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
self._reconcile_backsup_inventory = True
620
# not right yet - should be more semantically clear ?
622
self.control_store = control_store
623
self.control_weaves = control_store
624
# TODO: make sure to construct the right store classes, etc, depending
625
# on whether escaping is required.
626
self._warn_if_deprecated()
627
self._write_group = None
628
self.base = control_files._transport.base
631
return '%s(%r)' % (self.__class__.__name__,
634
def has_same_location(self, other):
635
"""Returns a boolean indicating if this repository is at the same
636
location as another repository.
638
This might return False even when two repository objects are accessing
639
the same physical repository via different URLs.
641
if self.__class__ is not other.__class__:
643
return (self.control_files._transport.base ==
644
other.control_files._transport.base)
646
def is_in_write_group(self):
647
"""Return True if there is an open write group.
649
:seealso: start_write_group.
651
return self._write_group is not None
654
return self.control_files.is_locked()
656
def is_write_locked(self):
657
"""Return True if this object is write locked."""
658
return self.is_locked() and self.control_files._lock_mode == 'w'
660
def lock_write(self, token=None):
661
"""Lock this repository for writing.
663
This causes caching within the repository obejct to start accumlating
664
data during reads, and allows a 'write_group' to be obtained. Write
665
groups must be used for actual data insertion.
667
:param token: if this is already locked, then lock_write will fail
668
unless the token matches the existing lock.
669
:returns: a token if this instance supports tokens, otherwise None.
670
:raises TokenLockingNotSupported: when a token is given but this
671
instance doesn't support using token locks.
672
:raises MismatchedToken: if the specified token doesn't match the token
673
of the existing lock.
674
:seealso: start_write_group.
676
A token should be passed in if you know that you have locked the object
677
some other way, and need to synchronise this object's state with that
680
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
682
result = self.control_files.lock_write(token=token)
687
self.control_files.lock_read()
690
def get_physical_lock_status(self):
691
return self.control_files.get_physical_lock_status()
693
def leave_lock_in_place(self):
694
"""Tell this repository not to release the physical lock when this
697
If lock_write doesn't return a token, then this method is not supported.
699
self.control_files.leave_in_place()
701
def dont_leave_lock_in_place(self):
702
"""Tell this repository to release the physical lock when this
703
object is unlocked, even if it didn't originally acquire it.
705
If lock_write doesn't return a token, then this method is not supported.
707
self.control_files.dont_leave_in_place()
710
def gather_stats(self, revid=None, committers=None):
711
"""Gather statistics from a revision id.
713
:param revid: The revision id to gather statistics from, if None, then
714
no revision specific statistics are gathered.
715
:param committers: Optional parameter controlling whether to grab
716
a count of committers from the revision specific statistics.
717
:return: A dictionary of statistics. Currently this contains:
718
committers: The number of committers if requested.
719
firstrev: A tuple with timestamp, timezone for the penultimate left
720
most ancestor of revid, if revid is not the NULL_REVISION.
721
latestrev: A tuple with timestamp, timezone for revid, if revid is
722
not the NULL_REVISION.
723
revisions: The total revision count in the repository.
724
size: An estimate disk size of the repository in bytes.
727
if revid and committers:
728
result['committers'] = 0
729
if revid and revid != _mod_revision.NULL_REVISION:
731
all_committers = set()
732
revisions = self.get_ancestry(revid)
733
# pop the leading None
735
first_revision = None
737
# ignore the revisions in the middle - just grab first and last
738
revisions = revisions[0], revisions[-1]
739
for revision in self.get_revisions(revisions):
740
if not first_revision:
741
first_revision = revision
743
all_committers.add(revision.committer)
744
last_revision = revision
746
result['committers'] = len(all_committers)
747
result['firstrev'] = (first_revision.timestamp,
748
first_revision.timezone)
749
result['latestrev'] = (last_revision.timestamp,
750
last_revision.timezone)
752
# now gather global repository information
753
if self.bzrdir.root_transport.listable():
754
c, t = self._revision_store.total_size(self.get_transaction())
755
result['revisions'] = c
759
def get_data_stream(self, revision_ids):
760
raise NotImplementedError(self.get_data_stream)
762
def insert_data_stream(self, stream):
763
"""XXX What does this really do?
765
Is it a substitute for fetch?
766
Should it manage its own write group ?
768
for item_key, bytes in stream:
769
if item_key[0] == 'file':
770
(file_id,) = item_key[1:]
771
knit = self.weave_store.get_weave_or_empty(
772
file_id, self.get_transaction())
773
elif item_key == ('inventory',):
774
knit = self.get_inventory_weave()
775
elif item_key == ('revisions',):
776
knit = self._revision_store.get_revision_file(
777
self.get_transaction())
778
elif item_key == ('signatures',):
779
knit = self._revision_store.get_signature_file(
780
self.get_transaction())
782
raise RepositoryDataStreamError(
783
"Unrecognised data stream key '%s'" % (item_key,))
784
decoded_list = bencode.bdecode(bytes)
785
format = decoded_list.pop(0)
788
for version, options, parents, some_bytes in decoded_list:
789
data_list.append((version, options, len(some_bytes), parents))
790
knit_bytes += some_bytes
791
knit.insert_data_stream(
792
(format, data_list, StringIO(knit_bytes).read))
795
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
796
"""Return the revision ids that other has that this does not.
798
These are returned in topological order.
800
revision_id: only return revision ids included by revision_id.
802
return InterRepository.get(other, self).missing_revision_ids(
803
revision_id, find_ghosts)
807
"""Open the repository rooted at base.
809
For instance, if the repository is at URL/.bzr/repository,
810
Repository.open(URL) -> a Repository instance.
812
control = bzrdir.BzrDir.open(base)
813
return control.open_repository()
815
def copy_content_into(self, destination, revision_id=None):
816
"""Make a complete copy of the content in self into destination.
818
This is a destructive operation! Do not use it on existing
821
return InterRepository.get(self, destination).copy_content(revision_id)
823
def commit_write_group(self):
824
"""Commit the contents accrued within the current write group.
826
:seealso: start_write_group.
828
if self._write_group is not self.get_transaction():
829
# has an unlock or relock occured ?
830
raise errors.BzrError('mismatched lock context %r and '
832
(self.get_transaction(), self._write_group))
833
self._commit_write_group()
834
self._write_group = None
836
def _commit_write_group(self):
837
"""Template method for per-repository write group cleanup.
839
This is called before the write group is considered to be
840
finished and should ensure that all data handed to the repository
841
for writing during the write group is safely committed (to the
842
extent possible considering file system caching etc).
845
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
846
"""Fetch the content required to construct revision_id from source.
848
If revision_id is None all content is copied.
849
:param find_ghosts: Find and copy revisions in the source that are
850
ghosts in the target (and not reachable directly by walking out to
851
the first-present revision in target from revision_id).
853
# fast path same-url fetch operations
854
if self.has_same_location(source):
855
# check that last_revision is in 'from' and then return a
857
if (revision_id is not None and
858
not _mod_revision.is_null(revision_id)):
859
self.get_revision(revision_id)
861
inter = InterRepository.get(source, self)
863
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
864
except NotImplementedError:
865
raise errors.IncompatibleRepositories(source, self)
867
def create_bundle(self, target, base, fileobj, format=None):
868
return serializer.write_bundle(self, target, base, fileobj, format)
870
def get_commit_builder(self, branch, parents, config, timestamp=None,
871
timezone=None, committer=None, revprops=None,
873
"""Obtain a CommitBuilder for this repository.
875
:param branch: Branch to commit to.
876
:param parents: Revision ids of the parents of the new revision.
877
:param config: Configuration to use.
878
:param timestamp: Optional timestamp recorded for commit.
879
:param timezone: Optional timezone for timestamp.
880
:param committer: Optional committer to set for commit.
881
:param revprops: Optional dictionary of revision properties.
882
:param revision_id: Optional revision id.
884
result = self._commit_builder_class(self, parents, config,
885
timestamp, timezone, committer, revprops, revision_id)
886
self.start_write_group()
890
if (self.control_files._lock_count == 1 and
891
self.control_files._lock_mode == 'w'):
892
if self._write_group is not None:
893
self.abort_write_group()
894
self.control_files.unlock()
895
raise errors.BzrError(
896
'Must end write groups before releasing write locks.')
