1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from cStringIO import StringIO
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
39
revision as _mod_revision,
44
from bzrlib.bundle import serializer
45
from bzrlib.revisiontree import RevisionTree
46
from bzrlib.store.versioned import VersionedFileStore
47
from bzrlib.store.text import TextStore
48
from bzrlib.testament import Testament
49
from bzrlib.util import bencode
52
from bzrlib.decorators import needs_read_lock, needs_write_lock
53
from bzrlib.inter import InterObject
54
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
55
from bzrlib.symbol_versioning import (
58
from bzrlib.trace import mutter, mutter_callsite, note, warning
61
# Old formats display a warning, but only once
62
_deprecation_warning_done = False
65
class CommitBuilder(object):
66
"""Provides an interface to build up a commit.
68
This allows describing a tree to be committed without needing to
69
know the internals of the format of the repository.
72
# all clients should supply tree roots.
73
record_root_entry = True
74
# the default CommitBuilder does not manage trees whose root is versioned.
75
_versioned_root = False
77
def __init__(self, repository, parents, config, timestamp=None,
78
timezone=None, committer=None, revprops=None,
80
"""Initiate a CommitBuilder.
82
:param repository: Repository to commit to.
83
:param parents: Revision ids of the parents of the new revision.
84
:param config: Configuration to use.
85
:param timestamp: Optional timestamp recorded for commit.
86
:param timezone: Optional timezone for timestamp.
87
:param committer: Optional committer to set for commit.
88
:param revprops: Optional dictionary of revision properties.
89
:param revision_id: Optional revision id.
94
self._committer = self._config.username()
96
assert isinstance(committer, basestring), type(committer)
97
self._committer = committer
99
self.new_inventory = Inventory(None)
100
self._new_revision_id = revision_id
101
self.parents = parents
102
self.repository = repository
105
if revprops is not None:
106
self._revprops.update(revprops)
108
if timestamp is None:
109
timestamp = time.time()
110
# Restrict resolution to 1ms
111
self._timestamp = round(timestamp, 3)
114
self._timezone = osutils.local_time_offset()
116
self._timezone = int(timezone)
118
self._generate_revision_if_needed()
119
self._heads = graph.HeadsCache(repository.get_graph()).heads
121
def commit(self, message):
122
"""Make the actual commit.
124
:return: The revision id of the recorded revision.
126
rev = _mod_revision.Revision(
127
timestamp=self._timestamp,
128
timezone=self._timezone,
129
committer=self._committer,
131
inventory_sha1=self.inv_sha1,
132
revision_id=self._new_revision_id,
133
properties=self._revprops)
134
rev.parent_ids = self.parents
135
self.repository.add_revision(self._new_revision_id, rev,
136
self.new_inventory, self._config)
137
self.repository.commit_write_group()
138
return self._new_revision_id
141
"""Abort the commit that is being built.
143
self.repository.abort_write_group()
145
def revision_tree(self):
146
"""Return the tree that was just committed.
148
After calling commit() this can be called to get a RevisionTree
149
representing the newly committed tree. This is preferred to
150
calling Repository.revision_tree() because that may require
151
deserializing the inventory, while we already have a copy in
154
return RevisionTree(self.repository, self.new_inventory,
155
self._new_revision_id)
157
def finish_inventory(self):
158
"""Tell the builder that the inventory is finished."""
159
if self.new_inventory.root is None:
160
raise AssertionError('Root entry should be supplied to'
161
' record_entry_contents, as of bzr 0.10.',
162
DeprecationWarning, stacklevel=2)
163
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
164
self.new_inventory.revision_id = self._new_revision_id
165
self.inv_sha1 = self.repository.add_inventory(
166
self._new_revision_id,
171
def _gen_revision_id(self):
172
"""Return new revision-id."""
173
return generate_ids.gen_revision_id(self._config.username(),
176
def _generate_revision_if_needed(self):
177
"""Create a revision id if None was supplied.
179
If the repository can not support user-specified revision ids
180
they should override this function and raise CannotSetRevisionId
181
if _new_revision_id is not None.
183
:raises: CannotSetRevisionId
185
if self._new_revision_id is None:
186
self._new_revision_id = self._gen_revision_id()
187
self.random_revid = True
189
self.random_revid = False
191
def _check_root(self, ie, parent_invs, tree):
192
"""Helper for record_entry_contents.
194
:param ie: An entry being added.
195
:param parent_invs: The inventories of the parent revisions of the
197
:param tree: The tree that is being committed.
199
# In this revision format, root entries have no knit or weave When
200
# serializing out to disk and back in root.revision is always
202
ie.revision = self._new_revision_id
204
def _get_delta(self, ie, basis_inv, path):
205
"""Get a delta against the basis inventory for ie."""
206
if ie.file_id not in basis_inv:
208
return (None, path, ie.file_id, ie)
209
elif ie != basis_inv[ie.file_id]:
211
# TODO: avoid tis id2path call.
212
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
217
def record_entry_contents(self, ie, parent_invs, path, tree,
219
"""Record the content of ie from tree into the commit if needed.
221
Side effect: sets ie.revision when unchanged
223
:param ie: An inventory entry present in the commit.
224
:param parent_invs: The inventories of the parent revisions of the
226
:param path: The path the entry is at in the tree.
227
:param tree: The tree which contains this entry and should be used to
229
:param content_summary: Summary data from the tree about the paths
230
content - stat, length, exec, sha/link target. This is only
231
accessed when the entry has a revision of None - that is when it is
232
a candidate to commit.
233
:return: A tuple (change_delta, version_recorded). change_delta is
234
an inventory_delta change for this entry against the basis tree of
235
the commit, or None if no change occured against the basis tree.
236
version_recorded is True if a new version of the entry has been
237
recorded. For instance, committing a merge where a file was only
238
changed on the other side will return (delta, False).
240
if self.new_inventory.root is None:
241
if ie.parent_id is not None:
242
raise errors.RootMissing()
243
self._check_root(ie, parent_invs, tree)
244
if ie.revision is None:
245
kind = content_summary[0]
247
# ie is carried over from a prior commit
249
# XXX: repository specific check for nested tree support goes here - if
250
# the repo doesn't want nested trees we skip it ?
251
if (kind == 'tree-reference' and
252
not self.repository._format.supports_tree_reference):
253
# mismatch between commit builder logic and repository:
254
# this needs the entry creation pushed down into the builder.
255
raise NotImplementedError('Missing repository subtree support.')
256
self.new_inventory.add(ie)
258
# TODO: slow, take it out of the inner loop.
260
basis_inv = parent_invs[0]
262
basis_inv = Inventory(root_id=None)
264
# ie.revision is always None if the InventoryEntry is considered
265
# for committing. We may record the previous parents revision if the
266
# content is actually unchanged against a sole head.
267
if ie.revision is not None:
268
if not self._versioned_root and path == '':
269
# repositories that do not version the root set the root's
270
# revision to the new commit even when no change occurs, and
271
# this masks when a change may have occurred against the basis,
272
# so calculate if one happened.
273
if ie.file_id in basis_inv:
274
delta = (basis_inv.id2path(ie.file_id), path,
278
delta = (None, path, ie.file_id, ie)
281
# we don't need to commit this, because the caller already
282
# determined that an existing revision of this file is
284
return None, (ie.revision == self._new_revision_id)
285
# XXX: Friction: parent_candidates should return a list not a dict
286
# so that we don't have to walk the inventories again.
287
parent_candiate_entries = ie.parent_candidates(parent_invs)
288
head_set = self._heads(parent_candiate_entries.keys())
290
for inv in parent_invs:
291
if ie.file_id in inv:
292
old_rev = inv[ie.file_id].revision
293
if old_rev in head_set:
294
heads.append(inv[ie.file_id].revision)
295
head_set.remove(inv[ie.file_id].revision)
298
# now we check to see if we need to write a new record to the
300
# We write a new entry unless there is one head to the ancestors, and
301
# the kind-derived content is unchanged.
303
# Cheapest check first: no ancestors, or more the one head in the
304
# ancestors, we write a new node.
308
# There is a single head, look it up for comparison
309
parent_entry = parent_candiate_entries[heads[0]]
310
# if the non-content specific data has changed, we'll be writing a
312
if (parent_entry.parent_id != ie.parent_id or
313
parent_entry.name != ie.name):
315
# now we need to do content specific checks:
317
# if the kind changed the content obviously has
318
if kind != parent_entry.kind:
321
assert content_summary[2] is not None, \
322
"Files must not have executable = None"
324
if (# if the file length changed we have to store:
325
parent_entry.text_size != content_summary[1] or
326
# if the exec bit has changed we have to store:
327
parent_entry.executable != content_summary[2]):
329
elif parent_entry.text_sha1 == content_summary[3]:
330
# all meta and content is unchanged (using a hash cache
331
# hit to check the sha)
332
ie.revision = parent_entry.revision
333
ie.text_size = parent_entry.text_size
334
ie.text_sha1 = parent_entry.text_sha1
335
ie.executable = parent_entry.executable
336
return self._get_delta(ie, basis_inv, path), False
338
# Either there is only a hash change(no hash cache entry,
339
# or same size content change), or there is no change on
341
# Provide the parent's hash to the store layer, so that the
342
# content is unchanged we will not store a new node.
343
nostore_sha = parent_entry.text_sha1
345
# We want to record a new node regardless of the presence or
346
# absence of a content change in the file.
348
ie.executable = content_summary[2]
349
lines = tree.get_file(ie.file_id, path).readlines()
351
ie.text_sha1, ie.text_size = self._add_text_to_weave(
352
ie.file_id, lines, heads, nostore_sha)
353
except errors.ExistingContent:
354
# Turns out that the file content was unchanged, and we were
355
# only going to store a new node if it was changed. Carry over
357
ie.revision = parent_entry.revision
358
ie.text_size = parent_entry.text_size
359
ie.text_sha1 = parent_entry.text_sha1
360
ie.executable = parent_entry.executable
361
return self._get_delta(ie, basis_inv, path), False
362
elif kind == 'directory':
364
# all data is meta here, nothing specific to directory, so
366
ie.revision = parent_entry.revision
367
return self._get_delta(ie, basis_inv, path), False
369
self._add_text_to_weave(ie.file_id, lines, heads, None)
370
elif kind == 'symlink':
371
current_link_target = content_summary[3]
373
# symlink target is not generic metadata, check if it has
375
if current_link_target != parent_entry.symlink_target:
