1
# Copyright (C) 2005, 2006, 2007 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,
44
from bzrlib.bundle import serializer
45
from bzrlib.revisiontree import RevisionTree
46
from bzrlib.store.versioned import VersionedFileStore
47
from bzrlib.store.text import TextStore
48
from bzrlib.testament import Testament
51
from bzrlib.decorators import needs_read_lock, needs_write_lock
52
from bzrlib.inter import InterObject
53
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
54
from bzrlib.symbol_versioning import (
57
from bzrlib.trace import mutter, mutter_callsite, note, warning
60
# Old formats display a warning, but only once
61
_deprecation_warning_done = False
64
class CommitBuilder(object):
65
"""Provides an interface to build up a commit.
67
This allows describing a tree to be committed without needing to
68
know the internals of the format of the repository.
71
# all clients should supply tree roots.
72
record_root_entry = True
73
# the default CommitBuilder does not manage trees whose root is versioned.
74
_versioned_root = False
76
def __init__(self, repository, parents, config, timestamp=None,
77
timezone=None, committer=None, revprops=None,
79
"""Initiate a CommitBuilder.
81
:param repository: Repository to commit to.
82
:param parents: Revision ids of the parents of the new revision.
83
:param config: Configuration to use.
84
:param timestamp: Optional timestamp recorded for commit.
85
:param timezone: Optional timezone for timestamp.
86
:param committer: Optional committer to set for commit.
87
:param revprops: Optional dictionary of revision properties.
88
:param revision_id: Optional revision id.
93
self._committer = self._config.username()
95
assert isinstance(committer, basestring), type(committer)
96
self._committer = committer
98
self.new_inventory = Inventory(None)
99
self._new_revision_id = osutils.safe_revision_id(revision_id)
100
self.parents = parents
101
self.repository = repository
104
if revprops is not None:
105
self._revprops.update(revprops)
107
if timestamp is None:
108
timestamp = time.time()
109
# Restrict resolution to 1ms
110
self._timestamp = round(timestamp, 3)
113
self._timezone = osutils.local_time_offset()
115
self._timezone = int(timezone)
117
self._generate_revision_if_needed()
119
def commit(self, message):
120
"""Make the actual commit.
122
:return: The revision id of the recorded revision.
124
rev = _mod_revision.Revision(
125
timestamp=self._timestamp,
126
timezone=self._timezone,
127
committer=self._committer,
129
inventory_sha1=self.inv_sha1,
130
revision_id=self._new_revision_id,
131
properties=self._revprops)
132
rev.parent_ids = self.parents
133
self.repository.add_revision(self._new_revision_id, rev,
134
self.new_inventory, self._config)
135
self.repository.commit_write_group()
136
return self._new_revision_id
139
"""Abort the commit that is being built.
141
self.repository.abort_write_group()
143
def revision_tree(self):
144
"""Return the tree that was just committed.
146
After calling commit() this can be called to get a RevisionTree
147
representing the newly committed tree. This is preferred to
148
calling Repository.revision_tree() because that may require
149
deserializing the inventory, while we already have a copy in
152
return RevisionTree(self.repository, self.new_inventory,
153
self._new_revision_id)
155
def finish_inventory(self):
156
"""Tell the builder that the inventory is finished."""
157
if self.new_inventory.root is None:
158
symbol_versioning.warn('Root entry should be supplied to'
159
' record_entry_contents, as of bzr 0.10.',
160
DeprecationWarning, stacklevel=2)
161
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
162
self.new_inventory.revision_id = self._new_revision_id
163
self.inv_sha1 = self.repository.add_inventory(
164
self._new_revision_id,
169
def _gen_revision_id(self):
170
"""Return new revision-id."""
171
return generate_ids.gen_revision_id(self._config.username(),
174
def _generate_revision_if_needed(self):
175
"""Create a revision id if None was supplied.
177
If the repository can not support user-specified revision ids
178
they should override this function and raise CannotSetRevisionId
179
if _new_revision_id is not None.
181
:raises: CannotSetRevisionId
183
if self._new_revision_id is None:
184
self._new_revision_id = self._gen_revision_id()
185
self.random_revid = True
187
self.random_revid = False
189
def _check_root(self, ie, parent_invs, tree):
190
"""Helper for record_entry_contents.
192
:param ie: An entry being added.
193
:param parent_invs: The inventories of the parent revisions of the
195
:param tree: The tree that is being committed.
197
if ie.parent_id is not None:
198
# if ie is not root, add a root automatically.
199
symbol_versioning.warn('Root entry should be supplied to'
200
' record_entry_contents, as of bzr 0.10.',
201
DeprecationWarning, stacklevel=2)
202
self.record_entry_contents(tree.inventory.root.copy(), parent_invs,
205
# In this revision format, root entries have no knit or weave When
206
# serializing out to disk and back in root.revision is always
208
ie.revision = self._new_revision_id
210
def record_entry_contents(self, ie, parent_invs, path, tree):
211
"""Record the content of ie from tree into the commit if needed.
213
Side effect: sets ie.revision when unchanged
215
:param ie: An inventory entry present in the commit.
216
:param parent_invs: The inventories of the parent revisions of the
218
:param path: The path the entry is at in the tree.
219
:param tree: The tree which contains this entry and should be used to
221
:return: True if a new version of the entry has been recorded.
222
(Committing a merge where a file was only changed on the other side
223
will not return True.)
225
if self.new_inventory.root is None:
226
self._check_root(ie, parent_invs, tree)
227
self.new_inventory.add(ie)
229
# ie.revision is always None if the InventoryEntry is considered
230
# for committing. ie.snapshot will record the correct revision
231
# which may be the sole parent if it is untouched.
232
if ie.revision is not None:
233
return ie.revision == self._new_revision_id and (path != '' or
234
self._versioned_root)
236
parent_candiate_entries = ie.parent_candidates(parent_invs)
237
heads = self.repository.get_graph().heads(parent_candiate_entries.keys())
238
# XXX: Note that this is unordered - and this is tolerable because
239
# the previous code was also unordered.
240
previous_entries = dict((head, parent_candiate_entries[head]) for head
242
# we are creating a new revision for ie in the history store and
244
ie.snapshot(self._new_revision_id, path, previous_entries, tree, self)
245
return ie.revision == self._new_revision_id and (path != '' or
246
self._versioned_root)
248
def modified_directory(self, file_id, file_parents):
249
"""Record the presence of a symbolic link.
251
:param file_id: The file_id of the link to record.
252
:param file_parents: The per-file parent revision ids.
254
self._add_text_to_weave(file_id, [], file_parents.keys())
256
def modified_reference(self, file_id, file_parents):
257
"""Record the modification of a reference.
259
:param file_id: The file_id of the link to record.
260
:param file_parents: The per-file parent revision ids.
262
self._add_text_to_weave(file_id, [], file_parents.keys())
264
def modified_file_text(self, file_id, file_parents,
265
get_content_byte_lines, text_sha1=None,
267
"""Record the text of file file_id
269
:param file_id: The file_id of the file to record the text of.
270
:param file_parents: The per-file parent revision ids.
271
:param get_content_byte_lines: A callable which will return the byte
273
:param text_sha1: Optional SHA1 of the file contents.
274
:param text_size: Optional size of the file contents.
276
# mutter('storing text of file {%s} in revision {%s} into %r',
277
# file_id, self._new_revision_id, self.repository.weave_store)
278
# special case to avoid diffing on renames or
280
if (len(file_parents) == 1
281
and text_sha1 == file_parents.values()[0].text_sha1
282
and text_size == file_parents.values()[0].text_size):
283
previous_ie = file_parents.values()[0]
284
versionedfile = self.repository.weave_store.get_weave(file_id,
285
self.repository.get_transaction())
286
versionedfile.clone_text(self._new_revision_id,
287
previous_ie.revision, file_parents.keys())
288
return text_sha1, text_size
290
new_lines = get_content_byte_lines()
291
return self._add_text_to_weave(file_id, new_lines,
294
def modified_link(self, file_id, file_parents, link_target):
295
"""Record the presence of a symbolic link.
297
:param file_id: The file_id of the link to record.
298
:param file_parents: The per-file parent revision ids.
299
:param link_target: Target location of this link.
301
self._add_text_to_weave(file_id, [], file_parents.keys())
303
def _add_text_to_weave(self, file_id, new_lines, parents):
304
versionedfile = self.repository.weave_store.get_weave_or_empty(
305
file_id, self.repository.get_transaction())
306
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
307
# than add_lines, and allows committing when a parent is ghosted for
309
# Note: as we read the content directly from the tree, we know its not
310
# been turned into unicode or badly split - but a broken tree
311
# implementation could give us bad output from readlines() so this is
312
# not a guarantee of safety. What would be better is always checking
313
# the content during test suite execution. RBC 20070912
314
result = versionedfile.add_lines_with_ghosts(
315
self._new_revision_id, parents, new_lines,
316
random_id=self.random_revid, check_content=False)[0:2]
317
versionedfile.clear_cache()
321
class RootCommitBuilder(CommitBuilder):
322
"""This commitbuilder actually records the root id"""
324
# the root entry gets versioned properly by this builder.
325
_versioned_root = True
327
def _check_root(self, ie, parent_invs, tree):
328
"""Helper for record_entry_contents.
