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
knit.insert_data_stream(
846
(format, data_list, StringIO(knit_bytes).read))
849
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
850
"""Return the revision ids that other has that this does not.
852
These are returned in topological order.
854
revision_id: only return revision ids included by revision_id.
856
return InterRepository.get(other, self).missing_revision_ids(
857
revision_id, find_ghosts)
861
"""Open the repository rooted at base.
863
For instance, if the repository is at URL/.bzr/repository,
864
Repository.open(URL) -> a Repository instance.
866
control = bzrdir.BzrDir.open(base)
867
return control.open_repository()
869
def copy_content_into(self, destination, revision_id=None):
870
"""Make a complete copy of the content in self into destination.
872
This is a destructive operation! Do not use it on existing
875
return InterRepository.get(self, destination).copy_content(revision_id)
877
def commit_write_group(self):
878
"""Commit the contents accrued within the current write group.
880
:seealso: start_write_group.
882
if self._write_group is not self.get_transaction():
883
# has an unlock or relock occured ?
884
raise errors.BzrError('mismatched lock context %r and '
886
(self.get_transaction(), self._write_group))
887
self._commit_write_group()
888
self._write_group = None
890
def _commit_write_group(self):
891
"""Template method for per-repository write group cleanup.
893
This is called before the write group is considered to be
894
finished and should ensure that all data handed to the repository
895
for writing during the write group is safely committed (to the
896
extent possible considering file system caching etc).
899
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
900
"""Fetch the content required to construct revision_id from source.
902
If revision_id is None all content is copied.
903
:param find_ghosts: Find and copy revisions in the source that are
904
ghosts in the target (and not reachable directly by walking out to
905
the first-present revision in target from revision_id).
907
# fast path same-url fetch operations
908
if self.has_same_location(source):
909
# check that last_revision is in 'from' and then return a
911
if (revision_id is not None and
912
not _mod_revision.is_null(revision_id)):
913
self.get_revision(revision_id)
915
inter = InterRepository.get(source, self)
917
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
918
except NotImplementedError:
919
raise errors.IncompatibleRepositories(source, self)
921
def create_bundle(self, target, base, fileobj, format=None):
922
return serializer.write_bundle(self, target, base, fileobj, format)
924
def get_commit_builder(self, branch, parents, config, timestamp=None,
925
timezone=None, committer=None, revprops=None,
927
"""Obtain a CommitBuilder for this repository.
929
:param branch: Branch to commit to.
930
:param parents: Revision ids of the parents of the new revision.
931
:param config: Configuration to use.
932
:param timestamp: Optional timestamp recorded for commit.
933
:param timezone: Optional timezone for timestamp.
934
:param committer: Optional committer to set for commit.
935
:param revprops: Optional dictionary of revision properties.
936
:param revision_id: Optional revision id.
938
result = self._commit_builder_class(self, parents, config,
939
timestamp, timezone, committer, revprops, revision_id)
940
self.start_write_group()
944
if (self.control_files._lock_count == 1 and
945
self.control_files._lock_mode == 'w'):
946
if self._write_group is not None:
947
self.abort_write_group()
948
self.control_files.unlock()
949
raise errors.BzrError(
950
'Must end write groups before releasing write locks.')
951
self.control_files.unlock()
954
def clone(self, a_bzrdir, revision_id=None):
955
"""Clone this repository into a_bzrdir using the current format.
957
Currently no check is made that the format of this repository and
958
the bzrdir format are compatible. FIXME RBC 20060201.
960
:return: The newly created destination repository.
962
# TODO: deprecate after 0.16; cloning this with all its settings is
963
# probably not very useful -- mbp 20070423
964
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
965
self.copy_content_into(dest_repo, revision_id)
968
def start_write_group(self):
969
"""Start a write group in the repository.
971
Write groups are used by repositories which do not have a 1:1 mapping
972
between file ids and backend store to manage the insertion of data from
973
both fetch and commit operations.
975
A write lock is required around the start_write_group/commit_write_group
976
for the support of lock-requiring repository formats.
978
One can only insert data into a repository inside a write group.
982
if not self.is_write_locked():
983
raise errors.NotWriteLocked(self)
984
if self._write_group:
985
raise errors.BzrError('already in a write group')
986
self._start_write_group()
987
# so we can detect unlock/relock - the write group is now entered.
988
self._write_group = self.get_transaction()
990
def _start_write_group(self):
991
"""Template method for per-repository write group startup.
993
This is called before the write group is considered to be
998
def sprout(self, to_bzrdir, revision_id=None):
999
"""Create a descendent repository for new development.
1001
Unlike clone, this does not copy the settings of the repository.
1003
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1004
dest_repo.fetch(self, revision_id=revision_id)
1007
def _create_sprouting_repo(self, a_bzrdir, shared):
1008
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1009
# use target default format.
1010
dest_repo = a_bzrdir.create_repository()
1012
# Most control formats need the repository to be specifically
1013
# created, but on some old all-in-one formats it's not needed
1015
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1016
except errors.UninitializableFormat:
1017
dest_repo = a_bzrdir.open_repository()
1021
def has_revision(self, revision_id):
1022
"""True if this repository has a copy of the revision."""
1023
return revision_id in self.has_revisions((revision_id,))
1025
def has_revisions(self, revision_ids):
1026
"""Probe to find out the presence of multiple revisions.
1028
:param revision_ids: An iterable of revision_ids.
1029
:return: A set of the revision_ids that were present.
1031
raise NotImplementedError(self.has_revisions)
1033
return self._revision_store.has_revision_id(revision_id,
1034
self.get_transaction())
1037
def get_revision(self, revision_id):
1038
"""Return the Revision object for a named revision."""
1039
return self.get_revisions([revision_id])[0]
1042
def get_revision_reconcile(self, revision_id):
1043
"""'reconcile' helper routine that allows access to a revision always.
1045
This variant of get_revision does not cross check the weave graph
1046
against the revision one as get_revision does: but it should only
1047
be used by reconcile, or reconcile-alike commands that are correcting
1048
or testing the revision graph.
1050
return self._get_revisions([revision_id])[0]
1053
def get_revisions(self, revision_ids):
1054
"""Get many revisions at once."""
1055
return self._get_revisions(revision_ids)
1058
def _get_revisions(self, revision_ids):
1059
"""Core work logic to get many revisions without sanity checks."""
1060
for rev_id in revision_ids:
1061
if not rev_id or not isinstance(rev_id, basestring):
1062
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1063
revs = self._revision_store.get_revisions(revision_ids,
1064
self.get_transaction())
1066
assert not isinstance(rev.revision_id, unicode)
1067
for parent_id in rev.parent_ids:
1068
assert not isinstance(parent_id, unicode)
1072
def get_revision_xml(self, revision_id):
1073
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1074
# would have already do it.
1075
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1076
rev = self.get_revision(revision_id)
1077
rev_tmp = StringIO()
1078
# the current serializer..
1079
self._revision_store._serializer.write_revision(rev, rev_tmp)
1081
return rev_tmp.getvalue()
1084
def get_deltas_for_revisions(self, revisions):
1085
"""Produce a generator of revision deltas.
1087
Note that the input is a sequence of REVISIONS, not revision_ids.
1088
Trees will be held in memory until the generator exits.
1089
Each delta is relative to the revision's lefthand predecessor.
1091
required_trees = set()
1092
for revision in revisions:
1093
required_trees.add(revision.revision_id)
1094
required_trees.update(revision.parent_ids[:1])
1095
trees = dict((t.get_revision_id(), t) for
1096
t in self.revision_trees(required_trees))
1097
for revision in revisions:
1098
if not revision.parent_ids:
1099
old_tree = self.revision_tree(None)
1101
old_tree = trees[revision.parent_ids[0]]
1102
yield trees[revision.revision_id].changes_from(old_tree)
1105
def get_revision_delta(self, revision_id):
1106
"""Return the delta for one revision.
1108
The delta is relative to the left-hand predecessor of the
1111
r = self.get_revision(revision_id)
1112
return list(self.get_deltas_for_revisions([r]))[0]
1115
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1116
signature = gpg_strategy.sign(plaintext)
1117
self.add_signature_text(revision_id, signature)
1120
def add_signature_text(self, revision_id, signature):
1121
self._revision_store.add_revision_signature_text(revision_id,
1123
self.get_transaction())
1125
def find_text_key_references(self):
1126
"""Find the text key references within the repository.
1128
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1129
revision_ids. Each altered file-ids has the exact revision_ids that
1130
altered it listed explicitly.
