1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from cStringIO import StringIO
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
40
revision as _mod_revision,
46
from bzrlib.bundle import serializer
47
from bzrlib.revisiontree import RevisionTree
48
from bzrlib.store.versioned import VersionedFileStore
49
from bzrlib.store.text import TextStore
50
from bzrlib.testament import Testament
51
from bzrlib.util import bencode
54
from bzrlib.decorators import needs_read_lock, needs_write_lock
55
from bzrlib.inter import InterObject
56
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
57
from bzrlib.symbol_versioning import (
60
from bzrlib.trace import mutter, mutter_callsite, note, warning
63
# Old formats display a warning, but only once
64
_deprecation_warning_done = False
67
class CommitBuilder(object):
68
"""Provides an interface to build up a commit.
70
This allows describing a tree to be committed without needing to
71
know the internals of the format of the repository.
74
# all clients should supply tree roots.
75
record_root_entry = True
76
# the default CommitBuilder does not manage trees whose root is versioned.
77
_versioned_root = False
79
def __init__(self, repository, parents, config, timestamp=None,
80
timezone=None, committer=None, revprops=None,
82
"""Initiate a CommitBuilder.
84
:param repository: Repository to commit to.
85
:param parents: Revision ids of the parents of the new revision.
86
:param config: Configuration to use.
87
:param timestamp: Optional timestamp recorded for commit.
88
:param timezone: Optional timezone for timestamp.
89
:param committer: Optional committer to set for commit.
90
:param revprops: Optional dictionary of revision properties.
91
:param revision_id: Optional revision id.
96
self._committer = self._config.username()
98
assert isinstance(committer, basestring), type(committer)
99
self._committer = committer
101
self.new_inventory = Inventory(None)
102
self._new_revision_id = revision_id
103
self.parents = parents
104
self.repository = repository
107
if revprops is not None:
108
self._revprops.update(revprops)
110
if timestamp is None:
111
timestamp = time.time()
112
# Restrict resolution to 1ms
113
self._timestamp = round(timestamp, 3)
116
self._timezone = osutils.local_time_offset()
118
self._timezone = int(timezone)
120
self._generate_revision_if_needed()
121
self.__heads = graph.HeadsCache(repository.get_graph()).heads
123
def commit(self, message):
124
"""Make the actual commit.
126
:return: The revision id of the recorded revision.
128
rev = _mod_revision.Revision(
129
timestamp=self._timestamp,
130
timezone=self._timezone,
131
committer=self._committer,
133
inventory_sha1=self.inv_sha1,
134
revision_id=self._new_revision_id,
135
properties=self._revprops)
136
rev.parent_ids = self.parents
137
self.repository.add_revision(self._new_revision_id, rev,
138
self.new_inventory, self._config)
139
self.repository.commit_write_group()
140
return self._new_revision_id
143
"""Abort the commit that is being built.
145
self.repository.abort_write_group()
147
def revision_tree(self):
148
"""Return the tree that was just committed.
150
After calling commit() this can be called to get a RevisionTree
151
representing the newly committed tree. This is preferred to
152
calling Repository.revision_tree() because that may require
153
deserializing the inventory, while we already have a copy in
156
return RevisionTree(self.repository, self.new_inventory,
157
self._new_revision_id)
159
def finish_inventory(self):
160
"""Tell the builder that the inventory is finished."""
161
if self.new_inventory.root is None:
162
raise AssertionError('Root entry should be supplied to'
163
' record_entry_contents, as of bzr 0.10.',
164
DeprecationWarning, stacklevel=2)
165
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
166
self.new_inventory.revision_id = self._new_revision_id
167
self.inv_sha1 = self.repository.add_inventory(
168
self._new_revision_id,
173
def _gen_revision_id(self):
174
"""Return new revision-id."""
175
return generate_ids.gen_revision_id(self._config.username(),
178
def _generate_revision_if_needed(self):
179
"""Create a revision id if None was supplied.
181
If the repository can not support user-specified revision ids
182
they should override this function and raise CannotSetRevisionId
183
if _new_revision_id is not None.
185
:raises: CannotSetRevisionId
187
if self._new_revision_id is None:
188
self._new_revision_id = self._gen_revision_id()
189
self.random_revid = True
191
self.random_revid = False
193
def _heads(self, file_id, revision_ids):
194
"""Calculate the graph heads for revision_ids in the graph of file_id.
196
This can use either a per-file graph or a global revision graph as we
197
have an identity relationship between the two graphs.
199
return self.__heads(revision_ids)
201
def _check_root(self, ie, parent_invs, tree):
202
"""Helper for record_entry_contents.
204
:param ie: An entry being added.
205
:param parent_invs: The inventories of the parent revisions of the
207
:param tree: The tree that is being committed.
209
# In this revision format, root entries have no knit or weave When
210
# serializing out to disk and back in root.revision is always
212
ie.revision = self._new_revision_id
214
def _get_delta(self, ie, basis_inv, path):
215
"""Get a delta against the basis inventory for ie."""
216
if ie.file_id not in basis_inv:
218
return (None, path, ie.file_id, ie)
219
elif ie != basis_inv[ie.file_id]:
221
# TODO: avoid tis id2path call.
222
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
227
def record_entry_contents(self, ie, parent_invs, path, tree,
229
"""Record the content of ie from tree into the commit if needed.
231
Side effect: sets ie.revision when unchanged
233
:param ie: An inventory entry present in the commit.
234
:param parent_invs: The inventories of the parent revisions of the
236
:param path: The path the entry is at in the tree.
237
:param tree: The tree which contains this entry and should be used to
239
:param content_summary: Summary data from the tree about the paths
240
content - stat, length, exec, sha/link target. This is only
241
accessed when the entry has a revision of None - that is when it is
242
a candidate to commit.
243
:return: A tuple (change_delta, version_recorded). change_delta is
244
an inventory_delta change for this entry against the basis tree of
245
the commit, or None if no change occured against the basis tree.
246
version_recorded is True if a new version of the entry has been
247
recorded. For instance, committing a merge where a file was only
248
changed on the other side will return (delta, False).
250
if self.new_inventory.root is None:
251
if ie.parent_id is not None:
252
raise errors.RootMissing()
253
self._check_root(ie, parent_invs, tree)
254
if ie.revision is None:
255
kind = content_summary[0]
257
# ie is carried over from a prior commit
259
# XXX: repository specific check for nested tree support goes here - if
260
# the repo doesn't want nested trees we skip it ?
261
if (kind == 'tree-reference' and
262
not self.repository._format.supports_tree_reference):
263
# mismatch between commit builder logic and repository:
264
# this needs the entry creation pushed down into the builder.
265
raise NotImplementedError('Missing repository subtree support.')
266
self.new_inventory.add(ie)
268
# TODO: slow, take it out of the inner loop.
270
basis_inv = parent_invs[0]
272
basis_inv = Inventory(root_id=None)
274
# ie.revision is always None if the InventoryEntry is considered
275
# for committing. We may record the previous parents revision if the
276
# content is actually unchanged against a sole head.
277
if ie.revision is not None:
278
if not self._versioned_root and path == '':
279
# repositories that do not version the root set the root's
280
# revision to the new commit even when no change occurs, and
281
# this masks when a change may have occurred against the basis,
282
# so calculate if one happened.
283
if ie.file_id in basis_inv:
284
delta = (basis_inv.id2path(ie.file_id), path,
288
delta = (None, path, ie.file_id, ie)
291
# we don't need to commit this, because the caller already
292
# determined that an existing revision of this file is
294
return None, (ie.revision == self._new_revision_id)
295
# XXX: Friction: parent_candidates should return a list not a dict
296
# so that we don't have to walk the inventories again.
297
parent_candiate_entries = ie.parent_candidates(parent_invs)
298
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
300
for inv in parent_invs:
301
if ie.file_id in inv:
302
old_rev = inv[ie.file_id].revision
303
if old_rev in head_set:
304
heads.append(inv[ie.file_id].revision)
305
head_set.remove(inv[ie.file_id].revision)
308
# now we check to see if we need to write a new record to the
310
# We write a new entry unless there is one head to the ancestors, and
311
# the kind-derived content is unchanged.
313
# Cheapest check first: no ancestors, or more the one head in the
314
# ancestors, we write a new node.
318
# There is a single head, look it up for comparison
319
parent_entry = parent_candiate_entries[heads[0]]
320
# if the non-content specific data has changed, we'll be writing a
322
if (parent_entry.parent_id != ie.parent_id or
323
parent_entry.name != ie.name):
325
# now we need to do content specific checks:
327
# if the kind changed the content obviously has
328
if kind != parent_entry.kind:
331
assert content_summary[2] is not None, \
332
"Files must not have executable = None"
334
if (# if the file length changed we have to store:
335
parent_entry.text_size != content_summary[1] or
336
# if the exec bit has changed we have to store:
337
parent_entry.executable != content_summary[2]):
339
elif parent_entry.text_sha1 == content_summary[3]:
340
# all meta and content is unchanged (using a hash cache
341
# hit to check the sha)
342
ie.revision = parent_entry.revision
343
ie.text_size = parent_entry.text_size
344
ie.text_sha1 = parent_entry.text_sha1
345
ie.executable = parent_entry.executable
346
return self._get_delta(ie, basis_inv, path), False
348
# Either there is only a hash change(no hash cache entry,
349
# or same size content change), or there is no change on
351
# Provide the parent's hash to the store layer, so that the
352
# content is unchanged we will not store a new node.
353
nostore_sha = parent_entry.text_sha1
355
# We want to record a new node regardless of the presence or
356
# absence of a content change in the file.
358
ie.executable = content_summary[2]
359
lines = tree.get_file(ie.file_id, path).readlines()
361
ie.text_sha1, ie.text_size = self._add_text_to_weave(
362
ie.file_id, lines, heads, nostore_sha)
363
except errors.ExistingContent:
364
# Turns out that the file content was unchanged, and we were
365
# only going to store a new node if it was changed. Carry over
367
ie.revision = parent_entry.revision
368
ie.text_size = parent_entry.text_size
369
ie.text_sha1 = parent_entry.text_sha1
370
ie.executable = parent_entry.executable
371
return self._get_delta(ie, basis_inv, path), False
372
elif kind == 'directory':
374
# all data is meta here, nothing specific to directory, so
376
ie.revision = parent_entry.revision
377
return self._get_delta(ie, basis_inv, path), False
379
self._add_text_to_weave(ie.file_id, lines, heads, None)
380
elif kind == 'symlink':
381
current_link_target = content_summary[3]
383
# symlink target is not generic metadata, check if it has
385
if current_link_target != parent_entry.symlink_target:
388
# unchanged, carry over.
389
ie.revision = parent_entry.revision
390
ie.symlink_target = parent_entry.symlink_target
391
return self._get_delta(ie, basis_inv, path), False
392
ie.symlink_target = current_link_target
394
self._add_text_to_weave(ie.file_id, lines, heads, None)
395
elif kind == 'tree-reference':
397
if content_summary[3] != parent_entry.reference_revision:
400
# unchanged, carry over.
401
ie.reference_revision = parent_entry.reference_revision
402
ie.revision = parent_entry.revision
403
return self._get_delta(ie, basis_inv, path), False
404
ie.reference_revision = content_summary[3]
406
self._add_text_to_weave(ie.file_id, lines, heads, None)
408
raise NotImplementedError('unknown kind')
409
ie.revision = self._new_revision_id
410
return self._get_delta(ie, basis_inv, path), True
412
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
413
versionedfile = self.repository.weave_store.get_weave_or_empty(
414
file_id, self.repository.get_transaction())
415
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
416
# than add_lines, and allows committing when a parent is ghosted for
418
# Note: as we read the content directly from the tree, we know its not
419
# been turned into unicode or badly split - but a broken tree
420
# implementation could give us bad output from readlines() so this is
421
# not a guarantee of safety. What would be better is always checking
422
# the content during test suite execution. RBC 20070912
424
return versionedfile.add_lines_with_ghosts(
425
self._new_revision_id, parents, new_lines,
426
nostore_sha=nostore_sha, random_id=self.random_revid,
427
check_content=False)[0:2]
429
versionedfile.clear_cache()
432
class RootCommitBuilder(CommitBuilder):
433
"""This commitbuilder actually records the root id"""
435
# the root entry gets versioned properly by this builder.
436
_versioned_root = True
438
def _check_root(self, ie, parent_invs, tree):
439
"""Helper for record_entry_contents.
