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 cStringIO import StringIO
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
39
revision as _mod_revision,
45
from bzrlib.bundle import serializer
46
from bzrlib.revisiontree import RevisionTree
47
from bzrlib.store.versioned import VersionedFileStore
48
from bzrlib.store.text import TextStore
49
from bzrlib.testament import Testament
50
from bzrlib.util import bencode
53
from bzrlib.decorators import needs_read_lock, needs_write_lock
54
from bzrlib.inter import InterObject
55
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
56
from bzrlib.symbol_versioning import (
63
from bzrlib.trace import mutter, mutter_callsite, note, warning
66
# Old formats display a warning, but only once
67
_deprecation_warning_done = False
70
class CommitBuilder(object):
71
"""Provides an interface to build up a commit.
73
This allows describing a tree to be committed without needing to
74
know the internals of the format of the repository.
77
# all clients should supply tree roots.
78
record_root_entry = True
79
# the default CommitBuilder does not manage trees whose root is versioned.
80
_versioned_root = False
82
def __init__(self, repository, parents, config, timestamp=None,
83
timezone=None, committer=None, revprops=None,
85
"""Initiate a CommitBuilder.
87
:param repository: Repository to commit to.
88
:param parents: Revision ids of the parents of the new revision.
89
:param config: Configuration to use.
90
:param timestamp: Optional timestamp recorded for commit.
91
:param timezone: Optional timezone for timestamp.
92
:param committer: Optional committer to set for commit.
93
:param revprops: Optional dictionary of revision properties.
94
:param revision_id: Optional revision id.
99
self._committer = self._config.username()
101
self._committer = committer
103
self.new_inventory = Inventory(None)
104
self._new_revision_id = revision_id
105
self.parents = parents
106
self.repository = repository
109
if revprops is not None:
110
self._revprops.update(revprops)
112
if timestamp is None:
113
timestamp = time.time()
114
# Restrict resolution to 1ms
115
self._timestamp = round(timestamp, 3)
118
self._timezone = osutils.local_time_offset()
120
self._timezone = int(timezone)
122
self._generate_revision_if_needed()
123
self.__heads = graph.HeadsCache(repository.get_graph()).heads
125
def commit(self, message):
126
"""Make the actual commit.
128
:return: The revision id of the recorded revision.
130
rev = _mod_revision.Revision(
131
timestamp=self._timestamp,
132
timezone=self._timezone,
133
committer=self._committer,
135
inventory_sha1=self.inv_sha1,
136
revision_id=self._new_revision_id,
137
properties=self._revprops)
138
rev.parent_ids = self.parents
139
self.repository.add_revision(self._new_revision_id, rev,
140
self.new_inventory, self._config)
141
self.repository.commit_write_group()
142
return self._new_revision_id
145
"""Abort the commit that is being built.
147
self.repository.abort_write_group()
149
def revision_tree(self):
150
"""Return the tree that was just committed.
152
After calling commit() this can be called to get a RevisionTree
153
representing the newly committed tree. This is preferred to
154
calling Repository.revision_tree() because that may require
155
deserializing the inventory, while we already have a copy in
158
return RevisionTree(self.repository, self.new_inventory,
159
self._new_revision_id)
161
def finish_inventory(self):
162
"""Tell the builder that the inventory is finished."""
163
if self.new_inventory.root is None:
164
raise AssertionError('Root entry should be supplied to'
165
' record_entry_contents, as of bzr 0.10.')
166
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
167
self.new_inventory.revision_id = self._new_revision_id
168
self.inv_sha1 = self.repository.add_inventory(
169
self._new_revision_id,
174
def _gen_revision_id(self):
175
"""Return new revision-id."""
176
return generate_ids.gen_revision_id(self._config.username(),
179
def _generate_revision_if_needed(self):
180
"""Create a revision id if None was supplied.
182
If the repository can not support user-specified revision ids
183
they should override this function and raise CannotSetRevisionId
184
if _new_revision_id is not None.
186
:raises: CannotSetRevisionId
188
if self._new_revision_id is None:
189
self._new_revision_id = self._gen_revision_id()
190
self.random_revid = True
192
self.random_revid = False
194
def _heads(self, file_id, revision_ids):
195
"""Calculate the graph heads for revision_ids in the graph of file_id.
197
This can use either a per-file graph or a global revision graph as we
198
have an identity relationship between the two graphs.
200
return self.__heads(revision_ids)
202
def _check_root(self, ie, parent_invs, tree):
203
"""Helper for record_entry_contents.
205
:param ie: An entry being added.
206
:param parent_invs: The inventories of the parent revisions of the
208
:param tree: The tree that is being committed.
210
# In this revision format, root entries have no knit or weave When
211
# serializing out to disk and back in root.revision is always
213
ie.revision = self._new_revision_id
215
def _get_delta(self, ie, basis_inv, path):
216
"""Get a delta against the basis inventory for ie."""
217
if ie.file_id not in basis_inv:
219
return (None, path, ie.file_id, ie)
220
elif ie != basis_inv[ie.file_id]:
222
# TODO: avoid tis id2path call.
223
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
228
def record_entry_contents(self, ie, parent_invs, path, tree,
230
"""Record the content of ie from tree into the commit if needed.
232
Side effect: sets ie.revision when unchanged
234
:param ie: An inventory entry present in the commit.
235
:param parent_invs: The inventories of the parent revisions of the
237
:param path: The path the entry is at in the tree.
238
:param tree: The tree which contains this entry and should be used to
240
:param content_summary: Summary data from the tree about the paths
241
content - stat, length, exec, sha/link target. This is only
242
accessed when the entry has a revision of None - that is when it is
243
a candidate to commit.
244
:return: A tuple (change_delta, version_recorded, fs_hash).
245
change_delta is an inventory_delta change for this entry against
246
the basis tree of the commit, or None if no change occured against
248
version_recorded is True if a new version of the entry has been
249
recorded. For instance, committing a merge where a file was only
250
changed on the other side will return (delta, False).
251
fs_hash is either None, or the hash details for the path (currently
252
a tuple of the contents sha1 and the statvalue returned by
253
tree.get_file_with_stat()).
255
if self.new_inventory.root is None:
256
if ie.parent_id is not None:
257
raise errors.RootMissing()
258
self._check_root(ie, parent_invs, tree)
259
if ie.revision is None:
260
kind = content_summary[0]
262
# ie is carried over from a prior commit
264
# XXX: repository specific check for nested tree support goes here - if
265
# the repo doesn't want nested trees we skip it ?
266
if (kind == 'tree-reference' and
267
not self.repository._format.supports_tree_reference):
268
# mismatch between commit builder logic and repository:
269
# this needs the entry creation pushed down into the builder.
270
raise NotImplementedError('Missing repository subtree support.')
271
self.new_inventory.add(ie)
273
# TODO: slow, take it out of the inner loop.
275
basis_inv = parent_invs[0]
277
basis_inv = Inventory(root_id=None)
279
# ie.revision is always None if the InventoryEntry is considered
280
# for committing. We may record the previous parents revision if the
281
# content is actually unchanged against a sole head.
282
if ie.revision is not None:
283
if not self._versioned_root and path == '':
284
# repositories that do not version the root set the root's
285
# revision to the new commit even when no change occurs, and
286
# this masks when a change may have occurred against the basis,
287
# so calculate if one happened.
288
if ie.file_id in basis_inv:
289
delta = (basis_inv.id2path(ie.file_id), path,
293
delta = (None, path, ie.file_id, ie)
294
return delta, False, None
296
# we don't need to commit this, because the caller already
297
# determined that an existing revision of this file is
298
# appropriate. If its not being considered for committing then
299
# it and all its parents to the root must be unaltered so
300
# no-change against the basis.
301
if ie.revision == self._new_revision_id:
302
raise AssertionError("Impossible situation, a skipped "
303
"inventory entry (%r) claims to be modified in this "
304
"commit (%r).", (ie, self._new_revision_id))
305
return None, False, None
306
# XXX: Friction: parent_candidates should return a list not a dict
307
# so that we don't have to walk the inventories again.
308
parent_candiate_entries = ie.parent_candidates(parent_invs)
309
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
311
for inv in parent_invs:
312
if ie.file_id in inv:
313
old_rev = inv[ie.file_id].revision
314
if old_rev in head_set:
315
heads.append(inv[ie.file_id].revision)
316
head_set.remove(inv[ie.file_id].revision)
319
# now we check to see if we need to write a new record to the
321
# We write a new entry unless there is one head to the ancestors, and
322
# the kind-derived content is unchanged.
324
# Cheapest check first: no ancestors, or more the one head in the
325
# ancestors, we write a new node.
329
# There is a single head, look it up for comparison
330
parent_entry = parent_candiate_entries[heads[0]]
331
# if the non-content specific data has changed, we'll be writing a
333
if (parent_entry.parent_id != ie.parent_id or
334
parent_entry.name != ie.name):
336
# now we need to do content specific checks:
338
# if the kind changed the content obviously has
339
if kind != parent_entry.kind:
341
# Stat cache fingerprint feedback for the caller - None as we usually
342
# don't generate one.
345
if content_summary[2] is None:
346
raise ValueError("Files must not have executable = None")
348
if (# if the file length changed we have to store:
349
parent_entry.text_size != content_summary[1] or
350
# if the exec bit has changed we have to store:
351
parent_entry.executable != content_summary[2]):
353
elif parent_entry.text_sha1 == content_summary[3]:
354
# all meta and content is unchanged (using a hash cache
355
# hit to check the sha)
356
ie.revision = parent_entry.revision
357
ie.text_size = parent_entry.text_size
358
ie.text_sha1 = parent_entry.text_sha1
359
ie.executable = parent_entry.executable
360
return self._get_delta(ie, basis_inv, path), False, None
362
# Either there is only a hash change(no hash cache entry,
363
# or same size content change), or there is no change on
365
# Provide the parent's hash to the store layer, so that the
366
# content is unchanged we will not store a new node.
367
nostore_sha = parent_entry.text_sha1
369
# We want to record a new node regardless of the presence or
370
# absence of a content change in the file.
372
ie.executable = content_summary[2]
373
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
375
lines = file_obj.readlines()
379
ie.text_sha1, ie.text_size = self._add_text_to_weave(
380
ie.file_id, lines, heads, nostore_sha)
381
# Let the caller know we generated a stat fingerprint.
382
fingerprint = (ie.text_sha1, stat_value)
383
except errors.ExistingContent:
384
# Turns out that the file content was unchanged, and we were
385
# only going to store a new node if it was changed. Carry over
387
ie.revision = parent_entry.revision
388
ie.text_size = parent_entry.text_size
389
ie.text_sha1 = parent_entry.text_sha1
390
ie.executable = parent_entry.executable
391
return self._get_delta(ie, basis_inv, path), False, None
392
elif kind == 'directory':
394
# all data is meta here, nothing specific to directory, so
396
ie.revision = parent_entry.revision
397
return self._get_delta(ie, basis_inv, path), False, None
399
self._add_text_to_weave(ie.file_id, lines, heads, None)
400
elif kind == 'symlink':
401
current_link_target = content_summary[3]
403
# symlink target is not generic metadata, check if it has
405
if current_link_target != parent_entry.symlink_target:
408
# unchanged, carry over.
409
ie.revision = parent_entry.revision
410
ie.symlink_target = parent_entry.symlink_target
411
return self._get_delta(ie, basis_inv, path), False, None
412
ie.symlink_target = current_link_target
414
self._add_text_to_weave(ie.file_id, lines, heads, None)
415
elif kind == 'tree-reference':
417
if content_summary[3] != parent_entry.reference_revision:
420
# unchanged, carry over.
421
ie.reference_revision = parent_entry.reference_revision
422
ie.revision = parent_entry.revision
423
return self._get_delta(ie, basis_inv, path), False, None
424
ie.reference_revision = content_summary[3]
426
self._add_text_to_weave(ie.file_id, lines, heads, None)
428
raise NotImplementedError('unknown kind')
429
ie.revision = self._new_revision_id
430
return self._get_delta(ie, basis_inv, path), True, fingerprint
432
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
433
# Note: as we read the content directly from the tree, we know its not
434
# been turned into unicode or badly split - but a broken tree
435
# implementation could give us bad output from readlines() so this is
436
# not a guarantee of safety. What would be better is always checking
437
# the content during test suite execution. RBC 20070912
438
parent_keys = tuple((file_id, parent) for parent in parents)
439
return self.repository.texts.add_lines(
440
(file_id, self._new_revision_id), parent_keys, new_lines,
441
nostore_sha=nostore_sha, random_id=self.random_revid,
442
check_content=False)[0:2]
445
class RootCommitBuilder(CommitBuilder):
446
"""This commitbuilder actually records the root id"""
448
# the root entry gets versioned properly by this builder.
449
_versioned_root = True
451
def _check_root(self, ie, parent_invs, tree):
452
"""Helper for record_entry_contents.