1131
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1132
to whether they were referred to by the inventory of the
1133
revision_id that they contain. The inventory texts from all present
1134
revision ids are assessed to generate this report.
1136
revision_ids = self.all_revision_ids()
1137
w = self.get_inventory_weave()
1138
pb = ui.ui_factory.nested_progress_bar()
1140
return self._find_text_key_references_from_xml_inventory_lines(
1141
w.iter_lines_added_or_present_in_versions(revision_ids, pb=pb))
1145
def _find_text_key_references_from_xml_inventory_lines(self,
1147
"""Core routine for extracting references to texts from inventories.
1149
This performs the translation of xml lines to revision ids.
1151
:param line_iterator: An iterator of lines, origin_version_id
1152
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1153
to whether they were referred to by the inventory of the
1154
revision_id that they contain. Note that if that revision_id was
1155
not part of the line_iterator's output then False will be given -
1156
even though it may actually refer to that key.
1158
if not self._serializer.support_altered_by_hack:
1159
raise AssertionError(
1160
"_find_text_key_references_from_xml_inventory_lines only "
1161
"supported for branches which store inventory as unnested xml"
1162
", not on %r" % self)
1165
# this code needs to read every new line in every inventory for the
1166
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1167
# not present in one of those inventories is unnecessary but not
1168
# harmful because we are filtering by the revision id marker in the
1169
# inventory lines : we only select file ids altered in one of those
1170
# revisions. We don't need to see all lines in the inventory because
1171
# only those added in an inventory in rev X can contain a revision=X
1173
unescape_revid_cache = {}
1174
unescape_fileid_cache = {}
1176
# jam 20061218 In a big fetch, this handles hundreds of thousands
1177
# of lines, so it has had a lot of inlining and optimizing done.
1178
# Sorry that it is a little bit messy.
1179
# Move several functions to be local variables, since this is a long
1181
search = self._file_ids_altered_regex.search
1182
unescape = _unescape_xml
1183
setdefault = result.setdefault
1184
for line, version_id in line_iterator:
1185
match = search(line)
1188
# One call to match.group() returning multiple items is quite a
1189
# bit faster than 2 calls to match.group() each returning 1
1190
file_id, revision_id = match.group('file_id', 'revision_id')
1192
# Inlining the cache lookups helps a lot when you make 170,000
1193
# lines and 350k ids, versus 8.4 unique ids.
1194
# Using a cache helps in 2 ways:
1195
# 1) Avoids unnecessary decoding calls
1196
# 2) Re-uses cached strings, which helps in future set and
1198
# (2) is enough that removing encoding entirely along with
1199
# the cache (so we are using plain strings) results in no
1200
# performance improvement.
1202
revision_id = unescape_revid_cache[revision_id]
1204
unescaped = unescape(revision_id)
1205
unescape_revid_cache[revision_id] = unescaped
1206
revision_id = unescaped
1208
# Note that unconditionally unescaping means that we deserialise
1209
# every fileid, which for general 'pull' is not great, but we don't
1210
# really want to have some many fulltexts that this matters anyway.
1213
file_id = unescape_fileid_cache[file_id]
1215
unescaped = unescape(file_id)
1216
unescape_fileid_cache[file_id] = unescaped
1219
key = (file_id, revision_id)
1220
setdefault(key, False)
1221
if revision_id == version_id:
1225
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1227
"""Helper routine for fileids_altered_by_revision_ids.
1229
This performs the translation of xml lines to revision ids.
1231
:param line_iterator: An iterator of lines, origin_version_id
1232
:param revision_ids: The revision ids to filter for. This should be a
1233
set or other type which supports efficient __contains__ lookups, as
1234
the revision id from each parsed line will be looked up in the
1235
revision_ids filter.
1236
:return: a dictionary mapping altered file-ids to an iterable of
1237
revision_ids. Each altered file-ids has the exact revision_ids that
1238
altered it listed explicitly.
1241
setdefault = result.setdefault
1242
for file_id, revision_id in \
1243
self._find_text_key_references_from_xml_inventory_lines(
1244
line_iterator).iterkeys():
1245
# once data is all ensured-consistent; then this is
1246
# if revision_id == version_id
1247
if revision_id in revision_ids:
1248
setdefault(file_id, set()).add(revision_id)
1251
def fileids_altered_by_revision_ids(self, revision_ids):
1252
"""Find the file ids and versions affected by revisions.
1254
:param revisions: an iterable containing revision ids.
1255
:return: a dictionary mapping altered file-ids to an iterable of
1256
revision_ids. Each altered file-ids has the exact revision_ids that
1257
altered it listed explicitly.
1259
selected_revision_ids = set(revision_ids)
1260
w = self.get_inventory_weave()
1261
pb = ui.ui_factory.nested_progress_bar()
1263
return self._find_file_ids_from_xml_inventory_lines(
1264
w.iter_lines_added_or_present_in_versions(
1265
selected_revision_ids, pb=pb),
1266
selected_revision_ids)
1270
def iter_files_bytes(self, desired_files):
1271
"""Iterate through file versions.
1273
Files will not necessarily be returned in the order they occur in
1274
desired_files. No specific order is guaranteed.
1276
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1277
value supplied by the caller as part of desired_files. It should
1278
uniquely identify the file version in the caller's context. (Examples:
1279
an index number or a TreeTransform trans_id.)
1281
bytes_iterator is an iterable of bytestrings for the file. The
1282
kind of iterable and length of the bytestrings are unspecified, but for
1283
this implementation, it is a list of lines produced by
1284
VersionedFile.get_lines().
1286
:param desired_files: a list of (file_id, revision_id, identifier)
1289
transaction = self.get_transaction()
1290
for file_id, revision_id, callable_data in desired_files:
1292
weave = self.weave_store.get_weave(file_id, transaction)
1293
except errors.NoSuchFile:
1294
raise errors.NoSuchIdInRepository(self, file_id)
1295
yield callable_data, weave.get_lines(revision_id)
1297
def _generate_text_key_index(self, text_key_references=None,
1299
"""Generate a new text key index for the repository.
1301
This is an expensive function that will take considerable time to run.
1303
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1304
list of parents, also text keys. When a given key has no parents,
1305
the parents list will be [NULL_REVISION].
1307
# All revisions, to find inventory parents.
1308
if ancestors is None:
1309
revision_graph = self.get_revision_graph_with_ghosts()
1310
ancestors = revision_graph.get_ancestors()
1311
if text_key_references is None:
1312
text_key_references = self.find_text_key_references()
1313
pb = ui.ui_factory.nested_progress_bar()
1315
return self._do_generate_text_key_index(ancestors,
1316
text_key_references, pb)
1320
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1321
"""Helper for _generate_text_key_index to avoid deep nesting."""
1322
revision_order = tsort.topo_sort(ancestors)
1323
invalid_keys = set()
1325
for revision_id in revision_order:
1326
revision_keys[revision_id] = set()
1327
text_count = len(text_key_references)
1328
# a cache of the text keys to allow reuse; costs a dict of all the
1329
# keys, but saves a 2-tuple for every child of a given key.
1331
for text_key, valid in text_key_references.iteritems():
1333
invalid_keys.add(text_key)
1335
revision_keys[text_key[1]].add(text_key)
1336
text_key_cache[text_key] = text_key
1337
del text_key_references
1339
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1340
NULL_REVISION = _mod_revision.NULL_REVISION
1341
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1342
# too small for large or very branchy trees. However, for 55K path
1343
# trees, it would be easy to use too much memory trivially. Ideally we
1344
# could gauge this by looking at available real memory etc, but this is
1345
# always a tricky proposition.
1346
inventory_cache = lru_cache.LRUCache(10)
1347
batch_size = 10 # should be ~150MB on a 55K path tree
1348
batch_count = len(revision_order) / batch_size + 1
1350
pb.update("Calculating text parents.", processed_texts, text_count)
1351
for offset in xrange(batch_count):
1352
to_query = revision_order[offset * batch_size:(offset + 1) *
1356
for rev_tree in self.revision_trees(to_query):
1357
revision_id = rev_tree.get_revision_id()
1358
parent_ids = ancestors[revision_id]
1359
for text_key in revision_keys[revision_id]:
1360
pb.update("Calculating text parents.", processed_texts)
1361
processed_texts += 1
1362
candidate_parents = []
1363
for parent_id in parent_ids:
1364
parent_text_key = (text_key[0], parent_id)
1366
check_parent = parent_text_key not in \
1367
revision_keys[parent_id]
1369
# the parent parent_id is a ghost:
1370
check_parent = False
1371
# truncate the derived graph against this ghost.
1372
parent_text_key = None
1374
# look at the parent commit details inventories to
1375
# determine possible candidates in the per file graph.
1378
inv = inventory_cache[parent_id]
1380
inv = self.revision_tree(parent_id).inventory
1381
inventory_cache[parent_id] = inv
1382
parent_entry = inv._byid.get(text_key[0], None)
1383
if parent_entry is not None:
1385
text_key[0], parent_entry.revision)
1387
parent_text_key = None
1388
if parent_text_key is not None:
1389
candidate_parents.append(
1390
text_key_cache[parent_text_key])
1391
parent_heads = text_graph.heads(candidate_parents)
1392
new_parents = list(parent_heads)
1393
new_parents.sort(key=lambda x:candidate_parents.index(x))
1394
if new_parents == []:
1395
new_parents = [NULL_REVISION]
1396
text_index[text_key] = new_parents
1398
for text_key in invalid_keys:
1399
text_index[text_key] = [NULL_REVISION]
1402
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1403
"""Get an iterable listing the keys of all the data introduced by a set
1406
The keys will be ordered so that the corresponding items can be safely
1407
fetched and inserted in that order.
