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)
640
self._fetch_order = 'topological'
642
def _check_fallback_repository(self, repository):
643
"""Check that this repository can fallback to repository safely.
645
Raise an error if not.
647
:param repository: A repository to fallback to.
649
return InterRepository._assert_same_model(self, repository)
651
def add_inventory(self, revision_id, inv, parents):
652
"""Add the inventory inv to the repository as revision_id.
654
:param parents: The revision ids of the parents that revision_id
655
is known to have and are in the repository already.
657
:returns: The validator(which is a sha1 digest, though what is sha'd is
658
repository format specific) of the serialized inventory.
660
if not self.is_in_write_group():
661
raise AssertionError("%r not in write group" % (self,))
662
_mod_revision.check_not_reserved_id(revision_id)
663
if not (inv.revision_id is None or inv.revision_id == revision_id):
664
raise AssertionError(
665
"Mismatch between inventory revision"
666
" id and insertion revid (%r, %r)"
667
% (inv.revision_id, revision_id))
669
raise AssertionError()
670
inv_lines = self._serialise_inventory_to_lines(inv)
671
return self._inventory_add_lines(revision_id, parents,
672
inv_lines, check_content=False)
674
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
676
"""Add a new inventory expressed as a delta against another revision.
678
:param basis_revision_id: The inventory id the delta was created
679
against. (This does not have to be a direct parent.)
680
:param delta: The inventory delta (see Inventory.apply_delta for
682
:param new_revision_id: The revision id that the inventory is being
684
:param parents: The revision ids of the parents that revision_id is
685
known to have and are in the repository already. These are supplied
686
for repositories that depend on the inventory graph for revision
687
graph access, as well as for those that pun ancestry with delta
690
:returns: (validator, new_inv)
691
The validator(which is a sha1 digest, though what is sha'd is
692
repository format specific) of the serialized inventory, and the
695
if not self.is_in_write_group():
696
raise AssertionError("%r not in write group" % (self,))
697
_mod_revision.check_not_reserved_id(new_revision_id)
698
basis_tree = self.revision_tree(basis_revision_id)
699
basis_tree.lock_read()
701
# Note that this mutates the inventory of basis_tree, which not all
702
# inventory implementations may support: A better idiom would be to
703
# return a new inventory, but as there is no revision tree cache in
704
# repository this is safe for now - RBC 20081013
705
basis_inv = basis_tree.inventory
706
basis_inv.apply_delta(delta)
707
basis_inv.revision_id = new_revision_id
708
return (self.add_inventory(new_revision_id, basis_inv, parents),
713
def _inventory_add_lines(self, revision_id, parents, lines,
715
"""Store lines in inv_vf and return the sha1 of the inventory."""
716
parents = [(parent,) for parent in parents]
717
return self.inventories.add_lines((revision_id,), parents, lines,
718
check_content=check_content)[0]
720
def add_revision(self, revision_id, rev, inv=None, config=None):
721
"""Add rev to the revision store as revision_id.
723
:param revision_id: the revision id to use.
724
:param rev: The revision object.
725
:param inv: The inventory for the revision. if None, it will be looked
726
up in the inventory storer
727
:param config: If None no digital signature will be created.
728
If supplied its signature_needed method will be used
729
to determine if a signature should be made.
731
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
733
_mod_revision.check_not_reserved_id(revision_id)
734
if config is not None and config.signature_needed():
736
inv = self.get_inventory(revision_id)
737
plaintext = Testament(rev, inv).as_short_text()
738
self.store_revision_signature(
739
gpg.GPGStrategy(config), plaintext, revision_id)
740
# check inventory present
741
if not self.inventories.get_parent_map([(revision_id,)]):
743
raise errors.WeaveRevisionNotPresent(revision_id,
746
# yes, this is not suitable for adding with ghosts.
747
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
751
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
752
self._add_revision(rev)
754
def _add_revision(self, revision):
755
text = self._serializer.write_revision_to_string(revision)
756
key = (revision.revision_id,)
757
parents = tuple((parent,) for parent in revision.parent_ids)
758
self.revisions.add_lines(key, parents, osutils.split_lines(text))
760
def all_revision_ids(self):
761
"""Returns a list of all the revision ids in the repository.
763
This is conceptually deprecated because code should generally work on
764
the graph reachable from a particular revision, and ignore any other
765
revisions that might be present. There is no direct replacement
768
if 'evil' in debug.debug_flags:
769
mutter_callsite(2, "all_revision_ids is linear with history.")
770
return self._all_revision_ids()
772
def _all_revision_ids(self):
773
"""Returns a list of all the revision ids in the repository.
775
These are in as much topological order as the underlying store can
778
raise NotImplementedError(self._all_revision_ids)
780
def break_lock(self):
781
"""Break a lock if one is present from another instance.
783
Uses the ui factory to ask for confirmation if the lock may be from
786
self.control_files.break_lock()
789
def _eliminate_revisions_not_present(self, revision_ids):
790
"""Check every revision id in revision_ids to see if we have it.
792
Returns a set of the present revisions.
795
graph = self.get_graph()
796
parent_map = graph.get_parent_map(revision_ids)
797
# The old API returned a list, should this actually be a set?
798
return parent_map.keys()
801
def create(a_bzrdir):
802
"""Construct the current default format repository in a_bzrdir."""
803
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
805
def __init__(self, _format, a_bzrdir, control_files):
806
"""instantiate a Repository.
808
:param _format: The format of the repository on disk.
809
:param a_bzrdir: The BzrDir of the repository.
811
In the future we will have a single api for all stores for
812
getting file texts, inventories and revisions, then
813
this construct will accept instances of those things.
815
super(Repository, self).__init__()
816
self._format = _format
817
# the following are part of the public API for Repository:
818
self.bzrdir = a_bzrdir
819
self.control_files = control_files
820
self._transport = control_files._transport
821
self.base = self._transport.base
823
self._reconcile_does_inventory_gc = True
824
self._reconcile_fixes_text_parents = False
825
self._reconcile_backsup_inventory = True
826
# not right yet - should be more semantically clear ?
828
# TODO: make sure to construct the right store classes, etc, depending
829
# on whether escaping is required.
830
self._warn_if_deprecated()
831
self._write_group = None
832
# Additional places to query for data.
833
self._fallback_repositories = []
834
# What order should fetch operations request streams in?
835
# The default is unordered as that is the cheapest for an origin to
837
self._fetch_order = 'unordered'
838
# Does this repository use deltas that can be fetched as-deltas ?
839
# (E.g. knits, where the knit deltas can be transplanted intact.
840
# We default to False, which will ensure that enough data to get
841
# a full text out of any fetch stream will be grabbed.
842
self._fetch_uses_deltas = False
843
# Should fetch trigger a reconcile after the fetch? Only needed for
844
# some repository formats that can suffer internal inconsistencies.
845
self._fetch_reconcile = False
846
# An InventoryEntry cache, used during deserialization
847
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
850
return '%s(%r)' % (self.__class__.__name__,
853
def has_same_location(self, other):
854
"""Returns a boolean indicating if this repository is at the same
855
location as another repository.
857
This might return False even when two repository objects are accessing
858
the same physical repository via different URLs.
860
if self.__class__ is not other.__class__:
862
return (self._transport.base == other._transport.base)
864
def is_in_write_group(self):
865
"""Return True if there is an open write group.
867
:seealso: start_write_group.
869
return self._write_group is not None
872
return self.control_files.is_locked()
874
def is_write_locked(self):
875
"""Return True if this object is write locked."""
876
return self.is_locked() and self.control_files._lock_mode == 'w'
878
def lock_write(self, token=None):
879
"""Lock this repository for writing.
881
This causes caching within the repository obejct to start accumlating
882
data during reads, and allows a 'write_group' to be obtained. Write
883
groups must be used for actual data insertion.
885
:param token: if this is already locked, then lock_write will fail
886
unless the token matches the existing lock.
887
:returns: a token if this instance supports tokens, otherwise None.
888
:raises TokenLockingNotSupported: when a token is given but this
889
instance doesn't support using token locks.
890
:raises MismatchedToken: if the specified token doesn't match the token
891
of the existing lock.
892
:seealso: start_write_group.
894
A token should be passed in if you know that you have locked the object
895
some other way, and need to synchronise this object's state with that
898
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
900
result = self.control_files.lock_write(token=token)
901
for repo in self._fallback_repositories:
902
# Writes don't affect fallback repos
908
self.control_files.lock_read()
909
for repo in self._fallback_repositories:
913
def get_physical_lock_status(self):
914
return self.control_files.get_physical_lock_status()
916
def leave_lock_in_place(self):
917
"""Tell this repository not to release the physical lock when this
920
If lock_write doesn't return a token, then this method is not supported.
922
self.control_files.leave_in_place()
924
def dont_leave_lock_in_place(self):
925
"""Tell this repository to release the physical lock when this
926
object is unlocked, even if it didn't originally acquire it.
928
If lock_write doesn't return a token, then this method is not supported.
930
self.control_files.dont_leave_in_place()
933
def gather_stats(self, revid=None, committers=None):
934
"""Gather statistics from a revision id.
936
:param revid: The revision id to gather statistics from, if None, then
937
no revision specific statistics are gathered.
938
:param committers: Optional parameter controlling whether to grab
939
a count of committers from the revision specific statistics.
940
:return: A dictionary of statistics. Currently this contains:
941
committers: The number of committers if requested.
942
firstrev: A tuple with timestamp, timezone for the penultimate left
943
most ancestor of revid, if revid is not the NULL_REVISION.
944
latestrev: A tuple with timestamp, timezone for revid, if revid is
945
not the NULL_REVISION.
946
revisions: The total revision count in the repository.
947
size: An estimate disk size of the repository in bytes.
950
if revid and committers:
951
result['committers'] = 0
952
if revid and revid != _mod_revision.NULL_REVISION:
954
all_committers = set()
955
revisions = self.get_ancestry(revid)
956
# pop the leading None
958
first_revision = None
960
# ignore the revisions in the middle - just grab first and last
961
revisions = revisions[0], revisions[-1]
962
for revision in self.get_revisions(revisions):
963
if not first_revision:
964
first_revision = revision
966
all_committers.add(revision.committer)
967
last_revision = revision
969
result['committers'] = len(all_committers)
970
result['firstrev'] = (first_revision.timestamp,
971
first_revision.timezone)
972
result['latestrev'] = (last_revision.timestamp,
973
last_revision.timezone)
975
# now gather global repository information
976
# XXX: This is available for many repos regardless of listability.
977
if self.bzrdir.root_transport.listable():
978
# XXX: do we want to __define len__() ?
979
# Maybe the versionedfiles object should provide a different
980
# method to get the number of keys.
981
result['revisions'] = len(self.revisions.keys())
985
def find_branches(self, using=False):
986
"""Find branches underneath this repository.
988
This will include branches inside other branches.
990
:param using: If True, list only branches using this repository.
992
if using and not self.is_shared():
994
return [self.bzrdir.open_branch()]
995
except errors.NotBranchError:
997
class Evaluator(object):
1000
self.first_call = True
1002
def __call__(self, bzrdir):
1003
# On the first call, the parameter is always the bzrdir
1004
# containing the current repo.
1005
if not self.first_call:
1007
repository = bzrdir.open_repository()
1008
except errors.NoRepositoryPresent:
1011
return False, (None, repository)
1012
self.first_call = False
1014
value = (bzrdir.open_branch(), None)
1015
except errors.NotBranchError:
1016
value = (None, None)
1020
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1021
self.bzrdir.root_transport, evaluate=Evaluator()):
1022
if branch is not None:
1023
branches.append(branch)
1024
if not using and repository is not None:
1025
branches.extend(repository.find_branches())
1029
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1030
"""Return the revision ids that other has that this does not.
1032
These are returned in topological order.
1034
revision_id: only return revision ids included by revision_id.
1036
return InterRepository.get(other, self).search_missing_revision_ids(
1037
revision_id, find_ghosts)
1039
@deprecated_method(one_two)
1041
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1042
"""Return the revision ids that other has that this does not.
1044
These are returned in topological order.
