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):
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(revision_id)
806
"""Open the repository rooted at base.
808
For instance, if the repository is at URL/.bzr/repository,
809
Repository.open(URL) -> a Repository instance.
811
control = bzrdir.BzrDir.open(base)
812
return control.open_repository()
814
def copy_content_into(self, destination, revision_id=None):
815
"""Make a complete copy of the content in self into destination.
817
This is a destructive operation! Do not use it on existing
820
return InterRepository.get(self, destination).copy_content(revision_id)
822
def commit_write_group(self):
823
"""Commit the contents accrued within the current write group.
825
:seealso: start_write_group.
827
if self._write_group is not self.get_transaction():
828
# has an unlock or relock occured ?
829
raise errors.BzrError('mismatched lock context %r and '
831
(self.get_transaction(), self._write_group))
832
self._commit_write_group()
833
self._write_group = None
835
def _commit_write_group(self):
836
"""Template method for per-repository write group cleanup.
838
This is called before the write group is considered to be
839
finished and should ensure that all data handed to the repository
840
for writing during the write group is safely committed (to the
841
extent possible considering file system caching etc).
844
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
845
"""Fetch the content required to construct revision_id from source.
847
If revision_id is None all content is copied.
848
:param find_ghosts: Find and copy revisions in the source that are
849
ghosts in the target (and not reachable directly by walking out to
850
the first-present revision in target from revision_id).
852
# fast path same-url fetch operations
853
if self.has_same_location(source):
854
# check that last_revision is in 'from' and then return a
856
if (revision_id is not None and
857
not _mod_revision.is_null(revision_id)):
858
self.get_revision(revision_id)
860
inter = InterRepository.get(source, self)
862
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
863
except NotImplementedError:
864
raise errors.IncompatibleRepositories(source, self)
866
def create_bundle(self, target, base, fileobj, format=None):
867
return serializer.write_bundle(self, target, base, fileobj, format)
869
def get_commit_builder(self, branch, parents, config, timestamp=None,
870
timezone=None, committer=None, revprops=None,
872
"""Obtain a CommitBuilder for this repository.
874
:param branch: Branch to commit to.
875
:param parents: Revision ids of the parents of the new revision.
876
:param config: Configuration to use.
877
:param timestamp: Optional timestamp recorded for commit.
878
:param timezone: Optional timezone for timestamp.
879
:param committer: Optional committer to set for commit.
880
:param revprops: Optional dictionary of revision properties.
881
:param revision_id: Optional revision id.
883
result = self._commit_builder_class(self, parents, config,
884
timestamp, timezone, committer, revprops, revision_id)
885
self.start_write_group()
889
if (self.control_files._lock_count == 1 and
890
self.control_files._lock_mode == 'w'):
891
if self._write_group is not None:
892
self.abort_write_group()
893
self.control_files.unlock()
894
raise errors.BzrError(
895
'Must end write groups before releasing write locks.')
896
self.control_files.unlock()
899
def clone(self, a_bzrdir, revision_id=None):
900
"""Clone this repository into a_bzrdir using the current format.
902
Currently no check is made that the format of this repository and
903
the bzrdir format are compatible. FIXME RBC 20060201.
905
:return: The newly created destination repository.
907
# TODO: deprecate after 0.16; cloning this with all its settings is
908
# probably not very useful -- mbp 20070423
909
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
910
self.copy_content_into(dest_repo, revision_id)
913
def start_write_group(self):
914
"""Start a write group in the repository.
916
Write groups are used by repositories which do not have a 1:1 mapping
917
between file ids and backend store to manage the insertion of data from
918
both fetch and commit operations.
920
A write lock is required around the start_write_group/commit_write_group
921
for the support of lock-requiring repository formats.
923
One can only insert data into a repository inside a write group.
927
if not self.is_write_locked():
928
raise errors.NotWriteLocked(self)
929
if self._write_group:
930
raise errors.BzrError('already in a write group')
931
self._start_write_group()
932
# so we can detect unlock/relock - the write group is now entered.
933
self._write_group = self.get_transaction()
935
def _start_write_group(self):
936
"""Template method for per-repository write group startup.
938
This is called before the write group is considered to be
943
def sprout(self, to_bzrdir, revision_id=None):
944
"""Create a descendent repository for new development.
946
Unlike clone, this does not copy the settings of the repository.
948
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
949
dest_repo.fetch(self, revision_id=revision_id)
952
def _create_sprouting_repo(self, a_bzrdir, shared):
953
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
954
# use target default format.
955
dest_repo = a_bzrdir.create_repository()
957
# Most control formats need the repository to be specifically
958
# created, but on some old all-in-one formats it's not needed
960
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
961
except errors.UninitializableFormat:
962
dest_repo = a_bzrdir.open_repository()
966
def has_revision(self, revision_id):
967
"""True if this repository has a copy of the revision."""
968
if 'evil' in debug.debug_flags:
969
mutter_callsite(3, "has_revision is a LBYL symptom.")
970
return self._revision_store.has_revision_id(revision_id,
971
self.get_transaction())
974
def get_revision(self, revision_id):
975
"""Return the Revision object for a named revision."""
976
return self.get_revisions([revision_id])[0]
979
def get_revision_reconcile(self, revision_id):
980
"""'reconcile' helper routine that allows access to a revision always.
982
This variant of get_revision does not cross check the weave graph
983
against the revision one as get_revision does: but it should only
984
be used by reconcile, or reconcile-alike commands that are correcting
985
or testing the revision graph.
987
return self._get_revisions([revision_id])[0]
990
def get_revisions(self, revision_ids):
991
"""Get many revisions at once."""
992
return self._get_revisions(revision_ids)
995
def _get_revisions(self, revision_ids):
996
"""Core work logic to get many revisions without sanity checks."""
997
for rev_id in revision_ids:
998
if not rev_id or not isinstance(rev_id, basestring):
999
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1000
revs = self._revision_store.get_revisions(revision_ids,
1001
self.get_transaction())
1003
assert not isinstance(rev.revision_id, unicode)
1004
for parent_id in rev.parent_ids:
1005
assert not isinstance(parent_id, unicode)
1009
def get_revision_xml(self, revision_id):
1010
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1011
# would have already do it.
1012
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1013
rev = self.get_revision(revision_id)
1014
rev_tmp = StringIO()
1015
# the current serializer..
1016
self._revision_store._serializer.write_revision(rev, rev_tmp)
1018
return rev_tmp.getvalue()
1021
def get_deltas_for_revisions(self, revisions):
1022
"""Produce a generator of revision deltas.
1024
Note that the input is a sequence of REVISIONS, not revision_ids.
1025
Trees will be held in memory until the generator exits.
1026
Each delta is relative to the revision's lefthand predecessor.
1028
required_trees = set()
1029
for revision in revisions:
1030
required_trees.add(revision.revision_id)
1031
required_trees.update(revision.parent_ids[:1])
1032
trees = dict((t.get_revision_id(), t) for
1033
t in self.revision_trees(required_trees))
1034
for revision in revisions:
1035
if not revision.parent_ids:
1036
old_tree = self.revision_tree(None)
1038
old_tree = trees[revision.parent_ids[0]]
1039
yield trees[revision.revision_id].changes_from(old_tree)
1042
def get_revision_delta(self, revision_id):
1043
"""Return the delta for one revision.
1045
The delta is relative to the left-hand predecessor of the
1048
r = self.get_revision(revision_id)
1049
return list(self.get_deltas_for_revisions([r]))[0]
1052
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1053
signature = gpg_strategy.sign(plaintext)
1054
self.add_signature_text(revision_id, signature)
1057
def add_signature_text(self, revision_id, signature):
1058
self._revision_store.add_revision_signature_text(revision_id,
1060
self.get_transaction())
1062
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1064
"""Helper routine for fileids_altered_by_revision_ids.