441
:param ie: An entry being added.
442
:param parent_invs: The inventories of the parent revisions of the
444
:param tree: The tree that is being committed.
448
######################################################################
451
class Repository(object):
452
"""Repository holding history for one or more branches.
454
The repository holds and retrieves historical information including
455
revisions and file history. It's normally accessed only by the Branch,
456
which views a particular line of development through that history.
458
The Repository builds on top of Stores and a Transport, which respectively
459
describe the disk data format and the way of accessing the (possibly
463
# What class to use for a CommitBuilder. Often its simpler to change this
464
# in a Repository class subclass rather than to override
465
# get_commit_builder.
466
_commit_builder_class = CommitBuilder
467
# The search regex used by xml based repositories to determine what things
468
# where changed in a single commit.
469
_file_ids_altered_regex = lazy_regex.lazy_compile(
470
r'file_id="(?P<file_id>[^"]+)"'
471
r'.* revision="(?P<revision_id>[^"]+)"'
474
def abort_write_group(self):
475
"""Commit the contents accrued within the current write group.
477
:seealso: start_write_group.
479
if self._write_group is not self.get_transaction():
480
# has an unlock or relock occured ?
481
raise errors.BzrError('mismatched lock context and write group.')
482
self._abort_write_group()
483
self._write_group = None
485
def _abort_write_group(self):
486
"""Template method for per-repository write group cleanup.
488
This is called during abort before the write group is considered to be
489
finished and should cleanup any internal state accrued during the write
490
group. There is no requirement that data handed to the repository be
491
*not* made available - this is not a rollback - but neither should any
492
attempt be made to ensure that data added is fully commited. Abort is
493
invoked when an error has occured so futher disk or network operations
494
may not be possible or may error and if possible should not be
499
def add_inventory(self, revision_id, inv, parents):
500
"""Add the inventory inv to the repository as revision_id.
502
:param parents: The revision ids of the parents that revision_id
503
is known to have and are in the repository already.
505
returns the sha1 of the serialized inventory.
507
assert self.is_in_write_group()
508
_mod_revision.check_not_reserved_id(revision_id)
509
assert inv.revision_id is None or inv.revision_id == revision_id, \
510
"Mismatch between inventory revision" \
511
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
512
assert inv.root is not None
513
inv_lines = self._serialise_inventory_to_lines(inv)
514
inv_vf = self.get_inventory_weave()
515
return self._inventory_add_lines(inv_vf, revision_id, parents,
516
inv_lines, check_content=False)
518
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
520
"""Store lines in inv_vf and return the sha1 of the inventory."""
522
for parent in parents:
524
final_parents.append(parent)
525
return inv_vf.add_lines(revision_id, final_parents, lines,
526
check_content=check_content)[0]
529
def add_revision(self, revision_id, rev, inv=None, config=None):
530
"""Add rev to the revision store as revision_id.
532
:param revision_id: the revision id to use.
533
:param rev: The revision object.
534
:param inv: The inventory for the revision. if None, it will be looked
535
up in the inventory storer
536
:param config: If None no digital signature will be created.
537
If supplied its signature_needed method will be used
538
to determine if a signature should be made.
540
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
542
_mod_revision.check_not_reserved_id(revision_id)
543
if config is not None and config.signature_needed():
545
inv = self.get_inventory(revision_id)
546
plaintext = Testament(rev, inv).as_short_text()
547
self.store_revision_signature(
548
gpg.GPGStrategy(config), plaintext, revision_id)
549
if not revision_id in self.get_inventory_weave():
551
raise errors.WeaveRevisionNotPresent(revision_id,
552
self.get_inventory_weave())
554
# yes, this is not suitable for adding with ghosts.
555
self.add_inventory(revision_id, inv, rev.parent_ids)
556
self._revision_store.add_revision(rev, self.get_transaction())
558
def _add_revision_text(self, revision_id, text):
559
revision = self._revision_store._serializer.read_revision_from_string(
561
self._revision_store._add_revision(revision, StringIO(text),
562
self.get_transaction())
564
def all_revision_ids(self):
565
"""Returns a list of all the revision ids in the repository.
567
This is deprecated because code should generally work on the graph
568
reachable from a particular revision, and ignore any other revisions
569
that might be present. There is no direct replacement method.
571
if 'evil' in debug.debug_flags:
572
mutter_callsite(2, "all_revision_ids is linear with history.")
573
return self._all_revision_ids()
575
def _all_revision_ids(self):
576
"""Returns a list of all the revision ids in the repository.
578
These are in as much topological order as the underlying store can
581
raise NotImplementedError(self._all_revision_ids)
583
def break_lock(self):
584
"""Break a lock if one is present from another instance.
586
Uses the ui factory to ask for confirmation if the lock may be from
589
self.control_files.break_lock()
592
def _eliminate_revisions_not_present(self, revision_ids):
593
"""Check every revision id in revision_ids to see if we have it.
595
Returns a set of the present revisions.
598
for id in revision_ids:
599
if self.has_revision(id):
604
def create(a_bzrdir):
605
"""Construct the current default format repository in a_bzrdir."""
606
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
608
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
609
"""instantiate a Repository.
611
:param _format: The format of the repository on disk.
612
:param a_bzrdir: The BzrDir of the repository.
614
In the future we will have a single api for all stores for
615
getting file texts, inventories and revisions, then
616
this construct will accept instances of those things.
618
super(Repository, self).__init__()
619
self._format = _format
620
# the following are part of the public API for Repository:
621
self.bzrdir = a_bzrdir
622
self.control_files = control_files
623
self._revision_store = _revision_store
624
# backwards compatibility
625
self.weave_store = text_store
627
self._reconcile_does_inventory_gc = True
628
self._reconcile_fixes_text_parents = False
629
self._reconcile_backsup_inventory = True
630
# not right yet - should be more semantically clear ?
632
self.control_store = control_store
633
self.control_weaves = control_store
634
# TODO: make sure to construct the right store classes, etc, depending
635
# on whether escaping is required.
636
self._warn_if_deprecated()
637
self._write_group = None
638
self.base = control_files._transport.base
641
return '%s(%r)' % (self.__class__.__name__,
644
def has_same_location(self, other):
645
"""Returns a boolean indicating if this repository is at the same
646
location as another repository.
648
This might return False even when two repository objects are accessing
649
the same physical repository via different URLs.
651
if self.__class__ is not other.__class__:
653
return (self.control_files._transport.base ==
654
other.control_files._transport.base)
656
def is_in_write_group(self):
657
"""Return True if there is an open write group.
659
:seealso: start_write_group.
661
return self._write_group is not None
664
return self.control_files.is_locked()
666
def is_write_locked(self):
667
"""Return True if this object is write locked."""
668
return self.is_locked() and self.control_files._lock_mode == 'w'
670
def lock_write(self, token=None):
671
"""Lock this repository for writing.
673
This causes caching within the repository obejct to start accumlating
674
data during reads, and allows a 'write_group' to be obtained. Write
675
groups must be used for actual data insertion.
677
:param token: if this is already locked, then lock_write will fail
678
unless the token matches the existing lock.
679
:returns: a token if this instance supports tokens, otherwise None.
680
:raises TokenLockingNotSupported: when a token is given but this
681
instance doesn't support using token locks.
682
:raises MismatchedToken: if the specified token doesn't match the token
683
of the existing lock.
684
:seealso: start_write_group.
686
A token should be passed in if you know that you have locked the object
687
some other way, and need to synchronise this object's state with that
690
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
692
result = self.control_files.lock_write(token=token)
697
self.control_files.lock_read()
700
def get_physical_lock_status(self):
701
return self.control_files.get_physical_lock_status()
703
def leave_lock_in_place(self):
704
"""Tell this repository not to release the physical lock when this
707
If lock_write doesn't return a token, then this method is not supported.
709
self.control_files.leave_in_place()
711
def dont_leave_lock_in_place(self):
712
"""Tell this repository to release the physical lock when this
713
object is unlocked, even if it didn't originally acquire it.
715
If lock_write doesn't return a token, then this method is not supported.
717
self.control_files.dont_leave_in_place()
720
def gather_stats(self, revid=None, committers=None):
721
"""Gather statistics from a revision id.
723
:param revid: The revision id to gather statistics from, if None, then
724
no revision specific statistics are gathered.
725
:param committers: Optional parameter controlling whether to grab
726
a count of committers from the revision specific statistics.
727
:return: A dictionary of statistics. Currently this contains:
728
committers: The number of committers if requested.
729
firstrev: A tuple with timestamp, timezone for the penultimate left
730
most ancestor of revid, if revid is not the NULL_REVISION.
731
latestrev: A tuple with timestamp, timezone for revid, if revid is
732
not the NULL_REVISION.
733
revisions: The total revision count in the repository.
734
size: An estimate disk size of the repository in bytes.
737
if revid and committers:
738
result['committers'] = 0
739
if revid and revid != _mod_revision.NULL_REVISION:
741
all_committers = set()
742
revisions = self.get_ancestry(revid)
743
# pop the leading None
745
first_revision = None
747
# ignore the revisions in the middle - just grab first and last
748
revisions = revisions[0], revisions[-1]
749
for revision in self.get_revisions(revisions):
750
if not first_revision:
751
first_revision = revision
753
all_committers.add(revision.committer)
754
last_revision = revision
756
result['committers'] = len(all_committers)
757
result['firstrev'] = (first_revision.timestamp,
758
first_revision.timezone)
759
result['latestrev'] = (last_revision.timestamp,
760
last_revision.timezone)
762
# now gather global repository information
763
if self.bzrdir.root_transport.listable():
764
c, t = self._revision_store.total_size(self.get_transaction())
765
result['revisions'] = c
769
def get_data_stream(self, revision_ids):
770
raise NotImplementedError(self.get_data_stream)
772
def insert_data_stream(self, stream):
773
"""XXX What does this really do?
