1
# Copyright (C) 2005, 2006, 2007, 2008 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from bzrlib.lazy_import import lazy_import
18
lazy_import(globals(), """
38
revision as _mod_revision,
44
from bzrlib.bundle import serializer
45
from bzrlib.revisiontree import RevisionTree
46
from bzrlib.store.versioned import VersionedFileStore
47
from bzrlib.testament import Testament
50
from bzrlib.decorators import needs_read_lock, needs_write_lock
51
from bzrlib.inter import InterObject
52
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
53
from bzrlib import registry
54
from bzrlib.symbol_versioning import (
60
from bzrlib.trace import (
61
log_exception_quietly, note, mutter, mutter_callsite, warning)
64
# Old formats display a warning, but only once
65
_deprecation_warning_done = False
68
class CommitBuilder(object):
69
"""Provides an interface to build up a commit.
71
This allows describing a tree to be committed without needing to
72
know the internals of the format of the repository.
75
# all clients should supply tree roots.
76
record_root_entry = True
77
# the default CommitBuilder does not manage trees whose root is versioned.
78
_versioned_root = False
80
def __init__(self, repository, parents, config, timestamp=None,
81
timezone=None, committer=None, revprops=None,
83
"""Initiate a CommitBuilder.
85
:param repository: Repository to commit to.
86
:param parents: Revision ids of the parents of the new revision.
87
:param config: Configuration to use.
88
:param timestamp: Optional timestamp recorded for commit.
89
:param timezone: Optional timezone for timestamp.
90
:param committer: Optional committer to set for commit.
91
:param revprops: Optional dictionary of revision properties.
92
:param revision_id: Optional revision id.
97
self._committer = self._config.username()
99
self._committer = committer
101
self.new_inventory = Inventory(None)
102
self._new_revision_id = revision_id
103
self.parents = parents
104
self.repository = repository
107
if revprops is not None:
108
self._validate_revprops(revprops)
109
self._revprops.update(revprops)
111
if timestamp is None:
112
timestamp = time.time()
113
# Restrict resolution to 1ms
114
self._timestamp = round(timestamp, 3)
117
self._timezone = osutils.local_time_offset()
119
self._timezone = int(timezone)
121
self._generate_revision_if_needed()
122
self.__heads = graph.HeadsCache(repository.get_graph()).heads
123
self._basis_delta = []
124
# API compatibility, older code that used CommitBuilder did not call
125
# .record_delete(), which means the delta that is computed would not be
126
# valid. Callers that will call record_delete() should call
127
# .will_record_deletes() to indicate that.
128
self._recording_deletes = False
130
def _validate_unicode_text(self, text, context):
131
"""Verify things like commit messages don't have bogus characters."""
133
raise ValueError('Invalid value for %s: %r' % (context, text))
135
def _validate_revprops(self, revprops):
136
for key, value in revprops.iteritems():
137
# We know that the XML serializers do not round trip '\r'
138
# correctly, so refuse to accept them
139
if not isinstance(value, basestring):
140
raise ValueError('revision property (%s) is not a valid'
141
' (unicode) string: %r' % (key, value))
142
self._validate_unicode_text(value,
143
'revision property (%s)' % (key,))
145
def commit(self, message):
146
"""Make the actual commit.
148
:return: The revision id of the recorded revision.
150
self._validate_unicode_text(message, 'commit message')
151
rev = _mod_revision.Revision(
152
timestamp=self._timestamp,
153
timezone=self._timezone,
154
committer=self._committer,
156
inventory_sha1=self.inv_sha1,
157
revision_id=self._new_revision_id,
158
properties=self._revprops)
159
rev.parent_ids = self.parents
160
self.repository.add_revision(self._new_revision_id, rev,
161
self.new_inventory, self._config)
162
self.repository.commit_write_group()
163
return self._new_revision_id
166
"""Abort the commit that is being built.
168
self.repository.abort_write_group()
170
def revision_tree(self):
171
"""Return the tree that was just committed.
173
After calling commit() this can be called to get a RevisionTree
174
representing the newly committed tree. This is preferred to
175
calling Repository.revision_tree() because that may require
176
deserializing the inventory, while we already have a copy in
179
return RevisionTree(self.repository, self.new_inventory,
180
self._new_revision_id)
182
def finish_inventory(self):
183
"""Tell the builder that the inventory is finished."""
184
if self.new_inventory.root is None:
185
raise AssertionError('Root entry should be supplied to'
186
' record_entry_contents, as of bzr 0.10.')
187
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
188
self.new_inventory.revision_id = self._new_revision_id
189
self.inv_sha1 = self.repository.add_inventory(
190
self._new_revision_id,
195
def _gen_revision_id(self):
196
"""Return new revision-id."""
197
return generate_ids.gen_revision_id(self._config.username(),
200
def _generate_revision_if_needed(self):
201
"""Create a revision id if None was supplied.
203
If the repository can not support user-specified revision ids
204
they should override this function and raise CannotSetRevisionId
205
if _new_revision_id is not None.
207
:raises: CannotSetRevisionId
209
if self._new_revision_id is None:
210
self._new_revision_id = self._gen_revision_id()
211
self.random_revid = True
213
self.random_revid = False
215
def _heads(self, file_id, revision_ids):
216
"""Calculate the graph heads for revision_ids in the graph of file_id.
218
This can use either a per-file graph or a global revision graph as we
219
have an identity relationship between the two graphs.
221
return self.__heads(revision_ids)
223
def _check_root(self, ie, parent_invs, tree):
224
"""Helper for record_entry_contents.
226
:param ie: An entry being added.
227
:param parent_invs: The inventories of the parent revisions of the
229
:param tree: The tree that is being committed.
231
# In this revision format, root entries have no knit or weave When
232
# serializing out to disk and back in root.revision is always
234
ie.revision = self._new_revision_id
236
def _get_delta(self, ie, basis_inv, path):
237
"""Get a delta against the basis inventory for ie."""
238
if ie.file_id not in basis_inv:
240
result = (None, path, ie.file_id, ie)
241
self._basis_delta.append(result)
243
elif ie != basis_inv[ie.file_id]:
245
# TODO: avoid tis id2path call.
246
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
247
self._basis_delta.append(result)
253
def get_basis_delta(self):
254
"""Return the complete inventory delta versus the basis inventory.
256
This has been built up with the calls to record_delete and
257
record_entry_contents. The client must have already called
258
will_record_deletes() to indicate that they will be generating a
261
:return: An inventory delta, suitable for use with apply_delta, or
262
Repository.add_inventory_by_delta, etc.
264
if not self._recording_deletes:
265
raise AssertionError("recording deletes not activated.")
266
return self._basis_delta
268
def record_delete(self, path, file_id):
269
"""Record that a delete occured against a basis tree.
271
This is an optional API - when used it adds items to the basis_delta
272
being accumulated by the commit builder. It cannot be called unless the
273
method will_record_deletes() has been called to inform the builder that
274
a delta is being supplied.
276
:param path: The path of the thing deleted.
277
:param file_id: The file id that was deleted.
279
if not self._recording_deletes:
280
raise AssertionError("recording deletes not activated.")
281
delta = (path, None, file_id, None)
282
self._basis_delta.append(delta)
285
def will_record_deletes(self):
286
"""Tell the commit builder that deletes are being notified.
288
This enables the accumulation of an inventory delta; for the resulting
289
commit to be valid, deletes against the basis MUST be recorded via
290
builder.record_delete().
292
self._recording_deletes = True
294
def record_entry_contents(self, ie, parent_invs, path, tree,
296
"""Record the content of ie from tree into the commit if needed.
298
Side effect: sets ie.revision when unchanged
300
:param ie: An inventory entry present in the commit.
301
:param parent_invs: The inventories of the parent revisions of the
303
:param path: The path the entry is at in the tree.
304
:param tree: The tree which contains this entry and should be used to
306
:param content_summary: Summary data from the tree about the paths
307
content - stat, length, exec, sha/link target. This is only
308
accessed when the entry has a revision of None - that is when it is
309
a candidate to commit.
310
:return: A tuple (change_delta, version_recorded, fs_hash).
311
change_delta is an inventory_delta change for this entry against
312
the basis tree of the commit, or None if no change occured against
314
version_recorded is True if a new version of the entry has been
315
recorded. For instance, committing a merge where a file was only
316
changed on the other side will return (delta, False).
317
fs_hash is either None, or the hash details for the path (currently
318
a tuple of the contents sha1 and the statvalue returned by
319
tree.get_file_with_stat()).
321
if self.new_inventory.root is None:
322
if ie.parent_id is not None:
323
raise errors.RootMissing()
324
self._check_root(ie, parent_invs, tree)
325
if ie.revision is None:
326
kind = content_summary[0]
328
# ie is carried over from a prior commit
330
# XXX: repository specific check for nested tree support goes here - if
331
# the repo doesn't want nested trees we skip it ?
332
if (kind == 'tree-reference' and
333
not self.repository._format.supports_tree_reference):
334
# mismatch between commit builder logic and repository:
335
# this needs the entry creation pushed down into the builder.
336
raise NotImplementedError('Missing repository subtree support.')
337
self.new_inventory.add(ie)
339
# TODO: slow, take it out of the inner loop.
341
basis_inv = parent_invs[0]
343
basis_inv = Inventory(root_id=None)
345
# ie.revision is always None if the InventoryEntry is considered
346
# for committing. We may record the previous parents revision if the
347
# content is actually unchanged against a sole head.
348
if ie.revision is not None:
349
if not self._versioned_root and path == '':
350
# repositories that do not version the root set the root's
351
# revision to the new commit even when no change occurs (more
352
# specifically, they do not record a revision on the root; and
353
# the rev id is assigned to the root during deserialisation -
354
# this masks when a change may have occurred against the basis.
355
# To match this we always issue a delta, because the revision
356
# of the root will always be changing.
357
if ie.file_id in basis_inv:
358
delta = (basis_inv.id2path(ie.file_id), path,
362
delta = (None, path, ie.file_id, ie)
363
self._basis_delta.append(delta)
364
return delta, False, None
366
# we don't need to commit this, because the caller already
367
# determined that an existing revision of this file is
368
# appropriate. If its not being considered for committing then
369
# it and all its parents to the root must be unaltered so
370
# no-change against the basis.
371
if ie.revision == self._new_revision_id:
372
raise AssertionError("Impossible situation, a skipped "
373
"inventory entry (%r) claims to be modified in this "
374
"commit (%r).", (ie, self._new_revision_id))
375
return None, False, None
376
# XXX: Friction: parent_candidates should return a list not a dict
377
# so that we don't have to walk the inventories again.
378
parent_candiate_entries = ie.parent_candidates(parent_invs)
379
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
381
for inv in parent_invs:
382
if ie.file_id in inv:
383
old_rev = inv[ie.file_id].revision
384
if old_rev in head_set:
385
heads.append(inv[ie.file_id].revision)
386
head_set.remove(inv[ie.file_id].revision)
389
# now we check to see if we need to write a new record to the
391
# We write a new entry unless there is one head to the ancestors, and
392
# the kind-derived content is unchanged.
394
# Cheapest check first: no ancestors, or more the one head in the
395
# ancestors, we write a new node.
399
# There is a single head, look it up for comparison
400
parent_entry = parent_candiate_entries[heads[0]]
401
# if the non-content specific data has changed, we'll be writing a
403
if (parent_entry.parent_id != ie.parent_id or
404
parent_entry.name != ie.name):
406
# now we need to do content specific checks:
408
# if the kind changed the content obviously has
409
if kind != parent_entry.kind:
411
# Stat cache fingerprint feedback for the caller - None as we usually
412
# don't generate one.
415
if content_summary[2] is None:
416
raise ValueError("Files must not have executable = None")
418
if (# if the file length changed we have to store:
419
parent_entry.text_size != content_summary[1] or
420
# if the exec bit has changed we have to store:
421
parent_entry.executable != content_summary[2]):
423
elif parent_entry.text_sha1 == content_summary[3]:
424
# all meta and content is unchanged (using a hash cache
425
# hit to check the sha)
426
ie.revision = parent_entry.revision
427
ie.text_size = parent_entry.text_size
428
ie.text_sha1 = parent_entry.text_sha1
429
ie.executable = parent_entry.executable
430
return self._get_delta(ie, basis_inv, path), False, None
432
# Either there is only a hash change(no hash cache entry,
433
# or same size content change), or there is no change on
435
# Provide the parent's hash to the store layer, so that the
436
# content is unchanged we will not store a new node.
437
nostore_sha = parent_entry.text_sha1
439
# We want to record a new node regardless of the presence or
440
# absence of a content change in the file.
442
ie.executable = content_summary[2]
443
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
445
lines = file_obj.readlines()
449
ie.text_sha1, ie.text_size = self._add_text_to_weave(
450
ie.file_id, lines, heads, nostore_sha)