1066
This performs the translation of xml lines to revision ids.
1068
:param line_iterator: An iterator of lines, origin_version_id
1069
:param revision_ids: The revision ids to filter for. This should be a
1070
set or other type which supports efficient __contains__ lookups, as
1071
the revision id from each parsed line will be looked up in the
1072
revision_ids filter.
1073
:return: a dictionary mapping altered file-ids to an iterable of
1074
revision_ids. Each altered file-ids has the exact revision_ids that
1075
altered it listed explicitly.
1079
# this code needs to read every new line in every inventory for the
1080
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1081
# not present in one of those inventories is unnecessary but not
1082
# harmful because we are filtering by the revision id marker in the
1083
# inventory lines : we only select file ids altered in one of those
1084
# revisions. We don't need to see all lines in the inventory because
1085
# only those added in an inventory in rev X can contain a revision=X
1087
unescape_revid_cache = {}
1088
unescape_fileid_cache = {}
1090
# jam 20061218 In a big fetch, this handles hundreds of thousands
1091
# of lines, so it has had a lot of inlining and optimizing done.
1092
# Sorry that it is a little bit messy.
1093
# Move several functions to be local variables, since this is a long
1095
search = self._file_ids_altered_regex.search
1096
unescape = _unescape_xml
1097
setdefault = result.setdefault
1098
for line, version_id in line_iterator:
1099
match = search(line)
1102
# One call to match.group() returning multiple items is quite a
1103
# bit faster than 2 calls to match.group() each returning 1
1104
file_id, revision_id = match.group('file_id', 'revision_id')
1106
# Inlining the cache lookups helps a lot when you make 170,000
1107
# lines and 350k ids, versus 8.4 unique ids.
1108
# Using a cache helps in 2 ways:
1109
# 1) Avoids unnecessary decoding calls
1110
# 2) Re-uses cached strings, which helps in future set and
1112
# (2) is enough that removing encoding entirely along with
1113
# the cache (so we are using plain strings) results in no
1114
# performance improvement.
1116
revision_id = unescape_revid_cache[revision_id]
1118
unescaped = unescape(revision_id)
1119
unescape_revid_cache[revision_id] = unescaped
1120
revision_id = unescaped
1122
# once data is all ensured-consistent; then this is
1123
# if revision_id == version_id
1124
if revision_id in revision_ids:
1126
file_id = unescape_fileid_cache[file_id]
1128
unescaped = unescape(file_id)
1129
unescape_fileid_cache[file_id] = unescaped
1131
setdefault(file_id, set()).add(revision_id)
1134
def fileids_altered_by_revision_ids(self, revision_ids):
1135
"""Find the file ids and versions affected by revisions.
1137
:param revisions: an iterable containing revision ids.
1138
:return: a dictionary mapping altered file-ids to an iterable of
1139
revision_ids. Each altered file-ids has the exact revision_ids that
1140
altered it listed explicitly.
1142
assert self._serializer.support_altered_by_hack, \
1143
("fileids_altered_by_revision_ids only supported for branches "
1144
"which store inventory as unnested xml, not on %r" % self)
1145
selected_revision_ids = set(revision_ids)
1146
w = self.get_inventory_weave()
1147
pb = ui.ui_factory.nested_progress_bar()
1149
return self._find_file_ids_from_xml_inventory_lines(
1150
w.iter_lines_added_or_present_in_versions(
1151
selected_revision_ids, pb=pb),
1152
selected_revision_ids)
1156
def iter_files_bytes(self, desired_files):
1157
"""Iterate through file versions.
1159
Files will not necessarily be returned in the order they occur in
1160
desired_files. No specific order is guaranteed.
1162
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1163
value supplied by the caller as part of desired_files. It should
1164
uniquely identify the file version in the caller's context. (Examples:
1165
an index number or a TreeTransform trans_id.)
1167
bytes_iterator is an iterable of bytestrings for the file. The
1168
kind of iterable and length of the bytestrings are unspecified, but for
1169
this implementation, it is a list of lines produced by
1170
VersionedFile.get_lines().
1172
:param desired_files: a list of (file_id, revision_id, identifier)
1175
transaction = self.get_transaction()
1176
for file_id, revision_id, callable_data in desired_files:
1178
weave = self.weave_store.get_weave(file_id, transaction)
1179
except errors.NoSuchFile:
1180
raise errors.NoSuchIdInRepository(self, file_id)
1181
yield callable_data, weave.get_lines(revision_id)
1183
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1184
"""Get an iterable listing the keys of all the data introduced by a set
1187
The keys will be ordered so that the corresponding items can be safely
1188
fetched and inserted in that order.
1190
:returns: An iterable producing tuples of (knit-kind, file-id,
1191
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1192
'revisions'. file-id is None unless knit-kind is 'file'.
1194
# XXX: it's a bit weird to control the inventory weave caching in this
1195
# generator. Ideally the caching would be done in fetch.py I think. Or
1196
# maybe this generator should explicitly have the contract that it
1197
# should not be iterated until the previously yielded item has been
1200
inv_w = self.get_inventory_weave()
1201
inv_w.enable_cache()
1203
# file ids that changed
1204
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1206
num_file_ids = len(file_ids)
1207
for file_id, altered_versions in file_ids.iteritems():
1208
if _files_pb is not None:
1209
_files_pb.update("fetch texts", count, num_file_ids)
1211
yield ("file", file_id, altered_versions)
1212
# We're done with the files_pb. Note that it finished by the caller,
1213
# just as it was created by the caller.
1217
yield ("inventory", None, revision_ids)
1221
revisions_with_signatures = set()
1222
for rev_id in revision_ids:
1224
self.get_signature_text(rev_id)
1225
except errors.NoSuchRevision:
1229
revisions_with_signatures.add(rev_id)
1231
yield ("signatures", None, revisions_with_signatures)
1234
yield ("revisions", None, revision_ids)
1237
def get_inventory_weave(self):
1238
return self.control_weaves.get_weave('inventory',
1239
self.get_transaction())
1242
def get_inventory(self, revision_id):
1243
"""Get Inventory object by hash."""
1244
return self.deserialise_inventory(
1245
revision_id, self.get_inventory_xml(revision_id))
1247
def deserialise_inventory(self, revision_id, xml):
1248
"""Transform the xml into an inventory object.
1250
:param revision_id: The expected revision id of the inventory.
1251
:param xml: A serialised inventory.
1253
return self._serializer.read_inventory_from_string(xml, revision_id)
1255
def serialise_inventory(self, inv):
1256
return self._serializer.write_inventory_to_string(inv)
1258
def _serialise_inventory_to_lines(self, inv):
1259
return self._serializer.write_inventory_to_lines(inv)
1261
def get_serializer_format(self):
1262
return self._serializer.format_num
1265
def get_inventory_xml(self, revision_id):
1266
"""Get inventory XML as a file object."""
1268
assert isinstance(revision_id, str), type(revision_id)
1269
iw = self.get_inventory_weave()
1270
return iw.get_text(revision_id)
1272
raise errors.HistoryMissing(self, 'inventory', revision_id)
1275
def get_inventory_sha1(self, revision_id):
1276
"""Return the sha1 hash of the inventory entry
1278
return self.get_revision(revision_id).inventory_sha1
1281
def get_revision_graph(self, revision_id=None):
1282
"""Return a dictionary containing the revision graph.
1284
NB: This method should not be used as it accesses the entire graph all
1285
at once, which is much more data than most operations should require.
1287
:param revision_id: The revision_id to get a graph from. If None, then
1288
the entire revision graph is returned. This is a deprecated mode of
1289
operation and will be removed in the future.
1290
:return: a dictionary of revision_id->revision_parents_list.
1292
raise NotImplementedError(self.get_revision_graph)
1295
def get_revision_graph_with_ghosts(self, revision_ids=None):
1296
"""Return a graph of the revisions with ghosts marked as applicable.
