1
# Copyright (C) 2005-2010 Canonical Ltd
1
# Copyright (C) 2005 Canonical Ltd
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
5
5
# the Free Software Foundation; either version 2 of the License, or
6
6
# (at your option) any later version.
8
8
# This program is distributed in the hope that it will be useful,
9
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
11
# GNU General Public License for more details.
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
from bzrlib.lazy_import import lazy_import
18
lazy_import(globals(), """
41
revision as _mod_revision,
49
from bzrlib.bundle import serializer
50
from bzrlib.revisiontree import RevisionTree
51
from bzrlib.store.versioned import VersionedFileStore
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from copy import deepcopy
18
from cStringIO import StringIO
19
from unittest import TestSuite
20
import xml.sax.saxutils
23
from bzrlib.decorators import needs_read_lock, needs_write_lock
24
import bzrlib.errors as errors
25
from bzrlib.errors import InvalidRevisionId
26
from bzrlib.lockable_files import LockableFiles
27
from bzrlib.osutils import safe_unicode
28
from bzrlib.revision import NULL_REVISION
29
from bzrlib.store import copy_all
30
from bzrlib.store.weave import WeaveStore
31
from bzrlib.store.text import TextStore
32
from bzrlib.symbol_versioning import *
33
from bzrlib.trace import mutter
34
from bzrlib.tree import RevisionTree
52
35
from bzrlib.testament import Testament
59
from bzrlib.decorators import needs_read_lock, needs_write_lock, only_raises
60
from bzrlib.inter import InterObject
61
from bzrlib.inventory import (
67
from bzrlib.lock import _RelockDebugMixin, LogicalLockResult
68
from bzrlib.trace import (
69
log_exception_quietly, note, mutter, mutter_callsite, warning)
72
# Old formats display a warning, but only once
73
_deprecation_warning_done = False
76
class IsInWriteGroupError(errors.InternalBzrError):
78
_fmt = "May not refresh_data of repo %(repo)s while in a write group."
80
def __init__(self, repo):
81
errors.InternalBzrError.__init__(self, repo=repo)
84
class CommitBuilder(object):
85
"""Provides an interface to build up a commit.
87
This allows describing a tree to be committed without needing to
88
know the internals of the format of the repository.
91
# all clients should supply tree roots.
92
record_root_entry = True
93
# the default CommitBuilder does not manage trees whose root is versioned.
94
_versioned_root = False
96
def __init__(self, repository, parents, config, timestamp=None,
97
timezone=None, committer=None, revprops=None,
99
"""Initiate a CommitBuilder.
101
:param repository: Repository to commit to.
102
:param parents: Revision ids of the parents of the new revision.
103
:param config: Configuration to use.
104
:param timestamp: Optional timestamp recorded for commit.
105
:param timezone: Optional timezone for timestamp.
106
:param committer: Optional committer to set for commit.
107
:param revprops: Optional dictionary of revision properties.
108
:param revision_id: Optional revision id.
110
self._config = config
112
if committer is None:
113
self._committer = self._config.username()
115
self._committer = committer
117
self.new_inventory = Inventory(None)
118
self._new_revision_id = revision_id
119
self.parents = parents
120
self.repository = repository
123
if revprops is not None:
124
self._validate_revprops(revprops)
125
self._revprops.update(revprops)
127
if timestamp is None:
128
timestamp = time.time()
129
# Restrict resolution to 1ms
130
self._timestamp = round(timestamp, 3)
133
self._timezone = osutils.local_time_offset()
135
self._timezone = int(timezone)
137
self._generate_revision_if_needed()
138
self.__heads = graph.HeadsCache(repository.get_graph()).heads
139
self._basis_delta = []
140
# API compatibility, older code that used CommitBuilder did not call
141
# .record_delete(), which means the delta that is computed would not be
142
# valid. Callers that will call record_delete() should call
143
# .will_record_deletes() to indicate that.
144
self._recording_deletes = False
145
# memo'd check for no-op commits.
146
self._any_changes = False
148
def any_changes(self):
149
"""Return True if any entries were changed.
151
This includes merge-only changes. It is the core for the --unchanged
154
:return: True if any changes have occured.
156
return self._any_changes
158
def _validate_unicode_text(self, text, context):
159
"""Verify things like commit messages don't have bogus characters."""
161
raise ValueError('Invalid value for %s: %r' % (context, text))
163
def _validate_revprops(self, revprops):
164
for key, value in revprops.iteritems():
165
# We know that the XML serializers do not round trip '\r'
166
# correctly, so refuse to accept them
167
if not isinstance(value, basestring):
168
raise ValueError('revision property (%s) is not a valid'
169
' (unicode) string: %r' % (key, value))
170
self._validate_unicode_text(value,
171
'revision property (%s)' % (key,))
173
def commit(self, message):
174
"""Make the actual commit.
176
:return: The revision id of the recorded revision.
178
self._validate_unicode_text(message, 'commit message')
179
rev = _mod_revision.Revision(
180
timestamp=self._timestamp,
181
timezone=self._timezone,
182
committer=self._committer,
184
inventory_sha1=self.inv_sha1,
185
revision_id=self._new_revision_id,
186
properties=self._revprops)
187
rev.parent_ids = self.parents
188
self.repository.add_revision(self._new_revision_id, rev,
189
self.new_inventory, self._config)
190
self.repository.commit_write_group()
191
return self._new_revision_id
194
"""Abort the commit that is being built.
196
self.repository.abort_write_group()
198
def revision_tree(self):
199
"""Return the tree that was just committed.
201
After calling commit() this can be called to get a RevisionTree
202
representing the newly committed tree. This is preferred to
203
calling Repository.revision_tree() because that may require
204
deserializing the inventory, while we already have a copy in
207
if self.new_inventory is None:
208
self.new_inventory = self.repository.get_inventory(
209
self._new_revision_id)
210
return RevisionTree(self.repository, self.new_inventory,
211
self._new_revision_id)
213
def finish_inventory(self):
214
"""Tell the builder that the inventory is finished.
216
:return: The inventory id in the repository, which can be used with
217
repository.get_inventory.
219
if self.new_inventory is None:
220
# an inventory delta was accumulated without creating a new
222
basis_id = self.basis_delta_revision
223
# We ignore the 'inventory' returned by add_inventory_by_delta
224
# because self.new_inventory is used to hint to the rest of the
225
# system what code path was taken
226
self.inv_sha1, _ = self.repository.add_inventory_by_delta(
227
basis_id, self._basis_delta, self._new_revision_id,
230
if self.new_inventory.root is None:
231
raise AssertionError('Root entry should be supplied to'
232
' record_entry_contents, as of bzr 0.10.')
233
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
234
self.new_inventory.revision_id = self._new_revision_id
235
self.inv_sha1 = self.repository.add_inventory(
236
self._new_revision_id,
240
return self._new_revision_id
242
def _gen_revision_id(self):
243
"""Return new revision-id."""
244
return generate_ids.gen_revision_id(self._config.username(),
247
def _generate_revision_if_needed(self):
248
"""Create a revision id if None was supplied.
250
If the repository can not support user-specified revision ids
251
they should override this function and raise CannotSetRevisionId
252
if _new_revision_id is not None.
254
:raises: CannotSetRevisionId
256
if self._new_revision_id is None:
257
self._new_revision_id = self._gen_revision_id()
258
self.random_revid = True
260
self.random_revid = False
262
def _heads(self, file_id, revision_ids):
263
"""Calculate the graph heads for revision_ids in the graph of file_id.
265
This can use either a per-file graph or a global revision graph as we
266
have an identity relationship between the two graphs.
268
return self.__heads(revision_ids)
270
def _check_root(self, ie, parent_invs, tree):
271
"""Helper for record_entry_contents.
273
:param ie: An entry being added.
274
:param parent_invs: The inventories of the parent revisions of the
276
:param tree: The tree that is being committed.
278
# In this revision format, root entries have no knit or weave When
279
# serializing out to disk and back in root.revision is always
281
ie.revision = self._new_revision_id
283
def _require_root_change(self, tree):
284
"""Enforce an appropriate root object change.
286
This is called once when record_iter_changes is called, if and only if
287
the root was not in the delta calculated by record_iter_changes.
289
:param tree: The tree which is being committed.
291
# NB: if there are no parents then this method is not called, so no
292
# need to guard on parents having length.
293
entry = entry_factory['directory'](tree.path2id(''), '',
295
entry.revision = self._new_revision_id
296
self._basis_delta.append(('', '', entry.file_id, entry))
298
def _get_delta(self, ie, basis_inv, path):
299
"""Get a delta against the basis inventory for ie."""
300
if ie.file_id not in basis_inv:
302
result = (None, path, ie.file_id, ie)
303
self._basis_delta.append(result)
305
elif ie != basis_inv[ie.file_id]:
307
# TODO: avoid tis id2path call.
308
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
309
self._basis_delta.append(result)
315
def get_basis_delta(self):
316
"""Return the complete inventory delta versus the basis inventory.
318
This has been built up with the calls to record_delete and
319
record_entry_contents. The client must have already called
320
will_record_deletes() to indicate that they will be generating a
323
:return: An inventory delta, suitable for use with apply_delta, or
324
Repository.add_inventory_by_delta, etc.
326
if not self._recording_deletes:
327
raise AssertionError("recording deletes not activated.")
328
return self._basis_delta
330
def record_delete(self, path, file_id):
331
"""Record that a delete occured against a basis tree.
333
This is an optional API - when used it adds items to the basis_delta
334
being accumulated by the commit builder. It cannot be called unless the
335
method will_record_deletes() has been called to inform the builder that
336
a delta is being supplied.
338
:param path: The path of the thing deleted.
339
:param file_id: The file id that was deleted.
341
if not self._recording_deletes:
342
raise AssertionError("recording deletes not activated.")
343
delta = (path, None, file_id, None)
344
self._basis_delta.append(delta)
345
self._any_changes = True
348
def will_record_deletes(self):
349
"""Tell the commit builder that deletes are being notified.
351
This enables the accumulation of an inventory delta; for the resulting
352
commit to be valid, deletes against the basis MUST be recorded via
353
builder.record_delete().
355
self._recording_deletes = True
357
basis_id = self.parents[0]
359
basis_id = _mod_revision.NULL_REVISION
360
self.basis_delta_revision = basis_id
362
def record_entry_contents(self, ie, parent_invs, path, tree,
364
"""Record the content of ie from tree into the commit if needed.
366
Side effect: sets ie.revision when unchanged
368
:param ie: An inventory entry present in the commit.
369
:param parent_invs: The inventories of the parent revisions of the
371
:param path: The path the entry is at in the tree.
372
:param tree: The tree which contains this entry and should be used to
374
:param content_summary: Summary data from the tree about the paths
375
content - stat, length, exec, sha/link target. This is only
376
accessed when the entry has a revision of None - that is when it is
377
a candidate to commit.
378
:return: A tuple (change_delta, version_recorded, fs_hash).
379
change_delta is an inventory_delta change for this entry against
380
the basis tree of the commit, or None if no change occured against
382
version_recorded is True if a new version of the entry has been
383
recorded. For instance, committing a merge where a file was only
384
changed on the other side will return (delta, False).
385
fs_hash is either None, or the hash details for the path (currently
386
a tuple of the contents sha1 and the statvalue returned by
387
tree.get_file_with_stat()).
389
if self.new_inventory.root is None:
390
if ie.parent_id is not None:
391
raise errors.RootMissing()
392
self._check_root(ie, parent_invs, tree)
393
if ie.revision is None:
394
kind = content_summary[0]
396
# ie is carried over from a prior commit
398
# XXX: repository specific check for nested tree support goes here - if
399
# the repo doesn't want nested trees we skip it ?
400
if (kind == 'tree-reference' and
401
not self.repository._format.supports_tree_reference):
402
# mismatch between commit builder logic and repository:
403
# this needs the entry creation pushed down into the builder.
404
raise NotImplementedError('Missing repository subtree support.')
405
self.new_inventory.add(ie)
407
# TODO: slow, take it out of the inner loop.
409
basis_inv = parent_invs[0]
411
basis_inv = Inventory(root_id=None)
413
# ie.revision is always None if the InventoryEntry is considered
414
# for committing. We may record the previous parents revision if the
415
# content is actually unchanged against a sole head.
416
if ie.revision is not None:
417
if not self._versioned_root and path == '':
418
# repositories that do not version the root set the root's
419
# revision to the new commit even when no change occurs (more
420
# specifically, they do not record a revision on the root; and
421
# the rev id is assigned to the root during deserialisation -
422
# this masks when a change may have occurred against the basis.
423
# To match this we always issue a delta, because the revision
424
# of the root will always be changing.
425
if ie.file_id in basis_inv:
426
delta = (basis_inv.id2path(ie.file_id), path,
430
delta = (None, path, ie.file_id, ie)
431
self._basis_delta.append(delta)
432
return delta, False, None
434
# we don't need to commit this, because the caller already
435
# determined that an existing revision of this file is
436
# appropriate. If its not being considered for committing then
437
# it and all its parents to the root must be unaltered so
438
# no-change against the basis.
439
if ie.revision == self._new_revision_id:
440
raise AssertionError("Impossible situation, a skipped "
441
"inventory entry (%r) claims to be modified in this "
442
"commit (%r).", (ie, self._new_revision_id))
443
return None, False, None
444
# XXX: Friction: parent_candidates should return a list not a dict
445
# so that we don't have to walk the inventories again.
446
parent_candiate_entries = ie.parent_candidates(parent_invs)
447
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
449
for inv in parent_invs:
450
if ie.file_id in inv:
451
old_rev = inv[ie.file_id].revision
452
if old_rev in head_set:
453
heads.append(inv[ie.file_id].revision)
454
head_set.remove(inv[ie.file_id].revision)
457
# now we check to see if we need to write a new record to the
459
# We write a new entry unless there is one head to the ancestors, and
460
# the kind-derived content is unchanged.
462
# Cheapest check first: no ancestors, or more the one head in the
463
# ancestors, we write a new node.
467
# There is a single head, look it up for comparison
468
parent_entry = parent_candiate_entries[heads[0]]
469
# if the non-content specific data has changed, we'll be writing a
471
if (parent_entry.parent_id != ie.parent_id or
472
parent_entry.name != ie.name):
474
# now we need to do content specific checks:
476
# if the kind changed the content obviously has
477
if kind != parent_entry.kind:
479
# Stat cache fingerprint feedback for the caller - None as we usually
480
# don't generate one.
483
if content_summary[2] is None:
484
raise ValueError("Files must not have executable = None")
486
# We can't trust a check of the file length because of content
488
if (# if the exec bit has changed we have to store:
489
parent_entry.executable != content_summary[2]):
491
elif parent_entry.text_sha1 == content_summary[3]:
492
# all meta and content is unchanged (using a hash cache
493
# hit to check the sha)
494
ie.revision = parent_entry.revision
495
ie.text_size = parent_entry.text_size
496
ie.text_sha1 = parent_entry.text_sha1
497
ie.executable = parent_entry.executable
498
return self._get_delta(ie, basis_inv, path), False, None
500
# Either there is only a hash change(no hash cache entry,
501
# or same size content change), or there is no change on
503
# Provide the parent's hash to the store layer, so that the
504
# content is unchanged we will not store a new node.
505
nostore_sha = parent_entry.text_sha1
507
# We want to record a new node regardless of the presence or
508
# absence of a content change in the file.
510
ie.executable = content_summary[2]
511
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
513
text = file_obj.read()
517
ie.text_sha1, ie.text_size = self._add_text_to_weave(
518
ie.file_id, text, heads, nostore_sha)
519
# Let the caller know we generated a stat fingerprint.
520
fingerprint = (ie.text_sha1, stat_value)
521
except errors.ExistingContent:
522
# Turns out that the file content was unchanged, and we were
523
# only going to store a new node if it was changed. Carry over
525
ie.revision = parent_entry.revision
526
ie.text_size = parent_entry.text_size
527
ie.text_sha1 = parent_entry.text_sha1
528
ie.executable = parent_entry.executable
529
return self._get_delta(ie, basis_inv, path), False, None
530
elif kind == 'directory':
532
# all data is meta here, nothing specific to directory, so
534
ie.revision = parent_entry.revision
535
return self._get_delta(ie, basis_inv, path), False, None
536
self._add_text_to_weave(ie.file_id, '', heads, None)
537
elif kind == 'symlink':
538
current_link_target = content_summary[3]
540
# symlink target is not generic metadata, check if it has
542
if current_link_target != parent_entry.symlink_target:
545
# unchanged, carry over.
546
ie.revision = parent_entry.revision
547
ie.symlink_target = parent_entry.symlink_target
548
return self._get_delta(ie, basis_inv, path), False, None
549
ie.symlink_target = current_link_target
550
self._add_text_to_weave(ie.file_id, '', heads, None)
551
elif kind == 'tree-reference':
553
if content_summary[3] != parent_entry.reference_revision:
556
# unchanged, carry over.
557
ie.reference_revision = parent_entry.reference_revision
558
ie.revision = parent_entry.revision
559
return self._get_delta(ie, basis_inv, path), False, None
560
ie.reference_revision = content_summary[3]
561
if ie.reference_revision is None:
562
raise AssertionError("invalid content_summary for nested tree: %r"
563
% (content_summary,))
564
self._add_text_to_weave(ie.file_id, '', heads, None)
566
raise NotImplementedError('unknown kind')
567
ie.revision = self._new_revision_id
568
self._any_changes = True
569
return self._get_delta(ie, basis_inv, path), True, fingerprint
571
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
572
_entry_factory=entry_factory):
573
"""Record a new tree via iter_changes.
575
:param tree: The tree to obtain text contents from for changed objects.
576
:param basis_revision_id: The revision id of the tree the iter_changes
577
has been generated against. Currently assumed to be the same
578
as self.parents[0] - if it is not, errors may occur.
579
:param iter_changes: An iter_changes iterator with the changes to apply
580
to basis_revision_id. The iterator must not include any items with
581
a current kind of None - missing items must be either filtered out
582
or errored-on beefore record_iter_changes sees the item.
583
:param _entry_factory: Private method to bind entry_factory locally for
585
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
588
# Create an inventory delta based on deltas between all the parents and
589
# deltas between all the parent inventories. We use inventory delta's
590
# between the inventory objects because iter_changes masks
591
# last-changed-field only changes.
593
# file_id -> change map, change is fileid, paths, changed, versioneds,
594
# parents, names, kinds, executables
596
# {file_id -> revision_id -> inventory entry, for entries in parent
597
# trees that are not parents[0]
601
revtrees = list(self.repository.revision_trees(self.parents))
602
except errors.NoSuchRevision:
603
# one or more ghosts, slow path.
605
for revision_id in self.parents:
607
revtrees.append(self.repository.revision_tree(revision_id))
608
except errors.NoSuchRevision:
610
basis_revision_id = _mod_revision.NULL_REVISION
612
revtrees.append(self.repository.revision_tree(
613
_mod_revision.NULL_REVISION))
614
# The basis inventory from a repository
616
basis_inv = revtrees[0].inventory
618
basis_inv = self.repository.revision_tree(
619
_mod_revision.NULL_REVISION).inventory
620
if len(self.parents) > 0:
621
if basis_revision_id != self.parents[0] and not ghost_basis:
623
"arbitrary basis parents not yet supported with merges")
624
for revtree in revtrees[1:]:
625
for change in revtree.inventory._make_delta(basis_inv):
626
if change[1] is None:
627
# Not present in this parent.
629
if change[2] not in merged_ids:
630
if change[0] is not None:
631
basis_entry = basis_inv[change[2]]
632
merged_ids[change[2]] = [
634
basis_entry.revision,
637
parent_entries[change[2]] = {
639
basis_entry.revision:basis_entry,
641
change[3].revision:change[3],
644
merged_ids[change[2]] = [change[3].revision]
645
parent_entries[change[2]] = {change[3].revision:change[3]}
647
merged_ids[change[2]].append(change[3].revision)
648
parent_entries[change[2]][change[3].revision] = change[3]
651
# Setup the changes from the tree:
652
# changes maps file_id -> (change, [parent revision_ids])
654
for change in iter_changes:
655
# This probably looks up in basis_inv way to much.
656
if change[1][0] is not None:
657
head_candidate = [basis_inv[change[0]].revision]
660
changes[change[0]] = change, merged_ids.get(change[0],
662
unchanged_merged = set(merged_ids) - set(changes)
663
# Extend the changes dict with synthetic changes to record merges of
665
for file_id in unchanged_merged:
666
# Record a merged version of these items that did not change vs the
667
# basis. This can be either identical parallel changes, or a revert
668
# of a specific file after a merge. The recorded content will be
669
# that of the current tree (which is the same as the basis), but
670
# the per-file graph will reflect a merge.
671
# NB:XXX: We are reconstructing path information we had, this
672
# should be preserved instead.
673
# inv delta change: (file_id, (path_in_source, path_in_target),
674
# changed_content, versioned, parent, name, kind,
677
basis_entry = basis_inv[file_id]
678
except errors.NoSuchId:
679
# a change from basis->some_parents but file_id isn't in basis
680
# so was new in the merge, which means it must have changed
681
# from basis -> current, and as it hasn't the add was reverted
682
# by the user. So we discard this change.
686
(basis_inv.id2path(file_id), tree.id2path(file_id)),
688
(basis_entry.parent_id, basis_entry.parent_id),
689
(basis_entry.name, basis_entry.name),
690
(basis_entry.kind, basis_entry.kind),
691
(basis_entry.executable, basis_entry.executable))
692
changes[file_id] = (change, merged_ids[file_id])
693
# changes contains tuples with the change and a set of inventory
694
# candidates for the file.
696
# old_path, new_path, file_id, new_inventory_entry
697
seen_root = False # Is the root in the basis delta?
698
inv_delta = self._basis_delta
699
modified_rev = self._new_revision_id
700
for change, head_candidates in changes.values():
701
if change[3][1]: # versioned in target.
