1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from cStringIO import StringIO
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
40
revision as _mod_revision,
46
from bzrlib.bundle import serializer
47
from bzrlib.revisiontree import RevisionTree
48
from bzrlib.store.versioned import VersionedFileStore
49
from bzrlib.store.text import TextStore
50
from bzrlib.testament import Testament
51
from bzrlib.util import bencode
54
from bzrlib.decorators import needs_read_lock, needs_write_lock
55
from bzrlib.inter import InterObject
56
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
57
from bzrlib.symbol_versioning import (
60
from bzrlib.trace import mutter, mutter_callsite, note, warning
63
# Old formats display a warning, but only once
64
_deprecation_warning_done = False
67
class CommitBuilder(object):
68
"""Provides an interface to build up a commit.
70
This allows describing a tree to be committed without needing to
71
know the internals of the format of the repository.
74
# all clients should supply tree roots.
75
record_root_entry = True
76
# the default CommitBuilder does not manage trees whose root is versioned.
77
_versioned_root = False
79
def __init__(self, repository, parents, config, timestamp=None,
80
timezone=None, committer=None, revprops=None,
82
"""Initiate a CommitBuilder.
84
:param repository: Repository to commit to.
85
:param parents: Revision ids of the parents of the new revision.
86
:param config: Configuration to use.
87
:param timestamp: Optional timestamp recorded for commit.
88
:param timezone: Optional timezone for timestamp.
89
:param committer: Optional committer to set for commit.
90
:param revprops: Optional dictionary of revision properties.
91
:param revision_id: Optional revision id.
96
self._committer = self._config.username()
98
assert isinstance(committer, basestring), type(committer)
99
self._committer = committer
101
self.new_inventory = Inventory(None)
102
self._new_revision_id = revision_id
103
self.parents = parents
104
self.repository = repository
107
if revprops is not None:
108
self._revprops.update(revprops)
110
if timestamp is None:
111
timestamp = time.time()
112
# Restrict resolution to 1ms
113
self._timestamp = round(timestamp, 3)
116
self._timezone = osutils.local_time_offset()
118
self._timezone = int(timezone)
120
self._generate_revision_if_needed()
121
self.__heads = graph.HeadsCache(repository.get_graph()).heads
123
def commit(self, message):
124
"""Make the actual commit.
126
:return: The revision id of the recorded revision.
128
rev = _mod_revision.Revision(
129
timestamp=self._timestamp,
130
timezone=self._timezone,
131
committer=self._committer,
133
inventory_sha1=self.inv_sha1,
134
revision_id=self._new_revision_id,
135
properties=self._revprops)
136
rev.parent_ids = self.parents
137
self.repository.add_revision(self._new_revision_id, rev,
138
self.new_inventory, self._config)
139
self.repository.commit_write_group()
140
return self._new_revision_id
143
"""Abort the commit that is being built.
145
self.repository.abort_write_group()
147
def revision_tree(self):
148
"""Return the tree that was just committed.
150
After calling commit() this can be called to get a RevisionTree
151
representing the newly committed tree. This is preferred to
152
calling Repository.revision_tree() because that may require
153
deserializing the inventory, while we already have a copy in
156
return RevisionTree(self.repository, self.new_inventory,
157
self._new_revision_id)
159
def finish_inventory(self):
160
"""Tell the builder that the inventory is finished."""
161
if self.new_inventory.root is None:
162
raise AssertionError('Root entry should be supplied to'
163
' record_entry_contents, as of bzr 0.10.',
164
DeprecationWarning, stacklevel=2)
165
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
166
self.new_inventory.revision_id = self._new_revision_id
167
self.inv_sha1 = self.repository.add_inventory(
168
self._new_revision_id,
173
def _gen_revision_id(self):
174
"""Return new revision-id."""
175
return generate_ids.gen_revision_id(self._config.username(),
178
def _generate_revision_if_needed(self):
179
"""Create a revision id if None was supplied.
181
If the repository can not support user-specified revision ids
182
they should override this function and raise CannotSetRevisionId
183
if _new_revision_id is not None.
185
:raises: CannotSetRevisionId
187
if self._new_revision_id is None:
188
self._new_revision_id = self._gen_revision_id()
189
self.random_revid = True
191
self.random_revid = False
193
def _heads(self, file_id, revision_ids):
194
"""Calculate the graph heads for revision_ids in the graph of file_id.
196
This can use either a per-file graph or a global revision graph as we
197
have an identity relationship between the two graphs.
199
return self.__heads(revision_ids)
201
def _check_root(self, ie, parent_invs, tree):
202
"""Helper for record_entry_contents.
204
:param ie: An entry being added.
205
:param parent_invs: The inventories of the parent revisions of the
207
:param tree: The tree that is being committed.
209
# In this revision format, root entries have no knit or weave When
210
# serializing out to disk and back in root.revision is always
212
ie.revision = self._new_revision_id
214
def _get_delta(self, ie, basis_inv, path):
215
"""Get a delta against the basis inventory for ie."""
216
if ie.file_id not in basis_inv:
218
return (None, path, ie.file_id, ie)
219
elif ie != basis_inv[ie.file_id]:
221
# TODO: avoid tis id2path call.
222
return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
227
def record_entry_contents(self, ie, parent_invs, path, tree,
229
"""Record the content of ie from tree into the commit if needed.
231
Side effect: sets ie.revision when unchanged
233
:param ie: An inventory entry present in the commit.
234
:param parent_invs: The inventories of the parent revisions of the
236
:param path: The path the entry is at in the tree.
237
:param tree: The tree which contains this entry and should be used to
239
:param content_summary: Summary data from the tree about the paths
240
content - stat, length, exec, sha/link target. This is only
241
accessed when the entry has a revision of None - that is when it is
242
a candidate to commit.
243
:return: A tuple (change_delta, version_recorded). change_delta is
244
an inventory_delta change for this entry against the basis tree of
245
the commit, or None if no change occured against the basis tree.
246
version_recorded is True if a new version of the entry has been
247
recorded. For instance, committing a merge where a file was only
248
changed on the other side will return (delta, False).
250
if self.new_inventory.root is None:
251
if ie.parent_id is not None:
252
raise errors.RootMissing()
253
self._check_root(ie, parent_invs, tree)
254
if ie.revision is None:
255
kind = content_summary[0]
257
# ie is carried over from a prior commit
259
# XXX: repository specific check for nested tree support goes here - if
260
# the repo doesn't want nested trees we skip it ?
261
if (kind == 'tree-reference' and
262
not self.repository._format.supports_tree_reference):
263
# mismatch between commit builder logic and repository:
264
# this needs the entry creation pushed down into the builder.
265
raise NotImplementedError('Missing repository subtree support.')
266
self.new_inventory.add(ie)
268
# TODO: slow, take it out of the inner loop.
270
basis_inv = parent_invs[0]
272
basis_inv = Inventory(root_id=None)
274
# ie.revision is always None if the InventoryEntry is considered
275
# for committing. We may record the previous parents revision if the
276
# content is actually unchanged against a sole head.
277
if ie.revision is not None:
278
if not self._versioned_root and path == '':
279
# repositories that do not version the root set the root's
280
# revision to the new commit even when no change occurs, and
281
# this masks when a change may have occurred against the basis,
282
# so calculate if one happened.
283
if ie.file_id in basis_inv:
284
delta = (basis_inv.id2path(ie.file_id), path,
288
delta = (None, path, ie.file_id, ie)
291
# we don't need to commit this, because the caller already
292
# determined that an existing revision of this file is
294
return None, (ie.revision == self._new_revision_id)
295
# XXX: Friction: parent_candidates should return a list not a dict
296
# so that we don't have to walk the inventories again.
297
parent_candiate_entries = ie.parent_candidates(parent_invs)
298
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
300
for inv in parent_invs:
301
if ie.file_id in inv:
302
old_rev = inv[ie.file_id].revision
303
if old_rev in head_set:
304
heads.append(inv[ie.file_id].revision)
305
head_set.remove(inv[ie.file_id].revision)
308
# now we check to see if we need to write a new record to the
310
# We write a new entry unless there is one head to the ancestors, and
311
# the kind-derived content is unchanged.
313
# Cheapest check first: no ancestors, or more the one head in the
314
# ancestors, we write a new node.
318
# There is a single head, look it up for comparison
319
parent_entry = parent_candiate_entries[heads[0]]
320
# if the non-content specific data has changed, we'll be writing a
322
if (parent_entry.parent_id != ie.parent_id or
323
parent_entry.name != ie.name):
325
# now we need to do content specific checks:
327
# if the kind changed the content obviously has
328
if kind != parent_entry.kind:
331
assert content_summary[2] is not None, \
332
"Files must not have executable = None"
334
if (# if the file length changed we have to store:
335
parent_entry.text_size != content_summary[1] or
336
# if the exec bit has changed we have to store:
337
parent_entry.executable != content_summary[2]):
339
elif parent_entry.text_sha1 == content_summary[3]:
340
# all meta and content is unchanged (using a hash cache
341
# hit to check the sha)
342
ie.revision = parent_entry.revision
343
ie.text_size = parent_entry.text_size
344
ie.text_sha1 = parent_entry.text_sha1
345
ie.executable = parent_entry.executable
346
return self._get_delta(ie, basis_inv, path), False
348
# Either there is only a hash change(no hash cache entry,
349
# or same size content change), or there is no change on
351
# Provide the parent's hash to the store layer, so that the
352
# content is unchanged we will not store a new node.
353
nostore_sha = parent_entry.text_sha1
355
# We want to record a new node regardless of the presence or
356
# absence of a content change in the file.
358
ie.executable = content_summary[2]
359
lines = tree.get_file(ie.file_id, path).readlines()
361
ie.text_sha1, ie.text_size = self._add_text_to_weave(
362
ie.file_id, lines, heads, nostore_sha)
363
except errors.ExistingContent:
364
# Turns out that the file content was unchanged, and we were
365
# only going to store a new node if it was changed. Carry over
367
ie.revision = parent_entry.revision
368
ie.text_size = parent_entry.text_size
369
ie.text_sha1 = parent_entry.text_sha1
370
ie.executable = parent_entry.executable
371
return self._get_delta(ie, basis_inv, path), False
372
elif kind == 'directory':
374
# all data is meta here, nothing specific to directory, so
376
ie.revision = parent_entry.revision
377
return self._get_delta(ie, basis_inv, path), False
379
self._add_text_to_weave(ie.file_id, lines, heads, None)
380
elif kind == 'symlink':
381
current_link_target = content_summary[3]
383
# symlink target is not generic metadata, check if it has
385
if current_link_target != parent_entry.symlink_target:
388
# unchanged, carry over.
389
ie.revision = parent_entry.revision
390
ie.symlink_target = parent_entry.symlink_target
391
return self._get_delta(ie, basis_inv, path), False
392
ie.symlink_target = current_link_target
394
self._add_text_to_weave(ie.file_id, lines, heads, None)
395
elif kind == 'tree-reference':
397
if content_summary[3] != parent_entry.reference_revision:
400
# unchanged, carry over.
401
ie.reference_revision = parent_entry.reference_revision
402
ie.revision = parent_entry.revision
403
return self._get_delta(ie, basis_inv, path), False
404
ie.reference_revision = content_summary[3]
406
self._add_text_to_weave(ie.file_id, lines, heads, None)
408
raise NotImplementedError('unknown kind')
409
ie.revision = self._new_revision_id
410
return self._get_delta(ie, basis_inv, path), True
412
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
413
versionedfile = self.repository.weave_store.get_weave_or_empty(
414
file_id, self.repository.get_transaction())
415
# Don't change this to add_lines - add_lines_with_ghosts is cheaper
416
# than add_lines, and allows committing when a parent is ghosted for
418
# Note: as we read the content directly from the tree, we know its not
419
# been turned into unicode or badly split - but a broken tree
420
# implementation could give us bad output from readlines() so this is
421
# not a guarantee of safety. What would be better is always checking
422
# the content during test suite execution. RBC 20070912
424
return versionedfile.add_lines_with_ghosts(
425
self._new_revision_id, parents, new_lines,
426
nostore_sha=nostore_sha, random_id=self.random_revid,
427
check_content=False)[0:2]
429
versionedfile.clear_cache()
432
class RootCommitBuilder(CommitBuilder):
433
"""This commitbuilder actually records the root id"""
435
# the root entry gets versioned properly by this builder.
436
_versioned_root = True
438
def _check_root(self, ie, parent_invs, tree):
439
"""Helper for record_entry_contents.