451
# Let the caller know we generated a stat fingerprint.
452
fingerprint = (ie.text_sha1, stat_value)
453
except errors.ExistingContent:
454
# Turns out that the file content was unchanged, and we were
455
# only going to store a new node if it was changed. Carry over
457
ie.revision = parent_entry.revision
458
ie.text_size = parent_entry.text_size
459
ie.text_sha1 = parent_entry.text_sha1
460
ie.executable = parent_entry.executable
461
return self._get_delta(ie, basis_inv, path), False, None
462
elif kind == 'directory':
464
# all data is meta here, nothing specific to directory, so
466
ie.revision = parent_entry.revision
467
return self._get_delta(ie, basis_inv, path), False, None
469
self._add_text_to_weave(ie.file_id, lines, heads, None)
470
elif kind == 'symlink':
471
current_link_target = content_summary[3]
473
# symlink target is not generic metadata, check if it has
475
if current_link_target != parent_entry.symlink_target:
478
# unchanged, carry over.
479
ie.revision = parent_entry.revision
480
ie.symlink_target = parent_entry.symlink_target
481
return self._get_delta(ie, basis_inv, path), False, None
482
ie.symlink_target = current_link_target
484
self._add_text_to_weave(ie.file_id, lines, heads, None)
485
elif kind == 'tree-reference':
487
if content_summary[3] != parent_entry.reference_revision:
490
# unchanged, carry over.
491
ie.reference_revision = parent_entry.reference_revision
492
ie.revision = parent_entry.revision
493
return self._get_delta(ie, basis_inv, path), False, None
494
ie.reference_revision = content_summary[3]
496
self._add_text_to_weave(ie.file_id, lines, heads, None)
498
raise NotImplementedError('unknown kind')
499
ie.revision = self._new_revision_id
500
return self._get_delta(ie, basis_inv, path), True, fingerprint
502
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
503
# Note: as we read the content directly from the tree, we know its not
504
# been turned into unicode or badly split - but a broken tree
505
# implementation could give us bad output from readlines() so this is
506
# not a guarantee of safety. What would be better is always checking
507
# the content during test suite execution. RBC 20070912
508
parent_keys = tuple((file_id, parent) for parent in parents)
509
return self.repository.texts.add_lines(
510
(file_id, self._new_revision_id), parent_keys, new_lines,
511
nostore_sha=nostore_sha, random_id=self.random_revid,
512
check_content=False)[0:2]
515
class RootCommitBuilder(CommitBuilder):
516
"""This commitbuilder actually records the root id"""
518
# the root entry gets versioned properly by this builder.
519
_versioned_root = True
521
def _check_root(self, ie, parent_invs, tree):
522
"""Helper for record_entry_contents.
524
:param ie: An entry being added.
525
:param parent_invs: The inventories of the parent revisions of the
527
:param tree: The tree that is being committed.
531
######################################################################
534
class Repository(object):
535
"""Repository holding history for one or more branches.
537
The repository holds and retrieves historical information including
538
revisions and file history. It's normally accessed only by the Branch,
539
which views a particular line of development through that history.
541
The Repository builds on top of some byte storage facilies (the revisions,
542
signatures, inventories and texts attributes) and a Transport, which
543
respectively provide byte storage and a means to access the (possibly
546
The byte storage facilities are addressed via tuples, which we refer to
547
as 'keys' throughout the code base. Revision_keys, inventory_keys and
548
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
549
(file_id, revision_id). We use this interface because it allows low
550
friction with the underlying code that implements disk indices, network
551
encoding and other parts of bzrlib.
553
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
554
the serialised revisions for the repository. This can be used to obtain
555
revision graph information or to access raw serialised revisions.
556
The result of trying to insert data into the repository via this store
557
is undefined: it should be considered read-only except for implementors
559
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
560
the serialised signatures for the repository. This can be used to
561
obtain access to raw serialised signatures. The result of trying to
562
insert data into the repository via this store is undefined: it should
563
be considered read-only except for implementors of repositories.
564
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
565
the serialised inventories for the repository. This can be used to
566
obtain unserialised inventories. The result of trying to insert data
567
into the repository via this store is undefined: it should be
568
considered read-only except for implementors of repositories.
569
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
570
texts of files and directories for the repository. This can be used to
571
obtain file texts or file graphs. Note that Repository.iter_file_bytes
572
is usually a better interface for accessing file texts.
573
The result of trying to insert data into the repository via this store
574
is undefined: it should be considered read-only except for implementors
576
:ivar _transport: Transport for file access to repository, typically
577
pointing to .bzr/repository.
580
# What class to use for a CommitBuilder. Often its simpler to change this
581
# in a Repository class subclass rather than to override
582
# get_commit_builder.
583
_commit_builder_class = CommitBuilder
584
# The search regex used by xml based repositories to determine what things
585
# where changed in a single commit.
586
_file_ids_altered_regex = lazy_regex.lazy_compile(
587
r'file_id="(?P<file_id>[^"]+)"'
588
r'.* revision="(?P<revision_id>[^"]+)"'
591
def abort_write_group(self, suppress_errors=False):
592
"""Commit the contents accrued within the current write group.
594
:param suppress_errors: if true, abort_write_group will catch and log
595
unexpected errors that happen during the abort, rather than
596
allowing them to propagate. Defaults to False.
598
:seealso: start_write_group.
600
if self._write_group is not self.get_transaction():
601
# has an unlock or relock occured ?
602
raise errors.BzrError('mismatched lock context and write group.')
604
self._abort_write_group()
605
except Exception, exc:
606
self._write_group = None
607
if not suppress_errors:
609
mutter('abort_write_group failed')
610
log_exception_quietly()
611
note('bzr: ERROR (ignored): %s', exc)
612
self._write_group = None
614
def _abort_write_group(self):
615
"""Template method for per-repository write group cleanup.
617
This is called during abort before the write group is considered to be
618
finished and should cleanup any internal state accrued during the write
619
group. There is no requirement that data handed to the repository be
620
*not* made available - this is not a rollback - but neither should any
621
attempt be made to ensure that data added is fully commited. Abort is
622
invoked when an error has occured so futher disk or network operations
623
may not be possible or may error and if possible should not be
627
def add_fallback_repository(self, repository):
628
"""Add a repository to use for looking up data not held locally.
630
:param repository: A repository.
632
if not self._format.supports_external_lookups:
633
raise errors.UnstackableRepositoryFormat(self._format, self.base)
634
self._check_fallback_repository(repository)
635
self._fallback_repositories.append(repository)
636
self.texts.add_fallback_versioned_files(repository.texts)
637
self.inventories.add_fallback_versioned_files(repository.inventories)
638
self.revisions.add_fallback_versioned_files(repository.revisions)
639
self.signatures.add_fallback_versioned_files(repository.signatures)
641
def _check_fallback_repository(self, repository):
642
"""Check that this repository can fallback to repository safely.
644
Raise an error if not.
646
:param repository: A repository to fallback to.
648
return InterRepository._assert_same_model(self, repository)
650
def add_inventory(self, revision_id, inv, parents):
651
"""Add the inventory inv to the repository as revision_id.
653
:param parents: The revision ids of the parents that revision_id
654
is known to have and are in the repository already.
656
:returns: The validator(which is a sha1 digest, though what is sha'd is
657
repository format specific) of the serialized inventory.
659
if not self.is_in_write_group():
660
raise AssertionError("%r not in write group" % (self,))
661
_mod_revision.check_not_reserved_id(revision_id)
662
if not (inv.revision_id is None or inv.revision_id == revision_id):
663
raise AssertionError(
664
"Mismatch between inventory revision"
665
" id and insertion revid (%r, %r)"
666
% (inv.revision_id, revision_id))
668
raise AssertionError()
669
inv_lines = self._serialise_inventory_to_lines(inv)
670
return self._inventory_add_lines(revision_id, parents,
671
inv_lines, check_content=False)
673
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
675
"""Add a new inventory expressed as a delta against another revision.
677
:param basis_revision_id: The inventory id the delta was created
678
against. (This does not have to be a direct parent.)
679
:param delta: The inventory delta (see Inventory.apply_delta for
681
:param new_revision_id: The revision id that the inventory is being
683
:param parents: The revision ids of the parents that revision_id is
684
known to have and are in the repository already. These are supplied
685
for repositories that depend on the inventory graph for revision
686
graph access, as well as for those that pun ancestry with delta
689
:returns: (validator, new_inv)
690
The validator(which is a sha1 digest, though what is sha'd is
691
repository format specific) of the serialized inventory, and the
694
if not self.is_in_write_group():
695
raise AssertionError("%r not in write group" % (self,))
696
_mod_revision.check_not_reserved_id(new_revision_id)
697
basis_tree = self.revision_tree(basis_revision_id)
698
basis_tree.lock_read()
700
# Note that this mutates the inventory of basis_tree, which not all
701
# inventory implementations may support: A better idiom would be to
702
# return a new inventory, but as there is no revision tree cache in
703
# repository this is safe for now - RBC 20081013
704
basis_inv = basis_tree.inventory
705
basis_inv.apply_delta(delta)
706
basis_inv.revision_id = new_revision_id
707
return (self.add_inventory(new_revision_id, basis_inv, parents),
712
def _inventory_add_lines(self, revision_id, parents, lines,
714
"""Store lines in inv_vf and return the sha1 of the inventory."""
715
parents = [(parent,) for parent in parents]
716
return self.inventories.add_lines((revision_id,), parents, lines,
717
check_content=check_content)[0]
719
def add_revision(self, revision_id, rev, inv=None, config=None):
720
"""Add rev to the revision store as revision_id.
722
:param revision_id: the revision id to use.
723
:param rev: The revision object.
724
:param inv: The inventory for the revision. if None, it will be looked
725
up in the inventory storer
726
:param config: If None no digital signature will be created.
727
If supplied its signature_needed method will be used
728
to determine if a signature should be made.
730
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
732
_mod_revision.check_not_reserved_id(revision_id)
733
if config is not None and config.signature_needed():
735
inv = self.get_inventory(revision_id)
736
plaintext = Testament(rev, inv).as_short_text()
737
self.store_revision_signature(
738
gpg.GPGStrategy(config), plaintext, revision_id)
739
# check inventory present
740
if not self.inventories.get_parent_map([(revision_id,)]):
742
raise errors.WeaveRevisionNotPresent(revision_id,
745
# yes, this is not suitable for adding with ghosts.
746
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
750
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
751
self._add_revision(rev)
753
def _add_revision(self, revision):
754
text = self._serializer.write_revision_to_string(revision)
755
key = (revision.revision_id,)
756
parents = tuple((parent,) for parent in revision.parent_ids)
757
self.revisions.add_lines(key, parents, osutils.split_lines(text))
759
def all_revision_ids(self):
760
"""Returns a list of all the revision ids in the repository.
762
This is conceptually deprecated because code should generally work on
763
the graph reachable from a particular revision, and ignore any other
764
revisions that might be present. There is no direct replacement
767
if 'evil' in debug.debug_flags:
768
mutter_callsite(2, "all_revision_ids is linear with history.")
769
return self._all_revision_ids()
771
def _all_revision_ids(self):
772
"""Returns a list of all the revision ids in the repository.
774
These are in as much topological order as the underlying store can
777
raise NotImplementedError(self._all_revision_ids)
779
def break_lock(self):
780
"""Break a lock if one is present from another instance.
782
Uses the ui factory to ask for confirmation if the lock may be from
785
self.control_files.break_lock()
788
def _eliminate_revisions_not_present(self, revision_ids):
789
"""Check every revision id in revision_ids to see if we have it.
791
Returns a set of the present revisions.
794
graph = self.get_graph()
795
parent_map = graph.get_parent_map(revision_ids)
796
# The old API returned a list, should this actually be a set?
797
return parent_map.keys()
800
def create(a_bzrdir):
801
"""Construct the current default format repository in a_bzrdir."""
802
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
804
def __init__(self, _format, a_bzrdir, control_files):
805
"""instantiate a Repository.
807
:param _format: The format of the repository on disk.
808
:param a_bzrdir: The BzrDir of the repository.
810
In the future we will have a single api for all stores for
811
getting file texts, inventories and revisions, then
812
this construct will accept instances of those things.
814
super(Repository, self).__init__()
815
self._format = _format
816
# the following are part of the public API for Repository:
817
self.bzrdir = a_bzrdir
818
self.control_files = control_files
819
self._transport = control_files._transport
820
self.base = self._transport.base
822
self._reconcile_does_inventory_gc = True
823
self._reconcile_fixes_text_parents = False
824
self._reconcile_backsup_inventory = True
825
# not right yet - should be more semantically clear ?
827
# TODO: make sure to construct the right store classes, etc, depending
828
# on whether escaping is required.
829
self._warn_if_deprecated()
830
self._write_group = None
831
# Additional places to query for data.
832
self._fallback_repositories = []
833
# An InventoryEntry cache, used during deserialization
834
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
837
return '%s(%r)' % (self.__class__.__name__,
840
def has_same_location(self, other):
841
"""Returns a boolean indicating if this repository is at the same
842
location as another repository.
844
This might return False even when two repository objects are accessing
845
the same physical repository via different URLs.
847
if self.__class__ is not other.__class__:
849
return (self._transport.base == other._transport.base)
851
def is_in_write_group(self):
852
"""Return True if there is an open write group.
854
:seealso: start_write_group.
