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(), """
40
revision as _mod_revision,
46
from bzrlib.bundle import serializer
47
from bzrlib.revisiontree import RevisionTree
48
from bzrlib.store.versioned import VersionedFileStore
49
from bzrlib.store.text import TextStore
50
from bzrlib.testament import Testament
51
from bzrlib.util import bencode
54
from bzrlib.decorators import needs_read_lock, needs_write_lock
55
from bzrlib.inter import InterObject
56
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
57
from bzrlib.symbol_versioning import (
60
from bzrlib.trace import mutter, mutter_callsite, note, warning
63
# Old formats display a warning, but only once
64
_deprecation_warning_done = False
67
class CommitBuilder(object):
68
"""Provides an interface to build up a commit.
70
This allows describing a tree to be committed without needing to
71
know the internals of the format of the repository.
74
# all clients should supply tree roots.
75
record_root_entry = True
76
# the default CommitBuilder does not manage trees whose root is versioned.
77
_versioned_root = False
79
def __init__(self, repository, parents, config, timestamp=None,
80
timezone=None, committer=None, revprops=None,
82
"""Initiate a CommitBuilder.
84
:param repository: Repository to commit to.
85
:param parents: Revision ids of the parents of the new revision.
86
:param config: Configuration to use.
87
:param timestamp: Optional timestamp recorded for commit.
88
:param timezone: Optional timezone for timestamp.
89
:param committer: Optional committer to set for commit.
90
:param revprops: Optional dictionary of revision properties.
91
:param revision_id: Optional revision id.
96
self._committer = self._config.username()
98
assert isinstance(committer, basestring), type(committer)
99
self._committer = committer
101
self.new_inventory = Inventory(None)
102
self._new_revision_id = revision_id
103
self.parents = parents
104
self.repository = repository
107
if revprops is not None:
108
self._revprops.update(revprops)
110
if timestamp is None:
111
timestamp = time.time()
112
# Restrict resolution to 1ms
113
self._timestamp = round(timestamp, 3)
116
self._timezone = osutils.local_time_offset()
118
self._timezone = int(timezone)
120
self._generate_revision_if_needed()
121
self.__heads = graph.HeadsCache(repository.get_graph()).heads
123
def commit(self, message):
124
"""Make the actual commit.
126
:return: The revision id of the recorded revision.
128
rev = _mod_revision.Revision(
129
timestamp=self._timestamp,
130
timezone=self._timezone,
131
committer=self._committer,
133
inventory_sha1=self.inv_sha1,
134
revision_id=self._new_revision_id,
135
properties=self._revprops)
136
rev.parent_ids = self.parents
137
self.repository.add_revision(self._new_revision_id, rev,
138
self.new_inventory, self._config)
139
self.repository.commit_write_group()
140
return self._new_revision_id
143
"""Abort the commit that is being built.
145
self.repository.abort_write_group()
147
def revision_tree(self):
148
"""Return the tree that was just committed.
150
After calling commit() this can be called to get a RevisionTree
151
representing the newly committed tree. This is preferred to
152
calling Repository.revision_tree() because that may require
153
deserializing the inventory, while we already have a copy in
156
return RevisionTree(self.repository, self.new_inventory,
157
self._new_revision_id)
159
def finish_inventory(self):
160
"""Tell the builder that the inventory is finished."""
161
if self.new_inventory.root is None:
162
raise AssertionError('Root entry should be supplied to'
163
' record_entry_contents, as of bzr 0.10.',
164
DeprecationWarning, stacklevel=2)
165
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
166
self.new_inventory.revision_id = self._new_revision_id
167
self.inv_sha1 = self.repository.add_inventory(
168
self._new_revision_id,
173
def _gen_revision_id(self):
174
"""Return new revision-id."""
175
return generate_ids.gen_revision_id(self._config.username(),
178
def _generate_revision_if_needed(self):
179
"""Create a revision id if None was supplied.
181
If the repository can not support user-specified revision ids
182
they should override this function and raise CannotSetRevisionId
183
if _new_revision_id is not None.
185
:raises: CannotSetRevisionId
187
if self._new_revision_id is None:
188
self._new_revision_id = self._gen_revision_id()
189
self.random_revid = True
191
self.random_revid = False
193
def _heads(self, file_id, revision_ids):
194
"""Calculate the graph heads for revision_ids in the graph of file_id.
196
This can use either a per-file graph or a global revision graph as we
197
have an identity relationship between the two graphs.
199
return self.__heads(revision_ids)
201
def _check_root(self, ie, parent_invs, tree):
202
"""Helper for record_entry_contents.
204
:param ie: An entry being added.
205
:param parent_invs: The inventories of the parent revisions of the
207
:param tree: The tree that is being committed.
209
# In this revision format, root entries have no knit or weave When
210
# serializing out to disk and back in root.revision is always
212
ie.revision = self._new_revision_id
214
def _get_delta(self, ie, basis_inv, path):
215
"""Get a delta against the basis inventory for ie."""
216
if ie.file_id not in basis_inv:
218
return (None, path, ie.file_id, ie)
219
elif ie != basis_inv[ie.file_id]:
221
# TODO: avoid tis id2path call.
222
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
227
def record_entry_contents(self, ie, parent_invs, path, tree,
229
"""Record the content of ie from tree into the commit if needed.
231
Side effect: sets ie.revision when unchanged
233
:param ie: An inventory entry present in the commit.
234
:param parent_invs: The inventories of the parent revisions of the
236
:param path: The path the entry is at in the tree.
237
:param tree: The tree which contains this entry and should be used to
239
:param content_summary: Summary data from the tree about the paths
240
content - stat, length, exec, sha/link target. This is only
241
accessed when the entry has a revision of None - that is when it is
242
a candidate to commit.
243
:return: A tuple (change_delta, version_recorded). change_delta is
244
an inventory_delta change for this entry against the basis tree of
245
the commit, or None if no change occured against the basis tree.
246
version_recorded is True if a new version of the entry has been
247
recorded. For instance, committing a merge where a file was only
248
changed on the other side will return (delta, False).
250
if self.new_inventory.root is None:
251
if ie.parent_id is not None:
252
raise errors.RootMissing()
253
self._check_root(ie, parent_invs, tree)
254
if ie.revision is None:
255
kind = content_summary[0]
257
# ie is carried over from a prior commit
259
# XXX: repository specific check for nested tree support goes here - if
260
# the repo doesn't want nested trees we skip it ?
261
if (kind == 'tree-reference' and
262
not self.repository._format.supports_tree_reference):
263
# mismatch between commit builder logic and repository:
264
# this needs the entry creation pushed down into the builder.
265
raise NotImplementedError('Missing repository subtree support.')
266
self.new_inventory.add(ie)
268
# TODO: slow, take it out of the inner loop.
270
basis_inv = parent_invs[0]
272
basis_inv = Inventory(root_id=None)
274
# ie.revision is always None if the InventoryEntry is considered
275
# for committing. We may record the previous parents revision if the
276
# content is actually unchanged against a sole head.
277
if ie.revision is not None:
278
if not self._versioned_root and path == '':
279
# repositories that do not version the root set the root's
280
# revision to the new commit even when no change occurs, and
281
# this masks when a change may have occurred against the basis,
282
# so calculate if one happened.
283
if ie.file_id in basis_inv:
284
delta = (basis_inv.id2path(ie.file_id), path,
288
delta = (None, path, ie.file_id, ie)
291
# we don't need to commit this, because the caller already
292
# determined that an existing revision of this file is
294
return None, (ie.revision == self._new_revision_id)
295
# XXX: Friction: parent_candidates should return a list not a dict
296
# so that we don't have to walk the inventories again.
297
parent_candiate_entries = ie.parent_candidates(parent_invs)
298
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
300
for inv in parent_invs:
301
if ie.file_id in inv:
302
old_rev = inv[ie.file_id].revision
303
if old_rev in head_set:
304
heads.append(inv[ie.file_id].revision)
305
head_set.remove(inv[ie.file_id].revision)
308
# now we check to see if we need to write a new record to the
310
# We write a new entry unless there is one head to the ancestors, and
311
# the kind-derived content is unchanged.
313
# Cheapest check first: no ancestors, or more the one head in the
314
# ancestors, we write a new node.
318
# There is a single head, look it up for comparison
319
parent_entry = parent_candiate_entries[heads[0]]
320
# if the non-content specific data has changed, we'll be writing a
322
if (parent_entry.parent_id != ie.parent_id or
323
parent_entry.name != ie.name):
325
# now we need to do content specific checks:
327
# if the kind changed the content obviously has
328
if kind != parent_entry.kind:
331
assert content_summary[2] is not None, \
332
"Files must not have executable = None"
334
if (# if the file length changed we have to store:
335
parent_entry.text_size != content_summary[1] or
336
# if the exec bit has changed we have to store:
337
parent_entry.executable != content_summary[2]):
339
elif parent_entry.text_sha1 == content_summary[3]:
340
# all meta and content is unchanged (using a hash cache
341
# hit to check the sha)
342
ie.revision = parent_entry.revision
343
ie.text_size = parent_entry.text_size
344
ie.text_sha1 = parent_entry.text_sha1
345
ie.executable = parent_entry.executable
346
return self._get_delta(ie, basis_inv, path), False
348
# Either there is only a hash change(no hash cache entry,
349
# or same size content change), or there is no change on
351
# Provide the parent's hash to the store layer, so that the
352
# content is unchanged we will not store a new node.
353
nostore_sha = parent_entry.text_sha1
355
# We want to record a new node regardless of the presence or
356
# absence of a content change in the file.
358
ie.executable = content_summary[2]
359
lines = tree.get_file(ie.file_id, path).readlines()
361
ie.text_sha1, ie.text_size = self._add_text_to_weave(
362
ie.file_id, lines, heads, nostore_sha)
363
except errors.ExistingContent:
364
# Turns out that the file content was unchanged, and we were
365
# only going to store a new node if it was changed. Carry over
367
ie.revision = parent_entry.revision
368
ie.text_size = parent_entry.text_size
369
ie.text_sha1 = parent_entry.text_sha1
370
ie.executable = parent_entry.executable
371
return self._get_delta(ie, basis_inv, path), False
372
elif kind == 'directory':
374
# all data is meta here, nothing specific to directory, so
376
ie.revision = parent_entry.revision
377
return self._get_delta(ie, basis_inv, path), False
379
self._add_text_to_weave(ie.file_id, lines, heads, None)
380
elif kind == 'symlink':
381
current_link_target = content_summary[3]
383
# symlink target is not generic metadata, check if it has
385
if current_link_target != parent_entry.symlink_target:
388
# unchanged, carry over.
389
ie.revision = parent_entry.revision
390
ie.symlink_target = parent_entry.symlink_target
391
return self._get_delta(ie, basis_inv, path), False
392
ie.symlink_target = current_link_target
394
self._add_text_to_weave(ie.file_id, lines, heads, None)
395
elif kind == 'tree-reference':
397
if content_summary[3] != parent_entry.reference_revision:
400
# unchanged, carry over.
401
ie.reference_revision = parent_entry.reference_revision
402
ie.revision = parent_entry.revision
403
return self._get_delta(ie, basis_inv, path), False
404
ie.reference_revision = content_summary[3]
406
self._add_text_to_weave(ie.file_id, lines, heads, None)
408
raise NotImplementedError('unknown kind')
409
ie.revision = self._new_revision_id
410
return self._get_delta(ie, basis_inv, path), True
412
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
413
versionedfile = self.repository.weave_store.get_weave_or_empty(
414
file_id, self.repository.get_transaction())
415
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
416
# than add_lines, and allows committing when a parent is ghosted for
418
# Note: as we read the content directly from the tree, we know its not
419
# been turned into unicode or badly split - but a broken tree
420
# implementation could give us bad output from readlines() so this is
421
# not a guarantee of safety. What would be better is always checking
422
# the content during test suite execution. RBC 20070912
424
return versionedfile.add_lines_with_ghosts(
425
self._new_revision_id, parents, new_lines,
426
nostore_sha=nostore_sha, random_id=self.random_revid,
427
check_content=False)[0:2]
429
versionedfile.clear_cache()
432
class RootCommitBuilder(CommitBuilder):
433
"""This commitbuilder actually records the root id"""
435
# the root entry gets versioned properly by this builder.
436
_versioned_root = True
438
def _check_root(self, ie, parent_invs, tree):
439
"""Helper for record_entry_contents.