330
:param ie: An entry being added.
331
:param parent_invs: The inventories of the parent revisions of the
333
:param tree: The tree that is being committed.
335
# ie must be root for this builder
336
assert ie.parent_id is None
339
######################################################################
342
class Repository(object):
343
"""Repository holding history for one or more branches.
345
The repository holds and retrieves historical information including
346
revisions and file history. It's normally accessed only by the Branch,
347
which views a particular line of development through that history.
349
The Repository builds on top of Stores and a Transport, which respectively
350
describe the disk data format and the way of accessing the (possibly
354
# What class to use for a CommitBuilder. Often its simpler to change this
355
# in a Repository class subclass rather than to override
356
# get_commit_builder.
357
_commit_builder_class = CommitBuilder
358
# The search regex used by xml based repositories to determine what things
359
# where changed in a single commit.
360
_file_ids_altered_regex = lazy_regex.lazy_compile(
361
r'file_id="(?P<file_id>[^"]+)"'
362
r'.* revision="(?P<revision_id>[^"]+)"'
365
def abort_write_group(self):
366
"""Commit the contents accrued within the current write group.
368
:seealso: start_write_group.
370
if self._write_group is not self.get_transaction():
371
# has an unlock or relock occured ?
372
raise errors.BzrError('mismatched lock context and write group.')
373
self._abort_write_group()
374
self._write_group = None
376
def _abort_write_group(self):
377
"""Template method for per-repository write group cleanup.
379
This is called during abort before the write group is considered to be
380
finished and should cleanup any internal state accrued during the write
381
group. There is no requirement that data handed to the repository be
382
*not* made available - this is not a rollback - but neither should any
383
attempt be made to ensure that data added is fully commited. Abort is
384
invoked when an error has occured so futher disk or network operations
385
may not be possible or may error and if possible should not be
390
def add_inventory(self, revision_id, inv, parents):
391
"""Add the inventory inv to the repository as revision_id.
393
:param parents: The revision ids of the parents that revision_id
394
is known to have and are in the repository already.
396
returns the sha1 of the serialized inventory.
398
revision_id = osutils.safe_revision_id(revision_id)
399
_mod_revision.check_not_reserved_id(revision_id)
400
assert inv.revision_id is None or inv.revision_id == revision_id, \
401
"Mismatch between inventory revision" \
402
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
403
assert inv.root is not None
404
inv_lines = self._serialise_inventory_to_lines(inv)
405
inv_vf = self.get_inventory_weave()
406
return self._inventory_add_lines(inv_vf, revision_id, parents,
407
inv_lines, check_content=False)
409
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
411
"""Store lines in inv_vf and return the sha1 of the inventory."""
413
for parent in parents:
415
final_parents.append(parent)
416
return inv_vf.add_lines(revision_id, final_parents, lines,
417
check_content=check_content)[0]
420
def add_revision(self, revision_id, rev, inv=None, config=None):
421
"""Add rev to the revision store as revision_id.
423
:param revision_id: the revision id to use.
424
:param rev: The revision object.
425
:param inv: The inventory for the revision. if None, it will be looked
426
up in the inventory storer
427
:param config: If None no digital signature will be created.
428
If supplied its signature_needed method will be used
429
to determine if a signature should be made.
431
revision_id = osutils.safe_revision_id(revision_id)
432
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
434
_mod_revision.check_not_reserved_id(revision_id)
435
if config is not None and config.signature_needed():
437
inv = self.get_inventory(revision_id)
438
plaintext = Testament(rev, inv).as_short_text()
439
self.store_revision_signature(
440
gpg.GPGStrategy(config), plaintext, revision_id)
441
if not revision_id in self.get_inventory_weave():
443
raise errors.WeaveRevisionNotPresent(revision_id,
444
self.get_inventory_weave())
446
# yes, this is not suitable for adding with ghosts.
447
self.add_inventory(revision_id, inv, rev.parent_ids)
448
self._revision_store.add_revision(rev, self.get_transaction())
450
def _add_revision_text(self, revision_id, text):
451
revision = self._revision_store._serializer.read_revision_from_string(
453
self._revision_store._add_revision(revision, StringIO(text),
454
self.get_transaction())
456
def all_revision_ids(self):
457
"""Returns a list of all the revision ids in the repository.
459
This is deprecated because code should generally work on the graph
460
reachable from a particular revision, and ignore any other revisions
461
that might be present. There is no direct replacement method.
463
if 'evil' in debug.debug_flags:
464
mutter_callsite(2, "all_revision_ids is linear with history.")
465
return self._all_revision_ids()
467
def _all_revision_ids(self):
468
"""Returns a list of all the revision ids in the repository.
470
These are in as much topological order as the underlying store can
473
raise NotImplementedError(self._all_revision_ids)
475
def break_lock(self):
476
"""Break a lock if one is present from another instance.
478
Uses the ui factory to ask for confirmation if the lock may be from
481
self.control_files.break_lock()
484
def _eliminate_revisions_not_present(self, revision_ids):
485
"""Check every revision id in revision_ids to see if we have it.
487
Returns a set of the present revisions.
490
for id in revision_ids:
491
if self.has_revision(id):
496
def create(a_bzrdir):
497
"""Construct the current default format repository in a_bzrdir."""
498
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
500
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
501
"""instantiate a Repository.
503
:param _format: The format of the repository on disk.
504
:param a_bzrdir: The BzrDir of the repository.
506
In the future we will have a single api for all stores for
507
getting file texts, inventories and revisions, then
508
this construct will accept instances of those things.
510
super(Repository, self).__init__()
511
self._format = _format
512
# the following are part of the public API for Repository:
513
self.bzrdir = a_bzrdir
514
self.control_files = control_files
515
self._revision_store = _revision_store
516
# backwards compatibility
517
self.weave_store = text_store
519
self._reconcile_does_inventory_gc = True
520
# not right yet - should be more semantically clear ?
522
self.control_store = control_store
523
self.control_weaves = control_store
524
# TODO: make sure to construct the right store classes, etc, depending
525
# on whether escaping is required.
526
self._warn_if_deprecated()
527
self._write_group = None
530
return '%s(%r)' % (self.__class__.__name__,
531
self.bzrdir.transport.base)
533
def has_same_location(self, other):
534
"""Returns a boolean indicating if this repository is at the same
535
location as another repository.
537
This might return False even when two repository objects are accessing
538
the same physical repository via different URLs.
540
if self.__class__ is not other.__class__:
542
return (self.control_files._transport.base ==
543
other.control_files._transport.base)
545
def is_in_write_group(self):
546
"""Return True if there is an open write group.
548
:seealso: start_write_group.
550
return self._write_group is not None
553
return self.control_files.is_locked()
555
def lock_write(self, token=None):
556
"""Lock this repository for writing.
558
This causes caching within the repository obejct to start accumlating
559
data during reads, and allows a 'write_group' to be obtained. Write
560
groups must be used for actual data insertion.
562
:param token: if this is already locked, then lock_write will fail
563
unless the token matches the existing lock.
564
:returns: a token if this instance supports tokens, otherwise None.
565
:raises TokenLockingNotSupported: when a token is given but this
566
instance doesn't support using token locks.
567
:raises MismatchedToken: if the specified token doesn't match the token
568
of the existing lock.
569
:seealso: start_write_group.
571
A token should be passed in if you know that you have locked the object
572
some other way, and need to synchronise this object's state with that
575
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
577
result = self.control_files.lock_write(token=token)
582
self.control_files.lock_read()
585
def get_physical_lock_status(self):
586
return self.control_files.get_physical_lock_status()
588
def leave_lock_in_place(self):
589
"""Tell this repository not to release the physical lock when this
592
If lock_write doesn't return a token, then this method is not supported.
594
self.control_files.leave_in_place()
596
def dont_leave_lock_in_place(self):
597
"""Tell this repository to release the physical lock when this
598
object is unlocked, even if it didn't originally acquire it.
600
If lock_write doesn't return a token, then this method is not supported.
602
self.control_files.dont_leave_in_place()
605
def gather_stats(self, revid=None, committers=None):
606
"""Gather statistics from a revision id.
608
:param revid: The revision id to gather statistics from, if None, then
609
no revision specific statistics are gathered.
610
:param committers: Optional parameter controlling whether to grab
611
a count of committers from the revision specific statistics.
612
:return: A dictionary of statistics. Currently this contains:
613
committers: The number of committers if requested.
