1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from cStringIO import StringIO
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
39
revision as _mod_revision,
44
from bzrlib.bundle import serializer
45
from bzrlib.revisiontree import RevisionTree
46
from bzrlib.store.versioned import VersionedFileStore
47
from bzrlib.store.text import TextStore
48
from bzrlib.testament import Testament
49
from bzrlib.util import bencode
52
from bzrlib.decorators import needs_read_lock, needs_write_lock
53
from bzrlib.inter import InterObject
54
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
55
from bzrlib.symbol_versioning import (
58
from bzrlib.trace import mutter, mutter_callsite, note, warning
61
# Old formats display a warning, but only once
62
_deprecation_warning_done = False
65
class CommitBuilder(object):
66
"""Provides an interface to build up a commit.
68
This allows describing a tree to be committed without needing to
69
know the internals of the format of the repository.
72
# all clients should supply tree roots.
73
record_root_entry = True
74
# the default CommitBuilder does not manage trees whose root is versioned.
75
_versioned_root = False
77
def __init__(self, repository, parents, config, timestamp=None,
78
timezone=None, committer=None, revprops=None,
80
"""Initiate a CommitBuilder.
82
:param repository: Repository to commit to.
83
:param parents: Revision ids of the parents of the new revision.
84
:param config: Configuration to use.
85
:param timestamp: Optional timestamp recorded for commit.
86
:param timezone: Optional timezone for timestamp.
87
:param committer: Optional committer to set for commit.
88
:param revprops: Optional dictionary of revision properties.
89
:param revision_id: Optional revision id.
94
self._committer = self._config.username()
96
assert isinstance(committer, basestring), type(committer)
97
self._committer = committer
99
self.new_inventory = Inventory(None)
100
self._new_revision_id = revision_id
101
self.parents = parents
102
self.repository = repository
105
if revprops is not None:
106
self._revprops.update(revprops)
108
if timestamp is None:
109
timestamp = time.time()
110
# Restrict resolution to 1ms
111
self._timestamp = round(timestamp, 3)
114
self._timezone = osutils.local_time_offset()
116
self._timezone = int(timezone)
118
self._generate_revision_if_needed()
119
self._heads = graph.HeadsCache(repository.get_graph()).heads
121
def commit(self, message):
122
"""Make the actual commit.
124
:return: The revision id of the recorded revision.
126
rev = _mod_revision.Revision(
127
timestamp=self._timestamp,
128
timezone=self._timezone,
129
committer=self._committer,
131
inventory_sha1=self.inv_sha1,
132
revision_id=self._new_revision_id,
133
properties=self._revprops)
134
rev.parent_ids = self.parents
135
self.repository.add_revision(self._new_revision_id, rev,
136
self.new_inventory, self._config)
137
self.repository.commit_write_group()
138
return self._new_revision_id
141
"""Abort the commit that is being built.
143
self.repository.abort_write_group()
145
def revision_tree(self):
146
"""Return the tree that was just committed.
148
After calling commit() this can be called to get a RevisionTree
149
representing the newly committed tree. This is preferred to
150
calling Repository.revision_tree() because that may require
151
deserializing the inventory, while we already have a copy in
154
return RevisionTree(self.repository, self.new_inventory,
155
self._new_revision_id)
157
def finish_inventory(self):
158
"""Tell the builder that the inventory is finished."""
159
if self.new_inventory.root is None:
160
raise AssertionError('Root entry should be supplied to'
161
' record_entry_contents, as of bzr 0.10.',
162
DeprecationWarning, stacklevel=2)
163
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
164
self.new_inventory.revision_id = self._new_revision_id
165
self.inv_sha1 = self.repository.add_inventory(
166
self._new_revision_id,
171
def _gen_revision_id(self):
172
"""Return new revision-id."""
173
return generate_ids.gen_revision_id(self._config.username(),
176
def _generate_revision_if_needed(self):
177
"""Create a revision id if None was supplied.
179
If the repository can not support user-specified revision ids
180
they should override this function and raise CannotSetRevisionId
181
if _new_revision_id is not None.
183
:raises: CannotSetRevisionId
185
if self._new_revision_id is None:
186
self._new_revision_id = self._gen_revision_id()
187
self.random_revid = True
189
self.random_revid = False
191
def _check_root(self, ie, parent_invs, tree):
192
"""Helper for record_entry_contents.
194
:param ie: An entry being added.
195
:param parent_invs: The inventories of the parent revisions of the
197
:param tree: The tree that is being committed.
199
# In this revision format, root entries have no knit or weave When
200
# serializing out to disk and back in root.revision is always
202
ie.revision = self._new_revision_id
204
def _get_delta(self, ie, basis_inv, path):
205
"""Get a delta against the basis inventory for ie."""
206
if ie.file_id not in basis_inv:
208
return (None, path, ie.file_id, ie)
209
elif ie != basis_inv[ie.file_id]:
211
# TODO: avoid tis id2path call.
212
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
217
def record_entry_contents(self, ie, parent_invs, path, tree,
219
"""Record the content of ie from tree into the commit if needed.
221
Side effect: sets ie.revision when unchanged
223
:param ie: An inventory entry present in the commit.
224
:param parent_invs: The inventories of the parent revisions of the
226
:param path: The path the entry is at in the tree.
227
:param tree: The tree which contains this entry and should be used to
229
:param content_summary: Summary data from the tree about the paths
230
content - stat, length, exec, sha/link target. This is only
231
accessed when the entry has a revision of None - that is when it is
232
a candidate to commit.
233
:return: A tuple (change_delta, version_recorded). change_delta is
234
an inventory_delta change for this entry against the basis tree of
235
the commit, or None if no change occured against the basis tree.
236
version_recorded is True if a new version of the entry has been
237
recorded. For instance, committing a merge where a file was only
238
changed on the other side will return (delta, False).
240
if self.new_inventory.root is None:
241
if ie.parent_id is not None:
242
raise errors.RootMissing()
243
self._check_root(ie, parent_invs, tree)
244
if ie.revision is None:
245
kind = content_summary[0]
247
# ie is carried over from a prior commit
249
# XXX: repository specific check for nested tree support goes here - if
250
# the repo doesn't want nested trees we skip it ?
251
if (kind == 'tree-reference' and
252
not self.repository._format.supports_tree_reference):
253
# mismatch between commit builder logic and repository:
254
# this needs the entry creation pushed down into the builder.
255
raise NotImplementedError('Missing repository subtree support.')
256
self.new_inventory.add(ie)
258
# TODO: slow, take it out of the inner loop.
260
basis_inv = parent_invs[0]
262
basis_inv = Inventory(root_id=None)
264
# ie.revision is always None if the InventoryEntry is considered
265
# for committing. We may record the previous parents revision if the
266
# content is actually unchanged against a sole head.
267
if ie.revision is not None:
268
if not self._versioned_root and path == '':
269
# repositories that do not version the root set the root's
270
# revision to the new commit even when no change occurs, and
271
# this masks when a change may have occurred against the basis,
272
# so calculate if one happened.
273
if ie.file_id in basis_inv:
274
delta = (basis_inv.id2path(ie.file_id), path,
278
delta = (None, path, ie.file_id, ie)
281
# we don't need to commit this, because the caller already
282
# determined that an existing revision of this file is
284
return None, (ie.revision == self._new_revision_id)
285
# XXX: Friction: parent_candidates should return a list not a dict
286
# so that we don't have to walk the inventories again.
287
parent_candiate_entries = ie.parent_candidates(parent_invs)
288
head_set = self._heads(parent_candiate_entries.keys())
290
for inv in parent_invs:
291
if ie.file_id in inv:
292
old_rev = inv[ie.file_id].revision
293
if old_rev in head_set:
294
heads.append(inv[ie.file_id].revision)
295
head_set.remove(inv[ie.file_id].revision)
298
# now we check to see if we need to write a new record to the
300
# We write a new entry unless there is one head to the ancestors, and
301
# the kind-derived content is unchanged.
303
# Cheapest check first: no ancestors, or more the one head in the
304
# ancestors, we write a new node.
308
# There is a single head, look it up for comparison
309
parent_entry = parent_candiate_entries[heads[0]]
310
# if the non-content specific data has changed, we'll be writing a
312
if (parent_entry.parent_id != ie.parent_id or
313
parent_entry.name != ie.name):
315
# now we need to do content specific checks:
317
# if the kind changed the content obviously has
318
if kind != parent_entry.kind:
321
assert content_summary[2] is not None, \
322
"Files must not have executable = None"
324
if (# if the file length changed we have to store:
325
parent_entry.text_size != content_summary[1] or
326
# if the exec bit has changed we have to store:
327
parent_entry.executable != content_summary[2]):
329
elif parent_entry.text_sha1 == content_summary[3]:
330
# all meta and content is unchanged (using a hash cache
331
# hit to check the sha)
332
ie.revision = parent_entry.revision
333
ie.text_size = parent_entry.text_size
334
ie.text_sha1 = parent_entry.text_sha1
335
ie.executable = parent_entry.executable
336
return self._get_delta(ie, basis_inv, path), False
338
# Either there is only a hash change(no hash cache entry,
339
# or same size content change), or there is no change on
341
# Provide the parent's hash to the store layer, so that the
342
# content is unchanged we will not store a new node.
343
nostore_sha = parent_entry.text_sha1
345
# We want to record a new node regardless of the presence or
346
# absence of a content change in the file.
348
ie.executable = content_summary[2]
349
lines = tree.get_file(ie.file_id, path).readlines()
351
ie.text_sha1, ie.text_size = self._add_text_to_weave(
352
ie.file_id, lines, heads, nostore_sha)
353
except errors.ExistingContent:
354
# Turns out that the file content was unchanged, and we were
355
# only going to store a new node if it was changed. Carry over
357
ie.revision = parent_entry.revision
358
ie.text_size = parent_entry.text_size
359
ie.text_sha1 = parent_entry.text_sha1
360
ie.executable = parent_entry.executable
361
return self._get_delta(ie, basis_inv, path), False
362
elif kind == 'directory':
364
# all data is meta here, nothing specific to directory, so
366
ie.revision = parent_entry.revision
367
return self._get_delta(ie, basis_inv, path), False
369
self._add_text_to_weave(ie.file_id, lines, heads, None)
370
elif kind == 'symlink':
371
current_link_target = content_summary[3]
373
# symlink target is not generic metadata, check if it has
375
if current_link_target != parent_entry.symlink_target:
378
# unchanged, carry over.
379
ie.revision = parent_entry.revision
380
ie.symlink_target = parent_entry.symlink_target
381
return self._get_delta(ie, basis_inv, path), False
382
ie.symlink_target = current_link_target
384
self._add_text_to_weave(ie.file_id, lines, heads, None)
385
elif kind == 'tree-reference':
387
if content_summary[3] != parent_entry.reference_revision:
390
# unchanged, carry over.
391
ie.reference_revision = parent_entry.reference_revision
392
ie.revision = parent_entry.revision
393
return self._get_delta(ie, basis_inv, path), False
394
ie.reference_revision = content_summary[3]
396
self._add_text_to_weave(ie.file_id, lines, heads, None)
398
raise NotImplementedError('unknown kind')
399
ie.revision = self._new_revision_id
400
return self._get_delta(ie, basis_inv, path), True
402
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
403
versionedfile = self.repository.weave_store.get_weave_or_empty(
404
file_id, self.repository.get_transaction())
405
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
406
# than add_lines, and allows committing when a parent is ghosted for
408
# Note: as we read the content directly from the tree, we know its not
409
# been turned into unicode or badly split - but a broken tree
410
# implementation could give us bad output from readlines() so this is
411
# not a guarantee of safety. What would be better is always checking
412
# the content during test suite execution. RBC 20070912
414
return versionedfile.add_lines_with_ghosts(
415
self._new_revision_id, parents, new_lines,
416
nostore_sha=nostore_sha, random_id=self.random_revid,
417
check_content=False)[0:2]
419
versionedfile.clear_cache()
422
class RootCommitBuilder(CommitBuilder):
423
"""This commitbuilder actually records the root id"""
425
# the root entry gets versioned properly by this builder.
426
_versioned_root = True
428
def _check_root(self, ie, parent_invs, tree):
429
"""Helper for record_entry_contents.
431
:param ie: An entry being added.
432
:param parent_invs: The inventories of the parent revisions of the
434
:param tree: The tree that is being committed.
438
######################################################################
441
class Repository(object):
442
"""Repository holding history for one or more branches.
444
The repository holds and retrieves historical information including
445
revisions and file history. It's normally accessed only by the Branch,
446
which views a particular line of development through that history.
448
The Repository builds on top of Stores and a Transport, which respectively
449
describe the disk data format and the way of accessing the (possibly
453
# What class to use for a CommitBuilder. Often its simpler to change this
454
# in a Repository class subclass rather than to override
455
# get_commit_builder.
456
_commit_builder_class = CommitBuilder
457
# The search regex used by xml based repositories to determine what things
458
# where changed in a single commit.
459
_file_ids_altered_regex = lazy_regex.lazy_compile(
460
r'file_id="(?P<file_id>[^"]+)"'
461
r'.* revision="(?P<revision_id>[^"]+)"'
464
def abort_write_group(self):
465
"""Commit the contents accrued within the current write group.
467
:seealso: start_write_group.
469
if self._write_group is not self.get_transaction():
470
# has an unlock or relock occured ?
471
raise errors.BzrError('mismatched lock context and write group.')
472
self._abort_write_group()
473
self._write_group = None
475
def _abort_write_group(self):
476
"""Template method for per-repository write group cleanup.
478
This is called during abort before the write group is considered to be
479
finished and should cleanup any internal state accrued during the write
480
group. There is no requirement that data handed to the repository be
481
*not* made available - this is not a rollback - but neither should any
482
attempt be made to ensure that data added is fully commited. Abort is
483
invoked when an error has occured so futher disk or network operations
484
may not be possible or may error and if possible should not be
489
def add_inventory(self, revision_id, inv, parents):
490
"""Add the inventory inv to the repository as revision_id.
492
:param parents: The revision ids of the parents that revision_id
493
is known to have and are in the repository already.
495
returns the sha1 of the serialized inventory.
497
assert self.is_in_write_group()
498
_mod_revision.check_not_reserved_id(revision_id)
499
assert inv.revision_id is None or inv.revision_id == revision_id, \
500
"Mismatch between inventory revision" \
501
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
502
assert inv.root is not None
503
inv_lines = self._serialise_inventory_to_lines(inv)
504
inv_vf = self.get_inventory_weave()
505
return self._inventory_add_lines(inv_vf, revision_id, parents,
506
inv_lines, check_content=False)
508
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
510
"""Store lines in inv_vf and return the sha1 of the inventory."""
512
for parent in parents:
514
final_parents.append(parent)
515
return inv_vf.add_lines(revision_id, final_parents, lines,
516
check_content=check_content)[0]
519
def add_revision(self, revision_id, rev, inv=None, config=None):
520
"""Add rev to the revision store as revision_id.
522
:param revision_id: the revision id to use.
523
:param rev: The revision object.
524
:param inv: The inventory for the revision. if None, it will be looked
525
up in the inventory storer
526
:param config: If None no digital signature will be created.
527
If supplied its signature_needed method will be used
528
to determine if a signature should be made.
530
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
532
_mod_revision.check_not_reserved_id(revision_id)
533
if config is not None and config.signature_needed():
535
inv = self.get_inventory(revision_id)
536
plaintext = Testament(rev, inv).as_short_text()
537
self.store_revision_signature(
538
gpg.GPGStrategy(config), plaintext, revision_id)
539
if not revision_id in self.get_inventory_weave():
541
raise errors.WeaveRevisionNotPresent(revision_id,
542
self.get_inventory_weave())
544
# yes, this is not suitable for adding with ghosts.
545
self.add_inventory(revision_id, inv, rev.parent_ids)
546
self._revision_store.add_revision(rev, self.get_transaction())
548
def _add_revision_text(self, revision_id, text):
549
revision = self._revision_store._serializer.read_revision_from_string(
551
self._revision_store._add_revision(revision, StringIO(text),
552
self.get_transaction())
554
def all_revision_ids(self):
555
"""Returns a list of all the revision ids in the repository.
557
This is deprecated because code should generally work on the graph
558
reachable from a particular revision, and ignore any other revisions
559
that might be present. There is no direct replacement method.
561
if 'evil' in debug.debug_flags:
562
mutter_callsite(2, "all_revision_ids is linear with history.")
563
return self._all_revision_ids()
565
def _all_revision_ids(self):
566
"""Returns a list of all the revision ids in the repository.
568
These are in as much topological order as the underlying store can
571
raise NotImplementedError(self._all_revision_ids)
573
def break_lock(self):
574
"""Break a lock if one is present from another instance.
576
Uses the ui factory to ask for confirmation if the lock may be from
579
self.control_files.break_lock()
582
def _eliminate_revisions_not_present(self, revision_ids):
583
"""Check every revision id in revision_ids to see if we have it.
585
Returns a set of the present revisions.
588
for id in revision_ids:
589
if self.has_revision(id):
594
def create(a_bzrdir):
595
"""Construct the current default format repository in a_bzrdir."""
596
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
598
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
599
"""instantiate a Repository.
601
:param _format: The format of the repository on disk.
602
:param a_bzrdir: The BzrDir of the repository.
604
In the future we will have a single api for all stores for
605
getting file texts, inventories and revisions, then
606
this construct will accept instances of those things.
608
super(Repository, self).__init__()
609
self._format = _format
610
# the following are part of the public API for Repository:
611
self.bzrdir = a_bzrdir
612
self.control_files = control_files
613
self._revision_store = _revision_store
614
# backwards compatibility
615
self.weave_store = text_store
617
self._reconcile_does_inventory_gc = True
618
self._reconcile_fixes_text_parents = False
619
self._reconcile_backsup_inventory = True
620
# not right yet - should be more semantically clear ?
622
self.control_store = control_store
623
self.control_weaves = control_store
624
# TODO: make sure to construct the right store classes, etc, depending
625
# on whether escaping is required.
626
self._warn_if_deprecated()
627
self._write_group = None
628
self.base = control_files._transport.base
631
return '%s(%r)' % (self.__class__.__name__,
634
def has_same_location(self, other):
635
"""Returns a boolean indicating if this repository is at the same
636
location as another repository.
638
This might return False even when two repository objects are accessing
639
the same physical repository via different URLs.
641
if self.__class__ is not other.__class__:
643
return (self.control_files._transport.base ==
644
other.control_files._transport.base)
646
def is_in_write_group(self):
647
"""Return True if there is an open write group.
649
:seealso: start_write_group.
651
return self._write_group is not None
654
return self.control_files.is_locked()
656
def is_write_locked(self):
657
"""Return True if this object is write locked."""
658
return self.is_locked() and self.control_files._lock_mode == 'w'
660
def lock_write(self, token=None):
661
"""Lock this repository for writing.
663
This causes caching within the repository obejct to start accumlating
664
data during reads, and allows a 'write_group' to be obtained. Write
665
groups must be used for actual data insertion.
667
:param token: if this is already locked, then lock_write will fail
668
unless the token matches the existing lock.
669
:returns: a token if this instance supports tokens, otherwise None.
670
:raises TokenLockingNotSupported: when a token is given but this
671
instance doesn't support using token locks.
672
:raises MismatchedToken: if the specified token doesn't match the token
673
of the existing lock.
674
:seealso: start_write_group.
676
A token should be passed in if you know that you have locked the object
677
some other way, and need to synchronise this object's state with that
680
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
682
result = self.control_files.lock_write(token=token)
687
self.control_files.lock_read()
690
def get_physical_lock_status(self):
691
return self.control_files.get_physical_lock_status()
693
def leave_lock_in_place(self):
694
"""Tell this repository not to release the physical lock when this
697
If lock_write doesn't return a token, then this method is not supported.
699
self.control_files.leave_in_place()
701
def dont_leave_lock_in_place(self):
702
"""Tell this repository to release the physical lock when this
703
object is unlocked, even if it didn't originally acquire it.
705
If lock_write doesn't return a token, then this method is not supported.
707
self.control_files.dont_leave_in_place()
710
def gather_stats(self, revid=None, committers=None):
711
"""Gather statistics from a revision id.
713
:param revid: The revision id to gather statistics from, if None, then
714
no revision specific statistics are gathered.
715
:param committers: Optional parameter controlling whether to grab
716
a count of committers from the revision specific statistics.
717
:return: A dictionary of statistics. Currently this contains:
718
committers: The number of committers if requested.
719
firstrev: A tuple with timestamp, timezone for the penultimate left
720
most ancestor of revid, if revid is not the NULL_REVISION.
721
latestrev: A tuple with timestamp, timezone for revid, if revid is
722
not the NULL_REVISION.
