clients should supply tree roots.
73
record_root_entry = True
74
# the default CommitBuilder does not manage trees whose root is versioned.
75
_versioned_root = False
77
def __init__(self, repository, parents, config, timestamp=None,
78
timezone=None, committer=None, revprops=None,
80
"""Initiate a CommitBuilder.
82
:param repository: Repository to commit to.
83
:param parents: Revision ids of the parents of the new revision.
84
:param config: Configuration to use.
85
:param timestamp: Optional timestamp recorded for commit.
86
:param timezone: Optional timezone for timestamp.
87
:param committer: Optional committer to set for commit.
88
:param revprops: Optional dictionary of revision properties.
89
:param revision_id: Optional revision id.
94
self._committer = self._config.username()
96
self._committer = committer
98
self.new_inventory = Inventory(None)
99
self._new_revision_id = revision_id
100
self.parents = parents
101
self.repository = repository
104
if revprops is not None:
105
self._validate_revprops(revprops)
106
self._revprops.update(revprops)
108
if timestamp is None:
109
timestamp = time.time()
110
# Restrict resolution to 1ms
111
self._timestamp = round(timestamp, 3)
114
self._timezone = osutils.local_time_offset()
116
self._timezone = int(timezone)
118
self._generate_revision_if_needed()
119
self.__heads = graph.HeadsCache(repository.get_graph()).heads
120
self._basis_delta = []
121
# API compatibility, older code that used CommitBuilder did not call
122
# .record_delete(), which means the delta that is computed would not be
123
# valid. Callers that will call record_delete() should call
124
# .will_record_deletes() to indicate that.
125
self._recording_deletes = False
127
def _validate_unicode_text(self, text, context):
128
"""Verify things like commit messages don't have bogus characters."""
130
raise ValueError('Invalid value for %s: %r' % (context, text))
132
def _validate_revprops(self, revprops):
133
for key, value in revprops.iteritems():
134
# We know that the XML serializers do not round trip '\r'
135
# correctly, so refuse to accept them
136
if not isinstance(value, basestring):
137
raise ValueError('revision property (%s) is not a valid'
138
' (unicode) string: %r' % (key, value))
139
self._validate_unicode_text(value,
140
'revision property (%s)' % (key,))
142
def commit(self, message):
143
"""Make the actual commit.
145
:return: The revision id of the recorded revision.
147
self._validate_unicode_text(message, 'commit message')
148
rev = _mod_revision.Revision(
149
timestamp=self._timestamp,
150
timezone=self._timezone,
151
committer=self._committer,
153
inventory_sha1=self.inv_sha1,
154
revision_id=self._new_revision_id,
155
properties=self._revprops)
156
rev.parent_ids = self.parents
157
self.repository.add_revision(self._new_revision_id, rev,
158
self.new_inventory, self._config)
159
self.repository.commit_write_group()
160
return self._new_revision_id
163
"""Abort the commit that is being built.
165
self.repository.abort_write_group()
167
def revision_tree(self):
168
"""Return the tree that was just committed.
170
After calling commit() this can be called to get a RevisionTree
171
representing the newly committed tree. This is preferred to
172
calling Repository.revision_tree() because that may require
173
deserializing the inventory, while we already have a copy in
176
return RevisionTree(self.repository, self.new_inventory,
177
self._new_revision_id)
179
def finish_inventory(self):
180
"""Tell the builder that the inventory is finished."""
181
if self.new_inventory.root is None:
182
raise AssertionError('Root entry should be supplied to'
183
' record_entry_contents, as of bzr 0.10.')
184
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
185
self.new_inventory.revision_id = self._new_revision_id
186
self.inv_sha1 = self.repository.add_inventory(
187
self._new_revision_id,
192
def _gen_revision_id(self):
193
"""Return new revision-id."""
194
return generate_ids.gen_revision_id(self._config.username(),
197
def _generate_revision_if_needed(self):
198
"""Create a revision id if None was supplied.
200
If the repository can not support user-specified revision ids
201
they should override this function and raise CannotSetRevisionId
202
if _new_revision_id is not None.
204
:raises: CannotSetRevisionId
206
if self._new_revision_id is None:
207
self._new_revision_id = self._gen_revision_id()
208
self.random_revid = True
210
self.random_revid = False
212
def _heads(self, file_id, revision_ids):
213
"""Calculate the graph heads for revision_ids in the graph of file_id.
215
This can use either a per-file graph or a global revision graph as we
216
have an identity relationship between the two graphs.
218
return self.__heads(revision_ids)
220
def _check_root(self, ie, parent_invs, tree):
221
"""Helper for record_entry_contents.
223
:param ie: An entry being added.
224
:param parent_invs: The inventories of the parent revisions of the
226
:param tree: The tree that is being committed.
228
# In this revision format, root entries have no knit or weave When
229
# serializing out to disk and back in root.revision is always
231
ie.revision = self._new_revision_id
233
def _get_delta(self, ie, basis_inv, path):
234
"""Get a delta against the basis inventory for ie."""
235
if ie.file_id not in basis_inv:
237
result = (None, path, ie.file_id, ie)
238
self._basis_delta.append(result)
240
elif ie != basis_inv[ie.file_id]:
242
# TODO: avoid tis id2path call.
243
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
244
self._basis_delta.append(result)
250
def get_basis_delta(self):
251
"""Return the complete inventory delta versus the basis inventory.
253
This has been built up with the calls to record_delete and
254
record_entry_contents. The client must have already called
255
will_record_deletes() to indicate that they will be generating a
258
:return: An inventory delta, suitable for use with apply_delta, or
259
Repository.add_inventory_by_delta, etc.
261
if not self._recording_deletes:
262
raise AssertionError("recording deletes not activated.")
263
return self._basis_delta
265
def record_delete(self, path, file_id):
266
"""Record that a delete occured against a basis tree.
268
This is an optional API - when used it adds items to the basis_delta
269
being accumulated by the commit builder. It cannot be called unless the
270
method will_record_deletes() has been called to inform the builder that
271
a delta is being supplied.
273
:param path: The path of the thing deleted.
274
:param file_id: The file id that was deleted.
276
if not self._recording_deletes:
277
raise AssertionError("recording deletes not activated.")
278
delta = (path, None, file_id, None)
279
self._basis_delta.append(delta)
282
def will_record_deletes(self):
283
"""Tell the commit builder that deletes are being notified.
285
This enables the accumulation of an inventory delta; for the resulting
286
commit to be valid, deletes against the basis MUST be recorded via
287
builder.record_delete().
289
self._recording_deletes = True
291
def record_entry_contents(self, ie, parent_invs, path, tree,
293
"""Record the content of ie from tree into the commit if needed.
295
Side effect: sets ie.revision when unchanged
297
:param ie: An inventory entry present in the commit.
298
:param parent_invs: The inventories of the parent revisions of the
300
:param path: The path the entry is at in the tree.
301
:param tree: The tree which contains this entry and should be used to
303
:param content_summary: Summary data from the tree about the paths
304
content - stat, length, exec, sha/link target. This is only
305
accessed when the entry has a revision of None - that is when it is
306
a candidate to commit.
307
:return: A tuple (change_delta, version_recorded, fs_hash).
308
change_delta is an inventory_delta change for this entry against
309
the basis tree of the commit, or None if no change occured against
311
version_recorded is True if a new version of the entry has been
312
recorded. For instance, committing a merge where a file was only
313
changed on the other side will return (delta, False).
314
fs_hash is either None, or the hash details for the path (currently
315
a tuple of the contents sha1 and the statvalue returned by
316
tree.get_file_with_stat()).
318
if self.new_inventory.root is None:
319
if ie.parent_id is not None:
320
raise errors.RootMissing()
321
self._check_root(ie, parent_invs, tree)
322
if ie.revision is None:
323
kind = content_summary[0]
325
# ie is carried over from a prior commit
327
# XXX: repository specific check for nested tree support goes here - if
328
# the repo doesn't want nested trees we skip it ?
329
if (kind == 'tree-reference' and
330
not self.repository._format.supports_tree_reference):
331
# mismatch between commit builder logic and repository:
332
# this needs the entry creation pushed down into the builder.
333
raise NotImplementedError('Missing repository subtree support.')
334
self.new_inventory.add(ie)
336
# TODO: slow, take it out of the inner loop.
338
basis_inv = parent_invs[0]
340
basis_inv = Inventory(root_id=None)
342
# ie.revision is always None if the InventoryEntry is considered
343
# for committing. We may record the previous parents revision if the
344
# content is actually unchanged against a sole head.
345
if ie.revision is not None:
346
if not self._versioned_root and path == '':
347
# repositories that do not version the root set the root's
348
# revision to the new commit even when no change occurs (more
349
# specifically, they do not record a revision on the root; and
350
# the rev id is assigned to the root during deserialisation -
351
# this masks when a change may have occurred against the basis.
352
# To match this we always issue a delta, because the revision
353
# of the root will always be changing.
354
if ie.file_id in basis_inv:
355
delta = (basis_inv.id2path(ie.file_id), path,
359
delta = (None, path, ie.file_id, ie)
360
self._basis_delta.append(delta)
361
return delta, False, None
363
# we don't need to commit this, because the caller already
364
# determined that an existing revision of this file is
365
# appropriate. If its not being considered for committing then
366
# it and all its parents to the root must be unaltered so
367
# no-change against the basis.
368
if ie.revision == self._new_revision_id:
369
raise AssertionError("Impossible situation, a skipped "
370
"inventory entry (%r) claims to be modified in this "
371
"commit (%r).", (ie, self._new_revision_id))
372
return None, False, None
373
# XXX: Friction: parent_candidates should return a list not a dict
374
# so that we don't have to walk the inventories again.
375
parent_candiate_entries = ie.parent_candidates(parent_invs)
376
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
378
for inv in parent_invs:
379
if ie.file_id in inv:
380
old_rev = inv[ie.file_id].revision
381
if old_rev in head_set:
382
heads.append(inv[ie.file_id].revision)
383
head_set.remove(inv[ie.file_id].revision)
386
# now we check to see if we need to write a new record to the
388
# We write a new entry unless there is one head to the ancestors, and
389
# the kind-derived content is unchanged.
391
# Cheapest check first: no ancestors, or more the one head in the
392
# ancestors, we write a new node.
396
# There is a single head, look it up for comparison
397
parent_entry = parent_candiate_entries[heads[0]]
398
# if the non-content specific data has changed, we'll be writing a
400
if (parent_entry.parent_id != ie.parent_id or
401
parent_entry.name != ie.name):
403
# now we need to do content specific checks:
405
# if the kind changed the content obviously has
406
if kind != parent_entry.kind:
408
# Stat cache fingerprint feedback for the caller - None as we usually
409
# don't generate one.
412
if content_summary[2] is None:
413
raise ValueError("Files must not have executable = None")
415
if (# if the file length changed we have to store:
416
parent_entry.text_size != content_summary[1] or
417
# if the exec bit has changed we have to store:
418
parent_entry.executable != content_summary[2]):
420
elif parent_entry.text_sha1 == content_summary[3]:
421
# all meta and content is unchanged (using a hash cache
422
# hit to check the sha)
423
ie.revision = parent_entry.revision
424
ie.text_size = parent_entry.text_size
425
ie.text_sha1 = parent_entry.text_sha1
426
ie.executable = parent_entry.executable
427
return self._get_delta(ie, basis_inv, path), False, None
429
# Either there is only a hash change(no hash cache entry,
430
# or same size content change), or there is no change on
432
# Provide the parent's hash to the store layer, so that the
433
# content is unchanged we will not store a new node.
434
nostore_sha = parent_entry.text_sha1
436
# We want to record a new node regardless of the presence or
437
# absence of a content change in the file.
439
ie.executable = content_summary[2]
440
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
442
lines = file_obj.readlines()
446
ie.text_sha1, ie.text_size = self._add_text_to_weave(
447
ie.file_id, lines, heads, nostore_sha)
448
# Let the caller know we generated a stat fingerprint.
449
fingerprint = (ie.text_sha1, stat_value)
450
except errors.ExistingContent:
451
# Turns out that the file content was unchanged, and we were
452
# only going to store a new node if it was changed. Carry over
454
ie.revision = parent_entry.revision
455
ie.text_size = parent_entry.text_size
456
ie.text_sha1 = parent_entry.text_sha1
457
ie.executable = parent_entry.executable
458
return self._get_delta(ie, basis_inv, path), False, None
459
elif kind == 'directory':
461
# all data is meta here, nothing specific to directory, so
463
ie.revision = parent_entry.revision
464
return self._get_delta(ie, basis_inv, path), False, None
466
self._add_text_to_weave(ie.file_id, lines, heads, None)
467
elif kind == 'symlink':
468
current_link_target = content_summary[3]
470
# symlink target is not generic metadata, check if it has
472
if current_link_target != parent_entry.symlink_target:
475
# unchanged, carry over.
476
ie.revision = parent_entry.revision
477
ie.symlink_target = parent_entry.symlink_target
478
return self._get_delta(ie, basis_inv, path), False, None
479
ie.symlink_target = current_link_target
481
self._add_text_to_weave(ie.file_id, lines, heads, None)
482
elif kind == 'tree-reference':
484
if content_summary[3] != parent_entry.reference_revision:
487
# unchanged, carry over.
488
ie.reference_revision = parent_entry.reference_revision
489
ie.revision = parent_entry.revision
490
return self._get_delta(ie, basis_inv, path), False, None
491
ie.reference_revision = content_summary[3]
493
self._add_text_to_weave(ie.file_id, lines, heads, None)
495
raise NotImplementedError('unknown kind')
496
ie.revision = self._new_revision_id
497
return self._get_delta(ie, basis_inv, path), True, fingerprint
499
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
500
# Note: as we read the content directly from the tree, we know its not
501
# been turned into unicode or badly split - but a broken tree
502
# implementation could give us bad output from readlines() so this is
503
# not a guarantee of safety. What would be better is always checking
504
# the content during test suite execution. RBC 20070912
505
parent_keys = tuple((file_id, parent) for parent in parents)
506
return self.repository.texts.add_lines(
507
(file_id, self._new_revision_id), parent_keys, new_lines,
508
nostore_sha=nostore_sha, random_id=self.random_revid,
509
check_content=False)[0:2]
512
class RootCommitBuilder(CommitBuilder):
513
"""This commitbuilder actually records the root id"""
515
# the root entry gets versioned properly by this builder.
516
_versioned_root = True
518
def _check_root(self, ie, parent_invs, tree):
519
"""Helper for record_entry_contents.
521
:param ie: An entry being added.
522
:param parent_invs: The inventories of the parent revisions of the
524
:param tree: The tree that is being committed.
528
######################################################################
531
class Repository(object):
532
"""Repository holding history for one or more branches.
534
The repository holds and retrieves historical information including
535
revisions and file history. It's normally accessed only by the Branch,
536
which views a particular line of development through that history.