1298
:param revision_ids: an iterable of revisions to graph or None for all.
1299
:return: a Graph object with the graph reachable from revision_ids.
1301
if 'evil' in debug.debug_flags:
1303
"get_revision_graph_with_ghosts scales with size of history.")
1304
result = deprecated_graph.Graph()
1305
if not revision_ids:
1306
pending = set(self.all_revision_ids())
1309
pending = set(revision_ids)
1310
# special case NULL_REVISION
1311
if _mod_revision.NULL_REVISION in pending:
1312
pending.remove(_mod_revision.NULL_REVISION)
1313
required = set(pending)
1316
revision_id = pending.pop()
1318
rev = self.get_revision(revision_id)
1319
except errors.NoSuchRevision:
1320
if revision_id in required:
1323
result.add_ghost(revision_id)
1325
for parent_id in rev.parent_ids:
1326
# is this queued or done ?
1327
if (parent_id not in pending and
1328
parent_id not in done):
1330
pending.add(parent_id)
1331
result.add_node(revision_id, rev.parent_ids)
1332
done.add(revision_id)
1335
def _get_history_vf(self):
1336
"""Get a versionedfile whose history graph reflects all revisions.
1338
For weave repositories, this is the inventory weave.
1340
return self.get_inventory_weave()
1342
def iter_reverse_revision_history(self, revision_id):
1343
"""Iterate backwards through revision ids in the lefthand history
1345
:param revision_id: The revision id to start with. All its lefthand
1346
ancestors will be traversed.
1348
if revision_id in (None, _mod_revision.NULL_REVISION):
1350
next_id = revision_id
1351
versionedfile = self._get_history_vf()
1354
parents = versionedfile.get_parents(next_id)
1355
if len(parents) == 0:
1358
next_id = parents[0]
1361
def get_revision_inventory(self, revision_id):
1362
"""Return inventory of a past revision."""
1363
# TODO: Unify this with get_inventory()
1364
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1365
# must be the same as its revision, so this is trivial.
1366
if revision_id is None:
1367
# This does not make sense: if there is no revision,
1368
# then it is the current tree inventory surely ?!
1369
# and thus get_root_id() is something that looks at the last
1370
# commit on the branch, and the get_root_id is an inventory check.
1371
raise NotImplementedError
1372
# return Inventory(self.get_root_id())
1374
return self.get_inventory(revision_id)
1377
def is_shared(self):
1378
"""Return True if this repository is flagged as a shared repository."""
1379
raise NotImplementedError(self.is_shared)
1382
def reconcile(self, other=None, thorough=False):
1383
"""Reconcile this repository."""
1384
from bzrlib.reconcile import RepoReconciler
1385
reconciler = RepoReconciler(self, thorough=thorough)
1386
reconciler.reconcile()
1389
def _refresh_data(self):
1390
"""Helper called from lock_* to ensure coherency with disk.
1392
The default implementation does nothing; it is however possible
1393
for repositories to maintain loaded indices across multiple locks
1394
by checking inside their implementation of this method to see
1395
whether their indices are still valid. This depends of course on
1396
the disk format being validatable in this manner.
1400
def revision_tree(self, revision_id):
1401
"""Return Tree for a revision on this branch.
1403
`revision_id` may be None for the empty tree revision.
1405
# TODO: refactor this to use an existing revision object
1406
# so we don't need to read it in twice.
1407
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1408
return RevisionTree(self, Inventory(root_id=None),
1409
_mod_revision.NULL_REVISION)
1411
inv = self.get_revision_inventory(revision_id)
1412
return RevisionTree(self, inv, revision_id)
1415
def revision_trees(self, revision_ids):
1416
"""Return Tree for a revision on this branch.
1418
`revision_id` may not be None or 'null:'"""
1419
assert None not in revision_ids
1420
assert _mod_revision.NULL_REVISION not in revision_ids
1421
texts = self.get_inventory_weave().get_texts(revision_ids)
1422
for text, revision_id in zip(texts, revision_ids):
1423
inv = self.deserialise_inventory(revision_id, text)
1424
yield RevisionTree(self, inv, revision_id)
1427
def get_ancestry(self, revision_id, topo_sorted=True):
1428
"""Return a list of revision-ids integrated by a revision.
1430
The first element of the list is always None, indicating the origin
1431
revision. This might change when we have history horizons, or
1432
perhaps we should have a new API.
1434
This is topologically sorted.
1436
if _mod_revision.is_null(revision_id):
1438
if not self.has_revision(revision_id):
1439
raise errors.NoSuchRevision(self, revision_id)
1440
w = self.get_inventory_weave()
1441
candidates = w.get_ancestry(revision_id, topo_sorted)
1442
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1445
"""Compress the data within the repository.
1447
This operation only makes sense for some repository types. For other
1448
types it should be a no-op that just returns.
1450
This stub method does not require a lock, but subclasses should use
1451
@needs_write_lock as this is a long running call its reasonable to
1452
implicitly lock for the user.
1456
def print_file(self, file, revision_id):
1457
"""Print `file` to stdout.
1459
FIXME RBC 20060125 as John Meinel points out this is a bad api
1460
- it writes to stdout, it assumes that that is valid etc. Fix
1461
by creating a new more flexible convenience function.
1463
tree = self.revision_tree(revision_id)
1464
# use inventory as it was in that revision
1465
file_id = tree.inventory.path2id(file)
1467
# TODO: jam 20060427 Write a test for this code path
1468
# it had a bug in it, and was raising the wrong
1470
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1471
tree.print_file(file_id)
1473
def get_transaction(self):
1474
return self.control_files.get_transaction()
1476
def revision_parents(self, revision_id):
1477
return self.get_inventory_weave().parent_names(revision_id)
1479
def get_parents(self, revision_ids):
1480
"""See StackedParentsProvider.get_parents"""
1482
for revision_id in revision_ids:
1483
if revision_id == _mod_revision.NULL_REVISION:
1487
parents = self.get_revision(revision_id).parent_ids
1488
except errors.NoSuchRevision:
1491
if len(parents) == 0:
1492
parents = [_mod_revision.NULL_REVISION]
1493
parents_list.append(parents)
1496
def _make_parents_provider(self):
1499
def get_graph(self, other_repository=None):
1500
"""Return the graph walker for this repository format"""
1501
parents_provider = self._make_parents_provider()
1502
if (other_repository is not None and
1503
other_repository.bzrdir.transport.base !=
1504
self.bzrdir.transport.base):
1505
parents_provider = graph._StackedParentsProvider(
1506
[parents_provider, other_repository._make_parents_provider()])
1507
return graph.Graph(parents_provider)
1509
def get_versioned_file_checker(self, revisions, revision_versions_cache):
1510
return VersionedFileChecker(revisions, revision_versions_cache, self)
1513
def set_make_working_trees(self, new_value):
1514
"""Set the policy flag for making working trees when creating branches.
1516
This only applies to branches that use this repository.
1518
The default is 'True'.
1519
:param new_value: True to restore the default, False to disable making
1522
raise NotImplementedError(self.set_make_working_trees)
1524
def make_working_trees(self):
1525
"""Returns the policy for making working trees on new branches."""
1526
raise NotImplementedError(self.make_working_trees)
1529
def sign_revision(self, revision_id, gpg_strategy):
1530
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1531
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1534
def has_signature_for_revision_id(self, revision_id):
1535
"""Query for a revision signature for revision_id in the repository."""
1536
return self._revision_store.has_signature(revision_id,
1537
self.get_transaction())
1540
def get_signature_text(self, revision_id):
1541
"""Return the text for a signature."""
1542
return self._revision_store.get_signature_text(revision_id,
1543
self.get_transaction())
1546
def check(self, revision_ids=None):
1547
"""Check consistency of all history of given revision_ids.
1549
Different repository implementations should override _check().
