1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from cStringIO import StringIO
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
39
revision as _mod_revision,
44
from bzrlib.bundle import serializer
45
from bzrlib.revisiontree import RevisionTree
46
from bzrlib.store.versioned import VersionedFileStore
47
from bzrlib.store.text import TextStore
48
from bzrlib.testament import Testament
49
from bzrlib.util import bencode
52
from bzrlib.decorators import needs_read_lock, needs_write_lock
53
from bzrlib.inter import InterObject
54
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
55
from bzrlib.symbol_versioning import (
58
from bzrlib.trace import mutter, mutter_callsite, note, warning
61
# Old formats display a warning, but only once
62
_deprecation_warning_done = False
65
class CommitBuilder(object):
66
"""Provides an interface to build up a commit.
68
This allows describing a tree to be committed without needing to
69
know the internals of the format of the repository.
72
# all clients should supply tree roots.
73
record_root_entry = True
74
# the default CommitBuilder does not manage trees whose root is versioned.
75
_versioned_root = False
77
def __init__(self, repository, parents, config, timestamp=None,
78
timezone=None, committer=None, revprops=None,
80
"""Initiate a CommitBuilder.
82
:param repository: Repository to commit to.
83
:param parents: Revision ids of the parents of the new revision.
84
:param config: Configuration to use.
85
:param timestamp: Optional timestamp recorded for commit.
86
:param timezone: Optional timezone for timestamp.
87
:param committer: Optional committer to set for commit.
88
:param revprops: Optional dictionary of revision properties.
89
:param revision_id: Optional revision id.
94
self._committer = self._config.username()
96
assert isinstance(committer, basestring), type(committer)
97
self._committer = committer
99
self.new_inventory = Inventory(None)
100
self._new_revision_id = revision_id
101
self.parents = parents
102
self.repository = repository
105
if revprops is not None:
106
self._revprops.update(revprops)
108
if timestamp is None:
109
timestamp = time.time()
110
# Restrict resolution to 1ms
111
self._timestamp = round(timestamp, 3)
114
self._timezone = osutils.local_time_offset()
116
self._timezone = int(timezone)
118
self._generate_revision_if_needed()
119
self._heads = graph.HeadsCache(repository.get_graph()).heads
121
def commit(self, message):
122
"""Make the actual commit.
124
:return: The revision id of the recorded revision.
126
rev = _mod_revision.Revision(
127
timestamp=self._timestamp,
128
timezone=self._timezone,
129
committer=self._committer,
131
inventory_sha1=self.inv_sha1,
132
revision_id=self._new_revision_id,
133
properties=self._revprops)
134
rev.parent_ids = self.parents
135
self.repository.add_revision(self._new_revision_id, rev,
136
self.new_inventory, self._config)
137
self.repository.commit_write_group()
138
return self._new_revision_id
141
"""Abort the commit that is being built.
143
self.repository.abort_write_group()
145
def revision_tree(self):
146
"""Return the tree that was just committed.
148
After calling commit() this can be called to get a RevisionTree
149
representing the newly committed tree. This is preferred to
150
calling Repository.revision_tree() because that may require
151
deserializing the inventory, while we already have a copy in
154
return RevisionTree(self.repository, self.new_inventory,
155
self._new_revision_id)
157
def finish_inventory(self):
158
"""Tell the builder that the inventory is finished."""
159
if self.new_inventory.root is None:
160
raise AssertionError('Root entry should be supplied to'
161
' record_entry_contents, as of bzr 0.10.',
162
DeprecationWarning, stacklevel=2)
163
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
164
self.new_inventory.revision_id = self._new_revision_id
165
self.inv_sha1 = self.repository.add_inventory(
166
self._new_revision_id,
171
def _gen_revision_id(self):
172
"""Return new revision-id."""
173
return generate_ids.gen_revision_id(self._config.username(),
176
def _generate_revision_if_needed(self):
177
"""Create a revision id if None was supplied.
179
If the repository can not support user-specified revision ids
180
they should override this function and raise CannotSetRevisionId
181
if _new_revision_id is not None.
183
:raises: CannotSetRevisionId
185
if self._new_revision_id is None:
186
self._new_revision_id = self._gen_revision_id()
187
self.random_revid = True
189
self.random_revid = False
191
def _check_root(self, ie, parent_invs, tree):
192
"""Helper for record_entry_contents.
194
:param ie: An entry being added.
195
:param parent_invs: The inventories of the parent revisions of the
197
:param tree: The tree that is being committed.
199
# In this revision format, root entries have no knit or weave When
200
# serializing out to disk and back in root.revision is always
202
ie.revision = self._new_revision_id
204
def _get_delta(self, ie, basis_inv, path):
205
"""Get a delta against the basis inventory for ie."""
206
if ie.file_id not in basis_inv:
208
return (None, path, ie.file_id, ie)
209
elif ie != basis_inv[ie.file_id]:
211
# TODO: avoid tis id2path call.
212
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
217
def record_entry_contents(self, ie, parent_invs, path, tree,
219
"""Record the content of ie from tree into the commit if needed.
221
Side effect: sets ie.revision when unchanged
223
:param ie: An inventory entry present in the commit.
224
:param parent_invs: The inventories of the parent revisions of the
226
:param path: The path the entry is at in the tree.
227
:param tree: The tree which contains this entry and should be used to
229
:param content_summary: Summary data from the tree about the paths
230
content - stat, length, exec, sha/link target. This is only
231
accessed when the entry has a revision of None - that is when it is
232
a candidate to commit.
233
:return: A tuple (change_delta, version_recorded). change_delta is
234
an inventory_delta change for this entry against the basis tree of
235
the commit, or None if no change occured against the basis tree.
236
version_recorded is True if a new version of the entry has been
237
recorded. For instance, committing a merge where a file was only
238
changed on the other side will return (delta, False).
240
if self.new_inventory.root is None:
241
if ie.parent_id is not None:
242
raise errors.RootMissing()
243
self._check_root(ie, parent_invs, tree)
244
if ie.revision is None:
245
kind = content_summary[0]
247
# ie is carried over from a prior commit
249
# XXX: repository specific check for nested tree support goes here - if
250
# the repo doesn't want nested trees we skip it ?
251
if (kind == 'tree-reference' and
252
not self.repository._format.supports_tree_reference):
253
# mismatch between commit builder logic and repository:
254
# this needs the entry creation pushed down into the builder.
255
raise NotImplementedError('Missing repository subtree support.')
256
self.new_inventory.add(ie)
258
# TODO: slow, take it out of the inner loop.
260
basis_inv = parent_invs[0]
262
basis_inv = Inventory(root_id=None)
264
# ie.revision is always None if the InventoryEntry is considered
265
# for committing. We may record the previous parents revision if the
266
# content is actually unchanged against a sole head.
267
if ie.revision is not None:
268
if not self._versioned_root and path == '':
269
# repositories that do not version the root set the root's
270
# revision to the new commit even when no change occurs, and
271
# this masks when a change may have occurred against the basis,
272
# so calculate if one happened.
273
if ie.file_id in basis_inv:
274
delta = (basis_inv.id2path(ie.file_id), path,
278
delta = (None, path, ie.file_id, ie)
281
# we don't need to commit this, because the caller already
282
# determined that an existing revision of this file is
284
return None, (ie.revision == self._new_revision_id)
285
# XXX: Friction: parent_candidates should return a list not a dict
286
# so that we don't have to walk the inventories again.
287
parent_candiate_entries = ie.parent_candidates(parent_invs)
288
head_set = self._heads(parent_candiate_entries.keys())
290
for inv in parent_invs:
291
if ie.file_id in inv:
292
old_rev = inv[ie.file_id].revision
293
if old_rev in head_set:
294
heads.append(inv[ie.file_id].revision)
295
head_set.remove(inv[ie.file_id].revision)
298
# now we check to see if we need to write a new record to the
300
# We write a new entry unless there is one head to the ancestors, and
301
# the kind-derived content is unchanged.
303
# Cheapest check first: no ancestors, or more the one head in the
304
# ancestors, we write a new node.
308
# There is a single head, look it up for comparison
309
parent_entry = parent_candiate_entries[heads[0]]
310
# if the non-content specific data has changed, we'll be writing a
312
if (parent_entry.parent_id != ie.parent_id or
313
parent_entry.name != ie.name):
315
# now we need to do content specific checks:
317
# if the kind changed the content obviously has
318
if kind != parent_entry.kind:
321
assert content_summary[2] is not None, \
322
"Files must not have executable = None"
324
if (# if the file length changed we have to store:
325
parent_entry.text_size != content_summary[1] or
326
# if the exec bit has changed we have to store:
327
parent_entry.executable != content_summary[2]):
329
elif parent_entry.text_sha1 == content_summary[3]:
330
# all meta and content is unchanged (using a hash cache
331
# hit to check the sha)
332
ie.revision = parent_entry.revision
333
ie.text_size = parent_entry.text_size
334
ie.text_sha1 = parent_entry.text_sha1
335
ie.executable = parent_entry.executable
336
return self._get_delta(ie, basis_inv, path), False
338
# Either there is only a hash change(no hash cache entry,
339
# or same size content change), or there is no change on
341
# Provide the parent's hash to the store layer, so that the
342
# content is unchanged we will not store a new node.
343
nostore_sha = parent_entry.text_sha1
345
# We want to record a new node regardless of the presence or
346
# absence of a content change in the file.
348
ie.executable = content_summary[2]
349
lines = tree.get_file(ie.file_id, path).readlines()
351
ie.text_sha1, ie.text_size = self._add_text_to_weave(
352
ie.file_id, lines, heads, nostore_sha)
353
except errors.ExistingContent:
354
# Turns out that the file content was unchanged, and we were
355
# only going to store a new node if it was changed. Carry over
357
ie.revision = parent_entry.revision
358
ie.text_size = parent_entry.text_size
359
ie.text_sha1 = parent_entry.text_sha1
360
ie.executable = parent_entry.executable
361
return self._get_delta(ie, basis_inv, path), False
362
elif kind == 'directory':
364
# all data is meta here, nothing specific to directory, so
366
ie.revision = parent_entry.revision
367
return self._get_delta(ie, basis_inv, path), False
369
self._add_text_to_weave(ie.file_id, lines, heads, None)
370
elif kind == 'symlink':
371
current_link_target = content_summary[3]
373
# symlink target is not generic metadata, check if it has
375
if current_link_target != parent_entry.symlink_target:
378
# unchanged, carry over.
379
ie.revision = parent_entry.revision
380
ie.symlink_target = parent_entry.symlink_target
381
return self._get_delta(ie, basis_inv, path), False
382
ie.symlink_target = current_link_target
384
self._add_text_to_weave(ie.file_id, lines, heads, None)
385
elif kind == 'tree-reference':
387
if content_summary[3] != parent_entry.reference_revision:
390
# unchanged, carry over.
391
ie.reference_revision = parent_entry.reference_revision
392
ie.revision = parent_entry.revision
393
return self._get_delta(ie, basis_inv, path), False
394
ie.reference_revision = content_summary[3]
396
self._add_text_to_weave(ie.file_id, lines, heads, None)
398
raise NotImplementedError('unknown kind')
399
ie.revision = self._new_revision_id
400
return self._get_delta(ie, basis_inv, path), True
402
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
403
versionedfile = self.repository.weave_store.get_weave_or_empty(
404
file_id, self.repository.get_transaction())
405
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
406
# than add_lines, and allows committing when a parent is ghosted for
408
# Note: as we read the content directly from the tree, we know its not
409
# been turned into unicode or badly split - but a broken tree
410
# implementation could give us bad output from readlines() so this is
411
# not a guarantee of safety. What would be better is always checking
412
# the content during test suite execution. RBC 20070912
414
return versionedfile.add_lines_with_ghosts(
415
self._new_revision_id, parents, new_lines,
416
nostore_sha=nostore_sha, random_id=self.random_revid,
417
check_content=False)[0:2]
419
versionedfile.clear_cache()
422
class RootCommitBuilder(CommitBuilder):
423
"""This commitbuilder actually records the root id"""
425
# the root entry gets versioned properly by this builder.
426
_versioned_root = True
428
def _check_root(self, ie, parent_invs, tree):
429
"""Helper for record_entry_contents.
431
:param ie: An entry being added.
432
:param parent_invs: The inventories of the parent revisions of the
434
:param tree: The tree that is being committed.
438
######################################################################
441
class Repository(object):
442
"""Repository holding history for one or more branches.
444
The repository holds and retrieves historical information including
445
revisions and file history. It's normally accessed only by the Branch,
446
which views a particular line of development through that history.
448
The Repository builds on top of Stores and a Transport, which respectively
449
describe the disk data format and the way of accessing the (possibly
453
# What class to use for a CommitBuilder. Often its simpler to change this
454
# in a Repository class subclass rather than to override
455
# get_commit_builder.
456
_commit_builder_class = CommitBuilder
457
# The search regex used by xml based repositories to determine what things
458
# where changed in a single commit.
459
_file_ids_altered_regex = lazy_regex.lazy_compile(
460
r'file_id="(?P<file_id>[^"]+)"'
461
r'.* revision="(?P<revision_id>[^"]+)"'
464
def abort_write_group(self):
465
"""Commit the contents accrued within the current write group.
467
:seealso: start_write_group.
469
if self._write_group is not self.get_transaction():
470
# has an unlock or relock occured ?
471
raise errors.BzrError('mismatched lock context and write group.')
472
self._abort_write_group()
473
self._write_group = None
475
def _abort_write_group(self):
476
"""Template method for per-repository write group cleanup.
478
This is called during abort before the write group is considered to be
479
finished and should cleanup any internal state accrued during the write
480
group. There is no requirement that data handed to the repository be
481
*not* made available - this is not a rollback - but neither should any
482
attempt be made to ensure that data added is fully commited. Abort is
483
invoked when an error has occured so futher disk or network operations
484
may not be possible or may error and if possible should not be
489
def add_inventory(self, revision_id, inv, parents):
490
"""Add the inventory inv to the repository as revision_id.
492
:param parents: The revision ids of the parents that revision_id
493
is known to have and are in the repository already.
495
returns the sha1 of the serialized inventory.
497
assert self.is_in_write_group()
498
_mod_revision.check_not_reserved_id(revision_id)
499
assert inv.revision_id is None or inv.revision_id == revision_id, \
500
"Mismatch between inventory revision" \
501
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
502
assert inv.root is not None
503
inv_lines = self._serialise_inventory_to_lines(inv)
504
inv_vf = self.get_inventory_weave()
505
return self._inventory_add_lines(inv_vf, revision_id, parents,
506
inv_lines, check_content=False)
508
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
510
"""Store lines in inv_vf and return the sha1 of the inventory."""
512
for parent in parents:
514
final_parents.append(parent)
515
return inv_vf.add_lines(revision_id, final_parents, lines,
516
check_content=check_content)[0]
519
def add_revision(self, revision_id, rev, inv=None, config=None):
520
"""Add rev to the revision store as revision_id.
522
:param revision_id: the revision id to use.
523
:param rev: The revision object.
524
:param inv: The inventory for the revision. if None, it will be looked
525
up in the inventory storer
526
:param config: If None no digital signature will be created.
527
If supplied its signature_needed method will be used
528
to determine if a signature should be made.
530
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
532
_mod_revision.check_not_reserved_id(revision_id)
533
if config is not None and config.signature_needed():
535
inv = self.get_inventory(revision_id)
536
plaintext = Testament(rev, inv).as_short_text()
537
self.store_revision_signature(
538
gpg.GPGStrategy(config), plaintext, revision_id)
539
if not revision_id in self.get_inventory_weave():
541
raise errors.WeaveRevisionNotPresent(revision_id,
542
self.get_inventory_weave())
544
# yes, this is not suitable for adding with ghosts.
545
self.add_inventory(revision_id, inv, rev.parent_ids)
546
self._revision_store.add_revision(rev, self.get_transaction())
548
def _add_revision_text(self, revision_id, text):
549
revision = self._revision_store._serializer.read_revision_from_string(
551
self._revision_store._add_revision(revision, StringIO(text),
552
self.get_transaction())
554
def all_revision_ids(self):
555
"""Returns a list of all the revision ids in the repository.
557
This is deprecated because code should generally work on the graph
558
reachable from a particular revision, and ignore any other revisions
559
that might be present. There is no direct replacement method.
561
if 'evil' in debug.debug_flags:
562
mutter_callsite(2, "all_revision_ids is linear with history.")
563
return self._all_revision_ids()
565
def _all_revision_ids(self):
566
"""Returns a list of all the revision ids in the repository.
568
These are in as much topological order as the underlying store can
571
raise NotImplementedError(self._all_revision_ids)
573
def break_lock(self):
574
"""Break a lock if one is present from another instance.
576
Uses the ui factory to ask for confirmation if the lock may be from
579
self.control_files.break_lock()
582
def _eliminate_revisions_not_present(self, revision_ids):
583
"""Check every revision id in revision_ids to see if we have it.
585
Returns a set of the present revisions.
588
for id in revision_ids:
589
if self.has_revision(id):
594
def create(a_bzrdir):
595
"""Construct the current default format repository in a_bzrdir."""
596
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
598
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
599
"""instantiate a Repository.
601
:param _format: The format of the repository on disk.
602
:param a_bzrdir: The BzrDir of the repository.
604
In the future we will have a single api for all stores for
605
getting file texts, inventories and revisions, then
606
this construct will accept instances of those things.
608
super(Repository, self).__init__()
609
self._format = _format
610
# the following are part of the public API for Repository:
611
self.bzrdir = a_bzrdir
612
self.control_files = control_files
613
self._revision_store = _revision_store
614
# backwards compatibility
615
self.weave_store = text_store
617
self._reconcile_does_inventory_gc = True
618
self._reconcile_fixes_text_parents = False
619
# not right yet - should be more semantically clear ?
621
self.control_store = control_store
622
self.control_weaves = control_store
623
# TODO: make sure to construct the right store classes, etc, depending
624
# on whether escaping is required.
625
self._warn_if_deprecated()
626
self._write_group = None
627
self.base = control_files._transport.base
630
return '%s(%r)' % (self.__class__.__name__,
633
def has_same_location(self, other):
634
"""Returns a boolean indicating if this repository is at the same
635
location as another repository.
637
This might return False even when two repository objects are accessing
638
the same physical repository via different URLs.
640
if self.__class__ is not other.__class__:
642
return (self.control_files._transport.base ==
643
other.control_files._transport.base)
645
def is_in_write_group(self):
646
"""Return True if there is an open write group.
648
:seealso: start_write_group.
650
return self._write_group is not None
653
return self.control_files.is_locked()
655
def is_write_locked(self):
656
"""Return True if this object is write locked."""
657
return self.is_locked() and self.control_files._lock_mode == 'w'
659
def lock_write(self, token=None):
660
"""Lock this repository for writing.
662
This causes caching within the repository obejct to start accumlating
663
data during reads, and allows a 'write_group' to be obtained. Write
664
groups must be used for actual data insertion.
666
:param token: if this is already locked, then lock_write will fail
667
unless the token matches the existing lock.
668
:returns: a token if this instance supports tokens, otherwise None.
669
:raises TokenLockingNotSupported: when a token is given but this
670
instance doesn't support using token locks.
671
:raises MismatchedToken: if the specified token doesn't match the token
672
of the existing lock.
673
:seealso: start_write_group.
675
A token should be passed in if you know that you have locked the object
676
some other way, and need to synchronise this object's state with that
679
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
681
result = self.control_files.lock_write(token=token)
686
self.control_files.lock_read()
689
def get_physical_lock_status(self):
690
return self.control_files.get_physical_lock_status()
692
def leave_lock_in_place(self):
693
"""Tell this repository not to release the physical lock when this
696
If lock_write doesn't return a token, then this method is not supported.
698
self.control_files.leave_in_place()
700
def dont_leave_lock_in_place(self):
701
"""Tell this repository to release the physical lock when this
702
object is unlocked, even if it didn't originally acquire it.
704
If lock_write doesn't return a token, then this method is not supported.
706
self.control_files.dont_leave_in_place()
709
def gather_stats(self, revid=None, committers=None):
710
"""Gather statistics from a revision id.
712
:param revid: The revision id to gather statistics from, if None, then
713
no revision specific statistics are gathered.