538
The Repository builds on top of some byte storage facilies (the revisions,
539
signatures, inventories and texts attributes) and a Transport, which
540
respectively provide byte storage and a means to access the (possibly
543
The byte storage facilities are addressed via tuples, which we refer to
544
as 'keys' throughout the code base. Revision_keys, inventory_keys and
545
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
546
(file_id, revision_id). We use this interface because it allows low
547
friction with the underlying code that implements disk indices, network
548
encoding and other parts of bzrlib.
550
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
551
the serialised revisions for the repository. This can be used to obtain
552
revision graph information or to access raw serialised revisions.
553
The result of trying to insert data into the repository via this store
554
is undefined: it should be considered read-only except for implementors
556
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
557
the serialised signatures for the repository. This can be used to
558
obtain access to raw serialised signatures. The result of trying to
559
insert data into the repository via this store is undefined: it should
560
be considered read-only except for implementors of repositories.
561
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
562
the serialised inventories for the repository. This can be used to
563
obtain unserialised inventories. The result of trying to insert data
564
into the repository via this store is undefined: it should be
565
considered read-only except for implementors of repositories.
566
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
567
texts of files and directories for the repository. This can be used to
568
obtain file texts or file graphs. Note that Repository.iter_file_bytes
569
is usually a better interface for accessing file texts.
570
The result of trying to insert data into the repository via this store
571
is undefined: it should be considered read-only except for implementors
573
:ivar _transport: Transport for file access to repository, typically
574
pointing to .bzr/repository.
577
# What class to use for a CommitBuilder. Often its simpler to change this
578
# in a Repository class subclass rather than to override
579
# get_commit_builder.
580
_commit_builder_class = CommitBuilder
581
# The search regex used by xml based repositories to determine what things
582
# where changed in a single commit.
583
_file_ids_altered_regex = lazy_regex.lazy_compile(
584
r'file_id="(?P<file_id>[^"]+)"'
585
r'.* revision="(?P<revision_id>[^"]+)"'
588
def abort_write_group(self, suppress_errors=False):
589
"""Commit the contents accrued within the current write group.
591
:param suppress_errors: if true, abort_write_group will catch and log
592
unexpected errors that happen during the abort, rather than
593
allowing them to propagate. Defaults to False.
595
:seealso: start_write_group.
597
if self._write_group is not self.get_transaction():
598
# has an unlock or relock occured ?
599
raise errors.BzrError('mismatched lock context and write group.')
601
self._abort_write_group()
602
except Exception, exc:
603
self._write_group = None
604
if not suppress_errors:
606
mutter('abort_write_group failed')
607
log_exception_quietly()
608
note('bzr: ERROR (ignored): %s', exc)
609
self._write_group = None
611
def _abort_write_group(self):
612
"""Template method for per-repository write group cleanup.
614
This is called during abort before the write group is considered to be
615
finished and should cleanup any internal state accrued during the write
616
group. There is no requirement that data handed to the repository be
617
*not* made available - this is not a rollback - but neither should any
618
attempt be made to ensure that data added is fully commited. Abort is
619
invoked when an error has occured so futher disk or network operations
620
may not be possible or may error and if possible should not be
624
def add_fallback_repository(self, repository):
625
"""Add a repository to use for looking up data not held locally.
627
:param repository: A repository.
629
if not self._format.supports_external_lookups:
630
raise errors.UnstackableRepositoryFormat(self._format, self.base)
631
self._check_fallback_repository(repository)
632
self._fallback_repositories.append(repository)
633
self.texts.add_fallback_versioned_files(repository.texts)
634
self.inventories.add_fallback_versioned_files(repository.inventories)
635
self.revisions.add_fallback_versioned_files(repository.revisions)
636
self.signatures.add_fallback_versioned_files(repository.signatures)
638
def _check_fallback_repository(self, repository):
639
"""Check that this repository can fallback to repository safely.
641
Raise an error if not.
643
:param repository: A repository to fallback to.
645
return InterRepository._assert_same_model(self, repository)
647
def add_inventory(self, revision_id, inv, parents):
648
"""Add the inventory inv to the repository as revision_id.
650
:param parents: The revision ids of the parents that revision_id
651
is known to have and are in the repository already.
653
:returns: The validator(which is a sha1 digest, though what is sha'd is
654
repository format specific) of the serialized inventory.
656
if not self.is_in_write_group():
657
raise AssertionError("%r not in write group" % (self,))
658
_mod_revision.check_not_reserved_id(revision_id)
659
if not (inv.revision_id is None or inv.revision_id == revision_id):
660
raise AssertionError(
661
"Mismatch between inventory revision"
662
" id and insertion revid (%r, %r)"
663
% (inv.revision_id, revision_id))
665
raise AssertionError()
666
inv_lines = self._serialise_inventory_to_lines(inv)
667
return self._inventory_add_lines(revision_id, parents,
668
inv_lines, check_content=False)
670
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
672
"""Add a new inventory expressed as a delta against another revision.
674
:param basis_revision_id: The inventory id the delta was created
675
against. (This does not have to be a direct parent.)
676
:param delta: The inventory delta (see Inventory.apply_delta for
678
:param new_revision_id: The revision id that the inventory is being
680
:param parents: The revision ids of the parents that revision_id is
681
known to have and are in the repository already. These are supplied
682
for repositories that depend on the inventory graph for revision
683
graph access, as well as for those that pun ancestry with delta
686
:returns: (validator, new_inv)
687
The validator(which is a sha1 digest, though what is sha'd is
688
repository format specific) of the serialized inventory, and the
691
if not self.is_in_write_group():
692
raise AssertionError("%r not in write group" % (self,))
693
_mod_revision.check_not_reserved_id(new_revision_id)
694
basis_tree = self.revision_tree(basis_revision_id)
695
basis_tree.lock_read()
697
# Note that this mutates the inventory of basis_tree, which not all
698
# inventory implementations may support: A better idiom would be to
699
# return a new inventory, but as there is no revision tree cache in
700
# repository this is safe for now - RBC 20081013
701
basis_inv = basis_tree.inventory
702
basis_inv.apply_delta(delta)
703
basis_inv.revision_id = new_revision_id
704
return (self.add_inventory(new_revision_id, basis_inv, parents),
709
def _inventory_add_lines(self, revision_id, parents, lines,
711
"""Store lines in inv_vf and return the sha1 of the inventory."""
712
parents = [(parent,) for parent in parents]
713
return self.inventories.add_lines((revision_id,), parents, lines,
714
check_content=check_content)[0]
716
def add_revision(self, revision_id, rev, inv=None, config=None):
717
"""Add rev to the revision store as revision_id.
719
:param revision_id: the revision id to use.
720
:param rev: The revision object.
721
:param inv: The inventory for the revision. if None, it will be looked
722
up in the inventory storer
723
:param config: If None no digital signature will be created.
724
If supplied its signature_needed method will be used
725
to determine if a signature should be made.
727
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
729
_mod_revision.check_not_reserved_id(revision_id)
730
if config is not None and config.signature_needed():
732
inv = self.get_inventory(revision_id)
733
plaintext = Testament(rev, inv).as_short_text()
734
self.store_revision_signature(
735
gpg.GPGStrategy(config), plaintext, revision_id)
736
# check inventory present
737
if not self.inventories.get_parent_map([(revision_id,)]):
739
raise errors.WeaveRevisionNotPresent(revision_id,
742
# yes, this is not suitable for adding with ghosts.
743
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
747
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
748
self._add_revision(rev)
750
def _add_revision(self, revision):
751
text = self._serializer.write_revision_to_string(revision)
752
key = (revision.revision_id,)
753
parents = tuple((parent,) for parent in revision.parent_ids)
754
self.revisions.add_lines(key, parents, osutils.split_lines(text))
756
def all_revision_ids(self):
757
"""Returns a list of all the revision ids in the repository.
759
This is conceptually deprecated because code should generally work on
760
the graph reachable from a particular revision, and ignore any other
761
revisions that might be present. There is no direct replacement
764
if 'evil' in debug.debug_flags:
765
mutter_callsite(2, "all_revision_ids is linear with history.")
766
return self._all_revision_ids()
768
def _all_revision_ids(self):
769
"""Returns a list of all the revision ids in the repository.
771
These are in as much topological order as the underlying store can
774
raise NotImplementedError(self._all_revision_ids)
776
def break_lock(self):
777
"""Break a lock if one is present from another instance.
779
Uses the ui factory to ask for confirmation if the lock may be from
782
self.control_files.break_lock()
785
def _eliminate_revisions_not_present(self, revision_ids):
786
"""Check every revision id in revision_ids to see if we have it.
788
Returns a set of the present revisions.
791
graph = self.get_graph()
792
parent_map = graph.get_parent_map(revision_ids)
793
# The old API returned a list, should this actually be a set?
794
return parent_map.keys()
797
def create(a_bzrdir):
798
"""Construct the current default format repository in a_bzrdir."""
799
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
801
def __init__(self, _format, a_bzrdir, control_files):
802
"""instantiate a Repository.
804
:param _format: The format of the repository on disk.
805
:param a_bzrdir: The BzrDir of the repository.
807
In the future we will have a single api for all stores for
808
getting file texts, inventories and revisions, then
809
this construct will accept instances of those things.
811
super(Repository, self).__init__()
812
self._format = _format
813
# the following are part of the public API for Repository:
814
self.bzrdir = a_bzrdir
815
self.control_files = control_files
816
self._transport = control_files._transport
817
self.base = self._transport.base
819
self._reconcile_does_inventory_gc = True
820
self._reconcile_fixes_text_parents = False
821
self._reconcile_backsup_inventory = True
822
# not right yet - should be more semantically clear ?
824
# TODO: make sure to construct the right store classes, etc, depending
825
# on whether escaping is required.
826
self._warn_if_deprecated()
827
self._write_group = None
828
# Additional places to query for data.
829
self._fallback_repositories = []
830
# An InventoryEntry cache, used during deserialization
831
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
834
return '%s(%r)' % (self.__class__.__name__,
837
def has_same_location(self, other):
838
"""Returns a boolean indicating if this repository is at the same
839
location as another repository.
841
This might return False even when two repository objects are accessing
842
the same physical repository via different URLs.
844
if self.__class__ is not other.__class__:
846
return (self._transport.base == other._transport.base)
848
def is_in_write_group(self):
849
"""Return True if there is an open write group.
851
:seealso: start_write_group.
853
return self._write_group is not None
856
return self.control_files.is_locked()
858
def is_write_locked(self):
859
"""Return True if this object is write locked."""
860
return self.is_locked() and self.control_files._lock_mode == 'w'
862
def lock_write(self, token=None):
863
"""Lock this repository for writing.
865
This causes caching within the repository obejct to start accumlating
866
data during reads, and allows a 'write_group' to be obtained. Write
867
groups must be used for actual data insertion.
869
:param token: if this is already locked, then lock_write will fail
870
unless the token matches the existing lock.
871
:returns: a token if this instance supports tokens, otherwise None.
872
:raises TokenLockingNotSupported: when a token is given but this
873
instance doesn't support using token locks.
874
:raises MismatchedToken: if the specified token doesn't match the token
875
of the existing lock.
876
:seealso: start_write_group.
878
A token should be passed in if you know that you have locked the object
879
some other way, and need to synchronise this object's state with that
882
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
884
locked = self.is_locked()
885
result = self.control_files.lock_write(token=token)
886
for repo in self._fallback_repositories:
887
# Writes don't affect fallback repos
894
locked = self.is_locked()
895
self.control_files.lock_read()
896
for repo in self._fallback_repositories:
901
def get_physical_lock_status(self):
902
return self.control_files.get_physical_lock_status()
904
def leave_lock_in_place(self):
905
"""Tell this repository not to release the physical lock when this
908
If lock_write doesn't return a token, then this method is not supported.
910
self.control_files.leave_in_place()
912
def dont_leave_lock_in_place(self):
913
"""Tell this repository to release the physical lock when this
914
object is unlocked, even if it didn't originally acquire it.
916
If lock_write doesn't return a token, then this method is not supported.
918
self.control_files.dont_leave_in_place()
921
def gather_stats(self, revid=None, committers=None):
922
"""Gather statistics from a revision id.
924
:param revid: The revision id to gather statistics from, if None, then
925
no revision specific statistics are gathered.
926
:param committers: Optional parameter controlling whether to grab
927
a count of committers from the revision specific statistics.
928
:return: A dictionary of statistics. Currently this contains:
929
committers: The number of committers if requested.
930
firstrev: A tuple with timestamp, timezone for the penultimate left
931
most ancestor of revid, if revid is not the NULL_REVISION.
932
latestrev: A tuple with timestamp, timezone for revid, if revid is
933
not the NULL_REVISION.
934
revisions: The total revision count in the repository.
935
size: An estimate disk size of the repository in bytes.
938
if revid and committers:
939
result['committers'] = 0
940
if revid and revid != _mod_revision.NULL_REVISION:
942
all_committers = set()
943
revisions = self.get_ancestry(revid)
944
# pop the leading None
946
first_revision = None
948
# ignore the revisions in the middle - just grab first and last
949
revisions = revisions[0], revisions[-1]
950
for revision in self.get_revisions(revisions):
951
if not first_revision:
952
first_revision = revision
954
all_committers.add(revision.committer)
955
last_revision = revision
957
result['committers'] = len(all_committers)
958
result['firstrev'] = (first_revision.timestamp,
959
first_revision.timezone)
960
result['latestrev'] = (last_revision.timestamp,
961
last_revision.timezone)
963
# now gather global repository information
964
# XXX: This is available for many repos regardless of listability.
965
if self.bzrdir.root_transport.listable():
966
# XXX: do we want to __define len__() ?
967
# Maybe the versionedfiles object should provide a different
968
# method to get the number of keys.
969
result['revisions'] = len(self.revisions.keys())
973
def find_branches(self, using=False):
974
"""Find branches underneath this repository.
976
This will include branches inside other branches.
978
:param using: If True, list only branches using this repository.
980
if using and not self.is_shared():
982
return [self.bzrdir.open_branch()]
983
except errors.NotBranchError:
985
class Evaluator(object):
988
self.first_call = True
990
def __call__(self, bzrdir):
991
# On the first call, the parameter is always the bzrdir
992
# containing the current repo.
993
if not self.first_call:
995
repository = bzrdir.open_repository()
996
except errors.NoRepositoryPresent:
999
return False, (None, repository)
1000
self.first_call = False
1002
value = (bzrdir.open_branch(), None)
1003
except errors.NotBranchError:
1004
value = (None, None)
1008
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1009
self.bzrdir.root_transport, evaluate=Evaluator()):
1010
if branch is not None:
1011
branches.append(branch)
1012
if not using and repository is not None:
1013
branches.extend(repository.find_branches())
1017
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1018
"""Return the revision ids that other has that this does not.
1020
These are returned in topological order.