1551
:param revision_ids: A non-empty list of revision_ids whose ancestry
1552
will be checked. Typically the last revision_id of a branch.
1554
return self._check(revision_ids)
1556
def _check(self, revision_ids):
1557
result = check.Check(self)
1561
def _warn_if_deprecated(self):
1562
global _deprecation_warning_done
1563
if _deprecation_warning_done:
1565
_deprecation_warning_done = True
1566
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1567
% (self._format, self.bzrdir.transport.base))
1569
def supports_rich_root(self):
1570
return self._format.rich_root_data
1572
def _check_ascii_revisionid(self, revision_id, method):
1573
"""Private helper for ascii-only repositories."""
1574
# weave repositories refuse to store revisionids that are non-ascii.
1575
if revision_id is not None:
1576
# weaves require ascii revision ids.
1577
if isinstance(revision_id, unicode):
1579
revision_id.encode('ascii')
1580
except UnicodeEncodeError:
1581
raise errors.NonAsciiRevisionId(method, self)
1584
revision_id.decode('ascii')
1585
except UnicodeDecodeError:
1586
raise errors.NonAsciiRevisionId(method, self)
1588
def revision_graph_can_have_wrong_parents(self):
1589
"""Is it possible for this repository to have a revision graph with
1592
If True, then this repository must also implement
1593
_find_inconsistent_revision_parents so that check and reconcile can
1594
check for inconsistencies before proceeding with other checks that may
1595
depend on the revision index being consistent.
1597
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1599
# remove these delegates a while after bzr 0.15
1600
def __make_delegated(name, from_module):
1601
def _deprecated_repository_forwarder():
1602
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1603
% (name, from_module),
1606
m = __import__(from_module, globals(), locals(), [name])
1608
return getattr(m, name)
1609
except AttributeError:
1610
raise AttributeError('module %s has no name %s'
1612
globals()[name] = _deprecated_repository_forwarder
1615
'AllInOneRepository',
1616
'WeaveMetaDirRepository',
1617
'PreSplitOutRepositoryFormat',
1618
'RepositoryFormat4',
1619
'RepositoryFormat5',
1620
'RepositoryFormat6',
1621
'RepositoryFormat7',
1623
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1627
'RepositoryFormatKnit',
1628
'RepositoryFormatKnit1',
1630
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1633
def install_revision(repository, rev, revision_tree):
1634
"""Install all revision data into a repository."""
1635
install_revisions(repository, [(rev, revision_tree, None)])
1638
def install_revisions(repository, iterable):
1639
"""Install all revision data into a repository.
1641
Accepts an iterable of revision, tree, signature tuples. The signature
1644
repository.start_write_group()
1646
for revision, revision_tree, signature in iterable:
1647
_install_revision(repository, revision, revision_tree, signature)
1649
repository.abort_write_group()
1652
repository.commit_write_group()
1655
def _install_revision(repository, rev, revision_tree, signature):
1656
"""Install all revision data into a repository."""
1657
present_parents = []
1659
for p_id in rev.parent_ids:
1660
if repository.has_revision(p_id):
1661
present_parents.append(p_id)
1662
parent_trees[p_id] = repository.revision_tree(p_id)
1664
parent_trees[p_id] = repository.revision_tree(None)
1666
inv = revision_tree.inventory
1667
entries = inv.iter_entries()
1668
# backwards compatibility hack: skip the root id.
1669
if not repository.supports_rich_root():
1670
path, root = entries.next()
1671
if root.revision != rev.revision_id:
1672
raise errors.IncompatibleRevision(repr(repository))
1673
# Add the texts that are not already present
1674
for path, ie in entries:
1675
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1676
repository.get_transaction())
1677
if ie.revision not in w:
1679
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1680
# with InventoryEntry.find_previous_heads(). if it is, then there
1681
# is a latent bug here where the parents may have ancestors of each
1683
for revision, tree in parent_trees.iteritems():
1684
if ie.file_id not in tree:
1686
parent_id = tree.inventory[ie.file_id].revision
1687
if parent_id in text_parents:
1689
text_parents.append(parent_id)
1691
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1692
repository.get_transaction())
1693
lines = revision_tree.get_file(ie.file_id).readlines()
1694
vfile.add_lines(rev.revision_id, text_parents, lines)
1696
# install the inventory
1697
repository.add_inventory(rev.revision_id, inv, present_parents)
1698
except errors.RevisionAlreadyPresent:
1700
if signature is not None:
1701
repository.add_signature(rev.revision_id, signature)
1702
repository.add_revision(rev.revision_id, rev, inv)
1705
class MetaDirRepository(Repository):
1706
"""Repositories in the new meta-dir layout."""
1708
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1709
super(MetaDirRepository, self).__init__(_format,
1715
dir_mode = self.control_files._dir_mode
1716
file_mode = self.control_files._file_mode
1719
def is_shared(self):
1720
"""Return True if this repository is flagged as a shared repository."""
1721
return self.control_files._transport.has('shared-storage')
1724
def set_make_working_trees(self, new_value):
1725
"""Set the policy flag for making working trees when creating branches.
1727
This only applies to branches that use this repository.
1729
The default is 'True'.
1730
:param new_value: True to restore the default, False to disable making
1735
self.control_files._transport.delete('no-working-trees')
1736
except errors.NoSuchFile:
1739
self.control_files.put_utf8('no-working-trees', '')
1741
def make_working_trees(self):
1742
"""Returns the policy for making working trees on new branches."""
1743
return not self.control_files._transport.has('no-working-trees')
1746
class RepositoryFormatRegistry(registry.Registry):
1747
"""Registry of RepositoryFormats."""
1749
def get(self, format_string):
1750
r = registry.Registry.get(self, format_string)
1756
format_registry = RepositoryFormatRegistry()
1757
"""Registry of formats, indexed by their identifying format string.
1759
This can contain either format instances themselves, or classes/factories that
1760
can be called to obtain one.
1764
#####################################################################
1765
# Repository Formats
1767
class RepositoryFormat(object):
1768
"""A repository format.
1770
Formats provide three things:
1771
* An initialization routine to construct repository data on disk.
1772
* a format string which is used when the BzrDir supports versioned
1774
* an open routine which returns a Repository instance.
1776
There is one and only one Format subclass for each on-disk format. But
1777
there can be one Repository subclass that is used for several different
1778
formats. The _format attribute on a Repository instance can be used to
1779
determine the disk format.
1781
Formats are placed in an dict by their format string for reference
1782
during opening. These should be subclasses of RepositoryFormat
1785
Once a format is deprecated, just deprecate the initialize and open
1786
methods on the format class. Do not deprecate the object, as the
1787
object will be created every system load.
1789
Common instance attributes:
1790
_matchingbzrdir - the bzrdir format that the repository format was
1791
originally written to work with. This can be used if manually
1792
constructing a bzrdir and repository, or more commonly for test suite
1796
# Set to True or False in derived classes. True indicates that the format
1797
# supports ghosts gracefully.
1798
supports_ghosts = None
1801
return "<%s>" % self.__class__.__name__
1803
def __eq__(self, other):
1804
# format objects are generally stateless
1805
return isinstance(other, self.__class__)
1807
def __ne__(self, other):
1808
return not self == other
1811
def find_format(klass, a_bzrdir):
1812
"""Return the format for the repository object in a_bzrdir.
1814
This is used by bzr native formats that have a "format" file in
1815
the repository. Other methods may be used by different types of
1819
transport = a_bzrdir.get_repository_transport(None)
1820
format_string = transport.get("format").read()
1821
return format_registry.get(format_string)
1822
except errors.NoSuchFile:
1823
raise errors.NoRepositoryPresent(a_bzrdir)
1825
raise errors.UnknownFormatError(format=format_string)
1828
def register_format(klass, format):
1829
format_registry.register(format.get_format_string(), format)
1832
def unregister_format(klass, format):
1833
format_registry.remove(format.get_format_string())
1836
def get_default_format(klass):
1837
"""Return the current default format."""
