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 validator(which is a sha1 digest, though what is sha'd is
506
repository format specific) of the serialized inventory.
508
assert self.is_in_write_group()
509
_mod_revision.check_not_reserved_id(revision_id)
510
assert inv.revision_id is None or inv.revision_id == revision_id, \
511
"Mismatch between inventory revision" \
512
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
513
assert inv.root is not None
514
inv_lines = self._serialise_inventory_to_lines(inv)
515
inv_vf = self.get_inventory_weave()
516
return self._inventory_add_lines(inv_vf, revision_id, parents,
517
inv_lines, check_content=False)
519
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
521
"""Store lines in inv_vf and return the sha1 of the inventory."""
523
for parent in parents:
525
final_parents.append(parent)
526
return inv_vf.add_lines(revision_id, final_parents, lines,
527
check_content=check_content)[0]
530
def add_revision(self, revision_id, rev, inv=None, config=None):
531
"""Add rev to the revision store as revision_id.
533
:param revision_id: the revision id to use.
534
:param rev: The revision object.
535
:param inv: The inventory for the revision. if None, it will be looked
536
up in the inventory storer
537
:param config: If None no digital signature will be created.
538
If supplied its signature_needed method will be used
539
to determine if a signature should be made.
541
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
543
_mod_revision.check_not_reserved_id(revision_id)
544
if config is not None and config.signature_needed():
546
inv = self.get_inventory(revision_id)
547
plaintext = Testament(rev, inv).as_short_text()
548
self.store_revision_signature(
549
gpg.GPGStrategy(config), plaintext, revision_id)
550
if not revision_id in self.get_inventory_weave():
552
raise errors.WeaveRevisionNotPresent(revision_id,
553
self.get_inventory_weave())
555
# yes, this is not suitable for adding with ghosts.
556
self.add_inventory(revision_id, inv, rev.parent_ids)
557
self._revision_store.add_revision(rev, self.get_transaction())
559
def _add_revision_text(self, revision_id, text):
560
revision = self._revision_store._serializer.read_revision_from_string(
562
self._revision_store._add_revision(revision, StringIO(text),
563
self.get_transaction())
565
def all_revision_ids(self):
566
"""Returns a list of all the revision ids in the repository.
568
This is deprecated because code should generally work on the graph
569
reachable from a particular revision, and ignore any other revisions
570
that might be present. There is no direct replacement method.
572
if 'evil' in debug.debug_flags:
573
mutter_callsite(2, "all_revision_ids is linear with history.")
574
return self._all_revision_ids()
576
def _all_revision_ids(self):
577
"""Returns a list of all the revision ids in the repository.
579
These are in as much topological order as the underlying store can
582
raise NotImplementedError(self._all_revision_ids)
584
def break_lock(self):
585
"""Break a lock if one is present from another instance.
587
Uses the ui factory to ask for confirmation if the lock may be from
590
self.control_files.break_lock()
593
def _eliminate_revisions_not_present(self, revision_ids):
594
"""Check every revision id in revision_ids to see if we have it.
596
Returns a set of the present revisions.
599
for id in revision_ids:
600
if self.has_revision(id):
605
def create(a_bzrdir):
606
"""Construct the current default format repository in a_bzrdir."""
607
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
609
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
610
"""instantiate a Repository.
612
:param _format: The format of the repository on disk.
613
:param a_bzrdir: The BzrDir of the repository.
615
In the future we will have a single api for all stores for
616
getting file texts, inventories and revisions, then
617
this construct will accept instances of those things.
619
super(Repository, self).__init__()
620
self._format = _format
621
# the following are part of the public API for Repository:
622
self.bzrdir = a_bzrdir
623
self.control_files = control_files
624
self._revision_store = _revision_store
625
# backwards compatibility
626
self.weave_store = text_store
628
self._reconcile_does_inventory_gc = True
629
self._reconcile_fixes_text_parents = False
630
self._reconcile_backsup_inventory = True
631
# not right yet - should be more semantically clear ?
633
self.control_store = control_store
634
self.control_weaves = control_store
635
# TODO: make sure to construct the right store classes, etc, depending
636
# on whether escaping is required.
637
self._warn_if_deprecated()
638
self._write_group = None
639
self.base = control_files._transport.base
642
return '%s(%r)' % (self.__class__.__name__,
645
def has_same_location(self, other):
646
"""Returns a boolean indicating if this repository is at the same
647
location as another repository.
649
This might return False even when two repository objects are accessing
650
the same physical repository via different URLs.
652
if self.__class__ is not other.__class__:
654
return (self.control_files._transport.base ==
655
other.control_files._transport.base)
657
def is_in_write_group(self):
658
"""Return True if there is an open write group.
660
:seealso: start_write_group.
662
return self._write_group is not None
665
return self.control_files.is_locked()
667
def is_write_locked(self):
668
"""Return True if this object is write locked."""
669
return self.is_locked() and self.control_files._lock_mode == 'w'
671
def lock_write(self, token=None):
672
"""Lock this repository for writing.
674
This causes caching within the repository obejct to start accumlating
675
data during reads, and allows a 'write_group' to be obtained. Write
676
groups must be used for actual data insertion.
678
:param token: if this is already locked, then lock_write will fail
679
unless the token matches the existing lock.
680
:returns: a token if this instance supports tokens, otherwise None.
681
:raises TokenLockingNotSupported: when a token is given but this
682
instance doesn't support using token locks.
683
:raises MismatchedToken: if the specified token doesn't match the token
684
of the existing lock.
685
:seealso: start_write_group.
687
A token should be passed in if you know that you have locked the object
688
some other way, and need to synchronise this object's state with that
691
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
693
result = self.control_files.lock_write(token=token)
698
self.control_files.lock_read()
701
def get_physical_lock_status(self):
702
return self.control_files.get_physical_lock_status()
704
def leave_lock_in_place(self):
705
"""Tell this repository not to release the physical lock when this
708
If lock_write doesn't return a token, then this method is not supported.
710
self.control_files.leave_in_place()
712
def dont_leave_lock_in_place(self):
713
"""Tell this repository to release the physical lock when this
714
object is unlocked, even if it didn't originally acquire it.
716
If lock_write doesn't return a token, then this method is not supported.
718
self.control_files.dont_leave_in_place()
721
def gather_stats(self, revid=None, committers=None):
722
"""Gather statistics from a revision id.
724
:param revid: The revision id to gather statistics from, if None, then
725
no revision specific statistics are gathered.
726
:param committers: Optional parameter controlling whether to grab
727
a count of committers from the revision specific statistics.
728
:return: A dictionary of statistics. Currently this contains:
729
committers: The number of committers if requested.
730
firstrev: A tuple with timestamp, timezone for the penultimate left
731
most ancestor of revid, if revid is not the NULL_REVISION.
732
latestrev: A tuple with timestamp, timezone for revid, if revid is
733
not the NULL_REVISION.
734
revisions: The total revision count in the repository.
735
size: An estimate disk size of the repository in bytes.
738
if revid and committers:
739
result['committers'] = 0
740
if revid and revid != _mod_revision.NULL_REVISION:
742
all_committers = set()
743
revisions = self.get_ancestry(revid)
744
# pop the leading None
746
first_revision = None
748
# ignore the revisions in the middle - just grab first and last
749
revisions = revisions[0], revisions[-1]
750
for revision in self.get_revisions(revisions):
751
if not first_revision:
752
first_revision = revision
754
all_committers.add(revision.committer)
755
last_revision = revision
757
result['committers'] = len(all_committers)
758
result['firstrev'] = (first_revision.timestamp,
759
first_revision.timezone)
760
result['latestrev'] = (last_revision.timestamp,
761
last_revision.timezone)
763
# now gather global repository information
764
if self.bzrdir.root_transport.listable():
765
c, t = self._revision_store.total_size(self.get_transaction())
766
result['revisions'] = c
770
def find_branches(self, using=False):
771
"""Find branches underneath this repository.
773
This will include branches inside other branches.
775
:param using: If True, list only branches using this repository.
777
if using and not self.is_shared():
779
return [self.bzrdir.open_branch()]
780
except errors.NotBranchError:
782
class Evaluator(object):
785
self.first_call = True
787
def __call__(self, bzrdir):
788
# On the first call, the parameter is always the bzrdir
789
# containing the current repo.
790
if not self.first_call:
792
repository = bzrdir.open_repository()
793
except errors.NoRepositoryPresent:
796
return False, (None, repository)
797
self.first_call = False
799
value = (bzrdir.open_branch(), None)
800
except errors.NotBranchError:
805
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
806
self.bzrdir.root_transport, evaluate=Evaluator()):
807
if branch is not None:
808
branches.append(branch)
809
if not using and repository is not None:
810
branches.extend(repository.find_branches())
813
def get_data_stream(self, revision_ids):
814
raise NotImplementedError(self.get_data_stream)
816
def insert_data_stream(self, stream):
817
"""XXX What does this really do?
819
Is it a substitute for fetch?
820
Should it manage its own write group ?
822
for item_key, bytes in stream:
823
if item_key[0] == 'file':
824
(file_id,) = item_key[1:]
825
knit = self.weave_store.get_weave_or_empty(
826
file_id, self.get_transaction())
827
elif item_key == ('inventory',):
828
knit = self.get_inventory_weave()
829
elif item_key == ('revisions',):
830
knit = self._revision_store.get_revision_file(
831
self.get_transaction())
832
elif item_key == ('signatures',):
833
knit = self._revision_store.get_signature_file(
834
self.get_transaction())
836
raise RepositoryDataStreamError(
837
"Unrecognised data stream key '%s'" % (item_key,))
838
decoded_list = bencode.bdecode(bytes)
839
format = decoded_list.pop(0)
842
for version, options, parents, some_bytes in decoded_list:
843
data_list.append((version, options, len(some_bytes), parents))
844
knit_bytes += some_bytes
845
buffer = StringIO(knit_bytes)
846
def reader_func(count):
850
return buffer.read(count)
851
knit.insert_data_stream(
852
(format, data_list, reader_func))
855
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
856
"""Return the revision ids that other has that this does not.
