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
delta = (path, None, file_id, None)
281
self._basis_delta.append(delta)
284
def will_record_deletes(self):
285
"""Tell the commit builder that deletes are being notified.
287
This enables the accumulation of an inventory delta; for the resulting
288
commit to be valid, deletes against the basis MUST be recorded via
289
builder.record_delete().
291
self._recording_deletes = True
293
def record_entry_contents(self, ie, parent_invs, path, tree,
295
"""Record the content of ie from tree into the commit if needed.
297
Side effect: sets ie.revision when unchanged
299
:param ie: An inventory entry present in the commit.
300
:param parent_invs: The inventories of the parent revisions of the
302
:param path: The path the entry is at in the tree.
303
:param tree: The tree which contains this entry and should be used to
305
:param content_summary: Summary data from the tree about the paths
306
content - stat, length, exec, sha/link target. This is only
307
accessed when the entry has a revision of None - that is when it is
308
a candidate to commit.
309
:return: A tuple (change_delta, version_recorded, fs_hash).
310
change_delta is an inventory_delta change for this entry against
311
the basis tree of the commit, or None if no change occured against
313
version_recorded is True if a new version of the entry has been
314
recorded. For instance, committing a merge where a file was only
315
changed on the other side will return (delta, False).
316
fs_hash is either None, or the hash details for the path (currently
317
a tuple of the contents sha1 and the statvalue returned by
318
tree.get_file_with_stat()).
320
if self.new_inventory.root is None:
321
if ie.parent_id is not None:
322
raise errors.RootMissing()
323
self._check_root(ie, parent_invs, tree)
324
if ie.revision is None:
325
kind = content_summary[0]
327
# ie is carried over from a prior commit
329
# XXX: repository specific check for nested tree support goes here - if
330
# the repo doesn't want nested trees we skip it ?
331
if (kind == 'tree-reference' and
332
not self.repository._format.supports_tree_reference):
333
# mismatch between commit builder logic and repository:
334
# this needs the entry creation pushed down into the builder.
335
raise NotImplementedError('Missing repository subtree support.')
336
self.new_inventory.add(ie)
338
# TODO: slow, take it out of the inner loop.
340
basis_inv = parent_invs[0]
342
basis_inv = Inventory(root_id=None)
344
# ie.revision is always None if the InventoryEntry is considered
345
# for committing. We may record the previous parents revision if the
346
# content is actually unchanged against a sole head.
347
if ie.revision is not None:
348
if not self._versioned_root and path == '':
349
# repositories that do not version the root set the root's
350
# revision to the new commit even when no change occurs (more
351
# specifically, they do not record a revision on the root; and
352
# the rev id is assigned to the root during deserialisation -
353
# this masks when a change may have occurred against the basis.
354
# To match this we always issue a delta, because the revision
355
# of the root will always be changing.
356
if ie.file_id in basis_inv:
357
delta = (basis_inv.id2path(ie.file_id), path,
361
delta = (None, path, ie.file_id, ie)
362
self._basis_delta.append(delta)
363
return delta, False, None
365
# we don't need to commit this, because the caller already
366
# determined that an existing revision of this file is
367
# appropriate. If its not being considered for committing then
368
# it and all its parents to the root must be unaltered so
369
# no-change against the basis.
370
if ie.revision == self._new_revision_id:
371
raise AssertionError("Impossible situation, a skipped "
372
"inventory entry (%r) claims to be modified in this "
373
"commit (%r).", (ie, self._new_revision_id))
374
return None, False, None
375
# XXX: Friction: parent_candidates should return a list not a dict
376
# so that we don't have to walk the inventories again.
377
parent_candiate_entries = ie.parent_candidates(parent_invs)
378
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
380
for inv in parent_invs:
381
if ie.file_id in inv:
382
old_rev = inv[ie.file_id].revision
383
if old_rev in head_set:
384
heads.append(inv[ie.file_id].revision)
385
head_set.remove(inv[ie.file_id].revision)
388
# now we check to see if we need to write a new record to the
390
# We write a new entry unless there is one head to the ancestors, and
391
# the kind-derived content is unchanged.
393
# Cheapest check first: no ancestors, or more the one head in the
394
# ancestors, we write a new node.
398
# There is a single head, look it up for comparison
399
parent_entry = parent_candiate_entries[heads[0]]
400
# if the non-content specific data has changed, we'll be writing a
402
if (parent_entry.parent_id != ie.parent_id or
403
parent_entry.name != ie.name):
405
# now we need to do content specific checks:
407
# if the kind changed the content obviously has
408
if kind != parent_entry.kind:
410
# Stat cache fingerprint feedback for the caller - None as we usually
411
# don't generate one.
414
if content_summary[2] is None:
415
raise ValueError("Files must not have executable = None")
417
if (# if the file length changed we have to store:
418
parent_entry.text_size != content_summary[1] or
419
# if the exec bit has changed we have to store:
420
parent_entry.executable != content_summary[2]):
422
elif parent_entry.text_sha1 == content_summary[3]:
423
# all meta and content is unchanged (using a hash cache
424
# hit to check the sha)
425
ie.revision = parent_entry.revision
426
ie.text_size = parent_entry.text_size
427
ie.text_sha1 = parent_entry.text_sha1
428
ie.executable = parent_entry.executable
429
return self._get_delta(ie, basis_inv, path), False, None
431
# Either there is only a hash change(no hash cache entry,
432
# or same size content change), or there is no change on
434
# Provide the parent's hash to the store layer, so that the
435
# content is unchanged we will not store a new node.
436
nostore_sha = parent_entry.text_sha1
438
# We want to record a new node regardless of the presence or
439
# absence of a content change in the file.
441
ie.executable = content_summary[2]
442
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
444
lines = file_obj.readlines()
448
ie.text_sha1, ie.text_size = self._add_text_to_weave(
449
ie.file_id, lines, heads, nostore_sha)
450
# Let the caller know we generated a stat fingerprint.
451
fingerprint = (ie.text_sha1, stat_value)
452
except errors.ExistingContent:
453
# Turns out that the file content was unchanged, and we were
454
# only going to store a new node if it was changed. Carry over
456
ie.revision = parent_entry.revision
457
ie.text_size = parent_entry.text_size
458
ie.text_sha1 = parent_entry.text_sha1
459
ie.executable = parent_entry.executable
460
return self._get_delta(ie, basis_inv, path), False, None
461
elif kind == 'directory':
463
# all data is meta here, nothing specific to directory, so
465
ie.revision = parent_entry.revision
466
return self._get_delta(ie, basis_inv, path), False, None
468
self._add_text_to_weave(ie.file_id, lines, heads, None)
469
elif kind == 'symlink':
470
current_link_target = content_summary[3]
472
# symlink target is not generic metadata, check if it has
474
if current_link_target != parent_entry.symlink_target:
477
# unchanged, carry over.
478
ie.revision = parent_entry.revision
479
ie.symlink_target = parent_entry.symlink_target
480
return self._get_delta(ie, basis_inv, path), False, None
481
ie.symlink_target = current_link_target
483
self._add_text_to_weave(ie.file_id, lines, heads, None)
484
elif kind == 'tree-reference':
486
if content_summary[3] != parent_entry.reference_revision:
489
# unchanged, carry over.
490
ie.reference_revision = parent_entry.reference_revision
491
ie.revision = parent_entry.revision
492
return self._get_delta(ie, basis_inv, path), False, None
493
ie.reference_revision = content_summary[3]
495
self._add_text_to_weave(ie.file_id, lines, heads, None)
497
raise NotImplementedError('unknown kind')
498
ie.revision = self._new_revision_id
499
return self._get_delta(ie, basis_inv, path), True, fingerprint
501
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
502
# Note: as we read the content directly from the tree, we know its not
503
# been turned into unicode or badly split - but a broken tree
504
# implementation could give us bad output from readlines() so this is
505
# not a guarantee of safety. What would be better is always checking
506
# the content during test suite execution. RBC 20070912
507
parent_keys = tuple((file_id, parent) for parent in parents)
508
return self.repository.texts.add_lines(
509
(file_id, self._new_revision_id), parent_keys, new_lines,
510
nostore_sha=nostore_sha, random_id=self.random_revid,
511
check_content=False)[0:2]
514
class RootCommitBuilder(CommitBuilder):
515
"""This commitbuilder actually records the root id"""
517
# the root entry gets versioned properly by this builder.
518
_versioned_root = True
520
def _check_root(self, ie, parent_invs, tree):
521
"""Helper for record_entry_contents.
523
:param ie: An entry being added.
524
:param parent_invs: The inventories of the parent revisions of the
526
:param tree: The tree that is being committed.
530
######################################################################
533
class Repository(object):
534
"""Repository holding history for one or more branches.
536
The repository holds and retrieves historical information including
537
revisions and file history. It's normally accessed only by the Branch,
538
which views a particular line of development through that history.
540
The Repository builds on top of some byte storage facilies (the revisions,
541
signatures, inventories and texts attributes) and a Transport, which
542
respectively provide byte storage and a means to access the (possibly
545
The byte storage facilities are addressed via tuples, which we refer to
546
as 'keys' throughout the code base. Revision_keys, inventory_keys and
547
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
548
(file_id, revision_id). We use this interface because it allows low
549
friction with the underlying code that implements disk indices, network
550
encoding and other parts of bzrlib.
552
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
553
the serialised revisions for the repository. This can be used to obtain
554
revision graph information or to access raw serialised revisions.
555
The result of trying to insert data into the repository via this store
556
is undefined: it should be considered read-only except for implementors
558
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
559
the serialised signatures for the repository. This can be used to
560
obtain access to raw serialised signatures. The result of trying to
561
insert data into the repository via this store is undefined: it should
562
be considered read-only except for implementors of repositories.
563
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
564
the serialised inventories for the repository. This can be used to
565
obtain unserialised inventories. The result of trying to insert data
566
into the repository via this store is undefined: it should be
567
considered read-only except for implementors of repositories.
568
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
569
texts of files and directories for the repository. This can be used to
570
obtain file texts or file graphs. Note that Repository.iter_file_bytes
571
is usually a better interface for accessing file texts.
572
The result of trying to insert data into the repository via this store
573
is undefined: it should be considered read-only except for implementors
575
:ivar _transport: Transport for file access to repository, typically
576
pointing to .bzr/repository.
579
# What class to use for a CommitBuilder. Often its simpler to change this
580
# in a Repository class subclass rather than to override
581
# get_commit_builder.
582
_commit_builder_class = CommitBuilder
583
# The search regex used by xml based repositories to determine what things
584
# where changed in a single commit.
585
_file_ids_altered_regex = lazy_regex.lazy_compile(
586
r'file_id="(?P<file_id>[^"]+)"'
587
r'.* revision="(?P<revision_id>[^"]+)"'
590
def abort_write_group(self, suppress_errors=False):
591
"""Commit the contents accrued within the current write group.
593
:param suppress_errors: if true, abort_write_group will catch and log
594
unexpected errors that happen during the abort, rather than
595
allowing them to propagate. Defaults to False.
597
:seealso: start_write_group.
599
if self._write_group is not self.get_transaction():
600
# has an unlock or relock occured ?
601
raise errors.BzrError('mismatched lock context and write group.')
603
self._abort_write_group()
604
except Exception, exc:
605
self._write_group = None
606
if not suppress_errors:
608
mutter('abort_write_group failed')
609
log_exception_quietly()
610
note('bzr: ERROR (ignored): %s', exc)
611
self._write_group = None
613
def _abort_write_group(self):
614
"""Template method for per-repository write group cleanup.
616
This is called during abort before the write group is considered to be
617
finished and should cleanup any internal state accrued during the write
618
group. There is no requirement that data handed to the repository be
619
*not* made available - this is not a rollback - but neither should any
620
attempt be made to ensure that data added is fully commited. Abort is
621
invoked when an error has occured so futher disk or network operations
622
may not be possible or may error and if possible should not be
626
def add_fallback_repository(self, repository):
627
"""Add a repository to use for looking up data not held locally.
629
:param repository: A repository.