723
revisions: The total revision count in the repository.
724
size: An estimate disk size of the repository in bytes.
727
if revid and committers:
728
result['committers'] = 0
729
if revid and revid != _mod_revision.NULL_REVISION:
731
all_committers = set()
732
revisions = self.get_ancestry(revid)
733
# pop the leading None
735
first_revision = None
737
# ignore the revisions in the middle - just grab first and last
738
revisions = revisions[0], revisions[-1]
739
for revision in self.get_revisions(revisions):
740
if not first_revision:
741
first_revision = revision
743
all_committers.add(revision.committer)
744
last_revision = revision
746
result['committers'] = len(all_committers)
747
result['firstrev'] = (first_revision.timestamp,
748
first_revision.timezone)
749
result['latestrev'] = (last_revision.timestamp,
750
last_revision.timezone)
752
# now gather global repository information
753
if self.bzrdir.root_transport.listable():
754
c, t = self._revision_store.total_size(self.get_transaction())
755
result['revisions'] = c
759
def get_data_stream(self, revision_ids):
760
raise NotImplementedError(self.get_data_stream)
762
def insert_data_stream(self, stream):
763
"""XXX What does this really do?
765
Is it a substitute for fetch?
766
Should it manage its own write group ?
768
for item_key, bytes in stream:
769
if item_key[0] == 'file':
770
(file_id,) = item_key[1:]
771
knit = self.weave_store.get_weave_or_empty(
772
file_id, self.get_transaction())
773
elif item_key == ('inventory',):
774
knit = self.get_inventory_weave()
775
elif item_key == ('revisions',):
776
knit = self._revision_store.get_revision_file(
777
self.get_transaction())
778
elif item_key == ('signatures',):
779
knit = self._revision_store.get_signature_file(
780
self.get_transaction())
782
raise RepositoryDataStreamError(
783
"Unrecognised data stream key '%s'" % (item_key,))
784
decoded_list = bencode.bdecode(bytes)
785
format = decoded_list.pop(0)
788
for version, options, parents, some_bytes in decoded_list:
789
data_list.append((version, options, len(some_bytes), parents))
790
knit_bytes += some_bytes
791
knit.insert_data_stream(
792
(format, data_list, StringIO(knit_bytes).read))
795
def missing_revision_ids(self, other, revision_id=None):
796
"""Return the revision ids that other has that this does not.
798
These are returned in topological order.
800
revision_id: only return revision ids included by revision_id.
802
return InterRepository.get(other, self).missing_revision_ids(revision_id)
806
"""Open the repository rooted at base.
808
For instance, if the repository is at URL/.bzr/repository,
809
Repository.open(URL) -> a Repository instance.
811
control = bzrdir.BzrDir.open(base)
812
return control.open_repository()
814
def copy_content_into(self, destination, revision_id=None):
815
"""Make a complete copy of the content in self into destination.
817
This is a destructive operation! Do not use it on existing
820
return InterRepository.get(self, destination).copy_content(revision_id)
822
def commit_write_group(self):
823
"""Commit the contents accrued within the current write group.
825
:seealso: start_write_group.
827
if self._write_group is not self.get_transaction():
828
# has an unlock or relock occured ?
829
raise errors.BzrError('mismatched lock context %r and '
831
(self.get_transaction(), self._write_group))
832
self._commit_write_group()
833
self._write_group = None
835
def _commit_write_group(self):
836
"""Template method for per-repository write group cleanup.
838
This is called before the write group is considered to be
839
finished and should ensure that all data handed to the repository
840
for writing during the write group is safely committed (to the
841
extent possible considering file system caching etc).
844
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
845
"""Fetch the content required to construct revision_id from source.
847
If revision_id is None all content is copied.
848
:param find_ghosts: Find and copy revisions in the source that are
849
ghosts in the target (and not reachable directly by walking out to
850
the first-present revision in target from revision_id).
852
# fast path same-url fetch operations
853
if self.has_same_location(source):
854
# check that last_revision is in 'from' and then return a
856
if (revision_id is not None and
857
not _mod_revision.is_null(revision_id)):
858
self.get_revision(revision_id)
860
inter = InterRepository.get(source, self)
862
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
863
except NotImplementedError:
864
raise errors.IncompatibleRepositories(source, self)
866
def create_bundle(self, target, base, fileobj, format=None):
867
return serializer.write_bundle(self, target, base, fileobj, format)
869
def get_commit_builder(self, branch, parents, config, timestamp=None,
870
timezone=None, committer=None, revprops=None,
872
"""Obtain a CommitBuilder for this repository.
874
:param branch: Branch to commit to.
875
:param parents: Revision ids of the parents of the new revision.
876
:param config: Configuration to use.
877
:param timestamp: Optional timestamp recorded for commit.
878
:param timezone: Optional timezone for timestamp.
879
:param committer: Optional committer to set for commit.
880
:param revprops: Optional dictionary of revision properties.
881
:param revision_id: Optional revision id.
883
result = self._commit_builder_class(self, parents, config,
884
timestamp, timezone, committer, revprops, revision_id)
885
self.start_write_group()
889
if (self.control_files._lock_count == 1 and
890
self.control_files._lock_mode == 'w'):
891
if self._write_group is not None:
892
self.abort_write_group()
893
self.control_files.unlock()
894
raise errors.BzrError(
895
'Must end write groups before releasing write locks.')
896
self.control_files.unlock()
899
def clone(self, a_bzrdir, revision_id=None):
900
"""Clone this repository into a_bzrdir using the current format.
902
Currently no check is made that the format of this repository and
903
the bzrdir format are compatible. FIXME RBC 20060201.
905
:return: The newly created destination repository.
907
# TODO: deprecate after 0.16; cloning this with all its settings is
908
# probably not very useful -- mbp 20070423
909
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
910
self.copy_content_into(dest_repo, revision_id)
913
def start_write_group(self):
914
"""Start a write group in the repository.
916
Write groups are used by repositories which do not have a 1:1 mapping
917
between file ids and backend store to manage the insertion of data from
918
both fetch and commit operations.
920
A write lock is required around the start_write_group/commit_write_group
921
for the support of lock-requiring repository formats.
923
One can only insert data into a repository inside a write group.
927
if not self.is_write_locked():
928
raise errors.NotWriteLocked(self)
929
if self._write_group:
930
raise errors.BzrError('already in a write group')
931
self._start_write_group()
932
# so we can detect unlock/relock - the write group is now entered.
933
self._write_group = self.get_transaction()
935
def _start_write_group(self):
936
"""Template method for per-repository write group startup.
938
This is called before the write group is considered to be
943
def sprout(self, to_bzrdir, revision_id=None):
944
"""Create a descendent repository for new development.
946
Unlike clone, this does not copy the settings of the repository.
948
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
949
dest_repo.fetch(self, revision_id=revision_id)
952
def _create_sprouting_repo(self, a_bzrdir, shared):
953
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
954
# use target default format.
955
dest_repo = a_bzrdir.create_repository()
957
# Most control formats need the repository to be specifically
958
# created, but on some old all-in-one formats it's not needed
960
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
961
except errors.UninitializableFormat:
962
dest_repo = a_bzrdir.open_repository()
966
def has_revision(self, revision_id):
967
"""True if this repository has a copy of the revision."""
968
if 'evil' in debug.debug_flags:
969
mutter_callsite(3, "has_revision is a LBYL symptom.")
970
return self._revision_store.has_revision_id(revision_id,
971
self.get_transaction())
974
def get_revision(self, revision_id):
975
"""Return the Revision object for a named revision."""
976
return self.get_revisions([revision_id])[0]
979
def get_revision_reconcile(self, revision_id):
980
"""'reconcile' helper routine that allows access to a revision always.
982
This variant of get_revision does not cross check the weave graph
983
against the revision one as get_revision does: but it should only
984
be used by reconcile, or reconcile-alike commands that are correcting
985
or testing the revision graph.
987
return self._get_revisions([revision_id])[0]
990
def get_revisions(self, revision_ids):
991
"""Get many revisions at once."""
992
return self._get_revisions(revision_ids)
995
def _get_revisions(self, revision_ids):
996
"""Core work logic to get many revisions without sanity checks."""
997
for rev_id in revision_ids:
998
if not rev_id or not isinstance(rev_id, basestring):
999
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1000
revs = self._revision_store.get_revisions(revision_ids,
1001
self.get_transaction())
1003
assert not isinstance(rev.revision_id, unicode)
1004
for parent_id in rev.parent_ids:
1005
assert not isinstance(parent_id, unicode)
1009
def get_revision_xml(self, revision_id):
1010
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1011
# would have already do it.
1012
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1013
rev = self.get_revision(revision_id)
1014
rev_tmp = StringIO()
1015
# the current serializer..
1016
self._revision_store._serializer.write_revision(rev, rev_tmp)
1018
return rev_tmp.getvalue()
1021
def get_deltas_for_revisions(self, revisions):
1022
"""Produce a generator of revision deltas.
1024
Note that the input is a sequence of REVISIONS, not revision_ids.
1025
Trees will be held in memory until the generator exits.
1026
Each delta is relative to the revision's lefthand predecessor.
1028
required_trees = set()
1029
for revision in revisions:
1030
required_trees.add(revision.revision_id)
1031
required_trees.update(revision.parent_ids[:1])
1032
trees = dict((t.get_revision_id(), t) for
1033
t in self.revision_trees(required_trees))
1034
for revision in revisions:
1035
if not revision.parent_ids:
1036
old_tree = self.revision_tree(None)
1038
old_tree = trees[revision.parent_ids[0]]
1039
yield trees[revision.revision_id].changes_from(old_tree)
1042
def get_revision_delta(self, revision_id):
1043
"""Return the delta for one revision.
1045
The delta is relative to the left-hand predecessor of the
1048
r = self.get_revision(revision_id)
1049
return list(self.get_deltas_for_revisions([r]))[0]
1052
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1053
signature = gpg_strategy.sign(plaintext)
1054
self._revision_store.add_revision_signature_text(revision_id,
1056
self.get_transaction())
1058
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1060
"""Helper routine for fileids_altered_by_revision_ids.
1062
This performs the translation of xml lines to revision ids.
1064
:param line_iterator: An iterator of lines
1065
:param revision_ids: The revision ids to filter for. This should be a
1066
set or other type which supports efficient __contains__ lookups, as
1067
the revision id from each parsed line will be looked up in the
1068
revision_ids filter.