1046
revision_id: only return revision ids included by revision_id.
1048
keys = self.search_missing_revision_ids(
1049
other, revision_id, find_ghosts).get_keys()
1052
parents = other.get_graph().get_parent_map(keys)
1055
return tsort.topo_sort(parents)
1059
"""Open the repository rooted at base.
1061
For instance, if the repository is at URL/.bzr/repository,
1062
Repository.open(URL) -> a Repository instance.
1064
control = bzrdir.BzrDir.open(base)
1065
return control.open_repository()
1067
def copy_content_into(self, destination, revision_id=None):
1068
"""Make a complete copy of the content in self into destination.
1070
This is a destructive operation! Do not use it on existing
1073
return InterRepository.get(self, destination).copy_content(revision_id)
1075
def commit_write_group(self):
1076
"""Commit the contents accrued within the current write group.
1078
:seealso: start_write_group.
1080
if self._write_group is not self.get_transaction():
1081
# has an unlock or relock occured ?
1082
raise errors.BzrError('mismatched lock context %r and '
1084
(self.get_transaction(), self._write_group))
1085
self._commit_write_group()
1086
self._write_group = None
1088
def _commit_write_group(self):
1089
"""Template method for per-repository write group cleanup.
1091
This is called before the write group is considered to be
1092
finished and should ensure that all data handed to the repository
1093
for writing during the write group is safely committed (to the
1094
extent possible considering file system caching etc).
1097
def suspend_write_group(self):
1098
raise errors.UnsuspendableWriteGroup(self)
1100
def resume_write_group(self, tokens):
1101
if not self.is_write_locked():
1102
raise errors.NotWriteLocked(self)
1103
if self._write_group:
1104
raise errors.BzrError('already in a write group')
1105
self._resume_write_group(tokens)
1106
# so we can detect unlock/relock - the write group is now entered.
1107
self._write_group = self.get_transaction()
1109
def _resume_write_group(self, tokens):
1110
raise errors.UnsuspendableWriteGroup(self)
1112
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1113
"""Fetch the content required to construct revision_id from source.
1115
If revision_id is None all content is copied.
1116
:param find_ghosts: Find and copy revisions in the source that are
1117
ghosts in the target (and not reachable directly by walking out to
1118
the first-present revision in target from revision_id).
1120
# fast path same-url fetch operations
1121
if self.has_same_location(source):
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)
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)
1243
def has_revision(self, revision_id):
1244
"""True if this repository has a copy of the revision."""
1245
return revision_id in self.has_revisions((revision_id,))
1248
def has_revisions(self, revision_ids):
1249
"""Probe to find out the presence of multiple revisions.
1251
:param revision_ids: An iterable of revision_ids.
1252
:return: A set of the revision_ids that were present.
1254
parent_map = self.revisions.get_parent_map(
1255
[(rev_id,) for rev_id in revision_ids])
1257
if _mod_revision.NULL_REVISION in revision_ids:
1258
result.add(_mod_revision.NULL_REVISION)
1259
result.update([key[0] for key in parent_map])
1263
def get_revision(self, revision_id):
1264
"""Return the Revision object for a named revision."""
1265
return self.get_revisions([revision_id])[0]
1268
def get_revision_reconcile(self, revision_id):
1269
"""'reconcile' helper routine that allows access to a revision always.
1271
This variant of get_revision does not cross check the weave graph
1272
against the revision one as get_revision does: but it should only
1273
be used by reconcile, or reconcile-alike commands that are correcting
1274
or testing the revision graph.
1276
return self._get_revisions([revision_id])[0]
1279
def get_revisions(self, revision_ids):
1280
"""Get many revisions at once."""
1281
return self._get_revisions(revision_ids)
1284
def _get_revisions(self, revision_ids):
1285
"""Core work logic to get many revisions without sanity checks."""
1286
for rev_id in revision_ids:
1287
if not rev_id or not isinstance(rev_id, basestring):
1288
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1289
keys = [(key,) for key in revision_ids]
1290
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1292
for record in stream:
1293
if record.storage_kind == 'absent':
1294
raise errors.NoSuchRevision(self, record.key[0])
1295
text = record.get_bytes_as('fulltext')
1296
rev = self._serializer.read_revision_from_string(text)
1297
revs[record.key[0]] = rev
1298
return [revs[revid] for revid in revision_ids]
1301
def get_revision_xml(self, revision_id):
1302
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1303
# would have already do it.
1304
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1305
rev = self.get_revision(revision_id)
1306
rev_tmp = cStringIO.StringIO()
1307
# the current serializer..
1308
self._serializer.write_revision(rev, rev_tmp)
1310
return rev_tmp.getvalue()
1312
def get_deltas_for_revisions(self, revisions):
1313
"""Produce a generator of revision deltas.
1315
Note that the input is a sequence of REVISIONS, not revision_ids.
1316
Trees will be held in memory until the generator exits.
1317
Each delta is relative to the revision's lefthand predecessor.
1319
required_trees = set()
1320
for revision in revisions:
1321
required_trees.add(revision.revision_id)
1322
required_trees.update(revision.parent_ids[:1])
1323
trees = dict((t.get_revision_id(), t) for
1324
t in self.revision_trees(required_trees))
1325
for revision in revisions:
1326
if not revision.parent_ids:
1327
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1329
old_tree = trees[revision.parent_ids[0]]
1330
yield trees[revision.revision_id].changes_from(old_tree)
1333
def get_revision_delta(self, revision_id):
1334
"""Return the delta for one revision.
1336
The delta is relative to the left-hand predecessor of the
1339
r = self.get_revision(revision_id)
1340
return list(self.get_deltas_for_revisions([r]))[0]
1343
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1344
signature = gpg_strategy.sign(plaintext)
1345
self.add_signature_text(revision_id, signature)
1348
def add_signature_text(self, revision_id, signature):
1349
self.signatures.add_lines((revision_id,), (),
1350
osutils.split_lines(signature))
1352
def find_text_key_references(self):
1353
"""Find the text key references within the repository.
1355
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1356
revision_ids. Each altered file-ids has the exact revision_ids that
1357
altered it listed explicitly.
1358
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1359
to whether they were referred to by the inventory of the
1360
revision_id that they contain. The inventory texts from all present
1361
revision ids are assessed to generate this report.
1363
revision_keys = self.revisions.keys()
1364
w = self.inventories
1365
pb = ui.ui_factory.nested_progress_bar()
1367
return self._find_text_key_references_from_xml_inventory_lines(
1368
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1372
def _find_text_key_references_from_xml_inventory_lines(self,
1374
"""Core routine for extracting references to texts from inventories.
1376
This performs the translation of xml lines to revision ids.
1378
:param line_iterator: An iterator of lines, origin_version_id
1379
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1380
to whether they were referred to by the inventory of the
1381
revision_id that they contain. Note that if that revision_id was
1382
not part of the line_iterator's output then False will be given -
1383
even though it may actually refer to that key.
1385
if not self._serializer.support_altered_by_hack:
1386
raise AssertionError(
1387
"_find_text_key_references_from_xml_inventory_lines only "
1388
"supported for branches which store inventory as unnested xml"
1389
", not on %r" % self)
1392
# this code needs to read every new line in every inventory for the
1393
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1394
# not present in one of those inventories is unnecessary but not
1395
# harmful because we are filtering by the revision id marker in the
1396
# inventory lines : we only select file ids altered in one of those
1397
# revisions. We don't need to see all lines in the inventory because
1398
# only those added in an inventory in rev X can contain a revision=X
1400
unescape_revid_cache = {}
1401
unescape_fileid_cache = {}
1403
# jam 20061218 In a big fetch, this handles hundreds of thousands
1404
# of lines, so it has had a lot of inlining and optimizing done.
1405
# Sorry that it is a little bit messy.
1406
# Move several functions to be local variables, since this is a long
1408
search = self._file_ids_altered_regex.search
1409
unescape = _unescape_xml
1410
setdefault = result.setdefault
1411
for line, line_key in line_iterator:
1412
match = search(line)
1415
# One call to match.group() returning multiple items is quite a
1416
# bit faster than 2 calls to match.group() each returning 1
1417
file_id, revision_id = match.group('file_id', 'revision_id')
1419
# Inlining the cache lookups helps a lot when you make 170,000
1420
# lines and 350k ids, versus 8.4 unique ids.
1421
# Using a cache helps in 2 ways:
1422
# 1) Avoids unnecessary decoding calls
1423
# 2) Re-uses cached strings, which helps in future set and
1425
# (2) is enough that removing encoding entirely along with
1426
# the cache (so we are using plain strings) results in no
1427
# performance improvement.
1429
revision_id = unescape_revid_cache[revision_id]
1431
unescaped = unescape(revision_id)
1432
unescape_revid_cache[revision_id] = unescaped
1433
revision_id = unescaped
1435
# Note that unconditionally unescaping means that we deserialise
1436
# every fileid, which for general 'pull' is not great, but we don't
1437
# really want to have some many fulltexts that this matters anyway.
1440
file_id = unescape_fileid_cache[file_id]
1442
unescaped = unescape(file_id)
1443
unescape_fileid_cache[file_id] = unescaped
1446
key = (file_id, revision_id)
1447
setdefault(key, False)
1448
if revision_id == line_key[-1]:
1452
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1454
"""Helper routine for fileids_altered_by_revision_ids.
1456
This performs the translation of xml lines to revision ids.
1458
:param line_iterator: An iterator of lines, origin_version_id
1459
:param revision_ids: The revision ids to filter for. This should be a
1460
set or other type which supports efficient __contains__ lookups, as
1461
the revision id from each parsed line will be looked up in the
1462
revision_ids filter.
1463
:return: a dictionary mapping altered file-ids to an iterable of
1464
revision_ids. Each altered file-ids has the exact revision_ids that
1465
altered it listed explicitly.
1468
setdefault = result.setdefault
1470
self._find_text_key_references_from_xml_inventory_lines(
1471
line_iterator).iterkeys():
1472
# once data is all ensured-consistent; then this is
1473
# if revision_id == version_id
1474
if key[-1:] in revision_ids:
1475
setdefault(key[0], set()).add(key[-1])
1478
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1479
"""Find the file ids and versions affected by revisions.
1481
:param revisions: an iterable containing revision ids.
1482
:param _inv_weave: The inventory weave from this repository or None.
1483
If None, the inventory weave will be opened automatically.
1484
:return: a dictionary mapping altered file-ids to an iterable of
1485
revision_ids. Each altered file-ids has the exact revision_ids that
1486
altered it listed explicitly.
1488
selected_keys = set((revid,) for revid in revision_ids)
1489
w = _inv_weave or self.inventories
1490
pb = ui.ui_factory.nested_progress_bar()
1492
return self._find_file_ids_from_xml_inventory_lines(
1493
w.iter_lines_added_or_present_in_keys(
1494
selected_keys, pb=pb),
1499
def iter_files_bytes(self, desired_files):
1500
"""Iterate through file versions.
1502
Files will not necessarily be returned in the order they occur in
1503
desired_files. No specific order is guaranteed.
1505
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1506
value supplied by the caller as part of desired_files. It should
1507
uniquely identify the file version in the caller's context. (Examples:
1508
an index number or a TreeTransform trans_id.)
1510
bytes_iterator is an iterable of bytestrings for the file. The
1511
kind of iterable and length of the bytestrings are unspecified, but for
1512
this implementation, it is a list of bytes produced by
1513
VersionedFile.get_record_stream().
1515
:param desired_files: a list of (file_id, revision_id, identifier)
1519
for file_id, revision_id, callable_data in desired_files:
1520
text_keys[(file_id, revision_id)] = callable_data
1521
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1522
if record.storage_kind == 'absent':
1523
raise errors.RevisionNotPresent(record.key, self)
1524
yield text_keys[record.key], record.get_bytes_as('fulltext')
1526
def _generate_text_key_index(self, text_key_references=None,
1528
"""Generate a new text key index for the repository.
1530
This is an expensive function that will take considerable time to run.
1532
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1533
list of parents, also text keys. When a given key has no parents,
1534
the parents list will be [NULL_REVISION].
