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(), """
37
revision as _mod_revision,
43
from bzrlib.bundle import serializer
44
from bzrlib.revisiontree import RevisionTree
45
from bzrlib.store.versioned import VersionedFileStore
46
from bzrlib.testament import Testament
49
from bzrlib import registry
50
from bzrlib.decorators import needs_read_lock, needs_write_lock
51
from bzrlib.inter import InterObject
52
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
53
from bzrlib.symbol_versioning import (
59
from bzrlib.trace import (
60
log_exception_quietly, note, mutter, mutter_callsite, warning)
63
# Old formats display a warning, but only once
64
_deprecation_warning_done = False
67
class CommitBuilder(object):
68
"""Provides an interface to build up a commit.
70
This allows describing a tree to be committed without needing to
71
know the internals of the format of the repository.
74
# all clients should supply tree roots.
75
record_root_entry = True
76
# the default CommitBuilder does not manage trees whose root is versioned.
77
_versioned_root = False
79
def __init__(self, repository, parents, config, timestamp=None,
80
timezone=None, committer=None, revprops=None,
82
"""Initiate a CommitBuilder.
84
:param repository: Repository to commit to.
85
:param parents: Revision ids of the parents of the new revision.
86
:param config: Configuration to use.
87
:param timestamp: Optional timestamp recorded for commit.
88
:param timezone: Optional timezone for timestamp.
89
:param committer: Optional committer to set for commit.
90
:param revprops: Optional dictionary of revision properties.
91
:param revision_id: Optional revision id.
96
self._committer = self._config.username()
98
self._committer = committer
100
self.new_inventory = Inventory(None)
101
self._new_revision_id = revision_id
102
self.parents = parents
103
self.repository = repository
106
if revprops is not None:
107
self._validate_revprops(revprops)
108
self._revprops.update(revprops)
110
if timestamp is None:
111
timestamp = time.time()
112
# Restrict resolution to 1ms
113
self._timestamp = round(timestamp, 3)
116
self._timezone = osutils.local_time_offset()
118
self._timezone = int(timezone)
120
self._generate_revision_if_needed()
121
self.__heads = graph.HeadsCache(repository.get_graph()).heads
122
self._basis_delta = []
123
# API compatibility, older code that used CommitBuilder did not call
124
# .record_delete(), which means the delta that is computed would not be
125
# valid. Callers that will call record_delete() should call
126
# .will_record_deletes() to indicate that.
127
self._recording_deletes = False
129
def _validate_unicode_text(self, text, context):
130
"""Verify things like commit messages don't have bogus characters."""
132
raise ValueError('Invalid value for %s: %r' % (context, text))
134
def _validate_revprops(self, revprops):
135
for key, value in revprops.iteritems():
136
# We know that the XML serializers do not round trip '\r'
137
# correctly, so refuse to accept them
138
if not isinstance(value, basestring):
139
raise ValueError('revision property (%s) is not a valid'
140
' (unicode) string: %r' % (key, value))
141
self._validate_unicode_text(value,
142
'revision property (%s)' % (key,))
144
def commit(self, message):
145
"""Make the actual commit.
147
:return: The revision id of the recorded revision.
149
self._validate_unicode_text(message, 'commit message')
150
rev = _mod_revision.Revision(
151
timestamp=self._timestamp,
152
timezone=self._timezone,
153
committer=self._committer,
155
inventory_sha1=self.inv_sha1,
156
revision_id=self._new_revision_id,
157
properties=self._revprops)
158
rev.parent_ids = self.parents
159
self.repository.add_revision(self._new_revision_id, rev,
160
self.new_inventory, self._config)
161
self.repository.commit_write_group()
162
return self._new_revision_id
165
"""Abort the commit that is being built.
167
self.repository.abort_write_group()
169
def revision_tree(self):
170
"""Return the tree that was just committed.
172
After calling commit() this can be called to get a RevisionTree
173
representing the newly committed tree. This is preferred to
174
calling Repository.revision_tree() because that may require
175
deserializing the inventory, while we already have a copy in
178
return RevisionTree(self.repository, self.new_inventory,
179
self._new_revision_id)
181
def finish_inventory(self):
182
"""Tell the builder that the inventory is finished."""
183
if self.new_inventory.root is None:
184
raise AssertionError('Root entry should be supplied to'
185
' record_entry_contents, as of bzr 0.10.')
186
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
187
self.new_inventory.revision_id = self._new_revision_id
188
self.inv_sha1 = self.repository.add_inventory(
189
self._new_revision_id,
194
def _gen_revision_id(self):
195
"""Return new revision-id."""
196
return generate_ids.gen_revision_id(self._config.username(),
199
def _generate_revision_if_needed(self):
200
"""Create a revision id if None was supplied.
202
If the repository can not support user-specified revision ids
203
they should override this function and raise CannotSetRevisionId
204
if _new_revision_id is not None.
206
:raises: CannotSetRevisionId
208
if self._new_revision_id is None:
209
self._new_revision_id = self._gen_revision_id()
210
self.random_revid = True
212
self.random_revid = False
214
def _heads(self, file_id, revision_ids):
215
"""Calculate the graph heads for revision_ids in the graph of file_id.
217
This can use either a per-file graph or a global revision graph as we
218
have an identity relationship between the two graphs.
220
return self.__heads(revision_ids)
222
def _check_root(self, ie, parent_invs, tree):
223
"""Helper for record_entry_contents.
225
:param ie: An entry being added.
226
:param parent_invs: The inventories of the parent revisions of the
228
:param tree: The tree that is being committed.
230
# In this revision format, root entries have no knit or weave When
231
# serializing out to disk and back in root.revision is always
233
ie.revision = self._new_revision_id
235
def _get_delta(self, ie, basis_inv, path):
236
"""Get a delta against the basis inventory for ie."""
237
if ie.file_id not in basis_inv:
239
result = (None, path, ie.file_id, ie)
240
self._basis_delta.append(result)
242
elif ie != basis_inv[ie.file_id]:
244
# TODO: avoid tis id2path call.
245
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
246
self._basis_delta.append(result)
252
def get_basis_delta(self):
253
"""Return the complete inventory delta versus the basis inventory.
255
This has been built up with the calls to record_delete and
256
record_entry_contents. The client must have already called
257
will_record_deletes() to indicate that they will be generating a
260
:return: An inventory delta, suitable for use with apply_delta, or
261
Repository.add_inventory_by_delta, etc.
263
if not self._recording_deletes:
264
raise AssertionError("recording deletes not activated.")
265
return self._basis_delta
267
def record_delete(self, path, file_id):
268
"""Record that a delete occured against a basis tree.
270
This is an optional API - when used it adds items to the basis_delta
271
being accumulated by the commit builder. It cannot be called unless the
272
method will_record_deletes() has been called to inform the builder that
273
a delta is being supplied.
275
:param path: The path of the thing deleted.
276
:param file_id: The file id that was deleted.
278
if not self._recording_deletes:
279
raise AssertionError("recording deletes not activated.")
280
self._basis_delta.append((path, None, file_id, None))
282
def will_record_deletes(self):
283
"""Tell the commit builder that deletes are being notified.
285
This enables the accumulation of an inventory delta; for the resulting
286
commit to be valid, deletes against the basis MUST be recorded via
287
builder.record_delete().
289
self._recording_deletes = True
291
def record_entry_contents(self, ie, parent_invs, path, tree,
293
"""Record the content of ie from tree into the commit if needed.
295
Side effect: sets ie.revision when unchanged
297
:param ie: An inventory entry present in the commit.
298
:param parent_invs: The inventories of the parent revisions of the
300
:param path: The path the entry is at in the tree.
301
:param tree: The tree which contains this entry and should be used to
303
:param content_summary: Summary data from the tree about the paths
304
content - stat, length, exec, sha/link target. This is only
305
accessed when the entry has a revision of None - that is when it is
306
a candidate to commit.
307
:return: A tuple (change_delta, version_recorded, fs_hash).
308
change_delta is an inventory_delta change for this entry against
309
the basis tree of the commit, or None if no change occured against
311
version_recorded is True if a new version of the entry has been
312
recorded. For instance, committing a merge where a file was only
313
changed on the other side will return (delta, False).
314
fs_hash is either None, or the hash details for the path (currently
315
a tuple of the contents sha1 and the statvalue returned by
316
tree.get_file_with_stat()).
318
if self.new_inventory.root is None:
319
if ie.parent_id is not None:
320
raise errors.RootMissing()
321
self._check_root(ie, parent_invs, tree)
322
if ie.revision is None:
323
kind = content_summary[0]
325
# ie is carried over from a prior commit
327
# XXX: repository specific check for nested tree support goes here - if
328
# the repo doesn't want nested trees we skip it ?
329
if (kind == 'tree-reference' and
330
not self.repository._format.supports_tree_reference):
331
# mismatch between commit builder logic and repository:
332
# this needs the entry creation pushed down into the builder.
333
raise NotImplementedError('Missing repository subtree support.')
334
self.new_inventory.add(ie)
336
# TODO: slow, take it out of the inner loop.
338
basis_inv = parent_invs[0]
340
basis_inv = Inventory(root_id=None)
342
# ie.revision is always None if the InventoryEntry is considered
343
# for committing. We may record the previous parents revision if the
344
# content is actually unchanged against a sole head.
345
if ie.revision is not None:
346
if not self._versioned_root and path == '':
347
# repositories that do not version the root set the root's
348
# revision to the new commit even when no change occurs (more
349
# specifically, they do not record a revision on the root; and
350
# the rev id is assigned to the root during deserialisation -
351
# this masks when a change may have occurred against the basis.
352
# To match this we always issue a delta, because the revision
353
# of the root will always be changing.
354
if ie.file_id in basis_inv:
355
delta = (basis_inv.id2path(ie.file_id), path,
359
delta = (None, path, ie.file_id, ie)
360
self._basis_delta.append(delta)
361
return delta, False, None
363
# we don't need to commit this, because the caller already
364
# determined that an existing revision of this file is
365
# appropriate. If its not being considered for committing then
366
# it and all its parents to the root must be unaltered so
367
# no-change against the basis.
368
if ie.revision == self._new_revision_id:
369
raise AssertionError("Impossible situation, a skipped "
370
"inventory entry (%r) claims to be modified in this "
371
"commit (%r).", (ie, self._new_revision_id))
372
return None, False, None
373
# XXX: Friction: parent_candidates should return a list not a dict
374
# so that we don't have to walk the inventories again.
375
parent_candiate_entries = ie.parent_candidates(parent_invs)
376
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
378
for inv in parent_invs:
379
if ie.file_id in inv:
380
old_rev = inv[ie.file_id].revision
381
if old_rev in head_set:
382
heads.append(inv[ie.file_id].revision)
383
head_set.remove(inv[ie.file_id].revision)
386
# now we check to see if we need to write a new record to the
388
# We write a new entry unless there is one head to the ancestors, and
389
# the kind-derived content is unchanged.
391
# Cheapest check first: no ancestors, or more the one head in the
392
# ancestors, we write a new node.
396
# There is a single head, look it up for comparison
397
parent_entry = parent_candiate_entries[heads[0]]
398
# if the non-content specific data has changed, we'll be writing a
400
if (parent_entry.parent_id != ie.parent_id or
401
parent_entry.name != ie.name):
403
# now we need to do content specific checks:
405
# if the kind changed the content obviously has
406
if kind != parent_entry.kind:
408
# Stat cache fingerprint feedback for the caller - None as we usually
409
# don't generate one.
412
if content_summary[2] is None:
413
raise ValueError("Files must not have executable = None")
415
if (# if the file length changed we have to store:
416
parent_entry.text_size != content_summary[1] or
417
# if the exec bit has changed we have to store:
418
parent_entry.executable != content_summary[2]):
420
elif parent_entry.text_sha1 == content_summary[3]:
421
# all meta and content is unchanged (using a hash cache
422
# hit to check the sha)
423
ie.revision = parent_entry.revision
424
ie.text_size = parent_entry.text_size
425
ie.text_sha1 = parent_entry.text_sha1
426
ie.executable = parent_entry.executable
427
return self._get_delta(ie, basis_inv, path), False, None
429
# Either there is only a hash change(no hash cache entry,
430
# or same size content change), or there is no change on
432
# Provide the parent's hash to the store layer, so that the
433
# content is unchanged we will not store a new node.
434
nostore_sha = parent_entry.text_sha1
436
# We want to record a new node regardless of the presence or
437
# absence of a content change in the file.
439
ie.executable = content_summary[2]
440
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
442
lines = file_obj.readlines()
446
ie.text_sha1, ie.text_size = self._add_text_to_weave(
447
ie.file_id, lines, heads, nostore_sha)
448
# Let the caller know we generated a stat fingerprint.
449
fingerprint = (ie.text_sha1, stat_value)
450
except errors.ExistingContent:
451
# Turns out that the file content was unchanged, and we were
452
# only going to store a new node if it was changed. Carry over
454
ie.revision = parent_entry.revision
455
ie.text_size = parent_entry.text_size
456
ie.text_sha1 = parent_entry.text_sha1
457
ie.executable = parent_entry.executable
458
return self._get_delta(ie, basis_inv, path), False, None
459
elif kind == 'directory':
461
# all data is meta here, nothing specific to directory, so
463
ie.revision = parent_entry.revision
464
return self._get_delta(ie, basis_inv, path), False, None
466
self._add_text_to_weave(ie.file_id, lines, heads, None)
467
elif kind == 'symlink':
468
current_link_target = content_summary[3]
470
# symlink target is not generic metadata, check if it has
472
if current_link_target != parent_entry.symlink_target:
475
# unchanged, carry over.
476
ie.revision = parent_entry.revision
477
ie.symlink_target = parent_entry.symlink_target
478
return self._get_delta(ie, basis_inv, path), False, None
479
ie.symlink_target = current_link_target
481
self._add_text_to_weave(ie.file_id, lines, heads, None)
482
elif kind == 'tree-reference':
484
if content_summary[3] != parent_entry.reference_revision:
487
# unchanged, carry over.
488
ie.reference_revision = parent_entry.reference_revision
489
ie.revision = parent_entry.revision
490
return self._get_delta(ie, basis_inv, path), False, None
491
ie.reference_revision = content_summary[3]
493
self._add_text_to_weave(ie.file_id, lines, heads, None)
495
raise NotImplementedError('unknown kind')
496
ie.revision = self._new_revision_id
497
return self._get_delta(ie, basis_inv, path), True, fingerprint
499
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
500
# Note: as we read the content directly from the tree, we know its not
501
# been turned into unicode or badly split - but a broken tree
502
# implementation could give us bad output from readlines() so this is
503
# not a guarantee of safety. What would be better is always checking
504
# the content during test suite execution. RBC 20070912
505
parent_keys = tuple((file_id, parent) for parent in parents)
506
return self.repository.texts.add_lines(
507
(file_id, self._new_revision_id), parent_keys, new_lines,
508
nostore_sha=nostore_sha, random_id=self.random_revid,
509
check_content=False)[0:2]
512
class RootCommitBuilder(CommitBuilder):
513
"""This commitbuilder actually records the root id"""
515
# the root entry gets versioned properly by this builder.
516
_versioned_root = True
518
def _check_root(self, ie, parent_invs, tree):
519
"""Helper for record_entry_contents.
