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):
1246
"""Generate a new text key index for the repository.
1248
This is an expensive function that will take considerable time to run.
1250
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1251
list of parents, also text keys. When a given key has no parents,
1252
the parents list will be [NULL_REVISION].
1254
# All revisions, to find inventory parents.
1255
revision_graph = self.get_revision_graph_with_ghosts()
1256
ancestors = revision_graph.get_ancestors()
1257
text_key_references = self.find_text_key_references()
1258
pb = ui.ui_factory.nested_progress_bar()
1260
return self._do_generate_text_key_index(ancestors,
1261
text_key_references, pb)
1265
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1266
"""Helper for _generate_text_key_index to avoid deep nesting."""
1267
revision_order = tsort.topo_sort(ancestors)
1268
invalid_keys = set()
1270
for revision_id in revision_order:
1271
revision_keys[revision_id] = set()
1272
text_count = len(text_key_references)
1273
# a cache of the text keys to allow reuse; costs a dict of all the
1274
# keys, but saves a 2-tuple for every child of a given key.
1276
for text_key, valid in text_key_references.iteritems():
1278
invalid_keys.add(text_key)
1280
revision_keys[text_key[1]].add(text_key)
1281
text_key_cache[text_key] = text_key
1282
del text_key_references
1284
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1285
NULL_REVISION = _mod_revision.NULL_REVISION
1286
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1287
# too small for large or very branchy trees. However, for 55K path
1288
# trees, it would be easy to use too much memory trivially. Ideally we
1289
# could gauge this by looking at available real memory etc, but this is
1290
# always a tricky proposition.
1291
inventory_cache = lru_cache.LRUCache(10)
1292
batch_size = 10 # should be ~150MB on a 55K path tree
1293
batch_count = len(revision_order) / batch_size + 1
1295
pb.update("Calculating text parents.", processed_texts, text_count)
1296
for offset in xrange(batch_count):
1297
to_query = revision_order[offset * batch_size:(offset + 1) *
1301
for rev_tree in self.revision_trees(to_query):
1302
revision_id = rev_tree.get_revision_id()
1303
parent_ids = ancestors[revision_id]
1304
for text_key in revision_keys[revision_id]:
1305
pb.update("Calculating text parents.", processed_texts)
1306
processed_texts += 1
1307
candidate_parents = []
1308
for parent_id in parent_ids:
1309
parent_text_key = (text_key[0], parent_id)
1311
check_parent = parent_text_key not in \
1312
revision_keys[parent_id]
1314
# the parent parent_id is a ghost:
1315
check_parent = False
1316
# truncate the derived graph against this ghost.
1317
parent_text_key = None
1319
# look at the parent commit details inventories to
1320
# determine possible candidates in the per file graph.
1323
inv = inventory_cache[parent_id]
1325
inv = self.revision_tree(parent_id).inventory
1326
inventory_cache[parent_id] = inv
1327
parent_entry = inv._byid.get(text_key[0], None)
1328
if parent_entry is not None:
1330
text_key[0], parent_entry.revision)
1332
parent_text_key = None
1333
if parent_text_key is not None:
1334
candidate_parents.append(
1335
text_key_cache[parent_text_key])
1336
parent_heads = text_graph.heads(candidate_parents)
1337
new_parents = list(parent_heads)
1338
new_parents.sort(key=lambda x:candidate_parents.index(x))
1339
if new_parents == []:
1340
new_parents = [NULL_REVISION]
1341
text_index[text_key] = new_parents
1343
for text_key in invalid_keys:
1344
text_index[text_key] = [NULL_REVISION]
1347
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1348
"""Get an iterable listing the keys of all the data introduced by a set
1351
The keys will be ordered so that the corresponding items can be safely
1352
fetched and inserted in that order.
1354
:returns: An iterable producing tuples of (knit-kind, file-id,
1355
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1356
'revisions'. file-id is None unless knit-kind is 'file'.
1358
# XXX: it's a bit weird to control the inventory weave caching in this
1359
# generator. Ideally the caching would be done in fetch.py I think. Or
1360
# maybe this generator should explicitly have the contract that it
1361
# should not be iterated until the previously yielded item has been
1364
inv_w = self.get_inventory_weave()
1365
inv_w.enable_cache()
1367
# file ids that changed
1368
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1370
num_file_ids = len(file_ids)
1371
for file_id, altered_versions in file_ids.iteritems():
1372
if _files_pb is not None:
1373
_files_pb.update("fetch texts", count, num_file_ids)
1375
yield ("file", file_id, altered_versions)
1376
# We're done with the files_pb. Note that it finished by the caller,
1377
# just as it was created by the caller.
1381
yield ("inventory", None, revision_ids)
1385
revisions_with_signatures = set()
1386
for rev_id in revision_ids:
1388
self.get_signature_text(rev_id)
1389
except errors.NoSuchRevision:
1393
revisions_with_signatures.add(rev_id)
1395
yield ("signatures", None, revisions_with_signatures)
1398
yield ("revisions", None, revision_ids)
1401
def get_inventory_weave(self):
1402
return self.control_weaves.get_weave('inventory',
1403
self.get_transaction())
1406
def get_inventory(self, revision_id):
1407
"""Get Inventory object by hash."""
1408
return self.deserialise_inventory(
1409
revision_id, self.get_inventory_xml(revision_id))
1411
def deserialise_inventory(self, revision_id, xml):
1412
"""Transform the xml into an inventory object.
1414
:param revision_id: The expected revision id of the inventory.
1415
:param xml: A serialised inventory.
1417
return self._serializer.read_inventory_from_string(xml, revision_id)
1419
def serialise_inventory(self, inv):
1420
return self._serializer.write_inventory_to_string(inv)
1422
def _serialise_inventory_to_lines(self, inv):
1423
return self._serializer.write_inventory_to_lines(inv)
1425
def get_serializer_format(self):
1426
return self._serializer.format_num
1429
def get_inventory_xml(self, revision_id):
1430
"""Get inventory XML as a file object."""
1432
assert isinstance(revision_id, str), type(revision_id)
1433
iw = self.get_inventory_weave()
1434
return iw.get_text(revision_id)
1436
raise errors.HistoryMissing(self, 'inventory', revision_id)
1439
def get_inventory_sha1(self, revision_id):
1440
"""Return the sha1 hash of the inventory entry
1442
return self.get_revision(revision_id).inventory_sha1
1445
def get_revision_graph(self, revision_id=None):
1446
"""Return a dictionary containing the revision graph.
