« back to all changes in this revision
Viewing changes to bzrlib/repository.py
- browse files at revision 5050.73.11
- compare with another revision
- download diff
- download tarball
- view history from revision 5050.73.11
- Committer: Vincent Ladeuil
- Date: 2011-07-06 09:22:00 UTC
- mfrom: (6008 +trunk)
- mto: (6012.1.1 trunk)
- mto: This revision was merged to the branch mainline in revision 6013.
- Revision ID: v.ladeuil+lp@free.fr-20110706092200-7iai2mwzc0sqdsvf
Merging in trunk
- files added:
- bzrlib/cethread.py
- bzrlib/plugins/changelog_merge
- bzrlib/plugins/changelog_merge/tests
- bzrlib/plugins/weave_fmt
- bzrlib/plugins/weave_fmt/tests
- bzrlib/tests/per_repository_vf
- po
- files removed:
- bzrlib/deprecated_graph.py
- files renamed:
- bzrlib/sign_my_commits.py => bzrlib/commit_signature_commands.py
- bzrlib/repofmt/weaverepo.py => bzrlib/plugins/weave_fmt/repository.py
- bzrlib/xml4.py => bzrlib/plugins/weave_fmt/xml4.py
- bzrlib/tests/per_repository/test_is_write_locked.py => bzrlib/tests/per_repository/test_locking.py
- bzrlib/tests/per_repository/helpers.py => bzrlib/tests/per_repository_vf/helpers.py
- bzrlib/tests/per_repository/test__generate_text_key_index.py => bzrlib/tests/per_repository_vf/test__generate_text_key_index.py
- bzrlib/tests/per_repository/test_add_inventory_by_delta.py => bzrlib/tests/per_repository_vf/test_add_inventory_by_delta.py
- bzrlib/tests/per_repository/test_check_reconcile.py => bzrlib/tests/per_repository_vf/test_check_reconcile.py
- bzrlib/tests/per_repository/test_find_text_key_references.py => bzrlib/tests/per_repository_vf/test_find_text_key_references.py
- files modified:
- Makefile
added
removed
bzrlib/repository.py
16
16
17
17
from bzrlib.lazy_import import lazy_import
18
18
lazy_import(globals(), """
19
import cStringIO
20
import re
19
import itertools
21
20
import time
22
21
23
22
from bzrlib import (
24
23
bzrdir,
25
check,
26
chk_map,
27
24
config,
28
25
controldir,
29
26
debug,
30
fetch as _mod_fetch,
31
fifo_cache,
32
27
generate_ids,
33
gpg,
34
28
graph,
35
inventory,
36
inventory_delta,
37
lazy_regex,
38
29
lockable_files,
39
30
lockdir,
40
lru_cache,
41
31
osutils,
42
pyutils,
43
32
revision as _mod_revision,
44
static_tuple,
45
symbol_versioning,
46
trace,
33
testament as _mod_testament,
47
34
tsort,
48
versionedfile,
35
gpg,
49
36
)
50
37
from bzrlib.bundle import serializer
51
from bzrlib.revisiontree import RevisionTree
52
from bzrlib.store.versioned import VersionedFileStore
53
from bzrlib.testament import Testament
54
38
""")
55
39
56
import sys
57
40
from bzrlib import (
58
41
errors,
59
42
registry,
43
symbol_versioning,
60
44
ui,
61
45
)
62
46
from bzrlib.decorators import needs_read_lock, needs_write_lock, only_raises
63
47
from bzrlib.inter import InterObject
64
from bzrlib.inventory import (
65
Inventory,
66
InventoryDirectory,
67
ROOT_ID,
68
entry_factory,
69
)
70
from bzrlib.recordcounter import RecordCounter
71
48
from bzrlib.lock import _RelockDebugMixin, LogicalLockResult
72
49
from bzrlib.trace import (
73
50
log_exception_quietly, note, mutter, mutter_callsite, warning)
94
71
95
72
# all clients should supply tree roots.
96
73
record_root_entry = True
97
# the default CommitBuilder does not manage trees whose root is versioned.
98
_versioned_root = False
74
# whether this commit builder supports the record_entry_contents interface
75
supports_record_entry_contents = False
99
76
100
77
def __init__(self, repository, parents, config, timestamp=None,
101
78
timezone=None, committer=None, revprops=None,
102
revision_id=None):
79
revision_id=None, lossy=False):
103
80
"""Initiate a CommitBuilder.
104
81
105
82
:param repository: Repository to commit to.
106
83
:param parents: Revision ids of the parents of the new revision.
107
:param config: Configuration to use.
108
84
:param timestamp: Optional timestamp recorded for commit.
109
85
:param timezone: Optional timezone for timestamp.
110
86
:param committer: Optional committer to set for commit.
111
87
:param revprops: Optional dictionary of revision properties.
112
88
:param revision_id: Optional revision id.
89
:param lossy: Whether to discard data that can not be natively
90
represented, when pushing to a foreign VCS
113
91
"""
114
92
self._config = config
93
self._lossy = lossy
115
94
116
95
if committer is None:
117
96
self._committer = self._config.username()
120
99
else:
121
100
self._committer = committer
122
101
123
self.new_inventory = Inventory(None)
124
102
self._new_revision_id = revision_id
125
103
self.parents = parents
126
104
self.repository = repository
141
119
self._timezone = int(timezone)
142
120
143
121
self._generate_revision_if_needed()
144
self.__heads = graph.HeadsCache(repository.get_graph()).heads
145
self._basis_delta = []
146
# API compatibility, older code that used CommitBuilder did not call
147
# .record_delete(), which means the delta that is computed would not be
148
# valid. Callers that will call record_delete() should call
149
# .will_record_deletes() to indicate that.
150
self._recording_deletes = False
151
# memo'd check for no-op commits.
152
self._any_changes = False
153
122
154
123
def any_changes(self):
155
124
"""Return True if any entries were changed.
156
125
157
126
This includes merge-only changes. It is the core for the --unchanged
158
127
detection in commit.
159
128
160
129
:return: True if any changes have occured.
161
130
"""
162
return self._any_changes
131
raise NotImplementedError(self.any_changes)
163
132
164
133
def _validate_unicode_text(self, text, context):
165
134
"""Verify things like commit messages don't have bogus characters."""
176
145
self._validate_unicode_text(value,
177
146
'revision property (%s)' % (key,))
178
147
179
def _ensure_fallback_inventories(self):
180
"""Ensure that appropriate inventories are available.
181
182
This only applies to repositories that are stacked, and is about
183
enusring the stacking invariants. Namely, that for any revision that is
184
present, we either have all of the file content, or we have the parent
185
inventory and the delta file content.
186
"""
187
if not self.repository._fallback_repositories:
188
return
189
if not self.repository._format.supports_chks:
190
raise errors.BzrError("Cannot commit directly to a stacked branch"
191
" in pre-2a formats. See "
192
"https://bugs.launchpad.net/bzr/+bug/375013 for details.")
193
# This is a stacked repo, we need to make sure we have the parent
194
# inventories for the parents.
195
parent_keys = [(p,) for p in self.parents]
196
parent_map = self.repository.inventories._index.get_parent_map(parent_keys)
197
missing_parent_keys = set([pk for pk in parent_keys
198
if pk not in parent_map])
199
fallback_repos = list(reversed(self.repository._fallback_repositories))
200
missing_keys = [('inventories', pk[0])
201
for pk in missing_parent_keys]
202
resume_tokens = []
203
while missing_keys and fallback_repos:
204
fallback_repo = fallback_repos.pop()
205
source = fallback_repo._get_source(self.repository._format)
206
sink = self.repository._get_sink()
207
stream = source.get_stream_for_missing_keys(missing_keys)
208
missing_keys = sink.insert_stream_without_locking(stream,
209
self.repository._format)
210
if missing_keys:
211
raise errors.BzrError('Unable to fill in parent inventories for a'
212
' stacked branch')
213
214
148
def commit(self, message):
215
149
"""Make the actual commit.
216
150
217
151
:return: The revision id of the recorded revision.
218
152
"""
219
self._validate_unicode_text(message, 'commit message')
220
rev = _mod_revision.Revision(
221
timestamp=self._timestamp,
222
timezone=self._timezone,
223
committer=self._committer,
224
message=message,
225
inventory_sha1=self.inv_sha1,
226
revision_id=self._new_revision_id,
227
properties=self._revprops)
228
rev.parent_ids = self.parents
229
self.repository.add_revision(self._new_revision_id, rev,
230
self.new_inventory, self._config)
231
self._ensure_fallback_inventories()
232
self.repository.commit_write_group()
233
return self._new_revision_id
153
raise NotImplementedError(self.commit)
234
154
235
155
def abort(self):
236
156
"""Abort the commit that is being built.
237
157
"""
238
self.repository.abort_write_group()
158
raise NotImplementedError(self.abort)
239
159
240
160
def revision_tree(self):
241
161
"""Return the tree that was just committed.
242
162
243
After calling commit() this can be called to get a RevisionTree
244
representing the newly committed tree. This is preferred to
245
calling Repository.revision_tree() because that may require
246
deserializing the inventory, while we already have a copy in
163
After calling commit() this can be called to get a
164
RevisionTree representing the newly committed tree. This is
165
preferred to calling Repository.revision_tree() because that may
166
require deserializing the inventory, while we already have a copy in
247
167
memory.
248
168
"""
249
if self.new_inventory is None:
250
self.new_inventory = self.repository.get_inventory(
251
self._new_revision_id)
252
return RevisionTree(self.repository, self.new_inventory,
253
self._new_revision_id)
169
raise NotImplementedError(self.revision_tree)
254
170
255
171
def finish_inventory(self):
256
172
"""Tell the builder that the inventory is finished.
258
174
:return: The inventory id in the repository, which can be used with
259
175
repository.get_inventory.
260
176
"""
261
if self.new_inventory is None:
262
# an inventory delta was accumulated without creating a new
263
# inventory.
264
basis_id = self.basis_delta_revision
265
# We ignore the 'inventory' returned by add_inventory_by_delta
266
# because self.new_inventory is used to hint to the rest of the
267
# system what code path was taken
268
self.inv_sha1, _ = self.repository.add_inventory_by_delta(
269
basis_id, self._basis_delta, self._new_revision_id,
270
self.parents)
271
else:
272
if self.new_inventory.root is None:
273
raise AssertionError('Root entry should be supplied to'
274
' record_entry_contents, as of bzr 0.10.')
275
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
276
self.new_inventory.revision_id = self._new_revision_id
277
self.inv_sha1 = self.repository.add_inventory(
278
self._new_revision_id,
279
self.new_inventory,
280
self.parents
281
)
282
return self._new_revision_id
177
raise NotImplementedError(self.finish_inventory)
283
178
284
179
def _gen_revision_id(self):
285
180
"""Return new revision-id."""
300
195
else:
301
196
self.random_revid = False
302
197
303
def _heads(self, file_id, revision_ids):
304
"""Calculate the graph heads for revision_ids in the graph of file_id.
305
306
This can use either a per-file graph or a global revision graph as we
307
have an identity relationship between the two graphs.
308
"""
309
return self.__heads(revision_ids)
310
311
def _check_root(self, ie, parent_invs, tree):
312
"""Helper for record_entry_contents.
313
314
:param ie: An entry being added.
