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(), """
40
revision as _mod_revision,
46
from bzrlib.bundle import serializer
47
from bzrlib.revisiontree import RevisionTree
48
from bzrlib.store.versioned import VersionedFileStore
49
from bzrlib.store.text import TextStore
50
from bzrlib.testament import Testament
51
from bzrlib.util import bencode
54
from bzrlib.decorators import needs_read_lock, needs_write_lock
55
from bzrlib.inter import InterObject
56
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
57
from bzrlib.symbol_versioning import (
60
from bzrlib.trace import mutter, mutter_callsite, note, warning
63
# Old formats display a warning, but only once
64
_deprecation_warning_done = False
67
class CommitBuilder(object):
68
"""Provides an interface to build up a commit.
70
This allows describing a tree to be committed without needing to
71
know the internals of the format of the repository.
74
# all clients should supply tree roots.
75
record_root_entry = True
76
# the default CommitBuilder does not manage trees whose root is versioned.
77
_versioned_root = False
79
def __init__(self, repository, parents, config, timestamp=None,
80
timezone=None, committer=None, revprops=None,
82
"""Initiate a CommitBuilder.
84
:param repository: Repository to commit to.
85
:param parents: Revision ids of the parents of the new revision.
86
:param config: Configuration to use.
87
:param timestamp: Optional timestamp recorded for commit.
88
:param timezone: Optional timezone for timestamp.
89
:param committer: Optional committer to set for commit.
90
:param revprops: Optional dictionary of revision properties.
91
:param revision_id: Optional revision id.
96
self._committer = self._config.username()
98
assert isinstance(committer, basestring), type(committer)
99
self._committer = committer
101
self.new_inventory = Inventory(None)
102
self._new_revision_id = revision_id
103
self.parents = parents
104
self.repository = repository
107
if revprops is not None:
108
self._revprops.update(revprops)
110
if timestamp is None:
111
timestamp = time.time()
112
# Restrict resolution to 1ms
113
self._timestamp = round(timestamp, 3)
116
self._timezone = osutils.local_time_offset()
118
self._timezone = int(timezone)
120
self._generate_revision_if_needed()
121
self.__heads = graph.HeadsCache(repository.get_graph()).heads
123
def commit(self, message):
124
"""Make the actual commit.
126
:return: The revision id of the recorded revision.
128
rev = _mod_revision.Revision(
129
timestamp=self._timestamp,
130
timezone=self._timezone,
131
committer=self._committer,
133
inventory_sha1=self.inv_sha1,
134
revision_id=self._new_revision_id,
135
properties=self._revprops)
136
rev.parent_ids = self.parents
137
self.repository.add_revision(self._new_revision_id, rev,
138
self.new_inventory, self._config)
139
self.repository.commit_write_group()
140
return self._new_revision_id
143
"""Abort the commit that is being built.
145
self.repository.abort_write_group()
147
def revision_tree(self):
148
"""Return the tree that was just committed.
150
After calling commit() this can be called to get a RevisionTree
151
representing the newly committed tree. This is preferred to
152
calling Repository.revision_tree() because that may require
153
deserializing the inventory, while we already have a copy in
156
return RevisionTree(self.repository, self.new_inventory,
157
self._new_revision_id)
159
def finish_inventory(self):
160
"""Tell the builder that the inventory is finished."""
161
if self.new_inventory.root is None:
162
raise AssertionError('Root entry should be supplied to'
163
' record_entry_contents, as of bzr 0.10.',
164
DeprecationWarning, stacklevel=2)
165
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
166
self.new_inventory.revision_id = self._new_revision_id
167
self.inv_sha1 = self.repository.add_inventory(
168
self._new_revision_id,
173
def _gen_revision_id(self):
174
"""Return new revision-id."""
175
return generate_ids.gen_revision_id(self._config.username(),
178
def _generate_revision_if_needed(self):
179
"""Create a revision id if None was supplied.
181
If the repository can not support user-specified revision ids
182
they should override this function and raise CannotSetRevisionId
183
if _new_revision_id is not None.
185
:raises: CannotSetRevisionId
187
if self._new_revision_id is None:
188
self._new_revision_id = self._gen_revision_id()
189
self.random_revid = True
191
self.random_revid = False
193
def _heads(self, file_id, revision_ids):
194
"""Calculate the graph heads for revision_ids in the graph of file_id.
196
This can use either a per-file graph or a global revision graph as we
197
have an identity relationship between the two graphs.
199
return self.__heads(revision_ids)
201
def _check_root(self, ie, parent_invs, tree):
202
"""Helper for record_entry_contents.
204
:param ie: An entry being added.
205
:param parent_invs: The inventories of the parent revisions of the
207
:param tree: The tree that is being committed.
209
# In this revision format, root entries have no knit or weave When
210
# serializing out to disk and back in root.revision is always
212
ie.revision = self._new_revision_id
214
def _get_delta(self, ie, basis_inv, path):
215
"""Get a delta against the basis inventory for ie."""
216
if ie.file_id not in basis_inv:
218
return (None, path, ie.file_id, ie)
219
elif ie != basis_inv[ie.file_id]:
221
# TODO: avoid tis id2path call.
222
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
227
def record_entry_contents(self, ie, parent_invs, path, tree,
229
"""Record the content of ie from tree into the commit if needed.
231
Side effect: sets ie.revision when unchanged
233
:param ie: An inventory entry present in the commit.
234
:param parent_invs: The inventories of the parent revisions of the
236
:param path: The path the entry is at in the tree.
237
:param tree: The tree which contains this entry and should be used to
239
:param content_summary: Summary data from the tree about the paths
240
content - stat, length, exec, sha/link target. This is only
241
accessed when the entry has a revision of None - that is when it is
242
a candidate to commit.
243
:return: A tuple (change_delta, version_recorded). change_delta is
244
an inventory_delta change for this entry against the basis tree of
245
the commit, or None if no change occured against the basis tree.
246
version_recorded is True if a new version of the entry has been
247
recorded. For instance, committing a merge where a file was only
248
changed on the other side will return (delta, False).
250
if self.new_inventory.root is None:
251
if ie.parent_id is not None:
252
raise errors.RootMissing()
253
self._check_root(ie, parent_invs, tree)
254
if ie.revision is None:
255
kind = content_summary[0]
257
# ie is carried over from a prior commit
259
# XXX: repository specific check for nested tree support goes here - if
260
# the repo doesn't want nested trees we skip it ?
261
if (kind == 'tree-reference' and
262
not self.repository._format.supports_tree_reference):
263
# mismatch between commit builder logic and repository:
264
# this needs the entry creation pushed down into the builder.
265
raise NotImplementedError('Missing repository subtree support.')
266
self.new_inventory.add(ie)
268
# TODO: slow, take it out of the inner loop.
270
basis_inv = parent_invs[0]
272
basis_inv = Inventory(root_id=None)
274
# ie.revision is always None if the InventoryEntry is considered
275
# for committing. We may record the previous parents revision if the
276
# content is actually unchanged against a sole head.
277
if ie.revision is not None:
278
if not self._versioned_root and path == '':
279
# repositories that do not version the root set the root's
280
# revision to the new commit even when no change occurs, and
281
# this masks when a change may have occurred against the basis,
282
# so calculate if one happened.
283
if ie.file_id in basis_inv:
284
delta = (basis_inv.id2path(ie.file_id), path,
288
delta = (None, path, ie.file_id, ie)
291
# we don't need to commit this, because the caller already
292
# determined that an existing revision of this file is
294
return None, (ie.revision == self._new_revision_id)
295
# XXX: Friction: parent_candidates should return a list not a dict
296
# so that we don't have to walk the inventories again.
297
parent_candiate_entries = ie.parent_candidates(parent_invs)
298
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
300
for inv in parent_invs:
301
if ie.file_id in inv:
302
old_rev = inv[ie.file_id].revision
303
if old_rev in head_set:
304
heads.append(inv[ie.file_id].revision)
305
head_set.remove(inv[ie.file_id].revision)
308
# now we check to see if we need to write a new record to the
310
# We write a new entry unless there is one head to the ancestors, and
311
# the kind-derived content is unchanged.
313
# Cheapest check first: no ancestors, or more the one head in the
314
# ancestors, we write a new node.
318
# There is a single head, look it up for comparison
319
parent_entry = parent_candiate_entries[heads[0]]
320
# if the non-content specific data has changed, we'll be writing a
322
if (parent_entry.parent_id != ie.parent_id or
323
parent_entry.name != ie.name):
325
# now we need to do content specific checks:
327
# if the kind changed the content obviously has
328
if kind != parent_entry.kind:
331
assert content_summary[2] is not None, \
332
"Files must not have executable = None"
334
if (# if the file length changed we have to store:
335
parent_entry.text_size != content_summary[1] or
336
# if the exec bit has changed we have to store:
337
parent_entry.executable != content_summary[2]):
339
elif parent_entry.text_sha1 == content_summary[3]:
340
# all meta and content is unchanged (using a hash cache
341
# hit to check the sha)
342
ie.revision = parent_entry.revision
343
ie.text_size = parent_entry.text_size
344
ie.text_sha1 = parent_entry.text_sha1
345
ie.executable = parent_entry.executable
346
return self._get_delta(ie, basis_inv, path), False
348
# Either there is only a hash change(no hash cache entry,
349
# or same size content change), or there is no change on
351
# Provide the parent's hash to the store layer, so that the
352
# content is unchanged we will not store a new node.
353
nostore_sha = parent_entry.text_sha1
355
# We want to record a new node regardless of the presence or
356
# absence of a content change in the file.
358
ie.executable = content_summary[2]
359
lines = tree.get_file(ie.file_id, path).readlines()
361
ie.text_sha1, ie.text_size = self._add_text_to_weave(
362
ie.file_id, lines, heads, nostore_sha)
363
except errors.ExistingContent:
364
# Turns out that the file content was unchanged, and we were
365
# only going to store a new node if it was changed. Carry over
367
ie.revision = parent_entry.revision
368
ie.text_size = parent_entry.text_size
369
ie.text_sha1 = parent_entry.text_sha1
370
ie.executable = parent_entry.executable
371
return self._get_delta(ie, basis_inv, path), False
372
elif kind == 'directory':
374
# all data is meta here, nothing specific to directory, so
376
ie.revision = parent_entry.revision
377
return self._get_delta(ie, basis_inv, path), False
379
self._add_text_to_weave(ie.file_id, lines, heads, None)
380
elif kind == 'symlink':
381
current_link_target = content_summary[3]
383
# symlink target is not generic metadata, check if it has
385
if current_link_target != parent_entry.symlink_target:
388
# unchanged, carry over.
389
ie.revision = parent_entry.revision
390
ie.symlink_target = parent_entry.symlink_target
391
return self._get_delta(ie, basis_inv, path), False
392
ie.symlink_target = current_link_target
394
self._add_text_to_weave(ie.file_id, lines, heads, None)
395
elif kind == 'tree-reference':
397
if content_summary[3] != parent_entry.reference_revision:
400
# unchanged, carry over.
401
ie.reference_revision = parent_entry.reference_revision
402
ie.revision = parent_entry.revision
403
return self._get_delta(ie, basis_inv, path), False
404
ie.reference_revision = content_summary[3]
406
self._add_text_to_weave(ie.file_id, lines, heads, None)
408
raise NotImplementedError('unknown kind')
409
ie.revision = self._new_revision_id
410
return self._get_delta(ie, basis_inv, path), True
412
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
413
versionedfile = self.repository.weave_store.get_weave_or_empty(
414
file_id, self.repository.get_transaction())
415
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
416
# than add_lines, and allows committing when a parent is ghosted for
418
# Note: as we read the content directly from the tree, we know its not
419
# been turned into unicode or badly split - but a broken tree
420
# implementation could give us bad output from readlines() so this is
421
# not a guarantee of safety. What would be better is always checking
422
# the content during test suite execution. RBC 20070912
424
return versionedfile.add_lines_with_ghosts(
425
self._new_revision_id, parents, new_lines,
426
nostore_sha=nostore_sha, random_id=self.random_revid,
427
check_content=False)[0:2]
429
versionedfile.clear_cache()
432
class RootCommitBuilder(CommitBuilder):
433
"""This commitbuilder actually records the root id"""
435
# the root entry gets versioned properly by this builder.
436
_versioned_root = True
438
def _check_root(self, ie, parent_invs, tree):
439
"""Helper for record_entry_contents.
