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: The validator(which is a sha1 digest, though what is sha'd is
689
repository format specific) of the serialized inventory.
691
if not self.is_in_write_group():
692
raise AssertionError("%r not in write group" % (self,))
693
_mod_revision.check_not_reserved_id(new_revision_id)
694
basis_tree = self.revision_tree(basis_revision_id)
695
basis_tree.lock_read()
697
# Note that this mutates the inventory of basis_tree, which not all
698
# inventory implementations may support: A better idiom would be to
699
# return a new inventory, but as there is no revision tree cache in
700
# repository this is safe for now - RBC 20081013
701
basis_inv = basis_tree.inventory
702
basis_inv.apply_delta(delta)
703
basis_inv.revision_id = new_revision_id
704
return self.add_inventory(new_revision_id, basis_inv, parents)
708
def _inventory_add_lines(self, revision_id, parents, lines,
710
"""Store lines in inv_vf and return the sha1 of the inventory."""
711
parents = [(parent,) for parent in parents]
712
return self.inventories.add_lines((revision_id,), parents, lines,
713
check_content=check_content)[0]
715
def add_revision(self, revision_id, rev, inv=None, config=None):
716
"""Add rev to the revision store as revision_id.
718
:param revision_id: the revision id to use.
719
:param rev: The revision object.
720
:param inv: The inventory for the revision. if None, it will be looked
721
up in the inventory storer
722
:param config: If None no digital signature will be created.
723
If supplied its signature_needed method will be used
724
to determine if a signature should be made.
726
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
728
_mod_revision.check_not_reserved_id(revision_id)
729
if config is not None and config.signature_needed():
731
inv = self.get_inventory(revision_id)
732
plaintext = Testament(rev, inv).as_short_text()
733
self.store_revision_signature(
734
gpg.GPGStrategy(config), plaintext, revision_id)
735
# check inventory present
736
if not self.inventories.get_parent_map([(revision_id,)]):
738
raise errors.WeaveRevisionNotPresent(revision_id,
741
# yes, this is not suitable for adding with ghosts.
742
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
746
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
747
self._add_revision(rev)
749
def _add_revision(self, revision):
750
text = self._serializer.write_revision_to_string(revision)
751
key = (revision.revision_id,)
752
parents = tuple((parent,) for parent in revision.parent_ids)
753
self.revisions.add_lines(key, parents, osutils.split_lines(text))
755
def all_revision_ids(self):
756
"""Returns a list of all the revision ids in the repository.
758
This is conceptually deprecated because code should generally work on
759
the graph reachable from a particular revision, and ignore any other
760
revisions that might be present. There is no direct replacement
763
if 'evil' in debug.debug_flags:
764
mutter_callsite(2, "all_revision_ids is linear with history.")
765
return self._all_revision_ids()
767
def _all_revision_ids(self):
768
"""Returns a list of all the revision ids in the repository.
770
These are in as much topological order as the underlying store can
773
raise NotImplementedError(self._all_revision_ids)
775
def break_lock(self):
776
"""Break a lock if one is present from another instance.
778
Uses the ui factory to ask for confirmation if the lock may be from
781
self.control_files.break_lock()
784
def _eliminate_revisions_not_present(self, revision_ids):
785
"""Check every revision id in revision_ids to see if we have it.
787
Returns a set of the present revisions.
790
graph = self.get_graph()
791
parent_map = graph.get_parent_map(revision_ids)
792
# The old API returned a list, should this actually be a set?
793
return parent_map.keys()
796
def create(a_bzrdir):
797
"""Construct the current default format repository in a_bzrdir."""
798
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
800
def __init__(self, _format, a_bzrdir, control_files):
801
"""instantiate a Repository.
803
:param _format: The format of the repository on disk.
804
:param a_bzrdir: The BzrDir of the repository.
806
In the future we will have a single api for all stores for
807
getting file texts, inventories and revisions, then
808
this construct will accept instances of those things.
810
super(Repository, self).__init__()
811
self._format = _format
812
# the following are part of the public API for Repository:
813
self.bzrdir = a_bzrdir
814
self.control_files = control_files
815
self._transport = control_files._transport
816
self.base = self._transport.base
818
self._reconcile_does_inventory_gc = True
819
self._reconcile_fixes_text_parents = False
820
self._reconcile_backsup_inventory = True
821
# not right yet - should be more semantically clear ?
823
# TODO: make sure to construct the right store classes, etc, depending
824
# on whether escaping is required.
825
self._warn_if_deprecated()
826
self._write_group = None
827
# Additional places to query for data.
828
self._fallback_repositories = []
829
# What order should fetch operations request streams in?
830
# The default is unordered as that is the cheapest for an origin to
832
self._fetch_order = 'unordered'
833
# Does this repository use deltas that can be fetched as-deltas ?
834
# (E.g. knits, where the knit deltas can be transplanted intact.
835
# We default to False, which will ensure that enough data to get
836
# a full text out of any fetch stream will be grabbed.
837
self._fetch_uses_deltas = False
838
# Should fetch trigger a reconcile after the fetch? Only needed for
839
# some repository formats that can suffer internal inconsistencies.
840
self._fetch_reconcile = False
843
return '%s(%r)' % (self.__class__.__name__,
846
def has_same_location(self, other):
847
"""Returns a boolean indicating if this repository is at the same
848
location as another repository.
850
This might return False even when two repository objects are accessing
851
the same physical repository via different URLs.
853
if self.__class__ is not other.__class__:
855
return (self._transport.base == other._transport.base)
857
def is_in_write_group(self):
858
"""Return True if there is an open write group.
860
:seealso: start_write_group.
862
return self._write_group is not None
865
return self.control_files.is_locked()
867
def is_write_locked(self):
868
"""Return True if this object is write locked."""
869
return self.is_locked() and self.control_files._lock_mode == 'w'
871
def lock_write(self, token=None):
872
"""Lock this repository for writing.
874
This causes caching within the repository obejct to start accumlating
875
data during reads, and allows a 'write_group' to be obtained. Write
876
groups must be used for actual data insertion.
878
:param token: if this is already locked, then lock_write will fail
879
unless the token matches the existing lock.
880
:returns: a token if this instance supports tokens, otherwise None.
881
:raises TokenLockingNotSupported: when a token is given but this
882
instance doesn't support using token locks.
883
:raises MismatchedToken: if the specified token doesn't match the token
884
of the existing lock.
885
:seealso: start_write_group.
887
A token should be passed in if you know that you have locked the object
888
some other way, and need to synchronise this object's state with that
891
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
893
result = self.control_files.lock_write(token=token)
894
for repo in self._fallback_repositories:
895
# Writes don't affect fallback repos
901
self.control_files.lock_read()
902
for repo in self._fallback_repositories:
906
def get_physical_lock_status(self):
907
return self.control_files.get_physical_lock_status()
909
def leave_lock_in_place(self):
910
"""Tell this repository not to release the physical lock when this
913
If lock_write doesn't return a token, then this method is not supported.
915
self.control_files.leave_in_place()
917
def dont_leave_lock_in_place(self):
918
"""Tell this repository to release the physical lock when this
919
object is unlocked, even if it didn't originally acquire it.
921
If lock_write doesn't return a token, then this method is not supported.
923
self.control_files.dont_leave_in_place()
926
def gather_stats(self, revid=None, committers=None):
927
"""Gather statistics from a revision id.
929
:param revid: The revision id to gather statistics from, if None, then
930
no revision specific statistics are gathered.
931
:param committers: Optional parameter controlling whether to grab
932
a count of committers from the revision specific statistics.
933
:return: A dictionary of statistics. Currently this contains:
934
committers: The number of committers if requested.
935
firstrev: A tuple with timestamp, timezone for the penultimate left
936
most ancestor of revid, if revid is not the NULL_REVISION.
937
latestrev: A tuple with timestamp, timezone for revid, if revid is
938
not the NULL_REVISION.
939
revisions: The total revision count in the repository.
940
size: An estimate disk size of the repository in bytes.
943
if revid and committers:
944
result['committers'] = 0
945
if revid and revid != _mod_revision.NULL_REVISION:
947
all_committers = set()
948
revisions = self.get_ancestry(revid)
949
# pop the leading None
951
first_revision = None
953
# ignore the revisions in the middle - just grab first and last
954
revisions = revisions[0], revisions[-1]
955
for revision in self.get_revisions(revisions):
956
if not first_revision:
957
first_revision = revision
959
all_committers.add(revision.committer)
960
last_revision = revision
962
result['committers'] = len(all_committers)
963
result['firstrev'] = (first_revision.timestamp,
964
first_revision.timezone)
965
result['latestrev'] = (last_revision.timestamp,
966
last_revision.timezone)
968
# now gather global repository information
969
# XXX: This is available for many repos regardless of listability.
970
if self.bzrdir.root_transport.listable():
971
# XXX: do we want to __define len__() ?
972
# Maybe the versionedfiles object should provide a different
973
# method to get the number of keys.
974
result['revisions'] = len(self.revisions.keys())
978
def find_branches(self, using=False):
979
"""Find branches underneath this repository.
981
This will include branches inside other branches.
983
:param using: If True, list only branches using this repository.
985
if using and not self.is_shared():
987
return [self.bzrdir.open_branch()]
988
except errors.NotBranchError:
990
class Evaluator(object):
993
self.first_call = True
995
def __call__(self, bzrdir):
996
# On the first call, the parameter is always the bzrdir
997
# containing the current repo.
998
if not self.first_call:
1000
repository = bzrdir.open_repository()
1001
except errors.NoRepositoryPresent:
1004
return False, (None, repository)
1005
self.first_call = False
1007
value = (bzrdir.open_branch(), None)
1008
except errors.NotBranchError:
1009
value = (None, None)
1013
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1014
self.bzrdir.root_transport, evaluate=Evaluator()):
1015
if branch is not None:
1016
branches.append(branch)
1017
if not using and repository is not None:
1018
branches.extend(repository.find_branches())
1022
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1023
"""Return the revision ids that other has that this does not.
1025
These are returned in topological order.
1027
revision_id: only return revision ids included by revision_id.
1029
return InterRepository.get(other, self).search_missing_revision_ids(
1030
revision_id, find_ghosts)
1032
@deprecated_method(one_two)
1034
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1035
"""Return the revision ids that other has that this does not.
1037
These are returned in topological order.
1039
revision_id: only return revision ids included by revision_id.
1041
keys = self.search_missing_revision_ids(
1042
other, revision_id, find_ghosts).get_keys()
1045
parents = other.get_graph().get_parent_map(keys)
1048
return tsort.topo_sort(parents)
1052
"""Open the repository rooted at base.
1054
For instance, if the repository is at URL/.bzr/repository,
1055
Repository.open(URL) -> a Repository instance.
1057
control = bzrdir.BzrDir.open(base)
1058
return control.open_repository()
1060
def copy_content_into(self, destination, revision_id=None):
1061
"""Make a complete copy of the content in self into destination.
1063
This is a destructive operation! Do not use it on existing
1066
return InterRepository.get(self, destination).copy_content(revision_id)
1068
def commit_write_group(self):
1069
"""Commit the contents accrued within the current write group.
1071
:seealso: start_write_group.