1022
revision_id: only return revision ids included by revision_id.
1024
return InterRepository.get(other, self).search_missing_revision_ids(
1025
revision_id, find_ghosts)
1029
"""Open the repository rooted at base.
1031
For instance, if the repository is at URL/.bzr/repository,
1032
Repository.open(URL) -> a Repository instance.
1034
control = bzrdir.BzrDir.open(base)
1035
return control.open_repository()
1037
def copy_content_into(self, destination, revision_id=None):
1038
"""Make a complete copy of the content in self into destination.
1040
This is a destructive operation! Do not use it on existing
1043
return InterRepository.get(self, destination).copy_content(revision_id)
1045
def commit_write_group(self):
1046
"""Commit the contents accrued within the current write group.
1048
:seealso: start_write_group.
1050
if self._write_group is not self.get_transaction():
1051
# has an unlock or relock occured ?
1052
raise errors.BzrError('mismatched lock context %r and '
1054
(self.get_transaction(), self._write_group))
1055
self._commit_write_group()
1056
self._write_group = None
1058
def _commit_write_group(self):
1059
"""Template method for per-repository write group cleanup.
1061
This is called before the write group is considered to be
1062
finished and should ensure that all data handed to the repository
1063
for writing during the write group is safely committed (to the
1064
extent possible considering file system caching etc).
1067
def suspend_write_group(self):
1068
raise errors.UnsuspendableWriteGroup(self)
1070
def refresh_data(self):
1071
"""Re-read any data needed to to synchronise with disk.
1073
This method is intended to be called after another repository instance
1074
(such as one used by a smart server) has inserted data into the
1075
repository. It may not be called during a write group, but may be
1076
called at any other time.
1078
if self.is_in_write_group():
1079
raise errors.InternalBzrError(
1080
"May not refresh_data while in a write group.")
1081
self._refresh_data()
1083
def resume_write_group(self, tokens):
1084
if not self.is_write_locked():
1085
raise errors.NotWriteLocked(self)
1086
if self._write_group:
1087
raise errors.BzrError('already in a write group')
1088
self._resume_write_group(tokens)
1089
# so we can detect unlock/relock - the write group is now entered.
1090
self._write_group = self.get_transaction()
1092
def _resume_write_group(self, tokens):
1093
raise errors.UnsuspendableWriteGroup(self)
1095
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
1097
"""Fetch the content required to construct revision_id from source.
1099
If revision_id is None and fetch_spec is None, then all content is
1102
fetch() may not be used when the repository is in a write group -
1103
either finish the current write group before using fetch, or use
1104
fetch before starting the write group.
1106
:param find_ghosts: Find and copy revisions in the source that are
1107
ghosts in the target (and not reachable directly by walking out to
1108
the first-present revision in target from revision_id).
1109
:param revision_id: If specified, all the content needed for this
1110
revision ID will be copied to the target. Fetch will determine for
1111
itself which content needs to be copied.
1112
:param fetch_spec: If specified, a SearchResult or
1113
PendingAncestryResult that describes which revisions to copy. This
1114
allows copying multiple heads at once. Mutually exclusive with
1117
if fetch_spec is not None and revision_id is not None:
1118
raise AssertionError(
1119
"fetch_spec and revision_id are mutually exclusive.")
1120
if self.is_in_write_group():
1121
raise errors.InternalBzrError(
1122
"May not fetch while in a write group.")
1123
# fast path same-url fetch operations
1124
if self.has_same_location(source) and fetch_spec is None:
1125
# check that last_revision is in 'from' and then return a
1127
if (revision_id is not None and
1128
not _mod_revision.is_null(revision_id)):
1129
self.get_revision(revision_id)
1131
# if there is no specific appropriate InterRepository, this will get
1132
# the InterRepository base class, which raises an
1133
# IncompatibleRepositories when asked to fetch.
1134
inter = InterRepository.get(source, self)
1135
return inter.fetch(revision_id=revision_id, pb=pb,
1136
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
1138
def create_bundle(self, target, base, fileobj, format=None):
1139
return serializer.write_bundle(self, target, base, fileobj, format)
1141
def get_commit_builder(self, branch, parents, config, timestamp=None,
1142
timezone=None, committer=None, revprops=None,
1144
"""Obtain a CommitBuilder for this repository.
1146
:param branch: Branch to commit to.
1147
:param parents: Revision ids of the parents of the new revision.
1148
:param config: Configuration to use.
1149
:param timestamp: Optional timestamp recorded for commit.
1150
:param timezone: Optional timezone for timestamp.
1151
:param committer: Optional committer to set for commit.
1152
:param revprops: Optional dictionary of revision properties.
1153
:param revision_id: Optional revision id.
1155
result = self._commit_builder_class(self, parents, config,
1156
timestamp, timezone, committer, revprops, revision_id)
1157
self.start_write_group()
1161
if (self.control_files._lock_count == 1 and
1162
self.control_files._lock_mode == 'w'):
1163
if self._write_group is not None:
1164
self.abort_write_group()
1165
self.control_files.unlock()
1166
raise errors.BzrError(
1167
'Must end write groups before releasing write locks.')
1168
self.control_files.unlock()
1169
if self.control_files._lock_count == 0:
1170
self._inventory_entry_cache.clear()
1171
for repo in self._fallback_repositories:
1175
def clone(self, a_bzrdir, revision_id=None):
1176
"""Clone this repository into a_bzrdir using the current format.
1178
Currently no check is made that the format of this repository and
1179
the bzrdir format are compatible. FIXME RBC 20060201.
1181
:return: The newly created destination repository.
1183
# TODO: deprecate after 0.16; cloning this with all its settings is
1184
# probably not very useful -- mbp 20070423
1185
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1186
self.copy_content_into(dest_repo, revision_id)
1189
def start_write_group(self):
1190
"""Start a write group in the repository.
1192
Write groups are used by repositories which do not have a 1:1 mapping
1193
between file ids and backend store to manage the insertion of data from
1194
both fetch and commit operations.
1196
A write lock is required around the start_write_group/commit_write_group
1197
for the support of lock-requiring repository formats.
1199
One can only insert data into a repository inside a write group.
1203
if not self.is_write_locked():
1204
raise errors.NotWriteLocked(self)
1205
if self._write_group:
1206
raise errors.BzrError('already in a write group')
1207
self._start_write_group()
1208
# so we can detect unlock/relock - the write group is now entered.
1209
self._write_group = self.get_transaction()
1211
def _start_write_group(self):
1212
"""Template method for per-repository write group startup.
1214
This is called before the write group is considered to be
1219
def sprout(self, to_bzrdir, revision_id=None):
1220
"""Create a descendent repository for new development.
1222
Unlike clone, this does not copy the settings of the repository.
1224
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1225
dest_repo.fetch(self, revision_id=revision_id)
1228
def _create_sprouting_repo(self, a_bzrdir, shared):
1229
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1230
# use target default format.
1231
dest_repo = a_bzrdir.create_repository()
1233
# Most control formats need the repository to be specifically
1234
# created, but on some old all-in-one formats it's not needed
1236
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1237
except errors.UninitializableFormat:
1238
dest_repo = a_bzrdir.open_repository()
1241
def _get_sink(self):
1242
"""Return a sink for streaming into this repository."""
1243
return StreamSink(self)
1245
def _get_source(self, to_format):
1246
"""Return a source for streaming from this repository."""
1247
return StreamSource(self, to_format)
1250
def has_revision(self, revision_id):
1251
"""True if this repository has a copy of the revision."""
1252
return revision_id in self.has_revisions((revision_id,))
1255
def has_revisions(self, revision_ids):
1256
"""Probe to find out the presence of multiple revisions.
1258
:param revision_ids: An iterable of revision_ids.
1259
:return: A set of the revision_ids that were present.
1261
parent_map = self.revisions.get_parent_map(
1262
[(rev_id,) for rev_id in revision_ids])
1264
if _mod_revision.NULL_REVISION in revision_ids:
1265
result.add(_mod_revision.NULL_REVISION)
1266
result.update([key[0] for key in parent_map])
1270
def get_revision(self, revision_id):
1271
"""Return the Revision object for a named revision."""
1272
return self.get_revisions([revision_id])[0]
1275
def get_revision_reconcile(self, revision_id):
1276
"""'reconcile' helper routine that allows access to a revision always.
1278
This variant of get_revision does not cross check the weave graph
1279
against the revision one as get_revision does: but it should only
1280
be used by reconcile, or reconcile-alike commands that are correcting
1281
or testing the revision graph.
1283
return self._get_revisions([revision_id])[0]
1286
def get_revisions(self, revision_ids):
1287
"""Get many revisions at once."""
1288
return self._get_revisions(revision_ids)
1291
def _get_revisions(self, revision_ids):
1292
"""Core work logic to get many revisions without sanity checks."""
1293
for rev_id in revision_ids:
1294
if not rev_id or not isinstance(rev_id, basestring):
1295
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1296
keys = [(key,) for key in revision_ids]
1297
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1299
for record in stream:
1300
if record.storage_kind == 'absent':
1301
raise errors.NoSuchRevision(self, record.key[0])
1302
text = record.get_bytes_as('fulltext')
1303
rev = self._serializer.read_revision_from_string(text)
1304
revs[record.key[0]] = rev
1305
return [revs[revid] for revid in revision_ids]
1308
def get_revision_xml(self, revision_id):
1309
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1310
# would have already do it.
1311
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1312
rev = self.get_revision(revision_id)
1313
rev_tmp = cStringIO.StringIO()
1314
# the current serializer..
1315
self._serializer.write_revision(rev, rev_tmp)
1317
return rev_tmp.getvalue()
1319
def get_deltas_for_revisions(self, revisions):
1320
"""Produce a generator of revision deltas.
1322
Note that the input is a sequence of REVISIONS, not revision_ids.
1323
Trees will be held in memory until the generator exits.
1324
Each delta is relative to the revision's lefthand predecessor.
1326
required_trees = set()
1327
for revision in revisions:
1328
required_trees.add(revision.revision_id)
1329
required_trees.update(revision.parent_ids[:1])
1330
trees = dict((t.get_revision_id(), t) for
1331
t in self.revision_trees(required_trees))
1332
for revision in revisions:
1333
if not revision.parent_ids:
1334
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1336
old_tree = trees[revision.parent_ids[0]]
1337
yield trees[revision.revision_id].changes_from(old_tree)
1340
def get_revision_delta(self, revision_id):
1341
"""Return the delta for one revision.
1343
The delta is relative to the left-hand predecessor of the
1346
r = self.get_revision(revision_id)
1347
return list(self.get_deltas_for_revisions([r]))[0]
1350
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1351
signature = gpg_strategy.sign(plaintext)
1352
self.add_signature_text(revision_id, signature)
1355
def add_signature_text(self, revision_id, signature):
1356
self.signatures.add_lines((revision_id,), (),
1357
osutils.split_lines(signature))
1359
def find_text_key_references(self):
1360
"""Find the text key references within the repository.
1362
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1363
to whether they were referred to by the inventory of the
1364
revision_id that they contain. The inventory texts from all present
1365
revision ids are assessed to generate this report.
1367
revision_keys = self.revisions.keys()
1368
w = self.inventories
1369
pb = ui.ui_factory.nested_progress_bar()
1371
return self._find_text_key_references_from_xml_inventory_lines(
1372
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1376
def _find_text_key_references_from_xml_inventory_lines(self,
1378
"""Core routine for extracting references to texts from inventories.
1380
This performs the translation of xml lines to revision ids.
1382
:param line_iterator: An iterator of lines, origin_version_id
1383
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1384
to whether they were referred to by the inventory of the
1385
revision_id that they contain. Note that if that revision_id was
1386
not part of the line_iterator's output then False will be given -
1387
even though it may actually refer to that key.
1389
if not self._serializer.support_altered_by_hack:
1390
raise AssertionError(
1391
"_find_text_key_references_from_xml_inventory_lines only "
1392
"supported for branches which store inventory as unnested xml"
1393
", not on %r" % self)
1396
# this code needs to read every new line in every inventory for the
1397
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1398
# not present in one of those inventories is unnecessary but not
1399
# harmful because we are filtering by the revision id marker in the
1400
# inventory lines : we only select file ids altered in one of those
1401
# revisions. We don't need to see all lines in the inventory because
1402
# only those added in an inventory in rev X can contain a revision=X
1404
unescape_revid_cache = {}
1405
unescape_fileid_cache = {}
1407
# jam 20061218 In a big fetch, this handles hundreds of thousands
1408
# of lines, so it has had a lot of inlining and optimizing done.
1409
# Sorry that it is a little bit messy.
1410
# Move several functions to be local variables, since this is a long
1412
search = self._file_ids_altered_regex.search
1413
unescape = _unescape_xml
1414
setdefault = result.setdefault
1415
for line, line_key in line_iterator:
1416
match = search(line)
1419
# One call to match.group() returning multiple items is quite a
1420
# bit faster than 2 calls to match.group() each returning 1
1421
file_id, revision_id = match.group('file_id', 'revision_id')
1423
# Inlining the cache lookups helps a lot when you make 170,000
1424
# lines and 350k ids, versus 8.4 unique ids.
1425
# Using a cache helps in 2 ways:
1426
# 1) Avoids unnecessary decoding calls
1427
# 2) Re-uses cached strings, which helps in future set and
1429
# (2) is enough that removing encoding entirely along with
1430
# the cache (so we are using plain strings) results in no
1431
# performance improvement.
1433
revision_id = unescape_revid_cache[revision_id]
1435
unescaped = unescape(revision_id)
1436
unescape_revid_cache[revision_id] = unescaped
1437
revision_id = unescaped
1439
# Note that unconditionally unescaping means that we deserialise
1440
# every fileid, which for general 'pull' is not great, but we don't
1441
# really want to have some many fulltexts that this matters anyway.
1444
file_id = unescape_fileid_cache[file_id]
1446
unescaped = unescape(file_id)
1447
unescape_fileid_cache[file_id] = unescaped
1450
key = (file_id, revision_id)
1451
setdefault(key, False)
1452
if revision_id == line_key[-1]:
1456
def _inventory_xml_lines_for_keys(self, keys):
1457
"""Get a line iterator of the sort needed for findind references.
1459
Not relevant for non-xml inventory repositories.
1461
Ghosts in revision_keys are ignored.
1463
:param revision_keys: The revision keys for the inventories to inspect.
1464
:return: An iterator over (inventory line, revid) for the fulltexts of
1465
all of the xml inventories specified by revision_keys.
1467
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1468
for record in stream:
1469
if record.storage_kind != 'absent':
1470
chunks = record.get_bytes_as('chunked')
1471
revid = record.key[-1]
1472
lines = osutils.chunks_to_lines(chunks)
1476
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1478
"""Helper routine for fileids_altered_by_revision_ids.
1480
This performs the translation of xml lines to revision ids.