441
:param ie: An entry being added.
442
:param parent_invs: The inventories of the parent revisions of the
444
:param tree: The tree that is being committed.
448
######################################################################
451
class Repository(object):
452
"""Repository holding history for one or more branches.
454
The repository holds and retrieves historical information including
455
revisions and file history. It's normally accessed only by the Branch,
456
which views a particular line of development through that history.
458
The Repository builds on top of Stores and a Transport, which respectively
459
describe the disk data format and the way of accessing the (possibly
463
# What class to use for a CommitBuilder. Often its simpler to change this
464
# in a Repository class subclass rather than to override
465
# get_commit_builder.
466
_commit_builder_class = CommitBuilder
467
# The search regex used by xml based repositories to determine what things
468
# where changed in a single commit.
469
_file_ids_altered_regex = lazy_regex.lazy_compile(
470
r'file_id="(?P<file_id>[^"]+)"'
471
r'.* revision="(?P<revision_id>[^"]+)"'
474
def abort_write_group(self):
475
"""Commit the contents accrued within the current write group.
477
:seealso: start_write_group.
479
if self._write_group is not self.get_transaction():
480
# has an unlock or relock occured ?
481
raise errors.BzrError('mismatched lock context and write group.')
482
self._abort_write_group()
483
self._write_group = None
485
def _abort_write_group(self):
486
"""Template method for per-repository write group cleanup.
488
This is called during abort before the write group is considered to be
489
finished and should cleanup any internal state accrued during the write
490
group. There is no requirement that data handed to the repository be
491
*not* made available - this is not a rollback - but neither should any
492
attempt be made to ensure that data added is fully commited. Abort is
493
invoked when an error has occured so futher disk or network operations
494
may not be possible or may error and if possible should not be
499
def add_inventory(self, revision_id, inv, parents):
500
"""Add the inventory inv to the repository as revision_id.
502
:param parents: The revision ids of the parents that revision_id
503
is known to have and are in the repository already.
505
returns the sha1 of the serialized inventory.
507
assert self.is_in_write_group()
508
_mod_revision.check_not_reserved_id(revision_id)
509
assert inv.revision_id is None or inv.revision_id == revision_id, \
510
"Mismatch between inventory revision" \
511
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
512
assert inv.root is not None
513
inv_lines = self._serialise_inventory_to_lines(inv)
514
inv_vf = self.get_inventory_weave()
515
return self._inventory_add_lines(inv_vf, revision_id, parents,
516
inv_lines, check_content=False)
518
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
520
"""Store lines in inv_vf and return the sha1 of the inventory."""
522
for parent in parents:
524
final_parents.append(parent)
525
return inv_vf.add_lines(revision_id, final_parents, lines,
526
check_content=check_content)[0]
529
def add_revision(self, revision_id, rev, inv=None, config=None):
530
"""Add rev to the revision store as revision_id.
532
:param revision_id: the revision id to use.
533
:param rev: The revision object.
534
:param inv: The inventory for the revision. if None, it will be looked
535
up in the inventory storer
536
:param config: If None no digital signature will be created.
537
If supplied its signature_needed method will be used
538
to determine if a signature should be made.
540
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
542
_mod_revision.check_not_reserved_id(revision_id)
543
if config is not None and config.signature_needed():
545
inv = self.get_inventory(revision_id)
546
plaintext = Testament(rev, inv).as_short_text()
547
self.store_revision_signature(
548
gpg.GPGStrategy(config), plaintext, revision_id)
549
if not revision_id in self.get_inventory_weave():
551
raise errors.WeaveRevisionNotPresent(revision_id,
552
self.get_inventory_weave())
554
# yes, this is not suitable for adding with ghosts.
555
self.add_inventory(revision_id, inv, rev.parent_ids)
556
self._revision_store.add_revision(rev, self.get_transaction())
558
def _add_revision_text(self, revision_id, text):
559
revision = self._revision_store._serializer.read_revision_from_string(
561
self._revision_store._add_revision(revision, StringIO(text),
562
self.get_transaction())
564
def all_revision_ids(self):
565
"""Returns a list of all the revision ids in the repository.
567
This is deprecated because code should generally work on the graph
568
reachable from a particular revision, and ignore any other revisions
569
that might be present. There is no direct replacement method.
571
if 'evil' in debug.debug_flags:
572
mutter_callsite(2, "all_revision_ids is linear with history.")
573
return self._all_revision_ids()
575
def _all_revision_ids(self):
576
"""Returns a list of all the revision ids in the repository.
578
These are in as much topological order as the underlying store can
581
raise NotImplementedError(self._all_revision_ids)
583
def break_lock(self):
584
"""Break a lock if one is present from another instance.
586
Uses the ui factory to ask for confirmation if the lock may be from
589
self.control_files.break_lock()
592
def _eliminate_revisions_not_present(self, revision_ids):
593
"""Check every revision id in revision_ids to see if we have it.
595
Returns a set of the present revisions.
598
for id in revision_ids:
599
if self.has_revision(id):
604
def create(a_bzrdir):
605
"""Construct the current default format repository in a_bzrdir."""
606
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
608
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
609
"""instantiate a Repository.
611
:param _format: The format of the repository on disk.
612
:param a_bzrdir: The BzrDir of the repository.
614
In the future we will have a single api for all stores for
615
getting file texts, inventories and revisions, then
616
this construct will accept instances of those things.
618
super(Repository, self).__init__()
619
self._format = _format
620
# the following are part of the public API for Repository:
621
self.bzrdir = a_bzrdir
622
self.control_files = control_files
623
self._revision_store = _revision_store
624
# backwards compatibility
625
self.weave_store = text_store
627
self._reconcile_does_inventory_gc = True
628
self._reconcile_fixes_text_parents = False
629
self._reconcile_backsup_inventory = True
630
# not right yet - should be more semantically clear ?
632
self.control_store = control_store
633
self.control_weaves = control_store
634
# TODO: make sure to construct the right store classes, etc, depending
635
# on whether escaping is required.
636
self._warn_if_deprecated()
637
self._write_group = None
638
self.base = control_files._transport.base
641
return '%s(%r)' % (self.__class__.__name__,
644
def has_same_location(self, other):
645
"""Returns a boolean indicating if this repository is at the same
646
location as another repository.
648
This might return False even when two repository objects are accessing
649
the same physical repository via different URLs.
651
if self.__class__ is not other.__class__:
653
return (self.control_files._transport.base ==
654
other.control_files._transport.base)
656
def is_in_write_group(self):
657
"""Return True if there is an open write group.
659
:seealso: start_write_group.
661
return self._write_group is not None
664
return self.control_files.is_locked()
666
def is_write_locked(self):
667
"""Return True if this object is write locked."""
668
return self.is_locked() and self.control_files._lock_mode == 'w'
670
def lock_write(self, token=None):
671
"""Lock this repository for writing.
673
This causes caching within the repository obejct to start accumlating
674
data during reads, and allows a 'write_group' to be obtained. Write
675
groups must be used for actual data insertion.
677
:param token: if this is already locked, then lock_write will fail
678
unless the token matches the existing lock.
679
:returns: a token if this instance supports tokens, otherwise None.
680
:raises TokenLockingNotSupported: when a token is given but this
681
instance doesn't support using token locks.
682
:raises MismatchedToken: if the specified token doesn't match the token
683
of the existing lock.
684
:seealso: start_write_group.
686
A token should be passed in if you know that you have locked the object
687
some other way, and need to synchronise this object's state with that
690
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
692
result = self.control_files.lock_write(token=token)
697
self.control_files.lock_read()
700
def get_physical_lock_status(self):
701
return self.control_files.get_physical_lock_status()
703
def leave_lock_in_place(self):
704
"""Tell this repository not to release the physical lock when this
707
If lock_write doesn't return a token, then this method is not supported.
709
self.control_files.leave_in_place()
711
def dont_leave_lock_in_place(self):
712
"""Tell this repository to release the physical lock when this
713
object is unlocked, even if it didn't originally acquire it.
715
If lock_write doesn't return a token, then this method is not supported.
717
self.control_files.dont_leave_in_place()
720
def gather_stats(self, revid=None, committers=None):
721
"""Gather statistics from a revision id.
723
:param revid: The revision id to gather statistics from, if None, then
724
no revision specific statistics are gathered.
725
:param committers: Optional parameter controlling whether to grab
726
a count of committers from the revision specific statistics.
727
:return: A dictionary of statistics. Currently this contains:
728
committers: The number of committers if requested.
729
firstrev: A tuple with timestamp, timezone for the penultimate left
730
most ancestor of revid, if revid is not the NULL_REVISION.
731
latestrev: A tuple with timestamp, timezone for revid, if revid is
732
not the NULL_REVISION.
733
revisions: The total revision count in the repository.
734
size: An estimate disk size of the repository in bytes.
737
if revid and committers:
738
result['committers'] = 0
739
if revid and revid != _mod_revision.NULL_REVISION:
741
all_committers = set()
742
revisions = self.get_ancestry(revid)
743
# pop the leading None
745
first_revision = None
747
# ignore the revisions in the middle - just grab first and last
748
revisions = revisions[0], revisions[-1]
749
for revision in self.get_revisions(revisions):
750
if not first_revision:
751
first_revision = revision
753
all_committers.add(revision.committer)
754
last_revision = revision
756
result['committers'] = len(all_committers)
757
result['firstrev'] = (first_revision.timestamp,
758
first_revision.timezone)
759
result['latestrev'] = (last_revision.timestamp,
760
last_revision.timezone)
762
# now gather global repository information
763
if self.bzrdir.root_transport.listable():
764
c, t = self._revision_store.total_size(self.get_transaction())
765
result['revisions'] = c
769
def find_branches(self, using=False):
770
"""Find branches underneath this repository.
772
This will include branches inside other branches.
774
:param using: If True, list only branches using this repository.