1073
if self._write_group is not self.get_transaction():
1074
# has an unlock or relock occured ?
1075
raise errors.BzrError('mismatched lock context %r and '
1077
(self.get_transaction(), self._write_group))
1078
self._commit_write_group()
1079
self._write_group = None
1081
def _commit_write_group(self):
1082
"""Template method for per-repository write group cleanup.
1084
This is called before the write group is considered to be
1085
finished and should ensure that all data handed to the repository
1086
for writing during the write group is safely committed (to the
1087
extent possible considering file system caching etc).
1090
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1091
"""Fetch the content required to construct revision_id from source.
1093
If revision_id is None all content is copied.
1094
:param find_ghosts: Find and copy revisions in the source that are
1095
ghosts in the target (and not reachable directly by walking out to
1096
the first-present revision in target from revision_id).
1098
# fast path same-url fetch operations
1099
if self.has_same_location(source):
1100
# check that last_revision is in 'from' and then return a
1102
if (revision_id is not None and
1103
not _mod_revision.is_null(revision_id)):
1104
self.get_revision(revision_id)
1106
# if there is no specific appropriate InterRepository, this will get
1107
# the InterRepository base class, which raises an
1108
# IncompatibleRepositories when asked to fetch.
1109
inter = InterRepository.get(source, self)
1110
return inter.fetch(revision_id=revision_id, pb=pb,
1111
find_ghosts=find_ghosts)
1113
def create_bundle(self, target, base, fileobj, format=None):
1114
return serializer.write_bundle(self, target, base, fileobj, format)
1116
def get_commit_builder(self, branch, parents, config, timestamp=None,
1117
timezone=None, committer=None, revprops=None,
1119
"""Obtain a CommitBuilder for this repository.
1121
:param branch: Branch to commit to.
1122
:param parents: Revision ids of the parents of the new revision.
1123
:param config: Configuration to use.
1124
:param timestamp: Optional timestamp recorded for commit.
1125
:param timezone: Optional timezone for timestamp.
1126
:param committer: Optional committer to set for commit.
1127
:param revprops: Optional dictionary of revision properties.
1128
:param revision_id: Optional revision id.
1130
result = self._commit_builder_class(self, parents, config,
1131
timestamp, timezone, committer, revprops, revision_id)
1132
self.start_write_group()
1136
if (self.control_files._lock_count == 1 and
1137
self.control_files._lock_mode == 'w'):
1138
if self._write_group is not None:
1139
self.abort_write_group()
1140
self.control_files.unlock()
1141
raise errors.BzrError(
1142
'Must end write groups before releasing write locks.')
1143
self.control_files.unlock()
1144
for repo in self._fallback_repositories:
1148
def clone(self, a_bzrdir, revision_id=None):
1149
"""Clone this repository into a_bzrdir using the current format.
1151
Currently no check is made that the format of this repository and
1152
the bzrdir format are compatible. FIXME RBC 20060201.
1154
:return: The newly created destination repository.
1156
# TODO: deprecate after 0.16; cloning this with all its settings is
1157
# probably not very useful -- mbp 20070423
1158
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1159
self.copy_content_into(dest_repo, revision_id)
1162
def start_write_group(self):
1163
"""Start a write group in the repository.
1165
Write groups are used by repositories which do not have a 1:1 mapping
1166
between file ids and backend store to manage the insertion of data from
1167
both fetch and commit operations.
1169
A write lock is required around the start_write_group/commit_write_group
1170
for the support of lock-requiring repository formats.
1172
One can only insert data into a repository inside a write group.
1176
if not self.is_write_locked():
1177
raise errors.NotWriteLocked(self)
1178
if self._write_group:
1179
raise errors.BzrError('already in a write group')
1180
self._start_write_group()
1181
# so we can detect unlock/relock - the write group is now entered.
1182
self._write_group = self.get_transaction()
1184
def _start_write_group(self):
1185
"""Template method for per-repository write group startup.
1187
This is called before the write group is considered to be
1192
def sprout(self, to_bzrdir, revision_id=None):
1193
"""Create a descendent repository for new development.
1195
Unlike clone, this does not copy the settings of the repository.
1197
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1198
dest_repo.fetch(self, revision_id=revision_id)
1201
def _create_sprouting_repo(self, a_bzrdir, shared):
1202
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1203
# use target default format.
1204
dest_repo = a_bzrdir.create_repository()
1206
# Most control formats need the repository to be specifically
1207
# created, but on some old all-in-one formats it's not needed
1209
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1210
except errors.UninitializableFormat:
1211
dest_repo = a_bzrdir.open_repository()
1215
def has_revision(self, revision_id):
1216
"""True if this repository has a copy of the revision."""
1217
return revision_id in self.has_revisions((revision_id,))
1220
def has_revisions(self, revision_ids):
1221
"""Probe to find out the presence of multiple revisions.
1223
:param revision_ids: An iterable of revision_ids.
1224
:return: A set of the revision_ids that were present.
1226
parent_map = self.revisions.get_parent_map(
1227
[(rev_id,) for rev_id in revision_ids])
1229
if _mod_revision.NULL_REVISION in revision_ids:
1230
result.add(_mod_revision.NULL_REVISION)
1231
result.update([key[0] for key in parent_map])
1235
def get_revision(self, revision_id):
1236
"""Return the Revision object for a named revision."""
1237
return self.get_revisions([revision_id])[0]
1240
def get_revision_reconcile(self, revision_id):
1241
"""'reconcile' helper routine that allows access to a revision always.
1243
This variant of get_revision does not cross check the weave graph
1244
against the revision one as get_revision does: but it should only
1245
be used by reconcile, or reconcile-alike commands that are correcting
1246
or testing the revision graph.
1248
return self._get_revisions([revision_id])[0]
1251
def get_revisions(self, revision_ids):
1252
"""Get many revisions at once."""
1253
return self._get_revisions(revision_ids)
1256
def _get_revisions(self, revision_ids):
1257
"""Core work logic to get many revisions without sanity checks."""
1258
for rev_id in revision_ids:
1259
if not rev_id or not isinstance(rev_id, basestring):
1260
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1261
keys = [(key,) for key in revision_ids]
1262
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1264
for record in stream:
1265
if record.storage_kind == 'absent':
1266
raise errors.NoSuchRevision(self, record.key[0])
1267
text = record.get_bytes_as('fulltext')
1268
rev = self._serializer.read_revision_from_string(text)
1269
revs[record.key[0]] = rev
1270
return [revs[revid] for revid in revision_ids]
1273
def get_revision_xml(self, revision_id):
1274
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1275
# would have already do it.
1276
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1277
rev = self.get_revision(revision_id)
1278
rev_tmp = cStringIO.StringIO()
1279
# the current serializer..
1280
self._serializer.write_revision(rev, rev_tmp)
1282
return rev_tmp.getvalue()
1284
def get_deltas_for_revisions(self, revisions):
1285
"""Produce a generator of revision deltas.
1287
Note that the input is a sequence of REVISIONS, not revision_ids.
1288
Trees will be held in memory until the generator exits.
1289
Each delta is relative to the revision's lefthand predecessor.
1291
required_trees = set()
1292
for revision in revisions:
1293
required_trees.add(revision.revision_id)
1294
required_trees.update(revision.parent_ids[:1])
1295
trees = dict((t.get_revision_id(), t) for
1296
t in self.revision_trees(required_trees))
1297
for revision in revisions:
1298
if not revision.parent_ids:
1299
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1301
old_tree = trees[revision.parent_ids[0]]
1302
yield trees[revision.revision_id].changes_from(old_tree)
1305
def get_revision_delta(self, revision_id):
1306
"""Return the delta for one revision.
1308
The delta is relative to the left-hand predecessor of the
1311
r = self.get_revision(revision_id)
1312
return list(self.get_deltas_for_revisions([r]))[0]
1315
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1316
signature = gpg_strategy.sign(plaintext)
1317
self.add_signature_text(revision_id, signature)
1320
def add_signature_text(self, revision_id, signature):
1321
self.signatures.add_lines((revision_id,), (),
1322
osutils.split_lines(signature))
1324
def find_text_key_references(self):
1325
"""Find the text key references within the repository.
1327
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1328
revision_ids. Each altered file-ids has the exact revision_ids that
1329
altered it listed explicitly.
1330
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1331
to whether they were referred to by the inventory of the
1332
revision_id that they contain. The inventory texts from all present
1333
revision ids are assessed to generate this report.
1335
revision_keys = self.revisions.keys()
1336
w = self.inventories
1337
pb = ui.ui_factory.nested_progress_bar()
1339
return self._find_text_key_references_from_xml_inventory_lines(
1340
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1344
def _find_text_key_references_from_xml_inventory_lines(self,
1346
"""Core routine for extracting references to texts from inventories.
1348
This performs the translation of xml lines to revision ids.
1350
:param line_iterator: An iterator of lines, origin_version_id
1351
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1352
to whether they were referred to by the inventory of the
1353
revision_id that they contain. Note that if that revision_id was
1354
not part of the line_iterator's output then False will be given -
1355
even though it may actually refer to that key.
1357
if not self._serializer.support_altered_by_hack:
1358
raise AssertionError(
1359
"_find_text_key_references_from_xml_inventory_lines only "
1360
"supported for branches which store inventory as unnested xml"
1361
", not on %r" % self)
1364
# this code needs to read every new line in every inventory for the
1365
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1366
# not present in one of those inventories is unnecessary but not
1367
# harmful because we are filtering by the revision id marker in the
1368
# inventory lines : we only select file ids altered in one of those
1369
# revisions. We don't need to see all lines in the inventory because
1370
# only those added in an inventory in rev X can contain a revision=X
1372
unescape_revid_cache = {}
1373
unescape_fileid_cache = {}
1375
# jam 20061218 In a big fetch, this handles hundreds of thousands
1376
# of lines, so it has had a lot of inlining and optimizing done.
1377
# Sorry that it is a little bit messy.
1378
# Move several functions to be local variables, since this is a long
1380
search = self._file_ids_altered_regex.search
1381
unescape = _unescape_xml
1382
setdefault = result.setdefault
1383
for line, line_key in line_iterator:
1384
match = search(line)
1387
# One call to match.group() returning multiple items is quite a
1388
# bit faster than 2 calls to match.group() each returning 1
1389
file_id, revision_id = match.group('file_id', 'revision_id')
1391
# Inlining the cache lookups helps a lot when you make 170,000
1392
# lines and 350k ids, versus 8.4 unique ids.
1393
# Using a cache helps in 2 ways:
1394
# 1) Avoids unnecessary decoding calls
1395
# 2) Re-uses cached strings, which helps in future set and
1397
# (2) is enough that removing encoding entirely along with
1398
# the cache (so we are using plain strings) results in no
1399
# performance improvement.
1401
revision_id = unescape_revid_cache[revision_id]
1403
unescaped = unescape(revision_id)
1404
unescape_revid_cache[revision_id] = unescaped
1405
revision_id = unescaped
1407
# Note that unconditionally unescaping means that we deserialise
1408
# every fileid, which for general 'pull' is not great, but we don't
1409
# really want to have some many fulltexts that this matters anyway.
1412
file_id = unescape_fileid_cache[file_id]
1414
unescaped = unescape(file_id)
1415
unescape_fileid_cache[file_id] = unescaped
1418
key = (file_id, revision_id)
1419
setdefault(key, False)
1420
if revision_id == line_key[-1]:
1424
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1426
"""Helper routine for fileids_altered_by_revision_ids.