1448
NB: This method should not be used as it accesses the entire graph all
1449
at once, which is much more data than most operations should require.
1451
:param revision_id: The revision_id to get a graph from. If None, then
1452
the entire revision graph is returned. This is a deprecated mode of
1453
operation and will be removed in the future.
1454
:return: a dictionary of revision_id->revision_parents_list.
1456
raise NotImplementedError(self.get_revision_graph)
1459
def get_revision_graph_with_ghosts(self, revision_ids=None):
1460
"""Return a graph of the revisions with ghosts marked as applicable.
1462
:param revision_ids: an iterable of revisions to graph or None for all.
1463
:return: a Graph object with the graph reachable from revision_ids.
1465
if 'evil' in debug.debug_flags:
1467
"get_revision_graph_with_ghosts scales with size of history.")
1468
result = deprecated_graph.Graph()
1469
if not revision_ids:
1470
pending = set(self.all_revision_ids())
1473
pending = set(revision_ids)
1474
# special case NULL_REVISION
1475
if _mod_revision.NULL_REVISION in pending:
1476
pending.remove(_mod_revision.NULL_REVISION)
1477
required = set(pending)
1480
revision_id = pending.pop()
1482
rev = self.get_revision(revision_id)
1483
except errors.NoSuchRevision:
1484
if revision_id in required:
1487
result.add_ghost(revision_id)
1489
for parent_id in rev.parent_ids:
1490
# is this queued or done ?
1491
if (parent_id not in pending and
1492
parent_id not in done):
1494
pending.add(parent_id)
1495
result.add_node(revision_id, rev.parent_ids)
1496
done.add(revision_id)
1499
def _get_history_vf(self):
1500
"""Get a versionedfile whose history graph reflects all revisions.
1502
For weave repositories, this is the inventory weave.
1504
return self.get_inventory_weave()
1506
def iter_reverse_revision_history(self, revision_id):
1507
"""Iterate backwards through revision ids in the lefthand history
1509
:param revision_id: The revision id to start with. All its lefthand
1510
ancestors will be traversed.
1512
if revision_id in (None, _mod_revision.NULL_REVISION):
1514
next_id = revision_id
1515
versionedfile = self._get_history_vf()
1518
parents = versionedfile.get_parents(next_id)
1519
if len(parents) == 0:
1522
next_id = parents[0]
1525
def get_revision_inventory(self, revision_id):
1526
"""Return inventory of a past revision."""
1527
# TODO: Unify this with get_inventory()
1528
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1529
# must be the same as its revision, so this is trivial.
1530
if revision_id is None:
1531
# This does not make sense: if there is no revision,
1532
# then it is the current tree inventory surely ?!
1533
# and thus get_root_id() is something that looks at the last
1534
# commit on the branch, and the get_root_id is an inventory check.
1535
raise NotImplementedError
1536
# return Inventory(self.get_root_id())
1538
return self.get_inventory(revision_id)
1541
def is_shared(self):
1542
"""Return True if this repository is flagged as a shared repository."""
1543
raise NotImplementedError(self.is_shared)
1546
def reconcile(self, other=None, thorough=False):
1547
"""Reconcile this repository."""
1548
from bzrlib.reconcile import RepoReconciler
1549
reconciler = RepoReconciler(self, thorough=thorough)
1550
reconciler.reconcile()
1553
def _refresh_data(self):
1554
"""Helper called from lock_* to ensure coherency with disk.
1556
The default implementation does nothing; it is however possible
1557
for repositories to maintain loaded indices across multiple locks
1558
by checking inside their implementation of this method to see
1559
whether their indices are still valid. This depends of course on
1560
the disk format being validatable in this manner.
1564
def revision_tree(self, revision_id):
1565
"""Return Tree for a revision on this branch.
1567
`revision_id` may be None for the empty tree revision.
1569
# TODO: refactor this to use an existing revision object
1570
# so we don't need to read it in twice.
1571
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1572
return RevisionTree(self, Inventory(root_id=None),
1573
_mod_revision.NULL_REVISION)
1575
inv = self.get_revision_inventory(revision_id)
1576
return RevisionTree(self, inv, revision_id)
1579
def revision_trees(self, revision_ids):
1580
"""Return Tree for a revision on this branch.
1582
`revision_id` may not be None or 'null:'"""
1583
assert None not in revision_ids
1584
assert _mod_revision.NULL_REVISION not in revision_ids
1585
texts = self.get_inventory_weave().get_texts(revision_ids)
1586
for text, revision_id in zip(texts, revision_ids):
1587
inv = self.deserialise_inventory(revision_id, text)
1588
yield RevisionTree(self, inv, revision_id)
1591
def get_ancestry(self, revision_id, topo_sorted=True):
1592
"""Return a list of revision-ids integrated by a revision.
1594
The first element of the list is always None, indicating the origin
1595
revision. This might change when we have history horizons, or
1596
perhaps we should have a new API.
1598
This is topologically sorted.
1600
if _mod_revision.is_null(revision_id):
1602
if not self.has_revision(revision_id):
1603
raise errors.NoSuchRevision(self, revision_id)
1604
w = self.get_inventory_weave()
1605
candidates = w.get_ancestry(revision_id, topo_sorted)
1606
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1609
"""Compress the data within the repository.
1611
This operation only makes sense for some repository types. For other
1612
types it should be a no-op that just returns.
1614
This stub method does not require a lock, but subclasses should use
1615
@needs_write_lock as this is a long running call its reasonable to
1616
implicitly lock for the user.
1620
def print_file(self, file, revision_id):
1621
"""Print `file` to stdout.
1623
FIXME RBC 20060125 as John Meinel points out this is a bad api
1624
- it writes to stdout, it assumes that that is valid etc. Fix
1625
by creating a new more flexible convenience function.
