1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from cStringIO import StringIO
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
39
revision as _mod_revision,
44
from bzrlib.bundle import serializer
45
from bzrlib.revisiontree import RevisionTree
46
from bzrlib.store.versioned import VersionedFileStore
47
from bzrlib.store.text import TextStore
48
from bzrlib.testament import Testament
49
from bzrlib.util import bencode
52
from bzrlib.decorators import needs_read_lock, needs_write_lock
53
from bzrlib.inter import InterObject
54
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
55
from bzrlib.symbol_versioning import (
58
from bzrlib.trace import mutter, mutter_callsite, note, warning
61
# Old formats display a warning, but only once
62
_deprecation_warning_done = False
65
class CommitBuilder(object):
66
"""Provides an interface to build up a commit.
68
This allows describing a tree to be committed without needing to
69
know the internals of the format of the repository.
72
# all clients should supply tree roots.
73
record_root_entry = True
74
# the default CommitBuilder does not manage trees whose root is versioned.
75
_versioned_root = False
77
def __init__(self, repository, parents, config, timestamp=None,
78
timezone=None, committer=None, revprops=None,
80
"""Initiate a CommitBuilder.
82
:param repository: Repository to commit to.
83
:param parents: Revision ids of the parents of the new revision.
84
:param config: Configuration to use.
85
:param timestamp: Optional timestamp recorded for commit.
86
:param timezone: Optional timezone for timestamp.
87
:param committer: Optional committer to set for commit.
88
:param revprops: Optional dictionary of revision properties.
89
:param revision_id: Optional revision id.
94
self._committer = self._config.username()
96
assert isinstance(committer, basestring), type(committer)
97
self._committer = committer
99
self.new_inventory = Inventory(None)
100
self._new_revision_id = revision_id
101
self.parents = parents
102
self.repository = repository
105
if revprops is not None:
106
self._revprops.update(revprops)
108
if timestamp is None:
109
timestamp = time.time()
110
# Restrict resolution to 1ms
111
self._timestamp = round(timestamp, 3)
114
self._timezone = osutils.local_time_offset()
116
self._timezone = int(timezone)
118
self._generate_revision_if_needed()
119
self._heads = graph.HeadsCache(repository.get_graph()).heads
121
def commit(self, message):
122
"""Make the actual commit.
124
:return: The revision id of the recorded revision.
126
rev = _mod_revision.Revision(
127
timestamp=self._timestamp,
128
timezone=self._timezone,
129
committer=self._committer,
131
inventory_sha1=self.inv_sha1,
132
revision_id=self._new_revision_id,
133
properties=self._revprops)
134
rev.parent_ids = self.parents
135
self.repository.add_revision(self._new_revision_id, rev,
136
self.new_inventory, self._config)
137
self.repository.commit_write_group()
138
return self._new_revision_id
141
"""Abort the commit that is being built.
143
self.repository.abort_write_group()
145
def revision_tree(self):
146
"""Return the tree that was just committed.
148
After calling commit() this can be called to get a RevisionTree
149
representing the newly committed tree. This is preferred to
150
calling Repository.revision_tree() because that may require
151
deserializing the inventory, while we already have a copy in
154
return RevisionTree(self.repository, self.new_inventory,
155
self._new_revision_id)
157
def finish_inventory(self):
158
"""Tell the builder that the inventory is finished."""
159
if self.new_inventory.root is None:
160
raise AssertionError('Root entry should be supplied to'
161
' record_entry_contents, as of bzr 0.10.',
162
DeprecationWarning, stacklevel=2)
163
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
164
self.new_inventory.revision_id = self._new_revision_id
165
self.inv_sha1 = self.repository.add_inventory(
166
self._new_revision_id,
171
def _gen_revision_id(self):
172
"""Return new revision-id."""
173
return generate_ids.gen_revision_id(self._config.username(),
176
def _generate_revision_if_needed(self):
177
"""Create a revision id if None was supplied.
179
If the repository can not support user-specified revision ids
180
they should override this function and raise CannotSetRevisionId
181
if _new_revision_id is not None.
183
:raises: CannotSetRevisionId
185
if self._new_revision_id is None:
186
self._new_revision_id = self._gen_revision_id()
187
self.random_revid = True
189
self.random_revid = False
191
def _check_root(self, ie, parent_invs, tree):
192
"""Helper for record_entry_contents.
194
:param ie: An entry being added.
195
:param parent_invs: The inventories of the parent revisions of the
197
:param tree: The tree that is being committed.
199
# In this revision format, root entries have no knit or weave When
200
# serializing out to disk and back in root.revision is always
202
ie.revision = self._new_revision_id
204
def _get_delta(self, ie, basis_inv, path):
205
"""Get a delta against the basis inventory for ie."""
206
if ie.file_id not in basis_inv:
208
return (None, path, ie.file_id, ie)
209
elif ie != basis_inv[ie.file_id]:
211
# TODO: avoid tis id2path call.
212
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
217
def record_entry_contents(self, ie, parent_invs, path, tree,
219
"""Record the content of ie from tree into the commit if needed.
221
Side effect: sets ie.revision when unchanged
223
:param ie: An inventory entry present in the commit.
224
:param parent_invs: The inventories of the parent revisions of the
226
:param path: The path the entry is at in the tree.
227
:param tree: The tree which contains this entry and should be used to
229
:param content_summary: Summary data from the tree about the paths
230
content - stat, length, exec, sha/link target. This is only
231
accessed when the entry has a revision of None - that is when it is
232
a candidate to commit.
233
:return: A tuple (change_delta, version_recorded). change_delta is
234
an inventory_delta change for this entry against the basis tree of
235
the commit, or None if no change occured against the basis tree.
236
version_recorded is True if a new version of the entry has been
237
recorded. For instance, committing a merge where a file was only
238
changed on the other side will return (delta, False).
240
if self.new_inventory.root is None:
241
if ie.parent_id is not None:
242
raise errors.RootMissing()
243
self._check_root(ie, parent_invs, tree)
244
if ie.revision is None:
245
kind = content_summary[0]
247
# ie is carried over from a prior commit
249
# XXX: repository specific check for nested tree support goes here - if
250
# the repo doesn't want nested trees we skip it ?
251
if (kind == 'tree-reference' and
252
not self.repository._format.supports_tree_reference):
253
# mismatch between commit builder logic and repository:
254
# this needs the entry creation pushed down into the builder.
255
raise NotImplementedError('Missing repository subtree support.')
256
self.new_inventory.add(ie)
258
# TODO: slow, take it out of the inner loop.
260
basis_inv = parent_invs[0]
262
basis_inv = Inventory(root_id=None)
264
# ie.revision is always None if the InventoryEntry is considered
265
# for committing. We may record the previous parents revision if the
266
# content is actually unchanged against a sole head.
267
if ie.revision is not None:
268
if not self._versioned_root and path == '':
269
# repositories that do not version the root set the root's
270
# revision to the new commit even when no change occurs, and
271
# this masks when a change may have occurred against the basis,
272
# so calculate if one happened.
273
if ie.file_id in basis_inv:
274
delta = (basis_inv.id2path(ie.file_id), path,
278
delta = (None, path, ie.file_id, ie)
281
# we don't need to commit this, because the caller already
282
# determined that an existing revision of this file is
284
return None, (ie.revision == self._new_revision_id)
285
# XXX: Friction: parent_candidates should return a list not a dict
286
# so that we don't have to walk the inventories again.
287
parent_candiate_entries = ie.parent_candidates(parent_invs)
288
head_set = self._heads(parent_candiate_entries.keys())
290
for inv in parent_invs:
291
if ie.file_id in inv:
292
old_rev = inv[ie.file_id].revision
293
if old_rev in head_set:
294
heads.append(inv[ie.file_id].revision)
295
head_set.remove(inv[ie.file_id].revision)
298
# now we check to see if we need to write a new record to the
300
# We write a new entry unless there is one head to the ancestors, and
301
# the kind-derived content is unchanged.
303
# Cheapest check first: no ancestors, or more the one head in the
304
# ancestors, we write a new node.
308
# There is a single head, look it up for comparison
309
parent_entry = parent_candiate_entries[heads[0]]
310
# if the non-content specific data has changed, we'll be writing a
312
if (parent_entry.parent_id != ie.parent_id or
313
parent_entry.name != ie.name):
315
# now we need to do content specific checks:
317
# if the kind changed the content obviously has
318
if kind != parent_entry.kind:
321
assert content_summary[2] is not None, \
322
"Files must not have executable = None"
324
if (# if the file length changed we have to store:
325
parent_entry.text_size != content_summary[1] or
326
# if the exec bit has changed we have to store:
327
parent_entry.executable != content_summary[2]):
329
elif parent_entry.text_sha1 == content_summary[3]:
330
# all meta and content is unchanged (using a hash cache
331
# hit to check the sha)
332
ie.revision = parent_entry.revision
333
ie.text_size = parent_entry.text_size
334
ie.text_sha1 = parent_entry.text_sha1
335
ie.executable = parent_entry.executable
336
return self._get_delta(ie, basis_inv, path), False
338
# Either there is only a hash change(no hash cache entry,
339
# or same size content change), or there is no change on
341
# Provide the parent's hash to the store layer, so that the
342
# content is unchanged we will not store a new node.
343
nostore_sha = parent_entry.text_sha1
345
# We want to record a new node regardless of the presence or
346
# absence of a content change in the file.
348
ie.executable = content_summary[2]
349
lines = tree.get_file(ie.file_id, path).readlines()
351
ie.text_sha1, ie.text_size = self._add_text_to_weave(
352
ie.file_id, lines, heads, nostore_sha)
353
except errors.ExistingContent:
354
# Turns out that the file content was unchanged, and we were
355
# only going to store a new node if it was changed. Carry over
357
ie.revision = parent_entry.revision
358
ie.text_size = parent_entry.text_size
359
ie.text_sha1 = parent_entry.text_sha1
360
ie.executable = parent_entry.executable
361
return self._get_delta(ie, basis_inv, path), False
362
elif kind == 'directory':
364
# all data is meta here, nothing specific to directory, so
366
ie.revision = parent_entry.revision
367
return self._get_delta(ie, basis_inv, path), False
369
self._add_text_to_weave(ie.file_id, lines, heads, None)
370
elif kind == 'symlink':
371
current_link_target = content_summary[3]
373
# symlink target is not generic metadata, check if it has
375
if current_link_target != parent_entry.symlink_target:
378
# unchanged, carry over.
379
ie.revision = parent_entry.revision
380
ie.symlink_target = parent_entry.symlink_target
381
return self._get_delta(ie, basis_inv, path), False
382
ie.symlink_target = current_link_target
384
self._add_text_to_weave(ie.file_id, lines, heads, None)
385
elif kind == 'tree-reference':
387
if content_summary[3] != parent_entry.reference_revision:
390
# unchanged, carry over.
391
ie.reference_revision = parent_entry.reference_revision
392
ie.revision = parent_entry.revision
393
return self._get_delta(ie, basis_inv, path), False
394
ie.reference_revision = content_summary[3]
396
self._add_text_to_weave(ie.file_id, lines, heads, None)
398
raise NotImplementedError('unknown kind')
399
ie.revision = self._new_revision_id
400
return self._get_delta(ie, basis_inv, path), True
402
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
403
versionedfile = self.repository.weave_store.get_weave_or_empty(
404
file_id, self.repository.get_transaction())
405
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
406
# than add_lines, and allows committing when a parent is ghosted for
408
# Note: as we read the content directly from the tree, we know its not
409
# been turned into unicode or badly split - but a broken tree
410
# implementation could give us bad output from readlines() so this is
411
# not a guarantee of safety. What would be better is always checking
412
# the content during test suite execution. RBC 20070912
414
return versionedfile.add_lines_with_ghosts(
415
self._new_revision_id, parents, new_lines,
416
nostore_sha=nostore_sha, random_id=self.random_revid,
417
check_content=False)[0:2]
419
versionedfile.clear_cache()
422
class RootCommitBuilder(CommitBuilder):
423
"""This commitbuilder actually records the root id"""
425
# the root entry gets versioned properly by this builder.
426
_versioned_root = True
428
def _check_root(self, ie, parent_invs, tree):
429
"""Helper for record_entry_contents.
431
:param ie: An entry being added.
432
:param parent_invs: The inventories of the parent revisions of the
434
:param tree: The tree that is being committed.
438
######################################################################
441
class Repository(object):
442
"""Repository holding history for one or more branches.
444
The repository holds and retrieves historical information including
445
revisions and file history. It's normally accessed only by the Branch,
446
which views a particular line of development through that history.
448
The Repository builds on top of Stores and a Transport, which respectively
449
describe the disk data format and the way of accessing the (possibly
453
# What class to use for a CommitBuilder. Often its simpler to change this
454
# in a Repository class subclass rather than to override
455
# get_commit_builder.
456
_commit_builder_class = CommitBuilder
457
# The search regex used by xml based repositories to determine what things
458
# where changed in a single commit.
459
_file_ids_altered_regex = lazy_regex.lazy_compile(
460
r'file_id="(?P<file_id>[^"]+)"'
461
r'.* revision="(?P<revision_id>[^"]+)"'
464
def abort_write_group(self):
465
"""Commit the contents accrued within the current write group.
467
:seealso: start_write_group.
469
if self._write_group is not self.get_transaction():
470
# has an unlock or relock occured ?
471
raise errors.BzrError('mismatched lock context and write group.')
472
self._abort_write_group()
473
self._write_group = None
475
def _abort_write_group(self):
476
"""Template method for per-repository write group cleanup.
478
This is called during abort before the write group is considered to be
479
finished and should cleanup any internal state accrued during the write
480
group. There is no requirement that data handed to the repository be
481
*not* made available - this is not a rollback - but neither should any
482
attempt be made to ensure that data added is fully commited. Abort is
483
invoked when an error has occured so futher disk or network operations
484
may not be possible or may error and if possible should not be
489
def add_inventory(self, revision_id, inv, parents):
490
"""Add the inventory inv to the repository as revision_id.
492
:param parents: The revision ids of the parents that revision_id
493
is known to have and are in the repository already.
495
returns the sha1 of the serialized inventory.
497
assert self.is_in_write_group()
498
_mod_revision.check_not_reserved_id(revision_id)
499
assert inv.revision_id is None or inv.revision_id == revision_id, \
500
"Mismatch between inventory revision" \
501
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
502
assert inv.root is not None
503
inv_lines = self._serialise_inventory_to_lines(inv)
504
inv_vf = self.get_inventory_weave()
505
return self._inventory_add_lines(inv_vf, revision_id, parents,
506
inv_lines, check_content=False)
508
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
510
"""Store lines in inv_vf and return the sha1 of the inventory."""
512
for parent in parents:
514
final_parents.append(parent)
515
return inv_vf.add_lines(revision_id, final_parents, lines,
516
check_content=check_content)[0]
519
def add_revision(self, revision_id, rev, inv=None, config=None):
520
"""Add rev to the revision store as revision_id.
522
:param revision_id: the revision id to use.
523
:param rev: The revision object.
524
:param inv: The inventory for the revision. if None, it will be looked
525
up in the inventory storer
526
:param config: If None no digital signature will be created.
527
If supplied its signature_needed method will be used
528
to determine if a signature should be made.
530
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
532
_mod_revision.check_not_reserved_id(revision_id)
533
if config is not None and config.signature_needed():
535
inv = self.get_inventory(revision_id)
536
plaintext = Testament(rev, inv).as_short_text()
537
self.store_revision_signature(
538
gpg.GPGStrategy(config), plaintext, revision_id)
539
if not revision_id in self.get_inventory_weave():
541
raise errors.WeaveRevisionNotPresent(revision_id,
542
self.get_inventory_weave())
544
# yes, this is not suitable for adding with ghosts.
545
self.add_inventory(revision_id, inv, rev.parent_ids)
546
self._revision_store.add_revision(rev, self.get_transaction())
548
def _add_revision_text(self, revision_id, text):
549
revision = self._revision_store._serializer.read_revision_from_string(
551
self._revision_store._add_revision(revision, StringIO(text),
552
self.get_transaction())
554
def all_revision_ids(self):
555
"""Returns a list of all the revision ids in the repository.
557
This is deprecated because code should generally work on the graph
558
reachable from a particular revision, and ignore any other revisions
559
that might be present. There is no direct replacement method.
561
if 'evil' in debug.debug_flags:
562
mutter_callsite(2, "all_revision_ids is linear with history.")
563
return self._all_revision_ids()
565
def _all_revision_ids(self):
566
"""Returns a list of all the revision ids in the repository.
568
These are in as much topological order as the underlying store can
571
raise NotImplementedError(self._all_revision_ids)
573
def break_lock(self):
574
"""Break a lock if one is present from another instance.
576
Uses the ui factory to ask for confirmation if the lock may be from
579
self.control_files.break_lock()
582
def _eliminate_revisions_not_present(self, revision_ids):
583
"""Check every revision id in revision_ids to see if we have it.
585
Returns a set of the present revisions.
588
for id in revision_ids:
589
if self.has_revision(id):
594
def create(a_bzrdir):
595
"""Construct the current default format repository in a_bzrdir."""
596
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
598
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
599
"""instantiate a Repository.
601
:param _format: The format of the repository on disk.
602
:param a_bzrdir: The BzrDir of the repository.
604
In the future we will have a single api for all stores for
605
getting file texts, inventories and revisions, then
606
this construct will accept instances of those things.
608
super(Repository, self).__init__()
609
self._format = _format
610
# the following are part of the public API for Repository:
611
self.bzrdir = a_bzrdir
612
self.control_files = control_files
613
self._revision_store = _revision_store
614
# backwards compatibility
615
self.weave_store = text_store
617
self._reconcile_does_inventory_gc = True
618
self._reconcile_fixes_text_parents = False
619
# not right yet - should be more semantically clear ?
621
self.control_store = control_store
622
self.control_weaves = control_store
623
# TODO: make sure to construct the right store classes, etc, depending
624
# on whether escaping is required.
625
self._warn_if_deprecated()
626
self._write_group = None
627
self.base = control_files._transport.base
630
return '%s(%r)' % (self.__class__.__name__,
633
def has_same_location(self, other):
634
"""Returns a boolean indicating if this repository is at the same
635
location as another repository.
637
This might return False even when two repository objects are accessing
638
the same physical repository via different URLs.
640
if self.__class__ is not other.__class__:
642
return (self.control_files._transport.base ==
643
other.control_files._transport.base)
645
def is_in_write_group(self):
646
"""Return True if there is an open write group.
648
:seealso: start_write_group.
650
return self._write_group is not None
653
return self.control_files.is_locked()
655
def is_write_locked(self):
656
"""Return True if this object is write locked."""
657
return self.is_locked() and self.control_files._lock_mode == 'w'
659
def lock_write(self, token=None):
660
"""Lock this repository for writing.
662
This causes caching within the repository obejct to start accumlating
663
data during reads, and allows a 'write_group' to be obtained. Write
664
groups must be used for actual data insertion.
666
:param token: if this is already locked, then lock_write will fail
667
unless the token matches the existing lock.
668
:returns: a token if this instance supports tokens, otherwise None.
669
:raises TokenLockingNotSupported: when a token is given but this
670
instance doesn't support using token locks.
671
:raises MismatchedToken: if the specified token doesn't match the token
672
of the existing lock.
673
:seealso: start_write_group.
675
A token should be passed in if you know that you have locked the object
676
some other way, and need to synchronise this object's state with that
679
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
681
result = self.control_files.lock_write(token=token)
686
self.control_files.lock_read()
689
def get_physical_lock_status(self):
690
return self.control_files.get_physical_lock_status()
692
def leave_lock_in_place(self):
693
"""Tell this repository not to release the physical lock when this
696
If lock_write doesn't return a token, then this method is not supported.
698
self.control_files.leave_in_place()
700
def dont_leave_lock_in_place(self):
701
"""Tell this repository to release the physical lock when this
702
object is unlocked, even if it didn't originally acquire it.
704
If lock_write doesn't return a token, then this method is not supported.
706
self.control_files.dont_leave_in_place()
709
def gather_stats(self, revid=None, committers=None):
710
"""Gather statistics from a revision id.
712
:param revid: The revision id to gather statistics from, if None, then
713
no revision specific statistics are gathered.
714
:param committers: Optional parameter controlling whether to grab
715
a count of committers from the revision specific statistics.
716
:return: A dictionary of statistics. Currently this contains:
717
committers: The number of committers if requested.
718
firstrev: A tuple with timestamp, timezone for the penultimate left
719
most ancestor of revid, if revid is not the NULL_REVISION.