1428
This performs the translation of xml lines to revision ids.
1430
:param line_iterator: An iterator of lines, origin_version_id
1431
:param revision_ids: The revision ids to filter for. This should be a
1432
set or other type which supports efficient __contains__ lookups, as
1433
the revision id from each parsed line will be looked up in the
1434
revision_ids filter.
1435
:return: a dictionary mapping altered file-ids to an iterable of
1436
revision_ids. Each altered file-ids has the exact revision_ids that
1437
altered it listed explicitly.
1440
setdefault = result.setdefault
1442
self._find_text_key_references_from_xml_inventory_lines(
1443
line_iterator).iterkeys():
1444
# once data is all ensured-consistent; then this is
1445
# if revision_id == version_id
1446
if key[-1:] in revision_ids:
1447
setdefault(key[0], set()).add(key[-1])
1450
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1451
"""Find the file ids and versions affected by revisions.
1453
:param revisions: an iterable containing revision ids.
1454
:param _inv_weave: The inventory weave from this repository or None.
1455
If None, the inventory weave will be opened automatically.
1456
:return: a dictionary mapping altered file-ids to an iterable of
1457
revision_ids. Each altered file-ids has the exact revision_ids that
1458
altered it listed explicitly.
1460
selected_keys = set((revid,) for revid in revision_ids)
1461
w = _inv_weave or self.inventories
1462
pb = ui.ui_factory.nested_progress_bar()
1464
return self._find_file_ids_from_xml_inventory_lines(
1465
w.iter_lines_added_or_present_in_keys(
1466
selected_keys, pb=pb),
1471
def iter_files_bytes(self, desired_files):
1472
"""Iterate through file versions.
1474
Files will not necessarily be returned in the order they occur in
1475
desired_files. No specific order is guaranteed.
1477
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1478
value supplied by the caller as part of desired_files. It should
1479
uniquely identify the file version in the caller's context. (Examples:
1480
an index number or a TreeTransform trans_id.)
1482
bytes_iterator is an iterable of bytestrings for the file. The
1483
kind of iterable and length of the bytestrings are unspecified, but for
1484
this implementation, it is a list of bytes produced by
1485
VersionedFile.get_record_stream().
1487
:param desired_files: a list of (file_id, revision_id, identifier)
1490
transaction = self.get_transaction()
1492
for file_id, revision_id, callable_data in desired_files:
1493
text_keys[(file_id, revision_id)] = callable_data
1494
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1495
if record.storage_kind == 'absent':
1496
raise errors.RevisionNotPresent(record.key, self)
1497
yield text_keys[record.key], record.get_bytes_as('fulltext')
1499
def _generate_text_key_index(self, text_key_references=None,
1501
"""Generate a new text key index for the repository.
1503
This is an expensive function that will take considerable time to run.
1505
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1506
list of parents, also text keys. When a given key has no parents,
1507
the parents list will be [NULL_REVISION].
1509
# All revisions, to find inventory parents.
1510
if ancestors is None:
1511
graph = self.get_graph()
1512
ancestors = graph.get_parent_map(self.all_revision_ids())
1513
if text_key_references is None:
1514
text_key_references = self.find_text_key_references()
1515
pb = ui.ui_factory.nested_progress_bar()
1517
return self._do_generate_text_key_index(ancestors,
1518
text_key_references, pb)
1522
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1523
"""Helper for _generate_text_key_index to avoid deep nesting."""
1524
revision_order = tsort.topo_sort(ancestors)
1525
invalid_keys = set()
1527
for revision_id in revision_order:
1528
revision_keys[revision_id] = set()
1529
text_count = len(text_key_references)
1530
# a cache of the text keys to allow reuse; costs a dict of all the
1531
# keys, but saves a 2-tuple for every child of a given key.
1533
for text_key, valid in text_key_references.iteritems():
1535
invalid_keys.add(text_key)
1537
revision_keys[text_key[1]].add(text_key)
1538
text_key_cache[text_key] = text_key
1539
del text_key_references
1541
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1542
NULL_REVISION = _mod_revision.NULL_REVISION
1543
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1544
# too small for large or very branchy trees. However, for 55K path
1545
# trees, it would be easy to use too much memory trivially. Ideally we
1546
# could gauge this by looking at available real memory etc, but this is
1547
# always a tricky proposition.
1548
inventory_cache = lru_cache.LRUCache(10)
1549
batch_size = 10 # should be ~150MB on a 55K path tree
1550
batch_count = len(revision_order) / batch_size + 1
1552
pb.update("Calculating text parents.", processed_texts, text_count)
1553
for offset in xrange(batch_count):
1554
to_query = revision_order[offset * batch_size:(offset + 1) *
1558
for rev_tree in self.revision_trees(to_query):
1559
revision_id = rev_tree.get_revision_id()
1560
parent_ids = ancestors[revision_id]
1561
for text_key in revision_keys[revision_id]:
1562
pb.update("Calculating text parents.", processed_texts)
1563
processed_texts += 1
1564
candidate_parents = []
1565
for parent_id in parent_ids:
1566
parent_text_key = (text_key[0], parent_id)
1568
check_parent = parent_text_key not in \
1569
revision_keys[parent_id]
1571
# the parent parent_id is a ghost:
1572
check_parent = False
1573
# truncate the derived graph against this ghost.
1574
parent_text_key = None
1576
# look at the parent commit details inventories to
1577
# determine possible candidates in the per file graph.
1580
inv = inventory_cache[parent_id]
1582
inv = self.revision_tree(parent_id).inventory
1583
inventory_cache[parent_id] = inv
1584
parent_entry = inv._byid.get(text_key[0], None)
1585
if parent_entry is not None:
1587
text_key[0], parent_entry.revision)
1589
parent_text_key = None
1590
if parent_text_key is not None:
1591
candidate_parents.append(
1592
text_key_cache[parent_text_key])
1593
parent_heads = text_graph.heads(candidate_parents)
1594
new_parents = list(parent_heads)
1595
new_parents.sort(key=lambda x:candidate_parents.index(x))
1596
if new_parents == []:
1597
new_parents = [NULL_REVISION]
1598
text_index[text_key] = new_parents
1600
for text_key in invalid_keys:
1601
text_index[text_key] = [NULL_REVISION]
1604
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1605
"""Get an iterable listing the keys of all the data introduced by a set
1608
The keys will be ordered so that the corresponding items can be safely
1609
fetched and inserted in that order.
1611
:returns: An iterable producing tuples of (knit-kind, file-id,
1612
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1613
'revisions'. file-id is None unless knit-kind is 'file'.
1615
# XXX: it's a bit weird to control the inventory weave caching in this
1616
# generator. Ideally the caching would be done in fetch.py I think. Or
1617
# maybe this generator should explicitly have the contract that it
1618
# should not be iterated until the previously yielded item has been
1620
inv_w = self.inventories
1622
# file ids that changed
1623
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1625
num_file_ids = len(file_ids)
1626
for file_id, altered_versions in file_ids.iteritems():
1627
if _files_pb is not None:
1628
_files_pb.update("fetch texts", count, num_file_ids)
1630
yield ("file", file_id, altered_versions)
1631
# We're done with the files_pb. Note that it finished by the caller,
1632
# just as it was created by the caller.
1636
yield ("inventory", None, revision_ids)
1639
# XXX: Note ATM no callers actually pay attention to this return
1640
# instead they just use the list of revision ids and ignore
1641
# missing sigs. Consider removing this work entirely
1642
revisions_with_signatures = set(self.signatures.get_parent_map(
1643
[(r,) for r in revision_ids]))
1644
revisions_with_signatures = set(
1645
[r for (r,) in revisions_with_signatures])
1646
revisions_with_signatures.intersection_update(revision_ids)
1647
yield ("signatures", None, revisions_with_signatures)
1650
yield ("revisions", None, revision_ids)
1653
def get_inventory(self, revision_id):
1654
"""Get Inventory object by revision id."""
1655
return self.iter_inventories([revision_id]).next()
1657
def iter_inventories(self, revision_ids):
1658
"""Get many inventories by revision_ids.
1660
This will buffer some or all of the texts used in constructing the
1661
inventories in memory, but will only parse a single inventory at a
1664
:return: An iterator of inventories.
1666
if ((None in revision_ids)
1667
or (_mod_revision.NULL_REVISION in revision_ids)):
1668
raise ValueError('cannot get null revision inventory')
1669
return self._iter_inventories(revision_ids)
1671
def _iter_inventories(self, revision_ids):
1672
"""single-document based inventory iteration."""
1673
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1674
yield self.deserialise_inventory(revision_id, text)
1676
def _iter_inventory_xmls(self, revision_ids):
1677
keys = [(revision_id,) for revision_id in revision_ids]
1678
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1680
for record in stream:
1681
if record.storage_kind != 'absent':
1682
texts[record.key] = record.get_bytes_as('fulltext')
1684
raise errors.NoSuchRevision(self, record.key)
1686
yield texts[key], key[-1]
1688
def deserialise_inventory(self, revision_id, xml):
1689
"""Transform the xml into an inventory object.
1691
:param revision_id: The expected revision id of the inventory.
1692
:param xml: A serialised inventory.
1694
result = self._serializer.read_inventory_from_string(xml, revision_id)
1695
if result.revision_id != revision_id:
1696
raise AssertionError('revision id mismatch %s != %s' % (
1697
result.revision_id, revision_id))
1700
def serialise_inventory(self, inv):
1701
return self._serializer.write_inventory_to_string(inv)
1703
def _serialise_inventory_to_lines(self, inv):
1704
return self._serializer.write_inventory_to_lines(inv)
1706
def get_serializer_format(self):
1707
return self._serializer.format_num
1710
def get_inventory_xml(self, revision_id):
1711
"""Get inventory XML as a file object."""
1712
texts = self._iter_inventory_xmls([revision_id])
1714
text, revision_id = texts.next()
1715
except StopIteration:
1716
raise errors.HistoryMissing(self, 'inventory', revision_id)
1720
def get_inventory_sha1(self, revision_id):
1721
"""Return the sha1 hash of the inventory entry
1723
return self.get_revision(revision_id).inventory_sha1
1725
def iter_reverse_revision_history(self, revision_id):
1726
"""Iterate backwards through revision ids in the lefthand history
1728
:param revision_id: The revision id to start with. All its lefthand
1729
ancestors will be traversed.
1731
graph = self.get_graph()
1732
next_id = revision_id
1734
if next_id in (None, _mod_revision.NULL_REVISION):
1737
# Note: The following line may raise KeyError in the event of
1738
# truncated history. We decided not to have a try:except:raise
1739
# RevisionNotPresent here until we see a use for it, because of the
1740
# cost in an inner loop that is by its very nature O(history).
1741
# Robert Collins 20080326
1742
parents = graph.get_parent_map([next_id])[next_id]
1743
if len(parents) == 0:
1746
next_id = parents[0]
1749
def get_revision_inventory(self, revision_id):
1750
"""Return inventory of a past revision."""