521
:param ie: An entry being added.
522
:param parent_invs: The inventories of the parent revisions of the
524
:param tree: The tree that is being committed.
528
######################################################################
531
class Repository(object):
532
"""Repository holding history for one or more branches.
534
The repository holds and retrieves historical information including
535
revisions and file history. It's normally accessed only by the Branch,
536
which views a particular line of development through that history.
538
The Repository builds on top of some byte storage facilies (the revisions,
539
signatures, inventories and texts attributes) and a Transport, which
540
respectively provide byte storage and a means to access the (possibly
543
The byte storage facilities are addressed via tuples, which we refer to
544
as 'keys' throughout the code base. Revision_keys, inventory_keys and
545
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
546
(file_id, revision_id). We use this interface because it allows low
547
friction with the underlying code that implements disk indices, network
548
encoding and other parts of bzrlib.
550
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
551
the serialised revisions for the repository. This can be used to obtain
552
revision graph information or to access raw serialised revisions.
553
The result of trying to insert data into the repository via this store
554
is undefined: it should be considered read-only except for implementors
556
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
557
the serialised signatures for the repository. This can be used to
558
obtain access to raw serialised signatures. The result of trying to
559
insert data into the repository via this store is undefined: it should
560
be considered read-only except for implementors of repositories.
561
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
562
the serialised inventories for the repository. This can be used to
563
obtain unserialised inventories. The result of trying to insert data
564
into the repository via this store is undefined: it should be
565
considered read-only except for implementors of repositories.
566
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
567
texts of files and directories for the repository. This can be used to
568
obtain file texts or file graphs. Note that Repository.iter_file_bytes
569
is usually a better interface for accessing file texts.
570
The result of trying to insert data into the repository via this store
571
is undefined: it should be considered read-only except for implementors
573
:ivar _transport: Transport for file access to repository, typically
574
pointing to .bzr/repository.
577
# What class to use for a CommitBuilder. Often its simpler to change this
578
# in a Repository class subclass rather than to override
579
# get_commit_builder.
580
_commit_builder_class = CommitBuilder
581
# The search regex used by xml based repositories to determine what things
582
# where changed in a single commit.
583
_file_ids_altered_regex = lazy_regex.lazy_compile(
584
r'file_id="(?P<file_id>[^"]+)"'
585
r'.* revision="(?P<revision_id>[^"]+)"'
588
def abort_write_group(self, suppress_errors=False):
589
"""Commit the contents accrued within the current write group.
591
:param suppress_errors: if true, abort_write_group will catch and log
592
unexpected errors that happen during the abort, rather than
593
allowing them to propagate. Defaults to False.
595
:seealso: start_write_group.
597
if self._write_group is not self.get_transaction():
598
# has an unlock or relock occured ?
599
raise errors.BzrError('mismatched lock context and write group.')
601
self._abort_write_group()
602
except Exception, exc:
603
self._write_group = None
604
if not suppress_errors:
606
mutter('abort_write_group failed')
607
log_exception_quietly()
608
note('bzr: ERROR (ignored): %s', exc)
609
self._write_group = None
611
def _abort_write_group(self):
612
"""Template method for per-repository write group cleanup.
614
This is called during abort before the write group is considered to be
615
finished and should cleanup any internal state accrued during the write
616
group. There is no requirement that data handed to the repository be
617
*not* made available - this is not a rollback - but neither should any
618
attempt be made to ensure that data added is fully commited. Abort is
619
invoked when an error has occured so futher disk or network operations
620
may not be possible or may error and if possible should not be
624
def add_fallback_repository(self, repository):
625
"""Add a repository to use for looking up data not held locally.
627
:param repository: A repository.
629
if not self._format.supports_external_lookups:
630
raise errors.UnstackableRepositoryFormat(self._format, self.base)
631
self._check_fallback_repository(repository)
632
self._fallback_repositories.append(repository)
633
self.texts.add_fallback_versioned_files(repository.texts)
634
self.inventories.add_fallback_versioned_files(repository.inventories)
635
self.revisions.add_fallback_versioned_files(repository.revisions)
636
self.signatures.add_fallback_versioned_files(repository.signatures)
638
def _check_fallback_repository(self, repository):
639
"""Check that this repository can fallback to repository safely.
641
Raise an error if not.
643
:param repository: A repository to fallback to.
645
return InterRepository._assert_same_model(self, repository)
647
def add_inventory(self, revision_id, inv, parents):
648
"""Add the inventory inv to the repository as revision_id.
650
:param parents: The revision ids of the parents that revision_id
651
is known to have and are in the repository already.
653
:returns: The validator(which is a sha1 digest, though what is sha'd is
654
repository format specific) of the serialized inventory.
656
if not self.is_in_write_group():
657
raise AssertionError("%r not in write group" % (self,))
658
_mod_revision.check_not_reserved_id(revision_id)
659
if not (inv.revision_id is None or inv.revision_id == revision_id):
660
raise AssertionError(
661
"Mismatch between inventory revision"
662
" id and insertion revid (%r, %r)"
663
% (inv.revision_id, revision_id))
665
raise AssertionError()
666
inv_lines = self._serialise_inventory_to_lines(inv)
667
return self._inventory_add_lines(revision_id, parents,
668
inv_lines, check_content=False)
670
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
672
"""Add a new inventory expressed as a delta against another revision.
674
:param basis_revision_id: The inventory id the delta was created
675
against. (This does not have to be a direct parent.)
676
:param delta: The inventory delta (see Inventory.apply_delta for
678
:param new_revision_id: The revision id that the inventory is being
680
:param parents: The revision ids of the parents that revision_id is
681
known to have and are in the repository already. These are supplied
682
for repositories that depend on the inventory graph for revision
683
graph access, as well as for those that pun ancestry with delta
686
:returns: The validator(which is a sha1 digest, though what is sha'd is
687
repository format specific) of the serialized inventory.
689
if not self.is_in_write_group():
690
raise AssertionError("%r not in write group" % (self,))
691
_mod_revision.check_not_reserved_id(new_revision_id)
692
basis_tree = self.revision_tree(basis_revision_id)
693
basis_tree.lock_read()
695
# Note that this mutates the inventory of basis_tree, which not all
696
# inventory implementations may support: A better idiom would be to
697
# return a new inventory, but as there is no revision tree cache in
698
# repository this is safe for now - RBC 20081013
699
basis_inv = basis_tree.inventory
700
basis_inv.apply_delta(delta)
701
basis_inv.revision_id = new_revision_id
702
return self.add_inventory(new_revision_id, basis_inv, parents)
706
def _inventory_add_lines(self, revision_id, parents, lines,
708
"""Store lines in inv_vf and return the sha1 of the inventory."""
709
parents = [(parent,) for parent in parents]
710
return self.inventories.add_lines((revision_id,), parents, lines,
711
check_content=check_content)[0]
713
def add_revision(self, revision_id, rev, inv=None, config=None):
714
"""Add rev to the revision store as revision_id.
716
:param revision_id: the revision id to use.
717
:param rev: The revision object.
718
:param inv: The inventory for the revision. if None, it will be looked
719
up in the inventory storer
720
:param config: If None no digital signature will be created.
721
If supplied its signature_needed method will be used
722
to determine if a signature should be made.
724
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
726
_mod_revision.check_not_reserved_id(revision_id)
727
if config is not None and config.signature_needed():
729
inv = self.get_inventory(revision_id)
730
plaintext = Testament(rev, inv).as_short_text()
731
self.store_revision_signature(
732
gpg.GPGStrategy(config), plaintext, revision_id)
733
# check inventory present
734
if not self.inventories.get_parent_map([(revision_id,)]):
736
raise errors.WeaveRevisionNotPresent(revision_id,
739
# yes, this is not suitable for adding with ghosts.
740
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
744
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
745
self._add_revision(rev)
747
def _add_revision(self, revision):
748
text = self._serializer.write_revision_to_string(revision)
749
key = (revision.revision_id,)
750
parents = tuple((parent,) for parent in revision.parent_ids)
751
self.revisions.add_lines(key, parents, osutils.split_lines(text))
753
def all_revision_ids(self):
754
"""Returns a list of all the revision ids in the repository.
756
This is conceptually deprecated because code should generally work on
757
the graph reachable from a particular revision, and ignore any other
758
revisions that might be present. There is no direct replacement
761
if 'evil' in debug.debug_flags:
762
mutter_callsite(2, "all_revision_ids is linear with history.")
763
return self._all_revision_ids()
765
def _all_revision_ids(self):
766
"""Returns a list of all the revision ids in the repository.
768
These are in as much topological order as the underlying store can
771
raise NotImplementedError(self._all_revision_ids)
773
def break_lock(self):
774
"""Break a lock if one is present from another instance.
776
Uses the ui factory to ask for confirmation if the lock may be from
779
self.control_files.break_lock()
782
def _eliminate_revisions_not_present(self, revision_ids):
783
"""Check every revision id in revision_ids to see if we have it.
785
Returns a set of the present revisions.
788
graph = self.get_graph()
789
parent_map = graph.get_parent_map(revision_ids)
790
# The old API returned a list, should this actually be a set?
791
return parent_map.keys()
794
def create(a_bzrdir):
795
"""Construct the current default format repository in a_bzrdir."""
796
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
798
def __init__(self, _format, a_bzrdir, control_files):
799
"""instantiate a Repository.
801
:param _format: The format of the repository on disk.
802
:param a_bzrdir: The BzrDir of the repository.
804
In the future we will have a single api for all stores for
805
getting file texts, inventories and revisions, then
806
this construct will accept instances of those things.
808
super(Repository, self).__init__()
809
self._format = _format
810
# the following are part of the public API for Repository:
811
self.bzrdir = a_bzrdir
812
self.control_files = control_files
813
self._transport = control_files._transport
814
self.base = self._transport.base
816
self._reconcile_does_inventory_gc = True
817
self._reconcile_fixes_text_parents = False
818
self._reconcile_backsup_inventory = True
819
# not right yet - should be more semantically clear ?
821
# TODO: make sure to construct the right store classes, etc, depending
822
# on whether escaping is required.
823
self._warn_if_deprecated()
824
self._write_group = None
825
# Additional places to query for data.
826
self._fallback_repositories = []
827
# What order should fetch operations request streams in?
828
# The default is unordered as that is the cheapest for an origin to
830
self._fetch_order = 'unordered'
831
# Does this repository use deltas that can be fetched as-deltas ?
832
# (E.g. knits, where the knit deltas can be transplanted intact.
833
# We default to False, which will ensure that enough data to get
834
# a full text out of any fetch stream will be grabbed.
835
self._fetch_uses_deltas = False
836
# Should fetch trigger a reconcile after the fetch? Only needed for
837
# some repository formats that can suffer internal inconsistencies.
838
self._fetch_reconcile = False
841
return '%s(%r)' % (self.__class__.__name__,
844
def has_same_location(self, other):
845
"""Returns a boolean indicating if this repository is at the same
846
location as another repository.
848
This might return False even when two repository objects are accessing
849
the same physical repository via different URLs.
851
if self.__class__ is not other.__class__:
853
return (self._transport.base == other._transport.base)
855
def is_in_write_group(self):
856
"""Return True if there is an open write group.
858
:seealso: start_write_group.
860
return self._write_group is not None
863
return self.control_files.is_locked()
865
def is_write_locked(self):
866
"""Return True if this object is write locked."""
867
return self.is_locked() and self.control_files._lock_mode == 'w'
869
def lock_write(self, token=None):
870
"""Lock this repository for writing.
872
This causes caching within the repository obejct to start accumlating
873
data during reads, and allows a 'write_group' to be obtained. Write
874
groups must be used for actual data insertion.
876
:param token: if this is already locked, then lock_write will fail
877
unless the token matches the existing lock.
878
:returns: a token if this instance supports tokens, otherwise None.
879
:raises TokenLockingNotSupported: when a token is given but this
880
instance doesn't support using token locks.
881
:raises MismatchedToken: if the specified token doesn't match the token
882
of the existing lock.
883
:seealso: start_write_group.