631
if not self._format.supports_external_lookups:
632
raise errors.UnstackableRepositoryFormat(self._format, self.base)
633
self._check_fallback_repository(repository)
634
self._fallback_repositories.append(repository)
635
self.texts.add_fallback_versioned_files(repository.texts)
636
self.inventories.add_fallback_versioned_files(repository.inventories)
637
self.revisions.add_fallback_versioned_files(repository.revisions)
638
self.signatures.add_fallback_versioned_files(repository.signatures)
639
self._fetch_order = 'topological'
641
def _check_fallback_repository(self, repository):
642
"""Check that this repository can fallback to repository safely.
644
Raise an error if not.
646
:param repository: A repository to fallback to.
648
return InterRepository._assert_same_model(self, repository)
650
def add_inventory(self, revision_id, inv, parents):
651
"""Add the inventory inv to the repository as revision_id.
653
:param parents: The revision ids of the parents that revision_id
654
is known to have and are in the repository already.
656
:returns: The validator(which is a sha1 digest, though what is sha'd is
657
repository format specific) of the serialized inventory.
659
if not self.is_in_write_group():
660
raise AssertionError("%r not in write group" % (self,))
661
_mod_revision.check_not_reserved_id(revision_id)
662
if not (inv.revision_id is None or inv.revision_id == revision_id):
663
raise AssertionError(
664
"Mismatch between inventory revision"
665
" id and insertion revid (%r, %r)"
666
% (inv.revision_id, revision_id))
668
raise AssertionError()
669
inv_lines = self._serialise_inventory_to_lines(inv)
670
return self._inventory_add_lines(revision_id, parents,
671
inv_lines, check_content=False)
673
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
675
"""Add a new inventory expressed as a delta against another revision.
677
:param basis_revision_id: The inventory id the delta was created
678
against. (This does not have to be a direct parent.)
679
:param delta: The inventory delta (see Inventory.apply_delta for
681
:param new_revision_id: The revision id that the inventory is being
683
:param parents: The revision ids of the parents that revision_id is
684
known to have and are in the repository already. These are supplied
685
for repositories that depend on the inventory graph for revision
686
graph access, as well as for those that pun ancestry with delta
689
:returns: (validator, new_inv)
690
The validator(which is a sha1 digest, though what is sha'd is
691
repository format specific) of the serialized inventory, and the
694
if not self.is_in_write_group():
695
raise AssertionError("%r not in write group" % (self,))
696
_mod_revision.check_not_reserved_id(new_revision_id)
697
basis_tree = self.revision_tree(basis_revision_id)
698
basis_tree.lock_read()
700
# Note that this mutates the inventory of basis_tree, which not all
701
# inventory implementations may support: A better idiom would be to
702
# return a new inventory, but as there is no revision tree cache in
703
# repository this is safe for now - RBC 20081013
704
basis_inv = basis_tree.inventory
705
basis_inv.apply_delta(delta)
706
basis_inv.revision_id = new_revision_id
707
return (self.add_inventory(new_revision_id, basis_inv, parents),
712
def _inventory_add_lines(self, revision_id, parents, lines,
714
"""Store lines in inv_vf and return the sha1 of the inventory."""
715
parents = [(parent,) for parent in parents]
716
return self.inventories.add_lines((revision_id,), parents, lines,
717
check_content=check_content)[0]
719
def add_revision(self, revision_id, rev, inv=None, config=None):
720
"""Add rev to the revision store as revision_id.
722
:param revision_id: the revision id to use.
723
:param rev: The revision object.
724
:param inv: The inventory for the revision. if None, it will be looked
725
up in the inventory storer
726
:param config: If None no digital signature will be created.
727
If supplied its signature_needed method will be used
728
to determine if a signature should be made.
730
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
732
_mod_revision.check_not_reserved_id(revision_id)
733
if config is not None and config.signature_needed():
735
inv = self.get_inventory(revision_id)
736
plaintext = Testament(rev, inv).as_short_text()
737
self.store_revision_signature(
738
gpg.GPGStrategy(config), plaintext, revision_id)
739
# check inventory present
740
if not self.inventories.get_parent_map([(revision_id,)]):
742
raise errors.WeaveRevisionNotPresent(revision_id,
745
# yes, this is not suitable for adding with ghosts.
746
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
750
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
751
self._add_revision(rev)
753
def _add_revision(self, revision):
754
text = self._serializer.write_revision_to_string(revision)
755
key = (revision.revision_id,)
756
parents = tuple((parent,) for parent in revision.parent_ids)
757
self.revisions.add_lines(key, parents, osutils.split_lines(text))
759
def all_revision_ids(self):
760
"""Returns a list of all the revision ids in the repository.
762
This is conceptually deprecated because code should generally work on
763
the graph reachable from a particular revision, and ignore any other
764
revisions that might be present. There is no direct replacement
767
if 'evil' in debug.debug_flags:
768
mutter_callsite(2, "all_revision_ids is linear with history.")
769
return self._all_revision_ids()
771
def _all_revision_ids(self):
772
"""Returns a list of all the revision ids in the repository.
774
These are in as much topological order as the underlying store can
777
raise NotImplementedError(self._all_revision_ids)
779
def break_lock(self):
780
"""Break a lock if one is present from another instance.
782
Uses the ui factory to ask for confirmation if the lock may be from
785
self.control_files.break_lock()
788
def _eliminate_revisions_not_present(self, revision_ids):
789
"""Check every revision id in revision_ids to see if we have it.
791
Returns a set of the present revisions.
794
graph = self.get_graph()
795
parent_map = graph.get_parent_map(revision_ids)
796
# The old API returned a list, should this actually be a set?
797
return parent_map.keys()
800
def create(a_bzrdir):
801
"""Construct the current default format repository in a_bzrdir."""
802
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
804
def __init__(self, _format, a_bzrdir, control_files):
805
"""instantiate a Repository.
807
:param _format: The format of the repository on disk.
808
:param a_bzrdir: The BzrDir of the repository.
810
In the future we will have a single api for all stores for
811
getting file texts, inventories and revisions, then
812
this construct will accept instances of those things.
814
super(Repository, self).__init__()
815
self._format = _format
816
# the following are part of the public API for Repository:
817
self.bzrdir = a_bzrdir
818
self.control_files = control_files
819
self._transport = control_files._transport
820
self.base = self._transport.base
822
self._reconcile_does_inventory_gc = True
823
self._reconcile_fixes_text_parents = False
824
self._reconcile_backsup_inventory = True
825
# not right yet - should be more semantically clear ?
827
# TODO: make sure to construct the right store classes, etc, depending
828
# on whether escaping is required.
829
self._warn_if_deprecated()
830
self._write_group = None
831
# Additional places to query for data.
832
self._fallback_repositories = []
833
# What order should fetch operations request streams in?
834
# The default is unordered as that is the cheapest for an origin to
836
self._fetch_order = 'unordered'
837
# Does this repository use deltas that can be fetched as-deltas ?
838
# (E.g. knits, where the knit deltas can be transplanted intact.
839
# We default to False, which will ensure that enough data to get
840
# a full text out of any fetch stream will be grabbed.
841
self._fetch_uses_deltas = False
842
# Should fetch trigger a reconcile after the fetch? Only needed for
843
# some repository formats that can suffer internal inconsistencies.
844
self._fetch_reconcile = False
847
return '%s(%r)' % (self.__class__.__name__,
850
def has_same_location(self, other):
851
"""Returns a boolean indicating if this repository is at the same
852
location as another repository.
854
This might return False even when two repository objects are accessing
855
the same physical repository via different URLs.
857
if self.__class__ is not other.__class__:
859
return (self._transport.base == other._transport.base)
861
def is_in_write_group(self):
862
"""Return True if there is an open write group.
864
:seealso: start_write_group.
866
return self._write_group is not None
869
return self.control_files.is_locked()
871
def is_write_locked(self):
872
"""Return True if this object is write locked."""
873
return self.is_locked() and self.control_files._lock_mode == 'w'
875
def lock_write(self, token=None):
876
"""Lock this repository for writing.
878
This causes caching within the repository obejct to start accumlating
879
data during reads, and allows a 'write_group' to be obtained. Write
880
groups must be used for actual data insertion.
882
:param token: if this is already locked, then lock_write will fail
883
unless the token matches the existing lock.
884
:returns: a token if this instance supports tokens, otherwise None.
885
:raises TokenLockingNotSupported: when a token is given but this
886
instance doesn't support using token locks.
887
:raises MismatchedToken: if the specified token doesn't match the token
888
of the existing lock.
889
:seealso: start_write_group.
891
A token should be passed in if you know that you have locked the object
892
some other way, and need to synchronise this object's state with that
895
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
897
result = self.control_files.lock_write(token=token)
898
for repo in self._fallback_repositories:
899
# Writes don't affect fallback repos
905
self.control_files.lock_read()
906
for repo in self._fallback_repositories:
910
def get_physical_lock_status(self):
911
return self.control_files.get_physical_lock_status()
913
def leave_lock_in_place(self):
914
"""Tell this repository not to release the physical lock when this
917
If lock_write doesn't return a token, then this method is not supported.
919
self.control_files.leave_in_place()
921
def dont_leave_lock_in_place(self):
922
"""Tell this repository to release the physical lock when this
923
object is unlocked, even if it didn't originally acquire it.
925
If lock_write doesn't return a token, then this method is not supported.
927
self.control_files.dont_leave_in_place()
930
def gather_stats(self, revid=None, committers=None):
931
"""Gather statistics from a revision id.
933
:param revid: The revision id to gather statistics from, if None, then
934
no revision specific statistics are gathered.
935
:param committers: Optional parameter controlling whether to grab
936
a count of committers from the revision specific statistics.
937
:return: A dictionary of statistics. Currently this contains:
938
committers: The number of committers if requested.
939
firstrev: A tuple with timestamp, timezone for the penultimate left
940
most ancestor of revid, if revid is not the NULL_REVISION.
941
latestrev: A tuple with timestamp, timezone for revid, if revid is
942
not the NULL_REVISION.
943
revisions: The total revision count in the repository.
944
size: An estimate disk size of the repository in bytes.
947
if revid and committers:
948
result['committers'] = 0
949
if revid and revid != _mod_revision.NULL_REVISION:
951
all_committers = set()
952
revisions = self.get_ancestry(revid)
953
# pop the leading None
955
first_revision = None
957
# ignore the revisions in the middle - just grab first and last
958
revisions = revisions[0], revisions[-1]
959
for revision in self.get_revisions(revisions):
960
if not first_revision:
961
first_revision = revision
963
all_committers.add(revision.committer)
964
last_revision = revision
966
result['committers'] = len(all_committers)
967
result['firstrev'] = (first_revision.timestamp,
968
first_revision.timezone)
969
result['latestrev'] = (last_revision.timestamp,
970
last_revision.timezone)
972
# now gather global repository information
973
# XXX: This is available for many repos regardless of listability.
974
if self.bzrdir.root_transport.listable():
975
# XXX: do we want to __define len__() ?
976
# Maybe the versionedfiles object should provide a different
977
# method to get the number of keys.
978
result['revisions'] = len(self.revisions.keys())
982
def find_branches(self, using=False):
983
"""Find branches underneath this repository.
985
This will include branches inside other branches.
987
:param using: If True, list only branches using this repository.
989
if using and not self.is_shared():
991
return [self.bzrdir.open_branch()]
992
except errors.NotBranchError:
994
class Evaluator(object):
997
self.first_call = True
999
def __call__(self, bzrdir):
1000
# On the first call, the parameter is always the bzrdir
1001
# containing the current repo.
1002
if not self.first_call:
1004
repository = bzrdir.open_repository()
1005
except errors.NoRepositoryPresent:
1008
return False, (None, repository)
1009
self.first_call = False
1011
value = (bzrdir.open_branch(), None)
1012
except errors.NotBranchError:
1013
value = (None, None)
1017
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1018
self.bzrdir.root_transport, evaluate=Evaluator()):
1019
if branch is not None:
1020
branches.append(branch)
1021
if not using and repository is not None:
1022
branches.extend(repository.find_branches())
1026
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1027
"""Return the revision ids that other has that this does not.
1029
These are returned in topological order.
1031
revision_id: only return revision ids included by revision_id.
1033
return InterRepository.get(other, self).search_missing_revision_ids(
1034
revision_id, find_ghosts)
1036
@deprecated_method(one_two)
1038
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1039
"""Return the revision ids that other has that this does not.