454
:param ie: An entry being added.
455
:param parent_invs: The inventories of the parent revisions of the
457
:param tree: The tree that is being committed.
461
######################################################################
464
class Repository(object):
465
"""Repository holding history for one or more branches.
467
The repository holds and retrieves historical information including
468
revisions and file history. It's normally accessed only by the Branch,
469
which views a particular line of development through that history.
471
The Repository builds on top of some byte storage facilies (the revisions,
472
signatures, inventories and texts attributes) and a Transport, which
473
respectively provide byte storage and a means to access the (possibly
476
The byte storage facilities are addressed via tuples, which we refer to
477
as 'keys' throughout the code base. Revision_keys, inventory_keys and
478
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
479
(file_id, revision_id). We use this interface because it allows low
480
friction with the underlying code that implements disk indices, network
481
encoding and other parts of bzrlib.
483
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
484
the serialised revisions for the repository. This can be used to obtain
485
revision graph information or to access raw serialised revisions.
486
The result of trying to insert data into the repository via this store
487
is undefined: it should be considered read-only except for implementors
489
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
490
the serialised signatures for the repository. This can be used to
491
obtain access to raw serialised signatures. The result of trying to
492
insert data into the repository via this store is undefined: it should
493
be considered read-only except for implementors of repositories.
494
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
495
the serialised inventories for the repository. This can be used to
496
obtain unserialised inventories. The result of trying to insert data
497
into the repository via this store is undefined: it should be
498
considered read-only except for implementors of repositories.
499
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
500
texts of files and directories for the repository. This can be used to
501
obtain file texts or file graphs. Note that Repository.iter_file_bytes
502
is usually a better interface for accessing file texts.
503
The result of trying to insert data into the repository via this store
504
is undefined: it should be considered read-only except for implementors
506
:ivar _transport: Transport for file access to repository, typically
507
pointing to .bzr/repository.
510
# What class to use for a CommitBuilder. Often its simpler to change this
511
# in a Repository class subclass rather than to override
512
# get_commit_builder.
513
_commit_builder_class = CommitBuilder
514
# The search regex used by xml based repositories to determine what things
515
# where changed in a single commit.
516
_file_ids_altered_regex = lazy_regex.lazy_compile(
517
r'file_id="(?P<file_id>[^"]+)"'
518
r'.* revision="(?P<revision_id>[^"]+)"'
521
def abort_write_group(self):
522
"""Commit the contents accrued within the current write group.
524
:seealso: start_write_group.
526
if self._write_group is not self.get_transaction():
527
# has an unlock or relock occured ?
528
raise errors.BzrError('mismatched lock context and write group.')
529
self._abort_write_group()
530
self._write_group = None
532
def _abort_write_group(self):
533
"""Template method for per-repository write group cleanup.
535
This is called during abort before the write group is considered to be
536
finished and should cleanup any internal state accrued during the write
537
group. There is no requirement that data handed to the repository be
538
*not* made available - this is not a rollback - but neither should any
539
attempt be made to ensure that data added is fully commited. Abort is
540
invoked when an error has occured so futher disk or network operations
541
may not be possible or may error and if possible should not be
545
def add_fallback_repository(self, repository):
546
"""Add a repository to use for looking up data not held locally.
548
:param repository: A repository.
550
if not self._format.supports_external_lookups:
551
raise errors.UnstackableRepositoryFormat(self._format, self.base)
552
self._check_fallback_repository(repository)
553
self._fallback_repositories.append(repository)
554
self.texts.add_fallback_versioned_files(repository.texts)
555
self.inventories.add_fallback_versioned_files(repository.inventories)
556
self.revisions.add_fallback_versioned_files(repository.revisions)
557
self.signatures.add_fallback_versioned_files(repository.signatures)
559
def _check_fallback_repository(self, repository):
560
"""Check that this repository can fallback to repository safely.
562
Raise an error if not.
564
:param repository: A repository to fallback to.
566
return InterRepository._assert_same_model(self, repository)
568
def add_inventory(self, revision_id, inv, parents):
569
"""Add the inventory inv to the repository as revision_id.
571
:param parents: The revision ids of the parents that revision_id
572
is known to have and are in the repository already.
574
:returns: The validator(which is a sha1 digest, though what is sha'd is
575
repository format specific) of the serialized inventory.
577
if not self.is_in_write_group():
578
raise AssertionError("%r not in write group" % (self,))
579
_mod_revision.check_not_reserved_id(revision_id)
580
if not (inv.revision_id is None or inv.revision_id == revision_id):
581
raise AssertionError(
582
"Mismatch between inventory revision"
583
" id and insertion revid (%r, %r)"
584
% (inv.revision_id, revision_id))
586
raise AssertionError()
587
inv_lines = self._serialise_inventory_to_lines(inv)
588
return self._inventory_add_lines(revision_id, parents,
589
inv_lines, check_content=False)
591
def _inventory_add_lines(self, revision_id, parents, lines,
593
"""Store lines in inv_vf and return the sha1 of the inventory."""
594
parents = [(parent,) for parent in parents]
595
return self.inventories.add_lines((revision_id,), parents, lines,
596
check_content=check_content)[0]
598
def add_revision(self, revision_id, rev, inv=None, config=None):
599
"""Add rev to the revision store as revision_id.
601
:param revision_id: the revision id to use.
602
:param rev: The revision object.
603
:param inv: The inventory for the revision. if None, it will be looked
604
up in the inventory storer
605
:param config: If None no digital signature will be created.
606
If supplied its signature_needed method will be used
607
to determine if a signature should be made.
609
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
611
_mod_revision.check_not_reserved_id(revision_id)
612
if config is not None and config.signature_needed():
614
inv = self.get_inventory(revision_id)
615
plaintext = Testament(rev, inv).as_short_text()
616
self.store_revision_signature(
617
gpg.GPGStrategy(config), plaintext, revision_id)
618
# check inventory present
619
if not self.inventories.get_parent_map([(revision_id,)]):
621
raise errors.WeaveRevisionNotPresent(revision_id,
624
# yes, this is not suitable for adding with ghosts.
625
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
629
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
630
self._add_revision(rev)
632
def _add_revision(self, revision):
633
text = self._serializer.write_revision_to_string(revision)
634
key = (revision.revision_id,)
635
parents = tuple((parent,) for parent in revision.parent_ids)
636
self.revisions.add_lines(key, parents, osutils.split_lines(text))
638
def all_revision_ids(self):
639
"""Returns a list of all the revision ids in the repository.
641
This is conceptually deprecated because code should generally work on
642
the graph reachable from a particular revision, and ignore any other
643
revisions that might be present. There is no direct replacement
646
if 'evil' in debug.debug_flags:
647
mutter_callsite(2, "all_revision_ids is linear with history.")
648
return self._all_revision_ids()
650
def _all_revision_ids(self):
651
"""Returns a list of all the revision ids in the repository.
653
These are in as much topological order as the underlying store can
656
raise NotImplementedError(self._all_revision_ids)
658
def break_lock(self):
659
"""Break a lock if one is present from another instance.
661
Uses the ui factory to ask for confirmation if the lock may be from
664
self.control_files.break_lock()
667
def _eliminate_revisions_not_present(self, revision_ids):
668
"""Check every revision id in revision_ids to see if we have it.
670
Returns a set of the present revisions.
673
graph = self.get_graph()
674
parent_map = graph.get_parent_map(revision_ids)
675
# The old API returned a list, should this actually be a set?
676
return parent_map.keys()
679
def create(a_bzrdir):
680
"""Construct the current default format repository in a_bzrdir."""
681
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
683
def __init__(self, _format, a_bzrdir, control_files):
684
"""instantiate a Repository.
686
:param _format: The format of the repository on disk.
687
:param a_bzrdir: The BzrDir of the repository.
689
In the future we will have a single api for all stores for
690
getting file texts, inventories and revisions, then
691
this construct will accept instances of those things.
693
super(Repository, self).__init__()
694
self._format = _format
695
# the following are part of the public API for Repository:
696
self.bzrdir = a_bzrdir
697
self.control_files = control_files
698
self._transport = control_files._transport
699
self.base = self._transport.base
701
self._reconcile_does_inventory_gc = True
702
self._reconcile_fixes_text_parents = False
703
self._reconcile_backsup_inventory = True
704
# not right yet - should be more semantically clear ?
706
# TODO: make sure to construct the right store classes, etc, depending
707
# on whether escaping is required.
708
self._warn_if_deprecated()
709
self._write_group = None
710
# Additional places to query for data.
711
self._fallback_repositories = []
712
# What order should fetch operations request streams in?
713
# The default is unordered as that is the cheapest for an origin to
715
self._fetch_order = 'unordered'
716
# Does this repository use deltas that can be fetched as-deltas ?
717
# (E.g. knits, where the knit deltas can be transplanted intact.
718
# We default to False, which will ensure that enough data to get
719
# a full text out of any fetch stream will be grabbed.
720
self._fetch_uses_deltas = False
721
# Should fetch trigger a reconcile after the fetch? Only needed for
722
# some repository formats that can suffer internal inconsistencies.
723
self._fetch_reconcile = False
726
return '%s(%r)' % (self.__class__.__name__,
729
def has_same_location(self, other):
730
"""Returns a boolean indicating if this repository is at the same
731
location as another repository.
733
This might return False even when two repository objects are accessing
734
the same physical repository via different URLs.
736
if self.__class__ is not other.__class__:
738
return (self._transport.base == other._transport.base)
740
def is_in_write_group(self):
741
"""Return True if there is an open write group.
743
:seealso: start_write_group.
745
return self._write_group is not None
748
return self.control_files.is_locked()
750
def is_write_locked(self):
751
"""Return True if this object is write locked."""
752
return self.is_locked() and self.control_files._lock_mode == 'w'
754
def lock_write(self, token=None):
755
"""Lock this repository for writing.
757
This causes caching within the repository obejct to start accumlating
758
data during reads, and allows a 'write_group' to be obtained. Write
759
groups must be used for actual data insertion.
761
:param token: if this is already locked, then lock_write will fail
762
unless the token matches the existing lock.
763
:returns: a token if this instance supports tokens, otherwise None.
764
:raises TokenLockingNotSupported: when a token is given but this
765
instance doesn't support using token locks.
766
:raises MismatchedToken: if the specified token doesn't match the token
767
of the existing lock.
768
:seealso: start_write_group.
770
A token should be passed in if you know that you have locked the object
771
some other way, and need to synchronise this object's state with that
774
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
776
result = self.control_files.lock_write(token=token)
777
for repo in self._fallback_repositories:
778
# Writes don't affect fallback repos
784
self.control_files.lock_read()
785
for repo in self._fallback_repositories:
789
def get_physical_lock_status(self):
790
return self.control_files.get_physical_lock_status()
792
def leave_lock_in_place(self):
793
"""Tell this repository not to release the physical lock when this
796
If lock_write doesn't return a token, then this method is not supported.
798
self.control_files.leave_in_place()
800
def dont_leave_lock_in_place(self):
801
"""Tell this repository to release the physical lock when this
802
object is unlocked, even if it didn't originally acquire it.
804
If lock_write doesn't return a token, then this method is not supported.
806
self.control_files.dont_leave_in_place()
809
def gather_stats(self, revid=None, committers=None):
810
"""Gather statistics from a revision id.
812
:param revid: The revision id to gather statistics from, if None, then
813
no revision specific statistics are gathered.
814
:param committers: Optional parameter controlling whether to grab
815
a count of committers from the revision specific statistics.
816
:return: A dictionary of statistics. Currently this contains:
817
committers: The number of committers if requested.
818
firstrev: A tuple with timestamp, timezone for the penultimate left
819
most ancestor of revid, if revid is not the NULL_REVISION.
820
latestrev: A tuple with timestamp, timezone for revid, if revid is
821
not the NULL_REVISION.
822
revisions: The total revision count in the repository.
823
size: An estimate disk size of the repository in bytes.
826
if revid and committers:
827
result['committers'] = 0
828
if revid and revid != _mod_revision.NULL_REVISION:
830
all_committers = set()
831
revisions = self.get_ancestry(revid)
832
# pop the leading None
834
first_revision = None
836
# ignore the revisions in the middle - just grab first and last
837
revisions = revisions[0], revisions[-1]
838
for revision in self.get_revisions(revisions):
839
if not first_revision:
840
first_revision = revision
842
all_committers.add(revision.committer)
843
last_revision = revision
845
result['committers'] = len(all_committers)
846
result['firstrev'] = (first_revision.timestamp,
847
first_revision.timezone)
848
result['latestrev'] = (last_revision.timestamp,
849
last_revision.timezone)
851
# now gather global repository information
852
# XXX: This is available for many repos regardless of listability.
853
if self.bzrdir.root_transport.listable():
854
# XXX: do we want to __define len__() ?
855
# Maybe the versionedfiles object should provide a different
856
# method to get the number of keys.
857
result['revisions'] = len(self.revisions.keys())
861
def find_branches(self, using=False):
862
"""Find branches underneath this repository.