858
These are returned in topological order.
860
revision_id: only return revision ids included by revision_id.
862
return InterRepository.get(other, self).search_missing_revision_ids(
863
revision_id, find_ghosts)
865
@deprecated_method(symbol_versioning.one_two)
867
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
868
"""Return the revision ids that other has that this does not.
870
These are returned in topological order.
872
revision_id: only return revision ids included by revision_id.
874
keys = self.search_missing_revision_ids(
875
other, revision_id, find_ghosts).get_keys()
878
parents = other.get_graph().get_parent_map(keys)
881
return tsort.topo_sort(parents)
885
"""Open the repository rooted at base.
887
For instance, if the repository is at URL/.bzr/repository,
888
Repository.open(URL) -> a Repository instance.
890
control = bzrdir.BzrDir.open(base)
891
return control.open_repository()
893
def copy_content_into(self, destination, revision_id=None):
894
"""Make a complete copy of the content in self into destination.
896
This is a destructive operation! Do not use it on existing
899
return InterRepository.get(self, destination).copy_content(revision_id)
901
def commit_write_group(self):
902
"""Commit the contents accrued within the current write group.
904
:seealso: start_write_group.
906
if self._write_group is not self.get_transaction():
907
# has an unlock or relock occured ?
908
raise errors.BzrError('mismatched lock context %r and '
910
(self.get_transaction(), self._write_group))
911
self._commit_write_group()
912
self._write_group = None
914
def _commit_write_group(self):
915
"""Template method for per-repository write group cleanup.
917
This is called before the write group is considered to be
918
finished and should ensure that all data handed to the repository
919
for writing during the write group is safely committed (to the
920
extent possible considering file system caching etc).
923
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
924
"""Fetch the content required to construct revision_id from source.
926
If revision_id is None all content is copied.
927
:param find_ghosts: Find and copy revisions in the source that are
928
ghosts in the target (and not reachable directly by walking out to
929
the first-present revision in target from revision_id).
931
# fast path same-url fetch operations
932
if self.has_same_location(source):
933
# check that last_revision is in 'from' and then return a
935
if (revision_id is not None and
936
not _mod_revision.is_null(revision_id)):
937
self.get_revision(revision_id)
939
inter = InterRepository.get(source, self)
941
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
942
except NotImplementedError:
943
raise errors.IncompatibleRepositories(source, self)
945
def create_bundle(self, target, base, fileobj, format=None):
946
return serializer.write_bundle(self, target, base, fileobj, format)
948
def get_commit_builder(self, branch, parents, config, timestamp=None,
949
timezone=None, committer=None, revprops=None,
951
"""Obtain a CommitBuilder for this repository.
953
:param branch: Branch to commit to.
954
:param parents: Revision ids of the parents of the new revision.
955
:param config: Configuration to use.
956
:param timestamp: Optional timestamp recorded for commit.
957
:param timezone: Optional timezone for timestamp.
958
:param committer: Optional committer to set for commit.
959
:param revprops: Optional dictionary of revision properties.
960
:param revision_id: Optional revision id.
962
result = self._commit_builder_class(self, parents, config,
963
timestamp, timezone, committer, revprops, revision_id)
964
self.start_write_group()
968
if (self.control_files._lock_count == 1 and
969
self.control_files._lock_mode == 'w'):
970
if self._write_group is not None:
971
self.abort_write_group()
972
self.control_files.unlock()
973
raise errors.BzrError(
974
'Must end write groups before releasing write locks.')
975
self.control_files.unlock()
978
def clone(self, a_bzrdir, revision_id=None):
979
"""Clone this repository into a_bzrdir using the current format.
981
Currently no check is made that the format of this repository and
982
the bzrdir format are compatible. FIXME RBC 20060201.
984
:return: The newly created destination repository.
986
# TODO: deprecate after 0.16; cloning this with all its settings is
987
# probably not very useful -- mbp 20070423
988
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
989
self.copy_content_into(dest_repo, revision_id)
992
def start_write_group(self):
993
"""Start a write group in the repository.
995
Write groups are used by repositories which do not have a 1:1 mapping
996
between file ids and backend store to manage the insertion of data from
997
both fetch and commit operations.
999
A write lock is required around the start_write_group/commit_write_group
1000
for the support of lock-requiring repository formats.
1002
One can only insert data into a repository inside a write group.
1006
if not self.is_write_locked():
1007
raise errors.NotWriteLocked(self)
1008
if self._write_group:
1009
raise errors.BzrError('already in a write group')
1010
self._start_write_group()
1011
# so we can detect unlock/relock - the write group is now entered.
1012
self._write_group = self.get_transaction()
1014
def _start_write_group(self):
1015
"""Template method for per-repository write group startup.
1017
This is called before the write group is considered to be
1022
def sprout(self, to_bzrdir, revision_id=None):
1023
"""Create a descendent repository for new development.
1025
Unlike clone, this does not copy the settings of the repository.
1027
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1028
dest_repo.fetch(self, revision_id=revision_id)
1031
def _create_sprouting_repo(self, a_bzrdir, shared):
1032
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1033
# use target default format.
1034
dest_repo = a_bzrdir.create_repository()
1036
# Most control formats need the repository to be specifically
1037
# created, but on some old all-in-one formats it's not needed
1039
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1040
except errors.UninitializableFormat:
1041
dest_repo = a_bzrdir.open_repository()
1045
def has_revision(self, revision_id):
1046
"""True if this repository has a copy of the revision."""
1047
return revision_id in self.has_revisions((revision_id,))
1049
def has_revisions(self, revision_ids):
1050
"""Probe to find out the presence of multiple revisions.
1052
:param revision_ids: An iterable of revision_ids.
1053
:return: A set of the revision_ids that were present.
1055
raise NotImplementedError(self.has_revisions)
1057
return self._revision_store.has_revision_id(revision_id,
1058
self.get_transaction())
1061
def get_revision(self, revision_id):
1062
"""Return the Revision object for a named revision."""
1063
return self.get_revisions([revision_id])[0]
1066
def get_revision_reconcile(self, revision_id):
1067
"""'reconcile' helper routine that allows access to a revision always.
1069
This variant of get_revision does not cross check the weave graph
1070
against the revision one as get_revision does: but it should only
1071
be used by reconcile, or reconcile-alike commands that are correcting
1072
or testing the revision graph.
1074
return self._get_revisions([revision_id])[0]
1077
def get_revisions(self, revision_ids):
1078
"""Get many revisions at once."""
1079
return self._get_revisions(revision_ids)
1082
def _get_revisions(self, revision_ids):
1083
"""Core work logic to get many revisions without sanity checks."""
1084
for rev_id in revision_ids:
1085
if not rev_id or not isinstance(rev_id, basestring):
1086
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1087
revs = self._revision_store.get_revisions(revision_ids,
1088
self.get_transaction())
1090
assert not isinstance(rev.revision_id, unicode)
1091
for parent_id in rev.parent_ids:
1092
assert not isinstance(parent_id, unicode)
1096
def get_revision_xml(self, revision_id):
1097
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1098
# would have already do it.
1099
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1100
rev = self.get_revision(revision_id)
1101
rev_tmp = StringIO()
1102
# the current serializer..
1103
self._revision_store._serializer.write_revision(rev, rev_tmp)
1105
return rev_tmp.getvalue()
1108
def get_deltas_for_revisions(self, revisions):
1109
"""Produce a generator of revision deltas.
1111
Note that the input is a sequence of REVISIONS, not revision_ids.
1112
Trees will be held in memory until the generator exits.
1113
Each delta is relative to the revision's lefthand predecessor.
1115
required_trees = set()
1116
for revision in revisions:
1117
required_trees.add(revision.revision_id)
1118
required_trees.update(revision.parent_ids[:1])
1119
trees = dict((t.get_revision_id(), t) for
1120
t in self.revision_trees(required_trees))
1121
for revision in revisions:
1122
if not revision.parent_ids:
1123
old_tree = self.revision_tree(None)
1125
old_tree = trees[revision.parent_ids[0]]
1126
yield trees[revision.revision_id].changes_from(old_tree)
1129
def get_revision_delta(self, revision_id):
1130
"""Return the delta for one revision.
1132
The delta is relative to the left-hand predecessor of the
1135
r = self.get_revision(revision_id)
1136
return list(self.get_deltas_for_revisions([r]))[0]
1139
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1140
signature = gpg_strategy.sign(plaintext)
1141
self.add_signature_text(revision_id, signature)
1144
def add_signature_text(self, revision_id, signature):
1145
self._revision_store.add_revision_signature_text(revision_id,
1147
self.get_transaction())
1149
def find_text_key_references(self):
1150
"""Find the text key references within the repository.
1152
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1153
revision_ids. Each altered file-ids has the exact revision_ids that
1154
altered it listed explicitly.
1155
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1156
to whether they were referred to by the inventory of the
1157
revision_id that they contain. The inventory texts from all present
1158
revision ids are assessed to generate this report.
1160
revision_ids = self.all_revision_ids()
1161
w = self.get_inventory_weave()
1162
pb = ui.ui_factory.nested_progress_bar()
1164
return self._find_text_key_references_from_xml_inventory_lines(
1165
w.iter_lines_added_or_present_in_versions(revision_ids, pb=pb))
1169
def _find_text_key_references_from_xml_inventory_lines(self,
1171
"""Core routine for extracting references to texts from inventories.
1173
This performs the translation of xml lines to revision ids.
1175
:param line_iterator: An iterator of lines, origin_version_id
1176
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1177
to whether they were referred to by the inventory of the
1178
revision_id that they contain. Note that if that revision_id was
1179
not part of the line_iterator's output then False will be given -
1180
even though it may actually refer to that key.
