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
123
def _validate_unicode_text(self, text, context):
124
"""Verify things like commit messages don't have bogus characters."""
126
raise ValueError('Invalid value for %s: %r' % (context, text))
128
def _validate_revprops(self, revprops):
129
for key, value in revprops.iteritems():
130
# We know that the XML serializers do not round trip '\r'
131
# correctly, so refuse to accept them
132
if not isinstance(value, basestring):
133
raise ValueError('revision property (%s) is not a valid'
134
' (unicode) string: %r' % (key, value))
135
self._validate_unicode_text(value,
136
'revision property (%s)' % (key,))
138
def commit(self, message):
139
"""Make the actual commit.
141
:return: The revision id of the recorded revision.
143
self._validate_unicode_text(message, 'commit message')
144
rev = _mod_revision.Revision(
145
timestamp=self._timestamp,
146
timezone=self._timezone,
147
committer=self._committer,
149
inventory_sha1=self.inv_sha1,
150
revision_id=self._new_revision_id,
151
properties=self._revprops)
152
rev.parent_ids = self.parents
153
self.repository.add_revision(self._new_revision_id, rev,
154
self.new_inventory, self._config)
155
self.repository.commit_write_group()
156
return self._new_revision_id
159
"""Abort the commit that is being built.
161
self.repository.abort_write_group()
163
def revision_tree(self):
164
"""Return the tree that was just committed.
166
After calling commit() this can be called to get a RevisionTree
167
representing the newly committed tree. This is preferred to
168
calling Repository.revision_tree() because that may require
169
deserializing the inventory, while we already have a copy in
172
return RevisionTree(self.repository, self.new_inventory,
173
self._new_revision_id)
175
def finish_inventory(self):
176
"""Tell the builder that the inventory is finished."""
177
if self.new_inventory.root is None:
178
raise AssertionError('Root entry should be supplied to'
179
' record_entry_contents, as of bzr 0.10.')
180
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
181
self.new_inventory.revision_id = self._new_revision_id
182
self.inv_sha1 = self.repository.add_inventory(
183
self._new_revision_id,
188
def _gen_revision_id(self):
189
"""Return new revision-id."""
190
return generate_ids.gen_revision_id(self._config.username(),
193
def _generate_revision_if_needed(self):
194
"""Create a revision id if None was supplied.
196
If the repository can not support user-specified revision ids
197
they should override this function and raise CannotSetRevisionId
198
if _new_revision_id is not None.
200
:raises: CannotSetRevisionId
202
if self._new_revision_id is None:
203
self._new_revision_id = self._gen_revision_id()
204
self.random_revid = True
206
self.random_revid = False
208
def _heads(self, file_id, revision_ids):
209
"""Calculate the graph heads for revision_ids in the graph of file_id.
211
This can use either a per-file graph or a global revision graph as we
212
have an identity relationship between the two graphs.
214
return self.__heads(revision_ids)
216
def _check_root(self, ie, parent_invs, tree):
217
"""Helper for record_entry_contents.
219
:param ie: An entry being added.
220
:param parent_invs: The inventories of the parent revisions of the
222
:param tree: The tree that is being committed.
224
# In this revision format, root entries have no knit or weave When
225
# serializing out to disk and back in root.revision is always
227
ie.revision = self._new_revision_id
229
def _get_delta(self, ie, basis_inv, path):
230
"""Get a delta against the basis inventory for ie."""
231
if ie.file_id not in basis_inv:
233
return (None, path, ie.file_id, ie)
234
elif ie != basis_inv[ie.file_id]:
236
# TODO: avoid tis id2path call.
237
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
242
def record_entry_contents(self, ie, parent_invs, path, tree,
244
"""Record the content of ie from tree into the commit if needed.
246
Side effect: sets ie.revision when unchanged
248
:param ie: An inventory entry present in the commit.
249
:param parent_invs: The inventories of the parent revisions of the
251
:param path: The path the entry is at in the tree.
252
:param tree: The tree which contains this entry and should be used to
254
:param content_summary: Summary data from the tree about the paths
255
content - stat, length, exec, sha/link target. This is only
256
accessed when the entry has a revision of None - that is when it is
257
a candidate to commit.
258
:return: A tuple (change_delta, version_recorded, fs_hash).
259
change_delta is an inventory_delta change for this entry against
260
the basis tree of the commit, or None if no change occured against
262
version_recorded is True if a new version of the entry has been
263
recorded. For instance, committing a merge where a file was only
264
changed on the other side will return (delta, False).
265
fs_hash is either None, or the hash details for the path (currently
266
a tuple of the contents sha1 and the statvalue returned by
267
tree.get_file_with_stat()).
269
if self.new_inventory.root is None:
270
if ie.parent_id is not None:
271
raise errors.RootMissing()
272
self._check_root(ie, parent_invs, tree)
273
if ie.revision is None:
274
kind = content_summary[0]
276
# ie is carried over from a prior commit
278
# XXX: repository specific check for nested tree support goes here - if
279
# the repo doesn't want nested trees we skip it ?
280
if (kind == 'tree-reference' and
281
not self.repository._format.supports_tree_reference):
282
# mismatch between commit builder logic and repository:
283
# this needs the entry creation pushed down into the builder.
284
raise NotImplementedError('Missing repository subtree support.')
285
self.new_inventory.add(ie)
287
# TODO: slow, take it out of the inner loop.
289
basis_inv = parent_invs[0]
291
basis_inv = Inventory(root_id=None)
293
# ie.revision is always None if the InventoryEntry is considered
294
# for committing. We may record the previous parents revision if the
295
# content is actually unchanged against a sole head.
296
if ie.revision is not None:
297
if not self._versioned_root and path == '':
298
# repositories that do not version the root set the root's
299
# revision to the new commit even when no change occurs, and
300
# this masks when a change may have occurred against the basis,
301
# so calculate if one happened.
302
if ie.file_id in basis_inv:
303
delta = (basis_inv.id2path(ie.file_id), path,
307
delta = (None, path, ie.file_id, ie)
308
return delta, False, None
310
# we don't need to commit this, because the caller already
311
# determined that an existing revision of this file is
312
# appropriate. If its not being considered for committing then
313
# it and all its parents to the root must be unaltered so
314
# no-change against the basis.
315
if ie.revision == self._new_revision_id:
316
raise AssertionError("Impossible situation, a skipped "
317
"inventory entry (%r) claims to be modified in this "
318
"commit (%r).", (ie, self._new_revision_id))
319
return None, False, None
320
# XXX: Friction: parent_candidates should return a list not a dict
321
# so that we don't have to walk the inventories again.
322
parent_candiate_entries = ie.parent_candidates(parent_invs)
323
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
325
for inv in parent_invs:
326
if ie.file_id in inv:
327
old_rev = inv[ie.file_id].revision
328
if old_rev in head_set:
329
heads.append(inv[ie.file_id].revision)
330
head_set.remove(inv[ie.file_id].revision)
333
# now we check to see if we need to write a new record to the
335
# We write a new entry unless there is one head to the ancestors, and
336
# the kind-derived content is unchanged.
338
# Cheapest check first: no ancestors, or more the one head in the
339
# ancestors, we write a new node.
343
# There is a single head, look it up for comparison
344
parent_entry = parent_candiate_entries[heads[0]]
345
# if the non-content specific data has changed, we'll be writing a
347
if (parent_entry.parent_id != ie.parent_id or
348
parent_entry.name != ie.name):
350
# now we need to do content specific checks:
352
# if the kind changed the content obviously has
353
if kind != parent_entry.kind:
355
# Stat cache fingerprint feedback for the caller - None as we usually
356
# don't generate one.
359
if content_summary[2] is None:
360
raise ValueError("Files must not have executable = None")
362
if (# if the file length changed we have to store:
363
parent_entry.text_size != content_summary[1] or
364
# if the exec bit has changed we have to store:
365
parent_entry.executable != content_summary[2]):
367
elif parent_entry.text_sha1 == content_summary[3]:
368
# all meta and content is unchanged (using a hash cache
369
# hit to check the sha)
370
ie.revision = parent_entry.revision
371
ie.text_size = parent_entry.text_size
372
ie.text_sha1 = parent_entry.text_sha1
373
ie.executable = parent_entry.executable
374
return self._get_delta(ie, basis_inv, path), False, None
376
# Either there is only a hash change(no hash cache entry,
377
# or same size content change), or there is no change on
379
# Provide the parent's hash to the store layer, so that the
380
# content is unchanged we will not store a new node.
381
nostore_sha = parent_entry.text_sha1
383
# We want to record a new node regardless of the presence or
384
# absence of a content change in the file.
386
ie.executable = content_summary[2]
387
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
389
lines = file_obj.readlines()
393
ie.text_sha1, ie.text_size = self._add_text_to_weave(
394
ie.file_id, lines, heads, nostore_sha)
395
# Let the caller know we generated a stat fingerprint.
396
fingerprint = (ie.text_sha1, stat_value)
397
except errors.ExistingContent:
398
# Turns out that the file content was unchanged, and we were
399
# only going to store a new node if it was changed. Carry over
401
ie.revision = parent_entry.revision
402
ie.text_size = parent_entry.text_size
403
ie.text_sha1 = parent_entry.text_sha1
404
ie.executable = parent_entry.executable
405
return self._get_delta(ie, basis_inv, path), False, None
406
elif kind == 'directory':
408
# all data is meta here, nothing specific to directory, so
410
ie.revision = parent_entry.revision
411
return self._get_delta(ie, basis_inv, path), False, None
413
self._add_text_to_weave(ie.file_id, lines, heads, None)
414
elif kind == 'symlink':
415
current_link_target = content_summary[3]
417
# symlink target is not generic metadata, check if it has
419
if current_link_target != parent_entry.symlink_target:
422
# unchanged, carry over.
423
ie.revision = parent_entry.revision
424
ie.symlink_target = parent_entry.symlink_target
425
return self._get_delta(ie, basis_inv, path), False, None
426
ie.symlink_target = current_link_target
428
self._add_text_to_weave(ie.file_id, lines, heads, None)
429
elif kind == 'tree-reference':
431
if content_summary[3] != parent_entry.reference_revision:
434
# unchanged, carry over.
435
ie.reference_revision = parent_entry.reference_revision
436
ie.revision = parent_entry.revision
437
return self._get_delta(ie, basis_inv, path), False, None
438
ie.reference_revision = content_summary[3]
440
self._add_text_to_weave(ie.file_id, lines, heads, None)
442
raise NotImplementedError('unknown kind')
443
ie.revision = self._new_revision_id
444
return self._get_delta(ie, basis_inv, path), True, fingerprint
446
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
447
# Note: as we read the content directly from the tree, we know its not
448
# been turned into unicode or badly split - but a broken tree
449
# implementation could give us bad output from readlines() so this is
450
# not a guarantee of safety. What would be better is always checking
451
# the content during test suite execution. RBC 20070912
452
parent_keys = tuple((file_id, parent) for parent in parents)
453
return self.repository.texts.add_lines(
454
(file_id, self._new_revision_id), parent_keys, new_lines,
455
nostore_sha=nostore_sha, random_id=self.random_revid,
456
check_content=False)[0:2]
459
class RootCommitBuilder(CommitBuilder):
460
"""This commitbuilder actually records the root id"""
462
# the root entry gets versioned properly by this builder.
463
_versioned_root = True
465
def _check_root(self, ie, parent_invs, tree):
466
"""Helper for record_entry_contents.
468
:param ie: An entry being added.
469
:param parent_invs: The inventories of the parent revisions of the
471
:param tree: The tree that is being committed.
475
######################################################################
478
class Repository(object):
479
"""Repository holding history for one or more branches.
481
The repository holds and retrieves historical information including
482
revisions and file history. It's normally accessed only by the Branch,
483
which views a particular line of development through that history.
485
The Repository builds on top of some byte storage facilies (the revisions,
486
signatures, inventories and texts attributes) and a Transport, which
487
respectively provide byte storage and a means to access the (possibly
490
The byte storage facilities are addressed via tuples, which we refer to
491
as 'keys' throughout the code base. Revision_keys, inventory_keys and
492
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
493
(file_id, revision_id). We use this interface because it allows low
494
friction with the underlying code that implements disk indices, network
495
encoding and other parts of bzrlib.
497
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
498
the serialised revisions for the repository. This can be used to obtain
499
revision graph information or to access raw serialised revisions.
500
The result of trying to insert data into the repository via this store
501
is undefined: it should be considered read-only except for implementors
503
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
504
the serialised signatures for the repository. This can be used to
505
obtain access to raw serialised signatures. The result of trying to
506
insert data into the repository via this store is undefined: it should
507
be considered read-only except for implementors of repositories.
508
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
509
the serialised inventories for the repository. This can be used to
510
obtain unserialised inventories. The result of trying to insert data
511
into the repository via this store is undefined: it should be
512
considered read-only except for implementors of repositories.
513
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
514
texts of files and directories for the repository. This can be used to
515
obtain file texts or file graphs. Note that Repository.iter_file_bytes
516
is usually a better interface for accessing file texts.
