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)
640
def _check_fallback_repository(self, repository):
641
"""Check that this repository can fallback to repository safely.
643
Raise an error if not.
645
:param repository: A repository to fallback to.
647
return InterRepository._assert_same_model(self, repository)
649
def add_inventory(self, revision_id, inv, parents):
650
"""Add the inventory inv to the repository as revision_id.
652
:param parents: The revision ids of the parents that revision_id
653
is known to have and are in the repository already.
655
:returns: The validator(which is a sha1 digest, though what is sha'd is
656
repository format specific) of the serialized inventory.
658
if not self.is_in_write_group():
659
raise AssertionError("%r not in write group" % (self,))
660
_mod_revision.check_not_reserved_id(revision_id)
661
if not (inv.revision_id is None or inv.revision_id == revision_id):
662
raise AssertionError(
663
"Mismatch between inventory revision"
664
" id and insertion revid (%r, %r)"
665
% (inv.revision_id, revision_id))
667
raise AssertionError()
668
inv_lines = self._serialise_inventory_to_lines(inv)
669
return self._inventory_add_lines(revision_id, parents,
670
inv_lines, check_content=False)
672
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
674
"""Add a new inventory expressed as a delta against another revision.
676
:param basis_revision_id: The inventory id the delta was created
677
against. (This does not have to be a direct parent.)
678
:param delta: The inventory delta (see Inventory.apply_delta for
680
:param new_revision_id: The revision id that the inventory is being
682
:param parents: The revision ids of the parents that revision_id is
683
known to have and are in the repository already. These are supplied
684
for repositories that depend on the inventory graph for revision
685
graph access, as well as for those that pun ancestry with delta
688
:returns: (validator, new_inv)
689
The validator(which is a sha1 digest, though what is sha'd is
690
repository format specific) of the serialized inventory, and the
693
if not self.is_in_write_group():
694
raise AssertionError("%r not in write group" % (self,))
695
_mod_revision.check_not_reserved_id(new_revision_id)
696
basis_tree = self.revision_tree(basis_revision_id)
697
basis_tree.lock_read()
699
# Note that this mutates the inventory of basis_tree, which not all
700
# inventory implementations may support: A better idiom would be to
701
# return a new inventory, but as there is no revision tree cache in
702
# repository this is safe for now - RBC 20081013
703
basis_inv = basis_tree.inventory
704
basis_inv.apply_delta(delta)
705
basis_inv.revision_id = new_revision_id
706
return (self.add_inventory(new_revision_id, basis_inv, parents),
711
def _inventory_add_lines(self, revision_id, parents, lines,
713
"""Store lines in inv_vf and return the sha1 of the inventory."""
714
parents = [(parent,) for parent in parents]
715
return self.inventories.add_lines((revision_id,), parents, lines,
716
check_content=check_content)[0]
718
def add_revision(self, revision_id, rev, inv=None, config=None):
719
"""Add rev to the revision store as revision_id.
721
:param revision_id: the revision id to use.
722
:param rev: The revision object.
723
:param inv: The inventory for the revision. if None, it will be looked
724
up in the inventory storer
725
:param config: If None no digital signature will be created.
726
If supplied its signature_needed method will be used
727
to determine if a signature should be made.
729
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
731
_mod_revision.check_not_reserved_id(revision_id)
732
if config is not None and config.signature_needed():
734
inv = self.get_inventory(revision_id)
735
plaintext = Testament(rev, inv).as_short_text()
736
self.store_revision_signature(
737
gpg.GPGStrategy(config), plaintext, revision_id)
738
# check inventory present
739
if not self.inventories.get_parent_map([(revision_id,)]):
741
raise errors.WeaveRevisionNotPresent(revision_id,
744
# yes, this is not suitable for adding with ghosts.
745
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
749
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
750
self._add_revision(rev)
752
def _add_revision(self, revision):
753
text = self._serializer.write_revision_to_string(revision)
754
key = (revision.revision_id,)
755
parents = tuple((parent,) for parent in revision.parent_ids)
756
self.revisions.add_lines(key, parents, osutils.split_lines(text))
758
def all_revision_ids(self):
759
"""Returns a list of all the revision ids in the repository.
761
This is conceptually deprecated because code should generally work on
762
the graph reachable from a particular revision, and ignore any other
763
revisions that might be present. There is no direct replacement
766
if 'evil' in debug.debug_flags:
767
mutter_callsite(2, "all_revision_ids is linear with history.")
768
return self._all_revision_ids()
770
def _all_revision_ids(self):
771
"""Returns a list of all the revision ids in the repository.
773
These are in as much topological order as the underlying store can
776
raise NotImplementedError(self._all_revision_ids)
778
def break_lock(self):
779
"""Break a lock if one is present from another instance.
781
Uses the ui factory to ask for confirmation if the lock may be from
784
self.control_files.break_lock()
787
def _eliminate_revisions_not_present(self, revision_ids):
788
"""Check every revision id in revision_ids to see if we have it.
790
Returns a set of the present revisions.
793
graph = self.get_graph()
794
parent_map = graph.get_parent_map(revision_ids)
795
# The old API returned a list, should this actually be a set?
796
return parent_map.keys()
799
def create(a_bzrdir):
800
"""Construct the current default format repository in a_bzrdir."""
801
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
803
def __init__(self, _format, a_bzrdir, control_files):
804
"""instantiate a Repository.
806
:param _format: The format of the repository on disk.
807
:param a_bzrdir: The BzrDir of the repository.
809
In the future we will have a single api for all stores for
810
getting file texts, inventories and revisions, then
811
this construct will accept instances of those things.
813
super(Repository, self).__init__()
814
self._format = _format
815
# the following are part of the public API for Repository:
816
self.bzrdir = a_bzrdir
817
self.control_files = control_files
818
self._transport = control_files._transport
819
self.base = self._transport.base
821
self._reconcile_does_inventory_gc = True
822
self._reconcile_fixes_text_parents = False
823
self._reconcile_backsup_inventory = True
824
# not right yet - should be more semantically clear ?
826
# TODO: make sure to construct the right store classes, etc, depending
827
# on whether escaping is required.
828
self._warn_if_deprecated()
829
self._write_group = None
830
# Additional places to query for data.
831
self._fallback_repositories = []
832
# What order should fetch operations request streams in?
833
# The default is unordered as that is the cheapest for an origin to
835
self._fetch_order = 'unordered'
836
# Does this repository use deltas that can be fetched as-deltas ?
837
# (E.g. knits, where the knit deltas can be transplanted intact.
838
# We default to False, which will ensure that enough data to get
839
# a full text out of any fetch stream will be grabbed.
840
self._fetch_uses_deltas = False
841
# Should fetch trigger a reconcile after the fetch? Only needed for
842
# some repository formats that can suffer internal inconsistencies.
843
self._fetch_reconcile = False
846
return '%s(%r)' % (self.__class__.__name__,
849
def has_same_location(self, other):
850
"""Returns a boolean indicating if this repository is at the same
851
location as another repository.
853
This might return False even when two repository objects are accessing
854
the same physical repository via different URLs.
856
if self.__class__ is not other.__class__:
858
return (self._transport.base == other._transport.base)
860
def is_in_write_group(self):
861
"""Return True if there is an open write group.
863
:seealso: start_write_group.
865
return self._write_group is not None
868
return self.control_files.is_locked()
870
def is_write_locked(self):
871
"""Return True if this object is write locked."""
872
return self.is_locked() and self.control_files._lock_mode == 'w'
874
def lock_write(self, token=None):
875
"""Lock this repository for writing.
877
This causes caching within the repository obejct to start accumlating
878
data during reads, and allows a 'write_group' to be obtained. Write
879
groups must be used for actual data insertion.
881
:param token: if this is already locked, then lock_write will fail
882
unless the token matches the existing lock.
883
:returns: a token if this instance supports tokens, otherwise None.
884
:raises TokenLockingNotSupported: when a token is given but this
885
instance doesn't support using token locks.
886
:raises MismatchedToken: if the specified token doesn't match the token
887
of the existing lock.
888
:seealso: start_write_group.
890
A token should be passed in if you know that you have locked the object
891
some other way, and need to synchronise this object's state with that
894
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
896
result = self.control_files.lock_write(token=token)
897
for repo in self._fallback_repositories:
898
# Writes don't affect fallback repos
904
self.control_files.lock_read()
905
for repo in self._fallback_repositories:
909
def get_physical_lock_status(self):
910
return self.control_files.get_physical_lock_status()
912
def leave_lock_in_place(self):
913
"""Tell this repository not to release the physical lock when this
916
If lock_write doesn't return a token, then this method is not supported.
918
self.control_files.leave_in_place()
920
def dont_leave_lock_in_place(self):
921
"""Tell this repository to release the physical lock when this
922
object is unlocked, even if it didn't originally acquire it.
924
If lock_write doesn't return a token, then this method is not supported.
926
self.control_files.dont_leave_in_place()
929
def gather_stats(self, revid=None, committers=None):
930
"""Gather statistics from a revision id.
932
:param revid: The revision id to gather statistics from, if None, then
933
no revision specific statistics are gathered.
934
:param committers: Optional parameter controlling whether to grab
935
a count of committers from the revision specific statistics.
936
:return: A dictionary of statistics. Currently this contains:
937
committers: The number of committers if requested.
938
firstrev: A tuple with timestamp, timezone for the penultimate left
939
most ancestor of revid, if revid is not the NULL_REVISION.
940
latestrev: A tuple with timestamp, timezone for revid, if revid is
941
not the NULL_REVISION.
942
revisions: The total revision count in the repository.
943
size: An estimate disk size of the repository in bytes.
946
if revid and committers:
947
result['committers'] = 0
948
if revid and revid != _mod_revision.NULL_REVISION:
950
all_committers = set()
951
revisions = self.get_ancestry(revid)
952
# pop the leading None
954
first_revision = None
956
# ignore the revisions in the middle - just grab first and last
957
revisions = revisions[0], revisions[-1]
958
for revision in self.get_revisions(revisions):
959
if not first_revision:
960
first_revision = revision
962
all_committers.add(revision.committer)
963
last_revision = revision
965
result['committers'] = len(all_committers)
966
result['firstrev'] = (first_revision.timestamp,
967
first_revision.timezone)
968
result['latestrev'] = (last_revision.timestamp,
969
last_revision.timezone)
971
# now gather global repository information
972
# XXX: This is available for many repos regardless of listability.
973
if self.bzrdir.root_transport.listable():
974
# XXX: do we want to __define len__() ?
975
# Maybe the versionedfiles object should provide a different
976
# method to get the number of keys.
977
result['revisions'] = len(self.revisions.keys())
981
def find_branches(self, using=False):
982
"""Find branches underneath this repository.
984
This will include branches inside other branches.
986
:param using: If True, list only branches using this repository.
988
if using and not self.is_shared():
990
return [self.bzrdir.open_branch()]
991
except errors.NotBranchError:
993
class Evaluator(object):
996
self.first_call = True
998
def __call__(self, bzrdir):
999
# On the first call, the parameter is always the bzrdir
1000
# containing the current repo.
1001
if not self.first_call:
1003
repository = bzrdir.open_repository()
1004
except errors.NoRepositoryPresent:
1007
return False, (None, repository)
1008
self.first_call = False
1010
value = (bzrdir.open_branch(), None)
1011
except errors.NotBranchError:
1012
value = (None, None)
1016
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1017
self.bzrdir.root_transport, evaluate=Evaluator()):
1018
if branch is not None:
1019
branches.append(branch)
1020
if not using and repository is not None:
1021
branches.extend(repository.find_branches())
1025
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1026
"""Return the revision ids that other has that this does not.
1028
These are returned in topological order.
1030
revision_id: only return revision ids included by revision_id.