378
# unchanged, carry over.
379
ie.revision = parent_entry.revision
380
ie.symlink_target = parent_entry.symlink_target
381
return self._get_delta(ie, basis_inv, path), False
382
ie.symlink_target = current_link_target
384
self._add_text_to_weave(ie.file_id, lines, heads, None)
385
elif kind == 'tree-reference':
387
if content_summary[3] != parent_entry.reference_revision:
390
# unchanged, carry over.
391
ie.reference_revision = parent_entry.reference_revision
392
ie.revision = parent_entry.revision
393
return self._get_delta(ie, basis_inv, path), False
394
ie.reference_revision = content_summary[3]
396
self._add_text_to_weave(ie.file_id, lines, heads, None)
398
raise NotImplementedError('unknown kind')
399
ie.revision = self._new_revision_id
400
return self._get_delta(ie, basis_inv, path), True
402
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
403
versionedfile = self.repository.weave_store.get_weave_or_empty(
404
file_id, self.repository.get_transaction())
405
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
406
# than add_lines, and allows committing when a parent is ghosted for
408
# Note: as we read the content directly from the tree, we know its not
409
# been turned into unicode or badly split - but a broken tree
410
# implementation could give us bad output from readlines() so this is
411
# not a guarantee of safety. What would be better is always checking
412
# the content during test suite execution. RBC 20070912
414
return versionedfile.add_lines_with_ghosts(
415
self._new_revision_id, parents, new_lines,
416
nostore_sha=nostore_sha, random_id=self.random_revid,
417
check_content=False)[0:2]
419
versionedfile.clear_cache()
422
class RootCommitBuilder(CommitBuilder):
423
"""This commitbuilder actually records the root id"""
425
# the root entry gets versioned properly by this builder.
426
_versioned_root = True
428
def _check_root(self, ie, parent_invs, tree):
429
"""Helper for record_entry_contents.
431
:param ie: An entry being added.
432
:param parent_invs: The inventories of the parent revisions of the
434
:param tree: The tree that is being committed.
438
######################################################################
441
class Repository(object):
442
"""Repository holding history for one or more branches.
444
The repository holds and retrieves historical information including
445
revisions and file history. It's normally accessed only by the Branch,
446
which views a particular line of development through that history.
448
The Repository builds on top of Stores and a Transport, which respectively
449
describe the disk data format and the way of accessing the (possibly
453
# What class to use for a CommitBuilder. Often its simpler to change this
454
# in a Repository class subclass rather than to override
455
# get_commit_builder.
456
_commit_builder_class = CommitBuilder
457
# The search regex used by xml based repositories to determine what things
458
# where changed in a single commit.
459
_file_ids_altered_regex = lazy_regex.lazy_compile(
460
r'file_id="(?P<file_id>[^"]+)"'
461
r'.* revision="(?P<revision_id>[^"]+)"'
464
def abort_write_group(self):
465
"""Commit the contents accrued within the current write group.
467
:seealso: start_write_group.
469
if self._write_group is not self.get_transaction():
470
# has an unlock or relock occured ?
471
raise errors.BzrError('mismatched lock context and write group.')
472
self._abort_write_group()
473
self._write_group = None
475
def _abort_write_group(self):
476
"""Template method for per-repository write group cleanup.
478
This is called during abort before the write group is considered to be
479
finished and should cleanup any internal state accrued during the write
480
group. There is no requirement that data handed to the repository be
481
*not* made available - this is not a rollback - but neither should any
482
attempt be made to ensure that data added is fully commited. Abort is
483
invoked when an error has occured so futher disk or network operations
484
may not be possible or may error and if possible should not be
489
def add_inventory(self, revision_id, inv, parents):
490
"""Add the inventory inv to the repository as revision_id.
492
:param parents: The revision ids of the parents that revision_id
493
is known to have and are in the repository already.
495
returns the sha1 of the serialized inventory.
497
assert self.is_in_write_group()
498
_mod_revision.check_not_reserved_id(revision_id)
499
assert inv.revision_id is None or inv.revision_id == revision_id, \
500
"Mismatch between inventory revision" \
501
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
502
assert inv.root is not None
503
inv_lines = self._serialise_inventory_to_lines(inv)
504
inv_vf = self.get_inventory_weave()
505
return self._inventory_add_lines(inv_vf, revision_id, parents,
506
inv_lines, check_content=False)
508
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
510
"""Store lines in inv_vf and return the sha1 of the inventory."""
512
for parent in parents:
514
final_parents.append(parent)
515
return inv_vf.add_lines(revision_id, final_parents, lines,
516
check_content=check_content)[0]
519
def add_revision(self, revision_id, rev, inv=None, config=None):
520
"""Add rev to the revision store as revision_id.
522
:param revision_id: the revision id to use.
523
:param rev: The revision object.
524
:param inv: The inventory for the revision. if None, it will be looked
525
up in the inventory storer
526
:param config: If None no digital signature will be created.
527
If supplied its signature_needed method will be used
528
to determine if a signature should be made.
530
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
532
_mod_revision.check_not_reserved_id(revision_id)
533
if config is not None and config.signature_needed():
535
inv = self.get_inventory(revision_id)
536
plaintext = Testament(rev, inv).as_short_text()
537
self.store_revision_signature(
538
gpg.GPGStrategy(config), plaintext, revision_id)
539
if not revision_id in self.get_inventory_weave():
541
raise errors.WeaveRevisionNotPresent(revision_id,
542
self.get_inventory_weave())
544
# yes, this is not suitable for adding with ghosts.
545
self.add_inventory(revision_id, inv, rev.parent_ids)
546
self._revision_store.add_revision(rev, self.get_transaction())
548
def _add_revision_text(self, revision_id, text):
549
revision = self._revision_store._serializer.read_revision_from_string(
551
self._revision_store._add_revision(revision, StringIO(text),
552
self.get_transaction())
554
def all_revision_ids(self):
555
"""Returns a list of all the revision ids in the repository.
557
This is deprecated because code should generally work on the graph
558
reachable from a particular revision, and ignore any other revisions
559
that might be present. There is no direct replacement method.
561
if 'evil' in debug.debug_flags:
562
mutter_callsite(2, "all_revision_ids is linear with history.")
563
return self._all_revision_ids()
565
def _all_revision_ids(self):
566
"""Returns a list of all the revision ids in the repository.
568
These are in as much topological order as the underlying store can
571
raise NotImplementedError(self._all_revision_ids)
573
def break_lock(self):
574
"""Break a lock if one is present from another instance.
576
Uses the ui factory to ask for confirmation if the lock may be from
579
self.control_files.break_lock()
582
def _eliminate_revisions_not_present(self, revision_ids):
583
"""Check every revision id in revision_ids to see if we have it.
585
Returns a set of the present revisions.
588
for id in revision_ids:
589
if self.has_revision(id):
594
def create(a_bzrdir):
595
"""Construct the current default format repository in a_bzrdir."""
596
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
598
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
599
"""instantiate a Repository.
601
:param _format: The format of the repository on disk.
602
:param a_bzrdir: The BzrDir of the repository.
604
In the future we will have a single api for all stores for
605
getting file texts, inventories and revisions, then
606
this construct will accept instances of those things.
608
super(Repository, self).__init__()
609
self._format = _format
610
# the following are part of the public API for Repository:
611
self.bzrdir = a_bzrdir
612
self.control_files = control_files
613
self._revision_store = _revision_store
614
# backwards compatibility
615
self.weave_store = text_store
617
self._reconcile_does_inventory_gc = True
618
self._reconcile_fixes_text_parents = False
619
# not right yet - should be more semantically clear ?
621
self.control_store = control_store
622
self.control_weaves = control_store
623
# TODO: make sure to construct the right store classes, etc, depending
624
# on whether escaping is required.
625
self._warn_if_deprecated()
626
self._write_group = None
627
self.base = control_files._transport.base
630
return '%s(%r)' % (self.__class__.__name__,
633
def has_same_location(self, other):
634
"""Returns a boolean indicating if this repository is at the same
635
location as another repository.
637
This might return False even when two repository objects are accessing
638
the same physical repository via different URLs.
640
if self.__class__ is not other.__class__:
642
return (self.control_files._transport.base ==
643
other.control_files._transport.base)
645
def is_in_write_group(self):
646
"""Return True if there is an open write group.
648
:seealso: start_write_group.
650
return self._write_group is not None
653
return self.control_files.is_locked()
655
def is_write_locked(self):
656
"""Return True if this object is write locked."""
657
return self.is_locked() and self.control_files._lock_mode == 'w'
659
def lock_write(self, token=None):
660
"""Lock this repository for writing.
662
This causes caching within the repository obejct to start accumlating
663
data during reads, and allows a 'write_group' to be obtained. Write
664
groups must be used for actual data insertion.
666
:param token: if this is already locked, then lock_write will fail
667
unless the token matches the existing lock.
668
:returns: a token if this instance supports tokens, otherwise None.
669
:raises TokenLockingNotSupported: when a token is given but this
670
instance doesn't support using token locks.
671
:raises MismatchedToken: if the specified token doesn't match the token
672
of the existing lock.
673
:seealso: start_write_group.
675
A token should be passed in if you know that you have locked the object
676
some other way, and need to synchronise this object's state with that
679
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
681
result = self.control_files.lock_write(token=token)
686
self.control_files.lock_read()
689
def get_physical_lock_status(self):
690
return self.control_files.get_physical_lock_status()
692
def leave_lock_in_place(self):
693
"""Tell this repository not to release the physical lock when this
696
If lock_write doesn't return a token, then this method is not supported.
698
self.control_files.leave_in_place()
700
def dont_leave_lock_in_place(self):
701
"""Tell this repository to release the physical lock when this
702
object is unlocked, even if it didn't originally acquire it.
704
If lock_write doesn't return a token, then this method is not supported.
706
self.control_files.dont_leave_in_place()
709
def gather_stats(self, revid=None, committers=None):
710
"""Gather statistics from a revision id.
712
:param revid: The revision id to gather statistics from, if None, then
713
no revision specific statistics are gathered.
714
:param committers: Optional parameter controlling whether to grab
715
a count of committers from the revision specific statistics.
716
:return: A dictionary of statistics. Currently this contains:
717
committers: The number of committers if requested.
718
firstrev: A tuple with timestamp, timezone for the penultimate left
719
most ancestor of revid, if revid is not the NULL_REVISION.
720
latestrev: A tuple with timestamp, timezone for revid, if revid is
721
not the NULL_REVISION.