441
:param ie: An entry being added.
442
:param parent_invs: The inventories of the parent revisions of the
444
:param tree: The tree that is being committed.
448
######################################################################
451
class Repository(object):
452
"""Repository holding history for one or more branches.
454
The repository holds and retrieves historical information including
455
revisions and file history. It's normally accessed only by the Branch,
456
which views a particular line of development through that history.
458
The Repository builds on top of Stores and a Transport, which respectively
459
describe the disk data format and the way of accessing the (possibly
463
# What class to use for a CommitBuilder. Often its simpler to change this
464
# in a Repository class subclass rather than to override
465
# get_commit_builder.
466
_commit_builder_class = CommitBuilder
467
# The search regex used by xml based repositories to determine what things
468
# where changed in a single commit.
469
_file_ids_altered_regex = lazy_regex.lazy_compile(
470
r'file_id="(?P<file_id>[^"]+)"'
471
r'.* revision="(?P<revision_id>[^"]+)"'
474
def abort_write_group(self):
475
"""Commit the contents accrued within the current write group.
477
:seealso: start_write_group.
479
if self._write_group is not self.get_transaction():
480
# has an unlock or relock occured ?
481
raise errors.BzrError('mismatched lock context and write group.')
482
self._abort_write_group()
483
self._write_group = None
485
def _abort_write_group(self):
486
"""Template method for per-repository write group cleanup.
488
This is called during abort before the write group is considered to be
489
finished and should cleanup any internal state accrued during the write
490
group. There is no requirement that data handed to the repository be
491
*not* made available - this is not a rollback - but neither should any
492
attempt be made to ensure that data added is fully commited. Abort is
493
invoked when an error has occured so futher disk or network operations
494
may not be possible or may error and if possible should not be
499
def add_inventory(self, revision_id, inv, parents):
500
"""Add the inventory inv to the repository as revision_id.
502
:param parents: The revision ids of the parents that revision_id
503
is known to have and are in the repository already.
505
returns the sha1 of the serialized inventory.
507
assert self.is_in_write_group()
508
_mod_revision.check_not_reserved_id(revision_id)
509
assert inv.revision_id is None or inv.revision_id == revision_id, \
510
"Mismatch between inventory revision" \
511
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
512
assert inv.root is not None
513
inv_lines = self._serialise_inventory_to_lines(inv)
514
inv_vf = self.get_inventory_weave()
515
return self._inventory_add_lines(inv_vf, revision_id, parents,
516
inv_lines, check_content=False)
518
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
520
"""Store lines in inv_vf and return the sha1 of the inventory."""
522
for parent in parents:
524
final_parents.append(parent)
525
return inv_vf.add_lines(revision_id, final_parents, lines,
526
check_content=check_content)[0]
529
def add_revision(self, revision_id, rev, inv=None, config=None):
530
"""Add rev to the revision store as revision_id.
532
:param revision_id: the revision id to use.
533
:param rev: The revision object.
534
:param inv: The inventory for the revision. if None, it will be looked
535
up in the inventory storer
536
:param config: If None no digital signature will be created.
537
If supplied its signature_needed method will be used
538
to determine if a signature should be made.
540
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
542
_mod_revision.check_not_reserved_id(revision_id)
543
if config is not None and config.signature_needed():
545
inv = self.get_inventory(revision_id)
546
plaintext = Testament(rev, inv).as_short_text()
547
self.store_revision_signature(
548
gpg.GPGStrategy(config), plaintext, revision_id)
549
if not revision_id in self.get_inventory_weave():
551
raise errors.WeaveRevisionNotPresent(revision_id,
552
self.get_inventory_weave())
554
# yes, this is not suitable for adding with ghosts.
555
self.add_inventory(revision_id, inv, rev.parent_ids)
556
self._revision_store.add_revision(rev, self.get_transaction())
558
def _add_revision_text(self, revision_id, text):
559
revision = self._revision_store._serializer.read_revision_from_string(
561
self._revision_store._add_revision(revision, StringIO(text),
562
self.get_transaction())
564
def all_revision_ids(self):
565
"""Returns a list of all the revision ids in the repository.
567
This is deprecated because code should generally work on the graph
568
reachable from a particular revision, and ignore any other revisions
569
that might be present. There is no direct replacement method.
571
if 'evil' in debug.debug_flags:
572
mutter_callsite(2, "all_revision_ids is linear with history.")
573
return self._all_revision_ids()
575
def _all_revision_ids(self):
576
"""Returns a list of all the revision ids in the repository.
578
These are in as much topological order as the underlying store can
581
raise NotImplementedError(self._all_revision_ids)
583
def break_lock(self):
584
"""Break a lock if one is present from another instance.
586
Uses the ui factory to ask for confirmation if the lock may be from
589
self.control_files.break_lock()
592
def _eliminate_revisions_not_present(self, revision_ids):
593
"""Check every revision id in revision_ids to see if we have it.
595
Returns a set of the present revisions.
598
for id in revision_ids:
599
if self.has_revision(id):
604
def create(a_bzrdir):
605
"""Construct the current default format repository in a_bzrdir."""
606
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
608
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
609
"""instantiate a Repository.
611
:param _format: The format of the repository on disk.
612
:param a_bzrdir: The BzrDir of the repository.
614
In the future we will have a single api for all stores for
615
getting file texts, inventories and revisions, then
616
this construct will accept instances of those things.
618
super(Repository, self).__init__()
619
self._format = _format
620
# the following are part of the public API for Repository:
621
self.bzrdir = a_bzrdir
622
self.control_files = control_files
623
self._revision_store = _revision_store
624
# backwards compatibility
625
self.weave_store = text_store
627
self._reconcile_does_inventory_gc = True
628
self._reconcile_fixes_text_parents = False
629
self._reconcile_backsup_inventory = True
630
# not right yet - should be more semantically clear ?
632
self.control_store = control_store
633
self.control_weaves = control_store
634
# TODO: make sure to construct the right store classes, etc, depending
635
# on whether escaping is required.
636
self._warn_if_deprecated()
637
self._write_group = None
638
self.base = control_files._transport.base
641
return '%s(%r)' % (self.__class__.__name__,
644
def has_same_location(self, other):
645
"""Returns a boolean indicating if this repository is at the same
646
location as another repository.
648
This might return False even when two repository objects are accessing
649
the same physical repository via different URLs.
651
if self.__class__ is not other.__class__:
653
return (self.control_files._transport.base ==
654
other.control_files._transport.base)
656
def is_in_write_group(self):
657
"""Return True if there is an open write group.
659
:seealso: start_write_group.
661
return self._write_group is not None
664
return self.control_files.is_locked()
666
def is_write_locked(self):
667
"""Return True if this object is write locked."""
668
return self.is_locked() and self.control_files._lock_mode == 'w'
670
def lock_write(self, token=None):
671
"""Lock this repository for writing.
673
This causes caching within the repository obejct to start accumlating
674
data during reads, and allows a 'write_group' to be obtained. Write
675
groups must be used for actual data insertion.
677
:param token: if this is already locked, then lock_write will fail
678
unless the token matches the existing lock.
679
:returns: a token if this instance supports tokens, otherwise None.
680
:raises TokenLockingNotSupported: when a token is given but this
681
instance doesn't support using token locks.
682
:raises MismatchedToken: if the specified token doesn't match the token
683
of the existing lock.
684
:seealso: start_write_group.
686
A token should be passed in if you know that you have locked the object
687
some other way, and need to synchronise this object's state with that
690
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
692
result = self.control_files.lock_write(token=token)
697
self.control_files.lock_read()
700
def get_physical_lock_status(self):
701
return self.control_files.get_physical_lock_status()
703
def leave_lock_in_place(self):
704
"""Tell this repository not to release the physical lock when this
707
If lock_write doesn't return a token, then this method is not supported.
709
self.control_files.leave_in_place()
711
def dont_leave_lock_in_place(self):
712
"""Tell this repository to release the physical lock when this
713
object is unlocked, even if it didn't originally acquire it.
715
If lock_write doesn't return a token, then this method is not supported.
717
self.control_files.dont_leave_in_place()
720
def gather_stats(self, revid=None, committers=None):
721
"""Gather statistics from a revision id.
723
:param revid: The revision id to gather statistics from, if None, then
724
no revision specific statistics are gathered.
725
:param committers: Optional parameter controlling whether to grab
726
a count of committers from the revision specific statistics.
727
:return: A dictionary of statistics. Currently this contains:
728
committers: The number of committers if requested.
729
firstrev: A tuple with timestamp, timezone for the penultimate left
730
most ancestor of revid, if revid is not the NULL_REVISION.
731
latestrev: A tuple with timestamp, timezone for revid, if revid is
732
not the NULL_REVISION.
733
revisions: The total revision count in the repository.
734
size: An estimate disk size of the repository in bytes.
737
if revid and committers:
738
result['committers'] = 0
739
if revid and revid != _mod_revision.NULL_REVISION:
741
all_committers = set()
742
revisions = self.get_ancestry(revid)
743
# pop the leading None
745
first_revision = None
747
# ignore the revisions in the middle - just grab first and last
748
revisions = revisions[0], revisions[-1]
749
for revision in self.get_revisions(revisions):
750
if not first_revision:
751
first_revision = revision
753
all_committers.add(revision.committer)
754
last_revision = revision
756
result['committers'] = len(all_committers)
757
result['firstrev'] = (first_revision.timestamp,
758
first_revision.timezone)
759
result['latestrev'] = (last_revision.timestamp,
760
last_revision.timezone)
762
# now gather global repository information
763
if self.bzrdir.root_transport.listable():
764
c, t = self._revision_store.total_size(self.get_transaction())
765
result['revisions'] = c
769
def find_branches(self, using=False):
770
"""Find branches underneath this repository.
772
This will include branches inside other branches.
774
:param using: If True, list only branches using this repository.