1069
:return: a dictionary mapping altered file-ids to an iterable of
1070
revision_ids. Each altered file-ids has the exact revision_ids that
1071
altered it listed explicitly.
1075
# this code needs to read every new line in every inventory for the
1076
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1077
# not present in one of those inventories is unnecessary but not
1078
# harmful because we are filtering by the revision id marker in the
1079
# inventory lines : we only select file ids altered in one of those
1080
# revisions. We don't need to see all lines in the inventory because
1081
# only those added in an inventory in rev X can contain a revision=X
1083
unescape_revid_cache = {}
1084
unescape_fileid_cache = {}
1086
# jam 20061218 In a big fetch, this handles hundreds of thousands
1087
# of lines, so it has had a lot of inlining and optimizing done.
1088
# Sorry that it is a little bit messy.
1089
# Move several functions to be local variables, since this is a long
1091
search = self._file_ids_altered_regex.search
1092
unescape = _unescape_xml
1093
setdefault = result.setdefault
1094
for line in line_iterator:
1095
match = search(line)
1098
# One call to match.group() returning multiple items is quite a
1099
# bit faster than 2 calls to match.group() each returning 1
1100
file_id, revision_id = match.group('file_id', 'revision_id')
1102
# Inlining the cache lookups helps a lot when you make 170,000
1103
# lines and 350k ids, versus 8.4 unique ids.
1104
# Using a cache helps in 2 ways:
1105
# 1) Avoids unnecessary decoding calls
1106
# 2) Re-uses cached strings, which helps in future set and
1108
# (2) is enough that removing encoding entirely along with
1109
# the cache (so we are using plain strings) results in no
1110
# performance improvement.
1112
revision_id = unescape_revid_cache[revision_id]
1114
unescaped = unescape(revision_id)
1115
unescape_revid_cache[revision_id] = unescaped
1116
revision_id = unescaped
1118
if revision_id in revision_ids:
1120
file_id = unescape_fileid_cache[file_id]
1122
unescaped = unescape(file_id)
1123
unescape_fileid_cache[file_id] = unescaped
1125
setdefault(file_id, set()).add(revision_id)
1128
def fileids_altered_by_revision_ids(self, revision_ids):
1129
"""Find the file ids and versions affected by revisions.
1131
:param revisions: an iterable containing revision ids.
1132
:return: a dictionary mapping altered file-ids to an iterable of
1133
revision_ids. Each altered file-ids has the exact revision_ids that
1134
altered it listed explicitly.
1136
assert self._serializer.support_altered_by_hack, \
1137
("fileids_altered_by_revision_ids only supported for branches "
1138
"which store inventory as unnested xml, not on %r" % self)
1139
selected_revision_ids = set(revision_ids)
1140
w = self.get_inventory_weave()
1141
pb = ui.ui_factory.nested_progress_bar()
1143
return self._find_file_ids_from_xml_inventory_lines(
1144
w.iter_lines_added_or_present_in_versions(
1145
selected_revision_ids, pb=pb),
1146
selected_revision_ids)
1150
def iter_files_bytes(self, desired_files):
1151
"""Iterate through file versions.
1153
Files will not necessarily be returned in the order they occur in
1154
desired_files. No specific order is guaranteed.
1156
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1157
value supplied by the caller as part of desired_files. It should
1158
uniquely identify the file version in the caller's context. (Examples:
1159
an index number or a TreeTransform trans_id.)
1161
bytes_iterator is an iterable of bytestrings for the file. The
1162
kind of iterable and length of the bytestrings are unspecified, but for
1163
this implementation, it is a list of lines produced by
1164
VersionedFile.get_lines().
1166
:param desired_files: a list of (file_id, revision_id, identifier)
1169
transaction = self.get_transaction()
1170
for file_id, revision_id, callable_data in desired_files:
1172
weave = self.weave_store.get_weave(file_id, transaction)
1173
except errors.NoSuchFile:
1174
raise errors.NoSuchIdInRepository(self, file_id)
1175
yield callable_data, weave.get_lines(revision_id)
1177
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1178
"""Get an iterable listing the keys of all the data introduced by a set
1181
The keys will be ordered so that the corresponding items can be safely
1182
fetched and inserted in that order.
1184
:returns: An iterable producing tuples of (knit-kind, file-id,
1185
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1186
'revisions'. file-id is None unless knit-kind is 'file'.
1188
# XXX: it's a bit weird to control the inventory weave caching in this
1189
# generator. Ideally the caching would be done in fetch.py I think. Or
1190
# maybe this generator should explicitly have the contract that it
1191
# should not be iterated until the previously yielded item has been
1194
inv_w = self.get_inventory_weave()
1195
inv_w.enable_cache()
1197
# file ids that changed
1198
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1200
num_file_ids = len(file_ids)
1201
for file_id, altered_versions in file_ids.iteritems():
1202
if _files_pb is not None:
1203
_files_pb.update("fetch texts", count, num_file_ids)
1205
yield ("file", file_id, altered_versions)
1206
# We're done with the files_pb. Note that it finished by the caller,
1207
# just as it was created by the caller.
1211
yield ("inventory", None, revision_ids)
1215
revisions_with_signatures = set()
1216
for rev_id in revision_ids:
1218
self.get_signature_text(rev_id)
1219
except errors.NoSuchRevision:
1223
revisions_with_signatures.add(rev_id)
1225
yield ("signatures", None, revisions_with_signatures)
1228
yield ("revisions", None, revision_ids)
1231
def get_inventory_weave(self):
1232
return self.control_weaves.get_weave('inventory',
1233
self.get_transaction())
1236
def get_inventory(self, revision_id):
1237
"""Get Inventory object by hash."""
1238
return self.deserialise_inventory(
1239
revision_id, self.get_inventory_xml(revision_id))
1241
def deserialise_inventory(self, revision_id, xml):
1242
"""Transform the xml into an inventory object.
1244
:param revision_id: The expected revision id of the inventory.
1245
:param xml: A serialised inventory.
1247
return self._serializer.read_inventory_from_string(xml, revision_id)
1249
def serialise_inventory(self, inv):
1250
return self._serializer.write_inventory_to_string(inv)
1252
def _serialise_inventory_to_lines(self, inv):
1253
return self._serializer.write_inventory_to_lines(inv)
1255
def get_serializer_format(self):
1256
return self._serializer.format_num
1259
def get_inventory_xml(self, revision_id):
1260
"""Get inventory XML as a file object."""
1262
assert isinstance(revision_id, str), type(revision_id)
1263
iw = self.get_inventory_weave()
1264
return iw.get_text(revision_id)
1266
raise errors.HistoryMissing(self, 'inventory', revision_id)
1269
def get_inventory_sha1(self, revision_id):
1270
"""Return the sha1 hash of the inventory entry
1272
return self.get_revision(revision_id).inventory_sha1
1275
def get_revision_graph(self, revision_id=None):
1276
"""Return a dictionary containing the revision graph.
1278
NB: This method should not be used as it accesses the entire graph all
1279
at once, which is much more data than most operations should require.
1281
:param revision_id: The revision_id to get a graph from. If None, then
1282
the entire revision graph is returned. This is a deprecated mode of
1283
operation and will be removed in the future.
1284
:return: a dictionary of revision_id->revision_parents_list.
1286
raise NotImplementedError(self.get_revision_graph)
1289
def get_revision_graph_with_ghosts(self, revision_ids=None):
1290
"""Return a graph of the revisions with ghosts marked as applicable.
1292
:param revision_ids: an iterable of revisions to graph or None for all.
1293
:return: a Graph object with the graph reachable from revision_ids.
1295
if 'evil' in debug.debug_flags:
1297
"get_revision_graph_with_ghosts scales with size of history.")
1298
result = deprecated_graph.Graph()
1299
if not revision_ids:
1300
pending = set(self.all_revision_ids())
1303
pending = set(revision_ids)
1304
# special case NULL_REVISION
1305
if _mod_revision.NULL_REVISION in pending:
1306
pending.remove(_mod_revision.NULL_REVISION)
1307
required = set(pending)
1310
revision_id = pending.pop()
1312
rev = self.get_revision(revision_id)
1313
except errors.NoSuchRevision:
1314
if revision_id in required:
1317
result.add_ghost(revision_id)
1319
for parent_id in rev.parent_ids:
1320
# is this queued or done ?
1321
if (parent_id not in pending and
1322
parent_id not in done):
1324
pending.add(parent_id)
1325
result.add_node(revision_id, rev.parent_ids)
1326
done.add(revision_id)
1329
def _get_history_vf(self):
1330
"""Get a versionedfile whose history graph reflects all revisions.
1332
For weave repositories, this is the inventory weave.
1334
return self.get_inventory_weave()
1336
def iter_reverse_revision_history(self, revision_id):
1337
"""Iterate backwards through revision ids in the lefthand history
1339
:param revision_id: The revision id to start with. All its lefthand
1340
ancestors will be traversed.
1342
if revision_id in (None, _mod_revision.NULL_REVISION):
1344
next_id = revision_id
1345
versionedfile = self._get_history_vf()
1348
parents = versionedfile.get_parents(next_id)
1349
if len(parents) == 0:
1352
next_id = parents[0]
1355
def get_revision_inventory(self, revision_id):
1356
"""Return inventory of a past revision."""
1357
# TODO: Unify this with get_inventory()
1358
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1359
# must be the same as its revision, so this is trivial.
1360
if revision_id is None:
1361
# This does not make sense: if there is no revision,
1362
# then it is the current tree inventory surely ?!
1363
# and thus get_root_id() is something that looks at the last
1364
# commit on the branch, and the get_root_id is an inventory check.
1365
raise NotImplementedError
1366
# return Inventory(self.get_root_id())
1368
return self.get_inventory(revision_id)
1371
def is_shared(self):
1372
"""Return True if this repository is flagged as a shared repository."""
1373
raise NotImplementedError(self.is_shared)
1376
def reconcile(self, other=None, thorough=False):
1377
"""Reconcile this repository."""
1378
from bzrlib.reconcile import RepoReconciler
1379
reconciler = RepoReconciler(self, thorough=thorough)
1380
reconciler.reconcile()
1383
def _refresh_data(self):
1384
"""Helper called from lock_* to ensure coherency with disk.
1386
The default implementation does nothing; it is however possible
1387
for repositories to maintain loaded indices across multiple locks
1388
by checking inside their implementation of this method to see
1389
whether their indices are still valid. This depends of course on
1390
the disk format being validatable in this manner.
1394
def revision_tree(self, revision_id):
1395
"""Return Tree for a revision on this branch.