517
The result of trying to insert data into the repository via this store
518
is undefined: it should be considered read-only except for implementors
520
:ivar _transport: Transport for file access to repository, typically
521
pointing to .bzr/repository.
524
# What class to use for a CommitBuilder. Often its simpler to change this
525
# in a Repository class subclass rather than to override
526
# get_commit_builder.
527
_commit_builder_class = CommitBuilder
528
# The search regex used by xml based repositories to determine what things
529
# where changed in a single commit.
530
_file_ids_altered_regex = lazy_regex.lazy_compile(
531
r'file_id="(?P<file_id>[^"]+)"'
532
r'.* revision="(?P<revision_id>[^"]+)"'
535
def abort_write_group(self, suppress_errors=False):
536
"""Commit the contents accrued within the current write group.
538
:param suppress_errors: if true, abort_write_group will catch and log
539
unexpected errors that happen during the abort, rather than
540
allowing them to propagate. Defaults to False.
542
:seealso: start_write_group.
544
if self._write_group is not self.get_transaction():
545
# has an unlock or relock occured ?
546
raise errors.BzrError('mismatched lock context and write group.')
548
self._abort_write_group()
549
except Exception, exc:
550
self._write_group = None
551
if not suppress_errors:
553
mutter('abort_write_group failed')
554
log_exception_quietly()
555
note('bzr: ERROR (ignored): %s', exc)
556
self._write_group = None
558
def _abort_write_group(self):
559
"""Template method for per-repository write group cleanup.
561
This is called during abort before the write group is considered to be
562
finished and should cleanup any internal state accrued during the write
563
group. There is no requirement that data handed to the repository be
564
*not* made available - this is not a rollback - but neither should any
565
attempt be made to ensure that data added is fully commited. Abort is
566
invoked when an error has occured so futher disk or network operations
567
may not be possible or may error and if possible should not be
571
def add_fallback_repository(self, repository):
572
"""Add a repository to use for looking up data not held locally.
574
:param repository: A repository.
576
if not self._format.supports_external_lookups:
577
raise errors.UnstackableRepositoryFormat(self._format, self.base)
578
self._check_fallback_repository(repository)
579
self._fallback_repositories.append(repository)
580
self.texts.add_fallback_versioned_files(repository.texts)
581
self.inventories.add_fallback_versioned_files(repository.inventories)
582
self.revisions.add_fallback_versioned_files(repository.revisions)
583
self.signatures.add_fallback_versioned_files(repository.signatures)
585
def _check_fallback_repository(self, repository):
586
"""Check that this repository can fallback to repository safely.
588
Raise an error if not.
590
:param repository: A repository to fallback to.
592
return InterRepository._assert_same_model(self, repository)
594
def add_inventory(self, revision_id, inv, parents):
595
"""Add the inventory inv to the repository as revision_id.
597
:param parents: The revision ids of the parents that revision_id
598
is known to have and are in the repository already.
600
:returns: The validator(which is a sha1 digest, though what is sha'd is
601
repository format specific) of the serialized inventory.
603
if not self.is_in_write_group():
604
raise AssertionError("%r not in write group" % (self,))
605
_mod_revision.check_not_reserved_id(revision_id)
606
if not (inv.revision_id is None or inv.revision_id == revision_id):
607
raise AssertionError(
608
"Mismatch between inventory revision"
609
" id and insertion revid (%r, %r)"
610
% (inv.revision_id, revision_id))
612
raise AssertionError()
613
inv_lines = self._serialise_inventory_to_lines(inv)
614
return self._inventory_add_lines(revision_id, parents,
615
inv_lines, check_content=False)
617
def _inventory_add_lines(self, revision_id, parents, lines,
619
"""Store lines in inv_vf and return the sha1 of the inventory."""
620
parents = [(parent,) for parent in parents]
621
return self.inventories.add_lines((revision_id,), parents, lines,
622
check_content=check_content)[0]
624
def add_revision(self, revision_id, rev, inv=None, config=None):
625
"""Add rev to the revision store as revision_id.
627
:param revision_id: the revision id to use.
628
:param rev: The revision object.
629
:param inv: The inventory for the revision. if None, it will be looked
630
up in the inventory storer
631
:param config: If None no digital signature will be created.
632
If supplied its signature_needed method will be used
633
to determine if a signature should be made.
635
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
637
_mod_revision.check_not_reserved_id(revision_id)
638
if config is not None and config.signature_needed():
640
inv = self.get_inventory(revision_id)
641
plaintext = Testament(rev, inv).as_short_text()
642
self.store_revision_signature(
643
gpg.GPGStrategy(config), plaintext, revision_id)
644
# check inventory present
645
if not self.inventories.get_parent_map([(revision_id,)]):
647
raise errors.WeaveRevisionNotPresent(revision_id,
650
# yes, this is not suitable for adding with ghosts.
651
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
655
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
656
self._add_revision(rev)
658
def _add_revision(self, revision):
659
text = self._serializer.write_revision_to_string(revision)
660
key = (revision.revision_id,)
661
parents = tuple((parent,) for parent in revision.parent_ids)
662
self.revisions.add_lines(key, parents, osutils.split_lines(text))
664
def all_revision_ids(self):
665
"""Returns a list of all the revision ids in the repository.
667
This is conceptually deprecated because code should generally work on
668
the graph reachable from a particular revision, and ignore any other
669
revisions that might be present. There is no direct replacement
672
if 'evil' in debug.debug_flags:
673
mutter_callsite(2, "all_revision_ids is linear with history.")
674
return self._all_revision_ids()
676
def _all_revision_ids(self):
677
"""Returns a list of all the revision ids in the repository.
679
These are in as much topological order as the underlying store can
682
raise NotImplementedError(self._all_revision_ids)
684
def break_lock(self):
685
"""Break a lock if one is present from another instance.
687
Uses the ui factory to ask for confirmation if the lock may be from
690
self.control_files.break_lock()
693
def _eliminate_revisions_not_present(self, revision_ids):
694
"""Check every revision id in revision_ids to see if we have it.
696
Returns a set of the present revisions.
699
graph = self.get_graph()
700
parent_map = graph.get_parent_map(revision_ids)
701
# The old API returned a list, should this actually be a set?
702
return parent_map.keys()
705
def create(a_bzrdir):
706
"""Construct the current default format repository in a_bzrdir."""
707
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
709
def __init__(self, _format, a_bzrdir, control_files):
710
"""instantiate a Repository.
712
:param _format: The format of the repository on disk.
713
:param a_bzrdir: The BzrDir of the repository.
715
In the future we will have a single api for all stores for
716
getting file texts, inventories and revisions, then
717
this construct will accept instances of those things.
719
super(Repository, self).__init__()
720
self._format = _format
721
# the following are part of the public API for Repository:
722
self.bzrdir = a_bzrdir
723
self.control_files = control_files
724
self._transport = control_files._transport
725
self.base = self._transport.base
727
self._reconcile_does_inventory_gc = True
728
self._reconcile_fixes_text_parents = False
729
self._reconcile_backsup_inventory = True
730
# not right yet - should be more semantically clear ?
732
# TODO: make sure to construct the right store classes, etc, depending
733
# on whether escaping is required.
734
self._warn_if_deprecated()
735
self._write_group = None
736
# Additional places to query for data.
737
self._fallback_repositories = []
738
# What order should fetch operations request streams in?
739
# The default is unordered as that is the cheapest for an origin to
741
self._fetch_order = 'unordered'
742
# Does this repository use deltas that can be fetched as-deltas ?
743
# (E.g. knits, where the knit deltas can be transplanted intact.
744
# We default to False, which will ensure that enough data to get
745
# a full text out of any fetch stream will be grabbed.
746
self._fetch_uses_deltas = False
747
# Should fetch trigger a reconcile after the fetch? Only needed for
748
# some repository formats that can suffer internal inconsistencies.
749
self._fetch_reconcile = False
752
return '%s(%r)' % (self.__class__.__name__,
755
def has_same_location(self, other):
756
"""Returns a boolean indicating if this repository is at the same
757
location as another repository.
759
This might return False even when two repository objects are accessing
760
the same physical repository via different URLs.
762
if self.__class__ is not other.__class__:
764
return (self._transport.base == other._transport.base)
766
def is_in_write_group(self):
767
"""Return True if there is an open write group.
769
:seealso: start_write_group.
771
return self._write_group is not None
774
return self.control_files.is_locked()
776
def is_write_locked(self):
777
"""Return True if this object is write locked."""
778
return self.is_locked() and self.control_files._lock_mode == 'w'
780
def lock_write(self, token=None):
781
"""Lock this repository for writing.
783
This causes caching within the repository obejct to start accumlating
784
data during reads, and allows a 'write_group' to be obtained. Write
785
groups must be used for actual data insertion.
787
:param token: if this is already locked, then lock_write will fail
788
unless the token matches the existing lock.
789
:returns: a token if this instance supports tokens, otherwise None.
790
:raises TokenLockingNotSupported: when a token is given but this
791
instance doesn't support using token locks.
792
:raises MismatchedToken: if the specified token doesn't match the token
793
of the existing lock.
794
:seealso: start_write_group.
796
A token should be passed in if you know that you have locked the object
797
some other way, and need to synchronise this object's state with that
800
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
802
result = self.control_files.lock_write(token=token)
803
for repo in self._fallback_repositories:
804
# Writes don't affect fallback repos
810
self.control_files.lock_read()
811
for repo in self._fallback_repositories:
815
def get_physical_lock_status(self):
816
return self.control_files.get_physical_lock_status()
818
def leave_lock_in_place(self):
819
"""Tell this repository not to release the physical lock when this
822
If lock_write doesn't return a token, then this method is not supported.
824
self.control_files.leave_in_place()
826
def dont_leave_lock_in_place(self):
827
"""Tell this repository to release the physical lock when this
828
object is unlocked, even if it didn't originally acquire it.
830
If lock_write doesn't return a token, then this method is not supported.
832
self.control_files.dont_leave_in_place()
835
def gather_stats(self, revid=None, committers=None):
836
"""Gather statistics from a revision id.
838
:param revid: The revision id to gather statistics from, if None, then
839
no revision specific statistics are gathered.
840
:param committers: Optional parameter controlling whether to grab
841
a count of committers from the revision specific statistics.
842
:return: A dictionary of statistics. Currently this contains:
843
committers: The number of committers if requested.
844
firstrev: A tuple with timestamp, timezone for the penultimate left
845
most ancestor of revid, if revid is not the NULL_REVISION.
846
latestrev: A tuple with timestamp, timezone for revid, if revid is
847
not the NULL_REVISION.
848
revisions: The total revision count in the repository.
849
size: An estimate disk size of the repository in bytes.
852
if revid and committers:
853
result['committers'] = 0
854
if revid and revid != _mod_revision.NULL_REVISION:
856
all_committers = set()
857
revisions = self.get_ancestry(revid)
858
# pop the leading None
860
first_revision = None
862
# ignore the revisions in the middle - just grab first and last
863
revisions = revisions[0], revisions[-1]
864
for revision in self.get_revisions(revisions):
865
if not first_revision:
866
first_revision = revision
868
all_committers.add(revision.committer)
869
last_revision = revision
871
result['committers'] = len(all_committers)
872
result['firstrev'] = (first_revision.timestamp,
873
first_revision.timezone)
874
result['latestrev'] = (last_revision.timestamp,
875
last_revision.timezone)
877
# now gather global repository information
878
# XXX: This is available for many repos regardless of listability.
879
if self.bzrdir.root_transport.listable():
880
# XXX: do we want to __define len__() ?
881
# Maybe the versionedfiles object should provide a different
882
# method to get the number of keys.
883
result['revisions'] = len(self.revisions.keys())
887
def find_branches(self, using=False):
888
"""Find branches underneath this repository.
890
This will include branches inside other branches.
892
:param using: If True, list only branches using this repository.
894
if using and not self.is_shared():
896
return [self.bzrdir.open_branch()]
897
except errors.NotBranchError:
899
class Evaluator(object):
902
self.first_call = True
904
def __call__(self, bzrdir):
905
# On the first call, the parameter is always the bzrdir
906
# containing the current repo.
907
if not self.first_call:
909
repository = bzrdir.open_repository()
910
except errors.NoRepositoryPresent:
913
return False, (None, repository)
914
self.first_call = False
916
value = (bzrdir.open_branch(), None)
917
except errors.NotBranchError:
922
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
923
self.bzrdir.root_transport, evaluate=Evaluator()):
924
if branch is not None:
925
branches.append(branch)
926
if not using and repository is not None:
927
branches.extend(repository.find_branches())
931
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
932
"""Return the revision ids that other has that this does not.
934
These are returned in topological order.
936
revision_id: only return revision ids included by revision_id.
938
return InterRepository.get(other, self).search_missing_revision_ids(
939
revision_id, find_ghosts)
941
@deprecated_method(one_two)
943
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
944
"""Return the revision ids that other has that this does not.
946
These are returned in topological order.
948
revision_id: only return revision ids included by revision_id.
950
keys = self.search_missing_revision_ids(
951
other, revision_id, find_ghosts).get_keys()
954
parents = other.get_graph().get_parent_map(keys)
957
return tsort.topo_sort(parents)
961
"""Open the repository rooted at base.
