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 import registry
51
from bzrlib.decorators import needs_read_lock, needs_write_lock
52
from bzrlib.inter import InterObject
53
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
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 fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1098
"""Fetch the content required to construct revision_id from source.
1100
If revision_id is None all content is copied.
1101
:param find_ghosts: Find and copy revisions in the source that are
1102
ghosts in the target (and not reachable directly by walking out to
1103
the first-present revision in target from revision_id).
1105
# fast path same-url fetch operations
1106
if self.has_same_location(source):
1107
# check that last_revision is in 'from' and then return a
1109
if (revision_id is not None and
1110
not _mod_revision.is_null(revision_id)):
1111
self.get_revision(revision_id)
1113
# if there is no specific appropriate InterRepository, this will get
1114
# the InterRepository base class, which raises an
1115
# IncompatibleRepositories when asked to fetch.
1116
inter = InterRepository.get(source, self)
1117
return inter.fetch(revision_id=revision_id, pb=pb,
1118
find_ghosts=find_ghosts)
1120
def create_bundle(self, target, base, fileobj, format=None):
1121
return serializer.write_bundle(self, target, base, fileobj, format)
1123
def get_commit_builder(self, branch, parents, config, timestamp=None,
1124
timezone=None, committer=None, revprops=None,
1126
"""Obtain a CommitBuilder for this repository.
1128
:param branch: Branch to commit to.
1129
:param parents: Revision ids of the parents of the new revision.
1130
:param config: Configuration to use.
1131
:param timestamp: Optional timestamp recorded for commit.
1132
:param timezone: Optional timezone for timestamp.
1133
:param committer: Optional committer to set for commit.
1134
:param revprops: Optional dictionary of revision properties.
1135
:param revision_id: Optional revision id.
1137
result = self._commit_builder_class(self, parents, config,
1138
timestamp, timezone, committer, revprops, revision_id)
1139
self.start_write_group()
1143
if (self.control_files._lock_count == 1 and
1144
self.control_files._lock_mode == 'w'):
1145
if self._write_group is not None:
1146
self.abort_write_group()
1147
self.control_files.unlock()
1148
raise errors.BzrError(
1149
'Must end write groups before releasing write locks.')
1150
self.control_files.unlock()
1151
if self.control_files._lock_count == 0:
1152
self._inventory_entry_cache.clear()
1153
for repo in self._fallback_repositories:
1157
def clone(self, a_bzrdir, revision_id=None):
1158
"""Clone this repository into a_bzrdir using the current format.
1160
Currently no check is made that the format of this repository and
1161
the bzrdir format are compatible. FIXME RBC 20060201.
1163
:return: The newly created destination repository.
1165
# TODO: deprecate after 0.16; cloning this with all its settings is
1166
# probably not very useful -- mbp 20070423
1167
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1168
self.copy_content_into(dest_repo, revision_id)
1171
def start_write_group(self):
1172
"""Start a write group in the repository.
1174
Write groups are used by repositories which do not have a 1:1 mapping
1175
between file ids and backend store to manage the insertion of data from
1176
both fetch and commit operations.
1178
A write lock is required around the start_write_group/commit_write_group
1179
for the support of lock-requiring repository formats.
1181
One can only insert data into a repository inside a write group.
1185
if not self.is_write_locked():
1186
raise errors.NotWriteLocked(self)
1187
if self._write_group:
1188
raise errors.BzrError('already in a write group')
1189
self._start_write_group()
1190
# so we can detect unlock/relock - the write group is now entered.
1191
self._write_group = self.get_transaction()
1193
def _start_write_group(self):
1194
"""Template method for per-repository write group startup.
1196
This is called before the write group is considered to be
1201
def sprout(self, to_bzrdir, revision_id=None):
1202
"""Create a descendent repository for new development.
1204
Unlike clone, this does not copy the settings of the repository.
1206
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1207
dest_repo.fetch(self, revision_id=revision_id)
1210
def _create_sprouting_repo(self, a_bzrdir, shared):
1211
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1212
# use target default format.
1213
dest_repo = a_bzrdir.create_repository()
1215
# Most control formats need the repository to be specifically
1216
# created, but on some old all-in-one formats it's not needed
1218
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1219
except errors.UninitializableFormat:
1220
dest_repo = a_bzrdir.open_repository()
1224
def has_revision(self, revision_id):
1225
"""True if this repository has a copy of the revision."""
1226
return revision_id in self.has_revisions((revision_id,))
1229
def has_revisions(self, revision_ids):
1230
"""Probe to find out the presence of multiple revisions.
1232
:param revision_ids: An iterable of revision_ids.
1233
:return: A set of the revision_ids that were present.
1235
parent_map = self.revisions.get_parent_map(
1236
[(rev_id,) for rev_id in revision_ids])
1238
if _mod_revision.NULL_REVISION in revision_ids:
1239
result.add(_mod_revision.NULL_REVISION)
1240
result.update([key[0] for key in parent_map])
1244
def get_revision(self, revision_id):
1245
"""Return the Revision object for a named revision."""
1246
return self.get_revisions([revision_id])[0]
1249
def get_revision_reconcile(self, revision_id):
1250
"""'reconcile' helper routine that allows access to a revision always.
1252
This variant of get_revision does not cross check the weave graph
1253
against the revision one as get_revision does: but it should only
1254
be used by reconcile, or reconcile-alike commands that are correcting
1255
or testing the revision graph.
1257
return self._get_revisions([revision_id])[0]
1260
def get_revisions(self, revision_ids):
1261
"""Get many revisions at once."""
1262
return self._get_revisions(revision_ids)
1265
def _get_revisions(self, revision_ids):
1266
"""Core work logic to get many revisions without sanity checks."""
1267
for rev_id in revision_ids:
1268
if not rev_id or not isinstance(rev_id, basestring):
1269
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1270
keys = [(key,) for key in revision_ids]
1271
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1273
for record in stream:
1274
if record.storage_kind == 'absent':
1275
raise errors.NoSuchRevision(self, record.key[0])
1276
text = record.get_bytes_as('fulltext')
1277
rev = self._serializer.read_revision_from_string(text)
1278
revs[record.key[0]] = rev
1279
return [revs[revid] for revid in revision_ids]
1282
def get_revision_xml(self, revision_id):
1283
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1284
# would have already do it.
1285
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1286
rev = self.get_revision(revision_id)
1287
rev_tmp = cStringIO.StringIO()
1288
# the current serializer..
1289
self._serializer.write_revision(rev, rev_tmp)
1291
return rev_tmp.getvalue()
1293
def get_deltas_for_revisions(self, revisions):
1294
"""Produce a generator of revision deltas.
1296
Note that the input is a sequence of REVISIONS, not revision_ids.
1297
Trees will be held in memory until the generator exits.
1298
Each delta is relative to the revision's lefthand predecessor.
1300
required_trees = set()
1301
for revision in revisions:
1302
required_trees.add(revision.revision_id)
1303
required_trees.update(revision.parent_ids[:1])
1304
trees = dict((t.get_revision_id(), t) for
1305
t in self.revision_trees(required_trees))
1306
for revision in revisions:
1307
if not revision.parent_ids:
1308
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1310
old_tree = trees[revision.parent_ids[0]]
1311
yield trees[revision.revision_id].changes_from(old_tree)
1314
def get_revision_delta(self, revision_id):
1315
"""Return the delta for one revision.
1317
The delta is relative to the left-hand predecessor of the
1320
r = self.get_revision(revision_id)
1321
return list(self.get_deltas_for_revisions([r]))[0]
1324
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1325
signature = gpg_strategy.sign(plaintext)
1326
self.add_signature_text(revision_id, signature)
1329
def add_signature_text(self, revision_id, signature):
1330
self.signatures.add_lines((revision_id,), (),
1331
osutils.split_lines(signature))
1333
def find_text_key_references(self):
1334
"""Find the text key references within the repository.
1336
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1337
revision_ids. Each altered file-ids has the exact revision_ids that
1338
altered it listed explicitly.
1339
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1340
to whether they were referred to by the inventory of the
1341
revision_id that they contain. The inventory texts from all present
1342
revision ids are assessed to generate this report.
1344
revision_keys = self.revisions.keys()
1345
w = self.inventories
1346
pb = ui.ui_factory.nested_progress_bar()
1348
return self._find_text_key_references_from_xml_inventory_lines(
1349
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1353
def _find_text_key_references_from_xml_inventory_lines(self,
1355
"""Core routine for extracting references to texts from inventories.
1357
This performs the translation of xml lines to revision ids.
1359
:param line_iterator: An iterator of lines, origin_version_id
1360
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1361
to whether they were referred to by the inventory of the
1362
revision_id that they contain. Note that if that revision_id was
1363
not part of the line_iterator's output then False will be given -
1364
even though it may actually refer to that key.
1366
if not self._serializer.support_altered_by_hack:
1367
raise AssertionError(
1368
"_find_text_key_references_from_xml_inventory_lines only "
1369
"supported for branches which store inventory as unnested xml"
1370
", not on %r" % self)
1373
# this code needs to read every new line in every inventory for the
1374
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1375
# not present in one of those inventories is unnecessary but not
1376
# harmful because we are filtering by the revision id marker in the
1377
# inventory lines : we only select file ids altered in one of those
1378
# revisions. We don't need to see all lines in the inventory because
1379
# only those added in an inventory in rev X can contain a revision=X
1381
unescape_revid_cache = {}
1382
unescape_fileid_cache = {}
1384
# jam 20061218 In a big fetch, this handles hundreds of thousands
1385
# of lines, so it has had a lot of inlining and optimizing done.
1386
# Sorry that it is a little bit messy.
1387
# Move several functions to be local variables, since this is a long
1389
search = self._file_ids_altered_regex.search
1390
unescape = _unescape_xml
1391
setdefault = result.setdefault
1392
for line, line_key in line_iterator:
1393
match = search(line)
1396
# One call to match.group() returning multiple items is quite a
1397
# bit faster than 2 calls to match.group() each returning 1
1398
file_id, revision_id = match.group('file_id', 'revision_id')
1400
# Inlining the cache lookups helps a lot when you make 170,000
1401
# lines and 350k ids, versus 8.4 unique ids.
1402
# Using a cache helps in 2 ways:
1403
# 1) Avoids unnecessary decoding calls
1404
# 2) Re-uses cached strings, which helps in future set and
1406
# (2) is enough that removing encoding entirely along with
1407
# the cache (so we are using plain strings) results in no
1408
# performance improvement.
1410
revision_id = unescape_revid_cache[revision_id]
1412
unescaped = unescape(revision_id)
1413
unescape_revid_cache[revision_id] = unescaped
1414
revision_id = unescaped
1416
# Note that unconditionally unescaping means that we deserialise
1417
# every fileid, which for general 'pull' is not great, but we don't
1418
# really want to have some many fulltexts that this matters anyway.
1421
file_id = unescape_fileid_cache[file_id]
1423
unescaped = unescape(file_id)
1424
unescape_fileid_cache[file_id] = unescaped
1427
key = (file_id, revision_id)
1428
setdefault(key, False)
1429
if revision_id == line_key[-1]:
1433
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1435
"""Helper routine for fileids_altered_by_revision_ids.