897
self.control_files.unlock()
900
def clone(self, a_bzrdir, revision_id=None):
901
"""Clone this repository into a_bzrdir using the current format.
903
Currently no check is made that the format of this repository and
904
the bzrdir format are compatible. FIXME RBC 20060201.
906
:return: The newly created destination repository.
908
# TODO: deprecate after 0.16; cloning this with all its settings is
909
# probably not very useful -- mbp 20070423
910
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
911
self.copy_content_into(dest_repo, revision_id)
914
def start_write_group(self):
915
"""Start a write group in the repository.
917
Write groups are used by repositories which do not have a 1:1 mapping
918
between file ids and backend store to manage the insertion of data from
919
both fetch and commit operations.
921
A write lock is required around the start_write_group/commit_write_group
922
for the support of lock-requiring repository formats.
924
One can only insert data into a repository inside a write group.
928
if not self.is_write_locked():
929
raise errors.NotWriteLocked(self)
930
if self._write_group:
931
raise errors.BzrError('already in a write group')
932
self._start_write_group()
933
# so we can detect unlock/relock - the write group is now entered.
934
self._write_group = self.get_transaction()
936
def _start_write_group(self):
937
"""Template method for per-repository write group startup.
939
This is called before the write group is considered to be
944
def sprout(self, to_bzrdir, revision_id=None):
945
"""Create a descendent repository for new development.
947
Unlike clone, this does not copy the settings of the repository.
949
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
950
dest_repo.fetch(self, revision_id=revision_id)
953
def _create_sprouting_repo(self, a_bzrdir, shared):
954
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
955
# use target default format.
956
dest_repo = a_bzrdir.create_repository()
958
# Most control formats need the repository to be specifically
959
# created, but on some old all-in-one formats it's not needed
961
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
962
except errors.UninitializableFormat:
963
dest_repo = a_bzrdir.open_repository()
967
def has_revision(self, revision_id):
968
"""True if this repository has a copy of the revision."""
969
if 'evil' in debug.debug_flags:
970
mutter_callsite(3, "has_revision is a LBYL symptom.")
971
return self._revision_store.has_revision_id(revision_id,
972
self.get_transaction())
975
def get_revision(self, revision_id):
976
"""Return the Revision object for a named revision."""
977
return self.get_revisions([revision_id])[0]
980
def get_revision_reconcile(self, revision_id):
981
"""'reconcile' helper routine that allows access to a revision always.
983
This variant of get_revision does not cross check the weave graph
984
against the revision one as get_revision does: but it should only
985
be used by reconcile, or reconcile-alike commands that are correcting
986
or testing the revision graph.
988
return self._get_revisions([revision_id])[0]
991
def get_revisions(self, revision_ids):
992
"""Get many revisions at once."""
993
return self._get_revisions(revision_ids)
996
def _get_revisions(self, revision_ids):
997
"""Core work logic to get many revisions without sanity checks."""
998
for rev_id in revision_ids:
999
if not rev_id or not isinstance(rev_id, basestring):
1000
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1001
revs = self._revision_store.get_revisions(revision_ids,
1002
self.get_transaction())
1004
assert not isinstance(rev.revision_id, unicode)
1005
for parent_id in rev.parent_ids:
1006
assert not isinstance(parent_id, unicode)
1010
def get_revision_xml(self, revision_id):
1011
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1012
# would have already do it.
1013
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1014
rev = self.get_revision(revision_id)
1015
rev_tmp = StringIO()
1016
# the current serializer..
1017
self._revision_store._serializer.write_revision(rev, rev_tmp)
1019
return rev_tmp.getvalue()
1022
def get_deltas_for_revisions(self, revisions):
1023
"""Produce a generator of revision deltas.
1025
Note that the input is a sequence of REVISIONS, not revision_ids.
1026
Trees will be held in memory until the generator exits.
1027
Each delta is relative to the revision's lefthand predecessor.
1029
required_trees = set()
1030
for revision in revisions:
1031
required_trees.add(revision.revision_id)
1032
required_trees.update(revision.parent_ids[:1])
1033
trees = dict((t.get_revision_id(), t) for
1034
t in self.revision_trees(required_trees))
1035
for revision in revisions:
1036
if not revision.parent_ids:
1037
old_tree = self.revision_tree(None)
1039
old_tree = trees[revision.parent_ids[0]]
1040
yield trees[revision.revision_id].changes_from(old_tree)
1043
def get_revision_delta(self, revision_id):
1044
"""Return the delta for one revision.
1046
The delta is relative to the left-hand predecessor of the
1049
r = self.get_revision(revision_id)
1050
return list(self.get_deltas_for_revisions([r]))[0]
1053
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1054
signature = gpg_strategy.sign(plaintext)
1055
self._revision_store.add_revision_signature_text(revision_id,
1057
self.get_transaction())
1059
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1061
"""Helper routine for fileids_altered_by_revision_ids.
1063
This performs the translation of xml lines to revision ids.
1065
:param line_iterator: An iterator of lines, origin_version_id
1066
:param revision_ids: The revision ids to filter for. This should be a
1067
set or other type which supports efficient __contains__ lookups, as
1068
the revision id from each parsed line will be looked up in the
1069
revision_ids filter.
1070
:return: a dictionary mapping altered file-ids to an iterable of
1071
revision_ids. Each altered file-ids has the exact revision_ids that
1072
altered it listed explicitly.
1076
# this code needs to read every new line in every inventory for the
1077
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1078
# not present in one of those inventories is unnecessary but not
1079
# harmful because we are filtering by the revision id marker in the
1080
# inventory lines : we only select file ids altered in one of those
1081
# revisions. We don't need to see all lines in the inventory because
1082
# only those added in an inventory in rev X can contain a revision=X
1084
unescape_revid_cache = {}
1085
unescape_fileid_cache = {}
1087
# jam 20061218 In a big fetch, this handles hundreds of thousands
1088
# of lines, so it has had a lot of inlining and optimizing done.
1089
# Sorry that it is a little bit messy.
1090
# Move several functions to be local variables, since this is a long
1092
search = self._file_ids_altered_regex.search
1093
unescape = _unescape_xml
1094
setdefault = result.setdefault
1095
for line, version_id in line_iterator:
1096
match = search(line)
1099
# One call to match.group() returning multiple items is quite a
1100
# bit faster than 2 calls to match.group() each returning 1
1101
file_id, revision_id = match.group('file_id', 'revision_id')
1103
# Inlining the cache lookups helps a lot when you make 170,000
1104
# lines and 350k ids, versus 8.4 unique ids.
1105
# Using a cache helps in 2 ways:
1106
# 1) Avoids unnecessary decoding calls
1107
# 2) Re-uses cached strings, which helps in future set and
1109
# (2) is enough that removing encoding entirely along with
1110
# the cache (so we are using plain strings) results in no
1111
# performance improvement.
1113
revision_id = unescape_revid_cache[revision_id]
1115
unescaped = unescape(revision_id)
1116
unescape_revid_cache[revision_id] = unescaped
1117
revision_id = unescaped
1119
# once data is all ensured-consistent; then this is
1120
# if revision_id == version_id
1121
if revision_id in revision_ids:
1123
file_id = unescape_fileid_cache[file_id]
1125
unescaped = unescape(file_id)
1126
unescape_fileid_cache[file_id] = unescaped
1128
setdefault(file_id, set()).add(revision_id)
1131
def fileids_altered_by_revision_ids(self, revision_ids):
1132
"""Find the file ids and versions affected by revisions.
1134
:param revisions: an iterable containing revision ids.
1135
:return: a dictionary mapping altered file-ids to an iterable of
1136
revision_ids. Each altered file-ids has the exact revision_ids that
1137
altered it listed explicitly.
1139
assert self._serializer.support_altered_by_hack, \
1140
("fileids_altered_by_revision_ids only supported for branches "
1141
"which store inventory as unnested xml, not on %r" % self)
1142
selected_revision_ids = set(revision_ids)
1143
w = self.get_inventory_weave()
1144
pb = ui.ui_factory.nested_progress_bar()
1146
return self._find_file_ids_from_xml_inventory_lines(
1147
w.iter_lines_added_or_present_in_versions(
1148
selected_revision_ids, pb=pb),
1149
selected_revision_ids)
1153
def iter_files_bytes(self, desired_files):
1154
"""Iterate through file versions.