702
# Several things may be happening here:
703
# We may have a fork in the per-file graph
704
# - record a change with the content from tree
705
# We may have a change against < all trees
706
# - carry over the tree that hasn't changed
707
# We may have a change against all trees
708
# - record the change with the content from tree
711
entry = _entry_factory[kind](file_id, change[5][1],
713
head_set = self._heads(change[0], set(head_candidates))
716
for head_candidate in head_candidates:
717
if head_candidate in head_set:
718
heads.append(head_candidate)
719
head_set.remove(head_candidate)
722
# Could be a carry-over situation:
723
parent_entry_revs = parent_entries.get(file_id, None)
724
if parent_entry_revs:
725
parent_entry = parent_entry_revs.get(heads[0], None)
728
if parent_entry is None:
729
# The parent iter_changes was called against is the one
730
# that is the per-file head, so any change is relevant
731
# iter_changes is valid.
732
carry_over_possible = False
734
# could be a carry over situation
735
# A change against the basis may just indicate a merge,
736
# we need to check the content against the source of the
737
# merge to determine if it was changed after the merge
739
if (parent_entry.kind != entry.kind or
740
parent_entry.parent_id != entry.parent_id or
741
parent_entry.name != entry.name):
742
# Metadata common to all entries has changed
743
# against per-file parent
744
carry_over_possible = False
746
carry_over_possible = True
747
# per-type checks for changes against the parent_entry
750
# Cannot be a carry-over situation
751
carry_over_possible = False
752
# Populate the entry in the delta
754
# XXX: There is still a small race here: If someone reverts the content of a file
755
# after iter_changes examines and decides it has changed,
756
# we will unconditionally record a new version even if some
757
# other process reverts it while commit is running (with
758
# the revert happening after iter_changes did it's
761
entry.executable = True
763
entry.executable = False
764
if (carry_over_possible and
765
parent_entry.executable == entry.executable):
766
# Check the file length, content hash after reading
768
nostore_sha = parent_entry.text_sha1
771
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
773
text = file_obj.read()
777
entry.text_sha1, entry.text_size = self._add_text_to_weave(
778
file_id, text, heads, nostore_sha)
779
yield file_id, change[1][1], (entry.text_sha1, stat_value)
780
except errors.ExistingContent:
781
# No content change against a carry_over parent
782
# Perhaps this should also yield a fs hash update?
784
entry.text_size = parent_entry.text_size
785
entry.text_sha1 = parent_entry.text_sha1
786
elif kind == 'symlink':
788
entry.symlink_target = tree.get_symlink_target(file_id)
789
if (carry_over_possible and
790
parent_entry.symlink_target == entry.symlink_target):
793
self._add_text_to_weave(change[0], '', heads, None)
794
elif kind == 'directory':
795
if carry_over_possible:
798
# Nothing to set on the entry.
799
# XXX: split into the Root and nonRoot versions.
800
if change[1][1] != '' or self.repository.supports_rich_root():
801
self._add_text_to_weave(change[0], '', heads, None)
802
elif kind == 'tree-reference':
803
if not self.repository._format.supports_tree_reference:
804
# This isn't quite sane as an error, but we shouldn't
805
# ever see this code path in practice: tree's don't
806
# permit references when the repo doesn't support tree
808
raise errors.UnsupportedOperation(tree.add_reference,
810
reference_revision = tree.get_reference_revision(change[0])
811
entry.reference_revision = reference_revision
812
if (carry_over_possible and
813
parent_entry.reference_revision == reference_revision):
816
self._add_text_to_weave(change[0], '', heads, None)
818
raise AssertionError('unknown kind %r' % kind)
820
entry.revision = modified_rev
822
entry.revision = parent_entry.revision
825
new_path = change[1][1]
826
inv_delta.append((change[1][0], new_path, change[0], entry))
829
self.new_inventory = None
831
# This should perhaps be guarded by a check that the basis we
832
# commit against is the basis for the commit and if not do a delta
834
self._any_changes = True
836
# housekeeping root entry changes do not affect no-change commits.
837
self._require_root_change(tree)
838
self.basis_delta_revision = basis_revision_id
840
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
841
parent_keys = tuple([(file_id, parent) for parent in parents])
842
return self.repository.texts._add_text(
843
(file_id, self._new_revision_id), parent_keys, new_text,
844
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
847
class RootCommitBuilder(CommitBuilder):
848
"""This commitbuilder actually records the root id"""
850
# the root entry gets versioned properly by this builder.
851
_versioned_root = True
853
def _check_root(self, ie, parent_invs, tree):
854
"""Helper for record_entry_contents.
856
:param ie: An entry being added.
857
:param parent_invs: The inventories of the parent revisions of the
859
:param tree: The tree that is being committed.
862
def _require_root_change(self, tree):
863
"""Enforce an appropriate root object change.
865
This is called once when record_iter_changes is called, if and only if
866
the root was not in the delta calculated by record_iter_changes.
868
:param tree: The tree which is being committed.
870
# versioned roots do not change unless the tree found a change.
873
class RepositoryWriteLockResult(LogicalLockResult):
874
"""The result of write locking a repository.
876
:ivar repository_token: The token obtained from the underlying lock, or
878
:ivar unlock: A callable which will unlock the lock.
881
def __init__(self, unlock, repository_token):
882
LogicalLockResult.__init__(self, unlock)
883
self.repository_token = repository_token
886
return "RepositoryWriteLockResult(%s, %s)" % (self.repository_token,
890
######################################################################
894
class Repository(_RelockDebugMixin, bzrdir.ControlComponent):
36
from bzrlib.tree import EmptyTree
41
class Repository(object):
895
42
"""Repository holding history for one or more branches.
897
44
The repository holds and retrieves historical information including
898
45
revisions and file history. It's normally accessed only by the Branch,
899
46
which views a particular line of development through that history.
901
The Repository builds on top of some byte storage facilies (the revisions,
902
signatures, inventories, texts and chk_bytes attributes) and a Transport,
903
which respectively provide byte storage and a means to access the (possibly
48
The Repository builds on top of Stores and a Transport, which respectively
49
describe the disk data format and the way of accessing the (possibly
906
The byte storage facilities are addressed via tuples, which we refer to
907
as 'keys' throughout the code base. Revision_keys, inventory_keys and
908
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
909
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
910
byte string made up of a hash identifier and a hash value.
911
We use this interface because it allows low friction with the underlying
912
code that implements disk indices, network encoding and other parts of
915
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
916
the serialised revisions for the repository. This can be used to obtain
917
revision graph information or to access raw serialised revisions.
918
The result of trying to insert data into the repository via this store
919
is undefined: it should be considered read-only except for implementors
921
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
922
the serialised signatures for the repository. This can be used to
923
obtain access to raw serialised signatures. The result of trying to
924
insert data into the repository via this store is undefined: it should
925
be considered read-only except for implementors of repositories.
926
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
927
the serialised inventories for the repository. This can be used to
928
obtain unserialised inventories. The result of trying to insert data
929
into the repository via this store is undefined: it should be
930
considered read-only except for implementors of repositories.
931
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
932
texts of files and directories for the repository. This can be used to
933
obtain file texts or file graphs. Note that Repository.iter_file_bytes
934
is usually a better interface for accessing file texts.
935
The result of trying to insert data into the repository via this store
936
is undefined: it should be considered read-only except for implementors
938
:ivar chk_bytes: A bzrlib.versionedfile.VersionedFiles instance containing
939
any data the repository chooses to store or have indexed by its hash.
940
The result of trying to insert data into the repository via this store
941
is undefined: it should be considered read-only except for implementors
943
:ivar _transport: Transport for file access to repository, typically
944
pointing to .bzr/repository.
947
# What class to use for a CommitBuilder. Often its simpler to change this
948
# in a Repository class subclass rather than to override
949
# get_commit_builder.
950
_commit_builder_class = CommitBuilder
951
# The search regex used by xml based repositories to determine what things
952
# where changed in a single commit.
953
_file_ids_altered_regex = lazy_regex.lazy_compile(
954
r'file_id="(?P<file_id>[^"]+)"'
955
r'.* revision="(?P<revision_id>[^"]+)"'
958
def abort_write_group(self, suppress_errors=False):
959
"""Commit the contents accrued within the current write group.
961
:param suppress_errors: if true, abort_write_group will catch and log
962
unexpected errors that happen during the abort, rather than
963
allowing them to propagate. Defaults to False.
965
:seealso: start_write_group.
967
if self._write_group is not self.get_transaction():
968
# has an unlock or relock occured ?
971
'(suppressed) mismatched lock context and write group. %r, %r',
972
self._write_group, self.get_transaction())
974
raise errors.BzrError(
975
'mismatched lock context and write group. %r, %r' %
976
(self._write_group, self.get_transaction()))
978
self._abort_write_group()
979
except Exception, exc:
980
self._write_group = None
981
if not suppress_errors:
983
mutter('abort_write_group failed')
984
log_exception_quietly()
985
note('bzr: ERROR (ignored): %s', exc)
986
self._write_group = None
988
def _abort_write_group(self):
989
"""Template method for per-repository write group cleanup.
991
This is called during abort before the write group is considered to be
992
finished and should cleanup any internal state accrued during the write
993
group. There is no requirement that data handed to the repository be
994
*not* made available - this is not a rollback - but neither should any
995
attempt be made to ensure that data added is fully commited. Abort is
996
invoked when an error has occured so futher disk or network operations
997
may not be possible or may error and if possible should not be
1001
def add_fallback_repository(self, repository):
1002
"""Add a repository to use for looking up data not held locally.
1004
:param repository: A repository.
1006
if not self._format.supports_external_lookups:
1007
raise errors.UnstackableRepositoryFormat(self._format, self.base)
1008
if self.is_locked():
1009
# This repository will call fallback.unlock() when we transition to
1010
# the unlocked state, so we make sure to increment the lock count
1011
repository.lock_read()
1012
self._check_fallback_repository(repository)
1013
self._fallback_repositories.append(repository)
1014
self.texts.add_fallback_versioned_files(repository.texts)
1015
self.inventories.add_fallback_versioned_files(repository.inventories)
1016
self.revisions.add_fallback_versioned_files(repository.revisions)
1017
self.signatures.add_fallback_versioned_files(repository.signatures)
1018
if self.chk_bytes is not None:
1019
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
1021
def _check_fallback_repository(self, repository):
1022
"""Check that this repository can fallback to repository safely.
1024
Raise an error if not.
1026
:param repository: A repository to fallback to.
1028
return InterRepository._assert_same_model(self, repository)
1030
def add_inventory(self, revision_id, inv, parents):
1031
"""Add the inventory inv to the repository as revision_id.
1033
:param parents: The revision ids of the parents that revision_id
1034
is known to have and are in the repository already.
1036
:returns: The validator(which is a sha1 digest, though what is sha'd is
1037
repository format specific) of the serialized inventory.
1039
if not self.is_in_write_group():
1040
raise AssertionError("%r not in write group" % (self,))
1041
_mod_revision.check_not_reserved_id(revision_id)
1042
if not (inv.revision_id is None or inv.revision_id == revision_id):
1043
raise AssertionError(
1044
"Mismatch between inventory revision"
1045
" id and insertion revid (%r, %r)"
1046
% (inv.revision_id, revision_id))
1047
if inv.root is None:
1048
raise AssertionError()
1049
return self._add_inventory_checked(revision_id, inv, parents)
1051
def _add_inventory_checked(self, revision_id, inv, parents):
1052
"""Add inv to the repository after checking the inputs.
1054
This function can be overridden to allow different inventory styles.
1056
:seealso: add_inventory, for the contract.
1058
inv_lines = self._serializer.write_inventory_to_lines(inv)
1059
return self._inventory_add_lines(revision_id, parents,
1060
inv_lines, check_content=False)
1062
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1063
parents, basis_inv=None, propagate_caches=False):
1064
"""Add a new inventory expressed as a delta against another revision.
1066
See the inventory developers documentation for the theory behind
1069
:param basis_revision_id: The inventory id the delta was created
1070
against. (This does not have to be a direct parent.)
1071
:param delta: The inventory delta (see Inventory.apply_delta for
1073
:param new_revision_id: The revision id that the inventory is being
1075
:param parents: The revision ids of the parents that revision_id is
1076
known to have and are in the repository already. These are supplied
1077
for repositories that depend on the inventory graph for revision
1078
graph access, as well as for those that pun ancestry with delta
1080
:param basis_inv: The basis inventory if it is already known,
1082
:param propagate_caches: If True, the caches for this inventory are
1083
copied to and updated for the result if possible.
1085
:returns: (validator, new_inv)
1086
The validator(which is a sha1 digest, though what is sha'd is
1087
repository format specific) of the serialized inventory, and the
1088
resulting inventory.
1090
if not self.is_in_write_group():
1091
raise AssertionError("%r not in write group" % (self,))
1092
_mod_revision.check_not_reserved_id(new_revision_id)
1093
basis_tree = self.revision_tree(basis_revision_id)
1094
basis_tree.lock_read()
1096
# Note that this mutates the inventory of basis_tree, which not all
1097
# inventory implementations may support: A better idiom would be to
1098
# return a new inventory, but as there is no revision tree cache in
1099
# repository this is safe for now - RBC 20081013
1100
if basis_inv is None:
1101
basis_inv = basis_tree.inventory
1102
basis_inv.apply_delta(delta)
1103
basis_inv.revision_id = new_revision_id
1104
return (self.add_inventory(new_revision_id, basis_inv, parents),
1109
def _inventory_add_lines(self, revision_id, parents, lines,
1110
check_content=True):
1111
"""Store lines in inv_vf and return the sha1 of the inventory."""
1112
parents = [(parent,) for parent in parents]
1113
result = self.inventories.add_lines((revision_id,), parents, lines,
1114
check_content=check_content)[0]
1115
self.inventories._access.flush()
1118
def add_revision(self, revision_id, rev, inv=None, config=None):
1119
"""Add rev to the revision store as revision_id.
1121
:param revision_id: the revision id to use.
1122
:param rev: The revision object.
1123
:param inv: The inventory for the revision. if None, it will be looked
1124
up in the inventory storer
1125
:param config: If None no digital signature will be created.
1126
If supplied its signature_needed method will be used
1127
to determine if a signature should be made.
1129
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
1131
_mod_revision.check_not_reserved_id(revision_id)
1132
if config is not None and config.signature_needed():
1134
inv = self.get_inventory(revision_id)
1135
plaintext = Testament(rev, inv).as_short_text()
1136
self.store_revision_signature(
1137
gpg.GPGStrategy(config), plaintext, revision_id)
1138
# check inventory present
1139
if not self.inventories.get_parent_map([(revision_id,)]):
1141
raise errors.WeaveRevisionNotPresent(revision_id,
1144
# yes, this is not suitable for adding with ghosts.
1145
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
1148
key = (revision_id,)
1149
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
1150
self._add_revision(rev)
1152
def _add_revision(self, revision):
1153
text = self._serializer.write_revision_to_string(revision)
1154
key = (revision.revision_id,)
1155
parents = tuple((parent,) for parent in revision.parent_ids)
1156
self.revisions.add_lines(key, parents, osutils.split_lines(text))
54
def _all_possible_ids(self):
55
"""Return all the possible revisions that we could find."""
56
return self.get_inventory_weave().names()
1158
59
def all_revision_ids(self):
1159
"""Returns a list of all the revision ids in the repository.
1161
This is conceptually deprecated because code should generally work on
1162
the graph reachable from a particular revision, and ignore any other
1163
revisions that might be present. There is no direct replacement
1166
if 'evil' in debug.debug_flags:
1167
mutter_callsite(2, "all_revision_ids is linear with history.")
1168
return self._all_revision_ids()
1170
def _all_revision_ids(self):
1171
"""Returns a list of all the revision ids in the repository.
1173
These are in as much topological order as the underlying store can
1176
raise NotImplementedError(self._all_revision_ids)
1178
def break_lock(self):
1179
"""Break a lock if one is present from another instance.
1181
Uses the ui factory to ask for confirmation if the lock may be from
1184
self.control_files.break_lock()
60
"""Returns a list of all the revision ids in the repository.
62
These are in as much topological order as the underlying store can
63
present: for weaves ghosts may lead to a lack of correctness until
64
the reweave updates the parents list.
66
result = self._all_possible_ids()
67
return self._eliminate_revisions_not_present(result)
1187
70
def _eliminate_revisions_not_present(self, revision_ids):
1190
73
Returns a set of the present revisions.
1193
graph = self.get_graph()
1194
parent_map = graph.get_parent_map(revision_ids)
1195
# The old API returned a list, should this actually be a set?
1196
return parent_map.keys()
1198
def _check_inventories(self, checker):
1199
"""Check the inventories found from the revision scan.
1201
This is responsible for verifying the sha1 of inventories and
1202
creating a pending_keys set that covers data referenced by inventories.
1204
bar = ui.ui_factory.nested_progress_bar()
1206
self._do_check_inventories(checker, bar)
1210
def _do_check_inventories(self, checker, bar):
1211
"""Helper for _check_inventories."""
1213
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1214
kinds = ['chk_bytes', 'texts']
1215
count = len(checker.pending_keys)
1216
bar.update("inventories", 0, 2)
1217
current_keys = checker.pending_keys
1218
checker.pending_keys = {}
1219
# Accumulate current checks.
1220
for key in current_keys:
1221
if key[0] != 'inventories' and key[0] not in kinds:
1222
checker._report_items.append('unknown key type %r' % (key,))
1223
keys[key[0]].add(key[1:])
1224
if keys['inventories']:
1225
# NB: output order *should* be roughly sorted - topo or
1226
# inverse topo depending on repository - either way decent
1227
# to just delta against. However, pre-CHK formats didn't
1228
# try to optimise inventory layout on disk. As such the
1229
# pre-CHK code path does not use inventory deltas.
1231
for record in self.inventories.check(keys=keys['inventories']):
1232
if record.storage_kind == 'absent':
1233
checker._report_items.append(
1234
'Missing inventory {%s}' % (record.key,))
1236
last_object = self._check_record('inventories', record,
1237
checker, last_object,
1238
current_keys[('inventories',) + record.key])
1239
del keys['inventories']
1242
bar.update("texts", 1)
1243
while (checker.pending_keys or keys['chk_bytes']
1245
# Something to check.
1246
current_keys = checker.pending_keys
1247
checker.pending_keys = {}
1248
# Accumulate current checks.
1249
for key in current_keys:
1250
if key[0] not in kinds:
1251
checker._report_items.append('unknown key type %r' % (key,))
1252
keys[key[0]].add(key[1:])
1253
# Check the outermost kind only - inventories || chk_bytes || texts
1257
for record in getattr(self, kind).check(keys=keys[kind]):
1258
if record.storage_kind == 'absent':
1259
checker._report_items.append(
1260
'Missing %s {%s}' % (kind, record.key,))
1262
last_object = self._check_record(kind, record,
1263
checker, last_object, current_keys[(kind,) + record.key])
1267
def _check_record(self, kind, record, checker, last_object, item_data):
1268
"""Check a single text from this repository."""
1269
if kind == 'inventories':
1270
rev_id = record.key[0]
1271
inv = self._deserialise_inventory(rev_id,
1272
record.get_bytes_as('fulltext'))
1273
if last_object is not None:
1274
delta = inv._make_delta(last_object)
1275
for old_path, path, file_id, ie in delta:
1278
ie.check(checker, rev_id, inv)
1280
for path, ie in inv.iter_entries():
1281
ie.check(checker, rev_id, inv)
1282
if self._format.fast_deltas:
1284
elif kind == 'chk_bytes':
1285
# No code written to check chk_bytes for this repo format.
1286
checker._report_items.append(
1287
'unsupported key type chk_bytes for %s' % (record.key,))
1288
elif kind == 'texts':
1289
self._check_text(record, checker, item_data)
1291
checker._report_items.append(
1292
'unknown key type %s for %s' % (kind, record.key))
1294
def _check_text(self, record, checker, item_data):
1295
"""Check a single text."""
1296
# Check it is extractable.
1297
# TODO: check length.
1298
if record.storage_kind == 'chunked':
1299
chunks = record.get_bytes_as(record.storage_kind)
1300
sha1 = osutils.sha_strings(chunks)
1301
length = sum(map(len, chunks))
1303
content = record.get_bytes_as('fulltext')
1304
sha1 = osutils.sha_string(content)
1305
length = len(content)
1306
if item_data and sha1 != item_data[1]:
1307
checker._report_items.append(
1308
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1309
(record.key, sha1, item_data[1], item_data[2]))
76
for id in revision_ids:
77
if self.has_revision(id):
1312
82
def create(a_bzrdir):
1313
83
"""Construct the current default format repository in a_bzrdir."""
1314
84
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
1316
def __init__(self, _format, a_bzrdir, control_files):
86
def __init__(self, _format, a_bzrdir, control_files, revision_store):
1317
87
"""instantiate a Repository.
1319
89
:param _format: The format of the repository on disk.
1320
90
:param a_bzrdir: The BzrDir of the repository.
92
In the future we will have a single api for all stores for
93
getting file texts, inventories and revisions, then
94
this construct will accept instances of those things.
1322
# In the future we will have a single api for all stores for
1323
# getting file texts, inventories and revisions, then
1324
# this construct will accept instances of those things.
1325
super(Repository, self).__init__()
1326
97
self._format = _format
1327
98
# the following are part of the public API for Repository:
1328
99
self.bzrdir = a_bzrdir
1329
100
self.control_files = control_files
1330
self._transport = control_files._transport
1331
self.base = self._transport.base
1333
self._reconcile_does_inventory_gc = True
1334
self._reconcile_fixes_text_parents = False
1335
self._reconcile_backsup_inventory = True
1336
self._write_group = None
1337
# Additional places to query for data.
1338
self._fallback_repositories = []
1339
# An InventoryEntry cache, used during deserialization
1340
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1341
# Is it safe to return inventory entries directly from the entry cache,
1342
# rather copying them?
1343
self._safe_to_return_from_cache = False
1346
def user_transport(self):
1347
return self.bzrdir.user_transport
1350
def control_transport(self):
1351
return self._transport
1354
if self._fallback_repositories:
1355
return '%s(%r, fallback_repositories=%r)' % (
1356
self.__class__.__name__,
1358
self._fallback_repositories)
1360
return '%s(%r)' % (self.__class__.__name__,
1363
def _has_same_fallbacks(self, other_repo):
1364
"""Returns true if the repositories have the same fallbacks."""
1365
my_fb = self._fallback_repositories
1366
other_fb = other_repo._fallback_repositories
1367
if len(my_fb) != len(other_fb):
1369
for f, g in zip(my_fb, other_fb):
1370
if not f.has_same_location(g):
1374
def has_same_location(self, other):
1375
"""Returns a boolean indicating if this repository is at the same
1376
location as another repository.
1378
This might return False even when two repository objects are accessing
1379
the same physical repository via different URLs.
1381
if self.__class__ is not other.__class__:
1383
return (self._transport.base == other._transport.base)
1385
def is_in_write_group(self):
1386
"""Return True if there is an open write group.
1388
:seealso: start_write_group.
1390
return self._write_group is not None
1392
def is_locked(self):
1393
return self.control_files.is_locked()
1395
def is_write_locked(self):
1396
"""Return True if this object is write locked."""
1397
return self.is_locked() and self.control_files._lock_mode == 'w'
1399
def lock_write(self, token=None):
1400
"""Lock this repository for writing.
1402
This causes caching within the repository obejct to start accumlating
1403
data during reads, and allows a 'write_group' to be obtained. Write
1404
groups must be used for actual data insertion.
1406
A token should be passed in if you know that you have locked the object
1407
some other way, and need to synchronise this object's state with that
1410
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1412
:param token: if this is already locked, then lock_write will fail
1413
unless the token matches the existing lock.
1414
:returns: a token if this instance supports tokens, otherwise None.
1415
:raises TokenLockingNotSupported: when a token is given but this
1416
instance doesn't support using token locks.
1417
:raises MismatchedToken: if the specified token doesn't match the token
1418
of the existing lock.
1419
:seealso: start_write_group.
1420
:return: A RepositoryWriteLockResult.
1422
locked = self.is_locked()
1423
token = self.control_files.lock_write(token=token)
1425
self._warn_if_deprecated()
1426
self._note_lock('w')
1427
for repo in self._fallback_repositories:
1428
# Writes don't affect fallback repos
1430
self._refresh_data()
1431
return RepositoryWriteLockResult(self.unlock, token)
101
self.revision_store = revision_store
103
def lock_write(self):
104
self.control_files.lock_write()
1433
106
def lock_read(self):
1434
"""Lock the repository for read operations.
1436
:return: An object with an unlock method which will release the lock
1439
locked = self.is_locked()
1440
107
self.control_files.lock_read()
1442
self._warn_if_deprecated()
1443
self._note_lock('r')
1444
for repo in self._fallback_repositories:
1446
self._refresh_data()
1447
return LogicalLockResult(self.unlock)
1449
def get_physical_lock_status(self):
1450
return self.control_files.get_physical_lock_status()
1452
def leave_lock_in_place(self):
1453
"""Tell this repository not to release the physical lock when this
1456
If lock_write doesn't return a token, then this method is not supported.
1458
self.control_files.leave_in_place()
1460
def dont_leave_lock_in_place(self):
1461
"""Tell this repository to release the physical lock when this
1462
object is unlocked, even if it didn't originally acquire it.