1482
:param line_iterator: An iterator of lines, origin_version_id
1483
:param revision_ids: The revision ids to filter for. This should be a
1484
set or other type which supports efficient __contains__ lookups, as
1485
the revision id from each parsed line will be looked up in the
1486
revision_ids filter.
1487
:return: a dictionary mapping altered file-ids to an iterable of
1488
revision_ids. Each altered file-ids has the exact revision_ids that
1489
altered it listed explicitly.
1491
seen = set(self._find_text_key_references_from_xml_inventory_lines(
1492
line_iterator).iterkeys())
1493
# Note that revision_ids are revision keys.
1494
parent_maps = self.revisions.get_parent_map(revision_ids)
1496
map(parents.update, parent_maps.itervalues())
1497
parents.difference_update(revision_ids)
1498
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
1499
self._inventory_xml_lines_for_keys(parents)))
1500
new_keys = seen - parent_seen
1502
setdefault = result.setdefault
1503
for key in new_keys:
1504
setdefault(key[0], set()).add(key[-1])
1507
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1508
"""Find the file ids and versions affected by revisions.
1510
:param revisions: an iterable containing revision ids.
1511
:param _inv_weave: The inventory weave from this repository or None.
1512
If None, the inventory weave will be opened automatically.
1513
:return: a dictionary mapping altered file-ids to an iterable of
1514
revision_ids. Each altered file-ids has the exact revision_ids that
1515
altered it listed explicitly.
1517
selected_keys = set((revid,) for revid in revision_ids)
1518
w = _inv_weave or self.inventories
1519
pb = ui.ui_factory.nested_progress_bar()
1521
return self._find_file_ids_from_xml_inventory_lines(
1522
w.iter_lines_added_or_present_in_keys(
1523
selected_keys, pb=pb),
1528
def iter_files_bytes(self, desired_files):
1529
"""Iterate through file versions.
1531
Files will not necessarily be returned in the order they occur in
1532
desired_files. No specific order is guaranteed.
1534
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1535
value supplied by the caller as part of desired_files. It should
1536
uniquely identify the file version in the caller's context. (Examples:
1537
an index number or a TreeTransform trans_id.)
1539
bytes_iterator is an iterable of bytestrings for the file. The
1540
kind of iterable and length of the bytestrings are unspecified, but for
1541
this implementation, it is a list of bytes produced by
1542
VersionedFile.get_record_stream().
1544
:param desired_files: a list of (file_id, revision_id, identifier)
1548
for file_id, revision_id, callable_data in desired_files:
1549
text_keys[(file_id, revision_id)] = callable_data
1550
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1551
if record.storage_kind == 'absent':
1552
raise errors.RevisionNotPresent(record.key, self)
1553
yield text_keys[record.key], record.get_bytes_as('fulltext')
1555
def _generate_text_key_index(self, text_key_references=None,
1557
"""Generate a new text key index for the repository.
1559
This is an expensive function that will take considerable time to run.
1561
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1562
list of parents, also text keys. When a given key has no parents,
1563
the parents list will be [NULL_REVISION].
1565
# All revisions, to find inventory parents.
1566
if ancestors is None:
1567
graph = self.get_graph()
1568
ancestors = graph.get_parent_map(self.all_revision_ids())
1569
if text_key_references is None:
1570
text_key_references = self.find_text_key_references()
1571
pb = ui.ui_factory.nested_progress_bar()
1573
return self._do_generate_text_key_index(ancestors,
1574
text_key_references, pb)
1578
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1579
"""Helper for _generate_text_key_index to avoid deep nesting."""
1580
revision_order = tsort.topo_sort(ancestors)
1581
invalid_keys = set()
1583
for revision_id in revision_order:
1584
revision_keys[revision_id] = set()
1585
text_count = len(text_key_references)
1586
# a cache of the text keys to allow reuse; costs a dict of all the
1587
# keys, but saves a 2-tuple for every child of a given key.
1589
for text_key, valid in text_key_references.iteritems():
1591
invalid_keys.add(text_key)
1593
revision_keys[text_key[1]].add(text_key)
1594
text_key_cache[text_key] = text_key
1595
del text_key_references
1597
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1598
NULL_REVISION = _mod_revision.NULL_REVISION
1599
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1600
# too small for large or very branchy trees. However, for 55K path
1601
# trees, it would be easy to use too much memory trivially. Ideally we
1602
# could gauge this by looking at available real memory etc, but this is
1603
# always a tricky proposition.
1604
inventory_cache = lru_cache.LRUCache(10)
1605
batch_size = 10 # should be ~150MB on a 55K path tree
1606
batch_count = len(revision_order) / batch_size + 1
1608
pb.update("Calculating text parents", processed_texts, text_count)
1609
for offset in xrange(batch_count):
1610
to_query = revision_order[offset * batch_size:(offset + 1) *
1614
for rev_tree in self.revision_trees(to_query):
1615
revision_id = rev_tree.get_revision_id()
1616
parent_ids = ancestors[revision_id]
1617
for text_key in revision_keys[revision_id]:
1618
pb.update("Calculating text parents", processed_texts)
1619
processed_texts += 1
1620
candidate_parents = []
1621
for parent_id in parent_ids:
1622
parent_text_key = (text_key[0], parent_id)
1624
check_parent = parent_text_key not in \
1625
revision_keys[parent_id]
1627
# the parent parent_id is a ghost:
1628
check_parent = False
1629
# truncate the derived graph against this ghost.
1630
parent_text_key = None
1632
# look at the parent commit details inventories to
1633
# determine possible candidates in the per file graph.
1636
inv = inventory_cache[parent_id]
1638
inv = self.revision_tree(parent_id).inventory
1639
inventory_cache[parent_id] = inv
1640
parent_entry = inv._byid.get(text_key[0], None)
1641
if parent_entry is not None:
1643
text_key[0], parent_entry.revision)
1645
parent_text_key = None
1646
if parent_text_key is not None:
1647
candidate_parents.append(
1648
text_key_cache[parent_text_key])
1649
parent_heads = text_graph.heads(candidate_parents)
1650
new_parents = list(parent_heads)
1651
new_parents.sort(key=lambda x:candidate_parents.index(x))
1652
if new_parents == []:
1653
new_parents = [NULL_REVISION]
1654
text_index[text_key] = new_parents
1656
for text_key in invalid_keys:
1657
text_index[text_key] = [NULL_REVISION]
1660
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1661
"""Get an iterable listing the keys of all the data introduced by a set
1664
The keys will be ordered so that the corresponding items can be safely
1665
fetched and inserted in that order.
1667
:returns: An iterable producing tuples of (knit-kind, file-id,
1668
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1669
'revisions'. file-id is None unless knit-kind is 'file'.
1671
# XXX: it's a bit weird to control the inventory weave caching in this
1672
# generator. Ideally the caching would be done in fetch.py I think. Or
1673
# maybe this generator should explicitly have the contract that it
1674
# should not be iterated until the previously yielded item has been
1676
inv_w = self.inventories
1678
# file ids that changed
1679
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1681
num_file_ids = len(file_ids)
1682
for file_id, altered_versions in file_ids.iteritems():
1683
if _files_pb is not None:
1684
_files_pb.update("fetch texts", count, num_file_ids)
1686
yield ("file", file_id, altered_versions)
1687
# We're done with the files_pb. Note that it finished by the caller,
1688
# just as it was created by the caller.
1692
yield ("inventory", None, revision_ids)
1695
# XXX: Note ATM no callers actually pay attention to this return
1696
# instead they just use the list of revision ids and ignore
1697
# missing sigs. Consider removing this work entirely
1698
revisions_with_signatures = set(self.signatures.get_parent_map(
1699
[(r,) for r in revision_ids]))
1700
revisions_with_signatures = set(
1701
[r for (r,) in revisions_with_signatures])
1702
revisions_with_signatures.intersection_update(revision_ids)
1703
yield ("signatures", None, revisions_with_signatures)
1706
yield ("revisions", None, revision_ids)
1709
def get_inventory(self, revision_id):
1710
"""Get Inventory object by revision id."""
1711
return self.iter_inventories([revision_id]).next()
1713
def iter_inventories(self, revision_ids):
1714
"""Get many inventories by revision_ids.
1716
This will buffer some or all of the texts used in constructing the
1717
inventories in memory, but will only parse a single inventory at a
1720
:return: An iterator of inventories.
1722
if ((None in revision_ids)
1723
or (_mod_revision.NULL_REVISION in revision_ids)):
1724
raise ValueError('cannot get null revision inventory')
1725
return self._iter_inventories(revision_ids)
1727
def _iter_inventories(self, revision_ids):
1728
"""single-document based inventory iteration."""
1729
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1730
yield self.deserialise_inventory(revision_id, text)
1732
def _iter_inventory_xmls(self, revision_ids):
1733
keys = [(revision_id,) for revision_id in revision_ids]
1734
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1736
for record in stream:
1737
if record.storage_kind != 'absent':
1738
text_chunks[record.key] = record.get_bytes_as('chunked')
1740
raise errors.NoSuchRevision(self, record.key)
1742
chunks = text_chunks.pop(key)
1743
yield ''.join(chunks), key[-1]
1745
def deserialise_inventory(self, revision_id, xml):
1746
"""Transform the xml into an inventory object.
1748
:param revision_id: The expected revision id of the inventory.
1749
:param xml: A serialised inventory.
1751
result = self._serializer.read_inventory_from_string(xml, revision_id,
1752
entry_cache=self._inventory_entry_cache)
1753
if result.revision_id != revision_id:
1754
raise AssertionError('revision id mismatch %s != %s' % (
1755
result.revision_id, revision_id))
1758
def serialise_inventory(self, inv):
1759
return self._serializer.write_inventory_to_string(inv)
1761
def _serialise_inventory_to_lines(self, inv):
1762
return self._serializer.write_inventory_to_lines(inv)
1764
def get_serializer_format(self):
1765
return self._serializer.format_num
1768
def get_inventory_xml(self, revision_id):
1769
"""Get inventory XML as a file object."""
1770
texts = self._iter_inventory_xmls([revision_id])
1772
text, revision_id = texts.next()
1773
except StopIteration:
1774
raise errors.HistoryMissing(self, 'inventory', revision_id)
1778
def get_inventory_sha1(self, revision_id):
1779
"""Return the sha1 hash of the inventory entry
1781
return self.get_revision(revision_id).inventory_sha1
1783
def iter_reverse_revision_history(self, revision_id):
1784
"""Iterate backwards through revision ids in the lefthand history
1786
:param revision_id: The revision id to start with. All its lefthand
1787
ancestors will be traversed.
1789
graph = self.get_graph()
1790
next_id = revision_id
1792
if next_id in (None, _mod_revision.NULL_REVISION):
1795
# Note: The following line may raise KeyError in the event of
1796
# truncated history. We decided not to have a try:except:raise
1797
# RevisionNotPresent here until we see a use for it, because of the
1798
# cost in an inner loop that is by its very nature O(history).
1799
# Robert Collins 20080326
1800
parents = graph.get_parent_map([next_id])[next_id]
1801
if len(parents) == 0:
1804
next_id = parents[0]
1807
def get_revision_inventory(self, revision_id):
1808
"""Return inventory of a past revision."""
1809
# TODO: Unify this with get_inventory()
1810
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1811
# must be the same as its revision, so this is trivial.
1812
if revision_id is None:
1813
# This does not make sense: if there is no revision,
1814
# then it is the current tree inventory surely ?!
1815
# and thus get_root_id() is something that looks at the last
1816
# commit on the branch, and the get_root_id is an inventory check.
1817
raise NotImplementedError
1818
# return Inventory(self.get_root_id())
1820
return self.get_inventory(revision_id)
1822
def is_shared(self):
1823
"""Return True if this repository is flagged as a shared repository."""
1824
raise NotImplementedError(self.is_shared)
1827
def reconcile(self, other=None, thorough=False):
1828
"""Reconcile this repository."""
1829
from bzrlib.reconcile import RepoReconciler
1830
reconciler = RepoReconciler(self, thorough=thorough)
1831
reconciler.reconcile()
1834
def _refresh_data(self):
1835
"""Helper called from lock_* to ensure coherency with disk.
1837
The default implementation does nothing; it is however possible
1838
for repositories to maintain loaded indices across multiple locks
1839
by checking inside their implementation of this method to see
1840
whether their indices are still valid. This depends of course on
1841
the disk format being validatable in this manner. This method is
1842
also called by the refresh_data() public interface to cause a refresh
1843
to occur while in a write lock so that data inserted by a smart server
1844
push operation is visible on the client's instance of the physical
1849
def revision_tree(self, revision_id):
1850
"""Return Tree for a revision on this branch.
1852
`revision_id` may be NULL_REVISION for the empty tree revision.
1854
revision_id = _mod_revision.ensure_null(revision_id)
1855
# TODO: refactor this to use an existing revision object
1856
# so we don't need to read it in twice.
1857
if revision_id == _mod_revision.NULL_REVISION:
1858
return RevisionTree(self, Inventory(root_id=None),
1859
_mod_revision.NULL_REVISION)
1861
inv = self.get_revision_inventory(revision_id)
1862
return RevisionTree(self, inv, revision_id)
1864
def revision_trees(self, revision_ids):
1865
"""Return Tree for a revision on this branch.
1867
`revision_id` may not be None or 'null:'"""
1868
inventories = self.iter_inventories(revision_ids)
1869
for inv in inventories:
1870
yield RevisionTree(self, inv, inv.revision_id)
1873
def get_ancestry(self, revision_id, topo_sorted=True):
1874
"""Return a list of revision-ids integrated by a revision.
1876
The first element of the list is always None, indicating the origin
1877
revision. This might change when we have history horizons, or
1878
perhaps we should have a new API.
1880
This is topologically sorted.
1882
if _mod_revision.is_null(revision_id):
1884
if not self.has_revision(revision_id):
1885
raise errors.NoSuchRevision(self, revision_id)
1886
graph = self.get_graph()
1888
search = graph._make_breadth_first_searcher([revision_id])
1891
found, ghosts = search.next_with_ghosts()
1892
except StopIteration:
1895
if _mod_revision.NULL_REVISION in keys:
1896
keys.remove(_mod_revision.NULL_REVISION)
1898
parent_map = graph.get_parent_map(keys)
1899
keys = tsort.topo_sort(parent_map)
1900
return [None] + list(keys)
1903
"""Compress the data within the repository.
1905
This operation only makes sense for some repository types. For other
1906
types it should be a no-op that just returns.
1908
This stub method does not require a lock, but subclasses should use
1909
@needs_write_lock as this is a long running call its reasonable to
1910
implicitly lock for the user.