1032
return InterRepository.get(other, self).search_missing_revision_ids(
1033
revision_id, find_ghosts)
1035
@deprecated_method(one_two)
1037
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1038
"""Return the revision ids that other has that this does not.
1040
These are returned in topological order.
1042
revision_id: only return revision ids included by revision_id.
1044
keys = self.search_missing_revision_ids(
1045
other, revision_id, find_ghosts).get_keys()
1048
parents = other.get_graph().get_parent_map(keys)
1051
return tsort.topo_sort(parents)
1055
"""Open the repository rooted at base.
1057
For instance, if the repository is at URL/.bzr/repository,
1058
Repository.open(URL) -> a Repository instance.
1060
control = bzrdir.BzrDir.open(base)
1061
return control.open_repository()
1063
def copy_content_into(self, destination, revision_id=None):
1064
"""Make a complete copy of the content in self into destination.
1066
This is a destructive operation! Do not use it on existing
1069
return InterRepository.get(self, destination).copy_content(revision_id)
1071
def commit_write_group(self):
1072
"""Commit the contents accrued within the current write group.
1074
:seealso: start_write_group.
1076
if self._write_group is not self.get_transaction():
1077
# has an unlock or relock occured ?
1078
raise errors.BzrError('mismatched lock context %r and '
1080
(self.get_transaction(), self._write_group))
1081
self._commit_write_group()
1082
self._write_group = None
1084
def _commit_write_group(self):
1085
"""Template method for per-repository write group cleanup.
1087
This is called before the write group is considered to be
1088
finished and should ensure that all data handed to the repository
1089
for writing during the write group is safely committed (to the
1090
extent possible considering file system caching etc).
1093
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1094
"""Fetch the content required to construct revision_id from source.
1096
If revision_id is None all content is copied.
1097
:param find_ghosts: Find and copy revisions in the source that are
1098
ghosts in the target (and not reachable directly by walking out to
1099
the first-present revision in target from revision_id).
1101
# fast path same-url fetch operations
1102
if self.has_same_location(source):
1103
# check that last_revision is in 'from' and then return a
1105
if (revision_id is not None and
1106
not _mod_revision.is_null(revision_id)):
1107
self.get_revision(revision_id)
1109
# if there is no specific appropriate InterRepository, this will get
1110
# the InterRepository base class, which raises an
1111
# IncompatibleRepositories when asked to fetch.
1112
inter = InterRepository.get(source, self)
1113
return inter.fetch(revision_id=revision_id, pb=pb,
1114
find_ghosts=find_ghosts)
1116
def create_bundle(self, target, base, fileobj, format=None):
1117
return serializer.write_bundle(self, target, base, fileobj, format)
1119
def get_commit_builder(self, branch, parents, config, timestamp=None,
1120
timezone=None, committer=None, revprops=None,
1122
"""Obtain a CommitBuilder for this repository.
1124
:param branch: Branch to commit to.
1125
:param parents: Revision ids of the parents of the new revision.
1126
:param config: Configuration to use.
1127
:param timestamp: Optional timestamp recorded for commit.
1128
:param timezone: Optional timezone for timestamp.
1129
:param committer: Optional committer to set for commit.
1130
:param revprops: Optional dictionary of revision properties.
1131
:param revision_id: Optional revision id.
1133
result = self._commit_builder_class(self, parents, config,
1134
timestamp, timezone, committer, revprops, revision_id)
1135
self.start_write_group()
1139
if (self.control_files._lock_count == 1 and
1140
self.control_files._lock_mode == 'w'):
1141
if self._write_group is not None:
1142
self.abort_write_group()
1143
self.control_files.unlock()
1144
raise errors.BzrError(
1145
'Must end write groups before releasing write locks.')
1146
self.control_files.unlock()
1147
for repo in self._fallback_repositories:
1151
def clone(self, a_bzrdir, revision_id=None):
1152
"""Clone this repository into a_bzrdir using the current format.
1154
Currently no check is made that the format of this repository and
1155
the bzrdir format are compatible. FIXME RBC 20060201.
1157
:return: The newly created destination repository.
1159
# TODO: deprecate after 0.16; cloning this with all its settings is
1160
# probably not very useful -- mbp 20070423
1161
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1162
self.copy_content_into(dest_repo, revision_id)
1165
def start_write_group(self):
1166
"""Start a write group in the repository.
1168
Write groups are used by repositories which do not have a 1:1 mapping
1169
between file ids and backend store to manage the insertion of data from
1170
both fetch and commit operations.
1172
A write lock is required around the start_write_group/commit_write_group
1173
for the support of lock-requiring repository formats.
1175
One can only insert data into a repository inside a write group.
1179
if not self.is_write_locked():
1180
raise errors.NotWriteLocked(self)
1181
if self._write_group:
1182
raise errors.BzrError('already in a write group')
1183
self._start_write_group()
1184
# so we can detect unlock/relock - the write group is now entered.
1185
self._write_group = self.get_transaction()
1187
def _start_write_group(self):
1188
"""Template method for per-repository write group startup.
1190
This is called before the write group is considered to be
1195
def sprout(self, to_bzrdir, revision_id=None):
1196
"""Create a descendent repository for new development.
1198
Unlike clone, this does not copy the settings of the repository.
1200
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1201
dest_repo.fetch(self, revision_id=revision_id)
1204
def _create_sprouting_repo(self, a_bzrdir, shared):
1205
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1206
# use target default format.
1207
dest_repo = a_bzrdir.create_repository()
1209
# Most control formats need the repository to be specifically
1210
# created, but on some old all-in-one formats it's not needed
1212
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1213
except errors.UninitializableFormat:
1214
dest_repo = a_bzrdir.open_repository()
1218
def has_revision(self, revision_id):
1219
"""True if this repository has a copy of the revision."""
1220
return revision_id in self.has_revisions((revision_id,))
1223
def has_revisions(self, revision_ids):
1224
"""Probe to find out the presence of multiple revisions.
1226
:param revision_ids: An iterable of revision_ids.
1227
:return: A set of the revision_ids that were present.
1229
parent_map = self.revisions.get_parent_map(
1230
[(rev_id,) for rev_id in revision_ids])
1232
if _mod_revision.NULL_REVISION in revision_ids:
1233
result.add(_mod_revision.NULL_REVISION)
1234
result.update([key[0] for key in parent_map])
1238
def get_revision(self, revision_id):
1239
"""Return the Revision object for a named revision."""
1240
return self.get_revisions([revision_id])[0]
1243
def get_revision_reconcile(self, revision_id):
1244
"""'reconcile' helper routine that allows access to a revision always.
1246
This variant of get_revision does not cross check the weave graph
1247
against the revision one as get_revision does: but it should only
1248
be used by reconcile, or reconcile-alike commands that are correcting
1249
or testing the revision graph.
1251
return self._get_revisions([revision_id])[0]
1254
def get_revisions(self, revision_ids):
1255
"""Get many revisions at once."""
1256
return self._get_revisions(revision_ids)
1259
def _get_revisions(self, revision_ids):
1260
"""Core work logic to get many revisions without sanity checks."""
1261
for rev_id in revision_ids:
1262
if not rev_id or not isinstance(rev_id, basestring):
1263
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1264
keys = [(key,) for key in revision_ids]
1265
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1267
for record in stream:
1268
if record.storage_kind == 'absent':
1269
raise errors.NoSuchRevision(self, record.key[0])
1270
text = record.get_bytes_as('fulltext')
1271
rev = self._serializer.read_revision_from_string(text)
1272
revs[record.key[0]] = rev
1273
return [revs[revid] for revid in revision_ids]
1276
def get_revision_xml(self, revision_id):
1277
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1278
# would have already do it.
1279
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1280
rev = self.get_revision(revision_id)
1281
rev_tmp = cStringIO.StringIO()
1282
# the current serializer..
1283
self._serializer.write_revision(rev, rev_tmp)
1285
return rev_tmp.getvalue()
1287
def get_deltas_for_revisions(self, revisions):
1288
"""Produce a generator of revision deltas.
1290
Note that the input is a sequence of REVISIONS, not revision_ids.
1291
Trees will be held in memory until the generator exits.
1292
Each delta is relative to the revision's lefthand predecessor.
1294
required_trees = set()
1295
for revision in revisions:
1296
required_trees.add(revision.revision_id)
1297
required_trees.update(revision.parent_ids[:1])
1298
trees = dict((t.get_revision_id(), t) for
1299
t in self.revision_trees(required_trees))
1300
for revision in revisions:
1301
if not revision.parent_ids:
1302
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1304
old_tree = trees[revision.parent_ids[0]]
1305
yield trees[revision.revision_id].changes_from(old_tree)
1308
def get_revision_delta(self, revision_id):
1309
"""Return the delta for one revision.
1311
The delta is relative to the left-hand predecessor of the
1314
r = self.get_revision(revision_id)
1315
return list(self.get_deltas_for_revisions([r]))[0]
1318
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1319
signature = gpg_strategy.sign(plaintext)
1320
self.add_signature_text(revision_id, signature)
1323
def add_signature_text(self, revision_id, signature):
1324
self.signatures.add_lines((revision_id,), (),
1325
osutils.split_lines(signature))
1327
def find_text_key_references(self):
1328
"""Find the text key references within the repository.
1330
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1331
revision_ids. Each altered file-ids has the exact revision_ids that
1332
altered it listed explicitly.
1333
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1334
to whether they were referred to by the inventory of the
1335
revision_id that they contain. The inventory texts from all present
1336
revision ids are assessed to generate this report.
1338
revision_keys = self.revisions.keys()
1339
w = self.inventories
1340
pb = ui.ui_factory.nested_progress_bar()
1342
return self._find_text_key_references_from_xml_inventory_lines(
1343
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1347
def _find_text_key_references_from_xml_inventory_lines(self,
1349
"""Core routine for extracting references to texts from inventories.
1351
This performs the translation of xml lines to revision ids.
1353
:param line_iterator: An iterator of lines, origin_version_id
1354
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1355
to whether they were referred to by the inventory of the
1356
revision_id that they contain. Note that if that revision_id was
1357
not part of the line_iterator's output then False will be given -
1358
even though it may actually refer to that key.
1360
if not self._serializer.support_altered_by_hack:
1361
raise AssertionError(
1362
"_find_text_key_references_from_xml_inventory_lines only "
1363
"supported for branches which store inventory as unnested xml"
1364
", not on %r" % self)
1367
# this code needs to read every new line in every inventory for the
1368
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1369
# not present in one of those inventories is unnecessary but not
1370
# harmful because we are filtering by the revision id marker in the
1371
# inventory lines : we only select file ids altered in one of those
1372
# revisions. We don't need to see all lines in the inventory because
1373
# only those added in an inventory in rev X can contain a revision=X
1375
unescape_revid_cache = {}
1376
unescape_fileid_cache = {}
1378
# jam 20061218 In a big fetch, this handles hundreds of thousands
1379
# of lines, so it has had a lot of inlining and optimizing done.
1380
# Sorry that it is a little bit messy.
1381
# Move several functions to be local variables, since this is a long
1383
search = self._file_ids_altered_regex.search
1384
unescape = _unescape_xml
1385
setdefault = result.setdefault
1386
for line, line_key in line_iterator:
1387
match = search(line)
1390
# One call to match.group() returning multiple items is quite a
1391
# bit faster than 2 calls to match.group() each returning 1
1392
file_id, revision_id = match.group('file_id', 'revision_id')
1394
# Inlining the cache lookups helps a lot when you make 170,000
1395
# lines and 350k ids, versus 8.4 unique ids.
1396
# Using a cache helps in 2 ways:
1397
# 1) Avoids unnecessary decoding calls
1398
# 2) Re-uses cached strings, which helps in future set and
1400
# (2) is enough that removing encoding entirely along with
1401
# the cache (so we are using plain strings) results in no
1402
# performance improvement.
1404
revision_id = unescape_revid_cache[revision_id]
1406
unescaped = unescape(revision_id)
1407
unescape_revid_cache[revision_id] = unescaped
1408
revision_id = unescaped
1410
# Note that unconditionally unescaping means that we deserialise
1411
# every fileid, which for general 'pull' is not great, but we don't
1412
# really want to have some many fulltexts that this matters anyway.
1415
file_id = unescape_fileid_cache[file_id]
1417
unescaped = unescape(file_id)
1418
unescape_fileid_cache[file_id] = unescaped
1421
key = (file_id, revision_id)
1422
setdefault(key, False)
1423
if revision_id == line_key[-1]:
1427
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1429
"""Helper routine for fileids_altered_by_revision_ids.