441
:param ie: An entry being added.
442
:param parent_invs: The inventories of the parent revisions of the
444
:param tree: The tree that is being committed.
448
######################################################################
451
class Repository(object):
452
"""Repository holding history for one or more branches.
454
The repository holds and retrieves historical information including
455
revisions and file history. It's normally accessed only by the Branch,
456
which views a particular line of development through that history.
458
The Repository builds on top of Stores and a Transport, which respectively
459
describe the disk data format and the way of accessing the (possibly
463
# What class to use for a CommitBuilder. Often its simpler to change this
464
# in a Repository class subclass rather than to override
465
# get_commit_builder.
466
_commit_builder_class = CommitBuilder
467
# The search regex used by xml based repositories to determine what things
468
# where changed in a single commit.
469
_file_ids_altered_regex = lazy_regex.lazy_compile(
470
r'file_id="(?P<file_id>[^"]+)"'
471
r'.* revision="(?P<revision_id>[^"]+)"'
474
def abort_write_group(self):
475
"""Commit the contents accrued within the current write group.
477
:seealso: start_write_group.
479
if self._write_group is not self.get_transaction():
480
# has an unlock or relock occured ?
481
raise errors.BzrError('mismatched lock context and write group.')
482
self._abort_write_group()
483
self._write_group = None
485
def _abort_write_group(self):
486
"""Template method for per-repository write group cleanup.
488
This is called during abort before the write group is considered to be
489
finished and should cleanup any internal state accrued during the write
490
group. There is no requirement that data handed to the repository be
491
*not* made available - this is not a rollback - but neither should any
492
attempt be made to ensure that data added is fully commited. Abort is
493
invoked when an error has occured so futher disk or network operations
494
may not be possible or may error and if possible should not be
499
def add_inventory(self, revision_id, inv, parents):
500
"""Add the inventory inv to the repository as revision_id.
502
:param parents: The revision ids of the parents that revision_id
503
is known to have and are in the repository already.
505
returns the sha1 of the serialized inventory.
507
assert self.is_in_write_group()
508
_mod_revision.check_not_reserved_id(revision_id)
509
assert inv.revision_id is None or inv.revision_id == revision_id, \
510
"Mismatch between inventory revision" \
511
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
512
assert inv.root is not None
513
inv_lines = self._serialise_inventory_to_lines(inv)
514
inv_vf = self.get_inventory_weave()
515
return self._inventory_add_lines(inv_vf, revision_id, parents,
516
inv_lines, check_content=False)
518
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
520
"""Store lines in inv_vf and return the sha1 of the inventory."""
522
for parent in parents:
524
final_parents.append(parent)
525
return inv_vf.add_lines(revision_id, final_parents, lines,
526
check_content=check_content)[0]
529
def add_revision(self, revision_id, rev, inv=None, config=None):
530
"""Add rev to the revision store as revision_id.
532
:param revision_id: the revision id to use.
533
:param rev: The revision object.
534
:param inv: The inventory for the revision. if None, it will be looked
535
up in the inventory storer
536
:param config: If None no digital signature will be created.
537
If supplied its signature_needed method will be used
538
to determine if a signature should be made.
540
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
542
_mod_revision.check_not_reserved_id(revision_id)
543
if config is not None and config.signature_needed():
545
inv = self.get_inventory(revision_id)
546
plaintext = Testament(rev, inv).as_short_text()
547
self.store_revision_signature(
548
gpg.GPGStrategy(config), plaintext, revision_id)
549
if not revision_id in self.get_inventory_weave():
551
raise errors.WeaveRevisionNotPresent(revision_id,
552
self.get_inventory_weave())
554
# yes, this is not suitable for adding with ghosts.
555
self.add_inventory(revision_id, inv, rev.parent_ids)
556
self._revision_store.add_revision(rev, self.get_transaction())
558
def _add_revision_text(self, revision_id, text):
559
revision = self._revision_store._serializer.read_revision_from_string(
561
self._revision_store._add_revision(revision, StringIO(text),
562
self.get_transaction())
564
def all_revision_ids(self):
565
"""Returns a list of all the revision ids in the repository.
567
This is deprecated because code should generally work on the graph
568
reachable from a particular revision, and ignore any other revisions
569
that might be present. There is no direct replacement method.
571
if 'evil' in debug.debug_flags:
572
mutter_callsite(2, "all_revision_ids is linear with history.")
573
return self._all_revision_ids()
575
def _all_revision_ids(self):
576
"""Returns a list of all the revision ids in the repository.
578
These are in as much topological order as the underlying store can
581
raise NotImplementedError(self._all_revision_ids)
583
def break_lock(self):
584
"""Break a lock if one is present from another instance.
586
Uses the ui factory to ask for confirmation if the lock may be from
589
self.control_files.break_lock()
592
def _eliminate_revisions_not_present(self, revision_ids):
593
"""Check every revision id in revision_ids to see if we have it.
595
Returns a set of the present revisions.
598
for id in revision_ids:
599
if self.has_revision(id):
604
def create(a_bzrdir):
605
"""Construct the current default format repository in a_bzrdir."""
606
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
608
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
609
"""instantiate a Repository.
611
:param _format: The format of the repository on disk.
612
:param a_bzrdir: The BzrDir of the repository.
614
In the future we will have a single api for all stores for
615
getting file texts, inventories and revisions, then
616
this construct will accept instances of those things.
618
super(Repository, self).__init__()
619
self._format = _format
620
# the following are part of the public API for Repository:
621
self.bzrdir = a_bzrdir
622
self.control_files = control_files
623
self._revision_store = _revision_store
624
# backwards compatibility
625
self.weave_store = text_store
627
self._reconcile_does_inventory_gc = True
628
self._reconcile_fixes_text_parents = False
629
self._reconcile_backsup_inventory = True
630
# not right yet - should be more semantically clear ?
632
self.control_store = control_store
633
self.control_weaves = control_store
634
# TODO: make sure to construct the right store classes, etc, depending
635
# on whether escaping is required.
636
self._warn_if_deprecated()
637
self._write_group = None
638
self.base = control_files._transport.base
641
return '%s(%r)' % (self.__class__.__name__,
644
def has_same_location(self, other):
645
"""Returns a boolean indicating if this repository is at the same
646
location as another repository.
648
This might return False even when two repository objects are accessing
649
the same physical repository via different URLs.
651
if self.__class__ is not other.__class__:
653
return (self.control_files._transport.base ==
654
other.control_files._transport.base)
656
def is_in_write_group(self):
657
"""Return True if there is an open write group.
659
:seealso: start_write_group.
661
return self._write_group is not None
664
return self.control_files.is_locked()
666
def is_write_locked(self):
667
"""Return True if this object is write locked."""
668
return self.is_locked() and self.control_files._lock_mode == 'w'
670
def lock_write(self, token=None):
671
"""Lock this repository for writing.
673
This causes caching within the repository obejct to start accumlating
674
data during reads, and allows a 'write_group' to be obtained. Write
675
groups must be used for actual data insertion.
677
:param token: if this is already locked, then lock_write will fail
678
unless the token matches the existing lock.
679
:returns: a token if this instance supports tokens, otherwise None.
680
:raises TokenLockingNotSupported: when a token is given but this
681
instance doesn't support using token locks.
682
:raises MismatchedToken: if the specified token doesn't match the token
683
of the existing lock.
684
:seealso: start_write_group.
686
A token should be passed in if you know that you have locked the object
687
some other way, and need to synchronise this object's state with that
690
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
692
result = self.control_files.lock_write(token=token)
697
self.control_files.lock_read()
700
def get_physical_lock_status(self):
701
return self.control_files.get_physical_lock_status()
703
def leave_lock_in_place(self):
704
"""Tell this repository not to release the physical lock when this
707
If lock_write doesn't return a token, then this method is not supported.
709
self.control_files.leave_in_place()
711
def dont_leave_lock_in_place(self):
712
"""Tell this repository to release the physical lock when this
713
object is unlocked, even if it didn't originally acquire it.
715
If lock_write doesn't return a token, then this method is not supported.
717
self.control_files.dont_leave_in_place()
720
def gather_stats(self, revid=None, committers=None):
721
"""Gather statistics from a revision id.
723
:param revid: The revision id to gather statistics from, if None, then
724
no revision specific statistics are gathered.
725
:param committers: Optional parameter controlling whether to grab
726
a count of committers from the revision specific statistics.
727
:return: A dictionary of statistics. Currently this contains:
728
committers: The number of committers if requested.
729
firstrev: A tuple with timestamp, timezone for the penultimate left
730
most ancestor of revid, if revid is not the NULL_REVISION.
731
latestrev: A tuple with timestamp, timezone for revid, if revid is
732
not the NULL_REVISION.
733
revisions: The total revision count in the repository.
734
size: An estimate disk size of the repository in bytes.
737
if revid and committers:
738
result['committers'] = 0
739
if revid and revid != _mod_revision.NULL_REVISION:
741
all_committers = set()
742
revisions = self.get_ancestry(revid)
743
# pop the leading None
745
first_revision = None
747
# ignore the revisions in the middle - just grab first and last
748
revisions = revisions[0], revisions[-1]
749
for revision in self.get_revisions(revisions):
750
if not first_revision:
751
first_revision = revision
753
all_committers.add(revision.committer)
754
last_revision = revision
756
result['committers'] = len(all_committers)
757
result['firstrev'] = (first_revision.timestamp,
758
first_revision.timezone)
759
result['latestrev'] = (last_revision.timestamp,
760
last_revision.timezone)
762
# now gather global repository information
763
if self.bzrdir.root_transport.listable():
764
c, t = self._revision_store.total_size(self.get_transaction())
765
result['revisions'] = c
769
def get_data_stream(self, revision_ids):
770
raise NotImplementedError(self.get_data_stream)
772
def insert_data_stream(self, stream):
773
"""XXX What does this really do?