1536
# All revisions, to find inventory parents.
1537
if ancestors is None:
1538
graph = self.get_graph()
1539
ancestors = graph.get_parent_map(self.all_revision_ids())
1540
if text_key_references is None:
1541
text_key_references = self.find_text_key_references()
1542
pb = ui.ui_factory.nested_progress_bar()
1544
return self._do_generate_text_key_index(ancestors,
1545
text_key_references, pb)
1549
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1550
"""Helper for _generate_text_key_index to avoid deep nesting."""
1551
revision_order = tsort.topo_sort(ancestors)
1552
invalid_keys = set()
1554
for revision_id in revision_order:
1555
revision_keys[revision_id] = set()
1556
text_count = len(text_key_references)
1557
# a cache of the text keys to allow reuse; costs a dict of all the
1558
# keys, but saves a 2-tuple for every child of a given key.
1560
for text_key, valid in text_key_references.iteritems():
1562
invalid_keys.add(text_key)
1564
revision_keys[text_key[1]].add(text_key)
1565
text_key_cache[text_key] = text_key
1566
del text_key_references
1568
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1569
NULL_REVISION = _mod_revision.NULL_REVISION
1570
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1571
# too small for large or very branchy trees. However, for 55K path
1572
# trees, it would be easy to use too much memory trivially. Ideally we
1573
# could gauge this by looking at available real memory etc, but this is
1574
# always a tricky proposition.
1575
inventory_cache = lru_cache.LRUCache(10)
1576
batch_size = 10 # should be ~150MB on a 55K path tree
1577
batch_count = len(revision_order) / batch_size + 1
1579
pb.update("Calculating text parents.", processed_texts, text_count)
1580
for offset in xrange(batch_count):
1581
to_query = revision_order[offset * batch_size:(offset + 1) *
1585
for rev_tree in self.revision_trees(to_query):
1586
revision_id = rev_tree.get_revision_id()
1587
parent_ids = ancestors[revision_id]
1588
for text_key in revision_keys[revision_id]:
1589
pb.update("Calculating text parents.", processed_texts)
1590
processed_texts += 1
1591
candidate_parents = []
1592
for parent_id in parent_ids:
1593
parent_text_key = (text_key[0], parent_id)
1595
check_parent = parent_text_key not in \
1596
revision_keys[parent_id]
1598
# the parent parent_id is a ghost:
1599
check_parent = False
1600
# truncate the derived graph against this ghost.
1601
parent_text_key = None
1603
# look at the parent commit details inventories to
1604
# determine possible candidates in the per file graph.
1607
inv = inventory_cache[parent_id]
1609
inv = self.revision_tree(parent_id).inventory
1610
inventory_cache[parent_id] = inv
1611
parent_entry = inv._byid.get(text_key[0], None)
1612
if parent_entry is not None:
1614
text_key[0], parent_entry.revision)
1616
parent_text_key = None
1617
if parent_text_key is not None:
1618
candidate_parents.append(
1619
text_key_cache[parent_text_key])
1620
parent_heads = text_graph.heads(candidate_parents)
1621
new_parents = list(parent_heads)
1622
new_parents.sort(key=lambda x:candidate_parents.index(x))
1623
if new_parents == []:
1624
new_parents = [NULL_REVISION]
1625
text_index[text_key] = new_parents
1627
for text_key in invalid_keys:
1628
text_index[text_key] = [NULL_REVISION]
1631
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1632
"""Get an iterable listing the keys of all the data introduced by a set
1635
The keys will be ordered so that the corresponding items can be safely
1636
fetched and inserted in that order.
1638
:returns: An iterable producing tuples of (knit-kind, file-id,
1639
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1640
'revisions'. file-id is None unless knit-kind is 'file'.
1642
# XXX: it's a bit weird to control the inventory weave caching in this
1643
# generator. Ideally the caching would be done in fetch.py I think. Or
1644
# maybe this generator should explicitly have the contract that it
1645
# should not be iterated until the previously yielded item has been
1647
inv_w = self.inventories
1649
# file ids that changed
1650
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1652
num_file_ids = len(file_ids)
1653
for file_id, altered_versions in file_ids.iteritems():
1654
if _files_pb is not None:
1655
_files_pb.update("fetch texts", count, num_file_ids)
1657
yield ("file", file_id, altered_versions)
1658
# We're done with the files_pb. Note that it finished by the caller,
1659
# just as it was created by the caller.
1663
yield ("inventory", None, revision_ids)
1666
# XXX: Note ATM no callers actually pay attention to this return
1667
# instead they just use the list of revision ids and ignore
1668
# missing sigs. Consider removing this work entirely
1669
revisions_with_signatures = set(self.signatures.get_parent_map(
1670
[(r,) for r in revision_ids]))
1671
revisions_with_signatures = set(
1672
[r for (r,) in revisions_with_signatures])
1673
revisions_with_signatures.intersection_update(revision_ids)
1674
yield ("signatures", None, revisions_with_signatures)
1677
yield ("revisions", None, revision_ids)
1680
def get_inventory(self, revision_id):
1681
"""Get Inventory object by revision id."""
1682
return self.iter_inventories([revision_id]).next()
1684
def iter_inventories(self, revision_ids):
1685
"""Get many inventories by revision_ids.
1687
This will buffer some or all of the texts used in constructing the
1688
inventories in memory, but will only parse a single inventory at a
1691
:return: An iterator of inventories.
1693
if ((None in revision_ids)
1694
or (_mod_revision.NULL_REVISION in revision_ids)):
1695
raise ValueError('cannot get null revision inventory')
1696
return self._iter_inventories(revision_ids)
1698
def _iter_inventories(self, revision_ids):
1699
"""single-document based inventory iteration."""
1700
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1701
yield self.deserialise_inventory(revision_id, text)
1703
def _iter_inventory_xmls(self, revision_ids):
1704
keys = [(revision_id,) for revision_id in revision_ids]
1705
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1707
for record in stream:
1708
if record.storage_kind != 'absent':
1709
text_chunks[record.key] = record.get_bytes_as('chunked')
1711
raise errors.NoSuchRevision(self, record.key)
1713
chunks = text_chunks.pop(key)
1714
yield ''.join(chunks), key[-1]
1716
def deserialise_inventory(self, revision_id, xml):
1717
"""Transform the xml into an inventory object.
1719
:param revision_id: The expected revision id of the inventory.
1720
:param xml: A serialised inventory.
1722
result = self._serializer.read_inventory_from_string(xml, revision_id,
1723
entry_cache=self._inventory_entry_cache)
1724
if result.revision_id != revision_id:
1725
raise AssertionError('revision id mismatch %s != %s' % (
1726
result.revision_id, revision_id))
1729
def serialise_inventory(self, inv):
1730
return self._serializer.write_inventory_to_string(inv)
1732
def _serialise_inventory_to_lines(self, inv):
1733
return self._serializer.write_inventory_to_lines(inv)
1735
def get_serializer_format(self):
1736
return self._serializer.format_num
1739
def get_inventory_xml(self, revision_id):
1740
"""Get inventory XML as a file object."""
1741
texts = self._iter_inventory_xmls([revision_id])
1743
text, revision_id = texts.next()
1744
except StopIteration:
1745
raise errors.HistoryMissing(self, 'inventory', revision_id)
1749
def get_inventory_sha1(self, revision_id):
1750
"""Return the sha1 hash of the inventory entry
1752
return self.get_revision(revision_id).inventory_sha1
1754
def iter_reverse_revision_history(self, revision_id):
1755
"""Iterate backwards through revision ids in the lefthand history
1757
:param revision_id: The revision id to start with. All its lefthand
1758
ancestors will be traversed.
1760
graph = self.get_graph()
1761
next_id = revision_id
1763
if next_id in (None, _mod_revision.NULL_REVISION):
1766
# Note: The following line may raise KeyError in the event of
1767
# truncated history. We decided not to have a try:except:raise
1768
# RevisionNotPresent here until we see a use for it, because of the
1769
# cost in an inner loop that is by its very nature O(history).
1770
# Robert Collins 20080326
1771
parents = graph.get_parent_map([next_id])[next_id]
1772
if len(parents) == 0:
1775
next_id = parents[0]
1778
def get_revision_inventory(self, revision_id):
1779
"""Return inventory of a past revision."""
1780
# TODO: Unify this with get_inventory()
1781
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1782
# must be the same as its revision, so this is trivial.
1783
if revision_id is None:
1784
# This does not make sense: if there is no revision,
1785
# then it is the current tree inventory surely ?!
1786
# and thus get_root_id() is something that looks at the last
1787
# commit on the branch, and the get_root_id is an inventory check.
1788
raise NotImplementedError
1789
# return Inventory(self.get_root_id())
1791
return self.get_inventory(revision_id)
1793
def is_shared(self):
1794
"""Return True if this repository is flagged as a shared repository."""
1795
raise NotImplementedError(self.is_shared)
1798
def reconcile(self, other=None, thorough=False):
1799
"""Reconcile this repository."""
1800
from bzrlib.reconcile import RepoReconciler
1801
reconciler = RepoReconciler(self, thorough=thorough)
1802
reconciler.reconcile()
1805
def _refresh_data(self):
1806
"""Helper called from lock_* to ensure coherency with disk.
1808
The default implementation does nothing; it is however possible
1809
for repositories to maintain loaded indices across multiple locks
1810
by checking inside their implementation of this method to see
1811
whether their indices are still valid. This depends of course on
1812
the disk format being validatable in this manner.
1816
def revision_tree(self, revision_id):
1817
"""Return Tree for a revision on this branch.
1819
`revision_id` may be NULL_REVISION for the empty tree revision.
1821
revision_id = _mod_revision.ensure_null(revision_id)
1822
# TODO: refactor this to use an existing revision object
1823
# so we don't need to read it in twice.
1824
if revision_id == _mod_revision.NULL_REVISION:
1825
return RevisionTree(self, Inventory(root_id=None),
1826
_mod_revision.NULL_REVISION)
1828
inv = self.get_revision_inventory(revision_id)
1829
return RevisionTree(self, inv, revision_id)
1831
def revision_trees(self, revision_ids):
1832
"""Return Tree for a revision on this branch.
1834
`revision_id` may not be None or 'null:'"""
1835
inventories = self.iter_inventories(revision_ids)
1836
for inv in inventories:
1837
yield RevisionTree(self, inv, inv.revision_id)
1840
def get_ancestry(self, revision_id, topo_sorted=True):
1841
"""Return a list of revision-ids integrated by a revision.
1843
The first element of the list is always None, indicating the origin
1844
revision. This might change when we have history horizons, or
1845
perhaps we should have a new API.
1847
This is topologically sorted.
1849
if _mod_revision.is_null(revision_id):
1851
if not self.has_revision(revision_id):
1852
raise errors.NoSuchRevision(self, revision_id)
1853
graph = self.get_graph()
1855
search = graph._make_breadth_first_searcher([revision_id])
1858
found, ghosts = search.next_with_ghosts()
1859
except StopIteration:
1862
if _mod_revision.NULL_REVISION in keys:
1863
keys.remove(_mod_revision.NULL_REVISION)
1865
parent_map = graph.get_parent_map(keys)
1866
keys = tsort.topo_sort(parent_map)
1867
return [None] + list(keys)
1870
"""Compress the data within the repository.
1872
This operation only makes sense for some repository types. For other
1873
types it should be a no-op that just returns.
1875
This stub method does not require a lock, but subclasses should use
1876
@needs_write_lock as this is a long running call its reasonable to
1877
implicitly lock for the user.
1881
@deprecated_method(one_six)
1882
def print_file(self, file, revision_id):
1883
"""Print `file` to stdout.
1885
FIXME RBC 20060125 as John Meinel points out this is a bad api
1886
- it writes to stdout, it assumes that that is valid etc. Fix
1887
by creating a new more flexible convenience function.