856
return self._write_group is not None
859
return self.control_files.is_locked()
861
def is_write_locked(self):
862
"""Return True if this object is write locked."""
863
return self.is_locked() and self.control_files._lock_mode == 'w'
865
def lock_write(self, token=None):
866
"""Lock this repository for writing.
868
This causes caching within the repository obejct to start accumlating
869
data during reads, and allows a 'write_group' to be obtained. Write
870
groups must be used for actual data insertion.
872
:param token: if this is already locked, then lock_write will fail
873
unless the token matches the existing lock.
874
:returns: a token if this instance supports tokens, otherwise None.
875
:raises TokenLockingNotSupported: when a token is given but this
876
instance doesn't support using token locks.
877
:raises MismatchedToken: if the specified token doesn't match the token
878
of the existing lock.
879
:seealso: start_write_group.
881
A token should be passed in if you know that you have locked the object
882
some other way, and need to synchronise this object's state with that
885
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
887
result = self.control_files.lock_write(token=token)
888
for repo in self._fallback_repositories:
889
# Writes don't affect fallback repos
895
self.control_files.lock_read()
896
for repo in self._fallback_repositories:
900
def get_physical_lock_status(self):
901
return self.control_files.get_physical_lock_status()
903
def leave_lock_in_place(self):
904
"""Tell this repository not to release the physical lock when this
907
If lock_write doesn't return a token, then this method is not supported.
909
self.control_files.leave_in_place()
911
def dont_leave_lock_in_place(self):
912
"""Tell this repository to release the physical lock when this
913
object is unlocked, even if it didn't originally acquire it.
915
If lock_write doesn't return a token, then this method is not supported.
917
self.control_files.dont_leave_in_place()
920
def gather_stats(self, revid=None, committers=None):
921
"""Gather statistics from a revision id.
923
:param revid: The revision id to gather statistics from, if None, then
924
no revision specific statistics are gathered.
925
:param committers: Optional parameter controlling whether to grab
926
a count of committers from the revision specific statistics.
927
:return: A dictionary of statistics. Currently this contains:
928
committers: The number of committers if requested.
929
firstrev: A tuple with timestamp, timezone for the penultimate left
930
most ancestor of revid, if revid is not the NULL_REVISION.
931
latestrev: A tuple with timestamp, timezone for revid, if revid is
932
not the NULL_REVISION.
933
revisions: The total revision count in the repository.
934
size: An estimate disk size of the repository in bytes.
937
if revid and committers:
938
result['committers'] = 0
939
if revid and revid != _mod_revision.NULL_REVISION:
941
all_committers = set()
942
revisions = self.get_ancestry(revid)
943
# pop the leading None
945
first_revision = None
947
# ignore the revisions in the middle - just grab first and last
948
revisions = revisions[0], revisions[-1]
949
for revision in self.get_revisions(revisions):
950
if not first_revision:
951
first_revision = revision
953
all_committers.add(revision.committer)
954
last_revision = revision
956
result['committers'] = len(all_committers)
957
result['firstrev'] = (first_revision.timestamp,
958
first_revision.timezone)
959
result['latestrev'] = (last_revision.timestamp,
960
last_revision.timezone)
962
# now gather global repository information
963
# XXX: This is available for many repos regardless of listability.
964
if self.bzrdir.root_transport.listable():
965
# XXX: do we want to __define len__() ?
966
# Maybe the versionedfiles object should provide a different
967
# method to get the number of keys.
968
result['revisions'] = len(self.revisions.keys())
972
def find_branches(self, using=False):
973
"""Find branches underneath this repository.
975
This will include branches inside other branches.
977
:param using: If True, list only branches using this repository.
979
if using and not self.is_shared():
981
return [self.bzrdir.open_branch()]
982
except errors.NotBranchError:
984
class Evaluator(object):
987
self.first_call = True
989
def __call__(self, bzrdir):
990
# On the first call, the parameter is always the bzrdir
991
# containing the current repo.
992
if not self.first_call:
994
repository = bzrdir.open_repository()
995
except errors.NoRepositoryPresent:
998
return False, (None, repository)
999
self.first_call = False
1001
value = (bzrdir.open_branch(), None)
1002
except errors.NotBranchError:
1003
value = (None, None)
1007
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1008
self.bzrdir.root_transport, evaluate=Evaluator()):
1009
if branch is not None:
1010
branches.append(branch)
1011
if not using and repository is not None:
1012
branches.extend(repository.find_branches())
1016
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1017
"""Return the revision ids that other has that this does not.
1019
These are returned in topological order.
1021
revision_id: only return revision ids included by revision_id.
1023
return InterRepository.get(other, self).search_missing_revision_ids(
1024
revision_id, find_ghosts)
1026
@deprecated_method(one_two)
1028
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1029
"""Return the revision ids that other has that this does not.
1031
These are returned in topological order.
1033
revision_id: only return revision ids included by revision_id.
1035
keys = self.search_missing_revision_ids(
1036
other, revision_id, find_ghosts).get_keys()
1039
parents = other.get_graph().get_parent_map(keys)
1042
return tsort.topo_sort(parents)
1046
"""Open the repository rooted at base.
1048
For instance, if the repository is at URL/.bzr/repository,
1049
Repository.open(URL) -> a Repository instance.
1051
control = bzrdir.BzrDir.open(base)
1052
return control.open_repository()
1054
def copy_content_into(self, destination, revision_id=None):
1055
"""Make a complete copy of the content in self into destination.
1057
This is a destructive operation! Do not use it on existing
1060
return InterRepository.get(self, destination).copy_content(revision_id)
1062
def commit_write_group(self):
1063
"""Commit the contents accrued within the current write group.
1065
:seealso: start_write_group.
1067
if self._write_group is not self.get_transaction():
1068
# has an unlock or relock occured ?
1069
raise errors.BzrError('mismatched lock context %r and '
1071
(self.get_transaction(), self._write_group))
1072
self._commit_write_group()
1073
self._write_group = None
1075
def _commit_write_group(self):
1076
"""Template method for per-repository write group cleanup.
1078
This is called before the write group is considered to be
1079
finished and should ensure that all data handed to the repository
1080
for writing during the write group is safely committed (to the
1081
extent possible considering file system caching etc).
1084
def suspend_write_group(self):
1085
raise errors.UnsuspendableWriteGroup(self)
1087
def resume_write_group(self, tokens):
1088
if not self.is_write_locked():
1089
raise errors.NotWriteLocked(self)
1090
if self._write_group:
1091
raise errors.BzrError('already in a write group')
1092
self._resume_write_group(tokens)
1093
# so we can detect unlock/relock - the write group is now entered.
1094
self._write_group = self.get_transaction()
1096
def _resume_write_group(self, tokens):
1097
raise errors.UnsuspendableWriteGroup(self)
1099
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
1101
"""Fetch the content required to construct revision_id from source.
1103
If revision_id is None and fetch_spec is None, then all content is
1106
: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
# fast path same-url fetch operations
1121
if self.has_same_location(source) and fetch_spec is None:
1122
# check that last_revision is in 'from' and then return a
1124
if (revision_id is not None and
1125
not _mod_revision.is_null(revision_id)):
1126
self.get_revision(revision_id)
1128
# if there is no specific appropriate InterRepository, this will get
1129
# the InterRepository base class, which raises an
1130
# IncompatibleRepositories when asked to fetch.
1131
inter = InterRepository.get(source, self)
1132
return inter.fetch(revision_id=revision_id, pb=pb,
1133
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
1135
def create_bundle(self, target, base, fileobj, format=None):
1136
return serializer.write_bundle(self, target, base, fileobj, format)
1138
def get_commit_builder(self, branch, parents, config, timestamp=None,
1139
timezone=None, committer=None, revprops=None,
1141
"""Obtain a CommitBuilder for this repository.
1143
:param branch: Branch to commit to.
1144
:param parents: Revision ids of the parents of the new revision.
1145
:param config: Configuration to use.
1146
:param timestamp: Optional timestamp recorded for commit.
1147
:param timezone: Optional timezone for timestamp.
1148
:param committer: Optional committer to set for commit.
1149
:param revprops: Optional dictionary of revision properties.
1150
:param revision_id: Optional revision id.
1152
result = self._commit_builder_class(self, parents, config,
1153
timestamp, timezone, committer, revprops, revision_id)
1154
self.start_write_group()
1158
if (self.control_files._lock_count == 1 and
1159
self.control_files._lock_mode == 'w'):
1160
if self._write_group is not None:
1161
self.abort_write_group()
1162
self.control_files.unlock()
1163
raise errors.BzrError(
1164
'Must end write groups before releasing write locks.')
1165
self.control_files.unlock()
1166
if self.control_files._lock_count == 0:
1167
self._inventory_entry_cache.clear()
1168
for repo in self._fallback_repositories:
1172
def clone(self, a_bzrdir, revision_id=None):
1173
"""Clone this repository into a_bzrdir using the current format.
1175
Currently no check is made that the format of this repository and
1176
the bzrdir format are compatible. FIXME RBC 20060201.
1178
:return: The newly created destination repository.
1180
# TODO: deprecate after 0.16; cloning this with all its settings is
1181
# probably not very useful -- mbp 20070423
1182
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1183
self.copy_content_into(dest_repo, revision_id)
1186
def start_write_group(self):
1187
"""Start a write group in the repository.
1189
Write groups are used by repositories which do not have a 1:1 mapping
1190
between file ids and backend store to manage the insertion of data from
1191
both fetch and commit operations.
1193
A write lock is required around the start_write_group/commit_write_group
1194
for the support of lock-requiring repository formats.
1196
One can only insert data into a repository inside a write group.
1200
if not self.is_write_locked():
1201
raise errors.NotWriteLocked(self)
1202
if self._write_group:
1203
raise errors.BzrError('already in a write group')
1204
self._start_write_group()
1205
# so we can detect unlock/relock - the write group is now entered.
1206
self._write_group = self.get_transaction()
1208
def _start_write_group(self):
1209
"""Template method for per-repository write group startup.
1211
This is called before the write group is considered to be
1216
def sprout(self, to_bzrdir, revision_id=None):
1217
"""Create a descendent repository for new development.
1219
Unlike clone, this does not copy the settings of the repository.
1221
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1222
dest_repo.fetch(self, revision_id=revision_id)
1225
def _create_sprouting_repo(self, a_bzrdir, shared):
1226
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1227
# use target default format.
1228
dest_repo = a_bzrdir.create_repository()
1230
# Most control formats need the repository to be specifically
1231
# created, but on some old all-in-one formats it's not needed
1233
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1234
except errors.UninitializableFormat:
1235
dest_repo = a_bzrdir.open_repository()
1238
def _get_sink(self):
1239
"""Return a sink for streaming into this repository."""
1240
return StreamSink(self)
1242
def _get_source(self, to_format):
1243
"""Return a source for streaming from this repository."""
1244
return StreamSource(self, to_format)
1247
def has_revision(self, revision_id):
1248
"""True if this repository has a copy of the revision."""
1249
return revision_id in self.has_revisions((revision_id,))
1252
def has_revisions(self, revision_ids):
1253
"""Probe to find out the presence of multiple revisions.
1255
:param revision_ids: An iterable of revision_ids.
1256
:return: A set of the revision_ids that were present.
1258
parent_map = self.revisions.get_parent_map(
1259
[(rev_id,) for rev_id in revision_ids])
1261
if _mod_revision.NULL_REVISION in revision_ids:
1262
result.add(_mod_revision.NULL_REVISION)
1263
result.update([key[0] for key in parent_map])
1267
def get_revision(self, revision_id):
1268
"""Return the Revision object for a named revision."""
1269
return self.get_revisions([revision_id])[0]
1272
def get_revision_reconcile(self, revision_id):
1273
"""'reconcile' helper routine that allows access to a revision always.
1275
This variant of get_revision does not cross check the weave graph
1276
against the revision one as get_revision does: but it should only
1277
be used by reconcile, or reconcile-alike commands that are correcting
1278
or testing the revision graph.
1280
return self._get_revisions([revision_id])[0]
1283
def get_revisions(self, revision_ids):
1284
"""Get many revisions at once."""
1285
return self._get_revisions(revision_ids)
1288
def _get_revisions(self, revision_ids):
1289
"""Core work logic to get many revisions without sanity checks."""
1290
for rev_id in revision_ids:
1291
if not rev_id or not isinstance(rev_id, basestring):
1292
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1293
keys = [(key,) for key in revision_ids]
1294
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1296
for record in stream:
1297
if record.storage_kind == 'absent':
1298
raise errors.NoSuchRevision(self, record.key[0])
1299
text = record.get_bytes_as('fulltext')
1300
rev = self._serializer.read_revision_from_string(text)
1301
revs[record.key[0]] = rev
1302
return [revs[revid] for revid in revision_ids]
1305
def get_revision_xml(self, revision_id):
1306
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1307
# would have already do it.
1308
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1309
rev = self.get_revision(revision_id)
1310
rev_tmp = cStringIO.StringIO()
1311
# the current serializer..
1312
self._serializer.write_revision(rev, rev_tmp)
1314
return rev_tmp.getvalue()
1316
def get_deltas_for_revisions(self, revisions):
1317
"""Produce a generator of revision deltas.