777
class Evaluator(object):
780
self.first_call = True
782
def __call__(self, bzrdir):
783
# On the first call, the parameter is always the bzrdir
784
# containing the current repo.
785
if not self.first_call:
787
repository = bzrdir.open_repository()
788
except errors.NoRepositoryPresent:
791
return False, (None, repository)
792
self.first_call = False
794
value = (bzrdir.open_branch(), None)
795
except errors.NotBranchError:
800
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
801
self.bzrdir.root_transport, evaluate=Evaluator()):
802
if branch is not None:
803
branches.append(branch)
804
if not using and repository is not None:
805
branches.extend(repository.find_branches())
808
def get_data_stream(self, revision_ids):
809
raise NotImplementedError(self.get_data_stream)
811
def insert_data_stream(self, stream):
812
"""XXX What does this really do?
814
Is it a substitute for fetch?
815
Should it manage its own write group ?
817
for item_key, bytes in stream:
818
if item_key[0] == 'file':
819
(file_id,) = item_key[1:]
820
knit = self.weave_store.get_weave_or_empty(
821
file_id, self.get_transaction())
822
elif item_key == ('inventory',):
823
knit = self.get_inventory_weave()
824
elif item_key == ('revisions',):
825
knit = self._revision_store.get_revision_file(
826
self.get_transaction())
827
elif item_key == ('signatures',):
828
knit = self._revision_store.get_signature_file(
829
self.get_transaction())
831
raise RepositoryDataStreamError(
832
"Unrecognised data stream key '%s'" % (item_key,))
833
decoded_list = bencode.bdecode(bytes)
834
format = decoded_list.pop(0)
837
for version, options, parents, some_bytes in decoded_list:
838
data_list.append((version, options, len(some_bytes), parents))
839
knit_bytes += some_bytes
840
knit.insert_data_stream(
841
(format, data_list, StringIO(knit_bytes).read))
844
def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
845
"""Return the revision ids that other has that this does not.
847
These are returned in topological order.
849
revision_id: only return revision ids included by revision_id.
851
return InterRepository.get(other, self).missing_revision_ids(
852
revision_id, find_ghosts)
856
"""Open the repository rooted at base.
858
For instance, if the repository is at URL/.bzr/repository,
859
Repository.open(URL) -> a Repository instance.
861
control = bzrdir.BzrDir.open(base)
862
return control.open_repository()
864
def copy_content_into(self, destination, revision_id=None):
865
"""Make a complete copy of the content in self into destination.
867
This is a destructive operation! Do not use it on existing
870
return InterRepository.get(self, destination).copy_content(revision_id)
872
def commit_write_group(self):
873
"""Commit the contents accrued within the current write group.
875
:seealso: start_write_group.
877
if self._write_group is not self.get_transaction():
878
# has an unlock or relock occured ?
879
raise errors.BzrError('mismatched lock context %r and '
881
(self.get_transaction(), self._write_group))
882
self._commit_write_group()
883
self._write_group = None
885
def _commit_write_group(self):
886
"""Template method for per-repository write group cleanup.
888
This is called before the write group is considered to be
889
finished and should ensure that all data handed to the repository
890
for writing during the write group is safely committed (to the
891
extent possible considering file system caching etc).
894
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
895
"""Fetch the content required to construct revision_id from source.
897
If revision_id is None all content is copied.
898
:param find_ghosts: Find and copy revisions in the source that are
899
ghosts in the target (and not reachable directly by walking out to
900
the first-present revision in target from revision_id).
902
# fast path same-url fetch operations
903
if self.has_same_location(source):
904
# check that last_revision is in 'from' and then return a
906
if (revision_id is not None and
907
not _mod_revision.is_null(revision_id)):
908
self.get_revision(revision_id)
910
inter = InterRepository.get(source, self)
912
return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
913
except NotImplementedError:
914
raise errors.IncompatibleRepositories(source, self)
916
def create_bundle(self, target, base, fileobj, format=None):
917
return serializer.write_bundle(self, target, base, fileobj, format)
919
def get_commit_builder(self, branch, parents, config, timestamp=None,
920
timezone=None, committer=None, revprops=None,
922
"""Obtain a CommitBuilder for this repository.
924
:param branch: Branch to commit to.
925
:param parents: Revision ids of the parents of the new revision.
926
:param config: Configuration to use.
927
:param timestamp: Optional timestamp recorded for commit.
928
:param timezone: Optional timezone for timestamp.
929
:param committer: Optional committer to set for commit.
930
:param revprops: Optional dictionary of revision properties.
931
:param revision_id: Optional revision id.
933
result = self._commit_builder_class(self, parents, config,
934
timestamp, timezone, committer, revprops, revision_id)
935
self.start_write_group()
939
if (self.control_files._lock_count == 1 and
940
self.control_files._lock_mode == 'w'):
941
if self._write_group is not None:
942
self.abort_write_group()
943
self.control_files.unlock()
944
raise errors.BzrError(
945
'Must end write groups before releasing write locks.')
946
self.control_files.unlock()
949
def clone(self, a_bzrdir, revision_id=None):
950
"""Clone this repository into a_bzrdir using the current format.
952
Currently no check is made that the format of this repository and
953
the bzrdir format are compatible. FIXME RBC 20060201.
955
:return: The newly created destination repository.
957
# TODO: deprecate after 0.16; cloning this with all its settings is
958
# probably not very useful -- mbp 20070423
959
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
960
self.copy_content_into(dest_repo, revision_id)
963
def start_write_group(self):
964
"""Start a write group in the repository.
966
Write groups are used by repositories which do not have a 1:1 mapping
967
between file ids and backend store to manage the insertion of data from
968
both fetch and commit operations.
970
A write lock is required around the start_write_group/commit_write_group
971
for the support of lock-requiring repository formats.
973
One can only insert data into a repository inside a write group.
977
if not self.is_write_locked():
978
raise errors.NotWriteLocked(self)
979
if self._write_group:
980
raise errors.BzrError('already in a write group')
981
self._start_write_group()
982
# so we can detect unlock/relock - the write group is now entered.
983
self._write_group = self.get_transaction()
985
def _start_write_group(self):
986
"""Template method for per-repository write group startup.
988
This is called before the write group is considered to be
993
def sprout(self, to_bzrdir, revision_id=None):
994
"""Create a descendent repository for new development.
996
Unlike clone, this does not copy the settings of the repository.
998
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
999
dest_repo.fetch(self, revision_id=revision_id)
1002
def _create_sprouting_repo(self, a_bzrdir, shared):
1003
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1004
# use target default format.
1005
dest_repo = a_bzrdir.create_repository()
1007
# Most control formats need the repository to be specifically
1008
# created, but on some old all-in-one formats it's not needed
1010
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1011
except errors.UninitializableFormat:
1012
dest_repo = a_bzrdir.open_repository()
1016
def has_revision(self, revision_id):
1017
"""True if this repository has a copy of the revision."""
1018
if 'evil' in debug.debug_flags:
1019
mutter_callsite(3, "has_revision is a LBYL symptom.")
1020
return self._revision_store.has_revision_id(revision_id,
1021
self.get_transaction())
1024
def get_revision(self, revision_id):
1025
"""Return the Revision object for a named revision."""
1026
return self.get_revisions([revision_id])[0]
1029
def get_revision_reconcile(self, revision_id):
1030
"""'reconcile' helper routine that allows access to a revision always.
1032
This variant of get_revision does not cross check the weave graph
1033
against the revision one as get_revision does: but it should only
1034
be used by reconcile, or reconcile-alike commands that are correcting
1035
or testing the revision graph.
1037
return self._get_revisions([revision_id])[0]
1040
def get_revisions(self, revision_ids):
1041
"""Get many revisions at once."""
1042
return self._get_revisions(revision_ids)
1045
def _get_revisions(self, revision_ids):
1046
"""Core work logic to get many revisions without sanity checks."""
1047
for rev_id in revision_ids:
1048
if not rev_id or not isinstance(rev_id, basestring):
1049
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1050
revs = self._revision_store.get_revisions(revision_ids,
1051
self.get_transaction())
1053
assert not isinstance(rev.revision_id, unicode)
1054
for parent_id in rev.parent_ids:
1055
assert not isinstance(parent_id, unicode)
1059
def get_revision_xml(self, revision_id):
1060
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1061
# would have already do it.
1062
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1063
rev = self.get_revision(revision_id)
1064
rev_tmp = StringIO()
1065
# the current serializer..
1066
self._revision_store._serializer.write_revision(rev, rev_tmp)
1068
return rev_tmp.getvalue()
1071
def get_deltas_for_revisions(self, revisions):
1072
"""Produce a generator of revision deltas.
1074
Note that the input is a sequence of REVISIONS, not revision_ids.
1075
Trees will be held in memory until the generator exits.
1076
Each delta is relative to the revision's lefthand predecessor.
1078
required_trees = set()
1079
for revision in revisions:
1080
required_trees.add(revision.revision_id)
1081
required_trees.update(revision.parent_ids[:1])
1082
trees = dict((t.get_revision_id(), t) for
1083
t in self.revision_trees(required_trees))
1084
for revision in revisions:
1085
if not revision.parent_ids:
1086
old_tree = self.revision_tree(None)
1088
old_tree = trees[revision.parent_ids[0]]
1089
yield trees[revision.revision_id].changes_from(old_tree)
1092
def get_revision_delta(self, revision_id):
1093
"""Return the delta for one revision.
1095
The delta is relative to the left-hand predecessor of the
1098
r = self.get_revision(revision_id)
1099
return list(self.get_deltas_for_revisions([r]))[0]
1102
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1103
signature = gpg_strategy.sign(plaintext)
1104
self.add_signature_text(revision_id, signature)
1107
def add_signature_text(self, revision_id, signature):
1108
self._revision_store.add_revision_signature_text(revision_id,
1110
self.get_transaction())
1112
def find_text_key_references(self):
1113
"""Find the text key references within the repository.
1115
:return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1116
revision_ids. Each altered file-ids has the exact revision_ids that
1117
altered it listed explicitly.
1118
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1119
to whether they were referred to by the inventory of the
1120
revision_id that they contain. The inventory texts from all present
1121
revision ids are assessed to generate this report.
1123
revision_ids = self.all_revision_ids()
1124
w = self.get_inventory_weave()
1125
pb = ui.ui_factory.nested_progress_bar()
1127
return self._find_text_key_references_from_xml_inventory_lines(
1128
w.iter_lines_added_or_present_in_versions(revision_ids, pb=pb))
1132
def _find_text_key_references_from_xml_inventory_lines(self,
1134
"""Core routine for extracting references to texts from inventories.
1136
This performs the translation of xml lines to revision ids.
1138
:param line_iterator: An iterator of lines, origin_version_id
1139
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1140
to whether they were referred to by the inventory of the
1141
revision_id that they contain. Note that if that revision_id was
1142
not part of the line_iterator's output then False will be given -
1143
even though it may actually refer to that key.