885
A token should be passed in if you know that you have locked the object
886
some other way, and need to synchronise this object's state with that
889
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
891
result = self.control_files.lock_write(token=token)
892
for repo in self._fallback_repositories:
893
# Writes don't affect fallback repos
899
self.control_files.lock_read()
900
for repo in self._fallback_repositories:
904
def get_physical_lock_status(self):
905
return self.control_files.get_physical_lock_status()
907
def leave_lock_in_place(self):
908
"""Tell this repository not to release the physical lock when this
911
If lock_write doesn't return a token, then this method is not supported.
913
self.control_files.leave_in_place()
915
def dont_leave_lock_in_place(self):
916
"""Tell this repository to release the physical lock when this
917
object is unlocked, even if it didn't originally acquire it.
919
If lock_write doesn't return a token, then this method is not supported.
921
self.control_files.dont_leave_in_place()
924
def gather_stats(self, revid=None, committers=None):
925
"""Gather statistics from a revision id.
927
:param revid: The revision id to gather statistics from, if None, then
928
no revision specific statistics are gathered.
929
:param committers: Optional parameter controlling whether to grab
930
a count of committers from the revision specific statistics.
931
:return: A dictionary of statistics. Currently this contains:
932
committers: The number of committers if requested.
933
firstrev: A tuple with timestamp, timezone for the penultimate left
934
most ancestor of revid, if revid is not the NULL_REVISION.
935
latestrev: A tuple with timestamp, timezone for revid, if revid is
936
not the NULL_REVISION.
937
revisions: The total revision count in the repository.
938
size: An estimate disk size of the repository in bytes.
941
if revid and committers:
942
result['committers'] = 0
943
if revid and revid != _mod_revision.NULL_REVISION:
945
all_committers = set()
946
revisions = self.get_ancestry(revid)
947
# pop the leading None
949
first_revision = None
951
# ignore the revisions in the middle - just grab first and last
952
revisions = revisions[0], revisions[-1]
953
for revision in self.get_revisions(revisions):
954
if not first_revision:
955
first_revision = revision
957
all_committers.add(revision.committer)
958
last_revision = revision
960
result['committers'] = len(all_committers)
961
result['firstrev'] = (first_revision.timestamp,
962
first_revision.timezone)
963
result['latestrev'] = (last_revision.timestamp,
964
last_revision.timezone)
966
# now gather global repository information
967
# XXX: This is available for many repos regardless of listability.
968
if self.bzrdir.root_transport.listable():
969
# XXX: do we want to __define len__() ?
970
# Maybe the versionedfiles object should provide a different
971
# method to get the number of keys.
972
result['revisions'] = len(self.revisions.keys())
976
def find_branches(self, using=False):
977
"""Find branches underneath this repository.
979
This will include branches inside other branches.
981
:param using: If True, list only branches using this repository.
983
if using and not self.is_shared():
985
return [self.bzrdir.open_branch()]
986
except errors.NotBranchError:
988
class Evaluator(object):
991
self.first_call = True
993
def __call__(self, bzrdir):
994
# On the first call, the parameter is always the bzrdir
995
# containing the current repo.
996
if not self.first_call:
998
repository = bzrdir.open_repository()
999
except errors.NoRepositoryPresent:
1002
return False, (None, repository)
1003
self.first_call = False
1005
value = (bzrdir.open_branch(), None)
1006
except errors.NotBranchError:
1007
value = (None, None)
1011
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1012
self.bzrdir.root_transport, evaluate=Evaluator()):
1013
if branch is not None:
1014
branches.append(branch)
1015
if not using and repository is not None:
1016
branches.extend(repository.find_branches())
1020
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1021
"""Return the revision ids that other has that this does not.
1023
These are returned in topological order.
1025
revision_id: only return revision ids included by revision_id.
1027
return InterRepository.get(other, self).search_missing_revision_ids(
1028
revision_id, find_ghosts)
1030
@deprecated_method(one_two)
1032
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1033
"""Return the revision ids that other has that this does not.
1035
These are returned in topological order.
1037
revision_id: only return revision ids included by revision_id.
1039
keys = self.search_missing_revision_ids(
1040
other, revision_id, find_ghosts).get_keys()
1043
parents = other.get_graph().get_parent_map(keys)
1046
return tsort.topo_sort(parents)
1050
"""Open the repository rooted at base.
1052
For instance, if the repository is at URL/.bzr/repository,
1053
Repository.open(URL) -> a Repository instance.
1055
control = bzrdir.BzrDir.open(base)
1056
return control.open_repository()
1058
def copy_content_into(self, destination, revision_id=None):
1059
"""Make a complete copy of the content in self into destination.
1061
This is a destructive operation! Do not use it on existing
1064
return InterRepository.get(self, destination).copy_content(revision_id)
1066
def commit_write_group(self):
1067
"""Commit the contents accrued within the current write group.
1069
:seealso: start_write_group.
1071
if self._write_group is not self.get_transaction():
1072
# has an unlock or relock occured ?
1073
raise errors.BzrError('mismatched lock context %r and '
1075
(self.get_transaction(), self._write_group))
1076
self._commit_write_group()
1077
self._write_group = None
1079
def _commit_write_group(self):
1080
"""Template method for per-repository write group cleanup.
1082
This is called before the write group is considered to be
1083
finished and should ensure that all data handed to the repository
1084
for writing during the write group is safely committed (to the
1085
extent possible considering file system caching etc).
1088
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1089
"""Fetch the content required to construct revision_id from source.
1091
If revision_id is None all content is copied.
1092
:param find_ghosts: Find and copy revisions in the source that are
1093
ghosts in the target (and not reachable directly by walking out to
1094
the first-present revision in target from revision_id).
1096
# fast path same-url fetch operations
1097
if self.has_same_location(source):
1098
# check that last_revision is in 'from' and then return a
1100
if (revision_id is not None and
1101
not _mod_revision.is_null(revision_id)):
1102
self.get_revision(revision_id)
1104
# if there is no specific appropriate InterRepository, this will get
1105
# the InterRepository base class, which raises an
1106
# IncompatibleRepositories when asked to fetch.
1107
inter = InterRepository.get(source, self)
1108
return inter.fetch(revision_id=revision_id, pb=pb,
1109
find_ghosts=find_ghosts)
1111
def create_bundle(self, target, base, fileobj, format=None):
1112
return serializer.write_bundle(self, target, base, fileobj, format)
1114
def get_commit_builder(self, branch, parents, config, timestamp=None,
1115
timezone=None, committer=None, revprops=None,
1117
"""Obtain a CommitBuilder for this repository.
1119
:param branch: Branch to commit to.
1120
:param parents: Revision ids of the parents of the new revision.
1121
:param config: Configuration to use.
1122
:param timestamp: Optional timestamp recorded for commit.
1123
:param timezone: Optional timezone for timestamp.
1124
:param committer: Optional committer to set for commit.
1125
:param revprops: Optional dictionary of revision properties.
1126
:param revision_id: Optional revision id.
1128
result = self._commit_builder_class(self, parents, config,
1129
timestamp, timezone, committer, revprops, revision_id)
1130
self.start_write_group()
1134
if (self.control_files._lock_count == 1 and
1135
self.control_files._lock_mode == 'w'):
1136
if self._write_group is not None:
1137
self.abort_write_group()
1138
self.control_files.unlock()
1139
raise errors.BzrError(
1140
'Must end write groups before releasing write locks.')
1141
self.control_files.unlock()
1142
for repo in self._fallback_repositories:
1146
def clone(self, a_bzrdir, revision_id=None):
1147
"""Clone this repository into a_bzrdir using the current format.
1149
Currently no check is made that the format of this repository and
1150
the bzrdir format are compatible. FIXME RBC 20060201.
1152
:return: The newly created destination repository.
1154
# TODO: deprecate after 0.16; cloning this with all its settings is
1155
# probably not very useful -- mbp 20070423
1156
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1157
self.copy_content_into(dest_repo, revision_id)
1160
def start_write_group(self):
1161
"""Start a write group in the repository.
1163
Write groups are used by repositories which do not have a 1:1 mapping
1164
between file ids and backend store to manage the insertion of data from
1165
both fetch and commit operations.
1167
A write lock is required around the start_write_group/commit_write_group
1168
for the support of lock-requiring repository formats.
1170
One can only insert data into a repository inside a write group.
1174
if not self.is_write_locked():
1175
raise errors.NotWriteLocked(self)
1176
if self._write_group:
1177
raise errors.BzrError('already in a write group')
1178
self._start_write_group()
1179
# so we can detect unlock/relock - the write group is now entered.
1180
self._write_group = self.get_transaction()
1182
def _start_write_group(self):
1183
"""Template method for per-repository write group startup.
1185
This is called before the write group is considered to be
1190
def sprout(self, to_bzrdir, revision_id=None):
1191
"""Create a descendent repository for new development.
1193
Unlike clone, this does not copy the settings of the repository.
1195
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1196
dest_repo.fetch(self, revision_id=revision_id)
1199
def _create_sprouting_repo(self, a_bzrdir, shared):
1200
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1201
# use target default format.
1202
dest_repo = a_bzrdir.create_repository()
1204
# Most control formats need the repository to be specifically
1205
# created, but on some old all-in-one formats it's not needed
1207
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1208
except errors.UninitializableFormat:
1209
dest_repo = a_bzrdir.open_repository()
1213
def has_revision(self, revision_id):
1214
"""True if this repository has a copy of the revision."""
1215
return revision_id in self.has_revisions((revision_id,))
1218
def has_revisions(self, revision_ids):
1219
"""Probe to find out the presence of multiple revisions.
1221
:param revision_ids: An iterable of revision_ids.
1222
:return: A set of the revision_ids that were present.
1224
parent_map = self.revisions.get_parent_map(
1225
[(rev_id,) for rev_id in revision_ids])
1227
if _mod_revision.NULL_REVISION in revision_ids:
1228
result.add(_mod_revision.NULL_REVISION)
1229
result.update([key[0] for key in parent_map])
1233
def get_revision(self, revision_id):
1234
"""Return the Revision object for a named revision."""
1235
return self.get_revisions([revision_id])[0]
1238
def get_revision_reconcile(self, revision_id):
1239
"""'reconcile' helper routine that allows access to a revision always.
1241
This variant of get_revision does not cross check the weave graph
1242
against the revision one as get_revision does: but it should only
1243
be used by reconcile, or reconcile-alike commands that are correcting
1244
or testing the revision graph.
1246
return self._get_revisions([revision_id])[0]
1249
def get_revisions(self, revision_ids):
1250
"""Get many revisions at once."""
1251
return self._get_revisions(revision_ids)
1254
def _get_revisions(self, revision_ids):
1255
"""Core work logic to get many revisions without sanity checks."""
1256
for rev_id in revision_ids:
1257
if not rev_id or not isinstance(rev_id, basestring):
1258
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1259
keys = [(key,) for key in revision_ids]
1260
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1262
for record in stream:
1263
if record.storage_kind == 'absent':
1264
raise errors.NoSuchRevision(self, record.key[0])
1265
text = record.get_bytes_as('fulltext')
1266
rev = self._serializer.read_revision_from_string(text)
1267
revs[record.key[0]] = rev
1268
return [revs[revid] for revid in revision_ids]
1271
def get_revision_xml(self, revision_id):
1272
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1273
# would have already do it.
1274
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1275
rev = self.get_revision(revision_id)
1276
rev_tmp = cStringIO.StringIO()
1277
# the current serializer..
1278
self._serializer.write_revision(rev, rev_tmp)
1280
return rev_tmp.getvalue()
1282
def get_deltas_for_revisions(self, revisions):
1283
"""Produce a generator of revision deltas.
1285
Note that the input is a sequence of REVISIONS, not revision_ids.
1286
Trees will be held in memory until the generator exits.
1287
Each delta is relative to the revision's lefthand predecessor.
1289
required_trees = set()
1290
for revision in revisions:
1291
required_trees.add(revision.revision_id)
1292
required_trees.update(revision.parent_ids[:1])
1293
trees = dict((t.get_revision_id(), t) for
1294
t in self.revision_trees(required_trees))
1295
for revision in revisions:
1296
if not revision.parent_ids:
1297
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1299
old_tree = trees[revision.parent_ids[0]]
1300
yield trees[revision.revision_id].changes_from(old_tree)
1303
def get_revision_delta(self, revision_id):
1304
"""Return the delta for one revision.
1306
The delta is relative to the left-hand predecessor of the
1309
r = self.get_revision(revision_id)
1310
return list(self.get_deltas_for_revisions([r]))[0]
1313
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1314
signature = gpg_strategy.sign(plaintext)
1315
self.add_signature_text(revision_id, signature)
1318
def add_signature_text(self, revision_id, signature):
1319
self.signatures.add_lines((revision_id,), (),
1320
osutils.split_lines(signature))
1322
def find_text_key_references(self):
1323
"""Find the text key references within the repository.
1325
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1326
revision_ids. Each altered file-ids has the exact revision_ids that
1327
altered it listed explicitly.
1328
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1329
to whether they were referred to by the inventory of the
1330
revision_id that they contain. The inventory texts from all present
1331
revision ids are assessed to generate this report.
1333
revision_keys = self.revisions.keys()
1334
w = self.inventories
1335
pb = ui.ui_factory.nested_progress_bar()
1337
return self._find_text_key_references_from_xml_inventory_lines(
1338
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1342
def _find_text_key_references_from_xml_inventory_lines(self,
1344
"""Core routine for extracting references to texts from inventories.
1346
This performs the translation of xml lines to revision ids.
1348
:param line_iterator: An iterator of lines, origin_version_id
1349
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1350
to whether they were referred to by the inventory of the
1351
revision_id that they contain. Note that if that revision_id was
1352
not part of the line_iterator's output then False will be given -
1353
even though it may actually refer to that key.