1913
def get_transaction(self):
1914
return self.control_files.get_transaction()
1916
def get_parent_map(self, revision_ids):
1917
"""See graph._StackedParentsProvider.get_parent_map"""
1918
# revisions index works in keys; this just works in revisions
1919
# therefore wrap and unwrap
1922
for revision_id in revision_ids:
1923
if revision_id == _mod_revision.NULL_REVISION:
1924
result[revision_id] = ()
1925
elif revision_id is None:
1926
raise ValueError('get_parent_map(None) is not valid')
1928
query_keys.append((revision_id ,))
1929
for ((revision_id,), parent_keys) in \
1930
self.revisions.get_parent_map(query_keys).iteritems():
1932
result[revision_id] = tuple(parent_revid
1933
for (parent_revid,) in parent_keys)
1935
result[revision_id] = (_mod_revision.NULL_REVISION,)
1938
def _make_parents_provider(self):
1941
def get_graph(self, other_repository=None):
1942
"""Return the graph walker for this repository format"""
1943
parents_provider = self._make_parents_provider()
1944
if (other_repository is not None and
1945
not self.has_same_location(other_repository)):
1946
parents_provider = graph._StackedParentsProvider(
1947
[parents_provider, other_repository._make_parents_provider()])
1948
return graph.Graph(parents_provider)
1950
def _get_versioned_file_checker(self, text_key_references=None):
1951
"""Return an object suitable for checking versioned files.
1953
:param text_key_references: if non-None, an already built
1954
dictionary mapping text keys ((fileid, revision_id) tuples)
1955
to whether they were referred to by the inventory of the
1956
revision_id that they contain. If None, this will be
1959
return _VersionedFileChecker(self,
1960
text_key_references=text_key_references)
1962
def revision_ids_to_search_result(self, result_set):
1963
"""Convert a set of revision ids to a graph SearchResult."""
1964
result_parents = set()
1965
for parents in self.get_graph().get_parent_map(
1966
result_set).itervalues():
1967
result_parents.update(parents)
1968
included_keys = result_set.intersection(result_parents)
1969
start_keys = result_set.difference(included_keys)
1970
exclude_keys = result_parents.difference(result_set)
1971
result = graph.SearchResult(start_keys, exclude_keys,
1972
len(result_set), result_set)
1976
def set_make_working_trees(self, new_value):
1977
"""Set the policy flag for making working trees when creating branches.
1979
This only applies to branches that use this repository.
1981
The default is 'True'.
1982
:param new_value: True to restore the default, False to disable making
1985
raise NotImplementedError(self.set_make_working_trees)
1987
def make_working_trees(self):
1988
"""Returns the policy for making working trees on new branches."""
1989
raise NotImplementedError(self.make_working_trees)
1992
def sign_revision(self, revision_id, gpg_strategy):
1993
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1994
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1997
def has_signature_for_revision_id(self, revision_id):
1998
"""Query for a revision signature for revision_id in the repository."""
1999
if not self.has_revision(revision_id):
2000
raise errors.NoSuchRevision(self, revision_id)
2001
sig_present = (1 == len(
2002
self.signatures.get_parent_map([(revision_id,)])))
2006
def get_signature_text(self, revision_id):
2007
"""Return the text for a signature."""
2008
stream = self.signatures.get_record_stream([(revision_id,)],
2010
record = stream.next()
2011
if record.storage_kind == 'absent':
2012
raise errors.NoSuchRevision(self, revision_id)
2013
return record.get_bytes_as('fulltext')
2016
def check(self, revision_ids=None):
2017
"""Check consistency of all history of given revision_ids.
2019
Different repository implementations should override _check().
2021
:param revision_ids: A non-empty list of revision_ids whose ancestry
2022
will be checked. Typically the last revision_id of a branch.
2024
return self._check(revision_ids)
2026
def _check(self, revision_ids):
2027
result = check.Check(self)
2031
def _warn_if_deprecated(self):
2032
global _deprecation_warning_done
2033
if _deprecation_warning_done:
2035
_deprecation_warning_done = True
2036
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2037
% (self._format, self.bzrdir.transport.base))
2039
def supports_rich_root(self):
2040
return self._format.rich_root_data
2042
def _check_ascii_revisionid(self, revision_id, method):
2043
"""Private helper for ascii-only repositories."""
2044
# weave repositories refuse to store revisionids that are non-ascii.
2045
if revision_id is not None:
2046
# weaves require ascii revision ids.
2047
if isinstance(revision_id, unicode):
2049
revision_id.encode('ascii')
2050
except UnicodeEncodeError:
2051
raise errors.NonAsciiRevisionId(method, self)
2054
revision_id.decode('ascii')
2055
except UnicodeDecodeError:
2056
raise errors.NonAsciiRevisionId(method, self)
2058
def revision_graph_can_have_wrong_parents(self):
2059
"""Is it possible for this repository to have a revision graph with
2062
If True, then this repository must also implement
2063
_find_inconsistent_revision_parents so that check and reconcile can
2064
check for inconsistencies before proceeding with other checks that may
2065
depend on the revision index being consistent.
2067
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2070
# remove these delegates a while after bzr 0.15
2071
def __make_delegated(name, from_module):
2072
def _deprecated_repository_forwarder():
2073
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2074
% (name, from_module),
2077
m = __import__(from_module, globals(), locals(), [name])
2079
return getattr(m, name)
2080
except AttributeError:
2081
raise AttributeError('module %s has no name %s'
2083
globals()[name] = _deprecated_repository_forwarder
2086
'AllInOneRepository',
2087
'WeaveMetaDirRepository',
2088
'PreSplitOutRepositoryFormat',
2089
'RepositoryFormat4',
2090
'RepositoryFormat5',
2091
'RepositoryFormat6',
2092
'RepositoryFormat7',
2094
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2098
'RepositoryFormatKnit',
2099
'RepositoryFormatKnit1',
2101
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2104
def install_revision(repository, rev, revision_tree):
2105
"""Install all revision data into a repository."""
2106
install_revisions(repository, [(rev, revision_tree, None)])
2109
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2110
"""Install all revision data into a repository.
2112
Accepts an iterable of revision, tree, signature tuples. The signature
2115
repository.start_write_group()
2117
for n, (revision, revision_tree, signature) in enumerate(iterable):
2118
_install_revision(repository, revision, revision_tree, signature)
2120
pb.update('Transferring revisions', n + 1, num_revisions)
2122
repository.abort_write_group()
2125
repository.commit_write_group()
2128
def _install_revision(repository, rev, revision_tree, signature):
2129
"""Install all revision data into a repository."""
2130
present_parents = []
2132
for p_id in rev.parent_ids:
2133
if repository.has_revision(p_id):
2134
present_parents.append(p_id)
2135
parent_trees[p_id] = repository.revision_tree(p_id)
2137
parent_trees[p_id] = repository.revision_tree(
2138
_mod_revision.NULL_REVISION)
2140
inv = revision_tree.inventory
2141
entries = inv.iter_entries()
2142
# backwards compatibility hack: skip the root id.
2143
if not repository.supports_rich_root():
2144
path, root = entries.next()
2145
if root.revision != rev.revision_id:
2146
raise errors.IncompatibleRevision(repr(repository))
2148
for path, ie in entries:
2149
text_keys[(ie.file_id, ie.revision)] = ie
2150
text_parent_map = repository.texts.get_parent_map(text_keys)
2151
missing_texts = set(text_keys) - set(text_parent_map)
2152
# Add the texts that are not already present
2153
for text_key in missing_texts:
2154
ie = text_keys[text_key]
2156
# FIXME: TODO: The following loop overlaps/duplicates that done by
2157
# commit to determine parents. There is a latent/real bug here where
2158
# the parents inserted are not those commit would do - in particular
2159
# they are not filtered by heads(). RBC, AB
2160
for revision, tree in parent_trees.iteritems():
2161
if ie.file_id not in tree:
2163
parent_id = tree.inventory[ie.file_id].revision
2164
if parent_id in text_parents:
2166
text_parents.append((ie.file_id, parent_id))
2167
lines = revision_tree.get_file(ie.file_id).readlines()
2168
repository.texts.add_lines(text_key, text_parents, lines)
2170
# install the inventory
2171
repository.add_inventory(rev.revision_id, inv, present_parents)
2172
except errors.RevisionAlreadyPresent:
2174
if signature is not None:
2175
repository.add_signature_text(rev.revision_id, signature)
2176
repository.add_revision(rev.revision_id, rev, inv)
2179
class MetaDirRepository(Repository):
2180
"""Repositories in the new meta-dir layout.
2182
:ivar _transport: Transport for access to repository control files,
2183
typically pointing to .bzr/repository.
2186
def __init__(self, _format, a_bzrdir, control_files):
2187
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2188
self._transport = control_files._transport
2190
def is_shared(self):
2191
"""Return True if this repository is flagged as a shared repository."""
2192
return self._transport.has('shared-storage')
2195
def set_make_working_trees(self, new_value):
2196
"""Set the policy flag for making working trees when creating branches.
2198
This only applies to branches that use this repository.
2200
The default is 'True'.
2201
:param new_value: True to restore the default, False to disable making
2206
self._transport.delete('no-working-trees')
2207
except errors.NoSuchFile:
2210
self._transport.put_bytes('no-working-trees', '',
2211
mode=self.bzrdir._get_file_mode())
2213
def make_working_trees(self):
2214
"""Returns the policy for making working trees on new branches."""
2215
return not self._transport.has('no-working-trees')
2218
class MetaDirVersionedFileRepository(MetaDirRepository):
2219
"""Repositories in a meta-dir, that work via versioned file objects."""
2221
def __init__(self, _format, a_bzrdir, control_files):
2222
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2226
network_format_registry = registry.FormatRegistry()
2227
"""Registry of formats indexed by their network name.
2229
The network name for a repository format is an identifier that can be used when
2230
referring to formats with smart server operations. See
2231
RepositoryFormat.network_name() for more detail.
2235
format_registry = registry.FormatRegistry(network_format_registry)
2236
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2238
This can contain either format instances themselves, or classes/factories that
2239
can be called to obtain one.
2243
#####################################################################
2244
# Repository Formats
2246
class RepositoryFormat(object):
2247
"""A repository format.
2249
Formats provide four things:
2250
* An initialization routine to construct repository data on disk.
2251
* a optional format string which is used when the BzrDir supports
2253
* an open routine which returns a Repository instance.
2254
* A network name for referring to the format in smart server RPC
2257
There is one and only one Format subclass for each on-disk format. But
2258
there can be one Repository subclass that is used for several different
2259
formats. The _format attribute on a Repository instance can be used to
2260
determine the disk format.
2262
Formats are placed in a registry by their format string for reference
2263
during opening. These should be subclasses of RepositoryFormat for
2266
Once a format is deprecated, just deprecate the initialize and open
2267
methods on the format class. Do not deprecate the object, as the
2268
object may be created even when a repository instnace hasn't been
2271
Common instance attributes:
2272
_matchingbzrdir - the bzrdir format that the repository format was
2273
originally written to work with. This can be used if manually
2274
constructing a bzrdir and repository, or more commonly for test suite
2278
# Set to True or False in derived classes. True indicates that the format
2279
# supports ghosts gracefully.
2280
supports_ghosts = None
2281
# Can this repository be given external locations to lookup additional
2282
# data. Set to True or False in derived classes.
2283
supports_external_lookups = None
2284
# What order should fetch operations request streams in?
2285
# The default is unordered as that is the cheapest for an origin to
2287
_fetch_order = 'unordered'
2288
# Does this repository format use deltas that can be fetched as-deltas ?
2289
# (E.g. knits, where the knit deltas can be transplanted intact.
2290
# We default to False, which will ensure that enough data to get
2291
# a full text out of any fetch stream will be grabbed.
2292
_fetch_uses_deltas = False
2293
# Should fetch trigger a reconcile after the fetch? Only needed for
2294
# some repository formats that can suffer internal inconsistencies.
2295
_fetch_reconcile = False
2298
return "<%s>" % self.__class__.__name__
2300
def __eq__(self, other):
2301
# format objects are generally stateless
2302
return isinstance(other, self.__class__)
2304
def __ne__(self, other):
2305
return not self == other
2308
def find_format(klass, a_bzrdir):
2309
"""Return the format for the repository object in a_bzrdir.
2311
This is used by bzr native formats that have a "format" file in
2312
the repository. Other methods may be used by different types of
2316
transport = a_bzrdir.get_repository_transport(None)
2317
format_string = transport.get("format").read()
2318
return format_registry.get(format_string)
2319
except errors.NoSuchFile:
2320
raise errors.NoRepositoryPresent(a_bzrdir)
2322
raise errors.UnknownFormatError(format=format_string,
2326
def register_format(klass, format):
2327
format_registry.register(format.get_format_string(), format)
2330
def unregister_format(klass, format):
2331
format_registry.remove(format.get_format_string())
2334
def get_default_format(klass):
2335
"""Return the current default format."""
2336
from bzrlib import bzrdir
2337
return bzrdir.format_registry.make_bzrdir('default').repository_format
2339
def get_format_string(self):
2340
"""Return the ASCII format string that identifies this format.
2342
Note that in pre format ?? repositories the format string is
2343
not permitted nor written to disk.
2345
raise NotImplementedError(self.get_format_string)
2347
def get_format_description(self):
2348
"""Return the short description for this format."""
2349
raise NotImplementedError(self.get_format_description)
2351
# TODO: this shouldn't be in the base class, it's specific to things that
2352
# use weaves or knits -- mbp 20070207
2353
def _get_versioned_file_store(self,
2358
versionedfile_class=None,
2359
versionedfile_kwargs={},
2361
if versionedfile_class is None:
2362
versionedfile_class = self._versionedfile_class
2363
weave_transport = control_files._transport.clone(name)
2364
dir_mode = control_files._dir_mode
2365
file_mode = control_files._file_mode
2366
return VersionedFileStore(weave_transport, prefixed=prefixed,
2368
file_mode=file_mode,
2369
versionedfile_class=versionedfile_class,
2370
versionedfile_kwargs=versionedfile_kwargs,
2373
def initialize(self, a_bzrdir, shared=False):
2374
"""Initialize a repository of this format in a_bzrdir.
2376
:param a_bzrdir: The bzrdir to put the new repository in it.
2377
:param shared: The repository should be initialized as a sharable one.
2378
:returns: The new repository object.
2380
This may raise UninitializableFormat if shared repository are not
2381
compatible the a_bzrdir.
2383
raise NotImplementedError(self.initialize)
2385
def is_supported(self):
2386
"""Is this format supported?
2388
Supported formats must be initializable and openable.
2389
Unsupported formats may not support initialization or committing or
2390
some other features depending on the reason for not being supported.
2394
def network_name(self):
2395
"""A simple byte string uniquely identifying this format for RPC calls.
2397
MetaDir repository formats use their disk format string to identify the
2398
repository over the wire. All in one formats such as bzr < 0.8, and
2399
foreign formats like svn/git and hg should use some marker which is
2400
unique and immutable.
2402
raise NotImplementedError(self.network_name)
2404
def check_conversion_target(self, target_format):
2405
raise NotImplementedError(self.check_conversion_target)
2407
def open(self, a_bzrdir, _found=False):
2408
"""Return an instance of this format for the bzrdir a_bzrdir.