864
This will include branches inside other branches.
866
:param using: If True, list only branches using this repository.
868
if using and not self.is_shared():
870
return [self.bzrdir.open_branch()]
871
except errors.NotBranchError:
873
class Evaluator(object):
876
self.first_call = True
878
def __call__(self, bzrdir):
879
# On the first call, the parameter is always the bzrdir
880
# containing the current repo.
881
if not self.first_call:
883
repository = bzrdir.open_repository()
884
except errors.NoRepositoryPresent:
887
return False, (None, repository)
888
self.first_call = False
890
value = (bzrdir.open_branch(), None)
891
except errors.NotBranchError:
896
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
897
self.bzrdir.root_transport, evaluate=Evaluator()):
898
if branch is not None:
899
branches.append(branch)
900
if not using and repository is not None:
901
branches.extend(repository.find_branches())
905
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
906
"""Return the revision ids that other has that this does not.
908
These are returned in topological order.
910
revision_id: only return revision ids included by revision_id.
912
return InterRepository.get(other, self).search_missing_revision_ids(
913
revision_id, find_ghosts)
915
@deprecated_method(one_two)
917
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
918
"""Return the revision ids that other has that this does not.
920
These are returned in topological order.
922
revision_id: only return revision ids included by revision_id.
924
keys = self.search_missing_revision_ids(
925
other, revision_id, find_ghosts).get_keys()
928
parents = other.get_graph().get_parent_map(keys)
931
return tsort.topo_sort(parents)
935
"""Open the repository rooted at base.
937
For instance, if the repository is at URL/.bzr/repository,
938
Repository.open(URL) -> a Repository instance.
940
control = bzrdir.BzrDir.open(base)
941
return control.open_repository()
943
def copy_content_into(self, destination, revision_id=None):
944
"""Make a complete copy of the content in self into destination.
946
This is a destructive operation! Do not use it on existing
949
return InterRepository.get(self, destination).copy_content(revision_id)
951
def commit_write_group(self):
952
"""Commit the contents accrued within the current write group.
954
:seealso: start_write_group.
956
if self._write_group is not self.get_transaction():
957
# has an unlock or relock occured ?
958
raise errors.BzrError('mismatched lock context %r and '
960
(self.get_transaction(), self._write_group))
961
self._commit_write_group()
962
self._write_group = None
964
def _commit_write_group(self):
965
"""Template method for per-repository write group cleanup.
967
This is called before the write group is considered to be
968
finished and should ensure that all data handed to the repository
969
for writing during the write group is safely committed (to the
970
extent possible considering file system caching etc).
973
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
974
"""Fetch the content required to construct revision_id from source.
976
If revision_id is None all content is copied.
977
:param find_ghosts: Find and copy revisions in the source that are
978
ghosts in the target (and not reachable directly by walking out to
979
the first-present revision in target from revision_id).
981
# fast path same-url fetch operations
982
if self.has_same_location(source):
983
# check that last_revision is in 'from' and then return a
985
if (revision_id is not None and
986
not _mod_revision.is_null(revision_id)):
987
self.get_revision(revision_id)
989
# if there is no specific appropriate InterRepository, this will get
990
# the InterRepository base class, which raises an
991
# IncompatibleRepositories when asked to fetch.
992
inter = InterRepository.get(source, self)
993
return inter.fetch(revision_id=revision_id, pb=pb,
994
find_ghosts=find_ghosts)
996
def create_bundle(self, target, base, fileobj, format=None):
997
return serializer.write_bundle(self, target, base, fileobj, format)
999
def get_commit_builder(self, branch, parents, config, timestamp=None,
1000
timezone=None, committer=None, revprops=None,
1002
"""Obtain a CommitBuilder for this repository.
1004
:param branch: Branch to commit to.
1005
:param parents: Revision ids of the parents of the new revision.
1006
:param config: Configuration to use.
1007
:param timestamp: Optional timestamp recorded for commit.
1008
:param timezone: Optional timezone for timestamp.
1009
:param committer: Optional committer to set for commit.
1010
:param revprops: Optional dictionary of revision properties.
1011
:param revision_id: Optional revision id.
1013
result = self._commit_builder_class(self, parents, config,
1014
timestamp, timezone, committer, revprops, revision_id)
1015
self.start_write_group()
1019
if (self.control_files._lock_count == 1 and
1020
self.control_files._lock_mode == 'w'):
1021
if self._write_group is not None:
1022
self.abort_write_group()
1023
self.control_files.unlock()
1024
raise errors.BzrError(
1025
'Must end write groups before releasing write locks.')
1026
self.control_files.unlock()
1027
for repo in self._fallback_repositories:
1031
def clone(self, a_bzrdir, revision_id=None):
1032
"""Clone this repository into a_bzrdir using the current format.
1034
Currently no check is made that the format of this repository and
1035
the bzrdir format are compatible. FIXME RBC 20060201.
1037
:return: The newly created destination repository.
1039
# TODO: deprecate after 0.16; cloning this with all its settings is
1040
# probably not very useful -- mbp 20070423
1041
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1042
self.copy_content_into(dest_repo, revision_id)
1045
def start_write_group(self):
1046
"""Start a write group in the repository.
1048
Write groups are used by repositories which do not have a 1:1 mapping
1049
between file ids and backend store to manage the insertion of data from
1050
both fetch and commit operations.
1052
A write lock is required around the start_write_group/commit_write_group
1053
for the support of lock-requiring repository formats.
1055
One can only insert data into a repository inside a write group.
1059
if not self.is_write_locked():
1060
raise errors.NotWriteLocked(self)
1061
if self._write_group:
1062
raise errors.BzrError('already in a write group')
1063
self._start_write_group()
1064
# so we can detect unlock/relock - the write group is now entered.
1065
self._write_group = self.get_transaction()
1067
def _start_write_group(self):
1068
"""Template method for per-repository write group startup.
1070
This is called before the write group is considered to be
1075
def sprout(self, to_bzrdir, revision_id=None):
1076
"""Create a descendent repository for new development.
1078
Unlike clone, this does not copy the settings of the repository.
1080
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1081
dest_repo.fetch(self, revision_id=revision_id)
1084
def _create_sprouting_repo(self, a_bzrdir, shared):
1085
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1086
# use target default format.
1087
dest_repo = a_bzrdir.create_repository()
1089
# Most control formats need the repository to be specifically
1090
# created, but on some old all-in-one formats it's not needed
1092
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1093
except errors.UninitializableFormat:
1094
dest_repo = a_bzrdir.open_repository()
1098
def has_revision(self, revision_id):
1099
"""True if this repository has a copy of the revision."""
1100
return revision_id in self.has_revisions((revision_id,))
1103
def has_revisions(self, revision_ids):
1104
"""Probe to find out the presence of multiple revisions.
1106
:param revision_ids: An iterable of revision_ids.
1107
:return: A set of the revision_ids that were present.
1109
parent_map = self.revisions.get_parent_map(
1110
[(rev_id,) for rev_id in revision_ids])
1112
if _mod_revision.NULL_REVISION in revision_ids:
1113
result.add(_mod_revision.NULL_REVISION)
1114
result.update([key[0] for key in parent_map])
1118
def get_revision(self, revision_id):
1119
"""Return the Revision object for a named revision."""
1120
return self.get_revisions([revision_id])[0]
1123
def get_revision_reconcile(self, revision_id):
1124
"""'reconcile' helper routine that allows access to a revision always.
1126
This variant of get_revision does not cross check the weave graph
1127
against the revision one as get_revision does: but it should only
1128
be used by reconcile, or reconcile-alike commands that are correcting
1129
or testing the revision graph.
1131
return self._get_revisions([revision_id])[0]
1134
def get_revisions(self, revision_ids):
1135
"""Get many revisions at once."""
1136
return self._get_revisions(revision_ids)
1139
def _get_revisions(self, revision_ids):
1140
"""Core work logic to get many revisions without sanity checks."""
1141
for rev_id in revision_ids:
1142
if not rev_id or not isinstance(rev_id, basestring):
1143
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1144
keys = [(key,) for key in revision_ids]
1145
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1147
for record in stream:
1148
if record.storage_kind == 'absent':
1149
raise errors.NoSuchRevision(self, record.key[0])
1150
text = record.get_bytes_as('fulltext')
1151
rev = self._serializer.read_revision_from_string(text)
1152
revs[record.key[0]] = rev
1153
return [revs[revid] for revid in revision_ids]
1156
def get_revision_xml(self, revision_id):
1157
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1158
# would have already do it.
1159
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1160
rev = self.get_revision(revision_id)
1161
rev_tmp = StringIO()
1162
# the current serializer..
1163
self._serializer.write_revision(rev, rev_tmp)
1165
return rev_tmp.getvalue()
1167
def get_deltas_for_revisions(self, revisions):
1168
"""Produce a generator of revision deltas.
1170
Note that the input is a sequence of REVISIONS, not revision_ids.
1171
Trees will be held in memory until the generator exits.
1172
Each delta is relative to the revision's lefthand predecessor.
1174
required_trees = set()
1175
for revision in revisions:
1176
required_trees.add(revision.revision_id)
1177
required_trees.update(revision.parent_ids[:1])
1178
trees = dict((t.get_revision_id(), t) for
1179
t in self.revision_trees(required_trees))
1180
for revision in revisions:
1181
if not revision.parent_ids:
1182
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1184
old_tree = trees[revision.parent_ids[0]]
1185
yield trees[revision.revision_id].changes_from(old_tree)
1188
def get_revision_delta(self, revision_id):
1189
"""Return the delta for one revision.
1191
The delta is relative to the left-hand predecessor of the
1194
r = self.get_revision(revision_id)
1195
return list(self.get_deltas_for_revisions([r]))[0]
1198
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1199
signature = gpg_strategy.sign(plaintext)
1200
self.add_signature_text(revision_id, signature)
1203
def add_signature_text(self, revision_id, signature):
1204
self.signatures.add_lines((revision_id,), (),
1205
osutils.split_lines(signature))
1207
def find_text_key_references(self):
1208
"""Find the text key references within the repository.
1210
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1211
revision_ids. Each altered file-ids has the exact revision_ids that
1212
altered it listed explicitly.
1213
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1214
to whether they were referred to by the inventory of the
1215
revision_id that they contain. The inventory texts from all present
1216
revision ids are assessed to generate this report.
1218
revision_keys = self.revisions.keys()
1219
w = self.inventories
1220
pb = ui.ui_factory.nested_progress_bar()
1222
return self._find_text_key_references_from_xml_inventory_lines(
1223
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1227
def _find_text_key_references_from_xml_inventory_lines(self,
1229
"""Core routine for extracting references to texts from inventories.
1231
This performs the translation of xml lines to revision ids.
1233
:param line_iterator: An iterator of lines, origin_version_id
1234
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1235
to whether they were referred to by the inventory of the
1236
revision_id that they contain. Note that if that revision_id was
1237
not part of the line_iterator's output then False will be given -
1238
even though it may actually refer to that key.
1240
if not self._serializer.support_altered_by_hack:
1241
raise AssertionError(
1242
"_find_text_key_references_from_xml_inventory_lines only "
1243
"supported for branches which store inventory as unnested xml"
1244
", not on %r" % self)
1247
# this code needs to read every new line in every inventory for the
1248
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1249
# not present in one of those inventories is unnecessary but not
1250
# harmful because we are filtering by the revision id marker in the
1251
# inventory lines : we only select file ids altered in one of those
1252
# revisions. We don't need to see all lines in the inventory because
1253
# only those added in an inventory in rev X can contain a revision=X
1255
unescape_revid_cache = {}
1256
unescape_fileid_cache = {}
1258
# jam 20061218 In a big fetch, this handles hundreds of thousands
1259
# of lines, so it has had a lot of inlining and optimizing done.
1260
# Sorry that it is a little bit messy.
1261
# Move several functions to be local variables, since this is a long
1263
search = self._file_ids_altered_regex.search
1264
unescape = _unescape_xml
1265
setdefault = result.setdefault
1266
for line, line_key in line_iterator:
1267
match = search(line)
1270
# One call to match.group() returning multiple items is quite a
1271
# bit faster than 2 calls to match.group() each returning 1
1272
file_id, revision_id = match.group('file_id', 'revision_id')
1274
# Inlining the cache lookups helps a lot when you make 170,000
1275
# lines and 350k ids, versus 8.4 unique ids.
1276
# Using a cache helps in 2 ways:
1277
# 1) Avoids unnecessary decoding calls
1278
# 2) Re-uses cached strings, which helps in future set and
1280
# (2) is enough that removing encoding entirely along with
1281
# the cache (so we are using plain strings) results in no
1282
# performance improvement.
1284
revision_id = unescape_revid_cache[revision_id]
1286
unescaped = unescape(revision_id)
1287
unescape_revid_cache[revision_id] = unescaped
1288
revision_id = unescaped
1290
# Note that unconditionally unescaping means that we deserialise
1291
# every fileid, which for general 'pull' is not great, but we don't
1292
# really want to have some many fulltexts that this matters anyway.
1295
file_id = unescape_fileid_cache[file_id]
1297
unescaped = unescape(file_id)
1298
unescape_fileid_cache[file_id] = unescaped
1301
key = (file_id, revision_id)
1302
setdefault(key, False)
1303
if revision_id == line_key[-1]:
1307
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1309
"""Helper routine for fileids_altered_by_revision_ids.