1355
if not self._serializer.support_altered_by_hack:
1356
raise AssertionError(
1357
"_find_text_key_references_from_xml_inventory_lines only "
1358
"supported for branches which store inventory as unnested xml"
1359
", not on %r" % self)
1362
# this code needs to read every new line in every inventory for the
1363
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1364
# not present in one of those inventories is unnecessary but not
1365
# harmful because we are filtering by the revision id marker in the
1366
# inventory lines : we only select file ids altered in one of those
1367
# revisions. We don't need to see all lines in the inventory because
1368
# only those added in an inventory in rev X can contain a revision=X
1370
unescape_revid_cache = {}
1371
unescape_fileid_cache = {}
1373
# jam 20061218 In a big fetch, this handles hundreds of thousands
1374
# of lines, so it has had a lot of inlining and optimizing done.
1375
# Sorry that it is a little bit messy.
1376
# Move several functions to be local variables, since this is a long
1378
search = self._file_ids_altered_regex.search
1379
unescape = _unescape_xml
1380
setdefault = result.setdefault
1381
for line, line_key in line_iterator:
1382
match = search(line)
1385
# One call to match.group() returning multiple items is quite a
1386
# bit faster than 2 calls to match.group() each returning 1
1387
file_id, revision_id = match.group('file_id', 'revision_id')
1389
# Inlining the cache lookups helps a lot when you make 170,000
1390
# lines and 350k ids, versus 8.4 unique ids.
1391
# Using a cache helps in 2 ways:
1392
# 1) Avoids unnecessary decoding calls
1393
# 2) Re-uses cached strings, which helps in future set and
1395
# (2) is enough that removing encoding entirely along with
1396
# the cache (so we are using plain strings) results in no
1397
# performance improvement.
1399
revision_id = unescape_revid_cache[revision_id]
1401
unescaped = unescape(revision_id)
1402
unescape_revid_cache[revision_id] = unescaped
1403
revision_id = unescaped
1405
# Note that unconditionally unescaping means that we deserialise
1406
# every fileid, which for general 'pull' is not great, but we don't
1407
# really want to have some many fulltexts that this matters anyway.
1410
file_id = unescape_fileid_cache[file_id]
1412
unescaped = unescape(file_id)
1413
unescape_fileid_cache[file_id] = unescaped
1416
key = (file_id, revision_id)
1417
setdefault(key, False)
1418
if revision_id == line_key[-1]:
1422
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1424
"""Helper routine for fileids_altered_by_revision_ids.
1426
This performs the translation of xml lines to revision ids.
1428
:param line_iterator: An iterator of lines, origin_version_id
1429
:param revision_ids: The revision ids to filter for. This should be a
1430
set or other type which supports efficient __contains__ lookups, as
1431
the revision id from each parsed line will be looked up in the
1432
revision_ids filter.
1433
:return: a dictionary mapping altered file-ids to an iterable of
1434
revision_ids. Each altered file-ids has the exact revision_ids that
1435
altered it listed explicitly.
1438
setdefault = result.setdefault
1440
self._find_text_key_references_from_xml_inventory_lines(
1441
line_iterator).iterkeys():
1442
# once data is all ensured-consistent; then this is
1443
# if revision_id == version_id
1444
if key[-1:] in revision_ids:
1445
setdefault(key[0], set()).add(key[-1])
1448
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1449
"""Find the file ids and versions affected by revisions.
1451
:param revisions: an iterable containing revision ids.
1452
:param _inv_weave: The inventory weave from this repository or None.
1453
If None, the inventory weave will be opened automatically.
1454
:return: a dictionary mapping altered file-ids to an iterable of
1455
revision_ids. Each altered file-ids has the exact revision_ids that
1456
altered it listed explicitly.
1458
selected_keys = set((revid,) for revid in revision_ids)
1459
w = _inv_weave or self.inventories
1460
pb = ui.ui_factory.nested_progress_bar()
1462
return self._find_file_ids_from_xml_inventory_lines(
1463
w.iter_lines_added_or_present_in_keys(
1464
selected_keys, pb=pb),
1469
def iter_files_bytes(self, desired_files):
1470
"""Iterate through file versions.
1472
Files will not necessarily be returned in the order they occur in
1473
desired_files. No specific order is guaranteed.
1475
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1476
value supplied by the caller as part of desired_files. It should
1477
uniquely identify the file version in the caller's context. (Examples:
1478
an index number or a TreeTransform trans_id.)
1480
bytes_iterator is an iterable of bytestrings for the file. The
1481
kind of iterable and length of the bytestrings are unspecified, but for
1482
this implementation, it is a list of bytes produced by
1483
VersionedFile.get_record_stream().
1485
:param desired_files: a list of (file_id, revision_id, identifier)
1488
transaction = self.get_transaction()
1490
for file_id, revision_id, callable_data in desired_files:
1491
text_keys[(file_id, revision_id)] = callable_data
1492
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1493
if record.storage_kind == 'absent':
1494
raise errors.RevisionNotPresent(record.key, self)
1495
yield text_keys[record.key], record.get_bytes_as('fulltext')
1497
def _generate_text_key_index(self, text_key_references=None,
1499
"""Generate a new text key index for the repository.
1501
This is an expensive function that will take considerable time to run.
1503
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1504
list of parents, also text keys. When a given key has no parents,
1505
the parents list will be [NULL_REVISION].
1507
# All revisions, to find inventory parents.
1508
if ancestors is None:
1509
graph = self.get_graph()
1510
ancestors = graph.get_parent_map(self.all_revision_ids())
1511
if text_key_references is None:
1512
text_key_references = self.find_text_key_references()
1513
pb = ui.ui_factory.nested_progress_bar()
1515
return self._do_generate_text_key_index(ancestors,
1516
text_key_references, pb)
1520
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1521
"""Helper for _generate_text_key_index to avoid deep nesting."""
1522
revision_order = tsort.topo_sort(ancestors)
1523
invalid_keys = set()
1525
for revision_id in revision_order:
1526
revision_keys[revision_id] = set()
1527
text_count = len(text_key_references)
1528
# a cache of the text keys to allow reuse; costs a dict of all the
1529
# keys, but saves a 2-tuple for every child of a given key.
1531
for text_key, valid in text_key_references.iteritems():
1533
invalid_keys.add(text_key)
1535
revision_keys[text_key[1]].add(text_key)
1536
text_key_cache[text_key] = text_key
1537
del text_key_references
1539
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1540
NULL_REVISION = _mod_revision.NULL_REVISION
1541
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1542
# too small for large or very branchy trees. However, for 55K path
1543
# trees, it would be easy to use too much memory trivially. Ideally we
1544
# could gauge this by looking at available real memory etc, but this is
1545
# always a tricky proposition.
1546
inventory_cache = lru_cache.LRUCache(10)
1547
batch_size = 10 # should be ~150MB on a 55K path tree
1548
batch_count = len(revision_order) / batch_size + 1
1550
pb.update("Calculating text parents.", processed_texts, text_count)
1551
for offset in xrange(batch_count):
1552
to_query = revision_order[offset * batch_size:(offset + 1) *
1556
for rev_tree in self.revision_trees(to_query):
1557
revision_id = rev_tree.get_revision_id()
1558
parent_ids = ancestors[revision_id]
1559
for text_key in revision_keys[revision_id]:
1560
pb.update("Calculating text parents.", processed_texts)
1561
processed_texts += 1
1562
candidate_parents = []
1563
for parent_id in parent_ids:
1564
parent_text_key = (text_key[0], parent_id)
1566
check_parent = parent_text_key not in \
1567
revision_keys[parent_id]
1569
# the parent parent_id is a ghost:
1570
check_parent = False
1571
# truncate the derived graph against this ghost.
1572
parent_text_key = None
1574
# look at the parent commit details inventories to
1575
# determine possible candidates in the per file graph.
1578
inv = inventory_cache[parent_id]
1580
inv = self.revision_tree(parent_id).inventory
1581
inventory_cache[parent_id] = inv
1582
parent_entry = inv._byid.get(text_key[0], None)
1583
if parent_entry is not None:
1585
text_key[0], parent_entry.revision)
1587
parent_text_key = None
1588
if parent_text_key is not None:
1589
candidate_parents.append(
1590
text_key_cache[parent_text_key])
1591
parent_heads = text_graph.heads(candidate_parents)
1592
new_parents = list(parent_heads)
1593
new_parents.sort(key=lambda x:candidate_parents.index(x))
1594
if new_parents == []:
1595
new_parents = [NULL_REVISION]
1596
text_index[text_key] = new_parents
1598
for text_key in invalid_keys:
1599
text_index[text_key] = [NULL_REVISION]
1602
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1603
"""Get an iterable listing the keys of all the data introduced by a set
1606
The keys will be ordered so that the corresponding items can be safely
1607
fetched and inserted in that order.
1609
:returns: An iterable producing tuples of (knit-kind, file-id,
1610
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1611
'revisions'. file-id is None unless knit-kind is 'file'.
1613
# XXX: it's a bit weird to control the inventory weave caching in this
1614
# generator. Ideally the caching would be done in fetch.py I think. Or
1615
# maybe this generator should explicitly have the contract that it
1616
# should not be iterated until the previously yielded item has been
1618
inv_w = self.inventories
1620
# file ids that changed
1621
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1623
num_file_ids = len(file_ids)
1624
for file_id, altered_versions in file_ids.iteritems():
1625
if _files_pb is not None:
1626
_files_pb.update("fetch texts", count, num_file_ids)
1628
yield ("file", file_id, altered_versions)
1629
# We're done with the files_pb. Note that it finished by the caller,
1630
# just as it was created by the caller.
1634
yield ("inventory", None, revision_ids)
1637
# XXX: Note ATM no callers actually pay attention to this return
1638
# instead they just use the list of revision ids and ignore
1639
# missing sigs. Consider removing this work entirely
1640
revisions_with_signatures = set(self.signatures.get_parent_map(
1641
[(r,) for r in revision_ids]))
1642
revisions_with_signatures = set(
1643
[r for (r,) in revisions_with_signatures])
1644
revisions_with_signatures.intersection_update(revision_ids)
1645
yield ("signatures", None, revisions_with_signatures)
1648
yield ("revisions", None, revision_ids)
1651
def get_inventory(self, revision_id):
1652
"""Get Inventory object by revision id."""
1653
return self.iter_inventories([revision_id]).next()
1655
def iter_inventories(self, revision_ids):
1656
"""Get many inventories by revision_ids.
1658
This will buffer some or all of the texts used in constructing the
1659
inventories in memory, but will only parse a single inventory at a
1662
:return: An iterator of inventories.
1664
if ((None in revision_ids)
1665
or (_mod_revision.NULL_REVISION in revision_ids)):
1666
raise ValueError('cannot get null revision inventory')
1667
return self._iter_inventories(revision_ids)
1669
def _iter_inventories(self, revision_ids):
1670
"""single-document based inventory iteration."""
1671
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1672
yield self.deserialise_inventory(revision_id, text)
1674
def _iter_inventory_xmls(self, revision_ids):
1675
keys = [(revision_id,) for revision_id in revision_ids]
1676
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1678
for record in stream:
1679
if record.storage_kind != 'absent':
1680
texts[record.key] = record.get_bytes_as('fulltext')
1682
raise errors.NoSuchRevision(self, record.key)
1684
yield texts[key], key[-1]
1686
def deserialise_inventory(self, revision_id, xml):
1687
"""Transform the xml into an inventory object.
1689
:param revision_id: The expected revision id of the inventory.
1690
:param xml: A serialised inventory.
1692
result = self._serializer.read_inventory_from_string(xml, revision_id)
1693
if result.revision_id != revision_id:
1694
raise AssertionError('revision id mismatch %s != %s' % (
1695
result.revision_id, revision_id))
1698
def serialise_inventory(self, inv):
1699
return self._serializer.write_inventory_to_string(inv)
1701
def _serialise_inventory_to_lines(self, inv):
1702
return self._serializer.write_inventory_to_lines(inv)
1704
def get_serializer_format(self):
1705
return self._serializer.format_num
1708
def get_inventory_xml(self, revision_id):
1709
"""Get inventory XML as a file object."""
1710
texts = self._iter_inventory_xmls([revision_id])
1712
text, revision_id = texts.next()
1713
except StopIteration:
1714
raise errors.HistoryMissing(self, 'inventory', revision_id)
1718
def get_inventory_sha1(self, revision_id):
1719
"""Return the sha1 hash of the inventory entry
1721
return self.get_revision(revision_id).inventory_sha1
1723
def iter_reverse_revision_history(self, revision_id):
1724
"""Iterate backwards through revision ids in the lefthand history
1726
:param revision_id: The revision id to start with. All its lefthand
1727
ancestors will be traversed.
1729
graph = self.get_graph()
1730
next_id = revision_id
1732
if next_id in (None, _mod_revision.NULL_REVISION):
1735
# Note: The following line may raise KeyError in the event of
1736
# truncated history. We decided not to have a try:except:raise
1737
# RevisionNotPresent here until we see a use for it, because of the
1738
# cost in an inner loop that is by its very nature O(history).
1739
# Robert Collins 20080326
1740
parents = graph.get_parent_map([next_id])[next_id]
1741
if len(parents) == 0:
1744
next_id = parents[0]
1747
def get_revision_inventory(self, revision_id):
1748
"""Return inventory of a past revision."""
1749
# TODO: Unify this with get_inventory()
1750
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1751
# must be the same as its revision, so this is trivial.
1752
if revision_id is None:
1753
# This does not make sense: if there is no revision,
1754
# then it is the current tree inventory surely ?!
1755
# and thus get_root_id() is something that looks at the last
1756
# commit on the branch, and the get_root_id is an inventory check.
1757
raise NotImplementedError
1758
# return Inventory(self.get_root_id())
1760
return self.get_inventory(revision_id)
1762
def is_shared(self):
1763
"""Return True if this repository is flagged as a shared repository."""