775
Is it a substitute for fetch?
776
Should it manage its own write group ?
778
for item_key, bytes in stream:
779
if item_key[0] == 'file':
780
(file_id,) = item_key[1:]
781
knit = self.weave_store.get_weave_or_empty(
782
file_id, self.get_transaction())
783
elif item_key == ('inventory',):
784
knit = self.get_inventory_weave()
785
elif item_key == ('revisions',):
786
knit = self._revision_store.get_revision_file(
787
self.get_transaction())
788
elif item_key == ('signatures',):
789
knit = self._revision_store.get_signature_file(
790
self.get_transaction())
792
raise RepositoryDataStreamError(
793
"Unrecognised data stream key '%s'" % (item_key,))
794
decoded_list = bencode.bdecode(bytes)
795
format = decoded_list.pop(0)
798
for version, options, parents, some_bytes in decoded_list:
799
data_list.append((version, options, len(some_bytes), parents))
800
knit_bytes += some_bytes
801
knit.insert_data_stream(
802
(format, data_list, StringIO(knit_bytes).read))
805
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
806
"""Return the revision ids that other has that this does not.
808
These are returned in topological order.
810
revision_id: only return revision ids included by revision_id.
812
return InterRepository.get(other, self).missing_revision_ids(
813
revision_id, find_ghosts)
817
"""Open the repository rooted at base.
819
For instance, if the repository is at URL/.bzr/repository,
820
Repository.open(URL) -> a Repository instance.
822
control = bzrdir.BzrDir.open(base)
823
return control.open_repository()
825
def copy_content_into(self, destination, revision_id=None):
826
"""Make a complete copy of the content in self into destination.
828
This is a destructive operation! Do not use it on existing
831
return InterRepository.get(self, destination).copy_content(revision_id)
833
def commit_write_group(self):
834
"""Commit the contents accrued within the current write group.
836
:seealso: start_write_group.
838
if self._write_group is not self.get_transaction():
839
# has an unlock or relock occured ?
840
raise errors.BzrError('mismatched lock context %r and '
842
(self.get_transaction(), self._write_group))
843
self._commit_write_group()
844
self._write_group = None
846
def _commit_write_group(self):
847
"""Template method for per-repository write group cleanup.
849
This is called before the write group is considered to be
850
finished and should ensure that all data handed to the repository
851
for writing during the write group is safely committed (to the
852
extent possible considering file system caching etc).
855
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
856
"""Fetch the content required to construct revision_id from source.
858
If revision_id is None all content is copied.
859
:param find_ghosts: Find and copy revisions in the source that are
860
ghosts in the target (and not reachable directly by walking out to
861
the first-present revision in target from revision_id).
863
# fast path same-url fetch operations
864
if self.has_same_location(source):
865
# check that last_revision is in 'from' and then return a
867
if (revision_id is not None and
868
not _mod_revision.is_null(revision_id)):
869
self.get_revision(revision_id)
871
inter = InterRepository.get(source, self)
873
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
874
except NotImplementedError:
875
raise errors.IncompatibleRepositories(source, self)
877
def create_bundle(self, target, base, fileobj, format=None):
878
return serializer.write_bundle(self, target, base, fileobj, format)
880
def get_commit_builder(self, branch, parents, config, timestamp=None,
881
timezone=None, committer=None, revprops=None,
883
"""Obtain a CommitBuilder for this repository.
885
:param branch: Branch to commit to.
886
:param parents: Revision ids of the parents of the new revision.
887
:param config: Configuration to use.
888
:param timestamp: Optional timestamp recorded for commit.
889
:param timezone: Optional timezone for timestamp.
890
:param committer: Optional committer to set for commit.
891
:param revprops: Optional dictionary of revision properties.
892
:param revision_id: Optional revision id.
894
result = self._commit_builder_class(self, parents, config,
895
timestamp, timezone, committer, revprops, revision_id)
896
self.start_write_group()
900
if (self.control_files._lock_count == 1 and
901
self.control_files._lock_mode == 'w'):
902
if self._write_group is not None:
903
self.abort_write_group()
904
self.control_files.unlock()
905
raise errors.BzrError(
906
'Must end write groups before releasing write locks.')
907
self.control_files.unlock()
910
def clone(self, a_bzrdir, revision_id=None):
911
"""Clone this repository into a_bzrdir using the current format.
913
Currently no check is made that the format of this repository and
914
the bzrdir format are compatible. FIXME RBC 20060201.
916
:return: The newly created destination repository.
918
# TODO: deprecate after 0.16; cloning this with all its settings is
919
# probably not very useful -- mbp 20070423
920
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
921
self.copy_content_into(dest_repo, revision_id)
924
def start_write_group(self):
925
"""Start a write group in the repository.
927
Write groups are used by repositories which do not have a 1:1 mapping
928
between file ids and backend store to manage the insertion of data from
929
both fetch and commit operations.
931
A write lock is required around the start_write_group/commit_write_group
932
for the support of lock-requiring repository formats.
934
One can only insert data into a repository inside a write group.
938
if not self.is_write_locked():
939
raise errors.NotWriteLocked(self)
940
if self._write_group:
941
raise errors.BzrError('already in a write group')
942
self._start_write_group()
943
# so we can detect unlock/relock - the write group is now entered.
944
self._write_group = self.get_transaction()
946
def _start_write_group(self):
947
"""Template method for per-repository write group startup.
949
This is called before the write group is considered to be
954
def sprout(self, to_bzrdir, revision_id=None):
955
"""Create a descendent repository for new development.
957
Unlike clone, this does not copy the settings of the repository.
959
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
960
dest_repo.fetch(self, revision_id=revision_id)
963
def _create_sprouting_repo(self, a_bzrdir, shared):
964
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
965
# use target default format.
966
dest_repo = a_bzrdir.create_repository()
968
# Most control formats need the repository to be specifically
969
# created, but on some old all-in-one formats it's not needed
971
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
972
except errors.UninitializableFormat:
973
dest_repo = a_bzrdir.open_repository()
977
def has_revision(self, revision_id):
978
"""True if this repository has a copy of the revision."""
979
if 'evil' in debug.debug_flags:
980
mutter_callsite(3, "has_revision is a LBYL symptom.")
981
return self._revision_store.has_revision_id(revision_id,
982
self.get_transaction())
985
def get_revision(self, revision_id):
986
"""Return the Revision object for a named revision."""
987
return self.get_revisions([revision_id])[0]
990
def get_revision_reconcile(self, revision_id):
991
"""'reconcile' helper routine that allows access to a revision always.
993
This variant of get_revision does not cross check the weave graph
994
against the revision one as get_revision does: but it should only
995
be used by reconcile, or reconcile-alike commands that are correcting
996
or testing the revision graph.
998
return self._get_revisions([revision_id])[0]
1001
def get_revisions(self, revision_ids):
1002
"""Get many revisions at once."""
1003
return self._get_revisions(revision_ids)
1006
def _get_revisions(self, revision_ids):
1007
"""Core work logic to get many revisions without sanity checks."""
1008
for rev_id in revision_ids:
1009
if not rev_id or not isinstance(rev_id, basestring):
1010
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1011
revs = self._revision_store.get_revisions(revision_ids,
1012
self.get_transaction())
1014
assert not isinstance(rev.revision_id, unicode)
1015
for parent_id in rev.parent_ids:
1016
assert not isinstance(parent_id, unicode)
1020
def get_revision_xml(self, revision_id):
1021
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1022
# would have already do it.
1023
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1024
rev = self.get_revision(revision_id)
1025
rev_tmp = StringIO()
1026
# the current serializer..
1027
self._revision_store._serializer.write_revision(rev, rev_tmp)
1029
return rev_tmp.getvalue()
1032
def get_deltas_for_revisions(self, revisions):
1033
"""Produce a generator of revision deltas.
1035
Note that the input is a sequence of REVISIONS, not revision_ids.
1036
Trees will be held in memory until the generator exits.
1037
Each delta is relative to the revision's lefthand predecessor.
1039
required_trees = set()
1040
for revision in revisions:
1041
required_trees.add(revision.revision_id)
1042
required_trees.update(revision.parent_ids[:1])
1043
trees = dict((t.get_revision_id(), t) for
1044
t in self.revision_trees(required_trees))
1045
for revision in revisions:
1046
if not revision.parent_ids:
1047
old_tree = self.revision_tree(None)
1049
old_tree = trees[revision.parent_ids[0]]
1050
yield trees[revision.revision_id].changes_from(old_tree)
1053
def get_revision_delta(self, revision_id):
1054
"""Return the delta for one revision.
1056
The delta is relative to the left-hand predecessor of the
1059
r = self.get_revision(revision_id)
1060
return list(self.get_deltas_for_revisions([r]))[0]
1063
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1064
signature = gpg_strategy.sign(plaintext)
1065
self.add_signature_text(revision_id, signature)
1068
def add_signature_text(self, revision_id, signature):
1069
self._revision_store.add_revision_signature_text(revision_id,
1071
self.get_transaction())
1073
def find_text_key_references(self):
1074
"""Find the text key references within the repository.
1076
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1077
revision_ids. Each altered file-ids has the exact revision_ids that
1078
altered it listed explicitly.
1079
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1080
to whether they were referred to by the inventory of the
1081
revision_id that they contain. The inventory texts from all present
1082
revision ids are assessed to generate this report.
1084
revision_ids = self.all_revision_ids()
1085
w = self.get_inventory_weave()
1086
pb = ui.ui_factory.nested_progress_bar()
1088
return self._find_text_key_references_from_xml_inventory_lines(
1089
w.iter_lines_added_or_present_in_versions(revision_ids, pb=pb))
1093
def _find_text_key_references_from_xml_inventory_lines(self,
1095
"""Core routine for extracting references to texts from inventories.
1097
This performs the translation of xml lines to revision ids.
1099
:param line_iterator: An iterator of lines, origin_version_id
1100
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1101
to whether they were referred to by the inventory of the
1102
revision_id that they contain. Note that if that revision_id was
1103
not part of the line_iterator's output then False will be given -
1104
even though it may actually refer to that key.