720
latestrev: A tuple with timestamp, timezone for revid, if revid is
721
not the NULL_REVISION.
722
revisions: The total revision count in the repository.
723
size: An estimate disk size of the repository in bytes.
726
if revid and committers:
727
result['committers'] = 0
728
if revid and revid != _mod_revision.NULL_REVISION:
730
all_committers = set()
731
revisions = self.get_ancestry(revid)
732
# pop the leading None
734
first_revision = None
736
# ignore the revisions in the middle - just grab first and last
737
revisions = revisions[0], revisions[-1]
738
for revision in self.get_revisions(revisions):
739
if not first_revision:
740
first_revision = revision
742
all_committers.add(revision.committer)
743
last_revision = revision
745
result['committers'] = len(all_committers)
746
result['firstrev'] = (first_revision.timestamp,
747
first_revision.timezone)
748
result['latestrev'] = (last_revision.timestamp,
749
last_revision.timezone)
751
# now gather global repository information
752
if self.bzrdir.root_transport.listable():
753
c, t = self._revision_store.total_size(self.get_transaction())
754
result['revisions'] = c
758
def get_data_stream(self, revision_ids):
759
raise NotImplementedError(self.get_data_stream)
761
def insert_data_stream(self, stream):
762
"""XXX What does this really do?
764
Is it a substitute for fetch?
765
Should it manage its own write group ?
767
for item_key, bytes in stream:
768
if item_key[0] == 'file':
769
(file_id,) = item_key[1:]
770
knit = self.weave_store.get_weave_or_empty(
771
file_id, self.get_transaction())
772
elif item_key == ('inventory',):
773
knit = self.get_inventory_weave()
774
elif item_key == ('revisions',):
775
knit = self._revision_store.get_revision_file(
776
self.get_transaction())
777
elif item_key == ('signatures',):
778
knit = self._revision_store.get_signature_file(
779
self.get_transaction())
781
raise RepositoryDataStreamError(
782
"Unrecognised data stream key '%s'" % (item_key,))
783
decoded_list = bencode.bdecode(bytes)
784
format = decoded_list.pop(0)
787
for version, options, parents, some_bytes in decoded_list:
788
data_list.append((version, options, len(some_bytes), parents))
789
knit_bytes += some_bytes
790
knit.insert_data_stream(
791
(format, data_list, StringIO(knit_bytes).read))
794
def missing_revision_ids(self, other, revision_id=None):
795
"""Return the revision ids that other has that this does not.
797
These are returned in topological order.
799
revision_id: only return revision ids included by revision_id.
801
return InterRepository.get(other, self).missing_revision_ids(revision_id)
805
"""Open the repository rooted at base.
807
For instance, if the repository is at URL/.bzr/repository,
808
Repository.open(URL) -> a Repository instance.
810
control = bzrdir.BzrDir.open(base)
811
return control.open_repository()
813
def copy_content_into(self, destination, revision_id=None):
814
"""Make a complete copy of the content in self into destination.
816
This is a destructive operation! Do not use it on existing
819
return InterRepository.get(self, destination).copy_content(revision_id)
821
def commit_write_group(self):
822
"""Commit the contents accrued within the current write group.
824
:seealso: start_write_group.
826
if self._write_group is not self.get_transaction():
827
# has an unlock or relock occured ?
828
raise errors.BzrError('mismatched lock context %r and '
830
(self.get_transaction(), self._write_group))
831
self._commit_write_group()
832
self._write_group = None
834
def _commit_write_group(self):
835
"""Template method for per-repository write group cleanup.
837
This is called before the write group is considered to be
838
finished and should ensure that all data handed to the repository
839
for writing during the write group is safely committed (to the
840
extent possible considering file system caching etc).
843
def fetch(self, source, revision_id=None, pb=None):
844
"""Fetch the content required to construct revision_id from source.
846
If revision_id is None all content is copied.
848
# fast path same-url fetch operations
849
if self.has_same_location(source):
850
# check that last_revision is in 'from' and then return a
852
if (revision_id is not None and
853
not _mod_revision.is_null(revision_id)):
854
self.get_revision(revision_id)
856
inter = InterRepository.get(source, self)
858
return inter.fetch(revision_id=revision_id, pb=pb)
859
except NotImplementedError:
860
raise errors.IncompatibleRepositories(source, self)
862
def create_bundle(self, target, base, fileobj, format=None):
863
return serializer.write_bundle(self, target, base, fileobj, format)
865
def get_commit_builder(self, branch, parents, config, timestamp=None,
866
timezone=None, committer=None, revprops=None,
868
"""Obtain a CommitBuilder for this repository.
870
:param branch: Branch to commit to.
871
:param parents: Revision ids of the parents of the new revision.
872
:param config: Configuration to use.
873
:param timestamp: Optional timestamp recorded for commit.
874
:param timezone: Optional timezone for timestamp.
875
:param committer: Optional committer to set for commit.
876
:param revprops: Optional dictionary of revision properties.
877
:param revision_id: Optional revision id.
879
result = self._commit_builder_class(self, parents, config,
880
timestamp, timezone, committer, revprops, revision_id)
881
self.start_write_group()
885
if (self.control_files._lock_count == 1 and
886
self.control_files._lock_mode == 'w'):
887
if self._write_group is not None:
888
self.abort_write_group()
889
self.control_files.unlock()
890
raise errors.BzrError(
891
'Must end write groups before releasing write locks.')
892
self.control_files.unlock()
895
def clone(self, a_bzrdir, revision_id=None):
896
"""Clone this repository into a_bzrdir using the current format.
898
Currently no check is made that the format of this repository and
899
the bzrdir format are compatible. FIXME RBC 20060201.
901
:return: The newly created destination repository.
903
# TODO: deprecate after 0.16; cloning this with all its settings is
904
# probably not very useful -- mbp 20070423
905
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
906
self.copy_content_into(dest_repo, revision_id)
909
def start_write_group(self):
910
"""Start a write group in the repository.
912
Write groups are used by repositories which do not have a 1:1 mapping
913
between file ids and backend store to manage the insertion of data from
914
both fetch and commit operations.
916
A write lock is required around the start_write_group/commit_write_group
917
for the support of lock-requiring repository formats.
919
One can only insert data into a repository inside a write group.
923
if not self.is_write_locked():
924
raise errors.NotWriteLocked(self)
925
if self._write_group:
926
raise errors.BzrError('already in a write group')
927
self._start_write_group()
928
# so we can detect unlock/relock - the write group is now entered.
929
self._write_group = self.get_transaction()
931
def _start_write_group(self):
932
"""Template method for per-repository write group startup.
934
This is called before the write group is considered to be
939
def sprout(self, to_bzrdir, revision_id=None):
940
"""Create a descendent repository for new development.
942
Unlike clone, this does not copy the settings of the repository.
944
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
945
dest_repo.fetch(self, revision_id=revision_id)
948
def _create_sprouting_repo(self, a_bzrdir, shared):
949
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
950
# use target default format.
951
dest_repo = a_bzrdir.create_repository()
953
# Most control formats need the repository to be specifically
954
# created, but on some old all-in-one formats it's not needed
956
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
957
except errors.UninitializableFormat:
958
dest_repo = a_bzrdir.open_repository()
962
def has_revision(self, revision_id):
963
"""True if this repository has a copy of the revision."""
964
if 'evil' in debug.debug_flags:
965
mutter_callsite(3, "has_revision is a LBYL symptom.")
966
return self._revision_store.has_revision_id(revision_id,
967
self.get_transaction())
970
def get_revision(self, revision_id):
971
"""Return the Revision object for a named revision."""
972
return self.get_revisions([revision_id])[0]
975
def get_revision_reconcile(self, revision_id):
976
"""'reconcile' helper routine that allows access to a revision always.
978
This variant of get_revision does not cross check the weave graph
979
against the revision one as get_revision does: but it should only
980
be used by reconcile, or reconcile-alike commands that are correcting
981
or testing the revision graph.
983
return self._get_revisions([revision_id])[0]
986
def get_revisions(self, revision_ids):
987
"""Get many revisions at once."""
988
return self._get_revisions(revision_ids)
991
def _get_revisions(self, revision_ids):
992
"""Core work logic to get many revisions without sanity checks."""
993
for rev_id in revision_ids:
994
if not rev_id or not isinstance(rev_id, basestring):
995
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
996
revs = self._revision_store.get_revisions(revision_ids,
997
self.get_transaction())
999
assert not isinstance(rev.revision_id, unicode)
1000
for parent_id in rev.parent_ids:
1001
assert not isinstance(parent_id, unicode)
1005
def get_revision_xml(self, revision_id):
1006
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1007
# would have already do it.
1008
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1009
rev = self.get_revision(revision_id)
1010
rev_tmp = StringIO()
1011
# the current serializer..
1012
self._revision_store._serializer.write_revision(rev, rev_tmp)
1014
return rev_tmp.getvalue()
1017
def get_deltas_for_revisions(self, revisions):
1018
"""Produce a generator of revision deltas.
1020
Note that the input is a sequence of REVISIONS, not revision_ids.
1021
Trees will be held in memory until the generator exits.
1022
Each delta is relative to the revision's lefthand predecessor.
1024
required_trees = set()
1025
for revision in revisions:
1026
required_trees.add(revision.revision_id)
1027
required_trees.update(revision.parent_ids[:1])
1028
trees = dict((t.get_revision_id(), t) for
1029
t in self.revision_trees(required_trees))
1030
for revision in revisions:
1031
if not revision.parent_ids:
1032
old_tree = self.revision_tree(None)
1034
old_tree = trees[revision.parent_ids[0]]
1035
yield trees[revision.revision_id].changes_from(old_tree)
1038
def get_revision_delta(self, revision_id):
1039
"""Return the delta for one revision.
1041
The delta is relative to the left-hand predecessor of the
1044
r = self.get_revision(revision_id)
1045
return list(self.get_deltas_for_revisions([r]))[0]
1048
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1049
signature = gpg_strategy.sign(plaintext)
1050
self._revision_store.add_revision_signature_text(revision_id,
1052
self.get_transaction())
1054
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1056
"""Helper routine for fileids_altered_by_revision_ids.
1058
This performs the translation of xml lines to revision ids.
1060
:param line_iterator: An iterator of lines
1061
:param revision_ids: The revision ids to filter for.