1437
This performs the translation of xml lines to revision ids.
1439
:param line_iterator: An iterator of lines, origin_version_id
1440
:param revision_ids: The revision ids to filter for. This should be a
1441
set or other type which supports efficient __contains__ lookups, as
1442
the revision id from each parsed line will be looked up in the
1443
revision_ids filter.
1444
:return: a dictionary mapping altered file-ids to an iterable of
1445
revision_ids. Each altered file-ids has the exact revision_ids that
1446
altered it listed explicitly.
1449
setdefault = result.setdefault
1451
self._find_text_key_references_from_xml_inventory_lines(
1452
line_iterator).iterkeys():
1453
# once data is all ensured-consistent; then this is
1454
# if revision_id == version_id
1455
if key[-1:] in revision_ids:
1456
setdefault(key[0], set()).add(key[-1])
1459
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1460
"""Find the file ids and versions affected by revisions.
1462
:param revisions: an iterable containing revision ids.
1463
:param _inv_weave: The inventory weave from this repository or None.
1464
If None, the inventory weave will be opened automatically.
1465
:return: a dictionary mapping altered file-ids to an iterable of
1466
revision_ids. Each altered file-ids has the exact revision_ids that
1467
altered it listed explicitly.
1469
selected_keys = set((revid,) for revid in revision_ids)
1470
w = _inv_weave or self.inventories
1471
pb = ui.ui_factory.nested_progress_bar()
1473
return self._find_file_ids_from_xml_inventory_lines(
1474
w.iter_lines_added_or_present_in_keys(
1475
selected_keys, pb=pb),
1480
def iter_files_bytes(self, desired_files):
1481
"""Iterate through file versions.
1483
Files will not necessarily be returned in the order they occur in
1484
desired_files. No specific order is guaranteed.
1486
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1487
value supplied by the caller as part of desired_files. It should
1488
uniquely identify the file version in the caller's context. (Examples:
1489
an index number or a TreeTransform trans_id.)
1491
bytes_iterator is an iterable of bytestrings for the file. The
1492
kind of iterable and length of the bytestrings are unspecified, but for
1493
this implementation, it is a list of bytes produced by
1494
VersionedFile.get_record_stream().
1496
:param desired_files: a list of (file_id, revision_id, identifier)
1500
for file_id, revision_id, callable_data in desired_files:
1501
text_keys[(file_id, revision_id)] = callable_data
1502
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1503
if record.storage_kind == 'absent':
1504
raise errors.RevisionNotPresent(record.key, self)
1505
yield text_keys[record.key], record.get_bytes_as('fulltext')
1507
def _generate_text_key_index(self, text_key_references=None,
1509
"""Generate a new text key index for the repository.
1511
This is an expensive function that will take considerable time to run.
1513
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1514
list of parents, also text keys. When a given key has no parents,
1515
the parents list will be [NULL_REVISION].
1517
# All revisions, to find inventory parents.
1518
if ancestors is None:
1519
graph = self.get_graph()
1520
ancestors = graph.get_parent_map(self.all_revision_ids())
1521
if text_key_references is None:
1522
text_key_references = self.find_text_key_references()
1523
pb = ui.ui_factory.nested_progress_bar()
1525
return self._do_generate_text_key_index(ancestors,
1526
text_key_references, pb)
1530
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1531
"""Helper for _generate_text_key_index to avoid deep nesting."""
1532
revision_order = tsort.topo_sort(ancestors)
1533
invalid_keys = set()
1535
for revision_id in revision_order:
1536
revision_keys[revision_id] = set()
1537
text_count = len(text_key_references)
1538
# a cache of the text keys to allow reuse; costs a dict of all the
1539
# keys, but saves a 2-tuple for every child of a given key.
1541
for text_key, valid in text_key_references.iteritems():
1543
invalid_keys.add(text_key)
1545
revision_keys[text_key[1]].add(text_key)
1546
text_key_cache[text_key] = text_key
1547
del text_key_references
1549
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1550
NULL_REVISION = _mod_revision.NULL_REVISION
1551
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1552
# too small for large or very branchy trees. However, for 55K path
1553
# trees, it would be easy to use too much memory trivially. Ideally we
1554
# could gauge this by looking at available real memory etc, but this is
1555
# always a tricky proposition.
1556
inventory_cache = lru_cache.LRUCache(10)
1557
batch_size = 10 # should be ~150MB on a 55K path tree
1558
batch_count = len(revision_order) / batch_size + 1
1560
pb.update("Calculating text parents.", processed_texts, text_count)
1561
for offset in xrange(batch_count):
1562
to_query = revision_order[offset * batch_size:(offset + 1) *
1566
for rev_tree in self.revision_trees(to_query):
1567
revision_id = rev_tree.get_revision_id()
1568
parent_ids = ancestors[revision_id]
1569
for text_key in revision_keys[revision_id]:
1570
pb.update("Calculating text parents.", processed_texts)
1571
processed_texts += 1
1572
candidate_parents = []
1573
for parent_id in parent_ids:
1574
parent_text_key = (text_key[0], parent_id)
1576
check_parent = parent_text_key not in \
1577
revision_keys[parent_id]
1579
# the parent parent_id is a ghost:
1580
check_parent = False
1581
# truncate the derived graph against this ghost.
1582
parent_text_key = None
1584
# look at the parent commit details inventories to
1585
# determine possible candidates in the per file graph.
1588
inv = inventory_cache[parent_id]
1590
inv = self.revision_tree(parent_id).inventory
1591
inventory_cache[parent_id] = inv
1592
parent_entry = inv._byid.get(text_key[0], None)
1593
if parent_entry is not None:
1595
text_key[0], parent_entry.revision)
1597
parent_text_key = None
1598
if parent_text_key is not None:
1599
candidate_parents.append(
1600
text_key_cache[parent_text_key])
1601
parent_heads = text_graph.heads(candidate_parents)
1602
new_parents = list(parent_heads)
1603
new_parents.sort(key=lambda x:candidate_parents.index(x))
1604
if new_parents == []:
1605
new_parents = [NULL_REVISION]
1606
text_index[text_key] = new_parents
1608
for text_key in invalid_keys:
1609
text_index[text_key] = [NULL_REVISION]
1612
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1613
"""Get an iterable listing the keys of all the data introduced by a set
1616
The keys will be ordered so that the corresponding items can be safely
1617
fetched and inserted in that order.
1619
:returns: An iterable producing tuples of (knit-kind, file-id,
1620
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1621
'revisions'. file-id is None unless knit-kind is 'file'.
1623
# XXX: it's a bit weird to control the inventory weave caching in this
1624
# generator. Ideally the caching would be done in fetch.py I think. Or
1625
# maybe this generator should explicitly have the contract that it
1626
# should not be iterated until the previously yielded item has been
1628
inv_w = self.inventories
1630
# file ids that changed
1631
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1633
num_file_ids = len(file_ids)
1634
for file_id, altered_versions in file_ids.iteritems():
1635
if _files_pb is not None:
1636
_files_pb.update("fetch texts", count, num_file_ids)
1638
yield ("file", file_id, altered_versions)
1639
# We're done with the files_pb. Note that it finished by the caller,
1640
# just as it was created by the caller.
1644
yield ("inventory", None, revision_ids)
1647
# XXX: Note ATM no callers actually pay attention to this return
1648
# instead they just use the list of revision ids and ignore
1649
# missing sigs. Consider removing this work entirely
1650
revisions_with_signatures = set(self.signatures.get_parent_map(
1651
[(r,) for r in revision_ids]))
1652
revisions_with_signatures = set(
1653
[r for (r,) in revisions_with_signatures])
1654
revisions_with_signatures.intersection_update(revision_ids)
1655
yield ("signatures", None, revisions_with_signatures)
1658
yield ("revisions", None, revision_ids)
1661
def get_inventory(self, revision_id):
1662
"""Get Inventory object by revision id."""
1663
return self.iter_inventories([revision_id]).next()
1665
def iter_inventories(self, revision_ids):
1666
"""Get many inventories by revision_ids.
1668
This will buffer some or all of the texts used in constructing the
1669
inventories in memory, but will only parse a single inventory at a
1672
:return: An iterator of inventories.
1674
if ((None in revision_ids)
1675
or (_mod_revision.NULL_REVISION in revision_ids)):
1676
raise ValueError('cannot get null revision inventory')
1677
return self._iter_inventories(revision_ids)
1679
def _iter_inventories(self, revision_ids):
1680
"""single-document based inventory iteration."""
1681
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1682
yield self.deserialise_inventory(revision_id, text)
1684
def _iter_inventory_xmls(self, revision_ids):
1685
keys = [(revision_id,) for revision_id in revision_ids]
1686
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1688
for record in stream:
1689
if record.storage_kind != 'absent':
1690
text_chunks[record.key] = record.get_bytes_as('chunked')
1692
raise errors.NoSuchRevision(self, record.key)
1694
chunks = text_chunks.pop(key)
1695
yield ''.join(chunks), key[-1]
1697
def deserialise_inventory(self, revision_id, xml):
1698
"""Transform the xml into an inventory object.