775
Is it a substitute for fetch?
776
Should it manage its own write group ?
778
for item_key, bytes in stream:
779
if item_key[0] == 'file':
780
(file_id,) = item_key[1:]
781
knit = self.weave_store.get_weave_or_empty(
782
file_id, self.get_transaction())
783
elif item_key == ('inventory',):
784
knit = self.get_inventory_weave()
785
elif item_key == ('revisions',):
786
knit = self._revision_store.get_revision_file(
787
self.get_transaction())
788
elif item_key == ('signatures',):
789
knit = self._revision_store.get_signature_file(
790
self.get_transaction())
792
raise RepositoryDataStreamError(
793
"Unrecognised data stream key '%s'" % (item_key,))
794
decoded_list = bencode.bdecode(bytes)
795
format = decoded_list.pop(0)
798
for version, options, parents, some_bytes in decoded_list:
799
data_list.append((version, options, len(some_bytes), parents))
800
knit_bytes += some_bytes
801
knit.insert_data_stream(
802
(format, data_list, StringIO(knit_bytes).read))
805
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
806
"""Return the revision ids that other has that this does not.
808
These are returned in topological order.
810
revision_id: only return revision ids included by revision_id.
812
return InterRepository.get(other, self).missing_revision_ids(
813
revision_id, find_ghosts)
817
"""Open the repository rooted at base.
819
For instance, if the repository is at URL/.bzr/repository,
820
Repository.open(URL) -> a Repository instance.
822
control = bzrdir.BzrDir.open(base)
823
return control.open_repository()
825
def copy_content_into(self, destination, revision_id=None):
826
"""Make a complete copy of the content in self into destination.
828
This is a destructive operation! Do not use it on existing
831
return InterRepository.get(self, destination).copy_content(revision_id)
833
def commit_write_group(self):
834
"""Commit the contents accrued within the current write group.
836
:seealso: start_write_group.
838
if self._write_group is not self.get_transaction():
839
# has an unlock or relock occured ?
840
raise errors.BzrError('mismatched lock context %r and '
842
(self.get_transaction(), self._write_group))
843
self._commit_write_group()
844
self._write_group = None
846
def _commit_write_group(self):
847
"""Template method for per-repository write group cleanup.
849
This is called before the write group is considered to be
850
finished and should ensure that all data handed to the repository
851
for writing during the write group is safely committed (to the
852
extent possible considering file system caching etc).
855
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
856
"""Fetch the content required to construct revision_id from source.
858
If revision_id is None all content is copied.
859
:param find_ghosts: Find and copy revisions in the source that are
860
ghosts in the target (and not reachable directly by walking out to
861
the first-present revision in target from revision_id).
863
# fast path same-url fetch operations
864
if self.has_same_location(source):
865
# check that last_revision is in 'from' and then return a
867
if (revision_id is not None and
868
not _mod_revision.is_null(revision_id)):
869
self.get_revision(revision_id)
871
inter = InterRepository.get(source, self)
873
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
874
except NotImplementedError:
875
raise errors.IncompatibleRepositories(source, self)
877
def create_bundle(self, target, base, fileobj, format=None):
878
return serializer.write_bundle(self, target, base, fileobj, format)
880
def get_commit_builder(self, branch, parents, config, timestamp=None,
881
timezone=None, committer=None, revprops=None,
883
"""Obtain a CommitBuilder for this repository.
885
:param branch: Branch to commit to.
886
:param parents: Revision ids of the parents of the new revision.
887
:param config: Configuration to use.
888
:param timestamp: Optional timestamp recorded for commit.
889
:param timezone: Optional timezone for timestamp.
890
:param committer: Optional committer to set for commit.
891
:param revprops: Optional dictionary of revision properties.
892
:param revision_id: Optional revision id.
894
result = self._commit_builder_class(self, parents, config,
895
timestamp, timezone, committer, revprops, revision_id)
896
self.start_write_group()
900
if (self.control_files._lock_count == 1 and
901
self.control_files._lock_mode == 'w'):
902
if self._write_group is not None:
903
self.abort_write_group()
904
self.control_files.unlock()
905
raise errors.BzrError(
906
'Must end write groups before releasing write locks.')
907
self.control_files.unlock()
910
def clone(self, a_bzrdir, revision_id=None):
911
"""Clone this repository into a_bzrdir using the current format.
913
Currently no check is made that the format of this repository and
914
the bzrdir format are compatible. FIXME RBC 20060201.
916
:return: The newly created destination repository.
918
# TODO: deprecate after 0.16; cloning this with all its settings is
919
# probably not very useful -- mbp 20070423
920
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
921
self.copy_content_into(dest_repo, revision_id)
924
def start_write_group(self):
925
"""Start a write group in the repository.
927
Write groups are used by repositories which do not have a 1:1 mapping
928
between file ids and backend store to manage the insertion of data from
929
both fetch and commit operations.
931
A write lock is required around the start_write_group/commit_write_group
932
for the support of lock-requiring repository formats.
934
One can only insert data into a repository inside a write group.
938
if not self.is_write_locked():
939
raise errors.NotWriteLocked(self)
940
if self._write_group:
941
raise errors.BzrError('already in a write group')
942
self._start_write_group()
943
# so we can detect unlock/relock - the write group is now entered.
944
self._write_group = self.get_transaction()
946
def _start_write_group(self):
947
"""Template method for per-repository write group startup.
949
This is called before the write group is considered to be
954
def sprout(self, to_bzrdir, revision_id=None):
955
"""Create a descendent repository for new development.
957
Unlike clone, this does not copy the settings of the repository.
959
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
960
dest_repo.fetch(self, revision_id=revision_id)
963
def _create_sprouting_repo(self, a_bzrdir, shared):
964
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
965
# use target default format.
966
dest_repo = a_bzrdir.create_repository()
968
# Most control formats need the repository to be specifically
969
# created, but on some old all-in-one formats it's not needed
971
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
972
except errors.UninitializableFormat:
973
dest_repo = a_bzrdir.open_repository()
977
def has_revision(self, revision_id):
978
"""True if this repository has a copy of the revision."""
979
if 'evil' in debug.debug_flags:
980
mutter_callsite(3, "has_revision is a LBYL symptom.")
981
return self._revision_store.has_revision_id(revision_id,
982
self.get_transaction())
985
def get_revision(self, revision_id):
986
"""Return the Revision object for a named revision."""
987
return self.get_revisions([revision_id])[0]
990
def get_revision_reconcile(self, revision_id):
991
"""'reconcile' helper routine that allows access to a revision always.
993
This variant of get_revision does not cross check the weave graph
994
against the revision one as get_revision does: but it should only
995
be used by reconcile, or reconcile-alike commands that are correcting
996
or testing the revision graph.
998
return self._get_revisions([revision_id])[0]
1001
def get_revisions(self, revision_ids):
1002
"""Get many revisions at once."""
1003
return self._get_revisions(revision_ids)
1006
def _get_revisions(self, revision_ids):
1007
"""Core work logic to get many revisions without sanity checks."""
1008
for rev_id in revision_ids:
1009
if not rev_id or not isinstance(rev_id, basestring):
1010
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1011
revs = self._revision_store.get_revisions(revision_ids,
1012
self.get_transaction())
1014
assert not isinstance(rev.revision_id, unicode)
1015
for parent_id in rev.parent_ids:
1016
assert not isinstance(parent_id, unicode)
1020
def get_revision_xml(self, revision_id):
1021
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1022
# would have already do it.
1023
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1024
rev = self.get_revision(revision_id)
1025
rev_tmp = StringIO()
1026
# the current serializer..
1027
self._revision_store._serializer.write_revision(rev, rev_tmp)
1029
return rev_tmp.getvalue()
1032
def get_deltas_for_revisions(self, revisions):
1033
"""Produce a generator of revision deltas.
1035
Note that the input is a sequence of REVISIONS, not revision_ids.
1036
Trees will be held in memory until the generator exits.
1037
Each delta is relative to the revision's lefthand predecessor.
1039
required_trees = set()
1040
for revision in revisions:
1041
required_trees.add(revision.revision_id)
1042
required_trees.update(revision.parent_ids[:1])
1043
trees = dict((t.get_revision_id(), t) for
1044
t in self.revision_trees(required_trees))
1045
for revision in revisions:
1046
if not revision.parent_ids:
1047
old_tree = self.revision_tree(None)
1049
old_tree = trees[revision.parent_ids[0]]
1050
yield trees[revision.revision_id].changes_from(old_tree)
1053
def get_revision_delta(self, revision_id):
1054
"""Return the delta for one revision.
1056
The delta is relative to the left-hand predecessor of the
1059
r = self.get_revision(revision_id)
1060
return list(self.get_deltas_for_revisions([r]))[0]
1063
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1064
signature = gpg_strategy.sign(plaintext)
1065
self.add_signature_text(revision_id, signature)
1068
def add_signature_text(self, revision_id, signature):
1069
self._revision_store.add_revision_signature_text(revision_id,
1071
self.get_transaction())
1073
def find_text_key_references(self):
1074
"""Find the text key references within the repository.
1076
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1077
revision_ids. Each altered file-ids has the exact revision_ids that
1078
altered it listed explicitly.
1079
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1080
to whether they were referred to by the inventory of the
1081
revision_id that they contain. The inventory texts from all present
1082
revision ids are assessed to generate this report.
1084
revision_ids = self.all_revision_ids()
1085
w = self.get_inventory_weave()
1086
pb = ui.ui_factory.nested_progress_bar()
1088
return self._find_text_key_references_from_xml_inventory_lines(
1089
w.iter_lines_added_or_present_in_versions(revision_ids, pb=pb))
1093
def _find_text_key_references_from_xml_inventory_lines(self,
1095
"""Core routine for extracting references to texts from inventories.
1097
This performs the translation of xml lines to revision ids.
1099
:param line_iterator: An iterator of lines, origin_version_id
1100
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1101
to whether they were referred to by the inventory of the
1102
revision_id that they contain. Note that if that revision_id was
1103
not part of the line_iterator's output then False will be given -
1104
even though it may actually refer to that key.