1062
:return: a dictionary mapping altered file-ids to an iterable of
1063
revision_ids. Each altered file-ids has the exact revision_ids that
1064
altered it listed explicitly.
1068
# this code needs to read every new line in every inventory for the
1069
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1070
# not present in one of those inventories is unnecessary but not
1071
# harmful because we are filtering by the revision id marker in the
1072
# inventory lines : we only select file ids altered in one of those
1073
# revisions. We don't need to see all lines in the inventory because
1074
# only those added in an inventory in rev X can contain a revision=X
1076
unescape_revid_cache = {}
1077
unescape_fileid_cache = {}
1079
# jam 20061218 In a big fetch, this handles hundreds of thousands
1080
# of lines, so it has had a lot of inlining and optimizing done.
1081
# Sorry that it is a little bit messy.
1082
# Move several functions to be local variables, since this is a long
1084
search = self._file_ids_altered_regex.search
1085
unescape = _unescape_xml
1086
setdefault = result.setdefault
1087
for line in line_iterator:
1088
match = search(line)
1091
# One call to match.group() returning multiple items is quite a
1092
# bit faster than 2 calls to match.group() each returning 1
1093
file_id, revision_id = match.group('file_id', 'revision_id')
1095
# Inlining the cache lookups helps a lot when you make 170,000
1096
# lines and 350k ids, versus 8.4 unique ids.
1097
# Using a cache helps in 2 ways:
1098
# 1) Avoids unnecessary decoding calls
1099
# 2) Re-uses cached strings, which helps in future set and
1101
# (2) is enough that removing encoding entirely along with
1102
# the cache (so we are using plain strings) results in no
1103
# performance improvement.
1105
revision_id = unescape_revid_cache[revision_id]
1107
unescaped = unescape(revision_id)
1108
unescape_revid_cache[revision_id] = unescaped
1109
revision_id = unescaped
1111
if revision_id in revision_ids:
1113
file_id = unescape_fileid_cache[file_id]
1115
unescaped = unescape(file_id)
1116
unescape_fileid_cache[file_id] = unescaped
1118
setdefault(file_id, set()).add(revision_id)
1121
def fileids_altered_by_revision_ids(self, revision_ids):
1122
"""Find the file ids and versions affected by revisions.
1124
:param revisions: an iterable containing revision ids.
1125
:return: a dictionary mapping altered file-ids to an iterable of
1126
revision_ids. Each altered file-ids has the exact revision_ids that
1127
altered it listed explicitly.
1129
assert self._serializer.support_altered_by_hack, \
1130
("fileids_altered_by_revision_ids only supported for branches "
1131
"which store inventory as unnested xml, not on %r" % self)
1132
selected_revision_ids = set(revision_ids)
1133
w = self.get_inventory_weave()
1134
pb = ui.ui_factory.nested_progress_bar()
1136
return self._find_file_ids_from_xml_inventory_lines(
1137
w.iter_lines_added_or_present_in_versions(
1138
selected_revision_ids, pb=pb),
1139
selected_revision_ids)
1143
def iter_files_bytes(self, desired_files):
1144
"""Iterate through file versions.
1146
Files will not necessarily be returned in the order they occur in
1147
desired_files. No specific order is guaranteed.
1149
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1150
value supplied by the caller as part of desired_files. It should
1151
uniquely identify the file version in the caller's context. (Examples:
1152
an index number or a TreeTransform trans_id.)
1154
bytes_iterator is an iterable of bytestrings for the file. The
1155
kind of iterable and length of the bytestrings are unspecified, but for
1156
this implementation, it is a list of lines produced by
1157
VersionedFile.get_lines().
1159
:param desired_files: a list of (file_id, revision_id, identifier)
1162
transaction = self.get_transaction()
1163
for file_id, revision_id, callable_data in desired_files:
1165
weave = self.weave_store.get_weave(file_id, transaction)
1166
except errors.NoSuchFile:
1167
raise errors.NoSuchIdInRepository(self, file_id)
1168
yield callable_data, weave.get_lines(revision_id)
1170
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1171
"""Get an iterable listing the keys of all the data introduced by a set
1174
The keys will be ordered so that the corresponding items can be safely
1175
fetched and inserted in that order.
1177
:returns: An iterable producing tuples of (knit-kind, file-id,
1178
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1179
'revisions'. file-id is None unless knit-kind is 'file'.
1181
# XXX: it's a bit weird to control the inventory weave caching in this
1182
# generator. Ideally the caching would be done in fetch.py I think. Or
1183
# maybe this generator should explicitly have the contract that it
1184
# should not be iterated until the previously yielded item has been
1187
inv_w = self.get_inventory_weave()
1188
inv_w.enable_cache()
1190
# file ids that changed
1191
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1193
num_file_ids = len(file_ids)
1194
for file_id, altered_versions in file_ids.iteritems():
1195
if _files_pb is not None:
1196
_files_pb.update("fetch texts", count, num_file_ids)
1198
yield ("file", file_id, altered_versions)
1199
# We're done with the files_pb. Note that it finished by the caller,
1200
# just as it was created by the caller.
1204
yield ("inventory", None, revision_ids)
1208
revisions_with_signatures = set()
1209
for rev_id in revision_ids:
1211
self.get_signature_text(rev_id)
1212
except errors.NoSuchRevision:
1216
revisions_with_signatures.add(rev_id)
1218
yield ("signatures", None, revisions_with_signatures)
1221
yield ("revisions", None, revision_ids)
1224
def get_inventory_weave(self):
1225
return self.control_weaves.get_weave('inventory',
1226
self.get_transaction())
1229
def get_inventory(self, revision_id):
1230
"""Get Inventory object by hash."""
1231
return self.deserialise_inventory(
1232
revision_id, self.get_inventory_xml(revision_id))
1234
def deserialise_inventory(self, revision_id, xml):
1235
"""Transform the xml into an inventory object.
1237
:param revision_id: The expected revision id of the inventory.
1238
:param xml: A serialised inventory.
1240
return self._serializer.read_inventory_from_string(xml, revision_id)
1242
def serialise_inventory(self, inv):
1243
return self._serializer.write_inventory_to_string(inv)
1245
def _serialise_inventory_to_lines(self, inv):
1246
return self._serializer.write_inventory_to_lines(inv)
1248
def get_serializer_format(self):
1249
return self._serializer.format_num
1252
def get_inventory_xml(self, revision_id):
1253
"""Get inventory XML as a file object."""
1255
assert isinstance(revision_id, str), type(revision_id)
1256
iw = self.get_inventory_weave()
1257
return iw.get_text(revision_id)
1259
raise errors.HistoryMissing(self, 'inventory', revision_id)
1262
def get_inventory_sha1(self, revision_id):
1263
"""Return the sha1 hash of the inventory entry
1265
return self.get_revision(revision_id).inventory_sha1
1268
def get_revision_graph(self, revision_id=None):
1269
"""Return a dictionary containing the revision graph.
1271
NB: This method should not be used as it accesses the entire graph all
1272
at once, which is much more data than most operations should require.
1274
:param revision_id: The revision_id to get a graph from. If None, then
1275
the entire revision graph is returned. This is a deprecated mode of
1276
operation and will be removed in the future.
1277
:return: a dictionary of revision_id->revision_parents_list.
1279
raise NotImplementedError(self.get_revision_graph)
1282
def get_revision_graph_with_ghosts(self, revision_ids=None):
1283
"""Return a graph of the revisions with ghosts marked as applicable.
1285
:param revision_ids: an iterable of revisions to graph or None for all.
1286
:return: a Graph object with the graph reachable from revision_ids.
1288
if 'evil' in debug.debug_flags:
1290
"get_revision_graph_with_ghosts scales with size of history.")
1291
result = deprecated_graph.Graph()
1292
if not revision_ids:
1293
pending = set(self.all_revision_ids())
1296
pending = set(revision_ids)
1297
# special case NULL_REVISION
1298
if _mod_revision.NULL_REVISION in pending:
1299
pending.remove(_mod_revision.NULL_REVISION)
1300
required = set(pending)
1303
revision_id = pending.pop()
1305
rev = self.get_revision(revision_id)
1306
except errors.NoSuchRevision:
1307
if revision_id in required:
1310
result.add_ghost(revision_id)
1312
for parent_id in rev.parent_ids:
1313
# is this queued or done ?
1314
if (parent_id not in pending and
1315
parent_id not in done):
1317
pending.add(parent_id)
1318
result.add_node(revision_id, rev.parent_ids)
1319
done.add(revision_id)
1322
def _get_history_vf(self):
1323
"""Get a versionedfile whose history graph reflects all revisions.
1325
For weave repositories, this is the inventory weave.
1327
return self.get_inventory_weave()
1329
def iter_reverse_revision_history(self, revision_id):
1330
"""Iterate backwards through revision ids in the lefthand history
1332
:param revision_id: The revision id to start with. All its lefthand
1333
ancestors will be traversed.
1335
if revision_id in (None, _mod_revision.NULL_REVISION):
1337
next_id = revision_id
1338
versionedfile = self._get_history_vf()
1341
parents = versionedfile.get_parents(next_id)
1342
if len(parents) == 0:
1345
next_id = parents[0]
1348
def get_revision_inventory(self, revision_id):
1349
"""Return inventory of a past revision."""
1350
# TODO: Unify this with get_inventory()
1351
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1352
# must be the same as its revision, so this is trivial.
1353
if revision_id is None:
1354
# This does not make sense: if there is no revision,
1355
# then it is the current tree inventory surely ?!
1356
# and thus get_root_id() is something that looks at the last
1357
# commit on the branch, and the get_root_id is an inventory check.
1358
raise NotImplementedError
1359
# return Inventory(self.get_root_id())
1361
return self.get_inventory(revision_id)
1364
def is_shared(self):
1365
"""Return True if this repository is flagged as a shared repository."""
1366
raise NotImplementedError(self.is_shared)
1369
def reconcile(self, other=None, thorough=False):
1370
"""Reconcile this repository."""
1371
from bzrlib.reconcile import RepoReconciler
1372
reconciler = RepoReconciler(self, thorough=thorough)
1373
reconciler.reconcile()
1376
def _refresh_data(self):
1377
"""Helper called from lock_* to ensure coherency with disk.
1379
The default implementation does nothing; it is however possible
1380
for repositories to maintain loaded indices across multiple locks
1381
by checking inside their implementation of this method to see
1382
whether their indices are still valid. This depends of course on
1383
the disk format being validatable in this manner.