1409
:returns: An iterable producing tuples of (knit-kind, file-id,
1410
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1411
'revisions'. file-id is None unless knit-kind is 'file'.
1413
# XXX: it's a bit weird to control the inventory weave caching in this
1414
# generator. Ideally the caching would be done in fetch.py I think. Or
1415
# maybe this generator should explicitly have the contract that it
1416
# should not be iterated until the previously yielded item has been
1419
inv_w = self.get_inventory_weave()
1420
inv_w.enable_cache()
1422
# file ids that changed
1423
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1425
num_file_ids = len(file_ids)
1426
for file_id, altered_versions in file_ids.iteritems():
1427
if _files_pb is not None:
1428
_files_pb.update("fetch texts", count, num_file_ids)
1430
yield ("file", file_id, altered_versions)
1431
# We're done with the files_pb. Note that it finished by the caller,
1432
# just as it was created by the caller.
1436
yield ("inventory", None, revision_ids)
1440
revisions_with_signatures = set()
1441
for rev_id in revision_ids:
1443
self.get_signature_text(rev_id)
1444
except errors.NoSuchRevision:
1448
revisions_with_signatures.add(rev_id)
1450
yield ("signatures", None, revisions_with_signatures)
1453
yield ("revisions", None, revision_ids)
1456
def get_inventory_weave(self):
1457
return self.control_weaves.get_weave('inventory',
1458
self.get_transaction())
1461
def get_inventory(self, revision_id):
1462
"""Get Inventory object by revision id."""
1463
return self.iter_inventories([revision_id]).next()
1465
def iter_inventories(self, revision_ids):
1466
"""Get many inventories by revision_ids.
1468
This will buffer some or all of the texts used in constructing the
1469
inventories in memory, but will only parse a single inventory at a
1472
:return: An iterator of inventories.
1474
assert None not in revision_ids
1475
assert _mod_revision.NULL_REVISION not in revision_ids
1476
return self._iter_inventories(revision_ids)
1478
def _iter_inventories(self, revision_ids):
1479
"""single-document based inventory iteration."""
1480
texts = self.get_inventory_weave().get_texts(revision_ids)
1481
for text, revision_id in zip(texts, revision_ids):
1482
yield self.deserialise_inventory(revision_id, text)
1484
def deserialise_inventory(self, revision_id, xml):
1485
"""Transform the xml into an inventory object.
1487
:param revision_id: The expected revision id of the inventory.
1488
:param xml: A serialised inventory.
1490
result = self._serializer.read_inventory_from_string(xml, revision_id)
1491
if result.revision_id != revision_id:
1492
raise AssertionError('revision id mismatch %s != %s' % (
1493
result.revision_id, revision_id))
1496
def serialise_inventory(self, inv):
1497
return self._serializer.write_inventory_to_string(inv)
1499
def _serialise_inventory_to_lines(self, inv):
1500
return self._serializer.write_inventory_to_lines(inv)
1502
def get_serializer_format(self):
1503
return self._serializer.format_num
1506
def get_inventory_xml(self, revision_id):
1507
"""Get inventory XML as a file object."""
1509
assert isinstance(revision_id, str), type(revision_id)
1510
iw = self.get_inventory_weave()
1511
return iw.get_text(revision_id)
1513
raise errors.HistoryMissing(self, 'inventory', revision_id)
1516
def get_inventory_sha1(self, revision_id):
1517
"""Return the sha1 hash of the inventory entry
1519
return self.get_revision(revision_id).inventory_sha1
1522
def get_revision_graph(self, revision_id=None):
1523
"""Return a dictionary containing the revision graph.
1525
NB: This method should not be used as it accesses the entire graph all
1526
at once, which is much more data than most operations should require.
1528
:param revision_id: The revision_id to get a graph from. If None, then
1529
the entire revision graph is returned. This is a deprecated mode of
1530
operation and will be removed in the future.
1531
:return: a dictionary of revision_id->revision_parents_list.
1533
raise NotImplementedError(self.get_revision_graph)
1536
def get_revision_graph_with_ghosts(self, revision_ids=None):
1537
"""Return a graph of the revisions with ghosts marked as applicable.
1539
:param revision_ids: an iterable of revisions to graph or None for all.
1540
:return: a Graph object with the graph reachable from revision_ids.
1542
if 'evil' in debug.debug_flags:
1544
"get_revision_graph_with_ghosts scales with size of history.")
1545
result = deprecated_graph.Graph()
1546
if not revision_ids:
1547
pending = set(self.all_revision_ids())
1550
pending = set(revision_ids)
1551
# special case NULL_REVISION
1552
if _mod_revision.NULL_REVISION in pending:
1553
pending.remove(_mod_revision.NULL_REVISION)
1554
required = set(pending)
1557
revision_id = pending.pop()
1559
rev = self.get_revision(revision_id)
1560
except errors.NoSuchRevision:
1561
if revision_id in required:
1564
result.add_ghost(revision_id)
1566
for parent_id in rev.parent_ids:
1567
# is this queued or done ?
1568
if (parent_id not in pending and
1569
parent_id not in done):
1571
pending.add(parent_id)
1572
result.add_node(revision_id, rev.parent_ids)
1573
done.add(revision_id)
1576
def _get_history_vf(self):
1577
"""Get a versionedfile whose history graph reflects all revisions.
1579
For weave repositories, this is the inventory weave.
1581
return self.get_inventory_weave()
1583
def iter_reverse_revision_history(self, revision_id):
1584
"""Iterate backwards through revision ids in the lefthand history
1586
:param revision_id: The revision id to start with. All its lefthand
1587
ancestors will be traversed.
1589
if revision_id in (None, _mod_revision.NULL_REVISION):
1591
next_id = revision_id
1592
versionedfile = self._get_history_vf()
1595
parents = versionedfile.get_parents(next_id)
1596
if len(parents) == 0:
1599
next_id = parents[0]
1602
def get_revision_inventory(self, revision_id):
1603
"""Return inventory of a past revision."""
1604
# TODO: Unify this with get_inventory()
1605
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1606
# must be the same as its revision, so this is trivial.
1607
if revision_id is None:
1608
# This does not make sense: if there is no revision,
1609
# then it is the current tree inventory surely ?!
1610
# and thus get_root_id() is something that looks at the last
1611
# commit on the branch, and the get_root_id is an inventory check.
1612
raise NotImplementedError
1613
# return Inventory(self.get_root_id())
1615
return self.get_inventory(revision_id)
1618
def is_shared(self):
1619
"""Return True if this repository is flagged as a shared repository."""
1620
raise NotImplementedError(self.is_shared)
1623
def reconcile(self, other=None, thorough=False):
1624
"""Reconcile this repository."""
1625
from bzrlib.reconcile import RepoReconciler
1626
reconciler = RepoReconciler(self, thorough=thorough)
1627
reconciler.reconcile()
1630
def _refresh_data(self):
1631
"""Helper called from lock_* to ensure coherency with disk.
1633
The default implementation does nothing; it is however possible
1634
for repositories to maintain loaded indices across multiple locks
1635
by checking inside their implementation of this method to see
1636
whether their indices are still valid. This depends of course on
1637
the disk format being validatable in this manner.
1641
def revision_tree(self, revision_id):
1642
"""Return Tree for a revision on this branch.