722
revisions: The total revision count in the repository.
723
size: An estimate disk size of the repository in bytes.
726
if revid and committers:
727
result['committers'] = 0
728
if revid and revid != _mod_revision.NULL_REVISION:
730
all_committers = set()
731
revisions = self.get_ancestry(revid)
732
# pop the leading None
734
first_revision = None
736
# ignore the revisions in the middle - just grab first and last
737
revisions = revisions[0], revisions[-1]
738
for revision in self.get_revisions(revisions):
739
if not first_revision:
740
first_revision = revision
742
all_committers.add(revision.committer)
743
last_revision = revision
745
result['committers'] = len(all_committers)
746
result['firstrev'] = (first_revision.timestamp,
747
first_revision.timezone)
748
result['latestrev'] = (last_revision.timestamp,
749
last_revision.timezone)
751
# now gather global repository information
752
if self.bzrdir.root_transport.listable():
753
c, t = self._revision_store.total_size(self.get_transaction())
754
result['revisions'] = c
758
def get_data_stream(self, revision_ids):
759
raise NotImplementedError(self.get_data_stream)
761
def insert_data_stream(self, stream):
762
"""XXX What does this really do?
764
Is it a substitute for fetch?
765
Should it manage its own write group ?
767
for item_key, bytes in stream:
768
if item_key[0] == 'file':
769
(file_id,) = item_key[1:]
770
knit = self.weave_store.get_weave_or_empty(
771
file_id, self.get_transaction())
772
elif item_key == ('inventory',):
773
knit = self.get_inventory_weave()
774
elif item_key == ('revisions',):
775
knit = self._revision_store.get_revision_file(
776
self.get_transaction())
777
elif item_key == ('signatures',):
778
knit = self._revision_store.get_signature_file(
779
self.get_transaction())
781
raise RepositoryDataStreamError(
782
"Unrecognised data stream key '%s'" % (item_key,))
783
decoded_list = bencode.bdecode(bytes)
784
format = decoded_list.pop(0)
787
for version, options, parents, some_bytes in decoded_list:
788
data_list.append((version, options, len(some_bytes), parents))
789
knit_bytes += some_bytes
790
knit.insert_data_stream(
791
(format, data_list, StringIO(knit_bytes).read))
794
def missing_revision_ids(self, other, revision_id=None):
795
"""Return the revision ids that other has that this does not.
797
These are returned in topological order.
799
revision_id: only return revision ids included by revision_id.
801
return InterRepository.get(other, self).missing_revision_ids(revision_id)
805
"""Open the repository rooted at base.
807
For instance, if the repository is at URL/.bzr/repository,
808
Repository.open(URL) -> a Repository instance.
810
control = bzrdir.BzrDir.open(base)
811
return control.open_repository()
813
def copy_content_into(self, destination, revision_id=None):
814
"""Make a complete copy of the content in self into destination.
816
This is a destructive operation! Do not use it on existing
819
return InterRepository.get(self, destination).copy_content(revision_id)
821
def commit_write_group(self):
822
"""Commit the contents accrued within the current write group.
824
:seealso: start_write_group.
826
if self._write_group is not self.get_transaction():
827
# has an unlock or relock occured ?
828
raise errors.BzrError('mismatched lock context %r and '
830
(self.get_transaction(), self._write_group))
831
self._commit_write_group()
832
self._write_group = None
834
def _commit_write_group(self):
835
"""Template method for per-repository write group cleanup.
837
This is called before the write group is considered to be
838
finished and should ensure that all data handed to the repository
839
for writing during the write group is safely committed (to the
840
extent possible considering file system caching etc).
843
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
844
"""Fetch the content required to construct revision_id from source.
846
If revision_id is None all content is copied.
847
:param find_ghosts: Find and copy revisions in the source that are
848
ghosts in the target (and not reachable directly by walking out to
849
the first-present revision in target from revision_id).
851
# fast path same-url fetch operations
852
if self.has_same_location(source):
853
# check that last_revision is in 'from' and then return a
855
if (revision_id is not None and
856
not _mod_revision.is_null(revision_id)):
857
self.get_revision(revision_id)
859
inter = InterRepository.get(source, self)
861
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
862
except NotImplementedError:
863
raise errors.IncompatibleRepositories(source, self)
865
def create_bundle(self, target, base, fileobj, format=None):
866
return serializer.write_bundle(self, target, base, fileobj, format)
868
def get_commit_builder(self, branch, parents, config, timestamp=None,
869
timezone=None, committer=None, revprops=None,
871
"""Obtain a CommitBuilder for this repository.
873
:param branch: Branch to commit to.
874
:param parents: Revision ids of the parents of the new revision.
875
:param config: Configuration to use.
876
:param timestamp: Optional timestamp recorded for commit.
877
:param timezone: Optional timezone for timestamp.
878
:param committer: Optional committer to set for commit.
879
:param revprops: Optional dictionary of revision properties.
880
:param revision_id: Optional revision id.
882
result = self._commit_builder_class(self, parents, config,
883
timestamp, timezone, committer, revprops, revision_id)
884
self.start_write_group()
888
if (self.control_files._lock_count == 1 and
889
self.control_files._lock_mode == 'w'):
890
if self._write_group is not None:
891
self.abort_write_group()
892
self.control_files.unlock()
893
raise errors.BzrError(
894
'Must end write groups before releasing write locks.')
895
self.control_files.unlock()
898
def clone(self, a_bzrdir, revision_id=None):
899
"""Clone this repository into a_bzrdir using the current format.
901
Currently no check is made that the format of this repository and
902
the bzrdir format are compatible. FIXME RBC 20060201.
904
:return: The newly created destination repository.
906
# TODO: deprecate after 0.16; cloning this with all its settings is
907
# probably not very useful -- mbp 20070423
908
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
909
self.copy_content_into(dest_repo, revision_id)
912
def start_write_group(self):
913
"""Start a write group in the repository.
915
Write groups are used by repositories which do not have a 1:1 mapping
916
between file ids and backend store to manage the insertion of data from
917
both fetch and commit operations.
919
A write lock is required around the start_write_group/commit_write_group
920
for the support of lock-requiring repository formats.
922
One can only insert data into a repository inside a write group.
926
if not self.is_write_locked():
927
raise errors.NotWriteLocked(self)
928
if self._write_group:
929
raise errors.BzrError('already in a write group')
930
self._start_write_group()
931
# so we can detect unlock/relock - the write group is now entered.
932
self._write_group = self.get_transaction()
934
def _start_write_group(self):
935
"""Template method for per-repository write group startup.
937
This is called before the write group is considered to be
942
def sprout(self, to_bzrdir, revision_id=None):
943
"""Create a descendent repository for new development.
945
Unlike clone, this does not copy the settings of the repository.
947
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
948
dest_repo.fetch(self, revision_id=revision_id)
951
def _create_sprouting_repo(self, a_bzrdir, shared):
952
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
953
# use target default format.
954
dest_repo = a_bzrdir.create_repository()
956
# Most control formats need the repository to be specifically
957
# created, but on some old all-in-one formats it's not needed
959
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
960
except errors.UninitializableFormat:
961
dest_repo = a_bzrdir.open_repository()
965
def has_revision(self, revision_id):
966
"""True if this repository has a copy of the revision."""
967
if 'evil' in debug.debug_flags:
968
mutter_callsite(3, "has_revision is a LBYL symptom.")
969
return self._revision_store.has_revision_id(revision_id,
970
self.get_transaction())
973
def get_revision(self, revision_id):
974
"""Return the Revision object for a named revision."""
975
return self.get_revisions([revision_id])[0]
978
def get_revision_reconcile(self, revision_id):
979
"""'reconcile' helper routine that allows access to a revision always.
981
This variant of get_revision does not cross check the weave graph
982
against the revision one as get_revision does: but it should only
983
be used by reconcile, or reconcile-alike commands that are correcting
984
or testing the revision graph.
986
return self._get_revisions([revision_id])[0]
989
def get_revisions(self, revision_ids):
990
"""Get many revisions at once."""
991
return self._get_revisions(revision_ids)
994
def _get_revisions(self, revision_ids):
995
"""Core work logic to get many revisions without sanity checks."""
996
for rev_id in revision_ids:
997
if not rev_id or not isinstance(rev_id, basestring):
998
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
999
revs = self._revision_store.get_revisions(revision_ids,
1000
self.get_transaction())
1002
assert not isinstance(rev.revision_id, unicode)
1003
for parent_id in rev.parent_ids:
1004
assert not isinstance(parent_id, unicode)
1008
def get_revision_xml(self, revision_id):
1009
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1010
# would have already do it.
1011
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1012
rev = self.get_revision(revision_id)
1013
rev_tmp = StringIO()
1014
# the current serializer..
1015
self._revision_store._serializer.write_revision(rev, rev_tmp)
1017
return rev_tmp.getvalue()
1020
def get_deltas_for_revisions(self, revisions):
1021
"""Produce a generator of revision deltas.
1023
Note that the input is a sequence of REVISIONS, not revision_ids.
1024
Trees will be held in memory until the generator exits.
1025
Each delta is relative to the revision's lefthand predecessor.
1027
required_trees = set()
1028
for revision in revisions:
1029
required_trees.add(revision.revision_id)
1030
required_trees.update(revision.parent_ids[:1])
1031
trees = dict((t.get_revision_id(), t) for
1032
t in self.revision_trees(required_trees))
1033
for revision in revisions:
1034
if not revision.parent_ids:
1035
old_tree = self.revision_tree(None)
1037
old_tree = trees[revision.parent_ids[0]]
1038
yield trees[revision.revision_id].changes_from(old_tree)
1041
def get_revision_delta(self, revision_id):
1042
"""Return the delta for one revision.
1044
The delta is relative to the left-hand predecessor of the
1047
r = self.get_revision(revision_id)
1048
return list(self.get_deltas_for_revisions([r]))[0]
1051
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1052
signature = gpg_strategy.sign(plaintext)
1053
self._revision_store.add_revision_signature_text(revision_id,
1055
self.get_transaction())
1057
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1059
"""Helper routine for fileids_altered_by_revision_ids.
1061
This performs the translation of xml lines to revision ids.
1063
:param line_iterator: An iterator of lines
1064
:param revision_ids: The revision ids to filter for. This should be a
1065
set or other type which supports efficient __contains__ lookups, as
1066
the revision id from each parsed line will be looked up in the
1067
revision_ids filter.