1387
def revision_tree(self, revision_id):
1388
"""Return Tree for a revision on this branch.
1390
`revision_id` may be None for the empty tree revision.
1392
# TODO: refactor this to use an existing revision object
1393
# so we don't need to read it in twice.
1394
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1395
return RevisionTree(self, Inventory(root_id=None),
1396
_mod_revision.NULL_REVISION)
1398
inv = self.get_revision_inventory(revision_id)
1399
return RevisionTree(self, inv, revision_id)
1402
def revision_trees(self, revision_ids):
1403
"""Return Tree for a revision on this branch.
1405
`revision_id` may not be None or 'null:'"""
1406
assert None not in revision_ids
1407
assert _mod_revision.NULL_REVISION not in revision_ids
1408
texts = self.get_inventory_weave().get_texts(revision_ids)
1409
for text, revision_id in zip(texts, revision_ids):
1410
inv = self.deserialise_inventory(revision_id, text)
1411
yield RevisionTree(self, inv, revision_id)
1414
def get_ancestry(self, revision_id, topo_sorted=True):
1415
"""Return a list of revision-ids integrated by a revision.
1417
The first element of the list is always None, indicating the origin
1418
revision. This might change when we have history horizons, or
1419
perhaps we should have a new API.
1421
This is topologically sorted.
1423
if _mod_revision.is_null(revision_id):
1425
if not self.has_revision(revision_id):
1426
raise errors.NoSuchRevision(self, revision_id)
1427
w = self.get_inventory_weave()
1428
candidates = w.get_ancestry(revision_id, topo_sorted)
1429
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1432
"""Compress the data within the repository.
1434
This operation only makes sense for some repository types. For other
1435
types it should be a no-op that just returns.
1437
This stub method does not require a lock, but subclasses should use
1438
@needs_write_lock as this is a long running call its reasonable to
1439
implicitly lock for the user.
1443
def print_file(self, file, revision_id):
1444
"""Print `file` to stdout.
1446
FIXME RBC 20060125 as John Meinel points out this is a bad api
1447
- it writes to stdout, it assumes that that is valid etc. Fix
1448
by creating a new more flexible convenience function.
1450
tree = self.revision_tree(revision_id)
1451
# use inventory as it was in that revision
1452
file_id = tree.inventory.path2id(file)
1454
# TODO: jam 20060427 Write a test for this code path
1455
# it had a bug in it, and was raising the wrong
1457
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1458
tree.print_file(file_id)
1460
def get_transaction(self):
1461
return self.control_files.get_transaction()
1463
def revision_parents(self, revision_id):
1464
return self.get_inventory_weave().parent_names(revision_id)
1466
def get_parents(self, revision_ids):
1467
"""See StackedParentsProvider.get_parents"""
1469
for revision_id in revision_ids:
1470
if revision_id == _mod_revision.NULL_REVISION:
1474
parents = self.get_revision(revision_id).parent_ids
1475
except errors.NoSuchRevision:
1478
if len(parents) == 0:
1479
parents = [_mod_revision.NULL_REVISION]
1480
parents_list.append(parents)
1483
def _make_parents_provider(self):
1486
def get_graph(self, other_repository=None):
1487
"""Return the graph walker for this repository format"""
1488
parents_provider = self._make_parents_provider()
1489
if (other_repository is not None and
1490
other_repository.bzrdir.transport.base !=
1491
self.bzrdir.transport.base):
1492
parents_provider = graph._StackedParentsProvider(
1493
[parents_provider, other_repository._make_parents_provider()])
1494
return graph.Graph(parents_provider)
1496
def get_versioned_file_checker(self, revisions, revision_versions_cache):
1497
return VersionedFileChecker(revisions, revision_versions_cache, self)
1500
def set_make_working_trees(self, new_value):
1501
"""Set the policy flag for making working trees when creating branches.
1503
This only applies to branches that use this repository.
1505
The default is 'True'.
1506
:param new_value: True to restore the default, False to disable making
1509
raise NotImplementedError(self.set_make_working_trees)
1511
def make_working_trees(self):
1512
"""Returns the policy for making working trees on new branches."""
1513
raise NotImplementedError(self.make_working_trees)
1516
def sign_revision(self, revision_id, gpg_strategy):
1517
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1518
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1521
def has_signature_for_revision_id(self, revision_id):
1522
"""Query for a revision signature for revision_id in the repository."""
1523
return self._revision_store.has_signature(revision_id,
1524
self.get_transaction())
1527
def get_signature_text(self, revision_id):
1528
"""Return the text for a signature."""
1529
return self._revision_store.get_signature_text(revision_id,
1530
self.get_transaction())
1533
def check(self, revision_ids=None):
1534
"""Check consistency of all history of given revision_ids.
1536
Different repository implementations should override _check().
1538
:param revision_ids: A non-empty list of revision_ids whose ancestry
1539
will be checked. Typically the last revision_id of a branch.
1541
return self._check(revision_ids)
1543
def _check(self, revision_ids):
1544
result = check.Check(self)
1548
def _warn_if_deprecated(self):
1549
global _deprecation_warning_done
1550
if _deprecation_warning_done:
1552
_deprecation_warning_done = True
1553
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1554
% (self._format, self.bzrdir.transport.base))
1556
def supports_rich_root(self):
1557
return self._format.rich_root_data
1559
def _check_ascii_revisionid(self, revision_id, method):
1560
"""Private helper for ascii-only repositories."""
1561
# weave repositories refuse to store revisionids that are non-ascii.
1562
if revision_id is not None:
1563
# weaves require ascii revision ids.
1564
if isinstance(revision_id, unicode):
1566
revision_id.encode('ascii')
1567
except UnicodeEncodeError:
1568
raise errors.NonAsciiRevisionId(method, self)
1571
revision_id.decode('ascii')
1572
except UnicodeDecodeError:
1573
raise errors.NonAsciiRevisionId(method, self)
1575
def revision_graph_can_have_wrong_parents(self):
1576
"""Is it possible for this repository to have a revision graph with
1579
If True, then this repository must also implement
1580
_find_inconsistent_revision_parents so that check and reconcile can
1581
check for inconsistencies before proceeding with other checks that may
1582
depend on the revision index being consistent.
1584
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1586
# remove these delegates a while after bzr 0.15
1587
def __make_delegated(name, from_module):
1588
def _deprecated_repository_forwarder():
1589
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1590
% (name, from_module),
1593
m = __import__(from_module, globals(), locals(), [name])
1595
return getattr(m, name)
1596
except AttributeError:
1597
raise AttributeError('module %s has no name %s'
1599
globals()[name] = _deprecated_repository_forwarder
1602
'AllInOneRepository',
1603
'WeaveMetaDirRepository',
1604
'PreSplitOutRepositoryFormat',
1605
'RepositoryFormat4',
1606
'RepositoryFormat5',
1607
'RepositoryFormat6',
1608
'RepositoryFormat7',
1610
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1614
'RepositoryFormatKnit',
1615
'RepositoryFormatKnit1',
1617
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1620
def install_revision(repository, rev, revision_tree):
1621
"""Install all revision data into a repository."""
1622
repository.start_write_group()
1624
_install_revision(repository, rev, revision_tree)
1626
repository.abort_write_group()
1629
repository.commit_write_group()
1632
def _install_revision(repository, rev, revision_tree):
1633
"""Install all revision data into a repository."""
1634
present_parents = []
1636
for p_id in rev.parent_ids:
1637
if repository.has_revision(p_id):
1638
present_parents.append(p_id)
1639
parent_trees[p_id] = repository.revision_tree(p_id)
1641
parent_trees[p_id] = repository.revision_tree(None)
1643
inv = revision_tree.inventory
1644
entries = inv.iter_entries()
1645
# backwards compatibility hack: skip the root id.
1646
if not repository.supports_rich_root():
1647
path, root = entries.next()
1648
if root.revision != rev.revision_id:
1649
raise errors.IncompatibleRevision(repr(repository))
1650
# Add the texts that are not already present
1651
for path, ie in entries:
1652
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1653
repository.get_transaction())
1654
if ie.revision not in w:
1656
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1657
# with InventoryEntry.find_previous_heads(). if it is, then there
1658
# is a latent bug here where the parents may have ancestors of each
1660
for revision, tree in parent_trees.iteritems():
1661
if ie.file_id not in tree:
1663
parent_id = tree.inventory[ie.file_id].revision
1664
if parent_id in text_parents:
1666
text_parents.append(parent_id)
1668
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1669
repository.get_transaction())
1670
lines = revision_tree.get_file(ie.file_id).readlines()
1671
vfile.add_lines(rev.revision_id, text_parents, lines)
1673
# install the inventory
1674
repository.add_inventory(rev.revision_id, inv, present_parents)
1675
except errors.RevisionAlreadyPresent:
1677
repository.add_revision(rev.revision_id, rev, inv)
1680
class MetaDirRepository(Repository):
1681
"""Repositories in the new meta-dir layout."""
1683
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1684
super(MetaDirRepository, self).__init__(_format,
1690
dir_mode = self.control_files._dir_mode
1691
file_mode = self.control_files._file_mode
1694
def is_shared(self):
1695
"""Return True if this repository is flagged as a shared repository."""
1696
return self.control_files._transport.has('shared-storage')
1699
def set_make_working_trees(self, new_value):
1700
"""Set the policy flag for making working trees when creating branches.
1702
This only applies to branches that use this repository.
1704
The default is 'True'.
1705
:param new_value: True to restore the default, False to disable making
1710
self.control_files._transport.delete('no-working-trees')
1711
except errors.NoSuchFile:
1714
self.control_files.put_utf8('no-working-trees', '')
1716
def make_working_trees(self):
1717
"""Returns the policy for making working trees on new branches."""
1718
return not self.control_files._transport.has('no-working-trees')
1721
class RepositoryFormatRegistry(registry.Registry):
1722
"""Registry of RepositoryFormats."""
1724
def get(self, format_string):
1725
r = registry.Registry.get(self, format_string)
1731
format_registry = RepositoryFormatRegistry()
1732
"""Registry of formats, indexed by their identifying format string.
1734
This can contain either format instances themselves, or classes/factories that
1735
can be called to obtain one.
1739
#####################################################################
1740
# Repository Formats
1742
class RepositoryFormat(object):
1743
"""A repository format.