1838
from bzrlib import bzrdir
1839
return bzrdir.format_registry.make_bzrdir('default').repository_format
1841
def _get_control_store(self, repo_transport, control_files):
1842
"""Return the control store for this repository."""
1843
raise NotImplementedError(self._get_control_store)
1845
def get_format_string(self):
1846
"""Return the ASCII format string that identifies this format.
1848
Note that in pre format ?? repositories the format string is
1849
not permitted nor written to disk.
1851
raise NotImplementedError(self.get_format_string)
1853
def get_format_description(self):
1854
"""Return the short description for this format."""
1855
raise NotImplementedError(self.get_format_description)
1857
def _get_revision_store(self, repo_transport, control_files):
1858
"""Return the revision store object for this a_bzrdir."""
1859
raise NotImplementedError(self._get_revision_store)
1861
def _get_text_rev_store(self,
1868
"""Common logic for getting a revision store for a repository.
1870
see self._get_revision_store for the subclass-overridable method to
1871
get the store for a repository.
1873
from bzrlib.store.revision.text import TextRevisionStore
1874
dir_mode = control_files._dir_mode
1875
file_mode = control_files._file_mode
1876
text_store = TextStore(transport.clone(name),
1878
compressed=compressed,
1880
file_mode=file_mode)
1881
_revision_store = TextRevisionStore(text_store, serializer)
1882
return _revision_store
1884
# TODO: this shouldn't be in the base class, it's specific to things that
1885
# use weaves or knits -- mbp 20070207
1886
def _get_versioned_file_store(self,
1891
versionedfile_class=None,
1892
versionedfile_kwargs={},
1894
if versionedfile_class is None:
1895
versionedfile_class = self._versionedfile_class
1896
weave_transport = control_files._transport.clone(name)
1897
dir_mode = control_files._dir_mode
1898
file_mode = control_files._file_mode
1899
return VersionedFileStore(weave_transport, prefixed=prefixed,
1901
file_mode=file_mode,
1902
versionedfile_class=versionedfile_class,
1903
versionedfile_kwargs=versionedfile_kwargs,
1906
def initialize(self, a_bzrdir, shared=False):
1907
"""Initialize a repository of this format in a_bzrdir.
1909
:param a_bzrdir: The bzrdir to put the new repository in it.
1910
:param shared: The repository should be initialized as a sharable one.
1911
:returns: The new repository object.
1913
This may raise UninitializableFormat if shared repository are not
1914
compatible the a_bzrdir.
1916
raise NotImplementedError(self.initialize)
1918
def is_supported(self):
1919
"""Is this format supported?
1921
Supported formats must be initializable and openable.
1922
Unsupported formats may not support initialization or committing or
1923
some other features depending on the reason for not being supported.
1927
def check_conversion_target(self, target_format):
1928
raise NotImplementedError(self.check_conversion_target)
1930
def open(self, a_bzrdir, _found=False):
1931
"""Return an instance of this format for the bzrdir a_bzrdir.
1933
_found is a private parameter, do not use it.
1935
raise NotImplementedError(self.open)
1938
class MetaDirRepositoryFormat(RepositoryFormat):
1939
"""Common base class for the new repositories using the metadir layout."""
1941
rich_root_data = False
1942
supports_tree_reference = False
1943
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
1946
super(MetaDirRepositoryFormat, self).__init__()
1948
def _create_control_files(self, a_bzrdir):
1949
"""Create the required files and the initial control_files object."""
1950
# FIXME: RBC 20060125 don't peek under the covers
1951
# NB: no need to escape relative paths that are url safe.
1952
repository_transport = a_bzrdir.get_repository_transport(self)
1953
control_files = lockable_files.LockableFiles(repository_transport,
1954
'lock', lockdir.LockDir)
1955
control_files.create_lock()
1956
return control_files
1958
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
1959
"""Upload the initial blank content."""
1960
control_files = self._create_control_files(a_bzrdir)
1961
control_files.lock_write()
1963
control_files._transport.mkdir_multi(dirs,
1964
mode=control_files._dir_mode)
1965
for file, content in files:
1966
control_files.put(file, content)
1967
for file, content in utf8_files:
1968
control_files.put_utf8(file, content)
1970
control_files.put_utf8('shared-storage', '')
1972
control_files.unlock()
1975
# formats which have no format string are not discoverable
1976
# and not independently creatable, so are not registered. They're
1977
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1978
# needed, it's constructed directly by the BzrDir. Non-native formats where
1979
# the repository is not separately opened are similar.
1981
format_registry.register_lazy(
1982
'Bazaar-NG Repository format 7',
1983
'bzrlib.repofmt.weaverepo',
1987
# KEEP in sync with bzrdir.format_registry default, which controls the overall
1988
# default control directory format
1989
format_registry.register_lazy(
1990
'Bazaar-NG Knit Repository Format 1',
1991
'bzrlib.repofmt.knitrepo',
1992
'RepositoryFormatKnit1',
1994
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
1996
format_registry.register_lazy(
1997
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
1998
'bzrlib.repofmt.knitrepo',
1999
'RepositoryFormatKnit3',
2002
format_registry.register_lazy(
2003
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2004
'bzrlib.repofmt.knitrepo',
2005
'RepositoryFormatKnit4',
2008
# Pack-based formats. There is one format for pre-subtrees, and one for
2009
# post-subtrees to allow ease of testing.
2010
# NOTE: These are experimental in 0.92.
2011
format_registry.register_lazy(
2012
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2013
'bzrlib.repofmt.pack_repo',
2014
'RepositoryFormatKnitPack1',
2016
format_registry.register_lazy(
2017
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2018
'bzrlib.repofmt.pack_repo',
2019
'RepositoryFormatKnitPack3',
2023
class InterRepository(InterObject):
2024
"""This class represents operations taking place between two repositories.
2026
Its instances have methods like copy_content and fetch, and contain
2027
references to the source and target repositories these operations can be
2030
Often we will provide convenience methods on 'repository' which carry out
2031
operations with another repository - they will always forward to
2032
InterRepository.get(other).method_name(parameters).
2036
"""The available optimised InterRepository types."""
2038
def copy_content(self, revision_id=None):
2039
raise NotImplementedError(self.copy_content)
2041
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2042
"""Fetch the content required to construct revision_id.
2044
The content is copied from self.source to self.target.
2046
:param revision_id: if None all content is copied, if NULL_REVISION no
2048
:param pb: optional progress bar to use for progress reports. If not
2049
provided a default one will be created.
2051
Returns the copied revision count and the failed revisions in a tuple:
2054
raise NotImplementedError(self.fetch)
2057
def missing_revision_ids(self, revision_id=None):
2058
"""Return the revision ids that source has that target does not.
2060
These are returned in topological order.
2062
:param revision_id: only return revision ids included by this
2065
# generic, possibly worst case, slow code path.
2066
target_ids = set(self.target.all_revision_ids())
2067
if revision_id is not None:
2068
source_ids = self.source.get_ancestry(revision_id)
2069
assert source_ids[0] is None
2072
source_ids = self.source.all_revision_ids()
2073
result_set = set(source_ids).difference(target_ids)
2074
# this may look like a no-op: its not. It preserves the ordering
2075
# other_ids had while only returning the members from other_ids
2076
# that we've decided we need.
2077
return [rev_id for rev_id in source_ids if rev_id in result_set]
2080
def _same_model(source, target):
2081
"""True if source and target have the same data representation."""
2082
if source.supports_rich_root() != target.supports_rich_root():
2084
if source._serializer != target._serializer:
2089
class InterSameDataRepository(InterRepository):
2090
"""Code for converting between repositories that represent the same data.
2092
Data format and model must match for this to work.