1106
if not self._serializer.support_altered_by_hack:
1107
raise AssertionError(
1108
"_find_text_key_references_from_xml_inventory_lines only "
1109
"supported for branches which store inventory as unnested xml"
1110
", not on %r" % self)
1113
# this code needs to read every new line in every inventory for the
1114
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1115
# not present in one of those inventories is unnecessary but not
1116
# harmful because we are filtering by the revision id marker in the
1117
# inventory lines : we only select file ids altered in one of those
1118
# revisions. We don't need to see all lines in the inventory because
1119
# only those added in an inventory in rev X can contain a revision=X
1121
unescape_revid_cache = {}
1122
unescape_fileid_cache = {}
1124
# jam 20061218 In a big fetch, this handles hundreds of thousands
1125
# of lines, so it has had a lot of inlining and optimizing done.
1126
# Sorry that it is a little bit messy.
1127
# Move several functions to be local variables, since this is a long
1129
search = self._file_ids_altered_regex.search
1130
unescape = _unescape_xml
1131
setdefault = result.setdefault
1132
for line, version_id in line_iterator:
1133
match = search(line)
1136
# One call to match.group() returning multiple items is quite a
1137
# bit faster than 2 calls to match.group() each returning 1
1138
file_id, revision_id = match.group('file_id', 'revision_id')
1140
# Inlining the cache lookups helps a lot when you make 170,000
1141
# lines and 350k ids, versus 8.4 unique ids.
1142
# Using a cache helps in 2 ways:
1143
# 1) Avoids unnecessary decoding calls
1144
# 2) Re-uses cached strings, which helps in future set and
1146
# (2) is enough that removing encoding entirely along with
1147
# the cache (so we are using plain strings) results in no
1148
# performance improvement.
1150
revision_id = unescape_revid_cache[revision_id]
1152
unescaped = unescape(revision_id)
1153
unescape_revid_cache[revision_id] = unescaped
1154
revision_id = unescaped
1156
# Note that unconditionally unescaping means that we deserialise
1157
# every fileid, which for general 'pull' is not great, but we don't
1158
# really want to have some many fulltexts that this matters anyway.
1161
file_id = unescape_fileid_cache[file_id]
1163
unescaped = unescape(file_id)
1164
unescape_fileid_cache[file_id] = unescaped
1167
key = (file_id, revision_id)
1168
setdefault(key, False)
1169
if revision_id == version_id:
1173
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1175
"""Helper routine for fileids_altered_by_revision_ids.
1177
This performs the translation of xml lines to revision ids.
1179
:param line_iterator: An iterator of lines, origin_version_id
1180
:param revision_ids: The revision ids to filter for. This should be a
1181
set or other type which supports efficient __contains__ lookups, as
1182
the revision id from each parsed line will be looked up in the
1183
revision_ids filter.
1184
:return: a dictionary mapping altered file-ids to an iterable of
1185
revision_ids. Each altered file-ids has the exact revision_ids that
1186
altered it listed explicitly.
1189
setdefault = result.setdefault
1190
for file_id, revision_id in \
1191
self._find_text_key_references_from_xml_inventory_lines(
1192
line_iterator).iterkeys():
1193
# once data is all ensured-consistent; then this is
1194
# if revision_id == version_id
1195
if revision_id in revision_ids:
1196
setdefault(file_id, set()).add(revision_id)
1199
def fileids_altered_by_revision_ids(self, revision_ids):
1200
"""Find the file ids and versions affected by revisions.
1202
:param revisions: an iterable containing revision ids.
1203
:return: a dictionary mapping altered file-ids to an iterable of
1204
revision_ids. Each altered file-ids has the exact revision_ids that
1205
altered it listed explicitly.
1207
selected_revision_ids = set(revision_ids)
1208
w = self.get_inventory_weave()
1209
pb = ui.ui_factory.nested_progress_bar()
1211
return self._find_file_ids_from_xml_inventory_lines(
1212
w.iter_lines_added_or_present_in_versions(
1213
selected_revision_ids, pb=pb),
1214
selected_revision_ids)
1218
def iter_files_bytes(self, desired_files):
1219
"""Iterate through file versions.
1221
Files will not necessarily be returned in the order they occur in
1222
desired_files. No specific order is guaranteed.
1224
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1225
value supplied by the caller as part of desired_files. It should
1226
uniquely identify the file version in the caller's context. (Examples:
1227
an index number or a TreeTransform trans_id.)
1229
bytes_iterator is an iterable of bytestrings for the file. The
1230
kind of iterable and length of the bytestrings are unspecified, but for
1231
this implementation, it is a list of lines produced by
1232
VersionedFile.get_lines().
1234
:param desired_files: a list of (file_id, revision_id, identifier)
1237
transaction = self.get_transaction()
1238
for file_id, revision_id, callable_data in desired_files:
1240
weave = self.weave_store.get_weave(file_id, transaction)
1241
except errors.NoSuchFile:
1242
raise errors.NoSuchIdInRepository(self, file_id)
1243
yield callable_data, weave.get_lines(revision_id)
1245
def _generate_text_key_index(self, text_key_references=None,
1247
"""Generate a new text key index for the repository.
1249
This is an expensive function that will take considerable time to run.
1251
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1252
list of parents, also text keys. When a given key has no parents,
1253
the parents list will be [NULL_REVISION].
1255
# All revisions, to find inventory parents.
1256
if ancestors is None:
1257
revision_graph = self.get_revision_graph_with_ghosts()
1258
ancestors = revision_graph.get_ancestors()
1259
if text_key_references is None:
1260
text_key_references = self.find_text_key_references()
1261
pb = ui.ui_factory.nested_progress_bar()
1263
return self._do_generate_text_key_index(ancestors,
1264
text_key_references, pb)
1268
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1269
"""Helper for _generate_text_key_index to avoid deep nesting."""
1270
revision_order = tsort.topo_sort(ancestors)
1271
invalid_keys = set()
1273
for revision_id in revision_order:
1274
revision_keys[revision_id] = set()
1275
text_count = len(text_key_references)
1276
# a cache of the text keys to allow reuse; costs a dict of all the
1277
# keys, but saves a 2-tuple for every child of a given key.
1279
for text_key, valid in text_key_references.iteritems():
1281
invalid_keys.add(text_key)
1283
revision_keys[text_key[1]].add(text_key)
1284
text_key_cache[text_key] = text_key
1285
del text_key_references
1287
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1288
NULL_REVISION = _mod_revision.NULL_REVISION
1289
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1290
# too small for large or very branchy trees. However, for 55K path
1291
# trees, it would be easy to use too much memory trivially. Ideally we
1292
# could gauge this by looking at available real memory etc, but this is
1293
# always a tricky proposition.
1294
inventory_cache = lru_cache.LRUCache(10)
1295
batch_size = 10 # should be ~150MB on a 55K path tree
1296
batch_count = len(revision_order) / batch_size + 1
1298
pb.update("Calculating text parents.", processed_texts, text_count)
1299
for offset in xrange(batch_count):
1300
to_query = revision_order[offset * batch_size:(offset + 1) *
1304
for rev_tree in self.revision_trees(to_query):
1305
revision_id = rev_tree.get_revision_id()
1306
parent_ids = ancestors[revision_id]
1307
for text_key in revision_keys[revision_id]:
1308
pb.update("Calculating text parents.", processed_texts)
1309
processed_texts += 1
1310
candidate_parents = []
1311
for parent_id in parent_ids:
1312
parent_text_key = (text_key[0], parent_id)
1314
check_parent = parent_text_key not in \
1315
revision_keys[parent_id]
1317
# the parent parent_id is a ghost:
1318
check_parent = False
1319
# truncate the derived graph against this ghost.
1320
parent_text_key = None
1322
# look at the parent commit details inventories to
1323
# determine possible candidates in the per file graph.
1326
inv = inventory_cache[parent_id]
1328
inv = self.revision_tree(parent_id).inventory
1329
inventory_cache[parent_id] = inv
1330
parent_entry = inv._byid.get(text_key[0], None)
1331
if parent_entry is not None:
1333
text_key[0], parent_entry.revision)
1335
parent_text_key = None
1336
if parent_text_key is not None:
1337
candidate_parents.append(
1338
text_key_cache[parent_text_key])
1339
parent_heads = text_graph.heads(candidate_parents)
1340
new_parents = list(parent_heads)
1341
new_parents.sort(key=lambda x:candidate_parents.index(x))
1342
if new_parents == []:
1343
new_parents = [NULL_REVISION]
1344
text_index[text_key] = new_parents
1346
for text_key in invalid_keys:
1347
text_index[text_key] = [NULL_REVISION]
1350
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1351
"""Get an iterable listing the keys of all the data introduced by a set
1354
The keys will be ordered so that the corresponding items can be safely
1355
fetched and inserted in that order.
1357
:returns: An iterable producing tuples of (knit-kind, file-id,
1358
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1359
'revisions'. file-id is None unless knit-kind is 'file'.
1361
# XXX: it's a bit weird to control the inventory weave caching in this
1362
# generator. Ideally the caching would be done in fetch.py I think. Or
1363
# maybe this generator should explicitly have the contract that it
1364
# should not be iterated until the previously yielded item has been
1367
inv_w = self.get_inventory_weave()
1368
inv_w.enable_cache()
1370
# file ids that changed
1371
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1373
num_file_ids = len(file_ids)
1374
for file_id, altered_versions in file_ids.iteritems():
1375
if _files_pb is not None:
1376
_files_pb.update("fetch texts", count, num_file_ids)
1378
yield ("file", file_id, altered_versions)
1379
# We're done with the files_pb. Note that it finished by the caller,
1380
# just as it was created by the caller.
1384
yield ("inventory", None, revision_ids)
1388
revisions_with_signatures = set()
1389
for rev_id in revision_ids:
1391
self.get_signature_text(rev_id)
1392
except errors.NoSuchRevision:
1396
revisions_with_signatures.add(rev_id)
1398
yield ("signatures", None, revisions_with_signatures)
1401
yield ("revisions", None, revision_ids)
1404
def get_inventory_weave(self):
1405
return self.control_weaves.get_weave('inventory',
1406
self.get_transaction())
1409
def get_inventory(self, revision_id):
1410
"""Get Inventory object by hash."""
1411
return self.deserialise_inventory(
1412
revision_id, self.get_inventory_xml(revision_id))
1414
def deserialise_inventory(self, revision_id, xml):
1415
"""Transform the xml into an inventory object.
1417
:param revision_id: The expected revision id of the inventory.
1418
:param xml: A serialised inventory.
1420
return self._serializer.read_inventory_from_string(xml, revision_id)
1422
def serialise_inventory(self, inv):
1423
return self._serializer.write_inventory_to_string(inv)
1425
def _serialise_inventory_to_lines(self, inv):
1426
return self._serializer.write_inventory_to_lines(inv)
1428
def get_serializer_format(self):
1429
return self._serializer.format_num
1432
def get_inventory_xml(self, revision_id):
1433
"""Get inventory XML as a file object."""
1435
assert isinstance(revision_id, str), type(revision_id)
1436
iw = self.get_inventory_weave()
1437
return iw.get_text(revision_id)
1439
raise errors.HistoryMissing(self, 'inventory', revision_id)
1442
def get_inventory_sha1(self, revision_id):
1443
"""Return the sha1 hash of the inventory entry
1445
return self.get_revision(revision_id).inventory_sha1
1448
def get_revision_graph(self, revision_id=None):
1449
"""Return a dictionary containing the revision graph.
1451
NB: This method should not be used as it accesses the entire graph all
1452
at once, which is much more data than most operations should require.
1454
:param revision_id: The revision_id to get a graph from. If None, then
1455
the entire revision graph is returned. This is a deprecated mode of
1456
operation and will be removed in the future.
