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 _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1458
"""Helper routine for fileids_altered_by_revision_ids.
1460
This performs the translation of xml lines to revision ids.
1462
:param line_iterator: An iterator of lines, origin_version_id
1463
:param revision_ids: The revision ids to filter for. This should be a
1464
set or other type which supports efficient __contains__ lookups, as
1465
the revision id from each parsed line will be looked up in the
1466
revision_ids filter.
1467
:return: a dictionary mapping altered file-ids to an iterable of
1468
revision_ids. Each altered file-ids has the exact revision_ids that
1469
altered it listed explicitly.
1472
setdefault = result.setdefault
1474
self._find_text_key_references_from_xml_inventory_lines(
1475
line_iterator).iterkeys():
1476
# once data is all ensured-consistent; then this is
1477
# if revision_id == version_id
1478
if key[-1:] in revision_ids:
1479
setdefault(key[0], set()).add(key[-1])
1482
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1483
"""Find the file ids and versions affected by revisions.
1485
:param revisions: an iterable containing revision ids.
1486
:param _inv_weave: The inventory weave from this repository or None.
1487
If None, the inventory weave will be opened automatically.
1488
:return: a dictionary mapping altered file-ids to an iterable of
1489
revision_ids. Each altered file-ids has the exact revision_ids that
1490
altered it listed explicitly.
1492
selected_keys = set((revid,) for revid in revision_ids)
1493
w = _inv_weave or self.inventories
1494
pb = ui.ui_factory.nested_progress_bar()
1496
return self._find_file_ids_from_xml_inventory_lines(
1497
w.iter_lines_added_or_present_in_keys(
1498
selected_keys, pb=pb),
1503
def iter_files_bytes(self, desired_files):
1504
"""Iterate through file versions.
1506
Files will not necessarily be returned in the order they occur in
1507
desired_files. No specific order is guaranteed.
1509
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1510
value supplied by the caller as part of desired_files. It should
1511
uniquely identify the file version in the caller's context. (Examples:
1512
an index number or a TreeTransform trans_id.)
1514
bytes_iterator is an iterable of bytestrings for the file. The
1515
kind of iterable and length of the bytestrings are unspecified, but for
1516
this implementation, it is a list of bytes produced by
1517
VersionedFile.get_record_stream().
1519
:param desired_files: a list of (file_id, revision_id, identifier)
1523
for file_id, revision_id, callable_data in desired_files:
1524
text_keys[(file_id, revision_id)] = callable_data
1525
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1526
if record.storage_kind == 'absent':
1527
raise errors.RevisionNotPresent(record.key, self)
1528
yield text_keys[record.key], record.get_bytes_as('fulltext')
1530
def _generate_text_key_index(self, text_key_references=None,
1532
"""Generate a new text key index for the repository.
1534
This is an expensive function that will take considerable time to run.
1536
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1537
list of parents, also text keys. When a given key has no parents,
1538
the parents list will be [NULL_REVISION].
1540
# All revisions, to find inventory parents.
1541
if ancestors is None:
1542
graph = self.get_graph()
1543
ancestors = graph.get_parent_map(self.all_revision_ids())
1544
if text_key_references is None:
1545
text_key_references = self.find_text_key_references()
1546
pb = ui.ui_factory.nested_progress_bar()
1548
return self._do_generate_text_key_index(ancestors,
1549
text_key_references, pb)
1553
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1554
"""Helper for _generate_text_key_index to avoid deep nesting."""
1555
revision_order = tsort.topo_sort(ancestors)
1556
invalid_keys = set()
1558
for revision_id in revision_order:
1559
revision_keys[revision_id] = set()
1560
text_count = len(text_key_references)
1561
# a cache of the text keys to allow reuse; costs a dict of all the
1562
# keys, but saves a 2-tuple for every child of a given key.
1564
for text_key, valid in text_key_references.iteritems():
1566
invalid_keys.add(text_key)
1568
revision_keys[text_key[1]].add(text_key)
1569
text_key_cache[text_key] = text_key
1570
del text_key_references
1572
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1573
NULL_REVISION = _mod_revision.NULL_REVISION
1574
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1575
# too small for large or very branchy trees. However, for 55K path
1576
# trees, it would be easy to use too much memory trivially. Ideally we
1577
# could gauge this by looking at available real memory etc, but this is
1578
# always a tricky proposition.
1579
inventory_cache = lru_cache.LRUCache(10)
1580
batch_size = 10 # should be ~150MB on a 55K path tree
1581
batch_count = len(revision_order) / batch_size + 1
1583
pb.update("Calculating text parents.", processed_texts, text_count)
1584
for offset in xrange(batch_count):
1585
to_query = revision_order[offset * batch_size:(offset + 1) *
1589
for rev_tree in self.revision_trees(to_query):
1590
revision_id = rev_tree.get_revision_id()
1591
parent_ids = ancestors[revision_id]
1592
for text_key in revision_keys[revision_id]:
1593
pb.update("Calculating text parents.", processed_texts)
1594
processed_texts += 1
1595
candidate_parents = []
1596
for parent_id in parent_ids:
1597
parent_text_key = (text_key[0], parent_id)
1599
check_parent = parent_text_key not in \
1600
revision_keys[parent_id]
1602
# the parent parent_id is a ghost:
1603
check_parent = False
1604
# truncate the derived graph against this ghost.
1605
parent_text_key = None
1607
# look at the parent commit details inventories to
1608
# determine possible candidates in the per file graph.
1611
inv = inventory_cache[parent_id]
1613
inv = self.revision_tree(parent_id).inventory
1614
inventory_cache[parent_id] = inv
1615
parent_entry = inv._byid.get(text_key[0], None)
1616
if parent_entry is not None:
1618
text_key[0], parent_entry.revision)
1620
parent_text_key = None
1621
if parent_text_key is not None:
1622
candidate_parents.append(
1623
text_key_cache[parent_text_key])
1624
parent_heads = text_graph.heads(candidate_parents)
1625
new_parents = list(parent_heads)
1626
new_parents.sort(key=lambda x:candidate_parents.index(x))
1627
if new_parents == []:
1628
new_parents = [NULL_REVISION]
1629
text_index[text_key] = new_parents
1631
for text_key in invalid_keys:
1632
text_index[text_key] = [NULL_REVISION]
1635
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1636
"""Get an iterable listing the keys of all the data introduced by a set
1639
The keys will be ordered so that the corresponding items can be safely
1640
fetched and inserted in that order.
1642
:returns: An iterable producing tuples of (knit-kind, file-id,
1643
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1644
'revisions'. file-id is None unless knit-kind is 'file'.
1646
# XXX: it's a bit weird to control the inventory weave caching in this
1647
# generator. Ideally the caching would be done in fetch.py I think. Or
1648
# maybe this generator should explicitly have the contract that it
1649
# should not be iterated until the previously yielded item has been
1651
inv_w = self.inventories
1653
# file ids that changed
1654
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1656
num_file_ids = len(file_ids)
1657
for file_id, altered_versions in file_ids.iteritems():
1658
if _files_pb is not None:
1659
_files_pb.update("fetch texts", count, num_file_ids)
1661
yield ("file", file_id, altered_versions)
1662
# We're done with the files_pb. Note that it finished by the caller,
1663
# just as it was created by the caller.
1667
yield ("inventory", None, revision_ids)
1670
# XXX: Note ATM no callers actually pay attention to this return
1671
# instead they just use the list of revision ids and ignore
1672
# missing sigs. Consider removing this work entirely
1673
revisions_with_signatures = set(self.signatures.get_parent_map(
1674
[(r,) for r in revision_ids]))
1675
revisions_with_signatures = set(
1676
[r for (r,) in revisions_with_signatures])
1677
revisions_with_signatures.intersection_update(revision_ids)
1678
yield ("signatures", None, revisions_with_signatures)
1681
yield ("revisions", None, revision_ids)
1684
def get_inventory(self, revision_id):
1685
"""Get Inventory object by revision id."""
1686
return self.iter_inventories([revision_id]).next()
1688
def iter_inventories(self, revision_ids):
1689
"""Get many inventories by revision_ids.
1691
This will buffer some or all of the texts used in constructing the
1692
inventories in memory, but will only parse a single inventory at a
1695
:return: An iterator of inventories.
1697
if ((None in revision_ids)
1698
or (_mod_revision.NULL_REVISION in revision_ids)):
1699
raise ValueError('cannot get null revision inventory')
1700
return self._iter_inventories(revision_ids)
1702
def _iter_inventories(self, revision_ids):
1703
"""single-document based inventory iteration."""
1704
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1705
yield self.deserialise_inventory(revision_id, text)
1707
def _iter_inventory_xmls(self, revision_ids):
1708
keys = [(revision_id,) for revision_id in revision_ids]
1709
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1711
for record in stream:
1712
if record.storage_kind != 'absent':
1713
text_chunks[record.key] = record.get_bytes_as('chunked')
1715
raise errors.NoSuchRevision(self, record.key)
1717
chunks = text_chunks.pop(key)
1718
yield ''.join(chunks), key[-1]
1720
def deserialise_inventory(self, revision_id, xml):
1721
"""Transform the xml into an inventory object.
1723
:param revision_id: The expected revision id of the inventory.
1724
:param xml: A serialised inventory.
1726
result = self._serializer.read_inventory_from_string(xml, revision_id,
1727
entry_cache=self._inventory_entry_cache)
1728
if result.revision_id != revision_id:
1729
raise AssertionError('revision id mismatch %s != %s' % (
1730
result.revision_id, revision_id))
1733
def serialise_inventory(self, inv):
1734
return self._serializer.write_inventory_to_string(inv)
1736
def _serialise_inventory_to_lines(self, inv):
1737
return self._serializer.write_inventory_to_lines(inv)
1739
def get_serializer_format(self):
1740
return self._serializer.format_num
1743
def get_inventory_xml(self, revision_id):
1744
"""Get inventory XML as a file object."""