1319
Note that the input is a sequence of REVISIONS, not revision_ids.
1320
Trees will be held in memory until the generator exits.
1321
Each delta is relative to the revision's lefthand predecessor.
1323
required_trees = set()
1324
for revision in revisions:
1325
required_trees.add(revision.revision_id)
1326
required_trees.update(revision.parent_ids[:1])
1327
trees = dict((t.get_revision_id(), t) for
1328
t in self.revision_trees(required_trees))
1329
for revision in revisions:
1330
if not revision.parent_ids:
1331
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1333
old_tree = trees[revision.parent_ids[0]]
1334
yield trees[revision.revision_id].changes_from(old_tree)
1337
def get_revision_delta(self, revision_id):
1338
"""Return the delta for one revision.
1340
The delta is relative to the left-hand predecessor of the
1343
r = self.get_revision(revision_id)
1344
return list(self.get_deltas_for_revisions([r]))[0]
1347
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1348
signature = gpg_strategy.sign(plaintext)
1349
self.add_signature_text(revision_id, signature)
1352
def add_signature_text(self, revision_id, signature):
1353
self.signatures.add_lines((revision_id,), (),
1354
osutils.split_lines(signature))
1356
def find_text_key_references(self):
1357
"""Find the text key references within the repository.
1359
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1360
revision_ids. Each altered file-ids has the exact revision_ids that
1361
altered it listed explicitly.
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.
1845
def revision_tree(self, revision_id):
1846
"""Return Tree for a revision on this branch.
1848
`revision_id` may be NULL_REVISION for the empty tree revision.
1850
revision_id = _mod_revision.ensure_null(revision_id)
1851
# TODO: refactor this to use an existing revision object
1852
# so we don't need to read it in twice.
1853
if revision_id == _mod_revision.NULL_REVISION:
1854
return RevisionTree(self, Inventory(root_id=None),
1855
_mod_revision.NULL_REVISION)
1857
inv = self.get_revision_inventory(revision_id)
1858
return RevisionTree(self, inv, revision_id)
1860
def revision_trees(self, revision_ids):
1861
"""Return Tree for a revision on this branch.
1863
`revision_id` may not be None or 'null:'"""
1864
inventories = self.iter_inventories(revision_ids)
1865
for inv in inventories:
1866
yield RevisionTree(self, inv, inv.revision_id)
1869
def get_ancestry(self, revision_id, topo_sorted=True):
1870
"""Return a list of revision-ids integrated by a revision.
1872
The first element of the list is always None, indicating the origin
1873
revision. This might change when we have history horizons, or
1874
perhaps we should have a new API.
1876
This is topologically sorted.
1878
if _mod_revision.is_null(revision_id):
1880
if not self.has_revision(revision_id):
1881
raise errors.NoSuchRevision(self, revision_id)
1882
graph = self.get_graph()
1884
search = graph._make_breadth_first_searcher([revision_id])
1887
found, ghosts = search.next_with_ghosts()
1888
except StopIteration:
1891
if _mod_revision.NULL_REVISION in keys:
1892
keys.remove(_mod_revision.NULL_REVISION)
1894
parent_map = graph.get_parent_map(keys)
1895
keys = tsort.topo_sort(parent_map)
1896
return [None] + list(keys)
1899
"""Compress the data within the repository.
1901
This operation only makes sense for some repository types. For other
1902
types it should be a no-op that just returns.
1904
This stub method does not require a lock, but subclasses should use
1905
@needs_write_lock as this is a long running call its reasonable to
1906
implicitly lock for the user.
1910
@deprecated_method(one_six)
1911
def print_file(self, file, revision_id):
1912
"""Print `file` to stdout.
1914
FIXME RBC 20060125 as John Meinel points out this is a bad api
1915
- it writes to stdout, it assumes that that is valid etc. Fix
1916
by creating a new more flexible convenience function.
1918
tree = self.revision_tree(revision_id)
1919
# use inventory as it was in that revision
1920
file_id = tree.inventory.path2id(file)
1922
# TODO: jam 20060427 Write a test for this code path
1923
# it had a bug in it, and was raising the wrong
1925
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1926
tree.print_file(file_id)
1928
def get_transaction(self):
1929
return self.control_files.get_transaction()
1931
@deprecated_method(one_one)
1932
def get_parents(self, revision_ids):
1933
"""See StackedParentsProvider.get_parents"""
1934
parent_map = self.get_parent_map(revision_ids)
1935
return [parent_map.get(r, None) for r in revision_ids]
1937
def get_parent_map(self, revision_ids):
1938
"""See graph._StackedParentsProvider.get_parent_map"""
1939
# revisions index works in keys; this just works in revisions
1940
# therefore wrap and unwrap
1943
for revision_id in revision_ids:
1944
if revision_id == _mod_revision.NULL_REVISION:
1945
result[revision_id] = ()
1946
elif revision_id is None:
1947
raise ValueError('get_parent_map(None) is not valid')
1949
query_keys.append((revision_id ,))
1950
for ((revision_id,), parent_keys) in \
1951
self.revisions.get_parent_map(query_keys).iteritems():
1953
result[revision_id] = tuple(parent_revid
1954
for (parent_revid,) in parent_keys)
1956
result[revision_id] = (_mod_revision.NULL_REVISION,)
1959
def _make_parents_provider(self):
1962
def get_graph(self, other_repository=None):
1963
"""Return the graph walker for this repository format"""
1964
parents_provider = self._make_parents_provider()
1965
if (other_repository is not None and
1966
not self.has_same_location(other_repository)):
1967
parents_provider = graph._StackedParentsProvider(
1968
[parents_provider, other_repository._make_parents_provider()])
1969
return graph.Graph(parents_provider)
1971
def _get_versioned_file_checker(self):
1972
"""Return an object suitable for checking versioned files."""
1973
return _VersionedFileChecker(self)
1975
def revision_ids_to_search_result(self, result_set):
1976
"""Convert a set of revision ids to a graph SearchResult."""
1977
result_parents = set()
1978
for parents in self.get_graph().get_parent_map(
1979
result_set).itervalues():
1980
result_parents.update(parents)
1981
included_keys = result_set.intersection(result_parents)
1982
start_keys = result_set.difference(included_keys)
1983
exclude_keys = result_parents.difference(result_set)
1984
result = graph.SearchResult(start_keys, exclude_keys,
1985
len(result_set), result_set)
1989
def set_make_working_trees(self, new_value):
1990
"""Set the policy flag for making working trees when creating branches.
1992
This only applies to branches that use this repository.
1994
The default is 'True'.
1995
:param new_value: True to restore the default, False to disable making
1998
raise NotImplementedError(self.set_make_working_trees)
2000
def make_working_trees(self):
2001
"""Returns the policy for making working trees on new branches."""
2002
raise NotImplementedError(self.make_working_trees)
2005
def sign_revision(self, revision_id, gpg_strategy):
2006
plaintext = Testament.from_revision(self, revision_id).as_short_text()
2007
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2010
def has_signature_for_revision_id(self, revision_id):
2011
"""Query for a revision signature for revision_id in the repository."""
2012
if not self.has_revision(revision_id):
2013
raise errors.NoSuchRevision(self, revision_id)
2014
sig_present = (1 == len(
2015
self.signatures.get_parent_map([(revision_id,)])))
2019
def get_signature_text(self, revision_id):
2020
"""Return the text for a signature."""
2021
stream = self.signatures.get_record_stream([(revision_id,)],
2023
record = stream.next()
2024
if record.storage_kind == 'absent':
2025
raise errors.NoSuchRevision(self, revision_id)
2026
return record.get_bytes_as('fulltext')
2029
def check(self, revision_ids=None):
2030
"""Check consistency of all history of given revision_ids.
2032
Different repository implementations should override _check().
2034
:param revision_ids: A non-empty list of revision_ids whose ancestry
2035
will be checked. Typically the last revision_id of a branch.
2037
return self._check(revision_ids)
2039
def _check(self, revision_ids):
2040
result = check.Check(self)
2044
def _warn_if_deprecated(self):
2045
global _deprecation_warning_done
2046
if _deprecation_warning_done:
2048
_deprecation_warning_done = True
2049
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2050
% (self._format, self.bzrdir.transport.base))
2052
def supports_rich_root(self):
2053
return self._format.rich_root_data
2055
def _check_ascii_revisionid(self, revision_id, method):
2056
"""Private helper for ascii-only repositories."""
2057
# weave repositories refuse to store revisionids that are non-ascii.
2058
if revision_id is not None:
2059
# weaves require ascii revision ids.
2060
if isinstance(revision_id, unicode):
2062
revision_id.encode('ascii')
2063
except UnicodeEncodeError:
2064
raise errors.NonAsciiRevisionId(method, self)
2067
revision_id.decode('ascii')
2068
except UnicodeDecodeError:
2069
raise errors.NonAsciiRevisionId(method, self)
2071
def revision_graph_can_have_wrong_parents(self):
2072
"""Is it possible for this repository to have a revision graph with
2075
If True, then this repository must also implement
2076
_find_inconsistent_revision_parents so that check and reconcile can
2077
check for inconsistencies before proceeding with other checks that may
2078
depend on the revision index being consistent.
2080
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2083
# remove these delegates a while after bzr 0.15
2084
def __make_delegated(name, from_module):
2085
def _deprecated_repository_forwarder():
2086
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2087
% (name, from_module),
2090
m = __import__(from_module, globals(), locals(), [name])
2092
return getattr(m, name)
2093
except AttributeError:
2094
raise AttributeError('module %s has no name %s'
2096
globals()[name] = _deprecated_repository_forwarder
2099
'AllInOneRepository',
2100
'WeaveMetaDirRepository',
2101
'PreSplitOutRepositoryFormat',
2102
'RepositoryFormat4',
2103
'RepositoryFormat5',
2104
'RepositoryFormat6',
2105
'RepositoryFormat7',
2107
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2111
'RepositoryFormatKnit',
2112
'RepositoryFormatKnit1',
2114
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2117
def install_revision(repository, rev, revision_tree):
2118
"""Install all revision data into a repository."""
2119
install_revisions(repository, [(rev, revision_tree, None)])
2122
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2123
"""Install all revision data into a repository.
2125
Accepts an iterable of revision, tree, signature tuples. The signature
2128
repository.start_write_group()
2130
for n, (revision, revision_tree, signature) in enumerate(iterable):
2131
_install_revision(repository, revision, revision_tree, signature)
2133
pb.update('Transferring revisions', n + 1, num_revisions)
2135
repository.abort_write_group()
2138
repository.commit_write_group()
2141
def _install_revision(repository, rev, revision_tree, signature):
2142
"""Install all revision data into a repository."""
2143
present_parents = []
2145
for p_id in rev.parent_ids:
2146
if repository.has_revision(p_id):
2147
present_parents.append(p_id)
2148
parent_trees[p_id] = repository.revision_tree(p_id)
2150
parent_trees[p_id] = repository.revision_tree(
2151
_mod_revision.NULL_REVISION)
2153
inv = revision_tree.inventory
2154
entries = inv.iter_entries()
2155
# backwards compatibility hack: skip the root id.
2156
if not repository.supports_rich_root():
2157
path, root = entries.next()
2158
if root.revision != rev.revision_id:
2159
raise errors.IncompatibleRevision(repr(repository))
2161
for path, ie in entries:
2162
text_keys[(ie.file_id, ie.revision)] = ie
2163
text_parent_map = repository.texts.get_parent_map(text_keys)
2164
missing_texts = set(text_keys) - set(text_parent_map)
2165
# Add the texts that are not already present
2166
for text_key in missing_texts:
2167
ie = text_keys[text_key]
2169
# FIXME: TODO: The following loop overlaps/duplicates that done by
2170
# commit to determine parents. There is a latent/real bug here where
2171
# the parents inserted are not those commit would do - in particular
2172
# they are not filtered by heads(). RBC, AB
2173
for revision, tree in parent_trees.iteritems():
2174
if ie.file_id not in tree:
2176
parent_id = tree.inventory[ie.file_id].revision
2177
if parent_id in text_parents:
2179
text_parents.append((ie.file_id, parent_id))
2180
lines = revision_tree.get_file(ie.file_id).readlines()
2181
repository.texts.add_lines(text_key, text_parents, lines)
2183
# install the inventory
2184
repository.add_inventory(rev.revision_id, inv, present_parents)
2185
except errors.RevisionAlreadyPresent:
2187
if signature is not None:
2188
repository.add_signature_text(rev.revision_id, signature)
2189
repository.add_revision(rev.revision_id, rev, inv)
2192
class MetaDirRepository(Repository):
2193
"""Repositories in the new meta-dir layout.
2195
:ivar _transport: Transport for access to repository control files,
2196
typically pointing to .bzr/repository.
2199
def __init__(self, _format, a_bzrdir, control_files):
2200
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2201
self._transport = control_files._transport
2203
def is_shared(self):
2204
"""Return True if this repository is flagged as a shared repository."""
2205
return self._transport.has('shared-storage')
2208
def set_make_working_trees(self, new_value):
2209
"""Set the policy flag for making working trees when creating branches.