1068
:return: a dictionary mapping altered file-ids to an iterable of
1069
revision_ids. Each altered file-ids has the exact revision_ids that
1070
altered it listed explicitly.
1074
# this code needs to read every new line in every inventory for the
1075
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1076
# not present in one of those inventories is unnecessary but not
1077
# harmful because we are filtering by the revision id marker in the
1078
# inventory lines : we only select file ids altered in one of those
1079
# revisions. We don't need to see all lines in the inventory because
1080
# only those added in an inventory in rev X can contain a revision=X
1082
unescape_revid_cache = {}
1083
unescape_fileid_cache = {}
1085
# jam 20061218 In a big fetch, this handles hundreds of thousands
1086
# of lines, so it has had a lot of inlining and optimizing done.
1087
# Sorry that it is a little bit messy.
1088
# Move several functions to be local variables, since this is a long
1090
search = self._file_ids_altered_regex.search
1091
unescape = _unescape_xml
1092
setdefault = result.setdefault
1093
for line in line_iterator:
1094
match = search(line)
1097
# One call to match.group() returning multiple items is quite a
1098
# bit faster than 2 calls to match.group() each returning 1
1099
file_id, revision_id = match.group('file_id', 'revision_id')
1101
# Inlining the cache lookups helps a lot when you make 170,000
1102
# lines and 350k ids, versus 8.4 unique ids.
1103
# Using a cache helps in 2 ways:
1104
# 1) Avoids unnecessary decoding calls
1105
# 2) Re-uses cached strings, which helps in future set and
1107
# (2) is enough that removing encoding entirely along with
1108
# the cache (so we are using plain strings) results in no
1109
# performance improvement.
1111
revision_id = unescape_revid_cache[revision_id]
1113
unescaped = unescape(revision_id)
1114
unescape_revid_cache[revision_id] = unescaped
1115
revision_id = unescaped
1117
if revision_id in revision_ids:
1119
file_id = unescape_fileid_cache[file_id]
1121
unescaped = unescape(file_id)
1122
unescape_fileid_cache[file_id] = unescaped
1124
setdefault(file_id, set()).add(revision_id)
1127
def fileids_altered_by_revision_ids(self, revision_ids):
1128
"""Find the file ids and versions affected by revisions.
1130
:param revisions: an iterable containing revision ids.
1131
:return: a dictionary mapping altered file-ids to an iterable of
1132
revision_ids. Each altered file-ids has the exact revision_ids that
1133
altered it listed explicitly.
1135
assert self._serializer.support_altered_by_hack, \
1136
("fileids_altered_by_revision_ids only supported for branches "
1137
"which store inventory as unnested xml, not on %r" % self)
1138
selected_revision_ids = set(revision_ids)
1139
w = self.get_inventory_weave()
1140
pb = ui.ui_factory.nested_progress_bar()
1142
return self._find_file_ids_from_xml_inventory_lines(
1143
w.iter_lines_added_or_present_in_versions(
1144
selected_revision_ids, pb=pb),
1145
selected_revision_ids)
1149
def iter_files_bytes(self, desired_files):
1150
"""Iterate through file versions.
1152
Files will not necessarily be returned in the order they occur in
1153
desired_files. No specific order is guaranteed.
1155
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1156
value supplied by the caller as part of desired_files. It should
1157
uniquely identify the file version in the caller's context. (Examples:
1158
an index number or a TreeTransform trans_id.)
1160
bytes_iterator is an iterable of bytestrings for the file. The
1161
kind of iterable and length of the bytestrings are unspecified, but for
1162
this implementation, it is a list of lines produced by
1163
VersionedFile.get_lines().
1165
:param desired_files: a list of (file_id, revision_id, identifier)
1168
transaction = self.get_transaction()
1169
for file_id, revision_id, callable_data in desired_files:
1171
weave = self.weave_store.get_weave(file_id, transaction)
1172
except errors.NoSuchFile:
1173
raise errors.NoSuchIdInRepository(self, file_id)
1174
yield callable_data, weave.get_lines(revision_id)
1176
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1177
"""Get an iterable listing the keys of all the data introduced by a set
1180
The keys will be ordered so that the corresponding items can be safely
1181
fetched and inserted in that order.
1183
:returns: An iterable producing tuples of (knit-kind, file-id,
1184
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1185
'revisions'. file-id is None unless knit-kind is 'file'.
1187
# XXX: it's a bit weird to control the inventory weave caching in this
1188
# generator. Ideally the caching would be done in fetch.py I think. Or
1189
# maybe this generator should explicitly have the contract that it
1190
# should not be iterated until the previously yielded item has been
1193
inv_w = self.get_inventory_weave()
1194
inv_w.enable_cache()
1196
# file ids that changed
1197
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1199
num_file_ids = len(file_ids)
1200
for file_id, altered_versions in file_ids.iteritems():
1201
if _files_pb is not None:
1202
_files_pb.update("fetch texts", count, num_file_ids)
1204
yield ("file", file_id, altered_versions)
1205
# We're done with the files_pb. Note that it finished by the caller,
1206
# just as it was created by the caller.
1210
yield ("inventory", None, revision_ids)
1214
revisions_with_signatures = set()
1215
for rev_id in revision_ids:
1217
self.get_signature_text(rev_id)
1218
except errors.NoSuchRevision:
1222
revisions_with_signatures.add(rev_id)
1224
yield ("signatures", None, revisions_with_signatures)
1227
yield ("revisions", None, revision_ids)
1230
def get_inventory_weave(self):
1231
return self.control_weaves.get_weave('inventory',
1232
self.get_transaction())
1235
def get_inventory(self, revision_id):
1236
"""Get Inventory object by hash."""
1237
return self.deserialise_inventory(
1238
revision_id, self.get_inventory_xml(revision_id))
1240
def deserialise_inventory(self, revision_id, xml):
1241
"""Transform the xml into an inventory object.
1243
:param revision_id: The expected revision id of the inventory.
1244
:param xml: A serialised inventory.
1246
return self._serializer.read_inventory_from_string(xml, revision_id)
1248
def serialise_inventory(self, inv):
1249
return self._serializer.write_inventory_to_string(inv)
1251
def _serialise_inventory_to_lines(self, inv):
1252
return self._serializer.write_inventory_to_lines(inv)
1254
def get_serializer_format(self):
1255
return self._serializer.format_num
1258
def get_inventory_xml(self, revision_id):
1259
"""Get inventory XML as a file object."""
1261
assert isinstance(revision_id, str), type(revision_id)
1262
iw = self.get_inventory_weave()
1263
return iw.get_text(revision_id)
1265
raise errors.HistoryMissing(self, 'inventory', revision_id)
1268
def get_inventory_sha1(self, revision_id):
1269
"""Return the sha1 hash of the inventory entry
1271
return self.get_revision(revision_id).inventory_sha1
1274
def get_revision_graph(self, revision_id=None):
1275
"""Return a dictionary containing the revision graph.
1277
NB: This method should not be used as it accesses the entire graph all
1278
at once, which is much more data than most operations should require.
1280
:param revision_id: The revision_id to get a graph from. If None, then
1281
the entire revision graph is returned. This is a deprecated mode of
1282
operation and will be removed in the future.
1283
:return: a dictionary of revision_id->revision_parents_list.
1285
raise NotImplementedError(self.get_revision_graph)
1288
def get_revision_graph_with_ghosts(self, revision_ids=None):
1289
"""Return a graph of the revisions with ghosts marked as applicable.
1291
:param revision_ids: an iterable of revisions to graph or None for all.
1292
:return: a Graph object with the graph reachable from revision_ids.
1294
if 'evil' in debug.debug_flags:
1296
"get_revision_graph_with_ghosts scales with size of history.")
1297
result = deprecated_graph.Graph()
1298
if not revision_ids:
1299
pending = set(self.all_revision_ids())
1302
pending = set(revision_ids)
1303
# special case NULL_REVISION
1304
if _mod_revision.NULL_REVISION in pending:
1305
pending.remove(_mod_revision.NULL_REVISION)
1306
required = set(pending)
1309
revision_id = pending.pop()
1311
rev = self.get_revision(revision_id)
1312
except errors.NoSuchRevision:
1313
if revision_id in required:
1316
result.add_ghost(revision_id)
1318
for parent_id in rev.parent_ids:
1319
# is this queued or done ?
1320
if (parent_id not in pending and
1321
parent_id not in done):
1323
pending.add(parent_id)
1324
result.add_node(revision_id, rev.parent_ids)
1325
done.add(revision_id)
1328
def _get_history_vf(self):
1329
"""Get a versionedfile whose history graph reflects all revisions.
1331
For weave repositories, this is the inventory weave.
1333
return self.get_inventory_weave()
1335
def iter_reverse_revision_history(self, revision_id):
1336
"""Iterate backwards through revision ids in the lefthand history
1338
:param revision_id: The revision id to start with. All its lefthand
1339
ancestors will be traversed.
1341
if revision_id in (None, _mod_revision.NULL_REVISION):
1343
next_id = revision_id
1344
versionedfile = self._get_history_vf()
1347
parents = versionedfile.get_parents(next_id)
1348
if len(parents) == 0:
1351
next_id = parents[0]
1354
def get_revision_inventory(self, revision_id):
1355
"""Return inventory of a past revision."""
1356
# TODO: Unify this with get_inventory()
1357
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1358
# must be the same as its revision, so this is trivial.
1359
if revision_id is None:
1360
# This does not make sense: if there is no revision,
1361
# then it is the current tree inventory surely ?!
1362
# and thus get_root_id() is something that looks at the last
1363
# commit on the branch, and the get_root_id is an inventory check.
1364
raise NotImplementedError
1365
# return Inventory(self.get_root_id())
1367
return self.get_inventory(revision_id)
1370
def is_shared(self):
1371
"""Return True if this repository is flagged as a shared repository."""
1372
raise NotImplementedError(self.is_shared)
1375
def reconcile(self, other=None, thorough=False):
1376
"""Reconcile this repository."""
1377
from bzrlib.reconcile import RepoReconciler
1378
reconciler = RepoReconciler(self, thorough=thorough)
1379
reconciler.reconcile()
1382
def _refresh_data(self):
1383
"""Helper called from lock_* to ensure coherency with disk.
1385
The default implementation does nothing; it is however possible
1386
for repositories to maintain loaded indices across multiple locks
1387
by checking inside their implementation of this method to see
1388
whether their indices are still valid. This depends of course on
1389
the disk format being validatable in this manner.