1700
:param revision_id: The expected revision id of the inventory.
1701
:param xml: A serialised inventory.
1703
result = self._serializer.read_inventory_from_string(xml, revision_id,
1704
entry_cache=self._inventory_entry_cache)
1705
if result.revision_id != revision_id:
1706
raise AssertionError('revision id mismatch %s != %s' % (
1707
result.revision_id, revision_id))
1710
def serialise_inventory(self, inv):
1711
return self._serializer.write_inventory_to_string(inv)
1713
def _serialise_inventory_to_lines(self, inv):
1714
return self._serializer.write_inventory_to_lines(inv)
1716
def get_serializer_format(self):
1717
return self._serializer.format_num
1720
def get_inventory_xml(self, revision_id):
1721
"""Get inventory XML as a file object."""
1722
texts = self._iter_inventory_xmls([revision_id])
1724
text, revision_id = texts.next()
1725
except StopIteration:
1726
raise errors.HistoryMissing(self, 'inventory', revision_id)
1730
def get_inventory_sha1(self, revision_id):
1731
"""Return the sha1 hash of the inventory entry
1733
return self.get_revision(revision_id).inventory_sha1
1735
def iter_reverse_revision_history(self, revision_id):
1736
"""Iterate backwards through revision ids in the lefthand history
1738
:param revision_id: The revision id to start with. All its lefthand
1739
ancestors will be traversed.
1741
graph = self.get_graph()
1742
next_id = revision_id
1744
if next_id in (None, _mod_revision.NULL_REVISION):
1747
# Note: The following line may raise KeyError in the event of
1748
# truncated history. We decided not to have a try:except:raise
1749
# RevisionNotPresent here until we see a use for it, because of the
1750
# cost in an inner loop that is by its very nature O(history).
1751
# Robert Collins 20080326
1752
parents = graph.get_parent_map([next_id])[next_id]
1753
if len(parents) == 0:
1756
next_id = parents[0]
1759
def get_revision_inventory(self, revision_id):
1760
"""Return inventory of a past revision."""
1761
# TODO: Unify this with get_inventory()
1762
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1763
# must be the same as its revision, so this is trivial.
1764
if revision_id is None:
1765
# This does not make sense: if there is no revision,
1766
# then it is the current tree inventory surely ?!
1767
# and thus get_root_id() is something that looks at the last
1768
# commit on the branch, and the get_root_id is an inventory check.
1769
raise NotImplementedError
1770
# return Inventory(self.get_root_id())
1772
return self.get_inventory(revision_id)
1774
def is_shared(self):
1775
"""Return True if this repository is flagged as a shared repository."""
1776
raise NotImplementedError(self.is_shared)
1779
def reconcile(self, other=None, thorough=False):
1780
"""Reconcile this repository."""
1781
from bzrlib.reconcile import RepoReconciler
1782
reconciler = RepoReconciler(self, thorough=thorough)
1783
reconciler.reconcile()
1786
def _refresh_data(self):
1787
"""Helper called from lock_* to ensure coherency with disk.
1789
The default implementation does nothing; it is however possible
1790
for repositories to maintain loaded indices across multiple locks
1791
by checking inside their implementation of this method to see
1792
whether their indices are still valid. This depends of course on
1793
the disk format being validatable in this manner.
1797
def revision_tree(self, revision_id):
1798
"""Return Tree for a revision on this branch.
1800
`revision_id` may be NULL_REVISION for the empty tree revision.
1802
revision_id = _mod_revision.ensure_null(revision_id)
1803
# TODO: refactor this to use an existing revision object
1804
# so we don't need to read it in twice.
1805
if revision_id == _mod_revision.NULL_REVISION:
1806
return RevisionTree(self, Inventory(root_id=None),
1807
_mod_revision.NULL_REVISION)
1809
inv = self.get_revision_inventory(revision_id)
1810
return RevisionTree(self, inv, revision_id)
1812
def revision_trees(self, revision_ids):
1813
"""Return Tree for a revision on this branch.
1815
`revision_id` may not be None or 'null:'"""
1816
inventories = self.iter_inventories(revision_ids)
1817
for inv in inventories:
1818
yield RevisionTree(self, inv, inv.revision_id)
1821
def get_ancestry(self, revision_id, topo_sorted=True):
1822
"""Return a list of revision-ids integrated by a revision.
1824
The first element of the list is always None, indicating the origin
1825
revision. This might change when we have history horizons, or
1826
perhaps we should have a new API.
1828
This is topologically sorted.
1830
if _mod_revision.is_null(revision_id):
1832
if not self.has_revision(revision_id):
1833
raise errors.NoSuchRevision(self, revision_id)
1834
graph = self.get_graph()
1836
search = graph._make_breadth_first_searcher([revision_id])
1839
found, ghosts = search.next_with_ghosts()
1840
except StopIteration:
1843
if _mod_revision.NULL_REVISION in keys:
1844
keys.remove(_mod_revision.NULL_REVISION)
1846
parent_map = graph.get_parent_map(keys)
1847
keys = tsort.topo_sort(parent_map)
1848
return [None] + list(keys)
1851
"""Compress the data within the repository.
1853
This operation only makes sense for some repository types. For other
1854
types it should be a no-op that just returns.
1856
This stub method does not require a lock, but subclasses should use
1857
@needs_write_lock as this is a long running call its reasonable to
1858
implicitly lock for the user.
1862
@deprecated_method(one_six)
1863
def print_file(self, file, revision_id):
1864
"""Print `file` to stdout.
1866
FIXME RBC 20060125 as John Meinel points out this is a bad api
1867
- it writes to stdout, it assumes that that is valid etc. Fix
1868
by creating a new more flexible convenience function.
1870
tree = self.revision_tree(revision_id)
1871
# use inventory as it was in that revision
1872
file_id = tree.inventory.path2id(file)
1874
# TODO: jam 20060427 Write a test for this code path
1875
# it had a bug in it, and was raising the wrong
1877
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1878
tree.print_file(file_id)
1880
def get_transaction(self):
1881
return self.control_files.get_transaction()
1883
@deprecated_method(one_one)
1884
def get_parents(self, revision_ids):
1885
"""See StackedParentsProvider.get_parents"""
1886
parent_map = self.get_parent_map(revision_ids)
1887
return [parent_map.get(r, None) for r in revision_ids]
1889
def get_parent_map(self, revision_ids):
1890
"""See graph._StackedParentsProvider.get_parent_map"""
1891
# revisions index works in keys; this just works in revisions
1892
# therefore wrap and unwrap
1895
for revision_id in revision_ids:
1896
if revision_id == _mod_revision.NULL_REVISION:
1897
result[revision_id] = ()
1898
elif revision_id is None:
1899
raise ValueError('get_parent_map(None) is not valid')
1901
query_keys.append((revision_id ,))
1902
for ((revision_id,), parent_keys) in \
1903
self.revisions.get_parent_map(query_keys).iteritems():
1905
result[revision_id] = tuple(parent_revid
1906
for (parent_revid,) in parent_keys)
1908
result[revision_id] = (_mod_revision.NULL_REVISION,)
1911
def _make_parents_provider(self):
1914
def get_graph(self, other_repository=None):
1915
"""Return the graph walker for this repository format"""
1916
parents_provider = self._make_parents_provider()
1917
if (other_repository is not None and
1918
not self.has_same_location(other_repository)):
1919
parents_provider = graph._StackedParentsProvider(
1920
[parents_provider, other_repository._make_parents_provider()])
1921
return graph.Graph(parents_provider)
1923
def _get_versioned_file_checker(self):
1924
"""Return an object suitable for checking versioned files."""
1925
return _VersionedFileChecker(self)
1927
def revision_ids_to_search_result(self, result_set):
1928
"""Convert a set of revision ids to a graph SearchResult."""
1929
result_parents = set()
1930
for parents in self.get_graph().get_parent_map(
1931
result_set).itervalues():
1932
result_parents.update(parents)
1933
included_keys = result_set.intersection(result_parents)
1934
start_keys = result_set.difference(included_keys)
1935
exclude_keys = result_parents.difference(result_set)
1936
result = graph.SearchResult(start_keys, exclude_keys,
1937
len(result_set), result_set)
1941
def set_make_working_trees(self, new_value):
1942
"""Set the policy flag for making working trees when creating branches.
1944
This only applies to branches that use this repository.
1946
The default is 'True'.
1947
:param new_value: True to restore the default, False to disable making
1950
raise NotImplementedError(self.set_make_working_trees)
1952
def make_working_trees(self):
1953
"""Returns the policy for making working trees on new branches."""
1954
raise NotImplementedError(self.make_working_trees)
1957
def sign_revision(self, revision_id, gpg_strategy):
1958
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1959
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1962
def has_signature_for_revision_id(self, revision_id):
1963
"""Query for a revision signature for revision_id in the repository."""
1964
if not self.has_revision(revision_id):
1965
raise errors.NoSuchRevision(self, revision_id)
1966
sig_present = (1 == len(
1967
self.signatures.get_parent_map([(revision_id,)])))
1971
def get_signature_text(self, revision_id):
1972
"""Return the text for a signature."""
1973
stream = self.signatures.get_record_stream([(revision_id,)],
1975
record = stream.next()
1976
if record.storage_kind == 'absent':
1977
raise errors.NoSuchRevision(self, revision_id)
1978
return record.get_bytes_as('fulltext')
1981
def check(self, revision_ids=None):
1982
"""Check consistency of all history of given revision_ids.
1984
Different repository implementations should override _check().
1986
:param revision_ids: A non-empty list of revision_ids whose ancestry
1987
will be checked. Typically the last revision_id of a branch.
1989
return self._check(revision_ids)
1991
def _check(self, revision_ids):
1992
result = check.Check(self)
1996
def _warn_if_deprecated(self):
1997
global _deprecation_warning_done
1998
if _deprecation_warning_done:
2000
_deprecation_warning_done = True
2001
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
2002
% (self._format, self.bzrdir.transport.base))
2004
def supports_rich_root(self):
2005
return self._format.rich_root_data
2007
def _check_ascii_revisionid(self, revision_id, method):
2008
"""Private helper for ascii-only repositories."""