1041
These are returned in topological order.
1043
revision_id: only return revision ids included by revision_id.
1045
keys = self.search_missing_revision_ids(
1046
other, revision_id, find_ghosts).get_keys()
1049
parents = other.get_graph().get_parent_map(keys)
1052
return tsort.topo_sort(parents)
1056
"""Open the repository rooted at base.
1058
For instance, if the repository is at URL/.bzr/repository,
1059
Repository.open(URL) -> a Repository instance.
1061
control = bzrdir.BzrDir.open(base)
1062
return control.open_repository()
1064
def copy_content_into(self, destination, revision_id=None):
1065
"""Make a complete copy of the content in self into destination.
1067
This is a destructive operation! Do not use it on existing
1070
return InterRepository.get(self, destination).copy_content(revision_id)
1072
def commit_write_group(self):
1073
"""Commit the contents accrued within the current write group.
1075
:seealso: start_write_group.
1077
if self._write_group is not self.get_transaction():
1078
# has an unlock or relock occured ?
1079
raise errors.BzrError('mismatched lock context %r and '
1081
(self.get_transaction(), self._write_group))
1082
self._commit_write_group()
1083
self._write_group = None
1085
def _commit_write_group(self):
1086
"""Template method for per-repository write group cleanup.
1088
This is called before the write group is considered to be
1089
finished and should ensure that all data handed to the repository
1090
for writing during the write group is safely committed (to the
1091
extent possible considering file system caching etc).
1094
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1095
"""Fetch the content required to construct revision_id from source.
1097
If revision_id is None all content is copied.
1098
:param find_ghosts: Find and copy revisions in the source that are
1099
ghosts in the target (and not reachable directly by walking out to
1100
the first-present revision in target from revision_id).
1102
# fast path same-url fetch operations
1103
if self.has_same_location(source):
1104
# check that last_revision is in 'from' and then return a
1106
if (revision_id is not None and
1107
not _mod_revision.is_null(revision_id)):
1108
self.get_revision(revision_id)
1110
# if there is no specific appropriate InterRepository, this will get
1111
# the InterRepository base class, which raises an
1112
# IncompatibleRepositories when asked to fetch.
1113
inter = InterRepository.get(source, self)
1114
return inter.fetch(revision_id=revision_id, pb=pb,
1115
find_ghosts=find_ghosts)
1117
def create_bundle(self, target, base, fileobj, format=None):
1118
return serializer.write_bundle(self, target, base, fileobj, format)
1120
def get_commit_builder(self, branch, parents, config, timestamp=None,
1121
timezone=None, committer=None, revprops=None,
1123
"""Obtain a CommitBuilder for this repository.
1125
:param branch: Branch to commit to.
1126
:param parents: Revision ids of the parents of the new revision.
1127
:param config: Configuration to use.
1128
:param timestamp: Optional timestamp recorded for commit.
1129
:param timezone: Optional timezone for timestamp.
1130
:param committer: Optional committer to set for commit.
1131
:param revprops: Optional dictionary of revision properties.
1132
:param revision_id: Optional revision id.
1134
result = self._commit_builder_class(self, parents, config,
1135
timestamp, timezone, committer, revprops, revision_id)
1136
self.start_write_group()
1140
if (self.control_files._lock_count == 1 and
1141
self.control_files._lock_mode == 'w'):
1142
if self._write_group is not None:
1143
self.abort_write_group()
1144
self.control_files.unlock()
1145
raise errors.BzrError(
1146
'Must end write groups before releasing write locks.')
1147
self.control_files.unlock()
1148
for repo in self._fallback_repositories:
1152
def clone(self, a_bzrdir, revision_id=None):
1153
"""Clone this repository into a_bzrdir using the current format.
1155
Currently no check is made that the format of this repository and
1156
the bzrdir format are compatible. FIXME RBC 20060201.
1158
:return: The newly created destination repository.
1160
# TODO: deprecate after 0.16; cloning this with all its settings is
1161
# probably not very useful -- mbp 20070423
1162
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1163
self.copy_content_into(dest_repo, revision_id)
1166
def start_write_group(self):
1167
"""Start a write group in the repository.
1169
Write groups are used by repositories which do not have a 1:1 mapping
1170
between file ids and backend store to manage the insertion of data from
1171
both fetch and commit operations.
1173
A write lock is required around the start_write_group/commit_write_group
1174
for the support of lock-requiring repository formats.
1176
One can only insert data into a repository inside a write group.
1180
if not self.is_write_locked():
1181
raise errors.NotWriteLocked(self)
1182
if self._write_group:
1183
raise errors.BzrError('already in a write group')
1184
self._start_write_group()
1185
# so we can detect unlock/relock - the write group is now entered.
1186
self._write_group = self.get_transaction()
1188
def _start_write_group(self):
1189
"""Template method for per-repository write group startup.
1191
This is called before the write group is considered to be
1196
def sprout(self, to_bzrdir, revision_id=None):
1197
"""Create a descendent repository for new development.
1199
Unlike clone, this does not copy the settings of the repository.
1201
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1202
dest_repo.fetch(self, revision_id=revision_id)
1205
def _create_sprouting_repo(self, a_bzrdir, shared):
1206
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1207
# use target default format.
1208
dest_repo = a_bzrdir.create_repository()
1210
# Most control formats need the repository to be specifically
1211
# created, but on some old all-in-one formats it's not needed
1213
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1214
except errors.UninitializableFormat:
1215
dest_repo = a_bzrdir.open_repository()
1219
def has_revision(self, revision_id):
1220
"""True if this repository has a copy of the revision."""
1221
return revision_id in self.has_revisions((revision_id,))
1224
def has_revisions(self, revision_ids):
1225
"""Probe to find out the presence of multiple revisions.
1227
:param revision_ids: An iterable of revision_ids.
1228
:return: A set of the revision_ids that were present.
1230
parent_map = self.revisions.get_parent_map(
1231
[(rev_id,) for rev_id in revision_ids])
1233
if _mod_revision.NULL_REVISION in revision_ids:
1234
result.add(_mod_revision.NULL_REVISION)
1235
result.update([key[0] for key in parent_map])
1239
def get_revision(self, revision_id):
1240
"""Return the Revision object for a named revision."""
1241
return self.get_revisions([revision_id])[0]
1244
def get_revision_reconcile(self, revision_id):
1245
"""'reconcile' helper routine that allows access to a revision always.
1247
This variant of get_revision does not cross check the weave graph
1248
against the revision one as get_revision does: but it should only
1249
be used by reconcile, or reconcile-alike commands that are correcting
1250
or testing the revision graph.
1252
return self._get_revisions([revision_id])[0]
1255
def get_revisions(self, revision_ids):
1256
"""Get many revisions at once."""
1257
return self._get_revisions(revision_ids)
1260
def _get_revisions(self, revision_ids):
1261
"""Core work logic to get many revisions without sanity checks."""
1262
for rev_id in revision_ids:
1263
if not rev_id or not isinstance(rev_id, basestring):
1264
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1265
keys = [(key,) for key in revision_ids]
1266
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1268
for record in stream:
1269
if record.storage_kind == 'absent':
1270
raise errors.NoSuchRevision(self, record.key[0])
1271
text = record.get_bytes_as('fulltext')
1272
rev = self._serializer.read_revision_from_string(text)
1273
revs[record.key[0]] = rev
1274
return [revs[revid] for revid in revision_ids]
1277
def get_revision_xml(self, revision_id):
1278
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1279
# would have already do it.
1280
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1281
rev = self.get_revision(revision_id)
1282
rev_tmp = cStringIO.StringIO()
1283
# the current serializer..
1284
self._serializer.write_revision(rev, rev_tmp)
1286
return rev_tmp.getvalue()
1288
def get_deltas_for_revisions(self, revisions):
1289
"""Produce a generator of revision deltas.
1291
Note that the input is a sequence of REVISIONS, not revision_ids.
1292
Trees will be held in memory until the generator exits.
1293
Each delta is relative to the revision's lefthand predecessor.
1295
required_trees = set()
1296
for revision in revisions:
1297
required_trees.add(revision.revision_id)
1298
required_trees.update(revision.parent_ids[:1])
1299
trees = dict((t.get_revision_id(), t) for
1300
t in self.revision_trees(required_trees))
1301
for revision in revisions:
1302
if not revision.parent_ids:
1303
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1305
old_tree = trees[revision.parent_ids[0]]
1306
yield trees[revision.revision_id].changes_from(old_tree)
1309
def get_revision_delta(self, revision_id):
1310
"""Return the delta for one revision.
1312
The delta is relative to the left-hand predecessor of the
1315
r = self.get_revision(revision_id)
1316
return list(self.get_deltas_for_revisions([r]))[0]
1319
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1320
signature = gpg_strategy.sign(plaintext)
1321
self.add_signature_text(revision_id, signature)
1324
def add_signature_text(self, revision_id, signature):
1325
self.signatures.add_lines((revision_id,), (),
1326
osutils.split_lines(signature))
1328
def find_text_key_references(self):
1329
"""Find the text key references within the repository.
1331
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1332
revision_ids. Each altered file-ids has the exact revision_ids that
1333
altered it listed explicitly.
1334
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1335
to whether they were referred to by the inventory of the
1336
revision_id that they contain. The inventory texts from all present
1337
revision ids are assessed to generate this report.
1339
revision_keys = self.revisions.keys()
1340
w = self.inventories
1341
pb = ui.ui_factory.nested_progress_bar()
1343
return self._find_text_key_references_from_xml_inventory_lines(
1344
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1348
def _find_text_key_references_from_xml_inventory_lines(self,
1350
"""Core routine for extracting references to texts from inventories.
1352
This performs the translation of xml lines to revision ids.
1354
:param line_iterator: An iterator of lines, origin_version_id
1355
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1356
to whether they were referred to by the inventory of the
1357
revision_id that they contain. Note that if that revision_id was
1358
not part of the line_iterator's output then False will be given -
1359
even though it may actually refer to that key.
1361
if not self._serializer.support_altered_by_hack:
1362
raise AssertionError(
1363
"_find_text_key_references_from_xml_inventory_lines only "
1364
"supported for branches which store inventory as unnested xml"
1365
", not on %r" % self)
1368
# this code needs to read every new line in every inventory for the
1369
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1370
# not present in one of those inventories is unnecessary but not
1371
# harmful because we are filtering by the revision id marker in the
1372
# inventory lines : we only select file ids altered in one of those
1373
# revisions. We don't need to see all lines in the inventory because
1374
# only those added in an inventory in rev X can contain a revision=X
1376
unescape_revid_cache = {}
1377
unescape_fileid_cache = {}
1379
# jam 20061218 In a big fetch, this handles hundreds of thousands
1380
# of lines, so it has had a lot of inlining and optimizing done.
1381
# Sorry that it is a little bit messy.
1382
# Move several functions to be local variables, since this is a long
1384
search = self._file_ids_altered_regex.search
1385
unescape = _unescape_xml
1386
setdefault = result.setdefault
1387
for line, line_key in line_iterator:
1388
match = search(line)
1391
# One call to match.group() returning multiple items is quite a
1392
# bit faster than 2 calls to match.group() each returning 1
1393
file_id, revision_id = match.group('file_id', 'revision_id')
1395
# Inlining the cache lookups helps a lot when you make 170,000
1396
# lines and 350k ids, versus 8.4 unique ids.
1397
# Using a cache helps in 2 ways:
1398
# 1) Avoids unnecessary decoding calls
1399
# 2) Re-uses cached strings, which helps in future set and
1401
# (2) is enough that removing encoding entirely along with
1402
# the cache (so we are using plain strings) results in no
1403
# performance improvement.
1405
revision_id = unescape_revid_cache[revision_id]
1407
unescaped = unescape(revision_id)
1408
unescape_revid_cache[revision_id] = unescaped
1409
revision_id = unescaped
1411
# Note that unconditionally unescaping means that we deserialise
1412
# every fileid, which for general 'pull' is not great, but we don't
1413
# really want to have some many fulltexts that this matters anyway.
1416
file_id = unescape_fileid_cache[file_id]
1418
unescaped = unescape(file_id)
1419
unescape_fileid_cache[file_id] = unescaped
1422
key = (file_id, revision_id)
1423
setdefault(key, False)
1424
if revision_id == line_key[-1]:
1428
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1430
"""Helper routine for fileids_altered_by_revision_ids.