315
:param parent_invs: The inventories of the parent revisions of the
316
commit.
317
:param tree: The tree that is being committed.
318
"""
319
# In this revision format, root entries have no knit or weave When
320
# serializing out to disk and back in root.revision is always
321
# _new_revision_id
322
ie.revision = self._new_revision_id
323
324
def _require_root_change(self, tree):
325
"""Enforce an appropriate root object change.
326
327
This is called once when record_iter_changes is called, if and only if
328
the root was not in the delta calculated by record_iter_changes.
329
330
:param tree: The tree which is being committed.
331
"""
332
if len(self.parents) == 0:
333
raise errors.RootMissing()
334
entry = entry_factory['directory'](tree.path2id(''), '',
335
None)
336
entry.revision = self._new_revision_id
337
self._basis_delta.append(('', '', entry.file_id, entry))
338
339
def _get_delta(self, ie, basis_inv, path):
340
"""Get a delta against the basis inventory for ie."""
341
if ie.file_id not in basis_inv:
342
# add
343
result = (None, path, ie.file_id, ie)
344
self._basis_delta.append(result)
345
return result
346
elif ie != basis_inv[ie.file_id]:
347
# common but altered
348
# TODO: avoid tis id2path call.
349
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
350
self._basis_delta.append(result)
351
return result
352
else:
353
# common, unaltered
354
return None
355
356
def get_basis_delta(self):
357
"""Return the complete inventory delta versus the basis inventory.
358
359
This has been built up with the calls to record_delete and
360
record_entry_contents. The client must have already called
361
will_record_deletes() to indicate that they will be generating a
362
complete delta.
363
364
:return: An inventory delta, suitable for use with apply_delta, or
365
Repository.add_inventory_by_delta, etc.
366
"""
367
if not self._recording_deletes:
368
raise AssertionError("recording deletes not activated.")
369
return self._basis_delta
370
371
def record_delete(self, path, file_id):
372
"""Record that a delete occured against a basis tree.
373
374
This is an optional API - when used it adds items to the basis_delta
375
being accumulated by the commit builder. It cannot be called unless the
376
method will_record_deletes() has been called to inform the builder that
377
a delta is being supplied.
378
379
:param path: The path of the thing deleted.
380
:param file_id: The file id that was deleted.
381
"""
382
if not self._recording_deletes:
383
raise AssertionError("recording deletes not activated.")
384
delta = (path, None, file_id, None)
385
self._basis_delta.append(delta)
386
self._any_changes = True
387
return delta
388
389
198
def will_record_deletes(self):
390
199
"""Tell the commit builder that deletes are being notified.
391
200
393
202
commit to be valid, deletes against the basis MUST be recorded via
394
203
builder.record_delete().
395
204
"""
396
self._recording_deletes = True
397
try:
398
basis_id = self.parents[0]
399
except IndexError:
400
basis_id = _mod_revision.NULL_REVISION
401
self.basis_delta_revision = basis_id
402
403
def record_entry_contents(self, ie, parent_invs, path, tree,
404
content_summary):
405
"""Record the content of ie from tree into the commit if needed.
406
407
Side effect: sets ie.revision when unchanged
408
409
:param ie: An inventory entry present in the commit.
410
:param parent_invs: The inventories of the parent revisions of the
411
commit.
412
:param path: The path the entry is at in the tree.
413
:param tree: The tree which contains this entry and should be used to
414
obtain content.
415
:param content_summary: Summary data from the tree about the paths
416
content - stat, length, exec, sha/link target. This is only
417
accessed when the entry has a revision of None - that is when it is
418
a candidate to commit.
419
:return: A tuple (change_delta, version_recorded, fs_hash).
420
change_delta is an inventory_delta change for this entry against
421
the basis tree of the commit, or None if no change occured against
422
the basis tree.
423
version_recorded is True if a new version of the entry has been
424
recorded. For instance, committing a merge where a file was only
425
changed on the other side will return (delta, False).
426
fs_hash is either None, or the hash details for the path (currently
427
a tuple of the contents sha1 and the statvalue returned by
428
tree.get_file_with_stat()).
429
"""
430
if self.new_inventory.root is None:
431
if ie.parent_id is not None:
432
raise errors.RootMissing()
433
self._check_root(ie, parent_invs, tree)
434
if ie.revision is None:
435
kind = content_summary[0]
436
else:
437
# ie is carried over from a prior commit
438
kind = ie.kind
439
# XXX: repository specific check for nested tree support goes here - if
440
# the repo doesn't want nested trees we skip it ?
441
if (kind == 'tree-reference' and
442
not self.repository._format.supports_tree_reference):
443
# mismatch between commit builder logic and repository:
444
# this needs the entry creation pushed down into the builder.
445
raise NotImplementedError('Missing repository subtree support.')
446
self.new_inventory.add(ie)
447
448
# TODO: slow, take it out of the inner loop.
449
try:
450
basis_inv = parent_invs[0]
451
except IndexError:
452
basis_inv = Inventory(root_id=None)
453
454
# ie.revision is always None if the InventoryEntry is considered
455
# for committing. We may record the previous parents revision if the
456
# content is actually unchanged against a sole head.
457
if ie.revision is not None:
458
if not self._versioned_root and path == '':
459
# repositories that do not version the root set the root's
460
# revision to the new commit even when no change occurs (more
461
# specifically, they do not record a revision on the root; and
462
# the rev id is assigned to the root during deserialisation -
463
# this masks when a change may have occurred against the basis.
464
# To match this we always issue a delta, because the revision
465
# of the root will always be changing.
466
if ie.file_id in basis_inv:
467
delta = (basis_inv.id2path(ie.file_id), path,
468
ie.file_id, ie)
469
else:
470
# add
471
delta = (None, path, ie.file_id, ie)
472
self._basis_delta.append(delta)
473
return delta, False, None
474
else:
475
# we don't need to commit this, because the caller already
476
# determined that an existing revision of this file is
477
# appropriate. If it's not being considered for committing then
478
# it and all its parents to the root must be unaltered so
479
# no-change against the basis.
480
if ie.revision == self._new_revision_id:
481
raise AssertionError("Impossible situation, a skipped "
482
"inventory entry (%r) claims to be modified in this "
483
"commit (%r).", (ie, self._new_revision_id))
484
return None, False, None
485
# XXX: Friction: parent_candidates should return a list not a dict
486
# so that we don't have to walk the inventories again.
487
parent_candiate_entries = ie.parent_candidates(parent_invs)
488
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
489
heads = []
490
for inv in parent_invs:
491
if ie.file_id in inv:
492
old_rev = inv[ie.file_id].revision
493
if old_rev in head_set:
494
heads.append(inv[ie.file_id].revision)
495
head_set.remove(inv[ie.file_id].revision)
496
497
store = False
498
# now we check to see if we need to write a new record to the
499
# file-graph.
500
# We write a new entry unless there is one head to the ancestors, and
501
# the kind-derived content is unchanged.
502
503
# Cheapest check first: no ancestors, or more the one head in the
504
# ancestors, we write a new node.
505
if len(heads) != 1:
506
store = True
507
if not store:
508
# There is a single head, look it up for comparison
509
parent_entry = parent_candiate_entries[heads[0]]
510
# if the non-content specific data has changed, we'll be writing a
511
# node:
512
if (parent_entry.parent_id != ie.parent_id or
513
parent_entry.name != ie.name):
514
store = True
515
# now we need to do content specific checks:
516
if not store:
517
# if the kind changed the content obviously has
518
if kind != parent_entry.kind:
519
store = True
520
# Stat cache fingerprint feedback for the caller - None as we usually
521
# don't generate one.
522
fingerprint = None
523
if kind == 'file':
524
if content_summary[2] is None:
525
raise ValueError("Files must not have executable = None")
526
if not store:
527
# We can't trust a check of the file length because of content
528
# filtering...
529
if (# if the exec bit has changed we have to store:
530
parent_entry.executable != content_summary[2]):
531
store = True
532
elif parent_entry.text_sha1 == content_summary[3]:
533
# all meta and content is unchanged (using a hash cache
534
# hit to check the sha)
535
ie.revision = parent_entry.revision
536
ie.text_size = parent_entry.text_size
537
ie.text_sha1 = parent_entry.text_sha1
538
ie.executable = parent_entry.executable
539
return self._get_delta(ie, basis_inv, path), False, None
540
else:
541
# Either there is only a hash change(no hash cache entry,
542
# or same size content change), or there is no change on
543
# this file at all.
544
# Provide the parent's hash to the store layer, so that the
545
# content is unchanged we will not store a new node.
546
nostore_sha = parent_entry.text_sha1
547
if store:
548
# We want to record a new node regardless of the presence or
549
# absence of a content change in the file.
550
nostore_sha = None
551
ie.executable = content_summary[2]
552
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
553
try:
554
text = file_obj.read()
555
finally:
556
file_obj.close()
557
try:
558
ie.text_sha1, ie.text_size = self._add_text_to_weave(
559
ie.file_id, text, heads, nostore_sha)
560
# Let the caller know we generated a stat fingerprint.
561
fingerprint = (ie.text_sha1, stat_value)
562
except errors.ExistingContent:
563
# Turns out that the file content was unchanged, and we were
564
# only going to store a new node if it was changed. Carry over
565
# the entry.
566
ie.revision = parent_entry.revision
567
ie.text_size = parent_entry.text_size
568
ie.text_sha1 = parent_entry.text_sha1
569
ie.executable = parent_entry.executable
570
return self._get_delta(ie, basis_inv, path), False, None
571
elif kind == 'directory':
572
if not store:
573
# all data is meta here, nothing specific to directory, so
574
# carry over:
575
ie.revision = parent_entry.revision
576
return self._get_delta(ie, basis_inv, path), False, None
577
self._add_text_to_weave(ie.file_id, '', heads, None)
578
elif kind == 'symlink':
579
current_link_target = content_summary[3]
580
if not store:
581
# symlink target is not generic metadata, check if it has
582
# changed.
583
if current_link_target != parent_entry.symlink_target:
584
store = True
585
if not store:
586
# unchanged, carry over.
587
ie.revision = parent_entry.revision
588
ie.symlink_target = parent_entry.symlink_target
589
return self._get_delta(ie, basis_inv, path), False, None
590
ie.symlink_target = current_link_target
591
self._add_text_to_weave(ie.file_id, '', heads, None)
592
elif kind == 'tree-reference':
593
if not store:
594
if content_summary[3] != parent_entry.reference_revision:
595
store = True
596
if not store:
597
# unchanged, carry over.
598
ie.reference_revision = parent_entry.reference_revision
599
ie.revision = parent_entry.revision
600
return self._get_delta(ie, basis_inv, path), False, None
601
ie.reference_revision = content_summary[3]
602
if ie.reference_revision is None:
603
raise AssertionError("invalid content_summary for nested tree: %r"
604
% (content_summary,))
605
self._add_text_to_weave(ie.file_id, '', heads, None)
606
else:
607
raise NotImplementedError('unknown kind')
608
ie.revision = self._new_revision_id
609
self._any_changes = True
610
return self._get_delta(ie, basis_inv, path), True, fingerprint
611
612
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
613
_entry_factory=entry_factory):
205
raise NotImplementedError(self.will_record_deletes)
206
207
def record_iter_changes(self, tree, basis_revision_id, iter_changes):
614
208
"""Record a new tree via iter_changes.