1156
Files will not necessarily be returned in the order they occur in
1157
desired_files. No specific order is guaranteed.
1159
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1160
value supplied by the caller as part of desired_files. It should
1161
uniquely identify the file version in the caller's context. (Examples:
1162
an index number or a TreeTransform trans_id.)
1164
bytes_iterator is an iterable of bytestrings for the file. The
1165
kind of iterable and length of the bytestrings are unspecified, but for
1166
this implementation, it is a list of lines produced by
1167
VersionedFile.get_lines().
1169
:param desired_files: a list of (file_id, revision_id, identifier)
1172
transaction = self.get_transaction()
1173
for file_id, revision_id, callable_data in desired_files:
1175
weave = self.weave_store.get_weave(file_id, transaction)
1176
except errors.NoSuchFile:
1177
raise errors.NoSuchIdInRepository(self, file_id)
1178
yield callable_data, weave.get_lines(revision_id)
1180
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1181
"""Get an iterable listing the keys of all the data introduced by a set
1184
The keys will be ordered so that the corresponding items can be safely
1185
fetched and inserted in that order.
1187
:returns: An iterable producing tuples of (knit-kind, file-id,
1188
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1189
'revisions'. file-id is None unless knit-kind is 'file'.
1191
# XXX: it's a bit weird to control the inventory weave caching in this
1192
# generator. Ideally the caching would be done in fetch.py I think. Or
1193
# maybe this generator should explicitly have the contract that it
1194
# should not be iterated until the previously yielded item has been
1197
inv_w = self.get_inventory_weave()
1198
inv_w.enable_cache()
1200
# file ids that changed
1201
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1203
num_file_ids = len(file_ids)
1204
for file_id, altered_versions in file_ids.iteritems():
1205
if _files_pb is not None:
1206
_files_pb.update("fetch texts", count, num_file_ids)
1208
yield ("file", file_id, altered_versions)
1209
# We're done with the files_pb. Note that it finished by the caller,
1210
# just as it was created by the caller.
1214
yield ("inventory", None, revision_ids)
1218
revisions_with_signatures = set()
1219
for rev_id in revision_ids:
1221
self.get_signature_text(rev_id)
1222
except errors.NoSuchRevision:
1226
revisions_with_signatures.add(rev_id)
1228
yield ("signatures", None, revisions_with_signatures)
1231
yield ("revisions", None, revision_ids)
1234
def get_inventory_weave(self):
1235
return self.control_weaves.get_weave('inventory',
1236
self.get_transaction())
1239
def get_inventory(self, revision_id):
1240
"""Get Inventory object by hash."""
1241
return self.deserialise_inventory(
1242
revision_id, self.get_inventory_xml(revision_id))
1244
def deserialise_inventory(self, revision_id, xml):
1245
"""Transform the xml into an inventory object.
1247
:param revision_id: The expected revision id of the inventory.
1248
:param xml: A serialised inventory.
1250
return self._serializer.read_inventory_from_string(xml, revision_id)
1252
def serialise_inventory(self, inv):
1253
return self._serializer.write_inventory_to_string(inv)
1255
def _serialise_inventory_to_lines(self, inv):
1256
return self._serializer.write_inventory_to_lines(inv)
1258
def get_serializer_format(self):
1259
return self._serializer.format_num
1262
def get_inventory_xml(self, revision_id):
1263
"""Get inventory XML as a file object."""
1265
assert isinstance(revision_id, str), type(revision_id)
1266
iw = self.get_inventory_weave()
1267
return iw.get_text(revision_id)
1269
raise errors.HistoryMissing(self, 'inventory', revision_id)
1272
def get_inventory_sha1(self, revision_id):
1273
"""Return the sha1 hash of the inventory entry
1275
return self.get_revision(revision_id).inventory_sha1
1278
def get_revision_graph(self, revision_id=None):
1279
"""Return a dictionary containing the revision graph.
1281
NB: This method should not be used as it accesses the entire graph all
1282
at once, which is much more data than most operations should require.
1284
:param revision_id: The revision_id to get a graph from. If None, then
1285
the entire revision graph is returned. This is a deprecated mode of
1286
operation and will be removed in the future.
1287
:return: a dictionary of revision_id->revision_parents_list.
1289
raise NotImplementedError(self.get_revision_graph)
1292
def get_revision_graph_with_ghosts(self, revision_ids=None):
1293
"""Return a graph of the revisions with ghosts marked as applicable.
1295
:param revision_ids: an iterable of revisions to graph or None for all.
1296
:return: a Graph object with the graph reachable from revision_ids.
1298
if 'evil' in debug.debug_flags:
1300
"get_revision_graph_with_ghosts scales with size of history.")
1301
result = deprecated_graph.Graph()
1302
if not revision_ids:
1303
pending = set(self.all_revision_ids())
1306
pending = set(revision_ids)
1307
# special case NULL_REVISION
1308
if _mod_revision.NULL_REVISION in pending:
1309
pending.remove(_mod_revision.NULL_REVISION)
1310
required = set(pending)
1313
revision_id = pending.pop()
1315
rev = self.get_revision(revision_id)
1316
except errors.NoSuchRevision:
1317
if revision_id in required:
1320
result.add_ghost(revision_id)
1322
for parent_id in rev.parent_ids:
1323
# is this queued or done ?
1324
if (parent_id not in pending and
1325
parent_id not in done):
1327
pending.add(parent_id)
1328
result.add_node(revision_id, rev.parent_ids)
1329
done.add(revision_id)
1332
def _get_history_vf(self):
1333
"""Get a versionedfile whose history graph reflects all revisions.
1335
For weave repositories, this is the inventory weave.
1337
return self.get_inventory_weave()
1339
def iter_reverse_revision_history(self, revision_id):
1340
"""Iterate backwards through revision ids in the lefthand history
1342
:param revision_id: The revision id to start with. All its lefthand
1343
ancestors will be traversed.
1345
if revision_id in (None, _mod_revision.NULL_REVISION):
1347
next_id = revision_id
1348
versionedfile = self._get_history_vf()
1351
parents = versionedfile.get_parents(next_id)
1352
if len(parents) == 0:
1355
next_id = parents[0]
1358
def get_revision_inventory(self, revision_id):
1359
"""Return inventory of a past revision."""
1360
# TODO: Unify this with get_inventory()
1361
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1362
# must be the same as its revision, so this is trivial.
1363
if revision_id is None:
1364
# This does not make sense: if there is no revision,
1365
# then it is the current tree inventory surely ?!
1366
# and thus get_root_id() is something that looks at the last
1367
# commit on the branch, and the get_root_id is an inventory check.
1368
raise NotImplementedError
1369
# return Inventory(self.get_root_id())
1371
return self.get_inventory(revision_id)
1374
def is_shared(self):
1375
"""Return True if this repository is flagged as a shared repository."""
1376
raise NotImplementedError(self.is_shared)
1379
def reconcile(self, other=None, thorough=False):
1380
"""Reconcile this repository."""
1381
from bzrlib.reconcile import RepoReconciler
1382
reconciler = RepoReconciler(self, thorough=thorough)
1383
reconciler.reconcile()
1386
def _refresh_data(self):
1387
"""Helper called from lock_* to ensure coherency with disk.
1389
The default implementation does nothing; it is however possible
1390
for repositories to maintain loaded indices across multiple locks
1391
by checking inside their implementation of this method to see
1392
whether their indices are still valid. This depends of course on
1393
the disk format being validatable in this manner.
1397
def revision_tree(self, revision_id):
1398
"""Return Tree for a revision on this branch.
1400
`revision_id` may be None for the empty tree revision.
1402
# TODO: refactor this to use an existing revision object
1403
# so we don't need to read it in twice.
1404
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1405
return RevisionTree(self, Inventory(root_id=None),
1406
_mod_revision.NULL_REVISION)
1408
inv = self.get_revision_inventory(revision_id)
1409
return RevisionTree(self, inv, revision_id)
1412
def revision_trees(self, revision_ids):
1413
"""Return Tree for a revision on this branch.