1431
This performs the translation of xml lines to revision ids.
1433
:param line_iterator: An iterator of lines, origin_version_id
1434
:param revision_ids: The revision ids to filter for. This should be a
1435
set or other type which supports efficient __contains__ lookups, as
1436
the revision id from each parsed line will be looked up in the
1437
revision_ids filter.
1438
:return: a dictionary mapping altered file-ids to an iterable of
1439
revision_ids. Each altered file-ids has the exact revision_ids that
1440
altered it listed explicitly.
1443
setdefault = result.setdefault
1445
self._find_text_key_references_from_xml_inventory_lines(
1446
line_iterator).iterkeys():
1447
# once data is all ensured-consistent; then this is
1448
# if revision_id == version_id
1449
if key[-1:] in revision_ids:
1450
setdefault(key[0], set()).add(key[-1])
1453
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1454
"""Find the file ids and versions affected by revisions.
1456
:param revisions: an iterable containing revision ids.
1457
:param _inv_weave: The inventory weave from this repository or None.
1458
If None, the inventory weave will be opened automatically.
1459
:return: a dictionary mapping altered file-ids to an iterable of
1460
revision_ids. Each altered file-ids has the exact revision_ids that
1461
altered it listed explicitly.
1463
selected_keys = set((revid,) for revid in revision_ids)
1464
w = _inv_weave or self.inventories
1465
pb = ui.ui_factory.nested_progress_bar()
1467
return self._find_file_ids_from_xml_inventory_lines(
1468
w.iter_lines_added_or_present_in_keys(
1469
selected_keys, pb=pb),
1474
def iter_files_bytes(self, desired_files):
1475
"""Iterate through file versions.
1477
Files will not necessarily be returned in the order they occur in
1478
desired_files. No specific order is guaranteed.
1480
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1481
value supplied by the caller as part of desired_files. It should
1482
uniquely identify the file version in the caller's context. (Examples:
1483
an index number or a TreeTransform trans_id.)
1485
bytes_iterator is an iterable of bytestrings for the file. The
1486
kind of iterable and length of the bytestrings are unspecified, but for
1487
this implementation, it is a list of bytes produced by
1488
VersionedFile.get_record_stream().
1490
:param desired_files: a list of (file_id, revision_id, identifier)
1493
transaction = self.get_transaction()
1495
for file_id, revision_id, callable_data in desired_files:
1496
text_keys[(file_id, revision_id)] = callable_data
1497
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1498
if record.storage_kind == 'absent':
1499
raise errors.RevisionNotPresent(record.key, self)
1500
yield text_keys[record.key], record.get_bytes_as('fulltext')
1502
def _generate_text_key_index(self, text_key_references=None,
1504
"""Generate a new text key index for the repository.
1506
This is an expensive function that will take considerable time to run.
1508
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1509
list of parents, also text keys. When a given key has no parents,
1510
the parents list will be [NULL_REVISION].
1512
# All revisions, to find inventory parents.
1513
if ancestors is None:
1514
graph = self.get_graph()
1515
ancestors = graph.get_parent_map(self.all_revision_ids())
1516
if text_key_references is None:
1517
text_key_references = self.find_text_key_references()
1518
pb = ui.ui_factory.nested_progress_bar()
1520
return self._do_generate_text_key_index(ancestors,
1521
text_key_references, pb)
1525
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1526
"""Helper for _generate_text_key_index to avoid deep nesting."""
1527
revision_order = tsort.topo_sort(ancestors)
1528
invalid_keys = set()
1530
for revision_id in revision_order:
1531
revision_keys[revision_id] = set()
1532
text_count = len(text_key_references)
1533
# a cache of the text keys to allow reuse; costs a dict of all the
1534
# keys, but saves a 2-tuple for every child of a given key.
1536
for text_key, valid in text_key_references.iteritems():
1538
invalid_keys.add(text_key)
1540
revision_keys[text_key[1]].add(text_key)
1541
text_key_cache[text_key] = text_key
1542
del text_key_references
1544
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1545
NULL_REVISION = _mod_revision.NULL_REVISION
1546
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1547
# too small for large or very branchy trees. However, for 55K path
1548
# trees, it would be easy to use too much memory trivially. Ideally we
1549
# could gauge this by looking at available real memory etc, but this is
1550
# always a tricky proposition.
1551
inventory_cache = lru_cache.LRUCache(10)
1552
batch_size = 10 # should be ~150MB on a 55K path tree
1553
batch_count = len(revision_order) / batch_size + 1
1555
pb.update("Calculating text parents.", processed_texts, text_count)
1556
for offset in xrange(batch_count):
1557
to_query = revision_order[offset * batch_size:(offset + 1) *
1561
for rev_tree in self.revision_trees(to_query):
1562
revision_id = rev_tree.get_revision_id()
1563
parent_ids = ancestors[revision_id]
1564
for text_key in revision_keys[revision_id]:
1565
pb.update("Calculating text parents.", processed_texts)
1566
processed_texts += 1
1567
candidate_parents = []
1568
for parent_id in parent_ids:
1569
parent_text_key = (text_key[0], parent_id)
1571
check_parent = parent_text_key not in \
1572
revision_keys[parent_id]
1574
# the parent parent_id is a ghost:
1575
check_parent = False
1576
# truncate the derived graph against this ghost.
1577
parent_text_key = None
1579
# look at the parent commit details inventories to
1580
# determine possible candidates in the per file graph.
1583
inv = inventory_cache[parent_id]
1585
inv = self.revision_tree(parent_id).inventory
1586
inventory_cache[parent_id] = inv
1587
parent_entry = inv._byid.get(text_key[0], None)
1588
if parent_entry is not None:
1590
text_key[0], parent_entry.revision)
1592
parent_text_key = None
1593
if parent_text_key is not None:
1594
candidate_parents.append(
1595
text_key_cache[parent_text_key])
1596
parent_heads = text_graph.heads(candidate_parents)
1597
new_parents = list(parent_heads)
1598
new_parents.sort(key=lambda x:candidate_parents.index(x))
1599
if new_parents == []:
1600
new_parents = [NULL_REVISION]
1601
text_index[text_key] = new_parents
1603
for text_key in invalid_keys:
1604
text_index[text_key] = [NULL_REVISION]
1607
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1608
"""Get an iterable listing the keys of all the data introduced by a set
1611
The keys will be ordered so that the corresponding items can be safely
1612
fetched and inserted in that order.
1614
:returns: An iterable producing tuples of (knit-kind, file-id,
1615
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1616
'revisions'. file-id is None unless knit-kind is 'file'.
1618
# XXX: it's a bit weird to control the inventory weave caching in this
1619
# generator. Ideally the caching would be done in fetch.py I think. Or
1620
# maybe this generator should explicitly have the contract that it
1621
# should not be iterated until the previously yielded item has been
1623
inv_w = self.inventories
1625
# file ids that changed
1626
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1628
num_file_ids = len(file_ids)
1629
for file_id, altered_versions in file_ids.iteritems():
1630
if _files_pb is not None:
1631
_files_pb.update("fetch texts", count, num_file_ids)
1633
yield ("file", file_id, altered_versions)
1634
# We're done with the files_pb. Note that it finished by the caller,
1635
# just as it was created by the caller.
1639
yield ("inventory", None, revision_ids)
1642
# XXX: Note ATM no callers actually pay attention to this return
1643
# instead they just use the list of revision ids and ignore
1644
# missing sigs. Consider removing this work entirely
1645
revisions_with_signatures = set(self.signatures.get_parent_map(
1646
[(r,) for r in revision_ids]))
1647
revisions_with_signatures = set(
1648
[r for (r,) in revisions_with_signatures])
1649
revisions_with_signatures.intersection_update(revision_ids)
1650
yield ("signatures", None, revisions_with_signatures)
1653
yield ("revisions", None, revision_ids)
1656
def get_inventory(self, revision_id):
1657
"""Get Inventory object by revision id."""
1658
return self.iter_inventories([revision_id]).next()
1660
def iter_inventories(self, revision_ids):
1661
"""Get many inventories by revision_ids.
1663
This will buffer some or all of the texts used in constructing the
1664
inventories in memory, but will only parse a single inventory at a
1667
:return: An iterator of inventories.
1669
if ((None in revision_ids)
1670
or (_mod_revision.NULL_REVISION in revision_ids)):
1671
raise ValueError('cannot get null revision inventory')
1672
return self._iter_inventories(revision_ids)
1674
def _iter_inventories(self, revision_ids):
1675
"""single-document based inventory iteration."""
1676
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1677
yield self.deserialise_inventory(revision_id, text)
1679
def _iter_inventory_xmls(self, revision_ids):
1680
keys = [(revision_id,) for revision_id in revision_ids]
1681
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1683
for record in stream:
1684
if record.storage_kind != 'absent':
1685
texts[record.key] = record.get_bytes_as('fulltext')
1687
raise errors.NoSuchRevision(self, record.key)
1689
yield texts[key], key[-1]
1691
def deserialise_inventory(self, revision_id, xml):
1692
"""Transform the xml into an inventory object.
1694
:param revision_id: The expected revision id of the inventory.
1695
:param xml: A serialised inventory.
1697
result = self._serializer.read_inventory_from_string(xml, revision_id)
1698
if result.revision_id != revision_id:
1699
raise AssertionError('revision id mismatch %s != %s' % (
1700
result.revision_id, revision_id))
1703
def serialise_inventory(self, inv):
1704
return self._serializer.write_inventory_to_string(inv)
1706
def _serialise_inventory_to_lines(self, inv):
1707
return self._serializer.write_inventory_to_lines(inv)
1709
def get_serializer_format(self):
1710
return self._serializer.format_num
1713
def get_inventory_xml(self, revision_id):
1714
"""Get inventory XML as a file object."""
1715
texts = self._iter_inventory_xmls([revision_id])
1717
text, revision_id = texts.next()
1718
except StopIteration:
1719
raise errors.HistoryMissing(self, 'inventory', revision_id)
1723
def get_inventory_sha1(self, revision_id):
1724
"""Return the sha1 hash of the inventory entry
1726
return self.get_revision(revision_id).inventory_sha1
1728
def iter_reverse_revision_history(self, revision_id):
1729
"""Iterate backwards through revision ids in the lefthand history
1731
:param revision_id: The revision id to start with. All its lefthand
1732
ancestors will be traversed.
1734
graph = self.get_graph()
1735
next_id = revision_id
1737
if next_id in (None, _mod_revision.NULL_REVISION):
1740
# Note: The following line may raise KeyError in the event of
1741
# truncated history. We decided not to have a try:except:raise
1742
# RevisionNotPresent here until we see a use for it, because of the
1743
# cost in an inner loop that is by its very nature O(history).
1744
# Robert Collins 20080326
1745
parents = graph.get_parent_map([next_id])[next_id]
1746
if len(parents) == 0:
1749
next_id = parents[0]
1752
def get_revision_inventory(self, revision_id):
1753
"""Return inventory of a past revision."""