776
if using and not self.is_shared():
778
return [self.bzrdir.open_branch()]
779
except errors.NotBranchError:
781
class Evaluator(object):
784
self.first_call = True
786
def __call__(self, bzrdir):
787
# On the first call, the parameter is always the bzrdir
788
# containing the current repo.
789
if not self.first_call:
791
repository = bzrdir.open_repository()
792
except errors.NoRepositoryPresent:
795
return False, (None, repository)
796
self.first_call = False
798
value = (bzrdir.open_branch(), None)
799
except errors.NotBranchError:
804
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
805
self.bzrdir.root_transport, evaluate=Evaluator()):
806
if branch is not None:
807
branches.append(branch)
808
if not using and repository is not None:
809
branches.extend(repository.find_branches())
812
def get_data_stream(self, revision_ids):
813
raise NotImplementedError(self.get_data_stream)
815
def insert_data_stream(self, stream):
816
"""XXX What does this really do?
818
Is it a substitute for fetch?
819
Should it manage its own write group ?
821
for item_key, bytes in stream:
822
if item_key[0] == 'file':
823
(file_id,) = item_key[1:]
824
knit = self.weave_store.get_weave_or_empty(
825
file_id, self.get_transaction())
826
elif item_key == ('inventory',):
827
knit = self.get_inventory_weave()
828
elif item_key == ('revisions',):
829
knit = self._revision_store.get_revision_file(
830
self.get_transaction())
831
elif item_key == ('signatures',):
832
knit = self._revision_store.get_signature_file(
833
self.get_transaction())
835
raise RepositoryDataStreamError(
836
"Unrecognised data stream key '%s'" % (item_key,))
837
decoded_list = bencode.bdecode(bytes)
838
format = decoded_list.pop(0)
841
for version, options, parents, some_bytes in decoded_list:
842
data_list.append((version, options, len(some_bytes), parents))
843
knit_bytes += some_bytes
844
knit.insert_data_stream(
845
(format, data_list, StringIO(knit_bytes).read))
848
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
849
"""Return the revision ids that other has that this does not.
851
These are returned in topological order.
853
revision_id: only return revision ids included by revision_id.
855
return InterRepository.get(other, self).missing_revision_ids(
856
revision_id, find_ghosts)
860
"""Open the repository rooted at base.
862
For instance, if the repository is at URL/.bzr/repository,
863
Repository.open(URL) -> a Repository instance.
865
control = bzrdir.BzrDir.open(base)
866
return control.open_repository()
868
def copy_content_into(self, destination, revision_id=None):
869
"""Make a complete copy of the content in self into destination.
871
This is a destructive operation! Do not use it on existing
874
return InterRepository.get(self, destination).copy_content(revision_id)
876
def commit_write_group(self):
877
"""Commit the contents accrued within the current write group.
879
:seealso: start_write_group.
881
if self._write_group is not self.get_transaction():
882
# has an unlock or relock occured ?
883
raise errors.BzrError('mismatched lock context %r and '
885
(self.get_transaction(), self._write_group))
886
self._commit_write_group()
887
self._write_group = None
889
def _commit_write_group(self):
890
"""Template method for per-repository write group cleanup.
892
This is called before the write group is considered to be
893
finished and should ensure that all data handed to the repository
894
for writing during the write group is safely committed (to the
895
extent possible considering file system caching etc).
898
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
899
"""Fetch the content required to construct revision_id from source.
901
If revision_id is None all content is copied.
902
:param find_ghosts: Find and copy revisions in the source that are
903
ghosts in the target (and not reachable directly by walking out to
904
the first-present revision in target from revision_id).
906
# fast path same-url fetch operations
907
if self.has_same_location(source):
908
# check that last_revision is in 'from' and then return a
910
if (revision_id is not None and
911
not _mod_revision.is_null(revision_id)):
912
self.get_revision(revision_id)
914
inter = InterRepository.get(source, self)
916
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
917
except NotImplementedError:
918
raise errors.IncompatibleRepositories(source, self)
920
def create_bundle(self, target, base, fileobj, format=None):
921
return serializer.write_bundle(self, target, base, fileobj, format)
923
def get_commit_builder(self, branch, parents, config, timestamp=None,
924
timezone=None, committer=None, revprops=None,
926
"""Obtain a CommitBuilder for this repository.
928
:param branch: Branch to commit to.
929
:param parents: Revision ids of the parents of the new revision.
930
:param config: Configuration to use.
931
:param timestamp: Optional timestamp recorded for commit.
932
:param timezone: Optional timezone for timestamp.
933
:param committer: Optional committer to set for commit.
934
:param revprops: Optional dictionary of revision properties.
935
:param revision_id: Optional revision id.
937
result = self._commit_builder_class(self, parents, config,
938
timestamp, timezone, committer, revprops, revision_id)
939
self.start_write_group()
943
if (self.control_files._lock_count == 1 and
944
self.control_files._lock_mode == 'w'):
945
if self._write_group is not None:
946
self.abort_write_group()
947
self.control_files.unlock()
948
raise errors.BzrError(
949
'Must end write groups before releasing write locks.')
950
self.control_files.unlock()
953
def clone(self, a_bzrdir, revision_id=None):
954
"""Clone this repository into a_bzrdir using the current format.
956
Currently no check is made that the format of this repository and
957
the bzrdir format are compatible. FIXME RBC 20060201.
959
:return: The newly created destination repository.
961
# TODO: deprecate after 0.16; cloning this with all its settings is
962
# probably not very useful -- mbp 20070423
963
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
964
self.copy_content_into(dest_repo, revision_id)
967
def start_write_group(self):
968
"""Start a write group in the repository.
970
Write groups are used by repositories which do not have a 1:1 mapping
971
between file ids and backend store to manage the insertion of data from
972
both fetch and commit operations.
974
A write lock is required around the start_write_group/commit_write_group
975
for the support of lock-requiring repository formats.
977
One can only insert data into a repository inside a write group.
981
if not self.is_write_locked():
982
raise errors.NotWriteLocked(self)
983
if self._write_group:
984
raise errors.BzrError('already in a write group')
985
self._start_write_group()
986
# so we can detect unlock/relock - the write group is now entered.
987
self._write_group = self.get_transaction()
989
def _start_write_group(self):
990
"""Template method for per-repository write group startup.
992
This is called before the write group is considered to be
997
def sprout(self, to_bzrdir, revision_id=None):
998
"""Create a descendent repository for new development.
1000
Unlike clone, this does not copy the settings of the repository.
1002
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1003
dest_repo.fetch(self, revision_id=revision_id)
1006
def _create_sprouting_repo(self, a_bzrdir, shared):
1007
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1008
# use target default format.
1009
dest_repo = a_bzrdir.create_repository()
1011
# Most control formats need the repository to be specifically
1012
# created, but on some old all-in-one formats it's not needed
1014
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1015
except errors.UninitializableFormat:
1016
dest_repo = a_bzrdir.open_repository()
1020
def has_revision(self, revision_id):
1021
"""True if this repository has a copy of the revision."""
1022
if 'evil' in debug.debug_flags:
1023
mutter_callsite(3, "has_revision is a LBYL symptom.")
1024
return self._revision_store.has_revision_id(revision_id,
1025
self.get_transaction())
1028
def get_revision(self, revision_id):
1029
"""Return the Revision object for a named revision."""
1030
return self.get_revisions([revision_id])[0]
1033
def get_revision_reconcile(self, revision_id):
1034
"""'reconcile' helper routine that allows access to a revision always.
1036
This variant of get_revision does not cross check the weave graph
1037
against the revision one as get_revision does: but it should only
1038
be used by reconcile, or reconcile-alike commands that are correcting
1039
or testing the revision graph.
1041
return self._get_revisions([revision_id])[0]
1044
def get_revisions(self, revision_ids):
1045
"""Get many revisions at once."""
1046
return self._get_revisions(revision_ids)
1049
def _get_revisions(self, revision_ids):
1050
"""Core work logic to get many revisions without sanity checks."""
1051
for rev_id in revision_ids:
1052
if not rev_id or not isinstance(rev_id, basestring):
1053
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1054
revs = self._revision_store.get_revisions(revision_ids,
1055
self.get_transaction())
1057
assert not isinstance(rev.revision_id, unicode)
1058
for parent_id in rev.parent_ids:
1059
assert not isinstance(parent_id, unicode)
1063
def get_revision_xml(self, revision_id):
1064
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1065
# would have already do it.
1066
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1067
rev = self.get_revision(revision_id)
1068
rev_tmp = StringIO()
1069
# the current serializer..
1070
self._revision_store._serializer.write_revision(rev, rev_tmp)
1072
return rev_tmp.getvalue()
1075
def get_deltas_for_revisions(self, revisions):
1076
"""Produce a generator of revision deltas.
1078
Note that the input is a sequence of REVISIONS, not revision_ids.
1079
Trees will be held in memory until the generator exits.
1080
Each delta is relative to the revision's lefthand predecessor.
1082
required_trees = set()
1083
for revision in revisions:
1084
required_trees.add(revision.revision_id)
1085
required_trees.update(revision.parent_ids[:1])
1086
trees = dict((t.get_revision_id(), t) for
1087
t in self.revision_trees(required_trees))
1088
for revision in revisions:
1089
if not revision.parent_ids:
1090
old_tree = self.revision_tree(None)
1092
old_tree = trees[revision.parent_ids[0]]
1093
yield trees[revision.revision_id].changes_from(old_tree)
1096
def get_revision_delta(self, revision_id):
1097
"""Return the delta for one revision.
1099
The delta is relative to the left-hand predecessor of the
1102
r = self.get_revision(revision_id)
1103
return list(self.get_deltas_for_revisions([r]))[0]
1106
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1107
signature = gpg_strategy.sign(plaintext)
1108
self.add_signature_text(revision_id, signature)
1111
def add_signature_text(self, revision_id, signature):
1112
self._revision_store.add_revision_signature_text(revision_id,
1114
self.get_transaction())
1116
def find_text_key_references(self):
1117
"""Find the text key references within the repository.
1119
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1120
revision_ids. Each altered file-ids has the exact revision_ids that
1121
altered it listed explicitly.
1122
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1123
to whether they were referred to by the inventory of the
1124
revision_id that they contain. The inventory texts from all present
1125
revision ids are assessed to generate this report.
1127
revision_ids = self.all_revision_ids()
1128
w = self.get_inventory_weave()
1129
pb = ui.ui_factory.nested_progress_bar()
1131
return self._find_text_key_references_from_xml_inventory_lines(
1132
w.iter_lines_added_or_present_in_versions(revision_ids, pb=pb))
1136
def _find_text_key_references_from_xml_inventory_lines(self,
1138
"""Core routine for extracting references to texts from inventories.
1140
This performs the translation of xml lines to revision ids.
1142
:param line_iterator: An iterator of lines, origin_version_id
1143
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1144
to whether they were referred to by the inventory of the
1145
revision_id that they contain. Note that if that revision_id was
1146
not part of the line_iterator's output then False will be given -
1147
even though it may actually refer to that key.