1627
tree = self.revision_tree(revision_id)
1628
# use inventory as it was in that revision
1629
file_id = tree.inventory.path2id(file)
1631
# TODO: jam 20060427 Write a test for this code path
1632
# it had a bug in it, and was raising the wrong
1634
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1635
tree.print_file(file_id)
1637
def get_transaction(self):
1638
return self.control_files.get_transaction()
1640
def revision_parents(self, revision_id):
1641
return self.get_inventory_weave().parent_names(revision_id)
1643
def get_parents(self, revision_ids):
1644
"""See StackedParentsProvider.get_parents"""
1646
for revision_id in revision_ids:
1647
if revision_id == _mod_revision.NULL_REVISION:
1651
parents = self.get_revision(revision_id).parent_ids
1652
except errors.NoSuchRevision:
1655
if len(parents) == 0:
1656
parents = [_mod_revision.NULL_REVISION]
1657
parents_list.append(parents)
1660
def _make_parents_provider(self):
1663
def get_graph(self, other_repository=None):
1664
"""Return the graph walker for this repository format"""
1665
parents_provider = self._make_parents_provider()
1666
if (other_repository is not None and
1667
other_repository.bzrdir.transport.base !=
1668
self.bzrdir.transport.base):
1669
parents_provider = graph._StackedParentsProvider(
1670
[parents_provider, other_repository._make_parents_provider()])
1671
return graph.Graph(parents_provider)
1673
def get_versioned_file_checker(self):
1674
"""Return an object suitable for checking versioned files."""
1675
return VersionedFileChecker(self)
1678
def set_make_working_trees(self, new_value):
1679
"""Set the policy flag for making working trees when creating branches.
1681
This only applies to branches that use this repository.
1683
The default is 'True'.
1684
:param new_value: True to restore the default, False to disable making
1687
raise NotImplementedError(self.set_make_working_trees)
1689
def make_working_trees(self):
1690
"""Returns the policy for making working trees on new branches."""
1691
raise NotImplementedError(self.make_working_trees)
1694
def sign_revision(self, revision_id, gpg_strategy):
1695
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1696
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1699
def has_signature_for_revision_id(self, revision_id):
1700
"""Query for a revision signature for revision_id in the repository."""
1701
return self._revision_store.has_signature(revision_id,
1702
self.get_transaction())
1705
def get_signature_text(self, revision_id):
1706
"""Return the text for a signature."""
1707
return self._revision_store.get_signature_text(revision_id,
1708
self.get_transaction())
1711
def check(self, revision_ids=None):
1712
"""Check consistency of all history of given revision_ids.
1714
Different repository implementations should override _check().
1716
:param revision_ids: A non-empty list of revision_ids whose ancestry
1717
will be checked. Typically the last revision_id of a branch.
1719
return self._check(revision_ids)
1721
def _check(self, revision_ids):
1722
result = check.Check(self)
1726
def _warn_if_deprecated(self):
1727
global _deprecation_warning_done
1728
if _deprecation_warning_done:
1730
_deprecation_warning_done = True
1731
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1732
% (self._format, self.bzrdir.transport.base))
1734
def supports_rich_root(self):
1735
return self._format.rich_root_data
1737
def _check_ascii_revisionid(self, revision_id, method):
1738
"""Private helper for ascii-only repositories."""
1739
# weave repositories refuse to store revisionids that are non-ascii.
1740
if revision_id is not None:
1741
# weaves require ascii revision ids.
1742
if isinstance(revision_id, unicode):
1744
revision_id.encode('ascii')
1745
except UnicodeEncodeError:
1746
raise errors.NonAsciiRevisionId(method, self)
1749
revision_id.decode('ascii')
1750
except UnicodeDecodeError:
1751
raise errors.NonAsciiRevisionId(method, self)
1753
def revision_graph_can_have_wrong_parents(self):
1754
"""Is it possible for this repository to have a revision graph with
1757
If True, then this repository must also implement
1758
_find_inconsistent_revision_parents so that check and reconcile can
1759
check for inconsistencies before proceeding with other checks that may
1760
depend on the revision index being consistent.
1762
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1764
# remove these delegates a while after bzr 0.15
1765
def __make_delegated(name, from_module):
1766
def _deprecated_repository_forwarder():
1767
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1768
% (name, from_module),
1771
m = __import__(from_module, globals(), locals(), [name])
1773
return getattr(m, name)
1774
except AttributeError:
1775
raise AttributeError('module %s has no name %s'
1777
globals()[name] = _deprecated_repository_forwarder
1780
'AllInOneRepository',
1781
'WeaveMetaDirRepository',
1782
'PreSplitOutRepositoryFormat',
1783
'RepositoryFormat4',
1784
'RepositoryFormat5',
1785
'RepositoryFormat6',
1786
'RepositoryFormat7',
1788
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1792
'RepositoryFormatKnit',
1793
'RepositoryFormatKnit1',
1795
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1798
def install_revision(repository, rev, revision_tree):
1799
"""Install all revision data into a repository."""
1800
install_revisions(repository, [(rev, revision_tree, None)])
1803
def install_revisions(repository, iterable):
1804
"""Install all revision data into a repository.
1806
Accepts an iterable of revision, tree, signature tuples. The signature
1809
repository.start_write_group()
1811
for revision, revision_tree, signature in iterable:
1812
_install_revision(repository, revision, revision_tree, signature)
1814
repository.abort_write_group()
1817
repository.commit_write_group()
1820
def _install_revision(repository, rev, revision_tree, signature):
1821
"""Install all revision data into a repository."""
1822
present_parents = []
1824
for p_id in rev.parent_ids:
1825
if repository.has_revision(p_id):
1826
present_parents.append(p_id)
1827
parent_trees[p_id] = repository.revision_tree(p_id)
1829
parent_trees[p_id] = repository.revision_tree(None)
1831
inv = revision_tree.inventory
1832
entries = inv.iter_entries()
1833
# backwards compatibility hack: skip the root id.
1834
if not repository.supports_rich_root():
1835
path, root = entries.next()
1836
if root.revision != rev.revision_id:
1837
raise errors.IncompatibleRevision(repr(repository))
1838
# Add the texts that are not already present
1839
for path, ie in entries:
1840
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1841
repository.get_transaction())
1842
if ie.revision not in w:
1844
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1845
# with InventoryEntry.find_previous_heads(). if it is, then there
1846
# is a latent bug here where the parents may have ancestors of each
1848
for revision, tree in parent_trees.iteritems():
1849
if ie.file_id not in tree:
1851
parent_id = tree.inventory[ie.file_id].revision
1852
if parent_id in text_parents:
1854
text_parents.append(parent_id)
1856
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1857
repository.get_transaction())
1858
lines = revision_tree.get_file(ie.file_id).readlines()
1859
vfile.add_lines(rev.revision_id, text_parents, lines)
1861
# install the inventory
1862
repository.add_inventory(rev.revision_id, inv, present_parents)
1863
except errors.RevisionAlreadyPresent:
1865
if signature is not None:
1866
repository.add_signature_text(rev.revision_id, signature)
1867
repository.add_revision(rev.revision_id, rev, inv)
1870
class MetaDirRepository(Repository):
1871
"""Repositories in the new meta-dir layout."""