1754
# TODO: Unify this with get_inventory()
1755
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1756
# must be the same as its revision, so this is trivial.
1757
if revision_id is None:
1758
# This does not make sense: if there is no revision,
1759
# then it is the current tree inventory surely ?!
1760
# and thus get_root_id() is something that looks at the last
1761
# commit on the branch, and the get_root_id is an inventory check.
1762
raise NotImplementedError
1763
# return Inventory(self.get_root_id())
1765
return self.get_inventory(revision_id)
1767
def is_shared(self):
1768
"""Return True if this repository is flagged as a shared repository."""
1769
raise NotImplementedError(self.is_shared)
1772
def reconcile(self, other=None, thorough=False):
1773
"""Reconcile this repository."""
1774
from bzrlib.reconcile import RepoReconciler
1775
reconciler = RepoReconciler(self, thorough=thorough)
1776
reconciler.reconcile()
1779
def _refresh_data(self):
1780
"""Helper called from lock_* to ensure coherency with disk.
1782
The default implementation does nothing; it is however possible
1783
for repositories to maintain loaded indices across multiple locks
1784
by checking inside their implementation of this method to see
1785
whether their indices are still valid. This depends of course on
1786
the disk format being validatable in this manner.
1790
def revision_tree(self, revision_id):
1791
"""Return Tree for a revision on this branch.
1793
`revision_id` may be NULL_REVISION for the empty tree revision.
1795
revision_id = _mod_revision.ensure_null(revision_id)
1796
# TODO: refactor this to use an existing revision object
1797
# so we don't need to read it in twice.
1798
if revision_id == _mod_revision.NULL_REVISION:
1799
return RevisionTree(self, Inventory(root_id=None),
1800
_mod_revision.NULL_REVISION)
1802
inv = self.get_revision_inventory(revision_id)
1803
return RevisionTree(self, inv, revision_id)
1805
def revision_trees(self, revision_ids):
1806
"""Return Tree for a revision on this branch.
1808
`revision_id` may not be None or 'null:'"""
1809
inventories = self.iter_inventories(revision_ids)
1810
for inv in inventories:
1811
yield RevisionTree(self, inv, inv.revision_id)
1814
def get_ancestry(self, revision_id, topo_sorted=True):
1815
"""Return a list of revision-ids integrated by a revision.
1817
The first element of the list is always None, indicating the origin
1818
revision. This might change when we have history horizons, or
1819
perhaps we should have a new API.
1821
This is topologically sorted.
1823
if _mod_revision.is_null(revision_id):
1825
if not self.has_revision(revision_id):
1826
raise errors.NoSuchRevision(self, revision_id)
1827
graph = self.get_graph()
1829
search = graph._make_breadth_first_searcher([revision_id])
1832
found, ghosts = search.next_with_ghosts()
1833
except StopIteration:
1836
if _mod_revision.NULL_REVISION in keys:
1837
keys.remove(_mod_revision.NULL_REVISION)
1839
parent_map = graph.get_parent_map(keys)
1840
keys = tsort.topo_sort(parent_map)
1841
return [None] + list(keys)
1844
"""Compress the data within the repository.
1846
This operation only makes sense for some repository types. For other
1847
types it should be a no-op that just returns.
1849
This stub method does not require a lock, but subclasses should use
1850
@needs_write_lock as this is a long running call its reasonable to
1851
implicitly lock for the user.
1855
@deprecated_method(one_six)
1856
def print_file(self, file, revision_id):
1857
"""Print `file` to stdout.
1859
FIXME RBC 20060125 as John Meinel points out this is a bad api
1860
- it writes to stdout, it assumes that that is valid etc. Fix
1861
by creating a new more flexible convenience function.
1863
tree = self.revision_tree(revision_id)
1864
# use inventory as it was in that revision
1865
file_id = tree.inventory.path2id(file)
1867
# TODO: jam 20060427 Write a test for this code path
1868
# it had a bug in it, and was raising the wrong
1870
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1871
tree.print_file(file_id)
1873
def get_transaction(self):
1874
return self.control_files.get_transaction()
1876
@deprecated_method(one_one)
1877
def get_parents(self, revision_ids):
1878
"""See StackedParentsProvider.get_parents"""
1879
parent_map = self.get_parent_map(revision_ids)
1880
return [parent_map.get(r, None) for r in revision_ids]
1882
def get_parent_map(self, revision_ids):
1883
"""See graph._StackedParentsProvider.get_parent_map"""
1884
# revisions index works in keys; this just works in revisions
1885
# therefore wrap and unwrap
1888
for revision_id in revision_ids:
1889
if revision_id == _mod_revision.NULL_REVISION:
1890
result[revision_id] = ()
1891
elif revision_id is None:
1892
raise ValueError('get_parent_map(None) is not valid')
1894
query_keys.append((revision_id ,))
1895
for ((revision_id,), parent_keys) in \
1896
self.revisions.get_parent_map(query_keys).iteritems():
1898
result[revision_id] = tuple(parent_revid
1899
for (parent_revid,) in parent_keys)
1901
result[revision_id] = (_mod_revision.NULL_REVISION,)
1904
def _make_parents_provider(self):
1907
def get_graph(self, other_repository=None):
1908
"""Return the graph walker for this repository format"""
1909
parents_provider = self._make_parents_provider()
1910
if (other_repository is not None and
1911
not self.has_same_location(other_repository)):
1912
parents_provider = graph._StackedParentsProvider(
1913
[parents_provider, other_repository._make_parents_provider()])
1914
return graph.Graph(parents_provider)
1916
def _get_versioned_file_checker(self):
1917
"""Return an object suitable for checking versioned files."""
1918
return _VersionedFileChecker(self)
1920
def revision_ids_to_search_result(self, result_set):
1921
"""Convert a set of revision ids to a graph SearchResult."""
1922
result_parents = set()
1923
for parents in self.get_graph().get_parent_map(
1924
result_set).itervalues():
1925
result_parents.update(parents)
1926
included_keys = result_set.intersection(result_parents)
1927
start_keys = result_set.difference(included_keys)
1928
exclude_keys = result_parents.difference(result_set)
1929
result = graph.SearchResult(start_keys, exclude_keys,
1930
len(result_set), result_set)
1934
def set_make_working_trees(self, new_value):
1935
"""Set the policy flag for making working trees when creating branches.
1937
This only applies to branches that use this repository.
1939
The default is 'True'.
1940
:param new_value: True to restore the default, False to disable making
1943
raise NotImplementedError(self.set_make_working_trees)
1945
def make_working_trees(self):
1946
"""Returns the policy for making working trees on new branches."""
1947
raise NotImplementedError(self.make_working_trees)
1950
def sign_revision(self, revision_id, gpg_strategy):
1951
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1952
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1955
def has_signature_for_revision_id(self, revision_id):
1956
"""Query for a revision signature for revision_id in the repository."""
1957
if not self.has_revision(revision_id):
1958
raise errors.NoSuchRevision(self, revision_id)
1959
sig_present = (1 == len(
1960
self.signatures.get_parent_map([(revision_id,)])))
1964
def get_signature_text(self, revision_id):
1965
"""Return the text for a signature."""
1966
stream = self.signatures.get_record_stream([(revision_id,)],
1968
record = stream.next()
1969
if record.storage_kind == 'absent':
1970
raise errors.NoSuchRevision(self, revision_id)
1971
return record.get_bytes_as('fulltext')
1974
def check(self, revision_ids=None):
1975
"""Check consistency of all history of given revision_ids.
1977
Different repository implementations should override _check().
1979
:param revision_ids: A non-empty list of revision_ids whose ancestry
1980
will be checked. Typically the last revision_id of a branch.
1982
return self._check(revision_ids)
1984
def _check(self, revision_ids):
1985
result = check.Check(self)
1989
def _warn_if_deprecated(self):
1990
global _deprecation_warning_done
1991
if _deprecation_warning_done:
1993
_deprecation_warning_done = True
1994
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1995
% (self._format, self.bzrdir.transport.base))
1997
def supports_rich_root(self):
1998
return self._format.rich_root_data
2000
def _check_ascii_revisionid(self, revision_id, method):
2001
"""Private helper for ascii-only repositories."""
2002
# weave repositories refuse to store revisionids that are non-ascii.
2003
if revision_id is not None:
2004
# weaves require ascii revision ids.
2005
if isinstance(revision_id, unicode):
2007
revision_id.encode('ascii')
2008
except UnicodeEncodeError:
2009
raise errors.NonAsciiRevisionId(method, self)
2012
revision_id.decode('ascii')
2013
except UnicodeDecodeError:
2014
raise errors.NonAsciiRevisionId(method, self)
2016
def revision_graph_can_have_wrong_parents(self):
2017
"""Is it possible for this repository to have a revision graph with
2020
If True, then this repository must also implement
2021
_find_inconsistent_revision_parents so that check and reconcile can
2022
check for inconsistencies before proceeding with other checks that may
2023
depend on the revision index being consistent.
2025
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2028
# remove these delegates a while after bzr 0.15
2029
def __make_delegated(name, from_module):
2030
def _deprecated_repository_forwarder():
2031
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2032
% (name, from_module),
2035
m = __import__(from_module, globals(), locals(), [name])
2037
return getattr(m, name)
2038
except AttributeError:
2039
raise AttributeError('module %s has no name %s'
2041
globals()[name] = _deprecated_repository_forwarder
2044
'AllInOneRepository',
2045
'WeaveMetaDirRepository',
2046
'PreSplitOutRepositoryFormat',
2047
'RepositoryFormat4',
2048
'RepositoryFormat5',
2049
'RepositoryFormat6',
2050
'RepositoryFormat7',
2052
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2056
'RepositoryFormatKnit',
2057
'RepositoryFormatKnit1',
2059
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2062
def install_revision(repository, rev, revision_tree):
2063
"""Install all revision data into a repository."""
2064
install_revisions(repository, [(rev, revision_tree, None)])
2067
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2068
"""Install all revision data into a repository.
2070
Accepts an iterable of revision, tree, signature tuples. The signature
2073
repository.start_write_group()
2075
for n, (revision, revision_tree, signature) in enumerate(iterable):
2076
_install_revision(repository, revision, revision_tree, signature)
2078
pb.update('Transferring revisions', n + 1, num_revisions)
2080
repository.abort_write_group()
2083
repository.commit_write_group()
2086
def _install_revision(repository, rev, revision_tree, signature):
2087
"""Install all revision data into a repository."""
2088
present_parents = []
2090
for p_id in rev.parent_ids:
2091
if repository.has_revision(p_id):
2092
present_parents.append(p_id)
2093
parent_trees[p_id] = repository.revision_tree(p_id)
2095
parent_trees[p_id] = repository.revision_tree(
2096
_mod_revision.NULL_REVISION)
2098
inv = revision_tree.inventory
2099
entries = inv.iter_entries()
2100
# backwards compatibility hack: skip the root id.
2101
if not repository.supports_rich_root():
2102
path, root = entries.next()
2103
if root.revision != rev.revision_id:
2104
raise errors.IncompatibleRevision(repr(repository))
2106
for path, ie in entries:
2107
text_keys[(ie.file_id, ie.revision)] = ie
2108
text_parent_map = repository.texts.get_parent_map(text_keys)
2109
missing_texts = set(text_keys) - set(text_parent_map)
2110
# Add the texts that are not already present
2111
for text_key in missing_texts:
2112
ie = text_keys[text_key]
2114
# FIXME: TODO: The following loop overlaps/duplicates that done by
2115
# commit to determine parents. There is a latent/real bug here where
2116
# the parents inserted are not those commit would do - in particular
2117
# they are not filtered by heads(). RBC, AB
2118
for revision, tree in parent_trees.iteritems():
2119
if ie.file_id not in tree:
2121
parent_id = tree.inventory[ie.file_id].revision
2122
if parent_id in text_parents:
2124
text_parents.append((ie.file_id, parent_id))
2125
lines = revision_tree.get_file(ie.file_id).readlines()
2126
repository.texts.add_lines(text_key, text_parents, lines)
2128
# install the inventory
2129
repository.add_inventory(rev.revision_id, inv, present_parents)
2130
except errors.RevisionAlreadyPresent:
2132
if signature is not None:
2133
repository.add_signature_text(rev.revision_id, signature)
2134
repository.add_revision(rev.revision_id, rev, inv)
2137
class MetaDirRepository(Repository):
2138
"""Repositories in the new meta-dir layout.