2211
This only applies to branches that use this repository.
2213
The default is 'True'.
2214
:param new_value: True to restore the default, False to disable making
2219
self._transport.delete('no-working-trees')
2220
except errors.NoSuchFile:
2223
self._transport.put_bytes('no-working-trees', '',
2224
mode=self.bzrdir._get_file_mode())
2226
def make_working_trees(self):
2227
"""Returns the policy for making working trees on new branches."""
2228
return not self._transport.has('no-working-trees')
2231
class MetaDirVersionedFileRepository(MetaDirRepository):
2232
"""Repositories in a meta-dir, that work via versioned file objects."""
2234
def __init__(self, _format, a_bzrdir, control_files):
2235
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2239
network_format_registry = registry.FormatRegistry()
2240
"""Registry of formats indexed by their network name.
2242
The network name for a repository format is an identifier that can be used when
2243
referring to formats with smart server operations. See
2244
RepositoryFormat.network_name() for more detail.
2248
format_registry = registry.FormatRegistry(network_format_registry)
2249
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2251
This can contain either format instances themselves, or classes/factories that
2252
can be called to obtain one.
2256
#####################################################################
2257
# Repository Formats
2259
class RepositoryFormat(object):
2260
"""A repository format.
2262
Formats provide four things:
2263
* An initialization routine to construct repository data on disk.
2264
* a optional format string which is used when the BzrDir supports
2266
* an open routine which returns a Repository instance.
2267
* A network name for referring to the format in smart server RPC
2270
There is one and only one Format subclass for each on-disk format. But
2271
there can be one Repository subclass that is used for several different
2272
formats. The _format attribute on a Repository instance can be used to
2273
determine the disk format.
2275
Formats are placed in a registry by their format string for reference
2276
during opening. These should be subclasses of RepositoryFormat for
2279
Once a format is deprecated, just deprecate the initialize and open
2280
methods on the format class. Do not deprecate the object, as the
2281
object may be created even when a repository instnace hasn't been
2284
Common instance attributes:
2285
_matchingbzrdir - the bzrdir format that the repository format was
2286
originally written to work with. This can be used if manually
2287
constructing a bzrdir and repository, or more commonly for test suite
2291
# Set to True or False in derived classes. True indicates that the format
2292
# supports ghosts gracefully.
2293
supports_ghosts = None
2294
# Can this repository be given external locations to lookup additional
2295
# data. Set to True or False in derived classes.
2296
supports_external_lookups = None
2297
# What order should fetch operations request streams in?
2298
# The default is unordered as that is the cheapest for an origin to
2300
_fetch_order = 'unordered'
2301
# Does this repository format use deltas that can be fetched as-deltas ?
2302
# (E.g. knits, where the knit deltas can be transplanted intact.
2303
# We default to False, which will ensure that enough data to get
2304
# a full text out of any fetch stream will be grabbed.
2305
_fetch_uses_deltas = False
2306
# Should fetch trigger a reconcile after the fetch? Only needed for
2307
# some repository formats that can suffer internal inconsistencies.
2308
_fetch_reconcile = False
2311
return "<%s>" % self.__class__.__name__
2313
def __eq__(self, other):
2314
# format objects are generally stateless
2315
return isinstance(other, self.__class__)
2317
def __ne__(self, other):
2318
return not self == other
2321
def find_format(klass, a_bzrdir):
2322
"""Return the format for the repository object in a_bzrdir.
2324
This is used by bzr native formats that have a "format" file in
2325
the repository. Other methods may be used by different types of
2329
transport = a_bzrdir.get_repository_transport(None)
2330
format_string = transport.get("format").read()
2331
return format_registry.get(format_string)
2332
except errors.NoSuchFile:
2333
raise errors.NoRepositoryPresent(a_bzrdir)
2335
raise errors.UnknownFormatError(format=format_string,
2339
def register_format(klass, format):
2340
format_registry.register(format.get_format_string(), format)
2343
def unregister_format(klass, format):
2344
format_registry.remove(format.get_format_string())
2347
def get_default_format(klass):
2348
"""Return the current default format."""
2349
from bzrlib import bzrdir
2350
return bzrdir.format_registry.make_bzrdir('default').repository_format
2352
def get_format_string(self):
2353
"""Return the ASCII format string that identifies this format.
2355
Note that in pre format ?? repositories the format string is
2356
not permitted nor written to disk.
2358
raise NotImplementedError(self.get_format_string)
2360
def get_format_description(self):
2361
"""Return the short description for this format."""
2362
raise NotImplementedError(self.get_format_description)
2364
# TODO: this shouldn't be in the base class, it's specific to things that
2365
# use weaves or knits -- mbp 20070207
2366
def _get_versioned_file_store(self,
2371
versionedfile_class=None,
2372
versionedfile_kwargs={},
2374
if versionedfile_class is None:
2375
versionedfile_class = self._versionedfile_class
2376
weave_transport = control_files._transport.clone(name)
2377
dir_mode = control_files._dir_mode
2378
file_mode = control_files._file_mode
2379
return VersionedFileStore(weave_transport, prefixed=prefixed,
2381
file_mode=file_mode,
2382
versionedfile_class=versionedfile_class,
2383
versionedfile_kwargs=versionedfile_kwargs,
2386
def initialize(self, a_bzrdir, shared=False):
2387
"""Initialize a repository of this format in a_bzrdir.
2389
:param a_bzrdir: The bzrdir to put the new repository in it.
2390
:param shared: The repository should be initialized as a sharable one.
2391
:returns: The new repository object.
2393
This may raise UninitializableFormat if shared repository are not
2394
compatible the a_bzrdir.
2396
raise NotImplementedError(self.initialize)
2398
def is_supported(self):
2399
"""Is this format supported?
2401
Supported formats must be initializable and openable.
2402
Unsupported formats may not support initialization or committing or
2403
some other features depending on the reason for not being supported.
2407
def network_name(self):
2408
"""A simple byte string uniquely identifying this format for RPC calls.
2410
MetaDir repository formats use their disk format string to identify the
2411
repository over the wire. All in one formats such as bzr < 0.8, and
2412
foreign formats like svn/git and hg should use some marker which is
2413
unique and immutable.
2415
raise NotImplementedError(self.network_name)
2417
def check_conversion_target(self, target_format):
2418
raise NotImplementedError(self.check_conversion_target)
2420
def open(self, a_bzrdir, _found=False):
2421
"""Return an instance of this format for the bzrdir a_bzrdir.
2423
_found is a private parameter, do not use it.
2425
raise NotImplementedError(self.open)
2428
class MetaDirRepositoryFormat(RepositoryFormat):
2429
"""Common base class for the new repositories using the metadir layout."""
2431
rich_root_data = False
2432
supports_tree_reference = False
2433
supports_external_lookups = False
2436
def _matchingbzrdir(self):
2437
matching = bzrdir.BzrDirMetaFormat1()
2438
matching.repository_format = self
2442
super(MetaDirRepositoryFormat, self).__init__()
2444
def _create_control_files(self, a_bzrdir):
2445
"""Create the required files and the initial control_files object."""
2446
# FIXME: RBC 20060125 don't peek under the covers
2447
# NB: no need to escape relative paths that are url safe.
2448
repository_transport = a_bzrdir.get_repository_transport(self)
2449
control_files = lockable_files.LockableFiles(repository_transport,
2450
'lock', lockdir.LockDir)
2451
control_files.create_lock()
2452
return control_files
2454
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2455
"""Upload the initial blank content."""
2456
control_files = self._create_control_files(a_bzrdir)
2457
control_files.lock_write()
2458
transport = control_files._transport
2460
utf8_files += [('shared-storage', '')]
2462
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2463
for (filename, content_stream) in files:
2464
transport.put_file(filename, content_stream,
2465
mode=a_bzrdir._get_file_mode())
2466
for (filename, content_bytes) in utf8_files:
2467
transport.put_bytes_non_atomic(filename, content_bytes,
2468
mode=a_bzrdir._get_file_mode())
2470
control_files.unlock()
2472
def network_name(self):
2473
"""Metadir formats have matching disk and network format strings."""
2474
return self.get_format_string()
2477
# Pre-0.8 formats that don't have a disk format string (because they are
2478
# versioned by the matching control directory). We use the control directories
2479
# disk format string as a key for the network_name because they meet the
2480
# constraints (simple string, unique, immmutable).
2481
network_format_registry.register_lazy(
2482
"Bazaar-NG branch, format 5\n",
2483
'bzrlib.repofmt.weaverepo',
2484
'RepositoryFormat5',
2486
network_format_registry.register_lazy(
2487
"Bazaar-NG branch, format 6\n",
2488
'bzrlib.repofmt.weaverepo',
2489
'RepositoryFormat6',
2492
# formats which have no format string are not discoverable or independently
2493
# creatable on disk, so are not registered in format_registry. They're
2494
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2495
# needed, it's constructed directly by the BzrDir. Non-native formats where
2496
# the repository is not separately opened are similar.
2498
format_registry.register_lazy(
2499
'Bazaar-NG Repository format 7',
2500
'bzrlib.repofmt.weaverepo',
2504
format_registry.register_lazy(
2505
'Bazaar-NG Knit Repository Format 1',
2506
'bzrlib.repofmt.knitrepo',
2507
'RepositoryFormatKnit1',
2510
format_registry.register_lazy(
2511
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2512
'bzrlib.repofmt.knitrepo',
2513
'RepositoryFormatKnit3',
2516
format_registry.register_lazy(
2517
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2518
'bzrlib.repofmt.knitrepo',
2519
'RepositoryFormatKnit4',
2522
# Pack-based formats. There is one format for pre-subtrees, and one for
2523
# post-subtrees to allow ease of testing.
2524
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2525
format_registry.register_lazy(
2526
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2527
'bzrlib.repofmt.pack_repo',
2528
'RepositoryFormatKnitPack1',
2530
format_registry.register_lazy(
2531
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2532
'bzrlib.repofmt.pack_repo',
2533
'RepositoryFormatKnitPack3',
2535
format_registry.register_lazy(
2536
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2537
'bzrlib.repofmt.pack_repo',
2538
'RepositoryFormatKnitPack4',
2540
format_registry.register_lazy(
2541
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2542
'bzrlib.repofmt.pack_repo',
2543
'RepositoryFormatKnitPack5',
2545
format_registry.register_lazy(
2546
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2547
'bzrlib.repofmt.pack_repo',
2548
'RepositoryFormatKnitPack5RichRoot',
2550
format_registry.register_lazy(
2551
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2552
'bzrlib.repofmt.pack_repo',
2553
'RepositoryFormatKnitPack5RichRootBroken',
2555
format_registry.register_lazy(
2556
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2557
'bzrlib.repofmt.pack_repo',
2558
'RepositoryFormatKnitPack6',
2560
format_registry.register_lazy(
2561
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2562
'bzrlib.repofmt.pack_repo',
2563
'RepositoryFormatKnitPack6RichRoot',
2566
# Development formats.
2567
# 1.7->1.8 go below here
2568
format_registry.register_lazy(
2569
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2570
'bzrlib.repofmt.pack_repo',
2571
'RepositoryFormatPackDevelopment2',
2573
format_registry.register_lazy(
2574
("Bazaar development format 2 with subtree support "
2575
"(needs bzr.dev from before 1.8)\n"),
2576
'bzrlib.repofmt.pack_repo',
2577
'RepositoryFormatPackDevelopment2Subtree',
2581
class InterRepository(InterObject):
2582
"""This class represents operations taking place between two repositories.
2584
Its instances have methods like copy_content and fetch, and contain
2585
references to the source and target repositories these operations can be
2588
Often we will provide convenience methods on 'repository' which carry out
2589
operations with another repository - they will always forward to
2590
InterRepository.get(other).method_name(parameters).
2593
_walk_to_common_revisions_batch_size = 1
2595
"""The available optimised InterRepository types."""
2597
def __init__(self, source, target):
2598
InterObject.__init__(self, source, target)
2599
# These two attributes may be overridden by e.g. InterOtherToRemote to
2600
# provide a faster implementation.
2601
self.target_get_graph = self.target.get_graph
2602
self.target_get_parent_map = self.target.get_parent_map
2605
def copy_content(self, revision_id=None):
2606
"""Make a complete copy of the content in self into destination.
2608
This is a destructive operation! Do not use it on existing
2611
:param revision_id: Only copy the content needed to construct
2612
revision_id and its parents.
2615
self.target.set_make_working_trees(self.source.make_working_trees())
2616
except NotImplementedError:
2618
self.target.fetch(self.source, revision_id=revision_id)
2620
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2622
"""Fetch the content required to construct revision_id.
2624
The content is copied from self.source to self.target.
2626
:param revision_id: if None all content is copied, if NULL_REVISION no
2628
:param pb: optional progress bar to use for progress reports. If not
2629
provided a default one will be created.
2632
from bzrlib.fetch import RepoFetcher
2633
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2634
self.source, self.source._format, self.target,
2635
self.target._format)
2636
f = RepoFetcher(to_repository=self.target,
2637
from_repository=self.source,
2638
last_revision=revision_id,
2639
fetch_spec=fetch_spec,
2640
pb=pb, find_ghosts=find_ghosts)
2642
def _walk_to_common_revisions(self, revision_ids):
2643
"""Walk out from revision_ids in source to revisions target has.