963
For instance, if the repository is at URL/.bzr/repository,
964
Repository.open(URL) -> a Repository instance.
966
control = bzrdir.BzrDir.open(base)
967
return control.open_repository()
969
def copy_content_into(self, destination, revision_id=None):
970
"""Make a complete copy of the content in self into destination.
972
This is a destructive operation! Do not use it on existing
975
return InterRepository.get(self, destination).copy_content(revision_id)
977
def commit_write_group(self):
978
"""Commit the contents accrued within the current write group.
980
:seealso: start_write_group.
982
if self._write_group is not self.get_transaction():
983
# has an unlock or relock occured ?
984
raise errors.BzrError('mismatched lock context %r and '
986
(self.get_transaction(), self._write_group))
987
self._commit_write_group()
988
self._write_group = None
990
def _commit_write_group(self):
991
"""Template method for per-repository write group cleanup.
993
This is called before the write group is considered to be
994
finished and should ensure that all data handed to the repository
995
for writing during the write group is safely committed (to the
996
extent possible considering file system caching etc).
999
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1000
"""Fetch the content required to construct revision_id from source.
1002
If revision_id is None all content is copied.
1003
:param find_ghosts: Find and copy revisions in the source that are
1004
ghosts in the target (and not reachable directly by walking out to
1005
the first-present revision in target from revision_id).
1007
# fast path same-url fetch operations
1008
if self.has_same_location(source):
1009
# check that last_revision is in 'from' and then return a
1011
if (revision_id is not None and
1012
not _mod_revision.is_null(revision_id)):
1013
self.get_revision(revision_id)
1015
# if there is no specific appropriate InterRepository, this will get
1016
# the InterRepository base class, which raises an
1017
# IncompatibleRepositories when asked to fetch.
1018
inter = InterRepository.get(source, self)
1019
return inter.fetch(revision_id=revision_id, pb=pb,
1020
find_ghosts=find_ghosts)
1022
def create_bundle(self, target, base, fileobj, format=None):
1023
return serializer.write_bundle(self, target, base, fileobj, format)
1025
def get_commit_builder(self, branch, parents, config, timestamp=None,
1026
timezone=None, committer=None, revprops=None,
1028
"""Obtain a CommitBuilder for this repository.
1030
:param branch: Branch to commit to.
1031
:param parents: Revision ids of the parents of the new revision.
1032
:param config: Configuration to use.
1033
:param timestamp: Optional timestamp recorded for commit.
1034
:param timezone: Optional timezone for timestamp.
1035
:param committer: Optional committer to set for commit.
1036
:param revprops: Optional dictionary of revision properties.
1037
:param revision_id: Optional revision id.
1039
result = self._commit_builder_class(self, parents, config,
1040
timestamp, timezone, committer, revprops, revision_id)
1041
self.start_write_group()
1045
if (self.control_files._lock_count == 1 and
1046
self.control_files._lock_mode == 'w'):
1047
if self._write_group is not None:
1048
self.abort_write_group()
1049
self.control_files.unlock()
1050
raise errors.BzrError(
1051
'Must end write groups before releasing write locks.')
1052
self.control_files.unlock()
1053
for repo in self._fallback_repositories:
1057
def clone(self, a_bzrdir, revision_id=None):
1058
"""Clone this repository into a_bzrdir using the current format.
1060
Currently no check is made that the format of this repository and
1061
the bzrdir format are compatible. FIXME RBC 20060201.
1063
:return: The newly created destination repository.
1065
# TODO: deprecate after 0.16; cloning this with all its settings is
1066
# probably not very useful -- mbp 20070423
1067
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1068
self.copy_content_into(dest_repo, revision_id)
1071
def start_write_group(self):
1072
"""Start a write group in the repository.
1074
Write groups are used by repositories which do not have a 1:1 mapping
1075
between file ids and backend store to manage the insertion of data from
1076
both fetch and commit operations.
1078
A write lock is required around the start_write_group/commit_write_group
1079
for the support of lock-requiring repository formats.
1081
One can only insert data into a repository inside a write group.
1085
if not self.is_write_locked():
1086
raise errors.NotWriteLocked(self)
1087
if self._write_group:
1088
raise errors.BzrError('already in a write group')
1089
self._start_write_group()
1090
# so we can detect unlock/relock - the write group is now entered.
1091
self._write_group = self.get_transaction()
1093
def _start_write_group(self):
1094
"""Template method for per-repository write group startup.
1096
This is called before the write group is considered to be
1101
def sprout(self, to_bzrdir, revision_id=None):
1102
"""Create a descendent repository for new development.
1104
Unlike clone, this does not copy the settings of the repository.
1106
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1107
dest_repo.fetch(self, revision_id=revision_id)
1110
def _create_sprouting_repo(self, a_bzrdir, shared):
1111
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1112
# use target default format.
1113
dest_repo = a_bzrdir.create_repository()
1115
# Most control formats need the repository to be specifically
1116
# created, but on some old all-in-one formats it's not needed
1118
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1119
except errors.UninitializableFormat:
1120
dest_repo = a_bzrdir.open_repository()
1124
def has_revision(self, revision_id):
1125
"""True if this repository has a copy of the revision."""
1126
return revision_id in self.has_revisions((revision_id,))
1129
def has_revisions(self, revision_ids):
1130
"""Probe to find out the presence of multiple revisions.
1132
:param revision_ids: An iterable of revision_ids.
1133
:return: A set of the revision_ids that were present.
1135
parent_map = self.revisions.get_parent_map(
1136
[(rev_id,) for rev_id in revision_ids])
1138
if _mod_revision.NULL_REVISION in revision_ids:
1139
result.add(_mod_revision.NULL_REVISION)
1140
result.update([key[0] for key in parent_map])
1144
def get_revision(self, revision_id):
1145
"""Return the Revision object for a named revision."""
1146
return self.get_revisions([revision_id])[0]
1149
def get_revision_reconcile(self, revision_id):
1150
"""'reconcile' helper routine that allows access to a revision always.
1152
This variant of get_revision does not cross check the weave graph
1153
against the revision one as get_revision does: but it should only
1154
be used by reconcile, or reconcile-alike commands that are correcting
1155
or testing the revision graph.
1157
return self._get_revisions([revision_id])[0]
1160
def get_revisions(self, revision_ids):
1161
"""Get many revisions at once."""
1162
return self._get_revisions(revision_ids)
1165
def _get_revisions(self, revision_ids):
1166
"""Core work logic to get many revisions without sanity checks."""
1167
for rev_id in revision_ids:
1168
if not rev_id or not isinstance(rev_id, basestring):
1169
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1170
keys = [(key,) for key in revision_ids]
1171
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1173
for record in stream:
1174
if record.storage_kind == 'absent':
1175
raise errors.NoSuchRevision(self, record.key[0])
1176
text = record.get_bytes_as('fulltext')
1177
rev = self._serializer.read_revision_from_string(text)
1178
revs[record.key[0]] = rev
1179
return [revs[revid] for revid in revision_ids]
1182
def get_revision_xml(self, revision_id):
1183
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1184
# would have already do it.
1185
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1186
rev = self.get_revision(revision_id)
1187
rev_tmp = cStringIO.StringIO()
1188
# the current serializer..
1189
self._serializer.write_revision(rev, rev_tmp)
1191
return rev_tmp.getvalue()
1193
def get_deltas_for_revisions(self, revisions):
1194
"""Produce a generator of revision deltas.
1196
Note that the input is a sequence of REVISIONS, not revision_ids.
1197
Trees will be held in memory until the generator exits.
1198
Each delta is relative to the revision's lefthand predecessor.
1200
required_trees = set()
1201
for revision in revisions:
1202
required_trees.add(revision.revision_id)
1203
required_trees.update(revision.parent_ids[:1])
1204
trees = dict((t.get_revision_id(), t) for
1205
t in self.revision_trees(required_trees))
1206
for revision in revisions:
1207
if not revision.parent_ids:
1208
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1210
old_tree = trees[revision.parent_ids[0]]
1211
yield trees[revision.revision_id].changes_from(old_tree)
1214
def get_revision_delta(self, revision_id):
1215
"""Return the delta for one revision.
1217
The delta is relative to the left-hand predecessor of the
1220
r = self.get_revision(revision_id)
1221
return list(self.get_deltas_for_revisions([r]))[0]
1224
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1225
signature = gpg_strategy.sign(plaintext)
1226
self.add_signature_text(revision_id, signature)
1229
def add_signature_text(self, revision_id, signature):
1230
self.signatures.add_lines((revision_id,), (),
1231
osutils.split_lines(signature))
1233
def find_text_key_references(self):
1234
"""Find the text key references within the repository.
1236
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1237
revision_ids. Each altered file-ids has the exact revision_ids that
1238
altered it listed explicitly.
1239
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1240
to whether they were referred to by the inventory of the
1241
revision_id that they contain. The inventory texts from all present
1242
revision ids are assessed to generate this report.
1244
revision_keys = self.revisions.keys()
1245
w = self.inventories
1246
pb = ui.ui_factory.nested_progress_bar()
1248
return self._find_text_key_references_from_xml_inventory_lines(
1249
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1253
def _find_text_key_references_from_xml_inventory_lines(self,
1255
"""Core routine for extracting references to texts from inventories.
1257
This performs the translation of xml lines to revision ids.
1259
:param line_iterator: An iterator of lines, origin_version_id
1260
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1261
to whether they were referred to by the inventory of the
1262
revision_id that they contain. Note that if that revision_id was
1263
not part of the line_iterator's output then False will be given -
1264
even though it may actually refer to that key.
1266
if not self._serializer.support_altered_by_hack:
1267
raise AssertionError(
1268
"_find_text_key_references_from_xml_inventory_lines only "
1269
"supported for branches which store inventory as unnested xml"
1270
", not on %r" % self)
1273
# this code needs to read every new line in every inventory for the
1274
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1275
# not present in one of those inventories is unnecessary but not
1276
# harmful because we are filtering by the revision id marker in the
1277
# inventory lines : we only select file ids altered in one of those
1278
# revisions. We don't need to see all lines in the inventory because
1279
# only those added in an inventory in rev X can contain a revision=X
1281
unescape_revid_cache = {}
1282
unescape_fileid_cache = {}
1284
# jam 20061218 In a big fetch, this handles hundreds of thousands
1285
# of lines, so it has had a lot of inlining and optimizing done.
1286
# Sorry that it is a little bit messy.
1287
# Move several functions to be local variables, since this is a long
1289
search = self._file_ids_altered_regex.search
1290
unescape = _unescape_xml
1291
setdefault = result.setdefault
1292
for line, line_key in line_iterator:
1293
match = search(line)
1296
# One call to match.group() returning multiple items is quite a
1297
# bit faster than 2 calls to match.group() each returning 1
1298
file_id, revision_id = match.group('file_id', 'revision_id')
1300
# Inlining the cache lookups helps a lot when you make 170,000
1301
# lines and 350k ids, versus 8.4 unique ids.
1302
# Using a cache helps in 2 ways:
1303
# 1) Avoids unnecessary decoding calls
1304
# 2) Re-uses cached strings, which helps in future set and
1306
# (2) is enough that removing encoding entirely along with
1307
# the cache (so we are using plain strings) results in no
1308
# performance improvement.
1310
revision_id = unescape_revid_cache[revision_id]
1312
unescaped = unescape(revision_id)
1313
unescape_revid_cache[revision_id] = unescaped
1314
revision_id = unescaped
1316
# Note that unconditionally unescaping means that we deserialise
1317
# every fileid, which for general 'pull' is not great, but we don't
1318
# really want to have some many fulltexts that this matters anyway.
1321
file_id = unescape_fileid_cache[file_id]
1323
unescaped = unescape(file_id)
1324
unescape_fileid_cache[file_id] = unescaped
1327
key = (file_id, revision_id)
1328
setdefault(key, False)
1329
if revision_id == line_key[-1]:
1333
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1335
"""Helper routine for fileids_altered_by_revision_ids.
1337
This performs the translation of xml lines to revision ids.
1339
:param line_iterator: An iterator of lines, origin_version_id
1340
:param revision_ids: The revision ids to filter for. This should be a
1341
set or other type which supports efficient __contains__ lookups, as
1342
the revision id from each parsed line will be looked up in the
1343
revision_ids filter.
1344
:return: a dictionary mapping altered file-ids to an iterable of
1345
revision_ids. Each altered file-ids has the exact revision_ids that
1346
altered it listed explicitly.
1349
setdefault = result.setdefault
1351
self._find_text_key_references_from_xml_inventory_lines(
1352
line_iterator).iterkeys():
1353
# once data is all ensured-consistent; then this is
1354
# if revision_id == version_id
1355
if key[-1:] in revision_ids:
1356
setdefault(key[0], set()).add(key[-1])
1359
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1360
"""Find the file ids and versions affected by revisions.
1362
:param revisions: an iterable containing revision ids.
1363
:param _inv_weave: The inventory weave from this repository or None.
1364
If None, the inventory weave will be opened automatically.