2140
:ivar _transport: Transport for access to repository control files,
2141
typically pointing to .bzr/repository.
2144
def __init__(self, _format, a_bzrdir, control_files):
2145
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2146
self._transport = control_files._transport
2148
def is_shared(self):
2149
"""Return True if this repository is flagged as a shared repository."""
2150
return self._transport.has('shared-storage')
2153
def set_make_working_trees(self, new_value):
2154
"""Set the policy flag for making working trees when creating branches.
2156
This only applies to branches that use this repository.
2158
The default is 'True'.
2159
:param new_value: True to restore the default, False to disable making
2164
self._transport.delete('no-working-trees')
2165
except errors.NoSuchFile:
2168
self._transport.put_bytes('no-working-trees', '',
2169
mode=self.bzrdir._get_file_mode())
2171
def make_working_trees(self):
2172
"""Returns the policy for making working trees on new branches."""
2173
return not self._transport.has('no-working-trees')
2176
class MetaDirVersionedFileRepository(MetaDirRepository):
2177
"""Repositories in a meta-dir, that work via versioned file objects."""
2179
def __init__(self, _format, a_bzrdir, control_files):
2180
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2184
class RepositoryFormatRegistry(registry.Registry):
2185
"""Registry of RepositoryFormats."""
2187
def get(self, format_string):
2188
r = registry.Registry.get(self, format_string)
2194
format_registry = RepositoryFormatRegistry()
2195
"""Registry of formats, indexed by their identifying format string.
2197
This can contain either format instances themselves, or classes/factories that
2198
can be called to obtain one.
2202
#####################################################################
2203
# Repository Formats
2205
class RepositoryFormat(object):
2206
"""A repository format.
2208
Formats provide three things:
2209
* An initialization routine to construct repository data on disk.
2210
* a format string which is used when the BzrDir supports versioned
2212
* an open routine which returns a Repository instance.
2214
There is one and only one Format subclass for each on-disk format. But
2215
there can be one Repository subclass that is used for several different
2216
formats. The _format attribute on a Repository instance can be used to
2217
determine the disk format.
2219
Formats are placed in an dict by their format string for reference
2220
during opening. These should be subclasses of RepositoryFormat
2223
Once a format is deprecated, just deprecate the initialize and open
2224
methods on the format class. Do not deprecate the object, as the
2225
object will be created every system load.
2227
Common instance attributes:
2228
_matchingbzrdir - the bzrdir format that the repository format was
2229
originally written to work with. This can be used if manually
2230
constructing a bzrdir and repository, or more commonly for test suite
2234
# Set to True or False in derived classes. True indicates that the format
2235
# supports ghosts gracefully.
2236
supports_ghosts = None
2237
# Can this repository be given external locations to lookup additional
2238
# data. Set to True or False in derived classes.
2239
supports_external_lookups = None
2242
return "<%s>" % self.__class__.__name__
2244
def __eq__(self, other):
2245
# format objects are generally stateless
2246
return isinstance(other, self.__class__)
2248
def __ne__(self, other):
2249
return not self == other
2252
def find_format(klass, a_bzrdir):
2253
"""Return the format for the repository object in a_bzrdir.
2255
This is used by bzr native formats that have a "format" file in
2256
the repository. Other methods may be used by different types of
2260
transport = a_bzrdir.get_repository_transport(None)
2261
format_string = transport.get("format").read()
2262
return format_registry.get(format_string)
2263
except errors.NoSuchFile:
2264
raise errors.NoRepositoryPresent(a_bzrdir)
2266
raise errors.UnknownFormatError(format=format_string,
2270
def register_format(klass, format):
2271
format_registry.register(format.get_format_string(), format)
2274
def unregister_format(klass, format):
2275
format_registry.remove(format.get_format_string())
2278
def get_default_format(klass):
2279
"""Return the current default format."""
2280
from bzrlib import bzrdir
2281
return bzrdir.format_registry.make_bzrdir('default').repository_format
2283
def get_format_string(self):
2284
"""Return the ASCII format string that identifies this format.
2286
Note that in pre format ?? repositories the format string is
2287
not permitted nor written to disk.
2289
raise NotImplementedError(self.get_format_string)
2291
def get_format_description(self):
2292
"""Return the short description for this format."""
2293
raise NotImplementedError(self.get_format_description)
2295
# TODO: this shouldn't be in the base class, it's specific to things that
2296
# use weaves or knits -- mbp 20070207
2297
def _get_versioned_file_store(self,
2302
versionedfile_class=None,
2303
versionedfile_kwargs={},
2305
if versionedfile_class is None:
2306
versionedfile_class = self._versionedfile_class
2307
weave_transport = control_files._transport.clone(name)
2308
dir_mode = control_files._dir_mode
2309
file_mode = control_files._file_mode
2310
return VersionedFileStore(weave_transport, prefixed=prefixed,
2312
file_mode=file_mode,
2313
versionedfile_class=versionedfile_class,
2314
versionedfile_kwargs=versionedfile_kwargs,
2317
def initialize(self, a_bzrdir, shared=False):
2318
"""Initialize a repository of this format in a_bzrdir.
2320
:param a_bzrdir: The bzrdir to put the new repository in it.
2321
:param shared: The repository should be initialized as a sharable one.
2322
:returns: The new repository object.
2324
This may raise UninitializableFormat if shared repository are not
2325
compatible the a_bzrdir.
2327
raise NotImplementedError(self.initialize)
2329
def is_supported(self):
2330
"""Is this format supported?
2332
Supported formats must be initializable and openable.
2333
Unsupported formats may not support initialization or committing or
2334
some other features depending on the reason for not being supported.
2338
def check_conversion_target(self, target_format):
2339
raise NotImplementedError(self.check_conversion_target)
2341
def open(self, a_bzrdir, _found=False):
2342
"""Return an instance of this format for the bzrdir a_bzrdir.
2344
_found is a private parameter, do not use it.
2346
raise NotImplementedError(self.open)
2349
class MetaDirRepositoryFormat(RepositoryFormat):
2350
"""Common base class for the new repositories using the metadir layout."""
2352
rich_root_data = False
2353
supports_tree_reference = False
2354
supports_external_lookups = False
2357
def _matchingbzrdir(self):
2358
matching = bzrdir.BzrDirMetaFormat1()
2359
matching.repository_format = self
2363
super(MetaDirRepositoryFormat, self).__init__()
2365
def _create_control_files(self, a_bzrdir):
2366
"""Create the required files and the initial control_files object."""
2367
# FIXME: RBC 20060125 don't peek under the covers
2368
# NB: no need to escape relative paths that are url safe.
2369
repository_transport = a_bzrdir.get_repository_transport(self)
2370
control_files = lockable_files.LockableFiles(repository_transport,
2371
'lock', lockdir.LockDir)
2372
control_files.create_lock()
2373
return control_files
2375
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2376
"""Upload the initial blank content."""
2377
control_files = self._create_control_files(a_bzrdir)
2378
control_files.lock_write()
2379
transport = control_files._transport
2381
utf8_files += [('shared-storage', '')]
2383
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2384
for (filename, content_stream) in files:
2385
transport.put_file(filename, content_stream,
2386
mode=a_bzrdir._get_file_mode())
2387
for (filename, content_bytes) in utf8_files:
2388
transport.put_bytes_non_atomic(filename, content_bytes,
2389
mode=a_bzrdir._get_file_mode())
2391
control_files.unlock()
2394
# formats which have no format string are not discoverable
2395
# and not independently creatable, so are not registered. They're
2396
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2397
# needed, it's constructed directly by the BzrDir. Non-native formats where
2398
# the repository is not separately opened are similar.
2400
format_registry.register_lazy(
2401
'Bazaar-NG Repository format 7',
2402
'bzrlib.repofmt.weaverepo',
2406
format_registry.register_lazy(
2407
'Bazaar-NG Knit Repository Format 1',
2408
'bzrlib.repofmt.knitrepo',
2409
'RepositoryFormatKnit1',
2412
format_registry.register_lazy(
2413
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2414
'bzrlib.repofmt.knitrepo',
2415
'RepositoryFormatKnit3',
2418
format_registry.register_lazy(
2419
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2420
'bzrlib.repofmt.knitrepo',
2421
'RepositoryFormatKnit4',
2424
# Pack-based formats. There is one format for pre-subtrees, and one for
2425
# post-subtrees to allow ease of testing.
2426
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2427
format_registry.register_lazy(
2428
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2429
'bzrlib.repofmt.pack_repo',
2430
'RepositoryFormatKnitPack1',
2432
format_registry.register_lazy(
2433
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2434
'bzrlib.repofmt.pack_repo',
2435
'RepositoryFormatKnitPack3',
2437
format_registry.register_lazy(
2438
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2439
'bzrlib.repofmt.pack_repo',
2440
'RepositoryFormatKnitPack4',
2442
format_registry.register_lazy(
2443
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2444
'bzrlib.repofmt.pack_repo',
2445
'RepositoryFormatKnitPack5',
2447
format_registry.register_lazy(
2448
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2449
'bzrlib.repofmt.pack_repo',
2450
'RepositoryFormatKnitPack5RichRoot',
2452
format_registry.register_lazy(
2453
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2454
'bzrlib.repofmt.pack_repo',
2455
'RepositoryFormatKnitPack5RichRootBroken',
2457
format_registry.register_lazy(
2458
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2459
'bzrlib.repofmt.pack_repo',
2460
'RepositoryFormatKnitPack6',
2462
format_registry.register_lazy(
2463
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2464
'bzrlib.repofmt.pack_repo',
2465
'RepositoryFormatKnitPack6RichRoot',
2468
# Development formats.
2469
# 1.7->1.8 go below here
2470
format_registry.register_lazy(
2471
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2472
'bzrlib.repofmt.pack_repo',
2473
'RepositoryFormatPackDevelopment2',
2475
format_registry.register_lazy(
2476
("Bazaar development format 2 with subtree support "
2477
"(needs bzr.dev from before 1.8)\n"),
2478
'bzrlib.repofmt.pack_repo',
2479
'RepositoryFormatPackDevelopment2Subtree',
2483
class InterRepository(InterObject):
2484
"""This class represents operations taking place between two repositories.
2486
Its instances have methods like copy_content and fetch, and contain
2487
references to the source and target repositories these operations can be
2490
Often we will provide convenience methods on 'repository' which carry out
2491
operations with another repository - they will always forward to
2492
InterRepository.get(other).method_name(parameters).
2495
_walk_to_common_revisions_batch_size = 1
2497
"""The available optimised InterRepository types."""
2499
def __init__(self, source, target):
2500
InterObject.__init__(self, source, target)
2501
# These two attributes may be overridden by e.g. InterOtherToRemote to
2502
# provide a faster implementation.
2503
self.target_get_graph = self.target.get_graph
2504
self.target_get_parent_map = self.target.get_parent_map
2506
def copy_content(self, revision_id=None):
2507
raise NotImplementedError(self.copy_content)
2509
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2510
"""Fetch the content required to construct revision_id.