2410
_found is a private parameter, do not use it.
2412
raise NotImplementedError(self.open)
2415
class MetaDirRepositoryFormat(RepositoryFormat):
2416
"""Common base class for the new repositories using the metadir layout."""
2418
rich_root_data = False
2419
supports_tree_reference = False
2420
supports_external_lookups = False
2423
def _matchingbzrdir(self):
2424
matching = bzrdir.BzrDirMetaFormat1()
2425
matching.repository_format = self
2429
super(MetaDirRepositoryFormat, self).__init__()
2431
def _create_control_files(self, a_bzrdir):
2432
"""Create the required files and the initial control_files object."""
2433
# FIXME: RBC 20060125 don't peek under the covers
2434
# NB: no need to escape relative paths that are url safe.
2435
repository_transport = a_bzrdir.get_repository_transport(self)
2436
control_files = lockable_files.LockableFiles(repository_transport,
2437
'lock', lockdir.LockDir)
2438
control_files.create_lock()
2439
return control_files
2441
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2442
"""Upload the initial blank content."""
2443
control_files = self._create_control_files(a_bzrdir)
2444
control_files.lock_write()
2445
transport = control_files._transport
2447
utf8_files += [('shared-storage', '')]
2449
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2450
for (filename, content_stream) in files:
2451
transport.put_file(filename, content_stream,
2452
mode=a_bzrdir._get_file_mode())
2453
for (filename, content_bytes) in utf8_files:
2454
transport.put_bytes_non_atomic(filename, content_bytes,
2455
mode=a_bzrdir._get_file_mode())
2457
control_files.unlock()
2459
def network_name(self):
2460
"""Metadir formats have matching disk and network format strings."""
2461
return self.get_format_string()
2464
# Pre-0.8 formats that don't have a disk format string (because they are
2465
# versioned by the matching control directory). We use the control directories
2466
# disk format string as a key for the network_name because they meet the
2467
# constraints (simple string, unique, immmutable).
2468
network_format_registry.register_lazy(
2469
"Bazaar-NG branch, format 5\n",
2470
'bzrlib.repofmt.weaverepo',
2471
'RepositoryFormat5',
2473
network_format_registry.register_lazy(
2474
"Bazaar-NG branch, format 6\n",
2475
'bzrlib.repofmt.weaverepo',
2476
'RepositoryFormat6',
2479
# formats which have no format string are not discoverable or independently
2480
# creatable on disk, so are not registered in format_registry. They're
2481
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2482
# needed, it's constructed directly by the BzrDir. Non-native formats where
2483
# the repository is not separately opened are similar.
2485
format_registry.register_lazy(
2486
'Bazaar-NG Repository format 7',
2487
'bzrlib.repofmt.weaverepo',
2491
format_registry.register_lazy(
2492
'Bazaar-NG Knit Repository Format 1',
2493
'bzrlib.repofmt.knitrepo',
2494
'RepositoryFormatKnit1',
2497
format_registry.register_lazy(
2498
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2499
'bzrlib.repofmt.knitrepo',
2500
'RepositoryFormatKnit3',
2503
format_registry.register_lazy(
2504
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2505
'bzrlib.repofmt.knitrepo',
2506
'RepositoryFormatKnit4',
2509
# Pack-based formats. There is one format for pre-subtrees, and one for
2510
# post-subtrees to allow ease of testing.
2511
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2512
format_registry.register_lazy(
2513
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2514
'bzrlib.repofmt.pack_repo',
2515
'RepositoryFormatKnitPack1',
2517
format_registry.register_lazy(
2518
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2519
'bzrlib.repofmt.pack_repo',
2520
'RepositoryFormatKnitPack3',
2522
format_registry.register_lazy(
2523
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2524
'bzrlib.repofmt.pack_repo',
2525
'RepositoryFormatKnitPack4',
2527
format_registry.register_lazy(
2528
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2529
'bzrlib.repofmt.pack_repo',
2530
'RepositoryFormatKnitPack5',
2532
format_registry.register_lazy(
2533
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2534
'bzrlib.repofmt.pack_repo',
2535
'RepositoryFormatKnitPack5RichRoot',
2537
format_registry.register_lazy(
2538
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2539
'bzrlib.repofmt.pack_repo',
2540
'RepositoryFormatKnitPack5RichRootBroken',
2542
format_registry.register_lazy(
2543
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2544
'bzrlib.repofmt.pack_repo',
2545
'RepositoryFormatKnitPack6',
2547
format_registry.register_lazy(
2548
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2549
'bzrlib.repofmt.pack_repo',
2550
'RepositoryFormatKnitPack6RichRoot',
2553
# Development formats.
2554
# 1.7->1.8 go below here
2555
format_registry.register_lazy(
2556
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2557
'bzrlib.repofmt.pack_repo',
2558
'RepositoryFormatPackDevelopment2',
2560
format_registry.register_lazy(
2561
("Bazaar development format 2 with subtree support "
2562
"(needs bzr.dev from before 1.8)\n"),
2563
'bzrlib.repofmt.pack_repo',
2564
'RepositoryFormatPackDevelopment2Subtree',
2568
class InterRepository(InterObject):
2569
"""This class represents operations taking place between two repositories.
2571
Its instances have methods like copy_content and fetch, and contain
2572
references to the source and target repositories these operations can be
2575
Often we will provide convenience methods on 'repository' which carry out
2576
operations with another repository - they will always forward to
2577
InterRepository.get(other).method_name(parameters).
2580
_walk_to_common_revisions_batch_size = 50
2582
"""The available optimised InterRepository types."""
2584
def __init__(self, source, target):
2585
InterObject.__init__(self, source, target)
2586
# These two attributes may be overridden by e.g. InterOtherToRemote to
2587
# provide a faster implementation.
2588
self.target_get_graph = self.target.get_graph
2589
self.target_get_parent_map = self.target.get_parent_map
2592
def copy_content(self, revision_id=None):
2593
"""Make a complete copy of the content in self into destination.
2595
This is a destructive operation! Do not use it on existing
2598
:param revision_id: Only copy the content needed to construct
2599
revision_id and its parents.
2602
self.target.set_make_working_trees(self.source.make_working_trees())
2603
except NotImplementedError:
2605
self.target.fetch(self.source, revision_id=revision_id)
2608
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2610
"""Fetch the content required to construct revision_id.
2612
The content is copied from self.source to self.target.
2614
:param revision_id: if None all content is copied, if NULL_REVISION no
2616
:param pb: optional progress bar to use for progress reports. If not
2617
provided a default one will be created.
2620
from bzrlib.fetch import RepoFetcher
2621
f = RepoFetcher(to_repository=self.target,
2622
from_repository=self.source,
2623
last_revision=revision_id,
2624
fetch_spec=fetch_spec,
2625
pb=pb, find_ghosts=find_ghosts)
2627
def _walk_to_common_revisions(self, revision_ids):
2628
"""Walk out from revision_ids in source to revisions target has.
2630
:param revision_ids: The start point for the search.
2631
:return: A set of revision ids.
2633
target_graph = self.target_get_graph()
2634
revision_ids = frozenset(revision_ids)
2635
# Fast path for the case where all the revisions are already in the
2637
# (Although this does incur an extra round trip for the
2638
# fairly common case where the target doesn't already have the revision
2640
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2641
return graph.SearchResult(revision_ids, set(), 0, set())
2642
missing_revs = set()
2643
source_graph = self.source.get_graph()
2644
# ensure we don't pay silly lookup costs.
2645
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2646
null_set = frozenset([_mod_revision.NULL_REVISION])
2647
searcher_exhausted = False
2651
# Iterate the searcher until we have enough next_revs
2652
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2654
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2655
next_revs.update(next_revs_part)
2656
ghosts.update(ghosts_part)
2657
except StopIteration:
2658
searcher_exhausted = True
2660
# If there are ghosts in the source graph, and the caller asked for
2661
# them, make sure that they are present in the target.
2662
# We don't care about other ghosts as we can't fetch them and
2663
# haven't been asked to.
2664
ghosts_to_check = set(revision_ids.intersection(ghosts))
2665
revs_to_get = set(next_revs).union(ghosts_to_check)
2667
have_revs = set(target_graph.get_parent_map(revs_to_get))
2668
# we always have NULL_REVISION present.
2669
have_revs = have_revs.union(null_set)
2670
# Check if the target is missing any ghosts we need.
2671
ghosts_to_check.difference_update(have_revs)
2673
# One of the caller's revision_ids is a ghost in both the
2674
# source and the target.
2675
raise errors.NoSuchRevision(
2676
self.source, ghosts_to_check.pop())
2677
missing_revs.update(next_revs - have_revs)
2678
# Because we may have walked past the original stop point, make
2679
# sure everything is stopped
2680
stop_revs = searcher.find_seen_ancestors(have_revs)
2681
searcher.stop_searching_any(stop_revs)
2682
if searcher_exhausted:
2684
return searcher.get_result()
2687
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2688
"""Return the revision ids that source has that target does not.
2690
:param revision_id: only return revision ids included by this
2692
:param find_ghosts: If True find missing revisions in deep history
2693
rather than just finding the surface difference.
2694
:return: A bzrlib.graph.SearchResult.
2696
# stop searching at found target revisions.
2697
if not find_ghosts and revision_id is not None:
2698
return self._walk_to_common_revisions([revision_id])
2699
# generic, possibly worst case, slow code path.
2700
target_ids = set(self.target.all_revision_ids())
2701
if revision_id is not None:
2702
source_ids = self.source.get_ancestry(revision_id)
2703
if source_ids[0] is not None:
2704
raise AssertionError()
2707
source_ids = self.source.all_revision_ids()
2708
result_set = set(source_ids).difference(target_ids)
2709
return self.source.revision_ids_to_search_result(result_set)
2712
def _same_model(source, target):
2713
"""True if source and target have the same data representation.
2715
Note: this is always called on the base class; overriding it in a
2716
subclass will have no effect.
2719
InterRepository._assert_same_model(source, target)
2721
except errors.IncompatibleRepositories, e:
2725
def _assert_same_model(source, target):
2726
"""Raise an exception if two repositories do not use the same model.
2728
if source.supports_rich_root() != target.supports_rich_root():
2729
raise errors.IncompatibleRepositories(source, target,
2730
"different rich-root support")
2731
if source._serializer != target._serializer:
2732
raise errors.IncompatibleRepositories(source, target,
2733
"different serializers")
2736
class InterSameDataRepository(InterRepository):
2737
"""Code for converting between repositories that represent the same data.
2739
Data format and model must match for this to work.
2743
def _get_repo_format_to_test(self):
2744
"""Repository format for testing with.
2746
InterSameData can pull from subtree to subtree and from non-subtree to
2747
non-subtree, so we test this with the richest repository format.
2749
from bzrlib.repofmt import knitrepo
2750
return knitrepo.RepositoryFormatKnit3()
2753
def is_compatible(source, target):
2754
return InterRepository._same_model(source, target)
2757
class InterWeaveRepo(InterSameDataRepository):
2758
"""Optimised code paths between Weave based repositories.
2760
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2761
implemented lazy inter-object optimisation.
2765
def _get_repo_format_to_test(self):
2766
from bzrlib.repofmt import weaverepo
2767
return weaverepo.RepositoryFormat7()
2770
def is_compatible(source, target):
2771
"""Be compatible with known Weave formats.
2773
We don't test for the stores being of specific types because that
2774
could lead to confusing results, and there is no need to be
2777
from bzrlib.repofmt.weaverepo import (
2783
return (isinstance(source._format, (RepositoryFormat5,
2785
RepositoryFormat7)) and
2786
isinstance(target._format, (RepositoryFormat5,
2788
RepositoryFormat7)))
2789
except AttributeError:
2793
def copy_content(self, revision_id=None):
2794
"""See InterRepository.copy_content()."""
2795
# weave specific optimised path:
2797
self.target.set_make_working_trees(self.source.make_working_trees())
2798
except (errors.RepositoryUpgradeRequired, NotImplemented):
2800
# FIXME do not peek!
2801
if self.source._transport.listable():
2802
pb = ui.ui_factory.nested_progress_bar()
2804
self.target.texts.insert_record_stream(
2805
self.source.texts.get_record_stream(
2806
self.source.texts.keys(), 'topological', False))
2807
pb.update('copying inventory', 0, 1)
2808
self.target.inventories.insert_record_stream(
2809
self.source.inventories.get_record_stream(
2810
self.source.inventories.keys(), 'topological', False))
2811
self.target.signatures.insert_record_stream(
2812
self.source.signatures.get_record_stream(
2813
self.source.signatures.keys(),
2815
self.target.revisions.insert_record_stream(
2816
self.source.revisions.get_record_stream(
2817
self.source.revisions.keys(),
2818
'topological', True))
2822
self.target.fetch(self.source, revision_id=revision_id)
2825
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2826
"""See InterRepository.missing_revision_ids()."""
2827
# we want all revisions to satisfy revision_id in source.
2828
# but we don't want to stat every file here and there.
2829
# we want then, all revisions other needs to satisfy revision_id
2830
# checked, but not those that we have locally.
2831
# so the first thing is to get a subset of the revisions to
2832
# satisfy revision_id in source, and then eliminate those that
2833
# we do already have.
2834
# this is slow on high latency connection to self, but as as this
2835
# disk format scales terribly for push anyway due to rewriting
2836
# inventory.weave, this is considered acceptable.
2838
if revision_id is not None:
2839
source_ids = self.source.get_ancestry(revision_id)
2840
if source_ids[0] is not None:
2841
raise AssertionError()
2844
source_ids = self.source._all_possible_ids()
2845
source_ids_set = set(source_ids)
2846
# source_ids is the worst possible case we may need to pull.
2847
# now we want to filter source_ids against what we actually
2848
# have in target, but don't try to check for existence where we know
2849
# we do not have a revision as that would be pointless.
2850
target_ids = set(self.target._all_possible_ids())
2851
possibly_present_revisions = target_ids.intersection(source_ids_set)
2852
actually_present_revisions = set(
2853
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2854
required_revisions = source_ids_set.difference(actually_present_revisions)
2855
if revision_id is not None:
2856
# we used get_ancestry to determine source_ids then we are assured all
2857
# revisions referenced are present as they are installed in topological order.
2858
# and the tip revision was validated by get_ancestry.
2859
result_set = required_revisions
2861
# if we just grabbed the possibly available ids, then
2862
# we only have an estimate of whats available and need to validate
2863
# that against the revision records.
2865
self.source._eliminate_revisions_not_present(required_revisions))
2866
return self.source.revision_ids_to_search_result(result_set)
2869
class InterKnitRepo(InterSameDataRepository):
2870
"""Optimised code paths between Knit based repositories."""