1145
if not self._serializer.support_altered_by_hack:
1146
raise AssertionError(
1147
"_find_text_key_references_from_xml_inventory_lines only "
1148
"supported for branches which store inventory as unnested xml"
1149
", not on %r" % self)
1152
# this code needs to read every new line in every inventory for the
1153
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1154
# not present in one of those inventories is unnecessary but not
1155
# harmful because we are filtering by the revision id marker in the
1156
# inventory lines : we only select file ids altered in one of those
1157
# revisions. We don't need to see all lines in the inventory because
1158
# only those added in an inventory in rev X can contain a revision=X
1160
unescape_revid_cache = {}
1161
unescape_fileid_cache = {}
1163
# jam 20061218 In a big fetch, this handles hundreds of thousands
1164
# of lines, so it has had a lot of inlining and optimizing done.
1165
# Sorry that it is a little bit messy.
1166
# Move several functions to be local variables, since this is a long
1168
search = self._file_ids_altered_regex.search
1169
unescape = _unescape_xml
1170
setdefault = result.setdefault
1171
for line, version_id in line_iterator:
1172
match = search(line)
1175
# One call to match.group() returning multiple items is quite a
1176
# bit faster than 2 calls to match.group() each returning 1
1177
file_id, revision_id = match.group('file_id', 'revision_id')
1179
# Inlining the cache lookups helps a lot when you make 170,000
1180
# lines and 350k ids, versus 8.4 unique ids.
1181
# Using a cache helps in 2 ways:
1182
# 1) Avoids unnecessary decoding calls
1183
# 2) Re-uses cached strings, which helps in future set and
1185
# (2) is enough that removing encoding entirely along with
1186
# the cache (so we are using plain strings) results in no
1187
# performance improvement.
1189
revision_id = unescape_revid_cache[revision_id]
1191
unescaped = unescape(revision_id)
1192
unescape_revid_cache[revision_id] = unescaped
1193
revision_id = unescaped
1195
# Note that unconditionally unescaping means that we deserialise
1196
# every fileid, which for general 'pull' is not great, but we don't
1197
# really want to have some many fulltexts that this matters anyway.
1200
file_id = unescape_fileid_cache[file_id]
1202
unescaped = unescape(file_id)
1203
unescape_fileid_cache[file_id] = unescaped
1206
key = (file_id, revision_id)
1207
setdefault(key, False)
1208
if revision_id == version_id:
1212
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1214
"""Helper routine for fileids_altered_by_revision_ids.
1216
This performs the translation of xml lines to revision ids.
1218
:param line_iterator: An iterator of lines, origin_version_id
1219
:param revision_ids: The revision ids to filter for. This should be a
1220
set or other type which supports efficient __contains__ lookups, as
1221
the revision id from each parsed line will be looked up in the
1222
revision_ids filter.
1223
:return: a dictionary mapping altered file-ids to an iterable of
1224
revision_ids. Each altered file-ids has the exact revision_ids that
1225
altered it listed explicitly.
1228
setdefault = result.setdefault
1229
for file_id, revision_id in \
1230
self._find_text_key_references_from_xml_inventory_lines(
1231
line_iterator).iterkeys():
1232
# once data is all ensured-consistent; then this is
1233
# if revision_id == version_id
1234
if revision_id in revision_ids:
1235
setdefault(file_id, set()).add(revision_id)
1238
def fileids_altered_by_revision_ids(self, revision_ids):
1239
"""Find the file ids and versions affected by revisions.
1241
:param revisions: an iterable containing revision ids.
1242
:return: a dictionary mapping altered file-ids to an iterable of
1243
revision_ids. Each altered file-ids has the exact revision_ids that
1244
altered it listed explicitly.
1246
selected_revision_ids = set(revision_ids)
1247
w = self.get_inventory_weave()
1248
pb = ui.ui_factory.nested_progress_bar()
1250
return self._find_file_ids_from_xml_inventory_lines(
1251
w.iter_lines_added_or_present_in_versions(
1252
selected_revision_ids, pb=pb),
1253
selected_revision_ids)
1257
def iter_files_bytes(self, desired_files):
1258
"""Iterate through file versions.
1260
Files will not necessarily be returned in the order they occur in
1261
desired_files. No specific order is guaranteed.
1263
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1264
value supplied by the caller as part of desired_files. It should
1265
uniquely identify the file version in the caller's context. (Examples:
1266
an index number or a TreeTransform trans_id.)
1268
bytes_iterator is an iterable of bytestrings for the file. The
1269
kind of iterable and length of the bytestrings are unspecified, but for
1270
this implementation, it is a list of lines produced by
1271
VersionedFile.get_lines().
1273
:param desired_files: a list of (file_id, revision_id, identifier)
1276
transaction = self.get_transaction()
1277
for file_id, revision_id, callable_data in desired_files:
1279
weave = self.weave_store.get_weave(file_id, transaction)
1280
except errors.NoSuchFile:
1281
raise errors.NoSuchIdInRepository(self, file_id)
1282
yield callable_data, weave.get_lines(revision_id)
1284
def _generate_text_key_index(self, text_key_references=None,
1286
"""Generate a new text key index for the repository.
1288
This is an expensive function that will take considerable time to run.
1290
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1291
list of parents, also text keys. When a given key has no parents,
1292
the parents list will be [NULL_REVISION].
1294
# All revisions, to find inventory parents.
1295
if ancestors is None:
1296
revision_graph = self.get_revision_graph_with_ghosts()
1297
ancestors = revision_graph.get_ancestors()
1298
if text_key_references is None:
1299
text_key_references = self.find_text_key_references()
1300
pb = ui.ui_factory.nested_progress_bar()
1302
return self._do_generate_text_key_index(ancestors,
1303
text_key_references, pb)
1307
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1308
"""Helper for _generate_text_key_index to avoid deep nesting."""
1309
revision_order = tsort.topo_sort(ancestors)
1310
invalid_keys = set()
1312
for revision_id in revision_order:
1313
revision_keys[revision_id] = set()
1314
text_count = len(text_key_references)
1315
# a cache of the text keys to allow reuse; costs a dict of all the
1316
# keys, but saves a 2-tuple for every child of a given key.
1318
for text_key, valid in text_key_references.iteritems():
1320
invalid_keys.add(text_key)
1322
revision_keys[text_key[1]].add(text_key)
1323
text_key_cache[text_key] = text_key
1324
del text_key_references
1326
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1327
NULL_REVISION = _mod_revision.NULL_REVISION
1328
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
1329
# too small for large or very branchy trees. However, for 55K path
1330
# trees, it would be easy to use too much memory trivially. Ideally we
1331
# could gauge this by looking at available real memory etc, but this is
1332
# always a tricky proposition.
1333
inventory_cache = lru_cache.LRUCache(10)
1334
batch_size = 10 # should be ~150MB on a 55K path tree
1335
batch_count = len(revision_order) / batch_size + 1
1337
pb.update("Calculating text parents.", processed_texts, text_count)
1338
for offset in xrange(batch_count):
1339
to_query = revision_order[offset * batch_size:(offset + 1) *
1343
for rev_tree in self.revision_trees(to_query):
1344
revision_id = rev_tree.get_revision_id()
1345
parent_ids = ancestors[revision_id]
1346
for text_key in revision_keys[revision_id]:
1347
pb.update("Calculating text parents.", processed_texts)
1348
processed_texts += 1
1349
candidate_parents = []
1350
for parent_id in parent_ids:
1351
parent_text_key = (text_key[0], parent_id)
1353
check_parent = parent_text_key not in \
1354
revision_keys[parent_id]
1356
# the parent parent_id is a ghost:
1357
check_parent = False
1358
# truncate the derived graph against this ghost.
1359
parent_text_key = None
1361
# look at the parent commit details inventories to
1362
# determine possible candidates in the per file graph.
1365
inv = inventory_cache[parent_id]
1367
inv = self.revision_tree(parent_id).inventory
1368
inventory_cache[parent_id] = inv
1369
parent_entry = inv._byid.get(text_key[0], None)
1370
if parent_entry is not None:
1372
text_key[0], parent_entry.revision)
1374
parent_text_key = None
1375
if parent_text_key is not None:
1376
candidate_parents.append(
1377
text_key_cache[parent_text_key])
1378
parent_heads = text_graph.heads(candidate_parents)
1379
new_parents = list(parent_heads)
1380
new_parents.sort(key=lambda x:candidate_parents.index(x))
1381
if new_parents == []:
1382
new_parents = [NULL_REVISION]
1383
text_index[text_key] = new_parents
1385
for text_key in invalid_keys:
1386
text_index[text_key] = [NULL_REVISION]
1389
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1390
"""Get an iterable listing the keys of all the data introduced by a set
1393
The keys will be ordered so that the corresponding items can be safely
1394
fetched and inserted in that order.
1396
:returns: An iterable producing tuples of (knit-kind, file-id,
1397
versions). knit-kind is one of 'file', 'inventory', 'signatures',
1398
'revisions'. file-id is None unless knit-kind is 'file'.
1400
# XXX: it's a bit weird to control the inventory weave caching in this
1401
# generator. Ideally the caching would be done in fetch.py I think. Or
1402
# maybe this generator should explicitly have the contract that it
1403
# should not be iterated until the previously yielded item has been
1406
inv_w = self.get_inventory_weave()
1407
inv_w.enable_cache()
1409
# file ids that changed
1410
file_ids = self.fileids_altered_by_revision_ids(revision_ids)
1412
num_file_ids = len(file_ids)
1413
for file_id, altered_versions in file_ids.iteritems():
1414
if _files_pb is not None:
1415
_files_pb.update("fetch texts", count, num_file_ids)
1417
yield ("file", file_id, altered_versions)
1418
# We're done with the files_pb. Note that it finished by the caller,
1419
# just as it was created by the caller.
1423
yield ("inventory", None, revision_ids)
1427
revisions_with_signatures = set()
1428
for rev_id in revision_ids:
1430
self.get_signature_text(rev_id)
1431
except errors.NoSuchRevision:
1435
revisions_with_signatures.add(rev_id)
1437
yield ("signatures", None, revisions_with_signatures)
1440
yield ("revisions", None, revision_ids)
1443
def get_inventory_weave(self):
1444
return self.control_weaves.get_weave('inventory',
1445
self.get_transaction())
1448
def get_inventory(self, revision_id):
1449
"""Get Inventory object by hash."""
1450
return self.deserialise_inventory(
1451
revision_id, self.get_inventory_xml(revision_id))
1453
def deserialise_inventory(self, revision_id, xml):
1454
"""Transform the xml into an inventory object.
1456
:param revision_id: The expected revision id of the inventory.
1457
:param xml: A serialised inventory.
1459
return self._serializer.read_inventory_from_string(xml, revision_id)
1461
def serialise_inventory(self, inv):
1462
return self._serializer.write_inventory_to_string(inv)
1464
def _serialise_inventory_to_lines(self, inv):
1465
return self._serializer.write_inventory_to_lines(inv)
1467
def get_serializer_format(self):
1468
return self._serializer.format_num
1471
def get_inventory_xml(self, revision_id):
1472
"""Get inventory XML as a file object."""
1474
assert isinstance(revision_id, str), type(revision_id)
1475
iw = self.get_inventory_weave()
1476
return iw.get_text(revision_id)
1478
raise errors.HistoryMissing(self, 'inventory', revision_id)
1481
def get_inventory_sha1(self, revision_id):
1482
"""Return the sha1 hash of the inventory entry
1484
return self.get_revision(revision_id).inventory_sha1
1487
def get_revision_graph(self, revision_id=None):
1488
"""Return a dictionary containing the revision graph.