2512
The content is copied from self.source to self.target.
2514
:param revision_id: if None all content is copied, if NULL_REVISION no
2516
:param pb: optional progress bar to use for progress reports. If not
2517
provided a default one will be created.
2519
:returns: (copied_revision_count, failures).
2521
# Normally we should find a specific InterRepository subclass to do
2522
# the fetch; if nothing else then at least InterSameDataRepository.
2523
# If none of them is suitable it looks like fetching is not possible;
2524
# we try to give a good message why. _assert_same_model will probably
2525
# give a helpful message; otherwise a generic one.
2526
self._assert_same_model(self.source, self.target)
2527
raise errors.IncompatibleRepositories(self.source, self.target,
2528
"no suitableInterRepository found")
2530
def _walk_to_common_revisions(self, revision_ids):
2531
"""Walk out from revision_ids in source to revisions target has.
2533
:param revision_ids: The start point for the search.
2534
:return: A set of revision ids.
2536
target_graph = self.target_get_graph()
2537
revision_ids = frozenset(revision_ids)
2538
# Fast path for the case where all the revisions are already in the
2540
# (Although this does incur an extra round trip for the
2541
# fairly common case where the target doesn't already have the revision
2543
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2544
return graph.SearchResult(revision_ids, set(), 0, set())
2545
missing_revs = set()
2546
source_graph = self.source.get_graph()
2547
# ensure we don't pay silly lookup costs.
2548
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2549
null_set = frozenset([_mod_revision.NULL_REVISION])
2550
searcher_exhausted = False
2554
# Iterate the searcher until we have enough next_revs
2555
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2557
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2558
next_revs.update(next_revs_part)
2559
ghosts.update(ghosts_part)
2560
except StopIteration:
2561
searcher_exhausted = True
2563
# If there are ghosts in the source graph, and the caller asked for
2564
# them, make sure that they are present in the target.
2565
# We don't care about other ghosts as we can't fetch them and
2566
# haven't been asked to.
2567
ghosts_to_check = set(revision_ids.intersection(ghosts))
2568
revs_to_get = set(next_revs).union(ghosts_to_check)
2570
have_revs = set(target_graph.get_parent_map(revs_to_get))
2571
# we always have NULL_REVISION present.
2572
have_revs = have_revs.union(null_set)
2573
# Check if the target is missing any ghosts we need.
2574
ghosts_to_check.difference_update(have_revs)
2576
# One of the caller's revision_ids is a ghost in both the
2577
# source and the target.
2578
raise errors.NoSuchRevision(
2579
self.source, ghosts_to_check.pop())
2580
missing_revs.update(next_revs - have_revs)
2581
# Because we may have walked past the original stop point, make
2582
# sure everything is stopped
2583
stop_revs = searcher.find_seen_ancestors(have_revs)
2584
searcher.stop_searching_any(stop_revs)
2585
if searcher_exhausted:
2587
return searcher.get_result()
2589
@deprecated_method(one_two)
2591
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2592
"""Return the revision ids that source has that target does not.
2594
These are returned in topological order.
2596
:param revision_id: only return revision ids included by this
2598
:param find_ghosts: If True find missing revisions in deep history
2599
rather than just finding the surface difference.
2601
return list(self.search_missing_revision_ids(
2602
revision_id, find_ghosts).get_keys())
2605
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2606
"""Return the revision ids that source has that target does not.
2608
:param revision_id: only return revision ids included by this
2610
:param find_ghosts: If True find missing revisions in deep history
2611
rather than just finding the surface difference.
2612
:return: A bzrlib.graph.SearchResult.
2614
# stop searching at found target revisions.
2615
if not find_ghosts and revision_id is not None:
2616
return self._walk_to_common_revisions([revision_id])
2617
# generic, possibly worst case, slow code path.
2618
target_ids = set(self.target.all_revision_ids())
2619
if revision_id is not None:
2620
source_ids = self.source.get_ancestry(revision_id)
2621
if source_ids[0] is not None:
2622
raise AssertionError()
2625
source_ids = self.source.all_revision_ids()
2626
result_set = set(source_ids).difference(target_ids)
2627
return self.source.revision_ids_to_search_result(result_set)
2630
def _same_model(source, target):
2631
"""True if source and target have the same data representation.
2633
Note: this is always called on the base class; overriding it in a
2634
subclass will have no effect.
2637
InterRepository._assert_same_model(source, target)
2639
except errors.IncompatibleRepositories, e:
2643
def _assert_same_model(source, target):
2644
"""Raise an exception if two repositories do not use the same model.
2646
if source.supports_rich_root() != target.supports_rich_root():
2647
raise errors.IncompatibleRepositories(source, target,
2648
"different rich-root support")
2649
if source._serializer != target._serializer:
2650
raise errors.IncompatibleRepositories(source, target,
2651
"different serializers")
2654
class InterSameDataRepository(InterRepository):
2655
"""Code for converting between repositories that represent the same data.
2657
Data format and model must match for this to work.
2661
def _get_repo_format_to_test(self):
2662
"""Repository format for testing with.
2664
InterSameData can pull from subtree to subtree and from non-subtree to
2665
non-subtree, so we test this with the richest repository format.
2667
from bzrlib.repofmt import knitrepo
2668
return knitrepo.RepositoryFormatKnit3()
2671
def is_compatible(source, target):
2672
return InterRepository._same_model(source, target)
2675
def copy_content(self, revision_id=None):
2676
"""Make a complete copy of the content in self into destination.
2678
This copies both the repository's revision data, and configuration information
2679
such as the make_working_trees setting.
2681
This is a destructive operation! Do not use it on existing
2684
:param revision_id: Only copy the content needed to construct
2685
revision_id and its parents.
2688
self.target.set_make_working_trees(self.source.make_working_trees())
2689
except NotImplementedError:
2691
# but don't bother fetching if we have the needed data now.
2692
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2693
self.target.has_revision(revision_id)):
2695
self.target.fetch(self.source, revision_id=revision_id)
2698
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2699
"""See InterRepository.fetch()."""
2700
from bzrlib.fetch import RepoFetcher
2701
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2702
self.source, self.source._format, self.target,
2703
self.target._format)
2704
f = RepoFetcher(to_repository=self.target,
2705
from_repository=self.source,
2706
last_revision=revision_id,
2707
pb=pb, find_ghosts=find_ghosts)
2708
return f.count_copied, f.failed_revisions
2711
class InterWeaveRepo(InterSameDataRepository):
2712
"""Optimised code paths between Weave based repositories.
2714
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2715
implemented lazy inter-object optimisation.
2719
def _get_repo_format_to_test(self):
2720
from bzrlib.repofmt import weaverepo
2721
return weaverepo.RepositoryFormat7()
2724
def is_compatible(source, target):
2725
"""Be compatible with known Weave formats.
2727
We don't test for the stores being of specific types because that
2728
could lead to confusing results, and there is no need to be
2731
from bzrlib.repofmt.weaverepo import (
2737
return (isinstance(source._format, (RepositoryFormat5,
2739
RepositoryFormat7)) and
2740
isinstance(target._format, (RepositoryFormat5,
2742
RepositoryFormat7)))
2743
except AttributeError:
2747
def copy_content(self, revision_id=None):
2748
"""See InterRepository.copy_content()."""
2749
# weave specific optimised path:
2751
self.target.set_make_working_trees(self.source.make_working_trees())
2752
except (errors.RepositoryUpgradeRequired, NotImplemented):
2754
# FIXME do not peek!
2755
if self.source._transport.listable():
2756
pb = ui.ui_factory.nested_progress_bar()
2758
self.target.texts.insert_record_stream(
2759
self.source.texts.get_record_stream(
2760
self.source.texts.keys(), 'topological', False))
2761
pb.update('copying inventory', 0, 1)
2762
self.target.inventories.insert_record_stream(
2763
self.source.inventories.get_record_stream(
2764
self.source.inventories.keys(), 'topological', False))
2765
self.target.signatures.insert_record_stream(
2766
self.source.signatures.get_record_stream(
2767
self.source.signatures.keys(),
2769
self.target.revisions.insert_record_stream(
2770
self.source.revisions.get_record_stream(
2771
self.source.revisions.keys(),
2772
'topological', True))
2776
self.target.fetch(self.source, revision_id=revision_id)
2779
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2780
"""See InterRepository.fetch()."""
2781
from bzrlib.fetch import RepoFetcher
2782
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2783
self.source, self.source._format, self.target, self.target._format)
2784
f = RepoFetcher(to_repository=self.target,
2785
from_repository=self.source,
2786
last_revision=revision_id,
2787
pb=pb, find_ghosts=find_ghosts)
2788
return f.count_copied, f.failed_revisions
2791
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2792
"""See InterRepository.missing_revision_ids()."""
2793
# we want all revisions to satisfy revision_id in source.
2794
# but we don't want to stat every file here and there.
2795
# we want then, all revisions other needs to satisfy revision_id
2796
# checked, but not those that we have locally.
2797
# so the first thing is to get a subset of the revisions to
2798
# satisfy revision_id in source, and then eliminate those that
2799
# we do already have.
2800
# this is slow on high latency connection to self, but as as this
2801
# disk format scales terribly for push anyway due to rewriting
2802
# inventory.weave, this is considered acceptable.
2804
if revision_id is not None:
2805
source_ids = self.source.get_ancestry(revision_id)
2806
if source_ids[0] is not None:
2807
raise AssertionError()
2810
source_ids = self.source._all_possible_ids()
2811
source_ids_set = set(source_ids)
2812
# source_ids is the worst possible case we may need to pull.
2813
# now we want to filter source_ids against what we actually
2814
# have in target, but don't try to check for existence where we know
2815
# we do not have a revision as that would be pointless.
2816
target_ids = set(self.target._all_possible_ids())
2817
possibly_present_revisions = target_ids.intersection(source_ids_set)
2818
actually_present_revisions = set(
2819
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2820
required_revisions = source_ids_set.difference(actually_present_revisions)
2821
if revision_id is not None:
2822
# we used get_ancestry to determine source_ids then we are assured all
2823
# revisions referenced are present as they are installed in topological order.
2824
# and the tip revision was validated by get_ancestry.
2825
result_set = required_revisions
2827
# if we just grabbed the possibly available ids, then
2828
# we only have an estimate of whats available and need to validate
2829
# that against the revision records.
2831
self.source._eliminate_revisions_not_present(required_revisions))
2832
return self.source.revision_ids_to_search_result(result_set)
2835
class InterKnitRepo(InterSameDataRepository):
2836
"""Optimised code paths between Knit based repositories."""
2839
def _get_repo_format_to_test(self):
2840
from bzrlib.repofmt import knitrepo
2841
return knitrepo.RepositoryFormatKnit1()
2844
def is_compatible(source, target):
2845
"""Be compatible with known Knit formats.
2847
We don't test for the stores being of specific types because that
2848
could lead to confusing results, and there is no need to be
2851
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2853
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2854
isinstance(target._format, RepositoryFormatKnit))
2855
except AttributeError:
2857
return are_knits and InterRepository._same_model(source, target)
2860
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2861
"""See InterRepository.fetch()."""
2862
from bzrlib.fetch import RepoFetcher
2863
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2864
self.source, self.source._format, self.target, self.target._format)
2865
f = RepoFetcher(to_repository=self.target,
2866
from_repository=self.source,
2867
last_revision=revision_id,
2868
pb=pb, find_ghosts=find_ghosts)
2869
return f.count_copied, f.failed_revisions
2872
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2873
"""See InterRepository.missing_revision_ids()."""