2873
def _get_repo_format_to_test(self):
2874
from bzrlib.repofmt import knitrepo
2875
return knitrepo.RepositoryFormatKnit1()
2878
def is_compatible(source, target):
2879
"""Be compatible with known Knit formats.
2881
We don't test for the stores being of specific types because that
2882
could lead to confusing results, and there is no need to be
2885
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2887
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2888
isinstance(target._format, RepositoryFormatKnit))
2889
except AttributeError:
2891
return are_knits and InterRepository._same_model(source, target)
2894
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2895
"""See InterRepository.missing_revision_ids()."""
2896
if revision_id is not None:
2897
source_ids = self.source.get_ancestry(revision_id)
2898
if source_ids[0] is not None:
2899
raise AssertionError()
2902
source_ids = self.source.all_revision_ids()
2903
source_ids_set = set(source_ids)
2904
# source_ids is the worst possible case we may need to pull.
2905
# now we want to filter source_ids against what we actually
2906
# have in target, but don't try to check for existence where we know
2907
# we do not have a revision as that would be pointless.
2908
target_ids = set(self.target.all_revision_ids())
2909
possibly_present_revisions = target_ids.intersection(source_ids_set)
2910
actually_present_revisions = set(
2911
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2912
required_revisions = source_ids_set.difference(actually_present_revisions)
2913
if revision_id is not None:
2914
# we used get_ancestry to determine source_ids then we are assured all
2915
# revisions referenced are present as they are installed in topological order.
2916
# and the tip revision was validated by get_ancestry.
2917
result_set = required_revisions
2919
# if we just grabbed the possibly available ids, then
2920
# we only have an estimate of whats available and need to validate
2921
# that against the revision records.
2923
self.source._eliminate_revisions_not_present(required_revisions))
2924
return self.source.revision_ids_to_search_result(result_set)
2927
class InterPackRepo(InterSameDataRepository):
2928
"""Optimised code paths between Pack based repositories."""
2931
def _get_repo_format_to_test(self):
2932
from bzrlib.repofmt import pack_repo
2933
return pack_repo.RepositoryFormatKnitPack1()
2936
def is_compatible(source, target):
2937
"""Be compatible with known Pack formats.
2939
We don't test for the stores being of specific types because that
2940
could lead to confusing results, and there is no need to be
2943
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2945
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2946
isinstance(target._format, RepositoryFormatPack))
2947
except AttributeError:
2949
return are_packs and InterRepository._same_model(source, target)
2952
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2954
"""See InterRepository.fetch()."""
2955
if (len(self.source._fallback_repositories) > 0 or
2956
len(self.target._fallback_repositories) > 0):
2957
# The pack layer is not aware of fallback repositories, so when
2958
# fetching from a stacked repository or into a stacked repository
2959
# we use the generic fetch logic which uses the VersionedFiles
2960
# attributes on repository.
2961
from bzrlib.fetch import RepoFetcher
2962
fetcher = RepoFetcher(self.target, self.source, revision_id,
2963
pb, find_ghosts, fetch_spec=fetch_spec)
2964
if fetch_spec is not None:
2965
if len(list(fetch_spec.heads)) != 1:
2966
raise AssertionError(
2967
"InterPackRepo.fetch doesn't support "
2968
"fetching multiple heads yet.")
2969
revision_id = fetch_spec.heads[0]
2971
if revision_id is None:
2973
# everything to do - use pack logic
2974
# to fetch from all packs to one without
2975
# inventory parsing etc, IFF nothing to be copied is in the target.
2977
source_revision_ids = frozenset(self.source.all_revision_ids())
2978
revision_ids = source_revision_ids - \
2979
frozenset(self.target_get_parent_map(source_revision_ids))
2980
revision_keys = [(revid,) for revid in revision_ids]
2981
target_pack_collection = self._get_target_pack_collection()
2982
index = target_pack_collection.revision_index.combined_index
2983
present_revision_ids = set(item[1][0] for item in
2984
index.iter_entries(revision_keys))
2985
revision_ids = set(revision_ids) - present_revision_ids
2986
# implementing the TODO will involve:
2987
# - detecting when all of a pack is selected
2988
# - avoiding as much as possible pre-selection, so the
2989
# more-core routines such as create_pack_from_packs can filter in
2990
# a just-in-time fashion. (though having a HEADS list on a
2991
# repository might make this a lot easier, because we could
2992
# sensibly detect 'new revisions' without doing a full index scan.
2993
elif _mod_revision.is_null(revision_id):
2998
revision_ids = self.search_missing_revision_ids(revision_id,
2999
find_ghosts=find_ghosts).get_keys()
3000
except errors.NoSuchRevision:
3001
raise errors.InstallFailed([revision_id])
3002
if len(revision_ids) == 0:
3004
return self._pack(self.source, self.target, revision_ids)
3006
def _pack(self, source, target, revision_ids):
3007
from bzrlib.repofmt.pack_repo import Packer
3008
target_pack_collection = self._get_target_pack_collection()
3009
packs = source._pack_collection.all_packs()
3010
pack = Packer(target_pack_collection, packs, '.fetch',
3011
revision_ids).pack()
3012
if pack is not None:
3013
target_pack_collection._save_pack_names()
3014
copied_revs = pack.get_revision_count()
3015
# Trigger an autopack. This may duplicate effort as we've just done
3016
# a pack creation, but for now it is simpler to think about as
3017
# 'upload data, then repack if needed'.
3019
return (copied_revs, [])
3023
def _autopack(self):
3024
self.target._pack_collection.autopack()
3026
def _get_target_pack_collection(self):
3027
return self.target._pack_collection
3030
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3031
"""See InterRepository.missing_revision_ids().
3033
:param find_ghosts: Find ghosts throughout the ancestry of
3036
if not find_ghosts and revision_id is not None:
3037
return self._walk_to_common_revisions([revision_id])
3038
elif revision_id is not None:
3039
# Find ghosts: search for revisions pointing from one repository to
3040
# the other, and vice versa, anywhere in the history of revision_id.
3041
graph = self.target_get_graph(other_repository=self.source)
3042
searcher = graph._make_breadth_first_searcher([revision_id])
3046
next_revs, ghosts = searcher.next_with_ghosts()
3047
except StopIteration:
3049
if revision_id in ghosts:
3050
raise errors.NoSuchRevision(self.source, revision_id)
3051
found_ids.update(next_revs)
3052
found_ids.update(ghosts)
3053
found_ids = frozenset(found_ids)
3054
# Double query here: should be able to avoid this by changing the
3055
# graph api further.
3056
result_set = found_ids - frozenset(
3057
self.target_get_parent_map(found_ids))
3059
source_ids = self.source.all_revision_ids()
3060
# source_ids is the worst possible case we may need to pull.
3061
# now we want to filter source_ids against what we actually
3062
# have in target, but don't try to check for existence where we know
3063
# we do not have a revision as that would be pointless.
3064
target_ids = set(self.target.all_revision_ids())
3065
result_set = set(source_ids).difference(target_ids)
3066
return self.source.revision_ids_to_search_result(result_set)
3069
class InterDifferingSerializer(InterKnitRepo):
3072
def _get_repo_format_to_test(self):
3076
def is_compatible(source, target):
3077
"""Be compatible with Knit2 source and Knit3 target"""
3078
if source.supports_rich_root() != target.supports_rich_root():
3080
# Ideally, we'd support fetching if the source had no tree references
3081
# even if it supported them...
3082
if (getattr(source, '_format.supports_tree_reference', False) and
3083
not getattr(target, '_format.supports_tree_reference', False)):
3087
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3088
"""Get the best delta and base for this revision.
3090
:return: (basis_id, delta)
3092
possible_trees = [(parent_id, cache[parent_id])
3093
for parent_id in parent_ids
3094
if parent_id in cache]
3095
if len(possible_trees) == 0:
3096
# There either aren't any parents, or the parents aren't in the
3097
# cache, so just use the last converted tree
3098
possible_trees.append((basis_id, cache[basis_id]))
3100
for basis_id, basis_tree in possible_trees:
3101
delta = tree.inventory._make_delta(basis_tree.inventory)
3102
deltas.append((len(delta), basis_id, delta))
3104
return deltas[0][1:]
3106
def _fetch_batch(self, revision_ids, basis_id, cache):
3107
"""Fetch across a few revisions.
3109
:param revision_ids: The revisions to copy
3110
:param basis_id: The revision_id of a tree that must be in cache, used
3111
as a basis for delta when no other base is available
3112
:param cache: A cache of RevisionTrees that we can use.
3113
:return: The revision_id of the last converted tree. The RevisionTree
3114
for it will be in cache
3116
# Walk though all revisions; get inventory deltas, copy referenced
3117
# texts that delta references, insert the delta, revision and
3121
pending_revisions = []
3122
parent_map = self.source.get_parent_map(revision_ids)
3123
for tree in self.source.revision_trees(revision_ids):
3124
current_revision_id = tree.get_revision_id()
3125
parent_ids = parent_map.get(current_revision_id, ())
3126
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3128
# Find text entries that need to be copied
3129
for old_path, new_path, file_id, entry in delta:
3130
if new_path is not None:
3131
if not (new_path or self.target.supports_rich_root()):
3132
# We don't copy the text for the root node unless the
3133
# target supports_rich_root.
3135
text_keys.add((file_id, entry.revision))
3136
revision = self.source.get_revision(current_revision_id)
3137
pending_deltas.append((basis_id, delta,
3138
current_revision_id, revision.parent_ids))
3139
pending_revisions.append(revision)
3140
cache[current_revision_id] = tree
3141
basis_id = current_revision_id
3143
from_texts = self.source.texts
3144
to_texts = self.target.texts
3145
to_texts.insert_record_stream(from_texts.get_record_stream(
3146
text_keys, self.target._format._fetch_order,
3147
not self.target._format._fetch_uses_deltas))
3149
for delta in pending_deltas:
3150
self.target.add_inventory_by_delta(*delta)
3151
# insert signatures and revisions
3152
for revision in pending_revisions:
3154
signature = self.source.get_signature_text(
3155
revision.revision_id)
3156
self.target.add_signature_text(revision.revision_id,
3158
except errors.NoSuchRevision:
3160
self.target.add_revision(revision.revision_id, revision)
3163
def _fetch_all_revisions(self, revision_ids, pb):
3164
"""Fetch everything for the list of revisions.
3166
:param revision_ids: The list of revisions to fetch. Must be in
3168
:param pb: A ProgressBar
3171
basis_id, basis_tree = self._get_basis(revision_ids[0])
3173
cache = lru_cache.LRUCache(100)
3174
cache[basis_id] = basis_tree
3175
del basis_tree # We don't want to hang on to it here
3176
for offset in range(0, len(revision_ids), batch_size):
3177
self.target.start_write_group()
3179
pb.update('Transferring revisions', offset,
3181
batch = revision_ids[offset:offset+batch_size]
3182
basis_id = self._fetch_batch(batch, basis_id, cache)
3184
self.target.abort_write_group()
3187
self.target.commit_write_group()
3188
pb.update('Transferring revisions', len(revision_ids),
3192
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3194
"""See InterRepository.fetch()."""
3195
if fetch_spec is not None:
3196
raise AssertionError("Not implemented yet...")
3197
revision_ids = self.target.search_missing_revision_ids(self.source,
3198
revision_id, find_ghosts=find_ghosts).get_keys()
3199
if not revision_ids:
3201
revision_ids = tsort.topo_sort(
3202
self.source.get_graph().get_parent_map(revision_ids))
3204
my_pb = ui.ui_factory.nested_progress_bar()
3207
symbol_versioning.warn(
3208
symbol_versioning.deprecated_in((1, 14, 0))
3209
% "pb parameter to fetch()")
3212
self._fetch_all_revisions(revision_ids, pb)
3214
if my_pb is not None:
3216
return len(revision_ids), 0
3218
def _get_basis(self, first_revision_id):
3219
"""Get a revision and tree which exists in the target.
3221
This assumes that first_revision_id is selected for transmission
3222
because all other ancestors are already present. If we can't find an
3223
ancestor we fall back to NULL_REVISION since we know that is safe.
3225
:return: (basis_id, basis_tree)
3227
first_rev = self.source.get_revision(first_revision_id)
3229
basis_id = first_rev.parent_ids[0]
3230
# only valid as a basis if the target has it
3231
self.target.get_revision(basis_id)
3232
# Try to get a basis tree - if its a ghost it will hit the
3233
# NoSuchRevision case.
3234
basis_tree = self.source.revision_tree(basis_id)
3235
except (IndexError, errors.NoSuchRevision):
3236
basis_id = _mod_revision.NULL_REVISION
3237
basis_tree = self.source.revision_tree(basis_id)
3238
return basis_id, basis_tree
3241
class InterOtherToRemote(InterRepository):
3242
"""An InterRepository that simply delegates to the 'real' InterRepository
3243
calculated for (source, target._real_repository).
3246
def __init__(self, source, target):
3247
InterRepository.__init__(self, source, target)
3248
self._real_inter = None
3251
def is_compatible(source, target):
3252
if isinstance(target, remote.RemoteRepository):
3256
def _ensure_real_inter(self):
3257
if self._real_inter is None:
3258
self.target._ensure_real()
3259
real_target = self.target._real_repository
3260
self._real_inter = InterRepository.get(self.source, real_target)
3261
# Make _real_inter use the RemoteRepository for get_parent_map
3262
self._real_inter.target_get_graph = self.target.get_graph
3263
self._real_inter.target_get_parent_map = self.target.get_parent_map
3265
def copy_content(self, revision_id=None):
3266
self._ensure_real_inter()
3267
self._real_inter.copy_content(revision_id=revision_id)
3269
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3271
self._ensure_real_inter()
3272
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3273
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
3276
def _get_repo_format_to_test(self):
3280
class InterRemoteToOther(InterRepository):
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 not isinstance(source, remote.RemoteRepository):
3290
return InterRepository._same_model(source, target)
3292
def _ensure_real_inter(self):
3293
if self._real_inter is None:
3294
self.source._ensure_real()
3295
real_source = self.source._real_repository
3296
self._real_inter = InterRepository.get(real_source, self.target)
3298
def copy_content(self, revision_id=None):
3299
self._ensure_real_inter()
3300
self._real_inter.copy_content(revision_id=revision_id)
3303
def _get_repo_format_to_test(self):
3308
class InterPackToRemotePack(InterPackRepo):
3309
"""A specialisation of InterPackRepo for a target that is a
3312
This will use the get_parent_map RPC rather than plain readvs, and also
3313
uses an RPC for autopacking.
3317
def is_compatible(source, target):
3318
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3319
if isinstance(source._format, RepositoryFormatPack):
3320
if isinstance(target, remote.RemoteRepository):
3321
target._format._ensure_real()
3322
if isinstance(target._format._custom_format,
3323
RepositoryFormatPack):
3324
if InterRepository._same_model(source, target):
3328
def _autopack(self):
3329
self.target.autopack()
3332
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3334
"""See InterRepository.fetch()."""