1464
If lock_write doesn't return a token, then this method is not supported.
1466
self.control_files.dont_leave_in_place()
1469
def gather_stats(self, revid=None, committers=None):
1470
"""Gather statistics from a revision id.
1472
:param revid: The revision id to gather statistics from, if None, then
1473
no revision specific statistics are gathered.
1474
:param committers: Optional parameter controlling whether to grab
1475
a count of committers from the revision specific statistics.
1476
:return: A dictionary of statistics. Currently this contains:
1477
committers: The number of committers if requested.
1478
firstrev: A tuple with timestamp, timezone for the penultimate left
1479
most ancestor of revid, if revid is not the NULL_REVISION.
1480
latestrev: A tuple with timestamp, timezone for revid, if revid is
1481
not the NULL_REVISION.
1482
revisions: The total revision count in the repository.
1483
size: An estimate disk size of the repository in bytes.
1486
if revid and committers:
1487
result['committers'] = 0
1488
if revid and revid != _mod_revision.NULL_REVISION:
1490
all_committers = set()
1491
revisions = self.get_ancestry(revid)
1492
# pop the leading None
1494
first_revision = None
1496
# ignore the revisions in the middle - just grab first and last
1497
revisions = revisions[0], revisions[-1]
1498
for revision in self.get_revisions(revisions):
1499
if not first_revision:
1500
first_revision = revision
1502
all_committers.add(revision.committer)
1503
last_revision = revision
1505
result['committers'] = len(all_committers)
1506
result['firstrev'] = (first_revision.timestamp,
1507
first_revision.timezone)
1508
result['latestrev'] = (last_revision.timestamp,
1509
last_revision.timezone)
1511
# now gather global repository information
1512
# XXX: This is available for many repos regardless of listability.
1513
if self.user_transport.listable():
1514
# XXX: do we want to __define len__() ?
1515
# Maybe the versionedfiles object should provide a different
1516
# method to get the number of keys.
1517
result['revisions'] = len(self.revisions.keys())
1518
# result['size'] = t
1521
def find_branches(self, using=False):
1522
"""Find branches underneath this repository.
1524
This will include branches inside other branches.
1526
:param using: If True, list only branches using this repository.
1528
if using and not self.is_shared():
1529
return self.bzrdir.list_branches()
1530
class Evaluator(object):
1533
self.first_call = True
1535
def __call__(self, bzrdir):
1536
# On the first call, the parameter is always the bzrdir
1537
# containing the current repo.
1538
if not self.first_call:
1540
repository = bzrdir.open_repository()
1541
except errors.NoRepositoryPresent:
1544
return False, ([], repository)
1545
self.first_call = False
1546
value = (bzrdir.list_branches(), None)
1550
for branches, repository in bzrdir.BzrDir.find_bzrdirs(
1551
self.user_transport, evaluate=Evaluator()):
1552
if branches is not None:
1553
ret.extend(branches)
1554
if not using and repository is not None:
1555
ret.extend(repository.find_branches())
1559
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
110
def missing_revision_ids(self, other, revision_id=None):
1560
111
"""Return the revision ids that other has that this does not.
1562
113
These are returned in topological order.
1564
115
revision_id: only return revision ids included by revision_id.
1566
return InterRepository.get(other, self).search_missing_revision_ids(
1567
revision_id, find_ghosts)
117
return InterRepository.get(other, self).missing_revision_ids(revision_id)
1573
123
For instance, if the repository is at URL/.bzr/repository,
1574
124
Repository.open(URL) -> a Repository instance.
1576
control = bzrdir.BzrDir.open(base)
126
control = bzrlib.bzrdir.BzrDir.open(base)
1577
127
return control.open_repository()
1579
def copy_content_into(self, destination, revision_id=None):
129
def copy_content_into(self, destination, revision_id=None, basis=None):
1580
130
"""Make a complete copy of the content in self into destination.
1582
This is a destructive operation! Do not use it on existing
132
This is a destructive operation! Do not use it on existing
1585
return InterRepository.get(self, destination).copy_content(revision_id)
1587
def commit_write_group(self):
1588
"""Commit the contents accrued within the current write group.
1590
:seealso: start_write_group.
1592
:return: it may return an opaque hint that can be passed to 'pack'.
1594
if self._write_group is not self.get_transaction():
1595
# has an unlock or relock occured ?
1596
raise errors.BzrError('mismatched lock context %r and '
1598
(self.get_transaction(), self._write_group))
1599
result = self._commit_write_group()
1600
self._write_group = None
1603
def _commit_write_group(self):
1604
"""Template method for per-repository write group cleanup.
1606
This is called before the write group is considered to be
1607
finished and should ensure that all data handed to the repository
1608
for writing during the write group is safely committed (to the
1609
extent possible considering file system caching etc).
1612
def suspend_write_group(self):
1613
raise errors.UnsuspendableWriteGroup(self)
1615
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1616
"""Return the keys of missing inventory parents for revisions added in
1619
A revision is not complete if the inventory delta for that revision
1620
cannot be calculated. Therefore if the parent inventories of a
1621
revision are not present, the revision is incomplete, and e.g. cannot
1622
be streamed by a smart server. This method finds missing inventory
1623
parents for revisions added in this write group.
1625
if not self._format.supports_external_lookups:
1626
# This is only an issue for stacked repositories
1628
if not self.is_in_write_group():
1629
raise AssertionError('not in a write group')
1631
# XXX: We assume that every added revision already has its
1632
# corresponding inventory, so we only check for parent inventories that
1633
# might be missing, rather than all inventories.
1634
parents = set(self.revisions._index.get_missing_parents())
1635
parents.discard(_mod_revision.NULL_REVISION)
1636
unstacked_inventories = self.inventories._index
1637
present_inventories = unstacked_inventories.get_parent_map(
1638
key[-1:] for key in parents)
1639
parents.difference_update(present_inventories)
1640
if len(parents) == 0:
1641
# No missing parent inventories.
1643
if not check_for_missing_texts:
1644
return set(('inventories', rev_id) for (rev_id,) in parents)
1645
# Ok, now we have a list of missing inventories. But these only matter
1646
# if the inventories that reference them are missing some texts they
1647
# appear to introduce.
1648
# XXX: Texts referenced by all added inventories need to be present,
1649
# but at the moment we're only checking for texts referenced by
1650
# inventories at the graph's edge.
1651
key_deps = self.revisions._index._key_dependencies
1652
key_deps.satisfy_refs_for_keys(present_inventories)
1653
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1654
file_ids = self.fileids_altered_by_revision_ids(referrers)
1655
missing_texts = set()
1656
for file_id, version_ids in file_ids.iteritems():
1657
missing_texts.update(
1658
(file_id, version_id) for version_id in version_ids)
1659
present_texts = self.texts.get_parent_map(missing_texts)
1660
missing_texts.difference_update(present_texts)
1661
if not missing_texts:
1662
# No texts are missing, so all revisions and their deltas are
1665
# Alternatively the text versions could be returned as the missing
1666
# keys, but this is likely to be less data.
1667
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1670
def refresh_data(self):
1671
"""Re-read any data needed to synchronise with disk.
1673
This method is intended to be called after another repository instance
1674
(such as one used by a smart server) has inserted data into the
1675
repository. On all repositories this will work outside of write groups.
1676
Some repository formats (pack and newer for bzrlib native formats)
1677
support refresh_data inside write groups. If called inside a write
1678
group on a repository that does not support refreshing in a write group
1679
IsInWriteGroupError will be raised.
1681
self._refresh_data()
1683
def resume_write_group(self, tokens):
1684
if not self.is_write_locked():
1685
raise errors.NotWriteLocked(self)
1686
if self._write_group:
1687
raise errors.BzrError('already in a write group')
1688
self._resume_write_group(tokens)
1689
# so we can detect unlock/relock - the write group is now entered.
1690
self._write_group = self.get_transaction()
1692
def _resume_write_group(self, tokens):
1693
raise errors.UnsuspendableWriteGroup(self)
1695
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
135
return InterRepository.get(self, destination).copy_content(revision_id, basis)
137
def fetch(self, source, revision_id=None, pb=None):
1697
138
"""Fetch the content required to construct revision_id from source.
1699
If revision_id is None and fetch_spec is None, then all content is
1702
fetch() may not be used when the repository is in a write group -
1703
either finish the current write group before using fetch, or use
1704
fetch before starting the write group.
1706
:param find_ghosts: Find and copy revisions in the source that are
1707
ghosts in the target (and not reachable directly by walking out to
1708
the first-present revision in target from revision_id).
1709
:param revision_id: If specified, all the content needed for this
1710
revision ID will be copied to the target. Fetch will determine for
1711
itself which content needs to be copied.
1712
:param fetch_spec: If specified, a SearchResult or
1713
PendingAncestryResult that describes which revisions to copy. This
1714
allows copying multiple heads at once. Mutually exclusive with
1717
if fetch_spec is not None and revision_id is not None:
1718
raise AssertionError(
1719
"fetch_spec and revision_id are mutually exclusive.")
1720
if self.is_in_write_group():
1721
raise errors.InternalBzrError(
1722
"May not fetch while in a write group.")
1723
# fast path same-url fetch operations
1724
# TODO: lift out to somewhere common with RemoteRepository
1725
# <https://bugs.edge.launchpad.net/bzr/+bug/401646>
1726
if (self.has_same_location(source)
1727
and fetch_spec is None
1728
and self._has_same_fallbacks(source)):
1729
# check that last_revision is in 'from' and then return a
1731
if (revision_id is not None and
1732
not _mod_revision.is_null(revision_id)):
1733
self.get_revision(revision_id)
1735
# if there is no specific appropriate InterRepository, this will get
1736
# the InterRepository base class, which raises an
1737
# IncompatibleRepositories when asked to fetch.
1738
inter = InterRepository.get(source, self)
1739
return inter.fetch(revision_id=revision_id, pb=pb,
1740
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
1742
def create_bundle(self, target, base, fileobj, format=None):
1743
return serializer.write_bundle(self, target, base, fileobj, format)
1745
def get_commit_builder(self, branch, parents, config, timestamp=None,
1746
timezone=None, committer=None, revprops=None,
1748
"""Obtain a CommitBuilder for this repository.
1750
:param branch: Branch to commit to.
1751
:param parents: Revision ids of the parents of the new revision.
1752
:param config: Configuration to use.
1753
:param timestamp: Optional timestamp recorded for commit.
1754
:param timezone: Optional timezone for timestamp.
1755
:param committer: Optional committer to set for commit.
1756
:param revprops: Optional dictionary of revision properties.
1757
:param revision_id: Optional revision id.
1759
if self._fallback_repositories:
1760
raise errors.BzrError("Cannot commit from a lightweight checkout "
1761
"to a stacked branch. See "
1762
"https://bugs.launchpad.net/bzr/+bug/375013 for details.")
1763
result = self._commit_builder_class(self, parents, config,
1764
timestamp, timezone, committer, revprops, revision_id)
1765
self.start_write_group()
1768
@only_raises(errors.LockNotHeld, errors.LockBroken)
140
If revision_id is None all content is copied.
142
return InterRepository.get(source, self).fetch(revision_id=revision_id,
1769
145
def unlock(self):
1770
if (self.control_files._lock_count == 1 and
1771
self.control_files._lock_mode == 'w'):
1772
if self._write_group is not None:
1773
self.abort_write_group()
1774
self.control_files.unlock()
1775
raise errors.BzrError(
1776
'Must end write groups before releasing write locks.')
1777
146
self.control_files.unlock()
1778
if self.control_files._lock_count == 0:
1779
self._inventory_entry_cache.clear()
1780
for repo in self._fallback_repositories:
1783
148
@needs_read_lock
1784
def clone(self, a_bzrdir, revision_id=None):
149
def clone(self, a_bzrdir, revision_id=None, basis=None):
1785
150
"""Clone this repository into a_bzrdir using the current format.
1787
152
Currently no check is made that the format of this repository and
1788
153
the bzrdir format are compatible. FIXME RBC 20060201.
1790
:return: The newly created destination repository.
1792
# TODO: deprecate after 0.16; cloning this with all its settings is
1793
# probably not very useful -- mbp 20070423
1794
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
1795
self.copy_content_into(dest_repo, revision_id)
1798
def start_write_group(self):
1799
"""Start a write group in the repository.
1801
Write groups are used by repositories which do not have a 1:1 mapping
1802
between file ids and backend store to manage the insertion of data from
1803
both fetch and commit operations.
1805
A write lock is required around the start_write_group/commit_write_group
1806
for the support of lock-requiring repository formats.
1808
One can only insert data into a repository inside a write group.
1812
if not self.is_write_locked():
1813
raise errors.NotWriteLocked(self)
1814
if self._write_group:
1815
raise errors.BzrError('already in a write group')
1816
self._start_write_group()
1817
# so we can detect unlock/relock - the write group is now entered.
1818
self._write_group = self.get_transaction()
1820
def _start_write_group(self):
1821
"""Template method for per-repository write group startup.
1823
This is called before the write group is considered to be
1828
def sprout(self, to_bzrdir, revision_id=None):
1829
"""Create a descendent repository for new development.
1831
Unlike clone, this does not copy the settings of the repository.
1833
dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
1834
dest_repo.fetch(self, revision_id=revision_id)
1837
def _create_sprouting_repo(self, a_bzrdir, shared):
1838
155
if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
1839
156
# use target default format.
1840
dest_repo = a_bzrdir.create_repository()
157
result = a_bzrdir.create_repository()
158
# FIXME RBC 20060209 split out the repository type to avoid this check ?
159
elif isinstance(a_bzrdir._format,
160
(bzrlib.bzrdir.BzrDirFormat4,
161
bzrlib.bzrdir.BzrDirFormat5,
162
bzrlib.bzrdir.BzrDirFormat6)):
163
result = a_bzrdir.open_repository()
1842
# Most control formats need the repository to be specifically
1843
# created, but on some old all-in-one formats it's not needed
1845
dest_repo = self._format.initialize(a_bzrdir, shared=shared)
1846
except errors.UninitializableFormat:
1847
dest_repo = a_bzrdir.open_repository()
1850
def _get_sink(self):
1851
"""Return a sink for streaming into this repository."""
1852
return StreamSink(self)
1854
def _get_source(self, to_format):
1855
"""Return a source for streaming from this repository."""
1856
return StreamSource(self, to_format)
165
result = self._format.initialize(a_bzrdir, shared=self.is_shared())
166
self.copy_content_into(result, revision_id, basis)
1859
169
def has_revision(self, revision_id):
1860
"""True if this repository has a copy of the revision."""
1861
return revision_id in self.has_revisions((revision_id,))
1864
def has_revisions(self, revision_ids):
1865
"""Probe to find out the presence of multiple revisions.
1867
:param revision_ids: An iterable of revision_ids.
1868
:return: A set of the revision_ids that were present.
1870
parent_map = self.revisions.get_parent_map(
1871
[(rev_id,) for rev_id in revision_ids])
1873
if _mod_revision.NULL_REVISION in revision_ids:
1874
result.add(_mod_revision.NULL_REVISION)
1875
result.update([key[0] for key in parent_map])
170
"""True if this branch has a copy of the revision.
172
This does not necessarily imply the revision is merge
173
or on the mainline."""
174
return (revision_id is None
175
or self.revision_store.has_id(revision_id))
178
def get_revision_xml_file(self, revision_id):
179
"""Return XML file object for revision object."""
180
if not revision_id or not isinstance(revision_id, basestring):
181
raise InvalidRevisionId(revision_id=revision_id, branch=self)
183
return self.revision_store.get(revision_id)
184
except (IndexError, KeyError):
185
raise bzrlib.errors.NoSuchRevision(self, revision_id)
188
def get_revision_xml(self, revision_id):
189
return self.get_revision_xml_file(revision_id).read()
1878
191
@needs_read_lock
1879
192
def get_revision(self, revision_id):
1880
"""Return the Revision object for a named revision."""
1881
return self.get_revisions([revision_id])[0]
1884
def get_revision_reconcile(self, revision_id):
1885
"""'reconcile' helper routine that allows access to a revision always.
1887
This variant of get_revision does not cross check the weave graph
1888
against the revision one as get_revision does: but it should only
1889
be used by reconcile, or reconcile-alike commands that are correcting
1890
or testing the revision graph.
1892
return self._get_revisions([revision_id])[0]
1895
def get_revisions(self, revision_ids):
1896
"""Get many revisions at once.
1898
Repositories that need to check data on every revision read should
1899
subclass this method.
1901
return self._get_revisions(revision_ids)
1904
def _get_revisions(self, revision_ids):
1905
"""Core work logic to get many revisions without sanity checks."""
1907
for revid, rev in self._iter_revisions(revision_ids):
1909
raise errors.NoSuchRevision(self, revid)
1911
return [revs[revid] for revid in revision_ids]
1913
def _iter_revisions(self, revision_ids):
1914
"""Iterate over revision objects.
1916
:param revision_ids: An iterable of revisions to examine. None may be
1917
passed to request all revisions known to the repository. Note that
1918
not all repositories can find unreferenced revisions; for those
1919
repositories only referenced ones will be returned.
1920
:return: An iterator of (revid, revision) tuples. Absent revisions (
1921
those asked for but not available) are returned as (revid, None).
1923
if revision_ids is None:
1924
revision_ids = self.all_revision_ids()
1926
for rev_id in revision_ids:
1927
if not rev_id or not isinstance(rev_id, basestring):
1928
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1929
keys = [(key,) for key in revision_ids]
1930
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1931
for record in stream:
1932
revid = record.key[0]
1933
if record.storage_kind == 'absent':
1936
text = record.get_bytes_as('fulltext')
1937
rev = self._serializer.read_revision_from_string(text)
1940
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1941
"""Produce a generator of revision deltas.
1943
Note that the input is a sequence of REVISIONS, not revision_ids.
1944
Trees will be held in memory until the generator exits.
1945
Each delta is relative to the revision's lefthand predecessor.
1947
:param specific_fileids: if not None, the result is filtered
1948
so that only those file-ids, their parents and their
1949
children are included.
1951
# Get the revision-ids of interest
1952
required_trees = set()
1953
for revision in revisions:
1954
required_trees.add(revision.revision_id)
1955
required_trees.update(revision.parent_ids[:1])
1957
# Get the matching filtered trees. Note that it's more
1958
# efficient to pass filtered trees to changes_from() rather
1959
# than doing the filtering afterwards. changes_from() could
1960
# arguably do the filtering itself but it's path-based, not
1961
# file-id based, so filtering before or afterwards is
1963
if specific_fileids is None:
1964
trees = dict((t.get_revision_id(), t) for
1965
t in self.revision_trees(required_trees))
1967
trees = dict((t.get_revision_id(), t) for
1968
t in self._filtered_revision_trees(required_trees,
1971
# Calculate the deltas
1972
for revision in revisions:
1973
if not revision.parent_ids:
1974
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1976
old_tree = trees[revision.parent_ids[0]]
1977
yield trees[revision.revision_id].changes_from(old_tree)
1980
def get_revision_delta(self, revision_id, specific_fileids=None):
1981
"""Return the delta for one revision.
1983
The delta is relative to the left-hand predecessor of the
1986
:param specific_fileids: if not None, the result is filtered
1987
so that only those file-ids, their parents and their
1988
children are included.
1990
r = self.get_revision(revision_id)
1991
return list(self.get_deltas_for_revisions([r],
1992
specific_fileids=specific_fileids))[0]
193
"""Return the Revision object for a named revision"""
194
xml_file = self.get_revision_xml_file(revision_id)
197
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
198
except SyntaxError, e:
199
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
203
assert r.revision_id == revision_id
207
def get_revision_sha1(self, revision_id):
208
"""Hash the stored value of a revision, and return it."""
209
# In the future, revision entries will be signed. At that
210
# point, it is probably best *not* to include the signature
211
# in the revision hash. Because that lets you re-sign
212
# the revision, (add signatures/remove signatures) and still
213
# have all hash pointers stay consistent.
214
# But for now, just hash the contents.
215
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
1994
217
@needs_write_lock
1995
218
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1996
signature = gpg_strategy.sign(plaintext)
1997
self.add_signature_text(revision_id, signature)
2000
def add_signature_text(self, revision_id, signature):
2001
self.signatures.add_lines((revision_id,), (),
2002
osutils.split_lines(signature))
2004
def find_text_key_references(self):
2005
"""Find the text key references within the repository.
2007
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2008
to whether they were referred to by the inventory of the
2009
revision_id that they contain. The inventory texts from all present
2010
revision ids are assessed to generate this report.
2012
revision_keys = self.revisions.keys()
2013
w = self.inventories
2014
pb = ui.ui_factory.nested_progress_bar()
2016
return self._find_text_key_references_from_xml_inventory_lines(
2017
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
2021
def _find_text_key_references_from_xml_inventory_lines(self,
2023
"""Core routine for extracting references to texts from inventories.
2025
This performs the translation of xml lines to revision ids.
2027
:param line_iterator: An iterator of lines, origin_version_id
2028
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2029
to whether they were referred to by the inventory of the
2030
revision_id that they contain. Note that if that revision_id was
2031
not part of the line_iterator's output then False will be given -
2032
even though it may actually refer to that key.
2034
if not self._serializer.support_altered_by_hack:
2035
raise AssertionError(
2036
"_find_text_key_references_from_xml_inventory_lines only "
2037
"supported for branches which store inventory as unnested xml"
2038
", not on %r" % self)
2041
# this code needs to read every new line in every inventory for the
2042
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
2043
# not present in one of those inventories is unnecessary but not
2044
# harmful because we are filtering by the revision id marker in the
2045
# inventory lines : we only select file ids altered in one of those
2046
# revisions. We don't need to see all lines in the inventory because
2047
# only those added in an inventory in rev X can contain a revision=X
2049
unescape_revid_cache = {}
2050
unescape_fileid_cache = {}
2052
# jam 20061218 In a big fetch, this handles hundreds of thousands
2053
# of lines, so it has had a lot of inlining and optimizing done.
2054
# Sorry that it is a little bit messy.
2055
# Move several functions to be local variables, since this is a long
2057
search = self._file_ids_altered_regex.search
2058
unescape = _unescape_xml
2059
setdefault = result.setdefault
2060
for line, line_key in line_iterator:
2061
match = search(line)
2064
# One call to match.group() returning multiple items is quite a
2065
# bit faster than 2 calls to match.group() each returning 1
2066
file_id, revision_id = match.group('file_id', 'revision_id')
2068
# Inlining the cache lookups helps a lot when you make 170,000
2069
# lines and 350k ids, versus 8.4 unique ids.
2070
# Using a cache helps in 2 ways:
2071
# 1) Avoids unnecessary decoding calls
2072
# 2) Re-uses cached strings, which helps in future set and
2074
# (2) is enough that removing encoding entirely along with
2075
# the cache (so we are using plain strings) results in no
2076
# performance improvement.
2078
revision_id = unescape_revid_cache[revision_id]
2080
unescaped = unescape(revision_id)
2081
unescape_revid_cache[revision_id] = unescaped
2082
revision_id = unescaped
2084
# Note that unconditionally unescaping means that we deserialise
2085
# every fileid, which for general 'pull' is not great, but we don't
2086
# really want to have some many fulltexts that this matters anyway.
2089
file_id = unescape_fileid_cache[file_id]
2091
unescaped = unescape(file_id)
2092
unescape_fileid_cache[file_id] = unescaped
2095
key = (file_id, revision_id)
2096
setdefault(key, False)
2097
if revision_id == line_key[-1]:
2101
def _inventory_xml_lines_for_keys(self, keys):
2102
"""Get a line iterator of the sort needed for findind references.