1311
This performs the translation of xml lines to revision ids.
1313
:param line_iterator: An iterator of lines, origin_version_id
1314
:param revision_ids: The revision ids to filter for. This should be a
1315
set or other type which supports efficient __contains__ lookups, as
1316
the revision id from each parsed line will be looked up in the
1317
revision_ids filter.
1318
:return: a dictionary mapping altered file-ids to an iterable of
1319
revision_ids. Each altered file-ids has the exact revision_ids that
1320
altered it listed explicitly.
1323
setdefault = result.setdefault
1325
self._find_text_key_references_from_xml_inventory_lines(
1326
line_iterator).iterkeys():
1327
# once data is all ensured-consistent; then this is
1328
# if revision_id == version_id
1329
if key[-1:] in revision_ids:
1330
setdefault(key[0], set()).add(key[-1])
1333
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1334
"""Find the file ids and versions affected by revisions.
1336
:param revisions: an iterable containing revision ids.
1337
:param _inv_weave: The inventory weave from this repository or None.
1338
If None, the inventory weave will be opened automatically.
1339
:return: a dictionary mapping altered file-ids to an iterable of
1340
revision_ids. Each altered file-ids has the exact revision_ids that
1341
altered it listed explicitly.
1343
selected_keys = set((revid,) for revid in revision_ids)
1344
w = _inv_weave or self.inventories
1345
pb = ui.ui_factory.nested_progress_bar()
1347
return self._find_file_ids_from_xml_inventory_lines(
1348
w.iter_lines_added_or_present_in_keys(
1349
selected_keys, pb=pb),
1354
def iter_files_bytes(self, desired_files):
1355
"""Iterate through file versions.
1357
Files will not necessarily be returned in the order they occur in
1358
desired_files. No specific order is guaranteed.
1360
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1361
value supplied by the caller as part of desired_files. It should
1362
uniquely identify the file version in the caller's context. (Examples:
1363
an index number or a TreeTransform trans_id.)
1365
bytes_iterator is an iterable of bytestrings for the file. The
1366
kind of iterable and length of the bytestrings are unspecified, but for
1367
this implementation, it is a list of bytes produced by
1368
VersionedFile.get_record_stream().
1370
:param desired_files: a list of (file_id, revision_id, identifier)
1373
transaction = self.get_transaction()
1375
for file_id, revision_id, callable_data in desired_files:
1376
text_keys[(file_id, revision_id)] = callable_data
1377
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
1378
if record.storage_kind == 'absent':
1379
raise errors.RevisionNotPresent(record.key, self)
1380
yield text_keys[record.key], record.get_bytes_as('fulltext')
1382
def _generate_text_key_index(self, text_key_references=None,
1384
"""Generate a new text key index for the repository.
1386
This is an expensive function that will take considerable time to run.
1388
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1389
list of parents, also text keys. When a given key has no parents,
1390
the parents list will be [NULL_REVISION].
1392
# All revisions, to find inventory parents.
1393
if ancestors is None:
1394
graph = self.get_graph()
1395
ancestors = graph.get_parent_map(self.all_revision_ids())
1396
if text_key_references is None:
1397
text_key_references = self.find_text_key_references()
1398
pb = ui.ui_factory.nested_progress_bar()
1400
return self._do_generate_text_key_index(ancestors,
1401
text_key_references, pb)
1405
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1406
"""Helper for _generate_text_key_index to avoid deep nesting."""
1407
revision_order = tsort.topo_sort(ancestors)
1408
invalid_keys = set()
1410
for revision_id in revision_order:
1411
revision_keys[revision_id] = set()
1412
text_count = len(text_key_references)
1413
# a cache of the text keys to allow reuse; costs a dict of all the
1414
# keys, but saves a 2-tuple for every child of a given key.
1416
for text_key, valid in text_key_references.iteritems():
1418
invalid_keys.add(text_key)
1420
revision_keys[text_key[1]].add(text_key)
1421
text_key_cache[text_key] = text_key
1422
del text_key_references
1424
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1425
NULL_REVISION = _mod_revision.NULL_REVISION
1426
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1427
# too small for large or very branchy trees. However, for 55K path
1428
# trees, it would be easy to use too much memory trivially. Ideally we
1429
# could gauge this by looking at available real memory etc, but this is
1430
# always a tricky proposition.
1431
inventory_cache = lru_cache.LRUCache(10)
1432
batch_size = 10 # should be ~150MB on a 55K path tree
1433
batch_count = len(revision_order) / batch_size + 1
1435
pb.update("Calculating text parents.", processed_texts, text_count)
1436
for offset in xrange(batch_count):
1437
to_query = revision_order[offset * batch_size:(offset + 1) *
1441
for rev_tree in self.revision_trees(to_query):
1442
revision_id = rev_tree.get_revision_id()
1443
parent_ids = ancestors[revision_id]
1444
for text_key in revision_keys[revision_id]:
1445
pb.update("Calculating text parents.", processed_texts)
1446
processed_texts += 1
1447
candidate_parents = []
1448
for parent_id in parent_ids:
1449
parent_text_key = (text_key[0], parent_id)
1451
check_parent = parent_text_key not in \
1452
revision_keys[parent_id]
1454
# the parent parent_id is a ghost:
1455
check_parent = False
1456
# truncate the derived graph against this ghost.
1457
parent_text_key = None
1459
# look at the parent commit details inventories to
1460
# determine possible candidates in the per file graph.
1463
inv = inventory_cache[parent_id]
1465
inv = self.revision_tree(parent_id).inventory
1466
inventory_cache[parent_id] = inv
1467
parent_entry = inv._byid.get(text_key[0], None)
1468
if parent_entry is not None:
1470
text_key[0], parent_entry.revision)
1472
parent_text_key = None
1473
if parent_text_key is not None:
1474
candidate_parents.append(
1475
text_key_cache[parent_text_key])
1476
parent_heads = text_graph.heads(candidate_parents)
1477
new_parents = list(parent_heads)
1478
new_parents.sort(key=lambda x:candidate_parents.index(x))
1479
if new_parents == []:
1480
new_parents = [NULL_REVISION]
1481
text_index[text_key] = new_parents
1483
for text_key in invalid_keys:
1484
text_index[text_key] = [NULL_REVISION]
1487
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1488
"""Get an iterable listing the keys of all the data introduced by a set
1491
The keys will be ordered so that the corresponding items can be safely
1492
fetched and inserted in that order.
1494
:returns: An iterable producing tuples of (knit-kind, file-id,
1495
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1496
'revisions'. file-id is None unless knit-kind is 'file'.
1498
# XXX: it's a bit weird to control the inventory weave caching in this
1499
# generator. Ideally the caching would be done in fetch.py I think. Or
1500
# maybe this generator should explicitly have the contract that it
1501
# should not be iterated until the previously yielded item has been
1503
inv_w = self.inventories
1505
# file ids that changed
1506
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
1508
num_file_ids = len(file_ids)
1509
for file_id, altered_versions in file_ids.iteritems():
1510
if _files_pb is not None:
1511
_files_pb.update("fetch texts", count, num_file_ids)
1513
yield ("file", file_id, altered_versions)
1514
# We're done with the files_pb. Note that it finished by the caller,
1515
# just as it was created by the caller.
1519
yield ("inventory", None, revision_ids)
1522
revisions_with_signatures = set()
1523
for rev_id in revision_ids:
1525
self.get_signature_text(rev_id)
1526
except errors.NoSuchRevision:
1530
revisions_with_signatures.add(rev_id)
1531
yield ("signatures", None, revisions_with_signatures)
1534
yield ("revisions", None, revision_ids)
1537
def get_inventory(self, revision_id):
1538
"""Get Inventory object by revision id."""
1539
return self.iter_inventories([revision_id]).next()
1541
def iter_inventories(self, revision_ids):
1542
"""Get many inventories by revision_ids.
1544
This will buffer some or all of the texts used in constructing the
1545
inventories in memory, but will only parse a single inventory at a
1548
:return: An iterator of inventories.
1550
if ((None in revision_ids)
1551
or (_mod_revision.NULL_REVISION in revision_ids)):
1552
raise ValueError('cannot get null revision inventory')
1553
return self._iter_inventories(revision_ids)
1555
def _iter_inventories(self, revision_ids):
1556
"""single-document based inventory iteration."""
1557
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1558
yield self.deserialise_inventory(revision_id, text)
1560
def _iter_inventory_xmls(self, revision_ids):
1561
keys = [(revision_id,) for revision_id in revision_ids]
1562
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1564
for record in stream:
1565
if record.storage_kind != 'absent':
1566
texts[record.key] = record.get_bytes_as('fulltext')
1568
raise errors.NoSuchRevision(self, record.key)
1570
yield texts[key], key[-1]
1572
def deserialise_inventory(self, revision_id, xml):
1573
"""Transform the xml into an inventory object.
1575
:param revision_id: The expected revision id of the inventory.
1576
:param xml: A serialised inventory.
1578
result = self._serializer.read_inventory_from_string(xml, revision_id)
1579
if result.revision_id != revision_id:
1580
raise AssertionError('revision id mismatch %s != %s' % (
1581
result.revision_id, revision_id))
1584
def serialise_inventory(self, inv):
1585
return self._serializer.write_inventory_to_string(inv)
1587
def _serialise_inventory_to_lines(self, inv):
1588
return self._serializer.write_inventory_to_lines(inv)
1590
def get_serializer_format(self):
1591
return self._serializer.format_num
1594
def get_inventory_xml(self, revision_id):
1595
"""Get inventory XML as a file object."""
1596
texts = self._iter_inventory_xmls([revision_id])
1598
text, revision_id = texts.next()
1599
except StopIteration:
1600
raise errors.HistoryMissing(self, 'inventory', revision_id)
1604
def get_inventory_sha1(self, revision_id):
1605
"""Return the sha1 hash of the inventory entry
1607
return self.get_revision(revision_id).inventory_sha1
1609
def iter_reverse_revision_history(self, revision_id):
1610
"""Iterate backwards through revision ids in the lefthand history
1612
:param revision_id: The revision id to start with. All its lefthand
1613
ancestors will be traversed.
1615
graph = self.get_graph()
1616
next_id = revision_id
1618
if next_id in (None, _mod_revision.NULL_REVISION):
1621
# Note: The following line may raise KeyError in the event of
1622
# truncated history. We decided not to have a try:except:raise
1623
# RevisionNotPresent here until we see a use for it, because of the
1624
# cost in an inner loop that is by its very nature O(history).
1625
# Robert Collins 20080326
1626
parents = graph.get_parent_map([next_id])[next_id]
1627
if len(parents) == 0:
1630
next_id = parents[0]
1633
def get_revision_inventory(self, revision_id):
1634
"""Return inventory of a past revision."""
1635
# TODO: Unify this with get_inventory()
1636
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1637
# must be the same as its revision, so this is trivial.
1638
if revision_id is None:
1639
# This does not make sense: if there is no revision,
1640
# then it is the current tree inventory surely ?!
1641
# and thus get_root_id() is something that looks at the last
1642
# commit on the branch, and the get_root_id is an inventory check.
1643
raise NotImplementedError
1644
# return Inventory(self.get_root_id())
1646
return self.get_inventory(revision_id)
1648
def is_shared(self):
1649
"""Return True if this repository is flagged as a shared repository."""
1650
raise NotImplementedError(self.is_shared)
1653
def reconcile(self, other=None, thorough=False):
1654
"""Reconcile this repository."""
1655
from bzrlib.reconcile import RepoReconciler
1656
reconciler = RepoReconciler(self, thorough=thorough)
1657
reconciler.reconcile()
1660
def _refresh_data(self):
1661
"""Helper called from lock_* to ensure coherency with disk.