1457
:return: a dictionary of revision_id->revision_parents_list.
1459
raise NotImplementedError(self.get_revision_graph)
1462
def get_revision_graph_with_ghosts(self, revision_ids=None):
1463
"""Return a graph of the revisions with ghosts marked as applicable.
1465
:param revision_ids: an iterable of revisions to graph or None for all.
1466
:return: a Graph object with the graph reachable from revision_ids.
1468
if 'evil' in debug.debug_flags:
1470
"get_revision_graph_with_ghosts scales with size of history.")
1471
result = deprecated_graph.Graph()
1472
if not revision_ids:
1473
pending = set(self.all_revision_ids())
1476
pending = set(revision_ids)
1477
# special case NULL_REVISION
1478
if _mod_revision.NULL_REVISION in pending:
1479
pending.remove(_mod_revision.NULL_REVISION)
1480
required = set(pending)
1483
revision_id = pending.pop()
1485
rev = self.get_revision(revision_id)
1486
except errors.NoSuchRevision:
1487
if revision_id in required:
1490
result.add_ghost(revision_id)
1492
for parent_id in rev.parent_ids:
1493
# is this queued or done ?
1494
if (parent_id not in pending and
1495
parent_id not in done):
1497
pending.add(parent_id)
1498
result.add_node(revision_id, rev.parent_ids)
1499
done.add(revision_id)
1502
def _get_history_vf(self):
1503
"""Get a versionedfile whose history graph reflects all revisions.
1505
For weave repositories, this is the inventory weave.
1507
return self.get_inventory_weave()
1509
def iter_reverse_revision_history(self, revision_id):
1510
"""Iterate backwards through revision ids in the lefthand history
1512
:param revision_id: The revision id to start with. All its lefthand
1513
ancestors will be traversed.
1515
if revision_id in (None, _mod_revision.NULL_REVISION):
1517
next_id = revision_id
1518
versionedfile = self._get_history_vf()
1521
parents = versionedfile.get_parents(next_id)
1522
if len(parents) == 0:
1525
next_id = parents[0]
1528
def get_revision_inventory(self, revision_id):
1529
"""Return inventory of a past revision."""
1530
# TODO: Unify this with get_inventory()
1531
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1532
# must be the same as its revision, so this is trivial.
1533
if revision_id is None:
1534
# This does not make sense: if there is no revision,
1535
# then it is the current tree inventory surely ?!
1536
# and thus get_root_id() is something that looks at the last
1537
# commit on the branch, and the get_root_id is an inventory check.
1538
raise NotImplementedError
1539
# return Inventory(self.get_root_id())
1541
return self.get_inventory(revision_id)
1544
def is_shared(self):
1545
"""Return True if this repository is flagged as a shared repository."""
1546
raise NotImplementedError(self.is_shared)
1549
def reconcile(self, other=None, thorough=False):
1550
"""Reconcile this repository."""
1551
from bzrlib.reconcile import RepoReconciler
1552
reconciler = RepoReconciler(self, thorough=thorough)
1553
reconciler.reconcile()
1556
def _refresh_data(self):
1557
"""Helper called from lock_* to ensure coherency with disk.
1559
The default implementation does nothing; it is however possible
1560
for repositories to maintain loaded indices across multiple locks
1561
by checking inside their implementation of this method to see
1562
whether their indices are still valid. This depends of course on
1563
the disk format being validatable in this manner.
1567
def revision_tree(self, revision_id):
1568
"""Return Tree for a revision on this branch.
1570
`revision_id` may be None for the empty tree revision.
1572
# TODO: refactor this to use an existing revision object
1573
# so we don't need to read it in twice.
1574
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1575
return RevisionTree(self, Inventory(root_id=None),
1576
_mod_revision.NULL_REVISION)
1578
inv = self.get_revision_inventory(revision_id)
1579
return RevisionTree(self, inv, revision_id)
1582
def revision_trees(self, revision_ids):
1583
"""Return Tree for a revision on this branch.
1585
`revision_id` may not be None or 'null:'"""
1586
assert None not in revision_ids
1587
assert _mod_revision.NULL_REVISION not in revision_ids
1588
texts = self.get_inventory_weave().get_texts(revision_ids)
1589
for text, revision_id in zip(texts, revision_ids):
1590
inv = self.deserialise_inventory(revision_id, text)
1591
yield RevisionTree(self, inv, revision_id)
1594
def get_ancestry(self, revision_id, topo_sorted=True):
1595
"""Return a list of revision-ids integrated by a revision.
1597
The first element of the list is always None, indicating the origin
1598
revision. This might change when we have history horizons, or
1599
perhaps we should have a new API.
1601
This is topologically sorted.
1603
if _mod_revision.is_null(revision_id):
1605
if not self.has_revision(revision_id):
1606
raise errors.NoSuchRevision(self, revision_id)
1607
w = self.get_inventory_weave()
1608
candidates = w.get_ancestry(revision_id, topo_sorted)
1609
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1612
"""Compress the data within the repository.
1614
This operation only makes sense for some repository types. For other
1615
types it should be a no-op that just returns.
1617
This stub method does not require a lock, but subclasses should use
1618
@needs_write_lock as this is a long running call its reasonable to
1619
implicitly lock for the user.
1623
def print_file(self, file, revision_id):
1624
"""Print `file` to stdout.
1626
FIXME RBC 20060125 as John Meinel points out this is a bad api
1627
- it writes to stdout, it assumes that that is valid etc. Fix
1628
by creating a new more flexible convenience function.
1630
tree = self.revision_tree(revision_id)
1631
# use inventory as it was in that revision
1632
file_id = tree.inventory.path2id(file)
1634
# TODO: jam 20060427 Write a test for this code path
1635
# it had a bug in it, and was raising the wrong
1637
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1638
tree.print_file(file_id)
1640
def get_transaction(self):
1641
return self.control_files.get_transaction()
1643
def revision_parents(self, revision_id):
1644
return self.get_inventory_weave().parent_names(revision_id)
1646
@deprecated_method(symbol_versioning.one_one)
1647
def get_parents(self, revision_ids):
1648
"""See StackedParentsProvider.get_parents"""
1649
parent_map = self.get_parent_map(revision_ids)
1650
return [parent_map.get(r, None) for r in revision_ids]
1652
def get_parent_map(self, keys):
1653
"""See graph._StackedParentsProvider.get_parent_map"""
1655
for revision_id in keys:
1656
if revision_id == _mod_revision.NULL_REVISION:
1657
parent_map[revision_id] = []
1660
parent_ids = self.get_revision(revision_id).parent_ids
1661
except errors.NoSuchRevision:
1664
if len(parent_ids) == 0:
1665
parent_ids = [_mod_revision.NULL_REVISION]
1666
parent_map[revision_id] = parent_ids
1669
def _make_parents_provider(self):
1672
def get_graph(self, other_repository=None):
1673
"""Return the graph walker for this repository format"""
1674
parents_provider = self._make_parents_provider()
1675
if (other_repository is not None and
1676
other_repository.bzrdir.transport.base !=
1677
self.bzrdir.transport.base):
1678
parents_provider = graph._StackedParentsProvider(
1679
[parents_provider, other_repository._make_parents_provider()])
1680
return graph.Graph(parents_provider)
1682
def _get_versioned_file_checker(self):
1683
"""Return an object suitable for checking versioned files."""
1684
return _VersionedFileChecker(self)
1687
def set_make_working_trees(self, new_value):
1688
"""Set the policy flag for making working trees when creating branches.
1690
This only applies to branches that use this repository.
1692
The default is 'True'.
1693
:param new_value: True to restore the default, False to disable making
1696
raise NotImplementedError(self.set_make_working_trees)
1698
def make_working_trees(self):
1699
"""Returns the policy for making working trees on new branches."""
1700
raise NotImplementedError(self.make_working_trees)
1703
def sign_revision(self, revision_id, gpg_strategy):
1704
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1705
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1708
def has_signature_for_revision_id(self, revision_id):
1709
"""Query for a revision signature for revision_id in the repository."""
1710
return self._revision_store.has_signature(revision_id,
1711
self.get_transaction())
1714
def get_signature_text(self, revision_id):
1715
"""Return the text for a signature."""
1716
return self._revision_store.get_signature_text(revision_id,
1717
self.get_transaction())
1720
def check(self, revision_ids=None):
1721
"""Check consistency of all history of given revision_ids.
1723
Different repository implementations should override _check().
1725
:param revision_ids: A non-empty list of revision_ids whose ancestry
1726
will be checked. Typically the last revision_id of a branch.
1728
return self._check(revision_ids)
1730
def _check(self, revision_ids):
1731
result = check.Check(self)
1735
def _warn_if_deprecated(self):
1736
global _deprecation_warning_done
1737
if _deprecation_warning_done:
1739
_deprecation_warning_done = True
1740
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1741
% (self._format, self.bzrdir.transport.base))
1743
def supports_rich_root(self):
1744
return self._format.rich_root_data
1746
def _check_ascii_revisionid(self, revision_id, method):
1747
"""Private helper for ascii-only repositories."""
1748
# weave repositories refuse to store revisionids that are non-ascii.
1749
if revision_id is not None:
1750
# weaves require ascii revision ids.
1751
if isinstance(revision_id, unicode):
1753
revision_id.encode('ascii')
1754
except UnicodeEncodeError:
1755
raise errors.NonAsciiRevisionId(method, self)
1758
revision_id.decode('ascii')
1759
except UnicodeDecodeError:
1760
raise errors.NonAsciiRevisionId(method, self)
1762
def revision_graph_can_have_wrong_parents(self):
1763
"""Is it possible for this repository to have a revision graph with
1766
If True, then this repository must also implement
1767
_find_inconsistent_revision_parents so that check and reconcile can
1768
check for inconsistencies before proceeding with other checks that may
1769
depend on the revision index being consistent.
1771
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1773
# remove these delegates a while after bzr 0.15
1774
def __make_delegated(name, from_module):
1775
def _deprecated_repository_forwarder():
1776
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1777
% (name, from_module),
1780
m = __import__(from_module, globals(), locals(), [name])
1782
return getattr(m, name)
1783
except AttributeError:
1784
raise AttributeError('module %s has no name %s'
1786
globals()[name] = _deprecated_repository_forwarder
1789
'AllInOneRepository',
1790
'WeaveMetaDirRepository',
1791
'PreSplitOutRepositoryFormat',
1792
'RepositoryFormat4',
1793
'RepositoryFormat5',
1794
'RepositoryFormat6',
1795
'RepositoryFormat7',
1797
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1801
'RepositoryFormatKnit',
1802
'RepositoryFormatKnit1',
1804
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1807
def install_revision(repository, rev, revision_tree):
1808
"""Install all revision data into a repository."""
1809
install_revisions(repository, [(rev, revision_tree, None)])
1812
def install_revisions(repository, iterable):
1813
"""Install all revision data into a repository.