714
:param committers: Optional parameter controlling whether to grab
715
a count of committers from the revision specific statistics.
716
:return: A dictionary of statistics. Currently this contains:
717
committers: The number of committers if requested.
718
firstrev: A tuple with timestamp, timezone for the penultimate left
719
most ancestor of revid, if revid is not the NULL_REVISION.
720
latestrev: A tuple with timestamp, timezone for revid, if revid is
721
not the NULL_REVISION.
722
revisions: The total revision count in the repository.
723
size: An estimate disk size of the repository in bytes.
726
if revid and committers:
727
result['committers'] = 0
728
if revid and revid != _mod_revision.NULL_REVISION:
730
all_committers = set()
731
revisions = self.get_ancestry(revid)
732
# pop the leading None
734
first_revision = None
736
# ignore the revisions in the middle - just grab first and last
737
revisions = revisions[0], revisions[-1]
738
for revision in self.get_revisions(revisions):
739
if not first_revision:
740
first_revision = revision
742
all_committers.add(revision.committer)
743
last_revision = revision
745
result['committers'] = len(all_committers)
746
result['firstrev'] = (first_revision.timestamp,
747
first_revision.timezone)
748
result['latestrev'] = (last_revision.timestamp,
749
last_revision.timezone)
751
# now gather global repository information
752
if self.bzrdir.root_transport.listable():
753
c, t = self._revision_store.total_size(self.get_transaction())
754
result['revisions'] = c
758
def get_data_stream(self, revision_ids):
759
raise NotImplementedError(self.get_data_stream)
761
def insert_data_stream(self, stream):
762
"""XXX What does this really do?
764
Is it a substitute for fetch?
765
Should it manage its own write group ?
767
for item_key, bytes in stream:
768
if item_key[0] == 'file':
769
(file_id,) = item_key[1:]
770
knit = self.weave_store.get_weave_or_empty(
771
file_id, self.get_transaction())
772
elif item_key == ('inventory',):
773
knit = self.get_inventory_weave()
774
elif item_key == ('revisions',):
775
knit = self._revision_store.get_revision_file(
776
self.get_transaction())
777
elif item_key == ('signatures',):
778
knit = self._revision_store.get_signature_file(
779
self.get_transaction())
781
raise RepositoryDataStreamError(
782
"Unrecognised data stream key '%s'" % (item_key,))
783
decoded_list = bencode.bdecode(bytes)
784
format = decoded_list.pop(0)
787
for version, options, parents, some_bytes in decoded_list:
788
data_list.append((version, options, len(some_bytes), parents))
789
knit_bytes += some_bytes
790
knit.insert_data_stream(
791
(format, data_list, StringIO(knit_bytes).read))
794
def missing_revision_ids(self, other, revision_id=None):
795
"""Return the revision ids that other has that this does not.
797
These are returned in topological order.
799
revision_id: only return revision ids included by revision_id.
801
return InterRepository.get(other, self).missing_revision_ids(revision_id)
805
"""Open the repository rooted at base.
807
For instance, if the repository is at URL/.bzr/repository,
808
Repository.open(URL) -> a Repository instance.
810
control = bzrdir.BzrDir.open(base)
811
return control.open_repository()
813
def copy_content_into(self, destination, revision_id=None):
814
"""Make a complete copy of the content in self into destination.
816
This is a destructive operation! Do not use it on existing
819
return InterRepository.get(self, destination).copy_content(revision_id)
821
def commit_write_group(self):
822
"""Commit the contents accrued within the current write group.
824
:seealso: start_write_group.
826
if self._write_group is not self.get_transaction():
827
# has an unlock or relock occured ?
828
raise errors.BzrError('mismatched lock context %r and '
830
(self.get_transaction(), self._write_group))
831
self._commit_write_group()
832
self._write_group = None
834
def _commit_write_group(self):
835
"""Template method for per-repository write group cleanup.
837
This is called before the write group is considered to be
838
finished and should ensure that all data handed to the repository
839
for writing during the write group is safely committed (to the
840
extent possible considering file system caching etc).
843
def fetch(self, source, revision_id=None, pb=None):
844
"""Fetch the content required to construct revision_id from source.
846
If revision_id is None all content is copied.
848
# fast path same-url fetch operations
849
if self.has_same_location(source):
850
# check that last_revision is in 'from' and then return a
852
if (revision_id is not None and
853
not _mod_revision.is_null(revision_id)):
854
self.get_revision(revision_id)
856
inter = InterRepository.get(source, self)
858
return inter.fetch(revision_id=revision_id, pb=pb)
859
except NotImplementedError:
860
raise errors.IncompatibleRepositories(source, self)
862
def create_bundle(self, target, base, fileobj, format=None):
863
return serializer.write_bundle(self, target, base, fileobj, format)
865
def get_commit_builder(self, branch, parents, config, timestamp=None,
866
timezone=None, committer=None, revprops=None,
868
"""Obtain a CommitBuilder for this repository.
870
:param branch: Branch to commit to.
871
:param parents: Revision ids of the parents of the new revision.
872
:param config: Configuration to use.
873
:param timestamp: Optional timestamp recorded for commit.
874
:param timezone: Optional timezone for timestamp.
875
:param committer: Optional committer to set for commit.
876
:param revprops: Optional dictionary of revision properties.
877
:param revision_id: Optional revision id.
879
result = self._commit_builder_class(self, parents, config,
880
timestamp, timezone, committer, revprops, revision_id)
881
self.start_write_group()
885
if (self.control_files._lock_count == 1 and
886
self.control_files._lock_mode == 'w'):
887
if self._write_group is not None:
888
raise errors.BzrError(
889
'Must end write groups before releasing write locks.')
890
self.control_files.unlock()
893
def clone(self, a_bzrdir, revision_id=None):
894
"""Clone this repository into a_bzrdir using the current format.
896
Currently no check is made that the format of this repository and
897
the bzrdir format are compatible. FIXME RBC 20060201.
899
:return: The newly created destination repository.
901
# TODO: deprecate after 0.16; cloning this with all its settings is
902
# probably not very useful -- mbp 20070423
903
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
904
self.copy_content_into(dest_repo, revision_id)
907
def start_write_group(self):
908
"""Start a write group in the repository.
910
Write groups are used by repositories which do not have a 1:1 mapping
911
between file ids and backend store to manage the insertion of data from
912
both fetch and commit operations.
914
A write lock is required around the start_write_group/commit_write_group
915
for the support of lock-requiring repository formats.
917
One can only insert data into a repository inside a write group.
921
if not self.is_write_locked():
922
raise errors.NotWriteLocked(self)
923
if self._write_group:
924
raise errors.BzrError('already in a write group')
925
self._start_write_group()
926
# so we can detect unlock/relock - the write group is now entered.
927
self._write_group = self.get_transaction()
929
def _start_write_group(self):
930
"""Template method for per-repository write group startup.
932
This is called before the write group is considered to be
937
def sprout(self, to_bzrdir, revision_id=None):
938
"""Create a descendent repository for new development.
940
Unlike clone, this does not copy the settings of the repository.
942
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
943
dest_repo.fetch(self, revision_id=revision_id)
946
def _create_sprouting_repo(self, a_bzrdir, shared):
947
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
948
# use target default format.
949
dest_repo = a_bzrdir.create_repository()
951
# Most control formats need the repository to be specifically
952
# created, but on some old all-in-one formats it's not needed
954
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
955
except errors.UninitializableFormat:
956
dest_repo = a_bzrdir.open_repository()
960
def has_revision(self, revision_id):
961
"""True if this repository has a copy of the revision."""
962
if 'evil' in debug.debug_flags:
963
mutter_callsite(3, "has_revision is a LBYL symptom.")
964
return self._revision_store.has_revision_id(revision_id,
965
self.get_transaction())
968
def get_revision(self, revision_id):
969
"""Return the Revision object for a named revision."""
970
return self.get_revisions([revision_id])[0]
973
def get_revision_reconcile(self, revision_id):
974
"""'reconcile' helper routine that allows access to a revision always.
976
This variant of get_revision does not cross check the weave graph
977
against the revision one as get_revision does: but it should only
978
be used by reconcile, or reconcile-alike commands that are correcting
979
or testing the revision graph.
981
return self._get_revisions([revision_id])[0]
984
def get_revisions(self, revision_ids):
985
"""Get many revisions at once."""
986
return self._get_revisions(revision_ids)
989
def _get_revisions(self, revision_ids):
990
"""Core work logic to get many revisions without sanity checks."""
991
for rev_id in revision_ids:
992
if not rev_id or not isinstance(rev_id, basestring):
993
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
994
revs = self._revision_store.get_revisions(revision_ids,
995
self.get_transaction())
997
assert not isinstance(rev.revision_id, unicode)
998
for parent_id in rev.parent_ids:
999
assert not isinstance(parent_id, unicode)
1003
def get_revision_xml(self, revision_id):
1004
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1005
# would have already do it.
1006
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1007
rev = self.get_revision(revision_id)
1008
rev_tmp = StringIO()
1009
# the current serializer..
1010
self._revision_store._serializer.write_revision(rev, rev_tmp)
1012
return rev_tmp.getvalue()
1015
def get_deltas_for_revisions(self, revisions):
1016
"""Produce a generator of revision deltas.
1018
Note that the input is a sequence of REVISIONS, not revision_ids.
1019
Trees will be held in memory until the generator exits.
1020
Each delta is relative to the revision's lefthand predecessor.
1022
required_trees = set()
1023
for revision in revisions:
1024
required_trees.add(revision.revision_id)
1025
required_trees.update(revision.parent_ids[:1])
1026
trees = dict((t.get_revision_id(), t) for
1027
t in self.revision_trees(required_trees))
1028
for revision in revisions:
1029
if not revision.parent_ids:
1030
old_tree = self.revision_tree(None)
1032
old_tree = trees[revision.parent_ids[0]]
1033
yield trees[revision.revision_id].changes_from(old_tree)
1036
def get_revision_delta(self, revision_id):
1037
"""Return the delta for one revision.