1745
texts = self._iter_inventory_xmls([revision_id])
1747
text, revision_id = texts.next()
1748
except StopIteration:
1749
raise errors.HistoryMissing(self, 'inventory', revision_id)
1753
def get_inventory_sha1(self, revision_id):
1754
"""Return the sha1 hash of the inventory entry
1756
return self.get_revision(revision_id).inventory_sha1
1758
def iter_reverse_revision_history(self, revision_id):
1759
"""Iterate backwards through revision ids in the lefthand history
1761
:param revision_id: The revision id to start with. All its lefthand
1762
ancestors will be traversed.
1764
graph = self.get_graph()
1765
next_id = revision_id
1767
if next_id in (None, _mod_revision.NULL_REVISION):
1770
# Note: The following line may raise KeyError in the event of
1771
# truncated history. We decided not to have a try:except:raise
1772
# RevisionNotPresent here until we see a use for it, because of the
1773
# cost in an inner loop that is by its very nature O(history).
1774
# Robert Collins 20080326
1775
parents = graph.get_parent_map([next_id])[next_id]
1776
if len(parents) == 0:
1779
next_id = parents[0]
1782
def get_revision_inventory(self, revision_id):
1783
"""Return inventory of a past revision."""
1784
# TODO: Unify this with get_inventory()
1785
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1786
# must be the same as its revision, so this is trivial.
1787
if revision_id is None:
1788
# This does not make sense: if there is no revision,
1789
# then it is the current tree inventory surely ?!
1790
# and thus get_root_id() is something that looks at the last
1791
# commit on the branch, and the get_root_id is an inventory check.
1792
raise NotImplementedError
1793
# return Inventory(self.get_root_id())
1795
return self.get_inventory(revision_id)
1797
def is_shared(self):
1798
"""Return True if this repository is flagged as a shared repository."""
1799
raise NotImplementedError(self.is_shared)
1802
def reconcile(self, other=None, thorough=False):
1803
"""Reconcile this repository."""
1804
from bzrlib.reconcile import RepoReconciler
1805
reconciler = RepoReconciler(self, thorough=thorough)
1806
reconciler.reconcile()
1809
def _refresh_data(self):
1810
"""Helper called from lock_* to ensure coherency with disk.
1812
The default implementation does nothing; it is however possible
1813
for repositories to maintain loaded indices across multiple locks
1814
by checking inside their implementation of this method to see
1815
whether their indices are still valid. This depends of course on
1816
the disk format being validatable in this manner.
1820
def revision_tree(self, revision_id):
1821
"""Return Tree for a revision on this branch.
1823
`revision_id` may be NULL_REVISION for the empty tree revision.
1825
revision_id = _mod_revision.ensure_null(revision_id)
1826
# TODO: refactor this to use an existing revision object
1827
# so we don't need to read it in twice.
1828
if revision_id == _mod_revision.NULL_REVISION:
1829
return RevisionTree(self, Inventory(root_id=None),
1830
_mod_revision.NULL_REVISION)
1832
inv = self.get_revision_inventory(revision_id)
1833
return RevisionTree(self, inv, revision_id)
1835
def revision_trees(self, revision_ids):
1836
"""Return Tree for a revision on this branch.
1838
`revision_id` may not be None or 'null:'"""
1839
inventories = self.iter_inventories(revision_ids)
1840
for inv in inventories:
1841
yield RevisionTree(self, inv, inv.revision_id)
1844
def get_ancestry(self, revision_id, topo_sorted=True):
1845
"""Return a list of revision-ids integrated by a revision.
1847
The first element of the list is always None, indicating the origin
1848
revision. This might change when we have history horizons, or
1849
perhaps we should have a new API.
1851
This is topologically sorted.
1853
if _mod_revision.is_null(revision_id):
1855
if not self.has_revision(revision_id):
1856
raise errors.NoSuchRevision(self, revision_id)
1857
graph = self.get_graph()
1859
search = graph._make_breadth_first_searcher([revision_id])
1862
found, ghosts = search.next_with_ghosts()
1863
except StopIteration:
1866
if _mod_revision.NULL_REVISION in keys:
1867
keys.remove(_mod_revision.NULL_REVISION)
1869
parent_map = graph.get_parent_map(keys)
1870
keys = tsort.topo_sort(parent_map)
1871
return [None] + list(keys)
1874
"""Compress the data within the repository.
1876
This operation only makes sense for some repository types. For other
1877
types it should be a no-op that just returns.
1879
This stub method does not require a lock, but subclasses should use
1880
@needs_write_lock as this is a long running call its reasonable to
1881
implicitly lock for the user.
1885
@deprecated_method(one_six)
1886
def print_file(self, file, revision_id):
1887
"""Print `file` to stdout.
1889
FIXME RBC 20060125 as John Meinel points out this is a bad api
1890
- it writes to stdout, it assumes that that is valid etc. Fix
1891
by creating a new more flexible convenience function.
1893
tree = self.revision_tree(revision_id)
1894
# use inventory as it was in that revision
1895
file_id = tree.inventory.path2id(file)
1897
# TODO: jam 20060427 Write a test for this code path
1898
# it had a bug in it, and was raising the wrong
1900
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1901
tree.print_file(file_id)
1903
def get_transaction(self):
1904
return self.control_files.get_transaction()
1906
@deprecated_method(one_one)
1907
def get_parents(self, revision_ids):
1908
"""See StackedParentsProvider.get_parents"""
1909
parent_map = self.get_parent_map(revision_ids)
1910
return [parent_map.get(r, None) for r in revision_ids]
1912
def get_parent_map(self, revision_ids):
1913
"""See graph._StackedParentsProvider.get_parent_map"""
1914
# revisions index works in keys; this just works in revisions
1915
# therefore wrap and unwrap
1918
for revision_id in revision_ids:
1919
if revision_id == _mod_revision.NULL_REVISION:
1920
result[revision_id] = ()
1921
elif revision_id is None:
1922
raise ValueError('get_parent_map(None) is not valid')
1924
query_keys.append((revision_id ,))
1925
for ((revision_id,), parent_keys) in \
1926
self.revisions.get_parent_map(query_keys).iteritems():
1928
result[revision_id] = tuple(parent_revid
1929
for (parent_revid,) in parent_keys)
1931
result[revision_id] = (_mod_revision.NULL_REVISION,)
1934
def _make_parents_provider(self):
1937
def get_graph(self, other_repository=None):
1938
"""Return the graph walker for this repository format"""
1939
parents_provider = self._make_parents_provider()
1940
if (other_repository is not None and
1941
not self.has_same_location(other_repository)):
1942
parents_provider = graph._StackedParentsProvider(
1943
[parents_provider, other_repository._make_parents_provider()])
1944
return graph.Graph(parents_provider)
1946
def _get_versioned_file_checker(self):
1947
"""Return an object suitable for checking versioned files."""
1948
return _VersionedFileChecker(self)
1950
def revision_ids_to_search_result(self, result_set):
1951
"""Convert a set of revision ids to a graph SearchResult."""
1952
result_parents = set()
1953
for parents in self.get_graph().get_parent_map(
1954
result_set).itervalues():
1955
result_parents.update(parents)
1956
included_keys = result_set.intersection(result_parents)
1957
start_keys = result_set.difference(included_keys)
1958
exclude_keys = result_parents.difference(result_set)
1959
result = graph.SearchResult(start_keys, exclude_keys,
1960
len(result_set), result_set)
1964
def set_make_working_trees(self, new_value):
1965
"""Set the policy flag for making working trees when creating branches.
1967
This only applies to branches that use this repository.
1969
The default is 'True'.
1970
:param new_value: True to restore the default, False to disable making
1973
raise NotImplementedError(self.set_make_working_trees)
1975
def make_working_trees(self):
1976
"""Returns the policy for making working trees on new branches."""
1977
raise NotImplementedError(self.make_working_trees)
1980
def sign_revision(self, revision_id, gpg_strategy):
1981
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1982
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1985
def has_signature_for_revision_id(self, revision_id):
1986
"""Query for a revision signature for revision_id in the repository."""
1987
if not self.has_revision(revision_id):
1988
raise errors.NoSuchRevision(self, revision_id)
1989
sig_present = (1 == len(
1990
self.signatures.get_parent_map([(revision_id,)])))
1994
def get_signature_text(self, revision_id):
1995
"""Return the text for a signature."""
1996
stream = self.signatures.get_record_stream([(revision_id,)],
1998
record = stream.next()
1999
if record.storage_kind == 'absent':
2000
raise errors.NoSuchRevision(self, revision_id)
2001
return record.get_bytes_as('fulltext')
2004
def check(self, revision_ids=None):
2005
"""Check consistency of all history of given revision_ids.
2007
Different repository implementations should override _check().
2009
:param revision_ids: A non-empty list of revision_ids whose ancestry
2010
will be checked. Typically the last revision_id of a branch.
2012
return self._check(revision_ids)
2014
def _check(self, revision_ids):
2015
result = check.Check(self)
2019
def _warn_if_deprecated(self):
2020
global _deprecation_warning_done
2021
if _deprecation_warning_done:
2023
_deprecation_warning_done = True
2024
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2025
% (self._format, self.bzrdir.transport.base))
2027
def supports_rich_root(self):
2028
return self._format.rich_root_data
2030
def _check_ascii_revisionid(self, revision_id, method):
2031
"""Private helper for ascii-only repositories."""