1365
:return: a dictionary mapping altered file-ids to an iterable of
1366
revision_ids. Each altered file-ids has the exact revision_ids that
1367
altered it listed explicitly.
1369
selected_keys = set((revid,) for revid in revision_ids)
1370
w = _inv_weave or self.inventories
1371
pb = ui.ui_factory.nested_progress_bar()
1373
return self._find_file_ids_from_xml_inventory_lines(
1374
w.iter_lines_added_or_present_in_keys(
1375
selected_keys, pb=pb),
1380
def iter_files_bytes(self, desired_files):
1381
"""Iterate through file versions.
1383
Files will not necessarily be returned in the order they occur in
1384
desired_files. No specific order is guaranteed.
1386
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1387
value supplied by the caller as part of desired_files. It should
1388
uniquely identify the file version in the caller's context. (Examples:
1389
an index number or a TreeTransform trans_id.)
1391
bytes_iterator is an iterable of bytestrings for the file. The
1392
kind of iterable and length of the bytestrings are unspecified, but for
1393
this implementation, it is a list of bytes produced by
1394
VersionedFile.get_record_stream().
1396
:param desired_files: a list of (file_id, revision_id, identifier)
1399
transaction = self.get_transaction()
1401
for file_id, revision_id, callable_data in desired_files:
1402
text_keys[(file_id, revision_id)] = callable_data
1403
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1404
if record.storage_kind == 'absent':
1405
raise errors.RevisionNotPresent(record.key, self)
1406
yield text_keys[record.key], record.get_bytes_as('fulltext')
1408
def _generate_text_key_index(self, text_key_references=None,
1410
"""Generate a new text key index for the repository.
1412
This is an expensive function that will take considerable time to run.
1414
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1415
list of parents, also text keys. When a given key has no parents,
1416
the parents list will be [NULL_REVISION].
1418
# All revisions, to find inventory parents.
1419
if ancestors is None:
1420
graph = self.get_graph()
1421
ancestors = graph.get_parent_map(self.all_revision_ids())
1422
if text_key_references is None:
1423
text_key_references = self.find_text_key_references()
1424
pb = ui.ui_factory.nested_progress_bar()
1426
return self._do_generate_text_key_index(ancestors,
1427
text_key_references, pb)
1431
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1432
"""Helper for _generate_text_key_index to avoid deep nesting."""
1433
revision_order = tsort.topo_sort(ancestors)
1434
invalid_keys = set()
1436
for revision_id in revision_order:
1437
revision_keys[revision_id] = set()
1438
text_count = len(text_key_references)
1439
# a cache of the text keys to allow reuse; costs a dict of all the
1440
# keys, but saves a 2-tuple for every child of a given key.
1442
for text_key, valid in text_key_references.iteritems():
1444
invalid_keys.add(text_key)
1446
revision_keys[text_key[1]].add(text_key)
1447
text_key_cache[text_key] = text_key
1448
del text_key_references
1450
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1451
NULL_REVISION = _mod_revision.NULL_REVISION
1452
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1453
# too small for large or very branchy trees. However, for 55K path
1454
# trees, it would be easy to use too much memory trivially. Ideally we
1455
# could gauge this by looking at available real memory etc, but this is
1456
# always a tricky proposition.
1457
inventory_cache = lru_cache.LRUCache(10)
1458
batch_size = 10 # should be ~150MB on a 55K path tree
1459
batch_count = len(revision_order) / batch_size + 1
1461
pb.update("Calculating text parents.", processed_texts, text_count)
1462
for offset in xrange(batch_count):
1463
to_query = revision_order[offset * batch_size:(offset + 1) *
1467
for rev_tree in self.revision_trees(to_query):
1468
revision_id = rev_tree.get_revision_id()
1469
parent_ids = ancestors[revision_id]
1470
for text_key in revision_keys[revision_id]:
1471
pb.update("Calculating text parents.", processed_texts)
1472
processed_texts += 1
1473
candidate_parents = []
1474
for parent_id in parent_ids:
1475
parent_text_key = (text_key[0], parent_id)
1477
check_parent = parent_text_key not in \
1478
revision_keys[parent_id]
1480
# the parent parent_id is a ghost:
1481
check_parent = False
1482
# truncate the derived graph against this ghost.
1483
parent_text_key = None
1485
# look at the parent commit details inventories to
1486
# determine possible candidates in the per file graph.
1489
inv = inventory_cache[parent_id]
1491
inv = self.revision_tree(parent_id).inventory
1492
inventory_cache[parent_id] = inv
1493
parent_entry = inv._byid.get(text_key[0], None)
1494
if parent_entry is not None:
1496
text_key[0], parent_entry.revision)
1498
parent_text_key = None
1499
if parent_text_key is not None:
1500
candidate_parents.append(
1501
text_key_cache[parent_text_key])
1502
parent_heads = text_graph.heads(candidate_parents)
1503
new_parents = list(parent_heads)
1504
new_parents.sort(key=lambda x:candidate_parents.index(x))
1505
if new_parents == []:
1506
new_parents = [NULL_REVISION]
1507
text_index[text_key] = new_parents
1509
for text_key in invalid_keys:
1510
text_index[text_key] = [NULL_REVISION]
1513
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1514
"""Get an iterable listing the keys of all the data introduced by a set
1517
The keys will be ordered so that the corresponding items can be safely
1518
fetched and inserted in that order.
1520
:returns: An iterable producing tuples of (knit-kind, file-id,
1521
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1522
'revisions'. file-id is None unless knit-kind is 'file'.
1524
# XXX: it's a bit weird to control the inventory weave caching in this
1525
# generator. Ideally the caching would be done in fetch.py I think. Or
1526
# maybe this generator should explicitly have the contract that it
1527
# should not be iterated until the previously yielded item has been
1529
inv_w = self.inventories
1531
# file ids that changed
1532
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1534
num_file_ids = len(file_ids)
1535
for file_id, altered_versions in file_ids.iteritems():
1536
if _files_pb is not None:
1537
_files_pb.update("fetch texts", count, num_file_ids)
1539
yield ("file", file_id, altered_versions)
1540
# We're done with the files_pb. Note that it finished by the caller,
1541
# just as it was created by the caller.
1545
yield ("inventory", None, revision_ids)
1548
# XXX: Note ATM no callers actually pay attention to this return
1549
# instead they just use the list of revision ids and ignore
1550
# missing sigs. Consider removing this work entirely
1551
revisions_with_signatures = set(self.signatures.get_parent_map(
1552
[(r,) for r in revision_ids]))
1553
revisions_with_signatures = set(
1554
[r for (r,) in revisions_with_signatures])
1555
revisions_with_signatures.intersection_update(revision_ids)
1556
yield ("signatures", None, revisions_with_signatures)
1559
yield ("revisions", None, revision_ids)
1562
def get_inventory(self, revision_id):
1563
"""Get Inventory object by revision id."""
1564
return self.iter_inventories([revision_id]).next()
1566
def iter_inventories(self, revision_ids):
1567
"""Get many inventories by revision_ids.
1569
This will buffer some or all of the texts used in constructing the
1570
inventories in memory, but will only parse a single inventory at a
1573
:return: An iterator of inventories.
1575
if ((None in revision_ids)
1576
or (_mod_revision.NULL_REVISION in revision_ids)):
1577
raise ValueError('cannot get null revision inventory')
1578
return self._iter_inventories(revision_ids)
1580
def _iter_inventories(self, revision_ids):
1581
"""single-document based inventory iteration."""
1582
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1583
yield self.deserialise_inventory(revision_id, text)
1585
def _iter_inventory_xmls(self, revision_ids):
1586
keys = [(revision_id,) for revision_id in revision_ids]
1587
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1589
for record in stream:
1590
if record.storage_kind != 'absent':
1591
texts[record.key] = record.get_bytes_as('fulltext')
1593
raise errors.NoSuchRevision(self, record.key)
1595
yield texts[key], key[-1]
1597
def deserialise_inventory(self, revision_id, xml):
1598
"""Transform the xml into an inventory object.
1600
:param revision_id: The expected revision id of the inventory.
1601
:param xml: A serialised inventory.
1603
result = self._serializer.read_inventory_from_string(xml, revision_id)
1604
if result.revision_id != revision_id:
1605
raise AssertionError('revision id mismatch %s != %s' % (
1606
result.revision_id, revision_id))
1609
def serialise_inventory(self, inv):
1610
return self._serializer.write_inventory_to_string(inv)
1612
def _serialise_inventory_to_lines(self, inv):
1613
return self._serializer.write_inventory_to_lines(inv)
1615
def get_serializer_format(self):
1616
return self._serializer.format_num
1619
def get_inventory_xml(self, revision_id):
1620
"""Get inventory XML as a file object."""
1621
texts = self._iter_inventory_xmls([revision_id])
1623
text, revision_id = texts.next()
1624
except StopIteration:
1625
raise errors.HistoryMissing(self, 'inventory', revision_id)
1629
def get_inventory_sha1(self, revision_id):
1630
"""Return the sha1 hash of the inventory entry
1632
return self.get_revision(revision_id).inventory_sha1
1634
def iter_reverse_revision_history(self, revision_id):
1635
"""Iterate backwards through revision ids in the lefthand history
1637
:param revision_id: The revision id to start with. All its lefthand
1638
ancestors will be traversed.
1640
graph = self.get_graph()
1641
next_id = revision_id
1643
if next_id in (None, _mod_revision.NULL_REVISION):
1646
# Note: The following line may raise KeyError in the event of
1647
# truncated history. We decided not to have a try:except:raise
1648
# RevisionNotPresent here until we see a use for it, because of the
1649
# cost in an inner loop that is by its very nature O(history).
1650
# Robert Collins 20080326
1651
parents = graph.get_parent_map([next_id])[next_id]
1652
if len(parents) == 0:
1655
next_id = parents[0]
1658
def get_revision_inventory(self, revision_id):
1659
"""Return inventory of a past revision."""
1660
# TODO: Unify this with get_inventory()
1661
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1662
# must be the same as its revision, so this is trivial.
1663
if revision_id is None:
1664
# This does not make sense: if there is no revision,
1665
# then it is the current tree inventory surely ?!
1666
# and thus get_root_id() is something that looks at the last
1667
# commit on the branch, and the get_root_id is an inventory check.
1668
raise NotImplementedError
1669
# return Inventory(self.get_root_id())
1671
return self.get_inventory(revision_id)
1673
def is_shared(self):
1674
"""Return True if this repository is flagged as a shared repository."""
1675
raise NotImplementedError(self.is_shared)
1678
def reconcile(self, other=None, thorough=False):
1679
"""Reconcile this repository."""
1680
from bzrlib.reconcile import RepoReconciler
1681
reconciler = RepoReconciler(self, thorough=thorough)
1682
reconciler.reconcile()
1685
def _refresh_data(self):
1686
"""Helper called from lock_* to ensure coherency with disk.
1688
The default implementation does nothing; it is however possible
1689
for repositories to maintain loaded indices across multiple locks
1690
by checking inside their implementation of this method to see
1691
whether their indices are still valid. This depends of course on
1692
the disk format being validatable in this manner.
1696
def revision_tree(self, revision_id):
1697
"""Return Tree for a revision on this branch.
1699
`revision_id` may be NULL_REVISION for the empty tree revision.
1701
revision_id = _mod_revision.ensure_null(revision_id)
1702
# TODO: refactor this to use an existing revision object
1703
# so we don't need to read it in twice.
1704
if revision_id == _mod_revision.NULL_REVISION:
1705
return RevisionTree(self, Inventory(root_id=None),
1706
_mod_revision.NULL_REVISION)
1708
inv = self.get_revision_inventory(revision_id)
1709
return RevisionTree(self, inv, revision_id)
1711
def revision_trees(self, revision_ids):
1712
"""Return Tree for a revision on this branch.
1714
`revision_id` may not be None or 'null:'"""
1715
inventories = self.iter_inventories(revision_ids)
1716
for inv in inventories:
1717
yield RevisionTree(self, inv, inv.revision_id)
1720
def get_ancestry(self, revision_id, topo_sorted=True):
1721
"""Return a list of revision-ids integrated by a revision.
1723
The first element of the list is always None, indicating the origin
1724
revision. This might change when we have history horizons, or
1725
perhaps we should have a new API.
1727
This is topologically sorted.
1729
if _mod_revision.is_null(revision_id):
1731
if not self.has_revision(revision_id):
1732
raise errors.NoSuchRevision(self, revision_id)
1733
graph = self.get_graph()
1735
search = graph._make_breadth_first_searcher([revision_id])
1738
found, ghosts = search.next_with_ghosts()
1739
except StopIteration:
1742
if _mod_revision.NULL_REVISION in keys:
1743
keys.remove(_mod_revision.NULL_REVISION)
1745
parent_map = graph.get_parent_map(keys)
1746
keys = tsort.topo_sort(parent_map)
1747
return [None] + list(keys)
1750
"""Compress the data within the repository.
1752
This operation only makes sense for some repository types. For other
1753
types it should be a no-op that just returns.
1755
This stub method does not require a lock, but subclasses should use
1756
@needs_write_lock as this is a long running call its reasonable to
1757
implicitly lock for the user.