615
209
616
210
:param tree: The tree to obtain text contents from for changed objects.
621
215
to basis_revision_id. The iterator must not include any items with
622
216
a current kind of None - missing items must be either filtered out
623
217
or errored-on beefore record_iter_changes sees the item.
624
:param _entry_factory: Private method to bind entry_factory locally for
625
performance.
626
218
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
627
219
tree._observed_sha1.
628
220
"""
629
# Create an inventory delta based on deltas between all the parents and
630
# deltas between all the parent inventories. We use inventory delta's
631
# between the inventory objects because iter_changes masks
632
# last-changed-field only changes.
633
# Working data:
634
# file_id -> change map, change is fileid, paths, changed, versioneds,
635
# parents, names, kinds, executables
636
merged_ids = {}
637
# {file_id -> revision_id -> inventory entry, for entries in parent
638
# trees that are not parents[0]
639
parent_entries = {}
640
ghost_basis = False
641
try:
642
revtrees = list(self.repository.revision_trees(self.parents))
643
except errors.NoSuchRevision:
644
# one or more ghosts, slow path.
645
revtrees = []
646
for revision_id in self.parents:
647
try:
648
revtrees.append(self.repository.revision_tree(revision_id))
649
except errors.NoSuchRevision:
650
if not revtrees:
651
basis_revision_id = _mod_revision.NULL_REVISION
652
ghost_basis = True
653
revtrees.append(self.repository.revision_tree(
654
_mod_revision.NULL_REVISION))
655
# The basis inventory from a repository
656
if revtrees:
657
basis_inv = revtrees[0].inventory
658
else:
659
basis_inv = self.repository.revision_tree(
660
_mod_revision.NULL_REVISION).inventory
661
if len(self.parents) > 0:
662
if basis_revision_id != self.parents[0] and not ghost_basis:
663
raise Exception(
664
"arbitrary basis parents not yet supported with merges")
665
for revtree in revtrees[1:]:
666
for change in revtree.inventory._make_delta(basis_inv):
667
if change[1] is None:
668
# Not present in this parent.
669
continue
670
if change[2] not in merged_ids:
671
if change[0] is not None:
672
basis_entry = basis_inv[change[2]]
673
merged_ids[change[2]] = [
674
# basis revid
675
basis_entry.revision,
676
# new tree revid
677
change[3].revision]
678
parent_entries[change[2]] = {
679
# basis parent
680
basis_entry.revision:basis_entry,
681
# this parent
682
change[3].revision:change[3],
683
}
684
else:
685
merged_ids[change[2]] = [change[3].revision]
686
parent_entries[change[2]] = {change[3].revision:change[3]}
687
else:
688
merged_ids[change[2]].append(change[3].revision)
689
parent_entries[change[2]][change[3].revision] = change[3]
690
else:
691
merged_ids = {}
692
# Setup the changes from the tree:
693
# changes maps file_id -> (change, [parent revision_ids])
694
changes= {}
695
for change in iter_changes:
696
# This probably looks up in basis_inv way to much.
697
if change[1][0] is not None:
698
head_candidate = [basis_inv[change[0]].revision]
699
else:
700
head_candidate = []
701
changes[change[0]] = change, merged_ids.get(change[0],
702
head_candidate)
703
unchanged_merged = set(merged_ids) - set(changes)
704
# Extend the changes dict with synthetic changes to record merges of
705
# texts.
706
for file_id in unchanged_merged:
707
# Record a merged version of these items that did not change vs the
708
# basis. This can be either identical parallel changes, or a revert
709
# of a specific file after a merge. The recorded content will be
710
# that of the current tree (which is the same as the basis), but
711
# the per-file graph will reflect a merge.
712
# NB:XXX: We are reconstructing path information we had, this
713
# should be preserved instead.
714
# inv delta change: (file_id, (path_in_source, path_in_target),
715
# changed_content, versioned, parent, name, kind,
716
# executable)
717
try:
718
basis_entry = basis_inv[file_id]
719
except errors.NoSuchId:
720
# a change from basis->some_parents but file_id isn't in basis
721
# so was new in the merge, which means it must have changed
722
# from basis -> current, and as it hasn't the add was reverted
723
# by the user. So we discard this change.
724
pass
725
else:
726
change = (file_id,
727
(basis_inv.id2path(file_id), tree.id2path(file_id)),
728
False, (True, True),
729
(basis_entry.parent_id, basis_entry.parent_id),
730
(basis_entry.name, basis_entry.name),
731
(basis_entry.kind, basis_entry.kind),
732
(basis_entry.executable, basis_entry.executable))
733
changes[file_id] = (change, merged_ids[file_id])
734
# changes contains tuples with the change and a set of inventory
735
# candidates for the file.
736
# inv delta is:
737
# old_path, new_path, file_id, new_inventory_entry
738
seen_root = False # Is the root in the basis delta?
739
inv_delta = self._basis_delta
740
modified_rev = self._new_revision_id
741
for change, head_candidates in changes.values():
742
if change[3][1]: # versioned in target.
743
# Several things may be happening here:
744
# We may have a fork in the per-file graph
745
# - record a change with the content from tree
746
# We may have a change against < all trees
747
# - carry over the tree that hasn't changed
748
# We may have a change against all trees
749
# - record the change with the content from tree
750
kind = change[6][1]
751
file_id = change[0]
752
entry = _entry_factory[kind](file_id, change[5][1],
753
change[4][1])
754
head_set = self._heads(change[0], set(head_candidates))
755
heads = []
756
# Preserve ordering.
757
for head_candidate in head_candidates:
758
if head_candidate in head_set:
759
heads.append(head_candidate)
760
head_set.remove(head_candidate)
761
carried_over = False
762
if len(heads) == 1:
763
# Could be a carry-over situation:
764
parent_entry_revs = parent_entries.get(file_id, None)
765
if parent_entry_revs:
766
parent_entry = parent_entry_revs.get(heads[0], None)
767
else:
768
parent_entry = None
769
if parent_entry is None:
770
# The parent iter_changes was called against is the one
771
# that is the per-file head, so any change is relevant
772
# iter_changes is valid.
773
carry_over_possible = False
774
else:
775
# could be a carry over situation
776
# A change against the basis may just indicate a merge,
777
# we need to check the content against the source of the
778
# merge to determine if it was changed after the merge
779
# or carried over.
780
if (parent_entry.kind != entry.kind or
781
parent_entry.parent_id != entry.parent_id or
782
parent_entry.name != entry.name):
783
# Metadata common to all entries has changed
784
# against per-file parent
785
carry_over_possible = False
786
else:
787
carry_over_possible = True
788
# per-type checks for changes against the parent_entry
789
# are done below.
790
else:
791
# Cannot be a carry-over situation
792
carry_over_possible = False
793
# Populate the entry in the delta
794
if kind == 'file':
795
# XXX: There is still a small race here: If someone reverts the content of a file
796
# after iter_changes examines and decides it has changed,
797
# we will unconditionally record a new version even if some
798
# other process reverts it while commit is running (with
799
# the revert happening after iter_changes did its
800
# examination).
801
if change[7][1]:
802
entry.executable = True
803
else:
804
entry.executable = False
805
if (carry_over_possible and
806
parent_entry.executable == entry.executable):
807
# Check the file length, content hash after reading
808
# the file.
809
nostore_sha = parent_entry.text_sha1
810
else:
811
nostore_sha = None
812
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
813
try:
814
text = file_obj.read()
815
finally:
816
file_obj.close()
817
try:
818
entry.text_sha1, entry.text_size = self._add_text_to_weave(
819
file_id, text, heads, nostore_sha)
820
yield file_id, change[1][1], (entry.text_sha1, stat_value)
821
except errors.ExistingContent:
822
# No content change against a carry_over parent
823
# Perhaps this should also yield a fs hash update?
824
carried_over = True
825
entry.text_size = parent_entry.text_size
826
entry.text_sha1 = parent_entry.text_sha1
827
elif kind == 'symlink':
828
# Wants a path hint?
829
entry.symlink_target = tree.get_symlink_target(file_id)
830
if (carry_over_possible and
831
parent_entry.symlink_target == entry.symlink_target):
832
carried_over = True
833
else:
834
self._add_text_to_weave(change[0], '', heads, None)
835
elif kind == 'directory':
836
if carry_over_possible:
837
carried_over = True
838
else:
839
# Nothing to set on the entry.
840
# XXX: split into the Root and nonRoot versions.
841
if change[1][1] != '' or self.repository.supports_rich_root():
842
self._add_text_to_weave(change[0], '', heads, None)
843
elif kind == 'tree-reference':
844
if not self.repository._format.supports_tree_reference:
845
# This isn't quite sane as an error, but we shouldn't
846
# ever see this code path in practice: tree's don't
847
# permit references when the repo doesn't support tree
848
# references.
849
raise errors.UnsupportedOperation(tree.add_reference,
850
self.repository)
851
reference_revision = tree.get_reference_revision(change[0])
852
entry.reference_revision = reference_revision
853
if (carry_over_possible and
854
parent_entry.reference_revision == reference_revision):
855
carried_over = True
856
else:
857
self._add_text_to_weave(change[0], '', heads, None)
858
else:
859
raise AssertionError('unknown kind %r' % kind)
860
if not carried_over:
861
entry.revision = modified_rev
862
else:
863
entry.revision = parent_entry.revision
864
else:
865
entry = None
866
new_path = change[1][1]
867
inv_delta.append((change[1][0], new_path, change[0], entry))
868
if new_path == '':
869
seen_root = True
870
self.new_inventory = None
871
if len(inv_delta):
872
# This should perhaps be guarded by a check that the basis we
873
# commit against is the basis for the commit and if not do a delta
874
# against the basis.
875
self._any_changes = True
876
if not seen_root:
877
# housekeeping root entry changes do not affect no-change commits.
878
self._require_root_change(tree)
879
self.basis_delta_revision = basis_revision_id
880
881
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
882
parent_keys = tuple([(file_id, parent) for parent in parents])
883
return self.repository.texts._add_text(
884
(file_id, self._new_revision_id), parent_keys, new_text,
885
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
886
887
888
class RootCommitBuilder(CommitBuilder):
889
"""This commitbuilder actually records the root id"""
890
891
# the root entry gets versioned properly by this builder.
892
_versioned_root = True
893
894
def _check_root(self, ie, parent_invs, tree):
895
"""Helper for record_entry_contents.
896
897
:param ie: An entry being added.
898
:param parent_invs: The inventories of the parent revisions of the
899
commit.
900
:param tree: The tree that is being committed.
901
"""
902
903
def _require_root_change(self, tree):
904
"""Enforce an appropriate root object change.
905
906
This is called once when record_iter_changes is called, if and only if
907
the root was not in the delta calculated by record_iter_changes.
908
909
:param tree: The tree which is being committed.
910
"""
911
# versioned roots do not change unless the tree found a change.
221
raise NotImplementedError(self.record_iter_changes)
912
222
913
223
914
224
class RepositoryWriteLockResult(LogicalLockResult):
939
249
revisions and file history. It's normally accessed only by the Branch,
940
250
which views a particular line of development through that history.