2096
def _get_repo_format_to_test(self):
2097
"""Repository format for testing with.
2099
InterSameData can pull from subtree to subtree and from non-subtree to
2100
non-subtree, so we test this with the richest repository format.
2102
from bzrlib.repofmt import knitrepo
2103
return knitrepo.RepositoryFormatKnit3()
2106
def is_compatible(source, target):
2107
return InterRepository._same_model(source, target)
2110
def copy_content(self, revision_id=None):
2111
"""Make a complete copy of the content in self into destination.
2113
This copies both the repository's revision data, and configuration information
2114
such as the make_working_trees setting.
2116
This is a destructive operation! Do not use it on existing
2119
:param revision_id: Only copy the content needed to construct
2120
revision_id and its parents.
2123
self.target.set_make_working_trees(self.source.make_working_trees())
2124
except NotImplementedError:
2126
# but don't bother fetching if we have the needed data now.
2127
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2128
self.target.has_revision(revision_id)):
2130
self.target.fetch(self.source, revision_id=revision_id)
2133
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2134
"""See InterRepository.fetch()."""
2135
from bzrlib.fetch import GenericRepoFetcher
2136
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2137
self.source, self.source._format, self.target,
2138
self.target._format)
2139
f = GenericRepoFetcher(to_repository=self.target,
2140
from_repository=self.source,
2141
last_revision=revision_id,
2143
return f.count_copied, f.failed_revisions
2146
class InterWeaveRepo(InterSameDataRepository):
2147
"""Optimised code paths between Weave based repositories.
2149
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2150
implemented lazy inter-object optimisation.
2154
def _get_repo_format_to_test(self):
2155
from bzrlib.repofmt import weaverepo
2156
return weaverepo.RepositoryFormat7()
2159
def is_compatible(source, target):
2160
"""Be compatible with known Weave formats.
2162
We don't test for the stores being of specific types because that
2163
could lead to confusing results, and there is no need to be
2166
from bzrlib.repofmt.weaverepo import (
2172
return (isinstance(source._format, (RepositoryFormat5,
2174
RepositoryFormat7)) and
2175
isinstance(target._format, (RepositoryFormat5,
2177
RepositoryFormat7)))
2178
except AttributeError:
2182
def copy_content(self, revision_id=None):
2183
"""See InterRepository.copy_content()."""
2184
# weave specific optimised path:
2186
self.target.set_make_working_trees(self.source.make_working_trees())
2187
except NotImplementedError:
2189
# FIXME do not peek!
2190
if self.source.control_files._transport.listable():
2191
pb = ui.ui_factory.nested_progress_bar()
2193
self.target.weave_store.copy_all_ids(
2194
self.source.weave_store,
2196
from_transaction=self.source.get_transaction(),
2197
to_transaction=self.target.get_transaction())
2198
pb.update('copying inventory', 0, 1)
2199
self.target.control_weaves.copy_multi(
2200
self.source.control_weaves, ['inventory'],
2201
from_transaction=self.source.get_transaction(),
2202
to_transaction=self.target.get_transaction())
2203
self.target._revision_store.text_store.copy_all_ids(
2204
self.source._revision_store.text_store,
2209
self.target.fetch(self.source, revision_id=revision_id)
2212
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2213
"""See InterRepository.fetch()."""
2214
from bzrlib.fetch import GenericRepoFetcher
2215
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2216
self.source, self.source._format, self.target, self.target._format)
2217
f = GenericRepoFetcher(to_repository=self.target,
2218
from_repository=self.source,
2219
last_revision=revision_id,
2221
return f.count_copied, f.failed_revisions
2224
def missing_revision_ids(self, revision_id=None):
2225
"""See InterRepository.missing_revision_ids()."""
2226
# we want all revisions to satisfy revision_id in source.
2227
# but we don't want to stat every file here and there.
2228
# we want then, all revisions other needs to satisfy revision_id
2229
# checked, but not those that we have locally.
2230
# so the first thing is to get a subset of the revisions to
2231
# satisfy revision_id in source, and then eliminate those that
2232
# we do already have.
2233
# this is slow on high latency connection to self, but as as this
2234
# disk format scales terribly for push anyway due to rewriting
2235
# inventory.weave, this is considered acceptable.
2237
if revision_id is not None:
2238
source_ids = self.source.get_ancestry(revision_id)
2239
assert source_ids[0] is None
2242
source_ids = self.source._all_possible_ids()
2243
source_ids_set = set(source_ids)
2244
# source_ids is the worst possible case we may need to pull.
2245
# now we want to filter source_ids against what we actually
2246
# have in target, but don't try to check for existence where we know
2247
# we do not have a revision as that would be pointless.
2248
target_ids = set(self.target._all_possible_ids())
2249
possibly_present_revisions = target_ids.intersection(source_ids_set)
2250
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2251
required_revisions = source_ids_set.difference(actually_present_revisions)
2252
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2253
if revision_id is not None:
2254
# we used get_ancestry to determine source_ids then we are assured all
2255
# revisions referenced are present as they are installed in topological order.
2256
# and the tip revision was validated by get_ancestry.
2257
return required_topo_revisions
2259
# if we just grabbed the possibly available ids, then
2260
# we only have an estimate of whats available and need to validate
2261
# that against the revision records.
2262
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2265
class InterKnitRepo(InterSameDataRepository):
2266
"""Optimised code paths between Knit based repositories."""
2269
def _get_repo_format_to_test(self):
2270
from bzrlib.repofmt import knitrepo
2271
return knitrepo.RepositoryFormatKnit1()
2274
def is_compatible(source, target):
2275
"""Be compatible with known Knit formats.
2277
We don't test for the stores being of specific types because that
2278
could lead to confusing results, and there is no need to be
2281
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2283
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2284
isinstance(target._format, RepositoryFormatKnit))
2285
except AttributeError:
2287
return are_knits and InterRepository._same_model(source, target)
2290
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2291
"""See InterRepository.fetch()."""
2292
from bzrlib.fetch import KnitRepoFetcher
2293
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2294
self.source, self.source._format, self.target, self.target._format)
2295
f = KnitRepoFetcher(to_repository=self.target,
2296
from_repository=self.source,
2297
last_revision=revision_id,
2299
return f.count_copied, f.failed_revisions
2302
def missing_revision_ids(self, revision_id=None):
2303
"""See InterRepository.missing_revision_ids()."""
2304
if revision_id is not None:
2305
source_ids = self.source.get_ancestry(revision_id)
2306
assert source_ids[0] is None
2309
source_ids = self.source.all_revision_ids()
2310
source_ids_set = set(source_ids)
2311
# source_ids is the worst possible case we may need to pull.
2312
# now we want to filter source_ids against what we actually
2313
# have in target, but don't try to check for existence where we know
2314
# we do not have a revision as that would be pointless.
2315
target_ids = set(self.target.all_revision_ids())
2316
possibly_present_revisions = target_ids.intersection(source_ids_set)
2317
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2318
required_revisions = source_ids_set.difference(actually_present_revisions)
2319
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2320
if revision_id is not None:
2321
# we used get_ancestry to determine source_ids then we are assured all
2322
# revisions referenced are present as they are installed in topological order.
2323
# and the tip revision was validated by get_ancestry.
2324
return required_topo_revisions
2326
# if we just grabbed the possibly available ids, then
2327
# we only have an estimate of whats available and need to validate
2328
# that against the revision records.
2329
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2332
class InterPackRepo(InterSameDataRepository):
2333
"""Optimised code paths between Pack based repositories."""
2336
def _get_repo_format_to_test(self):
2337
from bzrlib.repofmt import pack_repo
2338
return pack_repo.RepositoryFormatKnitPack1()
2341
def is_compatible(source, target):
2342
"""Be compatible with known Pack formats.