1149
if not self._serializer.support_altered_by_hack:
1150
raise AssertionError(
1151
"_find_text_key_references_from_xml_inventory_lines only "
1152
"supported for branches which store inventory as unnested xml"
1153
", not on %r" % self)
1156
# this code needs to read every new line in every inventory for the
1157
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1158
# not present in one of those inventories is unnecessary but not
1159
# harmful because we are filtering by the revision id marker in the
1160
# inventory lines : we only select file ids altered in one of those
1161
# revisions. We don't need to see all lines in the inventory because
1162
# only those added in an inventory in rev X can contain a revision=X
1164
unescape_revid_cache = {}
1165
unescape_fileid_cache = {}
1167
# jam 20061218 In a big fetch, this handles hundreds of thousands
1168
# of lines, so it has had a lot of inlining and optimizing done.
1169
# Sorry that it is a little bit messy.
1170
# Move several functions to be local variables, since this is a long
1172
search = self._file_ids_altered_regex.search
1173
unescape = _unescape_xml
1174
setdefault = result.setdefault
1175
for line, version_id in line_iterator:
1176
match = search(line)
1179
# One call to match.group() returning multiple items is quite a
1180
# bit faster than 2 calls to match.group() each returning 1
1181
file_id, revision_id = match.group('file_id', 'revision_id')
1183
# Inlining the cache lookups helps a lot when you make 170,000
1184
# lines and 350k ids, versus 8.4 unique ids.
1185
# Using a cache helps in 2 ways:
1186
# 1) Avoids unnecessary decoding calls
1187
# 2) Re-uses cached strings, which helps in future set and
1189
# (2) is enough that removing encoding entirely along with
1190
# the cache (so we are using plain strings) results in no
1191
# performance improvement.
1193
revision_id = unescape_revid_cache[revision_id]
1195
unescaped = unescape(revision_id)
1196
unescape_revid_cache[revision_id] = unescaped
1197
revision_id = unescaped
1199
# Note that unconditionally unescaping means that we deserialise
1200
# every fileid, which for general 'pull' is not great, but we don't
1201
# really want to have some many fulltexts that this matters anyway.
1204
file_id = unescape_fileid_cache[file_id]
1206
unescaped = unescape(file_id)
1207
unescape_fileid_cache[file_id] = unescaped
1210
key = (file_id, revision_id)
1211
setdefault(key, False)
1212
if revision_id == version_id:
1216
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1218
"""Helper routine for fileids_altered_by_revision_ids.
1220
This performs the translation of xml lines to revision ids.
1222
:param line_iterator: An iterator of lines, origin_version_id
1223
:param revision_ids: The revision ids to filter for. This should be a
1224
set or other type which supports efficient __contains__ lookups, as
1225
the revision id from each parsed line will be looked up in the
1226
revision_ids filter.
1227
:return: a dictionary mapping altered file-ids to an iterable of
1228
revision_ids. Each altered file-ids has the exact revision_ids that
1229
altered it listed explicitly.
1232
setdefault = result.setdefault
1233
for file_id, revision_id in \
1234
self._find_text_key_references_from_xml_inventory_lines(
1235
line_iterator).iterkeys():
1236
# once data is all ensured-consistent; then this is
1237
# if revision_id == version_id
1238
if revision_id in revision_ids:
1239
setdefault(file_id, set()).add(revision_id)
1242
def fileids_altered_by_revision_ids(self, revision_ids):
1243
"""Find the file ids and versions affected by revisions.
1245
:param revisions: an iterable containing revision ids.
1246
:return: a dictionary mapping altered file-ids to an iterable of
1247
revision_ids. Each altered file-ids has the exact revision_ids that
1248
altered it listed explicitly.
1250
selected_revision_ids = set(revision_ids)
1251
w = self.get_inventory_weave()
1252
pb = ui.ui_factory.nested_progress_bar()
1254
return self._find_file_ids_from_xml_inventory_lines(
1255
w.iter_lines_added_or_present_in_versions(
1256
selected_revision_ids, pb=pb),
1257
selected_revision_ids)
1261
def iter_files_bytes(self, desired_files):
1262
"""Iterate through file versions.
1264
Files will not necessarily be returned in the order they occur in
1265
desired_files. No specific order is guaranteed.
1267
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1268
value supplied by the caller as part of desired_files. It should
1269
uniquely identify the file version in the caller's context. (Examples:
1270
an index number or a TreeTransform trans_id.)
1272
bytes_iterator is an iterable of bytestrings for the file. The
1273
kind of iterable and length of the bytestrings are unspecified, but for
1274
this implementation, it is a list of lines produced by
1275
VersionedFile.get_lines().
1277
:param desired_files: a list of (file_id, revision_id, identifier)
1280
transaction = self.get_transaction()
1281
for file_id, revision_id, callable_data in desired_files:
1283
weave = self.weave_store.get_weave(file_id, transaction)
1284
except errors.NoSuchFile:
1285
raise errors.NoSuchIdInRepository(self, file_id)
1286
yield callable_data, weave.get_lines(revision_id)
1288
def _generate_text_key_index(self, text_key_references=None,
1290
"""Generate a new text key index for the repository.
1292
This is an expensive function that will take considerable time to run.
1294
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1295
list of parents, also text keys. When a given key has no parents,
1296
the parents list will be [NULL_REVISION].
1298
# All revisions, to find inventory parents.
1299
if ancestors is None:
1300
revision_graph = self.get_revision_graph_with_ghosts()
1301
ancestors = revision_graph.get_ancestors()
1302
if text_key_references is None:
1303
text_key_references = self.find_text_key_references()
1304
pb = ui.ui_factory.nested_progress_bar()
1306
return self._do_generate_text_key_index(ancestors,
1307
text_key_references, pb)
1311
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1312
"""Helper for _generate_text_key_index to avoid deep nesting."""
1313
revision_order = tsort.topo_sort(ancestors)
1314
invalid_keys = set()
1316
for revision_id in revision_order:
1317
revision_keys[revision_id] = set()
1318
text_count = len(text_key_references)
1319
# a cache of the text keys to allow reuse; costs a dict of all the
1320
# keys, but saves a 2-tuple for every child of a given key.
1322
for text_key, valid in text_key_references.iteritems():
1324
invalid_keys.add(text_key)
1326
revision_keys[text_key[1]].add(text_key)
1327
text_key_cache[text_key] = text_key
1328
del text_key_references
1330
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1331
NULL_REVISION = _mod_revision.NULL_REVISION
1332
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1333
# too small for large or very branchy trees. However, for 55K path
1334
# trees, it would be easy to use too much memory trivially. Ideally we
1335
# could gauge this by looking at available real memory etc, but this is
1336
# always a tricky proposition.
1337
inventory_cache = lru_cache.LRUCache(10)
1338
batch_size = 10 # should be ~150MB on a 55K path tree
1339
batch_count = len(revision_order) / batch_size + 1
1341
pb.update("Calculating text parents.", processed_texts, text_count)
1342
for offset in xrange(batch_count):
1343
to_query = revision_order[offset * batch_size:(offset + 1) *
1347
for rev_tree in self.revision_trees(to_query):
1348
revision_id = rev_tree.get_revision_id()
1349
parent_ids = ancestors[revision_id]
1350
for text_key in revision_keys[revision_id]:
1351
pb.update("Calculating text parents.", processed_texts)
1352
processed_texts += 1
1353
candidate_parents = []
1354
for parent_id in parent_ids:
1355
parent_text_key = (text_key[0], parent_id)
1357
check_parent = parent_text_key not in \
1358
revision_keys[parent_id]
1360
# the parent parent_id is a ghost:
1361
check_parent = False
1362
# truncate the derived graph against this ghost.
1363
parent_text_key = None
1365
# look at the parent commit details inventories to
1366
# determine possible candidates in the per file graph.
1369
inv = inventory_cache[parent_id]
1371
inv = self.revision_tree(parent_id).inventory
1372
inventory_cache[parent_id] = inv
1373
parent_entry = inv._byid.get(text_key[0], None)
1374
if parent_entry is not None:
1376
text_key[0], parent_entry.revision)
1378
parent_text_key = None
1379
if parent_text_key is not None:
1380
candidate_parents.append(
1381
text_key_cache[parent_text_key])
1382
parent_heads = text_graph.heads(candidate_parents)
1383
new_parents = list(parent_heads)
1384
new_parents.sort(key=lambda x:candidate_parents.index(x))
1385
if new_parents == []:
1386
new_parents = [NULL_REVISION]
1387
text_index[text_key] = new_parents
1389
for text_key in invalid_keys:
1390
text_index[text_key] = [NULL_REVISION]
1393
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1394
"""Get an iterable listing the keys of all the data introduced by a set
1397
The keys will be ordered so that the corresponding items can be safely
1398
fetched and inserted in that order.
1400
:returns: An iterable producing tuples of (knit-kind, file-id,
1401
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1402
'revisions'. file-id is None unless knit-kind is 'file'.
1404
# XXX: it's a bit weird to control the inventory weave caching in this
1405
# generator. Ideally the caching would be done in fetch.py I think. Or
1406
# maybe this generator should explicitly have the contract that it
1407
# should not be iterated until the previously yielded item has been
1410
inv_w = self.get_inventory_weave()
1411
inv_w.enable_cache()
1413
# file ids that changed
1414
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1416
num_file_ids = len(file_ids)
1417
for file_id, altered_versions in file_ids.iteritems():
1418
if _files_pb is not None:
1419
_files_pb.update("fetch texts", count, num_file_ids)
1421
yield ("file", file_id, altered_versions)
1422
# We're done with the files_pb. Note that it finished by the caller,
1423
# just as it was created by the caller.
1427
yield ("inventory", None, revision_ids)
1431
revisions_with_signatures = set()
1432
for rev_id in revision_ids:
1434
self.get_signature_text(rev_id)
1435
except errors.NoSuchRevision:
1439
revisions_with_signatures.add(rev_id)
1441
yield ("signatures", None, revisions_with_signatures)
1444
yield ("revisions", None, revision_ids)
1447
def get_inventory_weave(self):
1448
return self.control_weaves.get_weave('inventory',
1449
self.get_transaction())
1452
def get_inventory(self, revision_id):
1453
"""Get Inventory object by hash."""
1454
return self.deserialise_inventory(
1455
revision_id, self.get_inventory_xml(revision_id))
1457
def deserialise_inventory(self, revision_id, xml):
1458
"""Transform the xml into an inventory object.
1460
:param revision_id: The expected revision id of the inventory.
1461
:param xml: A serialised inventory.
1463
return self._serializer.read_inventory_from_string(xml, revision_id)
1465
def serialise_inventory(self, inv):
1466
return self._serializer.write_inventory_to_string(inv)
1468
def _serialise_inventory_to_lines(self, inv):
1469
return self._serializer.write_inventory_to_lines(inv)
1471
def get_serializer_format(self):
1472
return self._serializer.format_num
1475
def get_inventory_xml(self, revision_id):
1476
"""Get inventory XML as a file object."""
1478
assert isinstance(revision_id, str), type(revision_id)
1479
iw = self.get_inventory_weave()
1480
return iw.get_text(revision_id)
1482
raise errors.HistoryMissing(self, 'inventory', revision_id)
1485
def get_inventory_sha1(self, revision_id):
1486
"""Return the sha1 hash of the inventory entry
1488
return self.get_revision(revision_id).inventory_sha1
1491
def get_revision_graph(self, revision_id=None):
1492
"""Return a dictionary containing the revision graph.