1182
if not self._serializer.support_altered_by_hack:
1183
raise AssertionError(
1184
"_find_text_key_references_from_xml_inventory_lines only "
1185
"supported for branches which store inventory as unnested xml"
1186
", not on %r" % self)
1189
# this code needs to read every new line in every inventory for the
1190
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1191
# not present in one of those inventories is unnecessary but not
1192
# harmful because we are filtering by the revision id marker in the
1193
# inventory lines : we only select file ids altered in one of those
1194
# revisions. We don't need to see all lines in the inventory because
1195
# only those added in an inventory in rev X can contain a revision=X
1197
unescape_revid_cache = {}
1198
unescape_fileid_cache = {}
1200
# jam 20061218 In a big fetch, this handles hundreds of thousands
1201
# of lines, so it has had a lot of inlining and optimizing done.
1202
# Sorry that it is a little bit messy.
1203
# Move several functions to be local variables, since this is a long
1205
search = self._file_ids_altered_regex.search
1206
unescape = _unescape_xml
1207
setdefault = result.setdefault
1208
for line, version_id in line_iterator:
1209
match = search(line)
1212
# One call to match.group() returning multiple items is quite a
1213
# bit faster than 2 calls to match.group() each returning 1
1214
file_id, revision_id = match.group('file_id', 'revision_id')
1216
# Inlining the cache lookups helps a lot when you make 170,000
1217
# lines and 350k ids, versus 8.4 unique ids.
1218
# Using a cache helps in 2 ways:
1219
# 1) Avoids unnecessary decoding calls
1220
# 2) Re-uses cached strings, which helps in future set and
1222
# (2) is enough that removing encoding entirely along with
1223
# the cache (so we are using plain strings) results in no
1224
# performance improvement.
1226
revision_id = unescape_revid_cache[revision_id]
1228
unescaped = unescape(revision_id)
1229
unescape_revid_cache[revision_id] = unescaped
1230
revision_id = unescaped
1232
# Note that unconditionally unescaping means that we deserialise
1233
# every fileid, which for general 'pull' is not great, but we don't
1234
# really want to have some many fulltexts that this matters anyway.
1237
file_id = unescape_fileid_cache[file_id]
1239
unescaped = unescape(file_id)
1240
unescape_fileid_cache[file_id] = unescaped
1243
key = (file_id, revision_id)
1244
setdefault(key, False)
1245
if revision_id == version_id:
1249
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1251
"""Helper routine for fileids_altered_by_revision_ids.
1253
This performs the translation of xml lines to revision ids.
1255
:param line_iterator: An iterator of lines, origin_version_id
1256
:param revision_ids: The revision ids to filter for. This should be a
1257
set or other type which supports efficient __contains__ lookups, as
1258
the revision id from each parsed line will be looked up in the
1259
revision_ids filter.
1260
:return: a dictionary mapping altered file-ids to an iterable of
1261
revision_ids. Each altered file-ids has the exact revision_ids that
1262
altered it listed explicitly.
1265
setdefault = result.setdefault
1266
for file_id, revision_id in \
1267
self._find_text_key_references_from_xml_inventory_lines(
1268
line_iterator).iterkeys():
1269
# once data is all ensured-consistent; then this is
1270
# if revision_id == version_id
1271
if revision_id in revision_ids:
1272
setdefault(file_id, set()).add(revision_id)
1275
def fileids_altered_by_revision_ids(self, revision_ids):
1276
"""Find the file ids and versions affected by revisions.
1278
:param revisions: an iterable containing revision ids.
1279
:return: a dictionary mapping altered file-ids to an iterable of
1280
revision_ids. Each altered file-ids has the exact revision_ids that
1281
altered it listed explicitly.
1283
selected_revision_ids = set(revision_ids)
1284
w = self.get_inventory_weave()
1285
pb = ui.ui_factory.nested_progress_bar()
1287
return self._find_file_ids_from_xml_inventory_lines(
1288
w.iter_lines_added_or_present_in_versions(
1289
selected_revision_ids, pb=pb),
1290
selected_revision_ids)
1294
def iter_files_bytes(self, desired_files):
1295
"""Iterate through file versions.
1297
Files will not necessarily be returned in the order they occur in
1298
desired_files. No specific order is guaranteed.
1300
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1301
value supplied by the caller as part of desired_files. It should
1302
uniquely identify the file version in the caller's context. (Examples:
1303
an index number or a TreeTransform trans_id.)
1305
bytes_iterator is an iterable of bytestrings for the file. The
1306
kind of iterable and length of the bytestrings are unspecified, but for
1307
this implementation, it is a list of lines produced by
1308
VersionedFile.get_lines().
1310
:param desired_files: a list of (file_id, revision_id, identifier)
1313
transaction = self.get_transaction()
1314
for file_id, revision_id, callable_data in desired_files:
1316
weave = self.weave_store.get_weave(file_id, transaction)
1317
except errors.NoSuchFile:
1318
raise errors.NoSuchIdInRepository(self, file_id)
1319
yield callable_data, weave.get_lines(revision_id)
1321
def _generate_text_key_index(self, text_key_references=None,
1323
"""Generate a new text key index for the repository.
1325
This is an expensive function that will take considerable time to run.
1327
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1328
list of parents, also text keys. When a given key has no parents,
1329
the parents list will be [NULL_REVISION].
1331
# All revisions, to find inventory parents.
1332
if ancestors is None:
1333
revision_graph = self.get_revision_graph_with_ghosts()
1334
ancestors = revision_graph.get_ancestors()
1335
if text_key_references is None:
1336
text_key_references = self.find_text_key_references()
1337
pb = ui.ui_factory.nested_progress_bar()
1339
return self._do_generate_text_key_index(ancestors,
1340
text_key_references, pb)
1344
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1345
"""Helper for _generate_text_key_index to avoid deep nesting."""
1346
revision_order = tsort.topo_sort(ancestors)
1347
invalid_keys = set()
1349
for revision_id in revision_order:
1350
revision_keys[revision_id] = set()
1351
text_count = len(text_key_references)
1352
# a cache of the text keys to allow reuse; costs a dict of all the
1353
# keys, but saves a 2-tuple for every child of a given key.
1355
for text_key, valid in text_key_references.iteritems():
1357
invalid_keys.add(text_key)
1359
revision_keys[text_key[1]].add(text_key)
1360
text_key_cache[text_key] = text_key
1361
del text_key_references
1363
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1364
NULL_REVISION = _mod_revision.NULL_REVISION
1365
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1366
# too small for large or very branchy trees. However, for 55K path
1367
# trees, it would be easy to use too much memory trivially. Ideally we
1368
# could gauge this by looking at available real memory etc, but this is
1369
# always a tricky proposition.
1370
inventory_cache = lru_cache.LRUCache(10)
1371
batch_size = 10 # should be ~150MB on a 55K path tree
1372
batch_count = len(revision_order) / batch_size + 1
1374
pb.update("Calculating text parents.", processed_texts, text_count)
1375
for offset in xrange(batch_count):
1376
to_query = revision_order[offset * batch_size:(offset + 1) *
1380
for rev_tree in self.revision_trees(to_query):
1381
revision_id = rev_tree.get_revision_id()
1382
parent_ids = ancestors[revision_id]
1383
for text_key in revision_keys[revision_id]:
1384
pb.update("Calculating text parents.", processed_texts)
1385
processed_texts += 1
1386
candidate_parents = []
1387
for parent_id in parent_ids:
1388
parent_text_key = (text_key[0], parent_id)
1390
check_parent = parent_text_key not in \
1391
revision_keys[parent_id]
1393
# the parent parent_id is a ghost:
1394
check_parent = False
1395
# truncate the derived graph against this ghost.
1396
parent_text_key = None
1398
# look at the parent commit details inventories to
1399
# determine possible candidates in the per file graph.
1402
inv = inventory_cache[parent_id]
1404
inv = self.revision_tree(parent_id).inventory
1405
inventory_cache[parent_id] = inv
1406
parent_entry = inv._byid.get(text_key[0], None)
1407
if parent_entry is not None:
1409
text_key[0], parent_entry.revision)
1411
parent_text_key = None
1412
if parent_text_key is not None:
1413
candidate_parents.append(
1414
text_key_cache[parent_text_key])
1415
parent_heads = text_graph.heads(candidate_parents)
1416
new_parents = list(parent_heads)
1417
new_parents.sort(key=lambda x:candidate_parents.index(x))
1418
if new_parents == []:
1419
new_parents = [NULL_REVISION]
1420
text_index[text_key] = new_parents
1422
for text_key in invalid_keys:
1423
text_index[text_key] = [NULL_REVISION]
1426
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1427
"""Get an iterable listing the keys of all the data introduced by a set
1430
The keys will be ordered so that the corresponding items can be safely
1431
fetched and inserted in that order.
1433
:returns: An iterable producing tuples of (knit-kind, file-id,
1434
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1435
'revisions'. file-id is None unless knit-kind is 'file'.
1437
# XXX: it's a bit weird to control the inventory weave caching in this
1438
# generator. Ideally the caching would be done in fetch.py I think. Or
1439
# maybe this generator should explicitly have the contract that it
1440
# should not be iterated until the previously yielded item has been
1443
inv_w = self.get_inventory_weave()
1444
inv_w.enable_cache()
1446
# file ids that changed
1447
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1449
num_file_ids = len(file_ids)
1450
for file_id, altered_versions in file_ids.iteritems():
1451
if _files_pb is not None:
1452
_files_pb.update("fetch texts", count, num_file_ids)
1454
yield ("file", file_id, altered_versions)
1455
# We're done with the files_pb. Note that it finished by the caller,
1456
# just as it was created by the caller.
1460
yield ("inventory", None, revision_ids)
1464
revisions_with_signatures = set()
1465
for rev_id in revision_ids:
1467
self.get_signature_text(rev_id)
1468
except errors.NoSuchRevision:
1472
revisions_with_signatures.add(rev_id)
1474
yield ("signatures", None, revisions_with_signatures)
1477
yield ("revisions", None, revision_ids)
1480
def get_inventory_weave(self):
1481
return self.control_weaves.get_weave('inventory',
1482
self.get_transaction())
1485
def get_inventory(self, revision_id):
1486
"""Get Inventory object by revision id."""