1815
Accepts an iterable of revision, tree, signature tuples. The signature
1818
repository.start_write_group()
1820
for revision, revision_tree, signature in iterable:
1821
_install_revision(repository, revision, revision_tree, signature)
1823
repository.abort_write_group()
1826
repository.commit_write_group()
1829
def _install_revision(repository, rev, revision_tree, signature):
1830
"""Install all revision data into a repository."""
1831
present_parents = []
1833
for p_id in rev.parent_ids:
1834
if repository.has_revision(p_id):
1835
present_parents.append(p_id)
1836
parent_trees[p_id] = repository.revision_tree(p_id)
1838
parent_trees[p_id] = repository.revision_tree(None)
1840
inv = revision_tree.inventory
1841
entries = inv.iter_entries()
1842
# backwards compatibility hack: skip the root id.
1843
if not repository.supports_rich_root():
1844
path, root = entries.next()
1845
if root.revision != rev.revision_id:
1846
raise errors.IncompatibleRevision(repr(repository))
1847
# Add the texts that are not already present
1848
for path, ie in entries:
1849
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1850
repository.get_transaction())
1851
if ie.revision not in w:
1853
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1854
# with InventoryEntry.find_previous_heads(). if it is, then there
1855
# is a latent bug here where the parents may have ancestors of each
1857
for revision, tree in parent_trees.iteritems():
1858
if ie.file_id not in tree:
1860
parent_id = tree.inventory[ie.file_id].revision
1861
if parent_id in text_parents:
1863
text_parents.append(parent_id)
1865
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1866
repository.get_transaction())
1867
lines = revision_tree.get_file(ie.file_id).readlines()
1868
vfile.add_lines(rev.revision_id, text_parents, lines)
1870
# install the inventory
1871
repository.add_inventory(rev.revision_id, inv, present_parents)
1872
except errors.RevisionAlreadyPresent:
1874
if signature is not None:
1875
repository.add_signature_text(rev.revision_id, signature)
1876
repository.add_revision(rev.revision_id, rev, inv)
1879
class MetaDirRepository(Repository):
1880
"""Repositories in the new meta-dir layout."""
1882
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1883
super(MetaDirRepository, self).__init__(_format,
1889
dir_mode = self.control_files._dir_mode
1890
file_mode = self.control_files._file_mode
1893
def is_shared(self):
1894
"""Return True if this repository is flagged as a shared repository."""
1895
return self.control_files._transport.has('shared-storage')
1898
def set_make_working_trees(self, new_value):
1899
"""Set the policy flag for making working trees when creating branches.
1901
This only applies to branches that use this repository.
1903
The default is 'True'.
1904
:param new_value: True to restore the default, False to disable making
1909
self.control_files._transport.delete('no-working-trees')
1910
except errors.NoSuchFile:
1913
self.control_files.put_utf8('no-working-trees', '')
1915
def make_working_trees(self):
1916
"""Returns the policy for making working trees on new branches."""
1917
return not self.control_files._transport.has('no-working-trees')
1920
class RepositoryFormatRegistry(registry.Registry):
1921
"""Registry of RepositoryFormats."""
1923
def get(self, format_string):
1924
r = registry.Registry.get(self, format_string)
1930
format_registry = RepositoryFormatRegistry()
1931
"""Registry of formats, indexed by their identifying format string.
1933
This can contain either format instances themselves, or classes/factories that
1934
can be called to obtain one.
1938
#####################################################################
1939
# Repository Formats
1941
class RepositoryFormat(object):
1942
"""A repository format.
1944
Formats provide three things:
1945
* An initialization routine to construct repository data on disk.
1946
* a format string which is used when the BzrDir supports versioned
1948
* an open routine which returns a Repository instance.
1950
There is one and only one Format subclass for each on-disk format. But
1951
there can be one Repository subclass that is used for several different
1952
formats. The _format attribute on a Repository instance can be used to
1953
determine the disk format.
1955
Formats are placed in an dict by their format string for reference
1956
during opening. These should be subclasses of RepositoryFormat
1959
Once a format is deprecated, just deprecate the initialize and open
1960
methods on the format class. Do not deprecate the object, as the
1961
object will be created every system load.
1963
Common instance attributes:
1964
_matchingbzrdir - the bzrdir format that the repository format was
1965
originally written to work with. This can be used if manually
1966
constructing a bzrdir and repository, or more commonly for test suite
1970
# Set to True or False in derived classes. True indicates that the format
1971
# supports ghosts gracefully.
1972
supports_ghosts = None
1975
return "<%s>" % self.__class__.__name__
1977
def __eq__(self, other):
1978
# format objects are generally stateless
1979
return isinstance(other, self.__class__)
1981
def __ne__(self, other):
1982
return not self == other
1985
def find_format(klass, a_bzrdir):
1986
"""Return the format for the repository object in a_bzrdir.
1988
This is used by bzr native formats that have a "format" file in
1989
the repository. Other methods may be used by different types of
1993
transport = a_bzrdir.get_repository_transport(None)
1994
format_string = transport.get("format").read()
1995
return format_registry.get(format_string)
1996
except errors.NoSuchFile:
1997
raise errors.NoRepositoryPresent(a_bzrdir)
1999
raise errors.UnknownFormatError(format=format_string)
2002
def register_format(klass, format):
2003
format_registry.register(format.get_format_string(), format)
2006
def unregister_format(klass, format):
2007
format_registry.remove(format.get_format_string())
2010
def get_default_format(klass):
2011
"""Return the current default format."""
2012
from bzrlib import bzrdir
2013
return bzrdir.format_registry.make_bzrdir('default').repository_format
2015
def _get_control_store(self, repo_transport, control_files):
2016
"""Return the control store for this repository."""
2017
raise NotImplementedError(self._get_control_store)
2019
def get_format_string(self):
2020
"""Return the ASCII format string that identifies this format.
2022
Note that in pre format ?? repositories the format string is
2023
not permitted nor written to disk.
2025
raise NotImplementedError(self.get_format_string)
2027
def get_format_description(self):
2028
"""Return the short description for this format."""
2029
raise NotImplementedError(self.get_format_description)
2031
def _get_revision_store(self, repo_transport, control_files):
2032
"""Return the revision store object for this a_bzrdir."""
2033
raise NotImplementedError(self._get_revision_store)
2035
def _get_text_rev_store(self,
2042
"""Common logic for getting a revision store for a repository.
2044
see self._get_revision_store for the subclass-overridable method to
2045
get the store for a repository.
2047
from bzrlib.store.revision.text import TextRevisionStore
2048
dir_mode = control_files._dir_mode
2049
file_mode = control_files._file_mode
2050
text_store = TextStore(transport.clone(name),
2052
compressed=compressed,
2054
file_mode=file_mode)
2055
_revision_store = TextRevisionStore(text_store, serializer)
2056
return _revision_store
2058
# TODO: this shouldn't be in the base class, it's specific to things that
2059
# use weaves or knits -- mbp 20070207
2060
def _get_versioned_file_store(self,
2065
versionedfile_class=None,
2066
versionedfile_kwargs={},
2068
if versionedfile_class is None:
2069
versionedfile_class = self._versionedfile_class
2070
weave_transport = control_files._transport.clone(name)
2071
dir_mode = control_files._dir_mode
2072
file_mode = control_files._file_mode
2073
return VersionedFileStore(weave_transport, prefixed=prefixed,
2075
file_mode=file_mode,
2076
versionedfile_class=versionedfile_class,
2077
versionedfile_kwargs=versionedfile_kwargs,
2080
def initialize(self, a_bzrdir, shared=False):
2081
"""Initialize a repository of this format in a_bzrdir.
2083
:param a_bzrdir: The bzrdir to put the new repository in it.
2084
:param shared: The repository should be initialized as a sharable one.
2085
:returns: The new repository object.
2087
This may raise UninitializableFormat if shared repository are not
2088
compatible the a_bzrdir.
2090
raise NotImplementedError(self.initialize)
2092
def is_supported(self):
2093
"""Is this format supported?
2095
Supported formats must be initializable and openable.
2096
Unsupported formats may not support initialization or committing or
2097
some other features depending on the reason for not being supported.
2101
def check_conversion_target(self, target_format):
2102
raise NotImplementedError(self.check_conversion_target)
2104
def open(self, a_bzrdir, _found=False):
2105
"""Return an instance of this format for the bzrdir a_bzrdir.
2107
_found is a private parameter, do not use it.
2109
raise NotImplementedError(self.open)
2112
class MetaDirRepositoryFormat(RepositoryFormat):
2113
"""Common base class for the new repositories using the metadir layout."""
2115
rich_root_data = False
2116
supports_tree_reference = False
2117
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
2120
super(MetaDirRepositoryFormat, self).__init__()
2122
def _create_control_files(self, a_bzrdir):
2123
"""Create the required files and the initial control_files object."""
2124
# FIXME: RBC 20060125 don't peek under the covers
2125
# NB: no need to escape relative paths that are url safe.
2126
repository_transport = a_bzrdir.get_repository_transport(self)
2127
control_files = lockable_files.LockableFiles(repository_transport,
2128
'lock', lockdir.LockDir)
2129
control_files.create_lock()
2130
return control_files
2132
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2133
"""Upload the initial blank content."""
2134
control_files = self._create_control_files(a_bzrdir)
2135
control_files.lock_write()
2137
control_files._transport.mkdir_multi(dirs,
2138
mode=control_files._dir_mode)
2139
for file, content in files:
2140
control_files.put(file, content)
2141
for file, content in utf8_files:
2142
control_files.put_utf8(file, content)
2144
control_files.put_utf8('shared-storage', '')
2146
control_files.unlock()
2149
# formats which have no format string are not discoverable
2150
# and not independently creatable, so are not registered. They're
2151
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2152
# needed, it's constructed directly by the BzrDir. Non-native formats where
2153
# the repository is not separately opened are similar.
2155
format_registry.register_lazy(
2156
'Bazaar-NG Repository format 7',
2157
'bzrlib.repofmt.weaverepo',
2161
format_registry.register_lazy(
2162
'Bazaar-NG Knit Repository Format 1',
2163
'bzrlib.repofmt.knitrepo',
2164
'RepositoryFormatKnit1',
2167
format_registry.register_lazy(
2168
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2169
'bzrlib.repofmt.knitrepo',
2170
'RepositoryFormatKnit3',
2173
format_registry.register_lazy(
2174
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2175
'bzrlib.repofmt.knitrepo',
2176
'RepositoryFormatKnit4',
2179
# Pack-based formats. There is one format for pre-subtrees, and one for
2180
# post-subtrees to allow ease of testing.
2181
# NOTE: These are experimental in 0.92.
2182
format_registry.register_lazy(
2183
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2184
'bzrlib.repofmt.pack_repo',
2185
'RepositoryFormatKnitPack1',
2187
format_registry.register_lazy(
2188
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2189
'bzrlib.repofmt.pack_repo',
2190
'RepositoryFormatKnitPack3',
2192
format_registry.register_lazy(
2193
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2194
'bzrlib.repofmt.pack_repo',
2195
'RepositoryFormatKnitPack4',
2199
class InterRepository(InterObject):
2200
"""This class represents operations taking place between two repositories.