1494
NB: This method should not be used as it accesses the entire graph all
1495
at once, which is much more data than most operations should require.
1497
:param revision_id: The revision_id to get a graph from. If None, then
1498
the entire revision graph is returned. This is a deprecated mode of
1499
operation and will be removed in the future.
1500
:return: a dictionary of revision_id->revision_parents_list.
1502
raise NotImplementedError(self.get_revision_graph)
1505
def get_revision_graph_with_ghosts(self, revision_ids=None):
1506
"""Return a graph of the revisions with ghosts marked as applicable.
1508
:param revision_ids: an iterable of revisions to graph or None for all.
1509
:return: a Graph object with the graph reachable from revision_ids.
1511
if 'evil' in debug.debug_flags:
1513
"get_revision_graph_with_ghosts scales with size of history.")
1514
result = deprecated_graph.Graph()
1515
if not revision_ids:
1516
pending = set(self.all_revision_ids())
1519
pending = set(revision_ids)
1520
# special case NULL_REVISION
1521
if _mod_revision.NULL_REVISION in pending:
1522
pending.remove(_mod_revision.NULL_REVISION)
1523
required = set(pending)
1526
revision_id = pending.pop()
1528
rev = self.get_revision(revision_id)
1529
except errors.NoSuchRevision:
1530
if revision_id in required:
1533
result.add_ghost(revision_id)
1535
for parent_id in rev.parent_ids:
1536
# is this queued or done ?
1537
if (parent_id not in pending and
1538
parent_id not in done):
1540
pending.add(parent_id)
1541
result.add_node(revision_id, rev.parent_ids)
1542
done.add(revision_id)
1545
def _get_history_vf(self):
1546
"""Get a versionedfile whose history graph reflects all revisions.
1548
For weave repositories, this is the inventory weave.
1550
return self.get_inventory_weave()
1552
def iter_reverse_revision_history(self, revision_id):
1553
"""Iterate backwards through revision ids in the lefthand history
1555
:param revision_id: The revision id to start with. All its lefthand
1556
ancestors will be traversed.
1558
if revision_id in (None, _mod_revision.NULL_REVISION):
1560
next_id = revision_id
1561
versionedfile = self._get_history_vf()
1564
parents = versionedfile.get_parents(next_id)
1565
if len(parents) == 0:
1568
next_id = parents[0]
1571
def get_revision_inventory(self, revision_id):
1572
"""Return inventory of a past revision."""
1573
# TODO: Unify this with get_inventory()
1574
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1575
# must be the same as its revision, so this is trivial.
1576
if revision_id is None:
1577
# This does not make sense: if there is no revision,
1578
# then it is the current tree inventory surely ?!
1579
# and thus get_root_id() is something that looks at the last
1580
# commit on the branch, and the get_root_id is an inventory check.
1581
raise NotImplementedError
1582
# return Inventory(self.get_root_id())
1584
return self.get_inventory(revision_id)
1587
def is_shared(self):
1588
"""Return True if this repository is flagged as a shared repository."""
1589
raise NotImplementedError(self.is_shared)
1592
def reconcile(self, other=None, thorough=False):
1593
"""Reconcile this repository."""
1594
from bzrlib.reconcile import RepoReconciler
1595
reconciler = RepoReconciler(self, thorough=thorough)
1596
reconciler.reconcile()
1599
def _refresh_data(self):
1600
"""Helper called from lock_* to ensure coherency with disk.
1602
The default implementation does nothing; it is however possible
1603
for repositories to maintain loaded indices across multiple locks
1604
by checking inside their implementation of this method to see
1605
whether their indices are still valid. This depends of course on
1606
the disk format being validatable in this manner.
1610
def revision_tree(self, revision_id):
1611
"""Return Tree for a revision on this branch.
1613
`revision_id` may be None for the empty tree revision.
1615
# TODO: refactor this to use an existing revision object
1616
# so we don't need to read it in twice.
1617
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1618
return RevisionTree(self, Inventory(root_id=None),
1619
_mod_revision.NULL_REVISION)
1621
inv = self.get_revision_inventory(revision_id)
1622
return RevisionTree(self, inv, revision_id)
1625
def revision_trees(self, revision_ids):
1626
"""Return Tree for a revision on this branch.
1628
`revision_id` may not be None or 'null:'"""
1629
assert None not in revision_ids
1630
assert _mod_revision.NULL_REVISION not in revision_ids
1631
texts = self.get_inventory_weave().get_texts(revision_ids)
1632
for text, revision_id in zip(texts, revision_ids):
1633
inv = self.deserialise_inventory(revision_id, text)
1634
yield RevisionTree(self, inv, revision_id)
1637
def get_ancestry(self, revision_id, topo_sorted=True):
1638
"""Return a list of revision-ids integrated by a revision.
1640
The first element of the list is always None, indicating the origin
1641
revision. This might change when we have history horizons, or
1642
perhaps we should have a new API.
1644
This is topologically sorted.
1646
if _mod_revision.is_null(revision_id):
1648
if not self.has_revision(revision_id):
1649
raise errors.NoSuchRevision(self, revision_id)
1650
w = self.get_inventory_weave()
1651
candidates = w.get_ancestry(revision_id, topo_sorted)
1652
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1655
"""Compress the data within the repository.
1657
This operation only makes sense for some repository types. For other
1658
types it should be a no-op that just returns.
1660
This stub method does not require a lock, but subclasses should use
1661
@needs_write_lock as this is a long running call its reasonable to
1662
implicitly lock for the user.
1666
def print_file(self, file, revision_id):
1667
"""Print `file` to stdout.
1669
FIXME RBC 20060125 as John Meinel points out this is a bad api
1670
- it writes to stdout, it assumes that that is valid etc. Fix
1671
by creating a new more flexible convenience function.
1673
tree = self.revision_tree(revision_id)
1674
# use inventory as it was in that revision
1675
file_id = tree.inventory.path2id(file)
1677
# TODO: jam 20060427 Write a test for this code path
1678
# it had a bug in it, and was raising the wrong
1680
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1681
tree.print_file(file_id)
1683
def get_transaction(self):
1684
return self.control_files.get_transaction()
1686
def revision_parents(self, revision_id):
1687
return self.get_inventory_weave().parent_names(revision_id)
1689
@deprecated_method(symbol_versioning.one_one)
1690
def get_parents(self, revision_ids):
1691
"""See StackedParentsProvider.get_parents"""
1692
parent_map = self.get_parent_map(revision_ids)
1693
return [parent_map.get(r, None) for r in revision_ids]
1695
def get_parent_map(self, keys):
1696
"""See graph._StackedParentsProvider.get_parent_map"""
1698
for revision_id in keys:
1699
if revision_id == _mod_revision.NULL_REVISION:
1700
parent_map[revision_id] = ()
1703
parent_id_list = self.get_revision(revision_id).parent_ids
1704
except errors.NoSuchRevision:
1707
if len(parent_id_list) == 0:
1708
parent_ids = (_mod_revision.NULL_REVISION,)
1710
parent_ids = tuple(parent_id_list)
1711
parent_map[revision_id] = parent_ids
1714
def _make_parents_provider(self):
1717
def get_graph(self, other_repository=None):
1718
"""Return the graph walker for this repository format"""
1719
parents_provider = self._make_parents_provider()
1720
if (other_repository is not None and
1721
other_repository.bzrdir.transport.base !=
1722
self.bzrdir.transport.base):
1723
parents_provider = graph._StackedParentsProvider(
1724
[parents_provider, other_repository._make_parents_provider()])
1725
return graph.Graph(parents_provider)
1727
def _get_versioned_file_checker(self):
1728
"""Return an object suitable for checking versioned files."""
1729
return _VersionedFileChecker(self)
1732
def set_make_working_trees(self, new_value):
1733
"""Set the policy flag for making working trees when creating branches.
1735
This only applies to branches that use this repository.
1737
The default is 'True'.
1738
:param new_value: True to restore the default, False to disable making
1741
raise NotImplementedError(self.set_make_working_trees)
1743
def make_working_trees(self):
1744
"""Returns the policy for making working trees on new branches."""
1745
raise NotImplementedError(self.make_working_trees)
1748
def sign_revision(self, revision_id, gpg_strategy):
1749
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1750
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1753
def has_signature_for_revision_id(self, revision_id):
1754
"""Query for a revision signature for revision_id in the repository."""
1755
return self._revision_store.has_signature(revision_id,
1756
self.get_transaction())
1759
def get_signature_text(self, revision_id):
1760
"""Return the text for a signature."""
1761
return self._revision_store.get_signature_text(revision_id,
1762
self.get_transaction())
1765
def check(self, revision_ids=None):
1766
"""Check consistency of all history of given revision_ids.
1768
Different repository implementations should override _check().
1770
:param revision_ids: A non-empty list of revision_ids whose ancestry
1771
will be checked. Typically the last revision_id of a branch.
1773
return self._check(revision_ids)
1775
def _check(self, revision_ids):
1776
result = check.Check(self)
1780
def _warn_if_deprecated(self):
1781
global _deprecation_warning_done
1782
if _deprecation_warning_done:
1784
_deprecation_warning_done = True
1785
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1786
% (self._format, self.bzrdir.transport.base))
1788
def supports_rich_root(self):
1789
return self._format.rich_root_data
1791
def _check_ascii_revisionid(self, revision_id, method):
1792
"""Private helper for ascii-only repositories."""
1793
# weave repositories refuse to store revisionids that are non-ascii.
1794
if revision_id is not None:
1795
# weaves require ascii revision ids.
1796
if isinstance(revision_id, unicode):
1798
revision_id.encode('ascii')
1799
except UnicodeEncodeError:
1800
raise errors.NonAsciiRevisionId(method, self)
1803
revision_id.decode('ascii')
1804
except UnicodeDecodeError:
1805
raise errors.NonAsciiRevisionId(method, self)
1807
def revision_graph_can_have_wrong_parents(self):
1808
"""Is it possible for this repository to have a revision graph with
1811
If True, then this repository must also implement
1812
_find_inconsistent_revision_parents so that check and reconcile can
1813
check for inconsistencies before proceeding with other checks that may
1814
depend on the revision index being consistent.
1816
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1818
# remove these delegates a while after bzr 0.15
1819
def __make_delegated(name, from_module):
1820
def _deprecated_repository_forwarder():
1821
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1822
% (name, from_module),
1825
m = __import__(from_module, globals(), locals(), [name])
1827
return getattr(m, name)
1828
except AttributeError:
1829
raise AttributeError('module %s has no name %s'
1831
globals()[name] = _deprecated_repository_forwarder
1834
'AllInOneRepository',
1835
'WeaveMetaDirRepository',
1836
'PreSplitOutRepositoryFormat',
1837
'RepositoryFormat4',
1838
'RepositoryFormat5',
1839
'RepositoryFormat6',
1840
'RepositoryFormat7',
1842
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1846
'RepositoryFormatKnit',
1847
'RepositoryFormatKnit1',
1849
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1852
def install_revision(repository, rev, revision_tree):
1853
"""Install all revision data into a repository."""