1393
def revision_tree(self, revision_id):
1394
"""Return Tree for a revision on this branch.
1396
`revision_id` may be None for the empty tree revision.
1398
# TODO: refactor this to use an existing revision object
1399
# so we don't need to read it in twice.
1400
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1401
return RevisionTree(self, Inventory(root_id=None),
1402
_mod_revision.NULL_REVISION)
1404
inv = self.get_revision_inventory(revision_id)
1405
return RevisionTree(self, inv, revision_id)
1408
def revision_trees(self, revision_ids):
1409
"""Return Tree for a revision on this branch.
1411
`revision_id` may not be None or 'null:'"""
1412
assert None not in revision_ids
1413
assert _mod_revision.NULL_REVISION not in revision_ids
1414
texts = self.get_inventory_weave().get_texts(revision_ids)
1415
for text, revision_id in zip(texts, revision_ids):
1416
inv = self.deserialise_inventory(revision_id, text)
1417
yield RevisionTree(self, inv, revision_id)
1420
def get_ancestry(self, revision_id, topo_sorted=True):
1421
"""Return a list of revision-ids integrated by a revision.
1423
The first element of the list is always None, indicating the origin
1424
revision. This might change when we have history horizons, or
1425
perhaps we should have a new API.
1427
This is topologically sorted.
1429
if _mod_revision.is_null(revision_id):
1431
if not self.has_revision(revision_id):
1432
raise errors.NoSuchRevision(self, revision_id)
1433
w = self.get_inventory_weave()
1434
candidates = w.get_ancestry(revision_id, topo_sorted)
1435
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1438
"""Compress the data within the repository.
1440
This operation only makes sense for some repository types. For other
1441
types it should be a no-op that just returns.
1443
This stub method does not require a lock, but subclasses should use
1444
@needs_write_lock as this is a long running call its reasonable to
1445
implicitly lock for the user.
1449
def print_file(self, file, revision_id):
1450
"""Print `file` to stdout.
1452
FIXME RBC 20060125 as John Meinel points out this is a bad api
1453
- it writes to stdout, it assumes that that is valid etc. Fix
1454
by creating a new more flexible convenience function.
1456
tree = self.revision_tree(revision_id)
1457
# use inventory as it was in that revision
1458
file_id = tree.inventory.path2id(file)
1460
# TODO: jam 20060427 Write a test for this code path
1461
# it had a bug in it, and was raising the wrong
1463
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1464
tree.print_file(file_id)
1466
def get_transaction(self):
1467
return self.control_files.get_transaction()
1469
def revision_parents(self, revision_id):
1470
return self.get_inventory_weave().parent_names(revision_id)
1472
def get_parents(self, revision_ids):
1473
"""See StackedParentsProvider.get_parents"""
1475
for revision_id in revision_ids:
1476
if revision_id == _mod_revision.NULL_REVISION:
1480
parents = self.get_revision(revision_id).parent_ids
1481
except errors.NoSuchRevision:
1484
if len(parents) == 0:
1485
parents = [_mod_revision.NULL_REVISION]
1486
parents_list.append(parents)
1489
def _make_parents_provider(self):
1492
def get_graph(self, other_repository=None):
1493
"""Return the graph walker for this repository format"""
1494
parents_provider = self._make_parents_provider()
1495
if (other_repository is not None and
1496
other_repository.bzrdir.transport.base !=
1497
self.bzrdir.transport.base):
1498
parents_provider = graph._StackedParentsProvider(
1499
[parents_provider, other_repository._make_parents_provider()])
1500
return graph.Graph(parents_provider)
1502
def get_versioned_file_checker(self, revisions, revision_versions_cache):
1503
return VersionedFileChecker(revisions, revision_versions_cache, self)
1506
def set_make_working_trees(self, new_value):
1507
"""Set the policy flag for making working trees when creating branches.
1509
This only applies to branches that use this repository.
1511
The default is 'True'.
1512
:param new_value: True to restore the default, False to disable making
1515
raise NotImplementedError(self.set_make_working_trees)
1517
def make_working_trees(self):
1518
"""Returns the policy for making working trees on new branches."""
1519
raise NotImplementedError(self.make_working_trees)
1522
def sign_revision(self, revision_id, gpg_strategy):
1523
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1524
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1527
def has_signature_for_revision_id(self, revision_id):
1528
"""Query for a revision signature for revision_id in the repository."""
1529
return self._revision_store.has_signature(revision_id,
1530
self.get_transaction())
1533
def get_signature_text(self, revision_id):
1534
"""Return the text for a signature."""
1535
return self._revision_store.get_signature_text(revision_id,
1536
self.get_transaction())
1539
def check(self, revision_ids=None):
1540
"""Check consistency of all history of given revision_ids.
1542
Different repository implementations should override _check().
1544
:param revision_ids: A non-empty list of revision_ids whose ancestry
1545
will be checked. Typically the last revision_id of a branch.
1547
return self._check(revision_ids)
1549
def _check(self, revision_ids):
1550
result = check.Check(self)
1554
def _warn_if_deprecated(self):
1555
global _deprecation_warning_done
1556
if _deprecation_warning_done:
1558
_deprecation_warning_done = True
1559
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1560
% (self._format, self.bzrdir.transport.base))
1562
def supports_rich_root(self):
1563
return self._format.rich_root_data
1565
def _check_ascii_revisionid(self, revision_id, method):
1566
"""Private helper for ascii-only repositories."""
1567
# weave repositories refuse to store revisionids that are non-ascii.
1568
if revision_id is not None:
1569
# weaves require ascii revision ids.
1570
if isinstance(revision_id, unicode):
1572
revision_id.encode('ascii')
1573
except UnicodeEncodeError:
1574
raise errors.NonAsciiRevisionId(method, self)
1577
revision_id.decode('ascii')
1578
except UnicodeDecodeError:
1579
raise errors.NonAsciiRevisionId(method, self)
1581
def revision_graph_can_have_wrong_parents(self):
1582
"""Is it possible for this repository to have a revision graph with
1585
If True, then this repository must also implement
1586
_find_inconsistent_revision_parents so that check and reconcile can
1587
check for inconsistencies before proceeding with other checks that may
1588
depend on the revision index being consistent.
1590
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1592
# remove these delegates a while after bzr 0.15
1593
def __make_delegated(name, from_module):
1594
def _deprecated_repository_forwarder():
1595
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1596
% (name, from_module),
1599
m = __import__(from_module, globals(), locals(), [name])
1601
return getattr(m, name)
1602
except AttributeError:
1603
raise AttributeError('module %s has no name %s'
1605
globals()[name] = _deprecated_repository_forwarder
1608
'AllInOneRepository',
1609
'WeaveMetaDirRepository',
1610
'PreSplitOutRepositoryFormat',
1611
'RepositoryFormat4',
1612
'RepositoryFormat5',
1613
'RepositoryFormat6',
1614
'RepositoryFormat7',
1616
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1620
'RepositoryFormatKnit',
1621
'RepositoryFormatKnit1',
1623
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1626
def install_revision(repository, rev, revision_tree):
1627
"""Install all revision data into a repository."""
1628
repository.start_write_group()
1630
_install_revision(repository, rev, revision_tree)
1632
repository.abort_write_group()
1635
repository.commit_write_group()
1638
def _install_revision(repository, rev, revision_tree):
1639
"""Install all revision data into a repository."""
1640
present_parents = []
1642
for p_id in rev.parent_ids:
1643
if repository.has_revision(p_id):
1644
present_parents.append(p_id)
1645
parent_trees[p_id] = repository.revision_tree(p_id)
1647
parent_trees[p_id] = repository.revision_tree(None)
1649
inv = revision_tree.inventory
1650
entries = inv.iter_entries()
1651
# backwards compatibility hack: skip the root id.
1652
if not repository.supports_rich_root():
1653
path, root = entries.next()
1654
if root.revision != rev.revision_id:
1655
raise errors.IncompatibleRevision(repr(repository))
1656
# Add the texts that are not already present
1657
for path, ie in entries:
1658
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1659
repository.get_transaction())
1660
if ie.revision not in w:
1662
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1663
# with InventoryEntry.find_previous_heads(). if it is, then there
1664
# is a latent bug here where the parents may have ancestors of each
1666
for revision, tree in parent_trees.iteritems():
1667
if ie.file_id not in tree:
1669
parent_id = tree.inventory[ie.file_id].revision
1670
if parent_id in text_parents:
1672
text_parents.append(parent_id)
1674
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1675
repository.get_transaction())
1676
lines = revision_tree.get_file(ie.file_id).readlines()
1677
vfile.add_lines(rev.revision_id, text_parents, lines)
1679
# install the inventory
1680
repository.add_inventory(rev.revision_id, inv, present_parents)
1681
except errors.RevisionAlreadyPresent:
1683
repository.add_revision(rev.revision_id, rev, inv)
1686
class MetaDirRepository(Repository):
1687
"""Repositories in the new meta-dir layout."""
1689
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1690
super(MetaDirRepository, self).__init__(_format,
1696
dir_mode = self.control_files._dir_mode
1697
file_mode = self.control_files._file_mode
1700
def is_shared(self):
1701
"""Return True if this repository is flagged as a shared repository."""
1702
return self.control_files._transport.has('shared-storage')
1705
def set_make_working_trees(self, new_value):
1706
"""Set the policy flag for making working trees when creating branches.
1708
This only applies to branches that use this repository.
1710
The default is 'True'.
1711
:param new_value: True to restore the default, False to disable making
1716
self.control_files._transport.delete('no-working-trees')
1717
except errors.NoSuchFile:
1720
self.control_files.put_utf8('no-working-trees', '')
1722
def make_working_trees(self):
1723
"""Returns the policy for making working trees on new branches."""
1724
return not self.control_files._transport.has('no-working-trees')
1727
class RepositoryFormatRegistry(registry.Registry):
1728
"""Registry of RepositoryFormats."""
1730
def get(self, format_string):
1731
r = registry.Registry.get(self, format_string)
1737
format_registry = RepositoryFormatRegistry()
1738
"""Registry of formats, indexed by their identifying format string.
1740
This can contain either format instances themselves, or classes/factories that
1741
can be called to obtain one.
1745
#####################################################################
1746
# Repository Formats
1748
class RepositoryFormat(object):
1749
"""A repository format.
1751
Formats provide three things:
1752
* An initialization routine to construct repository data on disk.
1753
* a format string which is used when the BzrDir supports versioned
1755
* an open routine which returns a Repository instance.