1761
@deprecated_method(one_six)
1762
def print_file(self, file, revision_id):
1763
"""Print `file` to stdout.
1765
FIXME RBC 20060125 as John Meinel points out this is a bad api
1766
- it writes to stdout, it assumes that that is valid etc. Fix
1767
by creating a new more flexible convenience function.
1769
tree = self.revision_tree(revision_id)
1770
# use inventory as it was in that revision
1771
file_id = tree.inventory.path2id(file)
1773
# TODO: jam 20060427 Write a test for this code path
1774
# it had a bug in it, and was raising the wrong
1776
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1777
tree.print_file(file_id)
1779
def get_transaction(self):
1780
return self.control_files.get_transaction()
1782
@deprecated_method(one_one)
1783
def get_parents(self, revision_ids):
1784
"""See StackedParentsProvider.get_parents"""
1785
parent_map = self.get_parent_map(revision_ids)
1786
return [parent_map.get(r, None) for r in revision_ids]
1788
def get_parent_map(self, revision_ids):
1789
"""See graph._StackedParentsProvider.get_parent_map"""
1790
# revisions index works in keys; this just works in revisions
1791
# therefore wrap and unwrap
1794
for revision_id in revision_ids:
1795
if revision_id == _mod_revision.NULL_REVISION:
1796
result[revision_id] = ()
1797
elif revision_id is None:
1798
raise ValueError('get_parent_map(None) is not valid')
1800
query_keys.append((revision_id ,))
1801
for ((revision_id,), parent_keys) in \
1802
self.revisions.get_parent_map(query_keys).iteritems():
1804
result[revision_id] = tuple(parent_revid
1805
for (parent_revid,) in parent_keys)
1807
result[revision_id] = (_mod_revision.NULL_REVISION,)
1810
def _make_parents_provider(self):
1813
def get_graph(self, other_repository=None):
1814
"""Return the graph walker for this repository format"""
1815
parents_provider = self._make_parents_provider()
1816
if (other_repository is not None and
1817
not self.has_same_location(other_repository)):
1818
parents_provider = graph._StackedParentsProvider(
1819
[parents_provider, other_repository._make_parents_provider()])
1820
return graph.Graph(parents_provider)
1822
def _get_versioned_file_checker(self):
1823
"""Return an object suitable for checking versioned files."""
1824
return _VersionedFileChecker(self)
1826
def revision_ids_to_search_result(self, result_set):
1827
"""Convert a set of revision ids to a graph SearchResult."""
1828
result_parents = set()
1829
for parents in self.get_graph().get_parent_map(
1830
result_set).itervalues():
1831
result_parents.update(parents)
1832
included_keys = result_set.intersection(result_parents)
1833
start_keys = result_set.difference(included_keys)
1834
exclude_keys = result_parents.difference(result_set)
1835
result = graph.SearchResult(start_keys, exclude_keys,
1836
len(result_set), result_set)
1840
def set_make_working_trees(self, new_value):
1841
"""Set the policy flag for making working trees when creating branches.
1843
This only applies to branches that use this repository.
1845
The default is 'True'.
1846
:param new_value: True to restore the default, False to disable making
1849
raise NotImplementedError(self.set_make_working_trees)
1851
def make_working_trees(self):
1852
"""Returns the policy for making working trees on new branches."""
1853
raise NotImplementedError(self.make_working_trees)
1856
def sign_revision(self, revision_id, gpg_strategy):
1857
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1858
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1861
def has_signature_for_revision_id(self, revision_id):
1862
"""Query for a revision signature for revision_id in the repository."""
1863
if not self.has_revision(revision_id):
1864
raise errors.NoSuchRevision(self, revision_id)
1865
sig_present = (1 == len(
1866
self.signatures.get_parent_map([(revision_id,)])))
1870
def get_signature_text(self, revision_id):
1871
"""Return the text for a signature."""
1872
stream = self.signatures.get_record_stream([(revision_id,)],
1874
record = stream.next()
1875
if record.storage_kind == 'absent':
1876
raise errors.NoSuchRevision(self, revision_id)
1877
return record.get_bytes_as('fulltext')
1880
def check(self, revision_ids=None):
1881
"""Check consistency of all history of given revision_ids.
1883
Different repository implementations should override _check().
1885
:param revision_ids: A non-empty list of revision_ids whose ancestry
1886
will be checked. Typically the last revision_id of a branch.
1888
return self._check(revision_ids)
1890
def _check(self, revision_ids):
1891
result = check.Check(self)
1895
def _warn_if_deprecated(self):
1896
global _deprecation_warning_done
1897
if _deprecation_warning_done:
1899
_deprecation_warning_done = True
1900
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1901
% (self._format, self.bzrdir.transport.base))
1903
def supports_rich_root(self):
1904
return self._format.rich_root_data
1906
def _check_ascii_revisionid(self, revision_id, method):
1907
"""Private helper for ascii-only repositories."""
1908
# weave repositories refuse to store revisionids that are non-ascii.
1909
if revision_id is not None:
1910
# weaves require ascii revision ids.
1911
if isinstance(revision_id, unicode):
1913
revision_id.encode('ascii')
1914
except UnicodeEncodeError:
1915
raise errors.NonAsciiRevisionId(method, self)
1918
revision_id.decode('ascii')
1919
except UnicodeDecodeError:
1920
raise errors.NonAsciiRevisionId(method, self)
1922
def revision_graph_can_have_wrong_parents(self):
1923
"""Is it possible for this repository to have a revision graph with
1926
If True, then this repository must also implement
1927
_find_inconsistent_revision_parents so that check and reconcile can
1928
check for inconsistencies before proceeding with other checks that may
1929
depend on the revision index being consistent.
1931
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1934
# remove these delegates a while after bzr 0.15
1935
def __make_delegated(name, from_module):
1936
def _deprecated_repository_forwarder():
1937
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1938
% (name, from_module),
1941
m = __import__(from_module, globals(), locals(), [name])
1943
return getattr(m, name)
1944
except AttributeError:
1945
raise AttributeError('module %s has no name %s'
1947
globals()[name] = _deprecated_repository_forwarder
1950
'AllInOneRepository',
1951
'WeaveMetaDirRepository',
1952
'PreSplitOutRepositoryFormat',
1953
'RepositoryFormat4',
1954
'RepositoryFormat5',
1955
'RepositoryFormat6',
1956
'RepositoryFormat7',
1958
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1962
'RepositoryFormatKnit',
1963
'RepositoryFormatKnit1',
1965
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1968
def install_revision(repository, rev, revision_tree):
1969
"""Install all revision data into a repository."""
1970
install_revisions(repository, [(rev, revision_tree, None)])
1973
def install_revisions(repository, iterable, num_revisions=None, pb=None):
1974
"""Install all revision data into a repository.
1976
Accepts an iterable of revision, tree, signature tuples. The signature
1979
repository.start_write_group()
1981
for n, (revision, revision_tree, signature) in enumerate(iterable):
1982
_install_revision(repository, revision, revision_tree, signature)
1984
pb.update('Transferring revisions', n + 1, num_revisions)
1986
repository.abort_write_group()
1989
repository.commit_write_group()
1992
def _install_revision(repository, rev, revision_tree, signature):
1993
"""Install all revision data into a repository."""
1994
present_parents = []
1996
for p_id in rev.parent_ids:
1997
if repository.has_revision(p_id):
1998
present_parents.append(p_id)
1999
parent_trees[p_id] = repository.revision_tree(p_id)
2001
parent_trees[p_id] = repository.revision_tree(
2002
_mod_revision.NULL_REVISION)
2004
inv = revision_tree.inventory
2005
entries = inv.iter_entries()
2006
# backwards compatibility hack: skip the root id.
2007
if not repository.supports_rich_root():
2008
path, root = entries.next()
2009
if root.revision != rev.revision_id:
2010
raise errors.IncompatibleRevision(repr(repository))
2012
for path, ie in entries:
2013
text_keys[(ie.file_id, ie.revision)] = ie
2014
text_parent_map = repository.texts.get_parent_map(text_keys)
2015
missing_texts = set(text_keys) - set(text_parent_map)
2016
# Add the texts that are not already present
2017
for text_key in missing_texts:
2018
ie = text_keys[text_key]
2020
# FIXME: TODO: The following loop overlaps/duplicates that done by
2021
# commit to determine parents. There is a latent/real bug here where
2022
# the parents inserted are not those commit would do - in particular
2023
# they are not filtered by heads(). RBC, AB
2024
for revision, tree in parent_trees.iteritems():
2025
if ie.file_id not in tree:
2027
parent_id = tree.inventory[ie.file_id].revision
2028
if parent_id in text_parents:
2030
text_parents.append((ie.file_id, parent_id))
2031
lines = revision_tree.get_file(ie.file_id).readlines()
2032
repository.texts.add_lines(text_key, text_parents, lines)
2034
# install the inventory
2035
repository.add_inventory(rev.revision_id, inv, present_parents)
2036
except errors.RevisionAlreadyPresent:
2038
if signature is not None:
2039
repository.add_signature_text(rev.revision_id, signature)
2040
repository.add_revision(rev.revision_id, rev, inv)
2043
class MetaDirRepository(Repository):
2044
"""Repositories in the new meta-dir layout.
2046
:ivar _transport: Transport for access to repository control files,
2047
typically pointing to .bzr/repository.
2050
def __init__(self, _format, a_bzrdir, control_files):
2051
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2052
self._transport = control_files._transport
2054
def is_shared(self):
2055
"""Return True if this repository is flagged as a shared repository."""
2056
return self._transport.has('shared-storage')
2059
def set_make_working_trees(self, new_value):
2060
"""Set the policy flag for making working trees when creating branches.
2062
This only applies to branches that use this repository.
2064
The default is 'True'.
2065
:param new_value: True to restore the default, False to disable making
2070
self._transport.delete('no-working-trees')
2071
except errors.NoSuchFile:
2074
self._transport.put_bytes('no-working-trees', '',
2075
mode=self.bzrdir._get_file_mode())
2077
def make_working_trees(self):
2078
"""Returns the policy for making working trees on new branches."""
2079
return not self._transport.has('no-working-trees')
2082
class MetaDirVersionedFileRepository(MetaDirRepository):
2083
"""Repositories in a meta-dir, that work via versioned file objects."""
2085
def __init__(self, _format, a_bzrdir, control_files):
2086
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2090
class RepositoryFormatRegistry(registry.Registry):
2091
"""Registry of RepositoryFormats."""
2093
def get(self, format_string):
2094
r = registry.Registry.get(self, format_string)
2100
format_registry = RepositoryFormatRegistry()
2101
"""Registry of formats, indexed by their identifying format string.
2103
This can contain either format instances themselves, or classes/factories that
2104
can be called to obtain one.
2108
#####################################################################
2109
# Repository Formats
2111
class RepositoryFormat(object):
2112
"""A repository format.
2114
Formats provide three things:
2115
* An initialization routine to construct repository data on disk.
2116
* a format string which is used when the BzrDir supports versioned
2118
* an open routine which returns a Repository instance.
2120
There is one and only one Format subclass for each on-disk format. But
2121
there can be one Repository subclass that is used for several different
2122
formats. The _format attribute on a Repository instance can be used to
2123
determine the disk format.
2125
Formats are placed in an dict by their format string for reference
2126
during opening. These should be subclasses of RepositoryFormat
2129
Once a format is deprecated, just deprecate the initialize and open
2130
methods on the format class. Do not deprecate the object, as the
2131
object will be created every system load.
2133
Common instance attributes:
2134
_matchingbzrdir - the bzrdir format that the repository format was
2135
originally written to work with. This can be used if manually
2136
constructing a bzrdir and repository, or more commonly for test suite
2140
# Set to True or False in derived classes. True indicates that the format
2141
# supports ghosts gracefully.
2142
supports_ghosts = None
2143
# Can this repository be given external locations to lookup additional
2144
# data. Set to True or False in derived classes.
2145
supports_external_lookups = None
2148
return "<%s>" % self.__class__.__name__
2150
def __eq__(self, other):
2151
# format objects are generally stateless
2152
return isinstance(other, self.__class__)
2154
def __ne__(self, other):
2155
return not self == other
2158
def find_format(klass, a_bzrdir):
2159
"""Return the format for the repository object in a_bzrdir.
2161
This is used by bzr native formats that have a "format" file in
2162
the repository. Other methods may be used by different types of
2166
transport = a_bzrdir.get_repository_transport(None)
2167
format_string = transport.get("format").read()
2168
return format_registry.get(format_string)
2169
except errors.NoSuchFile:
2170
raise errors.NoRepositoryPresent(a_bzrdir)
2172
raise errors.UnknownFormatError(format=format_string,
2176
def register_format(klass, format):
2177
format_registry.register(format.get_format_string(), format)
2180
def unregister_format(klass, format):
2181
format_registry.remove(format.get_format_string())
2184
def get_default_format(klass):
2185
"""Return the current default format."""
2186
from bzrlib import bzrdir
2187
return bzrdir.format_registry.make_bzrdir('default').repository_format
2189
def get_format_string(self):
2190
"""Return the ASCII format string that identifies this format.