2032
# weave repositories refuse to store revisionids that are non-ascii.
2033
if revision_id is not None:
2034
# weaves require ascii revision ids.
2035
if isinstance(revision_id, unicode):
2037
revision_id.encode('ascii')
2038
except UnicodeEncodeError:
2039
raise errors.NonAsciiRevisionId(method, self)
2042
revision_id.decode('ascii')
2043
except UnicodeDecodeError:
2044
raise errors.NonAsciiRevisionId(method, self)
2046
def revision_graph_can_have_wrong_parents(self):
2047
"""Is it possible for this repository to have a revision graph with
2050
If True, then this repository must also implement
2051
_find_inconsistent_revision_parents so that check and reconcile can
2052
check for inconsistencies before proceeding with other checks that may
2053
depend on the revision index being consistent.
2055
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2058
# remove these delegates a while after bzr 0.15
2059
def __make_delegated(name, from_module):
2060
def _deprecated_repository_forwarder():
2061
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2062
% (name, from_module),
2065
m = __import__(from_module, globals(), locals(), [name])
2067
return getattr(m, name)
2068
except AttributeError:
2069
raise AttributeError('module %s has no name %s'
2071
globals()[name] = _deprecated_repository_forwarder
2074
'AllInOneRepository',
2075
'WeaveMetaDirRepository',
2076
'PreSplitOutRepositoryFormat',
2077
'RepositoryFormat4',
2078
'RepositoryFormat5',
2079
'RepositoryFormat6',
2080
'RepositoryFormat7',
2082
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2086
'RepositoryFormatKnit',
2087
'RepositoryFormatKnit1',
2089
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2092
def install_revision(repository, rev, revision_tree):
2093
"""Install all revision data into a repository."""
2094
install_revisions(repository, [(rev, revision_tree, None)])
2097
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2098
"""Install all revision data into a repository.
2100
Accepts an iterable of revision, tree, signature tuples. The signature
2103
repository.start_write_group()
2105
for n, (revision, revision_tree, signature) in enumerate(iterable):
2106
_install_revision(repository, revision, revision_tree, signature)
2108
pb.update('Transferring revisions', n + 1, num_revisions)
2110
repository.abort_write_group()
2113
repository.commit_write_group()
2116
def _install_revision(repository, rev, revision_tree, signature):
2117
"""Install all revision data into a repository."""
2118
present_parents = []
2120
for p_id in rev.parent_ids:
2121
if repository.has_revision(p_id):
2122
present_parents.append(p_id)
2123
parent_trees[p_id] = repository.revision_tree(p_id)
2125
parent_trees[p_id] = repository.revision_tree(
2126
_mod_revision.NULL_REVISION)
2128
inv = revision_tree.inventory
2129
entries = inv.iter_entries()
2130
# backwards compatibility hack: skip the root id.
2131
if not repository.supports_rich_root():
2132
path, root = entries.next()
2133
if root.revision != rev.revision_id:
2134
raise errors.IncompatibleRevision(repr(repository))
2136
for path, ie in entries:
2137
text_keys[(ie.file_id, ie.revision)] = ie
2138
text_parent_map = repository.texts.get_parent_map(text_keys)
2139
missing_texts = set(text_keys) - set(text_parent_map)
2140
# Add the texts that are not already present
2141
for text_key in missing_texts:
2142
ie = text_keys[text_key]
2144
# FIXME: TODO: The following loop overlaps/duplicates that done by
2145
# commit to determine parents. There is a latent/real bug here where
2146
# the parents inserted are not those commit would do - in particular
2147
# they are not filtered by heads(). RBC, AB
2148
for revision, tree in parent_trees.iteritems():
2149
if ie.file_id not in tree:
2151
parent_id = tree.inventory[ie.file_id].revision
2152
if parent_id in text_parents:
2154
text_parents.append((ie.file_id, parent_id))
2155
lines = revision_tree.get_file(ie.file_id).readlines()
2156
repository.texts.add_lines(text_key, text_parents, lines)
2158
# install the inventory
2159
repository.add_inventory(rev.revision_id, inv, present_parents)
2160
except errors.RevisionAlreadyPresent:
2162
if signature is not None:
2163
repository.add_signature_text(rev.revision_id, signature)
2164
repository.add_revision(rev.revision_id, rev, inv)
2167
class MetaDirRepository(Repository):
2168
"""Repositories in the new meta-dir layout.
2170
:ivar _transport: Transport for access to repository control files,
2171
typically pointing to .bzr/repository.
2174
def __init__(self, _format, a_bzrdir, control_files):
2175
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2176
self._transport = control_files._transport
2178
def is_shared(self):
2179
"""Return True if this repository is flagged as a shared repository."""
2180
return self._transport.has('shared-storage')
2183
def set_make_working_trees(self, new_value):
2184
"""Set the policy flag for making working trees when creating branches.
2186
This only applies to branches that use this repository.
2188
The default is 'True'.
2189
:param new_value: True to restore the default, False to disable making
2194
self._transport.delete('no-working-trees')
2195
except errors.NoSuchFile:
2198
self._transport.put_bytes('no-working-trees', '',
2199
mode=self.bzrdir._get_file_mode())
2201
def make_working_trees(self):
2202
"""Returns the policy for making working trees on new branches."""
2203
return not self._transport.has('no-working-trees')
2206
class MetaDirVersionedFileRepository(MetaDirRepository):
2207
"""Repositories in a meta-dir, that work via versioned file objects."""
2209
def __init__(self, _format, a_bzrdir, control_files):
2210
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2214
network_format_registry = registry.FormatRegistry()
2215
"""Registry of formats indexed by their network name.
2217
The network name for a repository format is an identifier that can be used when
2218
referring to formats with smart server operations. See
2219
RepositoryFormat.network_name() for more detail.
2223
format_registry = registry.FormatRegistry(network_format_registry)
2224
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2226
This can contain either format instances themselves, or classes/factories that
2227
can be called to obtain one.
2231
#####################################################################
2232
# Repository Formats
2234
class RepositoryFormat(object):
2235
"""A repository format.
2237
Formats provide four things:
2238
* An initialization routine to construct repository data on disk.
2239
* a optional format string which is used when the BzrDir supports
2241
* an open routine which returns a Repository instance.
2242
* A network name for referring to the format in smart server RPC
2245
There is one and only one Format subclass for each on-disk format. But
2246
there can be one Repository subclass that is used for several different
2247
formats. The _format attribute on a Repository instance can be used to
2248
determine the disk format.
2250
Formats are placed in a registry by their format string for reference
2251
during opening. These should be subclasses of RepositoryFormat for
2254
Once a format is deprecated, just deprecate the initialize and open
2255
methods on the format class. Do not deprecate the object, as the
2256
object may be created even when a repository instnace hasn't been
2259
Common instance attributes:
2260
_matchingbzrdir - the bzrdir format that the repository format was
2261
originally written to work with. This can be used if manually
2262
constructing a bzrdir and repository, or more commonly for test suite
2266
# Set to True or False in derived classes. True indicates that the format
2267
# supports ghosts gracefully.
2268
supports_ghosts = None
2269
# Can this repository be given external locations to lookup additional
2270
# data. Set to True or False in derived classes.
2271
supports_external_lookups = None
2272
# What order should fetch operations request streams in?
2273
# The default is unordered as that is the cheapest for an origin to
2275
_fetch_order = 'unordered'
2276
# Does this repository format use deltas that can be fetched as-deltas ?
2277
# (E.g. knits, where the knit deltas can be transplanted intact.
2278
# We default to False, which will ensure that enough data to get
2279
# a full text out of any fetch stream will be grabbed.
2280
_fetch_uses_deltas = False
2281
# Should fetch trigger a reconcile after the fetch? Only needed for
2282
# some repository formats that can suffer internal inconsistencies.
2283
_fetch_reconcile = False
2286
return "<%s>" % self.__class__.__name__
2288
def __eq__(self, other):
2289
# format objects are generally stateless
2290
return isinstance(other, self.__class__)
2292
def __ne__(self, other):
2293
return not self == other
2296
def find_format(klass, a_bzrdir):
2297
"""Return the format for the repository object in a_bzrdir.
2299
This is used by bzr native formats that have a "format" file in
2300
the repository. Other methods may be used by different types of
2304
transport = a_bzrdir.get_repository_transport(None)
2305
format_string = transport.get("format").read()
2306
return format_registry.get(format_string)
2307
except errors.NoSuchFile:
2308
raise errors.NoRepositoryPresent(a_bzrdir)
2310
raise errors.UnknownFormatError(format=format_string,
2314
def register_format(klass, format):
2315
format_registry.register(format.get_format_string(), format)
2318
def unregister_format(klass, format):
2319
format_registry.remove(format.get_format_string())
2322
def get_default_format(klass):
2323
"""Return the current default format."""
2324
from bzrlib import bzrdir
2325
return bzrdir.format_registry.make_bzrdir('default').repository_format
2327
def get_format_string(self):
2328
"""Return the ASCII format string that identifies this format.
2330
Note that in pre format ?? repositories the format string is
2331
not permitted nor written to disk.
2333
raise NotImplementedError(self.get_format_string)
2335
def get_format_description(self):
2336
"""Return the short description for this format."""