1039
The delta is relative to the left-hand predecessor of the
1042
r = self.get_revision(revision_id)
1043
return list(self.get_deltas_for_revisions([r]))[0]
1046
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1047
signature = gpg_strategy.sign(plaintext)
1048
self._revision_store.add_revision_signature_text(revision_id,
1050
self.get_transaction())
1052
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1054
"""Helper routine for fileids_altered_by_revision_ids.
1056
This performs the translation of xml lines to revision ids.
1058
:param line_iterator: An iterator of lines
1059
:param revision_ids: The revision ids to filter for.
1060
:return: a dictionary mapping altered file-ids to an iterable of
1061
revision_ids. Each altered file-ids has the exact revision_ids that
1062
altered it listed explicitly.
1066
# this code needs to read every new line in every inventory for the
1067
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1068
# not present in one of those inventories is unnecessary but not
1069
# harmful because we are filtering by the revision id marker in the
1070
# inventory lines : we only select file ids altered in one of those
1071
# revisions. We don't need to see all lines in the inventory because
1072
# only those added in an inventory in rev X can contain a revision=X
1074
unescape_revid_cache = {}
1075
unescape_fileid_cache = {}
1077
# jam 20061218 In a big fetch, this handles hundreds of thousands
1078
# of lines, so it has had a lot of inlining and optimizing done.
1079
# Sorry that it is a little bit messy.
1080
# Move several functions to be local variables, since this is a long
1082
search = self._file_ids_altered_regex.search
1083
unescape = _unescape_xml
1084
setdefault = result.setdefault
1085
for line in line_iterator:
1086
match = search(line)
1089
# One call to match.group() returning multiple items is quite a
1090
# bit faster than 2 calls to match.group() each returning 1
1091
file_id, revision_id = match.group('file_id', 'revision_id')
1093
# Inlining the cache lookups helps a lot when you make 170,000
1094
# lines and 350k ids, versus 8.4 unique ids.
1095
# Using a cache helps in 2 ways:
1096
# 1) Avoids unnecessary decoding calls
1097
# 2) Re-uses cached strings, which helps in future set and
1099
# (2) is enough that removing encoding entirely along with
1100
# the cache (so we are using plain strings) results in no
1101
# performance improvement.
1103
revision_id = unescape_revid_cache[revision_id]
1105
unescaped = unescape(revision_id)
1106
unescape_revid_cache[revision_id] = unescaped
1107
revision_id = unescaped
1109
if revision_id in revision_ids:
1111
file_id = unescape_fileid_cache[file_id]
1113
unescaped = unescape(file_id)
1114
unescape_fileid_cache[file_id] = unescaped
1116
setdefault(file_id, set()).add(revision_id)
1119
def fileids_altered_by_revision_ids(self, revision_ids):
1120
"""Find the file ids and versions affected by revisions.
1122
:param revisions: an iterable containing revision ids.
1123
:return: a dictionary mapping altered file-ids to an iterable of
1124
revision_ids. Each altered file-ids has the exact revision_ids that
1125
altered it listed explicitly.
1127
assert self._serializer.support_altered_by_hack, \
1128
("fileids_altered_by_revision_ids only supported for branches "
1129
"which store inventory as unnested xml, not on %r" % self)
1130
selected_revision_ids = set(revision_ids)
1131
w = self.get_inventory_weave()
1132
pb = ui.ui_factory.nested_progress_bar()
1134
return self._find_file_ids_from_xml_inventory_lines(
1135
w.iter_lines_added_or_present_in_versions(
1136
selected_revision_ids, pb=pb),
1137
selected_revision_ids)
1141
def iter_files_bytes(self, desired_files):
1142
"""Iterate through file versions.
1144
Files will not necessarily be returned in the order they occur in
1145
desired_files. No specific order is guaranteed.
1147
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1148
value supplied by the caller as part of desired_files. It should
1149
uniquely identify the file version in the caller's context. (Examples:
1150
an index number or a TreeTransform trans_id.)
1152
bytes_iterator is an iterable of bytestrings for the file. The
1153
kind of iterable and length of the bytestrings are unspecified, but for
1154
this implementation, it is a list of lines produced by
1155
VersionedFile.get_lines().
1157
:param desired_files: a list of (file_id, revision_id, identifier)
1160
transaction = self.get_transaction()
1161
for file_id, revision_id, callable_data in desired_files:
1163
weave = self.weave_store.get_weave(file_id, transaction)
1164
except errors.NoSuchFile:
1165
raise errors.NoSuchIdInRepository(self, file_id)
1166
yield callable_data, weave.get_lines(revision_id)
1168
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1169
"""Get an iterable listing the keys of all the data introduced by a set
1172
The keys will be ordered so that the corresponding items can be safely
1173
fetched and inserted in that order.
1175
:returns: An iterable producing tuples of (knit-kind, file-id,
1176
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1177
'revisions'. file-id is None unless knit-kind is 'file'.
1179
# XXX: it's a bit weird to control the inventory weave caching in this
1180
# generator. Ideally the caching would be done in fetch.py I think. Or
1181
# maybe this generator should explicitly have the contract that it
1182
# should not be iterated until the previously yielded item has been
1185
inv_w = self.get_inventory_weave()
1186
inv_w.enable_cache()
1188
# file ids that changed
1189
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1191
num_file_ids = len(file_ids)
1192
for file_id, altered_versions in file_ids.iteritems():
1193
if _files_pb is not None:
1194
_files_pb.update("fetch texts", count, num_file_ids)
1196
yield ("file", file_id, altered_versions)
1197
# We're done with the files_pb. Note that it finished by the caller,
1198
# just as it was created by the caller.
1202
yield ("inventory", None, revision_ids)
1206
revisions_with_signatures = set()
1207
for rev_id in revision_ids:
1209
self.get_signature_text(rev_id)
1210
except errors.NoSuchRevision:
1214
revisions_with_signatures.add(rev_id)
1216
yield ("signatures", None, revisions_with_signatures)
1219
yield ("revisions", None, revision_ids)
1222
def get_inventory_weave(self):
1223
return self.control_weaves.get_weave('inventory',
1224
self.get_transaction())
1227
def get_inventory(self, revision_id):
1228
"""Get Inventory object by hash."""
1229
return self.deserialise_inventory(
1230
revision_id, self.get_inventory_xml(revision_id))
1232
def deserialise_inventory(self, revision_id, xml):
1233
"""Transform the xml into an inventory object.
1235
:param revision_id: The expected revision id of the inventory.
1236
:param xml: A serialised inventory.
1238
return self._serializer.read_inventory_from_string(xml, revision_id)
1240
def serialise_inventory(self, inv):
1241
return self._serializer.write_inventory_to_string(inv)
1243
def _serialise_inventory_to_lines(self, inv):
1244
return self._serializer.write_inventory_to_lines(inv)
1246
def get_serializer_format(self):
1247
return self._serializer.format_num
1250
def get_inventory_xml(self, revision_id):
1251
"""Get inventory XML as a file object."""
1253
assert isinstance(revision_id, str), type(revision_id)
1254
iw = self.get_inventory_weave()
1255
return iw.get_text(revision_id)
1257
raise errors.HistoryMissing(self, 'inventory', revision_id)
1260
def get_inventory_sha1(self, revision_id):
1261
"""Return the sha1 hash of the inventory entry
1263
return self.get_revision(revision_id).inventory_sha1
1266
def get_revision_graph(self, revision_id=None):
1267
"""Return a dictionary containing the revision graph.
1269
NB: This method should not be used as it accesses the entire graph all
1270
at once, which is much more data than most operations should require.
1272
:param revision_id: The revision_id to get a graph from. If None, then
1273
the entire revision graph is returned. This is a deprecated mode of
1274
operation and will be removed in the future.
1275
:return: a dictionary of revision_id->revision_parents_list.
1277
raise NotImplementedError(self.get_revision_graph)
1280
def get_revision_graph_with_ghosts(self, revision_ids=None):
1281
"""Return a graph of the revisions with ghosts marked as applicable.
1283
:param revision_ids: an iterable of revisions to graph or None for all.
1284
:return: a Graph object with the graph reachable from revision_ids.
1286
if 'evil' in debug.debug_flags:
1288
"get_revision_graph_with_ghosts scales with size of history.")
1289
result = deprecated_graph.Graph()
1290
if not revision_ids:
1291
pending = set(self.all_revision_ids())
1294
pending = set(revision_ids)
1295
# special case NULL_REVISION
1296
if _mod_revision.NULL_REVISION in pending:
1297
pending.remove(_mod_revision.NULL_REVISION)
1298
required = set(pending)
1301
revision_id = pending.pop()
1303
rev = self.get_revision(revision_id)
1304
except errors.NoSuchRevision:
1305
if revision_id in required:
1308
result.add_ghost(revision_id)
1310
for parent_id in rev.parent_ids:
1311
# is this queued or done ?
1312
if (parent_id not in pending and
1313
parent_id not in done):
1315
pending.add(parent_id)
1316
result.add_node(revision_id, rev.parent_ids)
1317
done.add(revision_id)
1320
def _get_history_vf(self):
1321
"""Get a versionedfile whose history graph reflects all revisions.
1323
For weave repositories, this is the inventory weave.
1325
return self.get_inventory_weave()
1327
def iter_reverse_revision_history(self, revision_id):
1328
"""Iterate backwards through revision ids in the lefthand history
1330
:param revision_id: The revision id to start with. All its lefthand
1331
ancestors will be traversed.
1333
if revision_id in (None, _mod_revision.NULL_REVISION):
1335
next_id = revision_id
1336
versionedfile = self._get_history_vf()
1339
parents = versionedfile.get_parents(next_id)
1340
if len(parents) == 0:
1343
next_id = parents[0]
1346
def get_revision_inventory(self, revision_id):
1347
"""Return inventory of a past revision."""
1348
# TODO: Unify this with get_inventory()
1349
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1350
# must be the same as its revision, so this is trivial.
1351
if revision_id is None:
1352
# This does not make sense: if there is no revision,
1353
# then it is the current tree inventory surely ?!
1354
# and thus get_root_id() is something that looks at the last
1355
# commit on the branch, and the get_root_id is an inventory check.
1356
raise NotImplementedError
1357
# return Inventory(self.get_root_id())
1359
return self.get_inventory(revision_id)
1362
def is_shared(self):
1363
"""Return True if this repository is flagged as a shared repository."""