2192
Note that in pre format ?? repositories the format string is
2193
not permitted nor written to disk.
2195
raise NotImplementedError(self.get_format_string)
2197
def get_format_description(self):
2198
"""Return the short description for this format."""
2199
raise NotImplementedError(self.get_format_description)
2201
# TODO: this shouldn't be in the base class, it's specific to things that
2202
# use weaves or knits -- mbp 20070207
2203
def _get_versioned_file_store(self,
2208
versionedfile_class=None,
2209
versionedfile_kwargs={},
2211
if versionedfile_class is None:
2212
versionedfile_class = self._versionedfile_class
2213
weave_transport = control_files._transport.clone(name)
2214
dir_mode = control_files._dir_mode
2215
file_mode = control_files._file_mode
2216
return VersionedFileStore(weave_transport, prefixed=prefixed,
2218
file_mode=file_mode,
2219
versionedfile_class=versionedfile_class,
2220
versionedfile_kwargs=versionedfile_kwargs,
2223
def initialize(self, a_bzrdir, shared=False):
2224
"""Initialize a repository of this format in a_bzrdir.
2226
:param a_bzrdir: The bzrdir to put the new repository in it.
2227
:param shared: The repository should be initialized as a sharable one.
2228
:returns: The new repository object.
2230
This may raise UninitializableFormat if shared repository are not
2231
compatible the a_bzrdir.
2233
raise NotImplementedError(self.initialize)
2235
def is_supported(self):
2236
"""Is this format supported?
2238
Supported formats must be initializable and openable.
2239
Unsupported formats may not support initialization or committing or
2240
some other features depending on the reason for not being supported.
2244
def check_conversion_target(self, target_format):
2245
raise NotImplementedError(self.check_conversion_target)
2247
def open(self, a_bzrdir, _found=False):
2248
"""Return an instance of this format for the bzrdir a_bzrdir.
2250
_found is a private parameter, do not use it.
2252
raise NotImplementedError(self.open)
2255
class MetaDirRepositoryFormat(RepositoryFormat):
2256
"""Common base class for the new repositories using the metadir layout."""
2258
rich_root_data = False
2259
supports_tree_reference = False
2260
supports_external_lookups = False
2263
def _matchingbzrdir(self):
2264
matching = bzrdir.BzrDirMetaFormat1()
2265
matching.repository_format = self
2269
super(MetaDirRepositoryFormat, self).__init__()
2271
def _create_control_files(self, a_bzrdir):
2272
"""Create the required files and the initial control_files object."""
2273
# FIXME: RBC 20060125 don't peek under the covers
2274
# NB: no need to escape relative paths that are url safe.
2275
repository_transport = a_bzrdir.get_repository_transport(self)
2276
control_files = lockable_files.LockableFiles(repository_transport,
2277
'lock', lockdir.LockDir)
2278
control_files.create_lock()
2279
return control_files
2281
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2282
"""Upload the initial blank content."""
2283
control_files = self._create_control_files(a_bzrdir)
2284
control_files.lock_write()
2285
transport = control_files._transport
2287
utf8_files += [('shared-storage', '')]
2289
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2290
for (filename, content_stream) in files:
2291
transport.put_file(filename, content_stream,
2292
mode=a_bzrdir._get_file_mode())
2293
for (filename, content_bytes) in utf8_files:
2294
transport.put_bytes_non_atomic(filename, content_bytes,
2295
mode=a_bzrdir._get_file_mode())
2297
control_files.unlock()
2300
# formats which have no format string are not discoverable
2301
# and not independently creatable, so are not registered. They're
2302
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2303
# needed, it's constructed directly by the BzrDir. Non-native formats where
2304
# the repository is not separately opened are similar.
2306
format_registry.register_lazy(
2307
'Bazaar-NG Repository format 7',
2308
'bzrlib.repofmt.weaverepo',
2312
format_registry.register_lazy(
2313
'Bazaar-NG Knit Repository Format 1',
2314
'bzrlib.repofmt.knitrepo',
2315
'RepositoryFormatKnit1',
2318
format_registry.register_lazy(
2319
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2320
'bzrlib.repofmt.knitrepo',
2321
'RepositoryFormatKnit3',
2324
format_registry.register_lazy(
2325
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2326
'bzrlib.repofmt.knitrepo',
2327
'RepositoryFormatKnit4',
2330
# Pack-based formats. There is one format for pre-subtrees, and one for
2331
# post-subtrees to allow ease of testing.
2332
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2333
format_registry.register_lazy(
2334
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2335
'bzrlib.repofmt.pack_repo',
2336
'RepositoryFormatKnitPack1',
2338
format_registry.register_lazy(
2339
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2340
'bzrlib.repofmt.pack_repo',
2341
'RepositoryFormatKnitPack3',
2343
format_registry.register_lazy(
2344
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2345
'bzrlib.repofmt.pack_repo',
2346
'RepositoryFormatKnitPack4',
2348
format_registry.register_lazy(
2349
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2350
'bzrlib.repofmt.pack_repo',
2351
'RepositoryFormatKnitPack5',
2353
format_registry.register_lazy(
2354
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2355
'bzrlib.repofmt.pack_repo',
2356
'RepositoryFormatKnitPack5RichRoot',
2358
format_registry.register_lazy(
2359
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2360
'bzrlib.repofmt.pack_repo',
2361
'RepositoryFormatKnitPack5RichRootBroken',
2363
format_registry.register_lazy(
2364
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
2365
'bzrlib.repofmt.pack_repo',
2366
'RepositoryFormatKnitPack6',
2368
format_registry.register_lazy(
2369
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
2370
'bzrlib.repofmt.pack_repo',
2371
'RepositoryFormatKnitPack6RichRoot',
2374
# Development formats.
2375
# 1.7->1.8 go below here
2376
format_registry.register_lazy(
2377
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2378
'bzrlib.repofmt.pack_repo',
2379
'RepositoryFormatPackDevelopment2',
2381
format_registry.register_lazy(
2382
("Bazaar development format 2 with subtree support "
2383
"(needs bzr.dev from before 1.8)\n"),
2384
'bzrlib.repofmt.pack_repo',
2385
'RepositoryFormatPackDevelopment2Subtree',
2389
class InterRepository(InterObject):
2390
"""This class represents operations taking place between two repositories.
2392
Its instances have methods like copy_content and fetch, and contain
2393
references to the source and target repositories these operations can be
2396
Often we will provide convenience methods on 'repository' which carry out
2397
operations with another repository - they will always forward to
2398
InterRepository.get(other).method_name(parameters).
2401
_walk_to_common_revisions_batch_size = 1
2403
"""The available optimised InterRepository types."""
2405
def __init__(self, source, target):
2406
InterObject.__init__(self, source, target)
2407
# These two attributes may be overridden by e.g. InterOtherToRemote to
2408
# provide a faster implementation.
2409
self.target_get_graph = self.target.get_graph
2410
self.target_get_parent_map = self.target.get_parent_map
2412
def copy_content(self, revision_id=None):
2413
raise NotImplementedError(self.copy_content)
2415
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2416
"""Fetch the content required to construct revision_id.
2418
The content is copied from self.source to self.target.
2420
:param revision_id: if None all content is copied, if NULL_REVISION no
2422
:param pb: optional progress bar to use for progress reports. If not
2423
provided a default one will be created.
2425
:returns: (copied_revision_count, failures).
2427
# Normally we should find a specific InterRepository subclass to do
2428
# the fetch; if nothing else then at least InterSameDataRepository.
2429
# If none of them is suitable it looks like fetching is not possible;
2430
# we try to give a good message why. _assert_same_model will probably
2431
# give a helpful message; otherwise a generic one.
2432
self._assert_same_model(self.source, self.target)
2433
raise errors.IncompatibleRepositories(self.source, self.target,
2434
"no suitableInterRepository found")
2436
def _walk_to_common_revisions(self, revision_ids):
2437
"""Walk out from revision_ids in source to revisions target has.
2439
:param revision_ids: The start point for the search.
2440
:return: A set of revision ids.
2442
target_graph = self.target_get_graph()
2443
revision_ids = frozenset(revision_ids)
2444
# Fast path for the case where all the revisions are already in the
2446
# (Although this does incur an extra round trip for the
2447
# fairly common case where the target doesn't already have the revision
2449
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2450
return graph.SearchResult(revision_ids, set(), 0, set())
2451
missing_revs = set()
2452
source_graph = self.source.get_graph()
2453
# ensure we don't pay silly lookup costs.
2454
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2455
null_set = frozenset([_mod_revision.NULL_REVISION])
2456
searcher_exhausted = False
2460
# Iterate the searcher until we have enough next_revs
2461
while len(next_revs) < self._walk_to_common_revisions_batch_size:
2463
next_revs_part, ghosts_part = searcher.next_with_ghosts()
2464
next_revs.update(next_revs_part)
2465
ghosts.update(ghosts_part)
2466
except StopIteration:
2467
searcher_exhausted = True
2469
# If there are ghosts in the source graph, and the caller asked for
2470
# them, make sure that they are present in the target.
2471
# We don't care about other ghosts as we can't fetch them and
2472
# haven't been asked to.
2473
ghosts_to_check = set(revision_ids.intersection(ghosts))
2474
revs_to_get = set(next_revs).union(ghosts_to_check)
2476
have_revs = set(target_graph.get_parent_map(revs_to_get))
2477
# we always have NULL_REVISION present.
2478
have_revs = have_revs.union(null_set)
2479
# Check if the target is missing any ghosts we need.
2480
ghosts_to_check.difference_update(have_revs)
2482
# One of the caller's revision_ids is a ghost in both the
2483
# source and the target.
2484
raise errors.NoSuchRevision(
2485
self.source, ghosts_to_check.pop())
2486
missing_revs.update(next_revs - have_revs)
2487
# Because we may have walked past the original stop point, make
2488
# sure everything is stopped
2489
stop_revs = searcher.find_seen_ancestors(have_revs)
2490
searcher.stop_searching_any(stop_revs)
2491
if searcher_exhausted:
2493
return searcher.get_result()
2495
@deprecated_method(one_two)
2497
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2498
"""Return the revision ids that source has that target does not.
2500
These are returned in topological order.
2502
:param revision_id: only return revision ids included by this
2504
:param find_ghosts: If True find missing revisions in deep history
2505
rather than just finding the surface difference.
2507
return list(self.search_missing_revision_ids(
2508
revision_id, find_ghosts).get_keys())
2511
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2512
"""Return the revision ids that source has that target does not.
2514
:param revision_id: only return revision ids included by this
2516
:param find_ghosts: If True find missing revisions in deep history
2517
rather than just finding the surface difference.
2518
:return: A bzrlib.graph.SearchResult.
2520
# stop searching at found target revisions.
2521
if not find_ghosts and revision_id is not None:
2522
return self._walk_to_common_revisions([revision_id])
2523
# generic, possibly worst case, slow code path.
2524
target_ids = set(self.target.all_revision_ids())
2525
if revision_id is not None:
2526
source_ids = self.source.get_ancestry(revision_id)
2527
if source_ids[0] is not None:
2528
raise AssertionError()
2531
source_ids = self.source.all_revision_ids()
2532
result_set = set(source_ids).difference(target_ids)
2533
return self.source.revision_ids_to_search_result(result_set)
2536
def _same_model(source, target):
2537
"""True if source and target have the same data representation.
2539
Note: this is always called on the base class; overriding it in a
2540
subclass will have no effect.
2543
InterRepository._assert_same_model(source, target)
2545
except errors.IncompatibleRepositories, e:
2549
def _assert_same_model(source, target):
2550
"""Raise an exception if two repositories do not use the same model.
2552
if source.supports_rich_root() != target.supports_rich_root():
2553
raise errors.IncompatibleRepositories(source, target,
2554
"different rich-root support")
2555
if source._serializer != target._serializer:
2556
raise errors.IncompatibleRepositories(source, target,
2557
"different serializers")
2560
class InterSameDataRepository(InterRepository):
2561
"""Code for converting between repositories that represent the same data.
2563
Data format and model must match for this to work.
2567
def _get_repo_format_to_test(self):
2568
"""Repository format for testing with.
2570
InterSameData can pull from subtree to subtree and from non-subtree to
2571
non-subtree, so we test this with the richest repository format.
2573
from bzrlib.repofmt import knitrepo
2574
return knitrepo.RepositoryFormatKnit3()
2577
def is_compatible(source, target):
2578
return InterRepository._same_model(source, target)
2581
def copy_content(self, revision_id=None):
2582
"""Make a complete copy of the content in self into destination.
2584
This copies both the repository's revision data, and configuration information
2585
such as the make_working_trees setting.
2587
This is a destructive operation! Do not use it on existing
2590
:param revision_id: Only copy the content needed to construct
2591
revision_id and its parents.
2594
self.target.set_make_working_trees(self.source.make_working_trees())
2595
except NotImplementedError:
2597
# but don't bother fetching if we have the needed data now.
2598
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2599
self.target.has_revision(revision_id)):
2601
self.target.fetch(self.source, revision_id=revision_id)
2604
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2605
"""See InterRepository.fetch()."""