1751
# TODO: Unify this with get_inventory()
1752
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1753
# must be the same as its revision, so this is trivial.
1754
if revision_id is None:
1755
# This does not make sense: if there is no revision,
1756
# then it is the current tree inventory surely ?!
1757
# and thus get_root_id() is something that looks at the last
1758
# commit on the branch, and the get_root_id is an inventory check.
1759
raise NotImplementedError
1760
# return Inventory(self.get_root_id())
1762
return self.get_inventory(revision_id)
1764
def is_shared(self):
1765
"""Return True if this repository is flagged as a shared repository."""
1766
raise NotImplementedError(self.is_shared)
1769
def reconcile(self, other=None, thorough=False):
1770
"""Reconcile this repository."""
1771
from bzrlib.reconcile import RepoReconciler
1772
reconciler = RepoReconciler(self, thorough=thorough)
1773
reconciler.reconcile()
1776
def _refresh_data(self):
1777
"""Helper called from lock_* to ensure coherency with disk.
1779
The default implementation does nothing; it is however possible
1780
for repositories to maintain loaded indices across multiple locks
1781
by checking inside their implementation of this method to see
1782
whether their indices are still valid. This depends of course on
1783
the disk format being validatable in this manner.
1787
def revision_tree(self, revision_id):
1788
"""Return Tree for a revision on this branch.
1790
`revision_id` may be NULL_REVISION for the empty tree revision.
1792
revision_id = _mod_revision.ensure_null(revision_id)
1793
# TODO: refactor this to use an existing revision object
1794
# so we don't need to read it in twice.
1795
if revision_id == _mod_revision.NULL_REVISION:
1796
return RevisionTree(self, Inventory(root_id=None),
1797
_mod_revision.NULL_REVISION)
1799
inv = self.get_revision_inventory(revision_id)
1800
return RevisionTree(self, inv, revision_id)
1802
def revision_trees(self, revision_ids):
1803
"""Return Tree for a revision on this branch.
1805
`revision_id` may not be None or 'null:'"""
1806
inventories = self.iter_inventories(revision_ids)
1807
for inv in inventories:
1808
yield RevisionTree(self, inv, inv.revision_id)
1811
def get_ancestry(self, revision_id, topo_sorted=True):
1812
"""Return a list of revision-ids integrated by a revision.
1814
The first element of the list is always None, indicating the origin
1815
revision. This might change when we have history horizons, or
1816
perhaps we should have a new API.
1818
This is topologically sorted.
1820
if _mod_revision.is_null(revision_id):
1822
if not self.has_revision(revision_id):
1823
raise errors.NoSuchRevision(self, revision_id)
1824
graph = self.get_graph()
1826
search = graph._make_breadth_first_searcher([revision_id])
1829
found, ghosts = search.next_with_ghosts()
1830
except StopIteration:
1833
if _mod_revision.NULL_REVISION in keys:
1834
keys.remove(_mod_revision.NULL_REVISION)
1836
parent_map = graph.get_parent_map(keys)
1837
keys = tsort.topo_sort(parent_map)
1838
return [None] + list(keys)
1841
"""Compress the data within the repository.
1843
This operation only makes sense for some repository types. For other
1844
types it should be a no-op that just returns.
1846
This stub method does not require a lock, but subclasses should use
1847
@needs_write_lock as this is a long running call its reasonable to
1848
implicitly lock for the user.
1852
@deprecated_method(one_six)
1853
def print_file(self, file, revision_id):
1854
"""Print `file` to stdout.
1856
FIXME RBC 20060125 as John Meinel points out this is a bad api
1857
- it writes to stdout, it assumes that that is valid etc. Fix
1858
by creating a new more flexible convenience function.
1860
tree = self.revision_tree(revision_id)
1861
# use inventory as it was in that revision
1862
file_id = tree.inventory.path2id(file)
1864
# TODO: jam 20060427 Write a test for this code path
1865
# it had a bug in it, and was raising the wrong
1867
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1868
tree.print_file(file_id)
1870
def get_transaction(self):
1871
return self.control_files.get_transaction()
1873
@deprecated_method(one_one)
1874
def get_parents(self, revision_ids):
1875
"""See StackedParentsProvider.get_parents"""
1876
parent_map = self.get_parent_map(revision_ids)
1877
return [parent_map.get(r, None) for r in revision_ids]
1879
def get_parent_map(self, revision_ids):
1880
"""See graph._StackedParentsProvider.get_parent_map"""
1881
# revisions index works in keys; this just works in revisions
1882
# therefore wrap and unwrap
1885
for revision_id in revision_ids:
1886
if revision_id == _mod_revision.NULL_REVISION:
1887
result[revision_id] = ()
1888
elif revision_id is None:
1889
raise ValueError('get_parent_map(None) is not valid')
1891
query_keys.append((revision_id ,))
1892
for ((revision_id,), parent_keys) in \
1893
self.revisions.get_parent_map(query_keys).iteritems():
1895
result[revision_id] = tuple(parent_revid
1896
for (parent_revid,) in parent_keys)
1898
result[revision_id] = (_mod_revision.NULL_REVISION,)
1901
def _make_parents_provider(self):
1904
def get_graph(self, other_repository=None):
1905
"""Return the graph walker for this repository format"""
1906
parents_provider = self._make_parents_provider()
1907
if (other_repository is not None and
1908
not self.has_same_location(other_repository)):
1909
parents_provider = graph._StackedParentsProvider(
1910
[parents_provider, other_repository._make_parents_provider()])
1911
return graph.Graph(parents_provider)
1913
def _get_versioned_file_checker(self):
1914
"""Return an object suitable for checking versioned files."""
1915
return _VersionedFileChecker(self)
1917
def revision_ids_to_search_result(self, result_set):
1918
"""Convert a set of revision ids to a graph SearchResult."""
1919
result_parents = set()
1920
for parents in self.get_graph().get_parent_map(
1921
result_set).itervalues():
1922
result_parents.update(parents)
1923
included_keys = result_set.intersection(result_parents)
1924
start_keys = result_set.difference(included_keys)
1925
exclude_keys = result_parents.difference(result_set)
1926
result = graph.SearchResult(start_keys, exclude_keys,
1927
len(result_set), result_set)
1931
def set_make_working_trees(self, new_value):
1932
"""Set the policy flag for making working trees when creating branches.
1934
This only applies to branches that use this repository.
1936
The default is 'True'.
1937
:param new_value: True to restore the default, False to disable making
1940
raise NotImplementedError(self.set_make_working_trees)
1942
def make_working_trees(self):
1943
"""Returns the policy for making working trees on new branches."""
1944
raise NotImplementedError(self.make_working_trees)
1947
def sign_revision(self, revision_id, gpg_strategy):
1948
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1949
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1952
def has_signature_for_revision_id(self, revision_id):
1953
"""Query for a revision signature for revision_id in the repository."""
1954
if not self.has_revision(revision_id):
1955
raise errors.NoSuchRevision(self, revision_id)
1956
sig_present = (1 == len(
1957
self.signatures.get_parent_map([(revision_id,)])))
1961
def get_signature_text(self, revision_id):
1962
"""Return the text for a signature."""
1963
stream = self.signatures.get_record_stream([(revision_id,)],
1965
record = stream.next()
1966
if record.storage_kind == 'absent':
1967
raise errors.NoSuchRevision(self, revision_id)
1968
return record.get_bytes_as('fulltext')
1971
def check(self, revision_ids=None):
1972
"""Check consistency of all history of given revision_ids.
1974
Different repository implementations should override _check().
1976
:param revision_ids: A non-empty list of revision_ids whose ancestry
1977
will be checked. Typically the last revision_id of a branch.
1979
return self._check(revision_ids)
1981
def _check(self, revision_ids):
1982
result = check.Check(self)
1986
def _warn_if_deprecated(self):
1987
global _deprecation_warning_done
1988
if _deprecation_warning_done:
1990
_deprecation_warning_done = True
1991
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1992
% (self._format, self.bzrdir.transport.base))
1994
def supports_rich_root(self):
1995
return self._format.rich_root_data
1997
def _check_ascii_revisionid(self, revision_id, method):
1998
"""Private helper for ascii-only repositories."""
1999
# weave repositories refuse to store revisionids that are non-ascii.
2000
if revision_id is not None:
2001
# weaves require ascii revision ids.
2002
if isinstance(revision_id, unicode):
2004
revision_id.encode('ascii')
2005
except UnicodeEncodeError:
2006
raise errors.NonAsciiRevisionId(method, self)
2009
revision_id.decode('ascii')
2010
except UnicodeDecodeError:
2011
raise errors.NonAsciiRevisionId(method, self)
2013
def revision_graph_can_have_wrong_parents(self):
2014
"""Is it possible for this repository to have a revision graph with
2017
If True, then this repository must also implement
2018
_find_inconsistent_revision_parents so that check and reconcile can
2019
check for inconsistencies before proceeding with other checks that may
2020
depend on the revision index being consistent.
2022
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2025
# remove these delegates a while after bzr 0.15
2026
def __make_delegated(name, from_module):
2027
def _deprecated_repository_forwarder():
2028
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2029
% (name, from_module),
2032
m = __import__(from_module, globals(), locals(), [name])
2034
return getattr(m, name)
2035
except AttributeError:
2036
raise AttributeError('module %s has no name %s'
2038
globals()[name] = _deprecated_repository_forwarder
2041
'AllInOneRepository',
2042
'WeaveMetaDirRepository',
2043
'PreSplitOutRepositoryFormat',
2044
'RepositoryFormat4',
2045
'RepositoryFormat5',
2046
'RepositoryFormat6',
2047
'RepositoryFormat7',
2049
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2053
'RepositoryFormatKnit',
2054
'RepositoryFormatKnit1',
2056
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2059
def install_revision(repository, rev, revision_tree):
2060
"""Install all revision data into a repository."""
2061
install_revisions(repository, [(rev, revision_tree, None)])
2064
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2065
"""Install all revision data into a repository.
2067
Accepts an iterable of revision, tree, signature tuples. The signature
2070
repository.start_write_group()
2072
for n, (revision, revision_tree, signature) in enumerate(iterable):
2073
_install_revision(repository, revision, revision_tree, signature)
2075
pb.update('Transferring revisions', n + 1, num_revisions)
2077
repository.abort_write_group()
2080
repository.commit_write_group()
2083
def _install_revision(repository, rev, revision_tree, signature):
2084
"""Install all revision data into a repository."""