1873
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1874
super(MetaDirRepository, self).__init__(_format,
1880
dir_mode = self.control_files._dir_mode
1881
file_mode = self.control_files._file_mode
1884
def is_shared(self):
1885
"""Return True if this repository is flagged as a shared repository."""
1886
return self.control_files._transport.has('shared-storage')
1889
def set_make_working_trees(self, new_value):
1890
"""Set the policy flag for making working trees when creating branches.
1892
This only applies to branches that use this repository.
1894
The default is 'True'.
1895
:param new_value: True to restore the default, False to disable making
1900
self.control_files._transport.delete('no-working-trees')
1901
except errors.NoSuchFile:
1904
self.control_files.put_utf8('no-working-trees', '')
1906
def make_working_trees(self):
1907
"""Returns the policy for making working trees on new branches."""
1908
return not self.control_files._transport.has('no-working-trees')
1911
class RepositoryFormatRegistry(registry.Registry):
1912
"""Registry of RepositoryFormats."""
1914
def get(self, format_string):
1915
r = registry.Registry.get(self, format_string)
1921
format_registry = RepositoryFormatRegistry()
1922
"""Registry of formats, indexed by their identifying format string.
1924
This can contain either format instances themselves, or classes/factories that
1925
can be called to obtain one.
1929
#####################################################################
1930
# Repository Formats
1932
class RepositoryFormat(object):
1933
"""A repository format.
1935
Formats provide three things:
1936
* An initialization routine to construct repository data on disk.
1937
* a format string which is used when the BzrDir supports versioned
1939
* an open routine which returns a Repository instance.
1941
There is one and only one Format subclass for each on-disk format. But
1942
there can be one Repository subclass that is used for several different
1943
formats. The _format attribute on a Repository instance can be used to
1944
determine the disk format.
1946
Formats are placed in an dict by their format string for reference
1947
during opening. These should be subclasses of RepositoryFormat
1950
Once a format is deprecated, just deprecate the initialize and open
1951
methods on the format class. Do not deprecate the object, as the
1952
object will be created every system load.
1954
Common instance attributes:
1955
_matchingbzrdir - the bzrdir format that the repository format was
1956
originally written to work with. This can be used if manually
1957
constructing a bzrdir and repository, or more commonly for test suite
1961
# Set to True or False in derived classes. True indicates that the format
1962
# supports ghosts gracefully.
1963
supports_ghosts = None
1966
return "<%s>" % self.__class__.__name__
1968
def __eq__(self, other):
1969
# format objects are generally stateless
1970
return isinstance(other, self.__class__)
1972
def __ne__(self, other):
1973
return not self == other
1976
def find_format(klass, a_bzrdir):
1977
"""Return the format for the repository object in a_bzrdir.
1979
This is used by bzr native formats that have a "format" file in
1980
the repository. Other methods may be used by different types of
1984
transport = a_bzrdir.get_repository_transport(None)
1985
format_string = transport.get("format").read()
1986
return format_registry.get(format_string)
1987
except errors.NoSuchFile:
1988
raise errors.NoRepositoryPresent(a_bzrdir)
1990
raise errors.UnknownFormatError(format=format_string)
1993
def register_format(klass, format):
1994
format_registry.register(format.get_format_string(), format)
1997
def unregister_format(klass, format):
1998
format_registry.remove(format.get_format_string())
2001
def get_default_format(klass):
2002
"""Return the current default format."""
2003
from bzrlib import bzrdir
2004
return bzrdir.format_registry.make_bzrdir('default').repository_format
2006
def _get_control_store(self, repo_transport, control_files):
2007
"""Return the control store for this repository."""
2008
raise NotImplementedError(self._get_control_store)
2010
def get_format_string(self):
2011
"""Return the ASCII format string that identifies this format.
2013
Note that in pre format ?? repositories the format string is
2014
not permitted nor written to disk.
2016
raise NotImplementedError(self.get_format_string)
2018
def get_format_description(self):
2019
"""Return the short description for this format."""
2020
raise NotImplementedError(self.get_format_description)
2022
def _get_revision_store(self, repo_transport, control_files):
2023
"""Return the revision store object for this a_bzrdir."""
2024
raise NotImplementedError(self._get_revision_store)
2026
def _get_text_rev_store(self,
2033
"""Common logic for getting a revision store for a repository.
2035
see self._get_revision_store for the subclass-overridable method to
2036
get the store for a repository.
2038
from bzrlib.store.revision.text import TextRevisionStore
2039
dir_mode = control_files._dir_mode
2040
file_mode = control_files._file_mode
2041
text_store = TextStore(transport.clone(name),
2043
compressed=compressed,
2045
file_mode=file_mode)
2046
_revision_store = TextRevisionStore(text_store, serializer)
2047
return _revision_store
2049
# TODO: this shouldn't be in the base class, it's specific to things that
2050
# use weaves or knits -- mbp 20070207
2051
def _get_versioned_file_store(self,
2056
versionedfile_class=None,
2057
versionedfile_kwargs={},
2059
if versionedfile_class is None:
2060
versionedfile_class = self._versionedfile_class
2061
weave_transport = control_files._transport.clone(name)
2062
dir_mode = control_files._dir_mode
2063
file_mode = control_files._file_mode
2064
return VersionedFileStore(weave_transport, prefixed=prefixed,
2066
file_mode=file_mode,
2067
versionedfile_class=versionedfile_class,
2068
versionedfile_kwargs=versionedfile_kwargs,
2071
def initialize(self, a_bzrdir, shared=False):
2072
"""Initialize a repository of this format in a_bzrdir.
2074
:param a_bzrdir: The bzrdir to put the new repository in it.
2075
:param shared: The repository should be initialized as a sharable one.
2076
:returns: The new repository object.
2078
This may raise UninitializableFormat if shared repository are not
2079
compatible the a_bzrdir.
2081
raise NotImplementedError(self.initialize)
2083
def is_supported(self):
2084
"""Is this format supported?
2086
Supported formats must be initializable and openable.
2087
Unsupported formats may not support initialization or committing or
2088
some other features depending on the reason for not being supported.