2606
from bzrlib.fetch import RepoFetcher
2607
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2608
self.source, self.source._format, self.target,
2609
self.target._format)
2610
f = RepoFetcher(to_repository=self.target,
2611
from_repository=self.source,
2612
last_revision=revision_id,
2613
pb=pb, find_ghosts=find_ghosts)
2614
return f.count_copied, f.failed_revisions
2617
class InterWeaveRepo(InterSameDataRepository):
2618
"""Optimised code paths between Weave based repositories.
2620
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2621
implemented lazy inter-object optimisation.
2625
def _get_repo_format_to_test(self):
2626
from bzrlib.repofmt import weaverepo
2627
return weaverepo.RepositoryFormat7()
2630
def is_compatible(source, target):
2631
"""Be compatible with known Weave formats.
2633
We don't test for the stores being of specific types because that
2634
could lead to confusing results, and there is no need to be
2637
from bzrlib.repofmt.weaverepo import (
2643
return (isinstance(source._format, (RepositoryFormat5,
2645
RepositoryFormat7)) and
2646
isinstance(target._format, (RepositoryFormat5,
2648
RepositoryFormat7)))
2649
except AttributeError:
2653
def copy_content(self, revision_id=None):
2654
"""See InterRepository.copy_content()."""
2655
# weave specific optimised path:
2657
self.target.set_make_working_trees(self.source.make_working_trees())
2658
except (errors.RepositoryUpgradeRequired, NotImplemented):
2660
# FIXME do not peek!
2661
if self.source._transport.listable():
2662
pb = ui.ui_factory.nested_progress_bar()
2664
self.target.texts.insert_record_stream(
2665
self.source.texts.get_record_stream(
2666
self.source.texts.keys(), 'topological', False))
2667
pb.update('copying inventory', 0, 1)
2668
self.target.inventories.insert_record_stream(
2669
self.source.inventories.get_record_stream(
2670
self.source.inventories.keys(), 'topological', False))
2671
self.target.signatures.insert_record_stream(
2672
self.source.signatures.get_record_stream(
2673
self.source.signatures.keys(),
2675
self.target.revisions.insert_record_stream(
2676
self.source.revisions.get_record_stream(
2677
self.source.revisions.keys(),
2678
'topological', True))
2682
self.target.fetch(self.source, revision_id=revision_id)
2685
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2686
"""See InterRepository.fetch()."""
2687
from bzrlib.fetch import RepoFetcher
2688
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2689
self.source, self.source._format, self.target, self.target._format)
2690
f = RepoFetcher(to_repository=self.target,
2691
from_repository=self.source,
2692
last_revision=revision_id,
2693
pb=pb, find_ghosts=find_ghosts)
2694
return f.count_copied, f.failed_revisions
2697
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2698
"""See InterRepository.missing_revision_ids()."""
2699
# we want all revisions to satisfy revision_id in source.
2700
# but we don't want to stat every file here and there.
2701
# we want then, all revisions other needs to satisfy revision_id
2702
# checked, but not those that we have locally.
2703
# so the first thing is to get a subset of the revisions to
2704
# satisfy revision_id in source, and then eliminate those that
2705
# we do already have.
2706
# this is slow on high latency connection to self, but as as this
2707
# disk format scales terribly for push anyway due to rewriting
2708
# inventory.weave, this is considered acceptable.
2710
if revision_id is not None:
2711
source_ids = self.source.get_ancestry(revision_id)
2712
if source_ids[0] is not None:
2713
raise AssertionError()
2716
source_ids = self.source._all_possible_ids()
2717
source_ids_set = set(source_ids)
2718
# source_ids is the worst possible case we may need to pull.
2719
# now we want to filter source_ids against what we actually
2720
# have in target, but don't try to check for existence where we know
2721
# we do not have a revision as that would be pointless.
2722
target_ids = set(self.target._all_possible_ids())
2723
possibly_present_revisions = target_ids.intersection(source_ids_set)
2724
actually_present_revisions = set(
2725
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2726
required_revisions = source_ids_set.difference(actually_present_revisions)
2727
if revision_id is not None:
2728
# we used get_ancestry to determine source_ids then we are assured all
2729
# revisions referenced are present as they are installed in topological order.
2730
# and the tip revision was validated by get_ancestry.
2731
result_set = required_revisions
2733
# if we just grabbed the possibly available ids, then
2734
# we only have an estimate of whats available and need to validate
2735
# that against the revision records.
2737
self.source._eliminate_revisions_not_present(required_revisions))
2738
return self.source.revision_ids_to_search_result(result_set)
2741
class InterKnitRepo(InterSameDataRepository):
2742
"""Optimised code paths between Knit based repositories."""
2745
def _get_repo_format_to_test(self):
2746
from bzrlib.repofmt import knitrepo
2747
return knitrepo.RepositoryFormatKnit1()
2750
def is_compatible(source, target):
2751
"""Be compatible with known Knit formats.
2753
We don't test for the stores being of specific types because that
2754
could lead to confusing results, and there is no need to be
2757
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2759
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2760
isinstance(target._format, RepositoryFormatKnit))
2761
except AttributeError:
2763
return are_knits and InterRepository._same_model(source, target)
2766
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2767
"""See InterRepository.fetch()."""
2768
from bzrlib.fetch import RepoFetcher
2769
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2770
self.source, self.source._format, self.target, self.target._format)
2771
f = RepoFetcher(to_repository=self.target,
2772
from_repository=self.source,
2773
last_revision=revision_id,
2774
pb=pb, find_ghosts=find_ghosts)
2775
return f.count_copied, f.failed_revisions
2778
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2779
"""See InterRepository.missing_revision_ids()."""
2780
if revision_id is not None:
2781
source_ids = self.source.get_ancestry(revision_id)
2782
if source_ids[0] is not None:
2783
raise AssertionError()
2786
source_ids = self.source.all_revision_ids()
2787
source_ids_set = set(source_ids)
2788
# source_ids is the worst possible case we may need to pull.
2789
# now we want to filter source_ids against what we actually
2790
# have in target, but don't try to check for existence where we know
2791
# we do not have a revision as that would be pointless.
2792
target_ids = set(self.target.all_revision_ids())
2793
possibly_present_revisions = target_ids.intersection(source_ids_set)
2794
actually_present_revisions = set(
2795
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2796
required_revisions = source_ids_set.difference(actually_present_revisions)
2797
if revision_id is not None:
2798
# we used get_ancestry to determine source_ids then we are assured all
2799
# revisions referenced are present as they are installed in topological order.
2800
# and the tip revision was validated by get_ancestry.
2801
result_set = required_revisions
2803
# if we just grabbed the possibly available ids, then
2804
# we only have an estimate of whats available and need to validate
2805
# that against the revision records.
2807
self.source._eliminate_revisions_not_present(required_revisions))
2808
return self.source.revision_ids_to_search_result(result_set)
2811
class InterPackRepo(InterSameDataRepository):
2812
"""Optimised code paths between Pack based repositories."""
2815
def _get_repo_format_to_test(self):
2816
from bzrlib.repofmt import pack_repo
2817
return pack_repo.RepositoryFormatKnitPack1()
2820
def is_compatible(source, target):
2821
"""Be compatible with known Pack formats.
2823
We don't test for the stores being of specific types because that
2824
could lead to confusing results, and there is no need to be
2827
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2829
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2830
isinstance(target._format, RepositoryFormatPack))
2831
except AttributeError:
2833
return are_packs and InterRepository._same_model(source, target)
2836
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2837
"""See InterRepository.fetch()."""
2838
if (len(self.source._fallback_repositories) > 0 or
2839
len(self.target._fallback_repositories) > 0):
2840
# The pack layer is not aware of fallback repositories, so when
2841
# fetching from a stacked repository or into a stacked repository
2842
# we use the generic fetch logic which uses the VersionedFiles
2843
# attributes on repository.
2844
from bzrlib.fetch import RepoFetcher
2845
# Make sure the generic fetcher sets the write cache size on the
2846
# new pack (just like Packer.pack does) to avoid doing many tiny
2847
# writes (which can be slow over a network connection).
2848
# XXX: ideally the transport layer would do this automatically.
2849
pack_coll = self._get_target_pack_collection()
2851
lambda: pack_coll._new_pack.set_write_cache_size(1024*1024))
2852
fetcher = RepoFetcher(self.target, self.source, revision_id,
2853
pb, find_ghosts, set_cache_size)
2854
return fetcher.count_copied, fetcher.failed_revisions
2855
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2856
self.source, self.source._format, self.target, self.target._format)
2857
self.count_copied = 0
2858
if revision_id is None:
2860
# everything to do - use pack logic
2861
# to fetch from all packs to one without
2862
# inventory parsing etc, IFF nothing to be copied is in the target.
2864
source_revision_ids = frozenset(self.source.all_revision_ids())
2865
revision_ids = source_revision_ids - \
2866
frozenset(self.target_get_parent_map(source_revision_ids))
2867
revision_keys = [(revid,) for revid in revision_ids]
2868
target_pack_collection = self._get_target_pack_collection()
2869
index = target_pack_collection.revision_index.combined_index
2870
present_revision_ids = set(item[1][0] for item in
2871
index.iter_entries(revision_keys))
2872
revision_ids = set(revision_ids) - present_revision_ids
2873
# implementing the TODO will involve:
2874
# - detecting when all of a pack is selected
2875
# - avoiding as much as possible pre-selection, so the
2876
# more-core routines such as create_pack_from_packs can filter in
2877
# a just-in-time fashion. (though having a HEADS list on a
2878
# repository might make this a lot easier, because we could
2879
# sensibly detect 'new revisions' without doing a full index scan.
2880
elif _mod_revision.is_null(revision_id):
2885
revision_ids = self.search_missing_revision_ids(revision_id,
2886
find_ghosts=find_ghosts).get_keys()
2887
except errors.NoSuchRevision:
2888
raise errors.InstallFailed([revision_id])
2889
if len(revision_ids) == 0:
2891
return self._pack(self.source, self.target, revision_ids)
2893
def _pack(self, source, target, revision_ids):
2894
from bzrlib.repofmt.pack_repo import Packer
2895
target_pack_collection = self._get_target_pack_collection()
2896
packs = source._pack_collection.all_packs()
2897
pack = Packer(target_pack_collection, packs, '.fetch',
2898
revision_ids).pack()
2899
if pack is not None:
2900
target_pack_collection._save_pack_names()
2901
copied_revs = pack.get_revision_count()
2902
# Trigger an autopack. This may duplicate effort as we've just done
2903
# a pack creation, but for now it is simpler to think about as
2904
# 'upload data, then repack if needed'.
2906
return (copied_revs, [])
2910
def _autopack(self):
2911
self.target._pack_collection.autopack()
2913
def _get_target_pack_collection(self):
2914
return self.target._pack_collection
2917
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2918
"""See InterRepository.missing_revision_ids().
2920
:param find_ghosts: Find ghosts throughout the ancestry of
2923
if not find_ghosts and revision_id is not None:
2924
return self._walk_to_common_revisions([revision_id])
2925
elif revision_id is not None:
2926
# Find ghosts: search for revisions pointing from one repository to
2927
# the other, and vice versa, anywhere in the history of revision_id.
2928
graph = self.target_get_graph(other_repository=self.source)
2929
searcher = graph._make_breadth_first_searcher([revision_id])
2933
next_revs, ghosts = searcher.next_with_ghosts()
2934
except StopIteration:
2936
if revision_id in ghosts:
2937
raise errors.NoSuchRevision(self.source, revision_id)
2938
found_ids.update(next_revs)
2939
found_ids.update(ghosts)
2940
found_ids = frozenset(found_ids)
2941
# Double query here: should be able to avoid this by changing the
2942
# graph api further.
2943
result_set = found_ids - frozenset(
2944
self.target_get_parent_map(found_ids))
2946
source_ids = self.source.all_revision_ids()
2947
# source_ids is the worst possible case we may need to pull.
2948
# now we want to filter source_ids against what we actually
2949
# have in target, but don't try to check for existence where we know
2950
# we do not have a revision as that would be pointless.
2951
target_ids = set(self.target.all_revision_ids())
2952
result_set = set(source_ids).difference(target_ids)
2953
return self.source.revision_ids_to_search_result(result_set)
2956
class InterModel1and2(InterRepository):
2959
def _get_repo_format_to_test(self):
2963
def is_compatible(source, target):
2964
if not source.supports_rich_root() and target.supports_rich_root():
2970
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2971
"""See InterRepository.fetch()."""
2972
from bzrlib.fetch import Model1toKnit2Fetcher
2973
f = Model1toKnit2Fetcher(to_repository=self.target,
2974
from_repository=self.source,
2975
last_revision=revision_id,
2976
pb=pb, find_ghosts=find_ghosts)
2977
return f.count_copied, f.failed_revisions
2980
def copy_content(self, revision_id=None):
2981
"""Make a complete copy of the content in self into destination.
2983
This is a destructive operation! Do not use it on existing
2986
:param revision_id: Only copy the content needed to construct
2987
revision_id and its parents.