1487
return self.iter_inventories([revision_id]).next()
1489
def iter_inventories(self, revision_ids):
1490
"""Get many inventories by revision_ids.
1492
This will buffer some or all of the texts used in constructing the
1493
inventories in memory, but will only parse a single inventory at a
1496
:return: An iterator of inventories.
1498
assert None not in revision_ids
1499
assert _mod_revision.NULL_REVISION not in revision_ids
1500
return self._iter_inventories(revision_ids)
1502
def _iter_inventories(self, revision_ids):
1503
"""single-document based inventory iteration."""
1504
texts = self.get_inventory_weave().get_texts(revision_ids)
1505
for text, revision_id in zip(texts, revision_ids):
1506
yield self.deserialise_inventory(revision_id, text)
1508
def deserialise_inventory(self, revision_id, xml):
1509
"""Transform the xml into an inventory object.
1511
:param revision_id: The expected revision id of the inventory.
1512
:param xml: A serialised inventory.
1514
result = self._serializer.read_inventory_from_string(xml, revision_id)
1515
if result.revision_id != revision_id:
1516
raise AssertionError('revision id mismatch %s != %s' % (
1517
result.revision_id, revision_id))
1520
def serialise_inventory(self, inv):
1521
return self._serializer.write_inventory_to_string(inv)
1523
def _serialise_inventory_to_lines(self, inv):
1524
return self._serializer.write_inventory_to_lines(inv)
1526
def get_serializer_format(self):
1527
return self._serializer.format_num
1530
def get_inventory_xml(self, revision_id):
1531
"""Get inventory XML as a file object."""
1533
assert isinstance(revision_id, str), type(revision_id)
1534
iw = self.get_inventory_weave()
1535
return iw.get_text(revision_id)
1537
raise errors.HistoryMissing(self, 'inventory', revision_id)
1540
def get_inventory_sha1(self, revision_id):
1541
"""Return the sha1 hash of the inventory entry
1543
return self.get_revision(revision_id).inventory_sha1
1546
def get_revision_graph(self, revision_id=None):
1547
"""Return a dictionary containing the revision graph.
1549
NB: This method should not be used as it accesses the entire graph all
1550
at once, which is much more data than most operations should require.
1552
:param revision_id: The revision_id to get a graph from. If None, then
1553
the entire revision graph is returned. This is a deprecated mode of
1554
operation and will be removed in the future.
1555
:return: a dictionary of revision_id->revision_parents_list.
1557
raise NotImplementedError(self.get_revision_graph)
1560
def get_revision_graph_with_ghosts(self, revision_ids=None):
1561
"""Return a graph of the revisions with ghosts marked as applicable.
1563
:param revision_ids: an iterable of revisions to graph or None for all.
1564
:return: a Graph object with the graph reachable from revision_ids.
1566
if 'evil' in debug.debug_flags:
1568
"get_revision_graph_with_ghosts scales with size of history.")
1569
result = deprecated_graph.Graph()
1570
if not revision_ids:
1571
pending = set(self.all_revision_ids())
1574
pending = set(revision_ids)
1575
# special case NULL_REVISION
1576
if _mod_revision.NULL_REVISION in pending:
1577
pending.remove(_mod_revision.NULL_REVISION)
1578
required = set(pending)
1581
revision_id = pending.pop()
1583
rev = self.get_revision(revision_id)
1584
except errors.NoSuchRevision:
1585
if revision_id in required:
1588
result.add_ghost(revision_id)
1590
for parent_id in rev.parent_ids:
1591
# is this queued or done ?
1592
if (parent_id not in pending and
1593
parent_id not in done):
1595
pending.add(parent_id)
1596
result.add_node(revision_id, rev.parent_ids)
1597
done.add(revision_id)
1600
def _get_history_vf(self):
1601
"""Get a versionedfile whose history graph reflects all revisions.
1603
For weave repositories, this is the inventory weave.
1605
return self.get_inventory_weave()
1607
def iter_reverse_revision_history(self, revision_id):
1608
"""Iterate backwards through revision ids in the lefthand history
1610
:param revision_id: The revision id to start with. All its lefthand
1611
ancestors will be traversed.
1613
if revision_id in (None, _mod_revision.NULL_REVISION):
1615
next_id = revision_id
1616
versionedfile = self._get_history_vf()
1619
parents = versionedfile.get_parents(next_id)
1620
if len(parents) == 0:
1623
next_id = parents[0]
1626
def get_revision_inventory(self, revision_id):
1627
"""Return inventory of a past revision."""
1628
# TODO: Unify this with get_inventory()
1629
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1630
# must be the same as its revision, so this is trivial.
1631
if revision_id is None:
1632
# This does not make sense: if there is no revision,
1633
# then it is the current tree inventory surely ?!
1634
# and thus get_root_id() is something that looks at the last
1635
# commit on the branch, and the get_root_id is an inventory check.
1636
raise NotImplementedError
1637
# return Inventory(self.get_root_id())
1639
return self.get_inventory(revision_id)
1642
def is_shared(self):
1643
"""Return True if this repository is flagged as a shared repository."""
1644
raise NotImplementedError(self.is_shared)
1647
def reconcile(self, other=None, thorough=False):
1648
"""Reconcile this repository."""
1649
from bzrlib.reconcile import RepoReconciler
1650
reconciler = RepoReconciler(self, thorough=thorough)
1651
reconciler.reconcile()
1654
def _refresh_data(self):
1655
"""Helper called from lock_* to ensure coherency with disk.
1657
The default implementation does nothing; it is however possible
1658
for repositories to maintain loaded indices across multiple locks
1659
by checking inside their implementation of this method to see
1660
whether their indices are still valid. This depends of course on
1661
the disk format being validatable in this manner.
1665
def revision_tree(self, revision_id):
1666
"""Return Tree for a revision on this branch.
1668
`revision_id` may be None for the empty tree revision.
1670
# TODO: refactor this to use an existing revision object
1671
# so we don't need to read it in twice.
1672
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1673
return RevisionTree(self, Inventory(root_id=None),
1674
_mod_revision.NULL_REVISION)
1676
inv = self.get_revision_inventory(revision_id)
1677
return RevisionTree(self, inv, revision_id)
1680
def revision_trees(self, revision_ids):
1681
"""Return Tree for a revision on this branch.
1683
`revision_id` may not be None or 'null:'"""
1684
inventories = self.iter_inventories(revision_ids)
1685
for inv in inventories:
1686
yield RevisionTree(self, inv, inv.revision_id)
1689
def get_ancestry(self, revision_id, topo_sorted=True):
1690
"""Return a list of revision-ids integrated by a revision.
1692
The first element of the list is always None, indicating the origin
1693
revision. This might change when we have history horizons, or
1694
perhaps we should have a new API.
1696
This is topologically sorted.
1698
if _mod_revision.is_null(revision_id):
1700
if not self.has_revision(revision_id):
1701
raise errors.NoSuchRevision(self, revision_id)
1702
w = self.get_inventory_weave()
1703
candidates = w.get_ancestry(revision_id, topo_sorted)
1704
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1707
"""Compress the data within the repository.
1709
This operation only makes sense for some repository types. For other
1710
types it should be a no-op that just returns.
1712
This stub method does not require a lock, but subclasses should use
1713
@needs_write_lock as this is a long running call its reasonable to
1714
implicitly lock for the user.
1718
def print_file(self, file, revision_id):
1719
"""Print `file` to stdout.
1721
FIXME RBC 20060125 as John Meinel points out this is a bad api
1722
- it writes to stdout, it assumes that that is valid etc. Fix
1723
by creating a new more flexible convenience function.
1725
tree = self.revision_tree(revision_id)
1726
# use inventory as it was in that revision
1727
file_id = tree.inventory.path2id(file)
1729
# TODO: jam 20060427 Write a test for this code path
1730
# it had a bug in it, and was raising the wrong
1732
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1733
tree.print_file(file_id)
1735
def get_transaction(self):
1736
return self.control_files.get_transaction()
1738
def revision_parents(self, revision_id):
1739
return self.get_inventory_weave().parent_names(revision_id)
1741
@deprecated_method(symbol_versioning.one_one)
1742
def get_parents(self, revision_ids):
1743
"""See StackedParentsProvider.get_parents"""
1744
parent_map = self.get_parent_map(revision_ids)
1745
return [parent_map.get(r, None) for r in revision_ids]
1747
def get_parent_map(self, keys):
1748
"""See graph._StackedParentsProvider.get_parent_map"""
1750
for revision_id in keys:
1751
if revision_id == _mod_revision.NULL_REVISION:
1752
parent_map[revision_id] = ()
1755
parent_id_list = self.get_revision(revision_id).parent_ids
1756
except errors.NoSuchRevision:
1759
if len(parent_id_list) == 0:
1760
parent_ids = (_mod_revision.NULL_REVISION,)
1762
parent_ids = tuple(parent_id_list)
1763
parent_map[revision_id] = parent_ids
1766
def _make_parents_provider(self):
1769
def get_graph(self, other_repository=None):
1770
"""Return the graph walker for this repository format"""
1771
parents_provider = self._make_parents_provider()
1772
if (other_repository is not None and
1773
other_repository.bzrdir.transport.base !=
1774
self.bzrdir.transport.base):
1775
parents_provider = graph._StackedParentsProvider(
1776
[parents_provider, other_repository._make_parents_provider()])
1777
return graph.Graph(parents_provider)
1779
def _get_versioned_file_checker(self):
1780
"""Return an object suitable for checking versioned files."""