941
251
942
The Repository builds on top of some byte storage facilies (the revisions,
943
signatures, inventories, texts and chk_bytes attributes) and a Transport,
944
which respectively provide byte storage and a means to access the (possibly
945
remote) disk.
946
947
The byte storage facilities are addressed via tuples, which we refer to
948
as 'keys' throughout the code base. Revision_keys, inventory_keys and
949
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
950
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
951
byte string made up of a hash identifier and a hash value.
952
We use this interface because it allows low friction with the underlying
953
code that implements disk indices, network encoding and other parts of
954
bzrlib.
955
956
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
957
the serialised revisions for the repository. This can be used to obtain
958
revision graph information or to access raw serialised revisions.
959
The result of trying to insert data into the repository via this store
960
is undefined: it should be considered read-only except for implementors
961
of repositories.
962
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
963
the serialised signatures for the repository. This can be used to
964
obtain access to raw serialised signatures. The result of trying to
965
insert data into the repository via this store is undefined: it should
966
be considered read-only except for implementors of repositories.
967
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
968
the serialised inventories for the repository. This can be used to
969
obtain unserialised inventories. The result of trying to insert data
970
into the repository via this store is undefined: it should be
971
considered read-only except for implementors of repositories.
972
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
973
texts of files and directories for the repository. This can be used to
974
obtain file texts or file graphs. Note that Repository.iter_file_bytes
975
is usually a better interface for accessing file texts.
976
The result of trying to insert data into the repository via this store
977
is undefined: it should be considered read-only except for implementors
978
of repositories.
979
:ivar chk_bytes: A bzrlib.versionedfile.VersionedFiles instance containing
980
any data the repository chooses to store or have indexed by its hash.
981
The result of trying to insert data into the repository via this store
982
is undefined: it should be considered read-only except for implementors
983
of repositories.
984
:ivar _transport: Transport for file access to repository, typically
985
pointing to .bzr/repository.
252
See VersionedFileRepository in bzrlib.vf_repository for the
253
base class for most Bazaar repositories.
986
254
"""
987
255
988
# What class to use for a CommitBuilder. Often it's simpler to change this
989
# in a Repository class subclass rather than to override
990
# get_commit_builder.
991
_commit_builder_class = CommitBuilder
992
# The search regex used by xml based repositories to determine what things
993
# where changed in a single commit.
994
_file_ids_altered_regex = lazy_regex.lazy_compile(
995
r'file_id="(?P<file_id>[^"]+)"'
996
r'.* revision="(?P<revision_id>[^"]+)"'
997
)
998
999
256
def abort_write_group(self, suppress_errors=False):
1000
257
"""Commit the contents accrued within the current write group.
1001
258
1044
301
1045
302
:param repository: A repository.
1046
303
"""
1047
if not self._format.supports_external_lookups:
1048
raise errors.UnstackableRepositoryFormat(self._format, self.base)
1049
if self.is_locked():
1050
# This repository will call fallback.unlock() when we transition to
1051
# the unlocked state, so we make sure to increment the lock count
1052
repository.lock_read()
1053
self._check_fallback_repository(repository)
1054
self._fallback_repositories.append(repository)
1055
self.texts.add_fallback_versioned_files(repository.texts)
1056
self.inventories.add_fallback_versioned_files(repository.inventories)
1057
self.revisions.add_fallback_versioned_files(repository.revisions)
1058
self.signatures.add_fallback_versioned_files(repository.signatures)
1059
if self.chk_bytes is not None:
1060
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
304
raise NotImplementedError(self.add_fallback_repository)
1061
305
1062
306
def _check_fallback_repository(self, repository):
1063
307
"""Check that this repository can fallback to repository safely.
1068
312
"""
1069
313
return InterRepository._assert_same_model(self, repository)
1070
314
1071
def add_inventory(self, revision_id, inv, parents):
1072
"""Add the inventory inv to the repository as revision_id.
1073
1074
:param parents: The revision ids of the parents that revision_id
1075
is known to have and are in the repository already.
1076
1077
:returns: The validator(which is a sha1 digest, though what is sha'd is
1078
repository format specific) of the serialized inventory.
1079
"""
1080
if not self.is_in_write_group():
1081
raise AssertionError("%r not in write group" % (self,))
1082
_mod_revision.check_not_reserved_id(revision_id)
1083
if not (inv.revision_id is None or inv.revision_id == revision_id):
1084
raise AssertionError(
1085
"Mismatch between inventory revision"
1086
" id and insertion revid (%r, %r)"
1087
% (inv.revision_id, revision_id))
1088
if inv.root is None:
1089
raise errors.RootMissing()
1090
return self._add_inventory_checked(revision_id, inv, parents)
1091
1092
def _add_inventory_checked(self, revision_id, inv, parents):
1093
"""Add inv to the repository after checking the inputs.
1094
1095
This function can be overridden to allow different inventory styles.
1096
1097
:seealso: add_inventory, for the contract.
1098
"""
1099
inv_lines = self._serializer.write_inventory_to_lines(inv)
1100
return self._inventory_add_lines(revision_id, parents,
1101
inv_lines, check_content=False)
1102
1103
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1104
parents, basis_inv=None, propagate_caches=False):
1105
"""Add a new inventory expressed as a delta against another revision.
1106
1107
See the inventory developers documentation for the theory behind
1108
inventory deltas.
1109
1110
:param basis_revision_id: The inventory id the delta was created
1111
against. (This does not have to be a direct parent.)
1112
:param delta: The inventory delta (see Inventory.apply_delta for
1113
details).
1114
:param new_revision_id: The revision id that the inventory is being
1115
added for.
1116
:param parents: The revision ids of the parents that revision_id is
1117
known to have and are in the repository already. These are supplied
1118
for repositories that depend on the inventory graph for revision
1119
graph access, as well as for those that pun ancestry with delta
1120
compression.
1121
:param basis_inv: The basis inventory if it is already known,
1122
otherwise None.
1123
:param propagate_caches: If True, the caches for this inventory are
1124
copied to and updated for the result if possible.
1125
1126
:returns: (validator, new_inv)
1127
The validator(which is a sha1 digest, though what is sha'd is
1128
repository format specific) of the serialized inventory, and the
1129
resulting inventory.
1130
"""
1131
if not self.is_in_write_group():
1132
raise AssertionError("%r not in write group" % (self,))
1133
_mod_revision.check_not_reserved_id(new_revision_id)
1134
basis_tree = self.revision_tree(basis_revision_id)
1135
basis_tree.lock_read()
1136
try:
1137
# Note that this mutates the inventory of basis_tree, which not all
1138
# inventory implementations may support: A better idiom would be to
1139
# return a new inventory, but as there is no revision tree cache in
1140
# repository this is safe for now - RBC 20081013
1141
if basis_inv is None:
1142
basis_inv = basis_tree.inventory
1143
basis_inv.apply_delta(delta)
1144
basis_inv.revision_id = new_revision_id
1145
return (self.add_inventory(new_revision_id, basis_inv, parents),
1146
basis_inv)
1147
finally:
1148
basis_tree.unlock()
1149
1150
def _inventory_add_lines(self, revision_id, parents, lines,
1151
check_content=True):
1152
"""Store lines in inv_vf and return the sha1 of the inventory."""
1153
parents = [(parent,) for parent in parents]
1154
result = self.inventories.add_lines((revision_id,), parents, lines,
1155
check_content=check_content)[0]
1156
self.inventories._access.flush()
1157
return result
1158
1159
def add_revision(self, revision_id, rev, inv=None, config=None):
1160
"""Add rev to the revision store as revision_id.
1161
1162
:param revision_id: the revision id to use.
1163
:param rev: The revision object.
1164
:param inv: The inventory for the revision. if None, it will be looked
1165
up in the inventory storer
1166
:param config: If None no digital signature will be created.
1167
If supplied its signature_needed method will be used
1168
to determine if a signature should be made.
1169
"""
1170
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
1171
# rev.parent_ids?
1172
_mod_revision.check_not_reserved_id(revision_id)
1173
if config is not None and config.signature_needed():
1174
if inv is None:
1175
inv = self.get_inventory(revision_id)
1176
plaintext = Testament(rev, inv).as_short_text()
1177
self.store_revision_signature(
1178
gpg.GPGStrategy(config), plaintext, revision_id)
1179
# check inventory present
1180
if not self.inventories.get_parent_map([(revision_id,)]):
1181
if inv is None:
1182
raise errors.WeaveRevisionNotPresent(revision_id,
1183
self.inventories)
1184
else:
1185
# yes, this is not suitable for adding with ghosts.
1186
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
1187
rev.parent_ids)
1188
else:
1189
key = (revision_id,)
1190
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
1191
self._add_revision(rev)
1192
1193
def _add_revision(self, revision):
1194
text = self._serializer.write_revision_to_string(revision)
1195
key = (revision.revision_id,)
1196
parents = tuple((parent,) for parent in revision.parent_ids)
1197
self.revisions.add_lines(key, parents, osutils.split_lines(text))
1198
1199
315
def all_revision_ids(self):
1200
316
"""Returns a list of all the revision ids in the repository.
1201
317
1236
352
# The old API returned a list, should this actually be a set?
1237
353
return parent_map.keys()
1238
354
1239
def _check_inventories(self, checker):
1240
"""Check the inventories found from the revision scan.
1241
1242
This is responsible for verifying the sha1 of inventories and
1243
creating a pending_keys set that covers data referenced by inventories.
1244
"""
1245
bar = ui.ui_factory.nested_progress_bar()
1246
try:
1247
self._do_check_inventories(checker, bar)
1248
finally:
1249
bar.finished()
1250
1251
def _do_check_inventories(self, checker, bar):
1252
"""Helper for _check_inventories."""
1253
revno = 0
1254
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1255
kinds = ['chk_bytes', 'texts']
1256
count = len(checker.pending_keys)
1257
bar.update("inventories", 0, 2)
1258
current_keys = checker.pending_keys
1259
checker.pending_keys = {}
1260
# Accumulate current checks.
1261
for key in current_keys:
1262
if key[0] != 'inventories' and key[0] not in kinds:
1263
checker._report_items.append('unknown key type %r' % (key,))
1264
keys[key[0]].add(key[1:])
1265
if keys['inventories']:
1266
# NB: output order *should* be roughly sorted - topo or
1267
# inverse topo depending on repository - either way decent
1268
# to just delta against. However, pre-CHK formats didn't
1269
# try to optimise inventory layout on disk. As such the
1270
# pre-CHK code path does not use inventory deltas.
1271
last_object = None
1272
for record in self.inventories.check(keys=keys['inventories']):
1273
if record.storage_kind == 'absent':
1274
checker._report_items.append(
1275
'Missing inventory {%s}' % (record.key,))
1276
else:
1277
last_object = self._check_record('inventories', record,
1278
checker, last_object,
1279
current_keys[('inventories',) + record.key])
1280
del keys['inventories']
1281
else:
1282
return
1283
bar.update("texts", 1)
1284
while (checker.pending_keys or keys['chk_bytes']
1285
or keys['texts']):
1286
# Something to check.
1287
current_keys = checker.pending_keys
1288
checker.pending_keys = {}
1289
# Accumulate current checks.
1290
for key in current_keys:
1291
if key[0] not in kinds:
1292
checker._report_items.append('unknown key type %r' % (key,))
1293
keys[key[0]].add(key[1:])
1294
# Check the outermost kind only - inventories || chk_bytes || texts
1295
for kind in kinds:
1296
if keys[kind]:
1297
last_object = None
1298
for record in getattr(self, kind).check(keys=keys[kind]):
1299
if record.storage_kind == 'absent':
1300
checker._report_items.append(
1301
'Missing %s {%s}' % (kind, record.key,))
1302
else:
1303
last_object = self._check_record(kind, record,
1304
checker, last_object, current_keys[(kind,) + record.key])
1305
keys[kind] = set()
1306
break
1307
1308
def _check_record(self, kind, record, checker, last_object, item_data):
1309
"""Check a single text from this repository."""