2990
self.target.set_make_working_trees(self.source.make_working_trees())
2991
except NotImplementedError:
2993
# but don't bother fetching if we have the needed data now.
2994
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2995
self.target.has_revision(revision_id)):
2997
self.target.fetch(self.source, revision_id=revision_id)
3000
class InterKnit1and2(InterKnitRepo):
3003
def _get_repo_format_to_test(self):
3007
def is_compatible(source, target):
3008
"""Be compatible with Knit1 source and Knit3 target"""
3010
from bzrlib.repofmt.knitrepo import (
3011
RepositoryFormatKnit1,
3012
RepositoryFormatKnit3,
3014
from bzrlib.repofmt.pack_repo import (
3015
RepositoryFormatKnitPack1,
3016
RepositoryFormatKnitPack3,
3017
RepositoryFormatKnitPack4,
3018
RepositoryFormatKnitPack5,
3019
RepositoryFormatKnitPack5RichRoot,
3020
RepositoryFormatKnitPack6,
3021
RepositoryFormatKnitPack6RichRoot,
3022
RepositoryFormatPackDevelopment2,
3023
RepositoryFormatPackDevelopment2Subtree,
3026
RepositoryFormatKnit1, # no rr, no subtree
3027
RepositoryFormatKnitPack1, # no rr, no subtree
3028
RepositoryFormatPackDevelopment2, # no rr, no subtree
3029
RepositoryFormatKnitPack5, # no rr, no subtree
3030
RepositoryFormatKnitPack6, # no rr, no subtree
3033
RepositoryFormatKnit3, # rr, subtree
3034
RepositoryFormatKnitPack3, # rr, subtree
3035
RepositoryFormatKnitPack4, # rr, no subtree
3036
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
3037
RepositoryFormatKnitPack6RichRoot,# rr, no subtree
3038
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
3040
for format in norichroot:
3041
if format.rich_root_data:
3042
raise AssertionError('Format %s is a rich-root format'
3043
' but is included in the non-rich-root list'
3045
for format in richroot:
3046
if not format.rich_root_data:
3047
raise AssertionError('Format %s is not a rich-root format'
3048
' but is included in the rich-root list'
3050
# TODO: One alternative is to just check format.rich_root_data,
3051
# instead of keeping membership lists. However, the formats
3052
# *also* have to use the same 'Knit' style of storage
3053
# (line-deltas, fulltexts, etc.)
3054
return (isinstance(source._format, norichroot) and
3055
isinstance(target._format, richroot))
3056
except AttributeError:
3060
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3061
"""See InterRepository.fetch()."""
3062
from bzrlib.fetch import Knit1to2Fetcher
3063
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
3064
self.source, self.source._format, self.target,
3065
self.target._format)
3066
f = Knit1to2Fetcher(to_repository=self.target,
3067
from_repository=self.source,
3068
last_revision=revision_id,
3069
pb=pb, find_ghosts=find_ghosts)
3070
return f.count_copied, f.failed_revisions
3073
class InterDifferingSerializer(InterKnitRepo):
3076
def _get_repo_format_to_test(self):
3080
def is_compatible(source, target):
3081
"""Be compatible with Knit2 source and Knit3 target"""
3082
if source.supports_rich_root() != target.supports_rich_root():
3084
# Ideally, we'd support fetching if the source had no tree references
3085
# even if it supported them...
3086
if (getattr(source, '_format.supports_tree_reference', False) and
3087
not getattr(target, '_format.supports_tree_reference', False)):
3092
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3093
"""See InterRepository.fetch()."""
3094
revision_ids = self.target.search_missing_revision_ids(self.source,
3095
revision_id, find_ghosts=find_ghosts).get_keys()
3096
revision_ids = tsort.topo_sort(
3097
self.source.get_graph().get_parent_map(revision_ids))
3098
def revisions_iterator():
3099
rev_ids = list(revision_ids)
3100
for offset in xrange(0, len(rev_ids), 100):
3101
current_revids = rev_ids[offset:offset+100]
3102
revisions = self.source.get_revisions(current_revids)
3103
trees = self.source.revision_trees(current_revids)
3104
keys = [(r,) for r in current_revids]
3105
sig_stream = self.source.signatures.get_record_stream(
3106
keys, 'unordered', True)
3108
for record in versionedfile.filter_absent(sig_stream):
3109
sigs[record.key[0]] = record.get_bytes_as('fulltext')
3110
for rev, tree in zip(revisions, trees):
3111
yield rev, tree, sigs.get(rev.revision_id, None)
3113
my_pb = ui.ui_factory.nested_progress_bar()
3118
install_revisions(self.target, revisions_iterator(),
3119
len(revision_ids), pb)
3121
if my_pb is not None:
3123
return len(revision_ids), 0
3126
class InterOtherToRemote(InterRepository):
3127
"""An InterRepository that simply delegates to the 'real' InterRepository
3128
calculated for (source, target._real_repository).
3131
_walk_to_common_revisions_batch_size = 50
3133
def __init__(self, source, target):
3134
InterRepository.__init__(self, source, target)
3135
self._real_inter = None
3138
def is_compatible(source, target):
3139
if isinstance(target, remote.RemoteRepository):
3143
def _ensure_real_inter(self):
3144
if self._real_inter is None:
3145
self.target._ensure_real()
3146
real_target = self.target._real_repository
3147
self._real_inter = InterRepository.get(self.source, real_target)
3148
# Make _real_inter use the RemoteRepository for get_parent_map
3149
self._real_inter.target_get_graph = self.target.get_graph
3150
self._real_inter.target_get_parent_map = self.target.get_parent_map
3152
def copy_content(self, revision_id=None):
3153
self._ensure_real_inter()
3154
self._real_inter.copy_content(revision_id=revision_id)
3156
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3157
self._ensure_real_inter()
3158
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3159
find_ghosts=find_ghosts)
3162
def _get_repo_format_to_test(self):
3166
class InterRemoteToOther(InterRepository):
3168
def __init__(self, source, target):
3169
InterRepository.__init__(self, source, target)
3170
self._real_inter = None
3173
def is_compatible(source, target):
3174
if not isinstance(source, remote.RemoteRepository):
3176
# Is source's model compatible with target's model?
3177
source._ensure_real()
3178
real_source = source._real_repository
3179
if isinstance(real_source, remote.RemoteRepository):
3180
raise NotImplementedError(
3181
"We don't support remote repos backed by remote repos yet.")
3182
return InterRepository._same_model(real_source, target)
3184
def _ensure_real_inter(self):
3185
if self._real_inter is None:
3186
self.source._ensure_real()
3187
real_source = self.source._real_repository
3188
self._real_inter = InterRepository.get(real_source, self.target)
3190
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3191
self._ensure_real_inter()
3192
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3193
find_ghosts=find_ghosts)
3195
def copy_content(self, revision_id=None):
3196
self._ensure_real_inter()
3197
self._real_inter.copy_content(revision_id=revision_id)
3200
def _get_repo_format_to_test(self):
3205
class InterPackToRemotePack(InterPackRepo):
3206
"""A specialisation of InterPackRepo for a target that is a
3209
This will use the get_parent_map RPC rather than plain readvs, and also
3210
uses an RPC for autopacking.
3213
_walk_to_common_revisions_batch_size = 50
3216
def is_compatible(source, target):
3217
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
3218
if isinstance(source._format, RepositoryFormatPack):
3219
if isinstance(target, remote.RemoteRepository):
3220
target._ensure_real()
3221
if isinstance(target._real_repository._format,
3222
RepositoryFormatPack):
3223
if InterRepository._same_model(source, target):
3227
def _autopack(self):
3228
self.target.autopack()
3230
def _get_target_pack_collection(self):
3231
return self.target._real_repository._pack_collection
3234
def _get_repo_format_to_test(self):
3238
InterRepository.register_optimiser(InterDifferingSerializer)
3239
InterRepository.register_optimiser(InterSameDataRepository)
3240
InterRepository.register_optimiser(InterWeaveRepo)
3241
InterRepository.register_optimiser(InterKnitRepo)
3242
InterRepository.register_optimiser(InterModel1and2)
3243
InterRepository.register_optimiser(InterKnit1and2)
3244
InterRepository.register_optimiser(InterPackRepo)
3245
InterRepository.register_optimiser(InterOtherToRemote)
3246
InterRepository.register_optimiser(InterRemoteToOther)
3247
InterRepository.register_optimiser(InterPackToRemotePack)
3250
class CopyConverter(object):
3251
"""A repository conversion tool which just performs a copy of the content.
3253
This is slow but quite reliable.
3256
def __init__(self, target_format):
3257
"""Create a CopyConverter.
3259
:param target_format: The format the resulting repository should be.
3261
self.target_format = target_format
3263
def convert(self, repo, pb):
3264
"""Perform the conversion of to_convert, giving feedback via pb.
3266
:param to_convert: The disk object to convert.
3267
:param pb: a progress bar to use for progress information.
3272
# this is only useful with metadir layouts - separated repo content.
3273
# trigger an assertion if not such
3274
repo._format.get_format_string()
3275
self.repo_dir = repo.bzrdir
3276
self.step('Moving repository to repository.backup')
3277
self.repo_dir.transport.move('repository', 'repository.backup')
3278
backup_transport = self.repo_dir.transport.clone('repository.backup')
3279
repo._format.check_conversion_target(self.target_format)
3280
self.source_repo = repo._format.open(self.repo_dir,
3282
_override_transport=backup_transport)
3283
self.step('Creating new repository')
3284
converted = self.target_format.initialize(self.repo_dir,
3285
self.source_repo.is_shared())
3286
converted.lock_write()
3288
self.step('Copying content into repository.')
3289
self.source_repo.copy_content_into(converted)
3292
self.step('Deleting old repository content.')
3293
self.repo_dir.transport.delete_tree('repository.backup')
3294
self.pb.note('repository converted')
3296
def step(self, message):
3297
"""Update the pb by a step."""
3299
self.pb.update(message, self.count, self.total)
3311
def _unescaper(match, _map=_unescape_map):
3312
code = match.group(1)
3316
if not code.startswith('#'):
3318
return unichr(int(code[1:])).encode('utf8')
3324
def _unescape_xml(data):
3325
"""Unescape predefined XML entities in a string of data."""
3327
if _unescape_re is None:
3328
_unescape_re = re.compile('\&([^;]*);')
3329
return _unescape_re.sub(_unescaper, data)
3332
class _VersionedFileChecker(object):
3334
def __init__(self, repository):
3335
self.repository = repository
3336
self.text_index = self.repository._generate_text_key_index()
3338
def calculate_file_version_parents(self, text_key):
3339
"""Calculate the correct parents for a file version according to
3342
parent_keys = self.text_index[text_key]
3343
if parent_keys == [_mod_revision.NULL_REVISION]:
3345
return tuple(parent_keys)
3347
def check_file_version_parents(self, texts, progress_bar=None):
3348
"""Check the parents stored in a versioned file are correct.
3350
It also detects file versions that are not referenced by their
3351
corresponding revision's inventory.
3353
:returns: A tuple of (wrong_parents, dangling_file_versions).
3354
wrong_parents is a dict mapping {revision_id: (stored_parents,
3355
correct_parents)} for each revision_id where the stored parents
3356
are not correct. dangling_file_versions is a set of (file_id,
3357
revision_id) tuples for versions that are present in this versioned
3358
file, but not used by the corresponding inventory.
3361
self.file_ids = set([file_id for file_id, _ in
3362
self.text_index.iterkeys()])
3363
# text keys is now grouped by file_id
3364
n_weaves = len(self.file_ids)
3365
files_in_revisions = {}
3366
revisions_of_files = {}
3367
n_versions = len(self.text_index)
3368
progress_bar.update('loading text store', 0, n_versions)
3369
parent_map = self.repository.texts.get_parent_map(self.text_index)
3370
# On unlistable transports this could well be empty/error...
3371
text_keys = self.repository.texts.keys()
3372
unused_keys = frozenset(text_keys) - set(self.text_index)
3373
for num, key in enumerate(self.text_index.iterkeys()):
3374
if progress_bar is not None:
3375
progress_bar.update('checking text graph', num, n_versions)
3376
correct_parents = self.calculate_file_version_parents(key)
3378
knit_parents = parent_map[key]
3379
except errors.RevisionNotPresent:
3382
if correct_parents != knit_parents:
3383
wrong_parents[key] = (knit_parents, correct_parents)
3384
return wrong_parents, unused_keys
3387
def _old_get_graph(repository, revision_id):
3388
"""DO NOT USE. That is all. I'm serious."""
3389
graph = repository.get_graph()
3390
revision_graph = dict(((key, value) for key, value in
3391
graph.iter_ancestry([revision_id]) if value is not None))
3392
return _strip_NULL_ghosts(revision_graph)
3395
def _strip_NULL_ghosts(revision_graph):
3396
"""Also don't use this. more compatibility code for unmigrated clients."""
3397
# Filter ghosts, and null:
3398
if _mod_revision.NULL_REVISION in revision_graph:
3399
del revision_graph[_mod_revision.NULL_REVISION]
3400
for key, parents in revision_graph.items():
3401
revision_graph[key] = tuple(parent for parent in parents if parent
3403
return revision_graph