614
firstrev: A tuple with timestamp, timezone for the penultimate left
615
most ancestor of revid, if revid is not the NULL_REVISION.
616
latestrev: A tuple with timestamp, timezone for revid, if revid is
617
not the NULL_REVISION.
618
revisions: The total revision count in the repository.
619
size: An estimate disk size of the repository in bytes.
622
if revid and committers:
623
result['committers'] = 0
624
if revid and revid != _mod_revision.NULL_REVISION:
626
all_committers = set()
627
revisions = self.get_ancestry(revid)
628
# pop the leading None
630
first_revision = None
632
# ignore the revisions in the middle - just grab first and last
633
revisions = revisions[0], revisions[-1]
634
for revision in self.get_revisions(revisions):
635
if not first_revision:
636
first_revision = revision
638
all_committers.add(revision.committer)
639
last_revision = revision
641
result['committers'] = len(all_committers)
642
result['firstrev'] = (first_revision.timestamp,
643
first_revision.timezone)
644
result['latestrev'] = (last_revision.timestamp,
645
last_revision.timezone)
647
# now gather global repository information
648
if self.bzrdir.root_transport.listable():
649
c, t = self._revision_store.total_size(self.get_transaction())
650
result['revisions'] = c
655
def missing_revision_ids(self, other, revision_id=None):
656
"""Return the revision ids that other has that this does not.
658
These are returned in topological order.
660
revision_id: only return revision ids included by revision_id.
662
revision_id = osutils.safe_revision_id(revision_id)
663
return InterRepository.get(other, self).missing_revision_ids(revision_id)
667
"""Open the repository rooted at base.
669
For instance, if the repository is at URL/.bzr/repository,
670
Repository.open(URL) -> a Repository instance.
672
control = bzrdir.BzrDir.open(base)
673
return control.open_repository()
675
def copy_content_into(self, destination, revision_id=None):
676
"""Make a complete copy of the content in self into destination.
678
This is a destructive operation! Do not use it on existing
681
revision_id = osutils.safe_revision_id(revision_id)
682
return InterRepository.get(self, destination).copy_content(revision_id)
684
def commit_write_group(self):
685
"""Commit the contents accrued within the current write group.
687
:seealso: start_write_group.
689
if self._write_group is not self.get_transaction():
690
# has an unlock or relock occured ?
691
raise errors.BzrError('mismatched lock context and write group.')
692
self._commit_write_group()
693
self._write_group = None
695
def _commit_write_group(self):
696
"""Template method for per-repository write group cleanup.
698
This is called before the write group is considered to be
699
finished and should ensure that all data handed to the repository
700
for writing during the write group is safely committed (to the
701
extent possible considering file system caching etc).
704
def fetch(self, source, revision_id=None, pb=None):
705
"""Fetch the content required to construct revision_id from source.
707
If revision_id is None all content is copied.
709
revision_id = osutils.safe_revision_id(revision_id)
710
inter = InterRepository.get(source, self)
712
return inter.fetch(revision_id=revision_id, pb=pb)
713
except NotImplementedError:
714
raise errors.IncompatibleRepositories(source, self)
716
def create_bundle(self, target, base, fileobj, format=None):
717
return serializer.write_bundle(self, target, base, fileobj, format)
719
def get_commit_builder(self, branch, parents, config, timestamp=None,
720
timezone=None, committer=None, revprops=None,
722
"""Obtain a CommitBuilder for this repository.
724
:param branch: Branch to commit to.
725
:param parents: Revision ids of the parents of the new revision.
726
:param config: Configuration to use.
727
:param timestamp: Optional timestamp recorded for commit.
728
:param timezone: Optional timezone for timestamp.
729
:param committer: Optional committer to set for commit.
730
:param revprops: Optional dictionary of revision properties.
731
:param revision_id: Optional revision id.
733
revision_id = osutils.safe_revision_id(revision_id)
734
result = self._commit_builder_class(self, parents, config,
735
timestamp, timezone, committer, revprops, revision_id)
736
self.start_write_group()
740
if (self.control_files._lock_count == 1 and
741
self.control_files._lock_mode == 'w'):
742
if self._write_group is not None:
743
raise errors.BzrError(
744
'Must end write groups before releasing write locks.')
745
self.control_files.unlock()
748
def clone(self, a_bzrdir, revision_id=None):
749
"""Clone this repository into a_bzrdir using the current format.
751
Currently no check is made that the format of this repository and
752
the bzrdir format are compatible. FIXME RBC 20060201.
754
:return: The newly created destination repository.
756
# TODO: deprecate after 0.16; cloning this with all its settings is
757
# probably not very useful -- mbp 20070423
758
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
759
self.copy_content_into(dest_repo, revision_id)
762
def start_write_group(self):
763
"""Start a write group in the repository.
765
Write groups are used by repositories which do not have a 1:1 mapping
766
between file ids and backend store to manage the insertion of data from
767
both fetch and commit operations.
769
A write lock is required around the start_write_group/commit_write_group
770
for the support of lock-requiring repository formats.
772
One can only insert data into a repository inside a write group.
776
if not self.is_locked() or self.control_files._lock_mode != 'w':
777
raise errors.NotWriteLocked(self)
778
if self._write_group:
779
raise errors.BzrError('already in a write group')
780
self._start_write_group()
781
# so we can detect unlock/relock - the write group is now entered.
782
self._write_group = self.get_transaction()
784
def _start_write_group(self):
785
"""Template method for per-repository write group startup.
787
This is called before the write group is considered to be
792
def sprout(self, to_bzrdir, revision_id=None):
793
"""Create a descendent repository for new development.
795
Unlike clone, this does not copy the settings of the repository.
797
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
798
dest_repo.fetch(self, revision_id=revision_id)
801
def _create_sprouting_repo(self, a_bzrdir, shared):
802
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
803
# use target default format.
804
dest_repo = a_bzrdir.create_repository()
806
# Most control formats need the repository to be specifically
807
# created, but on some old all-in-one formats it's not needed
809
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
810
except errors.UninitializableFormat:
811
dest_repo = a_bzrdir.open_repository()
815
def has_revision(self, revision_id):
816
"""True if this repository has a copy of the revision."""
817
if 'evil' in debug.debug_flags:
818
mutter_callsite(2, "has_revision is a LBYL symptom.")
819
revision_id = osutils.safe_revision_id(revision_id)
820
return self._revision_store.has_revision_id(revision_id,
821
self.get_transaction())
824
def get_revision(self, revision_id):
825
"""Return the Revision object for a named revision."""
826
return self.get_revisions([revision_id])[0]
829
def get_revision_reconcile(self, revision_id):
830
"""'reconcile' helper routine that allows access to a revision always.
832
This variant of get_revision does not cross check the weave graph
833
against the revision one as get_revision does: but it should only
834
be used by reconcile, or reconcile-alike commands that are correcting
835
or testing the revision graph.
837
return self._get_revisions([revision_id])[0]
840
def get_revisions(self, revision_ids):
841
"""Get many revisions at once."""
842
return self._get_revisions(revision_ids)
845
def _get_revisions(self, revision_ids):
846
"""Core work logic to get many revisions without sanity checks."""
847
revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
848
for rev_id in revision_ids:
849
if not rev_id or not isinstance(rev_id, basestring):
850
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
851
revs = self._revision_store.get_revisions(revision_ids,
852
self.get_transaction())
854
assert not isinstance(rev.revision_id, unicode)
855
for parent_id in rev.parent_ids:
856
assert not isinstance(parent_id, unicode)
860
def get_revision_xml(self, revision_id):
861
# TODO: jam 20070210 This shouldn't be necessary since get_revision
862
# would have already do it.
863
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
864
revision_id = osutils.safe_revision_id(revision_id)
865
rev = self.get_revision(revision_id)
867
# the current serializer..
868
self._revision_store._serializer.write_revision(rev, rev_tmp)
870
return rev_tmp.getvalue()
873
def get_deltas_for_revisions(self, revisions):
874
"""Produce a generator of revision deltas.
876
Note that the input is a sequence of REVISIONS, not revision_ids.
877
Trees will be held in memory until the generator exits.
878
Each delta is relative to the revision's lefthand predecessor.
880
required_trees = set()
881
for revision in revisions:
882
required_trees.add(revision.revision_id)
883
required_trees.update(revision.parent_ids[:1])
884
trees = dict((t.get_revision_id(), t) for
885
t in self.revision_trees(required_trees))
886
for revision in revisions:
887
if not revision.parent_ids:
888
old_tree = self.revision_tree(None)
890
old_tree = trees[revision.parent_ids[0]]
891
yield trees[revision.revision_id].changes_from(old_tree)
894
def get_revision_delta(self, revision_id):
895
"""Return the delta for one revision.