1490
NB: This method should not be used as it accesses the entire graph all
1491
at once, which is much more data than most operations should require.
1493
:param revision_id: The revision_id to get a graph from. If None, then
1494
the entire revision graph is returned. This is a deprecated mode of
1495
operation and will be removed in the future.
1496
:return: a dictionary of revision_id->revision_parents_list.
1498
raise NotImplementedError(self.get_revision_graph)
1501
def get_revision_graph_with_ghosts(self, revision_ids=None):
1502
"""Return a graph of the revisions with ghosts marked as applicable.
1504
:param revision_ids: an iterable of revisions to graph or None for all.
1505
:return: a Graph object with the graph reachable from revision_ids.
1507
if 'evil' in debug.debug_flags:
1509
"get_revision_graph_with_ghosts scales with size of history.")
1510
result = deprecated_graph.Graph()
1511
if not revision_ids:
1512
pending = set(self.all_revision_ids())
1515
pending = set(revision_ids)
1516
# special case NULL_REVISION
1517
if _mod_revision.NULL_REVISION in pending:
1518
pending.remove(_mod_revision.NULL_REVISION)
1519
required = set(pending)
1522
revision_id = pending.pop()
1524
rev = self.get_revision(revision_id)
1525
except errors.NoSuchRevision:
1526
if revision_id in required:
1529
result.add_ghost(revision_id)
1531
for parent_id in rev.parent_ids:
1532
# is this queued or done ?
1533
if (parent_id not in pending and
1534
parent_id not in done):
1536
pending.add(parent_id)
1537
result.add_node(revision_id, rev.parent_ids)
1538
done.add(revision_id)
1541
def _get_history_vf(self):
1542
"""Get a versionedfile whose history graph reflects all revisions.
1544
For weave repositories, this is the inventory weave.
1546
return self.get_inventory_weave()
1548
def iter_reverse_revision_history(self, revision_id):
1549
"""Iterate backwards through revision ids in the lefthand history
1551
:param revision_id: The revision id to start with. All its lefthand
1552
ancestors will be traversed.
1554
if revision_id in (None, _mod_revision.NULL_REVISION):
1556
next_id = revision_id
1557
versionedfile = self._get_history_vf()
1560
parents = versionedfile.get_parents(next_id)
1561
if len(parents) == 0:
1564
next_id = parents[0]
1567
def get_revision_inventory(self, revision_id):
1568
"""Return inventory of a past revision."""
1569
# TODO: Unify this with get_inventory()
1570
# bzr 0.0.6 and later imposes the constraint that the inventory_id
1571
# must be the same as its revision, so this is trivial.
1572
if revision_id is None:
1573
# This does not make sense: if there is no revision,
1574
# then it is the current tree inventory surely ?!
1575
# and thus get_root_id() is something that looks at the last
1576
# commit on the branch, and the get_root_id is an inventory check.
1577
raise NotImplementedError
1578
# return Inventory(self.get_root_id())
1580
return self.get_inventory(revision_id)
1583
def is_shared(self):
1584
"""Return True if this repository is flagged as a shared repository."""
1585
raise NotImplementedError(self.is_shared)
1588
def reconcile(self, other=None, thorough=False):
1589
"""Reconcile this repository."""
1590
from bzrlib.reconcile import RepoReconciler
1591
reconciler = RepoReconciler(self, thorough=thorough)
1592
reconciler.reconcile()
1595
def _refresh_data(self):
1596
"""Helper called from lock_* to ensure coherency with disk.
1598
The default implementation does nothing; it is however possible
1599
for repositories to maintain loaded indices across multiple locks
1600
by checking inside their implementation of this method to see
1601
whether their indices are still valid. This depends of course on
1602
the disk format being validatable in this manner.
1606
def revision_tree(self, revision_id):
1607
"""Return Tree for a revision on this branch.
1609
`revision_id` may be None for the empty tree revision.
1611
# TODO: refactor this to use an existing revision object
1612
# so we don't need to read it in twice.
1613
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1614
return RevisionTree(self, Inventory(root_id=None),
1615
_mod_revision.NULL_REVISION)
1617
inv = self.get_revision_inventory(revision_id)
1618
return RevisionTree(self, inv, revision_id)
1621
def revision_trees(self, revision_ids):
1622
"""Return Tree for a revision on this branch.
1624
`revision_id` may not be None or 'null:'"""
1625
assert None not in revision_ids
1626
assert _mod_revision.NULL_REVISION not in revision_ids
1627
texts = self.get_inventory_weave().get_texts(revision_ids)
1628
for text, revision_id in zip(texts, revision_ids):
1629
inv = self.deserialise_inventory(revision_id, text)
1630
yield RevisionTree(self, inv, revision_id)
1633
def get_ancestry(self, revision_id, topo_sorted=True):
1634
"""Return a list of revision-ids integrated by a revision.
1636
The first element of the list is always None, indicating the origin
1637
revision. This might change when we have history horizons, or
1638
perhaps we should have a new API.
1640
This is topologically sorted.
1642
if _mod_revision.is_null(revision_id):
1644
if not self.has_revision(revision_id):
1645
raise errors.NoSuchRevision(self, revision_id)
1646
w = self.get_inventory_weave()
1647
candidates = w.get_ancestry(revision_id, topo_sorted)
1648
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1651
"""Compress the data within the repository.
1653
This operation only makes sense for some repository types. For other
1654
types it should be a no-op that just returns.
1656
This stub method does not require a lock, but subclasses should use
1657
@needs_write_lock as this is a long running call its reasonable to
1658
implicitly lock for the user.
1662
def print_file(self, file, revision_id):
1663
"""Print `file` to stdout.
1665
FIXME RBC 20060125 as John Meinel points out this is a bad api
1666
- it writes to stdout, it assumes that that is valid etc. Fix
1667
by creating a new more flexible convenience function.
1669
tree = self.revision_tree(revision_id)
1670
# use inventory as it was in that revision
1671
file_id = tree.inventory.path2id(file)
1673
# TODO: jam 20060427 Write a test for this code path
1674
# it had a bug in it, and was raising the wrong
1676
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1677
tree.print_file(file_id)
1679
def get_transaction(self):
1680
return self.control_files.get_transaction()
1682
def revision_parents(self, revision_id):
1683
return self.get_inventory_weave().parent_names(revision_id)
1685
@deprecated_method(symbol_versioning.one_one)
1686
def get_parents(self, revision_ids):
1687
"""See StackedParentsProvider.get_parents"""
1688
parent_map = self.get_parent_map(revision_ids)
1689
return [parent_map.get(r, None) for r in revision_ids]
1691
def get_parent_map(self, keys):
1692
"""See graph._StackedParentsProvider.get_parent_map"""
1694
for revision_id in keys:
1695
if revision_id == _mod_revision.NULL_REVISION:
1696
parent_map[revision_id] = []
1699
parent_ids = self.get_revision(revision_id).parent_ids
1700
except errors.NoSuchRevision:
1703
if len(parent_ids) == 0:
1704
parent_ids = [_mod_revision.NULL_REVISION]
1705
parent_map[revision_id] = parent_ids
1708
def _make_parents_provider(self):
1711
def get_graph(self, other_repository=None):
1712
"""Return the graph walker for this repository format"""
1713
parents_provider = self._make_parents_provider()
1714
if (other_repository is not None and
1715
other_repository.bzrdir.transport.base !=
1716
self.bzrdir.transport.base):
1717
parents_provider = graph._StackedParentsProvider(
1718
[parents_provider, other_repository._make_parents_provider()])
1719
return graph.Graph(parents_provider)
1721
def _get_versioned_file_checker(self):
1722
"""Return an object suitable for checking versioned files."""
1723
return _VersionedFileChecker(self)
1726
def set_make_working_trees(self, new_value):
1727
"""Set the policy flag for making working trees when creating branches.
1729
This only applies to branches that use this repository.
1731
The default is 'True'.
1732
:param new_value: True to restore the default, False to disable making
1735
raise NotImplementedError(self.set_make_working_trees)
1737
def make_working_trees(self):
1738
"""Returns the policy for making working trees on new branches."""
1739
raise NotImplementedError(self.make_working_trees)
1742
def sign_revision(self, revision_id, gpg_strategy):
1743
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1744
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1747
def has_signature_for_revision_id(self, revision_id):
1748
"""Query for a revision signature for revision_id in the repository."""
1749
return self._revision_store.has_signature(revision_id,
1750
self.get_transaction())
1753
def get_signature_text(self, revision_id):
1754
"""Return the text for a signature."""
1755
return self._revision_store.get_signature_text(revision_id,
1756
self.get_transaction())
1759
def check(self, revision_ids=None):
1760
"""Check consistency of all history of given revision_ids.
1762
Different repository implementations should override _check().
1764
:param revision_ids: A non-empty list of revision_ids whose ancestry
1765
will be checked. Typically the last revision_id of a branch.
1767
return self._check(revision_ids)
1769
def _check(self, revision_ids):
1770
result = check.Check(self)
1774
def _warn_if_deprecated(self):
1775
global _deprecation_warning_done
1776
if _deprecation_warning_done:
1778
_deprecation_warning_done = True
1779
warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1780
% (self._format, self.bzrdir.transport.base))
1782
def supports_rich_root(self):
1783
return self._format.rich_root_data
1785
def _check_ascii_revisionid(self, revision_id, method):
1786
"""Private helper for ascii-only repositories."""
1787
# weave repositories refuse to store revisionids that are non-ascii.
1788
if revision_id is not None:
1789
# weaves require ascii revision ids.
1790
if isinstance(revision_id, unicode):
1792
revision_id.encode('ascii')
1793
except UnicodeEncodeError:
1794
raise errors.NonAsciiRevisionId(method, self)
1797
revision_id.decode('ascii')
1798
except UnicodeDecodeError:
1799
raise errors.NonAsciiRevisionId(method, self)
1801
def revision_graph_can_have_wrong_parents(self):
1802
"""Is it possible for this repository to have a revision graph with
1805
If True, then this repository must also implement
1806
_find_inconsistent_revision_parents so that check and reconcile can
1807
check for inconsistencies before proceeding with other checks that may
1808
depend on the revision index being consistent.
1810
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1812
# remove these delegates a while after bzr 0.15
1813
def __make_delegated(name, from_module):
1814
def _deprecated_repository_forwarder():
1815
symbol_versioning.warn('%s moved to %s in bzr 0.15'
1816
% (name, from_module),
1819
m = __import__(from_module, globals(), locals(), [name])
1821
return getattr(m, name)
1822
except AttributeError:
1823
raise AttributeError('module %s has no name %s'
1825
globals()[name] = _deprecated_repository_forwarder
1828
'AllInOneRepository',
1829
'WeaveMetaDirRepository',
1830
'PreSplitOutRepositoryFormat',
1831
'RepositoryFormat4',
1832
'RepositoryFormat5',
1833
'RepositoryFormat6',
1834
'RepositoryFormat7',
1836
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1840
'RepositoryFormatKnit',
1841
'RepositoryFormatKnit1',
1843
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1846
def install_revision(repository, rev, revision_tree):
1847
"""Install all revision data into a repository."""