1644
`revision_id` may be None for the empty tree revision.
1646
# TODO: refactor this to use an existing revision object
1647
# so we don't need to read it in twice.
1648
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1649
return RevisionTree(self, Inventory(root_id=None),
1650
_mod_revision.NULL_REVISION)
1652
inv = self.get_revision_inventory(revision_id)
1653
return RevisionTree(self, inv, revision_id)
1656
def revision_trees(self, revision_ids):
1657
"""Return Tree for a revision on this branch.
1659
`revision_id` may not be None or 'null:'"""
1660
inventories = self.iter_inventories(revision_ids)
1661
for inv in inventories:
1662
yield RevisionTree(self, inv, inv.revision_id)
1665
def get_ancestry(self, revision_id, topo_sorted=True):
1666
"""Return a list of revision-ids integrated by a revision.
1668
The first element of the list is always None, indicating the origin
1669
revision. This might change when we have history horizons, or
1670
perhaps we should have a new API.
1672
This is topologically sorted.
1674
if _mod_revision.is_null(revision_id):
1676
if not self.has_revision(revision_id):
1677
raise errors.NoSuchRevision(self, revision_id)
1678
w = self.get_inventory_weave()
1679
candidates = w.get_ancestry(revision_id, topo_sorted)
1680
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1683
"""Compress the data within the repository.
1685
This operation only makes sense for some repository types. For other
1686
types it should be a no-op that just returns.
1688
This stub method does not require a lock, but subclasses should use
1689
@needs_write_lock as this is a long running call its reasonable to
1690
implicitly lock for the user.
1694
def print_file(self, file, revision_id):
1695
"""Print `file` to stdout.
1697
FIXME RBC 20060125 as John Meinel points out this is a bad api
1698
- it writes to stdout, it assumes that that is valid etc. Fix
1699
by creating a new more flexible convenience function.
1701
tree = self.revision_tree(revision_id)
1702
# use inventory as it was in that revision
1703
file_id = tree.inventory.path2id(file)
1705
# TODO: jam 20060427 Write a test for this code path
1706
# it had a bug in it, and was raising the wrong
1708
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1709
tree.print_file(file_id)
1711
def get_transaction(self):
1712
return self.control_files.get_transaction()
1714
def revision_parents(self, revision_id):
1715
return self.get_inventory_weave().parent_names(revision_id)
1717
@deprecated_method(symbol_versioning.one_one)
1718
def get_parents(self, revision_ids):
1719
"""See StackedParentsProvider.get_parents"""
1720
parent_map = self.get_parent_map(revision_ids)
1721
return [parent_map.get(r, None) for r in revision_ids]
1723
def get_parent_map(self, keys):
1724
"""See graph._StackedParentsProvider.get_parent_map"""
1726
for revision_id in keys:
1727
if revision_id == _mod_revision.NULL_REVISION:
1728
parent_map[revision_id] = ()
1731
parent_id_list = self.get_revision(revision_id).parent_ids
1732
except errors.NoSuchRevision:
1735
if len(parent_id_list) == 0:
1736
parent_ids = (_mod_revision.NULL_REVISION,)
1738
parent_ids = tuple(parent_id_list)
1739
parent_map[revision_id] = parent_ids
1742
def _make_parents_provider(self):
1745
def get_graph(self, other_repository=None):
1746
"""Return the graph walker for this repository format"""
1747
parents_provider = self._make_parents_provider()
1748
if (other_repository is not None and
1749
other_repository.bzrdir.transport.base !=
1750
self.bzrdir.transport.base):
1751
parents_provider = graph._StackedParentsProvider(
1752
[parents_provider, other_repository._make_parents_provider()])
1753
return graph.Graph(parents_provider)
1755
def _get_versioned_file_checker(self):
1756
"""Return an object suitable for checking versioned files."""
1757
return _VersionedFileChecker(self)
1760
def set_make_working_trees(self, new_value):
1761
"""Set the policy flag for making working trees when creating branches.
1763
This only applies to branches that use this repository.
1765
The default is 'True'.
1766
:param new_value: True to restore the default, False to disable making
1769
raise NotImplementedError(self.set_make_working_trees)
1771
def make_working_trees(self):
1772
"""Returns the policy for making working trees on new branches."""
1773
raise NotImplementedError(self.make_working_trees)
1776
def sign_revision(self, revision_id, gpg_strategy):
1777
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1778
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1781
def has_signature_for_revision_id(self, revision_id):
1782
"""Query for a revision signature for revision_id in the repository."""
1783
return self._revision_store.has_signature(revision_id,
1784
self.get_transaction())
1787
def get_signature_text(self, revision_id):
1788
"""Return the text for a signature."""
1789
return self._revision_store.get_signature_text(revision_id,
1790
self.get_transaction())
1793
def check(self, revision_ids=None):
1794
"""Check consistency of all history of given revision_ids.
1796
Different repository implementations should override _check().
1798
:param revision_ids: A non-empty list of revision_ids whose ancestry
1799
will be checked. Typically the last revision_id of a branch.
1801
return self._check(revision_ids)
1803
def _check(self, revision_ids):
1804
result = check.Check(self)
1808
def _warn_if_deprecated(self):
1809
global _deprecation_warning_done
1810
if _deprecation_warning_done:
1812
_deprecation_warning_done = True
1813
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1814
% (self._format, self.bzrdir.transport.base))
1816
def supports_rich_root(self):
1817
return self._format.rich_root_data
1819
def _check_ascii_revisionid(self, revision_id, method):
1820
"""Private helper for ascii-only repositories."""
1821
# weave repositories refuse to store revisionids that are non-ascii.
1822
if revision_id is not None:
1823
# weaves require ascii revision ids.
1824
if isinstance(revision_id, unicode):
1826
revision_id.encode('ascii')
1827
except UnicodeEncodeError:
1828
raise errors.NonAsciiRevisionId(method, self)
1831
revision_id.decode('ascii')
1832
except UnicodeDecodeError:
1833
raise errors.NonAsciiRevisionId(method, self)
1835
def revision_graph_can_have_wrong_parents(self):
1836
"""Is it possible for this repository to have a revision graph with
1839
If True, then this repository must also implement
1840
_find_inconsistent_revision_parents so that check and reconcile can
1841
check for inconsistencies before proceeding with other checks that may
1842
depend on the revision index being consistent.
1844
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1846
# remove these delegates a while after bzr 0.15
1847
def __make_delegated(name, from_module):
1848
def _deprecated_repository_forwarder():
1849
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1850
% (name, from_module),
1853
m = __import__(from_module, globals(), locals(), [name])
1855
return getattr(m, name)
1856
except AttributeError:
1857
raise AttributeError('module %s has no name %s'
1859
globals()[name] = _deprecated_repository_forwarder
1862
'AllInOneRepository',
1863
'WeaveMetaDirRepository',
1864
'PreSplitOutRepositoryFormat',
1865
'RepositoryFormat4',
1866
'RepositoryFormat5',
1867
'RepositoryFormat6',
1868
'RepositoryFormat7',
1870
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1874
'RepositoryFormatKnit',
1875
'RepositoryFormatKnit1',
1877
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1880
def install_revision(repository, rev, revision_tree):
1881
"""Install all revision data into a repository."""
1882
install_revisions(repository, [(rev, revision_tree, None)])
1885
def install_revisions(repository, iterable, num_revisions=None, pb=None):
1886
"""Install all revision data into a repository.
1888
Accepts an iterable of revision, tree, signature tuples. The signature
1891
repository.start_write_group()
1893
for n, (revision, revision_tree, signature) in enumerate(iterable):
1894
_install_revision(repository, revision, revision_tree, signature)
1896
pb.update('Transferring revisions', n + 1, num_revisions)
1898
repository.abort_write_group()
1901
repository.commit_write_group()
1904
def _install_revision(repository, rev, revision_tree, signature):
1905
"""Install all revision data into a repository."""