897
The delta is relative to the left-hand predecessor of the
900
r = self.get_revision(revision_id)
901
return list(self.get_deltas_for_revisions([r]))[0]
904
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
905
revision_id = osutils.safe_revision_id(revision_id)
906
signature = gpg_strategy.sign(plaintext)
907
self._revision_store.add_revision_signature_text(revision_id,
909
self.get_transaction())
911
def fileids_altered_by_revision_ids(self, revision_ids):
912
"""Find the file ids and versions affected by revisions.
914
:param revisions: an iterable containing revision ids.
915
:return: a dictionary mapping altered file-ids to an iterable of
916
revision_ids. Each altered file-ids has the exact revision_ids that
917
altered it listed explicitly.
919
assert self._serializer.support_altered_by_hack, \
920
("fileids_altered_by_revision_ids only supported for branches "
921
"which store inventory as unnested xml, not on %r" % self)
922
selected_revision_ids = set(osutils.safe_revision_id(r)
923
for r in revision_ids)
924
w = self.get_inventory_weave()
927
# this code needs to read every new line in every inventory for the
928
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
929
# not present in one of those inventories is unnecessary but not
930
# harmful because we are filtering by the revision id marker in the
931
# inventory lines : we only select file ids altered in one of those
932
# revisions. We don't need to see all lines in the inventory because
933
# only those added in an inventory in rev X can contain a revision=X
935
unescape_revid_cache = {}
936
unescape_fileid_cache = {}
938
# jam 20061218 In a big fetch, this handles hundreds of thousands
939
# of lines, so it has had a lot of inlining and optimizing done.
940
# Sorry that it is a little bit messy.
941
# Move several functions to be local variables, since this is a long
943
search = self._file_ids_altered_regex.search
944
unescape = _unescape_xml
945
setdefault = result.setdefault
946
pb = ui.ui_factory.nested_progress_bar()
948
for line in w.iter_lines_added_or_present_in_versions(
949
selected_revision_ids, pb=pb):
953
# One call to match.group() returning multiple items is quite a
954
# bit faster than 2 calls to match.group() each returning 1
955
file_id, revision_id = match.group('file_id', 'revision_id')
957
# Inlining the cache lookups helps a lot when you make 170,000
958
# lines and 350k ids, versus 8.4 unique ids.
959
# Using a cache helps in 2 ways:
960
# 1) Avoids unnecessary decoding calls
961
# 2) Re-uses cached strings, which helps in future set and
963
# (2) is enough that removing encoding entirely along with
964
# the cache (so we are using plain strings) results in no
965
# performance improvement.
967
revision_id = unescape_revid_cache[revision_id]
969
unescaped = unescape(revision_id)
970
unescape_revid_cache[revision_id] = unescaped
971
revision_id = unescaped
973
if revision_id in selected_revision_ids:
975
file_id = unescape_fileid_cache[file_id]
977
unescaped = unescape(file_id)
978
unescape_fileid_cache[file_id] = unescaped
980
setdefault(file_id, set()).add(revision_id)
985
def iter_files_bytes(self, desired_files):
986
"""Iterate through file versions.
988
Files will not necessarily be returned in the order they occur in
989
desired_files. No specific order is guaranteed.
991
Yields pairs of identifier, bytes_iterator. identifier is an opaque
992
value supplied by the caller as part of desired_files. It should
993
uniquely identify the file version in the caller's context. (Examples:
994
an index number or a TreeTransform trans_id.)
996
bytes_iterator is an iterable of bytestrings for the file. The
997
kind of iterable and length of the bytestrings are unspecified, but for
998
this implementation, it is a list of lines produced by
999
VersionedFile.get_lines().
1001
:param desired_files: a list of (file_id, revision_id, identifier)
1004
transaction = self.get_transaction()
1005
for file_id, revision_id, callable_data in desired_files:
1007
weave = self.weave_store.get_weave(file_id, transaction)
1008
except errors.NoSuchFile:
1009
raise errors.NoSuchIdInRepository(self, file_id)
1010
yield callable_data, weave.get_lines(revision_id)
1012
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1013
"""Get an iterable listing the keys of all the data introduced by a set
1016
The keys will be ordered so that the corresponding items can be safely
1017
fetched and inserted in that order.
1019
:returns: An iterable producing tuples of (knit-kind, file-id,
1020
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1021
'revisions'. file-id is None unless knit-kind is 'file'.
1023
# XXX: it's a bit weird to control the inventory weave caching in this
1024
# generator. Ideally the caching would be done in fetch.py I think. Or
1025
# maybe this generator should explicitly have the contract that it
1026
# should not be iterated until the previously yielded item has been
1028
inv_w = self.get_inventory_weave()
1029
inv_w.enable_cache()
1031
# file ids that changed
1032
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1034
num_file_ids = len(file_ids)
1035
for file_id, altered_versions in file_ids.iteritems():
1036
if _files_pb is not None:
1037
_files_pb.update("fetch texts", count, num_file_ids)
1039
yield ("file", file_id, altered_versions)
1040
# We're done with the files_pb. Note that it finished by the caller,
1041
# just as it was created by the caller.
1045
yield ("inventory", None, revision_ids)
1049
revisions_with_signatures = set()
1050
for rev_id in revision_ids:
1052
self.get_signature_text(rev_id)
1053
except errors.NoSuchRevision:
1057
revisions_with_signatures.add(rev_id)
1058
yield ("signatures", None, revisions_with_signatures)
1061
yield ("revisions", None, revision_ids)
1064
def get_inventory_weave(self):
1065
return self.control_weaves.get_weave('inventory',
1066
self.get_transaction())
1069
def get_inventory(self, revision_id):
1070
"""Get Inventory object by hash."""
1071
# TODO: jam 20070210 Technically we don't need to sanitize, since all
1072
# called functions must sanitize.
1073
revision_id = osutils.safe_revision_id(revision_id)
1074
return self.deserialise_inventory(
1075
revision_id, self.get_inventory_xml(revision_id))
1077
def deserialise_inventory(self, revision_id, xml):
1078
"""Transform the xml into an inventory object.
1080
:param revision_id: The expected revision id of the inventory.
1081
:param xml: A serialised inventory.
1083
revision_id = osutils.safe_revision_id(revision_id)
1084
result = self._serializer.read_inventory_from_string(xml)
1085
result.root.revision = revision_id
1088
def serialise_inventory(self, inv):
1089
return self._serializer.write_inventory_to_string(inv)
1091
def _serialise_inventory_to_lines(self, inv):
1092
return self._serializer.write_inventory_to_lines(inv)
1094
def get_serializer_format(self):
1095
return self._serializer.format_num
1098
def get_inventory_xml(self, revision_id):
1099
"""Get inventory XML as a file object."""
1100
revision_id = osutils.safe_revision_id(revision_id)
1102
assert isinstance(revision_id, str), type(revision_id)
1103
iw = self.get_inventory_weave()
1104
return iw.get_text(revision_id)
1106
raise errors.HistoryMissing(self, 'inventory', revision_id)
1109
def get_inventory_sha1(self, revision_id):
1110
"""Return the sha1 hash of the inventory entry
1112
# TODO: jam 20070210 Shouldn't this be deprecated / removed?
1113
revision_id = osutils.safe_revision_id(revision_id)
1114
return self.get_revision(revision_id).inventory_sha1
1117
def get_revision_graph(self, revision_id=None):
1118
"""Return a dictionary containing the revision graph.
1120
NB: This method should not be used as it accesses the entire graph all
1121
at once, which is much more data than most operations should require.
1123
:param revision_id: The revision_id to get a graph from. If None, then
1124
the entire revision graph is returned. This is a deprecated mode of
1125
operation and will be removed in the future.
1126
:return: a dictionary of revision_id->revision_parents_list.
1128
raise NotImplementedError(self.get_revision_graph)
1131
def get_revision_graph_with_ghosts(self, revision_ids=None):
1132
"""Return a graph of the revisions with ghosts marked as applicable.
1134
:param revision_ids: an iterable of revisions to graph or None for all.
1135
:return: a Graph object with the graph reachable from revision_ids.
1137
if 'evil' in debug.debug_flags:
1139
"get_revision_graph_with_ghosts scales with size of history.")
1140
result = deprecated_graph.Graph()
1141
if not revision_ids:
1142
pending = set(self.all_revision_ids())
1145
pending = set(osutils.safe_revision_id(r) for r in revision_ids)
1146
# special case NULL_REVISION
1147
if _mod_revision.NULL_REVISION in pending:
1148
pending.remove(_mod_revision.NULL_REVISION)
1149
required = set(pending)
1152
revision_id = pending.pop()
1154
rev = self.get_revision(revision_id)
1155
except errors.NoSuchRevision:
1156
if revision_id in required:
1159
result.add_ghost(revision_id)
1161
for parent_id in rev.parent_ids:
1162
# is this queued or done ?
1163
if (parent_id not in pending and
1164
parent_id not in done):
1166
pending.add(parent_id)
1167
result.add_node(revision_id, rev.parent_ids)
1168
done.add(revision_id)
1171
def _get_history_vf(self):
1172
"""Get a versionedfile whose history graph reflects all revisions.