1889
tree = self.revision_tree(revision_id)
1890
# use inventory as it was in that revision
1891
file_id = tree.inventory.path2id(file)
1893
# TODO: jam 20060427 Write a test for this code path
1894
# it had a bug in it, and was raising the wrong
1896
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1897
tree.print_file(file_id)
1899
def get_transaction(self):
1900
return self.control_files.get_transaction()
1902
@deprecated_method(one_one)
1903
def get_parents(self, revision_ids):
1904
"""See StackedParentsProvider.get_parents"""
1905
parent_map = self.get_parent_map(revision_ids)
1906
return [parent_map.get(r, None) for r in revision_ids]
1908
def get_parent_map(self, revision_ids):
1909
"""See graph._StackedParentsProvider.get_parent_map"""
1910
# revisions index works in keys; this just works in revisions
1911
# therefore wrap and unwrap
1914
for revision_id in revision_ids:
1915
if revision_id == _mod_revision.NULL_REVISION:
1916
result[revision_id] = ()
1917
elif revision_id is None:
1918
raise ValueError('get_parent_map(None) is not valid')
1920
query_keys.append((revision_id ,))
1921
for ((revision_id,), parent_keys) in \
1922
self.revisions.get_parent_map(query_keys).iteritems():
1924
result[revision_id] = tuple(parent_revid
1925
for (parent_revid,) in parent_keys)
1927
result[revision_id] = (_mod_revision.NULL_REVISION,)
1930
def _make_parents_provider(self):
1933
def get_graph(self, other_repository=None):
1934
"""Return the graph walker for this repository format"""
1935
parents_provider = self._make_parents_provider()
1936
if (other_repository is not None and
1937
not self.has_same_location(other_repository)):
1938
parents_provider = graph._StackedParentsProvider(
1939
[parents_provider, other_repository._make_parents_provider()])
1940
return graph.Graph(parents_provider)
1942
def _get_versioned_file_checker(self):
1943
"""Return an object suitable for checking versioned files."""
1944
return _VersionedFileChecker(self)
1946
def revision_ids_to_search_result(self, result_set):
1947
"""Convert a set of revision ids to a graph SearchResult."""
1948
result_parents = set()
1949
for parents in self.get_graph().get_parent_map(
1950
result_set).itervalues():
1951
result_parents.update(parents)
1952
included_keys = result_set.intersection(result_parents)
1953
start_keys = result_set.difference(included_keys)
1954
exclude_keys = result_parents.difference(result_set)
1955
result = graph.SearchResult(start_keys, exclude_keys,
1956
len(result_set), result_set)
1960
def set_make_working_trees(self, new_value):
1961
"""Set the policy flag for making working trees when creating branches.
1963
This only applies to branches that use this repository.
1965
The default is 'True'.
1966
:param new_value: True to restore the default, False to disable making
1969
raise NotImplementedError(self.set_make_working_trees)
1971
def make_working_trees(self):
1972
"""Returns the policy for making working trees on new branches."""
1973
raise NotImplementedError(self.make_working_trees)
1976
def sign_revision(self, revision_id, gpg_strategy):
1977
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1978
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1981
def has_signature_for_revision_id(self, revision_id):
1982
"""Query for a revision signature for revision_id in the repository."""
1983
if not self.has_revision(revision_id):
1984
raise errors.NoSuchRevision(self, revision_id)
1985
sig_present = (1 == len(
1986
self.signatures.get_parent_map([(revision_id,)])))
1990
def get_signature_text(self, revision_id):
1991
"""Return the text for a signature."""
1992
stream = self.signatures.get_record_stream([(revision_id,)],
1994
record = stream.next()
1995
if record.storage_kind == 'absent':
1996
raise errors.NoSuchRevision(self, revision_id)
1997
return record.get_bytes_as('fulltext')
2000
def check(self, revision_ids=None):
2001
"""Check consistency of all history of given revision_ids.
2003
Different repository implementations should override _check().
2005
:param revision_ids: A non-empty list of revision_ids whose ancestry
2006
will be checked. Typically the last revision_id of a branch.
2008
return self._check(revision_ids)
2010
def _check(self, revision_ids):
2011
result = check.Check(self)
2015
def _warn_if_deprecated(self):
2016
global _deprecation_warning_done
2017
if _deprecation_warning_done:
2019
_deprecation_warning_done = True
2020
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2021
% (self._format, self.bzrdir.transport.base))
2023
def supports_rich_root(self):
2024
return self._format.rich_root_data
2026
def _check_ascii_revisionid(self, revision_id, method):
2027
"""Private helper for ascii-only repositories."""
2028
# weave repositories refuse to store revisionids that are non-ascii.
2029
if revision_id is not None:
2030
# weaves require ascii revision ids.
2031
if isinstance(revision_id, unicode):
2033
revision_id.encode('ascii')
2034
except UnicodeEncodeError:
2035
raise errors.NonAsciiRevisionId(method, self)
2038
revision_id.decode('ascii')
2039
except UnicodeDecodeError:
2040
raise errors.NonAsciiRevisionId(method, self)
2042
def revision_graph_can_have_wrong_parents(self):
2043
"""Is it possible for this repository to have a revision graph with
2046
If True, then this repository must also implement
2047
_find_inconsistent_revision_parents so that check and reconcile can
2048
check for inconsistencies before proceeding with other checks that may
2049
depend on the revision index being consistent.
2051
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2054
# remove these delegates a while after bzr 0.15
2055
def __make_delegated(name, from_module):
2056
def _deprecated_repository_forwarder():
2057
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2058
% (name, from_module),
2061
m = __import__(from_module, globals(), locals(), [name])
2063
return getattr(m, name)
2064
except AttributeError:
2065
raise AttributeError('module %s has no name %s'
2067
globals()[name] = _deprecated_repository_forwarder
2070
'AllInOneRepository',
2071
'WeaveMetaDirRepository',
2072
'PreSplitOutRepositoryFormat',
2073
'RepositoryFormat4',
2074
'RepositoryFormat5',
2075
'RepositoryFormat6',
2076
'RepositoryFormat7',
2078
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2082
'RepositoryFormatKnit',
2083
'RepositoryFormatKnit1',
2085
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2088
def install_revision(repository, rev, revision_tree):
2089
"""Install all revision data into a repository."""
2090
install_revisions(repository, [(rev, revision_tree, None)])
2093
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2094
"""Install all revision data into a repository.
2096
Accepts an iterable of revision, tree, signature tuples. The signature
2099
repository.start_write_group()
2101
for n, (revision, revision_tree, signature) in enumerate(iterable):
2102
_install_revision(repository, revision, revision_tree, signature)
2104
pb.update('Transferring revisions', n + 1, num_revisions)
2106
repository.abort_write_group()
2109
repository.commit_write_group()
2112
def _install_revision(repository, rev, revision_tree, signature):
2113
"""Install all revision data into a repository."""
2114
present_parents = []
2116
for p_id in rev.parent_ids:
2117
if repository.has_revision(p_id):
2118
present_parents.append(p_id)
2119
parent_trees[p_id] = repository.revision_tree(p_id)
2121
parent_trees[p_id] = repository.revision_tree(
2122
_mod_revision.NULL_REVISION)
2124
inv = revision_tree.inventory
2125
entries = inv.iter_entries()
2126
# backwards compatibility hack: skip the root id.
2127
if not repository.supports_rich_root():
2128
path, root = entries.next()
2129
if root.revision != rev.revision_id:
2130
raise errors.IncompatibleRevision(repr(repository))
2132
for path, ie in entries:
2133
text_keys[(ie.file_id, ie.revision)] = ie
2134
text_parent_map = repository.texts.get_parent_map(text_keys)
2135
missing_texts = set(text_keys) - set(text_parent_map)
2136
# Add the texts that are not already present
2137
for text_key in missing_texts:
2138
ie = text_keys[text_key]
2140
# FIXME: TODO: The following loop overlaps/duplicates that done by
2141
# commit to determine parents. There is a latent/real bug here where
2142
# the parents inserted are not those commit would do - in particular
2143
# they are not filtered by heads(). RBC, AB
2144
for revision, tree in parent_trees.iteritems():
2145
if ie.file_id not in tree:
2147
parent_id = tree.inventory[ie.file_id].revision
2148
if parent_id in text_parents:
2150
text_parents.append((ie.file_id, parent_id))
2151
lines = revision_tree.get_file(ie.file_id).readlines()
2152
repository.texts.add_lines(text_key, text_parents, lines)
2154
# install the inventory
2155
repository.add_inventory(rev.revision_id, inv, present_parents)
2156
except errors.RevisionAlreadyPresent:
2158
if signature is not None:
2159
repository.add_signature_text(rev.revision_id, signature)
2160
repository.add_revision(rev.revision_id, rev, inv)
2163
class MetaDirRepository(Repository):
2164
"""Repositories in the new meta-dir layout.
2166
:ivar _transport: Transport for access to repository control files,
2167
typically pointing to .bzr/repository.
2170
def __init__(self, _format, a_bzrdir, control_files):
2171
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2172
self._transport = control_files._transport
2174
def is_shared(self):
2175
"""Return True if this repository is flagged as a shared repository."""
2176
return self._transport.has('shared-storage')
2179
def set_make_working_trees(self, new_value):
2180
"""Set the policy flag for making working trees when creating branches.
2182
This only applies to branches that use this repository.
2184
The default is 'True'.
2185
:param new_value: True to restore the default, False to disable making
2190
self._transport.delete('no-working-trees')
2191
except errors.NoSuchFile:
2194
self._transport.put_bytes('no-working-trees', '',
2195
mode=self.bzrdir._get_file_mode())
2197
def make_working_trees(self):
2198
"""Returns the policy for making working trees on new branches."""
2199
return not self._transport.has('no-working-trees')
2202
class MetaDirVersionedFileRepository(MetaDirRepository):
2203
"""Repositories in a meta-dir, that work via versioned file objects."""
2205
def __init__(self, _format, a_bzrdir, control_files):
2206
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2210
network_format_registry = registry.FormatRegistry()
2211
"""Registry of formats indexed by their network name.
2213
The network name for a repository format is an identifier that can be used when
2214
referring to formats with smart server operations. See
2215
RepositoryFormat.network_name() for more detail.
2219
format_registry = registry.FormatRegistry(network_format_registry)
2220
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
2222
This can contain either format instances themselves, or classes/factories that
2223
can be called to obtain one.
2227
#####################################################################
2228
# Repository Formats
2230
class RepositoryFormat(object):
2231
"""A repository format.
2233
Formats provide four things:
2234
* An initialization routine to construct repository data on disk.
2235
* a optional format string which is used when the BzrDir supports
2237
* an open routine which returns a Repository instance.
2238
* A network name for referring to the format in smart server RPC
2241
There is one and only one Format subclass for each on-disk format. But
2242
there can be one Repository subclass that is used for several different
2243
formats. The _format attribute on a Repository instance can be used to
2244
determine the disk format.
2246
Formats are placed in a registry by their format string for reference
2247
during opening. These should be subclasses of RepositoryFormat for
2250
Once a format is deprecated, just deprecate the initialize and open
2251
methods on the format class. Do not deprecate the object, as the
2252
object may be created even when a repository instnace hasn't been
2255
Common instance attributes:
2256
_matchingbzrdir - the bzrdir format that the repository format was
2257
originally written to work with. This can be used if manually
2258
constructing a bzrdir and repository, or more commonly for test suite
2262
# Set to True or False in derived classes. True indicates that the format
2263
# supports ghosts gracefully.
2264
supports_ghosts = None
2265
# Can this repository be given external locations to lookup additional
2266
# data. Set to True or False in derived classes.
2267
supports_external_lookups = None
2270
return "<%s>" % self.__class__.__name__
2272
def __eq__(self, other):
2273
# format objects are generally stateless
2274
return isinstance(other, self.__class__)
2276
def __ne__(self, other):
2277
return not self == other
2280
def find_format(klass, a_bzrdir):
2281
"""Return the format for the repository object in a_bzrdir.
2283
This is used by bzr native formats that have a "format" file in
2284
the repository. Other methods may be used by different types of
2288
transport = a_bzrdir.get_repository_transport(None)
2289
format_string = transport.get("format").read()
2290
return format_registry.get(format_string)
2291
except errors.NoSuchFile:
2292
raise errors.NoRepositoryPresent(a_bzrdir)
2294
raise errors.UnknownFormatError(format=format_string,
2298
def register_format(klass, format):
2299
format_registry.register(format.get_format_string(), format)
2302
def unregister_format(klass, format):
2303
format_registry.remove(format.get_format_string())
2306
def get_default_format(klass):
2307
"""Return the current default format."""