2344
We don't test for the stores being of specific types because that
2345
could lead to confusing results, and there is no need to be
2348
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2350
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2351
isinstance(target._format, RepositoryFormatPack))
2352
except AttributeError:
2354
return are_packs and InterRepository._same_model(source, target)
2357
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2358
"""See InterRepository.fetch()."""
2359
from bzrlib.repofmt.pack_repo import Packer
2360
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2361
self.source, self.source._format, self.target, self.target._format)
2362
self.count_copied = 0
2363
if revision_id is None:
2365
# everything to do - use pack logic
2366
# to fetch from all packs to one without
2367
# inventory parsing etc, IFF nothing to be copied is in the target.
2369
revision_ids = self.source.all_revision_ids()
2370
# implementing the TODO will involve:
2371
# - detecting when all of a pack is selected
2372
# - avoiding as much as possible pre-selection, so the
2373
# more-core routines such as create_pack_from_packs can filter in
2374
# a just-in-time fashion. (though having a HEADS list on a
2375
# repository might make this a lot easier, because we could
2376
# sensibly detect 'new revisions' without doing a full index scan.
2377
elif _mod_revision.is_null(revision_id):
2382
revision_ids = self.missing_revision_ids(revision_id,
2383
find_ghosts=find_ghosts)
2384
except errors.NoSuchRevision:
2385
raise errors.InstallFailed([revision_id])
2386
packs = self.source._pack_collection.all_packs()
2387
pack = Packer(self.target._pack_collection, packs, '.fetch',
2388
revision_ids).pack()
2389
if pack is not None:
2390
self.target._pack_collection._save_pack_names()
2391
# Trigger an autopack. This may duplicate effort as we've just done
2392
# a pack creation, but for now it is simpler to think about as
2393
# 'upload data, then repack if needed'.
2394
self.target._pack_collection.autopack()
2395
return pack.get_revision_count()
2400
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2401
"""See InterRepository.missing_revision_ids().
2403
:param find_ghosts: Find ghosts throughough the ancestry of
2406
if not find_ghosts and revision_id is not None:
2407
graph = self.source.get_graph()
2408
missing_revs = set()
2409
searcher = graph._make_breadth_first_searcher([revision_id])
2411
self.target._pack_collection.revision_index.combined_index
2412
null_set = frozenset([_mod_revision.NULL_REVISION])
2415
next_revs = set(searcher.next())
2416
except StopIteration:
2418
next_revs.difference_update(null_set)
2419
target_keys = [(key,) for key in next_revs]
2420
have_revs = frozenset(node[1][0] for node in
2421
target_index.iter_entries(target_keys))
2422
missing_revs.update(next_revs - have_revs)
2423
searcher.stop_searching_any(have_revs)
2425
elif revision_id is not None:
2426
source_ids = self.source.get_ancestry(revision_id)
2427
assert source_ids[0] is None
2430
source_ids = self.source.all_revision_ids()
2431
# source_ids is the worst possible case we may need to pull.
2432
# now we want to filter source_ids against what we actually
2433
# have in target, but don't try to check for existence where we know
2434
# we do not have a revision as that would be pointless.
2435
target_ids = set(self.target.all_revision_ids())
2436
return [r for r in source_ids if (r not in target_ids)]
2439
class InterModel1and2(InterRepository):
2442
def _get_repo_format_to_test(self):
2446
def is_compatible(source, target):
2447
if not source.supports_rich_root() and target.supports_rich_root():
2453
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2454
"""See InterRepository.fetch()."""
2455
from bzrlib.fetch import Model1toKnit2Fetcher
2456
f = Model1toKnit2Fetcher(to_repository=self.target,
2457
from_repository=self.source,
2458
last_revision=revision_id,
2460
return f.count_copied, f.failed_revisions
2463
def copy_content(self, revision_id=None):
2464
"""Make a complete copy of the content in self into destination.
2466
This is a destructive operation! Do not use it on existing
2469
:param revision_id: Only copy the content needed to construct
2470
revision_id and its parents.
2473
self.target.set_make_working_trees(self.source.make_working_trees())
2474
except NotImplementedError:
2476
# but don't bother fetching if we have the needed data now.
2477
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2478
self.target.has_revision(revision_id)):
2480
self.target.fetch(self.source, revision_id=revision_id)
2483
class InterKnit1and2(InterKnitRepo):
2486
def _get_repo_format_to_test(self):
2490
def is_compatible(source, target):
2491
"""Be compatible with Knit1 source and Knit3 target"""
2492
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2494
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2495
RepositoryFormatKnit3)
2496
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2497
RepositoryFormatKnitPack3)
2498
return (isinstance(source._format,
2499
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2500
isinstance(target._format,
2501
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2503
except AttributeError:
2507
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2508
"""See InterRepository.fetch()."""
2509
from bzrlib.fetch import Knit1to2Fetcher
2510
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2511
self.source, self.source._format, self.target,
2512
self.target._format)
2513
f = Knit1to2Fetcher(to_repository=self.target,
2514
from_repository=self.source,
2515
last_revision=revision_id,
2517
return f.count_copied, f.failed_revisions
2520
class InterDifferingSerializer(InterKnitRepo):
2523
def _get_repo_format_to_test(self):
2527
def is_compatible(source, target):
2528
"""Be compatible with Knit2 source and Knit3 target"""
2529
if source.supports_rich_root() != target.supports_rich_root():
2531
# Ideally, we'd support fetching if the source had no tree references
2532
# even if it supported them...
2533
if (getattr(source, '_format.supports_tree_reference', False) and
2534
not getattr(target, '_format.supports_tree_reference', False)):
2539
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2540
"""See InterRepository.fetch()."""
2541
revision_ids = self.target.missing_revision_ids(self.source,
2543
def revisions_iterator():
2544
for current_revision_id in revision_ids:
2545
revision = self.source.get_revision(current_revision_id)
2546
tree = self.source.revision_tree(current_revision_id)
2548
signature = self.source.get_signature_text(
2549
current_revision_id)
2550
except errors.NoSuchRevision:
2552
yield revision, tree, signature
2553
install_revisions(self.target, revisions_iterator())
2554
return len(revision_ids), 0
2557
class InterRemoteToOther(InterRepository):
2559
def __init__(self, source, target):
2560
InterRepository.__init__(self, source, target)
2561
self._real_inter = None
2564
def is_compatible(source, target):
2565
if not isinstance(source, remote.RemoteRepository):
2567
source._ensure_real()
2568
real_source = source._real_repository
2569
# Is source's model compatible with target's model, and are they the
2570
# same format? Currently we can only optimise fetching from an
2571
# identical model & format repo.
2572
assert not isinstance(real_source, remote.RemoteRepository), (
2573
"We don't support remote repos backed by remote repos yet.")
2574
return real_source._format == target._format
2577
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2578
"""See InterRepository.fetch()."""
2579
from bzrlib.fetch import RemoteToOtherFetcher
2580
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2581
self.source, self.target, self.target._format)
2582
# TODO: jam 20070210 This should be an assert, not a translate
2583
revision_id = osutils.safe_revision_id(revision_id)
2584
f = RemoteToOtherFetcher(to_repository=self.target,
2585
from_repository=self.source,
2586
last_revision=revision_id,
2588
return f.count_copied, f.failed_revisions
2591
def _get_repo_format_to_test(self):
2595
class InterOtherToRemote(InterRepository):
2597
def __init__(self, source, target):
2598
InterRepository.__init__(self, source, target)
2599
self._real_inter = None
2602
def is_compatible(source, target):
2603
if isinstance(target, remote.RemoteRepository):
2607
def _ensure_real_inter(self):
2608
if self._real_inter is None:
2609
self.target._ensure_real()
2610
real_target = self.target._real_repository
2611
self._real_inter = InterRepository.get(self.source, real_target)
2613
def copy_content(self, revision_id=None):
2614
self._ensure_real_inter()
2615
self._real_inter.copy_content(revision_id=revision_id)
2617
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2618
self._ensure_real_inter()
2619
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2622
def _get_repo_format_to_test(self):
2626
InterRepository.register_optimiser(InterDifferingSerializer)
2627
InterRepository.register_optimiser(InterSameDataRepository)
2628
InterRepository.register_optimiser(InterWeaveRepo)
2629
InterRepository.register_optimiser(InterKnitRepo)
2630
InterRepository.register_optimiser(InterModel1and2)
2631
InterRepository.register_optimiser(InterKnit1and2)
2632
InterRepository.register_optimiser(InterPackRepo)
2633
InterRepository.register_optimiser(InterRemoteToOther)
2634
InterRepository.register_optimiser(InterOtherToRemote)
2637
class CopyConverter(object):
2638
"""A repository conversion tool which just performs a copy of the content.