2337
raise NotImplementedError(self.get_format_description)
2339
# TODO: this shouldn't be in the base class, it's specific to things that
2340
# use weaves or knits -- mbp 20070207
2341
def _get_versioned_file_store(self,
2346
versionedfile_class=None,
2347
versionedfile_kwargs={},
2349
if versionedfile_class is None:
2350
versionedfile_class = self._versionedfile_class
2351
weave_transport = control_files._transport.clone(name)
2352
dir_mode = control_files._dir_mode
2353
file_mode = control_files._file_mode
2354
return VersionedFileStore(weave_transport, prefixed=prefixed,
2356
file_mode=file_mode,
2357
versionedfile_class=versionedfile_class,
2358
versionedfile_kwargs=versionedfile_kwargs,
2361
def initialize(self, a_bzrdir, shared=False):
2362
"""Initialize a repository of this format in a_bzrdir.
2364
:param a_bzrdir: The bzrdir to put the new repository in it.
2365
:param shared: The repository should be initialized as a sharable one.
2366
:returns: The new repository object.
2368
This may raise UninitializableFormat if shared repository are not
2369
compatible the a_bzrdir.
2371
raise NotImplementedError(self.initialize)
2373
def is_supported(self):
2374
"""Is this format supported?
2376
Supported formats must be initializable and openable.
2377
Unsupported formats may not support initialization or committing or
2378
some other features depending on the reason for not being supported.
2382
def network_name(self):
2383
"""A simple byte string uniquely identifying this format for RPC calls.
2385
MetaDir repository formats use their disk format string to identify the
2386
repository over the wire. All in one formats such as bzr < 0.8, and
2387
foreign formats like svn/git and hg should use some marker which is
2388
unique and immutable.
2390
raise NotImplementedError(self.network_name)
2392
def check_conversion_target(self, target_format):
2393
raise NotImplementedError(self.check_conversion_target)
2395
def open(self, a_bzrdir, _found=False):
2396
"""Return an instance of this format for the bzrdir a_bzrdir.
2398
_found is a private parameter, do not use it.
2400
raise NotImplementedError(self.open)
2403
class MetaDirRepositoryFormat(RepositoryFormat):
2404
"""Common base class for the new repositories using the metadir layout."""
2406
rich_root_data = False
2407
supports_tree_reference = False
2408
supports_external_lookups = False
2411
def _matchingbzrdir(self):
2412
matching = bzrdir.BzrDirMetaFormat1()
2413
matching.repository_format = self
2417
super(MetaDirRepositoryFormat, self).__init__()
2419
def _create_control_files(self, a_bzrdir):
2420
"""Create the required files and the initial control_files object."""
2421
# FIXME: RBC 20060125 don't peek under the covers
2422
# NB: no need to escape relative paths that are url safe.
2423
repository_transport = a_bzrdir.get_repository_transport(self)
2424
control_files = lockable_files.LockableFiles(repository_transport,
2425
'lock', lockdir.LockDir)
2426
control_files.create_lock()
2427
return control_files
2429
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2430
"""Upload the initial blank content."""
2431
control_files = self._create_control_files(a_bzrdir)
2432
control_files.lock_write()
2433
transport = control_files._transport
2435
utf8_files += [('shared-storage', '')]
2437
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2438
for (filename, content_stream) in files:
2439
transport.put_file(filename, content_stream,
2440
mode=a_bzrdir._get_file_mode())
2441
for (filename, content_bytes) in utf8_files:
2442
transport.put_bytes_non_atomic(filename, content_bytes,
2443
mode=a_bzrdir._get_file_mode())
2445
control_files.unlock()
2447
def network_name(self):
2448
"""Metadir formats have matching disk and network format strings."""
2449
return self.get_format_string()
2452
# Pre-0.8 formats that don't have a disk format string (because they are
2453
# versioned by the matching control directory). We use the control directories
2454
# disk format string as a key for the network_name because they meet the
2455
# constraints (simple string, unique, immmutable).
2456
network_format_registry.register_lazy(
2457
"Bazaar-NG branch, format 5\n",
2458
'bzrlib.repofmt.weaverepo',
2459
'RepositoryFormat5',
2461
network_format_registry.register_lazy(
2462
"Bazaar-NG branch, format 6\n",
2463
'bzrlib.repofmt.weaverepo',
2464
'RepositoryFormat6',
2467
# formats which have no format string are not discoverable or independently
2468
# creatable on disk, so are not registered in format_registry. They're
2469
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2470
# needed, it's constructed directly by the BzrDir. Non-native formats where
2471
# the repository is not separately opened are similar.
2473
format_registry.register_lazy(
2474
'Bazaar-NG Repository format 7',
2475
'bzrlib.repofmt.weaverepo',
2479
format_registry.register_lazy(
2480
'Bazaar-NG Knit Repository Format 1',
2481
'bzrlib.repofmt.knitrepo',
2482
'RepositoryFormatKnit1',
2485
format_registry.register_lazy(
2486
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2487
'bzrlib.repofmt.knitrepo',
2488
'RepositoryFormatKnit3',
2491
format_registry.register_lazy(
2492
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2493
'bzrlib.repofmt.knitrepo',
2494
'RepositoryFormatKnit4',
2497
# Pack-based formats. There is one format for pre-subtrees, and one for
2498
# post-subtrees to allow ease of testing.
2499
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2500
format_registry.register_lazy(
2501
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2502
'bzrlib.repofmt.pack_repo',
2503
'RepositoryFormatKnitPack1',
2505
format_registry.register_lazy(
2506
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2507
'bzrlib.repofmt.pack_repo',
2508
'RepositoryFormatKnitPack3',
2510
format_registry.register_lazy(
2511
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2512
'bzrlib.repofmt.pack_repo',
2513
'RepositoryFormatKnitPack4',
2515
format_registry.register_lazy(
2516
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2517
'bzrlib.repofmt.pack_repo',
2518
'RepositoryFormatKnitPack5',
2520
format_registry.register_lazy(
2521
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2522
'bzrlib.repofmt.pack_repo',
2523
'RepositoryFormatKnitPack5RichRoot',
2525
format_registry.register_lazy(
2526
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2527
'bzrlib.repofmt.pack_repo',
2528
'RepositoryFormatKnitPack5RichRootBroken',
2530
format_registry.register_lazy(
2531
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2532
'bzrlib.repofmt.pack_repo',
2533
'RepositoryFormatKnitPack6',
2535
format_registry.register_lazy(
2536
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2537
'bzrlib.repofmt.pack_repo',
2538
'RepositoryFormatKnitPack6RichRoot',
2541
# Development formats.
2542
# 1.7->1.8 go below here
2543
format_registry.register_lazy(
2544
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2545
'bzrlib.repofmt.pack_repo',
2546
'RepositoryFormatPackDevelopment2',
2548
format_registry.register_lazy(
2549
("Bazaar development format 2 with subtree support "
2550
"(needs bzr.dev from before 1.8)\n"),
2551
'bzrlib.repofmt.pack_repo',
2552
'RepositoryFormatPackDevelopment2Subtree',
2556
class InterRepository(InterObject):
2557
"""This class represents operations taking place between two repositories.
2559
Its instances have methods like copy_content and fetch, and contain
2560
references to the source and target repositories these operations can be
2563
Often we will provide convenience methods on 'repository' which carry out
2564
operations with another repository - they will always forward to
2565
InterRepository.get(other).method_name(parameters).
2568
_walk_to_common_revisions_batch_size = 1
2570
"""The available optimised InterRepository types."""
2572
def __init__(self, source, target):
2573
InterObject.__init__(self, source, target)
2574
# These two attributes may be overridden by e.g. InterOtherToRemote to
2575
# provide a faster implementation.
2576
self.target_get_graph = self.target.get_graph
2577
self.target_get_parent_map = self.target.get_parent_map
2580
def copy_content(self, revision_id=None):
2581
"""Make a complete copy of the content in self into destination.
2583
This is a destructive operation! Do not use it on existing
2586
:param revision_id: Only copy the content needed to construct
2587
revision_id and its parents.
2590
self.target.set_make_working_trees(self.source.make_working_trees())
2591
except NotImplementedError:
2593
self.target.fetch(self.source, revision_id=revision_id)
2595
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2597
"""Fetch the content required to construct revision_id.
2599
The content is copied from self.source to self.target.
2601
:param revision_id: if None all content is copied, if NULL_REVISION no
2603
:param pb: optional progress bar to use for progress reports. If not
2604
provided a default one will be created.
2607
from bzrlib.fetch import RepoFetcher
2608
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2609
self.source, self.source._format, self.target,
2610
self.target._format)
2611
f = RepoFetcher(to_repository=self.target,
2612
from_repository=self.source,
2613
last_revision=revision_id,
2614
fetch_spec=fetch_spec,
2615
pb=pb, find_ghosts=find_ghosts)
2617
def _walk_to_common_revisions(self, revision_ids):
2618
"""Walk out from revision_ids in source to revisions target has.
2620
:param revision_ids: The start point for the search.
2621
:return: A set of revision ids.