1106
if not self._serializer.support_altered_by_hack:
1107
raise AssertionError(
1108
"_find_text_key_references_from_xml_inventory_lines only "
1109
"supported for branches which store inventory as unnested xml"
1110
", not on %r" % self)
1113
# this code needs to read every new line in every inventory for the
1114
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1115
# not present in one of those inventories is unnecessary but not
1116
# harmful because we are filtering by the revision id marker in the
1117
# inventory lines : we only select file ids altered in one of those
1118
# revisions. We don't need to see all lines in the inventory because
1119
# only those added in an inventory in rev X can contain a revision=X
1121
unescape_revid_cache = {}
1122
unescape_fileid_cache = {}
1124
# jam 20061218 In a big fetch, this handles hundreds of thousands
1125
# of lines, so it has had a lot of inlining and optimizing done.
1126
# Sorry that it is a little bit messy.
1127
# Move several functions to be local variables, since this is a long
1129
search = self._file_ids_altered_regex.search
1130
unescape = _unescape_xml
1131
setdefault = result.setdefault
1132
for line, version_id in line_iterator:
1133
match = search(line)
1136
# One call to match.group() returning multiple items is quite a
1137
# bit faster than 2 calls to match.group() each returning 1
1138
file_id, revision_id = match.group('file_id', 'revision_id')
1140
# Inlining the cache lookups helps a lot when you make 170,000
1141
# lines and 350k ids, versus 8.4 unique ids.
1142
# Using a cache helps in 2 ways:
1143
# 1) Avoids unnecessary decoding calls
1144
# 2) Re-uses cached strings, which helps in future set and
1146
# (2) is enough that removing encoding entirely along with
1147
# the cache (so we are using plain strings) results in no
1148
# performance improvement.
1150
revision_id = unescape_revid_cache[revision_id]
1152
unescaped = unescape(revision_id)
1153
unescape_revid_cache[revision_id] = unescaped
1154
revision_id = unescaped
1156
# Note that unconditionally unescaping means that we deserialise
1157
# every fileid, which for general 'pull' is not great, but we don't
1158
# really want to have some many fulltexts that this matters anyway.
1161
file_id = unescape_fileid_cache[file_id]
1163
unescaped = unescape(file_id)
1164
unescape_fileid_cache[file_id] = unescaped
1167
key = (file_id, revision_id)
1168
setdefault(key, False)
1169
if revision_id == version_id:
1173
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1175
"""Helper routine for fileids_altered_by_revision_ids.
1177
This performs the translation of xml lines to revision ids.
1179
:param line_iterator: An iterator of lines, origin_version_id
1180
:param revision_ids: The revision ids to filter for. This should be a
1181
set or other type which supports efficient __contains__ lookups, as
1182
the revision id from each parsed line will be looked up in the
1183
revision_ids filter.
1184
:return: a dictionary mapping altered file-ids to an iterable of
1185
revision_ids. Each altered file-ids has the exact revision_ids that
1186
altered it listed explicitly.
1189
setdefault = result.setdefault
1190
for file_id, revision_id in \
1191
self._find_text_key_references_from_xml_inventory_lines(
1192
line_iterator).iterkeys():
1193
# once data is all ensured-consistent; then this is
1194
# if revision_id == version_id
1195
if revision_id in revision_ids:
1196
setdefault(file_id, set()).add(revision_id)
1199
def fileids_altered_by_revision_ids(self, revision_ids):
1200
"""Find the file ids and versions affected by revisions.
1202
:param revisions: an iterable containing revision ids.
1203
:return: a dictionary mapping altered file-ids to an iterable of
1204
revision_ids. Each altered file-ids has the exact revision_ids that
1205
altered it listed explicitly.
1207
selected_revision_ids = set(revision_ids)
1208
w = self.get_inventory_weave()
1209
pb = ui.ui_factory.nested_progress_bar()
1211
return self._find_file_ids_from_xml_inventory_lines(
1212
w.iter_lines_added_or_present_in_versions(
1213
selected_revision_ids, pb=pb),
1214
selected_revision_ids)
1218
def iter_files_bytes(self, desired_files):
1219
"""Iterate through file versions.
1221
Files will not necessarily be returned in the order they occur in
1222
desired_files. No specific order is guaranteed.
1224
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1225
value supplied by the caller as part of desired_files. It should
1226
uniquely identify the file version in the caller's context. (Examples:
1227
an index number or a TreeTransform trans_id.)
1229
bytes_iterator is an iterable of bytestrings for the file. The
1230
kind of iterable and length of the bytestrings are unspecified, but for
1231
this implementation, it is a list of lines produced by
1232
VersionedFile.get_lines().
1234
:param desired_files: a list of (file_id, revision_id, identifier)
1237
transaction = self.get_transaction()
1238
for file_id, revision_id, callable_data in desired_files:
1240
weave = self.weave_store.get_weave(file_id, transaction)
1241
except errors.NoSuchFile:
1242
raise errors.NoSuchIdInRepository(self, file_id)
1243
yield callable_data, weave.get_lines(revision_id)
1245
def _generate_text_key_index(self, text_key_references=None):
1246
"""Generate a new text key index for the repository.
1248
This is an expensive function that will take considerable time to run.
1250
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1251
list of parents, also text keys. When a given key has no parents,
1252
the parents list will be [NULL_REVISION].
1254
# All revisions, to find inventory parents.
1255
revision_graph = self.get_revision_graph_with_ghosts()
1256
ancestors = revision_graph.get_ancestors()
1257
if text_key_references is None:
1258
text_key_references = self.find_text_key_references()
1259
pb = ui.ui_factory.nested_progress_bar()
1261
return self._do_generate_text_key_index(ancestors,
1262
text_key_references, pb)
1266
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1267
"""Helper for _generate_text_key_index to avoid deep nesting."""
1268
revision_order = tsort.topo_sort(ancestors)
1269
invalid_keys = set()
1271
for revision_id in revision_order:
1272
revision_keys[revision_id] = set()
1273
text_count = len(text_key_references)
1274
# a cache of the text keys to allow reuse; costs a dict of all the
1275
# keys, but saves a 2-tuple for every child of a given key.
1277
for text_key, valid in text_key_references.iteritems():
1279
invalid_keys.add(text_key)
1281
revision_keys[text_key[1]].add(text_key)
1282
text_key_cache[text_key] = text_key
1283
del text_key_references
1285
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1286
NULL_REVISION = _mod_revision.NULL_REVISION
1287
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1288
# too small for large or very branchy trees. However, for 55K path
1289
# trees, it would be easy to use too much memory trivially. Ideally we
1290
# could gauge this by looking at available real memory etc, but this is
1291
# always a tricky proposition.
1292
inventory_cache = lru_cache.LRUCache(10)
1293
batch_size = 10 # should be ~150MB on a 55K path tree
1294
batch_count = len(revision_order) / batch_size + 1
1296
pb.update("Calculating text parents.", processed_texts, text_count)
1297
for offset in xrange(batch_count):
1298
to_query = revision_order[offset * batch_size:(offset + 1) *
1302
for rev_tree in self.revision_trees(to_query):
1303
revision_id = rev_tree.get_revision_id()
1304
parent_ids = ancestors[revision_id]
1305
for text_key in revision_keys[revision_id]:
1306
pb.update("Calculating text parents.", processed_texts)
1307
processed_texts += 1
1308
candidate_parents = []
1309
for parent_id in parent_ids:
1310
parent_text_key = (text_key[0], parent_id)
1312
check_parent = parent_text_key not in \
1313
revision_keys[parent_id]
1315
# the parent parent_id is a ghost:
1316
check_parent = False
1317
# truncate the derived graph against this ghost.
1318
parent_text_key = None
1320
# look at the parent commit details inventories to
1321
# determine possible candidates in the per file graph.
1324
inv = inventory_cache[parent_id]
1326
inv = self.revision_tree(parent_id).inventory
1327
inventory_cache[parent_id] = inv
1328
parent_entry = inv._byid.get(text_key[0], None)
1329
if parent_entry is not None:
1331
text_key[0], parent_entry.revision)
1333
parent_text_key = None
1334
if parent_text_key is not None:
1335
candidate_parents.append(
1336
text_key_cache[parent_text_key])
1337
parent_heads = text_graph.heads(candidate_parents)
1338
new_parents = list(parent_heads)
1339
new_parents.sort(key=lambda x:candidate_parents.index(x))
1340
if new_parents == []:
1341
new_parents = [NULL_REVISION]
1342
text_index[text_key] = new_parents
1344
for text_key in invalid_keys:
1345
text_index[text_key] = [NULL_REVISION]
1348
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1349
"""Get an iterable listing the keys of all the data introduced by a set
1352
The keys will be ordered so that the corresponding items can be safely
1353
fetched and inserted in that order.
1355
:returns: An iterable producing tuples of (knit-kind, file-id,
1356
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1357
'revisions'. file-id is None unless knit-kind is 'file'.
1359
# XXX: it's a bit weird to control the inventory weave caching in this
1360
# generator. Ideally the caching would be done in fetch.py I think. Or
1361
# maybe this generator should explicitly have the contract that it
1362
# should not be iterated until the previously yielded item has been
1365
inv_w = self.get_inventory_weave()
1366
inv_w.enable_cache()
1368
# file ids that changed
1369
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1371
num_file_ids = len(file_ids)
1372
for file_id, altered_versions in file_ids.iteritems():
1373
if _files_pb is not None:
1374
_files_pb.update("fetch texts", count, num_file_ids)
1376
yield ("file", file_id, altered_versions)
1377
# We're done with the files_pb. Note that it finished by the caller,
1378
# just as it was created by the caller.
1382
yield ("inventory", None, revision_ids)
1386
revisions_with_signatures = set()
1387
for rev_id in revision_ids:
1389
self.get_signature_text(rev_id)
1390
except errors.NoSuchRevision:
1394
revisions_with_signatures.add(rev_id)
1396
yield ("signatures", None, revisions_with_signatures)
1399
yield ("revisions", None, revision_ids)
1402
def get_inventory_weave(self):
1403
return self.control_weaves.get_weave('inventory',
1404
self.get_transaction())
1407
def get_inventory(self, revision_id):
1408
"""Get Inventory object by hash."""
1409
return self.deserialise_inventory(
1410
revision_id, self.get_inventory_xml(revision_id))
1412
def deserialise_inventory(self, revision_id, xml):
1413
"""Transform the xml into an inventory object.
1415
:param revision_id: The expected revision id of the inventory.
1416
:param xml: A serialised inventory.
1418
return self._serializer.read_inventory_from_string(xml, revision_id)
1420
def serialise_inventory(self, inv):
1421
return self._serializer.write_inventory_to_string(inv)
1423
def _serialise_inventory_to_lines(self, inv):
1424
return self._serializer.write_inventory_to_lines(inv)
1426
def get_serializer_format(self):
1427
return self._serializer.format_num
1430
def get_inventory_xml(self, revision_id):
1431
"""Get inventory XML as a file object."""
1433
assert isinstance(revision_id, str), type(revision_id)
1434
iw = self.get_inventory_weave()
1435
return iw.get_text(revision_id)
1437
raise errors.HistoryMissing(self, 'inventory', revision_id)
1440
def get_inventory_sha1(self, revision_id):
1441
"""Return the sha1 hash of the inventory entry
1443
return self.get_revision(revision_id).inventory_sha1
1446
def get_revision_graph(self, revision_id=None):
1447
"""Return a dictionary containing the revision graph.
1449
NB: This method should not be used as it accesses the entire graph all
1450
at once, which is much more data than most operations should require.