1397
`revision_id` may be None for the empty tree revision.
1399
# TODO: refactor this to use an existing revision object
1400
# so we don't need to read it in twice.
1401
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1402
return RevisionTree(self, Inventory(root_id=None),
1403
_mod_revision.NULL_REVISION)
1405
inv = self.get_revision_inventory(revision_id)
1406
return RevisionTree(self, inv, revision_id)
1409
def revision_trees(self, revision_ids):
1410
"""Return Tree for a revision on this branch.
1412
`revision_id` may not be None or 'null:'"""
1413
assert None not in revision_ids
1414
assert _mod_revision.NULL_REVISION not in revision_ids
1415
texts = self.get_inventory_weave().get_texts(revision_ids)
1416
for text, revision_id in zip(texts, revision_ids):
1417
inv = self.deserialise_inventory(revision_id, text)
1418
yield RevisionTree(self, inv, revision_id)
1421
def get_ancestry(self, revision_id, topo_sorted=True):
1422
"""Return a list of revision-ids integrated by a revision.
1424
The first element of the list is always None, indicating the origin
1425
revision. This might change when we have history horizons, or
1426
perhaps we should have a new API.
1428
This is topologically sorted.
1430
if _mod_revision.is_null(revision_id):
1432
if not self.has_revision(revision_id):
1433
raise errors.NoSuchRevision(self, revision_id)
1434
w = self.get_inventory_weave()
1435
candidates = w.get_ancestry(revision_id, topo_sorted)
1436
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1439
"""Compress the data within the repository.
1441
This operation only makes sense for some repository types. For other
1442
types it should be a no-op that just returns.
1444
This stub method does not require a lock, but subclasses should use
1445
@needs_write_lock as this is a long running call its reasonable to
1446
implicitly lock for the user.
1450
def print_file(self, file, revision_id):
1451
"""Print `file` to stdout.
1453
FIXME RBC 20060125 as John Meinel points out this is a bad api
1454
- it writes to stdout, it assumes that that is valid etc. Fix
1455
by creating a new more flexible convenience function.
1457
tree = self.revision_tree(revision_id)
1458
# use inventory as it was in that revision
1459
file_id = tree.inventory.path2id(file)
1461
# TODO: jam 20060427 Write a test for this code path
1462
# it had a bug in it, and was raising the wrong
1464
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1465
tree.print_file(file_id)
1467
def get_transaction(self):
1468
return self.control_files.get_transaction()
1470
def revision_parents(self, revision_id):
1471
return self.get_inventory_weave().parent_names(revision_id)
1473
def get_parents(self, revision_ids):
1474
"""See StackedParentsProvider.get_parents"""
1476
for revision_id in revision_ids:
1477
if revision_id == _mod_revision.NULL_REVISION:
1481
parents = self.get_revision(revision_id).parent_ids
1482
except errors.NoSuchRevision:
1485
if len(parents) == 0:
1486
parents = [_mod_revision.NULL_REVISION]
1487
parents_list.append(parents)
1490
def _make_parents_provider(self):
1493
def get_graph(self, other_repository=None):
1494
"""Return the graph walker for this repository format"""
1495
parents_provider = self._make_parents_provider()
1496
if (other_repository is not None and
1497
other_repository.bzrdir.transport.base !=
1498
self.bzrdir.transport.base):
1499
parents_provider = graph._StackedParentsProvider(
1500
[parents_provider, other_repository._make_parents_provider()])
1501
return graph.Graph(parents_provider)
1503
def get_versioned_file_checker(self, revisions, revision_versions_cache):
1504
return VersionedFileChecker(revisions, revision_versions_cache, self)
1507
def set_make_working_trees(self, new_value):
1508
"""Set the policy flag for making working trees when creating branches.
1510
This only applies to branches that use this repository.
1512
The default is 'True'.
1513
:param new_value: True to restore the default, False to disable making
1516
raise NotImplementedError(self.set_make_working_trees)
1518
def make_working_trees(self):
1519
"""Returns the policy for making working trees on new branches."""
1520
raise NotImplementedError(self.make_working_trees)
1523
def sign_revision(self, revision_id, gpg_strategy):
1524
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1525
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1528
def has_signature_for_revision_id(self, revision_id):
1529
"""Query for a revision signature for revision_id in the repository."""
1530
return self._revision_store.has_signature(revision_id,
1531
self.get_transaction())
1534
def get_signature_text(self, revision_id):
1535
"""Return the text for a signature."""
1536
return self._revision_store.get_signature_text(revision_id,
1537
self.get_transaction())
1540
def check(self, revision_ids=None):
1541
"""Check consistency of all history of given revision_ids.
1543
Different repository implementations should override _check().
1545
:param revision_ids: A non-empty list of revision_ids whose ancestry
1546
will be checked. Typically the last revision_id of a branch.
1548
return self._check(revision_ids)
1550
def _check(self, revision_ids):
1551
result = check.Check(self)
1555
def _warn_if_deprecated(self):
1556
global _deprecation_warning_done
1557
if _deprecation_warning_done:
1559
_deprecation_warning_done = True
1560
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1561
% (self._format, self.bzrdir.transport.base))
1563
def supports_rich_root(self):
1564
return self._format.rich_root_data
1566
def _check_ascii_revisionid(self, revision_id, method):
1567
"""Private helper for ascii-only repositories."""
1568
# weave repositories refuse to store revisionids that are non-ascii.
1569
if revision_id is not None:
1570
# weaves require ascii revision ids.
1571
if isinstance(revision_id, unicode):
1573
revision_id.encode('ascii')
1574
except UnicodeEncodeError:
1575
raise errors.NonAsciiRevisionId(method, self)
1578
revision_id.decode('ascii')
1579
except UnicodeDecodeError:
1580
raise errors.NonAsciiRevisionId(method, self)
1582
def revision_graph_can_have_wrong_parents(self):
1583
"""Is it possible for this repository to have a revision graph with
1586
If True, then this repository must also implement
1587
_find_inconsistent_revision_parents so that check and reconcile can
1588
check for inconsistencies before proceeding with other checks that may
1589
depend on the revision index being consistent.
1591
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1593
# remove these delegates a while after bzr 0.15
1594
def __make_delegated(name, from_module):
1595
def _deprecated_repository_forwarder():
1596
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1597
% (name, from_module),
1600
m = __import__(from_module, globals(), locals(), [name])
1602
return getattr(m, name)
1603
except AttributeError:
1604
raise AttributeError('module %s has no name %s'
1606
globals()[name] = _deprecated_repository_forwarder
1609
'AllInOneRepository',
1610
'WeaveMetaDirRepository',
1611
'PreSplitOutRepositoryFormat',
1612
'RepositoryFormat4',
1613
'RepositoryFormat5',
1614
'RepositoryFormat6',
1615
'RepositoryFormat7',
1617
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1621
'RepositoryFormatKnit',
1622
'RepositoryFormatKnit1',
1624
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1627
def install_revision(repository, rev, revision_tree):
1628
"""Install all revision data into a repository."""
1629
repository.start_write_group()
1631
_install_revision(repository, rev, revision_tree)
1633
repository.abort_write_group()
1636
repository.commit_write_group()
1639
def _install_revision(repository, rev, revision_tree):
1640
"""Install all revision data into a repository."""
1641
present_parents = []
1643
for p_id in rev.parent_ids:
1644
if repository.has_revision(p_id):
1645
present_parents.append(p_id)
1646
parent_trees[p_id] = repository.revision_tree(p_id)
1648
parent_trees[p_id] = repository.revision_tree(None)
1650
inv = revision_tree.inventory
1651
entries = inv.iter_entries()
1652
# backwards compatibility hack: skip the root id.
1653
if not repository.supports_rich_root():
1654
path, root = entries.next()
1655
if root.revision != rev.revision_id:
1656
raise errors.IncompatibleRevision(repr(repository))
1657
# Add the texts that are not already present
1658
for path, ie in entries:
1659
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1660
repository.get_transaction())
1661
if ie.revision not in w:
1663
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1664
# with InventoryEntry.find_previous_heads(). if it is, then there
1665
# is a latent bug here where the parents may have ancestors of each
1667
for revision, tree in parent_trees.iteritems():
1668
if ie.file_id not in tree:
1670
parent_id = tree.inventory[ie.file_id].revision
1671
if parent_id in text_parents:
1673
text_parents.append(parent_id)
1675
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1676
repository.get_transaction())
1677
lines = revision_tree.get_file(ie.file_id).readlines()
1678
vfile.add_lines(rev.revision_id, text_parents, lines)
1680
# install the inventory
1681
repository.add_inventory(rev.revision_id, inv, present_parents)
1682
except errors.RevisionAlreadyPresent:
1684
repository.add_revision(rev.revision_id, rev, inv)
1687
class MetaDirRepository(Repository):
1688
"""Repositories in the new meta-dir layout."""
1690
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1691
super(MetaDirRepository, self).__init__(_format,
1697
dir_mode = self.control_files._dir_mode
1698
file_mode = self.control_files._file_mode
1701
def is_shared(self):
1702
"""Return True if this repository is flagged as a shared repository."""
1703
return self.control_files._transport.has('shared-storage')
1706
def set_make_working_trees(self, new_value):
1707
"""Set the policy flag for making working trees when creating branches.
1709
This only applies to branches that use this repository.
1711
The default is 'True'.
1712
:param new_value: True to restore the default, False to disable making
1717
self.control_files._transport.delete('no-working-trees')
1718
except errors.NoSuchFile:
1721
self.control_files.put_utf8('no-working-trees', '')
1723
def make_working_trees(self):
1724
"""Returns the policy for making working trees on new branches."""
1725
return not self.control_files._transport.has('no-working-trees')
1728
class RepositoryFormatRegistry(registry.Registry):
1729
"""Registry of RepositoryFormats."""
1731
def get(self, format_string):
1732
r = registry.Registry.get(self, format_string)
1738
format_registry = RepositoryFormatRegistry()
1739
"""Registry of formats, indexed by their identifying format string.
1741
This can contain either format instances themselves, or classes/factories that
1742
can be called to obtain one.
1746
#####################################################################
1747
# Repository Formats
1749
class RepositoryFormat(object):
1750
"""A repository format.
1752
Formats provide three things:
1753
* An initialization routine to construct repository data on disk.
1754
* a format string which is used when the BzrDir supports versioned
1756
* an open routine which returns a Repository instance.