1781
return _VersionedFileChecker(self)
1784
def set_make_working_trees(self, new_value):
1785
"""Set the policy flag for making working trees when creating branches.
1787
This only applies to branches that use this repository.
1789
The default is 'True'.
1790
:param new_value: True to restore the default, False to disable making
1793
raise NotImplementedError(self.set_make_working_trees)
1795
def make_working_trees(self):
1796
"""Returns the policy for making working trees on new branches."""
1797
raise NotImplementedError(self.make_working_trees)
1800
def sign_revision(self, revision_id, gpg_strategy):
1801
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1802
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1805
def has_signature_for_revision_id(self, revision_id):
1806
"""Query for a revision signature for revision_id in the repository."""
1807
return self._revision_store.has_signature(revision_id,
1808
self.get_transaction())
1811
def get_signature_text(self, revision_id):
1812
"""Return the text for a signature."""
1813
return self._revision_store.get_signature_text(revision_id,
1814
self.get_transaction())
1817
def check(self, revision_ids=None):
1818
"""Check consistency of all history of given revision_ids.
1820
Different repository implementations should override _check().
1822
:param revision_ids: A non-empty list of revision_ids whose ancestry
1823
will be checked. Typically the last revision_id of a branch.
1825
return self._check(revision_ids)
1827
def _check(self, revision_ids):
1828
result = check.Check(self)
1832
def _warn_if_deprecated(self):
1833
global _deprecation_warning_done
1834
if _deprecation_warning_done:
1836
_deprecation_warning_done = True
1837
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1838
% (self._format, self.bzrdir.transport.base))
1840
def supports_rich_root(self):
1841
return self._format.rich_root_data
1843
def _check_ascii_revisionid(self, revision_id, method):
1844
"""Private helper for ascii-only repositories."""
1845
# weave repositories refuse to store revisionids that are non-ascii.
1846
if revision_id is not None:
1847
# weaves require ascii revision ids.
1848
if isinstance(revision_id, unicode):
1850
revision_id.encode('ascii')
1851
except UnicodeEncodeError:
1852
raise errors.NonAsciiRevisionId(method, self)
1855
revision_id.decode('ascii')
1856
except UnicodeDecodeError:
1857
raise errors.NonAsciiRevisionId(method, self)
1859
def revision_graph_can_have_wrong_parents(self):
1860
"""Is it possible for this repository to have a revision graph with
1863
If True, then this repository must also implement
1864
_find_inconsistent_revision_parents so that check and reconcile can
1865
check for inconsistencies before proceeding with other checks that may
1866
depend on the revision index being consistent.
1868
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1870
# remove these delegates a while after bzr 0.15
1871
def __make_delegated(name, from_module):
1872
def _deprecated_repository_forwarder():
1873
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1874
% (name, from_module),
1877
m = __import__(from_module, globals(), locals(), [name])
1879
return getattr(m, name)
1880
except AttributeError:
1881
raise AttributeError('module %s has no name %s'
1883
globals()[name] = _deprecated_repository_forwarder
1886
'AllInOneRepository',
1887
'WeaveMetaDirRepository',
1888
'PreSplitOutRepositoryFormat',
1889
'RepositoryFormat4',
1890
'RepositoryFormat5',
1891
'RepositoryFormat6',
1892
'RepositoryFormat7',
1894
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1898
'RepositoryFormatKnit',
1899
'RepositoryFormatKnit1',
1901
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1904
def install_revision(repository, rev, revision_tree):
1905
"""Install all revision data into a repository."""
1906
install_revisions(repository, [(rev, revision_tree, None)])
1909
def install_revisions(repository, iterable, num_revisions=None, pb=None):
1910
"""Install all revision data into a repository.
1912
Accepts an iterable of revision, tree, signature tuples. The signature
1915
repository.start_write_group()
1917
for n, (revision, revision_tree, signature) in enumerate(iterable):
1918
_install_revision(repository, revision, revision_tree, signature)
1920
pb.update('Transferring revisions', n + 1, num_revisions)
1922
repository.abort_write_group()
1925
repository.commit_write_group()
1928
def _install_revision(repository, rev, revision_tree, signature):
1929
"""Install all revision data into a repository."""
1930
present_parents = []
1932
for p_id in rev.parent_ids:
1933
if repository.has_revision(p_id):
1934
present_parents.append(p_id)
1935
parent_trees[p_id] = repository.revision_tree(p_id)
1937
parent_trees[p_id] = repository.revision_tree(None)
1939
inv = revision_tree.inventory
1940
entries = inv.iter_entries()
1941
# backwards compatibility hack: skip the root id.
1942
if not repository.supports_rich_root():
1943
path, root = entries.next()
1944
if root.revision != rev.revision_id:
1945
raise errors.IncompatibleRevision(repr(repository))
1946
# Add the texts that are not already present
1947
for path, ie in entries:
1948
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1949
repository.get_transaction())
1950
if ie.revision not in w:
1952
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1953
# with InventoryEntry.find_previous_heads(). if it is, then there
1954
# is a latent bug here where the parents may have ancestors of each
1956
for revision, tree in parent_trees.iteritems():
1957
if ie.file_id not in tree:
1959
parent_id = tree.inventory[ie.file_id].revision
1960
if parent_id in text_parents:
1962
text_parents.append(parent_id)
1964
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1965
repository.get_transaction())
1966
lines = revision_tree.get_file(ie.file_id).readlines()
1967
vfile.add_lines(rev.revision_id, text_parents, lines)
1969
# install the inventory
1970
repository.add_inventory(rev.revision_id, inv, present_parents)
1971
except errors.RevisionAlreadyPresent:
1973
if signature is not None:
1974
repository.add_signature_text(rev.revision_id, signature)
1975
repository.add_revision(rev.revision_id, rev, inv)
1978
class MetaDirRepository(Repository):
1979
"""Repositories in the new meta-dir layout."""
1981
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1982
super(MetaDirRepository, self).__init__(_format,
1988
dir_mode = self.control_files._dir_mode
1989
file_mode = self.control_files._file_mode
1992
def is_shared(self):
1993
"""Return True if this repository is flagged as a shared repository."""
1994
return self.control_files._transport.has('shared-storage')
1997
def set_make_working_trees(self, new_value):
1998
"""Set the policy flag for making working trees when creating branches.
2000
This only applies to branches that use this repository.
2002
The default is 'True'.
2003
:param new_value: True to restore the default, False to disable making
2008
self.control_files._transport.delete('no-working-trees')
2009
except errors.NoSuchFile:
2012
self.control_files.put_utf8('no-working-trees', '')
2014
def make_working_trees(self):
2015
"""Returns the policy for making working trees on new branches."""
2016
return not self.control_files._transport.has('no-working-trees')
2019
class RepositoryFormatRegistry(registry.Registry):
2020
"""Registry of RepositoryFormats."""
2022
def get(self, format_string):
2023
r = registry.Registry.get(self, format_string)
2029
format_registry = RepositoryFormatRegistry()
2030
"""Registry of formats, indexed by their identifying format string.
2032
This can contain either format instances themselves, or classes/factories that
2033
can be called to obtain one.
2037
#####################################################################
2038
# Repository Formats
2040
class RepositoryFormat(object):
2041
"""A repository format.
2043
Formats provide three things:
2044
* An initialization routine to construct repository data on disk.
2045
* a format string which is used when the BzrDir supports versioned
2047
* an open routine which returns a Repository instance.
2049
There is one and only one Format subclass for each on-disk format. But
2050
there can be one Repository subclass that is used for several different
2051
formats. The _format attribute on a Repository instance can be used to
2052
determine the disk format.
2054
Formats are placed in an dict by their format string for reference
2055
during opening. These should be subclasses of RepositoryFormat
2058
Once a format is deprecated, just deprecate the initialize and open
2059
methods on the format class. Do not deprecate the object, as the
2060
object will be created every system load.
2062
Common instance attributes:
2063
_matchingbzrdir - the bzrdir format that the repository format was
2064
originally written to work with. This can be used if manually
2065
constructing a bzrdir and repository, or more commonly for test suite
2069
# Set to True or False in derived classes. True indicates that the format
2070
# supports ghosts gracefully.
2071
supports_ghosts = None
2074
return "<%s>" % self.__class__.__name__
2076
def __eq__(self, other):
2077
# format objects are generally stateless
2078
return isinstance(other, self.__class__)
2080
def __ne__(self, other):
2081
return not self == other
2084
def find_format(klass, a_bzrdir):
2085
"""Return the format for the repository object in a_bzrdir.
2087
This is used by bzr native formats that have a "format" file in
2088
the repository. Other methods may be used by different types of
2092
transport = a_bzrdir.get_repository_transport(None)
2093
format_string = transport.get("format").read()
2094
return format_registry.get(format_string)
2095
except errors.NoSuchFile:
2096
raise errors.NoRepositoryPresent(a_bzrdir)
2098
raise errors.UnknownFormatError(format=format_string)
2101
def register_format(klass, format):
2102
format_registry.register(format.get_format_string(), format)
2105
def unregister_format(klass, format):
2106
format_registry.remove(format.get_format_string())
2109
def get_default_format(klass):
2110
"""Return the current default format."""
2111
from bzrlib import bzrdir
2112
return bzrdir.format_registry.make_bzrdir('default').repository_format
2114
def _get_control_store(self, repo_transport, control_files):
2115
"""Return the control store for this repository."""
2116
raise NotImplementedError(self._get_control_store)
2118
def get_format_string(self):
2119
"""Return the ASCII format string that identifies this format.
2121
Note that in pre format ?? repositories the format string is
2122
not permitted nor written to disk.
2124
raise NotImplementedError(self.get_format_string)
2126
def get_format_description(self):
2127
"""Return the short description for this format."""
2128
raise NotImplementedError(self.get_format_description)
2130
def _get_revision_store(self, repo_transport, control_files):
2131
"""Return the revision store object for this a_bzrdir."""