2623
target_graph = self.target_get_graph()
2624
revision_ids = frozenset(revision_ids)
2625
# Fast path for the case where all the revisions are already in the
2627
# (Although this does incur an extra round trip for the
2628
# fairly common case where the target doesn't already have the revision
2630
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2631
return graph.SearchResult(revision_ids, set(), 0, set())
2632
missing_revs = set()
2633
source_graph = self.source.get_graph()
2634
# ensure we don't pay silly lookup costs.
2635
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2636
null_set = frozenset([_mod_revision.NULL_REVISION])
2637
searcher_exhausted = False
2641
# Iterate the searcher until we have enough next_revs
2642
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2644
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2645
next_revs.update(next_revs_part)
2646
ghosts.update(ghosts_part)
2647
except StopIteration:
2648
searcher_exhausted = True
2650
# If there are ghosts in the source graph, and the caller asked for
2651
# them, make sure that they are present in the target.
2652
# We don't care about other ghosts as we can't fetch them and
2653
# haven't been asked to.
2654
ghosts_to_check = set(revision_ids.intersection(ghosts))
2655
revs_to_get = set(next_revs).union(ghosts_to_check)
2657
have_revs = set(target_graph.get_parent_map(revs_to_get))
2658
# we always have NULL_REVISION present.
2659
have_revs = have_revs.union(null_set)
2660
# Check if the target is missing any ghosts we need.
2661
ghosts_to_check.difference_update(have_revs)
2663
# One of the caller's revision_ids is a ghost in both the
2664
# source and the target.
2665
raise errors.NoSuchRevision(
2666
self.source, ghosts_to_check.pop())
2667
missing_revs.update(next_revs - have_revs)
2668
# Because we may have walked past the original stop point, make
2669
# sure everything is stopped
2670
stop_revs = searcher.find_seen_ancestors(have_revs)
2671
searcher.stop_searching_any(stop_revs)
2672
if searcher_exhausted:
2674
return searcher.get_result()
2676
@deprecated_method(one_two)
2678
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2679
"""Return the revision ids that source has that target does not.
2681
These are returned in topological order.
2683
:param revision_id: only return revision ids included by this
2685
:param find_ghosts: If True find missing revisions in deep history
2686
rather than just finding the surface difference.
2688
return list(self.search_missing_revision_ids(
2689
revision_id, find_ghosts).get_keys())
2692
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2693
"""Return the revision ids that source has that target does not.
2695
:param revision_id: only return revision ids included by this
2697
:param find_ghosts: If True find missing revisions in deep history
2698
rather than just finding the surface difference.
2699
:return: A bzrlib.graph.SearchResult.
2701
# stop searching at found target revisions.
2702
if not find_ghosts and revision_id is not None:
2703
return self._walk_to_common_revisions([revision_id])
2704
# generic, possibly worst case, slow code path.
2705
target_ids = set(self.target.all_revision_ids())
2706
if revision_id is not None:
2707
source_ids = self.source.get_ancestry(revision_id)
2708
if source_ids[0] is not None:
2709
raise AssertionError()
2712
source_ids = self.source.all_revision_ids()
2713
result_set = set(source_ids).difference(target_ids)
2714
return self.source.revision_ids_to_search_result(result_set)
2717
def _same_model(source, target):
2718
"""True if source and target have the same data representation.
2720
Note: this is always called on the base class; overriding it in a
2721
subclass will have no effect.
2724
InterRepository._assert_same_model(source, target)
2726
except errors.IncompatibleRepositories, e:
2730
def _assert_same_model(source, target):
2731
"""Raise an exception if two repositories do not use the same model.
2733
if source.supports_rich_root() != target.supports_rich_root():
2734
raise errors.IncompatibleRepositories(source, target,
2735
"different rich-root support")
2736
if source._serializer != target._serializer:
2737
raise errors.IncompatibleRepositories(source, target,
2738
"different serializers")
2741
class InterSameDataRepository(InterRepository):
2742
"""Code for converting between repositories that represent the same data.
2744
Data format and model must match for this to work.
2748
def _get_repo_format_to_test(self):
2749
"""Repository format for testing with.
2751
InterSameData can pull from subtree to subtree and from non-subtree to
2752
non-subtree, so we test this with the richest repository format.
2754
from bzrlib.repofmt import knitrepo
2755
return knitrepo.RepositoryFormatKnit3()
2758
def is_compatible(source, target):
2759
return InterRepository._same_model(source, target)
2762
class InterWeaveRepo(InterSameDataRepository):
2763
"""Optimised code paths between Weave based repositories.
2765
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2766
implemented lazy inter-object optimisation.
2770
def _get_repo_format_to_test(self):
2771
from bzrlib.repofmt import weaverepo
2772
return weaverepo.RepositoryFormat7()
2775
def is_compatible(source, target):
2776
"""Be compatible with known Weave formats.
2778
We don't test for the stores being of specific types because that
2779
could lead to confusing results, and there is no need to be
2782
from bzrlib.repofmt.weaverepo import (
2788
return (isinstance(source._format, (RepositoryFormat5,
2790
RepositoryFormat7)) and
2791
isinstance(target._format, (RepositoryFormat5,
2793
RepositoryFormat7)))
2794
except AttributeError:
2798
def copy_content(self, revision_id=None):
2799
"""See InterRepository.copy_content()."""
2800
# weave specific optimised path:
2802
self.target.set_make_working_trees(self.source.make_working_trees())
2803
except (errors.RepositoryUpgradeRequired, NotImplemented):
2805
# FIXME do not peek!
2806
if self.source._transport.listable():
2807
pb = ui.ui_factory.nested_progress_bar()
2809
self.target.texts.insert_record_stream(
2810
self.source.texts.get_record_stream(
2811
self.source.texts.keys(), 'topological', False))
2812
pb.update('copying inventory', 0, 1)
2813
self.target.inventories.insert_record_stream(
2814
self.source.inventories.get_record_stream(
2815
self.source.inventories.keys(), 'topological', False))
2816
self.target.signatures.insert_record_stream(
2817
self.source.signatures.get_record_stream(
2818
self.source.signatures.keys(),
2820
self.target.revisions.insert_record_stream(
2821
self.source.revisions.get_record_stream(
2822
self.source.revisions.keys(),
2823
'topological', True))
2827
self.target.fetch(self.source, revision_id=revision_id)
2830
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2832
"""See InterRepository.fetch()."""
2833
from bzrlib.fetch import RepoFetcher
2834
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2835
self.source, self.source._format, self.target, self.target._format)
2836
f = RepoFetcher(to_repository=self.target,
2837
from_repository=self.source,
2838
last_revision=revision_id,
2839
fetch_spec=fetch_spec,
2840
pb=pb, find_ghosts=find_ghosts)
2843
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2844
"""See InterRepository.missing_revision_ids()."""
2845
# we want all revisions to satisfy revision_id in source.
2846
# but we don't want to stat every file here and there.
2847
# we want then, all revisions other needs to satisfy revision_id
2848
# checked, but not those that we have locally.
2849
# so the first thing is to get a subset of the revisions to
2850
# satisfy revision_id in source, and then eliminate those that
2851
# we do already have.
2852
# this is slow on high latency connection to self, but as as this
2853
# disk format scales terribly for push anyway due to rewriting
2854
# inventory.weave, this is considered acceptable.
2856
if revision_id is not None:
2857
source_ids = self.source.get_ancestry(revision_id)
2858
if source_ids[0] is not None:
2859
raise AssertionError()
2862
source_ids = self.source._all_possible_ids()
2863
source_ids_set = set(source_ids)
2864
# source_ids is the worst possible case we may need to pull.
2865
# now we want to filter source_ids against what we actually
2866
# have in target, but don't try to check for existence where we know
2867
# we do not have a revision as that would be pointless.
2868
target_ids = set(self.target._all_possible_ids())
2869
possibly_present_revisions = target_ids.intersection(source_ids_set)
2870
actually_present_revisions = set(
2871
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2872
required_revisions = source_ids_set.difference(actually_present_revisions)
2873
if revision_id is not None:
2874
# we used get_ancestry to determine source_ids then we are assured all
2875
# revisions referenced are present as they are installed in topological order.
2876
# and the tip revision was validated by get_ancestry.
2877
result_set = required_revisions
2879
# if we just grabbed the possibly available ids, then
2880
# we only have an estimate of whats available and need to validate
2881
# that against the revision records.
2883
self.source._eliminate_revisions_not_present(required_revisions))
2884
return self.source.revision_ids_to_search_result(result_set)
2887
class InterKnitRepo(InterSameDataRepository):
2888
"""Optimised code paths between Knit based repositories."""
2891
def _get_repo_format_to_test(self):
2892
from bzrlib.repofmt import knitrepo
2893
return knitrepo.RepositoryFormatKnit1()
2896
def is_compatible(source, target):
2897
"""Be compatible with known Knit formats.
2899
We don't test for the stores being of specific types because that
2900
could lead to confusing results, and there is no need to be
2903
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2905
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2906
isinstance(target._format, RepositoryFormatKnit))
2907
except AttributeError:
2909
return are_knits and InterRepository._same_model(source, target)
2912
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2914
"""See InterRepository.fetch()."""
2915
from bzrlib.fetch import RepoFetcher
2916
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2917
self.source, self.source._format, self.target, self.target._format)
2918
f = RepoFetcher(to_repository=self.target,
2919
from_repository=self.source,
2920
last_revision=revision_id,
2921
fetch_spec=fetch_spec,
2922
pb=pb, find_ghosts=find_ghosts)
2925
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2926
"""See InterRepository.missing_revision_ids()."""
2927
if revision_id is not None:
2928
source_ids = self.source.get_ancestry(revision_id)
2929
if source_ids[0] is not None:
2930
raise AssertionError()
2933
source_ids = self.source.all_revision_ids()
2934
source_ids_set = set(source_ids)
2935
# source_ids is the worst possible case we may need to pull.
2936
# now we want to filter source_ids against what we actually
2937
# have in target, but don't try to check for existence where we know
2938
# we do not have a revision as that would be pointless.
2939
target_ids = set(self.target.all_revision_ids())
2940
possibly_present_revisions = target_ids.intersection(source_ids_set)
2941
actually_present_revisions = set(
2942
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2943
required_revisions = source_ids_set.difference(actually_present_revisions)
2944
if revision_id is not None:
2945
# we used get_ancestry to determine source_ids then we are assured all
2946
# revisions referenced are present as they are installed in topological order.
2947
# and the tip revision was validated by get_ancestry.
2948
result_set = required_revisions
2950
# if we just grabbed the possibly available ids, then
2951
# we only have an estimate of whats available and need to validate
2952
# that against the revision records.
2954
self.source._eliminate_revisions_not_present(required_revisions))
2955
return self.source.revision_ids_to_search_result(result_set)
2958
class InterPackRepo(InterSameDataRepository):
2959
"""Optimised code paths between Pack based repositories."""