1745
Formats provide three things:
1746
* An initialization routine to construct repository data on disk.
1747
* a format string which is used when the BzrDir supports versioned
1749
* an open routine which returns a Repository instance.
1751
There is one and only one Format subclass for each on-disk format. But
1752
there can be one Repository subclass that is used for several different
1753
formats. The _format attribute on a Repository instance can be used to
1754
determine the disk format.
1756
Formats are placed in an dict by their format string for reference
1757
during opening. These should be subclasses of RepositoryFormat
1760
Once a format is deprecated, just deprecate the initialize and open
1761
methods on the format class. Do not deprecate the object, as the
1762
object will be created every system load.
1764
Common instance attributes:
1765
_matchingbzrdir - the bzrdir format that the repository format was
1766
originally written to work with. This can be used if manually
1767
constructing a bzrdir and repository, or more commonly for test suite
1772
return "<%s>" % self.__class__.__name__
1774
def __eq__(self, other):
1775
# format objects are generally stateless
1776
return isinstance(other, self.__class__)
1778
def __ne__(self, other):
1779
return not self == other
1782
def find_format(klass, a_bzrdir):
1783
"""Return the format for the repository object in a_bzrdir.
1785
This is used by bzr native formats that have a "format" file in
1786
the repository. Other methods may be used by different types of
1790
transport = a_bzrdir.get_repository_transport(None)
1791
format_string = transport.get("format").read()
1792
return format_registry.get(format_string)
1793
except errors.NoSuchFile:
1794
raise errors.NoRepositoryPresent(a_bzrdir)
1796
raise errors.UnknownFormatError(format=format_string)
1799
def register_format(klass, format):
1800
format_registry.register(format.get_format_string(), format)
1803
def unregister_format(klass, format):
1804
format_registry.remove(format.get_format_string())
1807
def get_default_format(klass):
1808
"""Return the current default format."""
1809
from bzrlib import bzrdir
1810
return bzrdir.format_registry.make_bzrdir('default').repository_format
1812
def _get_control_store(self, repo_transport, control_files):
1813
"""Return the control store for this repository."""
1814
raise NotImplementedError(self._get_control_store)
1816
def get_format_string(self):
1817
"""Return the ASCII format string that identifies this format.
1819
Note that in pre format ?? repositories the format string is
1820
not permitted nor written to disk.
1822
raise NotImplementedError(self.get_format_string)
1824
def get_format_description(self):
1825
"""Return the short description for this format."""
1826
raise NotImplementedError(self.get_format_description)
1828
def _get_revision_store(self, repo_transport, control_files):
1829
"""Return the revision store object for this a_bzrdir."""
1830
raise NotImplementedError(self._get_revision_store)
1832
def _get_text_rev_store(self,
1839
"""Common logic for getting a revision store for a repository.
1841
see self._get_revision_store for the subclass-overridable method to
1842
get the store for a repository.
1844
from bzrlib.store.revision.text import TextRevisionStore
1845
dir_mode = control_files._dir_mode
1846
file_mode = control_files._file_mode
1847
text_store = TextStore(transport.clone(name),
1849
compressed=compressed,
1851
file_mode=file_mode)
1852
_revision_store = TextRevisionStore(text_store, serializer)
1853
return _revision_store
1855
# TODO: this shouldn't be in the base class, it's specific to things that
1856
# use weaves or knits -- mbp 20070207
1857
def _get_versioned_file_store(self,
1862
versionedfile_class=None,
1863
versionedfile_kwargs={},
1865
if versionedfile_class is None:
1866
versionedfile_class = self._versionedfile_class
1867
weave_transport = control_files._transport.clone(name)
1868
dir_mode = control_files._dir_mode
1869
file_mode = control_files._file_mode
1870
return VersionedFileStore(weave_transport, prefixed=prefixed,
1872
file_mode=file_mode,
1873
versionedfile_class=versionedfile_class,
1874
versionedfile_kwargs=versionedfile_kwargs,
1877
def initialize(self, a_bzrdir, shared=False):
1878
"""Initialize a repository of this format in a_bzrdir.
1880
:param a_bzrdir: The bzrdir to put the new repository in it.
1881
:param shared: The repository should be initialized as a sharable one.
1882
:returns: The new repository object.
1884
This may raise UninitializableFormat if shared repository are not
1885
compatible the a_bzrdir.
1887
raise NotImplementedError(self.initialize)
1889
def is_supported(self):
1890
"""Is this format supported?
1892
Supported formats must be initializable and openable.
1893
Unsupported formats may not support initialization or committing or
1894
some other features depending on the reason for not being supported.
1898
def check_conversion_target(self, target_format):
1899
raise NotImplementedError(self.check_conversion_target)
1901
def open(self, a_bzrdir, _found=False):
1902
"""Return an instance of this format for the bzrdir a_bzrdir.
1904
_found is a private parameter, do not use it.
1906
raise NotImplementedError(self.open)
1909
class MetaDirRepositoryFormat(RepositoryFormat):
1910
"""Common base class for the new repositories using the metadir layout."""
1912
rich_root_data = False
1913
supports_tree_reference = False
1914
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
1917
super(MetaDirRepositoryFormat, self).__init__()
1919
def _create_control_files(self, a_bzrdir):
1920
"""Create the required files and the initial control_files object."""
1921
# FIXME: RBC 20060125 don't peek under the covers
1922
# NB: no need to escape relative paths that are url safe.
1923
repository_transport = a_bzrdir.get_repository_transport(self)
1924
control_files = lockable_files.LockableFiles(repository_transport,
1925
'lock', lockdir.LockDir)
1926
control_files.create_lock()
1927
return control_files
1929
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
1930
"""Upload the initial blank content."""
1931
control_files = self._create_control_files(a_bzrdir)
1932
control_files.lock_write()
1934
control_files._transport.mkdir_multi(dirs,
1935
mode=control_files._dir_mode)
1936
for file, content in files:
1937
control_files.put(file, content)
1938
for file, content in utf8_files:
1939
control_files.put_utf8(file, content)
1941
control_files.put_utf8('shared-storage', '')
1943
control_files.unlock()
1946
# formats which have no format string are not discoverable
1947
# and not independently creatable, so are not registered. They're
1948
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1949
# needed, it's constructed directly by the BzrDir. Non-native formats where
1950
# the repository is not separately opened are similar.
1952
format_registry.register_lazy(
1953
'Bazaar-NG Repository format 7',
1954
'bzrlib.repofmt.weaverepo',
1958
# KEEP in sync with bzrdir.format_registry default, which controls the overall
1959
# default control directory format
1960
format_registry.register_lazy(
1961
'Bazaar-NG Knit Repository Format 1',
1962
'bzrlib.repofmt.knitrepo',
1963
'RepositoryFormatKnit1',
1965
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
1967
format_registry.register_lazy(
1968
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
1969
'bzrlib.repofmt.knitrepo',
1970
'RepositoryFormatKnit3',
1973
# Pack-based formats. There is one format for pre-subtrees, and one for
1974
# post-subtrees to allow ease of testing.
1975
# NOTE: These are experimental in 0.92.
1976
format_registry.register_lazy(
1977
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
1978
'bzrlib.repofmt.pack_repo',
1979
'RepositoryFormatKnitPack1',
1981
format_registry.register_lazy(
1982
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
1983
'bzrlib.repofmt.pack_repo',
1984
'RepositoryFormatKnitPack3',
1988
class InterRepository(InterObject):
1989
"""This class represents operations taking place between two repositories.
1991
Its instances have methods like copy_content and fetch, and contain
1992
references to the source and target repositories these operations can be
1995
Often we will provide convenience methods on 'repository' which carry out
1996
operations with another repository - they will always forward to
1997
InterRepository.get(other).method_name(parameters).
2001
"""The available optimised InterRepository types."""
2003
def copy_content(self, revision_id=None):
2004
raise NotImplementedError(self.copy_content)
2006
def fetch(self, revision_id=None, pb=None):
2007
"""Fetch the content required to construct revision_id.
2009
The content is copied from self.source to self.target.
2011
:param revision_id: if None all content is copied, if NULL_REVISION no
2013
:param pb: optional progress bar to use for progress reports. If not
2014
provided a default one will be created.
2016
Returns the copied revision count and the failed revisions in a tuple:
2019
raise NotImplementedError(self.fetch)
2022
def missing_revision_ids(self, revision_id=None):
2023
"""Return the revision ids that source has that target does not.
2025
These are returned in topological order.
2027
:param revision_id: only return revision ids included by this
2030
# generic, possibly worst case, slow code path.
2031
target_ids = set(self.target.all_revision_ids())
2032
if revision_id is not None:
2033
source_ids = self.source.get_ancestry(revision_id)
2034
assert source_ids[0] is None
2037
source_ids = self.source.all_revision_ids()
2038
result_set = set(source_ids).difference(target_ids)
2039
# this may look like a no-op: its not. It preserves the ordering
2040
# other_ids had while only returning the members from other_ids
2041
# that we've decided we need.
2042
return [rev_id for rev_id in source_ids if rev_id in result_set]
2045
def _same_model(source, target):
2046
"""True if source and target have the same data representation."""
2047
if source.supports_rich_root() != target.supports_rich_root():
2049
if source._serializer != target._serializer:
2054
class InterSameDataRepository(InterRepository):
2055
"""Code for converting between repositories that represent the same data.
2057
Data format and model must match for this to work.
2061
def _get_repo_format_to_test(self):
2062
"""Repository format for testing with.
2064
InterSameData can pull from subtree to subtree and from non-subtree to
2065
non-subtree, so we test this with the richest repository format.
2067
from bzrlib.repofmt import knitrepo
2068
return knitrepo.RepositoryFormatKnit3()
2071
def is_compatible(source, target):
2072
return InterRepository._same_model(source, target)
2075
def copy_content(self, revision_id=None):
2076
"""Make a complete copy of the content in self into destination.
2078
This copies both the repository's revision data, and configuration information
2079
such as the make_working_trees setting.
2081
This is a destructive operation! Do not use it on existing
2084
:param revision_id: Only copy the content needed to construct
2085
revision_id and its parents.
2088
self.target.set_make_working_trees(self.source.make_working_trees())
2089
except NotImplementedError:
2091
# but don't bother fetching if we have the needed data now.
2092
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2093
self.target.has_revision(revision_id)):
2095
self.target.fetch(self.source, revision_id=revision_id)
2098
def fetch(self, revision_id=None, pb=None):
2099
"""See InterRepository.fetch()."""