2645
:param revision_ids: The start point for the search.
2646
:return: A set of revision ids.
2648
target_graph = self.target_get_graph()
2649
revision_ids = frozenset(revision_ids)
2650
# Fast path for the case where all the revisions are already in the
2652
# (Although this does incur an extra round trip for the
2653
# fairly common case where the target doesn't already have the revision
2655
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2656
return graph.SearchResult(revision_ids, set(), 0, set())
2657
missing_revs = set()
2658
source_graph = self.source.get_graph()
2659
# ensure we don't pay silly lookup costs.
2660
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2661
null_set = frozenset([_mod_revision.NULL_REVISION])
2662
searcher_exhausted = False
2666
# Iterate the searcher until we have enough next_revs
2667
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2669
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2670
next_revs.update(next_revs_part)
2671
ghosts.update(ghosts_part)
2672
except StopIteration:
2673
searcher_exhausted = True
2675
# If there are ghosts in the source graph, and the caller asked for
2676
# them, make sure that they are present in the target.
2677
# We don't care about other ghosts as we can't fetch them and
2678
# haven't been asked to.
2679
ghosts_to_check = set(revision_ids.intersection(ghosts))
2680
revs_to_get = set(next_revs).union(ghosts_to_check)
2682
have_revs = set(target_graph.get_parent_map(revs_to_get))
2683
# we always have NULL_REVISION present.
2684
have_revs = have_revs.union(null_set)
2685
# Check if the target is missing any ghosts we need.
2686
ghosts_to_check.difference_update(have_revs)
2688
# One of the caller's revision_ids is a ghost in both the
2689
# source and the target.
2690
raise errors.NoSuchRevision(
2691
self.source, ghosts_to_check.pop())
2692
missing_revs.update(next_revs - have_revs)
2693
# Because we may have walked past the original stop point, make
2694
# sure everything is stopped
2695
stop_revs = searcher.find_seen_ancestors(have_revs)
2696
searcher.stop_searching_any(stop_revs)
2697
if searcher_exhausted:
2699
return searcher.get_result()
2701
@deprecated_method(one_two)
2703
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2704
"""Return the revision ids that source has that target does not.
2706
These are returned in topological order.
2708
:param revision_id: only return revision ids included by this
2710
:param find_ghosts: If True find missing revisions in deep history
2711
rather than just finding the surface difference.
2713
return list(self.search_missing_revision_ids(
2714
revision_id, find_ghosts).get_keys())
2717
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2718
"""Return the revision ids that source has that target does not.
2720
:param revision_id: only return revision ids included by this
2722
:param find_ghosts: If True find missing revisions in deep history
2723
rather than just finding the surface difference.
2724
:return: A bzrlib.graph.SearchResult.
2726
# stop searching at found target revisions.
2727
if not find_ghosts and revision_id is not None:
2728
return self._walk_to_common_revisions([revision_id])
2729
# generic, possibly worst case, slow code path.
2730
target_ids = set(self.target.all_revision_ids())
2731
if revision_id is not None:
2732
source_ids = self.source.get_ancestry(revision_id)
2733
if source_ids[0] is not None:
2734
raise AssertionError()
2737
source_ids = self.source.all_revision_ids()
2738
result_set = set(source_ids).difference(target_ids)
2739
return self.source.revision_ids_to_search_result(result_set)
2742
def _same_model(source, target):
2743
"""True if source and target have the same data representation.
2745
Note: this is always called on the base class; overriding it in a
2746
subclass will have no effect.
2749
InterRepository._assert_same_model(source, target)
2751
except errors.IncompatibleRepositories, e:
2755
def _assert_same_model(source, target):
2756
"""Raise an exception if two repositories do not use the same model.
2758
if source.supports_rich_root() != target.supports_rich_root():
2759
raise errors.IncompatibleRepositories(source, target,
2760
"different rich-root support")
2761
if source._serializer != target._serializer:
2762
raise errors.IncompatibleRepositories(source, target,
2763
"different serializers")
2766
class InterSameDataRepository(InterRepository):
2767
"""Code for converting between repositories that represent the same data.
2769
Data format and model must match for this to work.
2773
def _get_repo_format_to_test(self):
2774
"""Repository format for testing with.
2776
InterSameData can pull from subtree to subtree and from non-subtree to
2777
non-subtree, so we test this with the richest repository format.
2779
from bzrlib.repofmt import knitrepo
2780
return knitrepo.RepositoryFormatKnit3()
2783
def is_compatible(source, target):
2784
return InterRepository._same_model(source, target)
2787
class InterWeaveRepo(InterSameDataRepository):
2788
"""Optimised code paths between Weave based repositories.
2790
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2791
implemented lazy inter-object optimisation.
2795
def _get_repo_format_to_test(self):
2796
from bzrlib.repofmt import weaverepo
2797
return weaverepo.RepositoryFormat7()
2800
def is_compatible(source, target):
2801
"""Be compatible with known Weave formats.
2803
We don't test for the stores being of specific types because that
2804
could lead to confusing results, and there is no need to be
2807
from bzrlib.repofmt.weaverepo import (
2813
return (isinstance(source._format, (RepositoryFormat5,
2815
RepositoryFormat7)) and
2816
isinstance(target._format, (RepositoryFormat5,
2818
RepositoryFormat7)))
2819
except AttributeError:
2823
def copy_content(self, revision_id=None):
2824
"""See InterRepository.copy_content()."""
2825
# weave specific optimised path:
2827
self.target.set_make_working_trees(self.source.make_working_trees())
2828
except (errors.RepositoryUpgradeRequired, NotImplemented):
2830
# FIXME do not peek!
2831
if self.source._transport.listable():
2832
pb = ui.ui_factory.nested_progress_bar()
2834
self.target.texts.insert_record_stream(
2835
self.source.texts.get_record_stream(
2836
self.source.texts.keys(), 'topological', False))
2837
pb.update('copying inventory', 0, 1)
2838
self.target.inventories.insert_record_stream(
2839
self.source.inventories.get_record_stream(
2840
self.source.inventories.keys(), 'topological', False))
2841
self.target.signatures.insert_record_stream(
2842
self.source.signatures.get_record_stream(
2843
self.source.signatures.keys(),
2845
self.target.revisions.insert_record_stream(
2846
self.source.revisions.get_record_stream(
2847
self.source.revisions.keys(),
2848
'topological', True))
2852
self.target.fetch(self.source, revision_id=revision_id)
2855
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2857
"""See InterRepository.fetch()."""
2858
from bzrlib.fetch import RepoFetcher
2859
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2860
self.source, self.source._format, self.target, self.target._format)
2861
f = RepoFetcher(to_repository=self.target,
2862
from_repository=self.source,
2863
last_revision=revision_id,
2864
fetch_spec=fetch_spec,
2865
pb=pb, find_ghosts=find_ghosts)
2868
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2869
"""See InterRepository.missing_revision_ids()."""
2870
# we want all revisions to satisfy revision_id in source.
2871
# but we don't want to stat every file here and there.
2872
# we want then, all revisions other needs to satisfy revision_id
2873
# checked, but not those that we have locally.
2874
# so the first thing is to get a subset of the revisions to
2875
# satisfy revision_id in source, and then eliminate those that
2876
# we do already have.
2877
# this is slow on high latency connection to self, but as as this
2878
# disk format scales terribly for push anyway due to rewriting
2879
# inventory.weave, this is considered acceptable.
2881
if revision_id is not None:
2882
source_ids = self.source.get_ancestry(revision_id)
2883
if source_ids[0] is not None:
2884
raise AssertionError()
2887
source_ids = self.source._all_possible_ids()
2888
source_ids_set = set(source_ids)
2889
# source_ids is the worst possible case we may need to pull.
2890
# now we want to filter source_ids against what we actually
2891
# have in target, but don't try to check for existence where we know
2892
# we do not have a revision as that would be pointless.
2893
target_ids = set(self.target._all_possible_ids())
2894
possibly_present_revisions = target_ids.intersection(source_ids_set)
2895
actually_present_revisions = set(
2896
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2897
required_revisions = source_ids_set.difference(actually_present_revisions)
2898
if revision_id is not None:
2899
# we used get_ancestry to determine source_ids then we are assured all
2900
# revisions referenced are present as they are installed in topological order.
2901
# and the tip revision was validated by get_ancestry.
2902
result_set = required_revisions
2904
# if we just grabbed the possibly available ids, then
2905
# we only have an estimate of whats available and need to validate
2906
# that against the revision records.
2908
self.source._eliminate_revisions_not_present(required_revisions))
2909
return self.source.revision_ids_to_search_result(result_set)
2912
class InterKnitRepo(InterSameDataRepository):
2913
"""Optimised code paths between Knit based repositories."""
2916
def _get_repo_format_to_test(self):
2917
from bzrlib.repofmt import knitrepo
2918
return knitrepo.RepositoryFormatKnit1()
2921
def is_compatible(source, target):
2922
"""Be compatible with known Knit formats.
2924
We don't test for the stores being of specific types because that
2925
could lead to confusing results, and there is no need to be
2928
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2930
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2931
isinstance(target._format, RepositoryFormatKnit))
2932
except AttributeError:
2934
return are_knits and InterRepository._same_model(source, target)
2937
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2939
"""See InterRepository.fetch()."""
2940
from bzrlib.fetch import RepoFetcher
2941
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2942
self.source, self.source._format, self.target, self.target._format)
2943
f = RepoFetcher(to_repository=self.target,
2944
from_repository=self.source,
2945
last_revision=revision_id,
2946
fetch_spec=fetch_spec,
2947
pb=pb, find_ghosts=find_ghosts)
2950
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2951
"""See InterRepository.missing_revision_ids()."""
2952
if revision_id is not None:
2953
source_ids = self.source.get_ancestry(revision_id)
2954
if source_ids[0] is not None:
2955
raise AssertionError()
2958
source_ids = self.source.all_revision_ids()
2959
source_ids_set = set(source_ids)
2960
# source_ids is the worst possible case we may need to pull.
2961
# now we want to filter source_ids against what we actually
2962
# have in target, but don't try to check for existence where we know
2963
# we do not have a revision as that would be pointless.
2964
target_ids = set(self.target.all_revision_ids())
2965
possibly_present_revisions = target_ids.intersection(source_ids_set)
2966
actually_present_revisions = set(
2967
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2968
required_revisions = source_ids_set.difference(actually_present_revisions)
2969
if revision_id is not None:
2970
# we used get_ancestry to determine source_ids then we are assured all
2971
# revisions referenced are present as they are installed in topological order.
2972
# and the tip revision was validated by get_ancestry.
2973
result_set = required_revisions
2975
# if we just grabbed the possibly available ids, then
2976
# we only have an estimate of whats available and need to validate
2977
# that against the revision records.
2979
self.source._eliminate_revisions_not_present(required_revisions))
2980
return self.source.revision_ids_to_search_result(result_set)
2983
class InterPackRepo(InterSameDataRepository):
2984
"""Optimised code paths between Pack based repositories."""
2987
def _get_repo_format_to_test(self):
2988
from bzrlib.repofmt import pack_repo
2989
return pack_repo.RepositoryFormatKnitPack1()
2992
def is_compatible(source, target):
2993
"""Be compatible with known Pack formats.
2995
We don't test for the stores being of specific types because that
2996
could lead to confusing results, and there is no need to be
2999
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3001
are_packs = (isinstance(source._format, RepositoryFormatPack) and
3002
isinstance(target._format, RepositoryFormatPack))
3003
except AttributeError:
3005
return are_packs and InterRepository._same_model(source, target)
3008
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3010
"""See InterRepository.fetch()."""
3011
if (len(self.source._fallback_repositories) > 0 or
3012
len(self.target._fallback_repositories) > 0):
3013
# The pack layer is not aware of fallback repositories, so when
3014
# fetching from a stacked repository or into a stacked repository
3015
# we use the generic fetch logic which uses the VersionedFiles
3016
# attributes on repository.
3017
from bzrlib.fetch import RepoFetcher
3018
fetcher = RepoFetcher(self.target, self.source, revision_id,
3019
pb, find_ghosts, fetch_spec=fetch_spec)
3020
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3021
self.source, self.source._format, self.target, self.target._format)
3022
if fetch_spec is not None:
3023
if len(list(fetch_spec.heads)) != 1:
3024
raise AssertionError(
3025
"InterPackRepo.fetch doesn't support "
3026
"fetching multiple heads yet.")