3335
if self.target._client._medium._is_remote_before((1, 13)):
3336
# The server won't support the insert_stream RPC, so just use
3337
# regular InterPackRepo logic. This avoids a bug that causes many
3338
# round-trips for small append calls.
3339
return InterPackRepo.fetch(self, revision_id=revision_id, pb=pb,
3340
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
3341
# Always fetch using the generic streaming fetch code, to allow
3342
# streaming fetching into remote servers.
3343
from bzrlib.fetch import RepoFetcher
3344
fetcher = RepoFetcher(self.target, self.source, revision_id,
3345
pb, find_ghosts, fetch_spec=fetch_spec)
3347
def _get_target_pack_collection(self):
3348
return self.target._real_repository._pack_collection
3351
def _get_repo_format_to_test(self):
3355
InterRepository.register_optimiser(InterDifferingSerializer)
3356
InterRepository.register_optimiser(InterSameDataRepository)
3357
InterRepository.register_optimiser(InterWeaveRepo)
3358
InterRepository.register_optimiser(InterKnitRepo)
3359
InterRepository.register_optimiser(InterPackRepo)
3360
InterRepository.register_optimiser(InterOtherToRemote)
3361
InterRepository.register_optimiser(InterRemoteToOther)
3362
InterRepository.register_optimiser(InterPackToRemotePack)
3365
class CopyConverter(object):
3366
"""A repository conversion tool which just performs a copy of the content.
3368
This is slow but quite reliable.
3371
def __init__(self, target_format):
3372
"""Create a CopyConverter.
3374
:param target_format: The format the resulting repository should be.
3376
self.target_format = target_format
3378
def convert(self, repo, pb):
3379
"""Perform the conversion of to_convert, giving feedback via pb.
3381
:param to_convert: The disk object to convert.
3382
:param pb: a progress bar to use for progress information.
3387
# this is only useful with metadir layouts - separated repo content.
3388
# trigger an assertion if not such
3389
repo._format.get_format_string()
3390
self.repo_dir = repo.bzrdir
3391
self.step('Moving repository to repository.backup')
3392
self.repo_dir.transport.move('repository', 'repository.backup')
3393
backup_transport = self.repo_dir.transport.clone('repository.backup')
3394
repo._format.check_conversion_target(self.target_format)
3395
self.source_repo = repo._format.open(self.repo_dir,
3397
_override_transport=backup_transport)
3398
self.step('Creating new repository')
3399
converted = self.target_format.initialize(self.repo_dir,
3400
self.source_repo.is_shared())
3401
converted.lock_write()
3403
self.step('Copying content into repository.')
3404
self.source_repo.copy_content_into(converted)
3407
self.step('Deleting old repository content.')
3408
self.repo_dir.transport.delete_tree('repository.backup')
3409
self.pb.note('repository converted')
3411
def step(self, message):
3412
"""Update the pb by a step."""
3414
self.pb.update(message, self.count, self.total)
3426
def _unescaper(match, _map=_unescape_map):
3427
code = match.group(1)
3431
if not code.startswith('#'):
3433
return unichr(int(code[1:])).encode('utf8')
3439
def _unescape_xml(data):
3440
"""Unescape predefined XML entities in a string of data."""
3442
if _unescape_re is None:
3443
_unescape_re = re.compile('\&([^;]*);')
3444
return _unescape_re.sub(_unescaper, data)
3447
class _VersionedFileChecker(object):
3449
def __init__(self, repository, text_key_references=None):
3450
self.repository = repository
3451
self.text_index = self.repository._generate_text_key_index(
3452
text_key_references=text_key_references)
3454
def calculate_file_version_parents(self, text_key):
3455
"""Calculate the correct parents for a file version according to
3458
parent_keys = self.text_index[text_key]
3459
if parent_keys == [_mod_revision.NULL_REVISION]:
3461
return tuple(parent_keys)
3463
def check_file_version_parents(self, texts, progress_bar=None):
3464
"""Check the parents stored in a versioned file are correct.
3466
It also detects file versions that are not referenced by their
3467
corresponding revision's inventory.
3469
:returns: A tuple of (wrong_parents, dangling_file_versions).
3470
wrong_parents is a dict mapping {revision_id: (stored_parents,
3471
correct_parents)} for each revision_id where the stored parents
3472
are not correct. dangling_file_versions is a set of (file_id,
3473
revision_id) tuples for versions that are present in this versioned
3474
file, but not used by the corresponding inventory.
3477
self.file_ids = set([file_id for file_id, _ in
3478
self.text_index.iterkeys()])
3479
# text keys is now grouped by file_id
3480
n_weaves = len(self.file_ids)
3481
files_in_revisions = {}
3482
revisions_of_files = {}
3483
n_versions = len(self.text_index)
3484
progress_bar.update('loading text store', 0, n_versions)
3485
parent_map = self.repository.texts.get_parent_map(self.text_index)
3486
# On unlistable transports this could well be empty/error...
3487
text_keys = self.repository.texts.keys()
3488
unused_keys = frozenset(text_keys) - set(self.text_index)
3489
for num, key in enumerate(self.text_index.iterkeys()):
3490
if progress_bar is not None:
3491
progress_bar.update('checking text graph', num, n_versions)
3492
correct_parents = self.calculate_file_version_parents(key)
3494
knit_parents = parent_map[key]
3495
except errors.RevisionNotPresent:
3498
if correct_parents != knit_parents:
3499
wrong_parents[key] = (knit_parents, correct_parents)
3500
return wrong_parents, unused_keys
3503
def _old_get_graph(repository, revision_id):
3504
"""DO NOT USE. That is all. I'm serious."""
3505
graph = repository.get_graph()
3506
revision_graph = dict(((key, value) for key, value in
3507
graph.iter_ancestry([revision_id]) if value is not None))
3508
return _strip_NULL_ghosts(revision_graph)
3511
def _strip_NULL_ghosts(revision_graph):
3512
"""Also don't use this. more compatibility code for unmigrated clients."""
3513
# Filter ghosts, and null:
3514
if _mod_revision.NULL_REVISION in revision_graph:
3515
del revision_graph[_mod_revision.NULL_REVISION]
3516
for key, parents in revision_graph.items():
3517
revision_graph[key] = tuple(parent for parent in parents if parent
3519
return revision_graph
3522
class StreamSink(object):
3523
"""An object that can insert a stream into a repository.
3525
This interface handles the complexity of reserialising inventories and
3526
revisions from different formats, and allows unidirectional insertion into
3527
stacked repositories without looking for the missing basis parents
3531
def __init__(self, target_repo):
3532
self.target_repo = target_repo
3534
def insert_stream(self, stream, src_format, resume_tokens):
3535
"""Insert a stream's content into the target repository.
3537
:param src_format: a bzr repository format.
3539
:return: a list of resume tokens and an iterable of keys additional
3540
items required before the insertion can be completed.
3542
self.target_repo.lock_write()
3545
self.target_repo.resume_write_group(resume_tokens)
3547
self.target_repo.start_write_group()
3549
# locked_insert_stream performs a commit|suspend.
3550
return self._locked_insert_stream(stream, src_format)
3552
self.target_repo.abort_write_group(suppress_errors=True)
3555
self.target_repo.unlock()
3557
def _locked_insert_stream(self, stream, src_format):
3558
to_serializer = self.target_repo._format._serializer
3559
src_serializer = src_format._serializer
3560
for substream_type, substream in stream:
3561
if substream_type == 'texts':
3562
self.target_repo.texts.insert_record_stream(substream)
3563
elif substream_type == 'inventories':
3564
if src_serializer == to_serializer:
3565
self.target_repo.inventories.insert_record_stream(
3568
self._extract_and_insert_inventories(
3569
substream, src_serializer)
3570
elif substream_type == 'revisions':
3571
# This may fallback to extract-and-insert more often than
3572
# required if the serializers are different only in terms of
3574
if src_serializer == to_serializer:
3575
self.target_repo.revisions.insert_record_stream(
3578
self._extract_and_insert_revisions(substream,
3580
elif substream_type == 'signatures':
3581
self.target_repo.signatures.insert_record_stream(substream)
3583
raise AssertionError('kaboom! %s' % (substream_type,))
3585
missing_keys = set()
3586
for prefix, versioned_file in (
3587
('texts', self.target_repo.texts),
3588
('inventories', self.target_repo.inventories),
3589
('revisions', self.target_repo.revisions),
3590
('signatures', self.target_repo.signatures),
3592
missing_keys.update((prefix,) + key for key in
3593
versioned_file.get_missing_compression_parent_keys())
3594
except NotImplementedError:
3595
# cannot even attempt suspending, and missing would have failed
3596
# during stream insertion.
3597
missing_keys = set()
3600
# suspend the write group and tell the caller what we is
3601
# missing. We know we can suspend or else we would not have
3602
# entered this code path. (All repositories that can handle
3603
# missing keys can handle suspending a write group).
3604
write_group_tokens = self.target_repo.suspend_write_group()
3605
return write_group_tokens, missing_keys
3606
self.target_repo.commit_write_group()
3609
def _extract_and_insert_inventories(self, substream, serializer):
3610
"""Generate a new inventory versionedfile in target, converting data.
3612
The inventory is retrieved from the source, (deserializing it), and
3613
stored in the target (reserializing it in a different format).
3615
for record in substream:
3616
bytes = record.get_bytes_as('fulltext')
3617
revision_id = record.key[0]
3618
inv = serializer.read_inventory_from_string(bytes, revision_id)
3619
parents = [key[0] for key in record.parents]
3620
self.target_repo.add_inventory(revision_id, inv, parents)
3622
def _extract_and_insert_revisions(self, substream, serializer):
3623
for record in substream:
3624
bytes = record.get_bytes_as('fulltext')
3625
revision_id = record.key[0]
3626
rev = serializer.read_revision_from_string(bytes)
3627
if rev.revision_id != revision_id:
3628
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
3629
self.target_repo.add_revision(revision_id, rev)
3632
if self.target_repo._format._fetch_reconcile:
3633
self.target_repo.reconcile()
3636
class StreamSource(object):
3637
"""A source of a stream for fetching between repositories."""
3639
def __init__(self, from_repository, to_format):
3640
"""Create a StreamSource streaming from from_repository."""
3641
self.from_repository = from_repository
3642
self.to_format = to_format
3644
def delta_on_metadata(self):
3645
"""Return True if delta's are permitted on metadata streams.
3647
That is on revisions and signatures.
3649
src_serializer = self.from_repository._format._serializer
3650
target_serializer = self.to_format._serializer
3651
return (self.to_format._fetch_uses_deltas and
3652
src_serializer == target_serializer)
3654
def _fetch_revision_texts(self, revs):
3655
# fetch signatures first and then the revision texts
3656
# may need to be a InterRevisionStore call here.
3657
from_sf = self.from_repository.signatures
3658
# A missing signature is just skipped.
3659
keys = [(rev_id,) for rev_id in revs]
3660
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
3662
self.to_format._fetch_order,
3663
not self.to_format._fetch_uses_deltas))
3664
# If a revision has a delta, this is actually expanded inside the
3665
# insert_record_stream code now, which is an alternate fix for
3667
from_rf = self.from_repository.revisions
3668
revisions = from_rf.get_record_stream(
3670
self.to_format._fetch_order,
3671
not self.delta_on_metadata())
3672
return [('signatures', signatures), ('revisions', revisions)]
3674
def _generate_root_texts(self, revs):
3675
"""This will be called by __fetch between fetching weave texts and
3676
fetching the inventory weave.
3678
Subclasses should override this if they need to generate root texts
3679
after fetching weave texts.
3681
if self._rich_root_upgrade():
3683
return bzrlib.fetch.Inter1and2Helper(
3684
self.from_repository).generate_root_texts(revs)
3688
def get_stream(self, search):
3690
revs = search.get_keys()
3691
graph = self.from_repository.get_graph()
3692
revs = list(graph.iter_topo_order(revs))
3693
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
3695
for knit_kind, file_id, revisions in data_to_fetch:
3696
if knit_kind != phase:
3698
# Make a new progress bar for this phase
3699
if knit_kind == "file":
3700
# Accumulate file texts
3701
text_keys.extend([(file_id, revision) for revision in
3703
elif knit_kind == "inventory":
3704
# Now copy the file texts.
3705
from_texts = self.from_repository.texts
3706
yield ('texts', from_texts.get_record_stream(
3707
text_keys, self.to_format._fetch_order,
3708
not self.to_format._fetch_uses_deltas))
3709
# Cause an error if a text occurs after we have done the
3712
# Before we process the inventory we generate the root
3713
# texts (if necessary) so that the inventories references
3715
for _ in self._generate_root_texts(revs):
3717
# NB: This currently reopens the inventory weave in source;
3718
# using a single stream interface instead would avoid this.
3719
from_weave = self.from_repository.inventories
3720
# we fetch only the referenced inventories because we do not
3721
# know for unselected inventories whether all their required
3722
# texts are present in the other repository - it could be
3724
yield ('inventories', from_weave.get_record_stream(
3725
[(rev_id,) for rev_id in revs],
3726
self.inventory_fetch_order(),
3727
not self.delta_on_metadata()))
3728
elif knit_kind == "signatures":
3729
# Nothing to do here; this will be taken care of when
3730
# _fetch_revision_texts happens.
3732
elif knit_kind == "revisions":
3733
for record in self._fetch_revision_texts(revs):
3736
raise AssertionError("Unknown knit kind %r" % knit_kind)
3738
def get_stream_for_missing_keys(self, missing_keys):
3739
# missing keys can only occur when we are byte copying and not
3740
# translating (because translation means we don't send
3741
# unreconstructable deltas ever).
3743
keys['texts'] = set()
3744
keys['revisions'] = set()
3745
keys['inventories'] = set()
3746
keys['signatures'] = set()
3747
for key in missing_keys:
3748
keys[key[0]].add(key[1:])
3749
if len(keys['revisions']):
3750
# If we allowed copying revisions at this point, we could end up
3751
# copying a revision without copying its required texts: a
3752
# violation of the requirements for repository integrity.
3753
raise AssertionError(
3754
'cannot copy revisions to fill in missing deltas %s' % (
3755
keys['revisions'],))
3756
for substream_kind, keys in keys.iteritems():
3757
vf = getattr(self.from_repository, substream_kind)
3758
# Ask for full texts always so that we don't need more round trips
3759
# after this stream.
3760
stream = vf.get_record_stream(keys,
3761
self.to_format._fetch_order, True)
3762
yield substream_kind, stream
3764
def inventory_fetch_order(self):
3765
if self._rich_root_upgrade():
3766
return 'topological'
3768
return self.to_format._fetch_order
3770
def _rich_root_upgrade(self):
3771
return (not self.from_repository._format.rich_root_data and
3772
self.to_format.rich_root_data)
added
removed
bzrlib/repository.py