1764
raise NotImplementedError(self.is_shared)
1767
def reconcile(self, other=None, thorough=False):
1768
"""Reconcile this repository."""
1769
from bzrlib.reconcile import RepoReconciler
1770
reconciler = RepoReconciler(self, thorough=thorough)
1771
reconciler.reconcile()
1774
def _refresh_data(self):
1775
"""Helper called from lock_* to ensure coherency with disk.
1777
The default implementation does nothing; it is however possible
1778
for repositories to maintain loaded indices across multiple locks
1779
by checking inside their implementation of this method to see
1780
whether their indices are still valid. This depends of course on
1781
the disk format being validatable in this manner.
1785
def revision_tree(self, revision_id):
1786
"""Return Tree for a revision on this branch.
1788
`revision_id` may be NULL_REVISION for the empty tree revision.
1790
revision_id = _mod_revision.ensure_null(revision_id)
1791
# TODO: refactor this to use an existing revision object
1792
# so we don't need to read it in twice.
1793
if revision_id == _mod_revision.NULL_REVISION:
1794
return RevisionTree(self, Inventory(root_id=None),
1795
_mod_revision.NULL_REVISION)
1797
inv = self.get_revision_inventory(revision_id)
1798
return RevisionTree(self, inv, revision_id)
1800
def revision_trees(self, revision_ids):
1801
"""Return Tree for a revision on this branch.
1803
`revision_id` may not be None or 'null:'"""
1804
inventories = self.iter_inventories(revision_ids)
1805
for inv in inventories:
1806
yield RevisionTree(self, inv, inv.revision_id)
1809
def get_ancestry(self, revision_id, topo_sorted=True):
1810
"""Return a list of revision-ids integrated by a revision.
1812
The first element of the list is always None, indicating the origin
1813
revision. This might change when we have history horizons, or
1814
perhaps we should have a new API.
1816
This is topologically sorted.
1818
if _mod_revision.is_null(revision_id):
1820
if not self.has_revision(revision_id):
1821
raise errors.NoSuchRevision(self, revision_id)
1822
graph = self.get_graph()
1824
search = graph._make_breadth_first_searcher([revision_id])
1827
found, ghosts = search.next_with_ghosts()
1828
except StopIteration:
1831
if _mod_revision.NULL_REVISION in keys:
1832
keys.remove(_mod_revision.NULL_REVISION)
1834
parent_map = graph.get_parent_map(keys)
1835
keys = tsort.topo_sort(parent_map)
1836
return [None] + list(keys)
1839
"""Compress the data within the repository.
1841
This operation only makes sense for some repository types. For other
1842
types it should be a no-op that just returns.
1844
This stub method does not require a lock, but subclasses should use
1845
@needs_write_lock as this is a long running call its reasonable to
1846
implicitly lock for the user.
1850
@deprecated_method(one_six)
1851
def print_file(self, file, revision_id):
1852
"""Print `file` to stdout.
1854
FIXME RBC 20060125 as John Meinel points out this is a bad api
1855
- it writes to stdout, it assumes that that is valid etc. Fix
1856
by creating a new more flexible convenience function.
1858
tree = self.revision_tree(revision_id)
1859
# use inventory as it was in that revision
1860
file_id = tree.inventory.path2id(file)
1862
# TODO: jam 20060427 Write a test for this code path
1863
# it had a bug in it, and was raising the wrong
1865
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1866
tree.print_file(file_id)
1868
def get_transaction(self):
1869
return self.control_files.get_transaction()
1871
@deprecated_method(one_one)
1872
def get_parents(self, revision_ids):
1873
"""See StackedParentsProvider.get_parents"""
1874
parent_map = self.get_parent_map(revision_ids)
1875
return [parent_map.get(r, None) for r in revision_ids]
1877
def get_parent_map(self, revision_ids):
1878
"""See graph._StackedParentsProvider.get_parent_map"""
1879
# revisions index works in keys; this just works in revisions
1880
# therefore wrap and unwrap
1883
for revision_id in revision_ids:
1884
if revision_id == _mod_revision.NULL_REVISION:
1885
result[revision_id] = ()
1886
elif revision_id is None:
1887
raise ValueError('get_parent_map(None) is not valid')
1889
query_keys.append((revision_id ,))
1890
for ((revision_id,), parent_keys) in \
1891
self.revisions.get_parent_map(query_keys).iteritems():
1893
result[revision_id] = tuple(parent_revid
1894
for (parent_revid,) in parent_keys)
1896
result[revision_id] = (_mod_revision.NULL_REVISION,)
1899
def _make_parents_provider(self):
1902
def get_graph(self, other_repository=None):
1903
"""Return the graph walker for this repository format"""
1904
parents_provider = self._make_parents_provider()
1905
if (other_repository is not None and
1906
not self.has_same_location(other_repository)):
1907
parents_provider = graph._StackedParentsProvider(
1908
[parents_provider, other_repository._make_parents_provider()])
1909
return graph.Graph(parents_provider)
1911
def _get_versioned_file_checker(self):
1912
"""Return an object suitable for checking versioned files."""
1913
return _VersionedFileChecker(self)
1915
def revision_ids_to_search_result(self, result_set):
1916
"""Convert a set of revision ids to a graph SearchResult."""
1917
result_parents = set()
1918
for parents in self.get_graph().get_parent_map(
1919
result_set).itervalues():
1920
result_parents.update(parents)
1921
included_keys = result_set.intersection(result_parents)
1922
start_keys = result_set.difference(included_keys)
1923
exclude_keys = result_parents.difference(result_set)
1924
result = graph.SearchResult(start_keys, exclude_keys,
1925
len(result_set), result_set)
1929
def set_make_working_trees(self, new_value):
1930
"""Set the policy flag for making working trees when creating branches.
1932
This only applies to branches that use this repository.
1934
The default is 'True'.
1935
:param new_value: True to restore the default, False to disable making
1938
raise NotImplementedError(self.set_make_working_trees)
1940
def make_working_trees(self):
1941
"""Returns the policy for making working trees on new branches."""
1942
raise NotImplementedError(self.make_working_trees)
1945
def sign_revision(self, revision_id, gpg_strategy):
1946
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1947
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1950
def has_signature_for_revision_id(self, revision_id):
1951
"""Query for a revision signature for revision_id in the repository."""
1952
if not self.has_revision(revision_id):
1953
raise errors.NoSuchRevision(self, revision_id)
1954
sig_present = (1 == len(
1955
self.signatures.get_parent_map([(revision_id,)])))
1959
def get_signature_text(self, revision_id):
1960
"""Return the text for a signature."""
1961
stream = self.signatures.get_record_stream([(revision_id,)],
1963
record = stream.next()
1964
if record.storage_kind == 'absent':
1965
raise errors.NoSuchRevision(self, revision_id)
1966
return record.get_bytes_as('fulltext')
1969
def check(self, revision_ids=None):
1970
"""Check consistency of all history of given revision_ids.
1972
Different repository implementations should override _check().
1974
:param revision_ids: A non-empty list of revision_ids whose ancestry
1975
will be checked. Typically the last revision_id of a branch.
1977
return self._check(revision_ids)
1979
def _check(self, revision_ids):
1980
result = check.Check(self)
1984
def _warn_if_deprecated(self):
1985
global _deprecation_warning_done
1986
if _deprecation_warning_done:
1988
_deprecation_warning_done = True
1989
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1990
% (self._format, self.bzrdir.transport.base))
1992
def supports_rich_root(self):
1993
return self._format.rich_root_data
1995
def _check_ascii_revisionid(self, revision_id, method):
1996
"""Private helper for ascii-only repositories."""
1997
# weave repositories refuse to store revisionids that are non-ascii.
1998
if revision_id is not None:
1999
# weaves require ascii revision ids.
2000
if isinstance(revision_id, unicode):
2002
revision_id.encode('ascii')
2003
except UnicodeEncodeError:
2004
raise errors.NonAsciiRevisionId(method, self)
2007
revision_id.decode('ascii')
2008
except UnicodeDecodeError:
2009
raise errors.NonAsciiRevisionId(method, self)
2011
def revision_graph_can_have_wrong_parents(self):
2012
"""Is it possible for this repository to have a revision graph with
2015
If True, then this repository must also implement
2016
_find_inconsistent_revision_parents so that check and reconcile can
2017
check for inconsistencies before proceeding with other checks that may
2018
depend on the revision index being consistent.
2020
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2023
# remove these delegates a while after bzr 0.15
2024
def __make_delegated(name, from_module):
2025
def _deprecated_repository_forwarder():
2026
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2027
% (name, from_module),
2030
m = __import__(from_module, globals(), locals(), [name])
2032
return getattr(m, name)
2033
except AttributeError:
2034
raise AttributeError('module %s has no name %s'
2036
globals()[name] = _deprecated_repository_forwarder
2039
'AllInOneRepository',
2040
'WeaveMetaDirRepository',
2041
'PreSplitOutRepositoryFormat',
2042
'RepositoryFormat4',
2043
'RepositoryFormat5',
2044
'RepositoryFormat6',
2045
'RepositoryFormat7',
2047
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2051
'RepositoryFormatKnit',
2052
'RepositoryFormatKnit1',
2054
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2057
def install_revision(repository, rev, revision_tree):
2058
"""Install all revision data into a repository."""
2059
install_revisions(repository, [(rev, revision_tree, None)])
2062
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2063
"""Install all revision data into a repository.
2065
Accepts an iterable of revision, tree, signature tuples. The signature
2068
repository.start_write_group()
2070
for n, (revision, revision_tree, signature) in enumerate(iterable):
2071
_install_revision(repository, revision, revision_tree, signature)
2073
pb.update('Transferring revisions', n + 1, num_revisions)
2075
repository.abort_write_group()
2078
repository.commit_write_group()
2081
def _install_revision(repository, rev, revision_tree, signature):
2082
"""Install all revision data into a repository."""
2083
present_parents = []
2085
for p_id in rev.parent_ids:
2086
if repository.has_revision(p_id):
2087
present_parents.append(p_id)
2088
parent_trees[p_id] = repository.revision_tree(p_id)
2090
parent_trees[p_id] = repository.revision_tree(
2091
_mod_revision.NULL_REVISION)
2093
inv = revision_tree.inventory
2094
entries = inv.iter_entries()
2095
# backwards compatibility hack: skip the root id.
2096
if not repository.supports_rich_root():
2097
path, root = entries.next()
2098
if root.revision != rev.revision_id:
2099
raise errors.IncompatibleRevision(repr(repository))
2101
for path, ie in entries:
2102
text_keys[(ie.file_id, ie.revision)] = ie
2103
text_parent_map = repository.texts.get_parent_map(text_keys)
2104
missing_texts = set(text_keys) - set(text_parent_map)
2105
# Add the texts that are not already present
2106
for text_key in missing_texts:
2107
ie = text_keys[text_key]
2109
# FIXME: TODO: The following loop overlaps/duplicates that done by
2110
# commit to determine parents. There is a latent/real bug here where
2111
# the parents inserted are not those commit would do - in particular
2112
# they are not filtered by heads(). RBC, AB
2113
for revision, tree in parent_trees.iteritems():
2114
if ie.file_id not in tree:
2116
parent_id = tree.inventory[ie.file_id].revision
2117
if parent_id in text_parents:
2119
text_parents.append((ie.file_id, parent_id))
2120
lines = revision_tree.get_file(ie.file_id).readlines()
2121
repository.texts.add_lines(text_key, text_parents, lines)
2123
# install the inventory
2124
repository.add_inventory(rev.revision_id, inv, present_parents)
2125
except errors.RevisionAlreadyPresent:
2127
if signature is not None:
2128
repository.add_signature_text(rev.revision_id, signature)
2129
repository.add_revision(rev.revision_id, rev, inv)
2132
class MetaDirRepository(Repository):
2133
"""Repositories in the new meta-dir layout.
2135
:ivar _transport: Transport for access to repository control files,
2136
typically pointing to .bzr/repository.
2139
def __init__(self, _format, a_bzrdir, control_files):
2140
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2141
self._transport = control_files._transport
2143
def is_shared(self):
2144
"""Return True if this repository is flagged as a shared repository."""
2145
return self._transport.has('shared-storage')
2148
def set_make_working_trees(self, new_value):
2149
"""Set the policy flag for making working trees when creating branches.
2151
This only applies to branches that use this repository.
2153
The default is 'True'.
2154
:param new_value: True to restore the default, False to disable making
2159
self._transport.delete('no-working-trees')
2160
except errors.NoSuchFile:
2163
self._transport.put_bytes('no-working-trees', '',
2164
mode=self.bzrdir._get_file_mode())
2166
def make_working_trees(self):
2167
"""Returns the policy for making working trees on new branches."""
2168
return not self._transport.has('no-working-trees')
2171
class MetaDirVersionedFileRepository(MetaDirRepository):
2172
"""Repositories in a meta-dir, that work via versioned file objects."""
2174
def __init__(self, _format, a_bzrdir, control_files):
2175
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2179
class RepositoryFormatRegistry(registry.Registry):
2180
"""Registry of RepositoryFormats."""
2182
def get(self, format_string):
2183
r = registry.Registry.get(self, format_string)
2189
format_registry = RepositoryFormatRegistry()
2190
"""Registry of formats, indexed by their identifying format string.
2192
This can contain either format instances themselves, or classes/factories that
2193
can be called to obtain one.
2197
#####################################################################
2198
# Repository Formats
2200
class RepositoryFormat(object):
2201
"""A repository format.
2203
Formats provide three things:
2204
* An initialization routine to construct repository data on disk.
2205
* a format string which is used when the BzrDir supports versioned
2207
* an open routine which returns a Repository instance.