2962
def _get_repo_format_to_test(self):
2963
from bzrlib.repofmt import pack_repo
2964
return pack_repo.RepositoryFormatKnitPack1()
2967
def is_compatible(source, target):
2968
"""Be compatible with known Pack formats.
2970
We don't test for the stores being of specific types because that
2971
could lead to confusing results, and there is no need to be
2974
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2976
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2977
isinstance(target._format, RepositoryFormatPack))
2978
except AttributeError:
2980
return are_packs and InterRepository._same_model(source, target)
2983
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2985
"""See InterRepository.fetch()."""
2986
if (len(self.source._fallback_repositories) > 0 or
2987
len(self.target._fallback_repositories) > 0):
2988
# The pack layer is not aware of fallback repositories, so when
2989
# fetching from a stacked repository or into a stacked repository
2990
# we use the generic fetch logic which uses the VersionedFiles
2991
# attributes on repository.
2992
from bzrlib.fetch import RepoFetcher
2993
fetcher = RepoFetcher(self.target, self.source, revision_id,
2994
pb, find_ghosts, fetch_spec=fetch_spec)
2995
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2996
self.source, self.source._format, self.target, self.target._format)
2997
if fetch_spec is not None:
2998
if len(list(fetch_spec.heads)) != 1:
2999
raise AssertionError(
3000
"InterPackRepo.fetch doesn't support "
3001
"fetching multiple heads yet.")
3002
revision_id = fetch_spec.heads[0]
3004
if revision_id is None:
3006
# everything to do - use pack logic
3007
# to fetch from all packs to one without
3008
# inventory parsing etc, IFF nothing to be copied is in the target.
3010
source_revision_ids = frozenset(self.source.all_revision_ids())
3011
revision_ids = source_revision_ids - \
3012
frozenset(self.target_get_parent_map(source_revision_ids))
3013
revision_keys = [(revid,) for revid in revision_ids]
3014
target_pack_collection = self._get_target_pack_collection()
3015
index = target_pack_collection.revision_index.combined_index
3016
present_revision_ids = set(item[1][0] for item in
3017
index.iter_entries(revision_keys))
3018
revision_ids = set(revision_ids) - present_revision_ids
3019
# implementing the TODO will involve:
3020
# - detecting when all of a pack is selected
3021
# - avoiding as much as possible pre-selection, so the
3022
# more-core routines such as create_pack_from_packs can filter in
3023
# a just-in-time fashion. (though having a HEADS list on a
3024
# repository might make this a lot easier, because we could
3025
# sensibly detect 'new revisions' without doing a full index scan.
3026
elif _mod_revision.is_null(revision_id):
3031
revision_ids = self.search_missing_revision_ids(revision_id,
3032
find_ghosts=find_ghosts).get_keys()
3033
except errors.NoSuchRevision:
3034
raise errors.InstallFailed([revision_id])
3035
if len(revision_ids) == 0:
3037
return self._pack(self.source, self.target, revision_ids)
3039
def _pack(self, source, target, revision_ids):
3040
from bzrlib.repofmt.pack_repo import Packer
3041
target_pack_collection = self._get_target_pack_collection()
3042
packs = source._pack_collection.all_packs()
3043
pack = Packer(target_pack_collection, packs, '.fetch',
3044
revision_ids).pack()
3045
if pack is not None:
3046
target_pack_collection._save_pack_names()
3047
copied_revs = pack.get_revision_count()
3048
# Trigger an autopack. This may duplicate effort as we've just done
3049
# a pack creation, but for now it is simpler to think about as
3050
# 'upload data, then repack if needed'.
3052
return (copied_revs, [])
3056
def _autopack(self):
3057
self.target._pack_collection.autopack()
3059
def _get_target_pack_collection(self):
3060
return self.target._pack_collection
3063
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3064
"""See InterRepository.missing_revision_ids().
3066
:param find_ghosts: Find ghosts throughout the ancestry of
3069
if not find_ghosts and revision_id is not None:
3070
return self._walk_to_common_revisions([revision_id])
3071
elif revision_id is not None:
3072
# Find ghosts: search for revisions pointing from one repository to
3073
# the other, and vice versa, anywhere in the history of revision_id.
3074
graph = self.target_get_graph(other_repository=self.source)
3075
searcher = graph._make_breadth_first_searcher([revision_id])
3079
next_revs, ghosts = searcher.next_with_ghosts()
3080
except StopIteration:
3082
if revision_id in ghosts:
3083
raise errors.NoSuchRevision(self.source, revision_id)
3084
found_ids.update(next_revs)
3085
found_ids.update(ghosts)
3086
found_ids = frozenset(found_ids)
3087
# Double query here: should be able to avoid this by changing the
3088
# graph api further.
3089
result_set = found_ids - frozenset(
3090
self.target_get_parent_map(found_ids))
3092
source_ids = self.source.all_revision_ids()
3093
# source_ids is the worst possible case we may need to pull.
3094
# now we want to filter source_ids against what we actually
3095
# have in target, but don't try to check for existence where we know
3096
# we do not have a revision as that would be pointless.
3097
target_ids = set(self.target.all_revision_ids())
3098
result_set = set(source_ids).difference(target_ids)
3099
return self.source.revision_ids_to_search_result(result_set)
3102
class InterDifferingSerializer(InterKnitRepo):
3105
def _get_repo_format_to_test(self):
3109
def is_compatible(source, target):
3110
"""Be compatible with Knit2 source and Knit3 target"""
3111
if source.supports_rich_root() != target.supports_rich_root():
3113
# Ideally, we'd support fetching if the source had no tree references
3114
# even if it supported them...
3115
if (getattr(source, '_format.supports_tree_reference', False) and
3116
not getattr(target, '_format.supports_tree_reference', False)):
3120
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3121
"""Get the best delta and base for this revision.
3123
:return: (basis_id, delta)
3125
possible_trees = [(parent_id, cache[parent_id])
3126
for parent_id in parent_ids
3127
if parent_id in cache]
3128
if len(possible_trees) == 0:
3129
# There either aren't any parents, or the parents aren't in the
3130
# cache, so just use the last converted tree
3131
possible_trees.append((basis_id, cache[basis_id]))
3133
for basis_id, basis_tree in possible_trees:
3134
delta = tree.inventory._make_delta(basis_tree.inventory)
3135
deltas.append((len(delta), basis_id, delta))
3137
return deltas[0][1:]
3139
def _fetch_batch(self, revision_ids, basis_id, cache):
3140
"""Fetch across a few revisions.
3142
:param revision_ids: The revisions to copy
3143
:param basis_id: The revision_id of a tree that must be in cache, used
3144
as a basis for delta when no other base is available
3145
:param cache: A cache of RevisionTrees that we can use.
3146
:return: The revision_id of the last converted tree. The RevisionTree
3147
for it will be in cache
3149
# Walk though all revisions; get inventory deltas, copy referenced
3150
# texts that delta references, insert the delta, revision and
3154
pending_revisions = []
3155
parent_map = self.source.get_parent_map(revision_ids)
3156
for tree in self.source.revision_trees(revision_ids):
3157
current_revision_id = tree.get_revision_id()
3158
parent_ids = parent_map.get(current_revision_id, ())
3159
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3161
# Find text entries that need to be copied
3162
for old_path, new_path, file_id, entry in delta:
3163
if new_path is not None:
3164
if not (new_path or self.target.supports_rich_root()):
3165
# We don't copy the text for the root node unless the
3166
# target supports_rich_root.
3169
# "if entry.revision == current_revision_id" ?
3170
if entry.revision == current_revision_id:
3171
text_keys.add((file_id, entry.revision))
3172
revision = self.source.get_revision(current_revision_id)
3173
pending_deltas.append((basis_id, delta,
3174
current_revision_id, revision.parent_ids))
3175
pending_revisions.append(revision)
3176
cache[current_revision_id] = tree
3177
basis_id = current_revision_id
3179
from_texts = self.source.texts
3180
to_texts = self.target.texts
3181
to_texts.insert_record_stream(from_texts.get_record_stream(
3182
text_keys, self.target._format._fetch_order,
3183
not self.target._format._fetch_uses_deltas))
3185
for delta in pending_deltas:
3186
self.target.add_inventory_by_delta(*delta)
3187
# insert signatures and revisions
3188
for revision in pending_revisions:
3190
signature = self.source.get_signature_text(
3191
revision.revision_id)
3192
self.target.add_signature_text(revision.revision_id,
3194
except errors.NoSuchRevision:
3196
self.target.add_revision(revision.revision_id, revision)
3199
def _fetch_all_revisions(self, revision_ids, pb):
3200
"""Fetch everything for the list of revisions.
3202
:param revision_ids: The list of revisions to fetch. Must be in
3204
:param pb: A ProgressBar
3207
basis_id, basis_tree = self._get_basis(revision_ids[0])
3209
cache = lru_cache.LRUCache(100)
3210
cache[basis_id] = basis_tree
3211
del basis_tree # We don't want to hang on to it here
3212
for offset in range(0, len(revision_ids), batch_size):
3213
self.target.start_write_group()
3215
pb.update('Transferring revisions', offset,
3217
batch = revision_ids[offset:offset+batch_size]
3218
basis_id = self._fetch_batch(batch, basis_id, cache)
3220
self.target.abort_write_group()
3223
self.target.commit_write_group()
3224
pb.update('Transferring revisions', len(revision_ids),
3228
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3230
"""See InterRepository.fetch()."""