1415
`revision_id` may not be None or 'null:'"""
1416
assert None not in revision_ids
1417
assert _mod_revision.NULL_REVISION not in revision_ids
1418
texts = self.get_inventory_weave().get_texts(revision_ids)
1419
for text, revision_id in zip(texts, revision_ids):
1420
inv = self.deserialise_inventory(revision_id, text)
1421
yield RevisionTree(self, inv, revision_id)
1424
def get_ancestry(self, revision_id, topo_sorted=True):
1425
"""Return a list of revision-ids integrated by a revision.
1427
The first element of the list is always None, indicating the origin
1428
revision. This might change when we have history horizons, or
1429
perhaps we should have a new API.
1431
This is topologically sorted.
1433
if _mod_revision.is_null(revision_id):
1435
if not self.has_revision(revision_id):
1436
raise errors.NoSuchRevision(self, revision_id)
1437
w = self.get_inventory_weave()
1438
candidates = w.get_ancestry(revision_id, topo_sorted)
1439
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1442
"""Compress the data within the repository.
1444
This operation only makes sense for some repository types. For other
1445
types it should be a no-op that just returns.
1447
This stub method does not require a lock, but subclasses should use
1448
@needs_write_lock as this is a long running call its reasonable to
1449
implicitly lock for the user.
1453
def print_file(self, file, revision_id):
1454
"""Print `file` to stdout.
1456
FIXME RBC 20060125 as John Meinel points out this is a bad api
1457
- it writes to stdout, it assumes that that is valid etc. Fix
1458
by creating a new more flexible convenience function.
1460
tree = self.revision_tree(revision_id)
1461
# use inventory as it was in that revision
1462
file_id = tree.inventory.path2id(file)
1464
# TODO: jam 20060427 Write a test for this code path
1465
# it had a bug in it, and was raising the wrong
1467
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1468
tree.print_file(file_id)
1470
def get_transaction(self):
1471
return self.control_files.get_transaction()
1473
def revision_parents(self, revision_id):
1474
return self.get_inventory_weave().parent_names(revision_id)
1476
def get_parents(self, revision_ids):
1477
"""See StackedParentsProvider.get_parents"""
1479
for revision_id in revision_ids:
1480
if revision_id == _mod_revision.NULL_REVISION:
1484
parents = self.get_revision(revision_id).parent_ids
1485
except errors.NoSuchRevision:
1488
if len(parents) == 0:
1489
parents = [_mod_revision.NULL_REVISION]
1490
parents_list.append(parents)
1493
def _make_parents_provider(self):
1496
def get_graph(self, other_repository=None):
1497
"""Return the graph walker for this repository format"""
1498
parents_provider = self._make_parents_provider()
1499
if (other_repository is not None and
1500
other_repository.bzrdir.transport.base !=
1501
self.bzrdir.transport.base):
1502
parents_provider = graph._StackedParentsProvider(
1503
[parents_provider, other_repository._make_parents_provider()])
1504
return graph.Graph(parents_provider)
1506
def get_versioned_file_checker(self, revisions, revision_versions_cache):
1507
return VersionedFileChecker(revisions, revision_versions_cache, self)
1510
def set_make_working_trees(self, new_value):
1511
"""Set the policy flag for making working trees when creating branches.
1513
This only applies to branches that use this repository.
1515
The default is 'True'.
1516
:param new_value: True to restore the default, False to disable making
1519
raise NotImplementedError(self.set_make_working_trees)
1521
def make_working_trees(self):
1522
"""Returns the policy for making working trees on new branches."""
1523
raise NotImplementedError(self.make_working_trees)
1526
def sign_revision(self, revision_id, gpg_strategy):
1527
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1528
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1531
def has_signature_for_revision_id(self, revision_id):
1532
"""Query for a revision signature for revision_id in the repository."""
1533
return self._revision_store.has_signature(revision_id,
1534
self.get_transaction())
1537
def get_signature_text(self, revision_id):
1538
"""Return the text for a signature."""
1539
return self._revision_store.get_signature_text(revision_id,
1540
self.get_transaction())
1543
def check(self, revision_ids=None):
1544
"""Check consistency of all history of given revision_ids.
1546
Different repository implementations should override _check().
1548
:param revision_ids: A non-empty list of revision_ids whose ancestry
1549
will be checked. Typically the last revision_id of a branch.
1551
return self._check(revision_ids)
1553
def _check(self, revision_ids):
1554
result = check.Check(self)
1558
def _warn_if_deprecated(self):
1559
global _deprecation_warning_done
1560
if _deprecation_warning_done:
1562
_deprecation_warning_done = True
1563
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1564
% (self._format, self.bzrdir.transport.base))
1566
def supports_rich_root(self):
1567
return self._format.rich_root_data
1569
def _check_ascii_revisionid(self, revision_id, method):
1570
"""Private helper for ascii-only repositories."""
1571
# weave repositories refuse to store revisionids that are non-ascii.
1572
if revision_id is not None:
1573
# weaves require ascii revision ids.
1574
if isinstance(revision_id, unicode):
1576
revision_id.encode('ascii')
1577
except UnicodeEncodeError:
1578
raise errors.NonAsciiRevisionId(method, self)
1581
revision_id.decode('ascii')
1582
except UnicodeDecodeError:
1583
raise errors.NonAsciiRevisionId(method, self)
1585
def revision_graph_can_have_wrong_parents(self):
1586
"""Is it possible for this repository to have a revision graph with
1589
If True, then this repository must also implement
1590
_find_inconsistent_revision_parents so that check and reconcile can
1591
check for inconsistencies before proceeding with other checks that may
1592
depend on the revision index being consistent.
1594
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1596
# remove these delegates a while after bzr 0.15
1597
def __make_delegated(name, from_module):
1598
def _deprecated_repository_forwarder():
1599
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1600
% (name, from_module),
1603
m = __import__(from_module, globals(), locals(), [name])
1605
return getattr(m, name)
1606
except AttributeError:
1607
raise AttributeError('module %s has no name %s'
1609
globals()[name] = _deprecated_repository_forwarder
1612
'AllInOneRepository',
1613
'WeaveMetaDirRepository',
1614
'PreSplitOutRepositoryFormat',
1615
'RepositoryFormat4',
1616
'RepositoryFormat5',
1617
'RepositoryFormat6',
1618
'RepositoryFormat7',
1620
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1624
'RepositoryFormatKnit',
1625
'RepositoryFormatKnit1',
1627
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1630
def install_revision(repository, rev, revision_tree):
1631
"""Install all revision data into a repository."""
1632
repository.start_write_group()
1634
_install_revision(repository, rev, revision_tree)
1636
repository.abort_write_group()
1639
repository.commit_write_group()
1642
def _install_revision(repository, rev, revision_tree):
1643
"""Install all revision data into a repository."""
1644
present_parents = []
1646
for p_id in rev.parent_ids:
1647
if repository.has_revision(p_id):
1648
present_parents.append(p_id)
1649
parent_trees[p_id] = repository.revision_tree(p_id)
1651
parent_trees[p_id] = repository.revision_tree(None)
1653
inv = revision_tree.inventory
1654
entries = inv.iter_entries()
1655
# backwards compatibility hack: skip the root id.
1656
if not repository.supports_rich_root():
1657
path, root = entries.next()
1658
if root.revision != rev.revision_id:
1659
raise errors.IncompatibleRevision(repr(repository))
1660
# Add the texts that are not already present
1661
for path, ie in entries:
1662
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1663
repository.get_transaction())
1664
if ie.revision not in w:
1666
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1667
# with InventoryEntry.find_previous_heads(). if it is, then there
1668
# is a latent bug here where the parents may have ancestors of each
1670
for revision, tree in parent_trees.iteritems():
1671
if ie.file_id not in tree:
1673
parent_id = tree.inventory[ie.file_id].revision
1674
if parent_id in text_parents:
1676
text_parents.append(parent_id)
1678
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1679
repository.get_transaction())
1680
lines = revision_tree.get_file(ie.file_id).readlines()
1681
vfile.add_lines(rev.revision_id, text_parents, lines)
1683
# install the inventory
1684
repository.add_inventory(rev.revision_id, inv, present_parents)
1685
except errors.RevisionAlreadyPresent:
1687
repository.add_revision(rev.revision_id, rev, inv)
1690
class MetaDirRepository(Repository):
1691
"""Repositories in the new meta-dir layout."""
1693
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1694
super(MetaDirRepository, self).__init__(_format,
1700
dir_mode = self.control_files._dir_mode
1701
file_mode = self.control_files._file_mode
1704
def is_shared(self):
1705
"""Return True if this repository is flagged as a shared repository."""