1310
if kind == 'inventories':
1311
rev_id = record.key[0]
1312
inv = self._deserialise_inventory(rev_id,
1313
record.get_bytes_as('fulltext'))
1314
if last_object is not None:
1315
delta = inv._make_delta(last_object)
1316
for old_path, path, file_id, ie in delta:
1317
if ie is None:
1318
continue
1319
ie.check(checker, rev_id, inv)
1320
else:
1321
for path, ie in inv.iter_entries():
1322
ie.check(checker, rev_id, inv)
1323
if self._format.fast_deltas:
1324
return inv
1325
elif kind == 'chk_bytes':
1326
# No code written to check chk_bytes for this repo format.
1327
checker._report_items.append(
1328
'unsupported key type chk_bytes for %s' % (record.key,))
1329
elif kind == 'texts':
1330
self._check_text(record, checker, item_data)
1331
else:
1332
checker._report_items.append(
1333
'unknown key type %s for %s' % (kind, record.key))
1334
1335
def _check_text(self, record, checker, item_data):
1336
"""Check a single text."""
1337
# Check it is extractable.
1338
# TODO: check length.
1339
if record.storage_kind == 'chunked':
1340
chunks = record.get_bytes_as(record.storage_kind)
1341
sha1 = osutils.sha_strings(chunks)
1342
length = sum(map(len, chunks))
1343
else:
1344
content = record.get_bytes_as('fulltext')
1345
sha1 = osutils.sha_string(content)
1346
length = len(content)
1347
if item_data and sha1 != item_data[1]:
1348
checker._report_items.append(
1349
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1350
(record.key, sha1, item_data[1], item_data[2]))
1351
1352
355
@staticmethod
1353
356
def create(a_bzrdir):
1354
357
"""Construct the current default format repository in a_bzrdir."""
1359
362
1360
363
:param _format: The format of the repository on disk.
1361
364
:param a_bzrdir: The BzrDir of the repository.
365
:param control_files: Control files to use for locking, etc.
1362
366
"""
1363
367
# In the future we will have a single api for all stores for
1364
368
# getting file texts, inventories and revisions, then
1371
375
self._transport = control_files._transport
1372
376
self.base = self._transport.base
1373
377
# for tests
1374
self._reconcile_does_inventory_gc = True
1375
self._reconcile_fixes_text_parents = False
1376
self._reconcile_backsup_inventory = True
1377
378
self._write_group = None
1378
379
# Additional places to query for data.
1379
380
self._fallback_repositories = []
1380
# An InventoryEntry cache, used during deserialization
1381
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1382
# Is it safe to return inventory entries directly from the entry cache,
1383
# rather copying them?
1384
self._safe_to_return_from_cache = False
1385
381
1386
382
@property
1387
383
def user_transport(self):
1527
523
if revid and committers:
1528
524
result['committers'] = 0
1529
525
if revid and revid != _mod_revision.NULL_REVISION:
526
graph = self.get_graph()
1530
527
if committers:
1531
528
all_committers = set()
1532
revisions = self.get_ancestry(revid)
1533
# pop the leading None
1534
revisions.pop(0)
1535
first_revision = None
529
revisions = [r for (r, p) in graph.iter_ancestry([revid])
530
if r != _mod_revision.NULL_REVISION]
531
last_revision = None
1536
532
if not committers:
1537
533
# ignore the revisions in the middle - just grab first and last
1538
534
revisions = revisions[0], revisions[-1]
1539
535
for revision in self.get_revisions(revisions):
1540
if not first_revision:
1541
first_revision = revision
536
if not last_revision:
537
last_revision = revision
1542
538
if committers:
1543
539
all_committers.add(revision.committer)
1544
last_revision = revision
540
first_revision = revision
1545
541
if committers:
1546
542
result['committers'] = len(all_committers)
1547
543
result['firstrev'] = (first_revision.timestamp,
1548
544
first_revision.timezone)
1549
545
result['latestrev'] = (last_revision.timestamp,
1550
546
last_revision.timezone)
1551
1552
# now gather global repository information
1553
# XXX: This is available for many repos regardless of listability.
1554
if self.user_transport.listable():
1555
# XXX: do we want to __define len__() ?
1556
# Maybe the versionedfiles object should provide a different
1557
# method to get the number of keys.
1558
result['revisions'] = len(self.revisions.keys())
1559
# result['size'] = t
1560
547
return result
1561
548
1562
549
def find_branches(self, using=False):
1597
584
return ret
1598
585
1599
586
@needs_read_lock
1600
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
587
def search_missing_revision_ids(self, other,
588
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
589
find_ghosts=True, revision_ids=None, if_present_ids=None,
590
limit=None):
1601
591
"""Return the revision ids that other has that this does not.
1602
592
1603
593
These are returned in topological order.
1604
594
1605
595
revision_id: only return revision ids included by revision_id.
1606
596
"""
597
if symbol_versioning.deprecated_passed(revision_id):
598
symbol_versioning.warn(
599
'search_missing_revision_ids(revision_id=...) was '
600
'deprecated in 2.4. Use revision_ids=[...] instead.',
601
DeprecationWarning, stacklevel=3)
602
if revision_ids is not None:
603
raise AssertionError(
604
'revision_ids is mutually exclusive with revision_id')
605
if revision_id is not None:
606
revision_ids = [revision_id]
1607
607
return InterRepository.get(other, self).search_missing_revision_ids(
1608
revision_id, find_ghosts)
608
find_ghosts=find_ghosts, revision_ids=revision_ids,
609
if_present_ids=if_present_ids, limit=limit)
1609
610
1610
611
@staticmethod
1611
612
def open(base):
1653
654
def suspend_write_group(self):
1654
655
raise errors.UnsuspendableWriteGroup(self)
1655
656
1656
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1657
"""Return the keys of missing inventory parents for revisions added in
1658
this write group.
1659
1660
A revision is not complete if the inventory delta for that revision
1661
cannot be calculated. Therefore if the parent inventories of a
1662
revision are not present, the revision is incomplete, and e.g. cannot
1663
be streamed by a smart server. This method finds missing inventory
1664
parents for revisions added in this write group.
1665
"""
1666
if not self._format.supports_external_lookups:
1667
# This is only an issue for stacked repositories
1668
return set()
1669
if not self.is_in_write_group():
1670
raise AssertionError('not in a write group')
1671
1672
# XXX: We assume that every added revision already has its
1673
# corresponding inventory, so we only check for parent inventories that
1674
# might be missing, rather than all inventories.
1675
parents = set(self.revisions._index.get_missing_parents())
1676
parents.discard(_mod_revision.NULL_REVISION)
1677
unstacked_inventories = self.inventories._index
1678
present_inventories = unstacked_inventories.get_parent_map(
1679
key[-1:] for key in parents)
1680
parents.difference_update(present_inventories)
1681
if len(parents) == 0:
1682
# No missing parent inventories.
1683
return set()
1684
if not check_for_missing_texts:
1685
return set(('inventories', rev_id) for (rev_id,) in parents)
1686
# Ok, now we have a list of missing inventories. But these only matter
1687
# if the inventories that reference them are missing some texts they
1688
# appear to introduce.
1689
# XXX: Texts referenced by all added inventories need to be present,
1690
# but at the moment we're only checking for texts referenced by
1691
# inventories at the graph's edge.
1692
key_deps = self.revisions._index._key_dependencies
1693
key_deps.satisfy_refs_for_keys(present_inventories)
1694
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1695
file_ids = self.fileids_altered_by_revision_ids(referrers)
1696
missing_texts = set()
1697
for file_id, version_ids in file_ids.iteritems():
1698
missing_texts.update(
1699
(file_id, version_id) for version_id in version_ids)
1700
present_texts = self.texts.get_parent_map(missing_texts)
1701
missing_texts.difference_update(present_texts)
1702
if not missing_texts:
1703
# No texts are missing, so all revisions and their deltas are
1704
# reconstructable.
1705
return set()
1706
# Alternatively the text versions could be returned as the missing
1707
# keys, but this is likely to be less data.
1708
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1709
return missing_keys
1710
1711
657
def refresh_data(self):
1712
658
"""Re-read any data needed to synchronise with disk.
1713
659
1733
679
def _resume_write_group(self, tokens):
1734
680
raise errors.UnsuspendableWriteGroup(self)
1735
681
1736
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
682
def fetch(self, source, revision_id=None, find_ghosts=False,
1737
683
fetch_spec=None):
1738
684
"""Fetch the content required to construct revision_id from source.
1739
685
1773
719
not _mod_revision.is_null(revision_id)):
1774
720
self.get_revision(revision_id)
1775
721
return 0, []
1776
# if there is no specific appropriate InterRepository, this will get
1777
# the InterRepository base class, which raises an
1778
# IncompatibleRepositories when asked to fetch.
1779
722
inter = InterRepository.get(source, self)
1780
return inter.fetch(revision_id=revision_id, pb=pb,
723
return inter.fetch(revision_id=revision_id,
1781
724
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
1782
725
1783
726
def create_bundle(self, target, base, fileobj, format=None):
1785
728
1786
729
def get_commit_builder(self, branch, parents, config, timestamp=None,
1787
730
timezone=None, committer=None, revprops=None,
1788
revision_id=None):
731
revision_id=None, lossy=False):
1789
732
"""Obtain a CommitBuilder for this repository.
1790
733
1791
734
:param branch: Branch to commit to.
1796
739
:param committer: Optional committer to set for commit.
1797
740
:param revprops: Optional dictionary of revision properties.
1798
741
:param revision_id: Optional revision id.
742
:param lossy: Whether to discard data that can not be natively
743
represented, when pushing to a foreign VCS
1799
744
"""
1800
if self._fallback_repositories and not self._format.supports_chks:
1801
raise errors.BzrError("Cannot commit directly to a stacked branch"
1802
" in pre-2a formats. See "
1803
"https://bugs.launchpad.net/bzr/+bug/375013 for details.")
1804
result = self._commit_builder_class(self, parents, config,
1805
timestamp, timezone, committer, revprops, revision_id)
1806
self.start_write_group()
1807
return result
745
raise NotImplementedError(self.get_commit_builder)
1808
746
1809
747
@only_raises(errors.LockNotHeld, errors.LockBroken)
1810
748
def unlock(self):
1817
755
'Must end write groups before releasing write locks.')
1818
756
self.control_files.unlock()
1819
757
if self.control_files._lock_count == 0:
1820
self._inventory_entry_cache.clear()
1821
758
for repo in self._fallback_repositories:
1822
759
repo.unlock()
1823
760
1888
825
dest_repo = a_bzrdir.open_repository()
1889
826
return dest_repo
1890
827
1891
def _get_sink(self):
1892
"""Return a sink for streaming into this repository."""