1906
present_parents = []
1908
for p_id in rev.parent_ids:
1909
if repository.has_revision(p_id):
1910
present_parents.append(p_id)
1911
parent_trees[p_id] = repository.revision_tree(p_id)
1913
parent_trees[p_id] = repository.revision_tree(None)
1915
inv = revision_tree.inventory
1916
entries = inv.iter_entries()
1917
# backwards compatibility hack: skip the root id.
1918
if not repository.supports_rich_root():
1919
path, root = entries.next()
1920
if root.revision != rev.revision_id:
1921
raise errors.IncompatibleRevision(repr(repository))
1922
# Add the texts that are not already present
1923
for path, ie in entries:
1924
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1925
repository.get_transaction())
1926
if ie.revision not in w:
1928
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1929
# with InventoryEntry.find_previous_heads(). if it is, then there
1930
# is a latent bug here where the parents may have ancestors of each
1932
for revision, tree in parent_trees.iteritems():
1933
if ie.file_id not in tree:
1935
parent_id = tree.inventory[ie.file_id].revision
1936
if parent_id in text_parents:
1938
text_parents.append(parent_id)
1940
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1941
repository.get_transaction())
1942
lines = revision_tree.get_file(ie.file_id).readlines()
1943
vfile.add_lines(rev.revision_id, text_parents, lines)
1945
# install the inventory
1946
repository.add_inventory(rev.revision_id, inv, present_parents)
1947
except errors.RevisionAlreadyPresent:
1949
if signature is not None:
1950
repository.add_signature_text(rev.revision_id, signature)
1951
repository.add_revision(rev.revision_id, rev, inv)
1954
class MetaDirRepository(Repository):
1955
"""Repositories in the new meta-dir layout."""
1957
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1958
super(MetaDirRepository, self).__init__(_format,
1964
dir_mode = self.control_files._dir_mode
1965
file_mode = self.control_files._file_mode
1968
def is_shared(self):
1969
"""Return True if this repository is flagged as a shared repository."""
1970
return self.control_files._transport.has('shared-storage')
1973
def set_make_working_trees(self, new_value):
1974
"""Set the policy flag for making working trees when creating branches.
1976
This only applies to branches that use this repository.
1978
The default is 'True'.
1979
:param new_value: True to restore the default, False to disable making
1984
self.control_files._transport.delete('no-working-trees')
1985
except errors.NoSuchFile:
1988
self.control_files.put_utf8('no-working-trees', '')
1990
def make_working_trees(self):
1991
"""Returns the policy for making working trees on new branches."""
1992
return not self.control_files._transport.has('no-working-trees')
1995
class RepositoryFormatRegistry(registry.Registry):
1996
"""Registry of RepositoryFormats."""
1998
def get(self, format_string):
1999
r = registry.Registry.get(self, format_string)
2005
format_registry = RepositoryFormatRegistry()
2006
"""Registry of formats, indexed by their identifying format string.
2008
This can contain either format instances themselves, or classes/factories that
2009
can be called to obtain one.
2013
#####################################################################
2014
# Repository Formats
2016
class RepositoryFormat(object):
2017
"""A repository format.
2019
Formats provide three things:
2020
* An initialization routine to construct repository data on disk.
2021
* a format string which is used when the BzrDir supports versioned
2023
* an open routine which returns a Repository instance.
2025
There is one and only one Format subclass for each on-disk format. But
2026
there can be one Repository subclass that is used for several different
2027
formats. The _format attribute on a Repository instance can be used to
2028
determine the disk format.
2030
Formats are placed in an dict by their format string for reference
2031
during opening. These should be subclasses of RepositoryFormat
2034
Once a format is deprecated, just deprecate the initialize and open
2035
methods on the format class. Do not deprecate the object, as the
2036
object will be created every system load.
2038
Common instance attributes:
2039
_matchingbzrdir - the bzrdir format that the repository format was
2040
originally written to work with. This can be used if manually
2041
constructing a bzrdir and repository, or more commonly for test suite
2045
# Set to True or False in derived classes. True indicates that the format
2046
# supports ghosts gracefully.
2047
supports_ghosts = None
2050
return "<%s>" % self.__class__.__name__
2052
def __eq__(self, other):
2053
# format objects are generally stateless
2054
return isinstance(other, self.__class__)
2056
def __ne__(self, other):
2057
return not self == other
2060
def find_format(klass, a_bzrdir):
2061
"""Return the format for the repository object in a_bzrdir.
2063
This is used by bzr native formats that have a "format" file in
2064
the repository. Other methods may be used by different types of
2068
transport = a_bzrdir.get_repository_transport(None)
2069
format_string = transport.get("format").read()
2070
return format_registry.get(format_string)
2071
except errors.NoSuchFile:
2072
raise errors.NoRepositoryPresent(a_bzrdir)
2074
raise errors.UnknownFormatError(format=format_string)
2077
def register_format(klass, format):
2078
format_registry.register(format.get_format_string(), format)
2081
def unregister_format(klass, format):
2082
format_registry.remove(format.get_format_string())
2085
def get_default_format(klass):
2086
"""Return the current default format."""
2087
from bzrlib import bzrdir
2088
return bzrdir.format_registry.make_bzrdir('default').repository_format
2090
def _get_control_store(self, repo_transport, control_files):
2091
"""Return the control store for this repository."""
2092
raise NotImplementedError(self._get_control_store)
2094
def get_format_string(self):
2095
"""Return the ASCII format string that identifies this format.
2097
Note that in pre format ?? repositories the format string is
2098
not permitted nor written to disk.
2100
raise NotImplementedError(self.get_format_string)
2102
def get_format_description(self):
2103
"""Return the short description for this format."""
2104
raise NotImplementedError(self.get_format_description)
2106
def _get_revision_store(self, repo_transport, control_files):
2107
"""Return the revision store object for this a_bzrdir."""
2108
raise NotImplementedError(self._get_revision_store)
2110
def _get_text_rev_store(self,
2117
"""Common logic for getting a revision store for a repository.
2119
see self._get_revision_store for the subclass-overridable method to
2120
get the store for a repository.
2122
from bzrlib.store.revision.text import TextRevisionStore
2123
dir_mode = control_files._dir_mode
2124
file_mode = control_files._file_mode
2125
text_store = TextStore(transport.clone(name),
2127
compressed=compressed,
2129
file_mode=file_mode)
2130
_revision_store = TextRevisionStore(text_store, serializer)
2131
return _revision_store
2133
# TODO: this shouldn't be in the base class, it's specific to things that
2134
# use weaves or knits -- mbp 20070207
2135
def _get_versioned_file_store(self,
2140
versionedfile_class=None,
2141
versionedfile_kwargs={},
2143
if versionedfile_class is None:
2144
versionedfile_class = self._versionedfile_class
2145
weave_transport = control_files._transport.clone(name)
2146
dir_mode = control_files._dir_mode
2147
file_mode = control_files._file_mode
2148
return VersionedFileStore(weave_transport, prefixed=prefixed,
2150
file_mode=file_mode,
2151
versionedfile_class=versionedfile_class,
2152
versionedfile_kwargs=versionedfile_kwargs,
2155
def initialize(self, a_bzrdir, shared=False):
2156
"""Initialize a repository of this format in a_bzrdir.
2158
:param a_bzrdir: The bzrdir to put the new repository in it.
2159
:param shared: The repository should be initialized as a sharable one.
2160
:returns: The new repository object.
2162
This may raise UninitializableFormat if shared repository are not
2163
compatible the a_bzrdir.
2165
raise NotImplementedError(self.initialize)
2167
def is_supported(self):
2168
"""Is this format supported?
2170
Supported formats must be initializable and openable.
2171
Unsupported formats may not support initialization or committing or
2172
some other features depending on the reason for not being supported.
2176
def check_conversion_target(self, target_format):
2177
raise NotImplementedError(self.check_conversion_target)
2179
def open(self, a_bzrdir, _found=False):
2180
"""Return an instance of this format for the bzrdir a_bzrdir.
2182
_found is a private parameter, do not use it.
2184
raise NotImplementedError(self.open)
2187
class MetaDirRepositoryFormat(RepositoryFormat):
2188
"""Common base class for the new repositories using the metadir layout."""
2190
rich_root_data = False
2191
supports_tree_reference = False
2192
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
2195
super(MetaDirRepositoryFormat, self).__init__()
2197
def _create_control_files(self, a_bzrdir):
2198
"""Create the required files and the initial control_files object."""
2199
# FIXME: RBC 20060125 don't peek under the covers
2200
# NB: no need to escape relative paths that are url safe.
2201
repository_transport = a_bzrdir.get_repository_transport(self)
2202
control_files = lockable_files.LockableFiles(repository_transport,
2203
'lock', lockdir.LockDir)
2204
control_files.create_lock()
2205
return control_files
2207
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2208
"""Upload the initial blank content."""