2092
def check_conversion_target(self, target_format):
2093
raise NotImplementedError(self.check_conversion_target)
2095
def open(self, a_bzrdir, _found=False):
2096
"""Return an instance of this format for the bzrdir a_bzrdir.
2098
_found is a private parameter, do not use it.
2100
raise NotImplementedError(self.open)
2103
class MetaDirRepositoryFormat(RepositoryFormat):
2104
"""Common base class for the new repositories using the metadir layout."""
2106
rich_root_data = False
2107
supports_tree_reference = False
2108
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
2111
super(MetaDirRepositoryFormat, self).__init__()
2113
def _create_control_files(self, a_bzrdir):
2114
"""Create the required files and the initial control_files object."""
2115
# FIXME: RBC 20060125 don't peek under the covers
2116
# NB: no need to escape relative paths that are url safe.
2117
repository_transport = a_bzrdir.get_repository_transport(self)
2118
control_files = lockable_files.LockableFiles(repository_transport,
2119
'lock', lockdir.LockDir)
2120
control_files.create_lock()
2121
return control_files
2123
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2124
"""Upload the initial blank content."""
2125
control_files = self._create_control_files(a_bzrdir)
2126
control_files.lock_write()
2128
control_files._transport.mkdir_multi(dirs,
2129
mode=control_files._dir_mode)
2130
for file, content in files:
2131
control_files.put(file, content)
2132
for file, content in utf8_files:
2133
control_files.put_utf8(file, content)
2135
control_files.put_utf8('shared-storage', '')
2137
control_files.unlock()
2140
# formats which have no format string are not discoverable
2141
# and not independently creatable, so are not registered. They're
2142
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2143
# needed, it's constructed directly by the BzrDir. Non-native formats where
2144
# the repository is not separately opened are similar.
2146
format_registry.register_lazy(
2147
'Bazaar-NG Repository format 7',
2148
'bzrlib.repofmt.weaverepo',
2152
# KEEP in sync with bzrdir.format_registry default, which controls the overall
2153
# default control directory format
2154
format_registry.register_lazy(
2155
'Bazaar-NG Knit Repository Format 1',
2156
'bzrlib.repofmt.knitrepo',
2157
'RepositoryFormatKnit1',
2159
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
2161
format_registry.register_lazy(
2162
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2163
'bzrlib.repofmt.knitrepo',
2164
'RepositoryFormatKnit3',
2167
format_registry.register_lazy(
2168
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2169
'bzrlib.repofmt.knitrepo',
2170
'RepositoryFormatKnit4',
2173
# Pack-based formats. There is one format for pre-subtrees, and one for
2174
# post-subtrees to allow ease of testing.
2175
# NOTE: These are experimental in 0.92.
2176
format_registry.register_lazy(
2177
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2178
'bzrlib.repofmt.pack_repo',
2179
'RepositoryFormatKnitPack1',
2181
format_registry.register_lazy(
2182
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2183
'bzrlib.repofmt.pack_repo',
2184
'RepositoryFormatKnitPack3',
2186
format_registry.register_lazy(
2187
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2188
'bzrlib.repofmt.pack_repo',
2189
'RepositoryFormatKnitPack4',
2193
class InterRepository(InterObject):
2194
"""This class represents operations taking place between two repositories.
2196
Its instances have methods like copy_content and fetch, and contain
2197
references to the source and target repositories these operations can be
2200
Often we will provide convenience methods on 'repository' which carry out
2201
operations with another repository - they will always forward to
2202
InterRepository.get(other).method_name(parameters).
2206
"""The available optimised InterRepository types."""
2208
def copy_content(self, revision_id=None):
2209
raise NotImplementedError(self.copy_content)
2211
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2212
"""Fetch the content required to construct revision_id.
2214
The content is copied from self.source to self.target.
2216
:param revision_id: if None all content is copied, if NULL_REVISION no
2218
:param pb: optional progress bar to use for progress reports. If not
2219
provided a default one will be created.
2221
Returns the copied revision count and the failed revisions in a tuple:
2224
raise NotImplementedError(self.fetch)
2227
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2228
"""Return the revision ids that source has that target does not.
2230
These are returned in topological order.
2232
:param revision_id: only return revision ids included by this
2235
# generic, possibly worst case, slow code path.
2236
target_ids = set(self.target.all_revision_ids())
2237
if revision_id is not None:
2238
source_ids = self.source.get_ancestry(revision_id)
2239
assert source_ids[0] is None
2242
source_ids = self.source.all_revision_ids()
2243
result_set = set(source_ids).difference(target_ids)
2244
# this may look like a no-op: its not. It preserves the ordering
2245
# other_ids had while only returning the members from other_ids
2246
# that we've decided we need.
2247
return [rev_id for rev_id in source_ids if rev_id in result_set]
2250
def _same_model(source, target):
2251
"""True if source and target have the same data representation."""
2252
if source.supports_rich_root() != target.supports_rich_root():
2254
if source._serializer != target._serializer:
2259
class InterSameDataRepository(InterRepository):
2260
"""Code for converting between repositories that represent the same data.
2262
Data format and model must match for this to work.
2266
def _get_repo_format_to_test(self):
2267
"""Repository format for testing with.
2269
InterSameData can pull from subtree to subtree and from non-subtree to
2270
non-subtree, so we test this with the richest repository format.
2272
from bzrlib.repofmt import knitrepo
2273
return knitrepo.RepositoryFormatKnit3()
2276
def is_compatible(source, target):
2277
return InterRepository._same_model(source, target)
2280
def copy_content(self, revision_id=None):
2281
"""Make a complete copy of the content in self into destination.
2283
This copies both the repository's revision data, and configuration information
2284
such as the make_working_trees setting.
2286
This is a destructive operation! Do not use it on existing
2289
:param revision_id: Only copy the content needed to construct
2290
revision_id and its parents.
2293
self.target.set_make_working_trees(self.source.make_working_trees())
2294
except NotImplementedError:
2296
# but don't bother fetching if we have the needed data now.
2297
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2298
self.target.has_revision(revision_id)):
2300
self.target.fetch(self.source, revision_id=revision_id)
2303
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2304
"""See InterRepository.fetch()."""