1663
The default implementation does nothing; it is however possible
1664
for repositories to maintain loaded indices across multiple locks
1665
by checking inside their implementation of this method to see
1666
whether their indices are still valid. This depends of course on
1667
the disk format being validatable in this manner.
1671
def revision_tree(self, revision_id):
1672
"""Return Tree for a revision on this branch.
1674
`revision_id` may be NULL_REVISION for the empty tree revision.
1676
revision_id = _mod_revision.ensure_null(revision_id)
1677
# TODO: refactor this to use an existing revision object
1678
# so we don't need to read it in twice.
1679
if revision_id == _mod_revision.NULL_REVISION:
1680
return RevisionTree(self, Inventory(root_id=None),
1681
_mod_revision.NULL_REVISION)
1683
inv = self.get_revision_inventory(revision_id)
1684
return RevisionTree(self, inv, revision_id)
1686
def revision_trees(self, revision_ids):
1687
"""Return Tree for a revision on this branch.
1689
`revision_id` may not be None or 'null:'"""
1690
inventories = self.iter_inventories(revision_ids)
1691
for inv in inventories:
1692
yield RevisionTree(self, inv, inv.revision_id)
1695
def get_ancestry(self, revision_id, topo_sorted=True):
1696
"""Return a list of revision-ids integrated by a revision.
1698
The first element of the list is always None, indicating the origin
1699
revision. This might change when we have history horizons, or
1700
perhaps we should have a new API.
1702
This is topologically sorted.
1704
if _mod_revision.is_null(revision_id):
1706
if not self.has_revision(revision_id):
1707
raise errors.NoSuchRevision(self, revision_id)
1708
graph = self.get_graph()
1710
search = graph._make_breadth_first_searcher([revision_id])
1713
found, ghosts = search.next_with_ghosts()
1714
except StopIteration:
1717
if _mod_revision.NULL_REVISION in keys:
1718
keys.remove(_mod_revision.NULL_REVISION)
1720
parent_map = graph.get_parent_map(keys)
1721
keys = tsort.topo_sort(parent_map)
1722
return [None] + list(keys)
1725
"""Compress the data within the repository.
1727
This operation only makes sense for some repository types. For other
1728
types it should be a no-op that just returns.
1730
This stub method does not require a lock, but subclasses should use
1731
@needs_write_lock as this is a long running call its reasonable to
1732
implicitly lock for the user.
1736
@deprecated_method(one_six)
1737
def print_file(self, file, revision_id):
1738
"""Print `file` to stdout.
1740
FIXME RBC 20060125 as John Meinel points out this is a bad api
1741
- it writes to stdout, it assumes that that is valid etc. Fix
1742
by creating a new more flexible convenience function.
1744
tree = self.revision_tree(revision_id)
1745
# use inventory as it was in that revision
1746
file_id = tree.inventory.path2id(file)
1748
# TODO: jam 20060427 Write a test for this code path
1749
# it had a bug in it, and was raising the wrong
1751
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1752
tree.print_file(file_id)
1754
def get_transaction(self):
1755
return self.control_files.get_transaction()
1757
@deprecated_method(one_one)
1758
def get_parents(self, revision_ids):
1759
"""See StackedParentsProvider.get_parents"""
1760
parent_map = self.get_parent_map(revision_ids)
1761
return [parent_map.get(r, None) for r in revision_ids]
1763
def get_parent_map(self, revision_ids):
1764
"""See graph._StackedParentsProvider.get_parent_map"""
1765
# revisions index works in keys; this just works in revisions
1766
# therefore wrap and unwrap
1769
for revision_id in revision_ids:
1770
if revision_id == _mod_revision.NULL_REVISION:
1771
result[revision_id] = ()
1772
elif revision_id is None:
1773
raise ValueError('get_parent_map(None) is not valid')
1775
query_keys.append((revision_id ,))
1776
for ((revision_id,), parent_keys) in \
1777
self.revisions.get_parent_map(query_keys).iteritems():
1779
result[revision_id] = tuple(parent_revid
1780
for (parent_revid,) in parent_keys)
1782
result[revision_id] = (_mod_revision.NULL_REVISION,)
1785
def _make_parents_provider(self):
1788
def get_graph(self, other_repository=None):
1789
"""Return the graph walker for this repository format"""
1790
parents_provider = self._make_parents_provider()
1791
if (other_repository is not None and
1792
not self.has_same_location(other_repository)):
1793
parents_provider = graph._StackedParentsProvider(
1794
[parents_provider, other_repository._make_parents_provider()])
1795
return graph.Graph(parents_provider)
1797
def _get_versioned_file_checker(self):
1798
"""Return an object suitable for checking versioned files."""
1799
return _VersionedFileChecker(self)
1801
def revision_ids_to_search_result(self, result_set):
1802
"""Convert a set of revision ids to a graph SearchResult."""
1803
result_parents = set()
1804
for parents in self.get_graph().get_parent_map(
1805
result_set).itervalues():
1806
result_parents.update(parents)
1807
included_keys = result_set.intersection(result_parents)
1808
start_keys = result_set.difference(included_keys)
1809
exclude_keys = result_parents.difference(result_set)
1810
result = graph.SearchResult(start_keys, exclude_keys,
1811
len(result_set), result_set)
1815
def set_make_working_trees(self, new_value):
1816
"""Set the policy flag for making working trees when creating branches.
1818
This only applies to branches that use this repository.
1820
The default is 'True'.
1821
:param new_value: True to restore the default, False to disable making
1824
raise NotImplementedError(self.set_make_working_trees)
1826
def make_working_trees(self):
1827
"""Returns the policy for making working trees on new branches."""
1828
raise NotImplementedError(self.make_working_trees)
1831
def sign_revision(self, revision_id, gpg_strategy):
1832
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1833
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1836
def has_signature_for_revision_id(self, revision_id):
1837
"""Query for a revision signature for revision_id in the repository."""
1838
if not self.has_revision(revision_id):
1839
raise errors.NoSuchRevision(self, revision_id)
1840
sig_present = (1 == len(
1841
self.signatures.get_parent_map([(revision_id,)])))
1845
def get_signature_text(self, revision_id):
1846
"""Return the text for a signature."""
1847
stream = self.signatures.get_record_stream([(revision_id,)],
1849
record = stream.next()
1850
if record.storage_kind == 'absent':
1851
raise errors.NoSuchRevision(self, revision_id)
1852
return record.get_bytes_as('fulltext')
1855
def check(self, revision_ids=None):
1856
"""Check consistency of all history of given revision_ids.
1858
Different repository implementations should override _check().
1860
:param revision_ids: A non-empty list of revision_ids whose ancestry
1861
will be checked. Typically the last revision_id of a branch.
1863
return self._check(revision_ids)
1865
def _check(self, revision_ids):
1866
result = check.Check(self)
1870
def _warn_if_deprecated(self):
1871
global _deprecation_warning_done
1872
if _deprecation_warning_done:
1874
_deprecation_warning_done = True
1875
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1876
% (self._format, self.bzrdir.transport.base))
1878
def supports_rich_root(self):
1879
return self._format.rich_root_data
1881
def _check_ascii_revisionid(self, revision_id, method):
1882
"""Private helper for ascii-only repositories."""
1883
# weave repositories refuse to store revisionids that are non-ascii.
1884
if revision_id is not None:
1885
# weaves require ascii revision ids.
1886
if isinstance(revision_id, unicode):
1888
revision_id.encode('ascii')
1889
except UnicodeEncodeError:
1890
raise errors.NonAsciiRevisionId(method, self)
1893
revision_id.decode('ascii')
1894
except UnicodeDecodeError:
1895
raise errors.NonAsciiRevisionId(method, self)
1897
def revision_graph_can_have_wrong_parents(self):
1898
"""Is it possible for this repository to have a revision graph with
1901
If True, then this repository must also implement
1902
_find_inconsistent_revision_parents so that check and reconcile can
1903
check for inconsistencies before proceeding with other checks that may
1904
depend on the revision index being consistent.
1906
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1909
# remove these delegates a while after bzr 0.15
1910
def __make_delegated(name, from_module):
1911
def _deprecated_repository_forwarder():
1912
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1913
% (name, from_module),
1916
m = __import__(from_module, globals(), locals(), [name])
1918
return getattr(m, name)
1919
except AttributeError:
1920
raise AttributeError('module %s has no name %s'
1922
globals()[name] = _deprecated_repository_forwarder
1925
'AllInOneRepository',
1926
'WeaveMetaDirRepository',
1927
'PreSplitOutRepositoryFormat',
1928
'RepositoryFormat4',
1929
'RepositoryFormat5',
1930
'RepositoryFormat6',
1931
'RepositoryFormat7',
1933
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1937
'RepositoryFormatKnit',
1938
'RepositoryFormatKnit1',
1940
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1943
def install_revision(repository, rev, revision_tree):
1944
"""Install all revision data into a repository."""
1945
install_revisions(repository, [(rev, revision_tree, None)])
1948
def install_revisions(repository, iterable, num_revisions=None, pb=None):
1949
"""Install all revision data into a repository.
1951
Accepts an iterable of revision, tree, signature tuples. The signature
1954
repository.start_write_group()
1956
for n, (revision, revision_tree, signature) in enumerate(iterable):
1957
_install_revision(repository, revision, revision_tree, signature)
1959
pb.update('Transferring revisions', n + 1, num_revisions)
1961
repository.abort_write_group()
1964
repository.commit_write_group()
1967
def _install_revision(repository, rev, revision_tree, signature):
1968
"""Install all revision data into a repository."""
1969
present_parents = []
1971
for p_id in rev.parent_ids:
1972
if repository.has_revision(p_id):
1973
present_parents.append(p_id)
1974
parent_trees[p_id] = repository.revision_tree(p_id)
1976
parent_trees[p_id] = repository.revision_tree(
1977
_mod_revision.NULL_REVISION)
1979
inv = revision_tree.inventory
1980
entries = inv.iter_entries()
1981
# backwards compatibility hack: skip the root id.
1982
if not repository.supports_rich_root():
1983
path, root = entries.next()
1984
if root.revision != rev.revision_id:
1985
raise errors.IncompatibleRevision(repr(repository))
1987
for path, ie in entries:
1988
text_keys[(ie.file_id, ie.revision)] = ie
1989
text_parent_map = repository.texts.get_parent_map(text_keys)
1990
missing_texts = set(text_keys) - set(text_parent_map)
1991
# Add the texts that are not already present
1992
for text_key in missing_texts:
1993
ie = text_keys[text_key]
1995
# FIXME: TODO: The following loop overlaps/duplicates that done by
1996
# commit to determine parents. There is a latent/real bug here where
1997
# the parents inserted are not those commit would do - in particular
1998
# they are not filtered by heads(). RBC, AB
1999
for revision, tree in parent_trees.iteritems():
2000
if ie.file_id not in tree:
2002
parent_id = tree.inventory[ie.file_id].revision
2003
if parent_id in text_parents:
2005
text_parents.append((ie.file_id, parent_id))
2006
lines = revision_tree.get_file(ie.file_id).readlines()
2007
repository.texts.add_lines(text_key, text_parents, lines)
2009
# install the inventory
2010
repository.add_inventory(rev.revision_id, inv, present_parents)
2011
except errors.RevisionAlreadyPresent:
2013
if signature is not None:
2014
repository.add_signature_text(rev.revision_id, signature)
2015
repository.add_revision(rev.revision_id, rev, inv)
2018
class MetaDirRepository(Repository):
2019
"""Repositories in the new meta-dir layout.
2021
:ivar _transport: Transport for access to repository control files,
2022
typically pointing to .bzr/repository.
2025
def __init__(self, _format, a_bzrdir, control_files):
2026
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2027
self._transport = control_files._transport
2029
def is_shared(self):
2030
"""Return True if this repository is flagged as a shared repository."""
2031
return self._transport.has('shared-storage')
2034
def set_make_working_trees(self, new_value):
2035
"""Set the policy flag for making working trees when creating branches.
2037
This only applies to branches that use this repository.
2039
The default is 'True'.
2040
:param new_value: True to restore the default, False to disable making
2045
self._transport.delete('no-working-trees')
2046
except errors.NoSuchFile:
2049
self._transport.put_bytes('no-working-trees', '',
2050
mode=self.bzrdir._get_file_mode())
2052
def make_working_trees(self):
2053
"""Returns the policy for making working trees on new branches."""
2054
return not self._transport.has('no-working-trees')
2057
class MetaDirVersionedFileRepository(MetaDirRepository):
2058
"""Repositories in a meta-dir, that work via versioned file objects."""