2209
control_files = self._create_control_files(a_bzrdir)
2210
control_files.lock_write()
2212
control_files._transport.mkdir_multi(dirs,
2213
mode=control_files._dir_mode)
2214
for file, content in files:
2215
control_files.put(file, content)
2216
for file, content in utf8_files:
2217
control_files.put_utf8(file, content)
2219
control_files.put_utf8('shared-storage', '')
2221
control_files.unlock()
2224
# formats which have no format string are not discoverable
2225
# and not independently creatable, so are not registered. They're
2226
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2227
# needed, it's constructed directly by the BzrDir. Non-native formats where
2228
# the repository is not separately opened are similar.
2230
format_registry.register_lazy(
2231
'Bazaar-NG Repository format 7',
2232
'bzrlib.repofmt.weaverepo',
2236
format_registry.register_lazy(
2237
'Bazaar-NG Knit Repository Format 1',
2238
'bzrlib.repofmt.knitrepo',
2239
'RepositoryFormatKnit1',
2242
format_registry.register_lazy(
2243
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2244
'bzrlib.repofmt.knitrepo',
2245
'RepositoryFormatKnit3',
2248
format_registry.register_lazy(
2249
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2250
'bzrlib.repofmt.knitrepo',
2251
'RepositoryFormatKnit4',
2254
# Pack-based formats. There is one format for pre-subtrees, and one for
2255
# post-subtrees to allow ease of testing.
2256
# NOTE: These are experimental in 0.92.
2257
format_registry.register_lazy(
2258
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2259
'bzrlib.repofmt.pack_repo',
2260
'RepositoryFormatKnitPack1',
2262
format_registry.register_lazy(
2263
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2264
'bzrlib.repofmt.pack_repo',
2265
'RepositoryFormatKnitPack3',
2267
format_registry.register_lazy(
2268
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2269
'bzrlib.repofmt.pack_repo',
2270
'RepositoryFormatKnitPack4',
2274
class InterRepository(InterObject):
2275
"""This class represents operations taking place between two repositories.
2277
Its instances have methods like copy_content and fetch, and contain
2278
references to the source and target repositories these operations can be
2281
Often we will provide convenience methods on 'repository' which carry out
2282
operations with another repository - they will always forward to
2283
InterRepository.get(other).method_name(parameters).
2287
"""The available optimised InterRepository types."""
2289
def copy_content(self, revision_id=None):
2290
raise NotImplementedError(self.copy_content)
2292
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2293
"""Fetch the content required to construct revision_id.
2295
The content is copied from self.source to self.target.
2297
:param revision_id: if None all content is copied, if NULL_REVISION no
2299
:param pb: optional progress bar to use for progress reports. If not
2300
provided a default one will be created.
2302
Returns the copied revision count and the failed revisions in a tuple:
2305
raise NotImplementedError(self.fetch)
2308
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2309
"""Return the revision ids that source has that target does not.
2311
These are returned in topological order.
2313
:param revision_id: only return revision ids included by this
2315
:param find_ghosts: If True find missing revisions in deep history
2316
rather than just finding the surface difference.
2318
# stop searching at found target revisions.
2319
if not find_ghosts and revision_id is not None:
2320
graph = self.source.get_graph()
2321
missing_revs = set()
2322
searcher = graph._make_breadth_first_searcher([revision_id])
2323
null_set = frozenset([_mod_revision.NULL_REVISION])
2326
next_revs = set(searcher.next())
2327
except StopIteration:
2329
next_revs.difference_update(null_set)
2330
have_revs = self.target.has_revisions(next_revs)
2331
missing_revs.update(next_revs - have_revs)
2332
searcher.stop_searching_any(have_revs)
2333
if next_revs - have_revs == set([revision_id]):
2334
# we saw the start rev itself, but no parents from it (or
2335
# next_revs would have been updated to e.g. set(). We remove
2336
# have_revs because if we found revision_id locally we
2337
# stop_searching at the first time around.
2338
raise errors.NoSuchRevision(self.source, revision_id)
2340
# generic, possibly worst case, slow code path.
2341
target_ids = set(self.target.all_revision_ids())
2342
if revision_id is not None:
2343
source_ids = self.source.get_ancestry(revision_id)
2344
assert source_ids[0] is None
2347
source_ids = self.source.all_revision_ids()
2348
result_set = set(source_ids).difference(target_ids)
2349
# this may look like a no-op: its not. It preserves the ordering
2350
# other_ids had while only returning the members from other_ids
2351
# that we've decided we need.
2352
return [rev_id for rev_id in source_ids if rev_id in result_set]
2355
def _same_model(source, target):
2356
"""True if source and target have the same data representation."""
2357
if source.supports_rich_root() != target.supports_rich_root():
2359
if source._serializer != target._serializer:
2364
class InterSameDataRepository(InterRepository):
2365
"""Code for converting between repositories that represent the same data.
2367
Data format and model must match for this to work.
2371
def _get_repo_format_to_test(self):
2372
"""Repository format for testing with.
2374
InterSameData can pull from subtree to subtree and from non-subtree to
2375
non-subtree, so we test this with the richest repository format.
2377
from bzrlib.repofmt import knitrepo
2378
return knitrepo.RepositoryFormatKnit3()
2381
def is_compatible(source, target):
2382
return InterRepository._same_model(source, target)
2385
def copy_content(self, revision_id=None):
2386
"""Make a complete copy of the content in self into destination.
2388
This copies both the repository's revision data, and configuration information
2389
such as the make_working_trees setting.
2391
This is a destructive operation! Do not use it on existing
2394
:param revision_id: Only copy the content needed to construct
2395
revision_id and its parents.
2398
self.target.set_make_working_trees(self.source.make_working_trees())
2399
except NotImplementedError:
2401
# but don't bother fetching if we have the needed data now.
2402
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2403
self.target.has_revision(revision_id)):
2405
self.target.fetch(self.source, revision_id=revision_id)
2408
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2409
"""See InterRepository.fetch()."""
2410
from bzrlib.fetch import GenericRepoFetcher
2411
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2412
self.source, self.source._format, self.target,
2413
self.target._format)
2414
f = GenericRepoFetcher(to_repository=self.target,
2415
from_repository=self.source,
2416
last_revision=revision_id,
2417
pb=pb, find_ghosts=find_ghosts)
2418
return f.count_copied, f.failed_revisions
2421
class InterWeaveRepo(InterSameDataRepository):
2422
"""Optimised code paths between Weave based repositories.
2424
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2425
implemented lazy inter-object optimisation.
2429
def _get_repo_format_to_test(self):
2430
from bzrlib.repofmt import weaverepo
2431
return weaverepo.RepositoryFormat7()
2434
def is_compatible(source, target):
2435
"""Be compatible with known Weave formats.
2437
We don't test for the stores being of specific types because that
2438
could lead to confusing results, and there is no need to be
2441
from bzrlib.repofmt.weaverepo import (
2447
return (isinstance(source._format, (RepositoryFormat5,
2449
RepositoryFormat7)) and
2450
isinstance(target._format, (RepositoryFormat5,
2452
RepositoryFormat7)))
2453
except AttributeError:
2457
def copy_content(self, revision_id=None):
2458
"""See InterRepository.copy_content()."""
2459
# weave specific optimised path:
2461
self.target.set_make_working_trees(self.source.make_working_trees())
2462
except NotImplementedError:
2464
# FIXME do not peek!
2465
if self.source.control_files._transport.listable():
2466
pb = ui.ui_factory.nested_progress_bar()
2468
self.target.weave_store.copy_all_ids(
2469
self.source.weave_store,
2471
from_transaction=self.source.get_transaction(),
2472
to_transaction=self.target.get_transaction())
2473
pb.update('copying inventory', 0, 1)
2474
self.target.control_weaves.copy_multi(
2475
self.source.control_weaves, ['inventory'],
2476
from_transaction=self.source.get_transaction(),
2477
to_transaction=self.target.get_transaction())
2478
self.target._revision_store.text_store.copy_all_ids(
2479
self.source._revision_store.text_store,
2484
self.target.fetch(self.source, revision_id=revision_id)
2487
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2488
"""See InterRepository.fetch()."""
2489
from bzrlib.fetch import GenericRepoFetcher
2490
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2491
self.source, self.source._format, self.target, self.target._format)
2492
f = GenericRepoFetcher(to_repository=self.target,
2493
from_repository=self.source,
2494
last_revision=revision_id,
2495
pb=pb, find_ghosts=find_ghosts)
2496
return f.count_copied, f.failed_revisions
2499
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2500
"""See InterRepository.missing_revision_ids()."""