1452
:param revision_id: The revision_id to get a graph from. If None, then
1453
the entire revision graph is returned. This is a deprecated mode of
1454
operation and will be removed in the future.
1455
:return: a dictionary of revision_id->revision_parents_list.
1457
raise NotImplementedError(self.get_revision_graph)
1460
def get_revision_graph_with_ghosts(self, revision_ids=None):
1461
"""Return a graph of the revisions with ghosts marked as applicable.
1463
:param revision_ids: an iterable of revisions to graph or None for all.
1464
:return: a Graph object with the graph reachable from revision_ids.
1466
if 'evil' in debug.debug_flags:
1468
"get_revision_graph_with_ghosts scales with size of history.")
1469
result = deprecated_graph.Graph()
1470
if not revision_ids:
1471
pending = set(self.all_revision_ids())
1474
pending = set(revision_ids)
1475
# special case NULL_REVISION
1476
if _mod_revision.NULL_REVISION in pending:
1477
pending.remove(_mod_revision.NULL_REVISION)
1478
required = set(pending)
1481
revision_id = pending.pop()
1483
rev = self.get_revision(revision_id)
1484
except errors.NoSuchRevision:
1485
if revision_id in required:
1488
result.add_ghost(revision_id)
1490
for parent_id in rev.parent_ids:
1491
# is this queued or done ?
1492
if (parent_id not in pending and
1493
parent_id not in done):
1495
pending.add(parent_id)
1496
result.add_node(revision_id, rev.parent_ids)
1497
done.add(revision_id)
1500
def _get_history_vf(self):
1501
"""Get a versionedfile whose history graph reflects all revisions.
1503
For weave repositories, this is the inventory weave.
1505
return self.get_inventory_weave()
1507
def iter_reverse_revision_history(self, revision_id):
1508
"""Iterate backwards through revision ids in the lefthand history
1510
:param revision_id: The revision id to start with. All its lefthand
1511
ancestors will be traversed.
1513
if revision_id in (None, _mod_revision.NULL_REVISION):
1515
next_id = revision_id
1516
versionedfile = self._get_history_vf()
1519
parents = versionedfile.get_parents(next_id)
1520
if len(parents) == 0:
1523
next_id = parents[0]
1526
def get_revision_inventory(self, revision_id):
1527
"""Return inventory of a past revision."""
1528
# TODO: Unify this with get_inventory()
1529
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1530
# must be the same as its revision, so this is trivial.
1531
if revision_id is None:
1532
# This does not make sense: if there is no revision,
1533
# then it is the current tree inventory surely ?!
1534
# and thus get_root_id() is something that looks at the last
1535
# commit on the branch, and the get_root_id is an inventory check.
1536
raise NotImplementedError
1537
# return Inventory(self.get_root_id())
1539
return self.get_inventory(revision_id)
1542
def is_shared(self):
1543
"""Return True if this repository is flagged as a shared repository."""
1544
raise NotImplementedError(self.is_shared)
1547
def reconcile(self, other=None, thorough=False):
1548
"""Reconcile this repository."""
1549
from bzrlib.reconcile import RepoReconciler
1550
reconciler = RepoReconciler(self, thorough=thorough)
1551
reconciler.reconcile()
1554
def _refresh_data(self):
1555
"""Helper called from lock_* to ensure coherency with disk.
1557
The default implementation does nothing; it is however possible
1558
for repositories to maintain loaded indices across multiple locks
1559
by checking inside their implementation of this method to see
1560
whether their indices are still valid. This depends of course on
1561
the disk format being validatable in this manner.
1565
def revision_tree(self, revision_id):
1566
"""Return Tree for a revision on this branch.
1568
`revision_id` may be None for the empty tree revision.
1570
# TODO: refactor this to use an existing revision object
1571
# so we don't need to read it in twice.
1572
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1573
return RevisionTree(self, Inventory(root_id=None),
1574
_mod_revision.NULL_REVISION)
1576
inv = self.get_revision_inventory(revision_id)
1577
return RevisionTree(self, inv, revision_id)
1580
def revision_trees(self, revision_ids):
1581
"""Return Tree for a revision on this branch.
1583
`revision_id` may not be None or 'null:'"""
1584
assert None not in revision_ids
1585
assert _mod_revision.NULL_REVISION not in revision_ids
1586
texts = self.get_inventory_weave().get_texts(revision_ids)
1587
for text, revision_id in zip(texts, revision_ids):
1588
inv = self.deserialise_inventory(revision_id, text)
1589
yield RevisionTree(self, inv, revision_id)
1592
def get_ancestry(self, revision_id, topo_sorted=True):
1593
"""Return a list of revision-ids integrated by a revision.
1595
The first element of the list is always None, indicating the origin
1596
revision. This might change when we have history horizons, or
1597
perhaps we should have a new API.
1599
This is topologically sorted.
1601
if _mod_revision.is_null(revision_id):
1603
if not self.has_revision(revision_id):
1604
raise errors.NoSuchRevision(self, revision_id)
1605
w = self.get_inventory_weave()
1606
candidates = w.get_ancestry(revision_id, topo_sorted)
1607
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1610
"""Compress the data within the repository.
1612
This operation only makes sense for some repository types. For other
1613
types it should be a no-op that just returns.
1615
This stub method does not require a lock, but subclasses should use
1616
@needs_write_lock as this is a long running call its reasonable to
1617
implicitly lock for the user.
1621
def print_file(self, file, revision_id):
1622
"""Print `file` to stdout.
1624
FIXME RBC 20060125 as John Meinel points out this is a bad api
1625
- it writes to stdout, it assumes that that is valid etc. Fix
1626
by creating a new more flexible convenience function.
1628
tree = self.revision_tree(revision_id)
1629
# use inventory as it was in that revision
1630
file_id = tree.inventory.path2id(file)
1632
# TODO: jam 20060427 Write a test for this code path
1633
# it had a bug in it, and was raising the wrong
1635
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1636
tree.print_file(file_id)
1638
def get_transaction(self):
1639
return self.control_files.get_transaction()
1641
def revision_parents(self, revision_id):
1642
return self.get_inventory_weave().parent_names(revision_id)
1644
def get_parents(self, revision_ids):
1645
"""See StackedParentsProvider.get_parents"""
1647
for revision_id in revision_ids:
1648
if revision_id == _mod_revision.NULL_REVISION:
1652
parents = self.get_revision(revision_id).parent_ids
1653
except errors.NoSuchRevision:
1656
if len(parents) == 0:
1657
parents = [_mod_revision.NULL_REVISION]
1658
parents_list.append(parents)
1661
def _make_parents_provider(self):
1664
def get_graph(self, other_repository=None):
1665
"""Return the graph walker for this repository format"""
1666
parents_provider = self._make_parents_provider()
1667
if (other_repository is not None and
1668
other_repository.bzrdir.transport.base !=
1669
self.bzrdir.transport.base):
1670
parents_provider = graph._StackedParentsProvider(
1671
[parents_provider, other_repository._make_parents_provider()])
1672
return graph.Graph(parents_provider)
1674
def _get_versioned_file_checker(self):
1675
"""Return an object suitable for checking versioned files."""
1676
return _VersionedFileChecker(self)
1679
def set_make_working_trees(self, new_value):
1680
"""Set the policy flag for making working trees when creating branches.
1682
This only applies to branches that use this repository.
1684
The default is 'True'.
1685
:param new_value: True to restore the default, False to disable making
1688
raise NotImplementedError(self.set_make_working_trees)
1690
def make_working_trees(self):
1691
"""Returns the policy for making working trees on new branches."""
1692
raise NotImplementedError(self.make_working_trees)
1695
def sign_revision(self, revision_id, gpg_strategy):
1696
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1697
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1700
def has_signature_for_revision_id(self, revision_id):
1701
"""Query for a revision signature for revision_id in the repository."""
1702
return self._revision_store.has_signature(revision_id,
1703
self.get_transaction())
1706
def get_signature_text(self, revision_id):
1707
"""Return the text for a signature."""
1708
return self._revision_store.get_signature_text(revision_id,
1709
self.get_transaction())
1712
def check(self, revision_ids=None):
1713
"""Check consistency of all history of given revision_ids.
1715
Different repository implementations should override _check().
1717
:param revision_ids: A non-empty list of revision_ids whose ancestry
1718
will be checked. Typically the last revision_id of a branch.
1720
return self._check(revision_ids)
1722
def _check(self, revision_ids):
1723
result = check.Check(self)
1727
def _warn_if_deprecated(self):
1728
global _deprecation_warning_done
1729
if _deprecation_warning_done:
1731
_deprecation_warning_done = True
1732
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1733
% (self._format, self.bzrdir.transport.base))
1735
def supports_rich_root(self):
1736
return self._format.rich_root_data
1738
def _check_ascii_revisionid(self, revision_id, method):
1739
"""Private helper for ascii-only repositories."""
1740
# weave repositories refuse to store revisionids that are non-ascii.
1741
if revision_id is not None:
1742
# weaves require ascii revision ids.
1743
if isinstance(revision_id, unicode):
1745
revision_id.encode('ascii')
1746
except UnicodeEncodeError:
1747
raise errors.NonAsciiRevisionId(method, self)
1750
revision_id.decode('ascii')
1751
except UnicodeDecodeError:
1752
raise errors.NonAsciiRevisionId(method, self)
1754
def revision_graph_can_have_wrong_parents(self):
1755
"""Is it possible for this repository to have a revision graph with
1758
If True, then this repository must also implement
1759
_find_inconsistent_revision_parents so that check and reconcile can
1760
check for inconsistencies before proceeding with other checks that may
1761
depend on the revision index being consistent.
1763
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1765
# remove these delegates a while after bzr 0.15
1766
def __make_delegated(name, from_module):
1767
def _deprecated_repository_forwarder():
1768
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1769
% (name, from_module),
1772
m = __import__(from_module, globals(), locals(), [name])
1774
return getattr(m, name)
1775
except AttributeError:
1776
raise AttributeError('module %s has no name %s'
1778
globals()[name] = _deprecated_repository_forwarder
1781
'AllInOneRepository',
1782
'WeaveMetaDirRepository',
1783
'PreSplitOutRepositoryFormat',
1784
'RepositoryFormat4',
1785
'RepositoryFormat5',
1786
'RepositoryFormat6',
1787
'RepositoryFormat7',
1789
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1793
'RepositoryFormatKnit',
1794
'RepositoryFormatKnit1',
1796
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1799
def install_revision(repository, rev, revision_tree):
1800
"""Install all revision data into a repository."""
1801
install_revisions(repository, [(rev, revision_tree, None)])
1804
def install_revisions(repository, iterable):
1805
"""Install all revision data into a repository.
1807
Accepts an iterable of revision, tree, signature tuples. The signature
1810
repository.start_write_group()
1812
for revision, revision_tree, signature in iterable:
1813
_install_revision(repository, revision, revision_tree, signature)
1815
repository.abort_write_group()
1818
repository.commit_write_group()
1821
def _install_revision(repository, rev, revision_tree, signature):
1822
"""Install all revision data into a repository."""