1757
There is one and only one Format subclass for each on-disk format. But
1758
there can be one Repository subclass that is used for several different
1759
formats. The _format attribute on a Repository instance can be used to
1760
determine the disk format.
1762
Formats are placed in an dict by their format string for reference
1763
during opening. These should be subclasses of RepositoryFormat
1766
Once a format is deprecated, just deprecate the initialize and open
1767
methods on the format class. Do not deprecate the object, as the
1768
object will be created every system load.
1770
Common instance attributes:
1771
_matchingbzrdir - the bzrdir format that the repository format was
1772
originally written to work with. This can be used if manually
1773
constructing a bzrdir and repository, or more commonly for test suite
1777
# Set to True or False in derived classes. True indicates that the format
1778
# supports ghosts gracefully.
1779
supports_ghosts = None
1782
return "<%s>" % self.__class__.__name__
1784
def __eq__(self, other):
1785
# format objects are generally stateless
1786
return isinstance(other, self.__class__)
1788
def __ne__(self, other):
1789
return not self == other
1792
def find_format(klass, a_bzrdir):
1793
"""Return the format for the repository object in a_bzrdir.
1795
This is used by bzr native formats that have a "format" file in
1796
the repository. Other methods may be used by different types of
1800
transport = a_bzrdir.get_repository_transport(None)
1801
format_string = transport.get("format").read()
1802
return format_registry.get(format_string)
1803
except errors.NoSuchFile:
1804
raise errors.NoRepositoryPresent(a_bzrdir)
1806
raise errors.UnknownFormatError(format=format_string)
1809
def register_format(klass, format):
1810
format_registry.register(format.get_format_string(), format)
1813
def unregister_format(klass, format):
1814
format_registry.remove(format.get_format_string())
1817
def get_default_format(klass):
1818
"""Return the current default format."""
1819
from bzrlib import bzrdir
1820
return bzrdir.format_registry.make_bzrdir('default').repository_format
1822
def _get_control_store(self, repo_transport, control_files):
1823
"""Return the control store for this repository."""
1824
raise NotImplementedError(self._get_control_store)
1826
def get_format_string(self):
1827
"""Return the ASCII format string that identifies this format.
1829
Note that in pre format ?? repositories the format string is
1830
not permitted nor written to disk.
1832
raise NotImplementedError(self.get_format_string)
1834
def get_format_description(self):
1835
"""Return the short description for this format."""
1836
raise NotImplementedError(self.get_format_description)
1838
def _get_revision_store(self, repo_transport, control_files):
1839
"""Return the revision store object for this a_bzrdir."""
1840
raise NotImplementedError(self._get_revision_store)
1842
def _get_text_rev_store(self,
1849
"""Common logic for getting a revision store for a repository.
1851
see self._get_revision_store for the subclass-overridable method to
1852
get the store for a repository.
1854
from bzrlib.store.revision.text import TextRevisionStore
1855
dir_mode = control_files._dir_mode
1856
file_mode = control_files._file_mode
1857
text_store = TextStore(transport.clone(name),
1859
compressed=compressed,
1861
file_mode=file_mode)
1862
_revision_store = TextRevisionStore(text_store, serializer)
1863
return _revision_store
1865
# TODO: this shouldn't be in the base class, it's specific to things that
1866
# use weaves or knits -- mbp 20070207
1867
def _get_versioned_file_store(self,
1872
versionedfile_class=None,
1873
versionedfile_kwargs={},
1875
if versionedfile_class is None:
1876
versionedfile_class = self._versionedfile_class
1877
weave_transport = control_files._transport.clone(name)
1878
dir_mode = control_files._dir_mode
1879
file_mode = control_files._file_mode
1880
return VersionedFileStore(weave_transport, prefixed=prefixed,
1882
file_mode=file_mode,
1883
versionedfile_class=versionedfile_class,
1884
versionedfile_kwargs=versionedfile_kwargs,
1887
def initialize(self, a_bzrdir, shared=False):
1888
"""Initialize a repository of this format in a_bzrdir.
1890
:param a_bzrdir: The bzrdir to put the new repository in it.
1891
:param shared: The repository should be initialized as a sharable one.
1892
:returns: The new repository object.
1894
This may raise UninitializableFormat if shared repository are not
1895
compatible the a_bzrdir.
1897
raise NotImplementedError(self.initialize)
1899
def is_supported(self):
1900
"""Is this format supported?
1902
Supported formats must be initializable and openable.
1903
Unsupported formats may not support initialization or committing or
1904
some other features depending on the reason for not being supported.
1908
def check_conversion_target(self, target_format):
1909
raise NotImplementedError(self.check_conversion_target)
1911
def open(self, a_bzrdir, _found=False):
1912
"""Return an instance of this format for the bzrdir a_bzrdir.
1914
_found is a private parameter, do not use it.
1916
raise NotImplementedError(self.open)
1919
class MetaDirRepositoryFormat(RepositoryFormat):
1920
"""Common base class for the new repositories using the metadir layout."""
1922
rich_root_data = False
1923
supports_tree_reference = False
1924
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
1927
super(MetaDirRepositoryFormat, self).__init__()
1929
def _create_control_files(self, a_bzrdir):
1930
"""Create the required files and the initial control_files object."""
1931
# FIXME: RBC 20060125 don't peek under the covers
1932
# NB: no need to escape relative paths that are url safe.
1933
repository_transport = a_bzrdir.get_repository_transport(self)
1934
control_files = lockable_files.LockableFiles(repository_transport,
1935
'lock', lockdir.LockDir)
1936
control_files.create_lock()
1937
return control_files
1939
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
1940
"""Upload the initial blank content."""
1941
control_files = self._create_control_files(a_bzrdir)
1942
control_files.lock_write()
1944
control_files._transport.mkdir_multi(dirs,
1945
mode=control_files._dir_mode)
1946
for file, content in files:
1947
control_files.put(file, content)
1948
for file, content in utf8_files:
1949
control_files.put_utf8(file, content)
1951
control_files.put_utf8('shared-storage', '')
1953
control_files.unlock()
1956
# formats which have no format string are not discoverable
1957
# and not independently creatable, so are not registered. They're
1958
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1959
# needed, it's constructed directly by the BzrDir. Non-native formats where
1960
# the repository is not separately opened are similar.
1962
format_registry.register_lazy(
1963
'Bazaar-NG Repository format 7',
1964
'bzrlib.repofmt.weaverepo',
1968
# KEEP in sync with bzrdir.format_registry default, which controls the overall
1969
# default control directory format
1970
format_registry.register_lazy(
1971
'Bazaar-NG Knit Repository Format 1',
1972
'bzrlib.repofmt.knitrepo',
1973
'RepositoryFormatKnit1',
1975
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
1977
format_registry.register_lazy(
1978
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
1979
'bzrlib.repofmt.knitrepo',
1980
'RepositoryFormatKnit3',
1983
# Pack-based formats. There is one format for pre-subtrees, and one for
1984
# post-subtrees to allow ease of testing.
1985
# NOTE: These are experimental in 0.92.
1986
format_registry.register_lazy(
1987
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
1988
'bzrlib.repofmt.pack_repo',
1989
'RepositoryFormatKnitPack1',
1991
format_registry.register_lazy(
1992
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
1993
'bzrlib.repofmt.pack_repo',
1994
'RepositoryFormatKnitPack3',
1998
class InterRepository(InterObject):
1999
"""This class represents operations taking place between two repositories.
2001
Its instances have methods like copy_content and fetch, and contain
2002
references to the source and target repositories these operations can be
2005
Often we will provide convenience methods on 'repository' which carry out
2006
operations with another repository - they will always forward to
2007
InterRepository.get(other).method_name(parameters).
2011
"""The available optimised InterRepository types."""
2013
def copy_content(self, revision_id=None):
2014
raise NotImplementedError(self.copy_content)
2016
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2017
"""Fetch the content required to construct revision_id.
2019
The content is copied from self.source to self.target.
2021
:param revision_id: if None all content is copied, if NULL_REVISION no
2023
:param pb: optional progress bar to use for progress reports. If not
2024
provided a default one will be created.
2026
Returns the copied revision count and the failed revisions in a tuple:
2029
raise NotImplementedError(self.fetch)
2032
def missing_revision_ids(self, revision_id=None):
2033
"""Return the revision ids that source has that target does not.
2035
These are returned in topological order.
2037
:param revision_id: only return revision ids included by this
2040
# generic, possibly worst case, slow code path.
2041
target_ids = set(self.target.all_revision_ids())
2042
if revision_id is not None:
2043
source_ids = self.source.get_ancestry(revision_id)
2044
assert source_ids[0] is None
2047
source_ids = self.source.all_revision_ids()
2048
result_set = set(source_ids).difference(target_ids)
2049
# this may look like a no-op: its not. It preserves the ordering
2050
# other_ids had while only returning the members from other_ids
2051
# that we've decided we need.
2052
return [rev_id for rev_id in source_ids if rev_id in result_set]
2055
def _same_model(source, target):
2056
"""True if source and target have the same data representation."""
2057
if source.supports_rich_root() != target.supports_rich_root():
2059
if source._serializer != target._serializer:
2064
class InterSameDataRepository(InterRepository):
2065
"""Code for converting between repositories that represent the same data.
2067
Data format and model must match for this to work.
2071
def _get_repo_format_to_test(self):
2072
"""Repository format for testing with.
2074
InterSameData can pull from subtree to subtree and from non-subtree to
2075
non-subtree, so we test this with the richest repository format.
2077
from bzrlib.repofmt import knitrepo
2078
return knitrepo.RepositoryFormatKnit3()
2081
def is_compatible(source, target):
2082
return InterRepository._same_model(source, target)
2085
def copy_content(self, revision_id=None):
2086
"""Make a complete copy of the content in self into destination.
2088
This copies both the repository's revision data, and configuration information
2089
such as the make_working_trees setting.
2091
This is a destructive operation! Do not use it on existing
2094
:param revision_id: Only copy the content needed to construct
2095
revision_id and its parents.
2098
self.target.set_make_working_trees(self.source.make_working_trees())
2099
except NotImplementedError:
2101
# but don't bother fetching if we have the needed data now.
2102
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2103
self.target.has_revision(revision_id)):
2105
self.target.fetch(self.source, revision_id=revision_id)
2108
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2109
"""See InterRepository.fetch()."""