2009
# weave repositories refuse to store revisionids that are non-ascii.
2010
if revision_id is not None:
2011
# weaves require ascii revision ids.
2012
if isinstance(revision_id, unicode):
2014
revision_id.encode('ascii')
2015
except UnicodeEncodeError:
2016
raise errors.NonAsciiRevisionId(method, self)
2019
revision_id.decode('ascii')
2020
except UnicodeDecodeError:
2021
raise errors.NonAsciiRevisionId(method, self)
2023
def revision_graph_can_have_wrong_parents(self):
2024
"""Is it possible for this repository to have a revision graph with
2027
If True, then this repository must also implement
2028
_find_inconsistent_revision_parents so that check and reconcile can
2029
check for inconsistencies before proceeding with other checks that may
2030
depend on the revision index being consistent.
2032
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2035
# remove these delegates a while after bzr 0.15
2036
def __make_delegated(name, from_module):
2037
def _deprecated_repository_forwarder():
2038
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2039
% (name, from_module),
2042
m = __import__(from_module, globals(), locals(), [name])
2044
return getattr(m, name)
2045
except AttributeError:
2046
raise AttributeError('module %s has no name %s'
2048
globals()[name] = _deprecated_repository_forwarder
2051
'AllInOneRepository',
2052
'WeaveMetaDirRepository',
2053
'PreSplitOutRepositoryFormat',
2054
'RepositoryFormat4',
2055
'RepositoryFormat5',
2056
'RepositoryFormat6',
2057
'RepositoryFormat7',
2059
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2063
'RepositoryFormatKnit',
2064
'RepositoryFormatKnit1',
2066
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2069
def install_revision(repository, rev, revision_tree):
2070
"""Install all revision data into a repository."""
2071
install_revisions(repository, [(rev, revision_tree, None)])
2074
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2075
"""Install all revision data into a repository.
2077
Accepts an iterable of revision, tree, signature tuples. The signature
2080
repository.start_write_group()
2082
for n, (revision, revision_tree, signature) in enumerate(iterable):
2083
_install_revision(repository, revision, revision_tree, signature)
2085
pb.update('Transferring revisions', n + 1, num_revisions)
2087
repository.abort_write_group()
2090
repository.commit_write_group()
2093
def _install_revision(repository, rev, revision_tree, signature):
2094
"""Install all revision data into a repository."""
2095
present_parents = []
2097
for p_id in rev.parent_ids:
2098
if repository.has_revision(p_id):
2099
present_parents.append(p_id)
2100
parent_trees[p_id] = repository.revision_tree(p_id)
2102
parent_trees[p_id] = repository.revision_tree(
2103
_mod_revision.NULL_REVISION)
2105
inv = revision_tree.inventory
2106
entries = inv.iter_entries()
2107
# backwards compatibility hack: skip the root id.
2108
if not repository.supports_rich_root():
2109
path, root = entries.next()
2110
if root.revision != rev.revision_id:
2111
raise errors.IncompatibleRevision(repr(repository))
2113
for path, ie in entries:
2114
text_keys[(ie.file_id, ie.revision)] = ie
2115
text_parent_map = repository.texts.get_parent_map(text_keys)
2116
missing_texts = set(text_keys) - set(text_parent_map)
2117
# Add the texts that are not already present
2118
for text_key in missing_texts:
2119
ie = text_keys[text_key]
2121
# FIXME: TODO: The following loop overlaps/duplicates that done by
2122
# commit to determine parents. There is a latent/real bug here where
2123
# the parents inserted are not those commit would do - in particular
2124
# they are not filtered by heads(). RBC, AB
2125
for revision, tree in parent_trees.iteritems():
2126
if ie.file_id not in tree:
2128
parent_id = tree.inventory[ie.file_id].revision
2129
if parent_id in text_parents:
2131
text_parents.append((ie.file_id, parent_id))
2132
lines = revision_tree.get_file(ie.file_id).readlines()
2133
repository.texts.add_lines(text_key, text_parents, lines)
2135
# install the inventory
2136
repository.add_inventory(rev.revision_id, inv, present_parents)
2137
except errors.RevisionAlreadyPresent:
2139
if signature is not None:
2140
repository.add_signature_text(rev.revision_id, signature)
2141
repository.add_revision(rev.revision_id, rev, inv)
2144
class MetaDirRepository(Repository):
2145
"""Repositories in the new meta-dir layout.
2147
:ivar _transport: Transport for access to repository control files,
2148
typically pointing to .bzr/repository.
2151
def __init__(self, _format, a_bzrdir, control_files):
2152
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2153
self._transport = control_files._transport
2155
def is_shared(self):
2156
"""Return True if this repository is flagged as a shared repository."""
2157
return self._transport.has('shared-storage')
2160
def set_make_working_trees(self, new_value):
2161
"""Set the policy flag for making working trees when creating branches.
2163
This only applies to branches that use this repository.
2165
The default is 'True'.
2166
:param new_value: True to restore the default, False to disable making
2171
self._transport.delete('no-working-trees')
2172
except errors.NoSuchFile:
2175
self._transport.put_bytes('no-working-trees', '',
2176
mode=self.bzrdir._get_file_mode())
2178
def make_working_trees(self):
2179
"""Returns the policy for making working trees on new branches."""
2180
return not self._transport.has('no-working-trees')
2183
class MetaDirVersionedFileRepository(MetaDirRepository):
2184
"""Repositories in a meta-dir, that work via versioned file objects."""
2186
def __init__(self, _format, a_bzrdir, control_files):
2187
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2191
class RepositoryFormatRegistry(registry.Registry):
2192
"""Registry of RepositoryFormats."""
2194
def get(self, format_string):
2195
r = registry.Registry.get(self, format_string)
2201
format_registry = RepositoryFormatRegistry()
2202
"""Registry of formats, indexed by their identifying format string.
2204
This can contain either format instances themselves, or classes/factories that
2205
can be called to obtain one.
2209
#####################################################################
2210
# Repository Formats
2212
class RepositoryFormat(object):
2213
"""A repository format.
2215
Formats provide three things:
2216
* An initialization routine to construct repository data on disk.
2217
* a format string which is used when the BzrDir supports versioned
2219
* an open routine which returns a Repository instance.
2221
There is one and only one Format subclass for each on-disk format. But
2222
there can be one Repository subclass that is used for several different
2223
formats. The _format attribute on a Repository instance can be used to
2224
determine the disk format.
2226
Formats are placed in an dict by their format string for reference
2227
during opening. These should be subclasses of RepositoryFormat
2230
Once a format is deprecated, just deprecate the initialize and open
2231
methods on the format class. Do not deprecate the object, as the
2232
object will be created every system load.
2234
Common instance attributes:
2235
_matchingbzrdir - the bzrdir format that the repository format was
2236
originally written to work with. This can be used if manually
2237
constructing a bzrdir and repository, or more commonly for test suite
2241
# Set to True or False in derived classes. True indicates that the format
2242
# supports ghosts gracefully.
2243
supports_ghosts = None
2244
# Can this repository be given external locations to lookup additional
2245
# data. Set to True or False in derived classes.
2246
supports_external_lookups = None
2249
return "<%s>" % self.__class__.__name__
2251
def __eq__(self, other):
2252
# format objects are generally stateless
2253
return isinstance(other, self.__class__)
2255
def __ne__(self, other):
2256
return not self == other
2259
def find_format(klass, a_bzrdir):
2260
"""Return the format for the repository object in a_bzrdir.
2262
This is used by bzr native formats that have a "format" file in
2263
the repository. Other methods may be used by different types of
2267
transport = a_bzrdir.get_repository_transport(None)
2268
format_string = transport.get("format").read()
2269
return format_registry.get(format_string)
2270
except errors.NoSuchFile:
2271
raise errors.NoRepositoryPresent(a_bzrdir)
2273
raise errors.UnknownFormatError(format=format_string,
2277
def register_format(klass, format):
2278
format_registry.register(format.get_format_string(), format)
2281
def unregister_format(klass, format):
2282
format_registry.remove(format.get_format_string())
2285
def get_default_format(klass):
2286
"""Return the current default format."""
2287
from bzrlib import bzrdir
2288
return bzrdir.format_registry.make_bzrdir('default').repository_format
2290
def get_format_string(self):
2291
"""Return the ASCII format string that identifies this format.
2293
Note that in pre format ?? repositories the format string is
2294
not permitted nor written to disk.
2296
raise NotImplementedError(self.get_format_string)
2298
def get_format_description(self):
2299
"""Return the short description for this format."""
2300
raise NotImplementedError(self.get_format_description)
2302
# TODO: this shouldn't be in the base class, it's specific to things that
2303
# use weaves or knits -- mbp 20070207
2304
def _get_versioned_file_store(self,
2309
versionedfile_class=None,
2310
versionedfile_kwargs={},
2312
if versionedfile_class is None:
2313
versionedfile_class = self._versionedfile_class
2314
weave_transport = control_files._transport.clone(name)
2315
dir_mode = control_files._dir_mode
2316
file_mode = control_files._file_mode
2317
return VersionedFileStore(weave_transport, prefixed=prefixed,
2319
file_mode=file_mode,
2320
versionedfile_class=versionedfile_class,
2321
versionedfile_kwargs=versionedfile_kwargs,
2324
def initialize(self, a_bzrdir, shared=False):
2325
"""Initialize a repository of this format in a_bzrdir.
2327
:param a_bzrdir: The bzrdir to put the new repository in it.
2328
:param shared: The repository should be initialized as a sharable one.
2329
:returns: The new repository object.
2331
This may raise UninitializableFormat if shared repository are not
2332
compatible the a_bzrdir.
2334
raise NotImplementedError(self.initialize)
2336
def is_supported(self):
2337
"""Is this format supported?
2339
Supported formats must be initializable and openable.
2340
Unsupported formats may not support initialization or committing or
2341
some other features depending on the reason for not being supported.