2132
raise NotImplementedError(self._get_revision_store)
2134
def _get_text_rev_store(self,
2141
"""Common logic for getting a revision store for a repository.
2143
see self._get_revision_store for the subclass-overridable method to
2144
get the store for a repository.
2146
from bzrlib.store.revision.text import TextRevisionStore
2147
dir_mode = control_files._dir_mode
2148
file_mode = control_files._file_mode
2149
text_store = TextStore(transport.clone(name),
2151
compressed=compressed,
2153
file_mode=file_mode)
2154
_revision_store = TextRevisionStore(text_store, serializer)
2155
return _revision_store
2157
# TODO: this shouldn't be in the base class, it's specific to things that
2158
# use weaves or knits -- mbp 20070207
2159
def _get_versioned_file_store(self,
2164
versionedfile_class=None,
2165
versionedfile_kwargs={},
2167
if versionedfile_class is None:
2168
versionedfile_class = self._versionedfile_class
2169
weave_transport = control_files._transport.clone(name)
2170
dir_mode = control_files._dir_mode
2171
file_mode = control_files._file_mode
2172
return VersionedFileStore(weave_transport, prefixed=prefixed,
2174
file_mode=file_mode,
2175
versionedfile_class=versionedfile_class,
2176
versionedfile_kwargs=versionedfile_kwargs,
2179
def initialize(self, a_bzrdir, shared=False):
2180
"""Initialize a repository of this format in a_bzrdir.
2182
:param a_bzrdir: The bzrdir to put the new repository in it.
2183
:param shared: The repository should be initialized as a sharable one.
2184
:returns: The new repository object.
2186
This may raise UninitializableFormat if shared repository are not
2187
compatible the a_bzrdir.
2189
raise NotImplementedError(self.initialize)
2191
def is_supported(self):
2192
"""Is this format supported?
2194
Supported formats must be initializable and openable.
2195
Unsupported formats may not support initialization or committing or
2196
some other features depending on the reason for not being supported.
2200
def check_conversion_target(self, target_format):
2201
raise NotImplementedError(self.check_conversion_target)
2203
def open(self, a_bzrdir, _found=False):
2204
"""Return an instance of this format for the bzrdir a_bzrdir.
2206
_found is a private parameter, do not use it.
2208
raise NotImplementedError(self.open)
2211
class MetaDirRepositoryFormat(RepositoryFormat):
2212
"""Common base class for the new repositories using the metadir layout."""
2214
rich_root_data = False
2215
supports_tree_reference = False
2216
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
2219
super(MetaDirRepositoryFormat, self).__init__()
2221
def _create_control_files(self, a_bzrdir):
2222
"""Create the required files and the initial control_files object."""
2223
# FIXME: RBC 20060125 don't peek under the covers
2224
# NB: no need to escape relative paths that are url safe.
2225
repository_transport = a_bzrdir.get_repository_transport(self)
2226
control_files = lockable_files.LockableFiles(repository_transport,
2227
'lock', lockdir.LockDir)
2228
control_files.create_lock()
2229
return control_files
2231
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2232
"""Upload the initial blank content."""
2233
control_files = self._create_control_files(a_bzrdir)
2234
control_files.lock_write()
2236
control_files._transport.mkdir_multi(dirs,
2237
mode=control_files._dir_mode)
2238
for file, content in files:
2239
control_files.put(file, content)
2240
for file, content in utf8_files:
2241
control_files.put_utf8(file, content)
2243
control_files.put_utf8('shared-storage', '')
2245
control_files.unlock()
2248
# formats which have no format string are not discoverable
2249
# and not independently creatable, so are not registered. They're
2250
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2251
# needed, it's constructed directly by the BzrDir. Non-native formats where
2252
# the repository is not separately opened are similar.
2254
format_registry.register_lazy(
2255
'Bazaar-NG Repository format 7',
2256
'bzrlib.repofmt.weaverepo',
2260
format_registry.register_lazy(
2261
'Bazaar-NG Knit Repository Format 1',
2262
'bzrlib.repofmt.knitrepo',
2263
'RepositoryFormatKnit1',
2266
format_registry.register_lazy(
2267
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2268
'bzrlib.repofmt.knitrepo',
2269
'RepositoryFormatKnit3',
2272
format_registry.register_lazy(
2273
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2274
'bzrlib.repofmt.knitrepo',
2275
'RepositoryFormatKnit4',
2278
# Pack-based formats. There is one format for pre-subtrees, and one for
2279
# post-subtrees to allow ease of testing.
2280
# NOTE: These are experimental in 0.92.
2281
format_registry.register_lazy(
2282
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2283
'bzrlib.repofmt.pack_repo',
2284
'RepositoryFormatKnitPack1',
2286
format_registry.register_lazy(
2287
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2288
'bzrlib.repofmt.pack_repo',
2289
'RepositoryFormatKnitPack3',
2291
format_registry.register_lazy(
2292
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2293
'bzrlib.repofmt.pack_repo',
2294
'RepositoryFormatKnitPack4',
2298
class InterRepository(InterObject):
2299
"""This class represents operations taking place between two repositories.
2301
Its instances have methods like copy_content and fetch, and contain
2302
references to the source and target repositories these operations can be
2305
Often we will provide convenience methods on 'repository' which carry out
2306
operations with another repository - they will always forward to
2307
InterRepository.get(other).method_name(parameters).
2311
"""The available optimised InterRepository types."""
2313
def copy_content(self, revision_id=None):
2314
raise NotImplementedError(self.copy_content)
2316
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2317
"""Fetch the content required to construct revision_id.
2319
The content is copied from self.source to self.target.
2321
:param revision_id: if None all content is copied, if NULL_REVISION no
2323
:param pb: optional progress bar to use for progress reports. If not
2324
provided a default one will be created.
2326
Returns the copied revision count and the failed revisions in a tuple:
2329
raise NotImplementedError(self.fetch)
2331
def _walk_to_common_revisions(self, revision_ids):
2332
"""Walk out from revision_ids in source to revisions target has.
2334
:param revision_ids: The start point for the search.
2335
:return: A set of revision ids.
2337
graph = self.source.get_graph()
2338
missing_revs = set()
2339
# ensure we don't pay silly lookup costs.
2340
revision_ids = frozenset(revision_ids)
2341
searcher = graph._make_breadth_first_searcher(revision_ids)
2342
null_set = frozenset([_mod_revision.NULL_REVISION])
2345
next_revs, ghosts = searcher.next_with_ghosts()
2346
except StopIteration:
2348
if revision_ids.intersection(ghosts):
2349
absent_ids = set(revision_ids.intersection(ghosts))
2350
# If all absent_ids are present in target, no error is needed.
2351
absent_ids.difference_update(
2352
self.target.has_revisions(absent_ids))
2354
raise errors.NoSuchRevision(self.source, absent_ids.pop())
2355
# we don't care about other ghosts as we can't fetch them and
2356
# haven't been asked to.
2357
next_revs = set(next_revs)
2358
# we always have NULL_REVISION present.
2359
have_revs = self.target.has_revisions(next_revs).union(null_set)
2360
missing_revs.update(next_revs - have_revs)
2361
searcher.stop_searching_any(have_revs)
2362
return searcher.get_result()
2364
@deprecated_method(symbol_versioning.one_two)
2366
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2367
"""Return the revision ids that source has that target does not.
2369
These are returned in topological order.
2371
:param revision_id: only return revision ids included by this
2373
:param find_ghosts: If True find missing revisions in deep history
2374
rather than just finding the surface difference.
2376
return list(self.search_missing_revision_ids(
2377
revision_id, find_ghosts).get_keys())
2380
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2381
"""Return the revision ids that source has that target does not.
2383
:param revision_id: only return revision ids included by this
2385
:param find_ghosts: If True find missing revisions in deep history
2386
rather than just finding the surface difference.
2387
:return: A bzrlib.graph.SearchResult.
2389
# stop searching at found target revisions.
2390
if not find_ghosts and revision_id is not None:
2391
return self._walk_to_common_revisions([revision_id])
2392
# generic, possibly worst case, slow code path.
2393
target_ids = set(self.target.all_revision_ids())
2394
if revision_id is not None:
2395
source_ids = self.source.get_ancestry(revision_id)
2396
assert source_ids[0] is None
2399
source_ids = self.source.all_revision_ids()
2400
result_set = set(source_ids).difference(target_ids)
2401
return self._set_to_search_result(result_set, self.source)
2403
def _set_to_search_result(self, result_set, repository):
2404
"""Convert a set of revision ids to a graph SearchResult."""
2405
result_parents = set()
2406
for parents in repository.get_graph().get_parent_map(
2407
result_set).itervalues():
2408
result_parents.update(parents)
2409
included_keys = result_set.intersection(result_parents)
2410
start_keys = result_set.difference(included_keys)
2411
exclude_keys = result_parents.difference(result_set)
2412
result = graph.SearchResult(start_keys, exclude_keys,
2413
len(result_set), result_set)
2417
def _same_model(source, target):
2418
"""True if source and target have the same data representation."""
2419
if source.supports_rich_root() != target.supports_rich_root():
2421
if source._serializer != target._serializer:
2426
class InterSameDataRepository(InterRepository):
2427
"""Code for converting between repositories that represent the same data.
2429
Data format and model must match for this to work.
2433
def _get_repo_format_to_test(self):
2434
"""Repository format for testing with.
2436
InterSameData can pull from subtree to subtree and from non-subtree to
2437
non-subtree, so we test this with the richest repository format.
2439
from bzrlib.repofmt import knitrepo
2440
return knitrepo.RepositoryFormatKnit3()
2443
def is_compatible(source, target):
2444
return InterRepository._same_model(source, target)
2447
def copy_content(self, revision_id=None):
2448
"""Make a complete copy of the content in self into destination.
2450
This copies both the repository's revision data, and configuration information
2451
such as the make_working_trees setting.
2453
This is a destructive operation! Do not use it on existing
2456
:param revision_id: Only copy the content needed to construct
2457
revision_id and its parents.