2874
if revision_id is not None:
2875
source_ids = self.source.get_ancestry(revision_id)
2876
if source_ids[0] is not None:
2877
raise AssertionError()
2880
source_ids = self.source.all_revision_ids()
2881
source_ids_set = set(source_ids)
2882
# source_ids is the worst possible case we may need to pull.
2883
# now we want to filter source_ids against what we actually
2884
# have in target, but don't try to check for existence where we know
2885
# we do not have a revision as that would be pointless.
2886
target_ids = set(self.target.all_revision_ids())
2887
possibly_present_revisions = target_ids.intersection(source_ids_set)
2888
actually_present_revisions = set(
2889
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2890
required_revisions = source_ids_set.difference(actually_present_revisions)
2891
if revision_id is not None:
2892
# we used get_ancestry to determine source_ids then we are assured all
2893
# revisions referenced are present as they are installed in topological order.
2894
# and the tip revision was validated by get_ancestry.
2895
result_set = required_revisions
2897
# if we just grabbed the possibly available ids, then
2898
# we only have an estimate of whats available and need to validate
2899
# that against the revision records.
2901
self.source._eliminate_revisions_not_present(required_revisions))
2902
return self.source.revision_ids_to_search_result(result_set)
2905
class InterPackRepo(InterSameDataRepository):
2906
"""Optimised code paths between Pack based repositories."""
2909
def _get_repo_format_to_test(self):
2910
from bzrlib.repofmt import pack_repo
2911
return pack_repo.RepositoryFormatKnitPack1()
2914
def is_compatible(source, target):
2915
"""Be compatible with known Pack formats.
2917
We don't test for the stores being of specific types because that
2918
could lead to confusing results, and there is no need to be
2921
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2923
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2924
isinstance(target._format, RepositoryFormatPack))
2925
except AttributeError:
2927
return are_packs and InterRepository._same_model(source, target)
2930
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2931
"""See InterRepository.fetch()."""
2932
if (len(self.source._fallback_repositories) > 0 or
2933
len(self.target._fallback_repositories) > 0):
2934
# The pack layer is not aware of fallback repositories, so when
2935
# fetching from a stacked repository or into a stacked repository
2936
# we use the generic fetch logic which uses the VersionedFiles
2937
# attributes on repository.
2938
from bzrlib.fetch import RepoFetcher
2939
# Make sure the generic fetcher sets the write cache size on the
2940
# new pack (just like Packer.pack does) to avoid doing many tiny
2941
# writes (which can be slow over a network connection).
2942
# XXX: ideally the transport layer would do this automatically.
2943
pack_coll = self._get_target_pack_collection()
2945
lambda: pack_coll._new_pack.set_write_cache_size(1024*1024))
2946
fetcher = RepoFetcher(self.target, self.source, revision_id,
2947
pb, find_ghosts, set_cache_size)
2948
return fetcher.count_copied, fetcher.failed_revisions
2949
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2950
self.source, self.source._format, self.target, self.target._format)
2951
self.count_copied = 0
2952
if revision_id is None:
2954
# everything to do - use pack logic
2955
# to fetch from all packs to one without
2956
# inventory parsing etc, IFF nothing to be copied is in the target.
2958
source_revision_ids = frozenset(self.source.all_revision_ids())
2959
revision_ids = source_revision_ids - \
2960
frozenset(self.target_get_parent_map(source_revision_ids))
2961
revision_keys = [(revid,) for revid in revision_ids]
2962
target_pack_collection = self._get_target_pack_collection()
2963
index = target_pack_collection.revision_index.combined_index
2964
present_revision_ids = set(item[1][0] for item in
2965
index.iter_entries(revision_keys))
2966
revision_ids = set(revision_ids) - present_revision_ids
2967
# implementing the TODO will involve:
2968
# - detecting when all of a pack is selected
2969
# - avoiding as much as possible pre-selection, so the
2970
# more-core routines such as create_pack_from_packs can filter in
2971
# a just-in-time fashion. (though having a HEADS list on a
2972
# repository might make this a lot easier, because we could
2973
# sensibly detect 'new revisions' without doing a full index scan.
2974
elif _mod_revision.is_null(revision_id):
2979
revision_ids = self.search_missing_revision_ids(revision_id,
2980
find_ghosts=find_ghosts).get_keys()
2981
except errors.NoSuchRevision:
2982
raise errors.InstallFailed([revision_id])
2983
if len(revision_ids) == 0:
2985
return self._pack(self.source, self.target, revision_ids)
2987
def _pack(self, source, target, revision_ids):
2988
from bzrlib.repofmt.pack_repo import Packer
2989
target_pack_collection = self._get_target_pack_collection()
2990
packs = source._pack_collection.all_packs()
2991
pack = Packer(target_pack_collection, packs, '.fetch',
2992
revision_ids).pack()
2993
if pack is not None:
2994
target_pack_collection._save_pack_names()
2995
copied_revs = pack.get_revision_count()
2996
# Trigger an autopack. This may duplicate effort as we've just done
2997
# a pack creation, but for now it is simpler to think about as
2998
# 'upload data, then repack if needed'.
3000
return (copied_revs, [])
3004
def _autopack(self):
3005
self.target._pack_collection.autopack()
3007
def _get_target_pack_collection(self):
3008
return self.target._pack_collection
3011
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3012
"""See InterRepository.missing_revision_ids().
3014
:param find_ghosts: Find ghosts throughout the ancestry of
3017
if not find_ghosts and revision_id is not None:
3018
return self._walk_to_common_revisions([revision_id])
3019
elif revision_id is not None:
3020
# Find ghosts: search for revisions pointing from one repository to
3021
# the other, and vice versa, anywhere in the history of revision_id.
3022
graph = self.target_get_graph(other_repository=self.source)
3023
searcher = graph._make_breadth_first_searcher([revision_id])
3027
next_revs, ghosts = searcher.next_with_ghosts()
3028
except StopIteration:
3030
if revision_id in ghosts:
3031
raise errors.NoSuchRevision(self.source, revision_id)
3032
found_ids.update(next_revs)
3033
found_ids.update(ghosts)
3034
found_ids = frozenset(found_ids)
3035
# Double query here: should be able to avoid this by changing the
3036
# graph api further.
3037
result_set = found_ids - frozenset(
3038
self.target_get_parent_map(found_ids))
3040
source_ids = self.source.all_revision_ids()
3041
# source_ids is the worst possible case we may need to pull.
3042
# now we want to filter source_ids against what we actually
3043
# have in target, but don't try to check for existence where we know
3044
# we do not have a revision as that would be pointless.
3045
target_ids = set(self.target.all_revision_ids())
3046
result_set = set(source_ids).difference(target_ids)
3047
return self.source.revision_ids_to_search_result(result_set)
3050
class InterModel1and2(InterRepository):
3053
def _get_repo_format_to_test(self):
3057
def is_compatible(source, target):
3058
if not source.supports_rich_root() and target.supports_rich_root():
3064
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3065
"""See InterRepository.fetch()."""
3066
from bzrlib.fetch import Model1toKnit2Fetcher
3067
f = Model1toKnit2Fetcher(to_repository=self.target,
3068
from_repository=self.source,
3069
last_revision=revision_id,
3070
pb=pb, find_ghosts=find_ghosts)
3071
return f.count_copied, f.failed_revisions
3074
def copy_content(self, revision_id=None):
3075
"""Make a complete copy of the content in self into destination.
3077
This is a destructive operation! Do not use it on existing
3080
:param revision_id: Only copy the content needed to construct
3081
revision_id and its parents.
3084
self.target.set_make_working_trees(self.source.make_working_trees())
3085
except NotImplementedError:
3087
# but don't bother fetching if we have the needed data now.
3088
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
3089
self.target.has_revision(revision_id)):
3091
self.target.fetch(self.source, revision_id=revision_id)
3094
class InterKnit1and2(InterKnitRepo):
3097
def _get_repo_format_to_test(self):
3101
def is_compatible(source, target):
3102
"""Be compatible with Knit1 source and Knit3 target"""
3104
from bzrlib.repofmt.knitrepo import (
3105
RepositoryFormatKnit1,
3106
RepositoryFormatKnit3,
3108
from bzrlib.repofmt.pack_repo import (
3109
RepositoryFormatKnitPack1,
3110
RepositoryFormatKnitPack3,
3111
RepositoryFormatKnitPack4,
3112
RepositoryFormatKnitPack5,
3113
RepositoryFormatKnitPack5RichRoot,
3114
RepositoryFormatKnitPack6,
3115
RepositoryFormatKnitPack6RichRoot,
3116
RepositoryFormatPackDevelopment2,
3117
RepositoryFormatPackDevelopment2Subtree,
3120
RepositoryFormatKnit1, # no rr, no subtree
3121
RepositoryFormatKnitPack1, # no rr, no subtree
3122
RepositoryFormatPackDevelopment2, # no rr, no subtree
3123
RepositoryFormatKnitPack5, # no rr, no subtree
3124
RepositoryFormatKnitPack6, # no rr, no subtree
3127
RepositoryFormatKnit3, # rr, subtree
3128
RepositoryFormatKnitPack3, # rr, subtree
3129
RepositoryFormatKnitPack4, # rr, no subtree
3130
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
3131
RepositoryFormatKnitPack6RichRoot,# rr, no subtree
3132
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
3134
for format in norichroot:
3135
if format.rich_root_data:
3136
raise AssertionError('Format %s is a rich-root format'
3137
' but is included in the non-rich-root list'
3139
for format in richroot:
3140
if not format.rich_root_data:
3141
raise AssertionError('Format %s is not a rich-root format'
3142
' but is included in the rich-root list'
3144
# TODO: One alternative is to just check format.rich_root_data,
3145
# instead of keeping membership lists. However, the formats
3146
# *also* have to use the same 'Knit' style of storage
3147
# (line-deltas, fulltexts, etc.)
3148
return (isinstance(source._format, norichroot) and
3149
isinstance(target._format, richroot))
3150
except AttributeError:
3154
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3155
"""See InterRepository.fetch()."""
3156
from bzrlib.fetch import Knit1to2Fetcher
3157
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3158
self.source, self.source._format, self.target,
3159
self.target._format)
3160
f = Knit1to2Fetcher(to_repository=self.target,
3161
from_repository=self.source,
3162
last_revision=revision_id,
3163
pb=pb, find_ghosts=find_ghosts)
3164
return f.count_copied, f.failed_revisions
3167
class InterDifferingSerializer(InterKnitRepo):
3170
def _get_repo_format_to_test(self):
3174
def is_compatible(source, target):
3175
"""Be compatible with Knit2 source and Knit3 target"""
3176
if source.supports_rich_root() != target.supports_rich_root():
3178
# Ideally, we'd support fetching if the source had no tree references
3179
# even if it supported them...
3180
if (getattr(source, '_format.supports_tree_reference', False) and
3181
not getattr(target, '_format.supports_tree_reference', False)):
3185
def _fetch_batch(self, revision_ids, basis_id, basis_tree):
3186
"""Fetch across a few revisions.
3188
:param revision_ids: The revisions to copy
3189
:param basis_id: The revision_id of basis_tree
3190
:param basis_tree: A tree that is not in revision_ids which should
3191
already exist in the target.