2345
def check_conversion_target(self, target_format):
2346
raise NotImplementedError(self.check_conversion_target)
2348
def open(self, a_bzrdir, _found=False):
2349
"""Return an instance of this format for the bzrdir a_bzrdir.
2351
_found is a private parameter, do not use it.
2353
raise NotImplementedError(self.open)
2356
class MetaDirRepositoryFormat(RepositoryFormat):
2357
"""Common base class for the new repositories using the metadir layout."""
2359
rich_root_data = False
2360
supports_tree_reference = False
2361
supports_external_lookups = False
2364
def _matchingbzrdir(self):
2365
matching = bzrdir.BzrDirMetaFormat1()
2366
matching.repository_format = self
2370
super(MetaDirRepositoryFormat, self).__init__()
2372
def _create_control_files(self, a_bzrdir):
2373
"""Create the required files and the initial control_files object."""
2374
# FIXME: RBC 20060125 don't peek under the covers
2375
# NB: no need to escape relative paths that are url safe.
2376
repository_transport = a_bzrdir.get_repository_transport(self)
2377
control_files = lockable_files.LockableFiles(repository_transport,
2378
'lock', lockdir.LockDir)
2379
control_files.create_lock()
2380
return control_files
2382
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2383
"""Upload the initial blank content."""
2384
control_files = self._create_control_files(a_bzrdir)
2385
control_files.lock_write()
2386
transport = control_files._transport
2388
utf8_files += [('shared-storage', '')]
2390
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2391
for (filename, content_stream) in files:
2392
transport.put_file(filename, content_stream,
2393
mode=a_bzrdir._get_file_mode())
2394
for (filename, content_bytes) in utf8_files:
2395
transport.put_bytes_non_atomic(filename, content_bytes,
2396
mode=a_bzrdir._get_file_mode())
2398
control_files.unlock()
2401
# formats which have no format string are not discoverable
2402
# and not independently creatable, so are not registered. They're
2403
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2404
# needed, it's constructed directly by the BzrDir. Non-native formats where
2405
# the repository is not separately opened are similar.
2407
format_registry.register_lazy(
2408
'Bazaar-NG Repository format 7',
2409
'bzrlib.repofmt.weaverepo',
2413
format_registry.register_lazy(
2414
'Bazaar-NG Knit Repository Format 1',
2415
'bzrlib.repofmt.knitrepo',
2416
'RepositoryFormatKnit1',
2419
format_registry.register_lazy(
2420
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2421
'bzrlib.repofmt.knitrepo',
2422
'RepositoryFormatKnit3',
2425
format_registry.register_lazy(
2426
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2427
'bzrlib.repofmt.knitrepo',
2428
'RepositoryFormatKnit4',
2431
# Pack-based formats. There is one format for pre-subtrees, and one for
2432
# post-subtrees to allow ease of testing.
2433
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2434
format_registry.register_lazy(
2435
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2436
'bzrlib.repofmt.pack_repo',
2437
'RepositoryFormatKnitPack1',
2439
format_registry.register_lazy(
2440
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2441
'bzrlib.repofmt.pack_repo',
2442
'RepositoryFormatKnitPack3',
2444
format_registry.register_lazy(
2445
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2446
'bzrlib.repofmt.pack_repo',
2447
'RepositoryFormatKnitPack4',
2449
format_registry.register_lazy(
2450
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2451
'bzrlib.repofmt.pack_repo',
2452
'RepositoryFormatKnitPack5',
2454
format_registry.register_lazy(
2455
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2456
'bzrlib.repofmt.pack_repo',
2457
'RepositoryFormatKnitPack5RichRoot',
2459
format_registry.register_lazy(
2460
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2461
'bzrlib.repofmt.pack_repo',
2462
'RepositoryFormatKnitPack5RichRootBroken',
2464
format_registry.register_lazy(
2465
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2466
'bzrlib.repofmt.pack_repo',
2467
'RepositoryFormatKnitPack6',
2469
format_registry.register_lazy(
2470
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2471
'bzrlib.repofmt.pack_repo',
2472
'RepositoryFormatKnitPack6RichRoot',
2475
# Development formats.
2476
# 1.7->1.8 go below here
2477
format_registry.register_lazy(
2478
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2479
'bzrlib.repofmt.pack_repo',
2480
'RepositoryFormatPackDevelopment2',
2482
format_registry.register_lazy(
2483
("Bazaar development format 2 with subtree support "
2484
"(needs bzr.dev from before 1.8)\n"),
2485
'bzrlib.repofmt.pack_repo',
2486
'RepositoryFormatPackDevelopment2Subtree',
2490
class InterRepository(InterObject):
2491
"""This class represents operations taking place between two repositories.
2493
Its instances have methods like copy_content and fetch, and contain
2494
references to the source and target repositories these operations can be
2497
Often we will provide convenience methods on 'repository' which carry out
2498
operations with another repository - they will always forward to
2499
InterRepository.get(other).method_name(parameters).
2502
_walk_to_common_revisions_batch_size = 1
2504
"""The available optimised InterRepository types."""
2506
def __init__(self, source, target):
2507
InterObject.__init__(self, source, target)
2508
# These two attributes may be overridden by e.g. InterOtherToRemote to
2509
# provide a faster implementation.
2510
self.target_get_graph = self.target.get_graph
2511
self.target_get_parent_map = self.target.get_parent_map
2513
def copy_content(self, revision_id=None):
2514
raise NotImplementedError(self.copy_content)
2516
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2517
"""Fetch the content required to construct revision_id.
2519
The content is copied from self.source to self.target.
2521
:param revision_id: if None all content is copied, if NULL_REVISION no
2523
:param pb: optional progress bar to use for progress reports. If not
2524
provided a default one will be created.
2526
:returns: (copied_revision_count, failures).
2528
# Normally we should find a specific InterRepository subclass to do
2529
# the fetch; if nothing else then at least InterSameDataRepository.
2530
# If none of them is suitable it looks like fetching is not possible;
2531
# we try to give a good message why. _assert_same_model will probably
2532
# give a helpful message; otherwise a generic one.
2533
self._assert_same_model(self.source, self.target)
2534
raise errors.IncompatibleRepositories(self.source, self.target,
2535
"no suitableInterRepository found")
2537
def _walk_to_common_revisions(self, revision_ids):
2538
"""Walk out from revision_ids in source to revisions target has.
2540
:param revision_ids: The start point for the search.
2541
:return: A set of revision ids.
2543
target_graph = self.target_get_graph()
2544
revision_ids = frozenset(revision_ids)
2545
# Fast path for the case where all the revisions are already in the
2547
# (Although this does incur an extra round trip for the
2548
# fairly common case where the target doesn't already have the revision
2550
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2551
return graph.SearchResult(revision_ids, set(), 0, set())
2552
missing_revs = set()
2553
source_graph = self.source.get_graph()
2554
# ensure we don't pay silly lookup costs.
2555
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2556
null_set = frozenset([_mod_revision.NULL_REVISION])
2557
searcher_exhausted = False
2561
# Iterate the searcher until we have enough next_revs
2562
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2564
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2565
next_revs.update(next_revs_part)
2566
ghosts.update(ghosts_part)
2567
except StopIteration:
2568
searcher_exhausted = True
2570
# If there are ghosts in the source graph, and the caller asked for
2571
# them, make sure that they are present in the target.
2572
# We don't care about other ghosts as we can't fetch them and
2573
# haven't been asked to.
2574
ghosts_to_check = set(revision_ids.intersection(ghosts))
2575
revs_to_get = set(next_revs).union(ghosts_to_check)
2577
have_revs = set(target_graph.get_parent_map(revs_to_get))
2578
# we always have NULL_REVISION present.
2579
have_revs = have_revs.union(null_set)
2580
# Check if the target is missing any ghosts we need.
2581
ghosts_to_check.difference_update(have_revs)
2583
# One of the caller's revision_ids is a ghost in both the
2584
# source and the target.
2585
raise errors.NoSuchRevision(
2586
self.source, ghosts_to_check.pop())
2587
missing_revs.update(next_revs - have_revs)
2588
# Because we may have walked past the original stop point, make
2589
# sure everything is stopped
2590
stop_revs = searcher.find_seen_ancestors(have_revs)
2591
searcher.stop_searching_any(stop_revs)
2592
if searcher_exhausted:
2594
return searcher.get_result()
2596
@deprecated_method(one_two)
2598
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2599
"""Return the revision ids that source has that target does not.
2601
These are returned in topological order.
2603
:param revision_id: only return revision ids included by this
2605
:param find_ghosts: If True find missing revisions in deep history
2606
rather than just finding the surface difference.
2608
return list(self.search_missing_revision_ids(
2609
revision_id, find_ghosts).get_keys())
2612
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2613
"""Return the revision ids that source has that target does not.
2615
:param revision_id: only return revision ids included by this
2617
:param find_ghosts: If True find missing revisions in deep history
2618
rather than just finding the surface difference.
2619
:return: A bzrlib.graph.SearchResult.
2621
# stop searching at found target revisions.
2622
if not find_ghosts and revision_id is not None:
2623
return self._walk_to_common_revisions([revision_id])
2624
# generic, possibly worst case, slow code path.
2625
target_ids = set(self.target.all_revision_ids())
2626
if revision_id is not None:
2627
source_ids = self.source.get_ancestry(revision_id)
2628
if source_ids[0] is not None:
2629
raise AssertionError()
2632
source_ids = self.source.all_revision_ids()
2633
result_set = set(source_ids).difference(target_ids)
2634
return self.source.revision_ids_to_search_result(result_set)
2637
def _same_model(source, target):
2638
"""True if source and target have the same data representation.
2640
Note: this is always called on the base class; overriding it in a
2641
subclass will have no effect.
2644
InterRepository._assert_same_model(source, target)
2646
except errors.IncompatibleRepositories, e:
2650
def _assert_same_model(source, target):
2651
"""Raise an exception if two repositories do not use the same model.