1823
present_parents = []
1825
for p_id in rev.parent_ids:
1826
if repository.has_revision(p_id):
1827
present_parents.append(p_id)
1828
parent_trees[p_id] = repository.revision_tree(p_id)
1830
parent_trees[p_id] = repository.revision_tree(None)
1832
inv = revision_tree.inventory
1833
entries = inv.iter_entries()
1834
# backwards compatibility hack: skip the root id.
1835
if not repository.supports_rich_root():
1836
path, root = entries.next()
1837
if root.revision != rev.revision_id:
1838
raise errors.IncompatibleRevision(repr(repository))
1839
# Add the texts that are not already present
1840
for path, ie in entries:
1841
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1842
repository.get_transaction())
1843
if ie.revision not in w:
1845
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1846
# with InventoryEntry.find_previous_heads(). if it is, then there
1847
# is a latent bug here where the parents may have ancestors of each
1849
for revision, tree in parent_trees.iteritems():
1850
if ie.file_id not in tree:
1852
parent_id = tree.inventory[ie.file_id].revision
1853
if parent_id in text_parents:
1855
text_parents.append(parent_id)
1857
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1858
repository.get_transaction())
1859
lines = revision_tree.get_file(ie.file_id).readlines()
1860
vfile.add_lines(rev.revision_id, text_parents, lines)
1862
# install the inventory
1863
repository.add_inventory(rev.revision_id, inv, present_parents)
1864
except errors.RevisionAlreadyPresent:
1866
if signature is not None:
1867
repository.add_signature_text(rev.revision_id, signature)
1868
repository.add_revision(rev.revision_id, rev, inv)
1871
class MetaDirRepository(Repository):
1872
"""Repositories in the new meta-dir layout."""
1874
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1875
super(MetaDirRepository, self).__init__(_format,
1881
dir_mode = self.control_files._dir_mode
1882
file_mode = self.control_files._file_mode
1885
def is_shared(self):
1886
"""Return True if this repository is flagged as a shared repository."""
1887
return self.control_files._transport.has('shared-storage')
1890
def set_make_working_trees(self, new_value):
1891
"""Set the policy flag for making working trees when creating branches.
1893
This only applies to branches that use this repository.
1895
The default is 'True'.
1896
:param new_value: True to restore the default, False to disable making
1901
self.control_files._transport.delete('no-working-trees')
1902
except errors.NoSuchFile:
1905
self.control_files.put_utf8('no-working-trees', '')
1907
def make_working_trees(self):
1908
"""Returns the policy for making working trees on new branches."""
1909
return not self.control_files._transport.has('no-working-trees')
1912
class RepositoryFormatRegistry(registry.Registry):
1913
"""Registry of RepositoryFormats."""
1915
def get(self, format_string):
1916
r = registry.Registry.get(self, format_string)
1922
format_registry = RepositoryFormatRegistry()
1923
"""Registry of formats, indexed by their identifying format string.
1925
This can contain either format instances themselves, or classes/factories that
1926
can be called to obtain one.
1930
#####################################################################
1931
# Repository Formats
1933
class RepositoryFormat(object):
1934
"""A repository format.
1936
Formats provide three things:
1937
* An initialization routine to construct repository data on disk.
1938
* a format string which is used when the BzrDir supports versioned
1940
* an open routine which returns a Repository instance.
1942
There is one and only one Format subclass for each on-disk format. But
1943
there can be one Repository subclass that is used for several different
1944
formats. The _format attribute on a Repository instance can be used to
1945
determine the disk format.
1947
Formats are placed in an dict by their format string for reference
1948
during opening. These should be subclasses of RepositoryFormat
1951
Once a format is deprecated, just deprecate the initialize and open
1952
methods on the format class. Do not deprecate the object, as the
1953
object will be created every system load.
1955
Common instance attributes:
1956
_matchingbzrdir - the bzrdir format that the repository format was
1957
originally written to work with. This can be used if manually
1958
constructing a bzrdir and repository, or more commonly for test suite
1962
# Set to True or False in derived classes. True indicates that the format
1963
# supports ghosts gracefully.
1964
supports_ghosts = None
1967
return "<%s>" % self.__class__.__name__
1969
def __eq__(self, other):
1970
# format objects are generally stateless
1971
return isinstance(other, self.__class__)
1973
def __ne__(self, other):
1974
return not self == other
1977
def find_format(klass, a_bzrdir):
1978
"""Return the format for the repository object in a_bzrdir.
1980
This is used by bzr native formats that have a "format" file in
1981
the repository. Other methods may be used by different types of
1985
transport = a_bzrdir.get_repository_transport(None)
1986
format_string = transport.get("format").read()
1987
return format_registry.get(format_string)
1988
except errors.NoSuchFile:
1989
raise errors.NoRepositoryPresent(a_bzrdir)
1991
raise errors.UnknownFormatError(format=format_string)
1994
def register_format(klass, format):
1995
format_registry.register(format.get_format_string(), format)
1998
def unregister_format(klass, format):
1999
format_registry.remove(format.get_format_string())
2002
def get_default_format(klass):
2003
"""Return the current default format."""
2004
from bzrlib import bzrdir
2005
return bzrdir.format_registry.make_bzrdir('default').repository_format
2007
def _get_control_store(self, repo_transport, control_files):
2008
"""Return the control store for this repository."""
2009
raise NotImplementedError(self._get_control_store)
2011
def get_format_string(self):
2012
"""Return the ASCII format string that identifies this format.
2014
Note that in pre format ?? repositories the format string is
2015
not permitted nor written to disk.
2017
raise NotImplementedError(self.get_format_string)
2019
def get_format_description(self):
2020
"""Return the short description for this format."""
2021
raise NotImplementedError(self.get_format_description)
2023
def _get_revision_store(self, repo_transport, control_files):
2024
"""Return the revision store object for this a_bzrdir."""
2025
raise NotImplementedError(self._get_revision_store)
2027
def _get_text_rev_store(self,
2034
"""Common logic for getting a revision store for a repository.
2036
see self._get_revision_store for the subclass-overridable method to
2037
get the store for a repository.
2039
from bzrlib.store.revision.text import TextRevisionStore
2040
dir_mode = control_files._dir_mode
2041
file_mode = control_files._file_mode
2042
text_store = TextStore(transport.clone(name),
2044
compressed=compressed,
2046
file_mode=file_mode)
2047
_revision_store = TextRevisionStore(text_store, serializer)
2048
return _revision_store
2050
# TODO: this shouldn't be in the base class, it's specific to things that
2051
# use weaves or knits -- mbp 20070207
2052
def _get_versioned_file_store(self,
2057
versionedfile_class=None,
2058
versionedfile_kwargs={},
2060
if versionedfile_class is None:
2061
versionedfile_class = self._versionedfile_class
2062
weave_transport = control_files._transport.clone(name)
2063
dir_mode = control_files._dir_mode
2064
file_mode = control_files._file_mode
2065
return VersionedFileStore(weave_transport, prefixed=prefixed,
2067
file_mode=file_mode,
2068
versionedfile_class=versionedfile_class,
2069
versionedfile_kwargs=versionedfile_kwargs,
2072
def initialize(self, a_bzrdir, shared=False):
2073
"""Initialize a repository of this format in a_bzrdir.
2075
:param a_bzrdir: The bzrdir to put the new repository in it.
2076
:param shared: The repository should be initialized as a sharable one.
2077
:returns: The new repository object.
2079
This may raise UninitializableFormat if shared repository are not
2080
compatible the a_bzrdir.
2082
raise NotImplementedError(self.initialize)
2084
def is_supported(self):
2085
"""Is this format supported?
2087
Supported formats must be initializable and openable.
2088
Unsupported formats may not support initialization or committing or
2089
some other features depending on the reason for not being supported.
2093
def check_conversion_target(self, target_format):
2094
raise NotImplementedError(self.check_conversion_target)
2096
def open(self, a_bzrdir, _found=False):
2097
"""Return an instance of this format for the bzrdir a_bzrdir.
2099
_found is a private parameter, do not use it.
2101
raise NotImplementedError(self.open)
2104
class MetaDirRepositoryFormat(RepositoryFormat):
2105
"""Common base class for the new repositories using the metadir layout."""
2107
rich_root_data = False
2108
supports_tree_reference = False
2109
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
2112
super(MetaDirRepositoryFormat, self).__init__()
2114
def _create_control_files(self, a_bzrdir):
2115
"""Create the required files and the initial control_files object."""
2116
# FIXME: RBC 20060125 don't peek under the covers
2117
# NB: no need to escape relative paths that are url safe.
2118
repository_transport = a_bzrdir.get_repository_transport(self)
2119
control_files = lockable_files.LockableFiles(repository_transport,
2120
'lock', lockdir.LockDir)
2121
control_files.create_lock()
2122
return control_files
2124
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2125
"""Upload the initial blank content."""
2126
control_files = self._create_control_files(a_bzrdir)
2127
control_files.lock_write()
2129
control_files._transport.mkdir_multi(dirs,
2130
mode=control_files._dir_mode)
2131
for file, content in files:
2132
control_files.put(file, content)
2133
for file, content in utf8_files:
2134
control_files.put_utf8(file, content)
2136
control_files.put_utf8('shared-storage', '')
2138
control_files.unlock()
2141
# formats which have no format string are not discoverable
2142
# and not independently creatable, so are not registered. They're
2143
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2144
# needed, it's constructed directly by the BzrDir. Non-native formats where
2145
# the repository is not separately opened are similar.
2147
format_registry.register_lazy(
2148
'Bazaar-NG Repository format 7',
2149
'bzrlib.repofmt.weaverepo',
2153
format_registry.register_lazy(
2154
'Bazaar-NG Knit Repository Format 1',
2155
'bzrlib.repofmt.knitrepo',
2156
'RepositoryFormatKnit1',
2159
format_registry.register_lazy(
2160
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2161
'bzrlib.repofmt.knitrepo',
2162
'RepositoryFormatKnit3',
2165
format_registry.register_lazy(
2166
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2167
'bzrlib.repofmt.knitrepo',
2168
'RepositoryFormatKnit4',
2171
# Pack-based formats. There is one format for pre-subtrees, and one for
2172
# post-subtrees to allow ease of testing.
2173
# NOTE: These are experimental in 0.92.
2174
format_registry.register_lazy(
2175
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2176
'bzrlib.repofmt.pack_repo',
2177
'RepositoryFormatKnitPack1',
2179
format_registry.register_lazy(
2180
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2181
'bzrlib.repofmt.pack_repo',
2182
'RepositoryFormatKnitPack3',
2184
format_registry.register_lazy(
2185
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2186
'bzrlib.repofmt.pack_repo',
2187
'RepositoryFormatKnitPack4',
2191
class InterRepository(InterObject):
2192
"""This class represents operations taking place between two repositories.