2308
from bzrlib import bzrdir
2309
return bzrdir.format_registry.make_bzrdir('default').repository_format
2311
def get_format_string(self):
2312
"""Return the ASCII format string that identifies this format.
2314
Note that in pre format ?? repositories the format string is
2315
not permitted nor written to disk.
2317
raise NotImplementedError(self.get_format_string)
2319
def get_format_description(self):
2320
"""Return the short description for this format."""
2321
raise NotImplementedError(self.get_format_description)
2323
# TODO: this shouldn't be in the base class, it's specific to things that
2324
# use weaves or knits -- mbp 20070207
2325
def _get_versioned_file_store(self,
2330
versionedfile_class=None,
2331
versionedfile_kwargs={},
2333
if versionedfile_class is None:
2334
versionedfile_class = self._versionedfile_class
2335
weave_transport = control_files._transport.clone(name)
2336
dir_mode = control_files._dir_mode
2337
file_mode = control_files._file_mode
2338
return VersionedFileStore(weave_transport, prefixed=prefixed,
2340
file_mode=file_mode,
2341
versionedfile_class=versionedfile_class,
2342
versionedfile_kwargs=versionedfile_kwargs,
2345
def initialize(self, a_bzrdir, shared=False):
2346
"""Initialize a repository of this format in a_bzrdir.
2348
:param a_bzrdir: The bzrdir to put the new repository in it.
2349
:param shared: The repository should be initialized as a sharable one.
2350
:returns: The new repository object.
2352
This may raise UninitializableFormat if shared repository are not
2353
compatible the a_bzrdir.
2355
raise NotImplementedError(self.initialize)
2357
def is_supported(self):
2358
"""Is this format supported?
2360
Supported formats must be initializable and openable.
2361
Unsupported formats may not support initialization or committing or
2362
some other features depending on the reason for not being supported.
2366
def network_name(self):
2367
"""A simple byte string uniquely identifying this format for RPC calls.
2369
MetaDir repository formats use their disk format string to identify the
2370
repository over the wire. All in one formats such as bzr < 0.8, and
2371
foreign formats like svn/git and hg should use some marker which is
2372
unique and immutable.
2374
raise NotImplementedError(self.network_name)
2376
def check_conversion_target(self, target_format):
2377
raise NotImplementedError(self.check_conversion_target)
2379
def open(self, a_bzrdir, _found=False):
2380
"""Return an instance of this format for the bzrdir a_bzrdir.
2382
_found is a private parameter, do not use it.
2384
raise NotImplementedError(self.open)
2387
class MetaDirRepositoryFormat(RepositoryFormat):
2388
"""Common base class for the new repositories using the metadir layout."""
2390
rich_root_data = False
2391
supports_tree_reference = False
2392
supports_external_lookups = False
2395
def _matchingbzrdir(self):
2396
matching = bzrdir.BzrDirMetaFormat1()
2397
matching.repository_format = self
2401
super(MetaDirRepositoryFormat, self).__init__()
2403
def _create_control_files(self, a_bzrdir):
2404
"""Create the required files and the initial control_files object."""
2405
# FIXME: RBC 20060125 don't peek under the covers
2406
# NB: no need to escape relative paths that are url safe.
2407
repository_transport = a_bzrdir.get_repository_transport(self)
2408
control_files = lockable_files.LockableFiles(repository_transport,
2409
'lock', lockdir.LockDir)
2410
control_files.create_lock()
2411
return control_files
2413
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2414
"""Upload the initial blank content."""
2415
control_files = self._create_control_files(a_bzrdir)
2416
control_files.lock_write()
2417
transport = control_files._transport
2419
utf8_files += [('shared-storage', '')]
2421
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2422
for (filename, content_stream) in files:
2423
transport.put_file(filename, content_stream,
2424
mode=a_bzrdir._get_file_mode())
2425
for (filename, content_bytes) in utf8_files:
2426
transport.put_bytes_non_atomic(filename, content_bytes,
2427
mode=a_bzrdir._get_file_mode())
2429
control_files.unlock()
2431
def network_name(self):
2432
"""Metadir formats have matching disk and network format strings."""
2433
return self.get_format_string()
2436
# Pre-0.8 formats that don't have a disk format string (because they are
2437
# versioned by the matching control directory). We use the control directories
2438
# disk format string as a key for the network_name because they meet the
2439
# constraints (simple string, unique, immmutable).
2440
network_format_registry.register_lazy(
2441
"Bazaar-NG branch, format 5\n",
2442
'bzrlib.repofmt.weaverepo',
2443
'RepositoryFormat5',
2445
network_format_registry.register_lazy(
2446
"Bazaar-NG branch, format 6\n",
2447
'bzrlib.repofmt.weaverepo',
2448
'RepositoryFormat6',
2451
# formats which have no format string are not discoverable or independently
2452
# creatable on disk, so are not registered in format_registry. They're
2453
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2454
# needed, it's constructed directly by the BzrDir. Non-native formats where
2455
# the repository is not separately opened are similar.
2457
format_registry.register_lazy(
2458
'Bazaar-NG Repository format 7',
2459
'bzrlib.repofmt.weaverepo',
2463
format_registry.register_lazy(
2464
'Bazaar-NG Knit Repository Format 1',
2465
'bzrlib.repofmt.knitrepo',
2466
'RepositoryFormatKnit1',
2469
format_registry.register_lazy(
2470
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2471
'bzrlib.repofmt.knitrepo',
2472
'RepositoryFormatKnit3',
2475
format_registry.register_lazy(
2476
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2477
'bzrlib.repofmt.knitrepo',
2478
'RepositoryFormatKnit4',
2481
# Pack-based formats. There is one format for pre-subtrees, and one for
2482
# post-subtrees to allow ease of testing.
2483
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2484
format_registry.register_lazy(
2485
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2486
'bzrlib.repofmt.pack_repo',
2487
'RepositoryFormatKnitPack1',
2489
format_registry.register_lazy(
2490
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2491
'bzrlib.repofmt.pack_repo',
2492
'RepositoryFormatKnitPack3',
2494
format_registry.register_lazy(
2495
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2496
'bzrlib.repofmt.pack_repo',
2497
'RepositoryFormatKnitPack4',
2499
format_registry.register_lazy(
2500
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2501
'bzrlib.repofmt.pack_repo',
2502
'RepositoryFormatKnitPack5',
2504
format_registry.register_lazy(
2505
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2506
'bzrlib.repofmt.pack_repo',
2507
'RepositoryFormatKnitPack5RichRoot',
2509
format_registry.register_lazy(
2510
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2511
'bzrlib.repofmt.pack_repo',
2512
'RepositoryFormatKnitPack5RichRootBroken',
2514
format_registry.register_lazy(
2515
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2516
'bzrlib.repofmt.pack_repo',
2517
'RepositoryFormatKnitPack6',
2519
format_registry.register_lazy(
2520
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2521
'bzrlib.repofmt.pack_repo',
2522
'RepositoryFormatKnitPack6RichRoot',
2525
# Development formats.
2526
# 1.7->1.8 go below here
2527
format_registry.register_lazy(
2528
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2529
'bzrlib.repofmt.pack_repo',
2530
'RepositoryFormatPackDevelopment2',
2532
format_registry.register_lazy(
2533
("Bazaar development format 2 with subtree support "
2534
"(needs bzr.dev from before 1.8)\n"),
2535
'bzrlib.repofmt.pack_repo',
2536
'RepositoryFormatPackDevelopment2Subtree',
2540
class InterRepository(InterObject):
2541
"""This class represents operations taking place between two repositories.
2543
Its instances have methods like copy_content and fetch, and contain
2544
references to the source and target repositories these operations can be
2547
Often we will provide convenience methods on 'repository' which carry out
2548
operations with another repository - they will always forward to
2549
InterRepository.get(other).method_name(parameters).
2552
_walk_to_common_revisions_batch_size = 1
2554
"""The available optimised InterRepository types."""
2556
def __init__(self, source, target):
2557
InterObject.__init__(self, source, target)
2558
# These two attributes may be overridden by e.g. InterOtherToRemote to
2559
# provide a faster implementation.
2560
self.target_get_graph = self.target.get_graph
2561
self.target_get_parent_map = self.target.get_parent_map
2563
def copy_content(self, revision_id=None):
2564
raise NotImplementedError(self.copy_content)
2566
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2567
"""Fetch the content required to construct revision_id.
2569
The content is copied from self.source to self.target.
2571
:param revision_id: if None all content is copied, if NULL_REVISION no
2573
:param pb: optional progress bar to use for progress reports. If not
2574
provided a default one will be created.
2576
:returns: (copied_revision_count, failures).
2578
# Normally we should find a specific InterRepository subclass to do
2579
# the fetch; if nothing else then at least InterSameDataRepository.
2580
# If none of them is suitable it looks like fetching is not possible;
2581
# we try to give a good message why. _assert_same_model will probably
2582
# give a helpful message; otherwise a generic one.
2583
self._assert_same_model(self.source, self.target)
2584
raise errors.IncompatibleRepositories(self.source, self.target,
2585
"no suitableInterRepository found")
2587
def _walk_to_common_revisions(self, revision_ids):
2588
"""Walk out from revision_ids in source to revisions target has.
2590
:param revision_ids: The start point for the search.
2591
:return: A set of revision ids.
2593
target_graph = self.target_get_graph()
2594
revision_ids = frozenset(revision_ids)
2595
# Fast path for the case where all the revisions are already in the
2597
# (Although this does incur an extra round trip for the
2598
# fairly common case where the target doesn't already have the revision
2600
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2601
return graph.SearchResult(revision_ids, set(), 0, set())
2602
missing_revs = set()
2603
source_graph = self.source.get_graph()
2604
# ensure we don't pay silly lookup costs.
2605
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2606
null_set = frozenset([_mod_revision.NULL_REVISION])
2607
searcher_exhausted = False
2611
# Iterate the searcher until we have enough next_revs
2612
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2614
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2615
next_revs.update(next_revs_part)
2616
ghosts.update(ghosts_part)
2617
except StopIteration:
2618
searcher_exhausted = True
2620
# If there are ghosts in the source graph, and the caller asked for
2621
# them, make sure that they are present in the target.
2622
# We don't care about other ghosts as we can't fetch them and
2623
# haven't been asked to.
2624
ghosts_to_check = set(revision_ids.intersection(ghosts))
2625
revs_to_get = set(next_revs).union(ghosts_to_check)
2627
have_revs = set(target_graph.get_parent_map(revs_to_get))
2628
# we always have NULL_REVISION present.
2629
have_revs = have_revs.union(null_set)
2630
# Check if the target is missing any ghosts we need.
2631
ghosts_to_check.difference_update(have_revs)
2633
# One of the caller's revision_ids is a ghost in both the
2634
# source and the target.
2635
raise errors.NoSuchRevision(
2636
self.source, ghosts_to_check.pop())
2637
missing_revs.update(next_revs - have_revs)
2638
# Because we may have walked past the original stop point, make
2639
# sure everything is stopped
2640
stop_revs = searcher.find_seen_ancestors(have_revs)
2641
searcher.stop_searching_any(stop_revs)
2642
if searcher_exhausted:
2644
return searcher.get_result()
2646
@deprecated_method(one_two)
2648
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2649
"""Return the revision ids that source has that target does not.
2651
These are returned in topological order.
2653
:param revision_id: only return revision ids included by this
2655
:param find_ghosts: If True find missing revisions in deep history
2656
rather than just finding the surface difference.
2658
return list(self.search_missing_revision_ids(
2659
revision_id, find_ghosts).get_keys())
2662
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2663
"""Return the revision ids that source has that target does not.
2665
:param revision_id: only return revision ids included by this
2667
:param find_ghosts: If True find missing revisions in deep history
2668
rather than just finding the surface difference.
2669
:return: A bzrlib.graph.SearchResult.