2104
Not relevant for non-xml inventory repositories.
2106
Ghosts in revision_keys are ignored.
2108
:param revision_keys: The revision keys for the inventories to inspect.
2109
:return: An iterator over (inventory line, revid) for the fulltexts of
2110
all of the xml inventories specified by revision_keys.
2112
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2113
for record in stream:
2114
if record.storage_kind != 'absent':
2115
chunks = record.get_bytes_as('chunked')
2116
revid = record.key[-1]
2117
lines = osutils.chunks_to_lines(chunks)
2121
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2123
"""Helper routine for fileids_altered_by_revision_ids.
2125
This performs the translation of xml lines to revision ids.
2127
:param line_iterator: An iterator of lines, origin_version_id
2128
:param revision_keys: The revision ids to filter for. This should be a
2129
set or other type which supports efficient __contains__ lookups, as
2130
the revision key from each parsed line will be looked up in the
2131
revision_keys filter.
2132
:return: a dictionary mapping altered file-ids to an iterable of
2133
revision_ids. Each altered file-ids has the exact revision_ids that
2134
altered it listed explicitly.
2136
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2137
line_iterator).iterkeys())
2138
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2139
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2140
self._inventory_xml_lines_for_keys(parent_keys)))
2141
new_keys = seen - parent_seen
2143
setdefault = result.setdefault
2144
for key in new_keys:
2145
setdefault(key[0], set()).add(key[-1])
2148
def _find_parent_ids_of_revisions(self, revision_ids):
2149
"""Find all parent ids that are mentioned in the revision graph.
2151
:return: set of revisions that are parents of revision_ids which are
2152
not part of revision_ids themselves
2154
parent_map = self.get_parent_map(revision_ids)
2156
map(parent_ids.update, parent_map.itervalues())
2157
parent_ids.difference_update(revision_ids)
2158
parent_ids.discard(_mod_revision.NULL_REVISION)
2161
def _find_parent_keys_of_revisions(self, revision_keys):
2162
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2164
:param revision_keys: An iterable of revision_keys.
2165
:return: The parents of all revision_keys that are not already in
2168
parent_map = self.revisions.get_parent_map(revision_keys)
2170
map(parent_keys.update, parent_map.itervalues())
2171
parent_keys.difference_update(revision_keys)
2172
parent_keys.discard(_mod_revision.NULL_REVISION)
2175
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
2176
"""Find the file ids and versions affected by revisions.
2178
:param revisions: an iterable containing revision ids.
2179
:param _inv_weave: The inventory weave from this repository or None.
2180
If None, the inventory weave will be opened automatically.
2181
:return: a dictionary mapping altered file-ids to an iterable of
2182
revision_ids. Each altered file-ids has the exact revision_ids that
2183
altered it listed explicitly.
2185
selected_keys = set((revid,) for revid in revision_ids)
2186
w = _inv_weave or self.inventories
2187
return self._find_file_ids_from_xml_inventory_lines(
2188
w.iter_lines_added_or_present_in_keys(
2189
selected_keys, pb=None),
2192
def iter_files_bytes(self, desired_files):
2193
"""Iterate through file versions.
2195
Files will not necessarily be returned in the order they occur in
2196
desired_files. No specific order is guaranteed.
2198
Yields pairs of identifier, bytes_iterator. identifier is an opaque
2199
value supplied by the caller as part of desired_files. It should
2200
uniquely identify the file version in the caller's context. (Examples:
2201
an index number or a TreeTransform trans_id.)
2203
bytes_iterator is an iterable of bytestrings for the file. The
2204
kind of iterable and length of the bytestrings are unspecified, but for
2205
this implementation, it is a list of bytes produced by
2206
VersionedFile.get_record_stream().
2208
:param desired_files: a list of (file_id, revision_id, identifier)
2212
for file_id, revision_id, callable_data in desired_files:
2213
text_keys[(file_id, revision_id)] = callable_data
2214
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2215
if record.storage_kind == 'absent':
2216
raise errors.RevisionNotPresent(record.key, self)
2217
yield text_keys[record.key], record.get_bytes_as('chunked')
2219
def _generate_text_key_index(self, text_key_references=None,
2221
"""Generate a new text key index for the repository.
2223
This is an expensive function that will take considerable time to run.
2225
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2226
list of parents, also text keys. When a given key has no parents,
2227
the parents list will be [NULL_REVISION].
2229
# All revisions, to find inventory parents.
2230
if ancestors is None:
2231
graph = self.get_graph()
2232
ancestors = graph.get_parent_map(self.all_revision_ids())
2233
if text_key_references is None:
2234
text_key_references = self.find_text_key_references()
2235
pb = ui.ui_factory.nested_progress_bar()
2237
return self._do_generate_text_key_index(ancestors,
2238
text_key_references, pb)
2242
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2243
"""Helper for _generate_text_key_index to avoid deep nesting."""
2244
revision_order = tsort.topo_sort(ancestors)
2245
invalid_keys = set()
2247
for revision_id in revision_order:
2248
revision_keys[revision_id] = set()
2249
text_count = len(text_key_references)
2250
# a cache of the text keys to allow reuse; costs a dict of all the
2251
# keys, but saves a 2-tuple for every child of a given key.
2253
for text_key, valid in text_key_references.iteritems():
2255
invalid_keys.add(text_key)
2257
revision_keys[text_key[1]].add(text_key)
2258
text_key_cache[text_key] = text_key
2259
del text_key_references
2261
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2262
NULL_REVISION = _mod_revision.NULL_REVISION
2263
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2264
# too small for large or very branchy trees. However, for 55K path
2265
# trees, it would be easy to use too much memory trivially. Ideally we
2266
# could gauge this by looking at available real memory etc, but this is
2267
# always a tricky proposition.
2268
inventory_cache = lru_cache.LRUCache(10)
2269
batch_size = 10 # should be ~150MB on a 55K path tree
2270
batch_count = len(revision_order) / batch_size + 1
2272
pb.update("Calculating text parents", processed_texts, text_count)
2273
for offset in xrange(batch_count):
2274
to_query = revision_order[offset * batch_size:(offset + 1) *
2278
for revision_id in to_query:
2279
parent_ids = ancestors[revision_id]
2280
for text_key in revision_keys[revision_id]:
2281
pb.update("Calculating text parents", processed_texts)
2282
processed_texts += 1
2283
candidate_parents = []
2284
for parent_id in parent_ids:
2285
parent_text_key = (text_key[0], parent_id)
2287
check_parent = parent_text_key not in \
2288
revision_keys[parent_id]
2290
# the parent parent_id is a ghost:
2291
check_parent = False
2292
# truncate the derived graph against this ghost.
2293
parent_text_key = None
2295
# look at the parent commit details inventories to
2296
# determine possible candidates in the per file graph.
2299
inv = inventory_cache[parent_id]
2301
inv = self.revision_tree(parent_id).inventory
2302
inventory_cache[parent_id] = inv
2304
parent_entry = inv[text_key[0]]
2305
except (KeyError, errors.NoSuchId):
2307
if parent_entry is not None:
2309
text_key[0], parent_entry.revision)
2311
parent_text_key = None
2312
if parent_text_key is not None:
2313
candidate_parents.append(
2314
text_key_cache[parent_text_key])
2315
parent_heads = text_graph.heads(candidate_parents)
2316
new_parents = list(parent_heads)
2317
new_parents.sort(key=lambda x:candidate_parents.index(x))
2318
if new_parents == []:
2319
new_parents = [NULL_REVISION]
2320
text_index[text_key] = new_parents
2322
for text_key in invalid_keys:
2323
text_index[text_key] = [NULL_REVISION]
2326
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2327
"""Get an iterable listing the keys of all the data introduced by a set
2330
The keys will be ordered so that the corresponding items can be safely
2331
fetched and inserted in that order.
2333
:returns: An iterable producing tuples of (knit-kind, file-id,
2334
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2335
'revisions'. file-id is None unless knit-kind is 'file'.
2337
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2340
for result in self._find_non_file_keys_to_fetch(revision_ids):
2343
def _find_file_keys_to_fetch(self, revision_ids, pb):
2344
# XXX: it's a bit weird to control the inventory weave caching in this
2345
# generator. Ideally the caching would be done in fetch.py I think. Or
2346
# maybe this generator should explicitly have the contract that it
2347
# should not be iterated until the previously yielded item has been
2349
inv_w = self.inventories
2351
# file ids that changed
2352
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2354
num_file_ids = len(file_ids)
2355
for file_id, altered_versions in file_ids.iteritems():
2357
pb.update("Fetch texts", count, num_file_ids)
2359
yield ("file", file_id, altered_versions)
2361
def _find_non_file_keys_to_fetch(self, revision_ids):
2363
yield ("inventory", None, revision_ids)
2366
# XXX: Note ATM no callers actually pay attention to this return
2367
# instead they just use the list of revision ids and ignore
2368
# missing sigs. Consider removing this work entirely
2369
revisions_with_signatures = set(self.signatures.get_parent_map(
2370
[(r,) for r in revision_ids]))
2371
revisions_with_signatures = set(
2372
[r for (r,) in revisions_with_signatures])
2373
revisions_with_signatures.intersection_update(revision_ids)
2374
yield ("signatures", None, revisions_with_signatures)
2377
yield ("revisions", None, revision_ids)
219
self.revision_store.add(StringIO(gpg_strategy.sign(plaintext)),
222
def fileid_involved_between_revs(self, from_revid, to_revid):
223
"""Find file_id(s) which are involved in the changes between revisions.
225
This determines the set of revisions which are involved, and then
226
finds all file ids affected by those revisions.
228
# TODO: jam 20060119 This code assumes that w.inclusions will
229
# always be correct. But because of the presence of ghosts
230
# it is possible to be wrong.
231
# One specific example from Robert Collins:
232
# Two branches, with revisions ABC, and AD
233
# C is a ghost merge of D.
234
# Inclusions doesn't recognize D as an ancestor.
235
# If D is ever merged in the future, the weave
236
# won't be fixed, because AD never saw revision C
237
# to cause a conflict which would force a reweave.
238
w = self.get_inventory_weave()
239
from_set = set(w.inclusions([w.lookup(from_revid)]))
240
to_set = set(w.inclusions([w.lookup(to_revid)]))
241
included = to_set.difference(from_set)
242
changed = map(w.idx_to_name, included)
243
return self._fileid_involved_by_set(changed)
245
def fileid_involved(self, last_revid=None):
246
"""Find all file_ids modified in the ancestry of last_revid.
248
:param last_revid: If None, last_revision() will be used.
250
w = self.get_inventory_weave()
252
changed = set(w._names)
254
included = w.inclusions([w.lookup(last_revid)])
255
changed = map(w.idx_to_name, included)
256
return self._fileid_involved_by_set(changed)
258
def fileid_involved_by_set(self, changes):
259
"""Find all file_ids modified by the set of revisions passed in.
261
:param changes: A set() of revision ids
263
# TODO: jam 20060119 This line does *nothing*, remove it.
264
# or better yet, change _fileid_involved_by_set so
265
# that it takes the inventory weave, rather than
266
# pulling it out by itself.
267
return self._fileid_involved_by_set(changes)
269
def _fileid_involved_by_set(self, changes):
270
"""Find the set of file-ids affected by the set of revisions.
272
:param changes: A set() of revision ids.
273
:return: A set() of file ids.
275
This peaks at the Weave, interpreting each line, looking to
276
see if it mentions one of the revisions. And if so, includes
277
the file id mentioned.
278
This expects both the Weave format, and the serialization
279
to have a single line per file/directory, and to have
280
fileid="" and revision="" on that line.
282
assert isinstance(self._format, (RepositoryFormat5,
285
RepositoryFormatKnit1)), \
286
"fileid_involved only supported for branches which store inventory as unnested xml"
288
w = self.get_inventory_weave()
290
for line in w._weave:
292
# it is ugly, but it is due to the weave structure
293
if not isinstance(line, basestring): continue
295
start = line.find('file_id="')+9
296
if start < 9: continue
297
end = line.find('"', start)
299
file_id = xml.sax.saxutils.unescape(line[start:end])
301
# check if file_id is already present
302
if file_id in file_ids: continue
304
start = line.find('revision="')+10
305
if start < 10: continue
306
end = line.find('"', start)
308
revision_id = xml.sax.saxutils.unescape(line[start:end])
310
if revision_id in changes:
311
file_ids.add(file_id)
315
def get_inventory_weave(self):
316
return self.control_weaves.get_weave('inventory',
317
self.get_transaction())
2379
319
@needs_read_lock
2380
320
def get_inventory(self, revision_id):
2381
"""Get Inventory object by revision id."""
2382
return self.iter_inventories([revision_id]).next()
2384
def iter_inventories(self, revision_ids, ordering=None):
2385
"""Get many inventories by revision_ids.
2387
This will buffer some or all of the texts used in constructing the
2388
inventories in memory, but will only parse a single inventory at a
2391
:param revision_ids: The expected revision ids of the inventories.
2392
:param ordering: optional ordering, e.g. 'topological'. If not
2393
specified, the order of revision_ids will be preserved (by
2394
buffering if necessary).
2395
:return: An iterator of inventories.
321
"""Get Inventory object by hash."""
322
xml = self.get_inventory_xml(revision_id)
323
return bzrlib.xml5.serializer_v5.read_inventory_from_string(xml)
326
def get_inventory_xml(self, revision_id):
327
"""Get inventory XML as a file object."""
329
assert isinstance(revision_id, basestring), type(revision_id)
330
iw = self.get_inventory_weave()
331
return iw.get_text(iw.lookup(revision_id))
333
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
336
def get_inventory_sha1(self, revision_id):
337
"""Return the sha1 hash of the inventory entry
2397
if ((None in revision_ids)
2398
or (_mod_revision.NULL_REVISION in revision_ids)):
2399
raise ValueError('cannot get null revision inventory')
2400
return self._iter_inventories(revision_ids, ordering)
2402
def _iter_inventories(self, revision_ids, ordering):
2403
"""single-document based inventory iteration."""
2404
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2405
for text, revision_id in inv_xmls:
2406
yield self._deserialise_inventory(revision_id, text)
2408
def _iter_inventory_xmls(self, revision_ids, ordering):
2409
if ordering is None:
2410
order_as_requested = True
2411
ordering = 'unordered'
339
return self.get_revision(revision_id).inventory_sha1
342
def get_revision_inventory(self, revision_id):
343
"""Return inventory of a past revision."""
344
# TODO: Unify this with get_inventory()
345
# bzr 0.0.6 and later imposes the constraint that the inventory_id
346
# must be the same as its revision, so this is trivial.
347
if revision_id is None:
348
# This does not make sense: if there is no revision,
349
# then it is the current tree inventory surely ?!
350
# and thus get_root_id() is something that looks at the last
351
# commit on the branch, and the get_root_id is an inventory check.
352
raise NotImplementedError
353
# return Inventory(self.get_root_id())
2413
order_as_requested = False
2414
keys = [(revision_id,) for revision_id in revision_ids]
2417
if order_as_requested:
2418
key_iter = iter(keys)
2419
next_key = key_iter.next()
2420
stream = self.inventories.get_record_stream(keys, ordering, True)
2422
for record in stream:
2423
if record.storage_kind != 'absent':
2424
chunks = record.get_bytes_as('chunked')
2425
if order_as_requested:
2426
text_chunks[record.key] = chunks
2428
yield ''.join(chunks), record.key[-1]
2430
raise errors.NoSuchRevision(self, record.key)
2431
if order_as_requested:
2432
# Yield as many results as we can while preserving order.
2433
while next_key in text_chunks:
2434
chunks = text_chunks.pop(next_key)
2435
yield ''.join(chunks), next_key[-1]
2437
next_key = key_iter.next()
2438
except StopIteration:
2439
# We still want to fully consume the get_record_stream,
2440
# just in case it is not actually finished at this point
2444
def _deserialise_inventory(self, revision_id, xml):
2445
"""Transform the xml into an inventory object.
2447
:param revision_id: The expected revision id of the inventory.
2448
:param xml: A serialised inventory.
2450
result = self._serializer.read_inventory_from_string(xml, revision_id,
2451
entry_cache=self._inventory_entry_cache,
2452
return_from_cache=self._safe_to_return_from_cache)
2453
if result.revision_id != revision_id:
2454
raise AssertionError('revision id mismatch %s != %s' % (
2455
result.revision_id, revision_id))
2458
def get_serializer_format(self):
2459
return self._serializer.format_num
355
return self.get_inventory(revision_id)
2461
357
@needs_read_lock
2462
def _get_inventory_xml(self, revision_id):
2463
"""Get serialized inventory as a string."""
2464
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2466
text, revision_id = texts.next()
2467
except StopIteration:
2468
raise errors.HistoryMissing(self, 'inventory', revision_id)
2471
def get_rev_id_for_revno(self, revno, known_pair):
2472
"""Return the revision id of a revno, given a later (revno, revid)
2473
pair in the same history.
2475
:return: if found (True, revid). If the available history ran out
2476
before reaching the revno, then this returns
2477
(False, (closest_revno, closest_revid)).
2479
known_revno, known_revid = known_pair
2480
partial_history = [known_revid]
2481
distance_from_known = known_revno - revno
2482
if distance_from_known < 0:
2484
'requested revno (%d) is later than given known revno (%d)'
2485
% (revno, known_revno))
2488
self, partial_history, stop_index=distance_from_known)
2489
except errors.RevisionNotPresent, err:
2490
if err.revision_id == known_revid:
2491
# The start revision (known_revid) wasn't found.
2493
# This is a stacked repository with no fallbacks, or a there's a
2494
# left-hand ghost. Either way, even though the revision named in
2495
# the error isn't in this repo, we know it's the next step in this
2496
# left-hand history.
2497
partial_history.append(err.revision_id)
2498
if len(partial_history) <= distance_from_known:
2499
# Didn't find enough history to get a revid for the revno.
2500
earliest_revno = known_revno - len(partial_history) + 1
2501
return (False, (earliest_revno, partial_history[-1]))
2502
if len(partial_history) - 1 > distance_from_known:
2503
raise AssertionError('_iter_for_revno returned too much history')
2504
return (True, partial_history[-1])
2506
def iter_reverse_revision_history(self, revision_id):
2507
"""Iterate backwards through revision ids in the lefthand history
2509
:param revision_id: The revision id to start with. All its lefthand
2510
ancestors will be traversed.
2512
graph = self.get_graph()
2513
next_id = revision_id
2515
if next_id in (None, _mod_revision.NULL_REVISION):
2518
parents = graph.get_parent_map([next_id])[next_id]
2520
raise errors.RevisionNotPresent(next_id, self)
2522
if len(parents) == 0:
2525
next_id = parents[0]
2527
358
def is_shared(self):
2528
359
"""Return True if this repository is flagged as a shared repository."""
2529
raise NotImplementedError(self.is_shared)
2532
def reconcile(self, other=None, thorough=False):
2533
"""Reconcile this repository."""
2534
from bzrlib.reconcile import RepoReconciler
2535
reconciler = RepoReconciler(self, thorough=thorough)
2536
reconciler.reconcile()
2539
def _refresh_data(self):
2540
"""Helper called from lock_* to ensure coherency with disk.
2542
The default implementation does nothing; it is however possible
2543
for repositories to maintain loaded indices across multiple locks
2544
by checking inside their implementation of this method to see
2545
whether their indices are still valid. This depends of course on
2546
the disk format being validatable in this manner. This method is
2547
also called by the refresh_data() public interface to cause a refresh
2548
to occur while in a write lock so that data inserted by a smart server
2549
push operation is visible on the client's instance of the physical
360
# FIXME format 4-6 cannot be shared, this is technically faulty.
361
return self.control_files._transport.has('shared-storage')
2553
363
@needs_read_lock
2554
364
def revision_tree(self, revision_id):
2555
365
"""Return Tree for a revision on this branch.
2557
`revision_id` may be NULL_REVISION for the empty tree revision.
2559
revision_id = _mod_revision.ensure_null(revision_id)
367
`revision_id` may be None for the null revision, in which case
368
an `EmptyTree` is returned."""
2560
369
# TODO: refactor this to use an existing revision object
2561
370
# so we don't need to read it in twice.
2562
if revision_id == _mod_revision.NULL_REVISION:
2563
return RevisionTree(self, Inventory(root_id=None),
2564
_mod_revision.NULL_REVISION)
371
if revision_id is None or revision_id == NULL_REVISION:
2566
inv = self.get_inventory(revision_id)
374
inv = self.get_revision_inventory(revision_id)
2567
375
return RevisionTree(self, inv, revision_id)
2569
def revision_trees(self, revision_ids):
2570
"""Return Trees for revisions in this repository.
2572
:param revision_ids: a sequence of revision-ids;
2573
a revision-id may not be None or 'null:'
2575
inventories = self.iter_inventories(revision_ids)
2576
for inv in inventories:
2577
yield RevisionTree(self, inv, inv.revision_id)
2579
def _filtered_revision_trees(self, revision_ids, file_ids):
2580
"""Return Tree for a revision on this branch with only some files.
2582
:param revision_ids: a sequence of revision-ids;
2583
a revision-id may not be None or 'null:'
2584
:param file_ids: if not None, the result is filtered
2585
so that only those file-ids, their parents and their
2586
children are included.
2588
inventories = self.iter_inventories(revision_ids)
2589
for inv in inventories:
2590
# Should we introduce a FilteredRevisionTree class rather
2591
# than pre-filter the inventory here?
2592
filtered_inv = inv.filter(file_ids)
2593
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
2595
377
@needs_read_lock
2596
def get_ancestry(self, revision_id, topo_sorted=True):
378
def get_ancestry(self, revision_id):
2597
379
"""Return a list of revision-ids integrated by a revision.
2599
The first element of the list is always None, indicating the origin
2600
revision. This might change when we have history horizons, or
2601
perhaps we should have a new API.
2603
381
This is topologically sorted.
2605
if _mod_revision.is_null(revision_id):
383
if revision_id is None:
2607
385
if not self.has_revision(revision_id):
2608
386
raise errors.NoSuchRevision(self, revision_id)
2609
graph = self.get_graph()
2611
search = graph._make_breadth_first_searcher([revision_id])
387
w = self.get_inventory_weave()
388
return [None] + map(w.idx_to_name,
389
w.inclusions([w.lookup(revision_id)]))
392
def print_file(self, file, revision_id):
393
"""Print `file` to stdout.
395
FIXME RBC 20060125 as John Meinel points out this is a bad api
396
- it writes to stdout, it assumes that that is valid etc. Fix
397
by creating a new more flexible convenience function.
399
tree = self.revision_tree(revision_id)
400
# use inventory as it was in that revision
401
file_id = tree.inventory.path2id(file)
403
raise BzrError("%r is not present in revision %s" % (file, revno))
2614
found, ghosts = search.next_with_ghosts()
2615
except StopIteration:
2618
if _mod_revision.NULL_REVISION in keys:
2619
keys.remove(_mod_revision.NULL_REVISION)
2621
parent_map = graph.get_parent_map(keys)
2622
keys = tsort.topo_sort(parent_map)
2623
return [None] + list(keys)
2625
def pack(self, hint=None, clean_obsolete_packs=False):
2626
"""Compress the data within the repository.
2628
This operation only makes sense for some repository types. For other
2629
types it should be a no-op that just returns.
2631
This stub method does not require a lock, but subclasses should use
2632
@needs_write_lock as this is a long running call its reasonable to
2633
implicitly lock for the user.
2635
:param hint: If not supplied, the whole repository is packed.