3027
revision_id = fetch_spec.heads[0]
3029
if revision_id is None:
3031
# everything to do - use pack logic
3032
# to fetch from all packs to one without
3033
# inventory parsing etc, IFF nothing to be copied is in the target.
3035
source_revision_ids = frozenset(self.source.all_revision_ids())
3036
revision_ids = source_revision_ids - \
3037
frozenset(self.target_get_parent_map(source_revision_ids))
3038
revision_keys = [(revid,) for revid in revision_ids]
3039
target_pack_collection = self._get_target_pack_collection()
3040
index = target_pack_collection.revision_index.combined_index
3041
present_revision_ids = set(item[1][0] for item in
3042
index.iter_entries(revision_keys))
3043
revision_ids = set(revision_ids) - present_revision_ids
3044
# implementing the TODO will involve:
3045
# - detecting when all of a pack is selected
3046
# - avoiding as much as possible pre-selection, so the
3047
# more-core routines such as create_pack_from_packs can filter in
3048
# a just-in-time fashion. (though having a HEADS list on a
3049
# repository might make this a lot easier, because we could
3050
# sensibly detect 'new revisions' without doing a full index scan.
3051
elif _mod_revision.is_null(revision_id):
3056
revision_ids = self.search_missing_revision_ids(revision_id,
3057
find_ghosts=find_ghosts).get_keys()
3058
except errors.NoSuchRevision:
3059
raise errors.InstallFailed([revision_id])
3060
if len(revision_ids) == 0:
3062
return self._pack(self.source, self.target, revision_ids)
3064
def _pack(self, source, target, revision_ids):
3065
from bzrlib.repofmt.pack_repo import Packer
3066
target_pack_collection = self._get_target_pack_collection()
3067
packs = source._pack_collection.all_packs()
3068
pack = Packer(target_pack_collection, packs, '.fetch',
3069
revision_ids).pack()
3070
if pack is not None:
3071
target_pack_collection._save_pack_names()
3072
copied_revs = pack.get_revision_count()
3073
# Trigger an autopack. This may duplicate effort as we've just done
3074
# a pack creation, but for now it is simpler to think about as
3075
# 'upload data, then repack if needed'.
3077
return (copied_revs, [])
3081
def _autopack(self):
3082
self.target._pack_collection.autopack()
3084
def _get_target_pack_collection(self):
3085
return self.target._pack_collection
3088
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3089
"""See InterRepository.missing_revision_ids().
3091
:param find_ghosts: Find ghosts throughout the ancestry of
3094
if not find_ghosts and revision_id is not None:
3095
return self._walk_to_common_revisions([revision_id])
3096
elif revision_id is not None:
3097
# Find ghosts: search for revisions pointing from one repository to
3098
# the other, and vice versa, anywhere in the history of revision_id.
3099
graph = self.target_get_graph(other_repository=self.source)
3100
searcher = graph._make_breadth_first_searcher([revision_id])
3104
next_revs, ghosts = searcher.next_with_ghosts()
3105
except StopIteration:
3107
if revision_id in ghosts:
3108
raise errors.NoSuchRevision(self.source, revision_id)
3109
found_ids.update(next_revs)
3110
found_ids.update(ghosts)
3111
found_ids = frozenset(found_ids)
3112
# Double query here: should be able to avoid this by changing the
3113
# graph api further.
3114
result_set = found_ids - frozenset(
3115
self.target_get_parent_map(found_ids))
3117
source_ids = self.source.all_revision_ids()
3118
# source_ids is the worst possible case we may need to pull.
3119
# now we want to filter source_ids against what we actually
3120
# have in target, but don't try to check for existence where we know
3121
# we do not have a revision as that would be pointless.
3122
target_ids = set(self.target.all_revision_ids())
3123
result_set = set(source_ids).difference(target_ids)
3124
return self.source.revision_ids_to_search_result(result_set)
3127
class InterDifferingSerializer(InterKnitRepo):
3130
def _get_repo_format_to_test(self):
3134
def is_compatible(source, target):
3135
"""Be compatible with Knit2 source and Knit3 target"""
3136
if source.supports_rich_root() != target.supports_rich_root():
3138
# Ideally, we'd support fetching if the source had no tree references
3139
# even if it supported them...
3140
if (getattr(source, '_format.supports_tree_reference', False) and
3141
not getattr(target, '_format.supports_tree_reference', False)):
3145
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3146
"""Get the best delta and base for this revision.
3148
:return: (basis_id, delta)
3150
possible_trees = [(parent_id, cache[parent_id])
3151
for parent_id in parent_ids
3152
if parent_id in cache]
3153
if len(possible_trees) == 0:
3154
# There either aren't any parents, or the parents aren't in the
3155
# cache, so just use the last converted tree
3156
possible_trees.append((basis_id, cache[basis_id]))
3158
for basis_id, basis_tree in possible_trees:
3159
delta = tree.inventory._make_delta(basis_tree.inventory)
3160
deltas.append((len(delta), basis_id, delta))
3162
return deltas[0][1:]
3164
def _fetch_batch(self, revision_ids, basis_id, cache):
3165
"""Fetch across a few revisions.
3167
:param revision_ids: The revisions to copy
3168
:param basis_id: The revision_id of a tree that must be in cache, used
3169
as a basis for delta when no other base is available
3170
:param cache: A cache of RevisionTrees that we can use.
3171
:return: The revision_id of the last converted tree. The RevisionTree
3172
for it will be in cache
3174
# Walk though all revisions; get inventory deltas, copy referenced
3175
# texts that delta references, insert the delta, revision and
3179
pending_revisions = []
3180
parent_map = self.source.get_parent_map(revision_ids)
3181
for tree in self.source.revision_trees(revision_ids):
3182
current_revision_id = tree.get_revision_id()
3183
parent_ids = parent_map.get(current_revision_id, ())
3184
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3186
# Find text entries that need to be copied
3187
for old_path, new_path, file_id, entry in delta:
3188
if new_path is not None:
3189
if not (new_path or self.target.supports_rich_root()):
3190
# We don't copy the text for the root node unless the
3191
# target supports_rich_root.
3193
text_keys.add((file_id, entry.revision))
3194
revision = self.source.get_revision(current_revision_id)
3195
pending_deltas.append((basis_id, delta,
3196
current_revision_id, revision.parent_ids))
3197
pending_revisions.append(revision)
3198
cache[current_revision_id] = tree
3199
basis_id = current_revision_id
3201
from_texts = self.source.texts
3202
to_texts = self.target.texts
3203
to_texts.insert_record_stream(from_texts.get_record_stream(
3204
text_keys, self.target._format._fetch_order,
3205
not self.target._format._fetch_uses_deltas))
3207
for delta in pending_deltas:
3208
self.target.add_inventory_by_delta(*delta)
3209
# insert signatures and revisions
3210
for revision in pending_revisions:
3212
signature = self.source.get_signature_text(
3213
revision.revision_id)
3214
self.target.add_signature_text(revision.revision_id,
3216
except errors.NoSuchRevision:
3218
self.target.add_revision(revision.revision_id, revision)
3221
def _fetch_all_revisions(self, revision_ids, pb):
3222
"""Fetch everything for the list of revisions.
3224
:param revision_ids: The list of revisions to fetch. Must be in
3226
:param pb: A ProgressBar
3229
basis_id, basis_tree = self._get_basis(revision_ids[0])
3231
cache = lru_cache.LRUCache(100)
3232
cache[basis_id] = basis_tree
3233
del basis_tree # We don't want to hang on to it here
3234
for offset in range(0, len(revision_ids), batch_size):
3235
self.target.start_write_group()
3237
pb.update('Transferring revisions', offset,
3239
batch = revision_ids[offset:offset+batch_size]
3240
basis_id = self._fetch_batch(batch, basis_id, cache)
3242
self.target.abort_write_group()
3245
self.target.commit_write_group()
3246
pb.update('Transferring revisions', len(revision_ids),
3250
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3252
"""See InterRepository.fetch()."""
3253
if fetch_spec is not None:
3254
raise AssertionError("Not implemented yet...")
3255
revision_ids = self.target.search_missing_revision_ids(self.source,
3256
revision_id, find_ghosts=find_ghosts).get_keys()
3257
if not revision_ids:
3259
revision_ids = tsort.topo_sort(
3260
self.source.get_graph().get_parent_map(revision_ids))
3262
my_pb = ui.ui_factory.nested_progress_bar()
3267
self._fetch_all_revisions(revision_ids, pb)
3269
if my_pb is not None:
3271
return len(revision_ids), 0
3273
def _get_basis(self, first_revision_id):
3274
"""Get a revision and tree which exists in the target.
3276
This assumes that first_revision_id is selected for transmission
3277
because all other ancestors are already present. If we can't find an
3278
ancestor we fall back to NULL_REVISION since we know that is safe.
3280
:return: (basis_id, basis_tree)
3282
first_rev = self.source.get_revision(first_revision_id)
3284
basis_id = first_rev.parent_ids[0]
3285
# only valid as a basis if the target has it
3286
self.target.get_revision(basis_id)
3287
# Try to get a basis tree - if its a ghost it will hit the
3288
# NoSuchRevision case.
3289
basis_tree = self.source.revision_tree(basis_id)
3290
except (IndexError, errors.NoSuchRevision):
3291
basis_id = _mod_revision.NULL_REVISION
3292
basis_tree = self.source.revision_tree(basis_id)
3293
return basis_id, basis_tree
3296
class InterOtherToRemote(InterRepository):
3297
"""An InterRepository that simply delegates to the 'real' InterRepository
3298
calculated for (source, target._real_repository).
3301
_walk_to_common_revisions_batch_size = 50
3303
def __init__(self, source, target):
3304
InterRepository.__init__(self, source, target)
3305
self._real_inter = None
3308
def is_compatible(source, target):
3309
if isinstance(target, remote.RemoteRepository):
3313
def _ensure_real_inter(self):
3314
if self._real_inter is None:
3315
self.target._ensure_real()
3316
real_target = self.target._real_repository
3317
self._real_inter = InterRepository.get(self.source, real_target)
3318
# Make _real_inter use the RemoteRepository for get_parent_map
3319
self._real_inter.target_get_graph = self.target.get_graph
3320
self._real_inter.target_get_parent_map = self.target.get_parent_map
3322
def copy_content(self, revision_id=None):
3323
self._ensure_real_inter()
3324
self._real_inter.copy_content(revision_id=revision_id)
3326
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3328
self._ensure_real_inter()
3329
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3330
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
3333
def _get_repo_format_to_test(self):
3337
class InterRemoteToOther(InterRepository):
3339
def __init__(self, source, target):
3340
InterRepository.__init__(self, source, target)
3341
self._real_inter = None
3344
def is_compatible(source, target):
3345
if not isinstance(source, remote.RemoteRepository):
3347
return InterRepository._same_model(source, target)
3349
def _ensure_real_inter(self):
3350
if self._real_inter is None:
3351
self.source._ensure_real()
3352
real_source = self.source._real_repository
3353
self._real_inter = InterRepository.get(real_source, self.target)
3356
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3358
"""See InterRepository.fetch()."""
3359
# Always fetch using the generic streaming fetch code, to allow
3360
# streaming fetching from remote servers.
3361
from bzrlib.fetch import RepoFetcher
3362
fetcher = RepoFetcher(self.target, self.source, revision_id,
3363
pb, find_ghosts, fetch_spec=fetch_spec)
3365
def copy_content(self, revision_id=None):
3366
self._ensure_real_inter()
3367
self._real_inter.copy_content(revision_id=revision_id)
3370
def _get_repo_format_to_test(self):
3375
class InterPackToRemotePack(InterPackRepo):
3376
"""A specialisation of InterPackRepo for a target that is a
3379
This will use the get_parent_map RPC rather than plain readvs, and also
3380
uses an RPC for autopacking.
3383
_walk_to_common_revisions_batch_size = 50
3386
def is_compatible(source, target):
3387
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3388
if isinstance(source._format, RepositoryFormatPack):
3389
if isinstance(target, remote.RemoteRepository):
3390
target._format._ensure_real()
3391
if isinstance(target._format._custom_format,
3392
RepositoryFormatPack):
3393
if InterRepository._same_model(source, target):
3397
def _autopack(self):
3398
self.target.autopack()
3401
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3403
"""See InterRepository.fetch()."""
3404
# Always fetch using the generic streaming fetch code, to allow
3405
# streaming fetching into remote servers.
3406
from bzrlib.fetch import RepoFetcher
3407
fetcher = RepoFetcher(self.target, self.source, revision_id,
3408
pb, find_ghosts, fetch_spec=fetch_spec)
3410
def _get_target_pack_collection(self):
3411
return self.target._real_repository._pack_collection
3414
def _get_repo_format_to_test(self):
3418
InterRepository.register_optimiser(InterDifferingSerializer)
3419
InterRepository.register_optimiser(InterSameDataRepository)
3420
InterRepository.register_optimiser(InterWeaveRepo)
3421
InterRepository.register_optimiser(InterKnitRepo)
3422
InterRepository.register_optimiser(InterPackRepo)
3423
InterRepository.register_optimiser(InterOtherToRemote)
3424
InterRepository.register_optimiser(InterRemoteToOther)
3425
InterRepository.register_optimiser(InterPackToRemotePack)
3428
class CopyConverter(object):
3429
"""A repository conversion tool which just performs a copy of the content.
3431
This is slow but quite reliable.
3434
def __init__(self, target_format):
3435
"""Create a CopyConverter.
3437
:param target_format: The format the resulting repository should be.
3439
self.target_format = target_format
3441
def convert(self, repo, pb):
3442
"""Perform the conversion of to_convert, giving feedback via pb.