1174
For weave repositories, this is the inventory weave.
1176
return self.get_inventory_weave()
1178
def iter_reverse_revision_history(self, revision_id):
1179
"""Iterate backwards through revision ids in the lefthand history
1181
:param revision_id: The revision id to start with. All its lefthand
1182
ancestors will be traversed.
1184
revision_id = osutils.safe_revision_id(revision_id)
1185
if revision_id in (None, _mod_revision.NULL_REVISION):
1187
next_id = revision_id
1188
versionedfile = self._get_history_vf()
1191
parents = versionedfile.get_parents(next_id)
1192
if len(parents) == 0:
1195
next_id = parents[0]
1198
def get_revision_inventory(self, revision_id):
1199
"""Return inventory of a past revision."""
1200
# TODO: Unify this with get_inventory()
1201
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1202
# must be the same as its revision, so this is trivial.
1203
if revision_id is None:
1204
# This does not make sense: if there is no revision,
1205
# then it is the current tree inventory surely ?!
1206
# and thus get_root_id() is something that looks at the last
1207
# commit on the branch, and the get_root_id is an inventory check.
1208
raise NotImplementedError
1209
# return Inventory(self.get_root_id())
1211
return self.get_inventory(revision_id)
1214
def is_shared(self):
1215
"""Return True if this repository is flagged as a shared repository."""
1216
raise NotImplementedError(self.is_shared)
1219
def reconcile(self, other=None, thorough=False):
1220
"""Reconcile this repository."""
1221
from bzrlib.reconcile import RepoReconciler
1222
reconciler = RepoReconciler(self, thorough=thorough)
1223
reconciler.reconcile()
1226
def _refresh_data(self):
1227
"""Helper called from lock_* to ensure coherency with disk.
1229
The default implementation does nothing; it is however possible
1230
for repositories to maintain loaded indices across multiple locks
1231
by checking inside their implementation of this method to see
1232
whether their indices are still valid. This depends of course on
1233
the disk format being validatable in this manner.
1237
def revision_tree(self, revision_id):
1238
"""Return Tree for a revision on this branch.
1240
`revision_id` may be None for the empty tree revision.
1242
# TODO: refactor this to use an existing revision object
1243
# so we don't need to read it in twice.
1244
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1245
return RevisionTree(self, Inventory(root_id=None),
1246
_mod_revision.NULL_REVISION)
1248
revision_id = osutils.safe_revision_id(revision_id)
1249
inv = self.get_revision_inventory(revision_id)
1250
return RevisionTree(self, inv, revision_id)
1253
def revision_trees(self, revision_ids):
1254
"""Return Tree for a revision on this branch.
1256
`revision_id` may not be None or 'null:'"""
1257
assert None not in revision_ids
1258
assert _mod_revision.NULL_REVISION not in revision_ids
1259
texts = self.get_inventory_weave().get_texts(revision_ids)
1260
for text, revision_id in zip(texts, revision_ids):
1261
inv = self.deserialise_inventory(revision_id, text)
1262
yield RevisionTree(self, inv, revision_id)
1265
def get_ancestry(self, revision_id, topo_sorted=True):
1266
"""Return a list of revision-ids integrated by a revision.
1268
The first element of the list is always None, indicating the origin
1269
revision. This might change when we have history horizons, or
1270
perhaps we should have a new API.
1272
This is topologically sorted.
1274
if _mod_revision.is_null(revision_id):
1276
revision_id = osutils.safe_revision_id(revision_id)
1277
if not self.has_revision(revision_id):
1278
raise errors.NoSuchRevision(self, revision_id)
1279
w = self.get_inventory_weave()
1280
candidates = w.get_ancestry(revision_id, topo_sorted)
1281
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1284
"""Compress the data within the repository.
1286
This operation only makes sense for some repository types. For other
1287
types it should be a no-op that just returns.
1289
This stub method does not require a lock, but subclasses should use
1290
@needs_write_lock as this is a long running call its reasonable to
1291
implicitly lock for the user.
1295
def print_file(self, file, revision_id):
1296
"""Print `file` to stdout.
1298
FIXME RBC 20060125 as John Meinel points out this is a bad api
1299
- it writes to stdout, it assumes that that is valid etc. Fix
1300
by creating a new more flexible convenience function.
1302
revision_id = osutils.safe_revision_id(revision_id)
1303
tree = self.revision_tree(revision_id)
1304
# use inventory as it was in that revision
1305
file_id = tree.inventory.path2id(file)
1307
# TODO: jam 20060427 Write a test for this code path
1308
# it had a bug in it, and was raising the wrong
1310
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1311
tree.print_file(file_id)
1313
def get_transaction(self):
1314
return self.control_files.get_transaction()
1316
def revision_parents(self, revision_id):
1317
revision_id = osutils.safe_revision_id(revision_id)
1318
return self.get_inventory_weave().parent_names(revision_id)
1320
def get_parents(self, revision_ids):
1321
"""See StackedParentsProvider.get_parents"""
1323
for revision_id in revision_ids:
1324
if revision_id == _mod_revision.NULL_REVISION:
1328
parents = self.get_revision(revision_id).parent_ids
1329
except errors.NoSuchRevision:
1332
if len(parents) == 0:
1333
parents = [_mod_revision.NULL_REVISION]
1334
parents_list.append(parents)
1337
def _make_parents_provider(self):
1340
def get_graph(self, other_repository=None):
1341
"""Return the graph walker for this repository format"""
1342
parents_provider = self._make_parents_provider()
1343
if (other_repository is not None and
1344
other_repository.bzrdir.transport.base !=
1345
self.bzrdir.transport.base):
1346
parents_provider = graph._StackedParentsProvider(
1347
[parents_provider, other_repository._make_parents_provider()])
1348
return graph.Graph(parents_provider)
1351
def set_make_working_trees(self, new_value):
1352
"""Set the policy flag for making working trees when creating branches.
1354
This only applies to branches that use this repository.
1356
The default is 'True'.
1357
:param new_value: True to restore the default, False to disable making
1360
raise NotImplementedError(self.set_make_working_trees)
1362
def make_working_trees(self):
1363
"""Returns the policy for making working trees on new branches."""
1364
raise NotImplementedError(self.make_working_trees)
1367
def sign_revision(self, revision_id, gpg_strategy):
1368
revision_id = osutils.safe_revision_id(revision_id)
1369
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1370
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1373
def has_signature_for_revision_id(self, revision_id):
1374
"""Query for a revision signature for revision_id in the repository."""
1375
revision_id = osutils.safe_revision_id(revision_id)
1376
return self._revision_store.has_signature(revision_id,
1377
self.get_transaction())
1380
def get_signature_text(self, revision_id):
1381
"""Return the text for a signature."""
1382
revision_id = osutils.safe_revision_id(revision_id)
1383
return self._revision_store.get_signature_text(revision_id,
1384
self.get_transaction())
1387
def check(self, revision_ids):
1388
"""Check consistency of all history of given revision_ids.
1390
Different repository implementations should override _check().
1392
:param revision_ids: A non-empty list of revision_ids whose ancestry
1393
will be checked. Typically the last revision_id of a branch.
1395
if not revision_ids:
1396
raise ValueError("revision_ids must be non-empty in %s.check"
1398
revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1399
return self._check(revision_ids)
1401
def _check(self, revision_ids):
1402
result = check.Check(self)
1406
def _warn_if_deprecated(self):
1407
global _deprecation_warning_done
1408
if _deprecation_warning_done:
1410
_deprecation_warning_done = True
1411
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1412
% (self._format, self.bzrdir.transport.base))
1414
def supports_rich_root(self):
1415
return self._format.rich_root_data
1417
def _check_ascii_revisionid(self, revision_id, method):
1418
"""Private helper for ascii-only repositories."""
1419
# weave repositories refuse to store revisionids that are non-ascii.
1420
if revision_id is not None:
1421
# weaves require ascii revision ids.
1422
if isinstance(revision_id, unicode):
1424
revision_id.encode('ascii')
1425
except UnicodeEncodeError:
1426
raise errors.NonAsciiRevisionId(method, self)
1429
revision_id.decode('ascii')
1430
except UnicodeDecodeError:
1431
raise errors.NonAsciiRevisionId(method, self)
1435
# remove these delegates a while after bzr 0.15
1436
def __make_delegated(name, from_module):
1437
def _deprecated_repository_forwarder():
1438
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1439
% (name, from_module),
1442
m = __import__(from_module, globals(), locals(), [name])
1444
return getattr(m, name)
1445
except AttributeError:
1446
raise AttributeError('module %s has no name %s'
1448
globals()[name] = _deprecated_repository_forwarder
1451
'AllInOneRepository',
1452
'WeaveMetaDirRepository',
1453
'PreSplitOutRepositoryFormat',
1454
'RepositoryFormat4',
1455
'RepositoryFormat5',
1456
'RepositoryFormat6',
1457
'RepositoryFormat7',
1459
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1463
'RepositoryFormatKnit',
1464
'RepositoryFormatKnit1',
1466
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1469
def install_revision(repository, rev, revision_tree):
1470
"""Install all revision data into a repository."""