1432
This performs the translation of xml lines to revision ids.
1434
:param line_iterator: An iterator of lines, origin_version_id
1435
:param revision_ids: The revision ids to filter for. This should be a
1436
set or other type which supports efficient __contains__ lookups, as
1437
the revision id from each parsed line will be looked up in the
1438
revision_ids filter.
1439
:return: a dictionary mapping altered file-ids to an iterable of
1440
revision_ids. Each altered file-ids has the exact revision_ids that
1441
altered it listed explicitly.
1444
setdefault = result.setdefault
1446
self._find_text_key_references_from_xml_inventory_lines(
1447
line_iterator).iterkeys():
1448
# once data is all ensured-consistent; then this is
1449
# if revision_id == version_id
1450
if key[-1:] in revision_ids:
1451
setdefault(key[0], set()).add(key[-1])
1454
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1455
"""Find the file ids and versions affected by revisions.
1457
:param revisions: an iterable containing revision ids.
1458
:param _inv_weave: The inventory weave from this repository or None.
1459
If None, the inventory weave will be opened automatically.
1460
:return: a dictionary mapping altered file-ids to an iterable of
1461
revision_ids. Each altered file-ids has the exact revision_ids that
1462
altered it listed explicitly.
1464
selected_keys = set((revid,) for revid in revision_ids)
1465
w = _inv_weave or self.inventories
1466
pb = ui.ui_factory.nested_progress_bar()
1468
return self._find_file_ids_from_xml_inventory_lines(
1469
w.iter_lines_added_or_present_in_keys(
1470
selected_keys, pb=pb),
1475
def iter_files_bytes(self, desired_files):
1476
"""Iterate through file versions.
1478
Files will not necessarily be returned in the order they occur in
1479
desired_files. No specific order is guaranteed.
1481
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1482
value supplied by the caller as part of desired_files. It should
1483
uniquely identify the file version in the caller's context. (Examples:
1484
an index number or a TreeTransform trans_id.)
1486
bytes_iterator is an iterable of bytestrings for the file. The
1487
kind of iterable and length of the bytestrings are unspecified, but for
1488
this implementation, it is a list of bytes produced by
1489
VersionedFile.get_record_stream().
1491
:param desired_files: a list of (file_id, revision_id, identifier)
1494
transaction = self.get_transaction()
1496
for file_id, revision_id, callable_data in desired_files:
1497
text_keys[(file_id, revision_id)] = callable_data
1498
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1499
if record.storage_kind == 'absent':
1500
raise errors.RevisionNotPresent(record.key, self)
1501
yield text_keys[record.key], record.get_bytes_as('fulltext')
1503
def _generate_text_key_index(self, text_key_references=None,
1505
"""Generate a new text key index for the repository.
1507
This is an expensive function that will take considerable time to run.
1509
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1510
list of parents, also text keys. When a given key has no parents,
1511
the parents list will be [NULL_REVISION].
1513
# All revisions, to find inventory parents.
1514
if ancestors is None:
1515
graph = self.get_graph()
1516
ancestors = graph.get_parent_map(self.all_revision_ids())
1517
if text_key_references is None:
1518
text_key_references = self.find_text_key_references()
1519
pb = ui.ui_factory.nested_progress_bar()
1521
return self._do_generate_text_key_index(ancestors,
1522
text_key_references, pb)
1526
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1527
"""Helper for _generate_text_key_index to avoid deep nesting."""
1528
revision_order = tsort.topo_sort(ancestors)
1529
invalid_keys = set()
1531
for revision_id in revision_order:
1532
revision_keys[revision_id] = set()
1533
text_count = len(text_key_references)
1534
# a cache of the text keys to allow reuse; costs a dict of all the
1535
# keys, but saves a 2-tuple for every child of a given key.
1537
for text_key, valid in text_key_references.iteritems():
1539
invalid_keys.add(text_key)
1541
revision_keys[text_key[1]].add(text_key)
1542
text_key_cache[text_key] = text_key
1543
del text_key_references
1545
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1546
NULL_REVISION = _mod_revision.NULL_REVISION
1547
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1548
# too small for large or very branchy trees. However, for 55K path
1549
# trees, it would be easy to use too much memory trivially. Ideally we
1550
# could gauge this by looking at available real memory etc, but this is
1551
# always a tricky proposition.
1552
inventory_cache = lru_cache.LRUCache(10)
1553
batch_size = 10 # should be ~150MB on a 55K path tree
1554
batch_count = len(revision_order) / batch_size + 1
1556
pb.update("Calculating text parents.", processed_texts, text_count)
1557
for offset in xrange(batch_count):
1558
to_query = revision_order[offset * batch_size:(offset + 1) *
1562
for rev_tree in self.revision_trees(to_query):
1563
revision_id = rev_tree.get_revision_id()
1564
parent_ids = ancestors[revision_id]
1565
for text_key in revision_keys[revision_id]:
1566
pb.update("Calculating text parents.", processed_texts)
1567
processed_texts += 1
1568
candidate_parents = []
1569
for parent_id in parent_ids:
1570
parent_text_key = (text_key[0], parent_id)
1572
check_parent = parent_text_key not in \
1573
revision_keys[parent_id]
1575
# the parent parent_id is a ghost:
1576
check_parent = False
1577
# truncate the derived graph against this ghost.
1578
parent_text_key = None
1580
# look at the parent commit details inventories to
1581
# determine possible candidates in the per file graph.
1584
inv = inventory_cache[parent_id]
1586
inv = self.revision_tree(parent_id).inventory
1587
inventory_cache[parent_id] = inv
1588
parent_entry = inv._byid.get(text_key[0], None)
1589
if parent_entry is not None:
1591
text_key[0], parent_entry.revision)
1593
parent_text_key = None
1594
if parent_text_key is not None:
1595
candidate_parents.append(
1596
text_key_cache[parent_text_key])
1597
parent_heads = text_graph.heads(candidate_parents)
1598
new_parents = list(parent_heads)
1599
new_parents.sort(key=lambda x:candidate_parents.index(x))
1600
if new_parents == []:
1601
new_parents = [NULL_REVISION]
1602
text_index[text_key] = new_parents
1604
for text_key in invalid_keys:
1605
text_index[text_key] = [NULL_REVISION]
1608
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1609
"""Get an iterable listing the keys of all the data introduced by a set
1612
The keys will be ordered so that the corresponding items can be safely
1613
fetched and inserted in that order.
1615
:returns: An iterable producing tuples of (knit-kind, file-id,
1616
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1617
'revisions'. file-id is None unless knit-kind is 'file'.
1619
# XXX: it's a bit weird to control the inventory weave caching in this
1620
# generator. Ideally the caching would be done in fetch.py I think. Or
1621
# maybe this generator should explicitly have the contract that it
1622
# should not be iterated until the previously yielded item has been
1624
inv_w = self.inventories
1626
# file ids that changed
1627
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1629
num_file_ids = len(file_ids)
1630
for file_id, altered_versions in file_ids.iteritems():
1631
if _files_pb is not None:
1632
_files_pb.update("fetch texts", count, num_file_ids)
1634
yield ("file", file_id, altered_versions)
1635
# We're done with the files_pb. Note that it finished by the caller,
1636
# just as it was created by the caller.
1640
yield ("inventory", None, revision_ids)
1643
# XXX: Note ATM no callers actually pay attention to this return
1644
# instead they just use the list of revision ids and ignore
1645
# missing sigs. Consider removing this work entirely
1646
revisions_with_signatures = set(self.signatures.get_parent_map(
1647
[(r,) for r in revision_ids]))
1648
revisions_with_signatures = set(
1649
[r for (r,) in revisions_with_signatures])
1650
revisions_with_signatures.intersection_update(revision_ids)
1651
yield ("signatures", None, revisions_with_signatures)
1654
yield ("revisions", None, revision_ids)
1657
def get_inventory(self, revision_id):
1658
"""Get Inventory object by revision id."""
1659
return self.iter_inventories([revision_id]).next()
1661
def iter_inventories(self, revision_ids):
1662
"""Get many inventories by revision_ids.
1664
This will buffer some or all of the texts used in constructing the
1665
inventories in memory, but will only parse a single inventory at a
1668
:return: An iterator of inventories.
1670
if ((None in revision_ids)
1671
or (_mod_revision.NULL_REVISION in revision_ids)):
1672
raise ValueError('cannot get null revision inventory')
1673
return self._iter_inventories(revision_ids)
1675
def _iter_inventories(self, revision_ids):
1676
"""single-document based inventory iteration."""
1677
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1678
yield self.deserialise_inventory(revision_id, text)
1680
def _iter_inventory_xmls(self, revision_ids):
1681
keys = [(revision_id,) for revision_id in revision_ids]
1682
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1684
for record in stream:
1685
if record.storage_kind != 'absent':
1686
text_chunks[record.key] = record.get_bytes_as('chunked')
1688
raise errors.NoSuchRevision(self, record.key)
1690
chunks = text_chunks.pop(key)
1691
yield ''.join(chunks), key[-1]
1693
def deserialise_inventory(self, revision_id, xml):
1694
"""Transform the xml into an inventory object.
1696
:param revision_id: The expected revision id of the inventory.
1697
:param xml: A serialised inventory.
1699
result = self._serializer.read_inventory_from_string(xml, revision_id)
1700
if result.revision_id != revision_id:
1701
raise AssertionError('revision id mismatch %s != %s' % (
1702
result.revision_id, revision_id))
1705
def serialise_inventory(self, inv):
1706
return self._serializer.write_inventory_to_string(inv)
1708
def _serialise_inventory_to_lines(self, inv):
1709
return self._serializer.write_inventory_to_lines(inv)
1711
def get_serializer_format(self):
1712
return self._serializer.format_num
1715
def get_inventory_xml(self, revision_id):
1716
"""Get inventory XML as a file object."""
1717
texts = self._iter_inventory_xmls([revision_id])
1719
text, revision_id = texts.next()
1720
except StopIteration:
1721
raise errors.HistoryMissing(self, 'inventory', revision_id)
1725
def get_inventory_sha1(self, revision_id):
1726
"""Return the sha1 hash of the inventory entry
1728
return self.get_revision(revision_id).inventory_sha1
1730
def iter_reverse_revision_history(self, revision_id):
1731
"""Iterate backwards through revision ids in the lefthand history
1733
:param revision_id: The revision id to start with. All its lefthand
1734
ancestors will be traversed.
1736
graph = self.get_graph()
1737
next_id = revision_id
1739
if next_id in (None, _mod_revision.NULL_REVISION):
1742
# Note: The following line may raise KeyError in the event of
1743
# truncated history. We decided not to have a try:except:raise
1744
# RevisionNotPresent here until we see a use for it, because of the
1745
# cost in an inner loop that is by its very nature O(history).
1746
# Robert Collins 20080326
1747
parents = graph.get_parent_map([next_id])[next_id]
1748
if len(parents) == 0:
1751
next_id = parents[0]
1754
def get_revision_inventory(self, revision_id):
1755
"""Return inventory of a past revision."""
1756
# TODO: Unify this with get_inventory()
1757
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1758
# must be the same as its revision, so this is trivial.
1759
if revision_id is None:
1760
# This does not make sense: if there is no revision,
1761
# then it is the current tree inventory surely ?!
1762
# and thus get_root_id() is something that looks at the last
1763
# commit on the branch, and the get_root_id is an inventory check.
1764
raise NotImplementedError
1765
# return Inventory(self.get_root_id())
1767
return self.get_inventory(revision_id)
1769
def is_shared(self):
1770
"""Return True if this repository is flagged as a shared repository."""
1771
raise NotImplementedError(self.is_shared)
1774
def reconcile(self, other=None, thorough=False):
1775
"""Reconcile this repository."""
1776
from bzrlib.reconcile import RepoReconciler
1777
reconciler = RepoReconciler(self, thorough=thorough)
1778
reconciler.reconcile()
1781
def _refresh_data(self):
1782
"""Helper called from lock_* to ensure coherency with disk.
1784
The default implementation does nothing; it is however possible
1785
for repositories to maintain loaded indices across multiple locks
1786
by checking inside their implementation of this method to see
1787
whether their indices are still valid. This depends of course on
1788
the disk format being validatable in this manner.