1758
There is one and only one Format subclass for each on-disk format. But
1759
there can be one Repository subclass that is used for several different
1760
formats. The _format attribute on a Repository instance can be used to
1761
determine the disk format.
1763
Formats are placed in an dict by their format string for reference
1764
during opening. These should be subclasses of RepositoryFormat
1767
Once a format is deprecated, just deprecate the initialize and open
1768
methods on the format class. Do not deprecate the object, as the
1769
object will be created every system load.
1771
Common instance attributes:
1772
_matchingbzrdir - the bzrdir format that the repository format was
1773
originally written to work with. This can be used if manually
1774
constructing a bzrdir and repository, or more commonly for test suite
1778
# Set to True or False in derived classes. True indicates that the format
1779
# supports ghosts gracefully.
1780
supports_ghosts = None
1783
return "<%s>" % self.__class__.__name__
1785
def __eq__(self, other):
1786
# format objects are generally stateless
1787
return isinstance(other, self.__class__)
1789
def __ne__(self, other):
1790
return not self == other
1793
def find_format(klass, a_bzrdir):
1794
"""Return the format for the repository object in a_bzrdir.
1796
This is used by bzr native formats that have a "format" file in
1797
the repository. Other methods may be used by different types of
1801
transport = a_bzrdir.get_repository_transport(None)
1802
format_string = transport.get("format").read()
1803
return format_registry.get(format_string)
1804
except errors.NoSuchFile:
1805
raise errors.NoRepositoryPresent(a_bzrdir)
1807
raise errors.UnknownFormatError(format=format_string)
1810
def register_format(klass, format):
1811
format_registry.register(format.get_format_string(), format)
1814
def unregister_format(klass, format):
1815
format_registry.remove(format.get_format_string())
1818
def get_default_format(klass):
1819
"""Return the current default format."""
1820
from bzrlib import bzrdir
1821
return bzrdir.format_registry.make_bzrdir('default').repository_format
1823
def _get_control_store(self, repo_transport, control_files):
1824
"""Return the control store for this repository."""
1825
raise NotImplementedError(self._get_control_store)
1827
def get_format_string(self):
1828
"""Return the ASCII format string that identifies this format.
1830
Note that in pre format ?? repositories the format string is
1831
not permitted nor written to disk.
1833
raise NotImplementedError(self.get_format_string)
1835
def get_format_description(self):
1836
"""Return the short description for this format."""
1837
raise NotImplementedError(self.get_format_description)
1839
def _get_revision_store(self, repo_transport, control_files):
1840
"""Return the revision store object for this a_bzrdir."""
1841
raise NotImplementedError(self._get_revision_store)
1843
def _get_text_rev_store(self,
1850
"""Common logic for getting a revision store for a repository.
1852
see self._get_revision_store for the subclass-overridable method to
1853
get the store for a repository.
1855
from bzrlib.store.revision.text import TextRevisionStore
1856
dir_mode = control_files._dir_mode
1857
file_mode = control_files._file_mode
1858
text_store = TextStore(transport.clone(name),
1860
compressed=compressed,
1862
file_mode=file_mode)
1863
_revision_store = TextRevisionStore(text_store, serializer)
1864
return _revision_store
1866
# TODO: this shouldn't be in the base class, it's specific to things that
1867
# use weaves or knits -- mbp 20070207
1868
def _get_versioned_file_store(self,
1873
versionedfile_class=None,
1874
versionedfile_kwargs={},
1876
if versionedfile_class is None:
1877
versionedfile_class = self._versionedfile_class
1878
weave_transport = control_files._transport.clone(name)
1879
dir_mode = control_files._dir_mode
1880
file_mode = control_files._file_mode
1881
return VersionedFileStore(weave_transport, prefixed=prefixed,
1883
file_mode=file_mode,
1884
versionedfile_class=versionedfile_class,
1885
versionedfile_kwargs=versionedfile_kwargs,
1888
def initialize(self, a_bzrdir, shared=False):
1889
"""Initialize a repository of this format in a_bzrdir.
1891
:param a_bzrdir: The bzrdir to put the new repository in it.
1892
:param shared: The repository should be initialized as a sharable one.
1893
:returns: The new repository object.
1895
This may raise UninitializableFormat if shared repository are not
1896
compatible the a_bzrdir.
1898
raise NotImplementedError(self.initialize)
1900
def is_supported(self):
1901
"""Is this format supported?
1903
Supported formats must be initializable and openable.
1904
Unsupported formats may not support initialization or committing or
1905
some other features depending on the reason for not being supported.
1909
def check_conversion_target(self, target_format):
1910
raise NotImplementedError(self.check_conversion_target)
1912
def open(self, a_bzrdir, _found=False):
1913
"""Return an instance of this format for the bzrdir a_bzrdir.
1915
_found is a private parameter, do not use it.
1917
raise NotImplementedError(self.open)
1920
class MetaDirRepositoryFormat(RepositoryFormat):
1921
"""Common base class for the new repositories using the metadir layout."""
1923
rich_root_data = False
1924
supports_tree_reference = False
1925
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
1928
super(MetaDirRepositoryFormat, self).__init__()
1930
def _create_control_files(self, a_bzrdir):
1931
"""Create the required files and the initial control_files object."""
1932
# FIXME: RBC 20060125 don't peek under the covers
1933
# NB: no need to escape relative paths that are url safe.
1934
repository_transport = a_bzrdir.get_repository_transport(self)
1935
control_files = lockable_files.LockableFiles(repository_transport,
1936
'lock', lockdir.LockDir)
1937
control_files.create_lock()
1938
return control_files
1940
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
1941
"""Upload the initial blank content."""
1942
control_files = self._create_control_files(a_bzrdir)
1943
control_files.lock_write()
1945
control_files._transport.mkdir_multi(dirs,
1946
mode=control_files._dir_mode)
1947
for file, content in files:
1948
control_files.put(file, content)
1949
for file, content in utf8_files:
1950
control_files.put_utf8(file, content)
1952
control_files.put_utf8('shared-storage', '')
1954
control_files.unlock()
1957
# formats which have no format string are not discoverable
1958
# and not independently creatable, so are not registered. They're
1959
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1960
# needed, it's constructed directly by the BzrDir. Non-native formats where
1961
# the repository is not separately opened are similar.
1963
format_registry.register_lazy(
1964
'Bazaar-NG Repository format 7',
1965
'bzrlib.repofmt.weaverepo',
1969
# KEEP in sync with bzrdir.format_registry default, which controls the overall
1970
# default control directory format
1971
format_registry.register_lazy(
1972
'Bazaar-NG Knit Repository Format 1',
1973
'bzrlib.repofmt.knitrepo',
1974
'RepositoryFormatKnit1',
1976
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
1978
format_registry.register_lazy(
1979
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
1980
'bzrlib.repofmt.knitrepo',
1981
'RepositoryFormatKnit3',
1984
# Pack-based formats. There is one format for pre-subtrees, and one for
1985
# post-subtrees to allow ease of testing.
1986
# NOTE: These are experimental in 0.92.
1987
format_registry.register_lazy(
1988
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
1989
'bzrlib.repofmt.pack_repo',
1990
'RepositoryFormatKnitPack1',
1992
format_registry.register_lazy(
1993
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
1994
'bzrlib.repofmt.pack_repo',
1995
'RepositoryFormatKnitPack3',
1999
class InterRepository(InterObject):
2000
"""This class represents operations taking place between two repositories.
2002
Its instances have methods like copy_content and fetch, and contain
2003
references to the source and target repositories these operations can be
2006
Often we will provide convenience methods on 'repository' which carry out
2007
operations with another repository - they will always forward to
2008
InterRepository.get(other).method_name(parameters).
2012
"""The available optimised InterRepository types."""
2014
def copy_content(self, revision_id=None):
2015
raise NotImplementedError(self.copy_content)
2017
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2018
"""Fetch the content required to construct revision_id.
2020
The content is copied from self.source to self.target.
2022
:param revision_id: if None all content is copied, if NULL_REVISION no
2024
:param pb: optional progress bar to use for progress reports. If not
2025
provided a default one will be created.
2027
Returns the copied revision count and the failed revisions in a tuple:
2030
raise NotImplementedError(self.fetch)
2033
def missing_revision_ids(self, revision_id=None):
2034
"""Return the revision ids that source has that target does not.
2036
These are returned in topological order.
2038
:param revision_id: only return revision ids included by this
2041
# generic, possibly worst case, slow code path.
2042
target_ids = set(self.target.all_revision_ids())
2043
if revision_id is not None:
2044
source_ids = self.source.get_ancestry(revision_id)
2045
assert source_ids[0] is None
2048
source_ids = self.source.all_revision_ids()
2049
result_set = set(source_ids).difference(target_ids)
2050
# this may look like a no-op: its not. It preserves the ordering
2051
# other_ids had while only returning the members from other_ids
2052
# that we've decided we need.
2053
return [rev_id for rev_id in source_ids if rev_id in result_set]
2056
def _same_model(source, target):
2057
"""True if source and target have the same data representation."""
2058
if source.supports_rich_root() != target.supports_rich_root():
2060
if source._serializer != target._serializer:
2065
class InterSameDataRepository(InterRepository):
2066
"""Code for converting between repositories that represent the same data.
2068
Data format and model must match for this to work.
2072
def _get_repo_format_to_test(self):
2073
"""Repository format for testing with.
2075
InterSameData can pull from subtree to subtree and from non-subtree to
2076
non-subtree, so we test this with the richest repository format.
2078
from bzrlib.repofmt import knitrepo
2079
return knitrepo.RepositoryFormatKnit3()
2082
def is_compatible(source, target):
2083
return InterRepository._same_model(source, target)
2086
def copy_content(self, revision_id=None):
2087
"""Make a complete copy of the content in self into destination.
2089
This copies both the repository's revision data, and configuration information
2090
such as the make_working_trees setting.
2092
This is a destructive operation! Do not use it on existing
2095
:param revision_id: Only copy the content needed to construct
2096
revision_id and its parents.
2099
self.target.set_make_working_trees(self.source.make_working_trees())
2100
except NotImplementedError:
2102
# but don't bother fetching if we have the needed data now.
2103
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2104
self.target.has_revision(revision_id)):
2106
self.target.fetch(self.source, revision_id=revision_id)
2109
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2110
"""See InterRepository.fetch()."""