2636
If supplied, the repository may use the hint parameter as a
2637
hint for the parts of the repository to pack. A hint can be
2638
obtained from the result of commit_write_group(). Out of
2639
date hints are simply ignored, because concurrent operations
2640
can obsolete them rapidly.
2642
:param clean_obsolete_packs: Clean obsolete packs immediately after
405
revno = self.revision_id_to_revno(revision_id)
406
except errors.NoSuchRevision:
407
# TODO: This should not be BzrError,
408
# but NoSuchFile doesn't fit either
409
raise BzrError('%r is not present in revision %s'
410
% (file, revision_id))
412
raise BzrError('%r is not present in revision %s'
414
tree.print_file(file_id)
2646
416
def get_transaction(self):
2647
417
return self.control_files.get_transaction()
2649
def get_parent_map(self, revision_ids):
2650
"""See graph.StackedParentsProvider.get_parent_map"""
2651
# revisions index works in keys; this just works in revisions
2652
# therefore wrap and unwrap
2655
for revision_id in revision_ids:
2656
if revision_id == _mod_revision.NULL_REVISION:
2657
result[revision_id] = ()
2658
elif revision_id is None:
2659
raise ValueError('get_parent_map(None) is not valid')
2661
query_keys.append((revision_id ,))
2662
for ((revision_id,), parent_keys) in \
2663
self.revisions.get_parent_map(query_keys).iteritems():
2665
result[revision_id] = tuple([parent_revid
2666
for (parent_revid,) in parent_keys])
2668
result[revision_id] = (_mod_revision.NULL_REVISION,)
2671
def _make_parents_provider(self):
2675
def get_known_graph_ancestry(self, revision_ids):
2676
"""Return the known graph for a set of revision ids and their ancestors.
2678
st = static_tuple.StaticTuple
2679
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2680
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2681
return graph.GraphThunkIdsToKeys(known_graph)
2683
def get_graph(self, other_repository=None):
2684
"""Return the graph walker for this repository format"""
2685
parents_provider = self._make_parents_provider()
2686
if (other_repository is not None and
2687
not self.has_same_location(other_repository)):
2688
parents_provider = graph.StackedParentsProvider(
2689
[parents_provider, other_repository._make_parents_provider()])
2690
return graph.Graph(parents_provider)
2692
def _get_versioned_file_checker(self, text_key_references=None,
2694
"""Return an object suitable for checking versioned files.
2696
:param text_key_references: if non-None, an already built
2697
dictionary mapping text keys ((fileid, revision_id) tuples)
2698
to whether they were referred to by the inventory of the
2699
revision_id that they contain. If None, this will be
2701
:param ancestors: Optional result from
2702
self.get_graph().get_parent_map(self.all_revision_ids()) if already
2705
return _VersionedFileChecker(self,
2706
text_key_references=text_key_references, ancestors=ancestors)
2708
def revision_ids_to_search_result(self, result_set):
2709
"""Convert a set of revision ids to a graph SearchResult."""
2710
result_parents = set()
2711
for parents in self.get_graph().get_parent_map(
2712
result_set).itervalues():
2713
result_parents.update(parents)
2714
included_keys = result_set.intersection(result_parents)
2715
start_keys = result_set.difference(included_keys)
2716
exclude_keys = result_parents.difference(result_set)
2717
result = graph.SearchResult(start_keys, exclude_keys,
2718
len(result_set), result_set)
2721
419
@needs_write_lock
2722
420
def set_make_working_trees(self, new_value):
2723
421
"""Set the policy flag for making working trees when creating branches.
2728
426
:param new_value: True to restore the default, False to disable making
2731
raise NotImplementedError(self.set_make_working_trees)
429
# FIXME: split out into a new class/strategy ?
430
if isinstance(self._format, (RepositoryFormat4,
433
raise NotImplementedError(self.set_make_working_trees)
436
self.control_files._transport.delete('no-working-trees')
437
except errors.NoSuchFile:
440
self.control_files.put_utf8('no-working-trees', '')
2733
442
def make_working_trees(self):
2734
443
"""Returns the policy for making working trees on new branches."""
2735
raise NotImplementedError(self.make_working_trees)
444
# FIXME: split out into a new class/strategy ?
445
if isinstance(self._format, (RepositoryFormat4,
449
return not self.control_files._transport.has('no-working-trees')
2737
451
@needs_write_lock
2738
452
def sign_revision(self, revision_id, gpg_strategy):
2739
453
plaintext = Testament.from_revision(self, revision_id).as_short_text()
2740
454
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2743
def has_signature_for_revision_id(self, revision_id):
2744
"""Query for a revision signature for revision_id in the repository."""
2745
if not self.has_revision(revision_id):
2746
raise errors.NoSuchRevision(self, revision_id)
2747
sig_present = (1 == len(
2748
self.signatures.get_parent_map([(revision_id,)])))
2752
def get_signature_text(self, revision_id):
2753
"""Return the text for a signature."""
2754
stream = self.signatures.get_record_stream([(revision_id,)],
2756
record = stream.next()
2757
if record.storage_kind == 'absent':
2758
raise errors.NoSuchRevision(self, revision_id)
2759
return record.get_bytes_as('fulltext')
2762
def check(self, revision_ids=None, callback_refs=None, check_repo=True):
2763
"""Check consistency of all history of given revision_ids.
2765
Different repository implementations should override _check().
2767
:param revision_ids: A non-empty list of revision_ids whose ancestry
2768
will be checked. Typically the last revision_id of a branch.
2769
:param callback_refs: A dict of check-refs to resolve and callback
2770
the check/_check method on the items listed as wanting the ref.
2772
:param check_repo: If False do not check the repository contents, just
2773
calculate the data callback_refs requires and call them back.
2775
return self._check(revision_ids, callback_refs=callback_refs,
2776
check_repo=check_repo)
2778
def _check(self, revision_ids, callback_refs, check_repo):
2779
result = check.Check(self, check_repo=check_repo)
2780
result.check(callback_refs)
2783
def _warn_if_deprecated(self, branch=None):
2784
global _deprecation_warning_done
2785
if _deprecation_warning_done:
2789
conf = config.GlobalConfig()
2791
conf = branch.get_config()
2792
if conf.suppress_warning('format_deprecation'):
2794
warning("Format %s for %s is deprecated -"
2795
" please use 'bzr upgrade' to get better performance"
2796
% (self._format, self.bzrdir.transport.base))
2798
_deprecation_warning_done = True
2800
def supports_rich_root(self):
2801
return self._format.rich_root_data
2803
def _check_ascii_revisionid(self, revision_id, method):
2804
"""Private helper for ascii-only repositories."""
2805
# weave repositories refuse to store revisionids that are non-ascii.
2806
if revision_id is not None:
2807
# weaves require ascii revision ids.
2808
if isinstance(revision_id, unicode):
2810
revision_id.encode('ascii')
2811
except UnicodeEncodeError:
2812
raise errors.NonAsciiRevisionId(method, self)
2815
revision_id.decode('ascii')
2816
except UnicodeDecodeError:
2817
raise errors.NonAsciiRevisionId(method, self)
2819
def revision_graph_can_have_wrong_parents(self):
2820
"""Is it possible for this repository to have a revision graph with
2823
If True, then this repository must also implement
2824
_find_inconsistent_revision_parents so that check and reconcile can
2825
check for inconsistencies before proceeding with other checks that may
2826
depend on the revision index being consistent.
2828
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2831
# remove these delegates a while after bzr 0.15
2832
def __make_delegated(name, from_module):
2833
def _deprecated_repository_forwarder():
2834
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2835
% (name, from_module),
2838
m = __import__(from_module, globals(), locals(), [name])
2840
return getattr(m, name)
2841
except AttributeError:
2842
raise AttributeError('module %s has no name %s'
2844
globals()[name] = _deprecated_repository_forwarder
2847
'AllInOneRepository',
2848
'WeaveMetaDirRepository',
2849
'PreSplitOutRepositoryFormat',
2850
'RepositoryFormat4',
2851
'RepositoryFormat5',
2852
'RepositoryFormat6',
2853
'RepositoryFormat7',
2855
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2859
'RepositoryFormatKnit',
2860
'RepositoryFormatKnit1',
2862
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2865
def install_revision(repository, rev, revision_tree):
2866
"""Install all revision data into a repository."""
2867
install_revisions(repository, [(rev, revision_tree, None)])
2870
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2871
"""Install all revision data into a repository.
2873
Accepts an iterable of revision, tree, signature tuples. The signature
2876
repository.start_write_group()
2878
inventory_cache = lru_cache.LRUCache(10)
2879
for n, (revision, revision_tree, signature) in enumerate(iterable):
2880
_install_revision(repository, revision, revision_tree, signature,
2883
pb.update('Transferring revisions', n + 1, num_revisions)
2885
repository.abort_write_group()
2888
repository.commit_write_group()
2891
def _install_revision(repository, rev, revision_tree, signature,
2893
"""Install all revision data into a repository."""
2894
present_parents = []
2896
for p_id in rev.parent_ids:
2897
if repository.has_revision(p_id):
2898
present_parents.append(p_id)
2899
parent_trees[p_id] = repository.revision_tree(p_id)
2901
parent_trees[p_id] = repository.revision_tree(
2902
_mod_revision.NULL_REVISION)
2904
inv = revision_tree.inventory
2905
entries = inv.iter_entries()
2906
# backwards compatibility hack: skip the root id.
2907
if not repository.supports_rich_root():
2908
path, root = entries.next()
2909
if root.revision != rev.revision_id:
2910
raise errors.IncompatibleRevision(repr(repository))
2912
for path, ie in entries:
2913
text_keys[(ie.file_id, ie.revision)] = ie
2914
text_parent_map = repository.texts.get_parent_map(text_keys)
2915
missing_texts = set(text_keys) - set(text_parent_map)
2916
# Add the texts that are not already present
2917
for text_key in missing_texts:
2918
ie = text_keys[text_key]
2920
# FIXME: TODO: The following loop overlaps/duplicates that done by
2921
# commit to determine parents. There is a latent/real bug here where
2922
# the parents inserted are not those commit would do - in particular
2923
# they are not filtered by heads(). RBC, AB
2924
for revision, tree in parent_trees.iteritems():
2925
if ie.file_id not in tree:
2927
parent_id = tree.inventory[ie.file_id].revision
2928
if parent_id in text_parents:
2930
text_parents.append((ie.file_id, parent_id))
2931
lines = revision_tree.get_file(ie.file_id).readlines()
2932
repository.texts.add_lines(text_key, text_parents, lines)
2934
# install the inventory
2935
if repository._format._commit_inv_deltas and len(rev.parent_ids):
2936
# Cache this inventory
2937
inventory_cache[rev.revision_id] = inv
2939
basis_inv = inventory_cache[rev.parent_ids[0]]
2941
repository.add_inventory(rev.revision_id, inv, present_parents)
2943
delta = inv._make_delta(basis_inv)
2944
repository.add_inventory_by_delta(rev.parent_ids[0], delta,
2945
rev.revision_id, present_parents)
2947
repository.add_inventory(rev.revision_id, inv, present_parents)
2948
except errors.RevisionAlreadyPresent:
2950
if signature is not None:
2951
repository.add_signature_text(rev.revision_id, signature)
2952
repository.add_revision(rev.revision_id, rev, inv)
457
class AllInOneRepository(Repository):
458
"""Legacy support - the repository behaviour for all-in-one branches."""
460
def __init__(self, _format, a_bzrdir, revision_store):
461
# we reuse one control files instance.
462
dir_mode = a_bzrdir._control_files._dir_mode
463
file_mode = a_bzrdir._control_files._file_mode
465
def get_weave(name, prefixed=False):
467
name = safe_unicode(name)
470
relpath = a_bzrdir._control_files._escape(name)
471
weave_transport = a_bzrdir._control_files._transport.clone(relpath)
472
ws = WeaveStore(weave_transport, prefixed=prefixed,
475
if a_bzrdir._control_files._transport.should_cache():
476
ws.enable_cache = True
479
def get_store(name, compressed=True, prefixed=False):
480
# FIXME: This approach of assuming stores are all entirely compressed
481
# or entirely uncompressed is tidy, but breaks upgrade from
482
# some existing branches where there's a mixture; we probably
483
# still want the option to look for both.
484
relpath = a_bzrdir._control_files._escape(name)
485
store = TextStore(a_bzrdir._control_files._transport.clone(relpath),
486
prefixed=prefixed, compressed=compressed,
489
#if self._transport.should_cache():
490
# cache_path = os.path.join(self.cache_root, name)
491
# os.mkdir(cache_path)
492
# store = bzrlib.store.CachedStore(store, cache_path)
495
# not broken out yet because the controlweaves|inventory_store
496
# and text_store | weave_store bits are still different.
497
if isinstance(_format, RepositoryFormat4):
498
self.inventory_store = get_store('inventory-store')
499
self.text_store = get_store('text-store')
500
elif isinstance(_format, RepositoryFormat5):
501
self.control_weaves = get_weave('')
502
self.weave_store = get_weave('weaves')
503
elif isinstance(_format, RepositoryFormat6):
504
self.control_weaves = get_weave('')
505
self.weave_store = get_weave('weaves', prefixed=True)
507
raise errors.BzrError('unreachable code: unexpected repository'
509
revision_store.register_suffix('sig')
510
super(AllInOneRepository, self).__init__(_format, a_bzrdir, a_bzrdir._control_files, revision_store)
2955
513
class MetaDirRepository(Repository):
2956
"""Repositories in the new meta-dir layout.
2958
:ivar _transport: Transport for access to repository control files,
2959
typically pointing to .bzr/repository.
2962
def __init__(self, _format, a_bzrdir, control_files):
2963
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2964
self._transport = control_files._transport
2966
def is_shared(self):
2967
"""Return True if this repository is flagged as a shared repository."""
2968
return self._transport.has('shared-storage')
2971
def set_make_working_trees(self, new_value):
2972
"""Set the policy flag for making working trees when creating branches.
2974
This only applies to branches that use this repository.
2976
The default is 'True'.
2977
:param new_value: True to restore the default, False to disable making
2982
self._transport.delete('no-working-trees')
2983
except errors.NoSuchFile:
514
"""Repositories in the new meta-dir layout."""
516
def __init__(self, _format, a_bzrdir, control_files, revision_store):
517
super(MetaDirRepository, self).__init__(_format,
522
dir_mode = self.control_files._dir_mode
523
file_mode = self.control_files._file_mode
525
def get_weave(name, prefixed=False):
527
name = safe_unicode(name)
530
relpath = self.control_files._escape(name)
531
weave_transport = self.control_files._transport.clone(relpath)
532
ws = WeaveStore(weave_transport, prefixed=prefixed,
535
if self.control_files._transport.should_cache():
536
ws.enable_cache = True
539
if isinstance(self._format, RepositoryFormat7):
540
self.control_weaves = get_weave('')
541
self.weave_store = get_weave('weaves', prefixed=True)
542
elif isinstance(self._format, RepositoryFormatKnit1):
543
self.control_weaves = get_weave('')
544
self.weave_store = get_weave('knits', prefixed=True)
2986
self._transport.put_bytes('no-working-trees', '',
2987
mode=self.bzrdir._get_file_mode())
2989
def make_working_trees(self):
2990
"""Returns the policy for making working trees on new branches."""
2991
return not self._transport.has('no-working-trees')
2994
class MetaDirVersionedFileRepository(MetaDirRepository):
2995
"""Repositories in a meta-dir, that work via versioned file objects."""
2997
def __init__(self, _format, a_bzrdir, control_files):
2998
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
3002
network_format_registry = registry.FormatRegistry()
3003
"""Registry of formats indexed by their network name.
3005
The network name for a repository format is an identifier that can be used when
3006
referring to formats with smart server operations. See
3007
RepositoryFormat.network_name() for more detail.
3011
format_registry = registry.FormatRegistry(network_format_registry)
3012
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
3014
This can contain either format instances themselves, or classes/factories that
3015
can be called to obtain one.
3019
#####################################################################
3020
# Repository Formats
546
raise errors.BzrError('unreachable code: unexpected repository'
3022
550
class RepositoryFormat(object):
3023
551
"""A repository format.
3025
Formats provide four things:
553
Formats provide three things:
3026
554
* An initialization routine to construct repository data on disk.
3027
* a optional format string which is used when the BzrDir supports
555
* a format string which is used when the BzrDir supports versioned
3029
557
* an open routine which returns a Repository instance.
3030
* A network name for referring to the format in smart server RPC
3033
There is one and only one Format subclass for each on-disk format. But
3034
there can be one Repository subclass that is used for several different
3035
formats. The _format attribute on a Repository instance can be used to
3036
determine the disk format.
3038
Formats are placed in a registry by their format string for reference
3039
during opening. These should be subclasses of RepositoryFormat for
559
Formats are placed in an dict by their format string for reference
560
during opening. These should be subclasses of RepositoryFormat
3042
563
Once a format is deprecated, just deprecate the initialize and open
3043
methods on the format class. Do not deprecate the object, as the
3044
object may be created even when a repository instance hasn't been
564
methods on the format class. Do not deprecate the object, as the
565
object will be created every system load.
3047
567
Common instance attributes:
3048
568
_matchingbzrdir - the bzrdir format that the repository format was
3049
569
originally written to work with. This can be used if manually
3050
570
constructing a bzrdir and repository, or more commonly for test suite
3054
# Set to True or False in derived classes. True indicates that the format
3055
# supports ghosts gracefully.
3056
supports_ghosts = None
3057
# Can this repository be given external locations to lookup additional
3058
# data. Set to True or False in derived classes.
3059
supports_external_lookups = None
3060
# Does this format support CHK bytestring lookups. Set to True or False in
3062
supports_chks = None
3063
# Should commit add an inventory, or an inventory delta to the repository.
3064
_commit_inv_deltas = True
3065
# What order should fetch operations request streams in?
3066
# The default is unordered as that is the cheapest for an origin to
3068
_fetch_order = 'unordered'
3069
# Does this repository format use deltas that can be fetched as-deltas ?
3070
# (E.g. knits, where the knit deltas can be transplanted intact.
3071
# We default to False, which will ensure that enough data to get
3072
# a full text out of any fetch stream will be grabbed.
3073
_fetch_uses_deltas = False
3074
# Should fetch trigger a reconcile after the fetch? Only needed for
3075
# some repository formats that can suffer internal inconsistencies.
3076
_fetch_reconcile = False
3077
# Does this format have < O(tree_size) delta generation. Used to hint what
3078
# code path for commit, amongst other things.
3080
# Does doing a pack operation compress data? Useful for the pack UI command
3081
# (so if there is one pack, the operation can still proceed because it may
3082
# help), and for fetching when data won't have come from the same
3084
pack_compresses = False
3085
# Does the repository inventory storage understand references to trees?
3086
supports_tree_reference = None
3087
# Is the format experimental ?
3088
experimental = False
3091
return "%s()" % self.__class__.__name__
3093
def __eq__(self, other):
3094
# format objects are generally stateless
3095
return isinstance(other, self.__class__)
3097
def __ne__(self, other):
3098
return not self == other
574
_default_format = None
575
"""The default format used for new repositories."""
578
"""The known formats."""
3101
581
def find_format(klass, a_bzrdir):
3102
"""Return the format for the repository object in a_bzrdir.
3104
This is used by bzr native formats that have a "format" file in
3105
the repository. Other methods may be used by different types of
582
"""Return the format for the repository object in a_bzrdir."""
3109
584
transport = a_bzrdir.get_repository_transport(None)
3110
format_string = transport.get_bytes("format")
3111
return format_registry.get(format_string)
585
format_string = transport.get("format").read()
586
return klass._formats[format_string]
3112
587
except errors.NoSuchFile:
3113
588
raise errors.NoRepositoryPresent(a_bzrdir)
3114
589
except KeyError:
3115
raise errors.UnknownFormatError(format=format_string,
3119
def register_format(klass, format):
3120
format_registry.register(format.get_format_string(), format)
3123
def unregister_format(klass, format):
3124
format_registry.remove(format.get_format_string())
590
raise errors.UnknownFormatError(format_string)
3127
593
def get_default_format(klass):
3128
594
"""Return the current default format."""
3129
from bzrlib import bzrdir
3130
return bzrdir.format_registry.make_bzrdir('default').repository_format
595
return klass._default_format
3132
597
def get_format_string(self):
3133
598
"""Return the ASCII format string that identifies this format.
3135
Note that in pre format ?? repositories the format string is
600
Note that in pre format ?? repositories the format string is
3136
601
not permitted nor written to disk.
3138
603
raise NotImplementedError(self.get_format_string)
3140
def get_format_description(self):
3141
"""Return the short description for this format."""
3142
raise NotImplementedError(self.get_format_description)
605
def _get_revision_store(self, repo_transport, control_files):
606
"""Return the revision store object for this a_bzrdir."""
607
raise NotImplementedError(self._get_revision_store)
3144
# TODO: this shouldn't be in the base class, it's specific to things that
3145
# use weaves or knits -- mbp 20070207
3146
def _get_versioned_file_store(self,
3151
versionedfile_class=None,
3152
versionedfile_kwargs={},
3154
if versionedfile_class is None:
3155
versionedfile_class = self._versionedfile_class
3156
weave_transport = control_files._transport.clone(name)
609
def _get_rev_store(self,
615
"""Common logic for getting a revision store for a repository.
617
see self._get_revision_store for the method to
618
get the store for a repository.
621
name = safe_unicode(name)
3157
624
dir_mode = control_files._dir_mode
3158
625
file_mode = control_files._file_mode
3159
return VersionedFileStore(weave_transport, prefixed=prefixed,
626
revision_store =TextStore(transport.clone(name),
628
compressed=compressed,
3160
629
dir_mode=dir_mode,
3161
file_mode=file_mode,
3162
versionedfile_class=versionedfile_class,
3163
versionedfile_kwargs=versionedfile_kwargs,
631
revision_store.register_suffix('sig')
632
return revision_store
3166
634
def initialize(self, a_bzrdir, shared=False):
3167
635
"""Initialize a repository of this format in a_bzrdir.
3169
637
:param a_bzrdir: The bzrdir to put the new repository in it.
3170
638
:param shared: The repository should be initialized as a sharable one.
3171
:returns: The new repository object.
3173
640
This may raise UninitializableFormat if shared repository are not
3174
641
compatible the a_bzrdir.
3176
raise NotImplementedError(self.initialize)
3178
644
def is_supported(self):
3179
645
"""Is this format supported?
3181
647
Supported formats must be initializable and openable.
3182
Unsupported formats may not support initialization or committing or
648
Unsupported formats may not support initialization or committing or
3183
649
some other features depending on the reason for not being supported.
3187
def network_name(self):
3188
"""A simple byte string uniquely identifying this format for RPC calls.
3190
MetaDir repository formats use their disk format string to identify the
3191
repository over the wire. All in one formats such as bzr < 0.8, and
3192
foreign formats like svn/git and hg should use some marker which is
3193
unique and immutable.
3195
raise NotImplementedError(self.network_name)
3197
def check_conversion_target(self, target_format):
3198
if self.rich_root_data and not target_format.rich_root_data:
3199
raise errors.BadConversionTarget(
3200
'Does not support rich root data.', target_format,
3202
if (self.supports_tree_reference and
3203
not getattr(target_format, 'supports_tree_reference', False)):
3204
raise errors.BadConversionTarget(
3205
'Does not support nested trees', target_format,
3208
653
def open(self, a_bzrdir, _found=False):
3209
654
"""Return an instance of this format for the bzrdir a_bzrdir.
3211
656
_found is a private parameter, do not use it.