1706
return self.control_files._transport.has('shared-storage')
1709
def set_make_working_trees(self, new_value):
1710
"""Set the policy flag for making working trees when creating branches.
1712
This only applies to branches that use this repository.
1714
The default is 'True'.
1715
:param new_value: True to restore the default, False to disable making
1720
self.control_files._transport.delete('no-working-trees')
1721
except errors.NoSuchFile:
1724
self.control_files.put_utf8('no-working-trees', '')
1726
def make_working_trees(self):
1727
"""Returns the policy for making working trees on new branches."""
1728
return not self.control_files._transport.has('no-working-trees')
1731
class RepositoryFormatRegistry(registry.Registry):
1732
"""Registry of RepositoryFormats."""
1734
def get(self, format_string):
1735
r = registry.Registry.get(self, format_string)
1741
format_registry = RepositoryFormatRegistry()
1742
"""Registry of formats, indexed by their identifying format string.
1744
This can contain either format instances themselves, or classes/factories that
1745
can be called to obtain one.
1749
#####################################################################
1750
# Repository Formats
1752
class RepositoryFormat(object):
1753
"""A repository format.
1755
Formats provide three things:
1756
* An initialization routine to construct repository data on disk.
1757
* a format string which is used when the BzrDir supports versioned
1759
* an open routine which returns a Repository instance.
1761
There is one and only one Format subclass for each on-disk format. But
1762
there can be one Repository subclass that is used for several different
1763
formats. The _format attribute on a Repository instance can be used to
1764
determine the disk format.
1766
Formats are placed in an dict by their format string for reference
1767
during opening. These should be subclasses of RepositoryFormat
1770
Once a format is deprecated, just deprecate the initialize and open
1771
methods on the format class. Do not deprecate the object, as the
1772
object will be created every system load.
1774
Common instance attributes:
1775
_matchingbzrdir - the bzrdir format that the repository format was
1776
originally written to work with. This can be used if manually
1777
constructing a bzrdir and repository, or more commonly for test suite
1781
# Set to True or False in derived classes. True indicates that the format
1782
# supports ghosts gracefully.
1783
supports_ghosts = None
1786
return "<%s>" % self.__class__.__name__
1788
def __eq__(self, other):
1789
# format objects are generally stateless
1790
return isinstance(other, self.__class__)
1792
def __ne__(self, other):
1793
return not self == other
1796
def find_format(klass, a_bzrdir):
1797
"""Return the format for the repository object in a_bzrdir.
1799
This is used by bzr native formats that have a "format" file in
1800
the repository. Other methods may be used by different types of
1804
transport = a_bzrdir.get_repository_transport(None)
1805
format_string = transport.get("format").read()
1806
return format_registry.get(format_string)
1807
except errors.NoSuchFile:
1808
raise errors.NoRepositoryPresent(a_bzrdir)
1810
raise errors.UnknownFormatError(format=format_string)
1813
def register_format(klass, format):
1814
format_registry.register(format.get_format_string(), format)
1817
def unregister_format(klass, format):
1818
format_registry.remove(format.get_format_string())
1821
def get_default_format(klass):
1822
"""Return the current default format."""
1823
from bzrlib import bzrdir
1824
return bzrdir.format_registry.make_bzrdir('default').repository_format
1826
def _get_control_store(self, repo_transport, control_files):
1827
"""Return the control store for this repository."""
1828
raise NotImplementedError(self._get_control_store)
1830
def get_format_string(self):
1831
"""Return the ASCII format string that identifies this format.
1833
Note that in pre format ?? repositories the format string is
1834
not permitted nor written to disk.
1836
raise NotImplementedError(self.get_format_string)
1838
def get_format_description(self):
1839
"""Return the short description for this format."""
1840
raise NotImplementedError(self.get_format_description)
1842
def _get_revision_store(self, repo_transport, control_files):
1843
"""Return the revision store object for this a_bzrdir."""
1844
raise NotImplementedError(self._get_revision_store)
1846
def _get_text_rev_store(self,
1853
"""Common logic for getting a revision store for a repository.
1855
see self._get_revision_store for the subclass-overridable method to
1856
get the store for a repository.
1858
from bzrlib.store.revision.text import TextRevisionStore
1859
dir_mode = control_files._dir_mode
1860
file_mode = control_files._file_mode
1861
text_store = TextStore(transport.clone(name),
1863
compressed=compressed,
1865
file_mode=file_mode)
1866
_revision_store = TextRevisionStore(text_store, serializer)
1867
return _revision_store
1869
# TODO: this shouldn't be in the base class, it's specific to things that
1870
# use weaves or knits -- mbp 20070207
1871
def _get_versioned_file_store(self,
1876
versionedfile_class=None,
1877
versionedfile_kwargs={},
1879
if versionedfile_class is None:
1880
versionedfile_class = self._versionedfile_class
1881
weave_transport = control_files._transport.clone(name)
1882
dir_mode = control_files._dir_mode
1883
file_mode = control_files._file_mode
1884
return VersionedFileStore(weave_transport, prefixed=prefixed,
1886
file_mode=file_mode,
1887
versionedfile_class=versionedfile_class,
1888
versionedfile_kwargs=versionedfile_kwargs,
1891
def initialize(self, a_bzrdir, shared=False):
1892
"""Initialize a repository of this format in a_bzrdir.
1894
:param a_bzrdir: The bzrdir to put the new repository in it.
1895
:param shared: The repository should be initialized as a sharable one.
1896
:returns: The new repository object.
1898
This may raise UninitializableFormat if shared repository are not
1899
compatible the a_bzrdir.
1901
raise NotImplementedError(self.initialize)
1903
def is_supported(self):
1904
"""Is this format supported?
1906
Supported formats must be initializable and openable.
1907
Unsupported formats may not support initialization or committing or
1908
some other features depending on the reason for not being supported.
1912
def check_conversion_target(self, target_format):
1913
raise NotImplementedError(self.check_conversion_target)
1915
def open(self, a_bzrdir, _found=False):
1916
"""Return an instance of this format for the bzrdir a_bzrdir.
1918
_found is a private parameter, do not use it.
1920
raise NotImplementedError(self.open)
1923
class MetaDirRepositoryFormat(RepositoryFormat):
1924
"""Common base class for the new repositories using the metadir layout."""
1926
rich_root_data = False
1927
supports_tree_reference = False
1928
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
1931
super(MetaDirRepositoryFormat, self).__init__()
1933
def _create_control_files(self, a_bzrdir):
1934
"""Create the required files and the initial control_files object."""
1935
# FIXME: RBC 20060125 don't peek under the covers
1936
# NB: no need to escape relative paths that are url safe.
1937
repository_transport = a_bzrdir.get_repository_transport(self)
1938
control_files = lockable_files.LockableFiles(repository_transport,
1939
'lock', lockdir.LockDir)
1940
control_files.create_lock()
1941
return control_files
1943
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
1944
"""Upload the initial blank content."""
1945
control_files = self._create_control_files(a_bzrdir)
1946
control_files.lock_write()
1948
control_files._transport.mkdir_multi(dirs,
1949
mode=control_files._dir_mode)
1950
for file, content in files:
1951
control_files.put(file, content)
1952
for file, content in utf8_files:
1953
control_files.put_utf8(file, content)
1955
control_files.put_utf8('shared-storage', '')
1957
control_files.unlock()
1960
# formats which have no format string are not discoverable
1961
# and not independently creatable, so are not registered. They're
1962
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1963
# needed, it's constructed directly by the BzrDir. Non-native formats where
1964
# the repository is not separately opened are similar.
1966
format_registry.register_lazy(
1967
'Bazaar-NG Repository format 7',
1968
'bzrlib.repofmt.weaverepo',
1972
# KEEP in sync with bzrdir.format_registry default, which controls the overall
1973
# default control directory format
1974
format_registry.register_lazy(
1975
'Bazaar-NG Knit Repository Format 1',
1976
'bzrlib.repofmt.knitrepo',
1977
'RepositoryFormatKnit1',
1979
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
1981
format_registry.register_lazy(
1982
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
1983
'bzrlib.repofmt.knitrepo',
1984
'RepositoryFormatKnit3',
1987
# Pack-based formats. There is one format for pre-subtrees, and one for
1988
# post-subtrees to allow ease of testing.
1989
# NOTE: These are experimental in 0.92.
1990
format_registry.register_lazy(
1991
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
1992
'bzrlib.repofmt.pack_repo',
1993
'RepositoryFormatKnitPack1',
1995
format_registry.register_lazy(
1996
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
1997
'bzrlib.repofmt.pack_repo',
1998
'RepositoryFormatKnitPack3',
2002
class InterRepository(InterObject):
2003
"""This class represents operations taking place between two repositories.