2110
from bzrlib.fetch import GenericRepoFetcher
2111
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2112
self.source, self.source._format, self.target,
2113
self.target._format)
2114
f = GenericRepoFetcher(to_repository=self.target,
2115
from_repository=self.source,
2116
last_revision=revision_id,
2118
return f.count_copied, f.failed_revisions
2121
class InterWeaveRepo(InterSameDataRepository):
2122
"""Optimised code paths between Weave based repositories.
2124
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2125
implemented lazy inter-object optimisation.
2129
def _get_repo_format_to_test(self):
2130
from bzrlib.repofmt import weaverepo
2131
return weaverepo.RepositoryFormat7()
2134
def is_compatible(source, target):
2135
"""Be compatible with known Weave formats.
2137
We don't test for the stores being of specific types because that
2138
could lead to confusing results, and there is no need to be
2141
from bzrlib.repofmt.weaverepo import (
2147
return (isinstance(source._format, (RepositoryFormat5,
2149
RepositoryFormat7)) and
2150
isinstance(target._format, (RepositoryFormat5,
2152
RepositoryFormat7)))
2153
except AttributeError:
2157
def copy_content(self, revision_id=None):
2158
"""See InterRepository.copy_content()."""
2159
# weave specific optimised path:
2161
self.target.set_make_working_trees(self.source.make_working_trees())
2162
except NotImplementedError:
2164
# FIXME do not peek!
2165
if self.source.control_files._transport.listable():
2166
pb = ui.ui_factory.nested_progress_bar()
2168
self.target.weave_store.copy_all_ids(
2169
self.source.weave_store,
2171
from_transaction=self.source.get_transaction(),
2172
to_transaction=self.target.get_transaction())
2173
pb.update('copying inventory', 0, 1)
2174
self.target.control_weaves.copy_multi(
2175
self.source.control_weaves, ['inventory'],
2176
from_transaction=self.source.get_transaction(),
2177
to_transaction=self.target.get_transaction())
2178
self.target._revision_store.text_store.copy_all_ids(
2179
self.source._revision_store.text_store,
2184
self.target.fetch(self.source, revision_id=revision_id)
2187
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2188
"""See InterRepository.fetch()."""
2189
from bzrlib.fetch import GenericRepoFetcher
2190
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2191
self.source, self.source._format, self.target, self.target._format)
2192
f = GenericRepoFetcher(to_repository=self.target,
2193
from_repository=self.source,
2194
last_revision=revision_id,
2196
return f.count_copied, f.failed_revisions
2199
def missing_revision_ids(self, revision_id=None):
2200
"""See InterRepository.missing_revision_ids()."""
2201
# we want all revisions to satisfy revision_id in source.
2202
# but we don't want to stat every file here and there.
2203
# we want then, all revisions other needs to satisfy revision_id
2204
# checked, but not those that we have locally.
2205
# so the first thing is to get a subset of the revisions to
2206
# satisfy revision_id in source, and then eliminate those that
2207
# we do already have.
2208
# this is slow on high latency connection to self, but as as this
2209
# disk format scales terribly for push anyway due to rewriting
2210
# inventory.weave, this is considered acceptable.
2212
if revision_id is not None:
2213
source_ids = self.source.get_ancestry(revision_id)
2214
assert source_ids[0] is None
2217
source_ids = self.source._all_possible_ids()
2218
source_ids_set = set(source_ids)
2219
# source_ids is the worst possible case we may need to pull.
2220
# now we want to filter source_ids against what we actually
2221
# have in target, but don't try to check for existence where we know
2222
# we do not have a revision as that would be pointless.
2223
target_ids = set(self.target._all_possible_ids())
2224
possibly_present_revisions = target_ids.intersection(source_ids_set)
2225
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2226
required_revisions = source_ids_set.difference(actually_present_revisions)
2227
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2228
if revision_id is not None:
2229
# we used get_ancestry to determine source_ids then we are assured all
2230
# revisions referenced are present as they are installed in topological order.
2231
# and the tip revision was validated by get_ancestry.
2232
return required_topo_revisions
2234
# if we just grabbed the possibly available ids, then
2235
# we only have an estimate of whats available and need to validate
2236
# that against the revision records.
2237
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2240
class InterKnitRepo(InterSameDataRepository):
2241
"""Optimised code paths between Knit based repositories."""
2244
def _get_repo_format_to_test(self):
2245
from bzrlib.repofmt import knitrepo
2246
return knitrepo.RepositoryFormatKnit1()
2249
def is_compatible(source, target):
2250
"""Be compatible with known Knit formats.
2252
We don't test for the stores being of specific types because that
2253
could lead to confusing results, and there is no need to be
2256
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2258
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2259
isinstance(target._format, RepositoryFormatKnit))
2260
except AttributeError:
2262
return are_knits and InterRepository._same_model(source, target)
2265
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2266
"""See InterRepository.fetch()."""
2267
from bzrlib.fetch import KnitRepoFetcher
2268
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2269
self.source, self.source._format, self.target, self.target._format)
2270
f = KnitRepoFetcher(to_repository=self.target,
2271
from_repository=self.source,
2272
last_revision=revision_id,
2274
return f.count_copied, f.failed_revisions
2277
def missing_revision_ids(self, revision_id=None):
2278
"""See InterRepository.missing_revision_ids()."""
2279
if revision_id is not None:
2280
source_ids = self.source.get_ancestry(revision_id)
2281
assert source_ids[0] is None
2284
source_ids = self.source.all_revision_ids()
2285
source_ids_set = set(source_ids)
2286
# source_ids is the worst possible case we may need to pull.
2287
# now we want to filter source_ids against what we actually
2288
# have in target, but don't try to check for existence where we know
2289
# we do not have a revision as that would be pointless.
2290
target_ids = set(self.target.all_revision_ids())
2291
possibly_present_revisions = target_ids.intersection(source_ids_set)
2292
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2293
required_revisions = source_ids_set.difference(actually_present_revisions)
2294
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2295
if revision_id is not None:
2296
# we used get_ancestry to determine source_ids then we are assured all
2297
# revisions referenced are present as they are installed in topological order.
2298
# and the tip revision was validated by get_ancestry.
2299
return required_topo_revisions
2301
# if we just grabbed the possibly available ids, then
2302
# we only have an estimate of whats available and need to validate
2303
# that against the revision records.
2304
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2307
class InterPackRepo(InterSameDataRepository):
2308
"""Optimised code paths between Pack based repositories."""
2311
def _get_repo_format_to_test(self):
2312
from bzrlib.repofmt import pack_repo
2313
return pack_repo.RepositoryFormatKnitPack1()
2316
def is_compatible(source, target):
2317
"""Be compatible with known Pack formats.
2319
We don't test for the stores being of specific types because that
2320
could lead to confusing results, and there is no need to be
2323
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2325
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2326
isinstance(target._format, RepositoryFormatPack))
2327
except AttributeError:
2329
return are_packs and InterRepository._same_model(source, target)
2332
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2333
"""See InterRepository.fetch()."""
2334
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2335
self.source, self.source._format, self.target, self.target._format)
2336
self.count_copied = 0
2337
if revision_id is None:
2339
# everything to do - use pack logic
2340
# to fetch from all packs to one without
2341
# inventory parsing etc, IFF nothing to be copied is in the target.
2343
revision_ids = self.source.all_revision_ids()
2344
# implementing the TODO will involve:
2345
# - detecting when all of a pack is selected
2346
# - avoiding as much as possible pre-selection, so the
2347
# more-core routines such as create_pack_from_packs can filter in
2348
# a just-in-time fashion. (though having a HEADS list on a
2349
# repository might make this a lot easier, because we could
2350
# sensibly detect 'new revisions' without doing a full index scan.
2351
elif _mod_revision.is_null(revision_id):
2356
revision_ids = self.missing_revision_ids(revision_id,
2357
find_ghosts=find_ghosts)
2358
except errors.NoSuchRevision:
2359
raise errors.InstallFailed([revision_id])
2360
packs = self.source._pack_collection.all_packs()
2361
pack = self.target._pack_collection.create_pack_from_packs(
2362
packs, '.fetch', revision_ids,
2364
if pack is not None:
2365
self.target._pack_collection._save_pack_names()
2366
# Trigger an autopack. This may duplicate effort as we've just done
2367
# a pack creation, but for now it is simpler to think about as
2368
# 'upload data, then repack if needed'.
2369
self.target._pack_collection.autopack()
2370
return pack.get_revision_count()
2375
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2376
"""See InterRepository.missing_revision_ids().
2378
:param find_ghosts: Find ghosts throughough the ancestry of
2381
if not find_ghosts and revision_id is not None:
2382
graph = self.source.get_graph()
2383
missing_revs = set()
2384
searcher = graph._make_breadth_first_searcher([revision_id])
2386
self.target._pack_collection.revision_index.combined_index
2387
null_set = frozenset([_mod_revision.NULL_REVISION])
2390
next_revs = set(searcher.next())
2391
except StopIteration:
2393
next_revs.difference_update(null_set)
2394
target_keys = [(key,) for key in next_revs]
2395
have_revs = frozenset(node[1][0] for node in
2396
target_index.iter_entries(target_keys))
2397
missing_revs.update(next_revs - have_revs)
2398
searcher.stop_searching_any(have_revs)
2400
elif revision_id is not None:
2401
source_ids = self.source.get_ancestry(revision_id)
2402
assert source_ids[0] is None
2405
source_ids = self.source.all_revision_ids()
2406
# source_ids is the worst possible case we may need to pull.
2407
# now we want to filter source_ids against what we actually
2408
# have in target, but don't try to check for existence where we know
2409
# we do not have a revision as that would be pointless.
2410
target_ids = set(self.target.all_revision_ids())
2411
return [r for r in source_ids if (r not in target_ids)]
2414
class InterModel1and2(InterRepository):
2417
def _get_repo_format_to_test(self):
2421
def is_compatible(source, target):
2422
if not source.supports_rich_root() and target.supports_rich_root():
2428
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2429
"""See InterRepository.fetch()."""
2430
from bzrlib.fetch import Model1toKnit2Fetcher
2431
f = Model1toKnit2Fetcher(to_repository=self.target,
2432
from_repository=self.source,
2433
last_revision=revision_id,
2435
return f.count_copied, f.failed_revisions
2438
def copy_content(self, revision_id=None):
2439
"""Make a complete copy of the content in self into destination.
2441
This is a destructive operation! Do not use it on existing
2444
:param revision_id: Only copy the content needed to construct
2445
revision_id and its parents.