3213
658
raise NotImplementedError(self.open)
3215
def _run_post_repo_init_hooks(self, repository, a_bzrdir, shared):
3216
from bzrlib.bzrdir import BzrDir, RepoInitHookParams
3217
hooks = BzrDir.hooks['post_repo_init']
3220
params = RepoInitHookParams(repository, self, a_bzrdir, shared)
661
def register_format(klass, format):
662
klass._formats[format.get_format_string()] = format
665
def set_default_format(klass, format):
666
klass._default_format = format
669
def unregister_format(klass, format):
670
assert klass._formats[format.get_format_string()] is format
671
del klass._formats[format.get_format_string()]
674
class PreSplitOutRepositoryFormat(RepositoryFormat):
675
"""Base class for the pre split out repository formats."""
677
def initialize(self, a_bzrdir, shared=False, _internal=False):
678
"""Create a weave repository.
680
TODO: when creating split out bzr branch formats, move this to a common
681
base for Format5, Format6. or something like that.
683
from bzrlib.weavefile import write_weave_v5
684
from bzrlib.weave import Weave
687
raise errors.IncompatibleFormat(self, a_bzrdir._format)
690
# always initialized when the bzrdir is.
691
return self.open(a_bzrdir, _found=True)
693
# Create an empty weave
695
bzrlib.weavefile.write_weave_v5(Weave(), sio)
696
empty_weave = sio.getvalue()
698
mutter('creating repository in %s.', a_bzrdir.transport.base)
699
dirs = ['revision-store', 'weaves']
700
lock_file = 'branch-lock'
701
files = [('inventory.weave', StringIO(empty_weave)),
704
# FIXME: RBC 20060125 dont peek under the covers
705
# NB: no need to escape relative paths that are url safe.
706
control_files = LockableFiles(a_bzrdir.transport, 'branch-lock')
707
control_files.lock_write()
708
control_files._transport.mkdir_multi(dirs,
709
mode=control_files._dir_mode)
711
for file, content in files:
712
control_files.put(file, content)
714
control_files.unlock()
715
return self.open(a_bzrdir, _found=True)
717
def open(self, a_bzrdir, _found=False):
718
"""See RepositoryFormat.open()."""
720
# we are being called directly and must probe.
721
raise NotImplementedError
723
repo_transport = a_bzrdir.get_repository_transport(None)
724
control_files = a_bzrdir._control_files
725
revision_store = self._get_revision_store(repo_transport, control_files)
726
return AllInOneRepository(_format=self,
728
revision_store=revision_store)
731
class RepositoryFormat4(PreSplitOutRepositoryFormat):
732
"""Bzr repository format 4.
734
This repository format has:
736
- TextStores for texts, inventories,revisions.
738
This format is deprecated: it indexes texts using a text id which is
739
removed in format 5; initializationa and write support for this format
744
super(RepositoryFormat4, self).__init__()
745
self._matchingbzrdir = bzrlib.bzrdir.BzrDirFormat4()
747
def initialize(self, url, shared=False, _internal=False):
748
"""Format 4 branches cannot be created."""
749
raise errors.UninitializableFormat(self)
751
def is_supported(self):
752
"""Format 4 is not supported.
754
It is not supported because the model changed from 4 to 5 and the
755
conversion logic is expensive - so doing it on the fly was not
760
def _get_revision_store(self, repo_transport, control_files):
761
"""See RepositoryFormat._get_revision_store()."""
762
return self._get_rev_store(repo_transport,
767
class RepositoryFormat5(PreSplitOutRepositoryFormat):
768
"""Bzr control format 5.
770
This repository format has:
771
- weaves for file texts and inventory
773
- TextStores for revisions and signatures.
777
super(RepositoryFormat5, self).__init__()
778
self._matchingbzrdir = bzrlib.bzrdir.BzrDirFormat5()
780
def _get_revision_store(self, repo_transport, control_files):
781
"""See RepositoryFormat._get_revision_store()."""
782
"""Return the revision store object for this a_bzrdir."""
783
return self._get_rev_store(repo_transport,
789
class RepositoryFormat6(PreSplitOutRepositoryFormat):
790
"""Bzr control format 6.
792
This repository format has:
793
- weaves for file texts and inventory
794
- hash subdirectory based stores.
795
- TextStores for revisions and signatures.
799
super(RepositoryFormat6, self).__init__()
800
self._matchingbzrdir = bzrlib.bzrdir.BzrDirFormat6()
802
def _get_revision_store(self, repo_transport, control_files):
803
"""See RepositoryFormat._get_revision_store()."""
804
return self._get_rev_store(repo_transport,
3225
811
class MetaDirRepositoryFormat(RepositoryFormat):
3226
"""Common base class for the new repositories using the metadir layout."""
3228
rich_root_data = False
3229
supports_tree_reference = False
3230
supports_external_lookups = False
3233
def _matchingbzrdir(self):
3234
matching = bzrdir.BzrDirMetaFormat1()
3235
matching.repository_format = self
812
"""Common base class for the new repositories using the metadir layour."""
3238
814
def __init__(self):
3239
815
super(MetaDirRepositoryFormat, self).__init__()
816
self._matchingbzrdir = bzrlib.bzrdir.BzrDirMetaFormat1()
3241
818
def _create_control_files(self, a_bzrdir):
3242
819
"""Create the required files and the initial control_files object."""
3243
# FIXME: RBC 20060125 don't peek under the covers
820
# FIXME: RBC 20060125 dont peek under the covers
3244
821
# NB: no need to escape relative paths that are url safe.
3245
823
repository_transport = a_bzrdir.get_repository_transport(self)
3246
control_files = lockable_files.LockableFiles(repository_transport,
3247
'lock', lockdir.LockDir)
3248
control_files.create_lock()
824
repository_transport.put(lock_file, StringIO()) # TODO get the file mode from the bzrdir lock files., mode=file_mode)
825
control_files = LockableFiles(repository_transport, 'lock')
3249
826
return control_files
828
def _get_revision_store(self, repo_transport, control_files):
829
"""See RepositoryFormat._get_revision_store()."""
830
return self._get_rev_store(repo_transport,
837
def open(self, a_bzrdir, _found=False, _override_transport=None):
838
"""See RepositoryFormat.open().
840
:param _override_transport: INTERNAL USE ONLY. Allows opening the
841
repository at a slightly different url
842
than normal. I.e. during 'upgrade'.
845
format = RepositoryFormat.find_format(a_bzrdir)
846
assert format.__class__ == self.__class__
847
if _override_transport is not None:
848
repo_transport = _override_transport
850
repo_transport = a_bzrdir.get_repository_transport(None)
851
control_files = LockableFiles(repo_transport, 'lock')
852
revision_store = self._get_revision_store(repo_transport, control_files)
853
return MetaDirRepository(_format=self,
855
control_files=control_files,
856
revision_store=revision_store)
3251
858
def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
3252
859
"""Upload the initial blank content."""
3253
860
control_files = self._create_control_files(a_bzrdir)
3254
861
control_files.lock_write()
3255
transport = control_files._transport
3257
utf8_files += [('shared-storage', '')]
862
control_files._transport.mkdir_multi(dirs,
863
mode=control_files._dir_mode)
3259
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
3260
for (filename, content_stream) in files:
3261
transport.put_file(filename, content_stream,
3262
mode=a_bzrdir._get_file_mode())
3263
for (filename, content_bytes) in utf8_files:
3264
transport.put_bytes_non_atomic(filename, content_bytes,
3265
mode=a_bzrdir._get_file_mode())
865
for file, content in files:
866
control_files.put(file, content)
867
for file, content in utf8_files:
868
control_files.put_utf8(file, content)
870
control_files.put_utf8('shared-storage', '')
3267
872
control_files.unlock()
3269
def network_name(self):
3270
"""Metadir formats have matching disk and network format strings."""
3271
return self.get_format_string()
3274
# Pre-0.8 formats that don't have a disk format string (because they are
3275
# versioned by the matching control directory). We use the control directories
3276
# disk format string as a key for the network_name because they meet the
3277
# constraints (simple string, unique, immutable).
3278
network_format_registry.register_lazy(
3279
"Bazaar-NG branch, format 5\n",
3280
'bzrlib.repofmt.weaverepo',
3281
'RepositoryFormat5',
3283
network_format_registry.register_lazy(
3284
"Bazaar-NG branch, format 6\n",
3285
'bzrlib.repofmt.weaverepo',
3286
'RepositoryFormat6',
3289
# formats which have no format string are not discoverable or independently
3290
# creatable on disk, so are not registered in format_registry. They're
3291
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
3292
# needed, it's constructed directly by the BzrDir. Non-native formats where
3293
# the repository is not separately opened are similar.
3295
format_registry.register_lazy(
3296
'Bazaar-NG Repository format 7',
3297
'bzrlib.repofmt.weaverepo',
3301
format_registry.register_lazy(
3302
'Bazaar-NG Knit Repository Format 1',
3303
'bzrlib.repofmt.knitrepo',
3304
'RepositoryFormatKnit1',
3307
format_registry.register_lazy(
3308
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
3309
'bzrlib.repofmt.knitrepo',
3310
'RepositoryFormatKnit3',
3313
format_registry.register_lazy(
3314
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
3315
'bzrlib.repofmt.knitrepo',
3316
'RepositoryFormatKnit4',
3319
# Pack-based formats. There is one format for pre-subtrees, and one for
3320
# post-subtrees to allow ease of testing.
3321
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
3322
format_registry.register_lazy(
3323
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
3324
'bzrlib.repofmt.pack_repo',
3325
'RepositoryFormatKnitPack1',
3327
format_registry.register_lazy(
3328
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
3329
'bzrlib.repofmt.pack_repo',
3330
'RepositoryFormatKnitPack3',
3332
format_registry.register_lazy(
3333
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
3334
'bzrlib.repofmt.pack_repo',
3335
'RepositoryFormatKnitPack4',
3337
format_registry.register_lazy(
3338
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
3339
'bzrlib.repofmt.pack_repo',
3340
'RepositoryFormatKnitPack5',
3342
format_registry.register_lazy(
3343
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
3344
'bzrlib.repofmt.pack_repo',
3345
'RepositoryFormatKnitPack5RichRoot',
3347
format_registry.register_lazy(
3348
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
3349
'bzrlib.repofmt.pack_repo',
3350
'RepositoryFormatKnitPack5RichRootBroken',
3352
format_registry.register_lazy(
3353
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
3354
'bzrlib.repofmt.pack_repo',
3355
'RepositoryFormatKnitPack6',
3357
format_registry.register_lazy(
3358
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
3359
'bzrlib.repofmt.pack_repo',
3360
'RepositoryFormatKnitPack6RichRoot',
3363
# Development formats.
3364
# Obsolete but kept pending a CHK based subtree format.
3365
format_registry.register_lazy(
3366
("Bazaar development format 2 with subtree support "
3367
"(needs bzr.dev from before 1.8)\n"),
3368
'bzrlib.repofmt.pack_repo',
3369
'RepositoryFormatPackDevelopment2Subtree',
3372
# 1.14->1.16 go below here
3373
format_registry.register_lazy(
3374
'Bazaar development format - group compression and chk inventory'
3375
' (needs bzr.dev from 1.14)\n',
3376
'bzrlib.repofmt.groupcompress_repo',
3377
'RepositoryFormatCHK1',
3380
format_registry.register_lazy(
3381
'Bazaar development format - chk repository with bencode revision '
3382
'serialization (needs bzr.dev from 1.16)\n',
3383
'bzrlib.repofmt.groupcompress_repo',
3384
'RepositoryFormatCHK2',
3386
format_registry.register_lazy(
3387
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3388
'bzrlib.repofmt.groupcompress_repo',
3389
'RepositoryFormat2a',
3393
class InterRepository(InterObject):
875
class RepositoryFormat7(MetaDirRepositoryFormat):
878
This repository format has:
879
- weaves for file texts and inventory
880
- hash subdirectory based stores.
881
- TextStores for revisions and signatures.
882
- a format marker of its own
883
- an optional 'shared-storage' flag
884
- an optional 'no-working-trees' flag
887
def get_format_string(self):
888
"""See RepositoryFormat.get_format_string()."""
889
return "Bazaar-NG Repository format 7"
891
def initialize(self, a_bzrdir, shared=False):
892
"""Create a weave repository.
894
:param shared: If true the repository will be initialized as a shared
897
from bzrlib.weavefile import write_weave_v5
898
from bzrlib.weave import Weave
900
# Create an empty weave
902
bzrlib.weavefile.write_weave_v5(Weave(), sio)
903
empty_weave = sio.getvalue()
905
mutter('creating repository in %s.', a_bzrdir.transport.base)
906
dirs = ['revision-store', 'weaves']
907
files = [('inventory.weave', StringIO(empty_weave)),
909
utf8_files = [('format', self.get_format_string())]
911
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
912
return self.open(a_bzrdir=a_bzrdir, _found=True)
915
class RepositoryFormatKnit1(MetaDirRepositoryFormat):
916
"""Bzr repository knit format 1.
918
This repository format has:
919
- knits for file texts and inventory
920
- hash subdirectory based stores.
921
- knits for revisions and signatures
922
- TextStores for revisions and signatures.
923
- a format marker of its own
924
- an optional 'shared-storage' flag
925
- an optional 'no-working-trees' flag
928
def get_format_string(self):
929
"""See RepositoryFormat.get_format_string()."""
930
return "Bazaar-NG Knit Repository Format 1"
932
def initialize(self, a_bzrdir, shared=False):
933
"""Create a knit format 1 repository.
935
:param shared: If true the repository will be initialized as a shared
937
XXX NOTE that this current uses a Weave for testing and will become
938
A Knit in due course.
940
from bzrlib.weavefile import write_weave_v5
941
from bzrlib.weave import Weave
943
# Create an empty weave
945
bzrlib.weavefile.write_weave_v5(Weave(), sio)
946
empty_weave = sio.getvalue()
948
mutter('creating repository in %s.', a_bzrdir.transport.base)
949
dirs = ['revision-store', 'knits']
950
files = [('inventory.weave', StringIO(empty_weave)),
952
utf8_files = [('format', self.get_format_string())]
954
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
955
return self.open(a_bzrdir=a_bzrdir, _found=True)
958
# formats which have no format string are not discoverable
959
# and not independently creatable, so are not registered.
960
_default_format = RepositoryFormat7()
961
RepositoryFormat.register_format(_default_format)
962
RepositoryFormat.register_format(RepositoryFormatKnit1())
963
RepositoryFormat.set_default_format(_default_format)
964
_legacy_formats = [RepositoryFormat4(),
969
class InterRepository(object):
3394
970
"""This class represents operations taking place between two repositories.
3396
972
Its instances have methods like copy_content and fetch, and contain
3397
references to the source and target repositories these operations can be
973
references to the source and target repositories these operations can be
3400
976
Often we will provide convenience methods on 'repository' which carry out
3401
977
operations with another repository - they will always forward to
3402
978
InterRepository.get(other).method_name(parameters).
980
# XXX: FIXME: FUTURE: robertc
981
# testing of these probably requires a factory in optimiser type, and
982
# then a test adapter to test each type thoroughly.
3405
_walk_to_common_revisions_batch_size = 50
3407
986
"""The available optimised InterRepository types."""
988
def __init__(self, source, target):
989
"""Construct a default InterRepository instance. Please use 'get'.
991
Only subclasses of InterRepository should call
992
InterRepository.__init__ - clients should call InterRepository.get
993
instead which will create an optimised InterRepository if possible.
3409
998
@needs_write_lock
3410
def copy_content(self, revision_id=None):
999
def copy_content(self, revision_id=None, basis=None):
3411
1000
"""Make a complete copy of the content in self into destination.
3413
This is a destructive operation! Do not use it on existing
1002
This is a destructive operation! Do not use it on existing
3416
1005
:param revision_id: Only copy the content needed to construct
3417
1006
revision_id and its parents.
1007
:param basis: Copy the needed data preferentially from basis.
3420
1010
self.target.set_make_working_trees(self.source.make_working_trees())
3421
1011
except NotImplementedError:
1013
# grab the basis available data
1014
if basis is not None:
1015
self.target.fetch(basis, revision_id=revision_id)
1016
# but dont both fetching if we have the needed data now.
1017
if (revision_id not in (None, NULL_REVISION) and
1018
self.target.has_revision(revision_id)):
3423
1020
self.target.fetch(self.source, revision_id=revision_id)
1022
def _double_lock(self, lock_source, lock_target):
1023
"""Take out too locks, rolling back the first if the second throws."""
1028
# we want to ensure that we don't leave source locked by mistake.
1029
# and any error on target should not confuse source.
1030
self.source.unlock()
3425
1033
@needs_write_lock
3426
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
1034
def fetch(self, revision_id=None, pb=None):
3428
1035
"""Fetch the content required to construct revision_id.
3430
The content is copied from self.source to self.target.
1037
The content is copied from source to target.
3432
1039
:param revision_id: if None all content is copied, if NULL_REVISION no
3433
1040
content is copied.
1041
:param pb: optional progress bar to use for progress reports. If not
1042
provided a default one will be created.
1044
Returns the copied revision count and the failed revisions in a tuple:
3437
ui.ui_factory.warn_experimental_format_fetch(self)
3438
1047
from bzrlib.fetch import RepoFetcher
3439
# See <https://launchpad.net/bugs/456077> asking for a warning here
3440
if self.source._format.network_name() != self.target._format.network_name():
3441
ui.ui_factory.show_user_warning('cross_format_fetch',
3442
from_format=self.source._format,
3443
to_format=self.target._format)
1048
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1049
self.source, self.source._format, self.target, self.target._format)
3444
1050
f = RepoFetcher(to_repository=self.target,
3445
from_repository=self.source,
3446
last_revision=revision_id,
3447
fetch_spec=fetch_spec,
3448
find_ghosts=find_ghosts)
3450
def _walk_to_common_revisions(self, revision_ids):
3451
"""Walk out from revision_ids in source to revisions target has.
3453
:param revision_ids: The start point for the search.
3454
:return: A set of revision ids.
3456
target_graph = self.target.get_graph()
3457
revision_ids = frozenset(revision_ids)
3458
missing_revs = set()
3459
source_graph = self.source.get_graph()
3460
# ensure we don't pay silly lookup costs.
3461
searcher = source_graph._make_breadth_first_searcher(revision_ids)
3462
null_set = frozenset([_mod_revision.NULL_REVISION])
3463
searcher_exhausted = False
3467
# Iterate the searcher until we have enough next_revs
3468
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3470
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3471
next_revs.update(next_revs_part)
3472
ghosts.update(ghosts_part)
3473
except StopIteration:
3474
searcher_exhausted = True
3476
# If there are ghosts in the source graph, and the caller asked for
3477
# them, make sure that they are present in the target.
3478
# We don't care about other ghosts as we can't fetch them and
3479
# haven't been asked to.
3480
ghosts_to_check = set(revision_ids.intersection(ghosts))
3481
revs_to_get = set(next_revs).union(ghosts_to_check)
3483
have_revs = set(target_graph.get_parent_map(revs_to_get))
3484
# we always have NULL_REVISION present.
3485
have_revs = have_revs.union(null_set)
3486
# Check if the target is missing any ghosts we need.
3487
ghosts_to_check.difference_update(have_revs)
3489
# One of the caller's revision_ids is a ghost in both the
3490
# source and the target.
3491
raise errors.NoSuchRevision(
3492
self.source, ghosts_to_check.pop())
3493
missing_revs.update(next_revs - have_revs)
3494
# Because we may have walked past the original stop point, make
3495
# sure everything is stopped
3496
stop_revs = searcher.find_seen_ancestors(have_revs)
3497
searcher.stop_searching_any(stop_revs)
3498
if searcher_exhausted:
3500
return searcher.get_result()
1051
from_repository=self.source,
1052
last_revision=revision_id,
1054
return f.count_copied, f.failed_revisions
1057
def get(klass, repository_source, repository_target):
1058
"""Retrieve a InterRepository worker object for these repositories.
1060
:param repository_source: the repository to be the 'source' member of
1061
the InterRepository instance.
1062
:param repository_target: the repository to be the 'target' member of
1063
the InterRepository instance.
1064
If an optimised InterRepository worker exists it will be used otherwise
1065
a default InterRepository instance will be created.
1067
for provider in klass._optimisers:
1068
if provider.is_compatible(repository_source, repository_target):
1069
return provider(repository_source, repository_target)
1070
return InterRepository(repository_source, repository_target)
1072
def lock_read(self):
1073
"""Take out a logical read lock.
1075
This will lock the source branch and the target branch. The source gets
1076
a read lock and the target a read lock.
1078
self._double_lock(self.source.lock_read, self.target.lock_read)
1080
def lock_write(self):
1081
"""Take out a logical write lock.
1083
This will lock the source branch and the target branch. The source gets
1084
a read lock and the target a write lock.
1086
self._double_lock(self.source.lock_read, self.target.lock_write)
3502
1088
@needs_read_lock
3503
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
1089
def missing_revision_ids(self, revision_id=None):
3504
1090
"""Return the revision ids that source has that target does not.
1092
These are returned in topological order.
3506
1094
:param revision_id: only return revision ids included by this
3508
:param find_ghosts: If True find missing revisions in deep history
3509
rather than just finding the surface difference.
3510
:return: A bzrlib.graph.SearchResult.
3512
# stop searching at found target revisions.
3513
if not find_ghosts and revision_id is not None:
3514
return self._walk_to_common_revisions([revision_id])
3515
1097
# generic, possibly worst case, slow code path.
3516
1098
target_ids = set(self.target.all_revision_ids())
3517
1099
if revision_id is not None:
3518
1100
source_ids = self.source.get_ancestry(revision_id)
3519
if source_ids[0] is not None:
3520
raise AssertionError()
1101
assert source_ids.pop(0) == None
3523
1103
source_ids = self.source.all_revision_ids()
3524
1104
result_set = set(source_ids).difference(target_ids)
3525
return self.source.revision_ids_to_search_result(result_set)
3528
def _same_model(source, target):
3529
"""True if source and target have the same data representation.
3531
Note: this is always called on the base class; overriding it in a
3532
subclass will have no effect.
1105
# this may look like a no-op: its not. It preserves the ordering
1106
# other_ids had while only returning the members from other_ids
1107
# that we've decided we need.
1108
return [rev_id for rev_id in source_ids if rev_id in result_set]
1111
def register_optimiser(klass, optimiser):
1112
"""Register an InterRepository optimiser."""
1113
klass._optimisers.add(optimiser)
1116
"""Release the locks on source and target."""
3535
InterRepository._assert_same_model(source, target)
3537
except errors.IncompatibleRepositories, e:
3541
def _assert_same_model(source, target):
3542
"""Raise an exception if two repositories do not use the same model.
3544
if source.supports_rich_root() != target.supports_rich_root():
3545
raise errors.IncompatibleRepositories(source, target,
3546
"different rich-root support")
3547
if source._serializer != target._serializer:
3548
raise errors.IncompatibleRepositories(source, target,
3549
"different serializers")
3552
class InterSameDataRepository(InterRepository):
3553
"""Code for converting between repositories that represent the same data.
3555
Data format and model must match for this to work.
3559
def _get_repo_format_to_test(self):
3560
"""Repository format for testing with.
3562
InterSameData can pull from subtree to subtree and from non-subtree to
3563
non-subtree, so we test this with the richest repository format.
3565
from bzrlib.repofmt import knitrepo
3566
return knitrepo.RepositoryFormatKnit3()
3569
def is_compatible(source, target):
3570
return InterRepository._same_model(source, target)
3573
class InterWeaveRepo(InterSameDataRepository):
3574
"""Optimised code paths between Weave based repositories.