2653
if source.supports_rich_root() != target.supports_rich_root():
2654
raise errors.IncompatibleRepositories(source, target,
2655
"different rich-root support")
2656
if source._serializer != target._serializer:
2657
raise errors.IncompatibleRepositories(source, target,
2658
"different serializers")
2661
class InterSameDataRepository(InterRepository):
2662
"""Code for converting between repositories that represent the same data.
2664
Data format and model must match for this to work.
2668
def _get_repo_format_to_test(self):
2669
"""Repository format for testing with.
2671
InterSameData can pull from subtree to subtree and from non-subtree to
2672
non-subtree, so we test this with the richest repository format.
2674
from bzrlib.repofmt import knitrepo
2675
return knitrepo.RepositoryFormatKnit3()
2678
def is_compatible(source, target):
2679
return InterRepository._same_model(source, target)
2682
def copy_content(self, revision_id=None):
2683
"""Make a complete copy of the content in self into destination.
2685
This copies both the repository's revision data, and configuration information
2686
such as the make_working_trees setting.
2688
This is a destructive operation! Do not use it on existing
2691
:param revision_id: Only copy the content needed to construct
2692
revision_id and its parents.
2695
self.target.set_make_working_trees(self.source.make_working_trees())
2696
except NotImplementedError:
2698
# but don't bother fetching if we have the needed data now.
2699
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2700
self.target.has_revision(revision_id)):
2702
self.target.fetch(self.source, revision_id=revision_id)
2705
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2706
"""See InterRepository.fetch()."""
2707
from bzrlib.fetch import RepoFetcher
2708
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2709
self.source, self.source._format, self.target,
2710
self.target._format)
2711
f = RepoFetcher(to_repository=self.target,
2712
from_repository=self.source,
2713
last_revision=revision_id,
2714
pb=pb, find_ghosts=find_ghosts)
2715
return f.count_copied, f.failed_revisions
2718
class InterWeaveRepo(InterSameDataRepository):
2719
"""Optimised code paths between Weave based repositories.
2721
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2722
implemented lazy inter-object optimisation.
2726
def _get_repo_format_to_test(self):
2727
from bzrlib.repofmt import weaverepo
2728
return weaverepo.RepositoryFormat7()
2731
def is_compatible(source, target):
2732
"""Be compatible with known Weave formats.
2734
We don't test for the stores being of specific types because that
2735
could lead to confusing results, and there is no need to be
2738
from bzrlib.repofmt.weaverepo import (
2744
return (isinstance(source._format, (RepositoryFormat5,
2746
RepositoryFormat7)) and
2747
isinstance(target._format, (RepositoryFormat5,
2749
RepositoryFormat7)))
2750
except AttributeError:
2754
def copy_content(self, revision_id=None):
2755
"""See InterRepository.copy_content()."""
2756
# weave specific optimised path:
2758
self.target.set_make_working_trees(self.source.make_working_trees())
2759
except (errors.RepositoryUpgradeRequired, NotImplemented):
2761
# FIXME do not peek!
2762
if self.source._transport.listable():
2763
pb = ui.ui_factory.nested_progress_bar()
2765
self.target.texts.insert_record_stream(
2766
self.source.texts.get_record_stream(
2767
self.source.texts.keys(), 'topological', False))
2768
pb.update('copying inventory', 0, 1)
2769
self.target.inventories.insert_record_stream(
2770
self.source.inventories.get_record_stream(
2771
self.source.inventories.keys(), 'topological', False))
2772
self.target.signatures.insert_record_stream(
2773
self.source.signatures.get_record_stream(
2774
self.source.signatures.keys(),
2776
self.target.revisions.insert_record_stream(
2777
self.source.revisions.get_record_stream(
2778
self.source.revisions.keys(),
2779
'topological', True))
2783
self.target.fetch(self.source, revision_id=revision_id)
2786
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2787
"""See InterRepository.fetch()."""
2788
from bzrlib.fetch import RepoFetcher
2789
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2790
self.source, self.source._format, self.target, self.target._format)
2791
f = RepoFetcher(to_repository=self.target,
2792
from_repository=self.source,
2793
last_revision=revision_id,
2794
pb=pb, find_ghosts=find_ghosts)
2795
return f.count_copied, f.failed_revisions
2798
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2799
"""See InterRepository.missing_revision_ids()."""
2800
# we want all revisions to satisfy revision_id in source.
2801
# but we don't want to stat every file here and there.
2802
# we want then, all revisions other needs to satisfy revision_id
2803
# checked, but not those that we have locally.
2804
# so the first thing is to get a subset of the revisions to
2805
# satisfy revision_id in source, and then eliminate those that
2806
# we do already have.
2807
# this is slow on high latency connection to self, but as as this
2808
# disk format scales terribly for push anyway due to rewriting
2809
# inventory.weave, this is considered acceptable.
2811
if revision_id is not None:
2812
source_ids = self.source.get_ancestry(revision_id)
2813
if source_ids[0] is not None:
2814
raise AssertionError()
2817
source_ids = self.source._all_possible_ids()
2818
source_ids_set = set(source_ids)
2819
# source_ids is the worst possible case we may need to pull.
2820
# now we want to filter source_ids against what we actually
2821
# have in target, but don't try to check for existence where we know
2822
# we do not have a revision as that would be pointless.
2823
target_ids = set(self.target._all_possible_ids())
2824
possibly_present_revisions = target_ids.intersection(source_ids_set)
2825
actually_present_revisions = set(
2826
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2827
required_revisions = source_ids_set.difference(actually_present_revisions)
2828
if revision_id is not None:
2829
# we used get_ancestry to determine source_ids then we are assured all
2830
# revisions referenced are present as they are installed in topological order.
2831
# and the tip revision was validated by get_ancestry.
2832
result_set = required_revisions
2834
# if we just grabbed the possibly available ids, then
2835
# we only have an estimate of whats available and need to validate
2836
# that against the revision records.
2838
self.source._eliminate_revisions_not_present(required_revisions))
2839
return self.source.revision_ids_to_search_result(result_set)
2842
class InterKnitRepo(InterSameDataRepository):
2843
"""Optimised code paths between Knit based repositories."""
2846
def _get_repo_format_to_test(self):
2847
from bzrlib.repofmt import knitrepo
2848
return knitrepo.RepositoryFormatKnit1()
2851
def is_compatible(source, target):
2852
"""Be compatible with known Knit formats.
2854
We don't test for the stores being of specific types because that
2855
could lead to confusing results, and there is no need to be
2858
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2860
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2861
isinstance(target._format, RepositoryFormatKnit))
2862
except AttributeError:
2864
return are_knits and InterRepository._same_model(source, target)
2867
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2868
"""See InterRepository.fetch()."""
2869
from bzrlib.fetch import RepoFetcher
2870
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2871
self.source, self.source._format, self.target, self.target._format)
2872
f = RepoFetcher(to_repository=self.target,
2873
from_repository=self.source,
2874
last_revision=revision_id,
2875
pb=pb, find_ghosts=find_ghosts)
2876
return f.count_copied, f.failed_revisions
2879
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2880
"""See InterRepository.missing_revision_ids()."""
2881
if revision_id is not None:
2882
source_ids = self.source.get_ancestry(revision_id)
2883
if source_ids[0] is not None:
2884
raise AssertionError()
2887
source_ids = self.source.all_revision_ids()
2888
source_ids_set = set(source_ids)
2889
# source_ids is the worst possible case we may need to pull.
2890
# now we want to filter source_ids against what we actually
2891
# have in target, but don't try to check for existence where we know
2892
# we do not have a revision as that would be pointless.
2893
target_ids = set(self.target.all_revision_ids())
2894
possibly_present_revisions = target_ids.intersection(source_ids_set)
2895
actually_present_revisions = set(
2896
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2897
required_revisions = source_ids_set.difference(actually_present_revisions)
2898
if revision_id is not None:
2899
# we used get_ancestry to determine source_ids then we are assured all
2900
# revisions referenced are present as they are installed in topological order.
2901
# and the tip revision was validated by get_ancestry.
2902
result_set = required_revisions
2904
# if we just grabbed the possibly available ids, then
2905
# we only have an estimate of whats available and need to validate
2906
# that against the revision records.
2908
self.source._eliminate_revisions_not_present(required_revisions))
2909
return self.source.revision_ids_to_search_result(result_set)
2912
class InterPackRepo(InterSameDataRepository):
2913
"""Optimised code paths between Pack based repositories."""
2916
def _get_repo_format_to_test(self):
2917
from bzrlib.repofmt import pack_repo
2918
return pack_repo.RepositoryFormatKnitPack1()
2921
def is_compatible(source, target):
2922
"""Be compatible with known Pack formats.
2924
We don't test for the stores being of specific types because that
2925
could lead to confusing results, and there is no need to be
2928
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2930
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2931
isinstance(target._format, RepositoryFormatPack))
2932
except AttributeError:
2934
return are_packs and InterRepository._same_model(source, target)
2937
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2938
"""See InterRepository.fetch()."""