2202
Its instances have methods like copy_content and fetch, and contain
2203
references to the source and target repositories these operations can be
2206
Often we will provide convenience methods on 'repository' which carry out
2207
operations with another repository - they will always forward to
2208
InterRepository.get(other).method_name(parameters).
2212
"""The available optimised InterRepository types."""
2214
def copy_content(self, revision_id=None):
2215
raise NotImplementedError(self.copy_content)
2217
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2218
"""Fetch the content required to construct revision_id.
2220
The content is copied from self.source to self.target.
2222
:param revision_id: if None all content is copied, if NULL_REVISION no
2224
:param pb: optional progress bar to use for progress reports. If not
2225
provided a default one will be created.
2227
Returns the copied revision count and the failed revisions in a tuple:
2230
raise NotImplementedError(self.fetch)
2233
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2234
"""Return the revision ids that source has that target does not.
2236
These are returned in topological order.
2238
:param revision_id: only return revision ids included by this
2241
# generic, possibly worst case, slow code path.
2242
target_ids = set(self.target.all_revision_ids())
2243
if revision_id is not None:
2244
source_ids = self.source.get_ancestry(revision_id)
2245
assert source_ids[0] is None
2248
source_ids = self.source.all_revision_ids()
2249
result_set = set(source_ids).difference(target_ids)
2250
# this may look like a no-op: its not. It preserves the ordering
2251
# other_ids had while only returning the members from other_ids
2252
# that we've decided we need.
2253
return [rev_id for rev_id in source_ids if rev_id in result_set]
2256
def _same_model(source, target):
2257
"""True if source and target have the same data representation."""
2258
if source.supports_rich_root() != target.supports_rich_root():
2260
if source._serializer != target._serializer:
2265
class InterSameDataRepository(InterRepository):
2266
"""Code for converting between repositories that represent the same data.
2268
Data format and model must match for this to work.
2272
def _get_repo_format_to_test(self):
2273
"""Repository format for testing with.
2275
InterSameData can pull from subtree to subtree and from non-subtree to
2276
non-subtree, so we test this with the richest repository format.
2278
from bzrlib.repofmt import knitrepo
2279
return knitrepo.RepositoryFormatKnit3()
2282
def is_compatible(source, target):
2283
return InterRepository._same_model(source, target)
2286
def copy_content(self, revision_id=None):
2287
"""Make a complete copy of the content in self into destination.
2289
This copies both the repository's revision data, and configuration information
2290
such as the make_working_trees setting.
2292
This is a destructive operation! Do not use it on existing
2295
:param revision_id: Only copy the content needed to construct
2296
revision_id and its parents.
2299
self.target.set_make_working_trees(self.source.make_working_trees())
2300
except NotImplementedError:
2302
# but don't bother fetching if we have the needed data now.
2303
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2304
self.target.has_revision(revision_id)):
2306
self.target.fetch(self.source, revision_id=revision_id)
2309
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2310
"""See InterRepository.fetch()."""
2311
from bzrlib.fetch import GenericRepoFetcher
2312
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2313
self.source, self.source._format, self.target,
2314
self.target._format)
2315
f = GenericRepoFetcher(to_repository=self.target,
2316
from_repository=self.source,
2317
last_revision=revision_id,
2319
return f.count_copied, f.failed_revisions
2322
class InterWeaveRepo(InterSameDataRepository):
2323
"""Optimised code paths between Weave based repositories.
2325
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2326
implemented lazy inter-object optimisation.
2330
def _get_repo_format_to_test(self):
2331
from bzrlib.repofmt import weaverepo
2332
return weaverepo.RepositoryFormat7()
2335
def is_compatible(source, target):
2336
"""Be compatible with known Weave formats.
2338
We don't test for the stores being of specific types because that
2339
could lead to confusing results, and there is no need to be
2342
from bzrlib.repofmt.weaverepo import (
2348
return (isinstance(source._format, (RepositoryFormat5,
2350
RepositoryFormat7)) and
2351
isinstance(target._format, (RepositoryFormat5,
2353
RepositoryFormat7)))
2354
except AttributeError:
2358
def copy_content(self, revision_id=None):
2359
"""See InterRepository.copy_content()."""
2360
# weave specific optimised path:
2362
self.target.set_make_working_trees(self.source.make_working_trees())
2363
except NotImplementedError:
2365
# FIXME do not peek!
2366
if self.source.control_files._transport.listable():
2367
pb = ui.ui_factory.nested_progress_bar()
2369
self.target.weave_store.copy_all_ids(
2370
self.source.weave_store,
2372
from_transaction=self.source.get_transaction(),
2373
to_transaction=self.target.get_transaction())
2374
pb.update('copying inventory', 0, 1)
2375
self.target.control_weaves.copy_multi(
2376
self.source.control_weaves, ['inventory'],
2377
from_transaction=self.source.get_transaction(),
2378
to_transaction=self.target.get_transaction())
2379
self.target._revision_store.text_store.copy_all_ids(
2380
self.source._revision_store.text_store,
2385
self.target.fetch(self.source, revision_id=revision_id)
2388
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2389
"""See InterRepository.fetch()."""
2390
from bzrlib.fetch import GenericRepoFetcher
2391
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2392
self.source, self.source._format, self.target, self.target._format)
2393
f = GenericRepoFetcher(to_repository=self.target,
2394
from_repository=self.source,
2395
last_revision=revision_id,
2397
return f.count_copied, f.failed_revisions
2400
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2401
"""See InterRepository.missing_revision_ids()."""
2402
# we want all revisions to satisfy revision_id in source.
2403
# but we don't want to stat every file here and there.
2404
# we want then, all revisions other needs to satisfy revision_id
2405
# checked, but not those that we have locally.
2406
# so the first thing is to get a subset of the revisions to
2407
# satisfy revision_id in source, and then eliminate those that
2408
# we do already have.
2409
# this is slow on high latency connection to self, but as as this
2410
# disk format scales terribly for push anyway due to rewriting
2411
# inventory.weave, this is considered acceptable.
2413
if revision_id is not None:
2414
source_ids = self.source.get_ancestry(revision_id)
2415
assert source_ids[0] is None
2418
source_ids = self.source._all_possible_ids()
2419
source_ids_set = set(source_ids)
2420
# source_ids is the worst possible case we may need to pull.
2421
# now we want to filter source_ids against what we actually
2422
# have in target, but don't try to check for existence where we know
2423
# we do not have a revision as that would be pointless.
2424
target_ids = set(self.target._all_possible_ids())
2425
possibly_present_revisions = target_ids.intersection(source_ids_set)
2426
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2427
required_revisions = source_ids_set.difference(actually_present_revisions)
2428
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2429
if revision_id is not None:
2430
# we used get_ancestry to determine source_ids then we are assured all
2431
# revisions referenced are present as they are installed in topological order.
2432
# and the tip revision was validated by get_ancestry.
2433
return required_topo_revisions
2435
# if we just grabbed the possibly available ids, then
2436
# we only have an estimate of whats available and need to validate
2437
# that against the revision records.
2438
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2441
class InterKnitRepo(InterSameDataRepository):
2442
"""Optimised code paths between Knit based repositories."""
2445
def _get_repo_format_to_test(self):
2446
from bzrlib.repofmt import knitrepo
2447
return knitrepo.RepositoryFormatKnit1()
2450
def is_compatible(source, target):
2451
"""Be compatible with known Knit formats.
2453
We don't test for the stores being of specific types because that
2454
could lead to confusing results, and there is no need to be
2457
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2459
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2460
isinstance(target._format, RepositoryFormatKnit))
2461
except AttributeError:
2463
return are_knits and InterRepository._same_model(source, target)
2466
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2467
"""See InterRepository.fetch()."""
2468
from bzrlib.fetch import KnitRepoFetcher
2469
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2470
self.source, self.source._format, self.target, self.target._format)
2471
f = KnitRepoFetcher(to_repository=self.target,
2472
from_repository=self.source,
2473
last_revision=revision_id,
2475
return f.count_copied, f.failed_revisions
2478
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2479
"""See InterRepository.missing_revision_ids()."""
2480
if revision_id is not None:
2481
source_ids = self.source.get_ancestry(revision_id)
2482
assert source_ids[0] is None
2485
source_ids = self.source.all_revision_ids()
2486
source_ids_set = set(source_ids)
2487
# source_ids is the worst possible case we may need to pull.
2488
# now we want to filter source_ids against what we actually
2489
# have in target, but don't try to check for existence where we know
2490
# we do not have a revision as that would be pointless.
2491
target_ids = set(self.target.all_revision_ids())
2492
possibly_present_revisions = target_ids.intersection(source_ids_set)
2493
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2494
required_revisions = source_ids_set.difference(actually_present_revisions)
2495
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2496
if revision_id is not None:
2497
# we used get_ancestry to determine source_ids then we are assured all
2498
# revisions referenced are present as they are installed in topological order.
2499
# and the tip revision was validated by get_ancestry.
2500
return required_topo_revisions
2502
# if we just grabbed the possibly available ids, then
2503
# we only have an estimate of whats available and need to validate
2504
# that against the revision records.
2505
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2508
class InterPackRepo(InterSameDataRepository):
2509
"""Optimised code paths between Pack based repositories."""
2512
def _get_repo_format_to_test(self):
2513
from bzrlib.repofmt import pack_repo
2514
return pack_repo.RepositoryFormatKnitPack1()
2517
def is_compatible(source, target):
2518
"""Be compatible with known Pack formats.
2520
We don't test for the stores being of specific types because that
2521
could lead to confusing results, and there is no need to be
2524
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2526
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2527
isinstance(target._format, RepositoryFormatPack))
2528
except AttributeError:
2530
return are_packs and InterRepository._same_model(source, target)
2533
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2534
"""See InterRepository.fetch()."""
2535
from bzrlib.repofmt.pack_repo import Packer
2536
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2537
self.source, self.source._format, self.target, self.target._format)
2538
self.count_copied = 0
2539
if revision_id is None:
2541
# everything to do - use pack logic
2542
# to fetch from all packs to one without
2543
# inventory parsing etc, IFF nothing to be copied is in the target.
2545
revision_ids = self.source.all_revision_ids()
2546
# implementing the TODO will involve:
2547
# - detecting when all of a pack is selected
2548
# - avoiding as much as possible pre-selection, so the
2549
# more-core routines such as create_pack_from_packs can filter in
2550
# a just-in-time fashion. (though having a HEADS list on a
2551
# repository might make this a lot easier, because we could
2552
# sensibly detect 'new revisions' without doing a full index scan.
2553
elif _mod_revision.is_null(revision_id):
2558
revision_ids = self.missing_revision_ids(revision_id,
2559
find_ghosts=find_ghosts)
2560
except errors.NoSuchRevision:
2561
raise errors.InstallFailed([revision_id])
2562
packs = self.source._pack_collection.all_packs()
2563
pack = Packer(self.target._pack_collection, packs, '.fetch',
2564
revision_ids).pack()
2565
if pack is not None:
2566
self.target._pack_collection._save_pack_names()
2567
# Trigger an autopack. This may duplicate effort as we've just done
2568
# a pack creation, but for now it is simpler to think about as
2569
# 'upload data, then repack if needed'.
2570
self.target._pack_collection.autopack()
2571
return (pack.get_revision_count(), [])
2576
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2577
"""See InterRepository.missing_revision_ids().