2100
from bzrlib.fetch import GenericRepoFetcher
2101
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2102
self.source, self.source._format, self.target,
2103
self.target._format)
2104
f = GenericRepoFetcher(to_repository=self.target,
2105
from_repository=self.source,
2106
last_revision=revision_id,
2108
return f.count_copied, f.failed_revisions
2111
class InterWeaveRepo(InterSameDataRepository):
2112
"""Optimised code paths between Weave based repositories.
2114
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2115
implemented lazy inter-object optimisation.
2119
def _get_repo_format_to_test(self):
2120
from bzrlib.repofmt import weaverepo
2121
return weaverepo.RepositoryFormat7()
2124
def is_compatible(source, target):
2125
"""Be compatible with known Weave formats.
2127
We don't test for the stores being of specific types because that
2128
could lead to confusing results, and there is no need to be
2131
from bzrlib.repofmt.weaverepo import (
2137
return (isinstance(source._format, (RepositoryFormat5,
2139
RepositoryFormat7)) and
2140
isinstance(target._format, (RepositoryFormat5,
2142
RepositoryFormat7)))
2143
except AttributeError:
2147
def copy_content(self, revision_id=None):
2148
"""See InterRepository.copy_content()."""
2149
# weave specific optimised path:
2151
self.target.set_make_working_trees(self.source.make_working_trees())
2152
except NotImplementedError:
2154
# FIXME do not peek!
2155
if self.source.control_files._transport.listable():
2156
pb = ui.ui_factory.nested_progress_bar()
2158
self.target.weave_store.copy_all_ids(
2159
self.source.weave_store,
2161
from_transaction=self.source.get_transaction(),
2162
to_transaction=self.target.get_transaction())
2163
pb.update('copying inventory', 0, 1)
2164
self.target.control_weaves.copy_multi(
2165
self.source.control_weaves, ['inventory'],
2166
from_transaction=self.source.get_transaction(),
2167
to_transaction=self.target.get_transaction())
2168
self.target._revision_store.text_store.copy_all_ids(
2169
self.source._revision_store.text_store,
2174
self.target.fetch(self.source, revision_id=revision_id)
2177
def fetch(self, revision_id=None, pb=None):
2178
"""See InterRepository.fetch()."""
2179
from bzrlib.fetch import GenericRepoFetcher
2180
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2181
self.source, self.source._format, self.target, self.target._format)
2182
f = GenericRepoFetcher(to_repository=self.target,
2183
from_repository=self.source,
2184
last_revision=revision_id,
2186
return f.count_copied, f.failed_revisions
2189
def missing_revision_ids(self, revision_id=None):
2190
"""See InterRepository.missing_revision_ids()."""
2191
# we want all revisions to satisfy revision_id in source.
2192
# but we don't want to stat every file here and there.
2193
# we want then, all revisions other needs to satisfy revision_id
2194
# checked, but not those that we have locally.
2195
# so the first thing is to get a subset of the revisions to
2196
# satisfy revision_id in source, and then eliminate those that
2197
# we do already have.
2198
# this is slow on high latency connection to self, but as as this
2199
# disk format scales terribly for push anyway due to rewriting
2200
# inventory.weave, this is considered acceptable.
2202
if revision_id is not None:
2203
source_ids = self.source.get_ancestry(revision_id)
2204
assert source_ids[0] is None
2207
source_ids = self.source._all_possible_ids()
2208
source_ids_set = set(source_ids)
2209
# source_ids is the worst possible case we may need to pull.
2210
# now we want to filter source_ids against what we actually
2211
# have in target, but don't try to check for existence where we know
2212
# we do not have a revision as that would be pointless.
2213
target_ids = set(self.target._all_possible_ids())
2214
possibly_present_revisions = target_ids.intersection(source_ids_set)
2215
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2216
required_revisions = source_ids_set.difference(actually_present_revisions)
2217
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2218
if revision_id is not None:
2219
# we used get_ancestry to determine source_ids then we are assured all
2220
# revisions referenced are present as they are installed in topological order.
2221
# and the tip revision was validated by get_ancestry.
2222
return required_topo_revisions
2224
# if we just grabbed the possibly available ids, then
2225
# we only have an estimate of whats available and need to validate
2226
# that against the revision records.
2227
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2230
class InterKnitRepo(InterSameDataRepository):
2231
"""Optimised code paths between Knit based repositories."""
2234
def _get_repo_format_to_test(self):
2235
from bzrlib.repofmt import knitrepo
2236
return knitrepo.RepositoryFormatKnit1()
2239
def is_compatible(source, target):
2240
"""Be compatible with known Knit formats.
2242
We don't test for the stores being of specific types because that
2243
could lead to confusing results, and there is no need to be
2246
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2248
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2249
isinstance(target._format, RepositoryFormatKnit))
2250
except AttributeError:
2252
return are_knits and InterRepository._same_model(source, target)
2255
def fetch(self, revision_id=None, pb=None):
2256
"""See InterRepository.fetch()."""
2257
from bzrlib.fetch import KnitRepoFetcher
2258
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2259
self.source, self.source._format, self.target, self.target._format)
2260
f = KnitRepoFetcher(to_repository=self.target,
2261
from_repository=self.source,
2262
last_revision=revision_id,
2264
return f.count_copied, f.failed_revisions
2267
def missing_revision_ids(self, revision_id=None):
2268
"""See InterRepository.missing_revision_ids()."""
2269
if revision_id is not None:
2270
source_ids = self.source.get_ancestry(revision_id)
2271
assert source_ids[0] is None
2274
source_ids = self.source.all_revision_ids()
2275
source_ids_set = set(source_ids)
2276
# source_ids is the worst possible case we may need to pull.
2277
# now we want to filter source_ids against what we actually
2278
# have in target, but don't try to check for existence where we know
2279
# we do not have a revision as that would be pointless.
2280
target_ids = set(self.target.all_revision_ids())
2281
possibly_present_revisions = target_ids.intersection(source_ids_set)
2282
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2283
required_revisions = source_ids_set.difference(actually_present_revisions)
2284
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2285
if revision_id is not None:
2286
# we used get_ancestry to determine source_ids then we are assured all
2287
# revisions referenced are present as they are installed in topological order.
2288
# and the tip revision was validated by get_ancestry.
2289
return required_topo_revisions
2291
# if we just grabbed the possibly available ids, then
2292
# we only have an estimate of whats available and need to validate
2293
# that against the revision records.
2294
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2297
class InterPackRepo(InterSameDataRepository):
2298
"""Optimised code paths between Pack based repositories."""
2301
def _get_repo_format_to_test(self):
2302
from bzrlib.repofmt import pack_repo
2303
return pack_repo.RepositoryFormatKnitPack1()
2306
def is_compatible(source, target):
2307
"""Be compatible with known Pack formats.
2309
We don't test for the stores being of specific types because that
2310
could lead to confusing results, and there is no need to be
2313
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2315
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2316
isinstance(target._format, RepositoryFormatPack))
2317
except AttributeError:
2319
return are_packs and InterRepository._same_model(source, target)
2322
def fetch(self, revision_id=None, pb=None):
2323
"""See InterRepository.fetch()."""
2324
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2325
self.source, self.source._format, self.target, self.target._format)
2326
self.count_copied = 0
2327
if revision_id is None:
2329
# everything to do - use pack logic
2330
# to fetch from all packs to one without
2331
# inventory parsing etc, IFF nothing to be copied is in the target.
2333
revision_ids = self.source.all_revision_ids()
2334
# implementing the TODO will involve:
2335
# - detecting when all of a pack is selected
2336
# - avoiding as much as possible pre-selection, so the
2337
# more-core routines such as create_pack_from_packs can filter in
2338
# a just-in-time fashion. (though having a HEADS list on a
2339
# repository might make this a lot easier, because we could
2340
# sensibly detect 'new revisions' without doing a full index scan.
2341
elif _mod_revision.is_null(revision_id):
2346
revision_ids = self.missing_revision_ids(revision_id)
2347
except errors.NoSuchRevision:
2348
raise errors.InstallFailed([revision_id])
2349
packs = self.source._pack_collection.all_packs()
2350
pack = self.target._pack_collection.create_pack_from_packs(
2351
packs, '.fetch', revision_ids,
2353
if pack is not None:
2354
self.target._pack_collection._save_pack_names()
2355
# Trigger an autopack. This may duplicate effort as we've just done
2356
# a pack creation, but for now it is simpler to think about as
2357
# 'upload data, then repack if needed'.
2358
self.target._pack_collection.autopack()
2359
return pack.get_revision_count()
2364
def missing_revision_ids(self, revision_id=None):
2365
"""See InterRepository.missing_revision_ids()."""
2366
if revision_id is not None:
2367
source_ids = self.source.get_ancestry(revision_id)
2368
assert source_ids[0] is None
2371
source_ids = self.source.all_revision_ids()
2372
# source_ids is the worst possible case we may need to pull.
2373
# now we want to filter source_ids against what we actually
2374
# have in target, but don't try to check for existence where we know
2375
# we do not have a revision as that would be pointless.
2376
target_ids = set(self.target.all_revision_ids())
2377
return [r for r in source_ids if (r not in target_ids)]
2380
class InterModel1and2(InterRepository):
2383
def _get_repo_format_to_test(self):
2387
def is_compatible(source, target):
2388
if not source.supports_rich_root() and target.supports_rich_root():
2394
def fetch(self, revision_id=None, pb=None):
2395
"""See InterRepository.fetch()."""
2396
from bzrlib.fetch import Model1toKnit2Fetcher
2397
f = Model1toKnit2Fetcher(to_repository=self.target,
2398
from_repository=self.source,
2399
last_revision=revision_id,
2401
return f.count_copied, f.failed_revisions
2404
def copy_content(self, revision_id=None):
2405
"""Make a complete copy of the content in self into destination.
2407
This is a destructive operation! Do not use it on existing
2410
:param revision_id: Only copy the content needed to construct
2411
revision_id and its parents.