1893
return StreamSink(self)
1894
1895
def _get_source(self, to_format):
1896
"""Return a source for streaming from this repository."""
1897
return StreamSource(self, to_format)
1898
1899
828
@needs_read_lock
1900
829
def has_revision(self, revision_id):
1901
830
"""True if this repository has a copy of the revision."""
1908
837
:param revision_ids: An iterable of revision_ids.
1909
838
:return: A set of the revision_ids that were present.
1910
839
"""
1911
parent_map = self.revisions.get_parent_map(
1912
[(rev_id,) for rev_id in revision_ids])
1913
result = set()
1914
if _mod_revision.NULL_REVISION in revision_ids:
1915
result.add(_mod_revision.NULL_REVISION)
1916
result.update([key[0] for key in parent_map])
1917
return result
840
raise NotImplementedError(self.has_revisions)
1918
841
1919
842
@needs_read_lock
1920
843
def get_revision(self, revision_id):
1921
844
"""Return the Revision object for a named revision."""
1922
845
return self.get_revisions([revision_id])[0]
1923
846
1924
@needs_read_lock
1925
847
def get_revision_reconcile(self, revision_id):
1926
848
"""'reconcile' helper routine that allows access to a revision always.
1927
849
1930
852
be used by reconcile, or reconcile-alike commands that are correcting
1931
853
or testing the revision graph.
1932
854
"""
1933
return self._get_revisions([revision_id])[0]
855
raise NotImplementedError(self.get_revision_reconcile)
1934
856
1935
@needs_read_lock
1936
857
def get_revisions(self, revision_ids):
1937
858
"""Get many revisions at once.
1938
859
1939
860
Repositories that need to check data on every revision read should
1940
861
subclass this method.
1941
862
"""
1942
return self._get_revisions(revision_ids)
1943
1944
@needs_read_lock
1945
def _get_revisions(self, revision_ids):
1946
"""Core work logic to get many revisions without sanity checks."""
1947
revs = {}
1948
for revid, rev in self._iter_revisions(revision_ids):
1949
if rev is None:
1950
raise errors.NoSuchRevision(self, revid)
1951
revs[revid] = rev
1952
return [revs[revid] for revid in revision_ids]
1953
1954
def _iter_revisions(self, revision_ids):
1955
"""Iterate over revision objects.
1956
1957
:param revision_ids: An iterable of revisions to examine. None may be
1958
passed to request all revisions known to the repository. Note that
1959
not all repositories can find unreferenced revisions; for those
1960
repositories only referenced ones will be returned.
1961
:return: An iterator of (revid, revision) tuples. Absent revisions (
1962
those asked for but not available) are returned as (revid, None).
1963
"""
1964
if revision_ids is None:
1965
revision_ids = self.all_revision_ids()
1966
else:
1967
for rev_id in revision_ids:
1968
if not rev_id or not isinstance(rev_id, basestring):
1969
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1970
keys = [(key,) for key in revision_ids]
1971
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1972
for record in stream:
1973
revid = record.key[0]
1974
if record.storage_kind == 'absent':
1975
yield (revid, None)
1976
else:
1977
text = record.get_bytes_as('fulltext')
1978
rev = self._serializer.read_revision_from_string(text)
1979
yield (revid, rev)
863
raise NotImplementedError(self.get_revisions)
1980
864
1981
865
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1982
866
"""Produce a generator of revision deltas.
2037
921
signature = gpg_strategy.sign(plaintext)
2038
922
self.add_signature_text(revision_id, signature)
2039
923
2040
@needs_write_lock
2041
924
def add_signature_text(self, revision_id, signature):
2042
self.signatures.add_lines((revision_id,), (),
2043
osutils.split_lines(signature))
2044
2045
def find_text_key_references(self):
2046
"""Find the text key references within the repository.
2047
2048
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2049
to whether they were referred to by the inventory of the
2050
revision_id that they contain. The inventory texts from all present
2051
revision ids are assessed to generate this report.
2052
"""
2053
revision_keys = self.revisions.keys()
2054
w = self.inventories
2055
pb = ui.ui_factory.nested_progress_bar()
2056
try:
2057
return self._find_text_key_references_from_xml_inventory_lines(
2058
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
2059
finally:
2060
pb.finished()
2061
2062
def _find_text_key_references_from_xml_inventory_lines(self,
2063
line_iterator):
2064
"""Core routine for extracting references to texts from inventories.
2065
2066
This performs the translation of xml lines to revision ids.
2067
2068
:param line_iterator: An iterator of lines, origin_version_id
2069
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2070
to whether they were referred to by the inventory of the
2071
revision_id that they contain. Note that if that revision_id was
2072
not part of the line_iterator's output then False will be given -
2073
even though it may actually refer to that key.
2074
"""
2075
if not self._serializer.support_altered_by_hack:
2076
raise AssertionError(
2077
"_find_text_key_references_from_xml_inventory_lines only "
2078
"supported for branches which store inventory as unnested xml"
2079
", not on %r" % self)
2080
result = {}
2081
2082
# this code needs to read every new line in every inventory for the
2083
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
2084
# not present in one of those inventories is unnecessary but not
2085
# harmful because we are filtering by the revision id marker in the
2086
# inventory lines : we only select file ids altered in one of those
2087
# revisions. We don't need to see all lines in the inventory because
2088
# only those added in an inventory in rev X can contain a revision=X
2089
# line.
2090
unescape_revid_cache = {}
2091
unescape_fileid_cache = {}
2092
2093
# jam 20061218 In a big fetch, this handles hundreds of thousands
2094
# of lines, so it has had a lot of inlining and optimizing done.
2095
# Sorry that it is a little bit messy.
2096
# Move several functions to be local variables, since this is a long
2097
# running loop.
2098
search = self._file_ids_altered_regex.search
2099
unescape = _unescape_xml
2100
setdefault = result.setdefault
2101
for line, line_key in line_iterator:
2102
match = search(line)
2103
if match is None:
2104
continue
2105
# One call to match.group() returning multiple items is quite a
2106
# bit faster than 2 calls to match.group() each returning 1
2107
file_id, revision_id = match.group('file_id', 'revision_id')
2108
2109
# Inlining the cache lookups helps a lot when you make 170,000
2110
# lines and 350k ids, versus 8.4 unique ids.
2111
# Using a cache helps in 2 ways:
2112
# 1) Avoids unnecessary decoding calls
2113
# 2) Re-uses cached strings, which helps in future set and
2114
# equality checks.
2115
# (2) is enough that removing encoding entirely along with
2116
# the cache (so we are using plain strings) results in no
2117
# performance improvement.
2118
try:
2119
revision_id = unescape_revid_cache[revision_id]
2120
except KeyError:
2121
unescaped = unescape(revision_id)
2122
unescape_revid_cache[revision_id] = unescaped
2123
revision_id = unescaped
2124
2125
# Note that unconditionally unescaping means that we deserialise
2126
# every fileid, which for general 'pull' is not great, but we don't
2127
# really want to have some many fulltexts that this matters anyway.
2128
# RBC 20071114.
2129
try:
2130
file_id = unescape_fileid_cache[file_id]
2131
except KeyError:
2132
unescaped = unescape(file_id)
2133
unescape_fileid_cache[file_id] = unescaped
2134
file_id = unescaped
2135
2136
key = (file_id, revision_id)
2137
setdefault(key, False)
2138
if revision_id == line_key[-1]:
2139
result[key] = True
2140
return result
2141
2142
def _inventory_xml_lines_for_keys(self, keys):
2143
"""Get a line iterator of the sort needed for findind references.
2144
2145
Not relevant for non-xml inventory repositories.
2146
2147
Ghosts in revision_keys are ignored.
2148
2149
:param revision_keys: The revision keys for the inventories to inspect.
2150
:return: An iterator over (inventory line, revid) for the fulltexts of
2151
all of the xml inventories specified by revision_keys.
2152
"""
2153
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2154
for record in stream:
2155
if record.storage_kind != 'absent':
2156
chunks = record.get_bytes_as('chunked')
2157
revid = record.key[-1]
2158
lines = osutils.chunks_to_lines(chunks)
2159
for line in lines:
2160
yield line, revid
2161
2162
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2163
revision_keys):
2164
"""Helper routine for fileids_altered_by_revision_ids.
2165
2166
This performs the translation of xml lines to revision ids.
2167
2168
:param line_iterator: An iterator of lines, origin_version_id
2169
:param revision_keys: The revision ids to filter for. This should be a
2170
set or other type which supports efficient __contains__ lookups, as
2171
the revision key from each parsed line will be looked up in the
2172
revision_keys filter.
2173
:return: a dictionary mapping altered file-ids to an iterable of
2174
revision_ids. Each altered file-ids has the exact revision_ids that
2175
altered it listed explicitly.
2176
"""
2177
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2178
line_iterator).iterkeys())
2179
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2180
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2181
self._inventory_xml_lines_for_keys(parent_keys)))
2182
new_keys = seen - parent_seen
2183
result = {}
2184
setdefault = result.setdefault
2185
for key in new_keys:
2186
setdefault(key[0], set()).add(key[-1])
2187
return result
925
"""Store a signature text for a revision.
926
927
:param revision_id: Revision id of the revision
928
:param signature: Signature text.
929
"""
930
raise NotImplementedError(self.add_signature_text)
2188
931
2189
932
def _find_parent_ids_of_revisions(self, revision_ids):
2190
933
"""Find all parent ids that are mentioned in the revision graph.
2199
942
parent_ids.discard(_mod_revision.NULL_REVISION)
2200
943
return parent_ids
2201
944
2202
def _find_parent_keys_of_revisions(self, revision_keys):
2203
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2204
2205
:param revision_keys: An iterable of revision_keys.
2206
:return: The parents of all revision_keys that are not already in
2207
revision_keys
2208
"""
2209
parent_map = self.revisions.get_parent_map(revision_keys)
2210
parent_keys = set()
2211
map(parent_keys.update, parent_map.itervalues())
2212
parent_keys.difference_update(revision_keys)
2213
parent_keys.discard(_mod_revision.NULL_REVISION)
2214
return parent_keys
2215
2216
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
945
def fileids_altered_by_revision_ids(self, revision_ids):
2217
946
"""Find the file ids and versions affected by revisions.
2218
947
2219
948
:param revisions: an iterable containing revision ids.
2220
:param _inv_weave: The inventory weave from this repository or None.
2221
If None, the inventory weave will be opened automatically.
2222
949
:return: a dictionary mapping altered file-ids to an iterable of
2223
revision_ids. Each altered file-ids has the exact revision_ids that
2224
altered it listed explicitly.
950
revision_ids. Each altered file-ids has the exact revision_ids
951
that altered it listed explicitly.
2225
952
"""
2226
selected_keys = set((revid,) for revid in revision_ids)
2227
w = _inv_weave or self.inventories
2228
return self._find_file_ids_from_xml_inventory_lines(
2229
w.iter_lines_added_or_present_in_keys(
2230
selected_keys, pb=None),
2231
selected_keys)
953
raise NotImplementedError(self.fileids_altered_by_revision_ids)
2232
954
2233
955
def iter_files_bytes(self, desired_files):
2234
956
"""Iterate through file versions.