1364
raise NotImplementedError(self.is_shared)
1367
def reconcile(self, other=None, thorough=False):
1368
"""Reconcile this repository."""
1369
from bzrlib.reconcile import RepoReconciler
1370
reconciler = RepoReconciler(self, thorough=thorough)
1371
reconciler.reconcile()
1374
def _refresh_data(self):
1375
"""Helper called from lock_* to ensure coherency with disk.
1377
The default implementation does nothing; it is however possible
1378
for repositories to maintain loaded indices across multiple locks
1379
by checking inside their implementation of this method to see
1380
whether their indices are still valid. This depends of course on
1381
the disk format being validatable in this manner.
1385
def revision_tree(self, revision_id):
1386
"""Return Tree for a revision on this branch.
1388
`revision_id` may be None for the empty tree revision.
1390
# TODO: refactor this to use an existing revision object
1391
# so we don't need to read it in twice.
1392
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1393
return RevisionTree(self, Inventory(root_id=None),
1394
_mod_revision.NULL_REVISION)
1396
inv = self.get_revision_inventory(revision_id)
1397
return RevisionTree(self, inv, revision_id)
1400
def revision_trees(self, revision_ids):
1401
"""Return Tree for a revision on this branch.
1403
`revision_id` may not be None or 'null:'"""
1404
assert None not in revision_ids
1405
assert _mod_revision.NULL_REVISION not in revision_ids
1406
texts = self.get_inventory_weave().get_texts(revision_ids)
1407
for text, revision_id in zip(texts, revision_ids):
1408
inv = self.deserialise_inventory(revision_id, text)
1409
yield RevisionTree(self, inv, revision_id)
1412
def get_ancestry(self, revision_id, topo_sorted=True):
1413
"""Return a list of revision-ids integrated by a revision.
1415
The first element of the list is always None, indicating the origin
1416
revision. This might change when we have history horizons, or
1417
perhaps we should have a new API.
1419
This is topologically sorted.
1421
if _mod_revision.is_null(revision_id):
1423
if not self.has_revision(revision_id):
1424
raise errors.NoSuchRevision(self, revision_id)
1425
w = self.get_inventory_weave()
1426
candidates = w.get_ancestry(revision_id, topo_sorted)
1427
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1430
"""Compress the data within the repository.
1432
This operation only makes sense for some repository types. For other
1433
types it should be a no-op that just returns.
1435
This stub method does not require a lock, but subclasses should use
1436
@needs_write_lock as this is a long running call its reasonable to
1437
implicitly lock for the user.
1441
def print_file(self, file, revision_id):
1442
"""Print `file` to stdout.
1444
FIXME RBC 20060125 as John Meinel points out this is a bad api
1445
- it writes to stdout, it assumes that that is valid etc. Fix
1446
by creating a new more flexible convenience function.
1448
tree = self.revision_tree(revision_id)
1449
# use inventory as it was in that revision
1450
file_id = tree.inventory.path2id(file)
1452
# TODO: jam 20060427 Write a test for this code path
1453
# it had a bug in it, and was raising the wrong
1455
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1456
tree.print_file(file_id)
1458
def get_transaction(self):
1459
return self.control_files.get_transaction()
1461
def revision_parents(self, revision_id):
1462
return self.get_inventory_weave().parent_names(revision_id)
1464
def get_parents(self, revision_ids):
1465
"""See StackedParentsProvider.get_parents"""
1467
for revision_id in revision_ids:
1468
if revision_id == _mod_revision.NULL_REVISION:
1472
parents = self.get_revision(revision_id).parent_ids
1473
except errors.NoSuchRevision:
1476
if len(parents) == 0:
1477
parents = [_mod_revision.NULL_REVISION]
1478
parents_list.append(parents)
1481
def _make_parents_provider(self):
1484
def get_graph(self, other_repository=None):
1485
"""Return the graph walker for this repository format"""
1486
parents_provider = self._make_parents_provider()
1487
if (other_repository is not None and
1488
other_repository.bzrdir.transport.base !=
1489
self.bzrdir.transport.base):
1490
parents_provider = graph._StackedParentsProvider(
1491
[parents_provider, other_repository._make_parents_provider()])
1492
return graph.Graph(parents_provider)
1494
def get_versioned_file_checker(self, revisions, revision_versions_cache):
1495
return VersionedFileChecker(revisions, revision_versions_cache, self)
1498
def set_make_working_trees(self, new_value):
1499
"""Set the policy flag for making working trees when creating branches.
1501
This only applies to branches that use this repository.
1503
The default is 'True'.
1504
:param new_value: True to restore the default, False to disable making
1507
raise NotImplementedError(self.set_make_working_trees)
1509
def make_working_trees(self):
1510
"""Returns the policy for making working trees on new branches."""
1511
raise NotImplementedError(self.make_working_trees)
1514
def sign_revision(self, revision_id, gpg_strategy):
1515
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1516
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1519
def has_signature_for_revision_id(self, revision_id):
1520
"""Query for a revision signature for revision_id in the repository."""
1521
return self._revision_store.has_signature(revision_id,
1522
self.get_transaction())
1525
def get_signature_text(self, revision_id):
1526
"""Return the text for a signature."""
1527
return self._revision_store.get_signature_text(revision_id,
1528
self.get_transaction())
1531
def check(self, revision_ids=None):
1532
"""Check consistency of all history of given revision_ids.
1534
Different repository implementations should override _check().
1536
:param revision_ids: A non-empty list of revision_ids whose ancestry
1537
will be checked. Typically the last revision_id of a branch.
1539
return self._check(revision_ids)
1541
def _check(self, revision_ids):
1542
result = check.Check(self)
1546
def _warn_if_deprecated(self):
1547
global _deprecation_warning_done
1548
if _deprecation_warning_done:
1550
_deprecation_warning_done = True
1551
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1552
% (self._format, self.bzrdir.transport.base))
1554
def supports_rich_root(self):
1555
return self._format.rich_root_data
1557
def _check_ascii_revisionid(self, revision_id, method):
1558
"""Private helper for ascii-only repositories."""
1559
# weave repositories refuse to store revisionids that are non-ascii.
1560
if revision_id is not None:
1561
# weaves require ascii revision ids.
1562
if isinstance(revision_id, unicode):
1564
revision_id.encode('ascii')
1565
except UnicodeEncodeError:
1566
raise errors.NonAsciiRevisionId(method, self)
1569
revision_id.decode('ascii')
1570
except UnicodeDecodeError:
1571
raise errors.NonAsciiRevisionId(method, self)
1573
def revision_graph_can_have_wrong_parents(self):
1574
"""Is it possible for this repository to have a revision graph with
1577
If True, then this repository must also implement
1578
_find_inconsistent_revision_parents so that check and reconcile can
1579
check for inconsistencies before proceeding with other checks that may
1580
depend on the revision index being consistent.
1582
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1584
# remove these delegates a while after bzr 0.15
1585
def __make_delegated(name, from_module):
1586
def _deprecated_repository_forwarder():
1587
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1588
% (name, from_module),
1591
m = __import__(from_module, globals(), locals(), [name])
1593
return getattr(m, name)
1594
except AttributeError:
1595
raise AttributeError('module %s has no name %s'
1597
globals()[name] = _deprecated_repository_forwarder
1600
'AllInOneRepository',
1601
'WeaveMetaDirRepository',
1602
'PreSplitOutRepositoryFormat',
1603
'RepositoryFormat4',
1604
'RepositoryFormat5',
1605
'RepositoryFormat6',
1606
'RepositoryFormat7',
1608
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1612
'RepositoryFormatKnit',
1613
'RepositoryFormatKnit1',
1615
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1618
def install_revision(repository, rev, revision_tree):
1619
"""Install all revision data into a repository."""
1620
repository.start_write_group()
1622
_install_revision(repository, rev, revision_tree)
1624
repository.abort_write_group()
1627
repository.commit_write_group()
1630
def _install_revision(repository, rev, revision_tree):
1631
"""Install all revision data into a repository."""
1632
present_parents = []
1634
for p_id in rev.parent_ids:
1635
if repository.has_revision(p_id):
1636
present_parents.append(p_id)
1637
parent_trees[p_id] = repository.revision_tree(p_id)
1639
parent_trees[p_id] = repository.revision_tree(None)
1641
inv = revision_tree.inventory
1642
entries = inv.iter_entries()
1643
# backwards compatibility hack: skip the root id.
1644
if not repository.supports_rich_root():
1645
path, root = entries.next()
1646
if root.revision != rev.revision_id:
1647
raise errors.IncompatibleRevision(repr(repository))
1648
# Add the texts that are not already present
1649
for path, ie in entries:
1650
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1651
repository.get_transaction())
1652
if ie.revision not in w:
1654
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1655
# with InventoryEntry.find_previous_heads(). if it is, then there
1656
# is a latent bug here where the parents may have ancestors of each
1658
for revision, tree in parent_trees.iteritems():
1659
if ie.file_id not in tree:
1661
parent_id = tree.inventory[ie.file_id].revision
1662
if parent_id in text_parents:
1664
text_parents.append(parent_id)
1666
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1667
repository.get_transaction())
1668
lines = revision_tree.get_file(ie.file_id).readlines()
1669
vfile.add_lines(rev.revision_id, text_parents, lines)
1671
# install the inventory
1672
repository.add_inventory(rev.revision_id, inv, present_parents)
1673
except errors.RevisionAlreadyPresent:
1675
repository.add_revision(rev.revision_id, rev, inv)
1678
class MetaDirRepository(Repository):
1679
"""Repositories in the new meta-dir layout."""
1681
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1682
super(MetaDirRepository, self).__init__(_format,
1688
dir_mode = self.control_files._dir_mode
1689
file_mode = self.control_files._file_mode
1692
def is_shared(self):
1693
"""Return True if this repository is flagged as a shared repository."""
1694
return self.control_files._transport.has('shared-storage')
1697
def set_make_working_trees(self, new_value):
1698
"""Set the policy flag for making working trees when creating branches.
1700
This only applies to branches that use this repository.
1702
The default is 'True'.
1703
:param new_value: True to restore the default, False to disable making
1708
self.control_files._transport.delete('no-working-trees')
1709
except errors.NoSuchFile:
1712
self.control_files.put_utf8('no-working-trees', '')
1714
def make_working_trees(self):
1715
"""Returns the policy for making working trees on new branches."""