2414
self.target.set_make_working_trees(self.source.make_working_trees())
2415
except NotImplementedError:
2417
# but don't bother fetching if we have the needed data now.
2418
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2419
self.target.has_revision(revision_id)):
2421
self.target.fetch(self.source, revision_id=revision_id)
2424
class InterKnit1and2(InterKnitRepo):
2427
def _get_repo_format_to_test(self):
2431
def is_compatible(source, target):
2432
"""Be compatible with Knit1 source and Knit3 target"""
2433
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2435
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2436
RepositoryFormatKnit3)
2437
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2438
RepositoryFormatKnitPack3)
2439
return (isinstance(source._format,
2440
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2441
isinstance(target._format,
2442
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2444
except AttributeError:
2448
def fetch(self, revision_id=None, pb=None):
2449
"""See InterRepository.fetch()."""
2450
from bzrlib.fetch import Knit1to2Fetcher
2451
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2452
self.source, self.source._format, self.target,
2453
self.target._format)
2454
f = Knit1to2Fetcher(to_repository=self.target,
2455
from_repository=self.source,
2456
last_revision=revision_id,
2458
return f.count_copied, f.failed_revisions
2461
class InterRemoteToOther(InterRepository):
2463
def __init__(self, source, target):
2464
InterRepository.__init__(self, source, target)
2465
self._real_inter = None
2468
def is_compatible(source, target):
2469
if not isinstance(source, remote.RemoteRepository):
2471
source._ensure_real()
2472
real_source = source._real_repository
2473
# Is source's model compatible with target's model, and are they the
2474
# same format? Currently we can only optimise fetching from an
2475
# identical model & format repo.
2476
assert not isinstance(real_source, remote.RemoteRepository), (
2477
"We don't support remote repos backed by remote repos yet.")
2478
return real_source._format == target._format
2481
def fetch(self, revision_id=None, pb=None):
2482
"""See InterRepository.fetch()."""
2483
from bzrlib.fetch import RemoteToOtherFetcher
2484
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2485
self.source, self.target, self.target._format)
2486
# TODO: jam 20070210 This should be an assert, not a translate
2487
revision_id = osutils.safe_revision_id(revision_id)
2488
f = RemoteToOtherFetcher(to_repository=self.target,
2489
from_repository=self.source,
2490
last_revision=revision_id,
2492
return f.count_copied, f.failed_revisions
2495
def _get_repo_format_to_test(self):
2499
class InterOtherToRemote(InterRepository):
2501
def __init__(self, source, target):
2502
InterRepository.__init__(self, source, target)
2503
self._real_inter = None
2506
def is_compatible(source, target):
2507
if isinstance(target, remote.RemoteRepository):
2511
def _ensure_real_inter(self):
2512
if self._real_inter is None:
2513
self.target._ensure_real()
2514
real_target = self.target._real_repository
2515
self._real_inter = InterRepository.get(self.source, real_target)
2517
def copy_content(self, revision_id=None):
2518
self._ensure_real_inter()
2519
self._real_inter.copy_content(revision_id=revision_id)
2521
def fetch(self, revision_id=None, pb=None):
2522
self._ensure_real_inter()
2523
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2526
def _get_repo_format_to_test(self):
2530
InterRepository.register_optimiser(InterSameDataRepository)
2531
InterRepository.register_optimiser(InterWeaveRepo)
2532
InterRepository.register_optimiser(InterKnitRepo)
2533
InterRepository.register_optimiser(InterModel1and2)
2534
InterRepository.register_optimiser(InterKnit1and2)
2535
InterRepository.register_optimiser(InterPackRepo)
2536
InterRepository.register_optimiser(InterRemoteToOther)
2537
InterRepository.register_optimiser(InterOtherToRemote)
2540
class CopyConverter(object):
2541
"""A repository conversion tool which just performs a copy of the content.
2543
This is slow but quite reliable.
2546
def __init__(self, target_format):
2547
"""Create a CopyConverter.
2549
:param target_format: The format the resulting repository should be.
2551
self.target_format = target_format
2553
def convert(self, repo, pb):
2554
"""Perform the conversion of to_convert, giving feedback via pb.
2556
:param to_convert: The disk object to convert.
2557
:param pb: a progress bar to use for progress information.
2562
# this is only useful with metadir layouts - separated repo content.
2563
# trigger an assertion if not such
2564
repo._format.get_format_string()
2565
self.repo_dir = repo.bzrdir
2566
self.step('Moving repository to repository.backup')
2567
self.repo_dir.transport.move('repository', 'repository.backup')
2568
backup_transport = self.repo_dir.transport.clone('repository.backup')
2569
repo._format.check_conversion_target(self.target_format)
2570
self.source_repo = repo._format.open(self.repo_dir,
2572
_override_transport=backup_transport)
2573
self.step('Creating new repository')
2574
converted = self.target_format.initialize(self.repo_dir,
2575
self.source_repo.is_shared())
2576
converted.lock_write()
2578
self.step('Copying content into repository.')
2579
self.source_repo.copy_content_into(converted)
2582
self.step('Deleting old repository content.')
2583
self.repo_dir.transport.delete_tree('repository.backup')
2584
self.pb.note('repository converted')
2586
def step(self, message):
2587
"""Update the pb by a step."""
2589
self.pb.update(message, self.count, self.total)
2601
def _unescaper(match, _map=_unescape_map):
2602
code = match.group(1)
2606
if not code.startswith('#'):
2608
return unichr(int(code[1:])).encode('utf8')
2614
def _unescape_xml(data):
2615
"""Unescape predefined XML entities in a string of data."""
2617
if _unescape_re is None:
2618
_unescape_re = re.compile('\&([^;]*);')
2619
return _unescape_re.sub(_unescaper, data)
2622
class _RevisionTextVersionCache(object):
2623
"""A cache of the versionedfile versions for revision and file-id."""
2625
def __init__(self, repository):
2626
self.repository = repository
2627
self.revision_versions = {}
2628
self.revision_parents = {}
2629
self.repo_graph = self.repository.get_graph()
2630
# XXX: RBC: I haven't tracked down what uses this, but it would be
2631
# better to use the headscache directly I think.
2632
self.heads = graph.HeadsCache(self.repo_graph).heads
2634
def add_revision_text_versions(self, tree):
2635
"""Cache text version data from the supplied revision tree"""
2637
for path, entry in tree.iter_entries_by_dir():
2638
inv_revisions[entry.file_id] = entry.revision
2639
self.revision_versions[tree.get_revision_id()] = inv_revisions
2640
return inv_revisions
2642
def get_text_version(self, file_id, revision_id):
2643
"""Determine the text version for a given file-id and revision-id"""
2645
inv_revisions = self.revision_versions[revision_id]
2648
tree = self.repository.revision_tree(revision_id)
2649
except errors.RevisionNotPresent:
2650
self.revision_versions[revision_id] = inv_revisions = {}
2652
inv_revisions = self.add_revision_text_versions(tree)
2653
return inv_revisions.get(file_id)
2655
def prepopulate_revs(self, revision_ids):
2656
# Filter out versions that we don't have an inventory for, so that the
2657
# revision_trees() call won't fail.
2658
inv_weave = self.repository.get_inventory_weave()
2659
revs = [r for r in revision_ids if inv_weave.has_version(r)]
2660
# XXX: this loop is very similar to
2661
# bzrlib.fetch.Inter1and2Helper.iter_rev_trees.
2663
mutter('%d revisions left to prepopulate', len(revs))
2664
for tree in self.repository.revision_trees(revs[:100]):
2665
if tree.inventory.revision_id is None:
2666
tree.inventory.revision_id = tree.get_revision_id()
2667
self.add_revision_text_versions(tree)
2670
def get_parents(self, revision_id):
2672
return self.revision_parents[revision_id]
2674
parents = self.repository.get_parents([revision_id])[0]
2675
self.revision_parents[revision_id] = parents
2678
def used_file_versions(self):
2679
"""Return a set of (revision_id, file_id) pairs for each file version
2680
referenced by any inventory cached by this _RevisionTextVersionCache.
2682
If the entire repository has been cached, this can be used to find all
2683
file versions that are actually referenced by inventories. Thus any
2684
other file version is completely unused and can be removed safely.
2687
for inventory_summary in self.revision_versions.itervalues():
2688
result.update(inventory_summary.items())
2692
class VersionedFileChecker(object):
2694
def __init__(self, planned_revisions, revision_versions, repository):
2695
self.planned_revisions = planned_revisions
2696
self.revision_versions = revision_versions
2697
self.repository = repository
2699
def calculate_file_version_parents(self, revision_id, file_id):
2700
"""Calculate the correct parents for a file version according to
2703
text_revision = self.revision_versions.get_text_version(
2704
file_id, revision_id)
2705
if text_revision is None:
2707
parents_of_text_revision = self.revision_versions.get_parents(
2709
parents_from_inventories = []
2710
for parent in parents_of_text_revision:
2711
if parent == _mod_revision.NULL_REVISION:
2713
introduced_in = self.revision_versions.get_text_version(file_id,
2715
if introduced_in is not None:
2716
parents_from_inventories.append(introduced_in)
2717
heads = set(self.revision_versions.heads(parents_from_inventories))
2719
for parent in parents_from_inventories:
2720
if parent in heads and parent not in new_parents:
2721
new_parents.append(parent)
2722
return tuple(new_parents)
2724
def check_file_version_parents(self, weave, file_id):
2725
"""Check the parents stored in a versioned file are correct.
2727
It also detects file versions that are not referenced by their
2728
corresponding revision's inventory.
2730
:returns: A tuple of (wrong_parents, dangling_file_versions).
2731
wrong_parents is a dict mapping {revision_id: (stored_parents,
2732
correct_parents)} for each revision_id where the stored parents
2733
are not correct. dangling_file_versions is a set of (file_id,
2734
revision_id) tuples for versions that are present in this versioned
2735
file, but not used by the corresponding inventory.
2738
dangling_file_versions = set()
2739
for num, revision_id in enumerate(self.planned_revisions):
2740
correct_parents = self.calculate_file_version_parents(
2741
revision_id, file_id)
2742
if correct_parents is None:
2744
text_revision = self.revision_versions.get_text_version(
2745
file_id, revision_id)
2747
knit_parents = tuple(weave.get_parents(revision_id))
2748
except errors.RevisionNotPresent:
2750
if text_revision != revision_id:
2751
# This file version is not referenced by its corresponding
2753
dangling_file_versions.add((file_id, revision_id))
2754
if correct_parents != knit_parents:
2755
wrong_parents[revision_id] = (knit_parents, correct_parents)
2756
return wrong_parents, dangling_file_versions
added
removed
bzrlib/repository.py