1471
present_parents = []
1473
for p_id in rev.parent_ids:
1474
if repository.has_revision(p_id):
1475
present_parents.append(p_id)
1476
parent_trees[p_id] = repository.revision_tree(p_id)
1478
parent_trees[p_id] = repository.revision_tree(None)
1480
inv = revision_tree.inventory
1481
entries = inv.iter_entries()
1482
# backwards compatibility hack: skip the root id.
1483
if not repository.supports_rich_root():
1484
path, root = entries.next()
1485
if root.revision != rev.revision_id:
1486
raise errors.IncompatibleRevision(repr(repository))
1487
# Add the texts that are not already present
1488
for path, ie in entries:
1489
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1490
repository.get_transaction())
1491
if ie.revision not in w:
1493
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1494
# with InventoryEntry.find_previous_heads(). if it is, then there
1495
# is a latent bug here where the parents may have ancestors of each
1497
for revision, tree in parent_trees.iteritems():
1498
if ie.file_id not in tree:
1500
parent_id = tree.inventory[ie.file_id].revision
1501
if parent_id in text_parents:
1503
text_parents.append(parent_id)
1505
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1506
repository.get_transaction())
1507
lines = revision_tree.get_file(ie.file_id).readlines()
1508
vfile.add_lines(rev.revision_id, text_parents, lines)
1510
# install the inventory
1511
repository.add_inventory(rev.revision_id, inv, present_parents)
1512
except errors.RevisionAlreadyPresent:
1514
repository.add_revision(rev.revision_id, rev, inv)
1517
class MetaDirRepository(Repository):
1518
"""Repositories in the new meta-dir layout."""
1520
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1521
super(MetaDirRepository, self).__init__(_format,
1527
dir_mode = self.control_files._dir_mode
1528
file_mode = self.control_files._file_mode
1531
def is_shared(self):
1532
"""Return True if this repository is flagged as a shared repository."""
1533
return self.control_files._transport.has('shared-storage')
1536
def set_make_working_trees(self, new_value):
1537
"""Set the policy flag for making working trees when creating branches.
1539
This only applies to branches that use this repository.
1541
The default is 'True'.
1542
:param new_value: True to restore the default, False to disable making
1547
self.control_files._transport.delete('no-working-trees')
1548
except errors.NoSuchFile:
1551
self.control_files.put_utf8('no-working-trees', '')
1553
def make_working_trees(self):
1554
"""Returns the policy for making working trees on new branches."""
1555
return not self.control_files._transport.has('no-working-trees')
1558
class RepositoryFormatRegistry(registry.Registry):
1559
"""Registry of RepositoryFormats.
1562
def get(self, format_string):
1563
r = registry.Registry.get(self, format_string)
1569
format_registry = RepositoryFormatRegistry()
1570
"""Registry of formats, indexed by their identifying format string.
1572
This can contain either format instances themselves, or classes/factories that
1573
can be called to obtain one.
1577
#####################################################################
1578
# Repository Formats
1580
class RepositoryFormat(object):
1581
"""A repository format.
1583
Formats provide three things:
1584
* An initialization routine to construct repository data on disk.
1585
* a format string which is used when the BzrDir supports versioned
1587
* an open routine which returns a Repository instance.
1589
Formats are placed in an dict by their format string for reference
1590
during opening. These should be subclasses of RepositoryFormat
1593
Once a format is deprecated, just deprecate the initialize and open
1594
methods on the format class. Do not deprecate the object, as the
1595
object will be created every system load.
1597
Common instance attributes:
1598
_matchingbzrdir - the bzrdir format that the repository format was
1599
originally written to work with. This can be used if manually
1600
constructing a bzrdir and repository, or more commonly for test suite
1605
return "<%s>" % self.__class__.__name__
1607
def __eq__(self, other):
1608
# format objects are generally stateless
1609
return isinstance(other, self.__class__)
1611
def __ne__(self, other):
1612
return not self == other
1615
def find_format(klass, a_bzrdir):
1616
"""Return the format for the repository object in a_bzrdir.
1618
This is used by bzr native formats that have a "format" file in
1619
the repository. Other methods may be used by different types of
1623
transport = a_bzrdir.get_repository_transport(None)
1624
format_string = transport.get("format").read()
1625
return format_registry.get(format_string)
1626
except errors.NoSuchFile:
1627
raise errors.NoRepositoryPresent(a_bzrdir)
1629
raise errors.UnknownFormatError(format=format_string)
1632
def register_format(klass, format):
1633
format_registry.register(format.get_format_string(), format)
1636
def unregister_format(klass, format):
1637
format_registry.remove(format.get_format_string())
1640
def get_default_format(klass):
1641
"""Return the current default format."""
1642
from bzrlib import bzrdir
1643
return bzrdir.format_registry.make_bzrdir('default').repository_format
1645
def _get_control_store(self, repo_transport, control_files):
1646
"""Return the control store for this repository."""
1647
raise NotImplementedError(self._get_control_store)
1649
def get_format_string(self):
1650
"""Return the ASCII format string that identifies this format.
1652
Note that in pre format ?? repositories the format string is
1653
not permitted nor written to disk.
1655
raise NotImplementedError(self.get_format_string)
1657
def get_format_description(self):
1658
"""Return the short description for this format."""
1659
raise NotImplementedError(self.get_format_description)
1661
def _get_revision_store(self, repo_transport, control_files):
1662
"""Return the revision store object for this a_bzrdir."""
1663
raise NotImplementedError(self._get_revision_store)
1665
def _get_text_rev_store(self,
1672
"""Common logic for getting a revision store for a repository.
1674
see self._get_revision_store for the subclass-overridable method to
1675
get the store for a repository.
1677
from bzrlib.store.revision.text import TextRevisionStore
1678
dir_mode = control_files._dir_mode
1679
file_mode = control_files._file_mode
1680
text_store = TextStore(transport.clone(name),
1682
compressed=compressed,
1684
file_mode=file_mode)
1685
_revision_store = TextRevisionStore(text_store, serializer)
1686
return _revision_store
1688
# TODO: this shouldn't be in the base class, it's specific to things that
1689
# use weaves or knits -- mbp 20070207
1690
def _get_versioned_file_store(self,
1695
versionedfile_class=None,
1696
versionedfile_kwargs={},
1698
if versionedfile_class is None:
1699
versionedfile_class = self._versionedfile_class
1700
weave_transport = control_files._transport.clone(name)
1701
dir_mode = control_files._dir_mode
1702
file_mode = control_files._file_mode
1703
return VersionedFileStore(weave_transport, prefixed=prefixed,
1705
file_mode=file_mode,
1706
versionedfile_class=versionedfile_class,
1707
versionedfile_kwargs=versionedfile_kwargs,
1710
def initialize(self, a_bzrdir, shared=False):
1711
"""Initialize a repository of this format in a_bzrdir.
1713
:param a_bzrdir: The bzrdir to put the new repository in it.
1714
:param shared: The repository should be initialized as a sharable one.
1715
:returns: The new repository object.
1717
This may raise UninitializableFormat if shared repository are not
1718
compatible the a_bzrdir.
1720
raise NotImplementedError(self.initialize)
1722
def is_supported(self):
1723
"""Is this format supported?
1725
Supported formats must be initializable and openable.
1726
Unsupported formats may not support initialization or committing or
1727
some other features depending on the reason for not being supported.
1731
def check_conversion_target(self, target_format):
1732
raise NotImplementedError(self.check_conversion_target)
1734
def open(self, a_bzrdir, _found=False):
1735
"""Return an instance of this format for the bzrdir a_bzrdir.
1737
_found is a private parameter, do not use it.
1739
raise NotImplementedError(self.open)
1742
class MetaDirRepositoryFormat(RepositoryFormat):
1743
"""Common base class for the new repositories using the metadir layout."""
1745
rich_root_data = False
1746
supports_tree_reference = False
1747
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
1750
super(MetaDirRepositoryFormat, self).__init__()
1752
def _create_control_files(self, a_bzrdir):
1753
"""Create the required files and the initial control_files object."""