3576
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
3577
implemented lazy inter-object optimisation.
3581
def _get_repo_format_to_test(self):
3582
from bzrlib.repofmt import weaverepo
3583
return weaverepo.RepositoryFormat7()
1118
self.target.unlock()
1120
self.source.unlock()
1123
def unregister_optimiser(klass, optimiser):
1124
"""Unregister an InterRepository optimiser."""
1125
klass._optimisers.remove(optimiser)
1128
class InterWeaveRepo(InterRepository):
1129
"""Optimised code paths between Weave based repositories."""
1131
_matching_repo_format = _default_format
1132
"""Repository format for testing with."""
3586
1135
def is_compatible(source, target):
3587
1136
"""Be compatible with known Weave formats.
3589
We don't test for the stores being of specific types because that
3590
could lead to confusing results, and there is no need to be
1138
We dont test for the stores being of specific types becase that
1139
could lead to confusing results, and there is no need to be
3591
1140
overly general.
3593
from bzrlib.repofmt.weaverepo import (
3599
1143
return (isinstance(source._format, (RepositoryFormat5,
3600
1144
RepositoryFormat6,
3604
1148
RepositoryFormat7)))
3605
1149
except AttributeError:
3608
1152
@needs_write_lock
3609
def copy_content(self, revision_id=None):
1153
def copy_content(self, revision_id=None, basis=None):
3610
1154
"""See InterRepository.copy_content()."""
3611
1155
# weave specific optimised path:
3613
self.target.set_make_working_trees(self.source.make_working_trees())
3614
except (errors.RepositoryUpgradeRequired, NotImplemented):
3616
# FIXME do not peek!
3617
if self.source._transport.listable():
3618
pb = ui.ui_factory.nested_progress_bar()
1156
if basis is not None:
1157
# copy the basis in, then fetch remaining data.
1158
basis.copy_content_into(self.target, revision_id)
1159
# the basis copy_content_into could misset this.
3620
self.target.texts.insert_record_stream(
3621
self.source.texts.get_record_stream(
3622
self.source.texts.keys(), 'topological', False))
3623
pb.update('Copying inventory', 0, 1)
3624
self.target.inventories.insert_record_stream(
3625
self.source.inventories.get_record_stream(
3626
self.source.inventories.keys(), 'topological', False))
3627
self.target.signatures.insert_record_stream(
3628
self.source.signatures.get_record_stream(
3629
self.source.signatures.keys(),
3631
self.target.revisions.insert_record_stream(
3632
self.source.revisions.get_record_stream(
3633
self.source.revisions.keys(),
3634
'topological', True))
1161
self.target.set_make_working_trees(self.source.make_working_trees())
1162
except NotImplementedError:
3638
1164
self.target.fetch(self.source, revision_id=revision_id)
1167
self.target.set_make_working_trees(self.source.make_working_trees())
1168
except NotImplementedError:
1170
# FIXME do not peek!
1171
if self.source.control_files._transport.listable():
1172
pb = bzrlib.ui.ui_factory.progress_bar()
1173
copy_all(self.source.weave_store,
1174
self.target.weave_store, pb=pb)
1175
pb.update('copying inventory', 0, 1)
1176
self.target.control_weaves.copy_multi(
1177
self.source.control_weaves, ['inventory'])
1178
copy_all(self.source.revision_store,
1179
self.target.revision_store, pb=pb)
1181
self.target.fetch(self.source, revision_id=revision_id)
1184
def fetch(self, revision_id=None, pb=None):
1185
"""See InterRepository.fetch()."""
1186
from bzrlib.fetch import RepoFetcher
1187
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1188
self.source, self.source._format, self.target, self.target._format)
1189
f = RepoFetcher(to_repository=self.target,
1190
from_repository=self.source,
1191
last_revision=revision_id,
1193
return f.count_copied, f.failed_revisions
3640
1195
@needs_read_lock
3641
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
1196
def missing_revision_ids(self, revision_id=None):
3642
1197
"""See InterRepository.missing_revision_ids()."""
3643
1198
# we want all revisions to satisfy revision_id in source.
3644
# but we don't want to stat every file here and there.
3645
# we want then, all revisions other needs to satisfy revision_id
1199
# but we dont want to stat every file here and there.
1200
# we want then, all revisions other needs to satisfy revision_id
3646
1201
# checked, but not those that we have locally.
3647
# so the first thing is to get a subset of the revisions to
1202
# so the first thing is to get a subset of the revisions to
3648
1203
# satisfy revision_id in source, and then eliminate those that
3649
# we do already have.
3650
# this is slow on high latency connection to self, but as this
3651
# disk format scales terribly for push anyway due to rewriting
1204
# we do already have.
1205
# this is slow on high latency connection to self, but as as this
1206
# disk format scales terribly for push anyway due to rewriting
3652
1207
# inventory.weave, this is considered acceptable.
3653
1208
# - RBC 20060209
3654
1209
if revision_id is not None:
3655
1210
source_ids = self.source.get_ancestry(revision_id)
3656
if source_ids[0] is not None:
3657
raise AssertionError()
1211
assert source_ids.pop(0) == None
3660
1213
source_ids = self.source._all_possible_ids()
3661
1214
source_ids_set = set(source_ids)
3662
1215
# source_ids is the worst possible case we may need to pull.
3663
1216
# now we want to filter source_ids against what we actually
3664
# have in target, but don't try to check for existence where we know
1217
# have in target, but dont try to check for existence where we know
3665
1218
# we do not have a revision as that would be pointless.
3666
1219
target_ids = set(self.target._all_possible_ids())
3667
1220
possibly_present_revisions = target_ids.intersection(source_ids_set)
3668
actually_present_revisions = set(
3669
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3670
required_revisions = source_ids_set.difference(actually_present_revisions)
3671
if revision_id is not None:
3672
# we used get_ancestry to determine source_ids then we are assured all
3673
# revisions referenced are present as they are installed in topological order.
3674
# and the tip revision was validated by get_ancestry.
3675
result_set = required_revisions
3677
# if we just grabbed the possibly available ids, then
3678
# we only have an estimate of whats available and need to validate
3679
# that against the revision records.
3681
self.source._eliminate_revisions_not_present(required_revisions))
3682
return self.source.revision_ids_to_search_result(result_set)
3685
class InterKnitRepo(InterSameDataRepository):
3686
"""Optimised code paths between Knit based repositories."""
3689
def _get_repo_format_to_test(self):
3690
from bzrlib.repofmt import knitrepo
3691
return knitrepo.RepositoryFormatKnit1()
3694
def is_compatible(source, target):
3695
"""Be compatible with known Knit formats.
3697
We don't test for the stores being of specific types because that
3698
could lead to confusing results, and there is no need to be
3701
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
3703
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
3704
isinstance(target._format, RepositoryFormatKnit))
3705
except AttributeError:
3707
return are_knits and InterRepository._same_model(source, target)
3710
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3711
"""See InterRepository.missing_revision_ids()."""
3712
if revision_id is not None:
3713
source_ids = self.source.get_ancestry(revision_id)
3714
if source_ids[0] is not None:
3715
raise AssertionError()
3718
source_ids = self.source.all_revision_ids()
3719
source_ids_set = set(source_ids)
3720
# source_ids is the worst possible case we may need to pull.
3721
# now we want to filter source_ids against what we actually
3722
# have in target, but don't try to check for existence where we know
3723
# we do not have a revision as that would be pointless.
3724
target_ids = set(self.target.all_revision_ids())
3725
possibly_present_revisions = target_ids.intersection(source_ids_set)
3726
actually_present_revisions = set(
3727
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3728
required_revisions = source_ids_set.difference(actually_present_revisions)
3729
if revision_id is not None:
3730
# we used get_ancestry to determine source_ids then we are assured all
3731
# revisions referenced are present as they are installed in topological order.
3732
# and the tip revision was validated by get_ancestry.
3733
result_set = required_revisions
3735
# if we just grabbed the possibly available ids, then
3736
# we only have an estimate of whats available and need to validate
3737
# that against the revision records.
3739
self.source._eliminate_revisions_not_present(required_revisions))
3740
return self.source.revision_ids_to_search_result(result_set)
3743
class InterDifferingSerializer(InterRepository):
3746
def _get_repo_format_to_test(self):
3750
def is_compatible(source, target):
3751
"""Be compatible with Knit2 source and Knit3 target"""
3752
# This is redundant with format.check_conversion_target(), however that
3753
# raises an exception, and we just want to say "False" as in we won't
3754
# support converting between these formats.
3755
if 'IDS_never' in debug.debug_flags:
3757
if source.supports_rich_root() and not target.supports_rich_root():
3759
if (source._format.supports_tree_reference
3760
and not target._format.supports_tree_reference):
3762
if target._fallback_repositories and target._format.supports_chks:
3763
# IDS doesn't know how to copy CHKs for the parent inventories it
3764
# adds to stacked repos.
3766
if 'IDS_always' in debug.debug_flags:
3768
# Only use this code path for local source and target. IDS does far
3769
# too much IO (both bandwidth and roundtrips) over a network.
3770
if not source.bzrdir.transport.base.startswith('file:///'):
3772
if not target.bzrdir.transport.base.startswith('file:///'):
3776
def _get_trees(self, revision_ids, cache):
3778
for rev_id in revision_ids:
3780
possible_trees.append((rev_id, cache[rev_id]))
3782
# Not cached, but inventory might be present anyway.
3784
tree = self.source.revision_tree(rev_id)
3785
except errors.NoSuchRevision:
3786
# Nope, parent is ghost.
3789
cache[rev_id] = tree
3790
possible_trees.append((rev_id, tree))
3791
return possible_trees
3793
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3794
"""Get the best delta and base for this revision.
3796
:return: (basis_id, delta)
3799
# Generate deltas against each tree, to find the shortest.
3800
texts_possibly_new_in_tree = set()
3801
for basis_id, basis_tree in possible_trees:
3802
delta = tree.inventory._make_delta(basis_tree.inventory)
3803
for old_path, new_path, file_id, new_entry in delta:
3804
if new_path is None:
3805
# This file_id isn't present in the new rev, so we don't
3809
# Rich roots are handled elsewhere...
3811
kind = new_entry.kind
3812
if kind != 'directory' and kind != 'file':
3813
# No text record associated with this inventory entry.
3815
# This is a directory or file that has changed somehow.
3816
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3817
deltas.append((len(delta), basis_id, delta))
3819
return deltas[0][1:]
3821
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3822
"""Find all parent revisions that are absent, but for which the
3823
inventory is present, and copy those inventories.
3825
This is necessary to preserve correctness when the source is stacked
3826
without fallbacks configured. (Note that in cases like upgrade the
3827
source may be not have _fallback_repositories even though it is
3831
for parents in parent_map.values():
3832
parent_revs.update(parents)
3833
present_parents = self.source.get_parent_map(parent_revs)
3834
absent_parents = set(parent_revs).difference(present_parents)
3835
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3836
(rev_id,) for rev_id in absent_parents)
3837
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3838
for parent_tree in self.source.revision_trees(parent_inv_ids):
3839
current_revision_id = parent_tree.get_revision_id()
3840
parents_parents_keys = parent_invs_keys_for_stacking[
3841
(current_revision_id,)]
3842
parents_parents = [key[-1] for key in parents_parents_keys]
3843
basis_id = _mod_revision.NULL_REVISION
3844
basis_tree = self.source.revision_tree(basis_id)
3845
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3846
self.target.add_inventory_by_delta(
3847
basis_id, delta, current_revision_id, parents_parents)
3848
cache[current_revision_id] = parent_tree
3850
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3851
"""Fetch across a few revisions.
3853
:param revision_ids: The revisions to copy
3854
:param basis_id: The revision_id of a tree that must be in cache, used
3855
as a basis for delta when no other base is available
3856
:param cache: A cache of RevisionTrees that we can use.
3857
:param a_graph: A Graph object to determine the heads() of the
3858
rich-root data stream.
3859
:return: The revision_id of the last converted tree. The RevisionTree
3860
for it will be in cache
3862
# Walk though all revisions; get inventory deltas, copy referenced
3863
# texts that delta references, insert the delta, revision and
3865
root_keys_to_create = set()
3868
pending_revisions = []
3869
parent_map = self.source.get_parent_map(revision_ids)
3870
self._fetch_parent_invs_for_stacking(parent_map, cache)
3871
self.source._safe_to_return_from_cache = True
3872
for tree in self.source.revision_trees(revision_ids):
3873
# Find a inventory delta for this revision.
3874
# Find text entries that need to be copied, too.
3875
current_revision_id = tree.get_revision_id()
3876
parent_ids = parent_map.get(current_revision_id, ())
3877
parent_trees = self._get_trees(parent_ids, cache)
3878
possible_trees = list(parent_trees)
3879
if len(possible_trees) == 0:
3880
# There either aren't any parents, or the parents are ghosts,
3881
# so just use the last converted tree.
3882
possible_trees.append((basis_id, cache[basis_id]))
3883
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3885
revision = self.source.get_revision(current_revision_id)
3886
pending_deltas.append((basis_id, delta,
3887
current_revision_id, revision.parent_ids))
3888
if self._converting_to_rich_root:
3889
self._revision_id_to_root_id[current_revision_id] = \
3891
# Determine which texts are in present in this revision but not in
3892
# any of the available parents.
3893
texts_possibly_new_in_tree = set()
3894
for old_path, new_path, file_id, entry in delta:
3895
if new_path is None:
3896
# This file_id isn't present in the new rev
3900
if not self.target.supports_rich_root():
3901
# The target doesn't support rich root, so we don't
3904
if self._converting_to_rich_root:
3905
# This can't be copied normally, we have to insert
3907
root_keys_to_create.add((file_id, entry.revision))
3910
texts_possibly_new_in_tree.add((file_id, entry.revision))
3911
for basis_id, basis_tree in possible_trees:
3912
basis_inv = basis_tree.inventory
3913
for file_key in list(texts_possibly_new_in_tree):
3914
file_id, file_revision = file_key
3916
entry = basis_inv[file_id]
3917
except errors.NoSuchId:
3919
if entry.revision == file_revision:
3920
texts_possibly_new_in_tree.remove(file_key)
3921
text_keys.update(texts_possibly_new_in_tree)
3922
pending_revisions.append(revision)
3923
cache[current_revision_id] = tree
3924
basis_id = current_revision_id
3925
self.source._safe_to_return_from_cache = False
3927
from_texts = self.source.texts
3928
to_texts = self.target.texts
3929
if root_keys_to_create:
3930
root_stream = _mod_fetch._new_root_data_stream(
3931
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3932
self.source, graph=a_graph)
3933
to_texts.insert_record_stream(root_stream)
3934
to_texts.insert_record_stream(from_texts.get_record_stream(
3935
text_keys, self.target._format._fetch_order,
3936
not self.target._format._fetch_uses_deltas))
3937
# insert inventory deltas
3938
for delta in pending_deltas:
3939
self.target.add_inventory_by_delta(*delta)
3940
if self.target._fallback_repositories:
3941
# Make sure this stacked repository has all the parent inventories
3942
# for the new revisions that we are about to insert. We do this
3943
# before adding the revisions so that no revision is added until
3944
# all the inventories it may depend on are added.
3945
# Note that this is overzealous, as we may have fetched these in an
3948
revision_ids = set()
3949
for revision in pending_revisions:
3950
revision_ids.add(revision.revision_id)
3951
parent_ids.update(revision.parent_ids)
3952
parent_ids.difference_update(revision_ids)
3953
parent_ids.discard(_mod_revision.NULL_REVISION)
3954
parent_map = self.source.get_parent_map(parent_ids)
3955
# we iterate over parent_map and not parent_ids because we don't
3956
# want to try copying any revision which is a ghost
3957
for parent_tree in self.source.revision_trees(parent_map):
3958
current_revision_id = parent_tree.get_revision_id()
3959
parents_parents = parent_map[current_revision_id]
3960
possible_trees = self._get_trees(parents_parents, cache)
3961
if len(possible_trees) == 0:
3962
# There either aren't any parents, or the parents are
3963
# ghosts, so just use the last converted tree.
3964
possible_trees.append((basis_id, cache[basis_id]))
3965
basis_id, delta = self._get_delta_for_revision(parent_tree,
3966
parents_parents, possible_trees)
3967
self.target.add_inventory_by_delta(
3968
basis_id, delta, current_revision_id, parents_parents)
3969
# insert signatures and revisions
3970
for revision in pending_revisions:
3972
signature = self.source.get_signature_text(
3973
revision.revision_id)
3974
self.target.add_signature_text(revision.revision_id,
3976
except errors.NoSuchRevision:
3978
self.target.add_revision(revision.revision_id, revision)
3981
def _fetch_all_revisions(self, revision_ids, pb):
3982
"""Fetch everything for the list of revisions.
3984
:param revision_ids: The list of revisions to fetch. Must be in
3986
:param pb: A ProgressTask
3989
basis_id, basis_tree = self._get_basis(revision_ids[0])
3991
cache = lru_cache.LRUCache(100)
3992
cache[basis_id] = basis_tree
3993
del basis_tree # We don't want to hang on to it here
3995
if self._converting_to_rich_root and len(revision_ids) > 100:
3996
a_graph = _mod_fetch._get_rich_root_heads_graph(self.source,
4001
for offset in range(0, len(revision_ids), batch_size):
4002
self.target.start_write_group()
4004
pb.update('Transferring revisions', offset,
4006
batch = revision_ids[offset:offset+batch_size]
4007
basis_id = self._fetch_batch(batch, basis_id, cache,
4010
self.source._safe_to_return_from_cache = False
4011
self.target.abort_write_group()
4014
hint = self.target.commit_write_group()
4017
if hints and self.target._format.pack_compresses:
4018
self.target.pack(hint=hints)
4019
pb.update('Transferring revisions', len(revision_ids),
4023
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
4025
"""See InterRepository.fetch()."""
4026
if fetch_spec is not None:
4027
raise AssertionError("Not implemented yet...")
4028
ui.ui_factory.warn_experimental_format_fetch(self)
4029
if (not self.source.supports_rich_root()
4030
and self.target.supports_rich_root()):
4031
self._converting_to_rich_root = True
4032
self._revision_id_to_root_id = {}
4034
self._converting_to_rich_root = False
4035
# See <https://launchpad.net/bugs/456077> asking for a warning here
4036
if self.source._format.network_name() != self.target._format.network_name():
4037
ui.ui_factory.show_user_warning('cross_format_fetch',
4038
from_format=self.source._format,
4039
to_format=self.target._format)
4040
revision_ids = self.target.search_missing_revision_ids(self.source,
4041
revision_id, find_ghosts=find_ghosts).get_keys()
4042
if not revision_ids:
4044
revision_ids = tsort.topo_sort(
4045
self.source.get_graph().get_parent_map(revision_ids))
4046
if not revision_ids:
4048
# Walk though all revisions; get inventory deltas, copy referenced
4049
# texts that delta references, insert the delta, revision and
4052
my_pb = ui.ui_factory.nested_progress_bar()
4055
symbol_versioning.warn(
4056
symbol_versioning.deprecated_in((1, 14, 0))
4057
% "pb parameter to fetch()")
4060
self._fetch_all_revisions(revision_ids, pb)
4062
if my_pb is not None:
4064
return len(revision_ids), 0
4066
def _get_basis(self, first_revision_id):
4067
"""Get a revision and tree which exists in the target.
4069
This assumes that first_revision_id is selected for transmission
4070
because all other ancestors are already present. If we can't find an
4071
ancestor we fall back to NULL_REVISION since we know that is safe.
4073
:return: (basis_id, basis_tree)
4075
first_rev = self.source.get_revision(first_revision_id)
4077
basis_id = first_rev.parent_ids[0]
4078
# only valid as a basis if the target has it
4079
self.target.get_revision(basis_id)
4080
# Try to get a basis tree - if its a ghost it will hit the
4081
# NoSuchRevision case.
4082
basis_tree = self.source.revision_tree(basis_id)
4083
except (IndexError, errors.NoSuchRevision):
4084
basis_id = _mod_revision.NULL_REVISION
4085
basis_tree = self.source.revision_tree(basis_id)
4086
return basis_id, basis_tree
4089
InterRepository.register_optimiser(InterDifferingSerializer)
4090
InterRepository.register_optimiser(InterSameDataRepository)
1221
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
1222
required_revisions = source_ids_set.difference(actually_present_revisions)
1223
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
1224
if revision_id is not None:
1225
# we used get_ancestry to determine source_ids then we are assured all
1226
# revisions referenced are present as they are installed in topological order.
1227
# and the tip revision was validated by get_ancestry.
1228
return required_topo_revisions
1230
# if we just grabbed the possibly available ids, then
1231
# we only have an estimate of whats available and need to validate
1232
# that against the revision records.
1233
return self.source._eliminate_revisions_not_present(required_topo_revisions)
4091
1236
InterRepository.register_optimiser(InterWeaveRepo)
4092
InterRepository.register_optimiser(InterKnitRepo)
1239
class RepositoryTestProviderAdapter(object):
1240
"""A tool to generate a suite testing multiple repository formats at once.
1242
This is done by copying the test once for each transport and injecting
1243
the transport_server, transport_readonly_server, and bzrdir_format and
1244
repository_format classes into each copy. Each copy is also given a new id()
1245
to make it easy to identify.
1248
def __init__(self, transport_server, transport_readonly_server, formats):
1249
self._transport_server = transport_server
1250
self._transport_readonly_server = transport_readonly_server
1251
self._formats = formats
1253
def adapt(self, test):
1254
result = TestSuite()
1255
for repository_format, bzrdir_format in self._formats:
1256
new_test = deepcopy(test)
1257
new_test.transport_server = self._transport_server
1258
new_test.transport_readonly_server = self._transport_readonly_server
1259
new_test.bzrdir_format = bzrdir_format
1260
new_test.repository_format = repository_format
1261
def make_new_test_id():
1262
new_id = "%s(%s)" % (new_test.id(), repository_format.__class__.__name__)
1263
return lambda: new_id
1264
new_test.id = make_new_test_id()
1265
result.addTest(new_test)
1269
class InterRepositoryTestProviderAdapter(object):
1270
"""A tool to generate a suite testing multiple inter repository formats.
1272
This is done by copying the test once for each interrepo provider and injecting
1273
the transport_server, transport_readonly_server, repository_format and
1274
repository_to_format classes into each copy.
1275
Each copy is also given a new id() to make it easy to identify.
1278
def __init__(self, transport_server, transport_readonly_server, formats):
1279
self._transport_server = transport_server
1280
self._transport_readonly_server = transport_readonly_server
1281
self._formats = formats
1283
def adapt(self, test):
1284
result = TestSuite()
1285
for interrepo_class, repository_format, repository_format_to in self._formats:
1286
new_test = deepcopy(test)
1287
new_test.transport_server = self._transport_server
1288
new_test.transport_readonly_server = self._transport_readonly_server
1289
new_test.interrepo_class = interrepo_class
1290
new_test.repository_format = repository_format
1291
new_test.repository_format_to = repository_format_to
1292
def make_new_test_id():
1293
new_id = "%s(%s)" % (new_test.id(), interrepo_class.__name__)
1294
return lambda: new_id
1295
new_test.id = make_new_test_id()
1296
result.addTest(new_test)
1300
def default_test_list():
1301
"""Generate the default list of interrepo permutations to test."""