2209
There is one and only one Format subclass for each on-disk format. But
2210
there can be one Repository subclass that is used for several different
2211
formats. The _format attribute on a Repository instance can be used to
2212
determine the disk format.
2214
Formats are placed in an dict by their format string for reference
2215
during opening. These should be subclasses of RepositoryFormat
2218
Once a format is deprecated, just deprecate the initialize and open
2219
methods on the format class. Do not deprecate the object, as the
2220
object will be created every system load.
2222
Common instance attributes:
2223
_matchingbzrdir - the bzrdir format that the repository format was
2224
originally written to work with. This can be used if manually
2225
constructing a bzrdir and repository, or more commonly for test suite
2229
# Set to True or False in derived classes. True indicates that the format
2230
# supports ghosts gracefully.
2231
supports_ghosts = None
2232
# Can this repository be given external locations to lookup additional
2233
# data. Set to True or False in derived classes.
2234
supports_external_lookups = None
2237
return "<%s>" % self.__class__.__name__
2239
def __eq__(self, other):
2240
# format objects are generally stateless
2241
return isinstance(other, self.__class__)
2243
def __ne__(self, other):
2244
return not self == other
2247
def find_format(klass, a_bzrdir):
2248
"""Return the format for the repository object in a_bzrdir.
2250
This is used by bzr native formats that have a "format" file in
2251
the repository. Other methods may be used by different types of
2255
transport = a_bzrdir.get_repository_transport(None)
2256
format_string = transport.get("format").read()
2257
return format_registry.get(format_string)
2258
except errors.NoSuchFile:
2259
raise errors.NoRepositoryPresent(a_bzrdir)
2261
raise errors.UnknownFormatError(format=format_string,
2265
def register_format(klass, format):
2266
format_registry.register(format.get_format_string(), format)
2269
def unregister_format(klass, format):
2270
format_registry.remove(format.get_format_string())
2273
def get_default_format(klass):
2274
"""Return the current default format."""
2275
from bzrlib import bzrdir
2276
return bzrdir.format_registry.make_bzrdir('default').repository_format
2278
def get_format_string(self):
2279
"""Return the ASCII format string that identifies this format.
2281
Note that in pre format ?? repositories the format string is
2282
not permitted nor written to disk.
2284
raise NotImplementedError(self.get_format_string)
2286
def get_format_description(self):
2287
"""Return the short description for this format."""
2288
raise NotImplementedError(self.get_format_description)
2290
# TODO: this shouldn't be in the base class, it's specific to things that
2291
# use weaves or knits -- mbp 20070207
2292
def _get_versioned_file_store(self,
2297
versionedfile_class=None,
2298
versionedfile_kwargs={},
2300
if versionedfile_class is None:
2301
versionedfile_class = self._versionedfile_class
2302
weave_transport = control_files._transport.clone(name)
2303
dir_mode = control_files._dir_mode
2304
file_mode = control_files._file_mode
2305
return VersionedFileStore(weave_transport, prefixed=prefixed,
2307
file_mode=file_mode,
2308
versionedfile_class=versionedfile_class,
2309
versionedfile_kwargs=versionedfile_kwargs,
2312
def initialize(self, a_bzrdir, shared=False):
2313
"""Initialize a repository of this format in a_bzrdir.
2315
:param a_bzrdir: The bzrdir to put the new repository in it.
2316
:param shared: The repository should be initialized as a sharable one.
2317
:returns: The new repository object.
2319
This may raise UninitializableFormat if shared repository are not
2320
compatible the a_bzrdir.
2322
raise NotImplementedError(self.initialize)
2324
def is_supported(self):
2325
"""Is this format supported?
2327
Supported formats must be initializable and openable.
2328
Unsupported formats may not support initialization or committing or
2329
some other features depending on the reason for not being supported.
2333
def check_conversion_target(self, target_format):
2334
raise NotImplementedError(self.check_conversion_target)
2336
def open(self, a_bzrdir, _found=False):
2337
"""Return an instance of this format for the bzrdir a_bzrdir.
2339
_found is a private parameter, do not use it.
2341
raise NotImplementedError(self.open)
2344
class MetaDirRepositoryFormat(RepositoryFormat):
2345
"""Common base class for the new repositories using the metadir layout."""
2347
rich_root_data = False
2348
supports_tree_reference = False
2349
supports_external_lookups = False
2352
def _matchingbzrdir(self):
2353
matching = bzrdir.BzrDirMetaFormat1()
2354
matching.repository_format = self
2358
super(MetaDirRepositoryFormat, self).__init__()
2360
def _create_control_files(self, a_bzrdir):
2361
"""Create the required files and the initial control_files object."""
2362
# FIXME: RBC 20060125 don't peek under the covers
2363
# NB: no need to escape relative paths that are url safe.
2364
repository_transport = a_bzrdir.get_repository_transport(self)
2365
control_files = lockable_files.LockableFiles(repository_transport,
2366
'lock', lockdir.LockDir)
2367
control_files.create_lock()
2368
return control_files
2370
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2371
"""Upload the initial blank content."""
2372
control_files = self._create_control_files(a_bzrdir)
2373
control_files.lock_write()
2374
transport = control_files._transport
2376
utf8_files += [('shared-storage', '')]
2378
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2379
for (filename, content_stream) in files:
2380
transport.put_file(filename, content_stream,
2381
mode=a_bzrdir._get_file_mode())
2382
for (filename, content_bytes) in utf8_files:
2383
transport.put_bytes_non_atomic(filename, content_bytes,
2384
mode=a_bzrdir._get_file_mode())
2386
control_files.unlock()
2389
# formats which have no format string are not discoverable
2390
# and not independently creatable, so are not registered. They're
2391
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2392
# needed, it's constructed directly by the BzrDir. Non-native formats where
2393
# the repository is not separately opened are similar.
2395
format_registry.register_lazy(
2396
'Bazaar-NG Repository format 7',
2397
'bzrlib.repofmt.weaverepo',
2401
format_registry.register_lazy(
2402
'Bazaar-NG Knit Repository Format 1',
2403
'bzrlib.repofmt.knitrepo',
2404
'RepositoryFormatKnit1',
2407
format_registry.register_lazy(
2408
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2409
'bzrlib.repofmt.knitrepo',
2410
'RepositoryFormatKnit3',
2413
format_registry.register_lazy(
2414
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2415
'bzrlib.repofmt.knitrepo',
2416
'RepositoryFormatKnit4',
2419
# Pack-based formats. There is one format for pre-subtrees, and one for
2420
# post-subtrees to allow ease of testing.
2421
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2422
format_registry.register_lazy(
2423
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2424
'bzrlib.repofmt.pack_repo',
2425
'RepositoryFormatKnitPack1',
2427
format_registry.register_lazy(
2428
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2429
'bzrlib.repofmt.pack_repo',
2430
'RepositoryFormatKnitPack3',
2432
format_registry.register_lazy(
2433
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2434
'bzrlib.repofmt.pack_repo',
2435
'RepositoryFormatKnitPack4',
2437
format_registry.register_lazy(
2438
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2439
'bzrlib.repofmt.pack_repo',
2440
'RepositoryFormatKnitPack5',
2442
format_registry.register_lazy(
2443
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2444
'bzrlib.repofmt.pack_repo',
2445
'RepositoryFormatKnitPack5RichRoot',
2447
format_registry.register_lazy(
2448
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2449
'bzrlib.repofmt.pack_repo',
2450
'RepositoryFormatKnitPack5RichRootBroken',
2452
format_registry.register_lazy(
2453
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2454
'bzrlib.repofmt.pack_repo',
2455
'RepositoryFormatKnitPack6',
2457
format_registry.register_lazy(
2458
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2459
'bzrlib.repofmt.pack_repo',
2460
'RepositoryFormatKnitPack6RichRoot',
2463
# Development formats.
2464
# 1.7->1.8 go below here
2465
format_registry.register_lazy(
2466
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2467
'bzrlib.repofmt.pack_repo',
2468
'RepositoryFormatPackDevelopment2',
2470
format_registry.register_lazy(
2471
("Bazaar development format 2 with subtree support "
2472
"(needs bzr.dev from before 1.8)\n"),
2473
'bzrlib.repofmt.pack_repo',
2474
'RepositoryFormatPackDevelopment2Subtree',
2478
class InterRepository(InterObject):
2479
"""This class represents operations taking place between two repositories.
2481
Its instances have methods like copy_content and fetch, and contain
2482
references to the source and target repositories these operations can be
2485
Often we will provide convenience methods on 'repository' which carry out
2486
operations with another repository - they will always forward to
2487
InterRepository.get(other).method_name(parameters).
2490
_walk_to_common_revisions_batch_size = 1
2492
"""The available optimised InterRepository types."""
2494
def __init__(self, source, target):
2495
InterObject.__init__(self, source, target)
2496
# These two attributes may be overridden by e.g. InterOtherToRemote to
2497
# provide a faster implementation.
2498
self.target_get_graph = self.target.get_graph
2499
self.target_get_parent_map = self.target.get_parent_map
2501
def copy_content(self, revision_id=None):
2502
raise NotImplementedError(self.copy_content)
2504
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2505
"""Fetch the content required to construct revision_id.
2507
The content is copied from self.source to self.target.
2509
:param revision_id: if None all content is copied, if NULL_REVISION no
2511
:param pb: optional progress bar to use for progress reports. If not
2512
provided a default one will be created.
2514
:returns: (copied_revision_count, failures).
2516
# Normally we should find a specific InterRepository subclass to do
2517
# the fetch; if nothing else then at least InterSameDataRepository.
2518
# If none of them is suitable it looks like fetching is not possible;
2519
# we try to give a good message why. _assert_same_model will probably
2520
# give a helpful message; otherwise a generic one.
2521
self._assert_same_model(self.source, self.target)
2522
raise errors.IncompatibleRepositories(self.source, self.target,
2523
"no suitableInterRepository found")
2525
def _walk_to_common_revisions(self, revision_ids):
2526
"""Walk out from revision_ids in source to revisions target has.
2528
:param revision_ids: The start point for the search.
2529
:return: A set of revision ids.
2531
target_graph = self.target_get_graph()
2532
revision_ids = frozenset(revision_ids)
2533
# Fast path for the case where all the revisions are already in the
2535
# (Although this does incur an extra round trip for the
2536
# fairly common case where the target doesn't already have the revision
2538
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2539
return graph.SearchResult(revision_ids, set(), 0, set())
2540
missing_revs = set()
2541
source_graph = self.source.get_graph()
2542
# ensure we don't pay silly lookup costs.
2543
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2544
null_set = frozenset([_mod_revision.NULL_REVISION])
2545
searcher_exhausted = False
2549
# Iterate the searcher until we have enough next_revs
2550
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2552
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2553
next_revs.update(next_revs_part)
2554
ghosts.update(ghosts_part)
2555
except StopIteration:
2556
searcher_exhausted = True
2558
# If there are ghosts in the source graph, and the caller asked for
2559
# them, make sure that they are present in the target.
2560
# We don't care about other ghosts as we can't fetch them and
2561
# haven't been asked to.
2562
ghosts_to_check = set(revision_ids.intersection(ghosts))
2563
revs_to_get = set(next_revs).union(ghosts_to_check)
2565
have_revs = set(target_graph.get_parent_map(revs_to_get))
2566
# we always have NULL_REVISION present.
2567
have_revs = have_revs.union(null_set)
2568
# Check if the target is missing any ghosts we need.
2569
ghosts_to_check.difference_update(have_revs)
2571
# One of the caller's revision_ids is a ghost in both the
2572
# source and the target.
2573
raise errors.NoSuchRevision(
2574
self.source, ghosts_to_check.pop())
2575
missing_revs.update(next_revs - have_revs)
2576
# Because we may have walked past the original stop point, make
2577
# sure everything is stopped
2578
stop_revs = searcher.find_seen_ancestors(have_revs)
2579
searcher.stop_searching_any(stop_revs)
2580
if searcher_exhausted:
2582
return searcher.get_result()
2584
@deprecated_method(one_two)
2586
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2587
"""Return the revision ids that source has that target does not.
2589
These are returned in topological order.
2591
:param revision_id: only return revision ids included by this
2593
:param find_ghosts: If True find missing revisions in deep history
2594
rather than just finding the surface difference.
2596
return list(self.search_missing_revision_ids(
2597
revision_id, find_ghosts).get_keys())
2600
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2601
"""Return the revision ids that source has that target does not.
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.
2607
:return: A bzrlib.graph.SearchResult.
2609
# stop searching at found target revisions.
2610
if not find_ghosts and revision_id is not None:
2611
return self._walk_to_common_revisions([revision_id])
2612
# generic, possibly worst case, slow code path.
2613
target_ids = set(self.target.all_revision_ids())
2614
if revision_id is not None:
2615
source_ids = self.source.get_ancestry(revision_id)
2616
if source_ids[0] is not None:
2617
raise AssertionError()
2620
source_ids = self.source.all_revision_ids()
2621
result_set = set(source_ids).difference(target_ids)
2622
return self.source.revision_ids_to_search_result(result_set)
2625
def _same_model(source, target):
2626
"""True if source and target have the same data representation.
2628
Note: this is always called on the base class; overriding it in a
2629
subclass will have no effect.
2632
InterRepository._assert_same_model(source, target)
2634
except errors.IncompatibleRepositories, e:
2638
def _assert_same_model(source, target):
2639
"""Raise an exception if two repositories do not use the same model.
2641
if source.supports_rich_root() != target.supports_rich_root():
2642
raise errors.IncompatibleRepositories(source, target,
2643
"different rich-root support")
2644
if source._serializer != target._serializer:
2645
raise errors.IncompatibleRepositories(source, target,
2646
"different serializers")
2649
class InterSameDataRepository(InterRepository):
2650
"""Code for converting between repositories that represent the same data.