1792
def revision_tree(self, revision_id):
1793
"""Return Tree for a revision on this branch.
1795
`revision_id` may be NULL_REVISION for the empty tree revision.
1797
revision_id = _mod_revision.ensure_null(revision_id)
1798
# TODO: refactor this to use an existing revision object
1799
# so we don't need to read it in twice.
1800
if revision_id == _mod_revision.NULL_REVISION:
1801
return RevisionTree(self, Inventory(root_id=None),
1802
_mod_revision.NULL_REVISION)
1804
inv = self.get_revision_inventory(revision_id)
1805
return RevisionTree(self, inv, revision_id)
1807
def revision_trees(self, revision_ids):
1808
"""Return Tree for a revision on this branch.
1810
`revision_id` may not be None or 'null:'"""
1811
inventories = self.iter_inventories(revision_ids)
1812
for inv in inventories:
1813
yield RevisionTree(self, inv, inv.revision_id)
1816
def get_ancestry(self, revision_id, topo_sorted=True):
1817
"""Return a list of revision-ids integrated by a revision.
1819
The first element of the list is always None, indicating the origin
1820
revision. This might change when we have history horizons, or
1821
perhaps we should have a new API.
1823
This is topologically sorted.
1825
if _mod_revision.is_null(revision_id):
1827
if not self.has_revision(revision_id):
1828
raise errors.NoSuchRevision(self, revision_id)
1829
graph = self.get_graph()
1831
search = graph._make_breadth_first_searcher([revision_id])
1834
found, ghosts = search.next_with_ghosts()
1835
except StopIteration:
1838
if _mod_revision.NULL_REVISION in keys:
1839
keys.remove(_mod_revision.NULL_REVISION)
1841
parent_map = graph.get_parent_map(keys)
1842
keys = tsort.topo_sort(parent_map)
1843
return [None] + list(keys)
1846
"""Compress the data within the repository.
1848
This operation only makes sense for some repository types. For other
1849
types it should be a no-op that just returns.
1851
This stub method does not require a lock, but subclasses should use
1852
@needs_write_lock as this is a long running call its reasonable to
1853
implicitly lock for the user.
1857
@deprecated_method(one_six)
1858
def print_file(self, file, revision_id):
1859
"""Print `file` to stdout.
1861
FIXME RBC 20060125 as John Meinel points out this is a bad api
1862
- it writes to stdout, it assumes that that is valid etc. Fix
1863
by creating a new more flexible convenience function.
1865
tree = self.revision_tree(revision_id)
1866
# use inventory as it was in that revision
1867
file_id = tree.inventory.path2id(file)
1869
# TODO: jam 20060427 Write a test for this code path
1870
# it had a bug in it, and was raising the wrong
1872
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1873
tree.print_file(file_id)
1875
def get_transaction(self):
1876
return self.control_files.get_transaction()
1878
@deprecated_method(one_one)
1879
def get_parents(self, revision_ids):
1880
"""See StackedParentsProvider.get_parents"""
1881
parent_map = self.get_parent_map(revision_ids)
1882
return [parent_map.get(r, None) for r in revision_ids]
1884
def get_parent_map(self, revision_ids):
1885
"""See graph._StackedParentsProvider.get_parent_map"""
1886
# revisions index works in keys; this just works in revisions
1887
# therefore wrap and unwrap
1890
for revision_id in revision_ids:
1891
if revision_id == _mod_revision.NULL_REVISION:
1892
result[revision_id] = ()
1893
elif revision_id is None:
1894
raise ValueError('get_parent_map(None) is not valid')
1896
query_keys.append((revision_id ,))
1897
for ((revision_id,), parent_keys) in \
1898
self.revisions.get_parent_map(query_keys).iteritems():
1900
result[revision_id] = tuple(parent_revid
1901
for (parent_revid,) in parent_keys)
1903
result[revision_id] = (_mod_revision.NULL_REVISION,)
1906
def _make_parents_provider(self):
1909
def get_graph(self, other_repository=None):
1910
"""Return the graph walker for this repository format"""
1911
parents_provider = self._make_parents_provider()
1912
if (other_repository is not None and
1913
not self.has_same_location(other_repository)):
1914
parents_provider = graph._StackedParentsProvider(
1915
[parents_provider, other_repository._make_parents_provider()])
1916
return graph.Graph(parents_provider)
1918
def _get_versioned_file_checker(self):
1919
"""Return an object suitable for checking versioned files."""
1920
return _VersionedFileChecker(self)
1922
def revision_ids_to_search_result(self, result_set):
1923
"""Convert a set of revision ids to a graph SearchResult."""
1924
result_parents = set()
1925
for parents in self.get_graph().get_parent_map(
1926
result_set).itervalues():
1927
result_parents.update(parents)
1928
included_keys = result_set.intersection(result_parents)
1929
start_keys = result_set.difference(included_keys)
1930
exclude_keys = result_parents.difference(result_set)
1931
result = graph.SearchResult(start_keys, exclude_keys,
1932
len(result_set), result_set)
1936
def set_make_working_trees(self, new_value):
1937
"""Set the policy flag for making working trees when creating branches.
1939
This only applies to branches that use this repository.
1941
The default is 'True'.
1942
:param new_value: True to restore the default, False to disable making
1945
raise NotImplementedError(self.set_make_working_trees)
1947
def make_working_trees(self):
1948
"""Returns the policy for making working trees on new branches."""
1949
raise NotImplementedError(self.make_working_trees)
1952
def sign_revision(self, revision_id, gpg_strategy):
1953
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1954
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1957
def has_signature_for_revision_id(self, revision_id):
1958
"""Query for a revision signature for revision_id in the repository."""
1959
if not self.has_revision(revision_id):
1960
raise errors.NoSuchRevision(self, revision_id)
1961
sig_present = (1 == len(
1962
self.signatures.get_parent_map([(revision_id,)])))
1966
def get_signature_text(self, revision_id):
1967
"""Return the text for a signature."""
1968
stream = self.signatures.get_record_stream([(revision_id,)],
1970
record = stream.next()
1971
if record.storage_kind == 'absent':
1972
raise errors.NoSuchRevision(self, revision_id)
1973
return record.get_bytes_as('fulltext')
1976
def check(self, revision_ids=None):
1977
"""Check consistency of all history of given revision_ids.
1979
Different repository implementations should override _check().
1981
:param revision_ids: A non-empty list of revision_ids whose ancestry
1982
will be checked. Typically the last revision_id of a branch.
1984
return self._check(revision_ids)
1986
def _check(self, revision_ids):
1987
result = check.Check(self)
1991
def _warn_if_deprecated(self):
1992
global _deprecation_warning_done
1993
if _deprecation_warning_done:
1995
_deprecation_warning_done = True
1996
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1997
% (self._format, self.bzrdir.transport.base))
1999
def supports_rich_root(self):
2000
return self._format.rich_root_data
2002
def _check_ascii_revisionid(self, revision_id, method):
2003
"""Private helper for ascii-only repositories."""
2004
# weave repositories refuse to store revisionids that are non-ascii.
2005
if revision_id is not None:
2006
# weaves require ascii revision ids.
2007
if isinstance(revision_id, unicode):
2009
revision_id.encode('ascii')
2010
except UnicodeEncodeError:
2011
raise errors.NonAsciiRevisionId(method, self)
2014
revision_id.decode('ascii')
2015
except UnicodeDecodeError:
2016
raise errors.NonAsciiRevisionId(method, self)
2018
def revision_graph_can_have_wrong_parents(self):
2019
"""Is it possible for this repository to have a revision graph with
2022
If True, then this repository must also implement
2023
_find_inconsistent_revision_parents so that check and reconcile can
2024
check for inconsistencies before proceeding with other checks that may
2025
depend on the revision index being consistent.
2027
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2030
# remove these delegates a while after bzr 0.15
2031
def __make_delegated(name, from_module):
2032
def _deprecated_repository_forwarder():
2033
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2034
% (name, from_module),
2037
m = __import__(from_module, globals(), locals(), [name])
2039
return getattr(m, name)
2040
except AttributeError:
2041
raise AttributeError('module %s has no name %s'
2043
globals()[name] = _deprecated_repository_forwarder
2046
'AllInOneRepository',
2047
'WeaveMetaDirRepository',
2048
'PreSplitOutRepositoryFormat',
2049
'RepositoryFormat4',
2050
'RepositoryFormat5',
2051
'RepositoryFormat6',
2052
'RepositoryFormat7',
2054
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2058
'RepositoryFormatKnit',
2059
'RepositoryFormatKnit1',
2061
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2064
def install_revision(repository, rev, revision_tree):
2065
"""Install all revision data into a repository."""
2066
install_revisions(repository, [(rev, revision_tree, None)])
2069
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2070
"""Install all revision data into a repository.
2072
Accepts an iterable of revision, tree, signature tuples. The signature
2075
repository.start_write_group()
2077
for n, (revision, revision_tree, signature) in enumerate(iterable):
2078
_install_revision(repository, revision, revision_tree, signature)
2080
pb.update('Transferring revisions', n + 1, num_revisions)
2082
repository.abort_write_group()
2085
repository.commit_write_group()
2088
def _install_revision(repository, rev, revision_tree, signature):
2089
"""Install all revision data into a repository."""
2090
present_parents = []
2092
for p_id in rev.parent_ids:
2093
if repository.has_revision(p_id):
2094
present_parents.append(p_id)
2095
parent_trees[p_id] = repository.revision_tree(p_id)
2097
parent_trees[p_id] = repository.revision_tree(
2098
_mod_revision.NULL_REVISION)
2100
inv = revision_tree.inventory
2101
entries = inv.iter_entries()
2102
# backwards compatibility hack: skip the root id.
2103
if not repository.supports_rich_root():
2104
path, root = entries.next()
2105
if root.revision != rev.revision_id:
2106
raise errors.IncompatibleRevision(repr(repository))
2108
for path, ie in entries:
2109
text_keys[(ie.file_id, ie.revision)] = ie
2110
text_parent_map = repository.texts.get_parent_map(text_keys)
2111
missing_texts = set(text_keys) - set(text_parent_map)
2112
# Add the texts that are not already present
2113
for text_key in missing_texts:
2114
ie = text_keys[text_key]
2116
# FIXME: TODO: The following loop overlaps/duplicates that done by
2117
# commit to determine parents. There is a latent/real bug here where
2118
# the parents inserted are not those commit would do - in particular
2119
# they are not filtered by heads(). RBC, AB
2120
for revision, tree in parent_trees.iteritems():
2121
if ie.file_id not in tree:
2123
parent_id = tree.inventory[ie.file_id].revision
2124
if parent_id in text_parents:
2126
text_parents.append((ie.file_id, parent_id))
2127
lines = revision_tree.get_file(ie.file_id).readlines()
2128
repository.texts.add_lines(text_key, text_parents, lines)
2130
# install the inventory
2131
repository.add_inventory(rev.revision_id, inv, present_parents)
2132
except errors.RevisionAlreadyPresent:
2134
if signature is not None:
2135
repository.add_signature_text(rev.revision_id, signature)
2136
repository.add_revision(rev.revision_id, rev, inv)
2139
class MetaDirRepository(Repository):
2140
"""Repositories in the new meta-dir layout.
2142
:ivar _transport: Transport for access to repository control files,
2143
typically pointing to .bzr/repository.
2146
def __init__(self, _format, a_bzrdir, control_files):
2147
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2148
self._transport = control_files._transport
2150
def is_shared(self):
2151
"""Return True if this repository is flagged as a shared repository."""
2152
return self._transport.has('shared-storage')
2155
def set_make_working_trees(self, new_value):
2156
"""Set the policy flag for making working trees when creating branches.
2158
This only applies to branches that use this repository.
2160
The default is 'True'.
2161
:param new_value: True to restore the default, False to disable making
2166
self._transport.delete('no-working-trees')
2167
except errors.NoSuchFile:
2170
self._transport.put_bytes('no-working-trees', '',
2171
mode=self.bzrdir._get_file_mode())
2173
def make_working_trees(self):
2174
"""Returns the policy for making working trees on new branches."""