2194
Its instances have methods like copy_content and fetch, and contain
2195
references to the source and target repositories these operations can be
2198
Often we will provide convenience methods on 'repository' which carry out
2199
operations with another repository - they will always forward to
2200
InterRepository.get(other).method_name(parameters).
2204
"""The available optimised InterRepository types."""
2206
def copy_content(self, revision_id=None):
2207
raise NotImplementedError(self.copy_content)
2209
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2210
"""Fetch the content required to construct revision_id.
2212
The content is copied from self.source to self.target.
2214
:param revision_id: if None all content is copied, if NULL_REVISION no
2216
:param pb: optional progress bar to use for progress reports. If not
2217
provided a default one will be created.
2219
Returns the copied revision count and the failed revisions in a tuple:
2222
raise NotImplementedError(self.fetch)
2225
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2226
"""Return the revision ids that source has that target does not.
2228
These are returned in topological order.
2230
:param revision_id: only return revision ids included by this
2233
# generic, possibly worst case, slow code path.
2234
target_ids = set(self.target.all_revision_ids())
2235
if revision_id is not None:
2236
source_ids = self.source.get_ancestry(revision_id)
2237
assert source_ids[0] is None
2240
source_ids = self.source.all_revision_ids()
2241
result_set = set(source_ids).difference(target_ids)
2242
# this may look like a no-op: its not. It preserves the ordering
2243
# other_ids had while only returning the members from other_ids
2244
# that we've decided we need.
2245
return [rev_id for rev_id in source_ids if rev_id in result_set]
2248
def _same_model(source, target):
2249
"""True if source and target have the same data representation."""
2250
if source.supports_rich_root() != target.supports_rich_root():
2252
if source._serializer != target._serializer:
2257
class InterSameDataRepository(InterRepository):
2258
"""Code for converting between repositories that represent the same data.
2260
Data format and model must match for this to work.
2264
def _get_repo_format_to_test(self):
2265
"""Repository format for testing with.
2267
InterSameData can pull from subtree to subtree and from non-subtree to
2268
non-subtree, so we test this with the richest repository format.
2270
from bzrlib.repofmt import knitrepo
2271
return knitrepo.RepositoryFormatKnit3()
2274
def is_compatible(source, target):
2275
return InterRepository._same_model(source, target)
2278
def copy_content(self, revision_id=None):
2279
"""Make a complete copy of the content in self into destination.
2281
This copies both the repository's revision data, and configuration information
2282
such as the make_working_trees setting.
2284
This is a destructive operation! Do not use it on existing
2287
:param revision_id: Only copy the content needed to construct
2288
revision_id and its parents.
2291
self.target.set_make_working_trees(self.source.make_working_trees())
2292
except NotImplementedError:
2294
# but don't bother fetching if we have the needed data now.
2295
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2296
self.target.has_revision(revision_id)):
2298
self.target.fetch(self.source, revision_id=revision_id)
2301
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2302
"""See InterRepository.fetch()."""
2303
from bzrlib.fetch import GenericRepoFetcher
2304
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2305
self.source, self.source._format, self.target,
2306
self.target._format)
2307
f = GenericRepoFetcher(to_repository=self.target,
2308
from_repository=self.source,
2309
last_revision=revision_id,
2311
return f.count_copied, f.failed_revisions
2314
class InterWeaveRepo(InterSameDataRepository):
2315
"""Optimised code paths between Weave based repositories.
2317
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2318
implemented lazy inter-object optimisation.
2322
def _get_repo_format_to_test(self):
2323
from bzrlib.repofmt import weaverepo
2324
return weaverepo.RepositoryFormat7()
2327
def is_compatible(source, target):
2328
"""Be compatible with known Weave formats.
2330
We don't test for the stores being of specific types because that
2331
could lead to confusing results, and there is no need to be
2334
from bzrlib.repofmt.weaverepo import (
2340
return (isinstance(source._format, (RepositoryFormat5,
2342
RepositoryFormat7)) and
2343
isinstance(target._format, (RepositoryFormat5,
2345
RepositoryFormat7)))
2346
except AttributeError:
2350
def copy_content(self, revision_id=None):
2351
"""See InterRepository.copy_content()."""
2352
# weave specific optimised path:
2354
self.target.set_make_working_trees(self.source.make_working_trees())
2355
except NotImplementedError:
2357
# FIXME do not peek!
2358
if self.source.control_files._transport.listable():
2359
pb = ui.ui_factory.nested_progress_bar()
2361
self.target.weave_store.copy_all_ids(
2362
self.source.weave_store,
2364
from_transaction=self.source.get_transaction(),
2365
to_transaction=self.target.get_transaction())
2366
pb.update('copying inventory', 0, 1)
2367
self.target.control_weaves.copy_multi(
2368
self.source.control_weaves, ['inventory'],
2369
from_transaction=self.source.get_transaction(),
2370
to_transaction=self.target.get_transaction())
2371
self.target._revision_store.text_store.copy_all_ids(
2372
self.source._revision_store.text_store,
2377
self.target.fetch(self.source, revision_id=revision_id)
2380
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2381
"""See InterRepository.fetch()."""
2382
from bzrlib.fetch import GenericRepoFetcher
2383
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2384
self.source, self.source._format, self.target, self.target._format)
2385
f = GenericRepoFetcher(to_repository=self.target,
2386
from_repository=self.source,
2387
last_revision=revision_id,
2389
return f.count_copied, f.failed_revisions
2392
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2393
"""See InterRepository.missing_revision_ids()."""
2394
# we want all revisions to satisfy revision_id in source.
2395
# but we don't want to stat every file here and there.
2396
# we want then, all revisions other needs to satisfy revision_id
2397
# checked, but not those that we have locally.
2398
# so the first thing is to get a subset of the revisions to
2399
# satisfy revision_id in source, and then eliminate those that
2400
# we do already have.
2401
# this is slow on high latency connection to self, but as as this
2402
# disk format scales terribly for push anyway due to rewriting
2403
# inventory.weave, this is considered acceptable.
2405
if revision_id is not None:
2406
source_ids = self.source.get_ancestry(revision_id)
2407
assert source_ids[0] is None
2410
source_ids = self.source._all_possible_ids()
2411
source_ids_set = set(source_ids)
2412
# source_ids is the worst possible case we may need to pull.
2413
# now we want to filter source_ids against what we actually
2414
# have in target, but don't try to check for existence where we know
2415
# we do not have a revision as that would be pointless.
2416
target_ids = set(self.target._all_possible_ids())
2417
possibly_present_revisions = target_ids.intersection(source_ids_set)
2418
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2419
required_revisions = source_ids_set.difference(actually_present_revisions)
2420
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2421
if revision_id is not None:
2422
# we used get_ancestry to determine source_ids then we are assured all
2423
# revisions referenced are present as they are installed in topological order.
2424
# and the tip revision was validated by get_ancestry.
2425
return required_topo_revisions
2427
# if we just grabbed the possibly available ids, then
2428
# we only have an estimate of whats available and need to validate
2429
# that against the revision records.
2430
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2433
class InterKnitRepo(InterSameDataRepository):
2434
"""Optimised code paths between Knit based repositories."""
2437
def _get_repo_format_to_test(self):
2438
from bzrlib.repofmt import knitrepo
2439
return knitrepo.RepositoryFormatKnit1()
2442
def is_compatible(source, target):
2443
"""Be compatible with known Knit formats.
2445
We don't test for the stores being of specific types because that
2446
could lead to confusing results, and there is no need to be
2449
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2451
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2452
isinstance(target._format, RepositoryFormatKnit))
2453
except AttributeError:
2455
return are_knits and InterRepository._same_model(source, target)
2458
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2459
"""See InterRepository.fetch()."""
2460
from bzrlib.fetch import KnitRepoFetcher
2461
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2462
self.source, self.source._format, self.target, self.target._format)
2463
f = KnitRepoFetcher(to_repository=self.target,
2464
from_repository=self.source,
2465
last_revision=revision_id,
2467
return f.count_copied, f.failed_revisions
2470
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2471
"""See InterRepository.missing_revision_ids()."""
2472
if revision_id is not None:
2473
source_ids = self.source.get_ancestry(revision_id)
2474
assert source_ids[0] is None
2477
source_ids = self.source.all_revision_ids()
2478
source_ids_set = set(source_ids)
2479
# source_ids is the worst possible case we may need to pull.
2480
# now we want to filter source_ids against what we actually
2481
# have in target, but don't try to check for existence where we know
2482
# we do not have a revision as that would be pointless.
2483
target_ids = set(self.target.all_revision_ids())
2484
possibly_present_revisions = target_ids.intersection(source_ids_set)
2485
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2486
required_revisions = source_ids_set.difference(actually_present_revisions)
2487
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2488
if revision_id is not None:
2489
# we used get_ancestry to determine source_ids then we are assured all
2490
# revisions referenced are present as they are installed in topological order.
2491
# and the tip revision was validated by get_ancestry.
2492
return required_topo_revisions
2494
# if we just grabbed the possibly available ids, then
2495
# we only have an estimate of whats available and need to validate
2496
# that against the revision records.
2497
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2500
class InterPackRepo(InterSameDataRepository):
2501
"""Optimised code paths between Pack based repositories."""
2504
def _get_repo_format_to_test(self):
2505
from bzrlib.repofmt import pack_repo
2506
return pack_repo.RepositoryFormatKnitPack1()
2509
def is_compatible(source, target):
2510
"""Be compatible with known Pack formats.
2512
We don't test for the stores being of specific types because that
2513
could lead to confusing results, and there is no need to be
2516
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2518
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2519
isinstance(target._format, RepositoryFormatPack))
2520
except AttributeError:
2522
return are_packs and InterRepository._same_model(source, target)
2525
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2526
"""See InterRepository.fetch()."""
2527
from bzrlib.repofmt.pack_repo import Packer
2528
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2529
self.source, self.source._format, self.target, self.target._format)
2530
self.count_copied = 0
2531
if revision_id is None:
2533
# everything to do - use pack logic
2534
# to fetch from all packs to one without
2535
# inventory parsing etc, IFF nothing to be copied is in the target.
2537
revision_ids = self.source.all_revision_ids()
2538
# implementing the TODO will involve:
2539
# - detecting when all of a pack is selected
2540
# - avoiding as much as possible pre-selection, so the
2541
# more-core routines such as create_pack_from_packs can filter in
2542
# a just-in-time fashion. (though having a HEADS list on a
2543
# repository might make this a lot easier, because we could
2544
# sensibly detect 'new revisions' without doing a full index scan.
2545
elif _mod_revision.is_null(revision_id):
2550
revision_ids = self.missing_revision_ids(revision_id,
2551
find_ghosts=find_ghosts)
2552
except errors.NoSuchRevision:
2553
raise errors.InstallFailed([revision_id])
2554
packs = self.source._pack_collection.all_packs()
2555
pack = Packer(self.target._pack_collection, packs, '.fetch',
2556
revision_ids).pack()
2557
if pack is not None:
2558
self.target._pack_collection._save_pack_names()
2559
# Trigger an autopack. This may duplicate effort as we've just done
2560
# a pack creation, but for now it is simpler to think about as
2561
# 'upload data, then repack if needed'.
2562
self.target._pack_collection.autopack()
2563
return (pack.get_revision_count(), [])
2568
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2569
"""See InterRepository.missing_revision_ids().