2652
Data format and model must match for this to work.
2656
def _get_repo_format_to_test(self):
2657
"""Repository format for testing with.
2659
InterSameData can pull from subtree to subtree and from non-subtree to
2660
non-subtree, so we test this with the richest repository format.
2662
from bzrlib.repofmt import knitrepo
2663
return knitrepo.RepositoryFormatKnit3()
2666
def is_compatible(source, target):
2667
return InterRepository._same_model(source, target)
2670
def copy_content(self, revision_id=None):
2671
"""Make a complete copy of the content in self into destination.
2673
This copies both the repository's revision data, and configuration information
2674
such as the make_working_trees setting.
2676
This is a destructive operation! Do not use it on existing
2679
:param revision_id: Only copy the content needed to construct
2680
revision_id and its parents.
2683
self.target.set_make_working_trees(self.source.make_working_trees())
2684
except NotImplementedError:
2686
# but don't bother fetching if we have the needed data now.
2687
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2688
self.target.has_revision(revision_id)):
2690
self.target.fetch(self.source, revision_id=revision_id)
2693
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2694
"""See InterRepository.fetch()."""
2695
from bzrlib.fetch import RepoFetcher
2696
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2697
self.source, self.source._format, self.target,
2698
self.target._format)
2699
f = RepoFetcher(to_repository=self.target,
2700
from_repository=self.source,
2701
last_revision=revision_id,
2702
pb=pb, find_ghosts=find_ghosts)
2703
return f.count_copied, f.failed_revisions
2706
class InterWeaveRepo(InterSameDataRepository):
2707
"""Optimised code paths between Weave based repositories.
2709
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2710
implemented lazy inter-object optimisation.
2714
def _get_repo_format_to_test(self):
2715
from bzrlib.repofmt import weaverepo
2716
return weaverepo.RepositoryFormat7()
2719
def is_compatible(source, target):
2720
"""Be compatible with known Weave formats.
2722
We don't test for the stores being of specific types because that
2723
could lead to confusing results, and there is no need to be
2726
from bzrlib.repofmt.weaverepo import (
2732
return (isinstance(source._format, (RepositoryFormat5,
2734
RepositoryFormat7)) and
2735
isinstance(target._format, (RepositoryFormat5,
2737
RepositoryFormat7)))
2738
except AttributeError:
2742
def copy_content(self, revision_id=None):
2743
"""See InterRepository.copy_content()."""
2744
# weave specific optimised path:
2746
self.target.set_make_working_trees(self.source.make_working_trees())
2747
except (errors.RepositoryUpgradeRequired, NotImplemented):
2749
# FIXME do not peek!
2750
if self.source._transport.listable():
2751
pb = ui.ui_factory.nested_progress_bar()
2753
self.target.texts.insert_record_stream(
2754
self.source.texts.get_record_stream(
2755
self.source.texts.keys(), 'topological', False))
2756
pb.update('copying inventory', 0, 1)
2757
self.target.inventories.insert_record_stream(
2758
self.source.inventories.get_record_stream(
2759
self.source.inventories.keys(), 'topological', False))
2760
self.target.signatures.insert_record_stream(
2761
self.source.signatures.get_record_stream(
2762
self.source.signatures.keys(),
2764
self.target.revisions.insert_record_stream(
2765
self.source.revisions.get_record_stream(
2766
self.source.revisions.keys(),
2767
'topological', True))
2771
self.target.fetch(self.source, revision_id=revision_id)
2774
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2775
"""See InterRepository.fetch()."""
2776
from bzrlib.fetch import RepoFetcher
2777
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2778
self.source, self.source._format, self.target, self.target._format)
2779
f = RepoFetcher(to_repository=self.target,
2780
from_repository=self.source,
2781
last_revision=revision_id,
2782
pb=pb, find_ghosts=find_ghosts)
2783
return f.count_copied, f.failed_revisions
2786
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2787
"""See InterRepository.missing_revision_ids()."""
2788
# we want all revisions to satisfy revision_id in source.
2789
# but we don't want to stat every file here and there.
2790
# we want then, all revisions other needs to satisfy revision_id
2791
# checked, but not those that we have locally.
2792
# so the first thing is to get a subset of the revisions to
2793
# satisfy revision_id in source, and then eliminate those that
2794
# we do already have.
2795
# this is slow on high latency connection to self, but as as this
2796
# disk format scales terribly for push anyway due to rewriting
2797
# inventory.weave, this is considered acceptable.
2799
if revision_id is not None:
2800
source_ids = self.source.get_ancestry(revision_id)
2801
if source_ids[0] is not None:
2802
raise AssertionError()
2805
source_ids = self.source._all_possible_ids()
2806
source_ids_set = set(source_ids)
2807
# source_ids is the worst possible case we may need to pull.
2808
# now we want to filter source_ids against what we actually
2809
# have in target, but don't try to check for existence where we know
2810
# we do not have a revision as that would be pointless.
2811
target_ids = set(self.target._all_possible_ids())
2812
possibly_present_revisions = target_ids.intersection(source_ids_set)
2813
actually_present_revisions = set(
2814
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2815
required_revisions = source_ids_set.difference(actually_present_revisions)
2816
if revision_id is not None:
2817
# we used get_ancestry to determine source_ids then we are assured all
2818
# revisions referenced are present as they are installed in topological order.
2819
# and the tip revision was validated by get_ancestry.
2820
result_set = required_revisions
2822
# if we just grabbed the possibly available ids, then
2823
# we only have an estimate of whats available and need to validate
2824
# that against the revision records.
2826
self.source._eliminate_revisions_not_present(required_revisions))
2827
return self.source.revision_ids_to_search_result(result_set)
2830
class InterKnitRepo(InterSameDataRepository):
2831
"""Optimised code paths between Knit based repositories."""
2834
def _get_repo_format_to_test(self):
2835
from bzrlib.repofmt import knitrepo
2836
return knitrepo.RepositoryFormatKnit1()
2839
def is_compatible(source, target):
2840
"""Be compatible with known Knit formats.
2842
We don't test for the stores being of specific types because that
2843
could lead to confusing results, and there is no need to be
2846
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2848
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2849
isinstance(target._format, RepositoryFormatKnit))
2850
except AttributeError:
2852
return are_knits and InterRepository._same_model(source, target)
2855
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2856
"""See InterRepository.fetch()."""
2857
from bzrlib.fetch import RepoFetcher
2858
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2859
self.source, self.source._format, self.target, self.target._format)
2860
f = RepoFetcher(to_repository=self.target,
2861
from_repository=self.source,
2862
last_revision=revision_id,
2863
pb=pb, find_ghosts=find_ghosts)
2864
return f.count_copied, f.failed_revisions
2867
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2868
"""See InterRepository.missing_revision_ids()."""
2869
if revision_id is not None:
2870
source_ids = self.source.get_ancestry(revision_id)
2871
if source_ids[0] is not None:
2872
raise AssertionError()
2875
source_ids = self.source.all_revision_ids()
2876
source_ids_set = set(source_ids)
2877
# source_ids is the worst possible case we may need to pull.
2878
# now we want to filter source_ids against what we actually
2879
# have in target, but don't try to check for existence where we know
2880
# we do not have a revision as that would be pointless.
2881
target_ids = set(self.target.all_revision_ids())
2882
possibly_present_revisions = target_ids.intersection(source_ids_set)
2883
actually_present_revisions = set(
2884
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2885
required_revisions = source_ids_set.difference(actually_present_revisions)
2886
if revision_id is not None:
2887
# we used get_ancestry to determine source_ids then we are assured all
2888
# revisions referenced are present as they are installed in topological order.
2889
# and the tip revision was validated by get_ancestry.
2890
result_set = required_revisions
2892
# if we just grabbed the possibly available ids, then
2893
# we only have an estimate of whats available and need to validate
2894
# that against the revision records.
2896
self.source._eliminate_revisions_not_present(required_revisions))
2897
return self.source.revision_ids_to_search_result(result_set)
2900
class InterPackRepo(InterSameDataRepository):
2901
"""Optimised code paths between Pack based repositories."""
2904
def _get_repo_format_to_test(self):
2905
from bzrlib.repofmt import pack_repo
2906
return pack_repo.RepositoryFormatKnitPack1()
2909
def is_compatible(source, target):
2910
"""Be compatible with known Pack formats.
2912
We don't test for the stores being of specific types because that
2913
could lead to confusing results, and there is no need to be
2916
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2918
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2919
isinstance(target._format, RepositoryFormatPack))
2920
except AttributeError:
2922
return are_packs and InterRepository._same_model(source, target)
2925
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2926
"""See InterRepository.fetch()."""
2927
if (len(self.source._fallback_repositories) > 0 or
2928
len(self.target._fallback_repositories) > 0):
2929
# The pack layer is not aware of fallback repositories, so when
2930
# fetching from a stacked repository or into a stacked repository
2931
# we use the generic fetch logic which uses the VersionedFiles
2932
# attributes on repository.
2933
from bzrlib.fetch import RepoFetcher
2934
# Make sure the generic fetcher sets the write cache size on the
2935
# new pack (just like Packer.pack does) to avoid doing many tiny
2936
# writes (which can be slow over a network connection).
2937
# XXX: ideally the transport layer would do this automatically.
2938
pack_coll = self._get_target_pack_collection()
2940
lambda: pack_coll._new_pack.set_write_cache_size(1024*1024))
2941
fetcher = RepoFetcher(self.target, self.source, revision_id,
2942
pb, find_ghosts, set_cache_size)
2943
return fetcher.count_copied, fetcher.failed_revisions
2944
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2945
self.source, self.source._format, self.target, self.target._format)
2946
self.count_copied = 0
2947
if revision_id is None:
2949
# everything to do - use pack logic
2950
# to fetch from all packs to one without
2951
# inventory parsing etc, IFF nothing to be copied is in the target.
2953
source_revision_ids = frozenset(self.source.all_revision_ids())
2954
revision_ids = source_revision_ids - \
2955
frozenset(self.target_get_parent_map(source_revision_ids))
2956
revision_keys = [(revid,) for revid in revision_ids]
2957
target_pack_collection = self._get_target_pack_collection()
2958
index = target_pack_collection.revision_index.combined_index
2959
present_revision_ids = set(item[1][0] for item in
2960
index.iter_entries(revision_keys))
2961
revision_ids = set(revision_ids) - present_revision_ids
2962
# implementing the TODO will involve:
2963
# - detecting when all of a pack is selected
2964
# - avoiding as much as possible pre-selection, so the
2965
# more-core routines such as create_pack_from_packs can filter in
2966
# a just-in-time fashion. (though having a HEADS list on a
2967
# repository might make this a lot easier, because we could
2968
# sensibly detect 'new revisions' without doing a full index scan.
2969
elif _mod_revision.is_null(revision_id):
2974
revision_ids = self.search_missing_revision_ids(revision_id,
2975
find_ghosts=find_ghosts).get_keys()
2976
except errors.NoSuchRevision:
2977
raise errors.InstallFailed([revision_id])
2978
if len(revision_ids) == 0:
2980
return self._pack(self.source, self.target, revision_ids)
2982
def _pack(self, source, target, revision_ids):
2983
from bzrlib.repofmt.pack_repo import Packer
2984
target_pack_collection = self._get_target_pack_collection()
2985
packs = source._pack_collection.all_packs()
2986
pack = Packer(target_pack_collection, packs, '.fetch',
2987
revision_ids).pack()
2988
if pack is not None:
2989
target_pack_collection._save_pack_names()
2990
copied_revs = pack.get_revision_count()
2991
# Trigger an autopack. This may duplicate effort as we've just done
2992
# a pack creation, but for now it is simpler to think about as
2993
# 'upload data, then repack if needed'.
2995
return (copied_revs, [])
2999
def _autopack(self):
3000
self.target._pack_collection.autopack()
3002
def _get_target_pack_collection(self):
3003
return self.target._pack_collection
3006
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3007
"""See InterRepository.missing_revision_ids().
3009
:param find_ghosts: Find ghosts throughout the ancestry of
3012
if not find_ghosts and revision_id is not None:
3013
return self._walk_to_common_revisions([revision_id])
3014
elif revision_id is not None:
3015
# Find ghosts: search for revisions pointing from one repository to
3016
# the other, and vice versa, anywhere in the history of revision_id.
3017
graph = self.target_get_graph(other_repository=self.source)
3018
searcher = graph._make_breadth_first_searcher([revision_id])
3022
next_revs, ghosts = searcher.next_with_ghosts()
3023
except StopIteration:
3025
if revision_id in ghosts:
3026
raise errors.NoSuchRevision(self.source, revision_id)
3027
found_ids.update(next_revs)
3028
found_ids.update(ghosts)
3029
found_ids = frozenset(found_ids)
3030
# Double query here: should be able to avoid this by changing the
3031
# graph api further.
3032
result_set = found_ids - frozenset(
3033
self.target_get_parent_map(found_ids))
3035
source_ids = self.source.all_revision_ids()
3036
# source_ids is the worst possible case we may need to pull.
3037
# now we want to filter source_ids against what we actually
3038
# have in target, but don't try to check for existence where we know
3039
# we do not have a revision as that would be pointless.
3040
target_ids = set(self.target.all_revision_ids())
3041
result_set = set(source_ids).difference(target_ids)
3042
return self.source.revision_ids_to_search_result(result_set)
3045
class InterModel1and2(InterRepository):
3048
def _get_repo_format_to_test(self):
3052
def is_compatible(source, target):
3053
if not source.supports_rich_root() and target.supports_rich_root():
3059
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3060
"""See InterRepository.fetch()."""