3192
:return: (basis_id, basis_tree) A new basis to use now that these trees
3195
# Walk though all revisions; get inventory deltas, copy referenced
3196
# texts that delta references, insert the delta, revision and
3200
pending_revisions = []
3201
for tree in self.source.revision_trees(revision_ids):
3202
current_revision_id = tree.get_revision_id()
3203
delta = tree.inventory._make_delta(basis_tree.inventory)
3204
for old_path, new_path, file_id, entry in delta:
3205
if new_path is not None:
3206
if not (new_path or self.target.supports_rich_root()):
3207
# We leave the inventory delta in, because that
3208
# will have the deserialised inventory root
3212
# "if entry.revision == current_revision_id" ?
3213
if entry.revision == current_revision_id:
3214
text_keys.add((file_id, entry.revision))
3215
revision = self.source.get_revision(current_revision_id)
3216
pending_deltas.append((basis_id, delta,
3217
current_revision_id, revision.parent_ids))
3218
pending_revisions.append(revision)
3219
basis_id = current_revision_id
3222
from_texts = self.source.texts
3223
to_texts = self.target.texts
3224
to_texts.insert_record_stream(from_texts.get_record_stream(
3225
text_keys, self.target._fetch_order,
3226
not self.target._fetch_uses_deltas))
3228
for delta in pending_deltas:
3229
self.target.add_inventory_by_delta(*delta)
3230
# insert signatures and revisions
3231
for revision in pending_revisions:
3233
signature = self.source.get_signature_text(
3234
revision.revision_id)
3235
self.target.add_signature_text(revision.revision_id,
3237
except errors.NoSuchRevision:
3239
self.target.add_revision(revision.revision_id, revision)
3240
return basis_id, basis_tree
3242
def _fetch_all_revisions(self, revision_ids, pb):
3243
"""Fetch everything for the list of revisions.
3245
:param revision_ids: The list of revisions to fetch. Must be in
3247
:param pb: A ProgressBar
3250
basis_id, basis_tree = self._get_basis(revision_ids[0])
3252
for offset in range(0, len(revision_ids), batch_size):
3253
self.target.start_write_group()
3255
pb.update('Transferring revisions', offset,
3257
batch = revision_ids[offset:offset+batch_size]
3258
basis_id, basis_tree = self._fetch_batch(batch,
3259
basis_id, basis_tree)
3261
self.target.abort_write_group()
3264
self.target.commit_write_group()
3267
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3268
"""See InterRepository.fetch()."""
3269
revision_ids = self.target.search_missing_revision_ids(self.source,
3270
revision_id, find_ghosts=find_ghosts).get_keys()
3271
if not revision_ids:
3273
revision_ids = tsort.topo_sort(
3274
self.source.get_graph().get_parent_map(revision_ids))
3276
my_pb = ui.ui_factory.nested_progress_bar()
3281
self._fetch_all_revisions(revision_ids, pb)
3283
if my_pb is not None:
3285
return len(revision_ids), 0
3287
def _get_basis(self, first_revision_id):
3288
"""Get a revision and tree which exists in the target.
3290
This assumes that first_revision_id is selected for transmission
3291
because all other ancestors are already present. If we can't find an
3292
ancestor we fall back to NULL_REVISION since we know that is safe.
3294
:return: (basis_id, basis_tree)
3296
first_rev = self.source.get_revision(first_revision_id)
3298
basis_id = first_rev.parent_ids[0]
3299
# only valid as a basis if the target has it
3300
self.target.get_revision(basis_id)
3301
# Try to get a basis tree - if its a ghost it will hit the
3302
# NoSuchRevision case.
3303
basis_tree = self.source.revision_tree(basis_id)
3304
except (IndexError, errors.NoSuchRevision):
3305
basis_id = _mod_revision.NULL_REVISION
3306
basis_tree = self.source.revision_tree(basis_id)
3307
return basis_id, basis_tree
3310
class InterOtherToRemote(InterRepository):
3311
"""An InterRepository that simply delegates to the 'real' InterRepository
3312
calculated for (source, target._real_repository).
3315
_walk_to_common_revisions_batch_size = 50
3317
def __init__(self, source, target):
3318
InterRepository.__init__(self, source, target)
3319
self._real_inter = None
3322
def is_compatible(source, target):
3323
if isinstance(target, remote.RemoteRepository):
3327
def _ensure_real_inter(self):
3328
if self._real_inter is None:
3329
self.target._ensure_real()
3330
real_target = self.target._real_repository
3331
self._real_inter = InterRepository.get(self.source, real_target)
3332
# Make _real_inter use the RemoteRepository for get_parent_map
3333
self._real_inter.target_get_graph = self.target.get_graph
3334
self._real_inter.target_get_parent_map = self.target.get_parent_map
3336
def copy_content(self, revision_id=None):
3337
self._ensure_real_inter()
3338
self._real_inter.copy_content(revision_id=revision_id)
3340
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3341
self._ensure_real_inter()
3342
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3343
find_ghosts=find_ghosts)
3346
def _get_repo_format_to_test(self):
3350
class InterRemoteToOther(InterRepository):
3352
def __init__(self, source, target):
3353
InterRepository.__init__(self, source, target)
3354
self._real_inter = None
3357
def is_compatible(source, target):
3358
if not isinstance(source, remote.RemoteRepository):
3360
# Is source's model compatible with target's model?
3361
source._ensure_real()
3362
real_source = source._real_repository
3363
if isinstance(real_source, remote.RemoteRepository):
3364
raise NotImplementedError(
3365
"We don't support remote repos backed by remote repos yet.")
3366
return InterRepository._same_model(real_source, target)
3368
def _ensure_real_inter(self):
3369
if self._real_inter is None:
3370
self.source._ensure_real()
3371
real_source = self.source._real_repository
3372
self._real_inter = InterRepository.get(real_source, self.target)
3374
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3375
self._ensure_real_inter()
3376
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3377
find_ghosts=find_ghosts)
3379
def copy_content(self, revision_id=None):
3380
self._ensure_real_inter()
3381
self._real_inter.copy_content(revision_id=revision_id)
3384
def _get_repo_format_to_test(self):
3389
class InterPackToRemotePack(InterPackRepo):
3390
"""A specialisation of InterPackRepo for a target that is a
3393
This will use the get_parent_map RPC rather than plain readvs, and also
3394
uses an RPC for autopacking.
3397
_walk_to_common_revisions_batch_size = 50
3400
def is_compatible(source, target):
3401
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3402
if isinstance(source._format, RepositoryFormatPack):
3403
if isinstance(target, remote.RemoteRepository):
3404
target._ensure_real()
3405
if isinstance(target._real_repository._format,
3406
RepositoryFormatPack):
3407
if InterRepository._same_model(source, target):
3411
def _autopack(self):
3412
self.target.autopack()
3414
def _get_target_pack_collection(self):
3415
return self.target._real_repository._pack_collection
3418
def _get_repo_format_to_test(self):
3422
InterRepository.register_optimiser(InterDifferingSerializer)
3423
InterRepository.register_optimiser(InterSameDataRepository)
3424
InterRepository.register_optimiser(InterWeaveRepo)
3425
InterRepository.register_optimiser(InterKnitRepo)
3426
InterRepository.register_optimiser(InterModel1and2)
3427
InterRepository.register_optimiser(InterKnit1and2)
3428
InterRepository.register_optimiser(InterPackRepo)
3429
InterRepository.register_optimiser(InterOtherToRemote)
3430
InterRepository.register_optimiser(InterRemoteToOther)
3431
InterRepository.register_optimiser(InterPackToRemotePack)
3434
class CopyConverter(object):
3435
"""A repository conversion tool which just performs a copy of the content.
3437
This is slow but quite reliable.
3440
def __init__(self, target_format):
3441
"""Create a CopyConverter.
3443
:param target_format: The format the resulting repository should be.
3445
self.target_format = target_format
3447
def convert(self, repo, pb):
3448
"""Perform the conversion of to_convert, giving feedback via pb.
3450
:param to_convert: The disk object to convert.
3451
:param pb: a progress bar to use for progress information.
3456
# this is only useful with metadir layouts - separated repo content.
3457
# trigger an assertion if not such
3458
repo._format.get_format_string()
3459
self.repo_dir = repo.bzrdir
3460
self.step('Moving repository to repository.backup')
3461
self.repo_dir.transport.move('repository', 'repository.backup')
3462
backup_transport = self.repo_dir.transport.clone('repository.backup')
3463
repo._format.check_conversion_target(self.target_format)
3464
self.source_repo = repo._format.open(self.repo_dir,
3466
_override_transport=backup_transport)
3467
self.step('Creating new repository')
3468
converted = self.target_format.initialize(self.repo_dir,
3469
self.source_repo.is_shared())
3470
converted.lock_write()
3472
self.step('Copying content into repository.')
3473
self.source_repo.copy_content_into(converted)
3476
self.step('Deleting old repository content.')
3477
self.repo_dir.transport.delete_tree('repository.backup')
3478
self.pb.note('repository converted')
3480
def step(self, message):
3481
"""Update the pb by a step."""
3483
self.pb.update(message, self.count, self.total)
3495
def _unescaper(match, _map=_unescape_map):
3496
code = match.group(1)
3500
if not code.startswith('#'):
3502
return unichr(int(code[1:])).encode('utf8')
3508
def _unescape_xml(data):
3509
"""Unescape predefined XML entities in a string of data."""
3511
if _unescape_re is None:
3512
_unescape_re = re.compile('\&([^;]*);')
3513
return _unescape_re.sub(_unescaper, data)
3516
class _VersionedFileChecker(object):
3518
def __init__(self, repository):
3519
self.repository = repository
3520
self.text_index = self.repository._generate_text_key_index()
3522
def calculate_file_version_parents(self, text_key):
3523
"""Calculate the correct parents for a file version according to
3526
parent_keys = self.text_index[text_key]
3527
if parent_keys == [_mod_revision.NULL_REVISION]:
3529
return tuple(parent_keys)
3531
def check_file_version_parents(self, texts, progress_bar=None):
3532
"""Check the parents stored in a versioned file are correct.
3534
It also detects file versions that are not referenced by their
3535
corresponding revision's inventory.
3537
:returns: A tuple of (wrong_parents, dangling_file_versions).
3538
wrong_parents is a dict mapping {revision_id: (stored_parents,
3539
correct_parents)} for each revision_id where the stored parents
3540
are not correct. dangling_file_versions is a set of (file_id,
3541
revision_id) tuples for versions that are present in this versioned
3542
file, but not used by the corresponding inventory.
3545
self.file_ids = set([file_id for file_id, _ in
3546
self.text_index.iterkeys()])
3547
# text keys is now grouped by file_id
3548
n_weaves = len(self.file_ids)
3549
files_in_revisions = {}
3550
revisions_of_files = {}
3551
n_versions = len(self.text_index)
3552
progress_bar.update('loading text store', 0, n_versions)
3553
parent_map = self.repository.texts.get_parent_map(self.text_index)
3554
# On unlistable transports this could well be empty/error...
3555
text_keys = self.repository.texts.keys()
3556
unused_keys = frozenset(text_keys) - set(self.text_index)
3557
for num, key in enumerate(self.text_index.iterkeys()):
3558
if progress_bar is not None:
3559
progress_bar.update('checking text graph', num, n_versions)
3560
correct_parents = self.calculate_file_version_parents(key)
3562
knit_parents = parent_map[key]
3563
except errors.RevisionNotPresent:
3566
if correct_parents != knit_parents:
3567
wrong_parents[key] = (knit_parents, correct_parents)
3568
return wrong_parents, unused_keys
3571
def _old_get_graph(repository, revision_id):
3572
"""DO NOT USE. That is all. I'm serious."""
3573
graph = repository.get_graph()
3574
revision_graph = dict(((key, value) for key, value in
3575
graph.iter_ancestry([revision_id]) if value is not None))
3576
return _strip_NULL_ghosts(revision_graph)
3579
def _strip_NULL_ghosts(revision_graph):
3580
"""Also don't use this. more compatibility code for unmigrated clients."""
3581
# Filter ghosts, and null:
3582
if _mod_revision.NULL_REVISION in revision_graph:
3583
del revision_graph[_mod_revision.NULL_REVISION]
3584
for key, parents in revision_graph.items():
3585
revision_graph[key] = tuple(parent for parent in parents if parent
3587
return revision_graph
added
removed
bzrlib/repository.py