2060
def __init__(self, _format, a_bzrdir, control_files):
2061
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2065
class RepositoryFormatRegistry(registry.Registry):
2066
"""Registry of RepositoryFormats."""
2068
def get(self, format_string):
2069
r = registry.Registry.get(self, format_string)
2075
format_registry = RepositoryFormatRegistry()
2076
"""Registry of formats, indexed by their identifying format string.
2078
This can contain either format instances themselves, or classes/factories that
2079
can be called to obtain one.
2083
#####################################################################
2084
# Repository Formats
2086
class RepositoryFormat(object):
2087
"""A repository format.
2089
Formats provide three things:
2090
* An initialization routine to construct repository data on disk.
2091
* a format string which is used when the BzrDir supports versioned
2093
* an open routine which returns a Repository instance.
2095
There is one and only one Format subclass for each on-disk format. But
2096
there can be one Repository subclass that is used for several different
2097
formats. The _format attribute on a Repository instance can be used to
2098
determine the disk format.
2100
Formats are placed in an dict by their format string for reference
2101
during opening. These should be subclasses of RepositoryFormat
2104
Once a format is deprecated, just deprecate the initialize and open
2105
methods on the format class. Do not deprecate the object, as the
2106
object will be created every system load.
2108
Common instance attributes:
2109
_matchingbzrdir - the bzrdir format that the repository format was
2110
originally written to work with. This can be used if manually
2111
constructing a bzrdir and repository, or more commonly for test suite
2115
# Set to True or False in derived classes. True indicates that the format
2116
# supports ghosts gracefully.
2117
supports_ghosts = None
2118
# Can this repository be given external locations to lookup additional
2119
# data. Set to True or False in derived classes.
2120
supports_external_lookups = None
2123
return "<%s>" % self.__class__.__name__
2125
def __eq__(self, other):
2126
# format objects are generally stateless
2127
return isinstance(other, self.__class__)
2129
def __ne__(self, other):
2130
return not self == other
2133
def find_format(klass, a_bzrdir):
2134
"""Return the format for the repository object in a_bzrdir.
2136
This is used by bzr native formats that have a "format" file in
2137
the repository. Other methods may be used by different types of
2141
transport = a_bzrdir.get_repository_transport(None)
2142
format_string = transport.get("format").read()
2143
return format_registry.get(format_string)
2144
except errors.NoSuchFile:
2145
raise errors.NoRepositoryPresent(a_bzrdir)
2147
raise errors.UnknownFormatError(format=format_string,
2151
def register_format(klass, format):
2152
format_registry.register(format.get_format_string(), format)
2155
def unregister_format(klass, format):
2156
format_registry.remove(format.get_format_string())
2159
def get_default_format(klass):
2160
"""Return the current default format."""
2161
from bzrlib import bzrdir
2162
return bzrdir.format_registry.make_bzrdir('default').repository_format
2164
def get_format_string(self):
2165
"""Return the ASCII format string that identifies this format.
2167
Note that in pre format ?? repositories the format string is
2168
not permitted nor written to disk.
2170
raise NotImplementedError(self.get_format_string)
2172
def get_format_description(self):
2173
"""Return the short description for this format."""
2174
raise NotImplementedError(self.get_format_description)
2176
# TODO: this shouldn't be in the base class, it's specific to things that
2177
# use weaves or knits -- mbp 20070207
2178
def _get_versioned_file_store(self,
2183
versionedfile_class=None,
2184
versionedfile_kwargs={},
2186
if versionedfile_class is None:
2187
versionedfile_class = self._versionedfile_class
2188
weave_transport = control_files._transport.clone(name)
2189
dir_mode = control_files._dir_mode
2190
file_mode = control_files._file_mode
2191
return VersionedFileStore(weave_transport, prefixed=prefixed,
2193
file_mode=file_mode,
2194
versionedfile_class=versionedfile_class,
2195
versionedfile_kwargs=versionedfile_kwargs,
2198
def initialize(self, a_bzrdir, shared=False):
2199
"""Initialize a repository of this format in a_bzrdir.
2201
:param a_bzrdir: The bzrdir to put the new repository in it.
2202
:param shared: The repository should be initialized as a sharable one.
2203
:returns: The new repository object.
2205
This may raise UninitializableFormat if shared repository are not
2206
compatible the a_bzrdir.
2208
raise NotImplementedError(self.initialize)
2210
def is_supported(self):
2211
"""Is this format supported?
2213
Supported formats must be initializable and openable.
2214
Unsupported formats may not support initialization or committing or
2215
some other features depending on the reason for not being supported.
2219
def check_conversion_target(self, target_format):
2220
raise NotImplementedError(self.check_conversion_target)
2222
def open(self, a_bzrdir, _found=False):
2223
"""Return an instance of this format for the bzrdir a_bzrdir.
2225
_found is a private parameter, do not use it.
2227
raise NotImplementedError(self.open)
2230
class MetaDirRepositoryFormat(RepositoryFormat):
2231
"""Common base class for the new repositories using the metadir layout."""
2233
rich_root_data = False
2234
supports_tree_reference = False
2235
supports_external_lookups = False
2236
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
2239
super(MetaDirRepositoryFormat, self).__init__()
2241
def _create_control_files(self, a_bzrdir):
2242
"""Create the required files and the initial control_files object."""
2243
# FIXME: RBC 20060125 don't peek under the covers
2244
# NB: no need to escape relative paths that are url safe.
2245
repository_transport = a_bzrdir.get_repository_transport(self)
2246
control_files = lockable_files.LockableFiles(repository_transport,
2247
'lock', lockdir.LockDir)
2248
control_files.create_lock()
2249
return control_files
2251
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2252
"""Upload the initial blank content."""
2253
control_files = self._create_control_files(a_bzrdir)
2254
control_files.lock_write()
2255
transport = control_files._transport
2257
utf8_files += [('shared-storage', '')]
2259
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
2260
for (filename, content_stream) in files:
2261
transport.put_file(filename, content_stream,
2262
mode=a_bzrdir._get_file_mode())
2263
for (filename, content_bytes) in utf8_files:
2264
transport.put_bytes_non_atomic(filename, content_bytes,
2265
mode=a_bzrdir._get_file_mode())
2267
control_files.unlock()
2270
# formats which have no format string are not discoverable
2271
# and not independently creatable, so are not registered. They're
2272
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2273
# needed, it's constructed directly by the BzrDir. Non-native formats where
2274
# the repository is not separately opened are similar.
2276
format_registry.register_lazy(
2277
'Bazaar-NG Repository format 7',
2278
'bzrlib.repofmt.weaverepo',
2282
format_registry.register_lazy(
2283
'Bazaar-NG Knit Repository Format 1',
2284
'bzrlib.repofmt.knitrepo',
2285
'RepositoryFormatKnit1',
2288
format_registry.register_lazy(
2289
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2290
'bzrlib.repofmt.knitrepo',
2291
'RepositoryFormatKnit3',
2294
format_registry.register_lazy(
2295
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2296
'bzrlib.repofmt.knitrepo',
2297
'RepositoryFormatKnit4',
2300
# Pack-based formats. There is one format for pre-subtrees, and one for
2301
# post-subtrees to allow ease of testing.
2302
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
2303
format_registry.register_lazy(
2304
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2305
'bzrlib.repofmt.pack_repo',
2306
'RepositoryFormatKnitPack1',
2308
format_registry.register_lazy(
2309
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2310
'bzrlib.repofmt.pack_repo',
2311
'RepositoryFormatKnitPack3',
2313
format_registry.register_lazy(
2314
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2315
'bzrlib.repofmt.pack_repo',
2316
'RepositoryFormatKnitPack4',
2318
format_registry.register_lazy(
2319
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
2320
'bzrlib.repofmt.pack_repo',
2321
'RepositoryFormatKnitPack5',
2323
format_registry.register_lazy(
2324
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
2325
'bzrlib.repofmt.pack_repo',
2326
'RepositoryFormatKnitPack5RichRoot',
2328
format_registry.register_lazy(
2329
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
2330
'bzrlib.repofmt.pack_repo',
2331
'RepositoryFormatKnitPack5RichRootBroken',
2334
# Development formats.
2335
# 1.7->1.8 go below here
2336
format_registry.register_lazy(
2337
"Bazaar development format 2 (needs bzr.dev from before 1.8)\n",
2338
'bzrlib.repofmt.pack_repo',
2339
'RepositoryFormatPackDevelopment2',
2341
format_registry.register_lazy(
2342
("Bazaar development format 2 with subtree support "
2343
"(needs bzr.dev from before 1.8)\n"),
2344
'bzrlib.repofmt.pack_repo',
2345
'RepositoryFormatPackDevelopment2Subtree',
2349
class InterRepository(InterObject):
2350
"""This class represents operations taking place between two repositories.
2352
Its instances have methods like copy_content and fetch, and contain
2353
references to the source and target repositories these operations can be
2356
Often we will provide convenience methods on 'repository' which carry out
2357
operations with another repository - they will always forward to
2358
InterRepository.get(other).method_name(parameters).
2362
"""The available optimised InterRepository types."""
2364
def copy_content(self, revision_id=None):
2365
raise NotImplementedError(self.copy_content)
2367
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2368
"""Fetch the content required to construct revision_id.
2370
The content is copied from self.source to self.target.
2372
:param revision_id: if None all content is copied, if NULL_REVISION no
2374
:param pb: optional progress bar to use for progress reports. If not
2375
provided a default one will be created.
2377
:returns: (copied_revision_count, failures).
2379
# Normally we should find a specific InterRepository subclass to do
2380
# the fetch; if nothing else then at least InterSameDataRepository.
2381
# If none of them is suitable it looks like fetching is not possible;
2382
# we try to give a good message why. _assert_same_model will probably
2383
# give a helpful message; otherwise a generic one.
2384
self._assert_same_model(self.source, self.target)
2385
raise errors.IncompatibleRepositories(self.source, self.target,
2386
"no suitableInterRepository found")
2388
def _walk_to_common_revisions(self, revision_ids):
2389
"""Walk out from revision_ids in source to revisions target has.
2391
:param revision_ids: The start point for the search.
2392
:return: A set of revision ids.
2394
target_graph = self.target.get_graph()
2395
revision_ids = frozenset(revision_ids)
2396
if set(target_graph.get_parent_map(revision_ids)) == revision_ids:
2397
return graph.SearchResult(revision_ids, set(), 0, set())
2398
missing_revs = set()
2399
source_graph = self.source.get_graph()
2400
# ensure we don't pay silly lookup costs.
2401
searcher = source_graph._make_breadth_first_searcher(revision_ids)
2402
null_set = frozenset([_mod_revision.NULL_REVISION])
2405
next_revs, ghosts = searcher.next_with_ghosts()
2406
except StopIteration:
2408
if revision_ids.intersection(ghosts):
2409
absent_ids = set(revision_ids.intersection(ghosts))
2410
# If all absent_ids are present in target, no error is needed.
2411
absent_ids.difference_update(
2412
set(target_graph.get_parent_map(absent_ids)))
2414
raise errors.NoSuchRevision(self.source, absent_ids.pop())
2415
# we don't care about other ghosts as we can't fetch them and
2416
# haven't been asked to.
2417
next_revs = set(next_revs)
2418
# we always have NULL_REVISION present.
2419
have_revs = set(target_graph.get_parent_map(next_revs)).union(null_set)
2420
missing_revs.update(next_revs - have_revs)
2421
searcher.stop_searching_any(have_revs)
2422
return searcher.get_result()
2424
@deprecated_method(one_two)
2426
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2427
"""Return the revision ids that source has that target does not.
2429
These are returned in topological order.
2431
:param revision_id: only return revision ids included by this
2433
:param find_ghosts: If True find missing revisions in deep history
2434
rather than just finding the surface difference.
2436
return list(self.search_missing_revision_ids(
2437
revision_id, find_ghosts).get_keys())
2440
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2441
"""Return the revision ids that source has that target does not.
2443
:param revision_id: only return revision ids included by this
2445
:param find_ghosts: If True find missing revisions in deep history
2446
rather than just finding the surface difference.
2447
:return: A bzrlib.graph.SearchResult.
2449
# stop searching at found target revisions.
2450
if not find_ghosts and revision_id is not None:
2451
return self._walk_to_common_revisions([revision_id])
2452
# generic, possibly worst case, slow code path.
2453
target_ids = set(self.target.all_revision_ids())
2454
if revision_id is not None:
2455
source_ids = self.source.get_ancestry(revision_id)
2456
if source_ids[0] is not None:
2457
raise AssertionError()
2460
source_ids = self.source.all_revision_ids()
2461
result_set = set(source_ids).difference(target_ids)
2462
return self.source.revision_ids_to_search_result(result_set)
2465
def _same_model(source, target):
2466
"""True if source and target have the same data representation.