2305
from bzrlib.fetch import GenericRepoFetcher
2306
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2307
self.source, self.source._format, self.target,
2308
self.target._format)
2309
f = GenericRepoFetcher(to_repository=self.target,
2310
from_repository=self.source,
2311
last_revision=revision_id,
2313
return f.count_copied, f.failed_revisions
2316
class InterWeaveRepo(InterSameDataRepository):
2317
"""Optimised code paths between Weave based repositories.
2319
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2320
implemented lazy inter-object optimisation.
2324
def _get_repo_format_to_test(self):
2325
from bzrlib.repofmt import weaverepo
2326
return weaverepo.RepositoryFormat7()
2329
def is_compatible(source, target):
2330
"""Be compatible with known Weave formats.
2332
We don't test for the stores being of specific types because that
2333
could lead to confusing results, and there is no need to be
2336
from bzrlib.repofmt.weaverepo import (
2342
return (isinstance(source._format, (RepositoryFormat5,
2344
RepositoryFormat7)) and
2345
isinstance(target._format, (RepositoryFormat5,
2347
RepositoryFormat7)))
2348
except AttributeError:
2352
def copy_content(self, revision_id=None):
2353
"""See InterRepository.copy_content()."""
2354
# weave specific optimised path:
2356
self.target.set_make_working_trees(self.source.make_working_trees())
2357
except NotImplementedError:
2359
# FIXME do not peek!
2360
if self.source.control_files._transport.listable():
2361
pb = ui.ui_factory.nested_progress_bar()
2363
self.target.weave_store.copy_all_ids(
2364
self.source.weave_store,
2366
from_transaction=self.source.get_transaction(),
2367
to_transaction=self.target.get_transaction())
2368
pb.update('copying inventory', 0, 1)
2369
self.target.control_weaves.copy_multi(
2370
self.source.control_weaves, ['inventory'],
2371
from_transaction=self.source.get_transaction(),
2372
to_transaction=self.target.get_transaction())
2373
self.target._revision_store.text_store.copy_all_ids(
2374
self.source._revision_store.text_store,
2379
self.target.fetch(self.source, revision_id=revision_id)
2382
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2383
"""See InterRepository.fetch()."""
2384
from bzrlib.fetch import GenericRepoFetcher
2385
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2386
self.source, self.source._format, self.target, self.target._format)
2387
f = GenericRepoFetcher(to_repository=self.target,
2388
from_repository=self.source,
2389
last_revision=revision_id,
2391
return f.count_copied, f.failed_revisions
2394
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2395
"""See InterRepository.missing_revision_ids()."""
2396
# we want all revisions to satisfy revision_id in source.
2397
# but we don't want to stat every file here and there.
2398
# we want then, all revisions other needs to satisfy revision_id
2399
# checked, but not those that we have locally.
2400
# so the first thing is to get a subset of the revisions to
2401
# satisfy revision_id in source, and then eliminate those that
2402
# we do already have.
2403
# this is slow on high latency connection to self, but as as this
2404
# disk format scales terribly for push anyway due to rewriting
2405
# inventory.weave, this is considered acceptable.
2407
if revision_id is not None:
2408
source_ids = self.source.get_ancestry(revision_id)
2409
assert source_ids[0] is None
2412
source_ids = self.source._all_possible_ids()
2413
source_ids_set = set(source_ids)
2414
# source_ids is the worst possible case we may need to pull.
2415
# now we want to filter source_ids against what we actually
2416
# have in target, but don't try to check for existence where we know
2417
# we do not have a revision as that would be pointless.
2418
target_ids = set(self.target._all_possible_ids())
2419
possibly_present_revisions = target_ids.intersection(source_ids_set)
2420
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2421
required_revisions = source_ids_set.difference(actually_present_revisions)
2422
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2423
if revision_id is not None:
2424
# we used get_ancestry to determine source_ids then we are assured all
2425
# revisions referenced are present as they are installed in topological order.
2426
# and the tip revision was validated by get_ancestry.
2427
return required_topo_revisions
2429
# if we just grabbed the possibly available ids, then
2430
# we only have an estimate of whats available and need to validate
2431
# that against the revision records.
2432
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2435
class InterKnitRepo(InterSameDataRepository):
2436
"""Optimised code paths between Knit based repositories."""
2439
def _get_repo_format_to_test(self):
2440
from bzrlib.repofmt import knitrepo
2441
return knitrepo.RepositoryFormatKnit1()
2444
def is_compatible(source, target):
2445
"""Be compatible with known Knit formats.
2447
We don't test for the stores being of specific types because that
2448
could lead to confusing results, and there is no need to be
2451
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2453
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2454
isinstance(target._format, RepositoryFormatKnit))
2455
except AttributeError:
2457
return are_knits and InterRepository._same_model(source, target)
2460
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2461
"""See InterRepository.fetch()."""
2462
from bzrlib.fetch import KnitRepoFetcher
2463
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2464
self.source, self.source._format, self.target, self.target._format)
2465
f = KnitRepoFetcher(to_repository=self.target,
2466
from_repository=self.source,
2467
last_revision=revision_id,
2469
return f.count_copied, f.failed_revisions
2472
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2473
"""See InterRepository.missing_revision_ids()."""
2474
if revision_id is not None:
2475
source_ids = self.source.get_ancestry(revision_id)
2476
assert source_ids[0] is None
2479
source_ids = self.source.all_revision_ids()
2480
source_ids_set = set(source_ids)
2481
# source_ids is the worst possible case we may need to pull.
2482
# now we want to filter source_ids against what we actually
2483
# have in target, but don't try to check for existence where we know
2484
# we do not have a revision as that would be pointless.
2485
target_ids = set(self.target.all_revision_ids())
2486
possibly_present_revisions = target_ids.intersection(source_ids_set)
2487
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2488
required_revisions = source_ids_set.difference(actually_present_revisions)
2489
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2490
if revision_id is not None:
2491
# we used get_ancestry to determine source_ids then we are assured all
2492
# revisions referenced are present as they are installed in topological order.
2493
# and the tip revision was validated by get_ancestry.
2494
return required_topo_revisions
2496
# if we just grabbed the possibly available ids, then
2497
# we only have an estimate of whats available and need to validate
2498
# that against the revision records.
2499
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2502
class InterPackRepo(InterSameDataRepository):
2503
"""Optimised code paths between Pack based repositories."""
2506
def _get_repo_format_to_test(self):
2507
from bzrlib.repofmt import pack_repo
2508
return pack_repo.RepositoryFormatKnitPack1()
2511
def is_compatible(source, target):
2512
"""Be compatible with known Pack formats.