2111
from bzrlib.fetch import GenericRepoFetcher
2112
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2113
self.source, self.source._format, self.target,
2114
self.target._format)
2115
f = GenericRepoFetcher(to_repository=self.target,
2116
from_repository=self.source,
2117
last_revision=revision_id,
2119
return f.count_copied, f.failed_revisions
2122
class InterWeaveRepo(InterSameDataRepository):
2123
"""Optimised code paths between Weave based repositories.
2125
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2126
implemented lazy inter-object optimisation.
2130
def _get_repo_format_to_test(self):
2131
from bzrlib.repofmt import weaverepo
2132
return weaverepo.RepositoryFormat7()
2135
def is_compatible(source, target):
2136
"""Be compatible with known Weave formats.
2138
We don't test for the stores being of specific types because that
2139
could lead to confusing results, and there is no need to be
2142
from bzrlib.repofmt.weaverepo import (
2148
return (isinstance(source._format, (RepositoryFormat5,
2150
RepositoryFormat7)) and
2151
isinstance(target._format, (RepositoryFormat5,
2153
RepositoryFormat7)))
2154
except AttributeError:
2158
def copy_content(self, revision_id=None):
2159
"""See InterRepository.copy_content()."""
2160
# weave specific optimised path:
2162
self.target.set_make_working_trees(self.source.make_working_trees())
2163
except NotImplementedError:
2165
# FIXME do not peek!
2166
if self.source.control_files._transport.listable():
2167
pb = ui.ui_factory.nested_progress_bar()
2169
self.target.weave_store.copy_all_ids(
2170
self.source.weave_store,
2172
from_transaction=self.source.get_transaction(),
2173
to_transaction=self.target.get_transaction())
2174
pb.update('copying inventory', 0, 1)
2175
self.target.control_weaves.copy_multi(
2176
self.source.control_weaves, ['inventory'],
2177
from_transaction=self.source.get_transaction(),
2178
to_transaction=self.target.get_transaction())
2179
self.target._revision_store.text_store.copy_all_ids(
2180
self.source._revision_store.text_store,
2185
self.target.fetch(self.source, revision_id=revision_id)
2188
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2189
"""See InterRepository.fetch()."""
2190
from bzrlib.fetch import GenericRepoFetcher
2191
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2192
self.source, self.source._format, self.target, self.target._format)
2193
f = GenericRepoFetcher(to_repository=self.target,
2194
from_repository=self.source,
2195
last_revision=revision_id,
2197
return f.count_copied, f.failed_revisions
2200
def missing_revision_ids(self, revision_id=None):
2201
"""See InterRepository.missing_revision_ids()."""
2202
# we want all revisions to satisfy revision_id in source.
2203
# but we don't want to stat every file here and there.
2204
# we want then, all revisions other needs to satisfy revision_id
2205
# checked, but not those that we have locally.
2206
# so the first thing is to get a subset of the revisions to
2207
# satisfy revision_id in source, and then eliminate those that
2208
# we do already have.
2209
# this is slow on high latency connection to self, but as as this
2210
# disk format scales terribly for push anyway due to rewriting
2211
# inventory.weave, this is considered acceptable.
2213
if revision_id is not None:
2214
source_ids = self.source.get_ancestry(revision_id)
2215
assert source_ids[0] is None
2218
source_ids = self.source._all_possible_ids()
2219
source_ids_set = set(source_ids)
2220
# source_ids is the worst possible case we may need to pull.
2221
# now we want to filter source_ids against what we actually
2222
# have in target, but don't try to check for existence where we know
2223
# we do not have a revision as that would be pointless.
2224
target_ids = set(self.target._all_possible_ids())
2225
possibly_present_revisions = target_ids.intersection(source_ids_set)
2226
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2227
required_revisions = source_ids_set.difference(actually_present_revisions)
2228
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2229
if revision_id is not None:
2230
# we used get_ancestry to determine source_ids then we are assured all
2231
# revisions referenced are present as they are installed in topological order.
2232
# and the tip revision was validated by get_ancestry.
2233
return required_topo_revisions
2235
# if we just grabbed the possibly available ids, then
2236
# we only have an estimate of whats available and need to validate
2237
# that against the revision records.
2238
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2241
class InterKnitRepo(InterSameDataRepository):
2242
"""Optimised code paths between Knit based repositories."""
2245
def _get_repo_format_to_test(self):
2246
from bzrlib.repofmt import knitrepo
2247
return knitrepo.RepositoryFormatKnit1()
2250
def is_compatible(source, target):
2251
"""Be compatible with known Knit formats.
2253
We don't test for the stores being of specific types because that
2254
could lead to confusing results, and there is no need to be
2257
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2259
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2260
isinstance(target._format, RepositoryFormatKnit))
2261
except AttributeError:
2263
return are_knits and InterRepository._same_model(source, target)
2266
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2267
"""See InterRepository.fetch()."""
2268
from bzrlib.fetch import KnitRepoFetcher
2269
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2270
self.source, self.source._format, self.target, self.target._format)
2271
f = KnitRepoFetcher(to_repository=self.target,
2272
from_repository=self.source,
2273
last_revision=revision_id,
2275
return f.count_copied, f.failed_revisions
2278
def missing_revision_ids(self, revision_id=None):
2279
"""See InterRepository.missing_revision_ids()."""
2280
if revision_id is not None:
2281
source_ids = self.source.get_ancestry(revision_id)
2282
assert source_ids[0] is None
2285
source_ids = self.source.all_revision_ids()
2286
source_ids_set = set(source_ids)
2287
# source_ids is the worst possible case we may need to pull.
2288
# now we want to filter source_ids against what we actually
2289
# have in target, but don't try to check for existence where we know
2290
# we do not have a revision as that would be pointless.
2291
target_ids = set(self.target.all_revision_ids())
2292
possibly_present_revisions = target_ids.intersection(source_ids_set)
2293
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2294
required_revisions = source_ids_set.difference(actually_present_revisions)
2295
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2296
if revision_id is not None:
2297
# we used get_ancestry to determine source_ids then we are assured all
2298
# revisions referenced are present as they are installed in topological order.
2299
# and the tip revision was validated by get_ancestry.
2300
return required_topo_revisions
2302
# if we just grabbed the possibly available ids, then
2303
# we only have an estimate of whats available and need to validate
2304
# that against the revision records.
2305
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2308
class InterPackRepo(InterSameDataRepository):
2309
"""Optimised code paths between Pack based repositories."""
2312
def _get_repo_format_to_test(self):
2313
from bzrlib.repofmt import pack_repo
2314
return pack_repo.RepositoryFormatKnitPack1()
2317
def is_compatible(source, target):
2318
"""Be compatible with known Pack formats.
2320
We don't test for the stores being of specific types because that
2321
could lead to confusing results, and there is no need to be
2324
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2326
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2327
isinstance(target._format, RepositoryFormatPack))
2328
except AttributeError:
2330
return are_packs and InterRepository._same_model(source, target)
2333
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2334
"""See InterRepository.fetch()."""
2335
from bzrlib.repofmt.pack_repo import Packer
2336
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2337
self.source, self.source._format, self.target, self.target._format)
2338
self.count_copied = 0
2339
if revision_id is None:
2341
# everything to do - use pack logic
2342
# to fetch from all packs to one without
2343
# inventory parsing etc, IFF nothing to be copied is in the target.
2345
revision_ids = self.source.all_revision_ids()
2346
# implementing the TODO will involve:
2347
# - detecting when all of a pack is selected
2348
# - avoiding as much as possible pre-selection, so the
2349
# more-core routines such as create_pack_from_packs can filter in
2350
# a just-in-time fashion. (though having a HEADS list on a
2351
# repository might make this a lot easier, because we could
2352
# sensibly detect 'new revisions' without doing a full index scan.
2353
elif _mod_revision.is_null(revision_id):
2358
revision_ids = self.missing_revision_ids(revision_id,
2359
find_ghosts=find_ghosts)
2360
except errors.NoSuchRevision:
2361
raise errors.InstallFailed([revision_id])
2362
packs = self.source._pack_collection.all_packs()
2363
pack = Packer(self.target._pack_collection, packs, '.fetch',
2364
revision_ids).pack()
2365
if pack is not None:
2366
self.target._pack_collection._save_pack_names()
2367
# Trigger an autopack. This may duplicate effort as we've just done
2368
# a pack creation, but for now it is simpler to think about as
2369
# 'upload data, then repack if needed'.
2370
self.target._pack_collection.autopack()
2371
return pack.get_revision_count()
2376
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2377
"""See InterRepository.missing_revision_ids().
2379
:param find_ghosts: Find ghosts throughough the ancestry of
2382
if not find_ghosts and revision_id is not None:
2383
graph = self.source.get_graph()
2384
missing_revs = set()
2385
searcher = graph._make_breadth_first_searcher([revision_id])
2387
self.target._pack_collection.revision_index.combined_index
2388
null_set = frozenset([_mod_revision.NULL_REVISION])
2391
next_revs = set(searcher.next())
2392
except StopIteration:
2394
next_revs.difference_update(null_set)
2395
target_keys = [(key,) for key in next_revs]
2396
have_revs = frozenset(node[1][0] for node in
2397
target_index.iter_entries(target_keys))
2398
missing_revs.update(next_revs - have_revs)
2399
searcher.stop_searching_any(have_revs)
2401
elif revision_id is not None:
2402
source_ids = self.source.get_ancestry(revision_id)
2403
assert source_ids[0] is None
2406
source_ids = self.source.all_revision_ids()
2407
# source_ids is the worst possible case we may need to pull.
2408
# now we want to filter source_ids against what we actually
2409
# have in target, but don't try to check for existence where we know
2410
# we do not have a revision as that would be pointless.
2411
target_ids = set(self.target.all_revision_ids())
2412
return [r for r in source_ids if (r not in target_ids)]
2415
class InterModel1and2(InterRepository):
2418
def _get_repo_format_to_test(self):
2422
def is_compatible(source, target):
2423
if not source.supports_rich_root() and target.supports_rich_root():
2429
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2430
"""See InterRepository.fetch()."""
2431
from bzrlib.fetch import Model1toKnit2Fetcher
2432
f = Model1toKnit2Fetcher(to_repository=self.target,
2433
from_repository=self.source,
2434
last_revision=revision_id,
2436
return f.count_copied, f.failed_revisions
2439
def copy_content(self, revision_id=None):
2440
"""Make a complete copy of the content in self into destination.
2442
This is a destructive operation! Do not use it on existing
2445
:param revision_id: Only copy the content needed to construct
2446
revision_id and its parents.