2640
This is slow but quite reliable.
2643
def __init__(self, target_format):
2644
"""Create a CopyConverter.
2646
:param target_format: The format the resulting repository should be.
2648
self.target_format = target_format
2650
def convert(self, repo, pb):
2651
"""Perform the conversion of to_convert, giving feedback via pb.
2653
:param to_convert: The disk object to convert.
2654
:param pb: a progress bar to use for progress information.
2659
# this is only useful with metadir layouts - separated repo content.
2660
# trigger an assertion if not such
2661
repo._format.get_format_string()
2662
self.repo_dir = repo.bzrdir
2663
self.step('Moving repository to repository.backup')
2664
self.repo_dir.transport.move('repository', 'repository.backup')
2665
backup_transport = self.repo_dir.transport.clone('repository.backup')
2666
repo._format.check_conversion_target(self.target_format)
2667
self.source_repo = repo._format.open(self.repo_dir,
2669
_override_transport=backup_transport)
2670
self.step('Creating new repository')
2671
converted = self.target_format.initialize(self.repo_dir,
2672
self.source_repo.is_shared())
2673
converted.lock_write()
2675
self.step('Copying content into repository.')
2676
self.source_repo.copy_content_into(converted)
2679
self.step('Deleting old repository content.')
2680
self.repo_dir.transport.delete_tree('repository.backup')
2681
self.pb.note('repository converted')
2683
def step(self, message):
2684
"""Update the pb by a step."""
2686
self.pb.update(message, self.count, self.total)
2698
def _unescaper(match, _map=_unescape_map):
2699
code = match.group(1)
2703
if not code.startswith('#'):
2705
return unichr(int(code[1:])).encode('utf8')
2711
def _unescape_xml(data):
2712
"""Unescape predefined XML entities in a string of data."""
2714
if _unescape_re is None:
2715
_unescape_re = re.compile('\&([^;]*);')
2716
return _unescape_re.sub(_unescaper, data)
2719
class _RevisionTextVersionCache(object):
2720
"""A cache of the versionedfile versions for revision and file-id."""
2722
def __init__(self, repository):
2723
self.repository = repository
2724
self.revision_versions = {}
2725
self.revision_parents = {}
2726
self.repo_graph = self.repository.get_graph()
2727
# XXX: RBC: I haven't tracked down what uses this, but it would be
2728
# better to use the headscache directly I think.
2729
self.heads = graph.HeadsCache(self.repo_graph).heads
2731
def add_revision_text_versions(self, tree):
2732
"""Cache text version data from the supplied revision tree"""
2734
for path, entry in tree.iter_entries_by_dir():
2735
inv_revisions[entry.file_id] = entry.revision
2736
self.revision_versions[tree.get_revision_id()] = inv_revisions
2737
return inv_revisions
2739
def get_text_version(self, file_id, revision_id):
2740
"""Determine the text version for a given file-id and revision-id"""
2742
inv_revisions = self.revision_versions[revision_id]
2745
tree = self.repository.revision_tree(revision_id)
2746
except errors.RevisionNotPresent:
2747
self.revision_versions[revision_id] = inv_revisions = {}
2749
inv_revisions = self.add_revision_text_versions(tree)
2750
return inv_revisions.get(file_id)
2752
def prepopulate_revs(self, revision_ids):
2753
# Filter out versions that we don't have an inventory for, so that the
2754
# revision_trees() call won't fail.
2755
inv_weave = self.repository.get_inventory_weave()
2756
revs = [r for r in revision_ids if inv_weave.has_version(r)]
2757
# XXX: this loop is very similar to
2758
# bzrlib.fetch.Inter1and2Helper.iter_rev_trees.
2760
mutter('%d revisions left to prepopulate', len(revs))
2761
for tree in self.repository.revision_trees(revs[:100]):
2762
if tree.inventory.revision_id is None:
2763
tree.inventory.revision_id = tree.get_revision_id()
2764
self.add_revision_text_versions(tree)
2767
def get_parents(self, revision_id):
2769
return self.revision_parents[revision_id]
2771
parents = self.repository.get_parents([revision_id])[0]
2772
self.revision_parents[revision_id] = parents
2775
def used_file_versions(self):
2776
"""Return a set of (revision_id, file_id) pairs for each file version
2777
referenced by any inventory cached by this _RevisionTextVersionCache.
2779
If the entire repository has been cached, this can be used to find all
2780
file versions that are actually referenced by inventories. Thus any
2781
other file version is completely unused and can be removed safely.
2784
for inventory_summary in self.revision_versions.itervalues():
2785
result.update(inventory_summary.items())
2789
class VersionedFileChecker(object):
2791
def __init__(self, planned_revisions, revision_versions, repository):
2792
self.planned_revisions = planned_revisions
2793
self.revision_versions = revision_versions
2794
self.repository = repository
2796
def calculate_file_version_parents(self, revision_id, file_id):
2797
"""Calculate the correct parents for a file version according to
2800
text_revision = self.revision_versions.get_text_version(
2801
file_id, revision_id)
2802
if text_revision is None:
2804
parents_of_text_revision = self.revision_versions.get_parents(
2806
parents_from_inventories = []
2807
for parent in parents_of_text_revision:
2808
if parent == _mod_revision.NULL_REVISION:
2810
introduced_in = self.revision_versions.get_text_version(file_id,
2812
if introduced_in is not None:
2813
parents_from_inventories.append(introduced_in)
2814
heads = set(self.revision_versions.heads(parents_from_inventories))
2816
for parent in parents_from_inventories:
2817
if parent in heads and parent not in new_parents:
2818
new_parents.append(parent)
2819
return tuple(new_parents)
2821
def check_file_version_parents(self, weave, file_id):
2822
"""Check the parents stored in a versioned file are correct.
2824
It also detects file versions that are not referenced by their
2825
corresponding revision's inventory.
2827
:returns: A tuple of (wrong_parents, dangling_file_versions).
2828
wrong_parents is a dict mapping {revision_id: (stored_parents,
2829
correct_parents)} for each revision_id where the stored parents
2830
are not correct. dangling_file_versions is a set of (file_id,
2831
revision_id) tuples for versions that are present in this versioned
2832
file, but not used by the corresponding inventory.
2835
dangling_file_versions = set()
2836
for num, revision_id in enumerate(self.planned_revisions):
2837
correct_parents = self.calculate_file_version_parents(
2838
revision_id, file_id)
2839
if correct_parents is None:
2841
text_revision = self.revision_versions.get_text_version(
2842
file_id, revision_id)
2844
knit_parents = tuple(weave.get_parents(revision_id))
2845
except errors.RevisionNotPresent:
2847
if text_revision != revision_id:
2848
# This file version is not referenced by its corresponding
2850
dangling_file_versions.add((file_id, revision_id))
2851
if correct_parents != knit_parents:
2852
wrong_parents[revision_id] = (knit_parents, correct_parents)
2853
return wrong_parents, dangling_file_versions
added
removed
bzrlib/repository.py