2085
present_parents = []
2087
for p_id in rev.parent_ids:
2088
if repository.has_revision(p_id):
2089
present_parents.append(p_id)
2090
parent_trees[p_id] = repository.revision_tree(p_id)
2092
parent_trees[p_id] = repository.revision_tree(
2093
_mod_revision.NULL_REVISION)
2095
inv = revision_tree.inventory
2096
entries = inv.iter_entries()
2097
# backwards compatibility hack: skip the root id.
2098
if not repository.supports_rich_root():
2099
path, root = entries.next()
2100
if root.revision != rev.revision_id:
2101
raise errors.IncompatibleRevision(repr(repository))
2103
for path, ie in entries:
2104
text_keys[(ie.file_id, ie.revision)] = ie
2105
text_parent_map = repository.texts.get_parent_map(text_keys)
2106
missing_texts = set(text_keys) - set(text_parent_map)
2107
# Add the texts that are not already present
2108
for text_key in missing_texts:
2109
ie = text_keys[text_key]
2111
# FIXME: TODO: The following loop overlaps/duplicates that done by
2112
# commit to determine parents. There is a latent/real bug here where
2113
# the parents inserted are not those commit would do - in particular
2114
# they are not filtered by heads(). RBC, AB
2115
for revision, tree in parent_trees.iteritems():
2116
if ie.file_id not in tree:
2118
parent_id = tree.inventory[ie.file_id].revision
2119
if parent_id in text_parents:
2121
text_parents.append((ie.file_id, parent_id))
2122
lines = revision_tree.get_file(ie.file_id).readlines()
2123
repository.texts.add_lines(text_key, text_parents, lines)
2125
# install the inventory
2126
repository.add_inventory(rev.revision_id, inv, present_parents)
2127
except errors.RevisionAlreadyPresent:
2129
if signature is not None:
2130
repository.add_signature_text(rev.revision_id, signature)
2131
repository.add_revision(rev.revision_id, rev, inv)
2134
class MetaDirRepository(Repository):
2135
"""Repositories in the new meta-dir layout.
2137
:ivar _transport: Transport for access to repository control files,
2138
typically pointing to .bzr/repository.
2141
def __init__(self, _format, a_bzrdir, control_files):
2142
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2143
self._transport = control_files._transport
2145
def is_shared(self):
2146
"""Return True if this repository is flagged as a shared repository."""
2147
return self._transport.has('shared-storage')
2150
def set_make_working_trees(self, new_value):
2151
"""Set the policy flag for making working trees when creating branches.
2153
This only applies to branches that use this repository.
2155
The default is 'True'.
2156
:param new_value: True to restore the default, False to disable making
2161
self._transport.delete('no-working-trees')
2162
except errors.NoSuchFile:
2165
self._transport.put_bytes('no-working-trees', '',
2166
mode=self.bzrdir._get_file_mode())
2168
def make_working_trees(self):
2169
"""Returns the policy for making working trees on new branches."""
2170
return not self._transport.has('no-working-trees')
2173
class MetaDirVersionedFileRepository(MetaDirRepository):
2174
"""Repositories in a meta-dir, that work via versioned file objects."""
2176
def __init__(self, _format, a_bzrdir, control_files):
2177
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2181
class RepositoryFormatRegistry(registry.Registry):
2182
"""Registry of RepositoryFormats."""
2184
def get(self, format_string):
2185
r = registry.Registry.get(self, format_string)
2191
format_registry = RepositoryFormatRegistry()
2192
"""Registry of formats, indexed by their identifying format string.
2194
This can contain either format instances themselves, or classes/factories that
2195
can be called to obtain one.
2199
#####################################################################
2200
# Repository Formats
2202
class RepositoryFormat(object):
2203
"""A repository format.
2205
Formats provide three things:
2206
* An initialization routine to construct repository data on disk.
2207
* a format string which is used when the BzrDir supports versioned
2209
* an open routine which returns a Repository instance.
2211
There is one and only one Format subclass for each on-disk format. But
2212
there can be one Repository subclass that is used for several different
2213
formats. The _format attribute on a Repository instance can be used to
2214
determine the disk format.
2216
Formats are placed in an dict by their format string for reference
2217
during opening. These should be subclasses of RepositoryFormat
2220
Once a format is deprecated, just deprecate the initialize and open
2221
methods on the format class. Do not deprecate the object, as the
2222
object will be created every system load.
2224
Common instance attributes:
2225
_matchingbzrdir - the bzrdir format that the repository format was
2226
originally written to work with. This can be used if manually
2227
constructing a bzrdir and repository, or more commonly for test suite
2231
# Set to True or False in derived classes. True indicates that the format
2232
# supports ghosts gracefully.
2233
supports_ghosts = None
2234
# Can this repository be given external locations to lookup additional
2235
# data. Set to True or False in derived classes.
2236
supports_external_lookups = None
2239
return "<%s>" % self.__class__.__name__
2241
def __eq__(self, other):
2242
# format objects are generally stateless
2243
return isinstance(other, self.__class__)
2245
def __ne__(self, other):
2246
return not self == other
2249
def find_format(klass, a_bzrdir):
2250
"""Return the format for the repository object in a_bzrdir.
2252
This is used by bzr native formats that have a "format" file in
2253
the repository. Other methods may be used by different types of
2257
transport = a_bzrdir.get_repository_transport(None)
2258
format_string = transport.get("format").read()
2259
return format_registry.get(format_string)
2260
except errors.NoSuchFile:
2261
raise errors.NoRepositoryPresent(a_bzrdir)
2263
raise errors.UnknownFormatError(format=format_string,
2267
def register_format(klass, format):
2268
format_registry.register(format.get_format_string(), format)
2271
def unregister_format(klass, format):
2272
format_registry.remove(format.get_format_string())
2275
def get_default_format(klass):
2276
"""Return the current default format."""
2277
from bzrlib import bzrdir
2278
return bzrdir.format_registry.make_bzrdir('default').repository_format
2280
def get_format_string(self):
2281
"""Return the ASCII format string that identifies this format.
2283
Note that in pre format ?? repositories the format string is
2284
not permitted nor written to disk.
2286
raise NotImplementedError(self.get_format_string)
2288
def get_format_description(self):
2289
"""Return the short description for this format."""
2290
raise NotImplementedError(self.get_format_description)
2292
# TODO: this shouldn't be in the base class, it's specific to things that
2293
# use weaves or knits -- mbp 20070207
2294
def _get_versioned_file_store(self,
2299
versionedfile_class=None,
2300
versionedfile_kwargs={},
2302
if versionedfile_class is None:
2303
versionedfile_class = self._versionedfile_class
2304
weave_transport = control_files._transport.clone(name)
2305
dir_mode = control_files._dir_mode
2306
file_mode = control_files._file_mode
2307
return VersionedFileStore(weave_transport, prefixed=prefixed,
2309
file_mode=file_mode,
2310
versionedfile_class=versionedfile_class,
2311
versionedfile_kwargs=versionedfile_kwargs,
2314
def initialize(self, a_bzrdir, shared=False):
2315
"""Initialize a repository of this format in a_bzrdir.
2317
:param a_bzrdir: The bzrdir to put the new repository in it.
2318
:param shared: The repository should be initialized as a sharable one.
2319
:returns: The new repository object.
2321
This may raise UninitializableFormat if shared repository are not
2322
compatible the a_bzrdir.
2324
raise NotImplementedError(self.initialize)
2326
def is_supported(self):
2327
"""Is this format supported?
2329
Supported formats must be initializable and openable.
2330
Unsupported formats may not support initialization or committing or
2331
some other features depending on the reason for not being supported.
2335
def check_conversion_target(self, target_format):
2336
raise NotImplementedError(self.check_conversion_target)
2338
def open(self, a_bzrdir, _found=False):
2339
"""Return an instance of this format for the bzrdir a_bzrdir.
2341
_found is a private parameter, do not use it.
2343
raise NotImplementedError(self.open)
2346
class MetaDirRepositoryFormat(RepositoryFormat):
2347
"""Common base class for the new repositories using the metadir layout."""
2349
rich_root_data = False
2350
supports_tree_reference = False
2351
supports_external_lookups = False
2354
def _matchingbzrdir(self):
2355
matching = bzrdir.BzrDirMetaFormat1()
2356
matching.repository_format = self
2360
super(MetaDirRepositoryFormat, self).__init__()
2362
def _create_control_files(self, a_bzrdir):
2363
"""Create the required files and the initial control_files object."""
2364
# FIXME: RBC 20060125 don't peek under the covers
2365
# NB: no need to escape relative paths that are url safe.
2366
repository_transport = a_bzrdir.get_repository_transport(self)
2367
control_files = lockable_files.LockableFiles(repository_transport,
2368
'lock', lockdir.LockDir)
2369
control_files.create_lock()
2370
return control_files
2372
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2373
"""Upload the initial blank content."""
2374
control_files = self._create_control_files(a_bzrdir)
2375
control_files.lock_write()
2376
transport = control_files._transport
2378
utf8_files += [('shared-storage', '')]
2380
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2381
for (filename, content_stream) in files:
2382
transport.put_file(filename, content_stream,
2383
mode=a_bzrdir._get_file_mode())
2384
for (filename, content_bytes) in utf8_files:
2385
transport.put_bytes_non_atomic(filename, content_bytes,
2386
mode=a_bzrdir._get_file_mode())
2388
control_files.unlock()
2391
# formats which have no format string are not discoverable
2392
# and not independently creatable, so are not registered. They're
2393
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2394
# needed, it's constructed directly by the BzrDir. Non-native formats where
2395
# the repository is not separately opened are similar.
2397
format_registry.register_lazy(
2398
'Bazaar-NG Repository format 7',
2399
'bzrlib.repofmt.weaverepo',
2403
format_registry.register_lazy(
2404
'Bazaar-NG Knit Repository Format 1',
2405
'bzrlib.repofmt.knitrepo',
2406
'RepositoryFormatKnit1',
2409
format_registry.register_lazy(
2410
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2411
'bzrlib.repofmt.knitrepo',
2412
'RepositoryFormatKnit3',
2415
format_registry.register_lazy(
2416
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2417
'bzrlib.repofmt.knitrepo',
2418
'RepositoryFormatKnit4',
2421
# Pack-based formats. There is one format for pre-subtrees, and one for
2422
# post-subtrees to allow ease of testing.
2423
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2424
format_registry.register_lazy(
2425
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2426
'bzrlib.repofmt.pack_repo',
2427
'RepositoryFormatKnitPack1',
2429
format_registry.register_lazy(
2430
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2431
'bzrlib.repofmt.pack_repo',
2432
'RepositoryFormatKnitPack3',
2434
format_registry.register_lazy(
2435
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2436
'bzrlib.repofmt.pack_repo',
2437
'RepositoryFormatKnitPack4',
2439
format_registry.register_lazy(
2440
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2441
'bzrlib.repofmt.pack_repo',
2442
'RepositoryFormatKnitPack5',
2444
format_registry.register_lazy(
2445
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2446
'bzrlib.repofmt.pack_repo',
2447
'RepositoryFormatKnitPack5RichRoot',
2449
format_registry.register_lazy(
2450
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2451
'bzrlib.repofmt.pack_repo',
2452
'RepositoryFormatKnitPack5RichRootBroken',
2454
format_registry.register_lazy(
2455
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2456
'bzrlib.repofmt.pack_repo',
2457
'RepositoryFormatKnitPack6',
2459
format_registry.register_lazy(
2460
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2461
'bzrlib.repofmt.pack_repo',
2462
'RepositoryFormatKnitPack6RichRoot',
2465
# Development formats.
2466
# 1.7->1.8 go below here
2467
format_registry.register_lazy(
2468
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2469
'bzrlib.repofmt.pack_repo',
2470
'RepositoryFormatPackDevelopment2',
2472
format_registry.register_lazy(
2473
("Bazaar development format 2 with subtree support "
2474
"(needs bzr.dev from before 1.8)\n"),
2475
'bzrlib.repofmt.pack_repo',
2476
'RepositoryFormatPackDevelopment2Subtree',
2480
class InterRepository(InterObject):
2481
"""This class represents operations taking place between two repositories.