2005
Its instances have methods like copy_content and fetch, and contain
2006
references to the source and target repositories these operations can be
2009
Often we will provide convenience methods on 'repository' which carry out
2010
operations with another repository - they will always forward to
2011
InterRepository.get(other).method_name(parameters).
2015
"""The available optimised InterRepository types."""
2017
def copy_content(self, revision_id=None):
2018
raise NotImplementedError(self.copy_content)
2020
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2021
"""Fetch the content required to construct revision_id.
2023
The content is copied from self.source to self.target.
2025
:param revision_id: if None all content is copied, if NULL_REVISION no
2027
:param pb: optional progress bar to use for progress reports. If not
2028
provided a default one will be created.
2030
Returns the copied revision count and the failed revisions in a tuple:
2033
raise NotImplementedError(self.fetch)
2036
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2037
"""Return the revision ids that source has that target does not.
2039
These are returned in topological order.
2041
:param revision_id: only return revision ids included by this
2044
# generic, possibly worst case, slow code path.
2045
target_ids = set(self.target.all_revision_ids())
2046
if revision_id is not None:
2047
source_ids = self.source.get_ancestry(revision_id)
2048
assert source_ids[0] is None
2051
source_ids = self.source.all_revision_ids()
2052
result_set = set(source_ids).difference(target_ids)
2053
# this may look like a no-op: its not. It preserves the ordering
2054
# other_ids had while only returning the members from other_ids
2055
# that we've decided we need.
2056
return [rev_id for rev_id in source_ids if rev_id in result_set]
2059
def _same_model(source, target):
2060
"""True if source and target have the same data representation."""
2061
if source.supports_rich_root() != target.supports_rich_root():
2063
if source._serializer != target._serializer:
2068
class InterSameDataRepository(InterRepository):
2069
"""Code for converting between repositories that represent the same data.
2071
Data format and model must match for this to work.
2075
def _get_repo_format_to_test(self):
2076
"""Repository format for testing with.
2078
InterSameData can pull from subtree to subtree and from non-subtree to
2079
non-subtree, so we test this with the richest repository format.
2081
from bzrlib.repofmt import knitrepo
2082
return knitrepo.RepositoryFormatKnit3()
2085
def is_compatible(source, target):
2086
return InterRepository._same_model(source, target)
2089
def copy_content(self, revision_id=None):
2090
"""Make a complete copy of the content in self into destination.
2092
This copies both the repository's revision data, and configuration information
2093
such as the make_working_trees setting.
2095
This is a destructive operation! Do not use it on existing
2098
:param revision_id: Only copy the content needed to construct
2099
revision_id and its parents.
2102
self.target.set_make_working_trees(self.source.make_working_trees())
2103
except NotImplementedError:
2105
# but don't bother fetching if we have the needed data now.
2106
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2107
self.target.has_revision(revision_id)):
2109
self.target.fetch(self.source, revision_id=revision_id)
2112
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2113
"""See InterRepository.fetch()."""
2114
from bzrlib.fetch import GenericRepoFetcher
2115
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2116
self.source, self.source._format, self.target,
2117
self.target._format)
2118
f = GenericRepoFetcher(to_repository=self.target,
2119
from_repository=self.source,
2120
last_revision=revision_id,
2122
return f.count_copied, f.failed_revisions
2125
class InterWeaveRepo(InterSameDataRepository):
2126
"""Optimised code paths between Weave based repositories.
2128
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2129
implemented lazy inter-object optimisation.
2133
def _get_repo_format_to_test(self):
2134
from bzrlib.repofmt import weaverepo
2135
return weaverepo.RepositoryFormat7()
2138
def is_compatible(source, target):
2139
"""Be compatible with known Weave formats.
2141
We don't test for the stores being of specific types because that
2142
could lead to confusing results, and there is no need to be
2145
from bzrlib.repofmt.weaverepo import (
2151
return (isinstance(source._format, (RepositoryFormat5,
2153
RepositoryFormat7)) and
2154
isinstance(target._format, (RepositoryFormat5,
2156
RepositoryFormat7)))
2157
except AttributeError:
2161
def copy_content(self, revision_id=None):
2162
"""See InterRepository.copy_content()."""
2163
# weave specific optimised path:
2165
self.target.set_make_working_trees(self.source.make_working_trees())
2166
except NotImplementedError:
2168
# FIXME do not peek!
2169
if self.source.control_files._transport.listable():
2170
pb = ui.ui_factory.nested_progress_bar()
2172
self.target.weave_store.copy_all_ids(
2173
self.source.weave_store,
2175
from_transaction=self.source.get_transaction(),
2176
to_transaction=self.target.get_transaction())
2177
pb.update('copying inventory', 0, 1)
2178
self.target.control_weaves.copy_multi(
2179
self.source.control_weaves, ['inventory'],
2180
from_transaction=self.source.get_transaction(),
2181
to_transaction=self.target.get_transaction())
2182
self.target._revision_store.text_store.copy_all_ids(
2183
self.source._revision_store.text_store,
2188
self.target.fetch(self.source, revision_id=revision_id)
2191
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2192
"""See InterRepository.fetch()."""
2193
from bzrlib.fetch import GenericRepoFetcher
2194
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2195
self.source, self.source._format, self.target, self.target._format)
2196
f = GenericRepoFetcher(to_repository=self.target,
2197
from_repository=self.source,
2198
last_revision=revision_id,
2200
return f.count_copied, f.failed_revisions
2203
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2204
"""See InterRepository.missing_revision_ids()."""
2205
# we want all revisions to satisfy revision_id in source.
2206
# but we don't want to stat every file here and there.
2207
# we want then, all revisions other needs to satisfy revision_id
2208
# checked, but not those that we have locally.
2209
# so the first thing is to get a subset of the revisions to
2210
# satisfy revision_id in source, and then eliminate those that
2211
# we do already have.
2212
# this is slow on high latency connection to self, but as as this
2213
# disk format scales terribly for push anyway due to rewriting
2214
# inventory.weave, this is considered acceptable.
2216
if revision_id is not None:
2217
source_ids = self.source.get_ancestry(revision_id)
2218
assert source_ids[0] is None
2221
source_ids = self.source._all_possible_ids()
2222
source_ids_set = set(source_ids)
2223
# source_ids is the worst possible case we may need to pull.
2224
# now we want to filter source_ids against what we actually
2225
# have in target, but don't try to check for existence where we know
2226
# we do not have a revision as that would be pointless.
2227
target_ids = set(self.target._all_possible_ids())
2228
possibly_present_revisions = target_ids.intersection(source_ids_set)
2229
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2230
required_revisions = source_ids_set.difference(actually_present_revisions)
2231
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2232
if revision_id is not None:
2233
# we used get_ancestry to determine source_ids then we are assured all
2234
# revisions referenced are present as they are installed in topological order.
2235
# and the tip revision was validated by get_ancestry.
2236
return required_topo_revisions
2238
# if we just grabbed the possibly available ids, then
2239
# we only have an estimate of whats available and need to validate
2240
# that against the revision records.
2241
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2244
class InterKnitRepo(InterSameDataRepository):
2245
"""Optimised code paths between Knit based repositories."""
2248
def _get_repo_format_to_test(self):
2249
from bzrlib.repofmt import knitrepo
2250
return knitrepo.RepositoryFormatKnit1()
2253
def is_compatible(source, target):
2254
"""Be compatible with known Knit formats.
2256
We don't test for the stores being of specific types because that
2257
could lead to confusing results, and there is no need to be
2260
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2262
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2263
isinstance(target._format, RepositoryFormatKnit))
2264
except AttributeError:
2266
return are_knits and InterRepository._same_model(source, target)
2269
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2270
"""See InterRepository.fetch()."""