1716
return not self.control_files._transport.has('no-working-trees')
1719
class RepositoryFormatRegistry(registry.Registry):
1720
"""Registry of RepositoryFormats."""
1722
def get(self, format_string):
1723
r = registry.Registry.get(self, format_string)
1729
format_registry = RepositoryFormatRegistry()
1730
"""Registry of formats, indexed by their identifying format string.
1732
This can contain either format instances themselves, or classes/factories that
1733
can be called to obtain one.
1737
#####################################################################
1738
# Repository Formats
1740
class RepositoryFormat(object):
1741
"""A repository format.
1743
Formats provide three things:
1744
* An initialization routine to construct repository data on disk.
1745
* a format string which is used when the BzrDir supports versioned
1747
* an open routine which returns a Repository instance.
1749
There is one and only one Format subclass for each on-disk format. But
1750
there can be one Repository subclass that is used for several different
1751
formats. The _format attribute on a Repository instance can be used to
1752
determine the disk format.
1754
Formats are placed in an dict by their format string for reference
1755
during opening. These should be subclasses of RepositoryFormat
1758
Once a format is deprecated, just deprecate the initialize and open
1759
methods on the format class. Do not deprecate the object, as the
1760
object will be created every system load.
1762
Common instance attributes:
1763
_matchingbzrdir - the bzrdir format that the repository format was
1764
originally written to work with. This can be used if manually
1765
constructing a bzrdir and repository, or more commonly for test suite
1770
return "<%s>" % self.__class__.__name__
1772
def __eq__(self, other):
1773
# format objects are generally stateless
1774
return isinstance(other, self.__class__)
1776
def __ne__(self, other):
1777
return not self == other
1780
def find_format(klass, a_bzrdir):
1781
"""Return the format for the repository object in a_bzrdir.
1783
This is used by bzr native formats that have a "format" file in
1784
the repository. Other methods may be used by different types of
1788
transport = a_bzrdir.get_repository_transport(None)
1789
format_string = transport.get("format").read()
1790
return format_registry.get(format_string)
1791
except errors.NoSuchFile:
1792
raise errors.NoRepositoryPresent(a_bzrdir)
1794
raise errors.UnknownFormatError(format=format_string)
1797
def register_format(klass, format):
1798
format_registry.register(format.get_format_string(), format)
1801
def unregister_format(klass, format):
1802
format_registry.remove(format.get_format_string())
1805
def get_default_format(klass):
1806
"""Return the current default format."""
1807
from bzrlib import bzrdir
1808
return bzrdir.format_registry.make_bzrdir('default').repository_format
1810
def _get_control_store(self, repo_transport, control_files):
1811
"""Return the control store for this repository."""
1812
raise NotImplementedError(self._get_control_store)
1814
def get_format_string(self):
1815
"""Return the ASCII format string that identifies this format.
1817
Note that in pre format ?? repositories the format string is
1818
not permitted nor written to disk.
1820
raise NotImplementedError(self.get_format_string)
1822
def get_format_description(self):
1823
"""Return the short description for this format."""
1824
raise NotImplementedError(self.get_format_description)
1826
def _get_revision_store(self, repo_transport, control_files):
1827
"""Return the revision store object for this a_bzrdir."""
1828
raise NotImplementedError(self._get_revision_store)
1830
def _get_text_rev_store(self,
1837
"""Common logic for getting a revision store for a repository.
1839
see self._get_revision_store for the subclass-overridable method to
1840
get the store for a repository.
1842
from bzrlib.store.revision.text import TextRevisionStore
1843
dir_mode = control_files._dir_mode
1844
file_mode = control_files._file_mode
1845
text_store = TextStore(transport.clone(name),
1847
compressed=compressed,
1849
file_mode=file_mode)
1850
_revision_store = TextRevisionStore(text_store, serializer)
1851
return _revision_store
1853
# TODO: this shouldn't be in the base class, it's specific to things that
1854
# use weaves or knits -- mbp 20070207
1855
def _get_versioned_file_store(self,
1860
versionedfile_class=None,
1861
versionedfile_kwargs={},
1863
if versionedfile_class is None:
1864
versionedfile_class = self._versionedfile_class
1865
weave_transport = control_files._transport.clone(name)
1866
dir_mode = control_files._dir_mode
1867
file_mode = control_files._file_mode
1868
return VersionedFileStore(weave_transport, prefixed=prefixed,
1870
file_mode=file_mode,
1871
versionedfile_class=versionedfile_class,
1872
versionedfile_kwargs=versionedfile_kwargs,
1875
def initialize(self, a_bzrdir, shared=False):
1876
"""Initialize a repository of this format in a_bzrdir.
1878
:param a_bzrdir: The bzrdir to put the new repository in it.
1879
:param shared: The repository should be initialized as a sharable one.
1880
:returns: The new repository object.
1882
This may raise UninitializableFormat if shared repository are not
1883
compatible the a_bzrdir.
1885
raise NotImplementedError(self.initialize)
1887
def is_supported(self):
1888
"""Is this format supported?
1890
Supported formats must be initializable and openable.
1891
Unsupported formats may not support initialization or committing or
1892
some other features depending on the reason for not being supported.
1896
def check_conversion_target(self, target_format):
1897
raise NotImplementedError(self.check_conversion_target)
1899
def open(self, a_bzrdir, _found=False):
1900
"""Return an instance of this format for the bzrdir a_bzrdir.
1902
_found is a private parameter, do not use it.
1904
raise NotImplementedError(self.open)
1907
class MetaDirRepositoryFormat(RepositoryFormat):
1908
"""Common base class for the new repositories using the metadir layout."""
1910
rich_root_data = False
1911
supports_tree_reference = False
1912
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
1915
super(MetaDirRepositoryFormat, self).__init__()
1917
def _create_control_files(self, a_bzrdir):
1918
"""Create the required files and the initial control_files object."""
1919
# FIXME: RBC 20060125 don't peek under the covers
1920
# NB: no need to escape relative paths that are url safe.
1921
repository_transport = a_bzrdir.get_repository_transport(self)
1922
control_files = lockable_files.LockableFiles(repository_transport,
1923
'lock', lockdir.LockDir)
1924
control_files.create_lock()
1925
return control_files
1927
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
1928
"""Upload the initial blank content."""
1929
control_files = self._create_control_files(a_bzrdir)
1930
control_files.lock_write()
1932
control_files._transport.mkdir_multi(dirs,
1933
mode=control_files._dir_mode)
1934
for file, content in files:
1935
control_files.put(file, content)
1936
for file, content in utf8_files:
1937
control_files.put_utf8(file, content)
1939
control_files.put_utf8('shared-storage', '')
1941
control_files.unlock()
1944
# formats which have no format string are not discoverable
1945
# and not independently creatable, so are not registered. They're
1946
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1947
# needed, it's constructed directly by the BzrDir. Non-native formats where
1948
# the repository is not separately opened are similar.
1950
format_registry.register_lazy(
1951
'Bazaar-NG Repository format 7',
1952
'bzrlib.repofmt.weaverepo',
1956
# KEEP in sync with bzrdir.format_registry default, which controls the overall
1957
# default control directory format
1958
format_registry.register_lazy(
1959
'Bazaar-NG Knit Repository Format 1',
1960
'bzrlib.repofmt.knitrepo',
1961
'RepositoryFormatKnit1',
1963
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
1965
format_registry.register_lazy(
1966
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
1967
'bzrlib.repofmt.knitrepo',
1968
'RepositoryFormatKnit3',
1971
# Experimental formats. These make no guarantee about data stability.
1972
# There is one format for pre-subtrees, and one for post-subtrees to
1973
# allow ease of testing.
1974
format_registry.register_lazy(
1975
'Bazaar Experimental no-subtrees\n',
1976
'bzrlib.repofmt.pack_repo',
1977
'RepositoryFormatKnitPack1',
1979
format_registry.register_lazy(
1980
'Bazaar Experimental subtrees\n',
1981
'bzrlib.repofmt.pack_repo',
1982
'RepositoryFormatKnitPack3',
1986
class InterRepository(InterObject):
1987
"""This class represents operations taking place between two repositories.
1989
Its instances have methods like copy_content and fetch, and contain
1990
references to the source and target repositories these operations can be
1993
Often we will provide convenience methods on 'repository' which carry out
1994
operations with another repository - they will always forward to
1995
InterRepository.get(other).method_name(parameters).
1999
"""The available optimised InterRepository types."""
2001
def copy_content(self, revision_id=None):
2002
raise NotImplementedError(self.copy_content)
2004
def fetch(self, revision_id=None, pb=None):
2005
"""Fetch the content required to construct revision_id.
2007
The content is copied from self.source to self.target.
2009
:param revision_id: if None all content is copied, if NULL_REVISION no
2011
:param pb: optional progress bar to use for progress reports. If not
2012
provided a default one will be created.
2014
Returns the copied revision count and the failed revisions in a tuple:
2017
raise NotImplementedError(self.fetch)
2020
def missing_revision_ids(self, revision_id=None):
2021
"""Return the revision ids that source has that target does not.
2023
These are returned in topological order.
2025
:param revision_id: only return revision ids included by this
2028
# generic, possibly worst case, slow code path.
2029
target_ids = set(self.target.all_revision_ids())
2030
if revision_id is not None:
2031
source_ids = self.source.get_ancestry(revision_id)
2032
assert source_ids[0] is None
2035
source_ids = self.source.all_revision_ids()
2036
result_set = set(source_ids).difference(target_ids)
2037
# this may look like a no-op: its not. It preserves the ordering
2038
# other_ids had while only returning the members from other_ids
2039
# that we've decided we need.
2040
return [rev_id for rev_id in source_ids if rev_id in result_set]
2043
def _same_model(source, target):
2044
"""True if source and target have the same data representation."""
2045
if source.supports_rich_root() != target.supports_rich_root():
2047
if source._serializer != target._serializer:
2052
class InterSameDataRepository(InterRepository):
2053
"""Code for converting between repositories that represent the same data.
2055
Data format and model must match for this to work.
2059
def _get_repo_format_to_test(self):
2060
"""Repository format for testing with.