2939
if (len(self.source._fallback_repositories) > 0 or
2940
len(self.target._fallback_repositories) > 0):
2941
# The pack layer is not aware of fallback repositories, so when
2942
# fetching from a stacked repository or into a stacked repository
2943
# we use the generic fetch logic which uses the VersionedFiles
2944
# attributes on repository.
2945
from bzrlib.fetch import RepoFetcher
2946
fetcher = RepoFetcher(self.target, self.source, revision_id,
2948
return fetcher.count_copied, fetcher.failed_revisions
2949
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2950
self.source, self.source._format, self.target, self.target._format)
2951
self.count_copied = 0
2952
if revision_id is None:
2954
# everything to do - use pack logic
2955
# to fetch from all packs to one without
2956
# inventory parsing etc, IFF nothing to be copied is in the target.
2958
source_revision_ids = frozenset(self.source.all_revision_ids())
2959
revision_ids = source_revision_ids - \
2960
frozenset(self.target_get_parent_map(source_revision_ids))
2961
revision_keys = [(revid,) for revid in revision_ids]
2962
target_pack_collection = self._get_target_pack_collection()
2963
index = target_pack_collection.revision_index.combined_index
2964
present_revision_ids = set(item[1][0] for item in
2965
index.iter_entries(revision_keys))
2966
revision_ids = set(revision_ids) - present_revision_ids
2967
# implementing the TODO will involve:
2968
# - detecting when all of a pack is selected
2969
# - avoiding as much as possible pre-selection, so the
2970
# more-core routines such as create_pack_from_packs can filter in
2971
# a just-in-time fashion. (though having a HEADS list on a
2972
# repository might make this a lot easier, because we could
2973
# sensibly detect 'new revisions' without doing a full index scan.
2974
elif _mod_revision.is_null(revision_id):
2979
revision_ids = self.search_missing_revision_ids(revision_id,
2980
find_ghosts=find_ghosts).get_keys()
2981
except errors.NoSuchRevision:
2982
raise errors.InstallFailed([revision_id])
2983
if len(revision_ids) == 0:
2985
return self._pack(self.source, self.target, revision_ids)
2987
def _pack(self, source, target, revision_ids):
2988
from bzrlib.repofmt.pack_repo import Packer
2989
target_pack_collection = self._get_target_pack_collection()
2990
packs = source._pack_collection.all_packs()
2991
pack = Packer(target_pack_collection, packs, '.fetch',
2992
revision_ids).pack()
2993
if pack is not None:
2994
target_pack_collection._save_pack_names()
2995
copied_revs = pack.get_revision_count()
2996
# Trigger an autopack. This may duplicate effort as we've just done
2997
# a pack creation, but for now it is simpler to think about as
2998
# 'upload data, then repack if needed'.
3000
return (copied_revs, [])
3004
def _autopack(self):
3005
self.target._pack_collection.autopack()
3007
def _get_target_pack_collection(self):
3008
return self.target._pack_collection
3011
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3012
"""See InterRepository.missing_revision_ids().
3014
:param find_ghosts: Find ghosts throughout the ancestry of
3017
if not find_ghosts and revision_id is not None:
3018
return self._walk_to_common_revisions([revision_id])
3019
elif revision_id is not None:
3020
# Find ghosts: search for revisions pointing from one repository to
3021
# the other, and vice versa, anywhere in the history of revision_id.
3022
graph = self.target_get_graph(other_repository=self.source)
3023
searcher = graph._make_breadth_first_searcher([revision_id])
3027
next_revs, ghosts = searcher.next_with_ghosts()
3028
except StopIteration:
3030
if revision_id in ghosts:
3031
raise errors.NoSuchRevision(self.source, revision_id)
3032
found_ids.update(next_revs)
3033
found_ids.update(ghosts)
3034
found_ids = frozenset(found_ids)
3035
# Double query here: should be able to avoid this by changing the
3036
# graph api further.
3037
result_set = found_ids - frozenset(
3038
self.target_get_parent_map(found_ids))
3040
source_ids = self.source.all_revision_ids()
3041
# source_ids is the worst possible case we may need to pull.
3042
# now we want to filter source_ids against what we actually
3043
# have in target, but don't try to check for existence where we know
3044
# we do not have a revision as that would be pointless.
3045
target_ids = set(self.target.all_revision_ids())
3046
result_set = set(source_ids).difference(target_ids)
3047
return self.source.revision_ids_to_search_result(result_set)
3050
class InterModel1and2(InterRepository):
3053
def _get_repo_format_to_test(self):
3057
def is_compatible(source, target):
3058
if not source.supports_rich_root() and target.supports_rich_root():
3064
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3065
"""See InterRepository.fetch()."""
3066
from bzrlib.fetch import Model1toKnit2Fetcher
3067
f = Model1toKnit2Fetcher(to_repository=self.target,
3068
from_repository=self.source,
3069
last_revision=revision_id,
3070
pb=pb, find_ghosts=find_ghosts)
3071
return f.count_copied, f.failed_revisions
3074
def copy_content(self, revision_id=None):
3075
"""Make a complete copy of the content in self into destination.
3077
This is a destructive operation! Do not use it on existing
3080
:param revision_id: Only copy the content needed to construct
3081
revision_id and its parents.
3084
self.target.set_make_working_trees(self.source.make_working_trees())
3085
except NotImplementedError:
3087
# but don't bother fetching if we have the needed data now.
3088
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
3089
self.target.has_revision(revision_id)):
3091
self.target.fetch(self.source, revision_id=revision_id)
3094
class InterKnit1and2(InterKnitRepo):
3097
def _get_repo_format_to_test(self):
3101
def is_compatible(source, target):
3102
"""Be compatible with Knit1 source and Knit3 target"""
3104
from bzrlib.repofmt.knitrepo import (
3105
RepositoryFormatKnit1,
3106
RepositoryFormatKnit3,
3108
from bzrlib.repofmt.pack_repo import (
3109
RepositoryFormatKnitPack1,
3110
RepositoryFormatKnitPack3,
3111
RepositoryFormatKnitPack4,
3112
RepositoryFormatKnitPack5,
3113
RepositoryFormatKnitPack5RichRoot,
3114
RepositoryFormatKnitPack6,
3115
RepositoryFormatKnitPack6RichRoot,
3116
RepositoryFormatPackDevelopment2,
3117
RepositoryFormatPackDevelopment2Subtree,
3120
RepositoryFormatKnit1, # no rr, no subtree
3121
RepositoryFormatKnitPack1, # no rr, no subtree
3122
RepositoryFormatPackDevelopment2, # no rr, no subtree
3123
RepositoryFormatKnitPack5, # no rr, no subtree
3124
RepositoryFormatKnitPack6, # no rr, no subtree
3127
RepositoryFormatKnit3, # rr, subtree
3128
RepositoryFormatKnitPack3, # rr, subtree
3129
RepositoryFormatKnitPack4, # rr, no subtree
3130
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
3131
RepositoryFormatKnitPack6RichRoot,# rr, no subtree
3132
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
3134
for format in norichroot:
3135
if format.rich_root_data:
3136
raise AssertionError('Format %s is a rich-root format'
3137
' but is included in the non-rich-root list'
3139
for format in richroot:
3140
if not format.rich_root_data:
3141
raise AssertionError('Format %s is not a rich-root format'
3142
' but is included in the rich-root list'
3144
# TODO: One alternative is to just check format.rich_root_data,
3145
# instead of keeping membership lists. However, the formats
3146
# *also* have to use the same 'Knit' style of storage
3147
# (line-deltas, fulltexts, etc.)
3148
return (isinstance(source._format, norichroot) and
3149
isinstance(target._format, richroot))
3150
except AttributeError:
3154
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3155
"""See InterRepository.fetch()."""
3156
from bzrlib.fetch import Knit1to2Fetcher
3157
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3158
self.source, self.source._format, self.target,
3159
self.target._format)
3160
f = Knit1to2Fetcher(to_repository=self.target,
3161
from_repository=self.source,
3162
last_revision=revision_id,
3163
pb=pb, find_ghosts=find_ghosts)
3164
return f.count_copied, f.failed_revisions
3167
class InterDifferingSerializer(InterKnitRepo):
3170
def _get_repo_format_to_test(self):
3174
def is_compatible(source, target):
3175
"""Be compatible with Knit2 source and Knit3 target"""
3176
if source.supports_rich_root() != target.supports_rich_root():
3178
# Ideally, we'd support fetching if the source had no tree references
3179
# even if it supported them...
3180
if (getattr(source, '_format.supports_tree_reference', False) and
3181
not getattr(target, '_format.supports_tree_reference', False)):
3185
def _fetch_batch(self, revision_ids, basis_id, basis_tree):
3186
"""Fetch across a few revisions.
3188
:param revision_ids: The revisions to copy
3189
:param basis_id: The revision_id of basis_tree
3190
:param basis_tree: A tree that is not in revision_ids which should
3191
already exist in the target.
3192
:return: (basis_id, basis_tree) A new basis to use now that these trees
3195
# Walk though all revisions; get inventory deltas, copy referenced
3196
# texts that delta references, insert the delta, revision and
3200
pending_revisions = []
3201
for tree in self.source.revision_trees(revision_ids):
3202
current_revision_id = tree.get_revision_id()
3203
delta = tree.inventory._make_delta(basis_tree.inventory)
3204
for old_path, new_path, file_id, entry in delta:
3205
if new_path is not None:
3206
if not (new_path or self.target.supports_rich_root()):
3207
# We leave the inventory delta in, because that
3208
# will have the deserialised inventory root
3212
# "if entry.revision == current_revision_id" ?
3213
if entry.revision == current_revision_id:
3214
text_keys.add((file_id, entry.revision))
3215
revision = self.source.get_revision(current_revision_id)
3216
pending_deltas.append((basis_id, delta,
3217
current_revision_id, revision.parent_ids))
3218
pending_revisions.append(revision)
3219
basis_id = current_revision_id
3222
from_texts = self.source.texts
3223
to_texts = self.target.texts
3224
to_texts.insert_record_stream(from_texts.get_record_stream(
3225
text_keys, self.target._fetch_order,
3226
not self.target._fetch_uses_deltas))
3228
for delta in pending_deltas:
3229
self.target.add_inventory_by_delta(*delta)
3230
# insert signatures and revisions
3231
for revision in pending_revisions:
3233
signature = self.source.get_signature_text(
3234
revision.revision_id)
3235
self.target.add_signature_text(revision.revision_id,
3237
except errors.NoSuchRevision:
3239
self.target.add_revision(revision.revision_id, revision)
3240
return basis_id, basis_tree
3242
def _fetch_all_revisions(self, revision_ids, pb):
3243
"""Fetch everything for the list of revisions.
3245
:param revision_ids: The list of revisions to fetch. Must be in
3247
:param pb: A ProgressBar
3250
basis_id, basis_tree = self._get_basis(revision_ids[0])
3252
for offset in range(0, len(revision_ids), batch_size):
3253
self.target.start_write_group()
3255
pb.update('Transferring revisions', offset,
3257
batch = revision_ids[offset:offset+batch_size]
3258
basis_id, basis_tree = self._fetch_batch(batch,
3259
basis_id, basis_tree)
3261
self.target.abort_write_group()
3264
self.target.commit_write_group()
3265
pb.update('Transferring revisions', len(revision_ids),
3269
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3270
"""See InterRepository.fetch()."""