2460
self.target.set_make_working_trees(self.source.make_working_trees())
2461
except NotImplementedError:
2463
# but don't bother fetching if we have the needed data now.
2464
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2465
self.target.has_revision(revision_id)):
2467
self.target.fetch(self.source, revision_id=revision_id)
2470
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2471
"""See InterRepository.fetch()."""
2472
from bzrlib.fetch import GenericRepoFetcher
2473
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2474
self.source, self.source._format, self.target,
2475
self.target._format)
2476
f = GenericRepoFetcher(to_repository=self.target,
2477
from_repository=self.source,
2478
last_revision=revision_id,
2479
pb=pb, find_ghosts=find_ghosts)
2480
return f.count_copied, f.failed_revisions
2483
class InterWeaveRepo(InterSameDataRepository):
2484
"""Optimised code paths between Weave based repositories.
2486
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2487
implemented lazy inter-object optimisation.
2491
def _get_repo_format_to_test(self):
2492
from bzrlib.repofmt import weaverepo
2493
return weaverepo.RepositoryFormat7()
2496
def is_compatible(source, target):
2497
"""Be compatible with known Weave formats.
2499
We don't test for the stores being of specific types because that
2500
could lead to confusing results, and there is no need to be
2503
from bzrlib.repofmt.weaverepo import (
2509
return (isinstance(source._format, (RepositoryFormat5,
2511
RepositoryFormat7)) and
2512
isinstance(target._format, (RepositoryFormat5,
2514
RepositoryFormat7)))
2515
except AttributeError:
2519
def copy_content(self, revision_id=None):
2520
"""See InterRepository.copy_content()."""
2521
# weave specific optimised path:
2523
self.target.set_make_working_trees(self.source.make_working_trees())
2524
except NotImplementedError:
2526
# FIXME do not peek!
2527
if self.source.control_files._transport.listable():
2528
pb = ui.ui_factory.nested_progress_bar()
2530
self.target.weave_store.copy_all_ids(
2531
self.source.weave_store,
2533
from_transaction=self.source.get_transaction(),
2534
to_transaction=self.target.get_transaction())
2535
pb.update('copying inventory', 0, 1)
2536
self.target.control_weaves.copy_multi(
2537
self.source.control_weaves, ['inventory'],
2538
from_transaction=self.source.get_transaction(),
2539
to_transaction=self.target.get_transaction())
2540
self.target._revision_store.text_store.copy_all_ids(
2541
self.source._revision_store.text_store,
2546
self.target.fetch(self.source, revision_id=revision_id)
2549
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2550
"""See InterRepository.fetch()."""
2551
from bzrlib.fetch import GenericRepoFetcher
2552
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2553
self.source, self.source._format, self.target, self.target._format)
2554
f = GenericRepoFetcher(to_repository=self.target,
2555
from_repository=self.source,
2556
last_revision=revision_id,
2557
pb=pb, find_ghosts=find_ghosts)
2558
return f.count_copied, f.failed_revisions
2561
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2562
"""See InterRepository.missing_revision_ids()."""
2563
# we want all revisions to satisfy revision_id in source.
2564
# but we don't want to stat every file here and there.
2565
# we want then, all revisions other needs to satisfy revision_id
2566
# checked, but not those that we have locally.
2567
# so the first thing is to get a subset of the revisions to
2568
# satisfy revision_id in source, and then eliminate those that
2569
# we do already have.
2570
# this is slow on high latency connection to self, but as as this
2571
# disk format scales terribly for push anyway due to rewriting
2572
# inventory.weave, this is considered acceptable.
2574
if revision_id is not None:
2575
source_ids = self.source.get_ancestry(revision_id)
2576
assert source_ids[0] is None
2579
source_ids = self.source._all_possible_ids()
2580
source_ids_set = set(source_ids)
2581
# source_ids is the worst possible case we may need to pull.
2582
# now we want to filter source_ids against what we actually
2583
# have in target, but don't try to check for existence where we know
2584
# we do not have a revision as that would be pointless.
2585
target_ids = set(self.target._all_possible_ids())
2586
possibly_present_revisions = target_ids.intersection(source_ids_set)
2587
actually_present_revisions = set(
2588
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2589
required_revisions = source_ids_set.difference(actually_present_revisions)
2590
if revision_id is not None:
2591
# we used get_ancestry to determine source_ids then we are assured all
2592
# revisions referenced are present as they are installed in topological order.
2593
# and the tip revision was validated by get_ancestry.
2594
result_set = required_revisions
2596
# if we just grabbed the possibly available ids, then
2597
# we only have an estimate of whats available and need to validate
2598
# that against the revision records.
2600
self.source._eliminate_revisions_not_present(required_revisions))
2601
return self._set_to_search_result(result_set, self.source)
2604
class InterKnitRepo(InterSameDataRepository):
2605
"""Optimised code paths between Knit based repositories."""
2608
def _get_repo_format_to_test(self):
2609
from bzrlib.repofmt import knitrepo
2610
return knitrepo.RepositoryFormatKnit1()
2613
def is_compatible(source, target):
2614
"""Be compatible with known Knit formats.
2616
We don't test for the stores being of specific types because that
2617
could lead to confusing results, and there is no need to be
2620
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2622
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2623
isinstance(target._format, RepositoryFormatKnit))
2624
except AttributeError:
2626
return are_knits and InterRepository._same_model(source, target)
2629
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2630
"""See InterRepository.fetch()."""
2631
from bzrlib.fetch import KnitRepoFetcher
2632
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2633
self.source, self.source._format, self.target, self.target._format)
2634
f = KnitRepoFetcher(to_repository=self.target,
2635
from_repository=self.source,
2636
last_revision=revision_id,
2637
pb=pb, find_ghosts=find_ghosts)
2638
return f.count_copied, f.failed_revisions
2641
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2642
"""See InterRepository.missing_revision_ids()."""
2643
if revision_id is not None:
2644
source_ids = self.source.get_ancestry(revision_id)
2645
assert source_ids[0] is None
2648
source_ids = self.source.all_revision_ids()
2649
source_ids_set = set(source_ids)
2650
# source_ids is the worst possible case we may need to pull.
2651
# now we want to filter source_ids against what we actually
2652
# have in target, but don't try to check for existence where we know
2653
# we do not have a revision as that would be pointless.
2654
target_ids = set(self.target.all_revision_ids())
2655
possibly_present_revisions = target_ids.intersection(source_ids_set)
2656
actually_present_revisions = set(
2657
self.target._eliminate_revisions_not_present(possibly_present_revisions))
2658
required_revisions = source_ids_set.difference(actually_present_revisions)
2659
if revision_id is not None:
2660
# we used get_ancestry to determine source_ids then we are assured all
2661
# revisions referenced are present as they are installed in topological order.
2662
# and the tip revision was validated by get_ancestry.
2663
result_set = required_revisions
2665
# if we just grabbed the possibly available ids, then
2666
# we only have an estimate of whats available and need to validate
2667
# that against the revision records.
2669
self.source._eliminate_revisions_not_present(required_revisions))
2670
return self._set_to_search_result(result_set, self.source)
2673
class InterPackRepo(InterSameDataRepository):
2674
"""Optimised code paths between Pack based repositories."""
2677
def _get_repo_format_to_test(self):
2678
from bzrlib.repofmt import pack_repo
2679
return pack_repo.RepositoryFormatKnitPack1()
2682
def is_compatible(source, target):
2683
"""Be compatible with known Pack formats.
2685
We don't test for the stores being of specific types because that
2686
could lead to confusing results, and there is no need to be
2689
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2691
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2692
isinstance(target._format, RepositoryFormatPack))
2693
except AttributeError:
2695
return are_packs and InterRepository._same_model(source, target)
2698
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2699
"""See InterRepository.fetch()."""
2700
from bzrlib.repofmt.pack_repo import Packer
2701
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2702
self.source, self.source._format, self.target, self.target._format)
2703
self.count_copied = 0
2704
if revision_id is None:
2706
# everything to do - use pack logic
2707
# to fetch from all packs to one without
2708
# inventory parsing etc, IFF nothing to be copied is in the target.
2710
revision_ids = self.source.all_revision_ids()
2711
# implementing the TODO will involve:
2712
# - detecting when all of a pack is selected
2713
# - avoiding as much as possible pre-selection, so the
2714
# more-core routines such as create_pack_from_packs can filter in
2715
# a just-in-time fashion. (though having a HEADS list on a
2716
# repository might make this a lot easier, because we could
2717
# sensibly detect 'new revisions' without doing a full index scan.
2718
elif _mod_revision.is_null(revision_id):
2723
revision_ids = self.search_missing_revision_ids(revision_id,
2724
find_ghosts=find_ghosts).get_keys()
2725
except errors.NoSuchRevision:
2726
raise errors.InstallFailed([revision_id])
2727
packs = self.source._pack_collection.all_packs()
2728
pack = Packer(self.target._pack_collection, packs, '.fetch',
2729
revision_ids).pack()
2730
if pack is not None:
2731
self.target._pack_collection._save_pack_names()
2732
# Trigger an autopack. This may duplicate effort as we've just done
2733
# a pack creation, but for now it is simpler to think about as
2734
# 'upload data, then repack if needed'.
2735
self.target._pack_collection.autopack()
2736
return (pack.get_revision_count(), [])
2741
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2742
"""See InterRepository.missing_revision_ids().