1848
install_revisions(repository, [(rev, revision_tree, None)])
1851
def install_revisions(repository, iterable):
1852
"""Install all revision data into a repository.
1854
Accepts an iterable of revision, tree, signature tuples. The signature
1857
repository.start_write_group()
1859
for revision, revision_tree, signature in iterable:
1860
_install_revision(repository, revision, revision_tree, signature)
1862
repository.abort_write_group()
1865
repository.commit_write_group()
1868
def _install_revision(repository, rev, revision_tree, signature):
1869
"""Install all revision data into a repository."""
1870
present_parents = []
1872
for p_id in rev.parent_ids:
1873
if repository.has_revision(p_id):
1874
present_parents.append(p_id)
1875
parent_trees[p_id] = repository.revision_tree(p_id)
1877
parent_trees[p_id] = repository.revision_tree(None)
1879
inv = revision_tree.inventory
1880
entries = inv.iter_entries()
1881
# backwards compatibility hack: skip the root id.
1882
if not repository.supports_rich_root():
1883
path, root = entries.next()
1884
if root.revision != rev.revision_id:
1885
raise errors.IncompatibleRevision(repr(repository))
1886
# Add the texts that are not already present
1887
for path, ie in entries:
1888
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1889
repository.get_transaction())
1890
if ie.revision not in w:
1892
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1893
# with InventoryEntry.find_previous_heads(). if it is, then there
1894
# is a latent bug here where the parents may have ancestors of each
1896
for revision, tree in parent_trees.iteritems():
1897
if ie.file_id not in tree:
1899
parent_id = tree.inventory[ie.file_id].revision
1900
if parent_id in text_parents:
1902
text_parents.append(parent_id)
1904
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1905
repository.get_transaction())
1906
lines = revision_tree.get_file(ie.file_id).readlines()
1907
vfile.add_lines(rev.revision_id, text_parents, lines)
1909
# install the inventory
1910
repository.add_inventory(rev.revision_id, inv, present_parents)
1911
except errors.RevisionAlreadyPresent:
1913
if signature is not None:
1914
repository.add_signature_text(rev.revision_id, signature)
1915
repository.add_revision(rev.revision_id, rev, inv)
1918
class MetaDirRepository(Repository):
1919
"""Repositories in the new meta-dir layout."""
1921
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1922
super(MetaDirRepository, self).__init__(_format,
1928
dir_mode = self.control_files._dir_mode
1929
file_mode = self.control_files._file_mode
1932
def is_shared(self):
1933
"""Return True if this repository is flagged as a shared repository."""
1934
return self.control_files._transport.has('shared-storage')
1937
def set_make_working_trees(self, new_value):
1938
"""Set the policy flag for making working trees when creating branches.
1940
This only applies to branches that use this repository.
1942
The default is 'True'.
1943
:param new_value: True to restore the default, False to disable making
1948
self.control_files._transport.delete('no-working-trees')
1949
except errors.NoSuchFile:
1952
self.control_files.put_utf8('no-working-trees', '')
1954
def make_working_trees(self):
1955
"""Returns the policy for making working trees on new branches."""
1956
return not self.control_files._transport.has('no-working-trees')
1959
class RepositoryFormatRegistry(registry.Registry):
1960
"""Registry of RepositoryFormats."""
1962
def get(self, format_string):
1963
r = registry.Registry.get(self, format_string)
1969
format_registry = RepositoryFormatRegistry()
1970
"""Registry of formats, indexed by their identifying format string.
1972
This can contain either format instances themselves, or classes/factories that
1973
can be called to obtain one.
1977
#####################################################################
1978
# Repository Formats
1980
class RepositoryFormat(object):
1981
"""A repository format.
1983
Formats provide three things:
1984
* An initialization routine to construct repository data on disk.
1985
* a format string which is used when the BzrDir supports versioned
1987
* an open routine which returns a Repository instance.
1989
There is one and only one Format subclass for each on-disk format. But
1990
there can be one Repository subclass that is used for several different
1991
formats. The _format attribute on a Repository instance can be used to
1992
determine the disk format.
1994
Formats are placed in an dict by their format string for reference
1995
during opening. These should be subclasses of RepositoryFormat
1998
Once a format is deprecated, just deprecate the initialize and open
1999
methods on the format class. Do not deprecate the object, as the
2000
object will be created every system load.
2002
Common instance attributes:
2003
_matchingbzrdir - the bzrdir format that the repository format was
2004
originally written to work with. This can be used if manually
2005
constructing a bzrdir and repository, or more commonly for test suite
2009
# Set to True or False in derived classes. True indicates that the format
2010
# supports ghosts gracefully.
2011
supports_ghosts = None
2014
return "<%s>" % self.__class__.__name__
2016
def __eq__(self, other):
2017
# format objects are generally stateless
2018
return isinstance(other, self.__class__)
2020
def __ne__(self, other):
2021
return not self == other
2024
def find_format(klass, a_bzrdir):
2025
"""Return the format for the repository object in a_bzrdir.
2027
This is used by bzr native formats that have a "format" file in
2028
the repository. Other methods may be used by different types of
2032
transport = a_bzrdir.get_repository_transport(None)
2033
format_string = transport.get("format").read()
2034
return format_registry.get(format_string)
2035
except errors.NoSuchFile:
2036
raise errors.NoRepositoryPresent(a_bzrdir)
2038
raise errors.UnknownFormatError(format=format_string)
2041
def register_format(klass, format):
2042
format_registry.register(format.get_format_string(), format)
2045
def unregister_format(klass, format):
2046
format_registry.remove(format.get_format_string())
2049
def get_default_format(klass):
2050
"""Return the current default format."""
2051
from bzrlib import bzrdir
2052
return bzrdir.format_registry.make_bzrdir('default').repository_format
2054
def _get_control_store(self, repo_transport, control_files):
2055
"""Return the control store for this repository."""
2056
raise NotImplementedError(self._get_control_store)
2058
def get_format_string(self):
2059
"""Return the ASCII format string that identifies this format.
2061
Note that in pre format ?? repositories the format string is
2062
not permitted nor written to disk.
2064
raise NotImplementedError(self.get_format_string)
2066
def get_format_description(self):
2067
"""Return the short description for this format."""
2068
raise NotImplementedError(self.get_format_description)
2070
def _get_revision_store(self, repo_transport, control_files):
2071
"""Return the revision store object for this a_bzrdir."""
2072
raise NotImplementedError(self._get_revision_store)
2074
def _get_text_rev_store(self,
2081
"""Common logic for getting a revision store for a repository.
2083
see self._get_revision_store for the subclass-overridable method to
2084
get the store for a repository.
2086
from bzrlib.store.revision.text import TextRevisionStore
2087
dir_mode = control_files._dir_mode
2088
file_mode = control_files._file_mode
2089
text_store = TextStore(transport.clone(name),
2091
compressed=compressed,
2093
file_mode=file_mode)
2094
_revision_store = TextRevisionStore(text_store, serializer)
2095
return _revision_store
2097
# TODO: this shouldn't be in the base class, it's specific to things that
2098
# use weaves or knits -- mbp 20070207
2099
def _get_versioned_file_store(self,
2104
versionedfile_class=None,
2105
versionedfile_kwargs={},
2107
if versionedfile_class is None:
2108
versionedfile_class = self._versionedfile_class
2109
weave_transport = control_files._transport.clone(name)
2110
dir_mode = control_files._dir_mode
2111
file_mode = control_files._file_mode
2112
return VersionedFileStore(weave_transport, prefixed=prefixed,
2114
file_mode=file_mode,
2115
versionedfile_class=versionedfile_class,
2116
versionedfile_kwargs=versionedfile_kwargs,
2119
def initialize(self, a_bzrdir, shared=False):
2120
"""Initialize a repository of this format in a_bzrdir.
2122
:param a_bzrdir: The bzrdir to put the new repository in it.
2123
:param shared: The repository should be initialized as a sharable one.
2124
:returns: The new repository object.
2126
This may raise UninitializableFormat if shared repository are not
2127
compatible the a_bzrdir.
2129
raise NotImplementedError(self.initialize)
2131
def is_supported(self):
2132
"""Is this format supported?
2134
Supported formats must be initializable and openable.
2135
Unsupported formats may not support initialization or committing or
2136
some other features depending on the reason for not being supported.
2140
def check_conversion_target(self, target_format):
2141
raise NotImplementedError(self.check_conversion_target)
2143
def open(self, a_bzrdir, _found=False):
2144
"""Return an instance of this format for the bzrdir a_bzrdir.
2146
_found is a private parameter, do not use it.
2148
raise NotImplementedError(self.open)
2151
class MetaDirRepositoryFormat(RepositoryFormat):
2152
"""Common base class for the new repositories using the metadir layout."""
2154
rich_root_data = False
2155
supports_tree_reference = False
2156
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
2159
super(MetaDirRepositoryFormat, self).__init__()
2161
def _create_control_files(self, a_bzrdir):
2162
"""Create the required files and the initial control_files object."""
2163
# FIXME: RBC 20060125 don't peek under the covers
2164
# NB: no need to escape relative paths that are url safe.
2165
repository_transport = a_bzrdir.get_repository_transport(self)
2166
control_files = lockable_files.LockableFiles(repository_transport,
2167
'lock', lockdir.LockDir)
2168
control_files.create_lock()
2169
return control_files
2171
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2172
"""Upload the initial blank content."""
2173
control_files = self._create_control_files(a_bzrdir)
2174
control_files.lock_write()
2176
control_files._transport.mkdir_multi(dirs,
2177
mode=control_files._dir_mode)
2178
for file, content in files:
2179
control_files.put(file, content)
2180
for file, content in utf8_files:
2181
control_files.put_utf8(file, content)
2183
control_files.put_utf8('shared-storage', '')
2185
control_files.unlock()
2188
# formats which have no format string are not discoverable
2189
# and not independently creatable, so are not registered. They're
2190
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
2191
# needed, it's constructed directly by the BzrDir. Non-native formats where
2192
# the repository is not separately opened are similar.
2194
format_registry.register_lazy(
2195
'Bazaar-NG Repository format 7',
2196
'bzrlib.repofmt.weaverepo',
2200
format_registry.register_lazy(
2201
'Bazaar-NG Knit Repository Format 1',
2202
'bzrlib.repofmt.knitrepo',
2203
'RepositoryFormatKnit1',
2206
format_registry.register_lazy(
2207
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2208
'bzrlib.repofmt.knitrepo',
2209
'RepositoryFormatKnit3',
2212
format_registry.register_lazy(
2213
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2214
'bzrlib.repofmt.knitrepo',
2215
'RepositoryFormatKnit4',
2218
# Pack-based formats. There is one format for pre-subtrees, and one for
2219
# post-subtrees to allow ease of testing.
2220
# NOTE: These are experimental in 0.92.
2221
format_registry.register_lazy(
2222
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2223
'bzrlib.repofmt.pack_repo',
2224
'RepositoryFormatKnitPack1',
2226
format_registry.register_lazy(
2227
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2228
'bzrlib.repofmt.pack_repo',
2229
'RepositoryFormatKnitPack3',
2231
format_registry.register_lazy(
2232
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2233
'bzrlib.repofmt.pack_repo',
2234
'RepositoryFormatKnitPack4',
2238
class InterRepository(InterObject):
2239
"""This class represents operations taking place between two repositories.