2175
return not self._transport.has('no-working-trees')
2178
class MetaDirVersionedFileRepository(MetaDirRepository):
2179
"""Repositories in a meta-dir, that work via versioned file objects."""
2181
def __init__(self, _format, a_bzrdir, control_files):
2182
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2186
class RepositoryFormatRegistry(registry.Registry):
2187
"""Registry of RepositoryFormats."""
2189
def get(self, format_string):
2190
r = registry.Registry.get(self, format_string)
2196
format_registry = RepositoryFormatRegistry()
2197
"""Registry of formats, indexed by their identifying format string.
2199
This can contain either format instances themselves, or classes/factories that
2200
can be called to obtain one.
2204
#####################################################################
2205
# Repository Formats
2207
class RepositoryFormat(object):
2208
"""A repository format.
2210
Formats provide three things:
2211
* An initialization routine to construct repository data on disk.
2212
* a format string which is used when the BzrDir supports versioned
2214
* an open routine which returns a Repository instance.
2216
There is one and only one Format subclass for each on-disk format. But
2217
there can be one Repository subclass that is used for several different
2218
formats. The _format attribute on a Repository instance can be used to
2219
determine the disk format.
2221
Formats are placed in an dict by their format string for reference
2222
during opening. These should be subclasses of RepositoryFormat
2225
Once a format is deprecated, just deprecate the initialize and open
2226
methods on the format class. Do not deprecate the object, as the
2227
object will be created every system load.
2229
Common instance attributes:
2230
_matchingbzrdir - the bzrdir format that the repository format was
2231
originally written to work with. This can be used if manually
2232
constructing a bzrdir and repository, or more commonly for test suite
2236
# Set to True or False in derived classes. True indicates that the format
2237
# supports ghosts gracefully.
2238
supports_ghosts = None
2239
# Can this repository be given external locations to lookup additional
2240
# data. Set to True or False in derived classes.
2241
supports_external_lookups = None
2244
return "<%s>" % self.__class__.__name__
2246
def __eq__(self, other):
2247
# format objects are generally stateless
2248
return isinstance(other, self.__class__)
2250
def __ne__(self, other):
2251
return not self == other
2254
def find_format(klass, a_bzrdir):
2255
"""Return the format for the repository object in a_bzrdir.
2257
This is used by bzr native formats that have a "format" file in
2258
the repository. Other methods may be used by different types of
2262
transport = a_bzrdir.get_repository_transport(None)
2263
format_string = transport.get("format").read()
2264
return format_registry.get(format_string)
2265
except errors.NoSuchFile:
2266
raise errors.NoRepositoryPresent(a_bzrdir)
2268
raise errors.UnknownFormatError(format=format_string,
2272
def register_format(klass, format):
2273
format_registry.register(format.get_format_string(), format)
2276
def unregister_format(klass, format):
2277
format_registry.remove(format.get_format_string())
2280
def get_default_format(klass):
2281
"""Return the current default format."""
2282
from bzrlib import bzrdir
2283
return bzrdir.format_registry.make_bzrdir('default').repository_format
2285
def get_format_string(self):
2286
"""Return the ASCII format string that identifies this format.
2288
Note that in pre format ?? repositories the format string is
2289
not permitted nor written to disk.
2291
raise NotImplementedError(self.get_format_string)
2293
def get_format_description(self):
2294
"""Return the short description for this format."""
2295
raise NotImplementedError(self.get_format_description)
2297
# TODO: this shouldn't be in the base class, it's specific to things that
2298
# use weaves or knits -- mbp 20070207
2299
def _get_versioned_file_store(self,
2304
versionedfile_class=None,
2305
versionedfile_kwargs={},
2307
if versionedfile_class is None:
2308
versionedfile_class = self._versionedfile_class
2309
weave_transport = control_files._transport.clone(name)
2310
dir_mode = control_files._dir_mode
2311
file_mode = control_files._file_mode
2312
return VersionedFileStore(weave_transport, prefixed=prefixed,
2314
file_mode=file_mode,
2315
versionedfile_class=versionedfile_class,
2316
versionedfile_kwargs=versionedfile_kwargs,
2319
def initialize(self, a_bzrdir, shared=False):
2320
"""Initialize a repository of this format in a_bzrdir.
2322
:param a_bzrdir: The bzrdir to put the new repository in it.
2323
:param shared: The repository should be initialized as a sharable one.
2324
:returns: The new repository object.
2326
This may raise UninitializableFormat if shared repository are not
2327
compatible the a_bzrdir.
2329
raise NotImplementedError(self.initialize)
2331
def is_supported(self):
2332
"""Is this format supported?
2334
Supported formats must be initializable and openable.
2335
Unsupported formats may not support initialization or committing or
2336
some other features depending on the reason for not being supported.
2340
def check_conversion_target(self, target_format):
2341
raise NotImplementedError(self.check_conversion_target)
2343
def open(self, a_bzrdir, _found=False):
2344
"""Return an instance of this format for the bzrdir a_bzrdir.
2346
_found is a private parameter, do not use it.
2348
raise NotImplementedError(self.open)
2351
class MetaDirRepositoryFormat(RepositoryFormat):
2352
"""Common base class for the new repositories using the metadir layout."""
2354
rich_root_data = False
2355
supports_tree_reference = False
2356
supports_external_lookups = False
2359
def _matchingbzrdir(self):
2360
matching = bzrdir.BzrDirMetaFormat1()
2361
matching.repository_format = self
2365
super(MetaDirRepositoryFormat, self).__init__()
2367
def _create_control_files(self, a_bzrdir):
2368
"""Create the required files and the initial control_files object."""
2369
# FIXME: RBC 20060125 don't peek under the covers
2370
# NB: no need to escape relative paths that are url safe.
2371
repository_transport = a_bzrdir.get_repository_transport(self)
2372
control_files = lockable_files.LockableFiles(repository_transport,
2373
'lock', lockdir.LockDir)
2374
control_files.create_lock()
2375
return control_files
2377
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2378
"""Upload the initial blank content."""
2379
control_files = self._create_control_files(a_bzrdir)
2380
control_files.lock_write()
2381
transport = control_files._transport
2383
utf8_files += [('shared-storage', '')]
2385
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2386
for (filename, content_stream) in files:
2387
transport.put_file(filename, content_stream,
2388
mode=a_bzrdir._get_file_mode())
2389
for (filename, content_bytes) in utf8_files:
2390
transport.put_bytes_non_atomic(filename, content_bytes,
2391
mode=a_bzrdir._get_file_mode())
2393
control_files.unlock()
2396
# formats which have no format string are not discoverable
2397
# and not independently creatable, so are not registered. They're
2398
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2399
# needed, it's constructed directly by the BzrDir. Non-native formats where
2400
# the repository is not separately opened are similar.
2402
format_registry.register_lazy(
2403
'Bazaar-NG Repository format 7',
2404
'bzrlib.repofmt.weaverepo',
2408
format_registry.register_lazy(
2409
'Bazaar-NG Knit Repository Format 1',
2410
'bzrlib.repofmt.knitrepo',
2411
'RepositoryFormatKnit1',
2414
format_registry.register_lazy(
2415
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2416
'bzrlib.repofmt.knitrepo',
2417
'RepositoryFormatKnit3',
2420
format_registry.register_lazy(
2421
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2422
'bzrlib.repofmt.knitrepo',
2423
'RepositoryFormatKnit4',
2426
# Pack-based formats. There is one format for pre-subtrees, and one for
2427
# post-subtrees to allow ease of testing.
2428
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2429
format_registry.register_lazy(
2430
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2431
'bzrlib.repofmt.pack_repo',
2432
'RepositoryFormatKnitPack1',
2434
format_registry.register_lazy(
2435
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2436
'bzrlib.repofmt.pack_repo',
2437
'RepositoryFormatKnitPack3',
2439
format_registry.register_lazy(
2440
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2441
'bzrlib.repofmt.pack_repo',
2442
'RepositoryFormatKnitPack4',
2444
format_registry.register_lazy(
2445
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2446
'bzrlib.repofmt.pack_repo',
2447
'RepositoryFormatKnitPack5',
2449
format_registry.register_lazy(
2450
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2451
'bzrlib.repofmt.pack_repo',
2452
'RepositoryFormatKnitPack5RichRoot',
2454
format_registry.register_lazy(
2455
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2456
'bzrlib.repofmt.pack_repo',
2457
'RepositoryFormatKnitPack5RichRootBroken',
2459
format_registry.register_lazy(
2460
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2461
'bzrlib.repofmt.pack_repo',
2462
'RepositoryFormatKnitPack6',
2464
format_registry.register_lazy(
2465
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2466
'bzrlib.repofmt.pack_repo',
2467
'RepositoryFormatKnitPack6RichRoot',
2470
# Development formats.
2471
# 1.7->1.8 go below here
2472
format_registry.register_lazy(
2473
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2474
'bzrlib.repofmt.pack_repo',
2475
'RepositoryFormatPackDevelopment2',
2477
format_registry.register_lazy(
2478
("Bazaar development format 2 with subtree support "
2479
"(needs bzr.dev from before 1.8)\n"),
2480
'bzrlib.repofmt.pack_repo',
2481
'RepositoryFormatPackDevelopment2Subtree',
2485
class InterRepository(InterObject):
2486
"""This class represents operations taking place between two repositories.
2488
Its instances have methods like copy_content and fetch, and contain
2489
references to the source and target repositories these operations can be
2492
Often we will provide convenience methods on 'repository' which carry out
2493
operations with another repository - they will always forward to
2494
InterRepository.get(other).method_name(parameters).
2497
_walk_to_common_revisions_batch_size = 1
2499
"""The available optimised InterRepository types."""
2501
def __init__(self, source, target):
2502
InterObject.__init__(self, source, target)
2503
# These two attributes may be overridden by e.g. InterOtherToRemote to
2504
# provide a faster implementation.
2505
self.target_get_graph = self.target.get_graph
2506
self.target_get_parent_map = self.target.get_parent_map
2508
def copy_content(self, revision_id=None):
2509
raise NotImplementedError(self.copy_content)
2511
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2512
"""Fetch the content required to construct revision_id.
2514
The content is copied from self.source to self.target.
2516
:param revision_id: if None all content is copied, if NULL_REVISION no
2518
:param pb: optional progress bar to use for progress reports. If not
2519
provided a default one will be created.
2521
:returns: (copied_revision_count, failures).
2523
# Normally we should find a specific InterRepository subclass to do
2524
# the fetch; if nothing else then at least InterSameDataRepository.
2525
# If none of them is suitable it looks like fetching is not possible;
2526
# we try to give a good message why. _assert_same_model will probably
2527
# give a helpful message; otherwise a generic one.
2528
self._assert_same_model(self.source, self.target)
2529
raise errors.IncompatibleRepositories(self.source, self.target,
2530
"no suitableInterRepository found")
2532
def _walk_to_common_revisions(self, revision_ids):
2533
"""Walk out from revision_ids in source to revisions target has.
2535
:param revision_ids: The start point for the search.
2536
:return: A set of revision ids.
2538
target_graph = self.target_get_graph()
2539
revision_ids = frozenset(revision_ids)
2540
# Fast path for the case where all the revisions are already in the
2542
# (Although this does incur an extra round trip for the
2543
# fairly common case where the target doesn't already have the revision
2545
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2546
return graph.SearchResult(revision_ids, set(), 0, set())
2547
missing_revs = set()
2548
source_graph = self.source.get_graph()
2549
# ensure we don't pay silly lookup costs.
2550
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2551
null_set = frozenset([_mod_revision.NULL_REVISION])
2552
searcher_exhausted = False
2556
# Iterate the searcher until we have enough next_revs
2557
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2559
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2560
next_revs.update(next_revs_part)
2561
ghosts.update(ghosts_part)
2562
except StopIteration:
2563
searcher_exhausted = True
2565
# If there are ghosts in the source graph, and the caller asked for
2566
# them, make sure that they are present in the target.
2567
# We don't care about other ghosts as we can't fetch them and
2568
# haven't been asked to.
2569
ghosts_to_check = set(revision_ids.intersection(ghosts))
2570
revs_to_get = set(next_revs).union(ghosts_to_check)
2572
have_revs = set(target_graph.get_parent_map(revs_to_get))
2573
# we always have NULL_REVISION present.
2574
have_revs = have_revs.union(null_set)
2575
# Check if the target is missing any ghosts we need.
2576
ghosts_to_check.difference_update(have_revs)
2578
# One of the caller's revision_ids is a ghost in both the
2579
# source and the target.
2580
raise errors.NoSuchRevision(
2581
self.source, ghosts_to_check.pop())
2582
missing_revs.update(next_revs - have_revs)
2583
# Because we may have walked past the original stop point, make
2584
# sure everything is stopped
2585
stop_revs = searcher.find_seen_ancestors(have_revs)
2586
searcher.stop_searching_any(stop_revs)
2587
if searcher_exhausted:
2589
return searcher.get_result()
2591
@deprecated_method(one_two)
2593
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2594
"""Return the revision ids that source has that target does not.