2483
Its instances have methods like copy_content and fetch, and contain
2484
references to the source and target repositories these operations can be
2487
Often we will provide convenience methods on 'repository' which carry out
2488
operations with another repository - they will always forward to
2489
InterRepository.get(other).method_name(parameters).
2492
_walk_to_common_revisions_batch_size = 1
2494
"""The available optimised InterRepository types."""
2496
def __init__(self, source, target):
2497
InterObject.__init__(self, source, target)
2498
# These two attributes may be overridden by e.g. InterOtherToRemote to
2499
# provide a faster implementation.
2500
self.target_get_graph = self.target.get_graph
2501
self.target_get_parent_map = self.target.get_parent_map
2503
def copy_content(self, revision_id=None):
2504
raise NotImplementedError(self.copy_content)
2506
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2507
"""Fetch the content required to construct revision_id.
2509
The content is copied from self.source to self.target.
2511
:param revision_id: if None all content is copied, if NULL_REVISION no
2513
:param pb: optional progress bar to use for progress reports. If not
2514
provided a default one will be created.
2516
:returns: (copied_revision_count, failures).
2518
# Normally we should find a specific InterRepository subclass to do
2519
# the fetch; if nothing else then at least InterSameDataRepository.
2520
# If none of them is suitable it looks like fetching is not possible;
2521
# we try to give a good message why. _assert_same_model will probably
2522
# give a helpful message; otherwise a generic one.
2523
self._assert_same_model(self.source, self.target)
2524
raise errors.IncompatibleRepositories(self.source, self.target,
2525
"no suitableInterRepository found")
2527
def _walk_to_common_revisions(self, revision_ids):
2528
"""Walk out from revision_ids in source to revisions target has.
2530
:param revision_ids: The start point for the search.
2531
:return: A set of revision ids.
2533
target_graph = self.target_get_graph()
2534
revision_ids = frozenset(revision_ids)
2535
# Fast path for the case where all the revisions are already in the
2537
# (Although this does incur an extra round trip for the
2538
# fairly common case where the target doesn't already have the revision
2540
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2541
return graph.SearchResult(revision_ids, set(), 0, set())
2542
missing_revs = set()
2543
source_graph = self.source.get_graph()
2544
# ensure we don't pay silly lookup costs.
2545
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2546
null_set = frozenset([_mod_revision.NULL_REVISION])
2547
searcher_exhausted = False
2551
# Iterate the searcher until we have enough next_revs
2552
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2554
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2555
next_revs.update(next_revs_part)
2556
ghosts.update(ghosts_part)
2557
except StopIteration:
2558
searcher_exhausted = True
2560
# If there are ghosts in the source graph, and the caller asked for
2561
# them, make sure that they are present in the target.
2562
# We don't care about other ghosts as we can't fetch them and
2563
# haven't been asked to.
2564
ghosts_to_check = set(revision_ids.intersection(ghosts))
2565
revs_to_get = set(next_revs).union(ghosts_to_check)
2567
have_revs = set(target_graph.get_parent_map(revs_to_get))
2568
# we always have NULL_REVISION present.
2569
have_revs = have_revs.union(null_set)
2570
# Check if the target is missing any ghosts we need.
2571
ghosts_to_check.difference_update(have_revs)
2573
# One of the caller's revision_ids is a ghost in both the
2574
# source and the target.
2575
raise errors.NoSuchRevision(
2576
self.source, ghosts_to_check.pop())
2577
missing_revs.update(next_revs - have_revs)
2578
# Because we may have walked past the original stop point, make
2579
# sure everything is stopped
2580
stop_revs = searcher.find_seen_ancestors(have_revs)
2581
searcher.stop_searching_any(stop_revs)
2582
if searcher_exhausted:
2584
return searcher.get_result()
2586
@deprecated_method(one_two)
2588
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2589
"""Return the revision ids that source has that target does not.
2591
These are returned in topological order.
2593
:param revision_id: only return revision ids included by this
2595
:param find_ghosts: If True find missing revisions in deep history
2596
rather than just finding the surface difference.
2598
return list(self.search_missing_revision_ids(
2599
revision_id, find_ghosts).get_keys())
2602
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2603
"""Return the revision ids that source has that target does not.
2605
:param revision_id: only return revision ids included by this
2607
:param find_ghosts: If True find missing revisions in deep history
2608
rather than just finding the surface difference.
2609
:return: A bzrlib.graph.SearchResult.
2611
# stop searching at found target revisions.
2612
if not find_ghosts and revision_id is not None:
2613
return self._walk_to_common_revisions([revision_id])
2614
# generic, possibly worst case, slow code path.
2615
target_ids = set(self.target.all_revision_ids())
2616
if revision_id is not None:
2617
source_ids = self.source.get_ancestry(revision_id)
2618
if source_ids[0] is not None:
2619
raise AssertionError()
2622
source_ids = self.source.all_revision_ids()
2623
result_set = set(source_ids).difference(target_ids)
2624
return self.source.revision_ids_to_search_result(result_set)
2627
def _same_model(source, target):
2628
"""True if source and target have the same data representation.
2630
Note: this is always called on the base class; overriding it in a
2631
subclass will have no effect.
2634
InterRepository._assert_same_model(source, target)
2636
except errors.IncompatibleRepositories, e:
2640
def _assert_same_model(source, target):
2641
"""Raise an exception if two repositories do not use the same model.
2643
if source.supports_rich_root() != target.supports_rich_root():
2644
raise errors.IncompatibleRepositories(source, target,
2645
"different rich-root support")
2646
if source._serializer != target._serializer:
2647
raise errors.IncompatibleRepositories(source, target,
2648
"different serializers")
2651
class InterSameDataRepository(InterRepository):
2652
"""Code for converting between repositories that represent the same data.
2654
Data format and model must match for this to work.
2658
def _get_repo_format_to_test(self):
2659
"""Repository format for testing with.
2661
InterSameData can pull from subtree to subtree and from non-subtree to
2662
non-subtree, so we test this with the richest repository format.
2664
from bzrlib.repofmt import knitrepo
2665
return knitrepo.RepositoryFormatKnit3()
2668
def is_compatible(source, target):
2669
return InterRepository._same_model(source, target)
2672
def copy_content(self, revision_id=None):
2673
"""Make a complete copy of the content in self into destination.
2675
This copies both the repository's revision data, and configuration information
2676
such as the make_working_trees setting.
2678
This is a destructive operation! Do not use it on existing
2681
:param revision_id: Only copy the content needed to construct
2682
revision_id and its parents.
2685
self.target.set_make_working_trees(self.source.make_working_trees())
2686
except NotImplementedError:
2688
# but don't bother fetching if we have the needed data now.
2689
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2690
self.target.has_revision(revision_id)):
2692
self.target.fetch(self.source, revision_id=revision_id)
2695
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2696
"""See InterRepository.fetch()."""
2697
from bzrlib.fetch import RepoFetcher
2698
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2699
self.source, self.source._format, self.target,
2700
self.target._format)
2701
f = RepoFetcher(to_repository=self.target,
2702
from_repository=self.source,
2703
last_revision=revision_id,
2704
pb=pb, find_ghosts=find_ghosts)
2705
return f.count_copied, f.failed_revisions
2708
class InterWeaveRepo(InterSameDataRepository):
2709
"""Optimised code paths between Weave based repositories.
2711
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2712
implemented lazy inter-object optimisation.
2716
def _get_repo_format_to_test(self):
2717
from bzrlib.repofmt import weaverepo
2718
return weaverepo.RepositoryFormat7()
2721
def is_compatible(source, target):
2722
"""Be compatible with known Weave formats.
2724
We don't test for the stores being of specific types because that
2725
could lead to confusing results, and there is no need to be
2728
from bzrlib.repofmt.weaverepo import (
2734
return (isinstance(source._format, (RepositoryFormat5,
2736
RepositoryFormat7)) and
2737
isinstance(target._format, (RepositoryFormat5,
2739
RepositoryFormat7)))
2740
except AttributeError:
2744
def copy_content(self, revision_id=None):
2745
"""See InterRepository.copy_content()."""
2746
# weave specific optimised path:
2748
self.target.set_make_working_trees(self.source.make_working_trees())
2749
except (errors.RepositoryUpgradeRequired, NotImplemented):
2751
# FIXME do not peek!
2752
if self.source._transport.listable():
2753
pb = ui.ui_factory.nested_progress_bar()
2755
self.target.texts.insert_record_stream(
2756
self.source.texts.get_record_stream(
2757
self.source.texts.keys(), 'topological', False))
2758
pb.update('copying inventory', 0, 1)
2759
self.target.inventories.insert_record_stream(
2760
self.source.inventories.get_record_stream(
2761
self.source.inventories.keys(), 'topological', False))
2762
self.target.signatures.insert_record_stream(
2763
self.source.signatures.get_record_stream(
2764
self.source.signatures.keys(),
2766
self.target.revisions.insert_record_stream(
2767
self.source.revisions.get_record_stream(
2768
self.source.revisions.keys(),
2769
'topological', True))
2773
self.target.fetch(self.source, revision_id=revision_id)
2776
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2777
"""See InterRepository.fetch()."""
2778
from bzrlib.fetch import RepoFetcher
2779
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2780
self.source, self.source._format, self.target, self.target._format)
2781
f = RepoFetcher(to_repository=self.target,
2782
from_repository=self.source,
2783
last_revision=revision_id,
2784
pb=pb, find_ghosts=find_ghosts)
2785
return f.count_copied, f.failed_revisions
2788
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2789
"""See InterRepository.missing_revision_ids()."""
2790
# we want all revisions to satisfy revision_id in source.
2791
# but we don't want to stat every file here and there.
2792
# we want then, all revisions other needs to satisfy revision_id
2793
# checked, but not those that we have locally.
2794
# so the first thing is to get a subset of the revisions to
2795
# satisfy revision_id in source, and then eliminate those that
2796
# we do already have.
2797
# this is slow on high latency connection to self, but as as this
2798
# disk format scales terribly for push anyway due to rewriting
2799
# inventory.weave, this is considered acceptable.
2801
if revision_id is not None:
2802
source_ids = self.source.get_ancestry(revision_id)
2803
if source_ids[0] is not None:
2804
raise AssertionError()
2807
source_ids = self.source._all_possible_ids()
2808
source_ids_set = set(source_ids)
2809
# source_ids is the worst possible case we may need to pull.
2810
# now we want to filter source_ids against what we actually
2811
# have in target, but don't try to check for existence where we know
2812
# we do not have a revision as that would be pointless.
2813
target_ids = set(self.target._all_possible_ids())
2814
possibly_present_revisions = target_ids.intersection(source_ids_set)
2815
actually_present_revisions = set(
2816
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2817
required_revisions = source_ids_set.difference(actually_present_revisions)
2818
if revision_id is not None:
2819
# we used get_ancestry to determine source_ids then we are assured all
2820
# revisions referenced are present as they are installed in topological order.
2821
# and the tip revision was validated by get_ancestry.
2822
result_set = required_revisions
2824
# if we just grabbed the possibly available ids, then
2825
# we only have an estimate of whats available and need to validate
2826
# that against the revision records.
2828
self.source._eliminate_revisions_not_present(required_revisions))
2829
return self.source.revision_ids_to_search_result(result_set)
2832
class InterKnitRepo(InterSameDataRepository):
2833
"""Optimised code paths between Knit based repositories."""