2241
Its instances have methods like copy_content and fetch, and contain
2242
references to the source and target repositories these operations can be
2245
Often we will provide convenience methods on 'repository' which carry out
2246
operations with another repository - they will always forward to
2247
InterRepository.get(other).method_name(parameters).
2251
"""The available optimised InterRepository types."""
2253
def copy_content(self, revision_id=None):
2254
raise NotImplementedError(self.copy_content)
2256
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2257
"""Fetch the content required to construct revision_id.
2259
The content is copied from self.source to self.target.
2261
:param revision_id: if None all content is copied, if NULL_REVISION no
2263
:param pb: optional progress bar to use for progress reports. If not
2264
provided a default one will be created.
2266
Returns the copied revision count and the failed revisions in a tuple:
2269
raise NotImplementedError(self.fetch)
2272
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2273
"""Return the revision ids that source has that target does not.
2275
These are returned in topological order.
2277
:param revision_id: only return revision ids included by this
2280
# generic, possibly worst case, slow code path.
2281
target_ids = set(self.target.all_revision_ids())
2282
if revision_id is not None:
2283
source_ids = self.source.get_ancestry(revision_id)
2284
assert source_ids[0] is None
2287
source_ids = self.source.all_revision_ids()
2288
result_set = set(source_ids).difference(target_ids)
2289
# this may look like a no-op: its not. It preserves the ordering
2290
# other_ids had while only returning the members from other_ids
2291
# that we've decided we need.
2292
return [rev_id for rev_id in source_ids if rev_id in result_set]
2295
def _same_model(source, target):
2296
"""True if source and target have the same data representation."""
2297
if source.supports_rich_root() != target.supports_rich_root():
2299
if source._serializer != target._serializer:
2304
class InterSameDataRepository(InterRepository):
2305
"""Code for converting between repositories that represent the same data.
2307
Data format and model must match for this to work.
2311
def _get_repo_format_to_test(self):
2312
"""Repository format for testing with.
2314
InterSameData can pull from subtree to subtree and from non-subtree to
2315
non-subtree, so we test this with the richest repository format.
2317
from bzrlib.repofmt import knitrepo
2318
return knitrepo.RepositoryFormatKnit3()
2321
def is_compatible(source, target):
2322
return InterRepository._same_model(source, target)
2325
def copy_content(self, revision_id=None):
2326
"""Make a complete copy of the content in self into destination.
2328
This copies both the repository's revision data, and configuration information
2329
such as the make_working_trees setting.
2331
This is a destructive operation! Do not use it on existing
2334
:param revision_id: Only copy the content needed to construct
2335
revision_id and its parents.
2338
self.target.set_make_working_trees(self.source.make_working_trees())
2339
except NotImplementedError:
2341
# but don't bother fetching if we have the needed data now.
2342
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2343
self.target.has_revision(revision_id)):
2345
self.target.fetch(self.source, revision_id=revision_id)
2348
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2349
"""See InterRepository.fetch()."""
2350
from bzrlib.fetch import GenericRepoFetcher
2351
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2352
self.source, self.source._format, self.target,
2353
self.target._format)
2354
f = GenericRepoFetcher(to_repository=self.target,
2355
from_repository=self.source,
2356
last_revision=revision_id,
2358
return f.count_copied, f.failed_revisions
2361
class InterWeaveRepo(InterSameDataRepository):
2362
"""Optimised code paths between Weave based repositories.
2364
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2365
implemented lazy inter-object optimisation.
2369
def _get_repo_format_to_test(self):
2370
from bzrlib.repofmt import weaverepo
2371
return weaverepo.RepositoryFormat7()
2374
def is_compatible(source, target):
2375
"""Be compatible with known Weave formats.
2377
We don't test for the stores being of specific types because that
2378
could lead to confusing results, and there is no need to be
2381
from bzrlib.repofmt.weaverepo import (
2387
return (isinstance(source._format, (RepositoryFormat5,
2389
RepositoryFormat7)) and
2390
isinstance(target._format, (RepositoryFormat5,
2392
RepositoryFormat7)))
2393
except AttributeError:
2397
def copy_content(self, revision_id=None):
2398
"""See InterRepository.copy_content()."""
2399
# weave specific optimised path:
2401
self.target.set_make_working_trees(self.source.make_working_trees())
2402
except NotImplementedError:
2404
# FIXME do not peek!
2405
if self.source.control_files._transport.listable():
2406
pb = ui.ui_factory.nested_progress_bar()
2408
self.target.weave_store.copy_all_ids(
2409
self.source.weave_store,
2411
from_transaction=self.source.get_transaction(),
2412
to_transaction=self.target.get_transaction())
2413
pb.update('copying inventory', 0, 1)
2414
self.target.control_weaves.copy_multi(
2415
self.source.control_weaves, ['inventory'],
2416
from_transaction=self.source.get_transaction(),
2417
to_transaction=self.target.get_transaction())
2418
self.target._revision_store.text_store.copy_all_ids(
2419
self.source._revision_store.text_store,
2424
self.target.fetch(self.source, revision_id=revision_id)
2427
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2428
"""See InterRepository.fetch()."""
2429
from bzrlib.fetch import GenericRepoFetcher
2430
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2431
self.source, self.source._format, self.target, self.target._format)
2432
f = GenericRepoFetcher(to_repository=self.target,
2433
from_repository=self.source,
2434
last_revision=revision_id,
2436
return f.count_copied, f.failed_revisions
2439
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2440
"""See InterRepository.missing_revision_ids()."""
2441
# we want all revisions to satisfy revision_id in source.
2442
# but we don't want to stat every file here and there.
2443
# we want then, all revisions other needs to satisfy revision_id
2444
# checked, but not those that we have locally.
2445
# so the first thing is to get a subset of the revisions to
2446
# satisfy revision_id in source, and then eliminate those that
2447
# we do already have.
2448
# this is slow on high latency connection to self, but as as this
2449
# disk format scales terribly for push anyway due to rewriting
2450
# inventory.weave, this is considered acceptable.
2452
if revision_id is not None:
2453
source_ids = self.source.get_ancestry(revision_id)
2454
assert source_ids[0] is None
2457
source_ids = self.source._all_possible_ids()
2458
source_ids_set = set(source_ids)
2459
# source_ids is the worst possible case we may need to pull.
2460
# now we want to filter source_ids against what we actually
2461
# have in target, but don't try to check for existence where we know
2462
# we do not have a revision as that would be pointless.
2463
target_ids = set(self.target._all_possible_ids())
2464
possibly_present_revisions = target_ids.intersection(source_ids_set)
2465
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2466
required_revisions = source_ids_set.difference(actually_present_revisions)
2467
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2468
if revision_id is not None:
2469
# we used get_ancestry to determine source_ids then we are assured all
2470
# revisions referenced are present as they are installed in topological order.
2471
# and the tip revision was validated by get_ancestry.
2472
return required_topo_revisions
2474
# if we just grabbed the possibly available ids, then
2475
# we only have an estimate of whats available and need to validate
2476
# that against the revision records.
2477
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2480
class InterKnitRepo(InterSameDataRepository):
2481
"""Optimised code paths between Knit based repositories."""
2484
def _get_repo_format_to_test(self):
2485
from bzrlib.repofmt import knitrepo
2486
return knitrepo.RepositoryFormatKnit1()
2489
def is_compatible(source, target):
2490
"""Be compatible with known Knit formats.
2492
We don't test for the stores being of specific types because that
2493
could lead to confusing results, and there is no need to be
2496
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
2498
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2499
isinstance(target._format, RepositoryFormatKnit))
2500
except AttributeError:
2502
return are_knits and InterRepository._same_model(source, target)
2505
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2506
"""See InterRepository.fetch()."""
2507
from bzrlib.fetch import KnitRepoFetcher
2508
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2509
self.source, self.source._format, self.target, self.target._format)
2510
f = KnitRepoFetcher(to_repository=self.target,
2511
from_repository=self.source,
2512
last_revision=revision_id,
2514
return f.count_copied, f.failed_revisions
2517
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2518
"""See InterRepository.missing_revision_ids()."""
2519
if revision_id is not None:
2520
source_ids = self.source.get_ancestry(revision_id)
2521
assert source_ids[0] is None
2524
source_ids = self.source.all_revision_ids()
2525
source_ids_set = set(source_ids)
2526
# source_ids is the worst possible case we may need to pull.
2527
# now we want to filter source_ids against what we actually
2528
# have in target, but don't try to check for existence where we know
2529
# we do not have a revision as that would be pointless.
2530
target_ids = set(self.target.all_revision_ids())
2531
possibly_present_revisions = target_ids.intersection(source_ids_set)
2532
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2533
required_revisions = source_ids_set.difference(actually_present_revisions)
2534
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2535
if revision_id is not None:
2536
# we used get_ancestry to determine source_ids then we are assured all
2537
# revisions referenced are present as they are installed in topological order.
2538
# and the tip revision was validated by get_ancestry.
2539
return required_topo_revisions
2541
# if we just grabbed the possibly available ids, then
2542
# we only have an estimate of whats available and need to validate
2543
# that against the revision records.
2544
return self.source._eliminate_revisions_not_present(required_topo_revisions)
2547
class InterPackRepo(InterSameDataRepository):
2548
"""Optimised code paths between Pack based repositories."""
2551
def _get_repo_format_to_test(self):
2552
from bzrlib.repofmt import pack_repo
2553
return pack_repo.RepositoryFormatKnitPack1()
2556
def is_compatible(source, target):
2557
"""Be compatible with known Pack formats.
2559
We don't test for the stores being of specific types because that
2560
could lead to confusing results, and there is no need to be
2563
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2565
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2566
isinstance(target._format, RepositoryFormatPack))
2567
except AttributeError:
2569
return are_packs and InterRepository._same_model(source, target)
2572
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2573
"""See InterRepository.fetch()."""
2574
from bzrlib.repofmt.pack_repo import Packer
2575
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2576
self.source, self.source._format, self.target, self.target._format)
2577
self.count_copied = 0
2578
if revision_id is None:
2580
# everything to do - use pack logic
2581
# to fetch from all packs to one without
2582
# inventory parsing etc, IFF nothing to be copied is in the target.
2584
revision_ids = self.source.all_revision_ids()
2585
# implementing the TODO will involve:
2586
# - detecting when all of a pack is selected
2587
# - avoiding as much as possible pre-selection, so the
2588
# more-core routines such as create_pack_from_packs can filter in
2589
# a just-in-time fashion. (though having a HEADS list on a
2590
# repository might make this a lot easier, because we could
2591
# sensibly detect 'new revisions' without doing a full index scan.
2592
elif _mod_revision.is_null(revision_id):
2597
revision_ids = self.missing_revision_ids(revision_id,
2598
find_ghosts=find_ghosts)
2599
except errors.NoSuchRevision:
2600
raise errors.InstallFailed([revision_id])
2601
packs = self.source._pack_collection.all_packs()
2602
pack = Packer(self.target._pack_collection, packs, '.fetch',
2603
revision_ids).pack()
2604
if pack is not None:
2605
self.target._pack_collection._save_pack_names()
2606
# Trigger an autopack. This may duplicate effort as we've just done
2607
# a pack creation, but for now it is simpler to think about as
2608
# 'upload data, then repack if needed'.
2609
self.target._pack_collection.autopack()
2610
return (pack.get_revision_count(), [])
2615
def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2616
"""See InterRepository.missing_revision_ids().