1303
# test the default InterRepository between format 6 and the current
1305
# XXX: robertc 20060220 reinstate this when there are two supported
1306
# formats which do not have an optimal code path between them.
1307
result.append((InterRepository,
1308
RepositoryFormat6(),
1309
RepositoryFormatKnit1()))
1310
for optimiser in InterRepository._optimisers:
1311
result.append((optimiser,
1312
optimiser._matching_repo_format,
1313
optimiser._matching_repo_format
1315
# if there are specific combinations we want to use, we can add them
4095
1320
class CopyConverter(object):
4096
1321
"""A repository conversion tool which just performs a copy of the content.
4098
1323
This is slow but quite reliable.
4104
1329
:param target_format: The format the resulting repository should be.
4106
1331
self.target_format = target_format
4108
1333
def convert(self, repo, pb):
4109
1334
"""Perform the conversion of to_convert, giving feedback via pb.
4111
1336
:param to_convert: The disk object to convert.
4112
1337
:param pb: a progress bar to use for progress information.
4114
pb = ui.ui_factory.nested_progress_bar()
4117
1342
# this is only useful with metadir layouts - separated repo content.
4118
1343
# trigger an assertion if not such
4119
1344
repo._format.get_format_string()
4120
1345
self.repo_dir = repo.bzrdir
4121
pb.update('Moving repository to repository.backup')
1346
self.step('Moving repository to repository.backup')
4122
1347
self.repo_dir.transport.move('repository', 'repository.backup')
4123
1348
backup_transport = self.repo_dir.transport.clone('repository.backup')
4124
repo._format.check_conversion_target(self.target_format)
4125
1349
self.source_repo = repo._format.open(self.repo_dir,
4127
1351
_override_transport=backup_transport)
4128
pb.update('Creating new repository')
1352
self.step('Creating new repository')
4129
1353
converted = self.target_format.initialize(self.repo_dir,
4130
1354
self.source_repo.is_shared())
4131
1355
converted.lock_write()
4133
pb.update('Copying content')
1357
self.step('Copying content into repository.')
4134
1358
self.source_repo.copy_content_into(converted)
4136
1360
converted.unlock()
4137
pb.update('Deleting old repository content')
1361
self.step('Deleting old repository content.')
4138
1362
self.repo_dir.transport.delete_tree('repository.backup')
4139
ui.ui_factory.note('repository converted')
4152
def _unescaper(match, _map=_unescape_map):
4153
code = match.group(1)
4157
if not code.startswith('#'):
4159
return unichr(int(code[1:])).encode('utf8')
4165
def _unescape_xml(data):
4166
"""Unescape predefined XML entities in a string of data."""
4168
if _unescape_re is None:
4169
_unescape_re = re.compile('\&([^;]*);')
4170
return _unescape_re.sub(_unescaper, data)
4173
class _VersionedFileChecker(object):
4175
def __init__(self, repository, text_key_references=None, ancestors=None):
4176
self.repository = repository
4177
self.text_index = self.repository._generate_text_key_index(
4178
text_key_references=text_key_references, ancestors=ancestors)
4180
def calculate_file_version_parents(self, text_key):
4181
"""Calculate the correct parents for a file version according to
4184
parent_keys = self.text_index[text_key]
4185
if parent_keys == [_mod_revision.NULL_REVISION]:
4187
return tuple(parent_keys)
4189
def check_file_version_parents(self, texts, progress_bar=None):
4190
"""Check the parents stored in a versioned file are correct.
4192
It also detects file versions that are not referenced by their
4193
corresponding revision's inventory.
4195
:returns: A tuple of (wrong_parents, dangling_file_versions).
4196
wrong_parents is a dict mapping {revision_id: (stored_parents,
4197
correct_parents)} for each revision_id where the stored parents
4198
are not correct. dangling_file_versions is a set of (file_id,
4199
revision_id) tuples for versions that are present in this versioned
4200
file, but not used by the corresponding inventory.
4202
local_progress = None
4203
if progress_bar is None:
4204
local_progress = ui.ui_factory.nested_progress_bar()
4205
progress_bar = local_progress
4207
return self._check_file_version_parents(texts, progress_bar)
4210
local_progress.finished()
4212
def _check_file_version_parents(self, texts, progress_bar):
4213
"""See check_file_version_parents."""
4215
self.file_ids = set([file_id for file_id, _ in
4216
self.text_index.iterkeys()])
4217
# text keys is now grouped by file_id
4218
n_versions = len(self.text_index)
4219
progress_bar.update('loading text store', 0, n_versions)
4220
parent_map = self.repository.texts.get_parent_map(self.text_index)
4221
# On unlistable transports this could well be empty/error...
4222
text_keys = self.repository.texts.keys()
4223
unused_keys = frozenset(text_keys) - set(self.text_index)
4224
for num, key in enumerate(self.text_index.iterkeys()):
4225
progress_bar.update('checking text graph', num, n_versions)
4226
correct_parents = self.calculate_file_version_parents(key)
4228
knit_parents = parent_map[key]
4229
except errors.RevisionNotPresent:
4232
if correct_parents != knit_parents:
4233
wrong_parents[key] = (knit_parents, correct_parents)
4234
return wrong_parents, unused_keys
4237
def _old_get_graph(repository, revision_id):
4238
"""DO NOT USE. That is all. I'm serious."""
4239
graph = repository.get_graph()
4240
revision_graph = dict(((key, value) for key, value in
4241
graph.iter_ancestry([revision_id]) if value is not None))
4242
return _strip_NULL_ghosts(revision_graph)
4245
def _strip_NULL_ghosts(revision_graph):
4246
"""Also don't use this. more compatibility code for unmigrated clients."""
4247
# Filter ghosts, and null:
4248
if _mod_revision.NULL_REVISION in revision_graph:
4249
del revision_graph[_mod_revision.NULL_REVISION]
4250
for key, parents in revision_graph.items():
4251
revision_graph[key] = tuple(parent for parent in parents if parent
4253
return revision_graph
4256
class StreamSink(object):
4257
"""An object that can insert a stream into a repository.
4259
This interface handles the complexity of reserialising inventories and
4260
revisions from different formats, and allows unidirectional insertion into
4261
stacked repositories without looking for the missing basis parents
4265
def __init__(self, target_repo):
4266
self.target_repo = target_repo
4268
def insert_stream(self, stream, src_format, resume_tokens):
4269
"""Insert a stream's content into the target repository.
4271
:param src_format: a bzr repository format.
4273
:return: a list of resume tokens and an iterable of keys additional
4274
items required before the insertion can be completed.
4276
self.target_repo.lock_write()
4279
self.target_repo.resume_write_group(resume_tokens)
4282
self.target_repo.start_write_group()
4285
# locked_insert_stream performs a commit|suspend.
4286
return self._locked_insert_stream(stream, src_format, is_resume)
4288
self.target_repo.abort_write_group(suppress_errors=True)
4291
self.target_repo.unlock()
4293
def _locked_insert_stream(self, stream, src_format, is_resume):
4294
to_serializer = self.target_repo._format._serializer
4295
src_serializer = src_format._serializer
4297
if to_serializer == src_serializer:
4298
# If serializers match and the target is a pack repository, set the
4299
# write cache size on the new pack. This avoids poor performance
4300
# on transports where append is unbuffered (such as
4301
# RemoteTransport). This is safe to do because nothing should read
4302
# back from the target repository while a stream with matching
4303
# serialization is being inserted.
4304
# The exception is that a delta record from the source that should
4305
# be a fulltext may need to be expanded by the target (see
4306
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4307
# explicitly flush any buffered writes first in that rare case.
4309
new_pack = self.target_repo._pack_collection._new_pack
4310
except AttributeError:
4311
# Not a pack repository
4314
new_pack.set_write_cache_size(1024*1024)
4315
for substream_type, substream in stream:
4316
if 'stream' in debug.debug_flags:
4317
mutter('inserting substream: %s', substream_type)
4318
if substream_type == 'texts':
4319
self.target_repo.texts.insert_record_stream(substream)
4320
elif substream_type == 'inventories':
4321
if src_serializer == to_serializer:
4322
self.target_repo.inventories.insert_record_stream(
4325
self._extract_and_insert_inventories(
4326
substream, src_serializer)
4327
elif substream_type == 'inventory-deltas':
4328
self._extract_and_insert_inventory_deltas(
4329
substream, src_serializer)
4330
elif substream_type == 'chk_bytes':
4331
# XXX: This doesn't support conversions, as it assumes the
4332
# conversion was done in the fetch code.
4333
self.target_repo.chk_bytes.insert_record_stream(substream)
4334
elif substream_type == 'revisions':
4335
# This may fallback to extract-and-insert more often than
4336
# required if the serializers are different only in terms of
4338
if src_serializer == to_serializer:
4339
self.target_repo.revisions.insert_record_stream(
4342
self._extract_and_insert_revisions(substream,
4344
elif substream_type == 'signatures':
4345
self.target_repo.signatures.insert_record_stream(substream)
4347
raise AssertionError('kaboom! %s' % (substream_type,))
4348
# Done inserting data, and the missing_keys calculations will try to
4349
# read back from the inserted data, so flush the writes to the new pack
4350
# (if this is pack format).
4351
if new_pack is not None:
4352
new_pack._write_data('', flush=True)
4353
# Find all the new revisions (including ones from resume_tokens)
4354
missing_keys = self.target_repo.get_missing_parent_inventories(
4355
check_for_missing_texts=is_resume)
4357
for prefix, versioned_file in (
4358
('texts', self.target_repo.texts),
4359
('inventories', self.target_repo.inventories),
4360
('revisions', self.target_repo.revisions),
4361
('signatures', self.target_repo.signatures),
4362
('chk_bytes', self.target_repo.chk_bytes),
4364
if versioned_file is None:
4366
# TODO: key is often going to be a StaticTuple object
4367
# I don't believe we can define a method by which
4368
# (prefix,) + StaticTuple will work, though we could
4369
# define a StaticTuple.sq_concat that would allow you to
4370
# pass in either a tuple or a StaticTuple as the second
4371
# object, so instead we could have:
4372
# StaticTuple(prefix) + key here...
4373
missing_keys.update((prefix,) + key for key in
4374
versioned_file.get_missing_compression_parent_keys())
4375
except NotImplementedError:
4376
# cannot even attempt suspending, and missing would have failed
4377
# during stream insertion.
4378
missing_keys = set()
4381
# suspend the write group and tell the caller what we is
4382
# missing. We know we can suspend or else we would not have
4383
# entered this code path. (All repositories that can handle
4384
# missing keys can handle suspending a write group).
4385
write_group_tokens = self.target_repo.suspend_write_group()
4386
return write_group_tokens, missing_keys
4387
hint = self.target_repo.commit_write_group()
4388
if (to_serializer != src_serializer and
4389
self.target_repo._format.pack_compresses):
4390
self.target_repo.pack(hint=hint)
4393
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4394
target_rich_root = self.target_repo._format.rich_root_data
4395
target_tree_refs = self.target_repo._format.supports_tree_reference
4396
for record in substream:
4397
# Insert the delta directly
4398
inventory_delta_bytes = record.get_bytes_as('fulltext')
4399
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4401
parse_result = deserialiser.parse_text_bytes(
4402
inventory_delta_bytes)
4403
except inventory_delta.IncompatibleInventoryDelta, err:
4404
trace.mutter("Incompatible delta: %s", err.msg)
4405
raise errors.IncompatibleRevision(self.target_repo._format)
4406
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4407
revision_id = new_id
4408
parents = [key[0] for key in record.parents]
4409
self.target_repo.add_inventory_by_delta(
4410
basis_id, inv_delta, revision_id, parents)
4412
def _extract_and_insert_inventories(self, substream, serializer,
4414
"""Generate a new inventory versionedfile in target, converting data.
4416
The inventory is retrieved from the source, (deserializing it), and
4417
stored in the target (reserializing it in a different format).
4419
target_rich_root = self.target_repo._format.rich_root_data
4420
target_tree_refs = self.target_repo._format.supports_tree_reference
4421
for record in substream:
4422
# It's not a delta, so it must be a fulltext in the source
4423
# serializer's format.
4424
bytes = record.get_bytes_as('fulltext')
4425
revision_id = record.key[0]
4426
inv = serializer.read_inventory_from_string(bytes, revision_id)
4427
parents = [key[0] for key in record.parents]
4428
self.target_repo.add_inventory(revision_id, inv, parents)
4429
# No need to keep holding this full inv in memory when the rest of
4430
# the substream is likely to be all deltas.
4433
def _extract_and_insert_revisions(self, substream, serializer):
4434
for record in substream:
4435
bytes = record.get_bytes_as('fulltext')
4436
revision_id = record.key[0]
4437
rev = serializer.read_revision_from_string(bytes)
4438
if rev.revision_id != revision_id:
4439
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4440
self.target_repo.add_revision(revision_id, rev)
4443
if self.target_repo._format._fetch_reconcile:
4444
self.target_repo.reconcile()
4447
class StreamSource(object):
4448
"""A source of a stream for fetching between repositories."""
4450
def __init__(self, from_repository, to_format):
4451
"""Create a StreamSource streaming from from_repository."""
4452
self.from_repository = from_repository
4453
self.to_format = to_format
4455
def delta_on_metadata(self):
4456
"""Return True if delta's are permitted on metadata streams.
4458
That is on revisions and signatures.
4460
src_serializer = self.from_repository._format._serializer
4461
target_serializer = self.to_format._serializer
4462
return (self.to_format._fetch_uses_deltas and
4463
src_serializer == target_serializer)
4465
def _fetch_revision_texts(self, revs):
4466
# fetch signatures first and then the revision texts
4467
# may need to be a InterRevisionStore call here.
4468
from_sf = self.from_repository.signatures
4469
# A missing signature is just skipped.
4470
keys = [(rev_id,) for rev_id in revs]
4471
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4473
self.to_format._fetch_order,
4474
not self.to_format._fetch_uses_deltas))
4475
# If a revision has a delta, this is actually expanded inside the
4476
# insert_record_stream code now, which is an alternate fix for
4478
from_rf = self.from_repository.revisions
4479
revisions = from_rf.get_record_stream(
4481
self.to_format._fetch_order,
4482
not self.delta_on_metadata())
4483
return [('signatures', signatures), ('revisions', revisions)]
4485
def _generate_root_texts(self, revs):
4486
"""This will be called by get_stream between fetching weave texts and
4487
fetching the inventory weave.
4489
if self._rich_root_upgrade():
4490
return _mod_fetch.Inter1and2Helper(
4491
self.from_repository).generate_root_texts(revs)
4495
def get_stream(self, search):
4497
revs = search.get_keys()
4498
graph = self.from_repository.get_graph()
4499
revs = tsort.topo_sort(graph.get_parent_map(revs))
4500
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4502
for knit_kind, file_id, revisions in data_to_fetch:
4503
if knit_kind != phase:
4505
# Make a new progress bar for this phase
4506
if knit_kind == "file":
4507
# Accumulate file texts
4508
text_keys.extend([(file_id, revision) for revision in
4510
elif knit_kind == "inventory":
4511
# Now copy the file texts.
4512
from_texts = self.from_repository.texts
4513
yield ('texts', from_texts.get_record_stream(
4514
text_keys, self.to_format._fetch_order,
4515
not self.to_format._fetch_uses_deltas))
4516
# Cause an error if a text occurs after we have done the
4519
# Before we process the inventory we generate the root
4520
# texts (if necessary) so that the inventories references
4522
for _ in self._generate_root_texts(revs):
4524
# we fetch only the referenced inventories because we do not
4525
# know for unselected inventories whether all their required
4526
# texts are present in the other repository - it could be
4528
for info in self._get_inventory_stream(revs):
4530
elif knit_kind == "signatures":
4531
# Nothing to do here; this will be taken care of when
4532
# _fetch_revision_texts happens.
4534
elif knit_kind == "revisions":
4535
for record in self._fetch_revision_texts(revs):
4538
raise AssertionError("Unknown knit kind %r" % knit_kind)
4540
def get_stream_for_missing_keys(self, missing_keys):
4541
# missing keys can only occur when we are byte copying and not
4542
# translating (because translation means we don't send
4543
# unreconstructable deltas ever).
4545
keys['texts'] = set()
4546
keys['revisions'] = set()
4547
keys['inventories'] = set()
4548
keys['chk_bytes'] = set()
4549
keys['signatures'] = set()
4550
for key in missing_keys:
4551
keys[key[0]].add(key[1:])
4552
if len(keys['revisions']):
4553
# If we allowed copying revisions at this point, we could end up
4554
# copying a revision without copying its required texts: a
4555
# violation of the requirements for repository integrity.
4556
raise AssertionError(
4557
'cannot copy revisions to fill in missing deltas %s' % (
4558
keys['revisions'],))
4559
for substream_kind, keys in keys.iteritems():
4560
vf = getattr(self.from_repository, substream_kind)
4561
if vf is None and keys:
4562
raise AssertionError(
4563
"cannot fill in keys for a versioned file we don't"
4564
" have: %s needs %s" % (substream_kind, keys))
4566
# No need to stream something we don't have
4568
if substream_kind == 'inventories':
4569
# Some missing keys are genuinely ghosts, filter those out.
4570
present = self.from_repository.inventories.get_parent_map(keys)
4571
revs = [key[0] for key in present]
4572
# Get the inventory stream more-or-less as we do for the
4573
# original stream; there's no reason to assume that records
4574
# direct from the source will be suitable for the sink. (Think
4575
# e.g. 2a -> 1.9-rich-root).
4576
for info in self._get_inventory_stream(revs, missing=True):
4580
# Ask for full texts always so that we don't need more round trips
4581
# after this stream.
4582
# Some of the missing keys are genuinely ghosts, so filter absent
4583
# records. The Sink is responsible for doing another check to
4584
# ensure that ghosts don't introduce missing data for future
4586
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4587
self.to_format._fetch_order, True))
4588
yield substream_kind, stream
4590
def inventory_fetch_order(self):
4591
if self._rich_root_upgrade():
4592
return 'topological'
4594
return self.to_format._fetch_order
4596
def _rich_root_upgrade(self):
4597
return (not self.from_repository._format.rich_root_data and
4598
self.to_format.rich_root_data)
4600
def _get_inventory_stream(self, revision_ids, missing=False):
4601
from_format = self.from_repository._format
4602
if (from_format.supports_chks and self.to_format.supports_chks and
4603
from_format.network_name() == self.to_format.network_name()):
4604
raise AssertionError(
4605
"this case should be handled by GroupCHKStreamSource")
4606
elif 'forceinvdeltas' in debug.debug_flags:
4607
return self._get_convertable_inventory_stream(revision_ids,
4608
delta_versus_null=missing)
4609
elif from_format.network_name() == self.to_format.network_name():
4611
return self._get_simple_inventory_stream(revision_ids,
4613
elif (not from_format.supports_chks and not self.to_format.supports_chks
4614
and from_format._serializer == self.to_format._serializer):
4615
# Essentially the same format.
4616
return self._get_simple_inventory_stream(revision_ids,
4619
# Any time we switch serializations, we want to use an
4620
# inventory-delta based approach.
4621
return self._get_convertable_inventory_stream(revision_ids,
4622
delta_versus_null=missing)
4624
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4625
# NB: This currently reopens the inventory weave in source;
4626
# using a single stream interface instead would avoid this.
4627
from_weave = self.from_repository.inventories
4629
delta_closure = True
4631
delta_closure = not self.delta_on_metadata()
4632
yield ('inventories', from_weave.get_record_stream(
4633
[(rev_id,) for rev_id in revision_ids],
4634
self.inventory_fetch_order(), delta_closure))
4636
def _get_convertable_inventory_stream(self, revision_ids,
4637
delta_versus_null=False):
4638
# The two formats are sufficiently different that there is no fast
4639
# path, so we need to send just inventorydeltas, which any
4640
# sufficiently modern client can insert into any repository.
4641
# The StreamSink code expects to be able to
4642
# convert on the target, so we need to put bytes-on-the-wire that can
4643
# be converted. That means inventory deltas (if the remote is <1.19,
4644
# RemoteStreamSink will fallback to VFS to insert the deltas).
4645
yield ('inventory-deltas',
4646
self._stream_invs_as_deltas(revision_ids,
4647
delta_versus_null=delta_versus_null))
4649
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4650
"""Return a stream of inventory-deltas for the given rev ids.
4652
:param revision_ids: The list of inventories to transmit
4653
:param delta_versus_null: Don't try to find a minimal delta for this
4654
entry, instead compute the delta versus the NULL_REVISION. This
4655
effectively streams a complete inventory. Used for stuff like
4656
filling in missing parents, etc.
4658
from_repo = self.from_repository
4659
revision_keys = [(rev_id,) for rev_id in revision_ids]
4660
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4661
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4663
inventories = self.from_repository.iter_inventories(
4664
revision_ids, 'topological')
4665
format = from_repo._format
4666
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4667
inventory_cache = lru_cache.LRUCache(50)
4668
null_inventory = from_repo.revision_tree(
4669
_mod_revision.NULL_REVISION).inventory
4670
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4671
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4672
# repo back into a non-rich-root repo ought to be allowed)
4673
serializer = inventory_delta.InventoryDeltaSerializer(
4674
versioned_root=format.rich_root_data,
4675
tree_references=format.supports_tree_reference)
4676
for inv in inventories:
4677
key = (inv.revision_id,)
4678
parent_keys = parent_map.get(key, ())
4680
if not delta_versus_null and parent_keys:
4681
# The caller did not ask for complete inventories and we have
4682
# some parents that we can delta against. Make a delta against
4683
# each parent so that we can find the smallest.
4684
parent_ids = [parent_key[0] for parent_key in parent_keys]
4685
for parent_id in parent_ids:
4686
if parent_id not in invs_sent_so_far:
4687
# We don't know that the remote side has this basis, so
4690
if parent_id == _mod_revision.NULL_REVISION:
4691
parent_inv = null_inventory
4693
parent_inv = inventory_cache.get(parent_id, None)
4694
if parent_inv is None:
4695
parent_inv = from_repo.get_inventory(parent_id)
4696
candidate_delta = inv._make_delta(parent_inv)
4697
if (delta is None or
4698
len(delta) > len(candidate_delta)):
4699
delta = candidate_delta
4700
basis_id = parent_id
4702
# Either none of the parents ended up being suitable, or we
4703
# were asked to delta against NULL
4704
basis_id = _mod_revision.NULL_REVISION
4705
delta = inv._make_delta(null_inventory)
4706
invs_sent_so_far.add(inv.revision_id)
4707
inventory_cache[inv.revision_id] = inv
4708
delta_serialized = ''.join(
4709
serializer.delta_to_lines(basis_id, key[-1], delta))
4710
yield versionedfile.FulltextContentFactory(
4711
key, parent_keys, None, delta_serialized)
4714
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4715
stop_revision=None):
4716
"""Extend the partial history to include a given index
4718
If a stop_index is supplied, stop when that index has been reached.
4719
If a stop_revision is supplied, stop when that revision is
4720
encountered. Otherwise, stop when the beginning of history is
4723
:param stop_index: The index which should be present. When it is
4724
present, history extension will stop.
4725
:param stop_revision: The revision id which should be present. When
4726
it is encountered, history extension will stop.
4728
start_revision = partial_history_cache[-1]
4729
iterator = repo.iter_reverse_revision_history(start_revision)
4731
#skip the last revision in the list
4734
if (stop_index is not None and
4735
len(partial_history_cache) > stop_index):
4737
if partial_history_cache[-1] == stop_revision:
4739
revision_id = iterator.next()
4740
partial_history_cache.append(revision_id)
4741
except StopIteration:
1363
self.pb.note('repository converted')
1365
def step(self, message):
1366
"""Update the pb by a step."""
1368
self.pb.update(message, self.count, self.total)
added
removed
bzrlib/repository.py