2241
963
uniquely identify the file version in the caller's context. (Examples:
2242
964
an index number or a TreeTransform trans_id.)
2243
965
2244
bytes_iterator is an iterable of bytestrings for the file. The
2245
kind of iterable and length of the bytestrings are unspecified, but for
2246
this implementation, it is a list of bytes produced by
2247
VersionedFile.get_record_stream().
2248
2249
966
:param desired_files: a list of (file_id, revision_id, identifier)
2250
967
triples
2251
968
"""
2252
text_keys = {}
2253
for file_id, revision_id, callable_data in desired_files:
2254
text_keys[(file_id, revision_id)] = callable_data
2255
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2256
if record.storage_kind == 'absent':
2257
raise errors.RevisionNotPresent(record.key, self)
2258
yield text_keys[record.key], record.get_bytes_as('chunked')
2259
2260
def _generate_text_key_index(self, text_key_references=None,
2261
ancestors=None):
2262
"""Generate a new text key index for the repository.
2263
2264
This is an expensive function that will take considerable time to run.
2265
2266
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2267
list of parents, also text keys. When a given key has no parents,
2268
the parents list will be [NULL_REVISION].
2269
"""
2270
# All revisions, to find inventory parents.
2271
if ancestors is None:
2272
graph = self.get_graph()
2273
ancestors = graph.get_parent_map(self.all_revision_ids())
2274
if text_key_references is None:
2275
text_key_references = self.find_text_key_references()
2276
pb = ui.ui_factory.nested_progress_bar()
2277
try:
2278
return self._do_generate_text_key_index(ancestors,
2279
text_key_references, pb)
2280
finally:
2281
pb.finished()
2282
2283
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2284
"""Helper for _generate_text_key_index to avoid deep nesting."""
2285
revision_order = tsort.topo_sort(ancestors)
2286
invalid_keys = set()
2287
revision_keys = {}
2288
for revision_id in revision_order:
2289
revision_keys[revision_id] = set()
2290
text_count = len(text_key_references)
2291
# a cache of the text keys to allow reuse; costs a dict of all the
2292
# keys, but saves a 2-tuple for every child of a given key.
2293
text_key_cache = {}
2294
for text_key, valid in text_key_references.iteritems():
2295
if not valid:
2296
invalid_keys.add(text_key)
2297
else:
2298
revision_keys[text_key[1]].add(text_key)
2299
text_key_cache[text_key] = text_key
2300
del text_key_references
2301
text_index = {}
2302
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2303
NULL_REVISION = _mod_revision.NULL_REVISION
2304
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2305
# too small for large or very branchy trees. However, for 55K path
2306
# trees, it would be easy to use too much memory trivially. Ideally we
2307
# could gauge this by looking at available real memory etc, but this is
2308
# always a tricky proposition.
2309
inventory_cache = lru_cache.LRUCache(10)
2310
batch_size = 10 # should be ~150MB on a 55K path tree
2311
batch_count = len(revision_order) / batch_size + 1
2312
processed_texts = 0
2313
pb.update("Calculating text parents", processed_texts, text_count)
2314
for offset in xrange(batch_count):
2315
to_query = revision_order[offset * batch_size:(offset + 1) *
2316
batch_size]
2317
if not to_query:
2318
break
2319
for revision_id in to_query:
2320
parent_ids = ancestors[revision_id]
2321
for text_key in revision_keys[revision_id]:
2322
pb.update("Calculating text parents", processed_texts)
2323
processed_texts += 1
2324
candidate_parents = []
2325
for parent_id in parent_ids:
2326
parent_text_key = (text_key[0], parent_id)
2327
try:
2328
check_parent = parent_text_key not in \
2329
revision_keys[parent_id]
2330
except KeyError:
2331
# the parent parent_id is a ghost:
2332
check_parent = False
2333
# truncate the derived graph against this ghost.
2334
parent_text_key = None
2335
if check_parent:
2336
# look at the parent commit details inventories to
2337
# determine possible candidates in the per file graph.
2338
# TODO: cache here.
2339
try:
2340
inv = inventory_cache[parent_id]
2341
except KeyError:
2342
inv = self.revision_tree(parent_id).inventory
2343
inventory_cache[parent_id] = inv
2344
try:
2345
parent_entry = inv[text_key[0]]
2346
except (KeyError, errors.NoSuchId):
2347
parent_entry = None
2348
if parent_entry is not None:
2349
parent_text_key = (
2350
text_key[0], parent_entry.revision)
2351
else:
2352
parent_text_key = None
2353
if parent_text_key is not None:
2354
candidate_parents.append(
2355
text_key_cache[parent_text_key])
2356
parent_heads = text_graph.heads(candidate_parents)
2357
new_parents = list(parent_heads)
2358
new_parents.sort(key=lambda x:candidate_parents.index(x))
2359
if new_parents == []:
2360
new_parents = [NULL_REVISION]
2361
text_index[text_key] = new_parents
2362
2363
for text_key in invalid_keys:
2364
text_index[text_key] = [NULL_REVISION]
2365
return text_index
2366
2367
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2368
"""Get an iterable listing the keys of all the data introduced by a set
2369
of revision IDs.
2370
2371
The keys will be ordered so that the corresponding items can be safely
2372
fetched and inserted in that order.
2373
2374
:returns: An iterable producing tuples of (knit-kind, file-id,
2375
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2376
'revisions'. file-id is None unless knit-kind is 'file'.
2377
"""
2378
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2379
yield result
2380
del _files_pb
2381
for result in self._find_non_file_keys_to_fetch(revision_ids):
2382
yield result
2383
2384
def _find_file_keys_to_fetch(self, revision_ids, pb):
2385
# XXX: it's a bit weird to control the inventory weave caching in this
2386
# generator. Ideally the caching would be done in fetch.py I think. Or
2387
# maybe this generator should explicitly have the contract that it
2388
# should not be iterated until the previously yielded item has been
2389
# processed?
2390
inv_w = self.inventories
2391
2392
# file ids that changed
2393
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2394
count = 0
2395
num_file_ids = len(file_ids)
2396
for file_id, altered_versions in file_ids.iteritems():
2397
if pb is not None:
2398
pb.update("Fetch texts", count, num_file_ids)
2399
count += 1
2400
yield ("file", file_id, altered_versions)
2401
2402
def _find_non_file_keys_to_fetch(self, revision_ids):
2403
# inventory
2404
yield ("inventory", None, revision_ids)
2405
2406
# signatures
2407
# XXX: Note ATM no callers actually pay attention to this return
2408
# instead they just use the list of revision ids and ignore
2409
# missing sigs. Consider removing this work entirely
2410
revisions_with_signatures = set(self.signatures.get_parent_map(
2411
[(r,) for r in revision_ids]))
2412
revisions_with_signatures = set(
2413
[r for (r,) in revisions_with_signatures])
2414
revisions_with_signatures.intersection_update(revision_ids)
2415
yield ("signatures", None, revisions_with_signatures)
2416
2417
# revisions
2418
yield ("revisions", None, revision_ids)
2419
2420
@needs_read_lock
2421
def get_inventory(self, revision_id):
2422
"""Get Inventory object by revision id."""
2423
return self.iter_inventories([revision_id]).next()
2424
2425
def iter_inventories(self, revision_ids, ordering=None):
2426
"""Get many inventories by revision_ids.
2427
2428
This will buffer some or all of the texts used in constructing the
2429
inventories in memory, but will only parse a single inventory at a
2430
time.
2431
2432
:param revision_ids: The expected revision ids of the inventories.
2433
:param ordering: optional ordering, e.g. 'topological'. If not
2434
specified, the order of revision_ids will be preserved (by
2435
buffering if necessary).
2436
:return: An iterator of inventories.
2437
"""
2438
if ((None in revision_ids)
2439
or (_mod_revision.NULL_REVISION in revision_ids)):
2440
raise ValueError('cannot get null revision inventory')
2441
return self._iter_inventories(revision_ids, ordering)
2442
2443
def _iter_inventories(self, revision_ids, ordering):
2444
"""single-document based inventory iteration."""
2445
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2446
for text, revision_id in inv_xmls:
2447
yield self._deserialise_inventory(revision_id, text)
2448
2449
def _iter_inventory_xmls(self, revision_ids, ordering):
2450
if ordering is None:
2451
order_as_requested = True
2452
ordering = 'unordered'
2453
else:
2454
order_as_requested = False
2455
keys = [(revision_id,) for revision_id in revision_ids]
2456
if not keys:
2457
return
2458
if order_as_requested:
2459
key_iter = iter(keys)
2460
next_key = key_iter.next()
2461
stream = self.inventories.get_record_stream(keys, ordering, True)
2462
text_chunks = {}
2463
for record in stream:
2464
if record.storage_kind != 'absent':
2465
chunks = record.get_bytes_as('chunked')
2466
if order_as_requested:
2467
text_chunks[record.key] = chunks
2468
else:
2469
yield ''.join(chunks), record.key[-1]
2470
else:
2471
raise errors.NoSuchRevision(self, record.key)
2472
if order_as_requested:
2473
# Yield as many results as we can while preserving order.
2474
while next_key in text_chunks:
2475
chunks = text_chunks.pop(next_key)
2476
yield ''.join(chunks), next_key[-1]
2477
try:
2478
next_key = key_iter.next()
2479
except StopIteration:
2480
# We still want to fully consume the get_record_stream,
2481
# just in case it is not actually finished at this point
2482
next_key = None
2483
break
2484
2485
def _deserialise_inventory(self, revision_id, xml):
2486
"""Transform the xml into an inventory object.
2487
2488
:param revision_id: The expected revision id of the inventory.
2489
:param xml: A serialised inventory.
2490
"""
2491
result = self._serializer.read_inventory_from_string(xml, revision_id,
2492
entry_cache=self._inventory_entry_cache,
2493
return_from_cache=self._safe_to_return_from_cache)
2494
if result.revision_id != revision_id:
2495
raise AssertionError('revision id mismatch %s != %s' % (
2496
result.revision_id, revision_id))
2497
return result
2498
2499
def get_serializer_format(self):
2500
return self._serializer.format_num
2501
2502
@needs_read_lock
2503
def _get_inventory_xml(self, revision_id):
2504
"""Get serialized inventory as a string."""
2505
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2506
try:
2507
text, revision_id = texts.next()
2508
except StopIteration:
2509
raise errors.HistoryMissing(self, 'inventory', revision_id)
2510
return text
969
raise NotImplementedError(self.iter_files_bytes)
2511
970
2512
971
def get_rev_id_for_revno(self, revno, known_pair):
2513
972
"""Return the revision id of a revno, given a later (revno, revid)
2544
1003
raise AssertionError('_iter_for_revno returned too much history')
2545
1004
return (True, partial_history[-1])
2546
1005
1006
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
2547
1007
def iter_reverse_revision_history(self, revision_id):
2548
1008
"""Iterate backwards through revision ids in the lefthand history
2549
1009
2586
1046
2587
1047
`revision_id` may be NULL_REVISION for the empty tree revision.