2596
These are returned in topological order.
2598
:param revision_id: only return revision ids included by this
2600
:param find_ghosts: If True find missing revisions in deep history
2601
rather than just finding the surface difference.
2603
return list(self.search_missing_revision_ids(
2604
revision_id, find_ghosts).get_keys())
2607
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2608
"""Return the revision ids that source has that target does not.
2610
:param revision_id: only return revision ids included by this
2612
:param find_ghosts: If True find missing revisions in deep history
2613
rather than just finding the surface difference.
2614
:return: A bzrlib.graph.SearchResult.
2616
# stop searching at found target revisions.
2617
if not find_ghosts and revision_id is not None:
2618
return self._walk_to_common_revisions([revision_id])
2619
# generic, possibly worst case, slow code path.
2620
target_ids = set(self.target.all_revision_ids())
2621
if revision_id is not None:
2622
source_ids = self.source.get_ancestry(revision_id)
2623
if source_ids[0] is not None:
2624
raise AssertionError()
2627
source_ids = self.source.all_revision_ids()
2628
result_set = set(source_ids).difference(target_ids)
2629
return self.source.revision_ids_to_search_result(result_set)
2632
def _same_model(source, target):
2633
"""True if source and target have the same data representation.
2635
Note: this is always called on the base class; overriding it in a
2636
subclass will have no effect.
2639
InterRepository._assert_same_model(source, target)
2641
except errors.IncompatibleRepositories, e:
2645
def _assert_same_model(source, target):
2646
"""Raise an exception if two repositories do not use the same model.
2648
if source.supports_rich_root() != target.supports_rich_root():
2649
raise errors.IncompatibleRepositories(source, target,
2650
"different rich-root support")
2651
if source._serializer != target._serializer:
2652
raise errors.IncompatibleRepositories(source, target,
2653
"different serializers")
2656
class InterSameDataRepository(InterRepository):
2657
"""Code for converting between repositories that represent the same data.
2659
Data format and model must match for this to work.
2663
def _get_repo_format_to_test(self):
2664
"""Repository format for testing with.
2666
InterSameData can pull from subtree to subtree and from non-subtree to
2667
non-subtree, so we test this with the richest repository format.
2669
from bzrlib.repofmt import knitrepo
2670
return knitrepo.RepositoryFormatKnit3()
2673
def is_compatible(source, target):
2674
return InterRepository._same_model(source, target)
2677
def copy_content(self, revision_id=None):
2678
"""Make a complete copy of the content in self into destination.
2680
This copies both the repository's revision data, and configuration information
2681
such as the make_working_trees setting.
2683
This is a destructive operation! Do not use it on existing
2686
:param revision_id: Only copy the content needed to construct
2687
revision_id and its parents.
2690
self.target.set_make_working_trees(self.source.make_working_trees())
2691
except NotImplementedError:
2693
# but don't bother fetching if we have the needed data now.
2694
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2695
self.target.has_revision(revision_id)):
2697
self.target.fetch(self.source, revision_id=revision_id)
2700
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2701
"""See InterRepository.fetch()."""
2702
from bzrlib.fetch import RepoFetcher
2703
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2704
self.source, self.source._format, self.target,
2705
self.target._format)
2706
f = RepoFetcher(to_repository=self.target,
2707
from_repository=self.source,
2708
last_revision=revision_id,
2709
pb=pb, find_ghosts=find_ghosts)
2710
return f.count_copied, f.failed_revisions
2713
class InterWeaveRepo(InterSameDataRepository):
2714
"""Optimised code paths between Weave based repositories.
2716
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2717
implemented lazy inter-object optimisation.
2721
def _get_repo_format_to_test(self):
2722
from bzrlib.repofmt import weaverepo
2723
return weaverepo.RepositoryFormat7()
2726
def is_compatible(source, target):
2727
"""Be compatible with known Weave formats.
2729
We don't test for the stores being of specific types because that
2730
could lead to confusing results, and there is no need to be
2733
from bzrlib.repofmt.weaverepo import (
2739
return (isinstance(source._format, (RepositoryFormat5,
2741
RepositoryFormat7)) and
2742
isinstance(target._format, (RepositoryFormat5,
2744
RepositoryFormat7)))
2745
except AttributeError:
2749
def copy_content(self, revision_id=None):
2750
"""See InterRepository.copy_content()."""
2751
# weave specific optimised path:
2753
self.target.set_make_working_trees(self.source.make_working_trees())
2754
except (errors.RepositoryUpgradeRequired, NotImplemented):
2756
# FIXME do not peek!
2757
if self.source._transport.listable():
2758
pb = ui.ui_factory.nested_progress_bar()
2760
self.target.texts.insert_record_stream(
2761
self.source.texts.get_record_stream(
2762
self.source.texts.keys(), 'topological', False))
2763
pb.update('copying inventory', 0, 1)
2764
self.target.inventories.insert_record_stream(
2765
self.source.inventories.get_record_stream(
2766
self.source.inventories.keys(), 'topological', False))
2767
self.target.signatures.insert_record_stream(
2768
self.source.signatures.get_record_stream(
2769
self.source.signatures.keys(),
2771
self.target.revisions.insert_record_stream(
2772
self.source.revisions.get_record_stream(
2773
self.source.revisions.keys(),
2774
'topological', True))
2778
self.target.fetch(self.source, revision_id=revision_id)
2781
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2782
"""See InterRepository.fetch()."""
2783
from bzrlib.fetch import RepoFetcher
2784
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2785
self.source, self.source._format, self.target, self.target._format)
2786
f = RepoFetcher(to_repository=self.target,
2787
from_repository=self.source,
2788
last_revision=revision_id,
2789
pb=pb, find_ghosts=find_ghosts)
2790
return f.count_copied, f.failed_revisions
2793
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2794
"""See InterRepository.missing_revision_ids()."""
2795
# we want all revisions to satisfy revision_id in source.
2796
# but we don't want to stat every file here and there.
2797
# we want then, all revisions other needs to satisfy revision_id
2798
# checked, but not those that we have locally.
2799
# so the first thing is to get a subset of the revisions to
2800
# satisfy revision_id in source, and then eliminate those that
2801
# we do already have.
2802
# this is slow on high latency connection to self, but as as this
2803
# disk format scales terribly for push anyway due to rewriting
2804
# inventory.weave, this is considered acceptable.
2806
if revision_id is not None:
2807
source_ids = self.source.get_ancestry(revision_id)
2808
if source_ids[0] is not None:
2809
raise AssertionError()
2812
source_ids = self.source._all_possible_ids()
2813
source_ids_set = set(source_ids)
2814
# source_ids is the worst possible case we may need to pull.
2815
# now we want to filter source_ids against what we actually
2816
# have in target, but don't try to check for existence where we know
2817
# we do not have a revision as that would be pointless.
2818
target_ids = set(self.target._all_possible_ids())
2819
possibly_present_revisions = target_ids.intersection(source_ids_set)
2820
actually_present_revisions = set(
2821
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2822
required_revisions = source_ids_set.difference(actually_present_revisions)
2823
if revision_id is not None:
2824
# we used get_ancestry to determine source_ids then we are assured all
2825
# revisions referenced are present as they are installed in topological order.
2826
# and the tip revision was validated by get_ancestry.
2827
result_set = required_revisions
2829
# if we just grabbed the possibly available ids, then
2830
# we only have an estimate of whats available and need to validate
2831
# that against the revision records.
2833
self.source._eliminate_revisions_not_present(required_revisions))
2834
return self.source.revision_ids_to_search_result(result_set)
2837
class InterKnitRepo(InterSameDataRepository):
2838
"""Optimised code paths between Knit based repositories."""
2841
def _get_repo_format_to_test(self):
2842
from bzrlib.repofmt import knitrepo
2843
return knitrepo.RepositoryFormatKnit1()
2846
def is_compatible(source, target):
2847
"""Be compatible with known Knit formats.
2849
We don't test for the stores being of specific types because that
2850
could lead to confusing results, and there is no need to be
2853
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2855
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2856
isinstance(target._format, RepositoryFormatKnit))
2857
except AttributeError:
2859
return are_knits and InterRepository._same_model(source, target)
2862
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2863
"""See InterRepository.fetch()."""
2864
from bzrlib.fetch import RepoFetcher
2865
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2866
self.source, self.source._format, self.target, self.target._format)
2867
f = RepoFetcher(to_repository=self.target,
2868
from_repository=self.source,
2869
last_revision=revision_id,
2870
pb=pb, find_ghosts=find_ghosts)
2871
return f.count_copied, f.failed_revisions
2874
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2875
"""See InterRepository.missing_revision_ids()."""
2876
if revision_id is not None:
2877
source_ids = self.source.get_ancestry(revision_id)
2878
if source_ids[0] is not None:
2879
raise AssertionError()
2882
source_ids = self.source.all_revision_ids()
2883
source_ids_set = set(source_ids)
2884
# source_ids is the worst possible case we may need to pull.
2885
# now we want to filter source_ids against what we actually
2886
# have in target, but don't try to check for existence where we know
2887
# we do not have a revision as that would be pointless.
2888
target_ids = set(self.target.all_revision_ids())
2889
possibly_present_revisions = target_ids.intersection(source_ids_set)
2890
actually_present_revisions = set(
2891
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2892
required_revisions = source_ids_set.difference(actually_present_revisions)
2893
if revision_id is not None:
2894
# we used get_ancestry to determine source_ids then we are assured all
2895
# revisions referenced are present as they are installed in topological order.
2896
# and the tip revision was validated by get_ancestry.
2897
result_set = required_revisions
2899
# if we just grabbed the possibly available ids, then
2900
# we only have an estimate of whats available and need to validate
2901
# that against the revision records.
2903
self.source._eliminate_revisions_not_present(required_revisions))
2904
return self.source.revision_ids_to_search_result(result_set)
2907
class InterPackRepo(InterSameDataRepository):
2908
"""Optimised code paths between Pack based repositories."""
2911
def _get_repo_format_to_test(self):
2912
from bzrlib.repofmt import pack_repo
2913
return pack_repo.RepositoryFormatKnitPack1()
2916
def is_compatible(source, target):
2917
"""Be compatible with known Pack formats.
2919
We don't test for the stores being of specific types because that
2920
could lead to confusing results, and there is no need to be
2923
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2925
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2926
isinstance(target._format, RepositoryFormatPack))
2927
except AttributeError:
2929
return are_packs and InterRepository._same_model(source, target)
2932
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2933
"""See InterRepository.fetch()."""
2934
if (len(self.source._fallback_repositories) > 0 or
2935
len(self.target._fallback_repositories) > 0):
2936
# The pack layer is not aware of fallback repositories, so when
2937
# fetching from a stacked repository or into a stacked repository
2938
# we use the generic fetch logic which uses the VersionedFiles
2939
# attributes on repository.
2940
from bzrlib.fetch import RepoFetcher
2941
fetcher = RepoFetcher(self.target, self.source, revision_id,
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)):
3180
def _fetch_batch(self, revision_ids, basis_id, basis_tree):
3181
"""Fetch across a few revisions.
3183
:param revision_ids: The revisions to copy
3184
:param basis_id: The revision_id of basis_tree
3185
:param basis_tree: A tree that is not in revision_ids which should
3186
already exist in the target.