2671
# stop searching at found target revisions.
2672
if not find_ghosts and revision_id is not None:
2673
return self._walk_to_common_revisions([revision_id])
2674
# generic, possibly worst case, slow code path.
2675
target_ids = set(self.target.all_revision_ids())
2676
if revision_id is not None:
2677
source_ids = self.source.get_ancestry(revision_id)
2678
if source_ids[0] is not None:
2679
raise AssertionError()
2682
source_ids = self.source.all_revision_ids()
2683
result_set = set(source_ids).difference(target_ids)
2684
return self.source.revision_ids_to_search_result(result_set)
2687
def _same_model(source, target):
2688
"""True if source and target have the same data representation.
2690
Note: this is always called on the base class; overriding it in a
2691
subclass will have no effect.
2694
InterRepository._assert_same_model(source, target)
2696
except errors.IncompatibleRepositories, e:
2700
def _assert_same_model(source, target):
2701
"""Raise an exception if two repositories do not use the same model.
2703
if source.supports_rich_root() != target.supports_rich_root():
2704
raise errors.IncompatibleRepositories(source, target,
2705
"different rich-root support")
2706
if source._serializer != target._serializer:
2707
raise errors.IncompatibleRepositories(source, target,
2708
"different serializers")
2711
class InterSameDataRepository(InterRepository):
2712
"""Code for converting between repositories that represent the same data.
2714
Data format and model must match for this to work.
2718
def _get_repo_format_to_test(self):
2719
"""Repository format for testing with.
2721
InterSameData can pull from subtree to subtree and from non-subtree to
2722
non-subtree, so we test this with the richest repository format.
2724
from bzrlib.repofmt import knitrepo
2725
return knitrepo.RepositoryFormatKnit3()
2728
def is_compatible(source, target):
2729
return InterRepository._same_model(source, target)
2732
def copy_content(self, revision_id=None):
2733
"""Make a complete copy of the content in self into destination.
2735
This copies both the repository's revision data, and configuration information
2736
such as the make_working_trees setting.
2738
This is a destructive operation! Do not use it on existing
2741
:param revision_id: Only copy the content needed to construct
2742
revision_id and its parents.
2745
self.target.set_make_working_trees(self.source.make_working_trees())
2746
except NotImplementedError:
2748
# but don't bother fetching if we have the needed data now.
2749
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2750
self.target.has_revision(revision_id)):
2752
self.target.fetch(self.source, revision_id=revision_id)
2755
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2756
"""See InterRepository.fetch()."""
2757
from bzrlib.fetch import RepoFetcher
2758
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2759
self.source, self.source._format, self.target,
2760
self.target._format)
2761
f = RepoFetcher(to_repository=self.target,
2762
from_repository=self.source,
2763
last_revision=revision_id,
2764
pb=pb, find_ghosts=find_ghosts)
2765
return f.count_copied, f.failed_revisions
2768
class InterWeaveRepo(InterSameDataRepository):
2769
"""Optimised code paths between Weave based repositories.
2771
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2772
implemented lazy inter-object optimisation.
2776
def _get_repo_format_to_test(self):
2777
from bzrlib.repofmt import weaverepo
2778
return weaverepo.RepositoryFormat7()
2781
def is_compatible(source, target):
2782
"""Be compatible with known Weave formats.
2784
We don't test for the stores being of specific types because that
2785
could lead to confusing results, and there is no need to be
2788
from bzrlib.repofmt.weaverepo import (
2794
return (isinstance(source._format, (RepositoryFormat5,
2796
RepositoryFormat7)) and
2797
isinstance(target._format, (RepositoryFormat5,
2799
RepositoryFormat7)))
2800
except AttributeError:
2804
def copy_content(self, revision_id=None):
2805
"""See InterRepository.copy_content()."""
2806
# weave specific optimised path:
2808
self.target.set_make_working_trees(self.source.make_working_trees())
2809
except (errors.RepositoryUpgradeRequired, NotImplemented):
2811
# FIXME do not peek!
2812
if self.source._transport.listable():
2813
pb = ui.ui_factory.nested_progress_bar()
2815
self.target.texts.insert_record_stream(
2816
self.source.texts.get_record_stream(
2817
self.source.texts.keys(), 'topological', False))
2818
pb.update('copying inventory', 0, 1)
2819
self.target.inventories.insert_record_stream(
2820
self.source.inventories.get_record_stream(
2821
self.source.inventories.keys(), 'topological', False))
2822
self.target.signatures.insert_record_stream(
2823
self.source.signatures.get_record_stream(
2824
self.source.signatures.keys(),
2826
self.target.revisions.insert_record_stream(
2827
self.source.revisions.get_record_stream(
2828
self.source.revisions.keys(),
2829
'topological', True))
2833
self.target.fetch(self.source, revision_id=revision_id)
2836
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2837
"""See InterRepository.fetch()."""
2838
from bzrlib.fetch import RepoFetcher
2839
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2840
self.source, self.source._format, self.target, self.target._format)
2841
f = RepoFetcher(to_repository=self.target,
2842
from_repository=self.source,
2843
last_revision=revision_id,
2844
pb=pb, find_ghosts=find_ghosts)
2845
return f.count_copied, f.failed_revisions
2848
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2849
"""See InterRepository.missing_revision_ids()."""
2850
# we want all revisions to satisfy revision_id in source.
2851
# but we don't want to stat every file here and there.
2852
# we want then, all revisions other needs to satisfy revision_id
2853
# checked, but not those that we have locally.
2854
# so the first thing is to get a subset of the revisions to
2855
# satisfy revision_id in source, and then eliminate those that
2856
# we do already have.
2857
# this is slow on high latency connection to self, but as as this
2858
# disk format scales terribly for push anyway due to rewriting
2859
# inventory.weave, this is considered acceptable.
2861
if revision_id is not None:
2862
source_ids = self.source.get_ancestry(revision_id)
2863
if source_ids[0] is not None:
2864
raise AssertionError()
2867
source_ids = self.source._all_possible_ids()
2868
source_ids_set = set(source_ids)
2869
# source_ids is the worst possible case we may need to pull.
2870
# now we want to filter source_ids against what we actually
2871
# have in target, but don't try to check for existence where we know
2872
# we do not have a revision as that would be pointless.
2873
target_ids = set(self.target._all_possible_ids())
2874
possibly_present_revisions = target_ids.intersection(source_ids_set)
2875
actually_present_revisions = set(
2876
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2877
required_revisions = source_ids_set.difference(actually_present_revisions)
2878
if revision_id is not None:
2879
# we used get_ancestry to determine source_ids then we are assured all
2880
# revisions referenced are present as they are installed in topological order.
2881
# and the tip revision was validated by get_ancestry.
2882
result_set = required_revisions
2884
# if we just grabbed the possibly available ids, then
2885
# we only have an estimate of whats available and need to validate
2886
# that against the revision records.
2888
self.source._eliminate_revisions_not_present(required_revisions))
2889
return self.source.revision_ids_to_search_result(result_set)
2892
class InterKnitRepo(InterSameDataRepository):
2893
"""Optimised code paths between Knit based repositories."""
2896
def _get_repo_format_to_test(self):
2897
from bzrlib.repofmt import knitrepo
2898
return knitrepo.RepositoryFormatKnit1()
2901
def is_compatible(source, target):
2902
"""Be compatible with known Knit formats.
2904
We don't test for the stores being of specific types because that
2905
could lead to confusing results, and there is no need to be
2908
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2910
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2911
isinstance(target._format, RepositoryFormatKnit))
2912
except AttributeError:
2914
return are_knits and InterRepository._same_model(source, target)
2917
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2918
"""See InterRepository.fetch()."""
2919
from bzrlib.fetch import RepoFetcher
2920
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2921
self.source, self.source._format, self.target, self.target._format)
2922
f = RepoFetcher(to_repository=self.target,
2923
from_repository=self.source,
2924
last_revision=revision_id,
2925
pb=pb, find_ghosts=find_ghosts)
2926
return f.count_copied, f.failed_revisions
2929
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2930
"""See InterRepository.missing_revision_ids()."""
2931
if revision_id is not None:
2932
source_ids = self.source.get_ancestry(revision_id)
2933
if source_ids[0] is not None:
2934
raise AssertionError()
2937
source_ids = self.source.all_revision_ids()
2938
source_ids_set = set(source_ids)
2939
# source_ids is the worst possible case we may need to pull.
2940
# now we want to filter source_ids against what we actually
2941
# have in target, but don't try to check for existence where we know
2942
# we do not have a revision as that would be pointless.
2943
target_ids = set(self.target.all_revision_ids())
2944
possibly_present_revisions = target_ids.intersection(source_ids_set)
2945
actually_present_revisions = set(
2946
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2947
required_revisions = source_ids_set.difference(actually_present_revisions)
2948
if revision_id is not None:
2949
# we used get_ancestry to determine source_ids then we are assured all
2950
# revisions referenced are present as they are installed in topological order.
2951
# and the tip revision was validated by get_ancestry.
2952
result_set = required_revisions
2954
# if we just grabbed the possibly available ids, then
2955
# we only have an estimate of whats available and need to validate
2956
# that against the revision records.
2958
self.source._eliminate_revisions_not_present(required_revisions))
2959
return self.source.revision_ids_to_search_result(result_set)
2962
class InterPackRepo(InterSameDataRepository):
2963
"""Optimised code paths between Pack based repositories."""
2966
def _get_repo_format_to_test(self):
2967
from bzrlib.repofmt import pack_repo
2968
return pack_repo.RepositoryFormatKnitPack1()
2971
def is_compatible(source, target):
2972
"""Be compatible with known Pack formats.
2974
We don't test for the stores being of specific types because that
2975
could lead to confusing results, and there is no need to be
2978
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2980
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2981
isinstance(target._format, RepositoryFormatPack))
2982
except AttributeError:
2984
return are_packs and InterRepository._same_model(source, target)
2987
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2988
"""See InterRepository.fetch()."""
2989
if (len(self.source._fallback_repositories) > 0 or
2990
len(self.target._fallback_repositories) > 0):
2991
# The pack layer is not aware of fallback repositories, so when
2992
# fetching from a stacked repository or into a stacked repository
2993
# we use the generic fetch logic which uses the VersionedFiles
2994
# attributes on repository.
2995
from bzrlib.fetch import RepoFetcher
2996
fetcher = RepoFetcher(self.target, self.source, revision_id,
2998
return fetcher.count_copied, fetcher.failed_revisions
2999
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3000
self.source, self.source._format, self.target, self.target._format)
3001
self.count_copied = 0
3002
if revision_id is None:
3004
# everything to do - use pack logic
3005
# to fetch from all packs to one without
3006
# inventory parsing etc, IFF nothing to be copied is in the target.
3008
source_revision_ids = frozenset(self.source.all_revision_ids())
3009
revision_ids = source_revision_ids - \
3010
frozenset(self.target_get_parent_map(source_revision_ids))
3011
revision_keys = [(revid,) for revid in revision_ids]
3012
target_pack_collection = self._get_target_pack_collection()
3013
index = target_pack_collection.revision_index.combined_index
3014
present_revision_ids = set(item[1][0] for item in
3015
index.iter_entries(revision_keys))
3016
revision_ids = set(revision_ids) - present_revision_ids
3017
# implementing the TODO will involve:
3018
# - detecting when all of a pack is selected
3019
# - avoiding as much as possible pre-selection, so the
3020
# more-core routines such as create_pack_from_packs can filter in
3021
# a just-in-time fashion. (though having a HEADS list on a
3022
# repository might make this a lot easier, because we could
3023
# sensibly detect 'new revisions' without doing a full index scan.
3024
elif _mod_revision.is_null(revision_id):
3029
revision_ids = self.search_missing_revision_ids(revision_id,
3030
find_ghosts=find_ghosts).get_keys()
3031
except errors.NoSuchRevision:
3032
raise errors.InstallFailed([revision_id])
3033
if len(revision_ids) == 0:
3035
return self._pack(self.source, self.target, revision_ids)
3037
def _pack(self, source, target, revision_ids):
3038
from bzrlib.repofmt.pack_repo import Packer
3039
target_pack_collection = self._get_target_pack_collection()
3040
packs = source._pack_collection.all_packs()
3041
pack = Packer(target_pack_collection, packs, '.fetch',
3042
revision_ids).pack()
3043
if pack is not None:
3044
target_pack_collection._save_pack_names()
3045
copied_revs = pack.get_revision_count()
3046
# Trigger an autopack. This may duplicate effort as we've just done
3047
# a pack creation, but for now it is simpler to think about as
3048
# 'upload data, then repack if needed'.
3050
return (copied_revs, [])
3054
def _autopack(self):
3055
self.target._pack_collection.autopack()
3057
def _get_target_pack_collection(self):
3058
return self.target._pack_collection
3061
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3062
"""See InterRepository.missing_revision_ids().