3231
if fetch_spec is not None:
3232
raise AssertionError("Not implemented yet...")
3233
revision_ids = self.target.search_missing_revision_ids(self.source,
3234
revision_id, find_ghosts=find_ghosts).get_keys()
3235
if not revision_ids:
3237
revision_ids = tsort.topo_sort(
3238
self.source.get_graph().get_parent_map(revision_ids))
3240
my_pb = ui.ui_factory.nested_progress_bar()
3245
self._fetch_all_revisions(revision_ids, pb)
3247
if my_pb is not None:
3249
return len(revision_ids), 0
3251
def _get_basis(self, first_revision_id):
3252
"""Get a revision and tree which exists in the target.
3254
This assumes that first_revision_id is selected for transmission
3255
because all other ancestors are already present. If we can't find an
3256
ancestor we fall back to NULL_REVISION since we know that is safe.
3258
:return: (basis_id, basis_tree)
3260
first_rev = self.source.get_revision(first_revision_id)
3262
basis_id = first_rev.parent_ids[0]
3263
# only valid as a basis if the target has it
3264
self.target.get_revision(basis_id)
3265
# Try to get a basis tree - if its a ghost it will hit the
3266
# NoSuchRevision case.
3267
basis_tree = self.source.revision_tree(basis_id)
3268
except (IndexError, errors.NoSuchRevision):
3269
basis_id = _mod_revision.NULL_REVISION
3270
basis_tree = self.source.revision_tree(basis_id)
3271
return basis_id, basis_tree
3274
class InterOtherToRemote(InterRepository):
3275
"""An InterRepository that simply delegates to the 'real' InterRepository
3276
calculated for (source, target._real_repository).
3279
_walk_to_common_revisions_batch_size = 50
3281
def __init__(self, source, target):
3282
InterRepository.__init__(self, source, target)
3283
self._real_inter = None
3286
def is_compatible(source, target):
3287
if isinstance(target, remote.RemoteRepository):
3291
def _ensure_real_inter(self):
3292
if self._real_inter is None:
3293
self.target._ensure_real()
3294
real_target = self.target._real_repository
3295
self._real_inter = InterRepository.get(self.source, real_target)
3296
# Make _real_inter use the RemoteRepository for get_parent_map
3297
self._real_inter.target_get_graph = self.target.get_graph
3298
self._real_inter.target_get_parent_map = self.target.get_parent_map
3300
def copy_content(self, revision_id=None):
3301
self._ensure_real_inter()
3302
self._real_inter.copy_content(revision_id=revision_id)
3304
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3306
self._ensure_real_inter()
3307
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3308
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
3311
def _get_repo_format_to_test(self):
3315
class InterRemoteToOther(InterRepository):
3317
def __init__(self, source, target):
3318
InterRepository.__init__(self, source, target)
3319
self._real_inter = None
3322
def is_compatible(source, target):
3323
if not isinstance(source, remote.RemoteRepository):
3325
return InterRepository._same_model(source, target)
3327
def _ensure_real_inter(self):
3328
if self._real_inter is None:
3329
self.source._ensure_real()
3330
real_source = self.source._real_repository
3331
self._real_inter = InterRepository.get(real_source, self.target)
3334
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3336
"""See InterRepository.fetch()."""
3337
# Always fetch using the generic streaming fetch code, to allow
3338
# streaming fetching from remote servers.
3339
from bzrlib.fetch import RepoFetcher
3340
fetcher = RepoFetcher(self.target, self.source, revision_id,
3341
pb, find_ghosts, fetch_spec=fetch_spec)
3343
def copy_content(self, revision_id=None):
3344
self._ensure_real_inter()
3345
self._real_inter.copy_content(revision_id=revision_id)
3348
def _get_repo_format_to_test(self):
3353
class InterPackToRemotePack(InterPackRepo):
3354
"""A specialisation of InterPackRepo for a target that is a
3357
This will use the get_parent_map RPC rather than plain readvs, and also
3358
uses an RPC for autopacking.
3361
_walk_to_common_revisions_batch_size = 50
3364
def is_compatible(source, target):
3365
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3366
if isinstance(source._format, RepositoryFormatPack):
3367
if isinstance(target, remote.RemoteRepository):
3368
target._format._ensure_real()
3369
if isinstance(target._format._custom_format,
3370
RepositoryFormatPack):
3371
if InterRepository._same_model(source, target):
3375
def _autopack(self):
3376
self.target.autopack()
3379
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3381
"""See InterRepository.fetch()."""
3382
# Always fetch using the generic streaming fetch code, to allow
3383
# streaming fetching into remote servers.
3384
from bzrlib.fetch import RepoFetcher
3385
fetcher = RepoFetcher(self.target, self.source, revision_id,
3386
pb, find_ghosts, fetch_spec=fetch_spec)
3388
def _get_target_pack_collection(self):
3389
return self.target._real_repository._pack_collection
3392
def _get_repo_format_to_test(self):
3396
InterRepository.register_optimiser(InterDifferingSerializer)
3397
InterRepository.register_optimiser(InterSameDataRepository)
3398
InterRepository.register_optimiser(InterWeaveRepo)
3399
InterRepository.register_optimiser(InterKnitRepo)
3400
InterRepository.register_optimiser(InterPackRepo)
3401
InterRepository.register_optimiser(InterOtherToRemote)
3402
InterRepository.register_optimiser(InterRemoteToOther)
3403
InterRepository.register_optimiser(InterPackToRemotePack)
3406
class CopyConverter(object):
3407
"""A repository conversion tool which just performs a copy of the content.
3409
This is slow but quite reliable.
3412
def __init__(self, target_format):
3413
"""Create a CopyConverter.
3415
:param target_format: The format the resulting repository should be.
3417
self.target_format = target_format
3419
def convert(self, repo, pb):
3420
"""Perform the conversion of to_convert, giving feedback via pb.
3422
:param to_convert: The disk object to convert.
3423
:param pb: a progress bar to use for progress information.
3428
# this is only useful with metadir layouts - separated repo content.
3429
# trigger an assertion if not such
3430
repo._format.get_format_string()
3431
self.repo_dir = repo.bzrdir
3432
self.step('Moving repository to repository.backup')
3433
self.repo_dir.transport.move('repository', 'repository.backup')
3434
backup_transport = self.repo_dir.transport.clone('repository.backup')
3435
repo._format.check_conversion_target(self.target_format)
3436
self.source_repo = repo._format.open(self.repo_dir,
3438
_override_transport=backup_transport)
3439
self.step('Creating new repository')
3440
converted = self.target_format.initialize(self.repo_dir,
3441
self.source_repo.is_shared())
3442
converted.lock_write()
3444
self.step('Copying content into repository.')
3445
self.source_repo.copy_content_into(converted)
3448
self.step('Deleting old repository content.')
3449
self.repo_dir.transport.delete_tree('repository.backup')
3450
self.pb.note('repository converted')
3452
def step(self, message):
3453
"""Update the pb by a step."""
3455
self.pb.update(message, self.count, self.total)
3467
def _unescaper(match, _map=_unescape_map):
3468
code = match.group(1)
3472
if not code.startswith('#'):
3474
return unichr(int(code[1:])).encode('utf8')
3480
def _unescape_xml(data):
3481
"""Unescape predefined XML entities in a string of data."""
3483
if _unescape_re is None:
3484
_unescape_re = re.compile('\&([^;]*);')
3485
return _unescape_re.sub(_unescaper, data)
3488
class _VersionedFileChecker(object):
3490
def __init__(self, repository):
3491
self.repository = repository
3492
self.text_index = self.repository._generate_text_key_index()
3494
def calculate_file_version_parents(self, text_key):
3495
"""Calculate the correct parents for a file version according to
3498
parent_keys = self.text_index[text_key]
3499
if parent_keys == [_mod_revision.NULL_REVISION]:
3501
return tuple(parent_keys)
3503
def check_file_version_parents(self, texts, progress_bar=None):
3504
"""Check the parents stored in a versioned file are correct.
3506
It also detects file versions that are not referenced by their
3507
corresponding revision's inventory.
3509
:returns: A tuple of (wrong_parents, dangling_file_versions).
3510
wrong_parents is a dict mapping {revision_id: (stored_parents,
3511
correct_parents)} for each revision_id where the stored parents
3512
are not correct. dangling_file_versions is a set of (file_id,
3513
revision_id) tuples for versions that are present in this versioned
3514
file, but not used by the corresponding inventory.
3517
self.file_ids = set([file_id for file_id, _ in
3518
self.text_index.iterkeys()])
3519
# text keys is now grouped by file_id
3520
n_weaves = len(self.file_ids)
3521
files_in_revisions = {}
3522
revisions_of_files = {}
3523
n_versions = len(self.text_index)
3524
progress_bar.update('loading text store', 0, n_versions)
3525
parent_map = self.repository.texts.get_parent_map(self.text_index)
3526
# On unlistable transports this could well be empty/error...
3527
text_keys = self.repository.texts.keys()
3528
unused_keys = frozenset(text_keys) - set(self.text_index)
3529
for num, key in enumerate(self.text_index.iterkeys()):
3530
if progress_bar is not None:
3531
progress_bar.update('checking text graph', num, n_versions)
3532
correct_parents = self.calculate_file_version_parents(key)
3534
knit_parents = parent_map[key]
3535
except errors.RevisionNotPresent:
3538
if correct_parents != knit_parents:
3539
wrong_parents[key] = (knit_parents, correct_parents)
3540
return wrong_parents, unused_keys
3543
def _old_get_graph(repository, revision_id):
3544
"""DO NOT USE. That is all. I'm serious."""