2514
We don't test for the stores being of specific types because that
2515
could lead to confusing results, and there is no need to be
2518
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2520
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2521
isinstance(target._format, RepositoryFormatPack))
2522
except AttributeError:
2524
return are_packs and InterRepository._same_model(source, target)
2527
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2528
"""See InterRepository.fetch()."""
2529
from bzrlib.repofmt.pack_repo import Packer
2530
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2531
self.source, self.source._format, self.target, self.target._format)
2532
self.count_copied = 0
2533
if revision_id is None:
2535
# everything to do - use pack logic
2536
# to fetch from all packs to one without
2537
# inventory parsing etc, IFF nothing to be copied is in the target.
2539
revision_ids = self.source.all_revision_ids()
2540
# implementing the TODO will involve:
2541
# - detecting when all of a pack is selected
2542
# - avoiding as much as possible pre-selection, so the
2543
# more-core routines such as create_pack_from_packs can filter in
2544
# a just-in-time fashion. (though having a HEADS list on a
2545
# repository might make this a lot easier, because we could
2546
# sensibly detect 'new revisions' without doing a full index scan.
2547
elif _mod_revision.is_null(revision_id):
2552
revision_ids = self.missing_revision_ids(revision_id,
2553
find_ghosts=find_ghosts)
2554
except errors.NoSuchRevision:
2555
raise errors.InstallFailed([revision_id])
2556
packs = self.source._pack_collection.all_packs()
2557
pack = Packer(self.target._pack_collection, packs, '.fetch',
2558
revision_ids).pack()
2559
if pack is not None:
2560
self.target._pack_collection._save_pack_names()
2561
# Trigger an autopack. This may duplicate effort as we've just done
2562
# a pack creation, but for now it is simpler to think about as
2563
# 'upload data, then repack if needed'.
2564
self.target._pack_collection.autopack()
2565
return (pack.get_revision_count(), [])
2570
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2571
"""See InterRepository.missing_revision_ids().
2573
:param find_ghosts: Find ghosts throughough the ancestry of
2576
if not find_ghosts and revision_id is not None:
2577
graph = self.source.get_graph()
2578
missing_revs = set()
2579
searcher = graph._make_breadth_first_searcher([revision_id])
2581
self.target._pack_collection.revision_index.combined_index
2582
null_set = frozenset([_mod_revision.NULL_REVISION])
2585
next_revs = set(searcher.next())
2586
except StopIteration:
2588
next_revs.difference_update(null_set)
2589
target_keys = [(key,) for key in next_revs]
2590
have_revs = frozenset(node[1][0] for node in
2591
target_index.iter_entries(target_keys))
2592
missing_revs.update(next_revs - have_revs)
2593
searcher.stop_searching_any(have_revs)
2594
if next_revs - have_revs == set([revision_id]):
2595
# we saw the start rev itself, but no parents from it (or
2596
# next_revs would have been updated to e.g. set(). We remove
2597
# have_revs because if we found revision_id locally we
2598
# stop_searching at the first time around.
2599
raise errors.NoSuchRevision(self.source, revision_id)
2601
elif revision_id is not None:
2602
source_ids = self.source.get_ancestry(revision_id)
2603
assert source_ids[0] is None
2606
source_ids = self.source.all_revision_ids()
2607
# source_ids is the worst possible case we may need to pull.
2608
# now we want to filter source_ids against what we actually
2609
# have in target, but don't try to check for existence where we know
2610
# we do not have a revision as that would be pointless.
2611
target_ids = set(self.target.all_revision_ids())
2612
return [r for r in source_ids if (r not in target_ids)]
2615
class InterModel1and2(InterRepository):
2618
def _get_repo_format_to_test(self):
2622
def is_compatible(source, target):
2623
if not source.supports_rich_root() and target.supports_rich_root():
2629
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2630
"""See InterRepository.fetch()."""
2631
from bzrlib.fetch import Model1toKnit2Fetcher
2632
f = Model1toKnit2Fetcher(to_repository=self.target,
2633
from_repository=self.source,
2634
last_revision=revision_id,
2636
return f.count_copied, f.failed_revisions
2639
def copy_content(self, revision_id=None):
2640
"""Make a complete copy of the content in self into destination.
2642
This is a destructive operation! Do not use it on existing
2645
:param revision_id: Only copy the content needed to construct
2646
revision_id and its parents.
2649
self.target.set_make_working_trees(self.source.make_working_trees())
2650
except NotImplementedError:
2652
# but don't bother fetching if we have the needed data now.
2653
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2654
self.target.has_revision(revision_id)):
2656
self.target.fetch(self.source, revision_id=revision_id)
2659
class InterKnit1and2(InterKnitRepo):
2662
def _get_repo_format_to_test(self):
2666
def is_compatible(source, target):
2667
"""Be compatible with Knit1 source and Knit3 target"""
2668
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2670
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2671
RepositoryFormatKnit3)
2672
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2673
RepositoryFormatKnitPack3)
2674
return (isinstance(source._format,
2675
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2676
isinstance(target._format,
2677
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2679
except AttributeError:
2683
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2684
"""See InterRepository.fetch()."""
2685
from bzrlib.fetch import Knit1to2Fetcher
2686
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2687
self.source, self.source._format, self.target,
2688
self.target._format)
2689
f = Knit1to2Fetcher(to_repository=self.target,
2690
from_repository=self.source,
2691
last_revision=revision_id,
2693
return f.count_copied, f.failed_revisions
2696
class InterDifferingSerializer(InterKnitRepo):
2699
def _get_repo_format_to_test(self):
2703
def is_compatible(source, target):
2704
"""Be compatible with Knit2 source and Knit3 target"""
2705
if source.supports_rich_root() != target.supports_rich_root():
2707
# Ideally, we'd support fetching if the source had no tree references
2708
# even if it supported them...
2709
if (getattr(source, '_format.supports_tree_reference', False) and
2710
not getattr(target, '_format.supports_tree_reference', False)):
2715
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2716
"""See InterRepository.fetch()."""