2501
# we want all revisions to satisfy revision_id in source.
2502
# but we don't want to stat every file here and there.
2503
# we want then, all revisions other needs to satisfy revision_id
2504
# checked, but not those that we have locally.
2505
# so the first thing is to get a subset of the revisions to
2506
# satisfy revision_id in source, and then eliminate those that
2507
# we do already have.
2508
# this is slow on high latency connection to self, but as as this
2509
# disk format scales terribly for push anyway due to rewriting
2510
# inventory.weave, this is considered acceptable.
2512
if revision_id is not None:
2513
source_ids = self.source.get_ancestry(revision_id)
2514
assert source_ids[0] is None
2517
source_ids = self.source._all_possible_ids()
2518
source_ids_set = set(source_ids)
2519
# source_ids is the worst possible case we may need to pull.
2520
# now we want to filter source_ids against what we actually
2521
# have in target, but don't try to check for existence where we know
2522
# we do not have a revision as that would be pointless.
2523
target_ids = set(self.target._all_possible_ids())
2524
possibly_present_revisions = target_ids.intersection(source_ids_set)
2525
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2526
required_revisions = source_ids_set.difference(actually_present_revisions)
2527
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2528
if revision_id is not None:
2529
# we used get_ancestry to determine source_ids then we are assured all
2530
# revisions referenced are present as they are installed in topological order.
2531
# and the tip revision was validated by get_ancestry.
2532
return required_topo_revisions
2534
# if we just grabbed the possibly available ids, then
2535
# we only have an estimate of whats available and need to validate
2536
# that against the revision records.
2537
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2540
class InterKnitRepo(InterSameDataRepository):
2541
"""Optimised code paths between Knit based repositories."""
2544
def _get_repo_format_to_test(self):
2545
from bzrlib.repofmt import knitrepo
2546
return knitrepo.RepositoryFormatKnit1()
2549
def is_compatible(source, target):
2550
"""Be compatible with known Knit formats.
2552
We don't test for the stores being of specific types because that
2553
could lead to confusing results, and there is no need to be
2556
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2558
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2559
isinstance(target._format, RepositoryFormatKnit))
2560
except AttributeError:
2562
return are_knits and InterRepository._same_model(source, target)
2565
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2566
"""See InterRepository.fetch()."""
2567
from bzrlib.fetch import KnitRepoFetcher
2568
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2569
self.source, self.source._format, self.target, self.target._format)
2570
f = KnitRepoFetcher(to_repository=self.target,
2571
from_repository=self.source,
2572
last_revision=revision_id,
2573
pb=pb, find_ghosts=find_ghosts)
2574
return f.count_copied, f.failed_revisions
2577
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2578
"""See InterRepository.missing_revision_ids()."""
2579
if revision_id is not None:
2580
source_ids = self.source.get_ancestry(revision_id)
2581
assert source_ids[0] is None
2584
source_ids = self.source.all_revision_ids()
2585
source_ids_set = set(source_ids)
2586
# source_ids is the worst possible case we may need to pull.
2587
# now we want to filter source_ids against what we actually
2588
# have in target, but don't try to check for existence where we know
2589
# we do not have a revision as that would be pointless.
2590
target_ids = set(self.target.all_revision_ids())
2591
possibly_present_revisions = target_ids.intersection(source_ids_set)
2592
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2593
required_revisions = source_ids_set.difference(actually_present_revisions)
2594
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2595
if revision_id is not None:
2596
# we used get_ancestry to determine source_ids then we are assured all
2597
# revisions referenced are present as they are installed in topological order.
2598
# and the tip revision was validated by get_ancestry.
2599
return required_topo_revisions
2601
# if we just grabbed the possibly available ids, then
2602
# we only have an estimate of whats available and need to validate
2603
# that against the revision records.
2604
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2607
class InterPackRepo(InterSameDataRepository):
2608
"""Optimised code paths between Pack based repositories."""
2611
def _get_repo_format_to_test(self):
2612
from bzrlib.repofmt import pack_repo
2613
return pack_repo.RepositoryFormatKnitPack1()
2616
def is_compatible(source, target):
2617
"""Be compatible with known Pack formats.
2619
We don't test for the stores being of specific types because that
2620
could lead to confusing results, and there is no need to be
2623
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2625
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2626
isinstance(target._format, RepositoryFormatPack))
2627
except AttributeError:
2629
return are_packs and InterRepository._same_model(source, target)
2632
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2633
"""See InterRepository.fetch()."""
2634
from bzrlib.repofmt.pack_repo import Packer
2635
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2636
self.source, self.source._format, self.target, self.target._format)
2637
self.count_copied = 0
2638
if revision_id is None:
2640
# everything to do - use pack logic
2641
# to fetch from all packs to one without
2642
# inventory parsing etc, IFF nothing to be copied is in the target.
2644
revision_ids = self.source.all_revision_ids()
2645
# implementing the TODO will involve:
2646
# - detecting when all of a pack is selected
2647
# - avoiding as much as possible pre-selection, so the
2648
# more-core routines such as create_pack_from_packs can filter in
2649
# a just-in-time fashion. (though having a HEADS list on a
2650
# repository might make this a lot easier, because we could
2651
# sensibly detect 'new revisions' without doing a full index scan.
2652
elif _mod_revision.is_null(revision_id):
2657
revision_ids = self.missing_revision_ids(revision_id,
2658
find_ghosts=find_ghosts)
2659
except errors.NoSuchRevision:
2660
raise errors.InstallFailed([revision_id])
2661
packs = self.source._pack_collection.all_packs()
2662
pack = Packer(self.target._pack_collection, packs, '.fetch',
2663
revision_ids).pack()
2664
if pack is not None:
2665
self.target._pack_collection._save_pack_names()
2666
# Trigger an autopack. This may duplicate effort as we've just done
2667
# a pack creation, but for now it is simpler to think about as
2668
# 'upload data, then repack if needed'.
2669
self.target._pack_collection.autopack()
2670
return (pack.get_revision_count(), [])
2675
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2676
"""See InterRepository.missing_revision_ids().
2678
:param find_ghosts: Find ghosts throughout the ancestry of
2681
if not find_ghosts and revision_id is not None:
2682
graph = self.source.get_graph()
2683
missing_revs = set()
2684
searcher = graph._make_breadth_first_searcher([revision_id])
2685
null_set = frozenset([_mod_revision.NULL_REVISION])
2688
next_revs = set(searcher.next())
2689
except StopIteration:
2691
next_revs.difference_update(null_set)
2692
have_revs = self.target.has_revisions(next_revs)
2693
missing_revs.update(next_revs - have_revs)
2694
searcher.stop_searching_any(have_revs)
2695
if next_revs - have_revs == set([revision_id]):
2696
# we saw the start rev itself, but no parents from it (or
2697
# next_revs would have been updated to e.g. set(). We remove
2698
# have_revs because if we found revision_id locally we
2699
# stop_searching at the first time around.
2700
raise errors.NoSuchRevision(self.source, revision_id)
2702
elif revision_id is not None:
2703
source_ids = self.source.get_ancestry(revision_id)
2704
assert source_ids[0] is None
2707
source_ids = self.source.all_revision_ids()
2708
# source_ids is the worst possible case we may need to pull.
2709
# now we want to filter source_ids against what we actually
2710
# have in target, but don't try to check for existence where we know
2711
# we do not have a revision as that would be pointless.
2712
target_ids = set(self.target.all_revision_ids())
2713
return [r for r in source_ids if (r not in target_ids)]
2716
class InterModel1and2(InterRepository):
2719
def _get_repo_format_to_test(self):
2723
def is_compatible(source, target):
2724
if not source.supports_rich_root() and target.supports_rich_root():
2730
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2731
"""See InterRepository.fetch()."""
2732
from bzrlib.fetch import Model1toKnit2Fetcher
2733
f = Model1toKnit2Fetcher(to_repository=self.target,
2734
from_repository=self.source,
2735
last_revision=revision_id,
2736
pb=pb, find_ghosts=find_ghosts)
2737
return f.count_copied, f.failed_revisions
2740
def copy_content(self, revision_id=None):
2741
"""Make a complete copy of the content in self into destination.
2743
This is a destructive operation! Do not use it on existing
2746
:param revision_id: Only copy the content needed to construct
2747
revision_id and its parents.