2836
def _get_repo_format_to_test(self):
2837
from bzrlib.repofmt import knitrepo
2838
return knitrepo.RepositoryFormatKnit1()
2841
def is_compatible(source, target):
2842
"""Be compatible with known Knit formats.
2844
We don't test for the stores being of specific types because that
2845
could lead to confusing results, and there is no need to be
2848
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2850
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2851
isinstance(target._format, RepositoryFormatKnit))
2852
except AttributeError:
2854
return are_knits and InterRepository._same_model(source, target)
2857
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2858
"""See InterRepository.fetch()."""
2859
from bzrlib.fetch import RepoFetcher
2860
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2861
self.source, self.source._format, self.target, self.target._format)
2862
f = RepoFetcher(to_repository=self.target,
2863
from_repository=self.source,
2864
last_revision=revision_id,
2865
pb=pb, find_ghosts=find_ghosts)
2866
return f.count_copied, f.failed_revisions
2869
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2870
"""See InterRepository.missing_revision_ids()."""
2871
if revision_id is not None:
2872
source_ids = self.source.get_ancestry(revision_id)
2873
if source_ids[0] is not None:
2874
raise AssertionError()
2877
source_ids = self.source.all_revision_ids()
2878
source_ids_set = set(source_ids)
2879
# source_ids is the worst possible case we may need to pull.
2880
# now we want to filter source_ids against what we actually
2881
# have in target, but don't try to check for existence where we know
2882
# we do not have a revision as that would be pointless.
2883
target_ids = set(self.target.all_revision_ids())
2884
possibly_present_revisions = target_ids.intersection(source_ids_set)
2885
actually_present_revisions = set(
2886
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2887
required_revisions = source_ids_set.difference(actually_present_revisions)
2888
if revision_id is not None:
2889
# we used get_ancestry to determine source_ids then we are assured all
2890
# revisions referenced are present as they are installed in topological order.
2891
# and the tip revision was validated by get_ancestry.
2892
result_set = required_revisions
2894
# if we just grabbed the possibly available ids, then
2895
# we only have an estimate of whats available and need to validate
2896
# that against the revision records.
2898
self.source._eliminate_revisions_not_present(required_revisions))
2899
return self.source.revision_ids_to_search_result(result_set)
2902
class InterPackRepo(InterSameDataRepository):
2903
"""Optimised code paths between Pack based repositories."""
2906
def _get_repo_format_to_test(self):
2907
from bzrlib.repofmt import pack_repo
2908
return pack_repo.RepositoryFormatKnitPack1()
2911
def is_compatible(source, target):
2912
"""Be compatible with known Pack formats.
2914
We don't test for the stores being of specific types because that
2915
could lead to confusing results, and there is no need to be
2918
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2920
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2921
isinstance(target._format, RepositoryFormatPack))
2922
except AttributeError:
2924
return are_packs and InterRepository._same_model(source, target)
2927
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2928
"""See InterRepository.fetch()."""
2929
if (len(self.source._fallback_repositories) > 0 or
2930
len(self.target._fallback_repositories) > 0):
2931
# The pack layer is not aware of fallback repositories, so when
2932
# fetching from a stacked repository or into a stacked repository
2933
# we use the generic fetch logic which uses the VersionedFiles
2934
# attributes on repository.
2935
from bzrlib.fetch import RepoFetcher
2936
# Make sure the generic fetcher sets the write cache size on the
2937
# new pack (just like Packer.pack does) to avoid doing many tiny
2938
# writes (which can be slow over a network connection).
2939
# XXX: ideally the transport layer would do this automatically.
2940
pack_coll = self._get_target_pack_collection()
2942
lambda: pack_coll._new_pack.set_write_cache_size(1024*1024))
2943
fetcher = RepoFetcher(self.target, self.source, revision_id,
2944
pb, find_ghosts, set_cache_size)
2945
return fetcher.count_copied, fetcher.failed_revisions
2946
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2947
self.source, self.source._format, self.target, self.target._format)
2948
self.count_copied = 0
2949
if revision_id is None:
2951
# everything to do - use pack logic
2952
# to fetch from all packs to one without
2953
# inventory parsing etc, IFF nothing to be copied is in the target.
2955
source_revision_ids = frozenset(self.source.all_revision_ids())
2956
revision_ids = source_revision_ids - \
2957
frozenset(self.target_get_parent_map(source_revision_ids))
2958
revision_keys = [(revid,) for revid in revision_ids]
2959
target_pack_collection = self._get_target_pack_collection()
2960
index = target_pack_collection.revision_index.combined_index
2961
present_revision_ids = set(item[1][0] for item in
2962
index.iter_entries(revision_keys))
2963
revision_ids = set(revision_ids) - present_revision_ids
2964
# implementing the TODO will involve:
2965
# - detecting when all of a pack is selected
2966
# - avoiding as much as possible pre-selection, so the
2967
# more-core routines such as create_pack_from_packs can filter in
2968
# a just-in-time fashion. (though having a HEADS list on a
2969
# repository might make this a lot easier, because we could
2970
# sensibly detect 'new revisions' without doing a full index scan.
2971
elif _mod_revision.is_null(revision_id):
2976
revision_ids = self.search_missing_revision_ids(revision_id,
2977
find_ghosts=find_ghosts).get_keys()
2978
except errors.NoSuchRevision:
2979
raise errors.InstallFailed([revision_id])
2980
if len(revision_ids) == 0:
2982
return self._pack(self.source, self.target, revision_ids)
2984
def _pack(self, source, target, revision_ids):
2985
from bzrlib.repofmt.pack_repo import Packer
2986
target_pack_collection = self._get_target_pack_collection()
2987
packs = source._pack_collection.all_packs()
2988
pack = Packer(target_pack_collection, packs, '.fetch',
2989
revision_ids).pack()
2990
if pack is not None:
2991
target_pack_collection._save_pack_names()
2992
copied_revs = pack.get_revision_count()
2993
# Trigger an autopack. This may duplicate effort as we've just done
2994
# a pack creation, but for now it is simpler to think about as
2995
# 'upload data, then repack if needed'.
2997
return (copied_revs, [])
3001
def _autopack(self):
3002
self.target._pack_collection.autopack()
3004
def _get_target_pack_collection(self):
3005
return self.target._pack_collection
3008
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3009
"""See InterRepository.missing_revision_ids().
3011
:param find_ghosts: Find ghosts throughout the ancestry of
3014
if not find_ghosts and revision_id is not None:
3015
return self._walk_to_common_revisions([revision_id])
3016
elif revision_id is not None:
3017
# Find ghosts: search for revisions pointing from one repository to
3018
# the other, and vice versa, anywhere in the history of revision_id.
3019
graph = self.target_get_graph(other_repository=self.source)
3020
searcher = graph._make_breadth_first_searcher([revision_id])
3024
next_revs, ghosts = searcher.next_with_ghosts()
3025
except StopIteration:
3027
if revision_id in ghosts:
3028
raise errors.NoSuchRevision(self.source, revision_id)
3029
found_ids.update(next_revs)
3030
found_ids.update(ghosts)
3031
found_ids = frozenset(found_ids)
3032
# Double query here: should be able to avoid this by changing the
3033
# graph api further.
3034
result_set = found_ids - frozenset(
3035
self.target_get_parent_map(found_ids))
3037
source_ids = self.source.all_revision_ids()
3038
# source_ids is the worst possible case we may need to pull.
3039
# now we want to filter source_ids against what we actually
3040
# have in target, but don't try to check for existence where we know
3041
# we do not have a revision as that would be pointless.
3042
target_ids = set(self.target.all_revision_ids())
3043
result_set = set(source_ids).difference(target_ids)
3044
return self.source.revision_ids_to_search_result(result_set)
3047
class InterModel1and2(InterRepository):
3050
def _get_repo_format_to_test(self):
3054
def is_compatible(source, target):
3055
if not source.supports_rich_root() and target.supports_rich_root():
3061
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3062
"""See InterRepository.fetch()."""
3063
from bzrlib.fetch import Model1toKnit2Fetcher
3064
f = Model1toKnit2Fetcher(to_repository=self.target,
3065
from_repository=self.source,
3066
last_revision=revision_id,
3067
pb=pb, find_ghosts=find_ghosts)
3068
return f.count_copied, f.failed_revisions
3071
def copy_content(self, revision_id=None):
3072
"""Make a complete copy of the content in self into destination.
3074
This is a destructive operation! Do not use it on existing
3077
:param revision_id: Only copy the content needed to construct
3078
revision_id and its parents.
3081
self.target.set_make_working_trees(self.source.make_working_trees())
3082
except NotImplementedError:
3084
# but don't bother fetching if we have the needed data now.
3085
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
3086
self.target.has_revision(revision_id)):
3088
self.target.fetch(self.source, revision_id=revision_id)
3091
class InterKnit1and2(InterKnitRepo):
3094
def _get_repo_format_to_test(self):
3098
def is_compatible(source, target):
3099
"""Be compatible with Knit1 source and Knit3 target"""
3101
from bzrlib.repofmt.knitrepo import (
3102
RepositoryFormatKnit1,
3103
RepositoryFormatKnit3,
3105
from bzrlib.repofmt.pack_repo import (
3106
RepositoryFormatKnitPack1,
3107
RepositoryFormatKnitPack3,
3108
RepositoryFormatKnitPack4,
3109
RepositoryFormatKnitPack5,
3110
RepositoryFormatKnitPack5RichRoot,
3111
RepositoryFormatKnitPack6,
3112
RepositoryFormatKnitPack6RichRoot,
3113
RepositoryFormatPackDevelopment2,
3114
RepositoryFormatPackDevelopment2Subtree,
3117
RepositoryFormatKnit1, # no rr, no subtree
3118
RepositoryFormatKnitPack1, # no rr, no subtree
3119
RepositoryFormatPackDevelopment2, # no rr, no subtree
3120
RepositoryFormatKnitPack5, # no rr, no subtree
3121
RepositoryFormatKnitPack6, # no rr, no subtree
3124
RepositoryFormatKnit3, # rr, subtree
3125
RepositoryFormatKnitPack3, # rr, subtree
3126
RepositoryFormatKnitPack4, # rr, no subtree
3127
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
3128
RepositoryFormatKnitPack6RichRoot,# rr, no subtree
3129
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
3131
for format in norichroot:
3132
if format.rich_root_data:
3133
raise AssertionError('Format %s is a rich-root format'
3134
' but is included in the non-rich-root list'
3136
for format in richroot:
3137
if not format.rich_root_data:
3138
raise AssertionError('Format %s is not a rich-root format'
3139
' but is included in the rich-root list'
3141
# TODO: One alternative is to just check format.rich_root_data,
3142
# instead of keeping membership lists. However, the formats
3143
# *also* have to use the same 'Knit' style of storage
3144
# (line-deltas, fulltexts, etc.)
3145
return (isinstance(source._format, norichroot) and
3146
isinstance(target._format, richroot))
3147
except AttributeError:
3151
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3152
"""See InterRepository.fetch()."""