3271
revision_ids = self.target.search_missing_revision_ids(self.source,
3272
revision_id, find_ghosts=find_ghosts).get_keys()
3273
if not revision_ids:
3275
revision_ids = tsort.topo_sort(
3276
self.source.get_graph().get_parent_map(revision_ids))
3278
my_pb = ui.ui_factory.nested_progress_bar()
3283
self._fetch_all_revisions(revision_ids, pb)
3285
if my_pb is not None:
3287
return len(revision_ids), 0
3289
def _get_basis(self, first_revision_id):
3290
"""Get a revision and tree which exists in the target.
3292
This assumes that first_revision_id is selected for transmission
3293
because all other ancestors are already present. If we can't find an
3294
ancestor we fall back to NULL_REVISION since we know that is safe.
3296
:return: (basis_id, basis_tree)
3298
first_rev = self.source.get_revision(first_revision_id)
3300
basis_id = first_rev.parent_ids[0]
3301
# only valid as a basis if the target has it
3302
self.target.get_revision(basis_id)
3303
# Try to get a basis tree - if its a ghost it will hit the
3304
# NoSuchRevision case.
3305
basis_tree = self.source.revision_tree(basis_id)
3306
except (IndexError, errors.NoSuchRevision):
3307
basis_id = _mod_revision.NULL_REVISION
3308
basis_tree = self.source.revision_tree(basis_id)
3309
return basis_id, basis_tree
3312
class InterOtherToRemote(InterRepository):
3313
"""An InterRepository that simply delegates to the 'real' InterRepository
3314
calculated for (source, target._real_repository).
3317
_walk_to_common_revisions_batch_size = 50
3319
def __init__(self, source, target):
3320
InterRepository.__init__(self, source, target)
3321
self._real_inter = None
3324
def is_compatible(source, target):
3325
if isinstance(target, remote.RemoteRepository):
3329
def _ensure_real_inter(self):
3330
if self._real_inter is None:
3331
self.target._ensure_real()
3332
real_target = self.target._real_repository
3333
self._real_inter = InterRepository.get(self.source, real_target)
3334
# Make _real_inter use the RemoteRepository for get_parent_map
3335
self._real_inter.target_get_graph = self.target.get_graph
3336
self._real_inter.target_get_parent_map = self.target.get_parent_map
3338
def copy_content(self, revision_id=None):
3339
self._ensure_real_inter()
3340
self._real_inter.copy_content(revision_id=revision_id)
3342
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3343
self._ensure_real_inter()
3344
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3345
find_ghosts=find_ghosts)
3348
def _get_repo_format_to_test(self):
3352
class InterRemoteToOther(InterRepository):
3354
def __init__(self, source, target):
3355
InterRepository.__init__(self, source, target)
3356
self._real_inter = None
3359
def is_compatible(source, target):
3360
if not isinstance(source, remote.RemoteRepository):
3362
# Is source's model compatible with target's model?
3363
source._ensure_real()
3364
real_source = source._real_repository
3365
if isinstance(real_source, remote.RemoteRepository):
3366
raise NotImplementedError(
3367
"We don't support remote repos backed by remote repos yet.")
3368
return InterRepository._same_model(real_source, target)
3370
def _ensure_real_inter(self):
3371
if self._real_inter is None:
3372
self.source._ensure_real()
3373
real_source = self.source._real_repository
3374
self._real_inter = InterRepository.get(real_source, self.target)
3376
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3377
self._ensure_real_inter()
3378
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3379
find_ghosts=find_ghosts)
3381
def copy_content(self, revision_id=None):
3382
self._ensure_real_inter()
3383
self._real_inter.copy_content(revision_id=revision_id)
3386
def _get_repo_format_to_test(self):
3391
class InterPackToRemotePack(InterPackRepo):
3392
"""A specialisation of InterPackRepo for a target that is a
3395
This will use the get_parent_map RPC rather than plain readvs, and also
3396
uses an RPC for autopacking.
3399
_walk_to_common_revisions_batch_size = 50
3402
def is_compatible(source, target):
3403
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3404
if isinstance(source._format, RepositoryFormatPack):
3405
if isinstance(target, remote.RemoteRepository):
3406
target._ensure_real()
3407
if isinstance(target._real_repository._format,
3408
RepositoryFormatPack):
3409
if InterRepository._same_model(source, target):
3413
def _autopack(self):
3414
self.target.autopack()
3416
def _get_target_pack_collection(self):
3417
return self.target._real_repository._pack_collection
3420
def _get_repo_format_to_test(self):
3424
InterRepository.register_optimiser(InterDifferingSerializer)
3425
InterRepository.register_optimiser(InterSameDataRepository)
3426
InterRepository.register_optimiser(InterWeaveRepo)
3427
InterRepository.register_optimiser(InterKnitRepo)
3428
InterRepository.register_optimiser(InterModel1and2)
3429
InterRepository.register_optimiser(InterKnit1and2)
3430
InterRepository.register_optimiser(InterPackRepo)
3431
InterRepository.register_optimiser(InterOtherToRemote)
3432
InterRepository.register_optimiser(InterRemoteToOther)
3433
InterRepository.register_optimiser(InterPackToRemotePack)
3436
class CopyConverter(object):
3437
"""A repository conversion tool which just performs a copy of the content.
3439
This is slow but quite reliable.
3442
def __init__(self, target_format):
3443
"""Create a CopyConverter.
3445
:param target_format: The format the resulting repository should be.
3447
self.target_format = target_format
3449
def convert(self, repo, pb):
3450
"""Perform the conversion of to_convert, giving feedback via pb.
3452
:param to_convert: The disk object to convert.
3453
:param pb: a progress bar to use for progress information.
3458
# this is only useful with metadir layouts - separated repo content.
3459
# trigger an assertion if not such
3460
repo._format.get_format_string()
3461
self.repo_dir = repo.bzrdir
3462
self.step('Moving repository to repository.backup')
3463
self.repo_dir.transport.move('repository', 'repository.backup')
3464
backup_transport = self.repo_dir.transport.clone('repository.backup')
3465
repo._format.check_conversion_target(self.target_format)
3466
self.source_repo = repo._format.open(self.repo_dir,
3468
_override_transport=backup_transport)
3469
self.step('Creating new repository')
3470
converted = self.target_format.initialize(self.repo_dir,
3471
self.source_repo.is_shared())
3472
converted.lock_write()
3474
self.step('Copying content into repository.')
3475
self.source_repo.copy_content_into(converted)
3478
self.step('Deleting old repository content.')
3479
self.repo_dir.transport.delete_tree('repository.backup')
3480
self.pb.note('repository converted')
3482
def step(self, message):
3483
"""Update the pb by a step."""
3485
self.pb.update(message, self.count, self.total)
3497
def _unescaper(match, _map=_unescape_map):
3498
code = match.group(1)
3502
if not code.startswith('#'):
3504
return unichr(int(code[1:])).encode('utf8')
3510
def _unescape_xml(data):
3511
"""Unescape predefined XML entities in a string of data."""
3513
if _unescape_re is None:
3514
_unescape_re = re.compile('\&([^;]*);')
3515
return _unescape_re.sub(_unescaper, data)
3518
class _VersionedFileChecker(object):
3520
def __init__(self, repository):
3521
self.repository = repository
3522
self.text_index = self.repository._generate_text_key_index()
3524
def calculate_file_version_parents(self, text_key):
3525
"""Calculate the correct parents for a file version according to
3528
parent_keys = self.text_index[text_key]
3529
if parent_keys == [_mod_revision.NULL_REVISION]:
3531
return tuple(parent_keys)
3533
def check_file_version_parents(self, texts, progress_bar=None):
3534
"""Check the parents stored in a versioned file are correct.
3536
It also detects file versions that are not referenced by their
3537
corresponding revision's inventory.
3539
:returns: A tuple of (wrong_parents, dangling_file_versions).
3540
wrong_parents is a dict mapping {revision_id: (stored_parents,
3541
correct_parents)} for each revision_id where the stored parents
3542
are not correct. dangling_file_versions is a set of (file_id,
3543
revision_id) tuples for versions that are present in this versioned
3544
file, but not used by the corresponding inventory.
3547
self.file_ids = set([file_id for file_id, _ in
3548
self.text_index.iterkeys()])
3549
# text keys is now grouped by file_id
3550
n_weaves = len(self.file_ids)
3551
files_in_revisions = {}
3552
revisions_of_files = {}
3553
n_versions = len(self.text_index)
3554
progress_bar.update('loading text store', 0, n_versions)
3555
parent_map = self.repository.texts.get_parent_map(self.text_index)
3556
# On unlistable transports this could well be empty/error...
3557
text_keys = self.repository.texts.keys()
3558
unused_keys = frozenset(text_keys) - set(self.text_index)
3559
for num, key in enumerate(self.text_index.iterkeys()):
3560
if progress_bar is not None:
3561
progress_bar.update('checking text graph', num, n_versions)
3562
correct_parents = self.calculate_file_version_parents(key)
3564
knit_parents = parent_map[key]
3565
except errors.RevisionNotPresent:
3568
if correct_parents != knit_parents:
3569
wrong_parents[key] = (knit_parents, correct_parents)
3570
return wrong_parents, unused_keys
3573
def _old_get_graph(repository, revision_id):
3574
"""DO NOT USE. That is all. I'm serious."""
3575
graph = repository.get_graph()
3576
revision_graph = dict(((key, value) for key, value in
3577
graph.iter_ancestry([revision_id]) if value is not None))
3578
return _strip_NULL_ghosts(revision_graph)
3581
def _strip_NULL_ghosts(revision_graph):
3582
"""Also don't use this. more compatibility code for unmigrated clients."""
3583
# Filter ghosts, and null:
3584
if _mod_revision.NULL_REVISION in revision_graph:
3585
del revision_graph[_mod_revision.NULL_REVISION]
3586
for key, parents in revision_graph.items():
3587
revision_graph[key] = tuple(parent for parent in parents if parent
3589
return revision_graph
added
removed
bzrlib/repository.py