1854
install_revisions(repository, [(rev, revision_tree, None)])
1857
def install_revisions(repository, iterable):
1858
"""Install all revision data into a repository.
1860
Accepts an iterable of revision, tree, signature tuples. The signature
1863
repository.start_write_group()
1865
for revision, revision_tree, signature in iterable:
1866
_install_revision(repository, revision, revision_tree, signature)
1868
repository.abort_write_group()
1871
repository.commit_write_group()
1874
def _install_revision(repository, rev, revision_tree, signature):
1875
"""Install all revision data into a repository."""
1876
present_parents = []
1878
for p_id in rev.parent_ids:
1879
if repository.has_revision(p_id):
1880
present_parents.append(p_id)
1881
parent_trees[p_id] = repository.revision_tree(p_id)
1883
parent_trees[p_id] = repository.revision_tree(None)
1885
inv = revision_tree.inventory
1886
entries = inv.iter_entries()
1887
# backwards compatibility hack: skip the root id.
1888
if not repository.supports_rich_root():
1889
path, root = entries.next()
1890
if root.revision != rev.revision_id:
1891
raise errors.IncompatibleRevision(repr(repository))
1892
# Add the texts that are not already present
1893
for path, ie in entries:
1894
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1895
repository.get_transaction())
1896
if ie.revision not in w:
1898
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1899
# with InventoryEntry.find_previous_heads(). if it is, then there
1900
# is a latent bug here where the parents may have ancestors of each
1902
for revision, tree in parent_trees.iteritems():
1903
if ie.file_id not in tree:
1905
parent_id = tree.inventory[ie.file_id].revision
1906
if parent_id in text_parents:
1908
text_parents.append(parent_id)
1910
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1911
repository.get_transaction())
1912
lines = revision_tree.get_file(ie.file_id).readlines()
1913
vfile.add_lines(rev.revision_id, text_parents, lines)
1915
# install the inventory
1916
repository.add_inventory(rev.revision_id, inv, present_parents)
1917
except errors.RevisionAlreadyPresent:
1919
if signature is not None:
1920
repository.add_signature_text(rev.revision_id, signature)
1921
repository.add_revision(rev.revision_id, rev, inv)
1924
class MetaDirRepository(Repository):
1925
"""Repositories in the new meta-dir layout."""
1927
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1928
super(MetaDirRepository, self).__init__(_format,
1934
dir_mode = self.control_files._dir_mode
1935
file_mode = self.control_files._file_mode
1938
def is_shared(self):
1939
"""Return True if this repository is flagged as a shared repository."""
1940
return self.control_files._transport.has('shared-storage')
1943
def set_make_working_trees(self, new_value):
1944
"""Set the policy flag for making working trees when creating branches.
1946
This only applies to branches that use this repository.
1948
The default is 'True'.
1949
:param new_value: True to restore the default, False to disable making
1954
self.control_files._transport.delete('no-working-trees')
1955
except errors.NoSuchFile:
1958
self.control_files.put_utf8('no-working-trees', '')
1960
def make_working_trees(self):
1961
"""Returns the policy for making working trees on new branches."""
1962
return not self.control_files._transport.has('no-working-trees')
1965
class RepositoryFormatRegistry(registry.Registry):
1966
"""Registry of RepositoryFormats."""
1968
def get(self, format_string):
1969
r = registry.Registry.get(self, format_string)
1975
format_registry = RepositoryFormatRegistry()
1976
"""Registry of formats, indexed by their identifying format string.
1978
This can contain either format instances themselves, or classes/factories that
1979
can be called to obtain one.
1983
#####################################################################
1984
# Repository Formats
1986
class RepositoryFormat(object):
1987
"""A repository format.
1989
Formats provide three things:
1990
* An initialization routine to construct repository data on disk.
1991
* a format string which is used when the BzrDir supports versioned
1993
* an open routine which returns a Repository instance.
1995
There is one and only one Format subclass for each on-disk format. But
1996
there can be one Repository subclass that is used for several different
1997
formats. The _format attribute on a Repository instance can be used to
1998
determine the disk format.
2000
Formats are placed in an dict by their format string for reference
2001
during opening. These should be subclasses of RepositoryFormat
2004
Once a format is deprecated, just deprecate the initialize and open
2005
methods on the format class. Do not deprecate the object, as the
2006
object will be created every system load.
2008
Common instance attributes:
2009
_matchingbzrdir - the bzrdir format that the repository format was
2010
originally written to work with. This can be used if manually
2011
constructing a bzrdir and repository, or more commonly for test suite
2015
# Set to True or False in derived classes. True indicates that the format
2016
# supports ghosts gracefully.
2017
supports_ghosts = None
2020
return "<%s>" % self.__class__.__name__
2022
def __eq__(self, other):
2023
# format objects are generally stateless
2024
return isinstance(other, self.__class__)
2026
def __ne__(self, other):
2027
return not self == other
2030
def find_format(klass, a_bzrdir):
2031
"""Return the format for the repository object in a_bzrdir.
2033
This is used by bzr native formats that have a "format" file in
2034
the repository. Other methods may be used by different types of
2038
transport = a_bzrdir.get_repository_transport(None)
2039
format_string = transport.get("format").read()
2040
return format_registry.get(format_string)
2041
except errors.NoSuchFile:
2042
raise errors.NoRepositoryPresent(a_bzrdir)
2044
raise errors.UnknownFormatError(format=format_string)
2047
def register_format(klass, format):
2048
format_registry.register(format.get_format_string(), format)
2051
def unregister_format(klass, format):
2052
format_registry.remove(format.get_format_string())
2055
def get_default_format(klass):
2056
"""Return the current default format."""
2057
from bzrlib import bzrdir
2058
return bzrdir.format_registry.make_bzrdir('default').repository_format
2060
def _get_control_store(self, repo_transport, control_files):
2061
"""Return the control store for this repository."""
2062
raise NotImplementedError(self._get_control_store)
2064
def get_format_string(self):
2065
"""Return the ASCII format string that identifies this format.
2067
Note that in pre format ?? repositories the format string is
2068
not permitted nor written to disk.
2070
raise NotImplementedError(self.get_format_string)
2072
def get_format_description(self):
2073
"""Return the short description for this format."""
2074
raise NotImplementedError(self.get_format_description)
2076
def _get_revision_store(self, repo_transport, control_files):
2077
"""Return the revision store object for this a_bzrdir."""
2078
raise NotImplementedError(self._get_revision_store)
2080
def _get_text_rev_store(self,
2087
"""Common logic for getting a revision store for a repository.
2089
see self._get_revision_store for the subclass-overridable method to
2090
get the store for a repository.
2092
from bzrlib.store.revision.text import TextRevisionStore
2093
dir_mode = control_files._dir_mode
2094
file_mode = control_files._file_mode
2095
text_store = TextStore(transport.clone(name),
2097
compressed=compressed,
2099
file_mode=file_mode)
2100
_revision_store = TextRevisionStore(text_store, serializer)
2101
return _revision_store
2103
# TODO: this shouldn't be in the base class, it's specific to things that
2104
# use weaves or knits -- mbp 20070207
2105
def _get_versioned_file_store(self,
2110
versionedfile_class=None,
2111
versionedfile_kwargs={},
2113
if versionedfile_class is None:
2114
versionedfile_class = self._versionedfile_class
2115
weave_transport = control_files._transport.clone(name)
2116
dir_mode = control_files._dir_mode
2117
file_mode = control_files._file_mode
2118
return VersionedFileStore(weave_transport, prefixed=prefixed,
2120
file_mode=file_mode,
2121
versionedfile_class=versionedfile_class,
2122
versionedfile_kwargs=versionedfile_kwargs,
2125
def initialize(self, a_bzrdir, shared=False):
2126
"""Initialize a repository of this format in a_bzrdir.
2128
:param a_bzrdir: The bzrdir to put the new repository in it.
2129
:param shared: The repository should be initialized as a sharable one.
2130
:returns: The new repository object.
2132
This may raise UninitializableFormat if shared repository are not
2133
compatible the a_bzrdir.
2135
raise NotImplementedError(self.initialize)
2137
def is_supported(self):
2138
"""Is this format supported?
2140
Supported formats must be initializable and openable.
2141
Unsupported formats may not support initialization or committing or
2142
some other features depending on the reason for not being supported.
2146
def check_conversion_target(self, target_format):
2147
raise NotImplementedError(self.check_conversion_target)
2149
def open(self, a_bzrdir, _found=False):
2150
"""Return an instance of this format for the bzrdir a_bzrdir.
2152
_found is a private parameter, do not use it.
2154
raise NotImplementedError(self.open)
2157
class MetaDirRepositoryFormat(RepositoryFormat):
2158
"""Common base class for the new repositories using the metadir layout."""
2160
rich_root_data = False
2161
supports_tree_reference = False
2162
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
2165
super(MetaDirRepositoryFormat, self).__init__()
2167
def _create_control_files(self, a_bzrdir):
2168
"""Create the required files and the initial control_files object."""
2169
# FIXME: RBC 20060125 don't peek under the covers
2170
# NB: no need to escape relative paths that are url safe.
2171
repository_transport = a_bzrdir.get_repository_transport(self)
2172
control_files = lockable_files.LockableFiles(repository_transport,
2173
'lock', lockdir.LockDir)
2174
control_files.create_lock()
2175
return control_files
2177
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2178
"""Upload the initial blank content."""
2179
control_files = self._create_control_files(a_bzrdir)
2180
control_files.lock_write()
2182
control_files._transport.mkdir_multi(dirs,
2183
mode=control_files._dir_mode)
2184
for file, content in files:
2185
control_files.put(file, content)
2186
for file, content in utf8_files:
2187
control_files.put_utf8(file, content)
2189
control_files.put_utf8('shared-storage', '')
2191
control_files.unlock()
2194
# formats which have no format string are not discoverable
2195
# and not independently creatable, so are not registered. They're
2196
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2197
# needed, it's constructed directly by the BzrDir. Non-native formats where
2198
# the repository is not separately opened are similar.
2200
format_registry.register_lazy(
2201
'Bazaar-NG Repository format 7',
2202
'bzrlib.repofmt.weaverepo',
2206
format_registry.register_lazy(
2207
'Bazaar-NG Knit Repository Format 1',
2208
'bzrlib.repofmt.knitrepo',
2209
'RepositoryFormatKnit1',
2212
format_registry.register_lazy(
2213
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2214
'bzrlib.repofmt.knitrepo',
2215
'RepositoryFormatKnit3',
2218
format_registry.register_lazy(
2219
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2220
'bzrlib.repofmt.knitrepo',
2221
'RepositoryFormatKnit4',
2224
# Pack-based formats. There is one format for pre-subtrees, and one for
2225
# post-subtrees to allow ease of testing.
2226
# NOTE: These are experimental in 0.92.
2227
format_registry.register_lazy(
2228
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2229
'bzrlib.repofmt.pack_repo',
2230
'RepositoryFormatKnitPack1',
2232
format_registry.register_lazy(
2233
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2234
'bzrlib.repofmt.pack_repo',
2235
'RepositoryFormatKnitPack3',
2237
format_registry.register_lazy(
2238
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2239
'bzrlib.repofmt.pack_repo',
2240
'RepositoryFormatKnitPack4',
2244
class InterRepository(InterObject):
2245
"""This class represents operations taking place between two repositories.