1754
# FIXME: RBC 20060125 don't peek under the covers
1755
# NB: no need to escape relative paths that are url safe.
1756
repository_transport = a_bzrdir.get_repository_transport(self)
1757
control_files = lockable_files.LockableFiles(repository_transport,
1758
'lock', lockdir.LockDir)
1759
control_files.create_lock()
1760
return control_files
1762
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
1763
"""Upload the initial blank content."""
1764
control_files = self._create_control_files(a_bzrdir)
1765
control_files.lock_write()
1767
control_files._transport.mkdir_multi(dirs,
1768
mode=control_files._dir_mode)
1769
for file, content in files:
1770
control_files.put(file, content)
1771
for file, content in utf8_files:
1772
control_files.put_utf8(file, content)
1774
control_files.put_utf8('shared-storage', '')
1776
control_files.unlock()
1779
# formats which have no format string are not discoverable
1780
# and not independently creatable, so are not registered. They're
1781
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1782
# needed, it's constructed directly by the BzrDir. Non-native formats where
1783
# the repository is not separately opened are similar.
1785
format_registry.register_lazy(
1786
'Bazaar-NG Repository format 7',
1787
'bzrlib.repofmt.weaverepo',
1790
# KEEP in sync with bzrdir.format_registry default, which controls the overall
1791
# default control directory format
1793
format_registry.register_lazy(
1794
'Bazaar-NG Knit Repository Format 1',
1795
'bzrlib.repofmt.knitrepo',
1796
'RepositoryFormatKnit1',
1798
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
1800
format_registry.register_lazy(
1801
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
1802
'bzrlib.repofmt.knitrepo',
1803
'RepositoryFormatKnit3',
1807
class InterRepository(InterObject):
1808
"""This class represents operations taking place between two repositories.
1810
Its instances have methods like copy_content and fetch, and contain
1811
references to the source and target repositories these operations can be
1814
Often we will provide convenience methods on 'repository' which carry out
1815
operations with another repository - they will always forward to
1816
InterRepository.get(other).method_name(parameters).
1820
"""The available optimised InterRepository types."""
1822
def copy_content(self, revision_id=None):
1823
raise NotImplementedError(self.copy_content)
1825
def fetch(self, revision_id=None, pb=None):
1826
"""Fetch the content required to construct revision_id.
1828
The content is copied from self.source to self.target.
1830
:param revision_id: if None all content is copied, if NULL_REVISION no
1832
:param pb: optional progress bar to use for progress reports. If not
1833
provided a default one will be created.
1835
Returns the copied revision count and the failed revisions in a tuple:
1838
raise NotImplementedError(self.fetch)
1841
def missing_revision_ids(self, revision_id=None):
1842
"""Return the revision ids that source has that target does not.
1844
These are returned in topological order.
1846
:param revision_id: only return revision ids included by this
1849
# generic, possibly worst case, slow code path.
1850
target_ids = set(self.target.all_revision_ids())
1851
if revision_id is not None:
1852
# TODO: jam 20070210 InterRepository is internal enough that it
1853
# should assume revision_ids are already utf-8
1854
revision_id = osutils.safe_revision_id(revision_id)
1855
source_ids = self.source.get_ancestry(revision_id)
1856
assert source_ids[0] is None
1859
source_ids = self.source.all_revision_ids()
1860
result_set = set(source_ids).difference(target_ids)
1861
# this may look like a no-op: its not. It preserves the ordering
1862
# other_ids had while only returning the members from other_ids
1863
# that we've decided we need.
1864
return [rev_id for rev_id in source_ids if rev_id in result_set]
1867
class InterSameDataRepository(InterRepository):
1868
"""Code for converting between repositories that represent the same data.
1870
Data format and model must match for this to work.
1874
def _get_repo_format_to_test(self):
1875
"""Repository format for testing with.
1877
InterSameData can pull from subtree to subtree and from non-subtree to
1878
non-subtree, so we test this with the richest repository format.
1880
from bzrlib.repofmt import knitrepo
1881
return knitrepo.RepositoryFormatKnit3()
1884
def is_compatible(source, target):
1885
if source.supports_rich_root() != target.supports_rich_root():
1887
if source._serializer != target._serializer:
1892
def copy_content(self, revision_id=None):
1893
"""Make a complete copy of the content in self into destination.
1895
This copies both the repository's revision data, and configuration information
1896
such as the make_working_trees setting.
1898
This is a destructive operation! Do not use it on existing
1901
:param revision_id: Only copy the content needed to construct
1902
revision_id and its parents.
1905
self.target.set_make_working_trees(self.source.make_working_trees())
1906
except NotImplementedError:
1908
# TODO: jam 20070210 This is fairly internal, so we should probably
1909
# just assert that revision_id is not unicode.
1910
revision_id = osutils.safe_revision_id(revision_id)
1911
# but don't bother fetching if we have the needed data now.
1912
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
1913
self.target.has_revision(revision_id)):
1915
self.target.fetch(self.source, revision_id=revision_id)
1918
def fetch(self, revision_id=None, pb=None):
1919
"""See InterRepository.fetch()."""
1920
from bzrlib.fetch import GenericRepoFetcher
1921
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1922
self.source, self.source._format, self.target,
1923
self.target._format)
1924
# TODO: jam 20070210 This should be an assert, not a translate
1925
revision_id = osutils.safe_revision_id(revision_id)
1926
f = GenericRepoFetcher(to_repository=self.target,
1927
from_repository=self.source,
1928
last_revision=revision_id,
1930
return f.count_copied, f.failed_revisions
1933
class InterWeaveRepo(InterSameDataRepository):
1934
"""Optimised code paths between Weave based repositories.
1936
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
1937
implemented lazy inter-object optimisation.
1941
def _get_repo_format_to_test(self):
1942
from bzrlib.repofmt import weaverepo
1943
return weaverepo.RepositoryFormat7()
1946
def is_compatible(source, target):
1947
"""Be compatible with known Weave formats.
1949
We don't test for the stores being of specific types because that
1950
could lead to confusing results, and there is no need to be
1953
from bzrlib.repofmt.weaverepo import (
1959
return (isinstance(source._format, (RepositoryFormat5,
1961
RepositoryFormat7)) and
1962
isinstance(target._format, (RepositoryFormat5,
1964
RepositoryFormat7)))
1965
except AttributeError:
1969
def copy_content(self, revision_id=None):
1970
"""See InterRepository.copy_content()."""
1971
# weave specific optimised path:
1972
# TODO: jam 20070210 Internal, should be an assert, not translate
1973
revision_id = osutils.safe_revision_id(revision_id)
1975
self.target.set_make_working_trees(self.source.make_working_trees())
1976
except NotImplementedError:
1978
# FIXME do not peek!
1979
if self.source.control_files._transport.listable():
1980
pb = ui.ui_factory.nested_progress_bar()
1982
self.target.weave_store.copy_all_ids(
1983
self.source.weave_store,
1985
from_transaction=self.source.get_transaction(),
1986
to_transaction=self.target.get_transaction())
1987
pb.update('copying inventory', 0, 1)
1988
self.target.control_weaves.copy_multi(
1989
self.source.control_weaves, ['inventory'],
1990
from_transaction=self.source.get_transaction(),
1991
to_transaction=self.target.get_transaction())
1992
self.target._revision_store.text_store.copy_all_ids(
1993
self.source._revision_store.text_store,
1998
self.target.fetch(self.source, revision_id=revision_id)
2001
def fetch(self, revision_id=None, pb=None):
2002
"""See InterRepository.fetch()."""
2003
from bzrlib.fetch import GenericRepoFetcher
2004
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2005
self.source, self.source._format, self.target, self.target._format)
2006
# TODO: jam 20070210 This should be an assert, not a translate
2007
revision_id = osutils.safe_revision_id(revision_id)
2008
f = GenericRepoFetcher(to_repository=self.target,
2009
from_repository=self.source,
2010
last_revision=revision_id,
2012
return f.count_copied, f.failed_revisions
2015
def missing_revision_ids(self, revision_id=None):
2016
"""See InterRepository.missing_revision_ids()."""
2017
# we want all revisions to satisfy revision_id in source.
2018
# but we don't want to stat every file here and there.
2019
# we want then, all revisions other needs to satisfy revision_id
2020
# checked, but not those that we have locally.
2021
# so the first thing is to get a subset of the revisions to
2022
# satisfy revision_id in source, and then eliminate those that
2023
# we do already have.
2024
# this is slow on high latency connection to self, but as as this
2025
# disk format scales terribly for push anyway due to rewriting
2026
# inventory.weave, this is considered acceptable.
2028
if revision_id is not None:
2029
source_ids = self.source.get_ancestry(revision_id)
2030
assert source_ids[0] is None
2033
source_ids = self.source._all_possible_ids()
2034
source_ids_set = set(source_ids)
2035
# source_ids is the worst possible case we may need to pull.
2036
# now we want to filter source_ids against what we actually
2037
# have in target, but don't try to check for existence where we know
2038
# we do not have a revision as that would be pointless.
2039
target_ids = set(self.target._all_possible_ids())
2040
possibly_present_revisions = target_ids.intersection(source_ids_set)
2041
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2042
required_revisions = source_ids_set.difference(actually_present_revisions)
2043
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2044
if revision_id is not None:
2045
# we used get_ancestry to determine source_ids then we are assured all
2046
# revisions referenced are present as they are installed in topological order.
2047
# and the tip revision was validated by get_ancestry.
2048
return required_topo_revisions
2050
# if we just grabbed the possibly available ids, then
2051
# we only have an estimate of whats available and need to validate
2052
# that against the revision records.
2053
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2056
class InterKnitRepo(InterSameDataRepository):
2057
"""Optimised code paths between Knit based repositories."""