2571
:param find_ghosts: Find ghosts throughough the ancestry of
2574
if not find_ghosts and revision_id is not None:
2575
graph = self.source.get_graph()
2576
missing_revs = set()
2577
searcher = graph._make_breadth_first_searcher([revision_id])
2579
self.target._pack_collection.revision_index.combined_index
2580
null_set = frozenset([_mod_revision.NULL_REVISION])
2583
next_revs = set(searcher.next())
2584
except StopIteration:
2586
next_revs.difference_update(null_set)
2587
target_keys = [(key,) for key in next_revs]
2588
have_revs = frozenset(node[1][0] for node in
2589
target_index.iter_entries(target_keys))
2590
missing_revs.update(next_revs - have_revs)
2591
searcher.stop_searching_any(have_revs)
2592
if next_revs - have_revs == set([revision_id]):
2593
# we saw the start rev itself, but no parents from it (or
2594
# next_revs would have been updated to e.g. set(). We remove
2595
# have_revs because if we found revision_id locally we
2596
# stop_searching at the first time around.
2597
raise errors.NoSuchRevision(self.source, revision_id)
2599
elif revision_id is not None:
2600
source_ids = self.source.get_ancestry(revision_id)
2601
assert source_ids[0] is None
2604
source_ids = self.source.all_revision_ids()
2605
# source_ids is the worst possible case we may need to pull.
2606
# now we want to filter source_ids against what we actually
2607
# have in target, but don't try to check for existence where we know
2608
# we do not have a revision as that would be pointless.
2609
target_ids = set(self.target.all_revision_ids())
2610
return [r for r in source_ids if (r not in target_ids)]
2613
class InterModel1and2(InterRepository):
2616
def _get_repo_format_to_test(self):
2620
def is_compatible(source, target):
2621
if not source.supports_rich_root() and target.supports_rich_root():
2627
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2628
"""See InterRepository.fetch()."""
2629
from bzrlib.fetch import Model1toKnit2Fetcher
2630
f = Model1toKnit2Fetcher(to_repository=self.target,
2631
from_repository=self.source,
2632
last_revision=revision_id,
2634
return f.count_copied, f.failed_revisions
2637
def copy_content(self, revision_id=None):
2638
"""Make a complete copy of the content in self into destination.
2640
This is a destructive operation! Do not use it on existing
2643
:param revision_id: Only copy the content needed to construct
2644
revision_id and its parents.
2647
self.target.set_make_working_trees(self.source.make_working_trees())
2648
except NotImplementedError:
2650
# but don't bother fetching if we have the needed data now.
2651
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2652
self.target.has_revision(revision_id)):
2654
self.target.fetch(self.source, revision_id=revision_id)
2657
class InterKnit1and2(InterKnitRepo):
2660
def _get_repo_format_to_test(self):
2664
def is_compatible(source, target):
2665
"""Be compatible with Knit1 source and Knit3 target"""
2666
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2668
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2669
RepositoryFormatKnit3)
2670
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2671
RepositoryFormatKnitPack3)
2672
return (isinstance(source._format,
2673
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2674
isinstance(target._format,
2675
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2677
except AttributeError:
2681
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2682
"""See InterRepository.fetch()."""
2683
from bzrlib.fetch import Knit1to2Fetcher
2684
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2685
self.source, self.source._format, self.target,
2686
self.target._format)
2687
f = Knit1to2Fetcher(to_repository=self.target,
2688
from_repository=self.source,
2689
last_revision=revision_id,
2691
return f.count_copied, f.failed_revisions
2694
class InterDifferingSerializer(InterKnitRepo):
2697
def _get_repo_format_to_test(self):
2701
def is_compatible(source, target):
2702
"""Be compatible with Knit2 source and Knit3 target"""
2703
if source.supports_rich_root() != target.supports_rich_root():
2705
# Ideally, we'd support fetching if the source had no tree references
2706
# even if it supported them...
2707
if (getattr(source, '_format.supports_tree_reference', False) and
2708
not getattr(target, '_format.supports_tree_reference', False)):
2713
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2714
"""See InterRepository.fetch()."""
2715
revision_ids = self.target.missing_revision_ids(self.source,
2717
def revisions_iterator():
2718
for current_revision_id in revision_ids:
2719
revision = self.source.get_revision(current_revision_id)
2720
tree = self.source.revision_tree(current_revision_id)
2722
signature = self.source.get_signature_text(
2723
current_revision_id)
2724
except errors.NoSuchRevision:
2726
yield revision, tree, signature
2727
install_revisions(self.target, revisions_iterator())
2728
return len(revision_ids), 0
2731
class InterRemoteToOther(InterRepository):
2733
def __init__(self, source, target):
2734
InterRepository.__init__(self, source, target)
2735
self._real_inter = None
2738
def is_compatible(source, target):
2739
if not isinstance(source, remote.RemoteRepository):
2741
source._ensure_real()
2742
real_source = source._real_repository
2743
# Is source's model compatible with target's model, and are they the
2744
# same format? Currently we can only optimise fetching from an
2745
# identical model & format repo.
2746
assert not isinstance(real_source, remote.RemoteRepository), (
2747
"We don't support remote repos backed by remote repos yet.")
2748
return real_source._format == target._format
2751
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2752
"""See InterRepository.fetch()."""
2753
from bzrlib.fetch import RemoteToOtherFetcher
2754
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2755
self.source, self.target, self.target._format)
2756
# TODO: jam 20070210 This should be an assert, not a translate
2757
revision_id = osutils.safe_revision_id(revision_id)
2758
f = RemoteToOtherFetcher(to_repository=self.target,
2759
from_repository=self.source,
2760
last_revision=revision_id,
2762
return f.count_copied, f.failed_revisions
2765
def _get_repo_format_to_test(self):
2769
class InterOtherToRemote(InterRepository):
2771
def __init__(self, source, target):
2772
InterRepository.__init__(self, source, target)
2773
self._real_inter = None
2776
def is_compatible(source, target):
2777
if isinstance(target, remote.RemoteRepository):
2781
def _ensure_real_inter(self):
2782
if self._real_inter is None:
2783
self.target._ensure_real()
2784
real_target = self.target._real_repository
2785
self._real_inter = InterRepository.get(self.source, real_target)
2787
def copy_content(self, revision_id=None):
2788
self._ensure_real_inter()
2789
self._real_inter.copy_content(revision_id=revision_id)
2791
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2792
self._ensure_real_inter()
2793
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2796
def _get_repo_format_to_test(self):
2800
InterRepository.register_optimiser(InterDifferingSerializer)
2801
InterRepository.register_optimiser(InterSameDataRepository)
2802
InterRepository.register_optimiser(InterWeaveRepo)
2803
InterRepository.register_optimiser(InterKnitRepo)
2804
InterRepository.register_optimiser(InterModel1and2)
2805
InterRepository.register_optimiser(InterKnit1and2)
2806
InterRepository.register_optimiser(InterPackRepo)
2807
InterRepository.register_optimiser(InterRemoteToOther)
2808
InterRepository.register_optimiser(InterOtherToRemote)
2811
class CopyConverter(object):
2812
"""A repository conversion tool which just performs a copy of the content.
2814
This is slow but quite reliable.
2817
def __init__(self, target_format):
2818
"""Create a CopyConverter.
2820
:param target_format: The format the resulting repository should be.
2822
self.target_format = target_format
2824
def convert(self, repo, pb):
2825
"""Perform the conversion of to_convert, giving feedback via pb.
2827
:param to_convert: The disk object to convert.
2828
:param pb: a progress bar to use for progress information.
2833
# this is only useful with metadir layouts - separated repo content.
2834
# trigger an assertion if not such
2835
repo._format.get_format_string()
2836
self.repo_dir = repo.bzrdir
2837
self.step('Moving repository to repository.backup')
2838
self.repo_dir.transport.move('repository', 'repository.backup')
2839
backup_transport = self.repo_dir.transport.clone('repository.backup')
2840
repo._format.check_conversion_target(self.target_format)
2841
self.source_repo = repo._format.open(self.repo_dir,
2843
_override_transport=backup_transport)
2844
self.step('Creating new repository')
2845
converted = self.target_format.initialize(self.repo_dir,
2846
self.source_repo.is_shared())
2847
converted.lock_write()
2849
self.step('Copying content into repository.')
2850
self.source_repo.copy_content_into(converted)
2853
self.step('Deleting old repository content.')
2854
self.repo_dir.transport.delete_tree('repository.backup')
2855
self.pb.note('repository converted')
2857
def step(self, message):
2858
"""Update the pb by a step."""
2860
self.pb.update(message, self.count, self.total)
2872
def _unescaper(match, _map=_unescape_map):
2873
code = match.group(1)
2877
if not code.startswith('#'):
2879
return unichr(int(code[1:])).encode('utf8')
2885
def _unescape_xml(data):
2886
"""Unescape predefined XML entities in a string of data."""
2888
if _unescape_re is None:
2889
_unescape_re = re.compile('\&([^;]*);')
2890
return _unescape_re.sub(_unescaper, data)
2893
class _VersionedFileChecker(object):
2895
def __init__(self, repository):
2896
self.repository = repository
2897
self.text_index = self.repository._generate_text_key_index()
2899
def calculate_file_version_parents(self, revision_id, file_id):
2900
"""Calculate the correct parents for a file version according to
2903
parent_keys = self.text_index[(file_id, revision_id)]
2904
if parent_keys == [_mod_revision.NULL_REVISION]:
2906
# strip the file_id, for the weave api
2907
return tuple([revision_id for file_id, revision_id in parent_keys])
2909
def check_file_version_parents(self, weave, file_id):
2910
"""Check the parents stored in a versioned file are correct.
2912
It also detects file versions that are not referenced by their
2913
corresponding revision's inventory.
2915
:returns: A tuple of (wrong_parents, dangling_file_versions).
2916
wrong_parents is a dict mapping {revision_id: (stored_parents,
2917
correct_parents)} for each revision_id where the stored parents
2918
are not correct. dangling_file_versions is a set of (file_id,
2919
revision_id) tuples for versions that are present in this versioned
2920
file, but not used by the corresponding inventory.
2923
unused_versions = set()
2924
for num, revision_id in enumerate(weave.versions()):
2926
correct_parents = self.calculate_file_version_parents(
2927
revision_id, file_id)
2929
# The version is not part of the used keys.
2930
unused_versions.add(revision_id)
2933
knit_parents = tuple(weave.get_parents(revision_id))
2934
except errors.RevisionNotPresent:
2936
if correct_parents != knit_parents:
2937
wrong_parents[revision_id] = (knit_parents, correct_parents)
2938
return wrong_parents, unused_versions
added
removed
bzrlib/repository.py