3444
:param to_convert: The disk object to convert.
3445
:param pb: a progress bar to use for progress information.
3450
# this is only useful with metadir layouts - separated repo content.
3451
# trigger an assertion if not such
3452
repo._format.get_format_string()
3453
self.repo_dir = repo.bzrdir
3454
self.step('Moving repository to repository.backup')
3455
self.repo_dir.transport.move('repository', 'repository.backup')
3456
backup_transport = self.repo_dir.transport.clone('repository.backup')
3457
repo._format.check_conversion_target(self.target_format)
3458
self.source_repo = repo._format.open(self.repo_dir,
3460
_override_transport=backup_transport)
3461
self.step('Creating new repository')
3462
converted = self.target_format.initialize(self.repo_dir,
3463
self.source_repo.is_shared())
3464
converted.lock_write()
3466
self.step('Copying content into repository.')
3467
self.source_repo.copy_content_into(converted)
3470
self.step('Deleting old repository content.')
3471
self.repo_dir.transport.delete_tree('repository.backup')
3472
self.pb.note('repository converted')
3474
def step(self, message):
3475
"""Update the pb by a step."""
3477
self.pb.update(message, self.count, self.total)
3489
def _unescaper(match, _map=_unescape_map):
3490
code = match.group(1)
3494
if not code.startswith('#'):
3496
return unichr(int(code[1:])).encode('utf8')
3502
def _unescape_xml(data):
3503
"""Unescape predefined XML entities in a string of data."""
3505
if _unescape_re is None:
3506
_unescape_re = re.compile('\&([^;]*);')
3507
return _unescape_re.sub(_unescaper, data)
3510
class _VersionedFileChecker(object):
3512
def __init__(self, repository):
3513
self.repository = repository
3514
self.text_index = self.repository._generate_text_key_index()
3516
def calculate_file_version_parents(self, text_key):
3517
"""Calculate the correct parents for a file version according to
3520
parent_keys = self.text_index[text_key]
3521
if parent_keys == [_mod_revision.NULL_REVISION]:
3523
return tuple(parent_keys)
3525
def check_file_version_parents(self, texts, progress_bar=None):
3526
"""Check the parents stored in a versioned file are correct.
3528
It also detects file versions that are not referenced by their
3529
corresponding revision's inventory.
3531
:returns: A tuple of (wrong_parents, dangling_file_versions).
3532
wrong_parents is a dict mapping {revision_id: (stored_parents,
3533
correct_parents)} for each revision_id where the stored parents
3534
are not correct. dangling_file_versions is a set of (file_id,
3535
revision_id) tuples for versions that are present in this versioned
3536
file, but not used by the corresponding inventory.
3539
self.file_ids = set([file_id for file_id, _ in
3540
self.text_index.iterkeys()])
3541
# text keys is now grouped by file_id
3542
n_weaves = len(self.file_ids)
3543
files_in_revisions = {}
3544
revisions_of_files = {}
3545
n_versions = len(self.text_index)
3546
progress_bar.update('loading text store', 0, n_versions)
3547
parent_map = self.repository.texts.get_parent_map(self.text_index)
3548
# On unlistable transports this could well be empty/error...
3549
text_keys = self.repository.texts.keys()
3550
unused_keys = frozenset(text_keys) - set(self.text_index)
3551
for num, key in enumerate(self.text_index.iterkeys()):
3552
if progress_bar is not None:
3553
progress_bar.update('checking text graph', num, n_versions)
3554
correct_parents = self.calculate_file_version_parents(key)
3556
knit_parents = parent_map[key]
3557
except errors.RevisionNotPresent:
3560
if correct_parents != knit_parents:
3561
wrong_parents[key] = (knit_parents, correct_parents)
3562
return wrong_parents, unused_keys
3565
def _old_get_graph(repository, revision_id):
3566
"""DO NOT USE. That is all. I'm serious."""
3567
graph = repository.get_graph()
3568
revision_graph = dict(((key, value) for key, value in
3569
graph.iter_ancestry([revision_id]) if value is not None))
3570
return _strip_NULL_ghosts(revision_graph)
3573
def _strip_NULL_ghosts(revision_graph):
3574
"""Also don't use this. more compatibility code for unmigrated clients."""
3575
# Filter ghosts, and null:
3576
if _mod_revision.NULL_REVISION in revision_graph:
3577
del revision_graph[_mod_revision.NULL_REVISION]
3578
for key, parents in revision_graph.items():
3579
revision_graph[key] = tuple(parent for parent in parents if parent
3581
return revision_graph
3584
class StreamSink(object):
3585
"""An object that can insert a stream into a repository.
3587
This interface handles the complexity of reserialising inventories and
3588
revisions from different formats, and allows unidirectional insertion into
3589
stacked repositories without looking for the missing basis parents
3593
def __init__(self, target_repo):
3594
self.target_repo = target_repo
3596
def insert_stream(self, stream, src_format, resume_tokens):
3597
"""Insert a stream's content into the target repository.
3599
:param src_format: a bzr repository format.
3601
:return: a list of resume tokens and an iterable of keys additional
3602
items required before the insertion can be completed.
3604
self.target_repo.lock_write()
3607
self.target_repo.resume_write_group(resume_tokens)
3609
self.target_repo.start_write_group()
3611
# locked_insert_stream performs a commit|suspend.
3612
return self._locked_insert_stream(stream, src_format)
3614
self.target_repo.abort_write_group(suppress_errors=True)
3617
self.target_repo.unlock()
3619
def _locked_insert_stream(self, stream, src_format):
3620
to_serializer = self.target_repo._format._serializer
3621
src_serializer = src_format._serializer
3622
for substream_type, substream in stream:
3623
if substream_type == 'texts':
3624
self.target_repo.texts.insert_record_stream(substream)
3625
elif substream_type == 'inventories':
3626
if src_serializer == to_serializer:
3627
self.target_repo.inventories.insert_record_stream(
3630
self._extract_and_insert_inventories(
3631
substream, src_serializer)
3632
elif substream_type == 'revisions':
3633
# This may fallback to extract-and-insert more often than
3634
# required if the serializers are different only in terms of
3636
if src_serializer == to_serializer:
3637
self.target_repo.revisions.insert_record_stream(
3640
self._extract_and_insert_revisions(substream,
3642
elif substream_type == 'signatures':
3643
self.target_repo.signatures.insert_record_stream(substream)
3645
raise AssertionError('kaboom! %s' % (substream_type,))
3647
missing_keys = set()
3648
for prefix, versioned_file in (
3649
('texts', self.target_repo.texts),
3650
('inventories', self.target_repo.inventories),
3651
('revisions', self.target_repo.revisions),
3652
('signatures', self.target_repo.signatures),
3654
missing_keys.update((prefix,) + key for key in
3655
versioned_file.get_missing_compression_parent_keys())
3656
except NotImplementedError:
3657
# cannot even attempt suspending, and missing would have failed
3658
# during stream insertion.
3659
missing_keys = set()
3662
# suspend the write group and tell the caller what we is
3663
# missing. We know we can suspend or else we would not have
3664
# entered this code path. (All repositories that can handle
3665
# missing keys can handle suspending a write group).
3666
write_group_tokens = self.target_repo.suspend_write_group()
3667
return write_group_tokens, missing_keys
3668
self.target_repo.commit_write_group()
3671
def _extract_and_insert_inventories(self, substream, serializer):
3672
"""Generate a new inventory versionedfile in target, converting data.
3674
The inventory is retrieved from the source, (deserializing it), and
3675
stored in the target (reserializing it in a different format).
3677
for record in substream:
3678
bytes = record.get_bytes_as('fulltext')
3679
revision_id = record.key[0]
3680
inv = serializer.read_inventory_from_string(bytes, revision_id)
3681
parents = [key[0] for key in record.parents]
3682
self.target_repo.add_inventory(revision_id, inv, parents)
3684
def _extract_and_insert_revisions(self, substream, serializer):
3685
for record in substream:
3686
bytes = record.get_bytes_as('fulltext')
3687
revision_id = record.key[0]
3688
rev = serializer.read_revision_from_string(bytes)
3689
if rev.revision_id != revision_id:
3690
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
3691
self.target_repo.add_revision(revision_id, rev)
3694
if self.target_repo._format._fetch_reconcile:
3695
self.target_repo.reconcile()
3698
class StreamSource(object):
3699
"""A source of a stream for fetching between repositories."""
3701
def __init__(self, from_repository, to_format):
3702
"""Create a StreamSource streaming from from_repository."""
3703
self.from_repository = from_repository
3704
self.to_format = to_format
3706
def delta_on_metadata(self):
3707
"""Return True if delta's are permitted on metadata streams.
3709
That is on revisions and signatures.
3711
src_serializer = self.from_repository._format._serializer
3712
target_serializer = self.to_format._serializer
3713
return (self.to_format._fetch_uses_deltas and
3714
src_serializer == target_serializer)
3716
def _fetch_revision_texts(self, revs):
3717
# fetch signatures first and then the revision texts
3718
# may need to be a InterRevisionStore call here.
3719
from_sf = self.from_repository.signatures
3720
# A missing signature is just skipped.
3721
keys = [(rev_id,) for rev_id in revs]
3722
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
3724
self.to_format._fetch_order,
3725
not self.to_format._fetch_uses_deltas))
3726
# If a revision has a delta, this is actually expanded inside the
3727
# insert_record_stream code now, which is an alternate fix for
3729
from_rf = self.from_repository.revisions
3730
revisions = from_rf.get_record_stream(
3732
self.to_format._fetch_order,
3733
not self.delta_on_metadata())
3734
return [('signatures', signatures), ('revisions', revisions)]
3736
def _generate_root_texts(self, revs):
3737
"""This will be called by __fetch between fetching weave texts and
3738
fetching the inventory weave.
3740
Subclasses should override this if they need to generate root texts
3741
after fetching weave texts.
3743
if self._rich_root_upgrade():
3745
return bzrlib.fetch.Inter1and2Helper(
3746
self.from_repository).generate_root_texts(revs)
3750
def get_stream(self, search):
3752
revs = search.get_keys()
3753
graph = self.from_repository.get_graph()
3754
revs = list(graph.iter_topo_order(revs))
3755
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
3757
for knit_kind, file_id, revisions in data_to_fetch:
3758
if knit_kind != phase:
3760
# Make a new progress bar for this phase
3761
if knit_kind == "file":
3762
# Accumulate file texts
3763
text_keys.extend([(file_id, revision) for revision in
3765
elif knit_kind == "inventory":
3766
# Now copy the file texts.
3767
from_texts = self.from_repository.texts
3768
yield ('texts', from_texts.get_record_stream(
3769
text_keys, self.to_format._fetch_order,
3770
not self.to_format._fetch_uses_deltas))
3771
# Cause an error if a text occurs after we have done the
3774
# Before we process the inventory we generate the root
3775
# texts (if necessary) so that the inventories references
3777
for _ in self._generate_root_texts(revs):
3779
# NB: This currently reopens the inventory weave in source;
3780
# using a single stream interface instead would avoid this.
3781
from_weave = self.from_repository.inventories
3782
# we fetch only the referenced inventories because we do not
3783
# know for unselected inventories whether all their required
3784
# texts are present in the other repository - it could be
3786
yield ('inventories', from_weave.get_record_stream(
3787
[(rev_id,) for rev_id in revs],
3788
self.inventory_fetch_order(),
3789
not self.delta_on_metadata()))
3790
elif knit_kind == "signatures":
3791
# Nothing to do here; this will be taken care of when
3792
# _fetch_revision_texts happens.
3794
elif knit_kind == "revisions":
3795
for record in self._fetch_revision_texts(revs):
3798
raise AssertionError("Unknown knit kind %r" % knit_kind)
3800
def get_stream_for_missing_keys(self, missing_keys):
3801
# missing keys can only occur when we are byte copying and not
3802
# translating (because translation means we don't send
3803
# unreconstructable deltas ever).
3805
keys['texts'] = set()
3806
keys['revisions'] = set()
3807
keys['inventories'] = set()
3808
keys['signatures'] = set()
3809
for key in missing_keys:
3810
keys[key[0]].add(key[1:])
3811
if len(keys['revisions']):
3812
# If we allowed copying revisions at this point, we could end up
3813
# copying a revision without copying its required texts: a
3814
# violation of the requirements for repository integrity.
3815
raise AssertionError(
3816
'cannot copy revisions to fill in missing deltas %s' % (
3817
keys['revisions'],))
3818
for substream_kind, keys in keys.iteritems():
3819
vf = getattr(self.from_repository, substream_kind)
3820
# Ask for full texts always so that we don't need more round trips
3821
# after this stream.
3822
stream = vf.get_record_stream(keys,
3823
self.to_format._fetch_order, True)
3824
yield substream_kind, stream
3826
def inventory_fetch_order(self):
3827
if self._rich_root_upgrade():
3828
return 'topological'
3830
return self.to_format._fetch_order
3832
def _rich_root_upgrade(self):
3833
return (not self.from_repository._format.rich_root_data and
3834
self.to_format.rich_root_data)
added
removed
bzrlib/repository.py