2618
:param find_ghosts: Find ghosts throughough the ancestry of
2621
if not find_ghosts and revision_id is not None:
2622
graph = self.source.get_graph()
2623
missing_revs = set()
2624
searcher = graph._make_breadth_first_searcher([revision_id])
2626
self.target._pack_collection.revision_index.combined_index
2627
null_set = frozenset([_mod_revision.NULL_REVISION])
2630
next_revs = set(searcher.next())
2631
except StopIteration:
2633
next_revs.difference_update(null_set)
2634
target_keys = [(key,) for key in next_revs]
2635
have_revs = frozenset(node[1][0] for node in
2636
target_index.iter_entries(target_keys))
2637
missing_revs.update(next_revs - have_revs)
2638
searcher.stop_searching_any(have_revs)
2639
if next_revs - have_revs == set([revision_id]):
2640
# we saw the start rev itself, but no parents from it (or
2641
# next_revs would have been updated to e.g. set(). We remove
2642
# have_revs because if we found revision_id locally we
2643
# stop_searching at the first time around.
2644
raise errors.NoSuchRevision(self.source, revision_id)
2646
elif revision_id is not None:
2647
source_ids = self.source.get_ancestry(revision_id)
2648
assert source_ids[0] is None
2651
source_ids = self.source.all_revision_ids()
2652
# source_ids is the worst possible case we may need to pull.
2653
# now we want to filter source_ids against what we actually
2654
# have in target, but don't try to check for existence where we know
2655
# we do not have a revision as that would be pointless.
2656
target_ids = set(self.target.all_revision_ids())
2657
return [r for r in source_ids if (r not in target_ids)]
2660
class InterModel1and2(InterRepository):
2663
def _get_repo_format_to_test(self):
2667
def is_compatible(source, target):
2668
if not source.supports_rich_root() and target.supports_rich_root():
2674
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2675
"""See InterRepository.fetch()."""
2676
from bzrlib.fetch import Model1toKnit2Fetcher
2677
f = Model1toKnit2Fetcher(to_repository=self.target,
2678
from_repository=self.source,
2679
last_revision=revision_id,
2681
return f.count_copied, f.failed_revisions
2684
def copy_content(self, revision_id=None):
2685
"""Make a complete copy of the content in self into destination.
2687
This is a destructive operation! Do not use it on existing
2690
:param revision_id: Only copy the content needed to construct
2691
revision_id and its parents.
2694
self.target.set_make_working_trees(self.source.make_working_trees())
2695
except NotImplementedError:
2697
# but don't bother fetching if we have the needed data now.
2698
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2699
self.target.has_revision(revision_id)):
2701
self.target.fetch(self.source, revision_id=revision_id)
2704
class InterKnit1and2(InterKnitRepo):
2707
def _get_repo_format_to_test(self):
2711
def is_compatible(source, target):
2712
"""Be compatible with Knit1 source and Knit3 target"""
2713
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
2715
from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2716
RepositoryFormatKnit3)
2717
from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2718
RepositoryFormatKnitPack3)
2719
return (isinstance(source._format,
2720
(RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2721
isinstance(target._format,
2722
(RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2724
except AttributeError:
2728
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2729
"""See InterRepository.fetch()."""
2730
from bzrlib.fetch import Knit1to2Fetcher
2731
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2732
self.source, self.source._format, self.target,
2733
self.target._format)
2734
f = Knit1to2Fetcher(to_repository=self.target,
2735
from_repository=self.source,
2736
last_revision=revision_id,
2738
return f.count_copied, f.failed_revisions
2741
class InterDifferingSerializer(InterKnitRepo):
2744
def _get_repo_format_to_test(self):
2748
def is_compatible(source, target):
2749
"""Be compatible with Knit2 source and Knit3 target"""
2750
if source.supports_rich_root() != target.supports_rich_root():
2752
# Ideally, we'd support fetching if the source had no tree references
2753
# even if it supported them...
2754
if (getattr(source, '_format.supports_tree_reference', False) and
2755
not getattr(target, '_format.supports_tree_reference', False)):
2760
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2761
"""See InterRepository.fetch()."""
2762
revision_ids = self.target.missing_revision_ids(self.source,
2764
def revisions_iterator():
2765
for current_revision_id in revision_ids:
2766
revision = self.source.get_revision(current_revision_id)
2767
tree = self.source.revision_tree(current_revision_id)
2769
signature = self.source.get_signature_text(
2770
current_revision_id)
2771
except errors.NoSuchRevision:
2773
yield revision, tree, signature
2774
install_revisions(self.target, revisions_iterator())
2775
return len(revision_ids), 0
2778
class InterRemoteToOther(InterRepository):
2780
def __init__(self, source, target):
2781
InterRepository.__init__(self, source, target)
2782
self._real_inter = None
2785
def is_compatible(source, target):
2786
if not isinstance(source, remote.RemoteRepository):
2788
source._ensure_real()
2789
real_source = source._real_repository
2790
# Is source's model compatible with target's model, and are they the
2791
# same format? Currently we can only optimise fetching from an
2792
# identical model & format repo.
2793
assert not isinstance(real_source, remote.RemoteRepository), (
2794
"We don't support remote repos backed by remote repos yet.")
2795
return real_source._format == target._format
2798
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2799
"""See InterRepository.fetch()."""
2800
from bzrlib.fetch import RemoteToOtherFetcher
2801
mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2802
self.source, self.target, self.target._format)
2803
# TODO: jam 20070210 This should be an assert, not a translate
2804
revision_id = osutils.safe_revision_id(revision_id)
2805
f = RemoteToOtherFetcher(to_repository=self.target,
2806
from_repository=self.source,
2807
last_revision=revision_id,
2809
return f.count_copied, f.failed_revisions
2812
def _get_repo_format_to_test(self):
2816
class InterOtherToRemote(InterRepository):
2818
def __init__(self, source, target):
2819
InterRepository.__init__(self, source, target)
2820
self._real_inter = None
2823
def is_compatible(source, target):
2824
if isinstance(target, remote.RemoteRepository):
2828
def _ensure_real_inter(self):
2829
if self._real_inter is None:
2830
self.target._ensure_real()
2831
real_target = self.target._real_repository
2832
self._real_inter = InterRepository.get(self.source, real_target)
2834
def copy_content(self, revision_id=None):
2835
self._ensure_real_inter()
2836
self._real_inter.copy_content(revision_id=revision_id)
2838
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2839
self._ensure_real_inter()
2840
self._real_inter.fetch(revision_id=revision_id, pb=pb)
2843
def _get_repo_format_to_test(self):
2847
InterRepository.register_optimiser(InterDifferingSerializer)
2848
InterRepository.register_optimiser(InterSameDataRepository)
2849
InterRepository.register_optimiser(InterWeaveRepo)
2850
InterRepository.register_optimiser(InterKnitRepo)
2851
InterRepository.register_optimiser(InterModel1and2)
2852
InterRepository.register_optimiser(InterKnit1and2)
2853
InterRepository.register_optimiser(InterPackRepo)
2854
InterRepository.register_optimiser(InterRemoteToOther)
2855
InterRepository.register_optimiser(InterOtherToRemote)
2858
class CopyConverter(object):
2859
"""A repository conversion tool which just performs a copy of the content.
2861
This is slow but quite reliable.
2864
def __init__(self, target_format):
2865
"""Create a CopyConverter.
2867
:param target_format: The format the resulting repository should be.
2869
self.target_format = target_format
2871
def convert(self, repo, pb):
2872
"""Perform the conversion of to_convert, giving feedback via pb.
2874
:param to_convert: The disk object to convert.
2875
:param pb: a progress bar to use for progress information.
2880
# this is only useful with metadir layouts - separated repo content.
2881
# trigger an assertion if not such
2882
repo._format.get_format_string()
2883
self.repo_dir = repo.bzrdir
2884
self.step('Moving repository to repository.backup')
2885
self.repo_dir.transport.move('repository', 'repository.backup')
2886
backup_transport = self.repo_dir.transport.clone('repository.backup')
2887
repo._format.check_conversion_target(self.target_format)
2888
self.source_repo = repo._format.open(self.repo_dir,
2890
_override_transport=backup_transport)
2891
self.step('Creating new repository')
2892
converted = self.target_format.initialize(self.repo_dir,
2893
self.source_repo.is_shared())
2894
converted.lock_write()
2896
self.step('Copying content into repository.')
2897
self.source_repo.copy_content_into(converted)
2900
self.step('Deleting old repository content.')
2901
self.repo_dir.transport.delete_tree('repository.backup')
2902
self.pb.note('repository converted')
2904
def step(self, message):
2905
"""Update the pb by a step."""
2907
self.pb.update(message, self.count, self.total)
2919
def _unescaper(match, _map=_unescape_map):
2920
code = match.group(1)
2924
if not code.startswith('#'):
2926
return unichr(int(code[1:])).encode('utf8')
2932
def _unescape_xml(data):
2933
"""Unescape predefined XML entities in a string of data."""
2935
if _unescape_re is None:
2936
_unescape_re = re.compile('\&([^;]*);')
2937
return _unescape_re.sub(_unescaper, data)
2940
class _VersionedFileChecker(object):
2942
def __init__(self, repository):
2943
self.repository = repository
2944
self.text_index = self.repository._generate_text_key_index()
2946
def calculate_file_version_parents(self, revision_id, file_id):
2947
"""Calculate the correct parents for a file version according to
2950
parent_keys = self.text_index[(file_id, revision_id)]
2951
if parent_keys == [_mod_revision.NULL_REVISION]:
2953
# strip the file_id, for the weave api
2954
return tuple([revision_id for file_id, revision_id in parent_keys])
2956
def check_file_version_parents(self, weave, file_id):
2957
"""Check the parents stored in a versioned file are correct.
2959
It also detects file versions that are not referenced by their
2960
corresponding revision's inventory.
2962
:returns: A tuple of (wrong_parents, dangling_file_versions).
2963
wrong_parents is a dict mapping {revision_id: (stored_parents,
2964
correct_parents)} for each revision_id where the stored parents
2965
are not correct. dangling_file_versions is a set of (file_id,
2966
revision_id) tuples for versions that are present in this versioned
2967
file, but not used by the corresponding inventory.
2970
unused_versions = set()
2971
for num, revision_id in enumerate(weave.versions()):
2973
correct_parents = self.calculate_file_version_parents(
2974
revision_id, file_id)
2976
# The version is not part of the used keys.
2977
unused_versions.add(revision_id)
2980
knit_parents = tuple(weave.get_parents(revision_id))
2981
except errors.RevisionNotPresent:
2983
if correct_parents != knit_parents:
2984
wrong_parents[revision_id] = (knit_parents, correct_parents)
2985
return wrong_parents, unused_versions
added
removed
bzrlib/repository.py