3187
:return: (basis_id, basis_tree) A new basis to use now that these trees
3190
# Walk though all revisions; get inventory deltas, copy referenced
3191
# texts that delta references, insert the delta, revision and
3195
pending_revisions = []
3196
for tree in self.source.revision_trees(revision_ids):
3197
current_revision_id = tree.get_revision_id()
3198
delta = tree.inventory._make_delta(basis_tree.inventory)
3199
for old_path, new_path, file_id, entry in delta:
3200
if new_path is not None:
3201
if not (new_path or self.target.supports_rich_root()):
3202
# We leave the inventory delta in, because that
3203
# will have the deserialised inventory root
3207
# "if entry.revision == current_revision_id" ?
3208
if entry.revision == current_revision_id:
3209
text_keys.add((file_id, entry.revision))
3210
revision = self.source.get_revision(current_revision_id)
3211
pending_deltas.append((basis_id, delta,
3212
current_revision_id, revision.parent_ids))
3213
pending_revisions.append(revision)
3214
basis_id = current_revision_id
3217
from_texts = self.source.texts
3218
to_texts = self.target.texts
3219
to_texts.insert_record_stream(from_texts.get_record_stream(
3220
text_keys, self.target._fetch_order,
3221
not self.target._fetch_uses_deltas))
3223
for delta in pending_deltas:
3224
self.target.add_inventory_by_delta(*delta)
3225
# insert signatures and revisions
3226
for revision in pending_revisions:
3228
signature = self.source.get_signature_text(
3229
revision.revision_id)
3230
self.target.add_signature_text(revision.revision_id,
3232
except errors.NoSuchRevision:
3234
self.target.add_revision(revision.revision_id, revision)
3235
return basis_id, basis_tree
3237
def _fetch_all_revisions(self, revision_ids, pb):
3238
"""Fetch everything for the list of revisions.
3240
:param revision_ids: The list of revisions to fetch. Must be in
3242
:param pb: A ProgressBar
3245
basis_id, basis_tree = self._get_basis(revision_ids[0])
3247
for offset in range(0, len(revision_ids), batch_size):
3248
self.target.start_write_group()
3250
pb.update('Transferring revisions', offset,
3252
batch = revision_ids[offset:offset+batch_size]
3253
basis_id, basis_tree = self._fetch_batch(batch,
3254
basis_id, basis_tree)
3256
self.target.abort_write_group()
3259
self.target.commit_write_group()
3260
pb.update('Transferring revisions', len(revision_ids),
3264
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3265
"""See InterRepository.fetch()."""
3266
revision_ids = self.target.search_missing_revision_ids(self.source,
3267
revision_id, find_ghosts=find_ghosts).get_keys()
3268
if not revision_ids:
3270
revision_ids = tsort.topo_sort(
3271
self.source.get_graph().get_parent_map(revision_ids))
3273
my_pb = ui.ui_factory.nested_progress_bar()
3278
self._fetch_all_revisions(revision_ids, pb)
3280
if my_pb is not None:
3282
return len(revision_ids), 0
3284
def _get_basis(self, first_revision_id):
3285
"""Get a revision and tree which exists in the target.
3287
This assumes that first_revision_id is selected for transmission
3288
because all other ancestors are already present. If we can't find an
3289
ancestor we fall back to NULL_REVISION since we know that is safe.
3291
:return: (basis_id, basis_tree)
3293
first_rev = self.source.get_revision(first_revision_id)
3295
basis_id = first_rev.parent_ids[0]
3296
# only valid as a basis if the target has it
3297
self.target.get_revision(basis_id)
3298
# Try to get a basis tree - if its a ghost it will hit the
3299
# NoSuchRevision case.
3300
basis_tree = self.source.revision_tree(basis_id)
3301
except (IndexError, errors.NoSuchRevision):
3302
basis_id = _mod_revision.NULL_REVISION
3303
basis_tree = self.source.revision_tree(basis_id)
3304
return basis_id, basis_tree
3307
class InterOtherToRemote(InterRepository):
3308
"""An InterRepository that simply delegates to the 'real' InterRepository
3309
calculated for (source, target._real_repository).
3312
_walk_to_common_revisions_batch_size = 50
3314
def __init__(self, source, target):
3315
InterRepository.__init__(self, source, target)
3316
self._real_inter = None
3319
def is_compatible(source, target):
3320
if isinstance(target, remote.RemoteRepository):
3324
def _ensure_real_inter(self):
3325
if self._real_inter is None:
3326
self.target._ensure_real()
3327
real_target = self.target._real_repository
3328
self._real_inter = InterRepository.get(self.source, real_target)
3329
# Make _real_inter use the RemoteRepository for get_parent_map
3330
self._real_inter.target_get_graph = self.target.get_graph
3331
self._real_inter.target_get_parent_map = self.target.get_parent_map
3333
def copy_content(self, revision_id=None):
3334
self._ensure_real_inter()
3335
self._real_inter.copy_content(revision_id=revision_id)
3337
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3338
self._ensure_real_inter()
3339
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3340
find_ghosts=find_ghosts)
3343
def _get_repo_format_to_test(self):
3347
class InterRemoteToOther(InterRepository):
3349
def __init__(self, source, target):
3350
InterRepository.__init__(self, source, target)
3351
self._real_inter = None
3354
def is_compatible(source, target):
3355
if not isinstance(source, remote.RemoteRepository):
3357
# Is source's model compatible with target's model?
3358
source._ensure_real()
3359
real_source = source._real_repository
3360
if isinstance(real_source, remote.RemoteRepository):
3361
raise NotImplementedError(
3362
"We don't support remote repos backed by remote repos yet.")
3363
return InterRepository._same_model(real_source, target)
3365
def _ensure_real_inter(self):
3366
if self._real_inter is None:
3367
self.source._ensure_real()
3368
real_source = self.source._real_repository
3369
self._real_inter = InterRepository.get(real_source, self.target)
3371
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3372
self._ensure_real_inter()
3373
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3374
find_ghosts=find_ghosts)
3376
def copy_content(self, revision_id=None):
3377
self._ensure_real_inter()
3378
self._real_inter.copy_content(revision_id=revision_id)
3381
def _get_repo_format_to_test(self):
3386
class InterPackToRemotePack(InterPackRepo):
3387
"""A specialisation of InterPackRepo for a target that is a
3390
This will use the get_parent_map RPC rather than plain readvs, and also
3391
uses an RPC for autopacking.
3394
_walk_to_common_revisions_batch_size = 50
3397
def is_compatible(source, target):
3398
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3399
if isinstance(source._format, RepositoryFormatPack):
3400
if isinstance(target, remote.RemoteRepository):
3401
target._ensure_real()
3402
if isinstance(target._real_repository._format,
3403
RepositoryFormatPack):
3404
if InterRepository._same_model(source, target):
3408
def _autopack(self):
3409
self.target.autopack()
3411
def _get_target_pack_collection(self):
3412
return self.target._real_repository._pack_collection
3415
def _get_repo_format_to_test(self):
3419
InterRepository.register_optimiser(InterDifferingSerializer)
3420
InterRepository.register_optimiser(InterSameDataRepository)
3421
InterRepository.register_optimiser(InterWeaveRepo)
3422
InterRepository.register_optimiser(InterKnitRepo)
3423
InterRepository.register_optimiser(InterModel1and2)
3424
InterRepository.register_optimiser(InterKnit1and2)
3425
InterRepository.register_optimiser(InterPackRepo)
3426
InterRepository.register_optimiser(InterOtherToRemote)
3427
InterRepository.register_optimiser(InterRemoteToOther)
3428
InterRepository.register_optimiser(InterPackToRemotePack)
3431
class CopyConverter(object):
3432
"""A repository conversion tool which just performs a copy of the content.
3434
This is slow but quite reliable.
3437
def __init__(self, target_format):
3438
"""Create a CopyConverter.
3440
:param target_format: The format the resulting repository should be.
3442
self.target_format = target_format
3444
def convert(self, repo, pb):
3445
"""Perform the conversion of to_convert, giving feedback via pb.
3447
:param to_convert: The disk object to convert.
3448
:param pb: a progress bar to use for progress information.
3453
# this is only useful with metadir layouts - separated repo content.
3454
# trigger an assertion if not such
3455
repo._format.get_format_string()
3456
self.repo_dir = repo.bzrdir
3457
self.step('Moving repository to repository.backup')
3458
self.repo_dir.transport.move('repository', 'repository.backup')
3459
backup_transport = self.repo_dir.transport.clone('repository.backup')
3460
repo._format.check_conversion_target(self.target_format)
3461
self.source_repo = repo._format.open(self.repo_dir,
3463
_override_transport=backup_transport)
3464
self.step('Creating new repository')
3465
converted = self.target_format.initialize(self.repo_dir,
3466
self.source_repo.is_shared())
3467
converted.lock_write()
3469
self.step('Copying content into repository.')
3470
self.source_repo.copy_content_into(converted)
3473
self.step('Deleting old repository content.')
3474
self.repo_dir.transport.delete_tree('repository.backup')
3475
self.pb.note('repository converted')
3477
def step(self, message):
3478
"""Update the pb by a step."""
3480
self.pb.update(message, self.count, self.total)
3492
def _unescaper(match, _map=_unescape_map):
3493
code = match.group(1)
3497
if not code.startswith('#'):
3499
return unichr(int(code[1:])).encode('utf8')
3505
def _unescape_xml(data):
3506
"""Unescape predefined XML entities in a string of data."""
3508
if _unescape_re is None:
3509
_unescape_re = re.compile('\&([^;]*);')
3510
return _unescape_re.sub(_unescaper, data)
3513
class _VersionedFileChecker(object):
3515
def __init__(self, repository):
3516
self.repository = repository
3517
self.text_index = self.repository._generate_text_key_index()
3519
def calculate_file_version_parents(self, text_key):
3520
"""Calculate the correct parents for a file version according to
3523
parent_keys = self.text_index[text_key]
3524
if parent_keys == [_mod_revision.NULL_REVISION]:
3526
return tuple(parent_keys)
3528
def check_file_version_parents(self, texts, progress_bar=None):
3529
"""Check the parents stored in a versioned file are correct.
3531
It also detects file versions that are not referenced by their
3532
corresponding revision's inventory.
3534
:returns: A tuple of (wrong_parents, dangling_file_versions).
3535
wrong_parents is a dict mapping {revision_id: (stored_parents,
3536
correct_parents)} for each revision_id where the stored parents
3537
are not correct. dangling_file_versions is a set of (file_id,
3538
revision_id) tuples for versions that are present in this versioned
3539
file, but not used by the corresponding inventory.
3542
self.file_ids = set([file_id for file_id, _ in
3543
self.text_index.iterkeys()])
3544
# text keys is now grouped by file_id
3545
n_weaves = len(self.file_ids)
3546
files_in_revisions = {}
3547
revisions_of_files = {}
3548
n_versions = len(self.text_index)
3549
progress_bar.update('loading text store', 0, n_versions)
3550
parent_map = self.repository.texts.get_parent_map(self.text_index)
3551
# On unlistable transports this could well be empty/error...
3552
text_keys = self.repository.texts.keys()
3553
unused_keys = frozenset(text_keys) - set(self.text_index)
3554
for num, key in enumerate(self.text_index.iterkeys()):
3555
if progress_bar is not None:
3556
progress_bar.update('checking text graph', num, n_versions)
3557
correct_parents = self.calculate_file_version_parents(key)
3559
knit_parents = parent_map[key]
3560
except errors.RevisionNotPresent:
3563
if correct_parents != knit_parents:
3564
wrong_parents[key] = (knit_parents, correct_parents)
3565
return wrong_parents, unused_keys
3568
def _old_get_graph(repository, revision_id):
3569
"""DO NOT USE. That is all. I'm serious."""
3570
graph = repository.get_graph()
3571
revision_graph = dict(((key, value) for key, value in
3572
graph.iter_ancestry([revision_id]) if value is not None))
3573
return _strip_NULL_ghosts(revision_graph)
3576
def _strip_NULL_ghosts(revision_graph):
3577
"""Also don't use this. more compatibility code for unmigrated clients."""
3578
# Filter ghosts, and null:
3579
if _mod_revision.NULL_REVISION in revision_graph:
3580
del revision_graph[_mod_revision.NULL_REVISION]
3581
for key, parents in revision_graph.items():
3582
revision_graph[key] = tuple(parent for parent in parents if parent
3584
return revision_graph
added
removed
bzrlib/repository.py