2062
InterSameData can pull from subtree to subtree and from non-subtree to
2063
non-subtree, so we test this with the richest repository format.
2065
from bzrlib.repofmt import knitrepo
2066
return knitrepo.RepositoryFormatKnit3()
2069
def is_compatible(source, target):
2070
return InterRepository._same_model(source, target)
2073
def copy_content(self, revision_id=None):
2074
"""Make a complete copy of the content in self into destination.
2076
This copies both the repository's revision data, and configuration information
2077
such as the make_working_trees setting.
2079
This is a destructive operation! Do not use it on existing
2082
:param revision_id: Only copy the content needed to construct
2083
revision_id and its parents.
2086
self.target.set_make_working_trees(self.source.make_working_trees())
2087
except NotImplementedError:
2089
# but don't bother fetching if we have the needed data now.
2090
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2091
self.target.has_revision(revision_id)):
2093
self.target.fetch(self.source, revision_id=revision_id)
2096
def fetch(self, revision_id=None, pb=None):
2097
"""See InterRepository.fetch()."""
2098
from bzrlib.fetch import GenericRepoFetcher
2099
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2100
self.source, self.source._format, self.target,
2101
self.target._format)
2102
f = GenericRepoFetcher(to_repository=self.target,
2103
from_repository=self.source,
2104
last_revision=revision_id,
2106
return f.count_copied, f.failed_revisions
2109
class InterWeaveRepo(InterSameDataRepository):
2110
"""Optimised code paths between Weave based repositories.
2112
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2113
implemented lazy inter-object optimisation.
2117
def _get_repo_format_to_test(self):
2118
from bzrlib.repofmt import weaverepo
2119
return weaverepo.RepositoryFormat7()
2122
def is_compatible(source, target):
2123
"""Be compatible with known Weave formats.
2125
We don't test for the stores being of specific types because that
2126
could lead to confusing results, and there is no need to be
2129
from bzrlib.repofmt.weaverepo import (
2135
return (isinstance(source._format, (RepositoryFormat5,
2137
RepositoryFormat7)) and
2138
isinstance(target._format, (RepositoryFormat5,
2140
RepositoryFormat7)))
2141
except AttributeError:
2145
def copy_content(self, revision_id=None):
2146
"""See InterRepository.copy_content()."""
2147
# weave specific optimised path:
2149
self.target.set_make_working_trees(self.source.make_working_trees())
2150
except NotImplementedError:
2152
# FIXME do not peek!
2153
if self.source.control_files._transport.listable():
2154
pb = ui.ui_factory.nested_progress_bar()
2156
self.target.weave_store.copy_all_ids(
2157
self.source.weave_store,
2159
from_transaction=self.source.get_transaction(),
2160
to_transaction=self.target.get_transaction())
2161
pb.update('copying inventory', 0, 1)
2162
self.target.control_weaves.copy_multi(
2163
self.source.control_weaves, ['inventory'],
2164
from_transaction=self.source.get_transaction(),
2165
to_transaction=self.target.get_transaction())
2166
self.target._revision_store.text_store.copy_all_ids(
2167
self.source._revision_store.text_store,
2172
self.target.fetch(self.source, revision_id=revision_id)
2175
def fetch(self, revision_id=None, pb=None):
2176
"""See InterRepository.fetch()."""
2177
from bzrlib.fetch import GenericRepoFetcher
2178
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2179
self.source, self.source._format, self.target, self.target._format)
2180
f = GenericRepoFetcher(to_repository=self.target,
2181
from_repository=self.source,
2182
last_revision=revision_id,
2184
return f.count_copied, f.failed_revisions
2187
def missing_revision_ids(self, revision_id=None):
2188
"""See InterRepository.missing_revision_ids()."""
2189
# we want all revisions to satisfy revision_id in source.
2190
# but we don't want to stat every file here and there.
2191
# we want then, all revisions other needs to satisfy revision_id
2192
# checked, but not those that we have locally.
2193
# so the first thing is to get a subset of the revisions to
2194
# satisfy revision_id in source, and then eliminate those that
2195
# we do already have.
2196
# this is slow on high latency connection to self, but as as this
2197
# disk format scales terribly for push anyway due to rewriting
2198
# inventory.weave, this is considered acceptable.
2200
if revision_id is not None:
2201
source_ids = self.source.get_ancestry(revision_id)
2202
assert source_ids[0] is None
2205
source_ids = self.source._all_possible_ids()
2206
source_ids_set = set(source_ids)
2207
# source_ids is the worst possible case we may need to pull.
2208
# now we want to filter source_ids against what we actually
2209
# have in target, but don't try to check for existence where we know
2210
# we do not have a revision as that would be pointless.
2211
target_ids = set(self.target._all_possible_ids())
2212
possibly_present_revisions = target_ids.intersection(source_ids_set)
2213
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2214
required_revisions = source_ids_set.difference(actually_present_revisions)
2215
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2216
if revision_id is not None:
2217
# we used get_ancestry to determine source_ids then we are assured all
2218
# revisions referenced are present as they are installed in topological order.
2219
# and the tip revision was validated by get_ancestry.
2220
return required_topo_revisions
2222
# if we just grabbed the possibly available ids, then
2223
# we only have an estimate of whats available and need to validate
2224
# that against the revision records.
2225
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2228
class InterKnitRepo(InterSameDataRepository):
2229
"""Optimised code paths between Knit based repositories."""
2232
def _get_repo_format_to_test(self):
2233
from bzrlib.repofmt import knitrepo
2234
return knitrepo.RepositoryFormatKnit1()
2237
def is_compatible(source, target):
2238
"""Be compatible with known Knit formats.
2240
We don't test for the stores being of specific types because that
2241
could lead to confusing results, and there is no need to be
2244
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2246
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2247
isinstance(target._format, RepositoryFormatKnit))
2248
except AttributeError:
2250
return are_knits and InterRepository._same_model(source, target)
2253
def fetch(self, revision_id=None, pb=None):
2254
"""See InterRepository.fetch()."""
2255
from bzrlib.fetch import KnitRepoFetcher
2256
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2257
self.source, self.source._format, self.target, self.target._format)
2258
f = KnitRepoFetcher(to_repository=self.target,
2259
from_repository=self.source,
2260
last_revision=revision_id,
2262
return f.count_copied, f.failed_revisions
2265
def missing_revision_ids(self, revision_id=None):
2266
"""See InterRepository.missing_revision_ids()."""
2267
if revision_id is not None:
2268
source_ids = self.source.get_ancestry(revision_id)
2269
assert source_ids[0] is None
2272
source_ids = self.source.all_revision_ids()
2273
source_ids_set = set(source_ids)
2274
# source_ids is the worst possible case we may need to pull.
2275
# now we want to filter source_ids against what we actually
2276
# have in target, but don't try to check for existence where we know
2277
# we do not have a revision as that would be pointless.
2278
target_ids = set(self.target.all_revision_ids())
2279
possibly_present_revisions = target_ids.intersection(source_ids_set)
2280
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2281
required_revisions = source_ids_set.difference(actually_present_revisions)
2282
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2283
if revision_id is not None:
2284
# we used get_ancestry to determine source_ids then we are assured all
2285
# revisions referenced are present as they are installed in topological order.
2286
# and the tip revision was validated by get_ancestry.
2287
return required_topo_revisions
2289
# if we just grabbed the possibly available ids, then
2290
# we only have an estimate of whats available and need to validate
2291
# that against the revision records.
2292
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2295
class InterPackRepo(InterSameDataRepository):
2296
"""Optimised code paths between Pack based repositories."""
2299
def _get_repo_format_to_test(self):
2300
from bzrlib.repofmt import pack_repo
2301
return pack_repo.RepositoryFormatKnitPack1()
2304
def is_compatible(source, target):
2305
"""Be compatible with known Pack formats.
2307
We don't test for the stores being of specific types because that
2308
could lead to confusing results, and there is no need to be
2311
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2313
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2314
isinstance(target._format, RepositoryFormatPack))
2315
except AttributeError:
2317
return are_packs and InterRepository._same_model(source, target)
2320
def fetch(self, revision_id=None, pb=None):
2321
"""See InterRepository.fetch()."""
2322
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2323
self.source, self.source._format, self.target, self.target._format)
2324
self.count_copied = 0
2325
if revision_id is None:
2327
# everything to do - use pack logic
2328
# to fetch from all packs to one without
2329
# inventory parsing etc, IFF nothing to be copied is in the target.
2331
revision_ids = self.source.all_revision_ids()
2332
# implementing the TODO will involve:
2333
# - detecting when all of a pack is selected
2334
# - avoiding as much as possible pre-selection, so the
2335
# more-core routines such as create_pack_from_packs can filter in
2336
# a just-in-time fashion. (though having a HEADS list on a
2337
# repository might make this a lot easier, because we could
2338
# sensibly detect 'new revisions' without doing a full index scan.
2339
elif _mod_revision.is_null(revision_id):
2344
revision_ids = self.missing_revision_ids(revision_id)
2345
except errors.NoSuchRevision:
2346
raise errors.InstallFailed([revision_id])
2347
packs = self.source._packs.all_packs()
2348
pack = self.target._packs.create_pack_from_packs(
2349
packs, '.fetch', revision_ids,
2351
if pack is not None:
2352
self.target._packs._save_pack_names()
2353
# Trigger an autopack. This may duplicate effort as we've just done
2354
# a pack creation, but for now it is simpler to think about as
2355
# 'upload data, then repack if needed'.
2356
self.target._packs.autopack()
2357
return pack.get_revision_count()
2362
def missing_revision_ids(self, revision_id=None):
2363
"""See InterRepository.missing_revision_ids()."""
2364
if revision_id is not None:
2365
source_ids = self.source.get_ancestry(revision_id)
2366
assert source_ids[0] is None
2369
source_ids = self.source.all_revision_ids()
2370
source_ids_set = set(source_ids)
2371
# source_ids is the worst possible case we may need to pull.
2372
# now we want to filter source_ids against what we actually
2373
# have in target, but don't try to check for existence where we know
2374
# we do not have a revision as that would be pointless.
2375
target_ids = set(self.target.all_revision_ids())
2376
actually_present_revisions = target_ids.intersection(source_ids_set)
2377
required_revisions = source_ids_set.difference(actually_present_revisions)
2378
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2379
return required_topo_revisions
2382
class InterModel1and2(InterRepository):
2385
def _get_repo_format_to_test(self):
2389
def is_compatible(source, target):
2390
if not source.supports_rich_root() and target.supports_rich_root():
2396
def fetch(self, revision_id=None, pb=None):
2397
"""See InterRepository.fetch()."""