3153
from bzrlib.fetch import Knit1to2Fetcher
3154
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3155
self.source, self.source._format, self.target,
3156
self.target._format)
3157
f = Knit1to2Fetcher(to_repository=self.target,
3158
from_repository=self.source,
3159
last_revision=revision_id,
3160
pb=pb, find_ghosts=find_ghosts)
3161
return f.count_copied, f.failed_revisions
3164
class InterDifferingSerializer(InterKnitRepo):
3167
def _get_repo_format_to_test(self):
3171
def is_compatible(source, target):
3172
"""Be compatible with Knit2 source and Knit3 target"""
3173
if source.supports_rich_root() != target.supports_rich_root():
3175
# Ideally, we'd support fetching if the source had no tree references
3176
# even if it supported them...
3177
if (getattr(source, '_format.supports_tree_reference', False) and
3178
not getattr(target, '_format.supports_tree_reference', False)):
3183
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3184
"""See InterRepository.fetch()."""
3185
revision_ids = self.target.search_missing_revision_ids(self.source,
3186
revision_id, find_ghosts=find_ghosts).get_keys()
3187
revision_ids = tsort.topo_sort(
3188
self.source.get_graph().get_parent_map(revision_ids))
3189
def revisions_iterator():
3190
rev_ids = list(revision_ids)
3191
for offset in xrange(0, len(rev_ids), 100):
3192
current_revids = rev_ids[offset:offset+100]
3193
revisions = self.source.get_revisions(current_revids)
3194
trees = self.source.revision_trees(current_revids)
3195
keys = [(r,) for r in current_revids]
3196
sig_stream = self.source.signatures.get_record_stream(
3197
keys, 'unordered', True)
3199
for record in versionedfile.filter_absent(sig_stream):
3200
sigs[record.key[0]] = record.get_bytes_as('fulltext')
3201
for rev, tree in zip(revisions, trees):
3202
yield rev, tree, sigs.get(rev.revision_id, None)
3204
my_pb = ui.ui_factory.nested_progress_bar()
3209
install_revisions(self.target, revisions_iterator(),
3210
len(revision_ids), pb)
3212
if my_pb is not None:
3214
return len(revision_ids), 0
3217
class InterOtherToRemote(InterRepository):
3218
"""An InterRepository that simply delegates to the 'real' InterRepository
3219
calculated for (source, target._real_repository).
3222
_walk_to_common_revisions_batch_size = 50
3224
def __init__(self, source, target):
3225
InterRepository.__init__(self, source, target)
3226
self._real_inter = None
3229
def is_compatible(source, target):
3230
if isinstance(target, remote.RemoteRepository):
3234
def _ensure_real_inter(self):
3235
if self._real_inter is None:
3236
self.target._ensure_real()
3237
real_target = self.target._real_repository
3238
self._real_inter = InterRepository.get(self.source, real_target)
3239
# Make _real_inter use the RemoteRepository for get_parent_map
3240
self._real_inter.target_get_graph = self.target.get_graph
3241
self._real_inter.target_get_parent_map = self.target.get_parent_map
3243
def copy_content(self, revision_id=None):
3244
self._ensure_real_inter()
3245
self._real_inter.copy_content(revision_id=revision_id)
3247
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3248
self._ensure_real_inter()
3249
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3250
find_ghosts=find_ghosts)
3253
def _get_repo_format_to_test(self):
3257
class InterRemoteToOther(InterRepository):
3259
def __init__(self, source, target):
3260
InterRepository.__init__(self, source, target)
3261
self._real_inter = None
3264
def is_compatible(source, target):
3265
if not isinstance(source, remote.RemoteRepository):
3267
# Is source's model compatible with target's model?
3268
source._ensure_real()
3269
real_source = source._real_repository
3270
if isinstance(real_source, remote.RemoteRepository):
3271
raise NotImplementedError(
3272
"We don't support remote repos backed by remote repos yet.")
3273
return InterRepository._same_model(real_source, target)
3275
def _ensure_real_inter(self):
3276
if self._real_inter is None:
3277
self.source._ensure_real()
3278
real_source = self.source._real_repository
3279
self._real_inter = InterRepository.get(real_source, self.target)
3281
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3282
self._ensure_real_inter()
3283
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3284
find_ghosts=find_ghosts)
3286
def copy_content(self, revision_id=None):
3287
self._ensure_real_inter()
3288
self._real_inter.copy_content(revision_id=revision_id)
3291
def _get_repo_format_to_test(self):
3296
class InterPackToRemotePack(InterPackRepo):
3297
"""A specialisation of InterPackRepo for a target that is a
3300
This will use the get_parent_map RPC rather than plain readvs, and also
3301
uses an RPC for autopacking.
3304
_walk_to_common_revisions_batch_size = 50
3307
def is_compatible(source, target):
3308
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3309
if isinstance(source._format, RepositoryFormatPack):
3310
if isinstance(target, remote.RemoteRepository):
3311
target._ensure_real()
3312
if isinstance(target._real_repository._format,
3313
RepositoryFormatPack):
3314
if InterRepository._same_model(source, target):
3318
def _autopack(self):
3319
self.target.autopack()
3321
def _get_target_pack_collection(self):
3322
return self.target._real_repository._pack_collection
3325
def _get_repo_format_to_test(self):
3329
InterRepository.register_optimiser(InterDifferingSerializer)
3330
InterRepository.register_optimiser(InterSameDataRepository)
3331
InterRepository.register_optimiser(InterWeaveRepo)
3332
InterRepository.register_optimiser(InterKnitRepo)
3333
InterRepository.register_optimiser(InterModel1and2)
3334
InterRepository.register_optimiser(InterKnit1and2)
3335
InterRepository.register_optimiser(InterPackRepo)
3336
InterRepository.register_optimiser(InterOtherToRemote)
3337
InterRepository.register_optimiser(InterRemoteToOther)
3338
InterRepository.register_optimiser(InterPackToRemotePack)
3341
class CopyConverter(object):
3342
"""A repository conversion tool which just performs a copy of the content.
3344
This is slow but quite reliable.
3347
def __init__(self, target_format):
3348
"""Create a CopyConverter.
3350
:param target_format: The format the resulting repository should be.
3352
self.target_format = target_format
3354
def convert(self, repo, pb):
3355
"""Perform the conversion of to_convert, giving feedback via pb.
3357
:param to_convert: The disk object to convert.
3358
:param pb: a progress bar to use for progress information.
3363
# this is only useful with metadir layouts - separated repo content.
3364
# trigger an assertion if not such
3365
repo._format.get_format_string()
3366
self.repo_dir = repo.bzrdir
3367
self.step('Moving repository to repository.backup')
3368
self.repo_dir.transport.move('repository', 'repository.backup')
3369
backup_transport = self.repo_dir.transport.clone('repository.backup')
3370
repo._format.check_conversion_target(self.target_format)
3371
self.source_repo = repo._format.open(self.repo_dir,
3373
_override_transport=backup_transport)
3374
self.step('Creating new repository')
3375
converted = self.target_format.initialize(self.repo_dir,
3376
self.source_repo.is_shared())
3377
converted.lock_write()
3379
self.step('Copying content into repository.')
3380
self.source_repo.copy_content_into(converted)
3383
self.step('Deleting old repository content.')
3384
self.repo_dir.transport.delete_tree('repository.backup')
3385
self.pb.note('repository converted')
3387
def step(self, message):
3388
"""Update the pb by a step."""
3390
self.pb.update(message, self.count, self.total)
3402
def _unescaper(match, _map=_unescape_map):
3403
code = match.group(1)
3407
if not code.startswith('#'):
3409
return unichr(int(code[1:])).encode('utf8')
3415
def _unescape_xml(data):
3416
"""Unescape predefined XML entities in a string of data."""
3418
if _unescape_re is None:
3419
_unescape_re = re.compile('\&([^;]*);')
3420
return _unescape_re.sub(_unescaper, data)
3423
class _VersionedFileChecker(object):
3425
def __init__(self, repository):
3426
self.repository = repository
3427
self.text_index = self.repository._generate_text_key_index()
3429
def calculate_file_version_parents(self, text_key):
3430
"""Calculate the correct parents for a file version according to
3433
parent_keys = self.text_index[text_key]
3434
if parent_keys == [_mod_revision.NULL_REVISION]:
3436
return tuple(parent_keys)
3438
def check_file_version_parents(self, texts, progress_bar=None):
3439
"""Check the parents stored in a versioned file are correct.
3441
It also detects file versions that are not referenced by their
3442
corresponding revision's inventory.
3444
:returns: A tuple of (wrong_parents, dangling_file_versions).
3445
wrong_parents is a dict mapping {revision_id: (stored_parents,
3446
correct_parents)} for each revision_id where the stored parents
3447
are not correct. dangling_file_versions is a set of (file_id,
3448
revision_id) tuples for versions that are present in this versioned
3449
file, but not used by the corresponding inventory.
3452
self.file_ids = set([file_id for file_id, _ in
3453
self.text_index.iterkeys()])
3454
# text keys is now grouped by file_id
3455
n_weaves = len(self.file_ids)
3456
files_in_revisions = {}
3457
revisions_of_files = {}
3458
n_versions = len(self.text_index)
3459
progress_bar.update('loading text store', 0, n_versions)
3460
parent_map = self.repository.texts.get_parent_map(self.text_index)
3461
# On unlistable transports this could well be empty/error...
3462
text_keys = self.repository.texts.keys()
3463
unused_keys = frozenset(text_keys) - set(self.text_index)
3464
for num, key in enumerate(self.text_index.iterkeys()):
3465
if progress_bar is not None:
3466
progress_bar.update('checking text graph', num, n_versions)
3467
correct_parents = self.calculate_file_version_parents(key)
3469
knit_parents = parent_map[key]
3470
except errors.RevisionNotPresent:
3473
if correct_parents != knit_parents:
3474
wrong_parents[key] = (knit_parents, correct_parents)
3475
return wrong_parents, unused_keys
3478
def _old_get_graph(repository, revision_id):
3479
"""DO NOT USE. That is all. I'm serious."""
3480
graph = repository.get_graph()
3481
revision_graph = dict(((key, value) for key, value in
3482
graph.iter_ancestry([revision_id]) if value is not None))
3483
return _strip_NULL_ghosts(revision_graph)
3486
def _strip_NULL_ghosts(revision_graph):
3487
"""Also don't use this. more compatibility code for unmigrated clients."""
3488
# Filter ghosts, and null:
3489
if _mod_revision.NULL_REVISION in revision_graph:
3490
del revision_graph[_mod_revision.NULL_REVISION]
3491
for key, parents in revision_graph.items():
3492
revision_graph[key] = tuple(parent for parent in parents if parent
3494
return revision_graph
added
removed
bzrlib/repository.py