3545
graph = repository.get_graph()
3546
revision_graph = dict(((key, value) for key, value in
3547
graph.iter_ancestry([revision_id]) if value is not None))
3548
return _strip_NULL_ghosts(revision_graph)
3551
def _strip_NULL_ghosts(revision_graph):
3552
"""Also don't use this. more compatibility code for unmigrated clients."""
3553
# Filter ghosts, and null:
3554
if _mod_revision.NULL_REVISION in revision_graph:
3555
del revision_graph[_mod_revision.NULL_REVISION]
3556
for key, parents in revision_graph.items():
3557
revision_graph[key] = tuple(parent for parent in parents if parent
3559
return revision_graph
3562
class StreamSink(object):
3563
"""An object that can insert a stream into a repository.
3565
This interface handles the complexity of reserialising inventories and
3566
revisions from different formats, and allows unidirectional insertion into
3567
stacked repositories without looking for the missing basis parents
3571
def __init__(self, target_repo):
3572
self.target_repo = target_repo
3574
def insert_stream(self, stream, src_format, resume_tokens):
3575
"""Insert a stream's content into the target repository.
3577
:param src_format: a bzr repository format.
3579
:return: a list of resume tokens and an iterable of keys additional
3580
items required before the insertion can be completed.
3582
self.target_repo.lock_write()
3585
self.target_repo.resume_write_group(resume_tokens)
3587
self.target_repo.start_write_group()
3589
# locked_insert_stream performs a commit|suspend.
3590
return self._locked_insert_stream(stream, src_format)
3592
self.target_repo.abort_write_group(suppress_errors=True)
3595
self.target_repo.unlock()
3597
def _locked_insert_stream(self, stream, src_format):
3598
to_serializer = self.target_repo._format._serializer
3599
src_serializer = src_format._serializer
3600
for substream_type, substream in stream:
3601
if substream_type == 'texts':
3602
self.target_repo.texts.insert_record_stream(substream)
3603
elif substream_type == 'inventories':
3604
if src_serializer == to_serializer:
3605
self.target_repo.inventories.insert_record_stream(
3608
self._extract_and_insert_inventories(
3609
substream, src_serializer)
3610
elif substream_type == 'revisions':
3611
# This may fallback to extract-and-insert more often than
3612
# required if the serializers are different only in terms of
3614
if src_serializer == to_serializer:
3615
self.target_repo.revisions.insert_record_stream(
3618
self._extract_and_insert_revisions(substream,
3620
elif substream_type == 'signatures':
3621
self.target_repo.signatures.insert_record_stream(substream)
3623
raise AssertionError('kaboom! %s' % (substream_type,))
3625
missing_keys = set()
3626
for prefix, versioned_file in (
3627
('texts', self.target_repo.texts),
3628
('inventories', self.target_repo.inventories),
3629
('revisions', self.target_repo.revisions),
3630
('signatures', self.target_repo.signatures),
3632
missing_keys.update((prefix,) + key for key in
3633
versioned_file.get_missing_compression_parent_keys())
3634
except NotImplementedError:
3635
# cannot even attempt suspending, and missing would have failed
3636
# during stream insertion.
3637
missing_keys = set()
3640
# suspend the write group and tell the caller what we is
3641
# missing. We know we can suspend or else we would not have
3642
# entered this code path. (All repositories that can handle
3643
# missing keys can handle suspending a write group).
3644
write_group_tokens = self.target_repo.suspend_write_group()
3645
return write_group_tokens, missing_keys
3646
self.target_repo.commit_write_group()
3649
def _extract_and_insert_inventories(self, substream, serializer):
3650
"""Generate a new inventory versionedfile in target, converting data.
3652
The inventory is retrieved from the source, (deserializing it), and
3653
stored in the target (reserializing it in a different format).
3655
for record in substream:
3656
bytes = record.get_bytes_as('fulltext')
3657
revision_id = record.key[0]
3658
inv = serializer.read_inventory_from_string(bytes, revision_id)
3659
parents = [key[0] for key in record.parents]
3660
self.target_repo.add_inventory(revision_id, inv, parents)
3662
def _extract_and_insert_revisions(self, substream, serializer):
3663
for record in substream:
3664
bytes = record.get_bytes_as('fulltext')
3665
revision_id = record.key[0]
3666
rev = serializer.read_revision_from_string(bytes)
3667
if rev.revision_id != revision_id:
3668
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
3669
self.target_repo.add_revision(revision_id, rev)
3672
if self.target_repo._format._fetch_reconcile:
3673
self.target_repo.reconcile()
3676
class StreamSource(object):
3677
"""A source of a stream for fetching between repositories."""
3679
def __init__(self, from_repository, to_format):
3680
"""Create a StreamSource streaming from from_repository."""
3681
self.from_repository = from_repository
3682
self.to_format = to_format
3684
def delta_on_metadata(self):
3685
"""Return True if delta's are permitted on metadata streams.
3687
That is on revisions and signatures.
3689
src_serializer = self.from_repository._format._serializer
3690
target_serializer = self.to_format._serializer
3691
return (self.to_format._fetch_uses_deltas and
3692
src_serializer == target_serializer)
3694
def _fetch_revision_texts(self, revs):
3695
# fetch signatures first and then the revision texts
3696
# may need to be a InterRevisionStore call here.
3697
from_sf = self.from_repository.signatures
3698
# A missing signature is just skipped.
3699
keys = [(rev_id,) for rev_id in revs]
3700
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
3702
self.to_format._fetch_order,
3703
not self.to_format._fetch_uses_deltas))
3704
# If a revision has a delta, this is actually expanded inside the
3705
# insert_record_stream code now, which is an alternate fix for
3707
from_rf = self.from_repository.revisions
3708
revisions = from_rf.get_record_stream(
3710
self.to_format._fetch_order,
3711
not self.delta_on_metadata())
3712
return [('signatures', signatures), ('revisions', revisions)]
3714
def _generate_root_texts(self, revs):
3715
"""This will be called by __fetch between fetching weave texts and
3716
fetching the inventory weave.
3718
Subclasses should override this if they need to generate root texts
3719
after fetching weave texts.
3721
if self._rich_root_upgrade():
3723
return bzrlib.fetch.Inter1and2Helper(
3724
self.from_repository).generate_root_texts(revs)
3728
def get_stream(self, search):
3730
revs = search.get_keys()
3731
graph = self.from_repository.get_graph()
3732
revs = list(graph.iter_topo_order(revs))
3733
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
3735
for knit_kind, file_id, revisions in data_to_fetch:
3736
if knit_kind != phase:
3738
# Make a new progress bar for this phase
3739
if knit_kind == "file":
3740
# Accumulate file texts
3741
text_keys.extend([(file_id, revision) for revision in
3743
elif knit_kind == "inventory":
3744
# Now copy the file texts.
3745
from_texts = self.from_repository.texts
3746
yield ('texts', from_texts.get_record_stream(
3747
text_keys, self.to_format._fetch_order,
3748
not self.to_format._fetch_uses_deltas))
3749
# Cause an error if a text occurs after we have done the
3752
# Before we process the inventory we generate the root
3753
# texts (if necessary) so that the inventories references
3755
for _ in self._generate_root_texts(revs):
3757
# NB: This currently reopens the inventory weave in source;
3758
# using a single stream interface instead would avoid this.
3759
from_weave = self.from_repository.inventories
3760
# we fetch only the referenced inventories because we do not
3761
# know for unselected inventories whether all their required
3762
# texts are present in the other repository - it could be
3764
yield ('inventories', from_weave.get_record_stream(
3765
[(rev_id,) for rev_id in revs],
3766
self.inventory_fetch_order(),
3767
not self.delta_on_metadata()))
3768
elif knit_kind == "signatures":
3769
# Nothing to do here; this will be taken care of when
3770
# _fetch_revision_texts happens.
3772
elif knit_kind == "revisions":
3773
for record in self._fetch_revision_texts(revs):
3776
raise AssertionError("Unknown knit kind %r" % knit_kind)
3778
def get_stream_for_missing_keys(self, missing_keys):
3779
# missing keys can only occur when we are byte copying and not
3780
# translating (because translation means we don't send
3781
# unreconstructable deltas ever).
3783
keys['texts'] = set()
3784
keys['revisions'] = set()
3785
keys['inventories'] = set()
3786
keys['signatures'] = set()
3787
for key in missing_keys:
3788
keys[key[0]].add(key[1:])
3789
if len(keys['revisions']):
3790
# If we allowed copying revisions at this point, we could end up
3791
# copying a revision without copying its required texts: a
3792
# violation of the requirements for repository integrity.
3793
raise AssertionError(
3794
'cannot copy revisions to fill in missing deltas %s' % (
3795
keys['revisions'],))
3796
for substream_kind, keys in keys.iteritems():
3797
vf = getattr(self.from_repository, substream_kind)
3798
# Ask for full texts always so that we don't need more round trips
3799
# after this stream.
3800
stream = vf.get_record_stream(keys,
3801
self.to_format._fetch_order, True)
3802
yield substream_kind, stream
3804
def inventory_fetch_order(self):
3805
if self._rich_root_upgrade():
3806
return 'topological'
3808
return self.to_format._fetch_order
3810
def _rich_root_upgrade(self):
3811
return (not self.from_repository._format.rich_root_data and
3812
self.to_format.rich_root_data)
added
removed
bzrlib/repository.py