2271
from bzrlib.fetch import KnitRepoFetcher
2272
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2273
self.source, self.source._format, self.target, self.target._format)
2274
f = KnitRepoFetcher(to_repository=self.target,
2275
from_repository=self.source,
2276
last_revision=revision_id,
2278
return f.count_copied, f.failed_revisions
2281
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2282
"""See InterRepository.missing_revision_ids()."""
2283
if revision_id is not None:
2284
source_ids = self.source.get_ancestry(revision_id)
2285
assert source_ids[0] is None
2288
source_ids = self.source.all_revision_ids()
2289
source_ids_set = set(source_ids)
2290
# source_ids is the worst possible case we may need to pull.
2291
# now we want to filter source_ids against what we actually
2292
# have in target, but don't try to check for existence where we know
2293
# we do not have a revision as that would be pointless.
2294
target_ids = set(self.target.all_revision_ids())
2295
possibly_present_revisions = target_ids.intersection(source_ids_set)
2296
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2297
required_revisions = source_ids_set.difference(actually_present_revisions)
2298
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2299
if revision_id is not None:
2300
# we used get_ancestry to determine source_ids then we are assured all
2301
# revisions referenced are present as they are installed in topological order.
2302
# and the tip revision was validated by get_ancestry.
2303
return required_topo_revisions
2305
# if we just grabbed the possibly available ids, then
2306
# we only have an estimate of whats available and need to validate
2307
# that against the revision records.
2308
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2311
class InterPackRepo(InterSameDataRepository):
2312
"""Optimised code paths between Pack based repositories."""
2315
def _get_repo_format_to_test(self):
2316
from bzrlib.repofmt import pack_repo
2317
return pack_repo.RepositoryFormatKnitPack1()
2320
def is_compatible(source, target):
2321
"""Be compatible with known Pack formats.
2323
We don't test for the stores being of specific types because that
2324
could lead to confusing results, and there is no need to be
2327
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2329
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2330
isinstance(target._format, RepositoryFormatPack))
2331
except AttributeError:
2333
return are_packs and InterRepository._same_model(source, target)
2336
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2337
"""See InterRepository.fetch()."""
2338
from bzrlib.repofmt.pack_repo import Packer
2339
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2340
self.source, self.source._format, self.target, self.target._format)
2341
self.count_copied = 0
2342
if revision_id is None:
2344
# everything to do - use pack logic
2345
# to fetch from all packs to one without
2346
# inventory parsing etc, IFF nothing to be copied is in the target.
2348
revision_ids = self.source.all_revision_ids()
2349
# implementing the TODO will involve:
2350
# - detecting when all of a pack is selected
2351
# - avoiding as much as possible pre-selection, so the
2352
# more-core routines such as create_pack_from_packs can filter in
2353
# a just-in-time fashion. (though having a HEADS list on a
2354
# repository might make this a lot easier, because we could
2355
# sensibly detect 'new revisions' without doing a full index scan.
2356
elif _mod_revision.is_null(revision_id):
2361
revision_ids = self.missing_revision_ids(revision_id,
2362
find_ghosts=find_ghosts)
2363
except errors.NoSuchRevision:
2364
raise errors.InstallFailed([revision_id])
2365
packs = self.source._pack_collection.all_packs()
2366
pack = Packer(self.target._pack_collection, packs, '.fetch',
2367
revision_ids).pack()
2368
if pack is not None:
2369
self.target._pack_collection._save_pack_names()
2370
# Trigger an autopack. This may duplicate effort as we've just done
2371
# a pack creation, but for now it is simpler to think about as
2372
# 'upload data, then repack if needed'.
2373
self.target._pack_collection.autopack()
2374
return (pack.get_revision_count(), [])
2379
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2380
"""See InterRepository.missing_revision_ids().
2382
:param find_ghosts: Find ghosts throughough the ancestry of
2385
if not find_ghosts and revision_id is not None:
2386
graph = self.source.get_graph()
2387
missing_revs = set()
2388
searcher = graph._make_breadth_first_searcher([revision_id])
2390
self.target._pack_collection.revision_index.combined_index
2391
null_set = frozenset([_mod_revision.NULL_REVISION])
2394
next_revs = set(searcher.next())
2395
except StopIteration:
2397
next_revs.difference_update(null_set)
2398
target_keys = [(key,) for key in next_revs]
2399
have_revs = frozenset(node[1][0] for node in
2400
target_index.iter_entries(target_keys))
2401
missing_revs.update(next_revs - have_revs)
2402
searcher.stop_searching_any(have_revs)
2403
if next_revs - have_revs == set([revision_id]):
2404
# we saw the start rev itself, but no parents from it (or
2405
# next_revs would have been updated to e.g. set(). We remove
2406
# have_revs because if we found revision_id locally we
2407
# stop_searching at the first time around.
2408
raise errors.NoSuchRevision(self.source, revision_id)
2410
elif revision_id is not None:
2411
source_ids = self.source.get_ancestry(revision_id)
2412
assert source_ids[0] is None
2415
source_ids = self.source.all_revision_ids()
2416
# source_ids is the worst possible case we may need to pull.
2417
# now we want to filter source_ids against what we actually
2418
# have in target, but don't try to check for existence where we know
2419
# we do not have a revision as that would be pointless.
2420
target_ids = set(self.target.all_revision_ids())
2421
return [r for r in source_ids if (r not in target_ids)]
2424
class InterModel1and2(InterRepository):
2427
def _get_repo_format_to_test(self):
2431
def is_compatible(source, target):
2432
if not source.supports_rich_root() and target.supports_rich_root():
2438
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2439
"""See InterRepository.fetch()."""
2440
from bzrlib.fetch import Model1toKnit2Fetcher
2441
f = Model1toKnit2Fetcher(to_repository=self.target,
2442
from_repository=self.source,
2443
last_revision=revision_id,
2445
return f.count_copied, f.failed_revisions
2448
def copy_content(self, revision_id=None):
2449
"""Make a complete copy of the content in self into destination.
2451
This is a destructive operation! Do not use it on existing
2454
:param revision_id: Only copy the content needed to construct
2455
revision_id and its parents.
2458
self.target.set_make_working_trees(self.source.make_working_trees())
2459
except NotImplementedError:
2461
# but don't bother fetching if we have the needed data now.
2462
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2463
self.target.has_revision(revision_id)):
2465
self.target.fetch(self.source, revision_id=revision_id)
2468
class InterKnit1and2(InterKnitRepo):
2471
def _get_repo_format_to_test(self):
2475
def is_compatible(source, target):
2476
"""Be compatible with Knit1 source and Knit3 target"""
2477
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2479
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2480
RepositoryFormatKnit3)
2481
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2482
RepositoryFormatKnitPack3)
2483
return (isinstance(source._format,
2484
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2485
isinstance(target._format,
2486
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2488
except AttributeError:
2492
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2493
"""See InterRepository.fetch()."""
2494
from bzrlib.fetch import Knit1to2Fetcher
2495
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2496
self.source, self.source._format, self.target,
2497
self.target._format)
2498
f = Knit1to2Fetcher(to_repository=self.target,
2499
from_repository=self.source,
2500
last_revision=revision_id,
2502
return f.count_copied, f.failed_revisions
2505
class InterRemoteToOther(InterRepository):
2507
def __init__(self, source, target):
2508
InterRepository.__init__(self, source, target)
2509
self._real_inter = None
2512
def is_compatible(source, target):
2513
if not isinstance(source, remote.RemoteRepository):
2515
source._ensure_real()
2516
real_source = source._real_repository
2517
# Is source's model compatible with target's model, and are they the
2518
# same format? Currently we can only optimise fetching from an
2519
# identical model & format repo.
2520
assert not isinstance(real_source, remote.RemoteRepository), (
2521
"We don't support remote repos backed by remote repos yet.")
2522
return real_source._format == target._format
2525
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2526
"""See InterRepository.fetch()."""