2717
revision_ids = self.target.missing_revision_ids(self.source,
2719
def revisions_iterator():
2720
for current_revision_id in revision_ids:
2721
revision = self.source.get_revision(current_revision_id)
2722
tree = self.source.revision_tree(current_revision_id)
2724
signature = self.source.get_signature_text(
2725
current_revision_id)
2726
except errors.NoSuchRevision:
2728
yield revision, tree, signature
2729
install_revisions(self.target, revisions_iterator())
2730
return len(revision_ids), 0
2733
class InterRemoteToOther(InterRepository):
2735
def __init__(self, source, target):
2736
InterRepository.__init__(self, source, target)
2737
self._real_inter = None
2740
def is_compatible(source, target):
2741
if not isinstance(source, remote.RemoteRepository):
2743
source._ensure_real()
2744
real_source = source._real_repository
2745
# Is source's model compatible with target's model, and are they the
2746
# same format? Currently we can only optimise fetching from an
2747
# identical model & format repo.
2748
assert not isinstance(real_source, remote.RemoteRepository), (
2749
"We don't support remote repos backed by remote repos yet.")
2750
return real_source._format == target._format
2753
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2754
"""See InterRepository.fetch()."""
2755
from bzrlib.fetch import RemoteToOtherFetcher
2756
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2757
self.source, self.target, self.target._format)
2758
# TODO: jam 20070210 This should be an assert, not a translate
2759
revision_id = osutils.safe_revision_id(revision_id)
2760
f = RemoteToOtherFetcher(to_repository=self.target,
2761
from_repository=self.source,
2762
last_revision=revision_id,
2764
return f.count_copied, f.failed_revisions
2767
def _get_repo_format_to_test(self):
2771
class InterOtherToRemote(InterRepository):
2773
def __init__(self, source, target):
2774
InterRepository.__init__(self, source, target)
2775
self._real_inter = None
2778
def is_compatible(source, target):
2779
if isinstance(target, remote.RemoteRepository):
2783
def _ensure_real_inter(self):
2784
if self._real_inter is None:
2785
self.target._ensure_real()
2786
real_target = self.target._real_repository
2787
self._real_inter = InterRepository.get(self.source, real_target)
2789
def copy_content(self, revision_id=None):
2790
self._ensure_real_inter()
2791
self._real_inter.copy_content(revision_id=revision_id)
2793
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2794
self._ensure_real_inter()
2795
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2798
def _get_repo_format_to_test(self):
2802
InterRepository.register_optimiser(InterDifferingSerializer)
2803
InterRepository.register_optimiser(InterSameDataRepository)
2804
InterRepository.register_optimiser(InterWeaveRepo)
2805
InterRepository.register_optimiser(InterKnitRepo)
2806
InterRepository.register_optimiser(InterModel1and2)
2807
InterRepository.register_optimiser(InterKnit1and2)
2808
InterRepository.register_optimiser(InterPackRepo)
2809
InterRepository.register_optimiser(InterRemoteToOther)
2810
InterRepository.register_optimiser(InterOtherToRemote)
2813
class CopyConverter(object):
2814
"""A repository conversion tool which just performs a copy of the content.
2816
This is slow but quite reliable.
2819
def __init__(self, target_format):
2820
"""Create a CopyConverter.
2822
:param target_format: The format the resulting repository should be.
2824
self.target_format = target_format
2826
def convert(self, repo, pb):
2827
"""Perform the conversion of to_convert, giving feedback via pb.
2829
:param to_convert: The disk object to convert.
2830
:param pb: a progress bar to use for progress information.
2835
# this is only useful with metadir layouts - separated repo content.
2836
# trigger an assertion if not such
2837
repo._format.get_format_string()
2838
self.repo_dir = repo.bzrdir
2839
self.step('Moving repository to repository.backup')
2840
self.repo_dir.transport.move('repository', 'repository.backup')
2841
backup_transport = self.repo_dir.transport.clone('repository.backup')
2842
repo._format.check_conversion_target(self.target_format)
2843
self.source_repo = repo._format.open(self.repo_dir,
2845
_override_transport=backup_transport)
2846
self.step('Creating new repository')
2847
converted = self.target_format.initialize(self.repo_dir,
2848
self.source_repo.is_shared())
2849
converted.lock_write()
2851
self.step('Copying content into repository.')
2852
self.source_repo.copy_content_into(converted)
2855
self.step('Deleting old repository content.')
2856
self.repo_dir.transport.delete_tree('repository.backup')
2857
self.pb.note('repository converted')
2859
def step(self, message):
2860
"""Update the pb by a step."""
2862
self.pb.update(message, self.count, self.total)
2874
def _unescaper(match, _map=_unescape_map):
2875
code = match.group(1)
2879
if not code.startswith('#'):
2881
return unichr(int(code[1:])).encode('utf8')
2887
def _unescape_xml(data):
2888
"""Unescape predefined XML entities in a string of data."""
2890
if _unescape_re is None:
2891
_unescape_re = re.compile('\&([^;]*);')
2892
return _unescape_re.sub(_unescaper, data)
2895
class VersionedFileChecker(object):
2897
def __init__(self, repository):
2898
self.repository = repository
2899
self.text_index = self.repository._generate_text_key_index()
2901
def calculate_file_version_parents(self, revision_id, file_id):
2902
"""Calculate the correct parents for a file version according to
2905
parent_keys = self.text_index[(file_id, revision_id)]
2906
if parent_keys == [_mod_revision.NULL_REVISION]:
2908
# strip the file_id, for the weave api
2909
return tuple([revision_id for file_id, revision_id in parent_keys])
2911
def check_file_version_parents(self, weave, file_id, planned_revisions):
2912
"""Check the parents stored in a versioned file are correct.
2914
It also detects file versions that are not referenced by their
2915
corresponding revision's inventory.
2917
:returns: A tuple of (wrong_parents, dangling_file_versions).
2918
wrong_parents is a dict mapping {revision_id: (stored_parents,
2919
correct_parents)} for each revision_id where the stored parents
2920
are not correct. dangling_file_versions is a set of (file_id,
2921
revision_id) tuples for versions that are present in this versioned
2922
file, but not used by the corresponding inventory.
2925
unused_versions = set()
2926
for num, revision_id in enumerate(planned_revisions):
2928
correct_parents = self.calculate_file_version_parents(
2929
revision_id, file_id)
2931
# we were asked to investigate a non-existant version.
2932
unused_versions.add(revision_id)
2935
knit_parents = tuple(weave.get_parents(revision_id))
2936
except errors.RevisionNotPresent:
2938
if correct_parents != knit_parents:
2939
wrong_parents[revision_id] = (knit_parents, correct_parents)
2940
return wrong_parents, unused_versions
added
removed
bzrlib/repository.py