2468
Note: this is always called on the base class; overriding it in a
2469
subclass will have no effect.
2472
InterRepository._assert_same_model(source, target)
2474
except errors.IncompatibleRepositories, e:
2478
def _assert_same_model(source, target):
2479
"""Raise an exception if two repositories do not use the same model.
2481
if source.supports_rich_root() != target.supports_rich_root():
2482
raise errors.IncompatibleRepositories(source, target,
2483
"different rich-root support")
2484
if source._serializer != target._serializer:
2485
raise errors.IncompatibleRepositories(source, target,
2486
"different serializers")
2489
class InterSameDataRepository(InterRepository):
2490
"""Code for converting between repositories that represent the same data.
2492
Data format and model must match for this to work.
2496
def _get_repo_format_to_test(self):
2497
"""Repository format for testing with.
2499
InterSameData can pull from subtree to subtree and from non-subtree to
2500
non-subtree, so we test this with the richest repository format.
2502
from bzrlib.repofmt import knitrepo
2503
return knitrepo.RepositoryFormatKnit3()
2506
def is_compatible(source, target):
2507
return InterRepository._same_model(source, target)
2510
def copy_content(self, revision_id=None):
2511
"""Make a complete copy of the content in self into destination.
2513
This copies both the repository's revision data, and configuration information
2514
such as the make_working_trees setting.
2516
This is a destructive operation! Do not use it on existing
2519
:param revision_id: Only copy the content needed to construct
2520
revision_id and its parents.
2523
self.target.set_make_working_trees(self.source.make_working_trees())
2524
except NotImplementedError:
2526
# but don't bother fetching if we have the needed data now.
2527
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2528
self.target.has_revision(revision_id)):
2530
self.target.fetch(self.source, revision_id=revision_id)
2533
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2534
"""See InterRepository.fetch()."""
2535
from bzrlib.fetch import RepoFetcher
2536
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2537
self.source, self.source._format, self.target,
2538
self.target._format)
2539
f = RepoFetcher(to_repository=self.target,
2540
from_repository=self.source,
2541
last_revision=revision_id,
2542
pb=pb, find_ghosts=find_ghosts)
2543
return f.count_copied, f.failed_revisions
2546
class InterWeaveRepo(InterSameDataRepository):
2547
"""Optimised code paths between Weave based repositories.
2549
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2550
implemented lazy inter-object optimisation.
2554
def _get_repo_format_to_test(self):
2555
from bzrlib.repofmt import weaverepo
2556
return weaverepo.RepositoryFormat7()
2559
def is_compatible(source, target):
2560
"""Be compatible with known Weave formats.
2562
We don't test for the stores being of specific types because that
2563
could lead to confusing results, and there is no need to be
2566
from bzrlib.repofmt.weaverepo import (
2572
return (isinstance(source._format, (RepositoryFormat5,
2574
RepositoryFormat7)) and
2575
isinstance(target._format, (RepositoryFormat5,
2577
RepositoryFormat7)))
2578
except AttributeError:
2582
def copy_content(self, revision_id=None):
2583
"""See InterRepository.copy_content()."""
2584
# weave specific optimised path:
2586
self.target.set_make_working_trees(self.source.make_working_trees())
2587
except (errors.RepositoryUpgradeRequired, NotImplemented):
2589
# FIXME do not peek!
2590
if self.source._transport.listable():
2591
pb = ui.ui_factory.nested_progress_bar()
2593
self.target.texts.insert_record_stream(
2594
self.source.texts.get_record_stream(
2595
self.source.texts.keys(), 'topological', False))
2596
pb.update('copying inventory', 0, 1)
2597
self.target.inventories.insert_record_stream(
2598
self.source.inventories.get_record_stream(
2599
self.source.inventories.keys(), 'topological', False))
2600
self.target.signatures.insert_record_stream(
2601
self.source.signatures.get_record_stream(
2602
self.source.signatures.keys(),
2604
self.target.revisions.insert_record_stream(
2605
self.source.revisions.get_record_stream(
2606
self.source.revisions.keys(),
2607
'topological', True))
2611
self.target.fetch(self.source, revision_id=revision_id)
2614
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2615
"""See InterRepository.fetch()."""
2616
from bzrlib.fetch import RepoFetcher
2617
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2618
self.source, self.source._format, self.target, self.target._format)
2619
f = RepoFetcher(to_repository=self.target,
2620
from_repository=self.source,
2621
last_revision=revision_id,
2622
pb=pb, find_ghosts=find_ghosts)
2623
return f.count_copied, f.failed_revisions
2626
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2627
"""See InterRepository.missing_revision_ids()."""
2628
# we want all revisions to satisfy revision_id in source.
2629
# but we don't want to stat every file here and there.
2630
# we want then, all revisions other needs to satisfy revision_id
2631
# checked, but not those that we have locally.
2632
# so the first thing is to get a subset of the revisions to
2633
# satisfy revision_id in source, and then eliminate those that
2634
# we do already have.
2635
# this is slow on high latency connection to self, but as as this
2636
# disk format scales terribly for push anyway due to rewriting
2637
# inventory.weave, this is considered acceptable.
2639
if revision_id is not None:
2640
source_ids = self.source.get_ancestry(revision_id)
2641
if source_ids[0] is not None:
2642
raise AssertionError()
2645
source_ids = self.source._all_possible_ids()
2646
source_ids_set = set(source_ids)
2647
# source_ids is the worst possible case we may need to pull.
2648
# now we want to filter source_ids against what we actually
2649
# have in target, but don't try to check for existence where we know
2650
# we do not have a revision as that would be pointless.
2651
target_ids = set(self.target._all_possible_ids())
2652
possibly_present_revisions = target_ids.intersection(source_ids_set)
2653
actually_present_revisions = set(
2654
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2655
required_revisions = source_ids_set.difference(actually_present_revisions)
2656
if revision_id is not None:
2657
# we used get_ancestry to determine source_ids then we are assured all
2658
# revisions referenced are present as they are installed in topological order.
2659
# and the tip revision was validated by get_ancestry.
2660
result_set = required_revisions
2662
# if we just grabbed the possibly available ids, then
2663
# we only have an estimate of whats available and need to validate
2664
# that against the revision records.
2666
self.source._eliminate_revisions_not_present(required_revisions))
2667
return self.source.revision_ids_to_search_result(result_set)
2670
class InterKnitRepo(InterSameDataRepository):
2671
"""Optimised code paths between Knit based repositories."""
2674
def _get_repo_format_to_test(self):
2675
from bzrlib.repofmt import knitrepo
2676
return knitrepo.RepositoryFormatKnit1()
2679
def is_compatible(source, target):
2680
"""Be compatible with known Knit formats.
2682
We don't test for the stores being of specific types because that
2683
could lead to confusing results, and there is no need to be
2686
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2688
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2689
isinstance(target._format, RepositoryFormatKnit))
2690
except AttributeError:
2692
return are_knits and InterRepository._same_model(source, target)
2695
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2696
"""See InterRepository.fetch()."""
2697
from bzrlib.fetch import RepoFetcher
2698
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2699
self.source, self.source._format, self.target, self.target._format)
2700
f = RepoFetcher(to_repository=self.target,
2701
from_repository=self.source,
2702
last_revision=revision_id,
2703
pb=pb, find_ghosts=find_ghosts)
2704
return f.count_copied, f.failed_revisions
2707
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2708
"""See InterRepository.missing_revision_ids()."""
2709
if revision_id is not None:
2710
source_ids = self.source.get_ancestry(revision_id)
2711
if source_ids[0] is not None:
2712
raise AssertionError()
2715
source_ids = self.source.all_revision_ids()
2716
source_ids_set = set(source_ids)
2717
# source_ids is the worst possible case we may need to pull.
2718
# now we want to filter source_ids against what we actually
2719
# have in target, but don't try to check for existence where we know
2720
# we do not have a revision as that would be pointless.
2721
target_ids = set(self.target.all_revision_ids())
2722
possibly_present_revisions = target_ids.intersection(source_ids_set)
2723
actually_present_revisions = set(
2724
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2725
required_revisions = source_ids_set.difference(actually_present_revisions)
2726
if revision_id is not None:
2727
# we used get_ancestry to determine source_ids then we are assured all
2728
# revisions referenced are present as they are installed in topological order.
2729
# and the tip revision was validated by get_ancestry.
2730
result_set = required_revisions
2732
# if we just grabbed the possibly available ids, then
2733
# we only have an estimate of whats available and need to validate
2734
# that against the revision records.
2736
self.source._eliminate_revisions_not_present(required_revisions))
2737
return self.source.revision_ids_to_search_result(result_set)
2740
class InterPackRepo(InterSameDataRepository):
2741
"""Optimised code paths between Pack based repositories."""
2744
def _get_repo_format_to_test(self):
2745
from bzrlib.repofmt import pack_repo
2746
return pack_repo.RepositoryFormatKnitPack1()
2749
def is_compatible(source, target):
2750
"""Be compatible with known Pack formats.
2752
We don't test for the stores being of specific types because that
2753
could lead to confusing results, and there is no need to be
2756
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2758
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2759
isinstance(target._format, RepositoryFormatPack))
2760
except AttributeError:
2762
return are_packs and InterRepository._same_model(source, target)
2765
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2766
"""See InterRepository.fetch()."""
2767
if (len(self.source._fallback_repositories) > 0 or
2768
len(self.target._fallback_repositories) > 0):
2769
# The pack layer is not aware of fallback repositories, so when
2770
# fetching from a stacked repository or into a stacked repository
2771
# we use the generic fetch logic which uses the VersionedFiles
2772
# attributes on repository.
2773
from bzrlib.fetch import RepoFetcher
2774
fetcher = RepoFetcher(self.target, self.source, revision_id,
2776
return fetcher.count_copied, fetcher.failed_revisions
2777
from bzrlib.repofmt.pack_repo import Packer
2778
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2779
self.source, self.source._format, self.target, self.target._format)
2780
self.count_copied = 0
2781
if revision_id is None:
2783
# everything to do - use pack logic
2784
# to fetch from all packs to one without
2785
# inventory parsing etc, IFF nothing to be copied is in the target.
2787
source_revision_ids = frozenset(self.source.all_revision_ids())
2788
revision_ids = source_revision_ids - \
2789
frozenset(self.target.get_parent_map(source_revision_ids))
2790
revision_keys = [(revid,) for revid in revision_ids]
2791
index = self.target._pack_collection.revision_index.combined_index
2792
present_revision_ids = set(item[1][0] for item in
2793
index.iter_entries(revision_keys))
2794
revision_ids = set(revision_ids) - present_revision_ids
2795
# implementing the TODO will involve:
2796
# - detecting when all of a pack is selected
2797
# - avoiding as much as possible pre-selection, so the
2798
# more-core routines such as create_pack_from_packs can filter in
2799
# a just-in-time fashion. (though having a HEADS list on a
2800
# repository might make this a lot easier, because we could
2801
# sensibly detect 'new revisions' without doing a full index scan.
2802
elif _mod_revision.is_null(revision_id):
2807
revision_ids = self.search_missing_revision_ids(revision_id,
2808
find_ghosts=find_ghosts).get_keys()
2809
except errors.NoSuchRevision:
2810
raise errors.InstallFailed([revision_id])
2811
if len(revision_ids) == 0:
2813
packs = self.source._pack_collection.all_packs()
2814
pack = Packer(self.target._pack_collection, packs, '.fetch',
2815
revision_ids).pack()
2816
if pack is not None:
2817
self.target._pack_collection._save_pack_names()
2818
# Trigger an autopack. This may duplicate effort as we've just done
2819
# a pack creation, but for now it is simpler to think about as
2820
# 'upload data, then repack if needed'.
2821
self.target._pack_collection.autopack()
2822
return (pack.get_revision_count(), [])
2827
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2828
"""See InterRepository.missing_revision_ids().
2830
:param find_ghosts: Find ghosts throughout the ancestry of
2833
if not find_ghosts and revision_id is not None:
2834
return self._walk_to_common_revisions([revision_id])
2835
elif revision_id is not None:
2836
# Find ghosts: search for revisions pointing from one repository to
2837
# the other, and vice versa, anywhere in the history of revision_id.
2838
graph = self.target.get_graph(other_repository=self.source)
2839
searcher = graph._make_breadth_first_searcher([revision_id])
2843
next_revs, ghosts = searcher.next_with_ghosts()
2844
except StopIteration:
2846
if revision_id in ghosts:
2847
raise errors.NoSuchRevision(self.source, revision_id)
2848
found_ids.update(next_revs)
2849
found_ids.update(ghosts)
2850
found_ids = frozenset(found_ids)
2851
# Double query here: should be able to avoid this by changing the
2852
# graph api further.
2853
result_set = found_ids - frozenset(
2854
self.target.get_parent_map(found_ids))
2856
source_ids = self.source.all_revision_ids()
2857
# source_ids is the worst possible case we may need to pull.
2858
# now we want to filter source_ids against what we actually
2859
# have in target, but don't try to check for existence where we know
2860
# we do not have a revision as that would be pointless.
2861
target_ids = set(self.target.all_revision_ids())
2862
result_set = set(source_ids).difference(target_ids)
2863
return self.source.revision_ids_to_search_result(result_set)
2866
class InterModel1and2(InterRepository):
2869
def _get_repo_format_to_test(self):
2873
def is_compatible(source, target):
2874
if not source.supports_rich_root() and target.supports_rich_root():
2880
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2881
"""See InterRepository.fetch()."""