3064
:param find_ghosts: Find ghosts throughout the ancestry of
3067
if not find_ghosts and revision_id is not None:
3068
return self._walk_to_common_revisions([revision_id])
3069
elif revision_id is not None:
3070
# Find ghosts: search for revisions pointing from one repository to
3071
# the other, and vice versa, anywhere in the history of revision_id.
3072
graph = self.target_get_graph(other_repository=self.source)
3073
searcher = graph._make_breadth_first_searcher([revision_id])
3077
next_revs, ghosts = searcher.next_with_ghosts()
3078
except StopIteration:
3080
if revision_id in ghosts:
3081
raise errors.NoSuchRevision(self.source, revision_id)
3082
found_ids.update(next_revs)
3083
found_ids.update(ghosts)
3084
found_ids = frozenset(found_ids)
3085
# Double query here: should be able to avoid this by changing the
3086
# graph api further.
3087
result_set = found_ids - frozenset(
3088
self.target_get_parent_map(found_ids))
3090
source_ids = self.source.all_revision_ids()
3091
# source_ids is the worst possible case we may need to pull.
3092
# now we want to filter source_ids against what we actually
3093
# have in target, but don't try to check for existence where we know
3094
# we do not have a revision as that would be pointless.
3095
target_ids = set(self.target.all_revision_ids())
3096
result_set = set(source_ids).difference(target_ids)
3097
return self.source.revision_ids_to_search_result(result_set)
3100
class InterModel1and2(InterRepository):
3103
def _get_repo_format_to_test(self):
3107
def is_compatible(source, target):
3108
if not source.supports_rich_root() and target.supports_rich_root():
3114
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3115
"""See InterRepository.fetch()."""
3116
from bzrlib.fetch import Model1toKnit2Fetcher
3117
f = Model1toKnit2Fetcher(to_repository=self.target,
3118
from_repository=self.source,
3119
last_revision=revision_id,
3120
pb=pb, find_ghosts=find_ghosts)
3121
return f.count_copied, f.failed_revisions
3124
def copy_content(self, revision_id=None):
3125
"""Make a complete copy of the content in self into destination.
3127
This is a destructive operation! Do not use it on existing
3130
:param revision_id: Only copy the content needed to construct
3131
revision_id and its parents.
3134
self.target.set_make_working_trees(self.source.make_working_trees())
3135
except NotImplementedError:
3137
# but don't bother fetching if we have the needed data now.
3138
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
3139
self.target.has_revision(revision_id)):
3141
self.target.fetch(self.source, revision_id=revision_id)
3144
class InterKnit1and2(InterKnitRepo):
3147
def _get_repo_format_to_test(self):
3151
def is_compatible(source, target):
3152
"""Be compatible with Knit1 source and Knit3 target"""
3154
from bzrlib.repofmt.knitrepo import (
3155
RepositoryFormatKnit1,
3156
RepositoryFormatKnit3,
3158
from bzrlib.repofmt.pack_repo import (
3159
RepositoryFormatKnitPack1,
3160
RepositoryFormatKnitPack3,
3161
RepositoryFormatKnitPack4,
3162
RepositoryFormatKnitPack5,
3163
RepositoryFormatKnitPack5RichRoot,
3164
RepositoryFormatKnitPack6,
3165
RepositoryFormatKnitPack6RichRoot,
3166
RepositoryFormatPackDevelopment2,
3167
RepositoryFormatPackDevelopment2Subtree,
3170
RepositoryFormatKnit1, # no rr, no subtree
3171
RepositoryFormatKnitPack1, # no rr, no subtree
3172
RepositoryFormatPackDevelopment2, # no rr, no subtree
3173
RepositoryFormatKnitPack5, # no rr, no subtree
3174
RepositoryFormatKnitPack6, # no rr, no subtree
3177
RepositoryFormatKnit3, # rr, subtree
3178
RepositoryFormatKnitPack3, # rr, subtree
3179
RepositoryFormatKnitPack4, # rr, no subtree
3180
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
3181
RepositoryFormatKnitPack6RichRoot,# rr, no subtree
3182
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
3184
for format in norichroot:
3185
if format.rich_root_data:
3186
raise AssertionError('Format %s is a rich-root format'
3187
' but is included in the non-rich-root list'
3189
for format in richroot:
3190
if not format.rich_root_data:
3191
raise AssertionError('Format %s is not a rich-root format'
3192
' but is included in the rich-root list'
3194
# TODO: One alternative is to just check format.rich_root_data,
3195
# instead of keeping membership lists. However, the formats
3196
# *also* have to use the same 'Knit' style of storage
3197
# (line-deltas, fulltexts, etc.)
3198
return (isinstance(source._format, norichroot) and
3199
isinstance(target._format, richroot))
3200
except AttributeError:
3204
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3205
"""See InterRepository.fetch()."""
3206
from bzrlib.fetch import Knit1to2Fetcher
3207
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3208
self.source, self.source._format, self.target,
3209
self.target._format)
3210
f = Knit1to2Fetcher(to_repository=self.target,
3211
from_repository=self.source,
3212
last_revision=revision_id,
3213
pb=pb, find_ghosts=find_ghosts)
3214
return f.count_copied, f.failed_revisions
3217
class InterDifferingSerializer(InterKnitRepo):
3220
def _get_repo_format_to_test(self):
3224
def is_compatible(source, target):
3225
"""Be compatible with Knit2 source and Knit3 target"""
3226
if source.supports_rich_root() != target.supports_rich_root():
3228
# Ideally, we'd support fetching if the source had no tree references
3229
# even if it supported them...
3230
if (getattr(source, '_format.supports_tree_reference', False) and
3231
not getattr(target, '_format.supports_tree_reference', False)):
3235
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3236
"""Get the best delta and base for this revision.
3238
:return: (basis_id, delta)
3240
possible_trees = [(parent_id, cache[parent_id])
3241
for parent_id in parent_ids
3242
if parent_id in cache]
3243
if len(possible_trees) == 0:
3244
# There either aren't any parents, or the parents aren't in the
3245
# cache, so just use the last converted tree
3246
possible_trees.append((basis_id, cache[basis_id]))
3248
for basis_id, basis_tree in possible_trees:
3249
delta = tree.inventory._make_delta(basis_tree.inventory)
3250
deltas.append((len(delta), basis_id, delta))
3252
return deltas[0][1:]
3254
def _fetch_batch(self, revision_ids, basis_id, cache):
3255
"""Fetch across a few revisions.
3257
:param revision_ids: The revisions to copy
3258
:param basis_id: The revision_id of a tree that must be in cache, used
3259
as a basis for delta when no other base is available
3260
:param cache: A cache of RevisionTrees that we can use.
3261
:return: The revision_id of the last converted tree. The RevisionTree
3262
for it will be in cache
3264
# Walk though all revisions; get inventory deltas, copy referenced
3265
# texts that delta references, insert the delta, revision and
3269
pending_revisions = []
3270
parent_map = self.source.get_parent_map(revision_ids)
3271
for tree in self.source.revision_trees(revision_ids):
3272
current_revision_id = tree.get_revision_id()
3273
parent_ids = parent_map.get(current_revision_id, ())
3274
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3276
# Find text entries that need to be copied
3277
for old_path, new_path, file_id, entry in delta:
3278
if new_path is not None:
3279
if not (new_path or self.target.supports_rich_root()):
3280
# We don't copy the text for the root node unless the
3281
# target supports_rich_root.
3284
# "if entry.revision == current_revision_id" ?
3285
if entry.revision == current_revision_id:
3286
text_keys.add((file_id, entry.revision))
3287
revision = self.source.get_revision(current_revision_id)
3288
pending_deltas.append((basis_id, delta,
3289
current_revision_id, revision.parent_ids))
3290
pending_revisions.append(revision)
3291
cache[current_revision_id] = tree
3292
basis_id = current_revision_id
3294
from_texts = self.source.texts
3295
to_texts = self.target.texts
3296
to_texts.insert_record_stream(from_texts.get_record_stream(
3297
text_keys, self.target._fetch_order,
3298
not self.target._fetch_uses_deltas))
3300
for delta in pending_deltas:
3301
self.target.add_inventory_by_delta(*delta)
3302
# insert signatures and revisions
3303
for revision in pending_revisions:
3305
signature = self.source.get_signature_text(
3306
revision.revision_id)
3307
self.target.add_signature_text(revision.revision_id,
3309
except errors.NoSuchRevision:
3311
self.target.add_revision(revision.revision_id, revision)
3314
def _fetch_all_revisions(self, revision_ids, pb):
3315
"""Fetch everything for the list of revisions.
3317
:param revision_ids: The list of revisions to fetch. Must be in
3319
:param pb: A ProgressBar
3322
basis_id, basis_tree = self._get_basis(revision_ids[0])
3324
cache = lru_cache.LRUCache(100)
3325
cache[basis_id] = basis_tree
3326
del basis_tree # We don't want to hang on to it here
3327
for offset in range(0, len(revision_ids), batch_size):
3328
self.target.start_write_group()
3330
pb.update('Transferring revisions', offset,
3332
batch = revision_ids[offset:offset+batch_size]
3333
basis_id = self._fetch_batch(batch, basis_id, cache)
3335
self.target.abort_write_group()
3338
self.target.commit_write_group()
3339
pb.update('Transferring revisions', len(revision_ids),
3343
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3344
"""See InterRepository.fetch()."""
3345
revision_ids = self.target.search_missing_revision_ids(self.source,
3346
revision_id, find_ghosts=find_ghosts).get_keys()
3347
if not revision_ids:
3349
revision_ids = tsort.topo_sort(
3350
self.source.get_graph().get_parent_map(revision_ids))
3352
my_pb = ui.ui_factory.nested_progress_bar()
3357
self._fetch_all_revisions(revision_ids, pb)
3359
if my_pb is not None:
3361
return len(revision_ids), 0
3363
def _get_basis(self, first_revision_id):
3364
"""Get a revision and tree which exists in the target.
3366
This assumes that first_revision_id is selected for transmission
3367
because all other ancestors are already present. If we can't find an
3368
ancestor we fall back to NULL_REVISION since we know that is safe.
3370
:return: (basis_id, basis_tree)
3372
first_rev = self.source.get_revision(first_revision_id)
3374
basis_id = first_rev.parent_ids[0]
3375
# only valid as a basis if the target has it
3376
self.target.get_revision(basis_id)
3377
# Try to get a basis tree - if its a ghost it will hit the
3378
# NoSuchRevision case.
3379
basis_tree = self.source.revision_tree(basis_id)
3380
except (IndexError, errors.NoSuchRevision):
3381
basis_id = _mod_revision.NULL_REVISION
3382
basis_tree = self.source.revision_tree(basis_id)
3383
return basis_id, basis_tree
3386
class InterOtherToRemote(InterRepository):
3387
"""An InterRepository that simply delegates to the 'real' InterRepository
3388
calculated for (source, target._real_repository).