2398
from bzrlib.fetch import Model1toKnit2Fetcher
2399
f = Model1toKnit2Fetcher(to_repository=self.target,
2400
from_repository=self.source,
2401
last_revision=revision_id,
2403
return f.count_copied, f.failed_revisions
2406
def copy_content(self, revision_id=None):
2407
"""Make a complete copy of the content in self into destination.
2409
This is a destructive operation! Do not use it on existing
2412
:param revision_id: Only copy the content needed to construct
2413
revision_id and its parents.
2416
self.target.set_make_working_trees(self.source.make_working_trees())
2417
except NotImplementedError:
2419
# but don't bother fetching if we have the needed data now.
2420
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2421
self.target.has_revision(revision_id)):
2423
self.target.fetch(self.source, revision_id=revision_id)
2426
class InterKnit1and2(InterKnitRepo):
2429
def _get_repo_format_to_test(self):
2433
def is_compatible(source, target):
2434
"""Be compatible with Knit1 source and Knit3 target"""
2435
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2437
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2438
RepositoryFormatKnit3)
2439
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2440
RepositoryFormatKnitPack3)
2441
return (isinstance(source._format,
2442
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2443
isinstance(target._format,
2444
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2446
except AttributeError:
2450
def fetch(self, revision_id=None, pb=None):
2451
"""See InterRepository.fetch()."""
2452
from bzrlib.fetch import Knit1to2Fetcher
2453
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2454
self.source, self.source._format, self.target,
2455
self.target._format)
2456
f = Knit1to2Fetcher(to_repository=self.target,
2457
from_repository=self.source,
2458
last_revision=revision_id,
2460
return f.count_copied, f.failed_revisions
2463
class InterRemoteToOther(InterRepository):
2465
def __init__(self, source, target):
2466
InterRepository.__init__(self, source, target)
2467
self._real_inter = None
2470
def is_compatible(source, target):
2471
if not isinstance(source, remote.RemoteRepository):
2473
source._ensure_real()
2474
real_source = source._real_repository
2475
# Is source's model compatible with target's model, and are they the
2476
# same format? Currently we can only optimise fetching from an
2477
# identical model & format repo.
2478
assert not isinstance(real_source, remote.RemoteRepository), (
2479
"We don't support remote repos backed by remote repos yet.")
2480
return real_source._format == target._format
2483
def fetch(self, revision_id=None, pb=None):
2484
"""See InterRepository.fetch()."""
2485
from bzrlib.fetch import RemoteToOtherFetcher
2486
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2487
self.source, self.target, self.target._format)
2488
# TODO: jam 20070210 This should be an assert, not a translate
2489
revision_id = osutils.safe_revision_id(revision_id)
2490
f = RemoteToOtherFetcher(to_repository=self.target,
2491
from_repository=self.source,
2492
last_revision=revision_id,
2494
return f.count_copied, f.failed_revisions
2497
def _get_repo_format_to_test(self):
2501
class InterOtherToRemote(InterRepository):
2503
def __init__(self, source, target):
2504
InterRepository.__init__(self, source, target)
2505
self._real_inter = None
2508
def is_compatible(source, target):
2509
if isinstance(target, remote.RemoteRepository):
2513
def _ensure_real_inter(self):
2514
if self._real_inter is None:
2515
self.target._ensure_real()
2516
real_target = self.target._real_repository
2517
self._real_inter = InterRepository.get(self.source, real_target)
2519
def copy_content(self, revision_id=None):
2520
self._ensure_real_inter()
2521
self._real_inter.copy_content(revision_id=revision_id)
2523
def fetch(self, revision_id=None, pb=None):
2524
self._ensure_real_inter()
2525
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2528
def _get_repo_format_to_test(self):
2532
InterRepository.register_optimiser(InterSameDataRepository)
2533
InterRepository.register_optimiser(InterWeaveRepo)
2534
InterRepository.register_optimiser(InterKnitRepo)
2535
InterRepository.register_optimiser(InterModel1and2)
2536
InterRepository.register_optimiser(InterKnit1and2)
2537
InterRepository.register_optimiser(InterPackRepo)
2538
InterRepository.register_optimiser(InterRemoteToOther)
2539
InterRepository.register_optimiser(InterOtherToRemote)
2542
class CopyConverter(object):
2543
"""A repository conversion tool which just performs a copy of the content.
2545
This is slow but quite reliable.
2548
def __init__(self, target_format):
2549
"""Create a CopyConverter.
2551
:param target_format: The format the resulting repository should be.
2553
self.target_format = target_format
2555
def convert(self, repo, pb):
2556
"""Perform the conversion of to_convert, giving feedback via pb.
2558
:param to_convert: The disk object to convert.
2559
:param pb: a progress bar to use for progress information.
2564
# this is only useful with metadir layouts - separated repo content.
2565
# trigger an assertion if not such
2566
repo._format.get_format_string()
2567
self.repo_dir = repo.bzrdir
2568
self.step('Moving repository to repository.backup')
2569
self.repo_dir.transport.move('repository', 'repository.backup')
2570
backup_transport = self.repo_dir.transport.clone('repository.backup')
2571
repo._format.check_conversion_target(self.target_format)
2572
self.source_repo = repo._format.open(self.repo_dir,
2574
_override_transport=backup_transport)
2575
self.step('Creating new repository')
2576
converted = self.target_format.initialize(self.repo_dir,
2577
self.source_repo.is_shared())
2578
converted.lock_write()
2580
self.step('Copying content into repository.')
2581
self.source_repo.copy_content_into(converted)
2584
self.step('Deleting old repository content.')
2585
self.repo_dir.transport.delete_tree('repository.backup')
2586
self.pb.note('repository converted')
2588
def step(self, message):
2589
"""Update the pb by a step."""
2591
self.pb.update(message, self.count, self.total)
2603
def _unescaper(match, _map=_unescape_map):
2604
code = match.group(1)
2608
if not code.startswith('#'):
2610
return unichr(int(code[1:])).encode('utf8')
2616
def _unescape_xml(data):
2617
"""Unescape predefined XML entities in a string of data."""
2619
if _unescape_re is None:
2620
_unescape_re = re.compile('\&([^;]*);')
2621
return _unescape_re.sub(_unescaper, data)
2624
class _RevisionTextVersionCache(object):
2625
"""A cache of the versionedfile versions for revision and file-id."""
2627
def __init__(self, repository):
2628
self.repository = repository
2629
self.revision_versions = {}
2630
self.revision_parents = {}
2631
self.repo_graph = self.repository.get_graph()
2632
# XXX: RBC: I haven't tracked down what uses this, but it would be
2633
# better to use the headscache directly I think.
2634
self.heads = graph.HeadsCache(self.repo_graph).heads
2636
def add_revision_text_versions(self, tree):
2637
"""Cache text version data from the supplied revision tree"""
2639
for path, entry in tree.iter_entries_by_dir():
2640
inv_revisions[entry.file_id] = entry.revision
2641
self.revision_versions[tree.get_revision_id()] = inv_revisions
2642
return inv_revisions
2644
def get_text_version(self, file_id, revision_id):
2645
"""Determine the text version for a given file-id and revision-id"""
2647
inv_revisions = self.revision_versions[revision_id]
2650
tree = self.repository.revision_tree(revision_id)
2651
except errors.RevisionNotPresent:
2652
self.revision_versions[revision_id] = inv_revisions = {}
2654
inv_revisions = self.add_revision_text_versions(tree)
2655
return inv_revisions.get(file_id)
2657
def prepopulate_revs(self, revision_ids):
2658
# Filter out versions that we don't have an inventory for, so that the
2659
# revision_trees() call won't fail.
2660
inv_weave = self.repository.get_inventory_weave()
2661
revs = [r for r in revision_ids if inv_weave.has_version(r)]
2662
# XXX: this loop is very similar to
2663
# bzrlib.fetch.Inter1and2Helper.iter_rev_trees.
2665
for tree in self.repository.revision_trees(revs[:100]):
2666
if tree.inventory.revision_id is None:
2667
tree.inventory.revision_id = tree.get_revision_id()
2668
self.add_revision_text_versions(tree)
2671
def get_parents(self, revision_id):
2673
return self.revision_parents[revision_id]
2675
parents = self.repository.get_parents([revision_id])[0]
2676
self.revision_parents[revision_id] = parents
2680
class VersionedFileChecker(object):
2682
def __init__(self, planned_revisions, revision_versions, repository):
2683
self.planned_revisions = planned_revisions
2684
self.revision_versions = revision_versions
2685
self.repository = repository
2687
def calculate_file_version_parents(self, revision_id, file_id):
2688
text_revision = self.revision_versions.get_text_version(
2689
file_id, revision_id)
2690
if text_revision is None:
2692
parents_of_text_revision = self.revision_versions.get_parents(
2694
parents_from_inventories = []
2695
for parent in parents_of_text_revision:
2696
if parent == _mod_revision.NULL_REVISION:
2698
introduced_in = self.revision_versions.get_text_version(file_id,
2700
if introduced_in is not None:
2701
parents_from_inventories.append(introduced_in)
2702
heads = set(self.revision_versions.heads(parents_from_inventories))
2704
for parent in parents_from_inventories:
2705
if parent in heads and parent not in new_parents:
2706
new_parents.append(parent)
2707
return tuple(new_parents)
2709
def check_file_version_parents(self, weave, file_id):
2711
for num, revision_id in enumerate(self.planned_revisions):
2712
correct_parents = self.calculate_file_version_parents(
2713
revision_id, file_id)
2714
if correct_parents is None:
2716
text_revision = self.revision_versions.get_text_version(
2717
file_id, revision_id)
2718
knit_parents = tuple(weave.get_parents(text_revision))
2719
if correct_parents != knit_parents:
2720
result[revision_id] = (knit_parents, correct_parents)
added
removed
bzrlib/repository.py