2247
Its instances have methods like copy_content and fetch, and contain
2248
references to the source and target repositories these operations can be
2251
Often we will provide convenience methods on 'repository' which carry out
2252
operations with another repository - they will always forward to
2253
InterRepository.get(other).method_name(parameters).
2257
"""The available optimised InterRepository types."""
2259
def copy_content(self, revision_id=None):
2260
raise NotImplementedError(self.copy_content)
2262
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2263
"""Fetch the content required to construct revision_id.
2265
The content is copied from self.source to self.target.
2267
:param revision_id: if None all content is copied, if NULL_REVISION no
2269
:param pb: optional progress bar to use for progress reports. If not
2270
provided a default one will be created.
2272
Returns the copied revision count and the failed revisions in a tuple:
2275
raise NotImplementedError(self.fetch)
2278
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2279
"""Return the revision ids that source has that target does not.
2281
These are returned in topological order.
2283
:param revision_id: only return revision ids included by this
2286
# generic, possibly worst case, slow code path.
2287
target_ids = set(self.target.all_revision_ids())
2288
if revision_id is not None:
2289
source_ids = self.source.get_ancestry(revision_id)
2290
assert source_ids[0] is None
2293
source_ids = self.source.all_revision_ids()
2294
result_set = set(source_ids).difference(target_ids)
2295
# this may look like a no-op: its not. It preserves the ordering
2296
# other_ids had while only returning the members from other_ids
2297
# that we've decided we need.
2298
return [rev_id for rev_id in source_ids if rev_id in result_set]
2301
def _same_model(source, target):
2302
"""True if source and target have the same data representation."""
2303
if source.supports_rich_root() != target.supports_rich_root():
2305
if source._serializer != target._serializer:
2310
class InterSameDataRepository(InterRepository):
2311
"""Code for converting between repositories that represent the same data.
2313
Data format and model must match for this to work.
2317
def _get_repo_format_to_test(self):
2318
"""Repository format for testing with.
2320
InterSameData can pull from subtree to subtree and from non-subtree to
2321
non-subtree, so we test this with the richest repository format.
2323
from bzrlib.repofmt import knitrepo
2324
return knitrepo.RepositoryFormatKnit3()
2327
def is_compatible(source, target):
2328
return InterRepository._same_model(source, target)
2331
def copy_content(self, revision_id=None):
2332
"""Make a complete copy of the content in self into destination.
2334
This copies both the repository's revision data, and configuration information
2335
such as the make_working_trees setting.
2337
This is a destructive operation! Do not use it on existing
2340
:param revision_id: Only copy the content needed to construct
2341
revision_id and its parents.
2344
self.target.set_make_working_trees(self.source.make_working_trees())
2345
except NotImplementedError:
2347
# but don't bother fetching if we have the needed data now.
2348
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2349
self.target.has_revision(revision_id)):
2351
self.target.fetch(self.source, revision_id=revision_id)
2354
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2355
"""See InterRepository.fetch()."""
2356
from bzrlib.fetch import GenericRepoFetcher
2357
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2358
self.source, self.source._format, self.target,
2359
self.target._format)
2360
f = GenericRepoFetcher(to_repository=self.target,
2361
from_repository=self.source,
2362
last_revision=revision_id,
2364
return f.count_copied, f.failed_revisions
2367
class InterWeaveRepo(InterSameDataRepository):
2368
"""Optimised code paths between Weave based repositories.
2370
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2371
implemented lazy inter-object optimisation.
2375
def _get_repo_format_to_test(self):
2376
from bzrlib.repofmt import weaverepo
2377
return weaverepo.RepositoryFormat7()
2380
def is_compatible(source, target):
2381
"""Be compatible with known Weave formats.
2383
We don't test for the stores being of specific types because that
2384
could lead to confusing results, and there is no need to be
2387
from bzrlib.repofmt.weaverepo import (
2393
return (isinstance(source._format, (RepositoryFormat5,
2395
RepositoryFormat7)) and
2396
isinstance(target._format, (RepositoryFormat5,
2398
RepositoryFormat7)))
2399
except AttributeError:
2403
def copy_content(self, revision_id=None):
2404
"""See InterRepository.copy_content()."""
2405
# weave specific optimised path:
2407
self.target.set_make_working_trees(self.source.make_working_trees())
2408
except NotImplementedError:
2410
# FIXME do not peek!
2411
if self.source.control_files._transport.listable():
2412
pb = ui.ui_factory.nested_progress_bar()
2414
self.target.weave_store.copy_all_ids(
2415
self.source.weave_store,
2417
from_transaction=self.source.get_transaction(),
2418
to_transaction=self.target.get_transaction())
2419
pb.update('copying inventory', 0, 1)
2420
self.target.control_weaves.copy_multi(
2421
self.source.control_weaves, ['inventory'],
2422
from_transaction=self.source.get_transaction(),
2423
to_transaction=self.target.get_transaction())
2424
self.target._revision_store.text_store.copy_all_ids(
2425
self.source._revision_store.text_store,
2430
self.target.fetch(self.source, revision_id=revision_id)
2433
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2434
"""See InterRepository.fetch()."""
2435
from bzrlib.fetch import GenericRepoFetcher
2436
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2437
self.source, self.source._format, self.target, self.target._format)
2438
f = GenericRepoFetcher(to_repository=self.target,
2439
from_repository=self.source,
2440
last_revision=revision_id,
2442
return f.count_copied, f.failed_revisions
2445
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2446
"""See InterRepository.missing_revision_ids()."""
2447
# we want all revisions to satisfy revision_id in source.
2448
# but we don't want to stat every file here and there.
2449
# we want then, all revisions other needs to satisfy revision_id
2450
# checked, but not those that we have locally.
2451
# so the first thing is to get a subset of the revisions to
2452
# satisfy revision_id in source, and then eliminate those that
2453
# we do already have.
2454
# this is slow on high latency connection to self, but as as this
2455
# disk format scales terribly for push anyway due to rewriting
2456
# inventory.weave, this is considered acceptable.
2458
if revision_id is not None:
2459
source_ids = self.source.get_ancestry(revision_id)
2460
assert source_ids[0] is None
2463
source_ids = self.source._all_possible_ids()
2464
source_ids_set = set(source_ids)
2465
# source_ids is the worst possible case we may need to pull.
2466
# now we want to filter source_ids against what we actually
2467
# have in target, but don't try to check for existence where we know
2468
# we do not have a revision as that would be pointless.
2469
target_ids = set(self.target._all_possible_ids())
2470
possibly_present_revisions = target_ids.intersection(source_ids_set)
2471
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2472
required_revisions = source_ids_set.difference(actually_present_revisions)
2473
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2474
if revision_id is not None:
2475
# we used get_ancestry to determine source_ids then we are assured all
2476
# revisions referenced are present as they are installed in topological order.
2477
# and the tip revision was validated by get_ancestry.
2478
return required_topo_revisions
2480
# if we just grabbed the possibly available ids, then
2481
# we only have an estimate of whats available and need to validate
2482
# that against the revision records.
2483
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2486
class InterKnitRepo(InterSameDataRepository):
2487
"""Optimised code paths between Knit based repositories."""
2490
def _get_repo_format_to_test(self):
2491
from bzrlib.repofmt import knitrepo
2492
return knitrepo.RepositoryFormatKnit1()
2495
def is_compatible(source, target):
2496
"""Be compatible with known Knit formats.
2498
We don't test for the stores being of specific types because that
2499
could lead to confusing results, and there is no need to be
2502
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2504
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2505
isinstance(target._format, RepositoryFormatKnit))
2506
except AttributeError:
2508
return are_knits and InterRepository._same_model(source, target)
2511
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2512
"""See InterRepository.fetch()."""
2513
from bzrlib.fetch import KnitRepoFetcher
2514
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2515
self.source, self.source._format, self.target, self.target._format)
2516
f = KnitRepoFetcher(to_repository=self.target,
2517
from_repository=self.source,
2518
last_revision=revision_id,
2520
return f.count_copied, f.failed_revisions
2523
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2524
"""See InterRepository.missing_revision_ids()."""
2525
if revision_id is not None:
2526
source_ids = self.source.get_ancestry(revision_id)
2527
assert source_ids[0] is None
2530
source_ids = self.source.all_revision_ids()
2531
source_ids_set = set(source_ids)
2532
# source_ids is the worst possible case we may need to pull.
2533
# now we want to filter source_ids against what we actually
2534
# have in target, but don't try to check for existence where we know
2535
# we do not have a revision as that would be pointless.
2536
target_ids = set(self.target.all_revision_ids())
2537
possibly_present_revisions = target_ids.intersection(source_ids_set)
2538
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2539
required_revisions = source_ids_set.difference(actually_present_revisions)
2540
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2541
if revision_id is not None:
2542
# we used get_ancestry to determine source_ids then we are assured all
2543
# revisions referenced are present as they are installed in topological order.
2544
# and the tip revision was validated by get_ancestry.
2545
return required_topo_revisions
2547
# if we just grabbed the possibly available ids, then
2548
# we only have an estimate of whats available and need to validate
2549
# that against the revision records.
2550
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2553
class InterPackRepo(InterSameDataRepository):
2554
"""Optimised code paths between Pack based repositories."""
2557
def _get_repo_format_to_test(self):
2558
from bzrlib.repofmt import pack_repo
2559
return pack_repo.RepositoryFormatKnitPack1()
2562
def is_compatible(source, target):
2563
"""Be compatible with known Pack formats.
2565
We don't test for the stores being of specific types because that
2566
could lead to confusing results, and there is no need to be
2569
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2571
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2572
isinstance(target._format, RepositoryFormatPack))
2573
except AttributeError:
2575
return are_packs and InterRepository._same_model(source, target)
2578
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2579
"""See InterRepository.fetch()."""
2580
from bzrlib.repofmt.pack_repo import Packer
2581
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2582
self.source, self.source._format, self.target, self.target._format)
2583
self.count_copied = 0
2584
if revision_id is None:
2586
# everything to do - use pack logic
2587
# to fetch from all packs to one without
2588
# inventory parsing etc, IFF nothing to be copied is in the target.
2590
revision_ids = self.source.all_revision_ids()
2591
# implementing the TODO will involve:
2592
# - detecting when all of a pack is selected
2593
# - avoiding as much as possible pre-selection, so the
2594
# more-core routines such as create_pack_from_packs can filter in
2595
# a just-in-time fashion. (though having a HEADS list on a
2596
# repository might make this a lot easier, because we could
2597
# sensibly detect 'new revisions' without doing a full index scan.
2598
elif _mod_revision.is_null(revision_id):
2603
revision_ids = self.missing_revision_ids(revision_id,
2604
find_ghosts=find_ghosts)
2605
except errors.NoSuchRevision:
2606
raise errors.InstallFailed([revision_id])
2607
packs = self.source._pack_collection.all_packs()
2608
pack = Packer(self.target._pack_collection, packs, '.fetch',
2609
revision_ids).pack()
2610
if pack is not None:
2611
self.target._pack_collection._save_pack_names()
2612
# Trigger an autopack. This may duplicate effort as we've just done
2613
# a pack creation, but for now it is simpler to think about as
2614
# 'upload data, then repack if needed'.
2615
self.target._pack_collection.autopack()
2616
return (pack.get_revision_count(), [])
2621
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2622
"""See InterRepository.missing_revision_ids().