2579
:param find_ghosts: Find ghosts throughough the ancestry of
2582
if not find_ghosts and revision_id is not None:
2583
graph = self.source.get_graph()
2584
missing_revs = set()
2585
searcher = graph._make_breadth_first_searcher([revision_id])
2587
self.target._pack_collection.revision_index.combined_index
2588
null_set = frozenset([_mod_revision.NULL_REVISION])
2591
next_revs = set(searcher.next())
2592
except StopIteration:
2594
next_revs.difference_update(null_set)
2595
target_keys = [(key,) for key in next_revs]
2596
have_revs = frozenset(node[1][0] for node in
2597
target_index.iter_entries(target_keys))
2598
missing_revs.update(next_revs - have_revs)
2599
searcher.stop_searching_any(have_revs)
2600
if next_revs - have_revs == set([revision_id]):
2601
# we saw the start rev itself, but no parents from it (or
2602
# next_revs would have been updated to e.g. set(). We remove
2603
# have_revs because if we found revision_id locally we
2604
# stop_searching at the first time around.
2605
raise errors.NoSuchRevision(self.source, revision_id)
2607
elif revision_id is not None:
2608
source_ids = self.source.get_ancestry(revision_id)
2609
assert source_ids[0] is None
2612
source_ids = self.source.all_revision_ids()
2613
# source_ids is the worst possible case we may need to pull.
2614
# now we want to filter source_ids against what we actually
2615
# have in target, but don't try to check for existence where we know
2616
# we do not have a revision as that would be pointless.
2617
target_ids = set(self.target.all_revision_ids())
2618
return [r for r in source_ids if (r not in target_ids)]
2621
class InterModel1and2(InterRepository):
2624
def _get_repo_format_to_test(self):
2628
def is_compatible(source, target):
2629
if not source.supports_rich_root() and target.supports_rich_root():
2635
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2636
"""See InterRepository.fetch()."""
2637
from bzrlib.fetch import Model1toKnit2Fetcher
2638
f = Model1toKnit2Fetcher(to_repository=self.target,
2639
from_repository=self.source,
2640
last_revision=revision_id,
2642
return f.count_copied, f.failed_revisions
2645
def copy_content(self, revision_id=None):
2646
"""Make a complete copy of the content in self into destination.
2648
This is a destructive operation! Do not use it on existing
2651
:param revision_id: Only copy the content needed to construct
2652
revision_id and its parents.
2655
self.target.set_make_working_trees(self.source.make_working_trees())
2656
except NotImplementedError:
2658
# but don't bother fetching if we have the needed data now.
2659
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2660
self.target.has_revision(revision_id)):
2662
self.target.fetch(self.source, revision_id=revision_id)
2665
class InterKnit1and2(InterKnitRepo):
2668
def _get_repo_format_to_test(self):
2672
def is_compatible(source, target):
2673
"""Be compatible with Knit1 source and Knit3 target"""
2674
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2676
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2677
RepositoryFormatKnit3)
2678
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2679
RepositoryFormatKnitPack3)
2680
return (isinstance(source._format,
2681
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2682
isinstance(target._format,
2683
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2685
except AttributeError:
2689
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2690
"""See InterRepository.fetch()."""
2691
from bzrlib.fetch import Knit1to2Fetcher
2692
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2693
self.source, self.source._format, self.target,
2694
self.target._format)
2695
f = Knit1to2Fetcher(to_repository=self.target,
2696
from_repository=self.source,
2697
last_revision=revision_id,
2699
return f.count_copied, f.failed_revisions
2702
class InterDifferingSerializer(InterKnitRepo):
2705
def _get_repo_format_to_test(self):
2709
def is_compatible(source, target):
2710
"""Be compatible with Knit2 source and Knit3 target"""
2711
if source.supports_rich_root() != target.supports_rich_root():
2713
# Ideally, we'd support fetching if the source had no tree references
2714
# even if it supported them...
2715
if (getattr(source, '_format.supports_tree_reference', False) and
2716
not getattr(target, '_format.supports_tree_reference', False)):
2721
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2722
"""See InterRepository.fetch()."""
2723
revision_ids = self.target.missing_revision_ids(self.source,
2725
def revisions_iterator():
2726
for current_revision_id in revision_ids:
2727
revision = self.source.get_revision(current_revision_id)
2728
tree = self.source.revision_tree(current_revision_id)
2730
signature = self.source.get_signature_text(
2731
current_revision_id)
2732
except errors.NoSuchRevision:
2734
yield revision, tree, signature
2735
install_revisions(self.target, revisions_iterator())
2736
return len(revision_ids), 0
2739
class InterRemoteToOther(InterRepository):
2741
def __init__(self, source, target):
2742
InterRepository.__init__(self, source, target)
2743
self._real_inter = None
2746
def is_compatible(source, target):
2747
if not isinstance(source, remote.RemoteRepository):
2749
source._ensure_real()
2750
real_source = source._real_repository
2751
# Is source's model compatible with target's model, and are they the
2752
# same format? Currently we can only optimise fetching from an
2753
# identical model & format repo.
2754
assert not isinstance(real_source, remote.RemoteRepository), (
2755
"We don't support remote repos backed by remote repos yet.")
2756
return real_source._format == target._format
2759
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2760
"""See InterRepository.fetch()."""
2761
from bzrlib.fetch import RemoteToOtherFetcher
2762
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2763
self.source, self.target, self.target._format)
2764
# TODO: jam 20070210 This should be an assert, not a translate
2765
revision_id = osutils.safe_revision_id(revision_id)
2766
f = RemoteToOtherFetcher(to_repository=self.target,
2767
from_repository=self.source,
2768
last_revision=revision_id,
2770
return f.count_copied, f.failed_revisions
2773
def _get_repo_format_to_test(self):
2777
class InterOtherToRemote(InterRepository):
2779
def __init__(self, source, target):
2780
InterRepository.__init__(self, source, target)
2781
self._real_inter = None
2784
def is_compatible(source, target):
2785
if isinstance(target, remote.RemoteRepository):
2789
def _ensure_real_inter(self):
2790
if self._real_inter is None:
2791
self.target._ensure_real()
2792
real_target = self.target._real_repository
2793
self._real_inter = InterRepository.get(self.source, real_target)
2795
def copy_content(self, revision_id=None):
2796
self._ensure_real_inter()
2797
self._real_inter.copy_content(revision_id=revision_id)
2799
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2800
self._ensure_real_inter()
2801
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2804
def _get_repo_format_to_test(self):
2808
InterRepository.register_optimiser(InterDifferingSerializer)
2809
InterRepository.register_optimiser(InterSameDataRepository)
2810
InterRepository.register_optimiser(InterWeaveRepo)
2811
InterRepository.register_optimiser(InterKnitRepo)
2812
InterRepository.register_optimiser(InterModel1and2)
2813
InterRepository.register_optimiser(InterKnit1and2)
2814
InterRepository.register_optimiser(InterPackRepo)
2815
InterRepository.register_optimiser(InterRemoteToOther)
2816
InterRepository.register_optimiser(InterOtherToRemote)
2819
class CopyConverter(object):
2820
"""A repository conversion tool which just performs a copy of the content.
2822
This is slow but quite reliable.
2825
def __init__(self, target_format):
2826
"""Create a CopyConverter.
2828
:param target_format: The format the resulting repository should be.
2830
self.target_format = target_format
2832
def convert(self, repo, pb):
2833
"""Perform the conversion of to_convert, giving feedback via pb.
2835
:param to_convert: The disk object to convert.
2836
:param pb: a progress bar to use for progress information.
2841
# this is only useful with metadir layouts - separated repo content.
2842
# trigger an assertion if not such
2843
repo._format.get_format_string()
2844
self.repo_dir = repo.bzrdir
2845
self.step('Moving repository to repository.backup')
2846
self.repo_dir.transport.move('repository', 'repository.backup')
2847
backup_transport = self.repo_dir.transport.clone('repository.backup')
2848
repo._format.check_conversion_target(self.target_format)
2849
self.source_repo = repo._format.open(self.repo_dir,
2851
_override_transport=backup_transport)
2852
self.step('Creating new repository')
2853
converted = self.target_format.initialize(self.repo_dir,
2854
self.source_repo.is_shared())
2855
converted.lock_write()
2857
self.step('Copying content into repository.')
2858
self.source_repo.copy_content_into(converted)
2861
self.step('Deleting old repository content.')
2862
self.repo_dir.transport.delete_tree('repository.backup')
2863
self.pb.note('repository converted')
2865
def step(self, message):
2866
"""Update the pb by a step."""
2868
self.pb.update(message, self.count, self.total)
2880
def _unescaper(match, _map=_unescape_map):
2881
code = match.group(1)
2885
if not code.startswith('#'):
2887
return unichr(int(code[1:])).encode('utf8')
2893
def _unescape_xml(data):
2894
"""Unescape predefined XML entities in a string of data."""
2896
if _unescape_re is None:
2897
_unescape_re = re.compile('\&([^;]*);')
2898
return _unescape_re.sub(_unescaper, data)
2901
class _VersionedFileChecker(object):
2903
def __init__(self, repository):
2904
self.repository = repository
2905
self.text_index = self.repository._generate_text_key_index()
2907
def calculate_file_version_parents(self, revision_id, file_id):
2908
"""Calculate the correct parents for a file version according to
2911
parent_keys = self.text_index[(file_id, revision_id)]
2912
if parent_keys == [_mod_revision.NULL_REVISION]:
2914
# strip the file_id, for the weave api
2915
return tuple([revision_id for file_id, revision_id in parent_keys])
2917
def check_file_version_parents(self, weave, file_id):
2918
"""Check the parents stored in a versioned file are correct.
2920
It also detects file versions that are not referenced by their
2921
corresponding revision's inventory.
2923
:returns: A tuple of (wrong_parents, dangling_file_versions).
2924
wrong_parents is a dict mapping {revision_id: (stored_parents,
2925
correct_parents)} for each revision_id where the stored parents
2926
are not correct. dangling_file_versions is a set of (file_id,
2927
revision_id) tuples for versions that are present in this versioned
2928
file, but not used by the corresponding inventory.
2931
unused_versions = set()
2932
for num, revision_id in enumerate(weave.versions()):
2934
correct_parents = self.calculate_file_version_parents(
2935
revision_id, file_id)
2937
# The version is not part of the used keys.
2938
unused_versions.add(revision_id)
2941
knit_parents = tuple(weave.get_parents(revision_id))
2942
except errors.RevisionNotPresent:
2944
if correct_parents != knit_parents:
2945
wrong_parents[revision_id] = (knit_parents, correct_parents)
2946
return wrong_parents, unused_versions
added
removed
bzrlib/repository.py