2448
self.target.set_make_working_trees(self.source.make_working_trees())
2449
except NotImplementedError:
2451
# but don't bother fetching if we have the needed data now.
2452
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2453
self.target.has_revision(revision_id)):
2455
self.target.fetch(self.source, revision_id=revision_id)
2458
class InterKnit1and2(InterKnitRepo):
2461
def _get_repo_format_to_test(self):
2465
def is_compatible(source, target):
2466
"""Be compatible with Knit1 source and Knit3 target"""
2467
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2469
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2470
RepositoryFormatKnit3)
2471
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2472
RepositoryFormatKnitPack3)
2473
return (isinstance(source._format,
2474
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2475
isinstance(target._format,
2476
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2478
except AttributeError:
2482
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2483
"""See InterRepository.fetch()."""
2484
from bzrlib.fetch import Knit1to2Fetcher
2485
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2486
self.source, self.source._format, self.target,
2487
self.target._format)
2488
f = Knit1to2Fetcher(to_repository=self.target,
2489
from_repository=self.source,
2490
last_revision=revision_id,
2492
return f.count_copied, f.failed_revisions
2495
class InterRemoteToOther(InterRepository):
2497
def __init__(self, source, target):
2498
InterRepository.__init__(self, source, target)
2499
self._real_inter = None
2502
def is_compatible(source, target):
2503
if not isinstance(source, remote.RemoteRepository):
2505
source._ensure_real()
2506
real_source = source._real_repository
2507
# Is source's model compatible with target's model, and are they the
2508
# same format? Currently we can only optimise fetching from an
2509
# identical model & format repo.
2510
assert not isinstance(real_source, remote.RemoteRepository), (
2511
"We don't support remote repos backed by remote repos yet.")
2512
return real_source._format == target._format
2515
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2516
"""See InterRepository.fetch()."""
2517
from bzrlib.fetch import RemoteToOtherFetcher
2518
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2519
self.source, self.target, self.target._format)
2520
# TODO: jam 20070210 This should be an assert, not a translate
2521
revision_id = osutils.safe_revision_id(revision_id)
2522
f = RemoteToOtherFetcher(to_repository=self.target,
2523
from_repository=self.source,
2524
last_revision=revision_id,
2526
return f.count_copied, f.failed_revisions
2529
def _get_repo_format_to_test(self):
2533
class InterOtherToRemote(InterRepository):
2535
def __init__(self, source, target):
2536
InterRepository.__init__(self, source, target)
2537
self._real_inter = None
2540
def is_compatible(source, target):
2541
if isinstance(target, remote.RemoteRepository):
2545
def _ensure_real_inter(self):
2546
if self._real_inter is None:
2547
self.target._ensure_real()
2548
real_target = self.target._real_repository
2549
self._real_inter = InterRepository.get(self.source, real_target)
2551
def copy_content(self, revision_id=None):
2552
self._ensure_real_inter()
2553
self._real_inter.copy_content(revision_id=revision_id)
2555
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2556
self._ensure_real_inter()
2557
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2560
def _get_repo_format_to_test(self):
2564
InterRepository.register_optimiser(InterSameDataRepository)
2565
InterRepository.register_optimiser(InterWeaveRepo)
2566
InterRepository.register_optimiser(InterKnitRepo)
2567
InterRepository.register_optimiser(InterModel1and2)
2568
InterRepository.register_optimiser(InterKnit1and2)
2569
InterRepository.register_optimiser(InterPackRepo)
2570
InterRepository.register_optimiser(InterRemoteToOther)
2571
InterRepository.register_optimiser(InterOtherToRemote)
2574
class CopyConverter(object):
2575
"""A repository conversion tool which just performs a copy of the content.
2577
This is slow but quite reliable.
2580
def __init__(self, target_format):
2581
"""Create a CopyConverter.
2583
:param target_format: The format the resulting repository should be.
2585
self.target_format = target_format
2587
def convert(self, repo, pb):
2588
"""Perform the conversion of to_convert, giving feedback via pb.
2590
:param to_convert: The disk object to convert.
2591
:param pb: a progress bar to use for progress information.
2596
# this is only useful with metadir layouts - separated repo content.
2597
# trigger an assertion if not such
2598
repo._format.get_format_string()
2599
self.repo_dir = repo.bzrdir
2600
self.step('Moving repository to repository.backup')
2601
self.repo_dir.transport.move('repository', 'repository.backup')
2602
backup_transport = self.repo_dir.transport.clone('repository.backup')
2603
repo._format.check_conversion_target(self.target_format)
2604
self.source_repo = repo._format.open(self.repo_dir,
2606
_override_transport=backup_transport)
2607
self.step('Creating new repository')
2608
converted = self.target_format.initialize(self.repo_dir,
2609
self.source_repo.is_shared())
2610
converted.lock_write()
2612
self.step('Copying content into repository.')
2613
self.source_repo.copy_content_into(converted)
2616
self.step('Deleting old repository content.')
2617
self.repo_dir.transport.delete_tree('repository.backup')
2618
self.pb.note('repository converted')
2620
def step(self, message):
2621
"""Update the pb by a step."""
2623
self.pb.update(message, self.count, self.total)
2635
def _unescaper(match, _map=_unescape_map):
2636
code = match.group(1)
2640
if not code.startswith('#'):
2642
return unichr(int(code[1:])).encode('utf8')
2648
def _unescape_xml(data):
2649
"""Unescape predefined XML entities in a string of data."""
2651
if _unescape_re is None:
2652
_unescape_re = re.compile('\&([^;]*);')
2653
return _unescape_re.sub(_unescaper, data)
2656
class _RevisionTextVersionCache(object):
2657
"""A cache of the versionedfile versions for revision and file-id."""
2659
def __init__(self, repository):
2660
self.repository = repository
2661
self.revision_versions = {}
2662
self.revision_parents = {}
2663
self.repo_graph = self.repository.get_graph()
2664
# XXX: RBC: I haven't tracked down what uses this, but it would be
2665
# better to use the headscache directly I think.
2666
self.heads = graph.HeadsCache(self.repo_graph).heads
2668
def add_revision_text_versions(self, tree):
2669
"""Cache text version data from the supplied revision tree"""
2671
for path, entry in tree.iter_entries_by_dir():
2672
inv_revisions[entry.file_id] = entry.revision
2673
self.revision_versions[tree.get_revision_id()] = inv_revisions
2674
return inv_revisions
2676
def get_text_version(self, file_id, revision_id):
2677
"""Determine the text version for a given file-id and revision-id"""
2679
inv_revisions = self.revision_versions[revision_id]
2682
tree = self.repository.revision_tree(revision_id)
2683
except errors.RevisionNotPresent:
2684
self.revision_versions[revision_id] = inv_revisions = {}
2686
inv_revisions = self.add_revision_text_versions(tree)
2687
return inv_revisions.get(file_id)
2689
def prepopulate_revs(self, revision_ids):
2690
# Filter out versions that we don't have an inventory for, so that the
2691
# revision_trees() call won't fail.
2692
inv_weave = self.repository.get_inventory_weave()
2693
revs = [r for r in revision_ids if inv_weave.has_version(r)]
2694
# XXX: this loop is very similar to
2695
# bzrlib.fetch.Inter1and2Helper.iter_rev_trees.
2697
mutter('%d revisions left to prepopulate', len(revs))
2698
for tree in self.repository.revision_trees(revs[:100]):
2699
if tree.inventory.revision_id is None:
2700
tree.inventory.revision_id = tree.get_revision_id()
2701
self.add_revision_text_versions(tree)
2704
def get_parents(self, revision_id):
2706
return self.revision_parents[revision_id]
2708
parents = self.repository.get_parents([revision_id])[0]
2709
self.revision_parents[revision_id] = parents
2712
def used_file_versions(self):
2713
"""Return a set of (revision_id, file_id) pairs for each file version
2714
referenced by any inventory cached by this _RevisionTextVersionCache.
2716
If the entire repository has been cached, this can be used to find all
2717
file versions that are actually referenced by inventories. Thus any
2718
other file version is completely unused and can be removed safely.
2721
for inventory_summary in self.revision_versions.itervalues():
2722
result.update(inventory_summary.items())
2726
class VersionedFileChecker(object):
2728
def __init__(self, planned_revisions, revision_versions, repository):
2729
self.planned_revisions = planned_revisions
2730
self.revision_versions = revision_versions
2731
self.repository = repository
2733
def calculate_file_version_parents(self, revision_id, file_id):
2734
"""Calculate the correct parents for a file version according to
2737
text_revision = self.revision_versions.get_text_version(
2738
file_id, revision_id)
2739
if text_revision is None:
2741
parents_of_text_revision = self.revision_versions.get_parents(
2743
parents_from_inventories = []
2744
for parent in parents_of_text_revision:
2745
if parent == _mod_revision.NULL_REVISION:
2747
introduced_in = self.revision_versions.get_text_version(file_id,
2749
if introduced_in is not None:
2750
parents_from_inventories.append(introduced_in)
2751
heads = set(self.revision_versions.heads(parents_from_inventories))
2753
for parent in parents_from_inventories:
2754
if parent in heads and parent not in new_parents:
2755
new_parents.append(parent)
2756
return tuple(new_parents)
2758
def check_file_version_parents(self, weave, file_id):
2759
"""Check the parents stored in a versioned file are correct.
2761
It also detects file versions that are not referenced by their
2762
corresponding revision's inventory.
2764
:returns: A tuple of (wrong_parents, dangling_file_versions).
2765
wrong_parents is a dict mapping {revision_id: (stored_parents,
2766
correct_parents)} for each revision_id where the stored parents
2767
are not correct. dangling_file_versions is a set of (file_id,
2768
revision_id) tuples for versions that are present in this versioned
2769
file, but not used by the corresponding inventory.
2772
dangling_file_versions = set()
2773
for num, revision_id in enumerate(self.planned_revisions):
2774
correct_parents = self.calculate_file_version_parents(
2775
revision_id, file_id)
2776
if correct_parents is None:
2778
text_revision = self.revision_versions.get_text_version(
2779
file_id, revision_id)
2781
knit_parents = tuple(weave.get_parents(revision_id))
2782
except errors.RevisionNotPresent:
2784
if text_revision != revision_id:
2785
# This file version is not referenced by its corresponding
2787
dangling_file_versions.add((file_id, revision_id))
2788
if correct_parents != knit_parents:
2789
wrong_parents[revision_id] = (knit_parents, correct_parents)
2790
return wrong_parents, dangling_file_versions
added
removed
bzrlib/repository.py