2527
from bzrlib.fetch import RemoteToOtherFetcher
2528
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2529
self.source, self.target, self.target._format)
2530
# TODO: jam 20070210 This should be an assert, not a translate
2531
revision_id = osutils.safe_revision_id(revision_id)
2532
f = RemoteToOtherFetcher(to_repository=self.target,
2533
from_repository=self.source,
2534
last_revision=revision_id,
2536
return f.count_copied, f.failed_revisions
2539
def _get_repo_format_to_test(self):
2543
class InterOtherToRemote(InterRepository):
2545
def __init__(self, source, target):
2546
InterRepository.__init__(self, source, target)
2547
self._real_inter = None
2550
def is_compatible(source, target):
2551
if isinstance(target, remote.RemoteRepository):
2555
def _ensure_real_inter(self):
2556
if self._real_inter is None:
2557
self.target._ensure_real()
2558
real_target = self.target._real_repository
2559
self._real_inter = InterRepository.get(self.source, real_target)
2561
def copy_content(self, revision_id=None):
2562
self._ensure_real_inter()
2563
self._real_inter.copy_content(revision_id=revision_id)
2565
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2566
self._ensure_real_inter()
2567
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2570
def _get_repo_format_to_test(self):
2574
InterRepository.register_optimiser(InterSameDataRepository)
2575
InterRepository.register_optimiser(InterWeaveRepo)
2576
InterRepository.register_optimiser(InterKnitRepo)
2577
InterRepository.register_optimiser(InterModel1and2)
2578
InterRepository.register_optimiser(InterKnit1and2)
2579
InterRepository.register_optimiser(InterPackRepo)
2580
InterRepository.register_optimiser(InterRemoteToOther)
2581
InterRepository.register_optimiser(InterOtherToRemote)
2584
class CopyConverter(object):
2585
"""A repository conversion tool which just performs a copy of the content.
2587
This is slow but quite reliable.
2590
def __init__(self, target_format):
2591
"""Create a CopyConverter.
2593
:param target_format: The format the resulting repository should be.
2595
self.target_format = target_format
2597
def convert(self, repo, pb):
2598
"""Perform the conversion of to_convert, giving feedback via pb.
2600
:param to_convert: The disk object to convert.
2601
:param pb: a progress bar to use for progress information.
2606
# this is only useful with metadir layouts - separated repo content.
2607
# trigger an assertion if not such
2608
repo._format.get_format_string()
2609
self.repo_dir = repo.bzrdir
2610
self.step('Moving repository to repository.backup')
2611
self.repo_dir.transport.move('repository', 'repository.backup')
2612
backup_transport = self.repo_dir.transport.clone('repository.backup')
2613
repo._format.check_conversion_target(self.target_format)
2614
self.source_repo = repo._format.open(self.repo_dir,
2616
_override_transport=backup_transport)
2617
self.step('Creating new repository')
2618
converted = self.target_format.initialize(self.repo_dir,
2619
self.source_repo.is_shared())
2620
converted.lock_write()
2622
self.step('Copying content into repository.')
2623
self.source_repo.copy_content_into(converted)
2626
self.step('Deleting old repository content.')
2627
self.repo_dir.transport.delete_tree('repository.backup')
2628
self.pb.note('repository converted')
2630
def step(self, message):
2631
"""Update the pb by a step."""
2633
self.pb.update(message, self.count, self.total)
2645
def _unescaper(match, _map=_unescape_map):
2646
code = match.group(1)
2650
if not code.startswith('#'):
2652
return unichr(int(code[1:])).encode('utf8')
2658
def _unescape_xml(data):
2659
"""Unescape predefined XML entities in a string of data."""
2661
if _unescape_re is None:
2662
_unescape_re = re.compile('\&([^;]*);')
2663
return _unescape_re.sub(_unescaper, data)
2666
class _RevisionTextVersionCache(object):
2667
"""A cache of the versionedfile versions for revision and file-id."""
2669
def __init__(self, repository):
2670
self.repository = repository
2671
self.revision_versions = {}
2672
self.revision_parents = {}
2673
self.repo_graph = self.repository.get_graph()
2674
# XXX: RBC: I haven't tracked down what uses this, but it would be
2675
# better to use the headscache directly I think.
2676
self.heads = graph.HeadsCache(self.repo_graph).heads
2678
def add_revision_text_versions(self, tree):
2679
"""Cache text version data from the supplied revision tree"""
2681
for path, entry in tree.iter_entries_by_dir():
2682
inv_revisions[entry.file_id] = entry.revision
2683
self.revision_versions[tree.get_revision_id()] = inv_revisions
2684
return inv_revisions
2686
def get_text_version(self, file_id, revision_id):
2687
"""Determine the text version for a given file-id and revision-id"""
2689
inv_revisions = self.revision_versions[revision_id]
2692
tree = self.repository.revision_tree(revision_id)
2693
except errors.RevisionNotPresent:
2694
self.revision_versions[revision_id] = inv_revisions = {}
2696
inv_revisions = self.add_revision_text_versions(tree)
2697
return inv_revisions.get(file_id)
2699
def prepopulate_revs(self, revision_ids):
2700
# Filter out versions that we don't have an inventory for, so that the
2701
# revision_trees() call won't fail.
2702
inv_weave = self.repository.get_inventory_weave()
2703
revs = [r for r in revision_ids if inv_weave.has_version(r)]
2704
# XXX: this loop is very similar to
2705
# bzrlib.fetch.Inter1and2Helper.iter_rev_trees.
2707
mutter('%d revisions left to prepopulate', len(revs))
2708
for tree in self.repository.revision_trees(revs[:100]):
2709
if tree.inventory.revision_id is None:
2710
tree.inventory.revision_id = tree.get_revision_id()
2711
self.add_revision_text_versions(tree)
2714
def get_parents(self, revision_id):
2716
return self.revision_parents[revision_id]
2718
parents = self.repository.get_parents([revision_id])[0]
2719
self.revision_parents[revision_id] = parents
2722
def used_file_versions(self):
2723
"""Return a set of (revision_id, file_id) pairs for each file version
2724
referenced by any inventory cached by this _RevisionTextVersionCache.
2726
If the entire repository has been cached, this can be used to find all
2727
file versions that are actually referenced by inventories. Thus any
2728
other file version is completely unused and can be removed safely.
2731
for inventory_summary in self.revision_versions.itervalues():
2732
result.update(inventory_summary.items())
2736
class VersionedFileChecker(object):
2738
def __init__(self, planned_revisions, revision_versions, repository):
2739
self.planned_revisions = planned_revisions
2740
self.revision_versions = revision_versions
2741
self.repository = repository
2743
def calculate_file_version_parents(self, revision_id, file_id):
2744
"""Calculate the correct parents for a file version according to
2747
text_revision = self.revision_versions.get_text_version(
2748
file_id, revision_id)
2749
if text_revision is None:
2751
parents_of_text_revision = self.revision_versions.get_parents(
2753
parents_from_inventories = []
2754
for parent in parents_of_text_revision:
2755
if parent == _mod_revision.NULL_REVISION:
2757
introduced_in = self.revision_versions.get_text_version(file_id,
2759
if introduced_in is not None:
2760
parents_from_inventories.append(introduced_in)
2761
heads = set(self.revision_versions.heads(parents_from_inventories))
2763
for parent in parents_from_inventories:
2764
if parent in heads and parent not in new_parents:
2765
new_parents.append(parent)
2766
return tuple(new_parents)
2768
def check_file_version_parents(self, weave, file_id):
2769
"""Check the parents stored in a versioned file are correct.
2771
It also detects file versions that are not referenced by their
2772
corresponding revision's inventory.
2774
:returns: A tuple of (wrong_parents, dangling_file_versions).
2775
wrong_parents is a dict mapping {revision_id: (stored_parents,
2776
correct_parents)} for each revision_id where the stored parents
2777
are not correct. dangling_file_versions is a set of (file_id,
2778
revision_id) tuples for versions that are present in this versioned
2779
file, but not used by the corresponding inventory.
2782
dangling_file_versions = set()
2783
for num, revision_id in enumerate(self.planned_revisions):
2784
correct_parents = self.calculate_file_version_parents(
2785
revision_id, file_id)
2786
if correct_parents is None:
2788
text_revision = self.revision_versions.get_text_version(
2789
file_id, revision_id)
2791
knit_parents = tuple(weave.get_parents(revision_id))
2792
except errors.RevisionNotPresent:
2794
if text_revision != revision_id:
2795
# This file version is not referenced by its corresponding
2797
dangling_file_versions.add((file_id, revision_id))
2798
if correct_parents != knit_parents:
2799
wrong_parents[revision_id] = (knit_parents, correct_parents)
2800
return wrong_parents, dangling_file_versions
added
removed
bzrlib/repository.py