2882
from bzrlib.fetch import Model1toKnit2Fetcher
2883
f = Model1toKnit2Fetcher(to_repository=self.target,
2884
from_repository=self.source,
2885
last_revision=revision_id,
2886
pb=pb, find_ghosts=find_ghosts)
2887
return f.count_copied, f.failed_revisions
2890
def copy_content(self, revision_id=None):
2891
"""Make a complete copy of the content in self into destination.
2893
This is a destructive operation! Do not use it on existing
2896
:param revision_id: Only copy the content needed to construct
2897
revision_id and its parents.
2900
self.target.set_make_working_trees(self.source.make_working_trees())
2901
except NotImplementedError:
2903
# but don't bother fetching if we have the needed data now.
2904
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2905
self.target.has_revision(revision_id)):
2907
self.target.fetch(self.source, revision_id=revision_id)
2910
class InterKnit1and2(InterKnitRepo):
2913
def _get_repo_format_to_test(self):
2917
def is_compatible(source, target):
2918
"""Be compatible with Knit1 source and Knit3 target"""
2920
from bzrlib.repofmt.knitrepo import (
2921
RepositoryFormatKnit1,
2922
RepositoryFormatKnit3,
2924
from bzrlib.repofmt.pack_repo import (
2925
RepositoryFormatKnitPack1,
2926
RepositoryFormatKnitPack3,
2927
RepositoryFormatKnitPack4,
2928
RepositoryFormatKnitPack5,
2929
RepositoryFormatKnitPack5RichRoot,
2930
RepositoryFormatPackDevelopment2,
2931
RepositoryFormatPackDevelopment2Subtree,
2934
RepositoryFormatKnit1, # no rr, no subtree
2935
RepositoryFormatKnitPack1, # no rr, no subtree
2936
RepositoryFormatPackDevelopment2, # no rr, no subtree
2937
RepositoryFormatKnitPack5, # no rr, no subtree
2940
RepositoryFormatKnit3, # rr, subtree
2941
RepositoryFormatKnitPack3, # rr, subtree
2942
RepositoryFormatKnitPack4, # rr, no subtree
2943
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
2944
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
2946
for format in norichroot:
2947
if format.rich_root_data:
2948
raise AssertionError('Format %s is a rich-root format'
2949
' but is included in the non-rich-root list'
2951
for format in richroot:
2952
if not format.rich_root_data:
2953
raise AssertionError('Format %s is not a rich-root format'
2954
' but is included in the rich-root list'
2956
# TODO: One alternative is to just check format.rich_root_data,
2957
# instead of keeping membership lists. However, the formats
2958
# *also* have to use the same 'Knit' style of storage
2959
# (line-deltas, fulltexts, etc.)
2960
return (isinstance(source._format, norichroot) and
2961
isinstance(target._format, richroot))
2962
except AttributeError:
2966
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2967
"""See InterRepository.fetch()."""
2968
from bzrlib.fetch import Knit1to2Fetcher
2969
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2970
self.source, self.source._format, self.target,
2971
self.target._format)
2972
f = Knit1to2Fetcher(to_repository=self.target,
2973
from_repository=self.source,
2974
last_revision=revision_id,
2975
pb=pb, find_ghosts=find_ghosts)
2976
return f.count_copied, f.failed_revisions
2979
class InterDifferingSerializer(InterKnitRepo):
2982
def _get_repo_format_to_test(self):
2986
def is_compatible(source, target):
2987
"""Be compatible with Knit2 source and Knit3 target"""
2988
if source.supports_rich_root() != target.supports_rich_root():
2990
# Ideally, we'd support fetching if the source had no tree references
2991
# even if it supported them...
2992
if (getattr(source, '_format.supports_tree_reference', False) and
2993
not getattr(target, '_format.supports_tree_reference', False)):
2998
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2999
"""See InterRepository.fetch()."""
3000
revision_ids = self.target.search_missing_revision_ids(self.source,
3001
revision_id, find_ghosts=find_ghosts).get_keys()
3002
revision_ids = tsort.topo_sort(
3003
self.source.get_graph().get_parent_map(revision_ids))
3004
def revisions_iterator():
3005
for current_revision_id in revision_ids:
3006
revision = self.source.get_revision(current_revision_id)
3007
tree = self.source.revision_tree(current_revision_id)
3009
signature = self.source.get_signature_text(
3010
current_revision_id)
3011
except errors.NoSuchRevision:
3013
yield revision, tree, signature
3015
my_pb = ui.ui_factory.nested_progress_bar()
3020
install_revisions(self.target, revisions_iterator(),
3021
len(revision_ids), pb)
3023
if my_pb is not None:
3025
return len(revision_ids), 0
3028
class InterOtherToRemote(InterRepository):
3030
def __init__(self, source, target):
3031
InterRepository.__init__(self, source, target)
3032
self._real_inter = None
3035
def is_compatible(source, target):
3036
if isinstance(target, remote.RemoteRepository):
3040
def _ensure_real_inter(self):
3041
if self._real_inter is None:
3042
self.target._ensure_real()
3043
real_target = self.target._real_repository
3044
self._real_inter = InterRepository.get(self.source, real_target)
3046
def copy_content(self, revision_id=None):
3047
self._ensure_real_inter()
3048
self._real_inter.copy_content(revision_id=revision_id)
3050
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3051
self._ensure_real_inter()
3052
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3053
find_ghosts=find_ghosts)
3056
def _get_repo_format_to_test(self):
3060
class InterRemoteToOther(InterRepository):
3062
def __init__(self, source, target):
3063
InterRepository.__init__(self, source, target)
3064
self._real_inter = None
3067
def is_compatible(source, target):
3068
if not isinstance(source, remote.RemoteRepository):
3070
# Is source's model compatible with target's model?
3071
source._ensure_real()
3072
real_source = source._real_repository
3073
if isinstance(real_source, remote.RemoteRepository):
3074
raise NotImplementedError(
3075
"We don't support remote repos backed by remote repos yet.")
3076
return InterRepository._same_model(real_source, target)
3078
def _ensure_real_inter(self):
3079
if self._real_inter is None:
3080
self.source._ensure_real()
3081
real_source = self.source._real_repository
3082
self._real_inter = InterRepository.get(real_source, self.target)
3084
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3085
self._ensure_real_inter()
3086
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3087
find_ghosts=find_ghosts)
3089
def copy_content(self, revision_id=None):
3090
self._ensure_real_inter()
3091
self._real_inter.copy_content(revision_id=revision_id)
3094
def _get_repo_format_to_test(self):
3099
InterRepository.register_optimiser(InterDifferingSerializer)
3100
InterRepository.register_optimiser(InterSameDataRepository)
3101
InterRepository.register_optimiser(InterWeaveRepo)
3102
InterRepository.register_optimiser(InterKnitRepo)
3103
InterRepository.register_optimiser(InterModel1and2)
3104
InterRepository.register_optimiser(InterKnit1and2)
3105
InterRepository.register_optimiser(InterPackRepo)
3106
InterRepository.register_optimiser(InterOtherToRemote)
3107
InterRepository.register_optimiser(InterRemoteToOther)
3110
class CopyConverter(object):
3111
"""A repository conversion tool which just performs a copy of the content.
3113
This is slow but quite reliable.
3116
def __init__(self, target_format):
3117
"""Create a CopyConverter.
3119
:param target_format: The format the resulting repository should be.
3121
self.target_format = target_format
3123
def convert(self, repo, pb):
3124
"""Perform the conversion of to_convert, giving feedback via pb.
3126
:param to_convert: The disk object to convert.
3127
:param pb: a progress bar to use for progress information.
3132
# this is only useful with metadir layouts - separated repo content.
3133
# trigger an assertion if not such
3134
repo._format.get_format_string()
3135
self.repo_dir = repo.bzrdir
3136
self.step('Moving repository to repository.backup')
3137
self.repo_dir.transport.move('repository', 'repository.backup')
3138
backup_transport = self.repo_dir.transport.clone('repository.backup')
3139
repo._format.check_conversion_target(self.target_format)
3140
self.source_repo = repo._format.open(self.repo_dir,
3142
_override_transport=backup_transport)
3143
self.step('Creating new repository')
3144
converted = self.target_format.initialize(self.repo_dir,
3145
self.source_repo.is_shared())
3146
converted.lock_write()
3148
self.step('Copying content into repository.')
3149
self.source_repo.copy_content_into(converted)
3152
self.step('Deleting old repository content.')
3153
self.repo_dir.transport.delete_tree('repository.backup')
3154
self.pb.note('repository converted')
3156
def step(self, message):
3157
"""Update the pb by a step."""
3159
self.pb.update(message, self.count, self.total)
3171
def _unescaper(match, _map=_unescape_map):
3172
code = match.group(1)
3176
if not code.startswith('#'):
3178
return unichr(int(code[1:])).encode('utf8')
3184
def _unescape_xml(data):
3185
"""Unescape predefined XML entities in a string of data."""
3187
if _unescape_re is None:
3188
_unescape_re = re.compile('\&([^;]*);')
3189
return _unescape_re.sub(_unescaper, data)
3192
class _VersionedFileChecker(object):
3194
def __init__(self, repository):
3195
self.repository = repository
3196
self.text_index = self.repository._generate_text_key_index()
3198
def calculate_file_version_parents(self, text_key):
3199
"""Calculate the correct parents for a file version according to
3202
parent_keys = self.text_index[text_key]
3203
if parent_keys == [_mod_revision.NULL_REVISION]:
3205
return tuple(parent_keys)
3207
def check_file_version_parents(self, texts, progress_bar=None):
3208
"""Check the parents stored in a versioned file are correct.
3210
It also detects file versions that are not referenced by their
3211
corresponding revision's inventory.
3213
:returns: A tuple of (wrong_parents, dangling_file_versions).
3214
wrong_parents is a dict mapping {revision_id: (stored_parents,
3215
correct_parents)} for each revision_id where the stored parents
3216
are not correct. dangling_file_versions is a set of (file_id,
3217
revision_id) tuples for versions that are present in this versioned
3218
file, but not used by the corresponding inventory.
3221
self.file_ids = set([file_id for file_id, _ in
3222
self.text_index.iterkeys()])
3223
# text keys is now grouped by file_id
3224
n_weaves = len(self.file_ids)
3225
files_in_revisions = {}
3226
revisions_of_files = {}
3227
n_versions = len(self.text_index)
3228
progress_bar.update('loading text store', 0, n_versions)
3229
parent_map = self.repository.texts.get_parent_map(self.text_index)
3230
# On unlistable transports this could well be empty/error...
3231
text_keys = self.repository.texts.keys()
3232
unused_keys = frozenset(text_keys) - set(self.text_index)
3233
for num, key in enumerate(self.text_index.iterkeys()):
3234
if progress_bar is not None:
3235
progress_bar.update('checking text graph', num, n_versions)
3236
correct_parents = self.calculate_file_version_parents(key)
3238
knit_parents = parent_map[key]
3239
except errors.RevisionNotPresent:
3242
if correct_parents != knit_parents:
3243
wrong_parents[key] = (knit_parents, correct_parents)
3244
return wrong_parents, unused_keys
3247
def _old_get_graph(repository, revision_id):
3248
"""DO NOT USE. That is all. I'm serious."""
3249
graph = repository.get_graph()
3250
revision_graph = dict(((key, value) for key, value in
3251
graph.iter_ancestry([revision_id]) if value is not None))
3252
return _strip_NULL_ghosts(revision_graph)
3255
def _strip_NULL_ghosts(revision_graph):
3256
"""Also don't use this. more compatibility code for unmigrated clients."""
3257
# Filter ghosts, and null:
3258
if _mod_revision.NULL_REVISION in revision_graph:
3259
del revision_graph[_mod_revision.NULL_REVISION]
3260
for key, parents in revision_graph.items():
3261
revision_graph[key] = tuple(parent for parent in parents if parent
3263
return revision_graph
added
removed
bzrlib/repository.py