2750
self.target.set_make_working_trees(self.source.make_working_trees())
2751
except NotImplementedError:
2753
# but don't bother fetching if we have the needed data now.
2754
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2755
self.target.has_revision(revision_id)):
2757
self.target.fetch(self.source, revision_id=revision_id)
2760
class InterKnit1and2(InterKnitRepo):
2763
def _get_repo_format_to_test(self):
2767
def is_compatible(source, target):
2768
"""Be compatible with Knit1 source and Knit3 target"""
2769
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2771
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2772
RepositoryFormatKnit3)
2773
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2774
RepositoryFormatKnitPack3)
2775
return (isinstance(source._format,
2776
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2777
isinstance(target._format,
2778
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2780
except AttributeError:
2784
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2785
"""See InterRepository.fetch()."""
2786
from bzrlib.fetch import Knit1to2Fetcher
2787
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2788
self.source, self.source._format, self.target,
2789
self.target._format)
2790
f = Knit1to2Fetcher(to_repository=self.target,
2791
from_repository=self.source,
2792
last_revision=revision_id,
2793
pb=pb, find_ghosts=find_ghosts)
2794
return f.count_copied, f.failed_revisions
2797
class InterDifferingSerializer(InterKnitRepo):
2800
def _get_repo_format_to_test(self):
2804
def is_compatible(source, target):
2805
"""Be compatible with Knit2 source and Knit3 target"""
2806
if source.supports_rich_root() != target.supports_rich_root():
2808
# Ideally, we'd support fetching if the source had no tree references
2809
# even if it supported them...
2810
if (getattr(source, '_format.supports_tree_reference', False) and
2811
not getattr(target, '_format.supports_tree_reference', False)):
2816
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2817
"""See InterRepository.fetch()."""
2818
revision_ids = self.target.missing_revision_ids(self.source,
2819
revision_id, find_ghosts=find_ghosts)
2820
def revisions_iterator():
2821
for current_revision_id in revision_ids:
2822
revision = self.source.get_revision(current_revision_id)
2823
tree = self.source.revision_tree(current_revision_id)
2825
signature = self.source.get_signature_text(
2826
current_revision_id)
2827
except errors.NoSuchRevision:
2829
yield revision, tree, signature
2831
my_pb = ui.ui_factory.nested_progress_bar()
2836
install_revisions(self.target, revisions_iterator(),
2837
len(revision_ids), pb)
2839
if my_pb is not None:
2841
return len(revision_ids), 0
2844
class InterRemoteToOther(InterRepository):
2846
def __init__(self, source, target):
2847
InterRepository.__init__(self, source, target)
2848
self._real_inter = None
2851
def is_compatible(source, target):
2852
if not isinstance(source, remote.RemoteRepository):
2854
source._ensure_real()
2855
real_source = source._real_repository
2856
# Is source's model compatible with target's model, and are they the
2857
# same format? Currently we can only optimise fetching from an
2858
# identical model & format repo.
2859
assert not isinstance(real_source, remote.RemoteRepository), (
2860
"We don't support remote repos backed by remote repos yet.")
2861
return real_source._format == target._format
2864
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2865
"""See InterRepository.fetch()."""
2866
from bzrlib.fetch import RemoteToOtherFetcher
2867
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2868
self.source, self.target, self.target._format)
2869
# TODO: jam 20070210 This should be an assert, not a translate
2870
revision_id = osutils.safe_revision_id(revision_id)
2871
f = RemoteToOtherFetcher(to_repository=self.target,
2872
from_repository=self.source,
2873
last_revision=revision_id,
2874
pb=pb, find_ghosts=find_ghosts)
2875
return f.count_copied, f.failed_revisions
2878
def _get_repo_format_to_test(self):
2882
class InterOtherToRemote(InterRepository):
2884
def __init__(self, source, target):
2885
InterRepository.__init__(self, source, target)
2886
self._real_inter = None
2889
def is_compatible(source, target):
2890
if isinstance(target, remote.RemoteRepository):
2894
def _ensure_real_inter(self):
2895
if self._real_inter is None:
2896
self.target._ensure_real()
2897
real_target = self.target._real_repository
2898
self._real_inter = InterRepository.get(self.source, real_target)
2900
def copy_content(self, revision_id=None):
2901
self._ensure_real_inter()
2902
self._real_inter.copy_content(revision_id=revision_id)
2904
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2905
self._ensure_real_inter()
2906
self._real_inter.fetch(revision_id=revision_id, pb=pb,
2907
find_ghosts=find_ghosts)
2910
def _get_repo_format_to_test(self):
2914
InterRepository.register_optimiser(InterDifferingSerializer)
2915
InterRepository.register_optimiser(InterSameDataRepository)
2916
InterRepository.register_optimiser(InterWeaveRepo)
2917
InterRepository.register_optimiser(InterKnitRepo)
2918
InterRepository.register_optimiser(InterModel1and2)
2919
InterRepository.register_optimiser(InterKnit1and2)
2920
InterRepository.register_optimiser(InterPackRepo)
2921
InterRepository.register_optimiser(InterRemoteToOther)
2922
InterRepository.register_optimiser(InterOtherToRemote)
2925
class CopyConverter(object):
2926
"""A repository conversion tool which just performs a copy of the content.
2928
This is slow but quite reliable.
2931
def __init__(self, target_format):
2932
"""Create a CopyConverter.
2934
:param target_format: The format the resulting repository should be.
2936
self.target_format = target_format
2938
def convert(self, repo, pb):
2939
"""Perform the conversion of to_convert, giving feedback via pb.
2941
:param to_convert: The disk object to convert.
2942
:param pb: a progress bar to use for progress information.
2947
# this is only useful with metadir layouts - separated repo content.
2948
# trigger an assertion if not such
2949
repo._format.get_format_string()
2950
self.repo_dir = repo.bzrdir
2951
self.step('Moving repository to repository.backup')
2952
self.repo_dir.transport.move('repository', 'repository.backup')
2953
backup_transport = self.repo_dir.transport.clone('repository.backup')
2954
repo._format.check_conversion_target(self.target_format)
2955
self.source_repo = repo._format.open(self.repo_dir,
2957
_override_transport=backup_transport)
2958
self.step('Creating new repository')
2959
converted = self.target_format.initialize(self.repo_dir,
2960
self.source_repo.is_shared())
2961
converted.lock_write()
2963
self.step('Copying content into repository.')
2964
self.source_repo.copy_content_into(converted)
2967
self.step('Deleting old repository content.')
2968
self.repo_dir.transport.delete_tree('repository.backup')
2969
self.pb.note('repository converted')
2971
def step(self, message):
2972
"""Update the pb by a step."""
2974
self.pb.update(message, self.count, self.total)
2986
def _unescaper(match, _map=_unescape_map):
2987
code = match.group(1)
2991
if not code.startswith('#'):
2993
return unichr(int(code[1:])).encode('utf8')
2999
def _unescape_xml(data):
3000
"""Unescape predefined XML entities in a string of data."""
3002
if _unescape_re is None:
3003
_unescape_re = re.compile('\&([^;]*);')
3004
return _unescape_re.sub(_unescaper, data)
3007
class _VersionedFileChecker(object):
3009
def __init__(self, repository):
3010
self.repository = repository
3011
self.text_index = self.repository._generate_text_key_index()
3013
def calculate_file_version_parents(self, revision_id, file_id):
3014
"""Calculate the correct parents for a file version according to
3017
parent_keys = self.text_index[(file_id, revision_id)]
3018
if parent_keys == [_mod_revision.NULL_REVISION]:
3020
# strip the file_id, for the weave api
3021
return tuple([revision_id for file_id, revision_id in parent_keys])
3023
def check_file_version_parents(self, weave, file_id):
3024
"""Check the parents stored in a versioned file are correct.
3026
It also detects file versions that are not referenced by their
3027
corresponding revision's inventory.
3029
:returns: A tuple of (wrong_parents, dangling_file_versions).
3030
wrong_parents is a dict mapping {revision_id: (stored_parents,
3031
correct_parents)} for each revision_id where the stored parents
3032
are not correct. dangling_file_versions is a set of (file_id,
3033
revision_id) tuples for versions that are present in this versioned
3034
file, but not used by the corresponding inventory.
3037
unused_versions = set()
3038
for num, revision_id in enumerate(weave.versions()):
3040
correct_parents = self.calculate_file_version_parents(
3041
revision_id, file_id)
3043
# The version is not part of the used keys.
3044
unused_versions.add(revision_id)
3047
knit_parents = tuple(weave.get_parents(revision_id))
3048
except errors.RevisionNotPresent:
3050
if correct_parents != knit_parents:
3051
wrong_parents[revision_id] = (knit_parents, correct_parents)
3052
return wrong_parents, unused_versions
added
removed
bzrlib/repository.py