2449
self.target.set_make_working_trees(self.source.make_working_trees())
2450
except NotImplementedError:
2452
# but don't bother fetching if we have the needed data now.
2453
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2454
self.target.has_revision(revision_id)):
2456
self.target.fetch(self.source, revision_id=revision_id)
2459
class InterKnit1and2(InterKnitRepo):
2462
def _get_repo_format_to_test(self):
2466
def is_compatible(source, target):
2467
"""Be compatible with Knit1 source and Knit3 target"""
2468
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2470
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2471
RepositoryFormatKnit3)
2472
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2473
RepositoryFormatKnitPack3)
2474
return (isinstance(source._format,
2475
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2476
isinstance(target._format,
2477
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2479
except AttributeError:
2483
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2484
"""See InterRepository.fetch()."""
2485
from bzrlib.fetch import Knit1to2Fetcher
2486
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2487
self.source, self.source._format, self.target,
2488
self.target._format)
2489
f = Knit1to2Fetcher(to_repository=self.target,
2490
from_repository=self.source,
2491
last_revision=revision_id,
2493
return f.count_copied, f.failed_revisions
2496
class InterRemoteToOther(InterRepository):
2498
def __init__(self, source, target):
2499
InterRepository.__init__(self, source, target)
2500
self._real_inter = None
2503
def is_compatible(source, target):
2504
if not isinstance(source, remote.RemoteRepository):
2506
source._ensure_real()
2507
real_source = source._real_repository
2508
# Is source's model compatible with target's model, and are they the
2509
# same format? Currently we can only optimise fetching from an
2510
# identical model & format repo.
2511
assert not isinstance(real_source, remote.RemoteRepository), (
2512
"We don't support remote repos backed by remote repos yet.")
2513
return real_source._format == target._format
2516
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2517
"""See InterRepository.fetch()."""
2518
from bzrlib.fetch import RemoteToOtherFetcher
2519
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2520
self.source, self.target, self.target._format)
2521
# TODO: jam 20070210 This should be an assert, not a translate
2522
revision_id = osutils.safe_revision_id(revision_id)
2523
f = RemoteToOtherFetcher(to_repository=self.target,
2524
from_repository=self.source,
2525
last_revision=revision_id,
2527
return f.count_copied, f.failed_revisions
2530
def _get_repo_format_to_test(self):
2534
class InterOtherToRemote(InterRepository):
2536
def __init__(self, source, target):
2537
InterRepository.__init__(self, source, target)
2538
self._real_inter = None
2541
def is_compatible(source, target):
2542
if isinstance(target, remote.RemoteRepository):
2546
def _ensure_real_inter(self):
2547
if self._real_inter is None:
2548
self.target._ensure_real()
2549
real_target = self.target._real_repository
2550
self._real_inter = InterRepository.get(self.source, real_target)
2552
def copy_content(self, revision_id=None):
2553
self._ensure_real_inter()
2554
self._real_inter.copy_content(revision_id=revision_id)
2556
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2557
self._ensure_real_inter()
2558
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2561
def _get_repo_format_to_test(self):
2565
InterRepository.register_optimiser(InterSameDataRepository)
2566
InterRepository.register_optimiser(InterWeaveRepo)
2567
InterRepository.register_optimiser(InterKnitRepo)
2568
InterRepository.register_optimiser(InterModel1and2)
2569
InterRepository.register_optimiser(InterKnit1and2)
2570
InterRepository.register_optimiser(InterPackRepo)
2571
InterRepository.register_optimiser(InterRemoteToOther)
2572
InterRepository.register_optimiser(InterOtherToRemote)
2575
class CopyConverter(object):
2576
"""A repository conversion tool which just performs a copy of the content.
2578
This is slow but quite reliable.
2581
def __init__(self, target_format):
2582
"""Create a CopyConverter.
2584
:param target_format: The format the resulting repository should be.
2586
self.target_format = target_format
2588
def convert(self, repo, pb):
2589
"""Perform the conversion of to_convert, giving feedback via pb.
2591
:param to_convert: The disk object to convert.
2592
:param pb: a progress bar to use for progress information.
2597
# this is only useful with metadir layouts - separated repo content.
2598
# trigger an assertion if not such
2599
repo._format.get_format_string()
2600
self.repo_dir = repo.bzrdir
2601
self.step('Moving repository to repository.backup')
2602
self.repo_dir.transport.move('repository', 'repository.backup')
2603
backup_transport = self.repo_dir.transport.clone('repository.backup')
2604
repo._format.check_conversion_target(self.target_format)
2605
self.source_repo = repo._format.open(self.repo_dir,
2607
_override_transport=backup_transport)
2608
self.step('Creating new repository')
2609
converted = self.target_format.initialize(self.repo_dir,
2610
self.source_repo.is_shared())
2611
converted.lock_write()
2613
self.step('Copying content into repository.')
2614
self.source_repo.copy_content_into(converted)
2617
self.step('Deleting old repository content.')
2618
self.repo_dir.transport.delete_tree('repository.backup')
2619
self.pb.note('repository converted')
2621
def step(self, message):
2622
"""Update the pb by a step."""
2624
self.pb.update(message, self.count, self.total)
2636
def _unescaper(match, _map=_unescape_map):
2637
code = match.group(1)
2641
if not code.startswith('#'):
2643
return unichr(int(code[1:])).encode('utf8')
2649
def _unescape_xml(data):
2650
"""Unescape predefined XML entities in a string of data."""
2652
if _unescape_re is None:
2653
_unescape_re = re.compile('\&([^;]*);')
2654
return _unescape_re.sub(_unescaper, data)
2657
class _RevisionTextVersionCache(object):
2658
"""A cache of the versionedfile versions for revision and file-id."""
2660
def __init__(self, repository):
2661
self.repository = repository
2662
self.revision_versions = {}
2663
self.revision_parents = {}
2664
self.repo_graph = self.repository.get_graph()
2665
# XXX: RBC: I haven't tracked down what uses this, but it would be
2666
# better to use the headscache directly I think.
2667
self.heads = graph.HeadsCache(self.repo_graph).heads
2669
def add_revision_text_versions(self, tree):
2670
"""Cache text version data from the supplied revision tree"""
2672
for path, entry in tree.iter_entries_by_dir():
2673
inv_revisions[entry.file_id] = entry.revision
2674
self.revision_versions[tree.get_revision_id()] = inv_revisions
2675
return inv_revisions
2677
def get_text_version(self, file_id, revision_id):
2678
"""Determine the text version for a given file-id and revision-id"""
2680
inv_revisions = self.revision_versions[revision_id]
2683
tree = self.repository.revision_tree(revision_id)
2684
except errors.RevisionNotPresent:
2685
self.revision_versions[revision_id] = inv_revisions = {}
2687
inv_revisions = self.add_revision_text_versions(tree)
2688
return inv_revisions.get(file_id)
2690
def prepopulate_revs(self, revision_ids):
2691
# Filter out versions that we don't have an inventory for, so that the
2692
# revision_trees() call won't fail.
2693
inv_weave = self.repository.get_inventory_weave()
2694
revs = [r for r in revision_ids if inv_weave.has_version(r)]
2695
# XXX: this loop is very similar to
2696
# bzrlib.fetch.Inter1and2Helper.iter_rev_trees.
2698
mutter('%d revisions left to prepopulate', len(revs))
2699
for tree in self.repository.revision_trees(revs[:100]):
2700
if tree.inventory.revision_id is None:
2701
tree.inventory.revision_id = tree.get_revision_id()
2702
self.add_revision_text_versions(tree)
2705
def get_parents(self, revision_id):
2707
return self.revision_parents[revision_id]
2709
parents = self.repository.get_parents([revision_id])[0]
2710
self.revision_parents[revision_id] = parents
2713
def used_file_versions(self):
2714
"""Return a set of (revision_id, file_id) pairs for each file version
2715
referenced by any inventory cached by this _RevisionTextVersionCache.
2717
If the entire repository has been cached, this can be used to find all
2718
file versions that are actually referenced by inventories. Thus any
2719
other file version is completely unused and can be removed safely.
2722
for inventory_summary in self.revision_versions.itervalues():
2723
result.update(inventory_summary.items())
2727
class VersionedFileChecker(object):
2729
def __init__(self, planned_revisions, revision_versions, repository):
2730
self.planned_revisions = planned_revisions
2731
self.revision_versions = revision_versions
2732
self.repository = repository
2734
def calculate_file_version_parents(self, revision_id, file_id):
2735
"""Calculate the correct parents for a file version according to
2738
text_revision = self.revision_versions.get_text_version(
2739
file_id, revision_id)
2740
if text_revision is None:
2742
parents_of_text_revision = self.revision_versions.get_parents(
2744
parents_from_inventories = []
2745
for parent in parents_of_text_revision:
2746
if parent == _mod_revision.NULL_REVISION:
2748
introduced_in = self.revision_versions.get_text_version(file_id,
2750
if introduced_in is not None:
2751
parents_from_inventories.append(introduced_in)
2752
heads = set(self.revision_versions.heads(parents_from_inventories))
2754
for parent in parents_from_inventories:
2755
if parent in heads and parent not in new_parents:
2756
new_parents.append(parent)
2757
return tuple(new_parents)
2759
def check_file_version_parents(self, weave, file_id):
2760
"""Check the parents stored in a versioned file are correct.
2762
It also detects file versions that are not referenced by their
2763
corresponding revision's inventory.
2765
:returns: A tuple of (wrong_parents, dangling_file_versions).
2766
wrong_parents is a dict mapping {revision_id: (stored_parents,
2767
correct_parents)} for each revision_id where the stored parents
2768
are not correct. dangling_file_versions is a set of (file_id,
2769
revision_id) tuples for versions that are present in this versioned
2770
file, but not used by the corresponding inventory.
2773
dangling_file_versions = set()
2774
for num, revision_id in enumerate(self.planned_revisions):
2775
correct_parents = self.calculate_file_version_parents(
2776
revision_id, file_id)
2777
if correct_parents is None:
2779
text_revision = self.revision_versions.get_text_version(
2780
file_id, revision_id)
2782
knit_parents = tuple(weave.get_parents(revision_id))
2783
except errors.RevisionNotPresent:
2785
if text_revision != revision_id:
2786
# This file version is not referenced by its corresponding
2788
dangling_file_versions.add((file_id, revision_id))
2789
if correct_parents != knit_parents:
2790
wrong_parents[revision_id] = (knit_parents, correct_parents)
2791
return wrong_parents, dangling_file_versions
added
removed
bzrlib/repository.py