2060
def _get_repo_format_to_test(self):
2061
from bzrlib.repofmt import knitrepo
2062
return knitrepo.RepositoryFormatKnit1()
2065
def is_compatible(source, target):
2066
"""Be compatible with known Knit formats.
2068
We don't test for the stores being of specific types because that
2069
could lead to confusing results, and there is no need to be
2072
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit1
2074
return (isinstance(source._format, (RepositoryFormatKnit1)) and
2075
isinstance(target._format, (RepositoryFormatKnit1)))
2076
except AttributeError:
2080
def fetch(self, revision_id=None, pb=None):
2081
"""See InterRepository.fetch()."""
2082
from bzrlib.fetch import KnitRepoFetcher
2083
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2084
self.source, self.source._format, self.target, self.target._format)
2085
# TODO: jam 20070210 This should be an assert, not a translate
2086
revision_id = osutils.safe_revision_id(revision_id)
2087
f = KnitRepoFetcher(to_repository=self.target,
2088
from_repository=self.source,
2089
last_revision=revision_id,
2091
return f.count_copied, f.failed_revisions
2094
def missing_revision_ids(self, revision_id=None):
2095
"""See InterRepository.missing_revision_ids()."""
2096
if revision_id is not None:
2097
source_ids = self.source.get_ancestry(revision_id)
2098
assert source_ids[0] is None
2101
source_ids = self.source.all_revision_ids()
2102
source_ids_set = set(source_ids)
2103
# source_ids is the worst possible case we may need to pull.
2104
# now we want to filter source_ids against what we actually
2105
# have in target, but don't try to check for existence where we know
2106
# we do not have a revision as that would be pointless.
2107
target_ids = set(self.target.all_revision_ids())
2108
possibly_present_revisions = target_ids.intersection(source_ids_set)
2109
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2110
required_revisions = source_ids_set.difference(actually_present_revisions)
2111
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2112
if revision_id is not None:
2113
# we used get_ancestry to determine source_ids then we are assured all
2114
# revisions referenced are present as they are installed in topological order.
2115
# and the tip revision was validated by get_ancestry.
2116
return required_topo_revisions
2118
# if we just grabbed the possibly available ids, then
2119
# we only have an estimate of whats available and need to validate
2120
# that against the revision records.
2121
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2124
class InterModel1and2(InterRepository):
2127
def _get_repo_format_to_test(self):
2131
def is_compatible(source, target):
2132
if not source.supports_rich_root() and target.supports_rich_root():
2138
def fetch(self, revision_id=None, pb=None):
2139
"""See InterRepository.fetch()."""
2140
from bzrlib.fetch import Model1toKnit2Fetcher
2141
# TODO: jam 20070210 This should be an assert, not a translate
2142
revision_id = osutils.safe_revision_id(revision_id)
2143
f = Model1toKnit2Fetcher(to_repository=self.target,
2144
from_repository=self.source,
2145
last_revision=revision_id,
2147
return f.count_copied, f.failed_revisions
2150
def copy_content(self, revision_id=None):
2151
"""Make a complete copy of the content in self into destination.
2153
This is a destructive operation! Do not use it on existing
2156
:param revision_id: Only copy the content needed to construct
2157
revision_id and its parents.
2160
self.target.set_make_working_trees(self.source.make_working_trees())
2161
except NotImplementedError:
2163
# TODO: jam 20070210 Internal, assert, don't translate
2164
revision_id = osutils.safe_revision_id(revision_id)
2165
# but don't bother fetching if we have the needed data now.
2166
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2167
self.target.has_revision(revision_id)):
2169
self.target.fetch(self.source, revision_id=revision_id)
2172
class InterKnit1and2(InterKnitRepo):
2175
def _get_repo_format_to_test(self):
2179
def is_compatible(source, target):
2180
"""Be compatible with Knit1 source and Knit3 target"""
2181
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2183
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit1, \
2184
RepositoryFormatKnit3
2185
return (isinstance(source._format, (RepositoryFormatKnit1)) and
2186
isinstance(target._format, (RepositoryFormatKnit3)))
2187
except AttributeError:
2191
def fetch(self, revision_id=None, pb=None):
2192
"""See InterRepository.fetch()."""
2193
from bzrlib.fetch import Knit1to2Fetcher
2194
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2195
self.source, self.source._format, self.target,
2196
self.target._format)
2197
# TODO: jam 20070210 This should be an assert, not a translate
2198
revision_id = osutils.safe_revision_id(revision_id)
2199
f = Knit1to2Fetcher(to_repository=self.target,
2200
from_repository=self.source,
2201
last_revision=revision_id,
2203
return f.count_copied, f.failed_revisions
2206
class InterRemoteRepository(InterRepository):
2207
"""Code for converting between RemoteRepository objects.
2209
This just gets an non-remote repository from the RemoteRepository, and calls
2210
InterRepository.get again.
2213
def __init__(self, source, target):
2214
if isinstance(source, remote.RemoteRepository):
2215
source._ensure_real()
2216
real_source = source._real_repository
2218
real_source = source
2219
if isinstance(target, remote.RemoteRepository):
2220
target._ensure_real()
2221
real_target = target._real_repository
2223
real_target = target
2224
self.real_inter = InterRepository.get(real_source, real_target)
2227
def is_compatible(source, target):
2228
if isinstance(source, remote.RemoteRepository):
2230
if isinstance(target, remote.RemoteRepository):
2234
def copy_content(self, revision_id=None):
2235
self.real_inter.copy_content(revision_id=revision_id)
2237
def fetch(self, revision_id=None, pb=None):
2238
self.real_inter.fetch(revision_id=revision_id, pb=pb)
2241
def _get_repo_format_to_test(self):
2245
InterRepository.register_optimiser(InterSameDataRepository)
2246
InterRepository.register_optimiser(InterWeaveRepo)
2247
InterRepository.register_optimiser(InterKnitRepo)
2248
InterRepository.register_optimiser(InterModel1and2)
2249
InterRepository.register_optimiser(InterKnit1and2)
2250
InterRepository.register_optimiser(InterRemoteRepository)
2253
class CopyConverter(object):
2254
"""A repository conversion tool which just performs a copy of the content.
2256
This is slow but quite reliable.
2259
def __init__(self, target_format):
2260
"""Create a CopyConverter.
2262
:param target_format: The format the resulting repository should be.
2264
self.target_format = target_format
2266
def convert(self, repo, pb):
2267
"""Perform the conversion of to_convert, giving feedback via pb.
2269
:param to_convert: The disk object to convert.
2270
:param pb: a progress bar to use for progress information.
2275
# this is only useful with metadir layouts - separated repo content.
2276
# trigger an assertion if not such
2277
repo._format.get_format_string()
2278
self.repo_dir = repo.bzrdir
2279
self.step('Moving repository to repository.backup')
2280
self.repo_dir.transport.move('repository', 'repository.backup')
2281
backup_transport = self.repo_dir.transport.clone('repository.backup')
2282
repo._format.check_conversion_target(self.target_format)
2283
self.source_repo = repo._format.open(self.repo_dir,
2285
_override_transport=backup_transport)
2286
self.step('Creating new repository')
2287
converted = self.target_format.initialize(self.repo_dir,
2288
self.source_repo.is_shared())
2289
converted.lock_write()
2291
self.step('Copying content into repository.')
2292
self.source_repo.copy_content_into(converted)
2295
self.step('Deleting old repository content.')
2296
self.repo_dir.transport.delete_tree('repository.backup')
2297
self.pb.note('repository converted')
2299
def step(self, message):
2300
"""Update the pb by a step."""
2302
self.pb.update(message, self.count, self.total)
2314
def _unescaper(match, _map=_unescape_map):
2315
code = match.group(1)
2319
if not code.startswith('#'):
2321
return unichr(int(code[1:])).encode('utf8')
2327
def _unescape_xml(data):
2328
"""Unescape predefined XML entities in a string of data."""
2330
if _unescape_re is None:
2331
_unescape_re = re.compile('\&([^;]*);')
2332
return _unescape_re.sub(_unescaper, data)
added
removed
bzrlib/repository.py