2744
:param find_ghosts: Find ghosts throughout the ancestry of
2747
if not find_ghosts and revision_id is not None:
2748
return self._walk_to_common_revisions([revision_id])
2749
elif revision_id is not None:
2750
source_ids = self.source.get_ancestry(revision_id)
2751
assert source_ids[0] is None
2754
source_ids = self.source.all_revision_ids()
2755
# source_ids is the worst possible case we may need to pull.
2756
# now we want to filter source_ids against what we actually
2757
# have in target, but don't try to check for existence where we know
2758
# we do not have a revision as that would be pointless.
2759
target_ids = set(self.target.all_revision_ids())
2760
result_set = set(source_ids).difference(target_ids)
2761
return self._set_to_search_result(result_set, self.source)
2764
class InterModel1and2(InterRepository):
2767
def _get_repo_format_to_test(self):
2771
def is_compatible(source, target):
2772
if not source.supports_rich_root() and target.supports_rich_root():
2778
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2779
"""See InterRepository.fetch()."""
2780
from bzrlib.fetch import Model1toKnit2Fetcher
2781
f = Model1toKnit2Fetcher(to_repository=self.target,
2782
from_repository=self.source,
2783
last_revision=revision_id,
2784
pb=pb, find_ghosts=find_ghosts)
2785
return f.count_copied, f.failed_revisions
2788
def copy_content(self, revision_id=None):
2789
"""Make a complete copy of the content in self into destination.
2791
This is a destructive operation! Do not use it on existing
2794
:param revision_id: Only copy the content needed to construct
2795
revision_id and its parents.
2798
self.target.set_make_working_trees(self.source.make_working_trees())
2799
except NotImplementedError:
2801
# but don't bother fetching if we have the needed data now.
2802
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2803
self.target.has_revision(revision_id)):
2805
self.target.fetch(self.source, revision_id=revision_id)
2808
class InterKnit1and2(InterKnitRepo):
2811
def _get_repo_format_to_test(self):
2815
def is_compatible(source, target):
2816
"""Be compatible with Knit1 source and Knit3 target"""
2817
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2819
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2820
RepositoryFormatKnit3)
2821
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2822
RepositoryFormatKnitPack3)
2823
return (isinstance(source._format,
2824
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2825
isinstance(target._format,
2826
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2828
except AttributeError:
2832
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2833
"""See InterRepository.fetch()."""
2834
from bzrlib.fetch import Knit1to2Fetcher
2835
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2836
self.source, self.source._format, self.target,
2837
self.target._format)
2838
f = Knit1to2Fetcher(to_repository=self.target,
2839
from_repository=self.source,
2840
last_revision=revision_id,
2841
pb=pb, find_ghosts=find_ghosts)
2842
return f.count_copied, f.failed_revisions
2845
class InterDifferingSerializer(InterKnitRepo):
2848
def _get_repo_format_to_test(self):
2852
def is_compatible(source, target):
2853
"""Be compatible with Knit2 source and Knit3 target"""
2854
if source.supports_rich_root() != target.supports_rich_root():
2856
# Ideally, we'd support fetching if the source had no tree references
2857
# even if it supported them...
2858
if (getattr(source, '_format.supports_tree_reference', False) and
2859
not getattr(target, '_format.supports_tree_reference', False)):
2864
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2865
"""See InterRepository.fetch()."""
2866
revision_ids = self.target.missing_revision_ids(self.source,
2867
revision_id, find_ghosts=find_ghosts).get_keys()
2868
revision_ids = tsort.topo_sort(
2869
self.get_graph().get_parent_map(revision_ids))
2870
def revisions_iterator():
2871
for current_revision_id in revision_ids:
2872
revision = self.source.get_revision(current_revision_id)
2873
tree = self.source.revision_tree(current_revision_id)
2875
signature = self.source.get_signature_text(
2876
current_revision_id)
2877
except errors.NoSuchRevision:
2879
yield revision, tree, signature
2881
my_pb = ui.ui_factory.nested_progress_bar()
2886
install_revisions(self.target, revisions_iterator(),
2887
len(revision_ids), pb)
2889
if my_pb is not None:
2891
return len(revision_ids), 0
2894
class InterRemoteToOther(InterRepository):
2896
def __init__(self, source, target):
2897
InterRepository.__init__(self, source, target)
2898
self._real_inter = None
2901
def is_compatible(source, target):
2902
if not isinstance(source, remote.RemoteRepository):
2904
# Is source's model compatible with target's model?
2905
source._ensure_real()
2906
real_source = source._real_repository
2907
assert not isinstance(real_source, remote.RemoteRepository), (
2908
"We don't support remote repos backed by remote repos yet.")
2909
return InterRepository._same_model(real_source, target)
2912
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2913
"""See InterRepository.fetch()."""
2914
from bzrlib.fetch import RemoteToOtherFetcher
2915
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2916
self.source, self.target, self.target._format)
2917
# TODO: jam 20070210 This should be an assert, not a translate
2918
revision_id = osutils.safe_revision_id(revision_id)
2919
f = RemoteToOtherFetcher(to_repository=self.target,
2920
from_repository=self.source,
2921
last_revision=revision_id,
2922
pb=pb, find_ghosts=find_ghosts)
2923
return f.count_copied, f.failed_revisions
2926
def _get_repo_format_to_test(self):
2930
class InterOtherToRemote(InterRepository):
2932
def __init__(self, source, target):
2933
InterRepository.__init__(self, source, target)
2934
self._real_inter = None
2937
def is_compatible(source, target):
2938
if isinstance(target, remote.RemoteRepository):
2942
def _ensure_real_inter(self):
2943
if self._real_inter is None:
2944
self.target._ensure_real()
2945
real_target = self.target._real_repository
2946
self._real_inter = InterRepository.get(self.source, real_target)
2948
def copy_content(self, revision_id=None):
2949
self._ensure_real_inter()
2950
self._real_inter.copy_content(revision_id=revision_id)
2952
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2953
self._ensure_real_inter()
2954
self._real_inter.fetch(revision_id=revision_id, pb=pb,
2955
find_ghosts=find_ghosts)
2958
def _get_repo_format_to_test(self):
2962
InterRepository.register_optimiser(InterDifferingSerializer)
2963
InterRepository.register_optimiser(InterSameDataRepository)
2964
InterRepository.register_optimiser(InterWeaveRepo)
2965
InterRepository.register_optimiser(InterKnitRepo)
2966
InterRepository.register_optimiser(InterModel1and2)
2967
InterRepository.register_optimiser(InterKnit1and2)
2968
InterRepository.register_optimiser(InterPackRepo)
2969
InterRepository.register_optimiser(InterRemoteToOther)
2970
InterRepository.register_optimiser(InterOtherToRemote)
2973
class CopyConverter(object):
2974
"""A repository conversion tool which just performs a copy of the content.
2976
This is slow but quite reliable.
2979
def __init__(self, target_format):
2980
"""Create a CopyConverter.
2982
:param target_format: The format the resulting repository should be.
2984
self.target_format = target_format
2986
def convert(self, repo, pb):
2987
"""Perform the conversion of to_convert, giving feedback via pb.
2989
:param to_convert: The disk object to convert.
2990
:param pb: a progress bar to use for progress information.
2995
# this is only useful with metadir layouts - separated repo content.
2996
# trigger an assertion if not such
2997
repo._format.get_format_string()
2998
self.repo_dir = repo.bzrdir
2999
self.step('Moving repository to repository.backup')
3000
self.repo_dir.transport.move('repository', 'repository.backup')
3001
backup_transport = self.repo_dir.transport.clone('repository.backup')
3002
repo._format.check_conversion_target(self.target_format)
3003
self.source_repo = repo._format.open(self.repo_dir,
3005
_override_transport=backup_transport)
3006
self.step('Creating new repository')
3007
converted = self.target_format.initialize(self.repo_dir,
3008
self.source_repo.is_shared())
3009
converted.lock_write()
3011
self.step('Copying content into repository.')
3012
self.source_repo.copy_content_into(converted)
3015
self.step('Deleting old repository content.')
3016
self.repo_dir.transport.delete_tree('repository.backup')
3017
self.pb.note('repository converted')
3019
def step(self, message):
3020
"""Update the pb by a step."""
3022
self.pb.update(message, self.count, self.total)
3034
def _unescaper(match, _map=_unescape_map):
3035
code = match.group(1)
3039
if not code.startswith('#'):
3041
return unichr(int(code[1:])).encode('utf8')
3047
def _unescape_xml(data):
3048
"""Unescape predefined XML entities in a string of data."""
3050
if _unescape_re is None:
3051
_unescape_re = re.compile('\&([^;]*);')
3052
return _unescape_re.sub(_unescaper, data)
3055
class _VersionedFileChecker(object):
3057
def __init__(self, repository):
3058
self.repository = repository
3059
self.text_index = self.repository._generate_text_key_index()
3061
def calculate_file_version_parents(self, revision_id, file_id):
3062
"""Calculate the correct parents for a file version according to
3065
parent_keys = self.text_index[(file_id, revision_id)]
3066
if parent_keys == [_mod_revision.NULL_REVISION]:
3068
# strip the file_id, for the weave api
3069
return tuple([revision_id for file_id, revision_id in parent_keys])
3071
def check_file_version_parents(self, weave, file_id):
3072
"""Check the parents stored in a versioned file are correct.
3074
It also detects file versions that are not referenced by their
3075
corresponding revision's inventory.
3077
:returns: A tuple of (wrong_parents, dangling_file_versions).
3078
wrong_parents is a dict mapping {revision_id: (stored_parents,
3079
correct_parents)} for each revision_id where the stored parents
3080
are not correct. dangling_file_versions is a set of (file_id,
3081
revision_id) tuples for versions that are present in this versioned
3082
file, but not used by the corresponding inventory.
3085
unused_versions = set()
3086
for num, revision_id in enumerate(weave.versions()):
3088
correct_parents = self.calculate_file_version_parents(
3089
revision_id, file_id)
3091
# The version is not part of the used keys.
3092
unused_versions.add(revision_id)
3095
knit_parents = tuple(weave.get_parents(revision_id))
3096
except errors.RevisionNotPresent:
3098
if correct_parents != knit_parents:
3099
wrong_parents[revision_id] = (knit_parents, correct_parents)
3100
return wrong_parents, unused_versions
added
removed
bzrlib/repository.py