3391
_walk_to_common_revisions_batch_size = 50
3393
def __init__(self, source, target):
3394
InterRepository.__init__(self, source, target)
3395
self._real_inter = None
3398
def is_compatible(source, target):
3399
if isinstance(target, remote.RemoteRepository):
3403
def _ensure_real_inter(self):
3404
if self._real_inter is None:
3405
self.target._ensure_real()
3406
real_target = self.target._real_repository
3407
self._real_inter = InterRepository.get(self.source, real_target)
3408
# Make _real_inter use the RemoteRepository for get_parent_map
3409
self._real_inter.target_get_graph = self.target.get_graph
3410
self._real_inter.target_get_parent_map = self.target.get_parent_map
3412
def copy_content(self, revision_id=None):
3413
self._ensure_real_inter()
3414
self._real_inter.copy_content(revision_id=revision_id)
3416
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3417
self._ensure_real_inter()
3418
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3419
find_ghosts=find_ghosts)
3422
def _get_repo_format_to_test(self):
3426
class InterRemoteToOther(InterRepository):
3428
def __init__(self, source, target):
3429
InterRepository.__init__(self, source, target)
3430
self._real_inter = None
3433
def is_compatible(source, target):
3434
if not isinstance(source, remote.RemoteRepository):
3436
# Is source's model compatible with target's model?
3437
source._ensure_real()
3438
real_source = source._real_repository
3439
if isinstance(real_source, remote.RemoteRepository):
3440
raise NotImplementedError(
3441
"We don't support remote repos backed by remote repos yet.")
3442
return InterRepository._same_model(real_source, target)
3444
def _ensure_real_inter(self):
3445
if self._real_inter is None:
3446
self.source._ensure_real()
3447
real_source = self.source._real_repository
3448
self._real_inter = InterRepository.get(real_source, self.target)
3450
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3451
self._ensure_real_inter()
3452
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3453
find_ghosts=find_ghosts)
3455
def copy_content(self, revision_id=None):
3456
self._ensure_real_inter()
3457
self._real_inter.copy_content(revision_id=revision_id)
3460
def _get_repo_format_to_test(self):
3465
class InterPackToRemotePack(InterPackRepo):
3466
"""A specialisation of InterPackRepo for a target that is a
3469
This will use the get_parent_map RPC rather than plain readvs, and also
3470
uses an RPC for autopacking.
3473
_walk_to_common_revisions_batch_size = 50
3476
def is_compatible(source, target):
3477
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3478
if isinstance(source._format, RepositoryFormatPack):
3479
if isinstance(target, remote.RemoteRepository):
3480
target._ensure_real()
3481
if isinstance(target._real_repository._format,
3482
RepositoryFormatPack):
3483
if InterRepository._same_model(source, target):
3487
def _autopack(self):
3488
self.target.autopack()
3491
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3492
"""See InterRepository.fetch()."""
3493
# Always fetch using the generic streaming fetch code, to allow
3494
# streaming fetching into remote servers.
3495
from bzrlib.fetch import RepoFetcher
3496
fetcher = RepoFetcher(self.target, self.source, revision_id,
3498
return fetcher.count_copied, fetcher.failed_revisions
3500
def _get_target_pack_collection(self):
3501
return self.target._real_repository._pack_collection
3504
def _get_repo_format_to_test(self):
3508
InterRepository.register_optimiser(InterDifferingSerializer)
3509
InterRepository.register_optimiser(InterSameDataRepository)
3510
InterRepository.register_optimiser(InterWeaveRepo)
3511
InterRepository.register_optimiser(InterKnitRepo)
3512
InterRepository.register_optimiser(InterModel1and2)
3513
InterRepository.register_optimiser(InterKnit1and2)
3514
InterRepository.register_optimiser(InterPackRepo)
3515
InterRepository.register_optimiser(InterOtherToRemote)
3516
InterRepository.register_optimiser(InterRemoteToOther)
3517
InterRepository.register_optimiser(InterPackToRemotePack)
3520
class CopyConverter(object):
3521
"""A repository conversion tool which just performs a copy of the content.
3523
This is slow but quite reliable.
3526
def __init__(self, target_format):
3527
"""Create a CopyConverter.
3529
:param target_format: The format the resulting repository should be.
3531
self.target_format = target_format
3533
def convert(self, repo, pb):
3534
"""Perform the conversion of to_convert, giving feedback via pb.
3536
:param to_convert: The disk object to convert.
3537
:param pb: a progress bar to use for progress information.
3542
# this is only useful with metadir layouts - separated repo content.
3543
# trigger an assertion if not such
3544
repo._format.get_format_string()
3545
self.repo_dir = repo.bzrdir
3546
self.step('Moving repository to repository.backup')
3547
self.repo_dir.transport.move('repository', 'repository.backup')
3548
backup_transport = self.repo_dir.transport.clone('repository.backup')
3549
repo._format.check_conversion_target(self.target_format)
3550
self.source_repo = repo._format.open(self.repo_dir,
3552
_override_transport=backup_transport)
3553
self.step('Creating new repository')
3554
converted = self.target_format.initialize(self.repo_dir,
3555
self.source_repo.is_shared())
3556
converted.lock_write()
3558
self.step('Copying content into repository.')
3559
self.source_repo.copy_content_into(converted)
3562
self.step('Deleting old repository content.')
3563
self.repo_dir.transport.delete_tree('repository.backup')
3564
self.pb.note('repository converted')
3566
def step(self, message):
3567
"""Update the pb by a step."""
3569
self.pb.update(message, self.count, self.total)
3581
def _unescaper(match, _map=_unescape_map):
3582
code = match.group(1)
3586
if not code.startswith('#'):
3588
return unichr(int(code[1:])).encode('utf8')
3594
def _unescape_xml(data):
3595
"""Unescape predefined XML entities in a string of data."""
3597
if _unescape_re is None:
3598
_unescape_re = re.compile('\&([^;]*);')
3599
return _unescape_re.sub(_unescaper, data)
3602
class _VersionedFileChecker(object):
3604
def __init__(self, repository):
3605
self.repository = repository
3606
self.text_index = self.repository._generate_text_key_index()
3608
def calculate_file_version_parents(self, text_key):
3609
"""Calculate the correct parents for a file version according to
3612
parent_keys = self.text_index[text_key]
3613
if parent_keys == [_mod_revision.NULL_REVISION]:
3615
return tuple(parent_keys)
3617
def check_file_version_parents(self, texts, progress_bar=None):
3618
"""Check the parents stored in a versioned file are correct.
3620
It also detects file versions that are not referenced by their
3621
corresponding revision's inventory.
3623
:returns: A tuple of (wrong_parents, dangling_file_versions).
3624
wrong_parents is a dict mapping {revision_id: (stored_parents,
3625
correct_parents)} for each revision_id where the stored parents
3626
are not correct. dangling_file_versions is a set of (file_id,
3627
revision_id) tuples for versions that are present in this versioned
3628
file, but not used by the corresponding inventory.
3631
self.file_ids = set([file_id for file_id, _ in
3632
self.text_index.iterkeys()])
3633
# text keys is now grouped by file_id
3634
n_weaves = len(self.file_ids)
3635
files_in_revisions = {}
3636
revisions_of_files = {}
3637
n_versions = len(self.text_index)
3638
progress_bar.update('loading text store', 0, n_versions)
3639
parent_map = self.repository.texts.get_parent_map(self.text_index)
3640
# On unlistable transports this could well be empty/error...
3641
text_keys = self.repository.texts.keys()
3642
unused_keys = frozenset(text_keys) - set(self.text_index)
3643
for num, key in enumerate(self.text_index.iterkeys()):
3644
if progress_bar is not None:
3645
progress_bar.update('checking text graph', num, n_versions)
3646
correct_parents = self.calculate_file_version_parents(key)
3648
knit_parents = parent_map[key]
3649
except errors.RevisionNotPresent:
3652
if correct_parents != knit_parents:
3653
wrong_parents[key] = (knit_parents, correct_parents)
3654
return wrong_parents, unused_keys
3657
def _old_get_graph(repository, revision_id):
3658
"""DO NOT USE. That is all. I'm serious."""
3659
graph = repository.get_graph()
3660
revision_graph = dict(((key, value) for key, value in
3661
graph.iter_ancestry([revision_id]) if value is not None))
3662
return _strip_NULL_ghosts(revision_graph)
3665
def _strip_NULL_ghosts(revision_graph):
3666
"""Also don't use this. more compatibility code for unmigrated clients."""
3667
# Filter ghosts, and null:
3668
if _mod_revision.NULL_REVISION in revision_graph:
3669
del revision_graph[_mod_revision.NULL_REVISION]
3670
for key, parents in revision_graph.items():
3671
revision_graph[key] = tuple(parent for parent in parents if parent
3673
return revision_graph
3676
class StreamSink(object):
3677
"""An object that can insert a stream into a repository.
3679
This interface handles the complexity of reserialising inventories and
3680
revisions from different formats, and allows unidirectional insertion into
3681
stacked repositories without looking for the missing basis parents
3685
def __init__(self, target_repo):
3686
self.target_repo = target_repo
3688
def insert_stream(self, stream, src_format, resume_tokens):
3689
"""Insert a stream's content into the target repository.
3691
:param src_format: a bzr repository format.
3693
:return: a list of resume tokens and an iterable of keys additional
3694
items required before the insertion can be completed.
3696
self.target_repo.lock_write()
3699
self.target_repo.resume_write_group(resume_tokens)
3701
self.target_repo.start_write_group()
3703
# locked_insert_stream performs a commit|suspend.
3704
return self._locked_insert_stream(stream, src_format)
3706
self.target_repo.abort_write_group(suppress_errors=True)
3709
self.target_repo.unlock()
3711
def _locked_insert_stream(self, stream, src_format):
3712
to_serializer = self.target_repo._format._serializer
3713
src_serializer = src_format._serializer
3714
for substream_type, substream in stream:
3715
if substream_type == 'texts':
3716
self.target_repo.texts.insert_record_stream(substream)
3717
elif substream_type == 'inventories':
3718
if src_serializer == to_serializer:
3719
self.target_repo.inventories.insert_record_stream(
3722
self._extract_and_insert_inventories(
3723
substream, src_serializer)
3724
elif substream_type == 'revisions':
3725
# This may fallback to extract-and-insert more often than
3726
# required if the serializers are different only in terms of
3728
if src_serializer == to_serializer:
3729
self.target_repo.revisions.insert_record_stream(
3732
self._extract_and_insert_revisions(substream,
3734
elif substream_type == 'signatures':
3735
self.target_repo.signatures.insert_record_stream(substream)
3737
raise AssertionError('kaboom! %s' % (substream_type,))
3739
missing_keys = set()
3740
for prefix, versioned_file in (
3741
('texts', self.target_repo.texts),
3742
('inventories', self.target_repo.inventories),
3743
('revisions', self.target_repo.revisions),
3744
('signatures', self.target_repo.signatures),
3746
missing_keys.update((prefix,) + key for key in
3747
versioned_file.get_missing_compression_parent_keys())
3748
except NotImplementedError:
3749
# cannot even attempt suspending, and missing would have failed
3750
# during stream insertion.
3751
missing_keys = set()
3754
# suspend the write group and tell the caller what we is
3755
# missing. We know we can suspend or else we would not have
3756
# entered this code path. (All repositories that can handle
3757
# missing keys can handle suspending a write group).
3758
write_group_tokens = self.target_repo.suspend_write_group()
3759
return write_group_tokens, missing_keys
3760
self.target_repo.commit_write_group()
3763
def _extract_and_insert_inventories(self, substream, serializer):
3764
"""Generate a new inventory versionedfile in target, converting data.
3766
The inventory is retrieved from the source, (deserializing it), and
3767
stored in the target (reserializing it in a different format).
3769
for record in substream:
3770
bytes = record.get_bytes_as('fulltext')
3771
revision_id = record.key[0]
3772
inv = serializer.read_inventory_from_string(bytes, revision_id)
3773
parents = [key[0] for key in record.parents]
3774
self.target_repo.add_inventory(revision_id, inv, parents)
3776
def _extract_and_insert_revisions(self, substream, serializer):
3777
for record in substream:
3778
bytes = record.get_bytes_as('fulltext')
3779
revision_id = record.key[0]
3780
rev = serializer.read_revision_from_string(bytes)
3781
if rev.revision_id != revision_id:
3782
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
3783
self.target_repo.add_revision(revision_id, rev)
3786
if self.target_repo._fetch_reconcile:
3787
self.target_repo.reconcile()
added
removed
bzrlib/repository.py