3061
from bzrlib.fetch import Model1toKnit2Fetcher
3062
f = Model1toKnit2Fetcher(to_repository=self.target,
3063
from_repository=self.source,
3064
last_revision=revision_id,
3065
pb=pb, find_ghosts=find_ghosts)
3066
return f.count_copied, f.failed_revisions
3069
def copy_content(self, revision_id=None):
3070
"""Make a complete copy of the content in self into destination.
3072
This is a destructive operation! Do not use it on existing
3075
:param revision_id: Only copy the content needed to construct
3076
revision_id and its parents.
3079
self.target.set_make_working_trees(self.source.make_working_trees())
3080
except NotImplementedError:
3082
# but don't bother fetching if we have the needed data now.
3083
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
3084
self.target.has_revision(revision_id)):
3086
self.target.fetch(self.source, revision_id=revision_id)
3089
class InterKnit1and2(InterKnitRepo):
3092
def _get_repo_format_to_test(self):
3096
def is_compatible(source, target):
3097
"""Be compatible with Knit1 source and Knit3 target"""
3099
from bzrlib.repofmt.knitrepo import (
3100
RepositoryFormatKnit1,
3101
RepositoryFormatKnit3,
3103
from bzrlib.repofmt.pack_repo import (
3104
RepositoryFormatKnitPack1,
3105
RepositoryFormatKnitPack3,
3106
RepositoryFormatKnitPack4,
3107
RepositoryFormatKnitPack5,
3108
RepositoryFormatKnitPack5RichRoot,
3109
RepositoryFormatKnitPack6,
3110
RepositoryFormatKnitPack6RichRoot,
3111
RepositoryFormatPackDevelopment2,
3112
RepositoryFormatPackDevelopment2Subtree,
3115
RepositoryFormatKnit1, # no rr, no subtree
3116
RepositoryFormatKnitPack1, # no rr, no subtree
3117
RepositoryFormatPackDevelopment2, # no rr, no subtree
3118
RepositoryFormatKnitPack5, # no rr, no subtree
3119
RepositoryFormatKnitPack6, # no rr, no subtree
3122
RepositoryFormatKnit3, # rr, subtree
3123
RepositoryFormatKnitPack3, # rr, subtree
3124
RepositoryFormatKnitPack4, # rr, no subtree
3125
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
3126
RepositoryFormatKnitPack6RichRoot,# rr, no subtree
3127
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
3129
for format in norichroot:
3130
if format.rich_root_data:
3131
raise AssertionError('Format %s is a rich-root format'
3132
' but is included in the non-rich-root list'
3134
for format in richroot:
3135
if not format.rich_root_data:
3136
raise AssertionError('Format %s is not a rich-root format'
3137
' but is included in the rich-root list'
3139
# TODO: One alternative is to just check format.rich_root_data,
3140
# instead of keeping membership lists. However, the formats
3141
# *also* have to use the same 'Knit' style of storage
3142
# (line-deltas, fulltexts, etc.)
3143
return (isinstance(source._format, norichroot) and
3144
isinstance(target._format, richroot))
3145
except AttributeError:
3149
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3150
"""See InterRepository.fetch()."""
3151
from bzrlib.fetch import Knit1to2Fetcher
3152
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3153
self.source, self.source._format, self.target,
3154
self.target._format)
3155
f = Knit1to2Fetcher(to_repository=self.target,
3156
from_repository=self.source,
3157
last_revision=revision_id,
3158
pb=pb, find_ghosts=find_ghosts)
3159
return f.count_copied, f.failed_revisions
3162
class InterDifferingSerializer(InterKnitRepo):
3165
def _get_repo_format_to_test(self):
3169
def is_compatible(source, target):
3170
"""Be compatible with Knit2 source and Knit3 target"""
3171
if source.supports_rich_root() != target.supports_rich_root():
3173
# Ideally, we'd support fetching if the source had no tree references
3174
# even if it supported them...
3175
if (getattr(source, '_format.supports_tree_reference', False) and
3176
not getattr(target, '_format.supports_tree_reference', False)):
3181
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3182
"""See InterRepository.fetch()."""
3183
revision_ids = self.target.search_missing_revision_ids(self.source,
3184
revision_id, find_ghosts=find_ghosts).get_keys()
3185
revision_ids = tsort.topo_sort(
3186
self.source.get_graph().get_parent_map(revision_ids))
3187
def revisions_iterator():
3188
rev_ids = list(revision_ids)
3189
for offset in xrange(0, len(rev_ids), 100):
3190
current_revids = rev_ids[offset:offset+100]
3191
revisions = self.source.get_revisions(current_revids)
3192
trees = self.source.revision_trees(current_revids)
3193
keys = [(r,) for r in current_revids]
3194
sig_stream = self.source.signatures.get_record_stream(
3195
keys, 'unordered', True)
3197
for record in versionedfile.filter_absent(sig_stream):
3198
sigs[record.key[0]] = record.get_bytes_as('fulltext')
3199
for rev, tree in zip(revisions, trees):
3200
yield rev, tree, sigs.get(rev.revision_id, None)
3202
my_pb = ui.ui_factory.nested_progress_bar()
3207
install_revisions(self.target, revisions_iterator(),
3208
len(revision_ids), pb)
3210
if my_pb is not None:
3212
return len(revision_ids), 0
3215
class InterOtherToRemote(InterRepository):
3216
"""An InterRepository that simply delegates to the 'real' InterRepository
3217
calculated for (source, target._real_repository).
3220
_walk_to_common_revisions_batch_size = 50
3222
def __init__(self, source, target):
3223
InterRepository.__init__(self, source, target)
3224
self._real_inter = None
3227
def is_compatible(source, target):
3228
if isinstance(target, remote.RemoteRepository):
3232
def _ensure_real_inter(self):
3233
if self._real_inter is None:
3234
self.target._ensure_real()
3235
real_target = self.target._real_repository
3236
self._real_inter = InterRepository.get(self.source, real_target)
3237
# Make _real_inter use the RemoteRepository for get_parent_map
3238
self._real_inter.target_get_graph = self.target.get_graph
3239
self._real_inter.target_get_parent_map = self.target.get_parent_map
3241
def copy_content(self, revision_id=None):
3242
self._ensure_real_inter()
3243
self._real_inter.copy_content(revision_id=revision_id)
3245
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3246
self._ensure_real_inter()
3247
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3248
find_ghosts=find_ghosts)
3251
def _get_repo_format_to_test(self):
3255
class InterRemoteToOther(InterRepository):
3257
def __init__(self, source, target):
3258
InterRepository.__init__(self, source, target)
3259
self._real_inter = None
3262
def is_compatible(source, target):
3263
if not isinstance(source, remote.RemoteRepository):
3265
# Is source's model compatible with target's model?
3266
source._ensure_real()
3267
real_source = source._real_repository
3268
if isinstance(real_source, remote.RemoteRepository):
3269
raise NotImplementedError(
3270
"We don't support remote repos backed by remote repos yet.")
3271
return InterRepository._same_model(real_source, target)
3273
def _ensure_real_inter(self):
3274
if self._real_inter is None:
3275
self.source._ensure_real()
3276
real_source = self.source._real_repository
3277
self._real_inter = InterRepository.get(real_source, self.target)
3279
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3280
self._ensure_real_inter()
3281
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3282
find_ghosts=find_ghosts)
3284
def copy_content(self, revision_id=None):
3285
self._ensure_real_inter()
3286
self._real_inter.copy_content(revision_id=revision_id)
3289
def _get_repo_format_to_test(self):
3294
class InterPackToRemotePack(InterPackRepo):
3295
"""A specialisation of InterPackRepo for a target that is a
3298
This will use the get_parent_map RPC rather than plain readvs, and also
3299
uses an RPC for autopacking.
3302
_walk_to_common_revisions_batch_size = 50
3305
def is_compatible(source, target):
3306
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3307
if isinstance(source._format, RepositoryFormatPack):
3308
if isinstance(target, remote.RemoteRepository):
3309
target._ensure_real()
3310
if isinstance(target._real_repository._format,
3311
RepositoryFormatPack):
3312
if InterRepository._same_model(source, target):
3316
def _autopack(self):
3317
self.target.autopack()
3319
def _get_target_pack_collection(self):
3320
return self.target._real_repository._pack_collection
3323
def _get_repo_format_to_test(self):
3327
InterRepository.register_optimiser(InterDifferingSerializer)
3328
InterRepository.register_optimiser(InterSameDataRepository)
3329
InterRepository.register_optimiser(InterWeaveRepo)
3330
InterRepository.register_optimiser(InterKnitRepo)
3331
InterRepository.register_optimiser(InterModel1and2)
3332
InterRepository.register_optimiser(InterKnit1and2)
3333
InterRepository.register_optimiser(InterPackRepo)
3334
InterRepository.register_optimiser(InterOtherToRemote)
3335
InterRepository.register_optimiser(InterRemoteToOther)
3336
InterRepository.register_optimiser(InterPackToRemotePack)
3339
class CopyConverter(object):
3340
"""A repository conversion tool which just performs a copy of the content.
3342
This is slow but quite reliable.
3345
def __init__(self, target_format):
3346
"""Create a CopyConverter.
3348
:param target_format: The format the resulting repository should be.
3350
self.target_format = target_format
3352
def convert(self, repo, pb):
3353
"""Perform the conversion of to_convert, giving feedback via pb.
3355
:param to_convert: The disk object to convert.
3356
:param pb: a progress bar to use for progress information.
3361
# this is only useful with metadir layouts - separated repo content.
3362
# trigger an assertion if not such
3363
repo._format.get_format_string()
3364
self.repo_dir = repo.bzrdir
3365
self.step('Moving repository to repository.backup')
3366
self.repo_dir.transport.move('repository', 'repository.backup')
3367
backup_transport = self.repo_dir.transport.clone('repository.backup')
3368
repo._format.check_conversion_target(self.target_format)
3369
self.source_repo = repo._format.open(self.repo_dir,
3371
_override_transport=backup_transport)
3372
self.step('Creating new repository')
3373
converted = self.target_format.initialize(self.repo_dir,
3374
self.source_repo.is_shared())
3375
converted.lock_write()
3377
self.step('Copying content into repository.')
3378
self.source_repo.copy_content_into(converted)
3381
self.step('Deleting old repository content.')
3382
self.repo_dir.transport.delete_tree('repository.backup')
3383
self.pb.note('repository converted')
3385
def step(self, message):
3386
"""Update the pb by a step."""
3388
self.pb.update(message, self.count, self.total)
3400
def _unescaper(match, _map=_unescape_map):
3401
code = match.group(1)
3405
if not code.startswith('#'):
3407
return unichr(int(code[1:])).encode('utf8')
3413
def _unescape_xml(data):
3414
"""Unescape predefined XML entities in a string of data."""
3416
if _unescape_re is None:
3417
_unescape_re = re.compile('\&([^;]*);')
3418
return _unescape_re.sub(_unescaper, data)
3421
class _VersionedFileChecker(object):
3423
def __init__(self, repository):
3424
self.repository = repository
3425
self.text_index = self.repository._generate_text_key_index()
3427
def calculate_file_version_parents(self, text_key):
3428
"""Calculate the correct parents for a file version according to
3431
parent_keys = self.text_index[text_key]
3432
if parent_keys == [_mod_revision.NULL_REVISION]:
3434
return tuple(parent_keys)
3436
def check_file_version_parents(self, texts, progress_bar=None):
3437
"""Check the parents stored in a versioned file are correct.
3439
It also detects file versions that are not referenced by their
3440
corresponding revision's inventory.
3442
:returns: A tuple of (wrong_parents, dangling_file_versions).
3443
wrong_parents is a dict mapping {revision_id: (stored_parents,
3444
correct_parents)} for each revision_id where the stored parents
3445
are not correct. dangling_file_versions is a set of (file_id,
3446
revision_id) tuples for versions that are present in this versioned
3447
file, but not used by the corresponding inventory.
3450
self.file_ids = set([file_id for file_id, _ in
3451
self.text_index.iterkeys()])
3452
# text keys is now grouped by file_id
3453
n_weaves = len(self.file_ids)
3454
files_in_revisions = {}
3455
revisions_of_files = {}
3456
n_versions = len(self.text_index)
3457
progress_bar.update('loading text store', 0, n_versions)
3458
parent_map = self.repository.texts.get_parent_map(self.text_index)
3459
# On unlistable transports this could well be empty/error...
3460
text_keys = self.repository.texts.keys()
3461
unused_keys = frozenset(text_keys) - set(self.text_index)
3462
for num, key in enumerate(self.text_index.iterkeys()):
3463
if progress_bar is not None:
3464
progress_bar.update('checking text graph', num, n_versions)
3465
correct_parents = self.calculate_file_version_parents(key)
3467
knit_parents = parent_map[key]
3468
except errors.RevisionNotPresent:
3471
if correct_parents != knit_parents:
3472
wrong_parents[key] = (knit_parents, correct_parents)
3473
return wrong_parents, unused_keys
3476
def _old_get_graph(repository, revision_id):
3477
"""DO NOT USE. That is all. I'm serious."""
3478
graph = repository.get_graph()
3479
revision_graph = dict(((key, value) for key, value in
3480
graph.iter_ancestry([revision_id]) if value is not None))
3481
return _strip_NULL_ghosts(revision_graph)
3484
def _strip_NULL_ghosts(revision_graph):
3485
"""Also don't use this. more compatibility code for unmigrated clients."""
3486
# Filter ghosts, and null:
3487
if _mod_revision.NULL_REVISION in revision_graph:
3488
del revision_graph[_mod_revision.NULL_REVISION]
3489
for key, parents in revision_graph.items():
3490
revision_graph[key] = tuple(parent for parent in parents if parent
3492
return revision_graph
added
removed
bzrlib/repository.py