2624
:param find_ghosts: Find ghosts throughough the ancestry of
2627
if not find_ghosts and revision_id is not None:
2628
graph = self.source.get_graph()
2629
missing_revs = set()
2630
searcher = graph._make_breadth_first_searcher([revision_id])
2632
self.target._pack_collection.revision_index.combined_index
2633
null_set = frozenset([_mod_revision.NULL_REVISION])
2636
next_revs = set(searcher.next())
2637
except StopIteration:
2639
next_revs.difference_update(null_set)
2640
target_keys = [(key,) for key in next_revs]
2641
have_revs = frozenset(node[1][0] for node in
2642
target_index.iter_entries(target_keys))
2643
missing_revs.update(next_revs - have_revs)
2644
searcher.stop_searching_any(have_revs)
2645
if next_revs - have_revs == set([revision_id]):
2646
# we saw the start rev itself, but no parents from it (or
2647
# next_revs would have been updated to e.g. set(). We remove
2648
# have_revs because if we found revision_id locally we
2649
# stop_searching at the first time around.
2650
raise errors.NoSuchRevision(self.source, revision_id)
2652
elif revision_id is not None:
2653
source_ids = self.source.get_ancestry(revision_id)
2654
assert source_ids[0] is None
2657
source_ids = self.source.all_revision_ids()
2658
# source_ids is the worst possible case we may need to pull.
2659
# now we want to filter source_ids against what we actually
2660
# have in target, but don't try to check for existence where we know
2661
# we do not have a revision as that would be pointless.
2662
target_ids = set(self.target.all_revision_ids())
2663
return [r for r in source_ids if (r not in target_ids)]
2666
class InterModel1and2(InterRepository):
2669
def _get_repo_format_to_test(self):
2673
def is_compatible(source, target):
2674
if not source.supports_rich_root() and target.supports_rich_root():
2680
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2681
"""See InterRepository.fetch()."""
2682
from bzrlib.fetch import Model1toKnit2Fetcher
2683
f = Model1toKnit2Fetcher(to_repository=self.target,
2684
from_repository=self.source,
2685
last_revision=revision_id,
2687
return f.count_copied, f.failed_revisions
2690
def copy_content(self, revision_id=None):
2691
"""Make a complete copy of the content in self into destination.
2693
This is a destructive operation! Do not use it on existing
2696
:param revision_id: Only copy the content needed to construct
2697
revision_id and its parents.
2700
self.target.set_make_working_trees(self.source.make_working_trees())
2701
except NotImplementedError:
2703
# but don't bother fetching if we have the needed data now.
2704
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2705
self.target.has_revision(revision_id)):
2707
self.target.fetch(self.source, revision_id=revision_id)
2710
class InterKnit1and2(InterKnitRepo):
2713
def _get_repo_format_to_test(self):
2717
def is_compatible(source, target):
2718
"""Be compatible with Knit1 source and Knit3 target"""
2719
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2721
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2722
RepositoryFormatKnit3)
2723
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2724
RepositoryFormatKnitPack3)
2725
return (isinstance(source._format,
2726
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2727
isinstance(target._format,
2728
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2730
except AttributeError:
2734
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2735
"""See InterRepository.fetch()."""
2736
from bzrlib.fetch import Knit1to2Fetcher
2737
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2738
self.source, self.source._format, self.target,
2739
self.target._format)
2740
f = Knit1to2Fetcher(to_repository=self.target,
2741
from_repository=self.source,
2742
last_revision=revision_id,
2744
return f.count_copied, f.failed_revisions
2747
class InterDifferingSerializer(InterKnitRepo):
2750
def _get_repo_format_to_test(self):
2754
def is_compatible(source, target):
2755
"""Be compatible with Knit2 source and Knit3 target"""
2756
if source.supports_rich_root() != target.supports_rich_root():
2758
# Ideally, we'd support fetching if the source had no tree references
2759
# even if it supported them...
2760
if (getattr(source, '_format.supports_tree_reference', False) and
2761
not getattr(target, '_format.supports_tree_reference', False)):
2766
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2767
"""See InterRepository.fetch()."""
2768
revision_ids = self.target.missing_revision_ids(self.source,
2770
def revisions_iterator():
2771
for current_revision_id in revision_ids:
2772
revision = self.source.get_revision(current_revision_id)
2773
tree = self.source.revision_tree(current_revision_id)
2775
signature = self.source.get_signature_text(
2776
current_revision_id)
2777
except errors.NoSuchRevision:
2779
yield revision, tree, signature
2780
install_revisions(self.target, revisions_iterator())
2781
return len(revision_ids), 0
2784
class InterRemoteToOther(InterRepository):
2786
def __init__(self, source, target):
2787
InterRepository.__init__(self, source, target)
2788
self._real_inter = None
2791
def is_compatible(source, target):
2792
if not isinstance(source, remote.RemoteRepository):
2794
source._ensure_real()
2795
real_source = source._real_repository
2796
# Is source's model compatible with target's model, and are they the
2797
# same format? Currently we can only optimise fetching from an
2798
# identical model & format repo.
2799
assert not isinstance(real_source, remote.RemoteRepository), (
2800
"We don't support remote repos backed by remote repos yet.")
2801
return real_source._format == target._format
2804
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2805
"""See InterRepository.fetch()."""
2806
from bzrlib.fetch import RemoteToOtherFetcher
2807
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2808
self.source, self.target, self.target._format)
2809
# TODO: jam 20070210 This should be an assert, not a translate
2810
revision_id = osutils.safe_revision_id(revision_id)
2811
f = RemoteToOtherFetcher(to_repository=self.target,
2812
from_repository=self.source,
2813
last_revision=revision_id,
2815
return f.count_copied, f.failed_revisions
2818
def _get_repo_format_to_test(self):
2822
class InterOtherToRemote(InterRepository):
2824
def __init__(self, source, target):
2825
InterRepository.__init__(self, source, target)
2826
self._real_inter = None
2829
def is_compatible(source, target):
2830
if isinstance(target, remote.RemoteRepository):
2834
def _ensure_real_inter(self):
2835
if self._real_inter is None:
2836
self.target._ensure_real()
2837
real_target = self.target._real_repository
2838
self._real_inter = InterRepository.get(self.source, real_target)
2840
def copy_content(self, revision_id=None):
2841
self._ensure_real_inter()
2842
self._real_inter.copy_content(revision_id=revision_id)
2844
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2845
self._ensure_real_inter()
2846
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2849
def _get_repo_format_to_test(self):
2853
InterRepository.register_optimiser(InterDifferingSerializer)
2854
InterRepository.register_optimiser(InterSameDataRepository)
2855
InterRepository.register_optimiser(InterWeaveRepo)
2856
InterRepository.register_optimiser(InterKnitRepo)
2857
InterRepository.register_optimiser(InterModel1and2)
2858
InterRepository.register_optimiser(InterKnit1and2)
2859
InterRepository.register_optimiser(InterPackRepo)
2860
InterRepository.register_optimiser(InterRemoteToOther)
2861
InterRepository.register_optimiser(InterOtherToRemote)
2864
class CopyConverter(object):
2865
"""A repository conversion tool which just performs a copy of the content.
2867
This is slow but quite reliable.
2870
def __init__(self, target_format):
2871
"""Create a CopyConverter.
2873
:param target_format: The format the resulting repository should be.
2875
self.target_format = target_format
2877
def convert(self, repo, pb):
2878
"""Perform the conversion of to_convert, giving feedback via pb.
2880
:param to_convert: The disk object to convert.
2881
:param pb: a progress bar to use for progress information.
2886
# this is only useful with metadir layouts - separated repo content.
2887
# trigger an assertion if not such
2888
repo._format.get_format_string()
2889
self.repo_dir = repo.bzrdir
2890
self.step('Moving repository to repository.backup')
2891
self.repo_dir.transport.move('repository', 'repository.backup')
2892
backup_transport = self.repo_dir.transport.clone('repository.backup')
2893
repo._format.check_conversion_target(self.target_format)
2894
self.source_repo = repo._format.open(self.repo_dir,
2896
_override_transport=backup_transport)
2897
self.step('Creating new repository')
2898
converted = self.target_format.initialize(self.repo_dir,
2899
self.source_repo.is_shared())
2900
converted.lock_write()
2902
self.step('Copying content into repository.')
2903
self.source_repo.copy_content_into(converted)
2906
self.step('Deleting old repository content.')
2907
self.repo_dir.transport.delete_tree('repository.backup')
2908
self.pb.note('repository converted')
2910
def step(self, message):
2911
"""Update the pb by a step."""
2913
self.pb.update(message, self.count, self.total)
2925
def _unescaper(match, _map=_unescape_map):
2926
code = match.group(1)
2930
if not code.startswith('#'):
2932
return unichr(int(code[1:])).encode('utf8')
2938
def _unescape_xml(data):
2939
"""Unescape predefined XML entities in a string of data."""
2941
if _unescape_re is None:
2942
_unescape_re = re.compile('\&([^;]*);')
2943
return _unescape_re.sub(_unescaper, data)
2946
class _VersionedFileChecker(object):
2948
def __init__(self, repository):
2949
self.repository = repository
2950
self.text_index = self.repository._generate_text_key_index()
2952
def calculate_file_version_parents(self, revision_id, file_id):
2953
"""Calculate the correct parents for a file version according to
2956
parent_keys = self.text_index[(file_id, revision_id)]
2957
if parent_keys == [_mod_revision.NULL_REVISION]:
2959
# strip the file_id, for the weave api
2960
return tuple([revision_id for file_id, revision_id in parent_keys])
2962
def check_file_version_parents(self, weave, file_id):
2963
"""Check the parents stored in a versioned file are correct.
2965
It also detects file versions that are not referenced by their
2966
corresponding revision's inventory.
2968
:returns: A tuple of (wrong_parents, dangling_file_versions).
2969
wrong_parents is a dict mapping {revision_id: (stored_parents,
2970
correct_parents)} for each revision_id where the stored parents
2971
are not correct. dangling_file_versions is a set of (file_id,
2972
revision_id) tuples for versions that are present in this versioned
2973
file, but not used by the corresponding inventory.
2976
unused_versions = set()
2977
for num, revision_id in enumerate(weave.versions()):
2979
correct_parents = self.calculate_file_version_parents(
2980
revision_id, file_id)
2982
# The version is not part of the used keys.
2983
unused_versions.add(revision_id)
2986
knit_parents = tuple(weave.get_parents(revision_id))
2987
except errors.RevisionNotPresent:
2989
if correct_parents != knit_parents:
2990
wrong_parents[revision_id] = (knit_parents, correct_parents)
2991
return wrong_parents, unused_versions
added
removed
bzrlib/repository.py