2588
1048
"""
2589
revision_id = _mod_revision.ensure_null(revision_id)
2590
# TODO: refactor this to use an existing revision object
2591
# so we don't need to read it in twice.
2592
if revision_id == _mod_revision.NULL_REVISION:
2593
return RevisionTree(self, Inventory(root_id=None),
2594
_mod_revision.NULL_REVISION)
2595
else:
2596
inv = self.get_inventory(revision_id)
2597
return RevisionTree(self, inv, revision_id)
1049
raise NotImplementedError(self.revision_tree)
2598
1050
2599
1051
def revision_trees(self, revision_ids):
2600
1052
"""Return Trees for revisions in this repository.
2602
1054
:param revision_ids: a sequence of revision-ids;
2603
1055
a revision-id may not be None or 'null:'
2604
1056
"""
2605
inventories = self.iter_inventories(revision_ids)
2606
for inv in inventories:
2607
yield RevisionTree(self, inv, inv.revision_id)
2608
2609
def _filtered_revision_trees(self, revision_ids, file_ids):
2610
"""Return Tree for a revision on this branch with only some files.
2611
2612
:param revision_ids: a sequence of revision-ids;
2613
a revision-id may not be None or 'null:'
2614
:param file_ids: if not None, the result is filtered
2615
so that only those file-ids, their parents and their
2616
children are included.
2617
"""
2618
inventories = self.iter_inventories(revision_ids)
2619
for inv in inventories:
2620
# Should we introduce a FilteredRevisionTree class rather
2621
# than pre-filter the inventory here?
2622
filtered_inv = inv.filter(file_ids)
2623
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
1057
raise NotImplementedError(self.revision_trees)
2624
1058
2625
1059
@needs_read_lock
1060
@symbol_versioning.deprecated_method(
1061
symbol_versioning.deprecated_in((2, 4, 0)))
2626
1062
def get_ancestry(self, revision_id, topo_sorted=True):
2627
1063
"""Return a list of revision-ids integrated by a revision.
2628
1064
2632
1068
2633
1069
This is topologically sorted.
2634
1070
"""
1071
if 'evil' in debug.debug_flags:
1072
mutter_callsite(2, "get_ancestry is linear with history.")
2635
1073
if _mod_revision.is_null(revision_id):
2636
1074
return [None]
2637
1075
if not self.has_revision(revision_id):
2678
1116
2679
1117
def get_parent_map(self, revision_ids):
2680
1118
"""See graph.StackedParentsProvider.get_parent_map"""
1119
raise NotImplementedError(self.get_parent_map)
1120
1121
def _get_parent_map_no_fallbacks(self, revision_ids):
1122
"""Same as Repository.get_parent_map except doesn't query fallbacks."""
2681
1123
# revisions index works in keys; this just works in revisions
2682
1124
# therefore wrap and unwrap
2683
1125
query_keys = []
2689
1131
raise ValueError('get_parent_map(None) is not valid')
2690
1132
else:
2691
1133
query_keys.append((revision_id ,))
1134
vf = self.revisions.without_fallbacks()
2692
1135
for ((revision_id,), parent_keys) in \
2693
self.revisions.get_parent_map(query_keys).iteritems():
1136
vf.get_parent_map(query_keys).iteritems():
2694
1137
if parent_keys:
2695
1138
result[revision_id] = tuple([parent_revid
2696
1139
for (parent_revid,) in parent_keys])
2699
1142
return result
2700
1143
2701
1144
def _make_parents_provider(self):
2702
return self
1145
if not self._format.supports_external_lookups:
1146
return self
1147
return graph.StackedParentsProvider(_LazyListJoin(
1148
[self._make_parents_provider_unstacked()],
1149
self._fallback_repositories))
1150
1151
def _make_parents_provider_unstacked(self):
1152
return graph.CallableToParentsProviderAdapter(
1153
self._get_parent_map_no_fallbacks)
2703
1154
2704
1155
@needs_read_lock
2705
1156
def get_known_graph_ancestry(self, revision_ids):
2706
1157
"""Return the known graph for a set of revision ids and their ancestors.
2707
1158
"""
2708
st = static_tuple.StaticTuple
2709
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2710
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2711
return graph.GraphThunkIdsToKeys(known_graph)
1159
raise NotImplementedError(self.get_known_graph_ancestry)
1160
1161
def get_file_graph(self):
1162
"""Return the graph walker for files."""
1163
raise NotImplementedError(self.get_file_graph)
2712
1164
2713
1165
def get_graph(self, other_repository=None):
2714
1166
"""Return the graph walker for this repository format"""
2719
1171
[parents_provider, other_repository._make_parents_provider()])
2720
1172
return graph.Graph(parents_provider)
2721
1173
2722
def _get_versioned_file_checker(self, text_key_references=None,
2723
ancestors=None):
2724
"""Return an object suitable for checking versioned files.
2725
2726
:param text_key_references: if non-None, an already built
2727
dictionary mapping text keys ((fileid, revision_id) tuples)
2728
to whether they were referred to by the inventory of the
2729
revision_id that they contain. If None, this will be
2730
calculated.
2731
:param ancestors: Optional result from
2732
self.get_graph().get_parent_map(self.all_revision_ids()) if already
2733
available.
2734
"""
2735
return _VersionedFileChecker(self,
2736
text_key_references=text_key_references, ancestors=ancestors)
2737
2738
1174
def revision_ids_to_search_result(self, result_set):
2739
1175
"""Convert a set of revision ids to a graph SearchResult."""
2740
1176
result_parents = set()
2766
1202
2767
1203
@needs_write_lock
2768
1204
def sign_revision(self, revision_id, gpg_strategy):
2769
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1205
testament = _mod_testament.Testament.from_revision(self, revision_id)
1206
plaintext = testament.as_short_text()
2770
1207
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2771
1208
2772
1209
@needs_read_lock
1210
def verify_revision(self, revision_id, gpg_strategy):
1211
"""Verify the signature on a revision.
1212
1213
:param revision_id: the revision to verify
1214
:gpg_strategy: the GPGStrategy object to used
1215
1216
:return: gpg.SIGNATURE_VALID or a failed SIGNATURE_ value
1217
"""
1218
if not self.has_signature_for_revision_id(revision_id):
1219
return gpg.SIGNATURE_NOT_SIGNED, None
1220
signature = self.get_signature_text(revision_id)
1221
1222
testament = _mod_testament.Testament.from_revision(self, revision_id)
1223
plaintext = testament.as_short_text()
1224
1225
return gpg_strategy.verify(signature, plaintext)
1226
2773
1227
def has_signature_for_revision_id(self, revision_id):
2774
1228
"""Query for a revision signature for revision_id in the repository."""
2775
if not self.has_revision(revision_id):
2776
raise errors.NoSuchRevision(self, revision_id)
2777
sig_present = (1 == len(
2778
self.signatures.get_parent_map([(revision_id,)])))
2779
return sig_present
1229
raise NotImplementedError(self.has_signature_for_revision_id)
2780
1230
2781
@needs_read_lock
2782
1231
def get_signature_text(self, revision_id):
2783
1232
"""Return the text for a signature."""
2784
stream = self.signatures.get_record_stream([(revision_id,)],
2785
'unordered', True)
2786
record = stream.next()
2787
if record.storage_kind == 'absent':
2788
raise errors.NoSuchRevision(self, revision_id)
2789
return record.get_bytes_as('fulltext')
1233
raise NotImplementedError(self.get_signature_text)
2790
1234
2791
@needs_read_lock
2792
1235
def check(self, revision_ids=None, callback_refs=None, check_repo=True):
2793
1236
"""Check consistency of all history of given revision_ids.
2794
1237
2802
1245
:param check_repo: If False do not check the repository contents, just
2803
1246
calculate the data callback_refs requires and call them back.
2804
1247
"""
2805
return self._check(revision_ids, callback_refs=callback_refs,
1248
return self._check(revision_ids=revision_ids, callback_refs=callback_refs,
2806
1249
check_repo=check_repo)
2807
1250
2808
def _check(self, revision_ids, callback_refs, check_repo):
2809
result = check.Check(self, check_repo=check_repo)
2810
result.check(callback_refs)
2811
return result
1251
def _check(self, revision_ids=None, callback_refs=None, check_repo=True):
1252
raise NotImplementedError(self.check)
2812
1253
2813
1254
def _warn_if_deprecated(self, branch=None):
1255
if not self._format.is_deprecated():
1256
return
2814
1257
global _deprecation_warning_done
2815
1258
if _deprecation_warning_done:
2816
1259
return
2846
1289
except UnicodeDecodeError:
2847
1290
raise errors.NonAsciiRevisionId(method, self)
2848
1291
2849
def revision_graph_can_have_wrong_parents(self):
2850
"""Is it possible for this repository to have a revision graph with
2851
incorrect parents?
2852
2853
If True, then this repository must also implement
2854
_find_inconsistent_revision_parents so that check and reconcile can
2855
check for inconsistencies before proceeding with other checks that may
2856
depend on the revision index being consistent.
2857
"""
2858
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2859
2860
2861
def install_revision(repository, rev, revision_tree):
2862
"""Install all revision data into a repository."""
2863
install_revisions(repository, [(rev, revision_tree, None)])
2864
2865
2866
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2867
"""Install all revision data into a repository.
2868
2869
Accepts an iterable of revision, tree, signature tuples. The signature
2870
may be None.
2871
"""
2872
repository.start_write_group()
2873
try:
2874
inventory_cache = lru_cache.LRUCache(10)
2875
for n, (revision, revision_tree, signature) in enumerate(iterable):
2876
_install_revision(repository, revision, revision_tree, signature,
2877
inventory_cache)
2878
if pb is not None:
2879
pb.update('Transferring revisions', n + 1, num_revisions)
2880
except:
2881
repository.abort_write_group()
2882
raise
2883
else:
2884
repository.commit_write_group()
2885
2886
2887
def _install_revision(repository, rev, revision_tree, signature,
2888
inventory_cache):
2889
"""Install all revision data into a repository."""
2890
present_parents = []
2891
parent_trees = {}
2892
for p_id in rev.parent_ids:
2893
if repository.has_revision(p_id):
2894
present_parents.append(p_id)
2895
parent_trees[p_id] = repository.revision_tree(p_id)
2896
else:
2897
parent_trees[p_id] = repository.revision_tree(
2898
_mod_revision.NULL_REVISION)
2899
2900
inv = revision_tree.inventory
2901
entries = inv.iter_entries()
2902
# backwards compatibility hack: skip the root id.
2903
if not repository.supports_rich_root():
2904
path, root = entries.next()
2905
if root.revision != rev.revision_id:
2906
raise errors.IncompatibleRevision(repr(repository))
2907
text_keys = {}
2908
for path, ie in entries:
2909
text_keys[(ie.file_id, ie.revision)] = ie
2910
text_parent_map = repository.texts.get_parent_map(text_keys)
2911
missing_texts = set(text_keys) - set(text_parent_map)
2912
# Add the texts that are not already present
2913
for text_key in missing_texts:
2914
ie = text_keys[text_key]
2915
text_parents = []
2916
# FIXME: TODO: The following loop overlaps/duplicates that done by
2917
# commit to determine parents. There is a latent/real bug here where
2918
# the parents inserted are not those commit would do - in particular
2919
# they are not filtered by heads(). RBC, AB
2920
for revision, tree in parent_trees.iteritems():
2921
if ie.file_id not in tree:
2922
continue
2923
parent_id = tree.inventory[ie.file_id].revision
2924
if parent_id in text_parents:
2925
continue
2926
text_parents.append((ie.file_id, parent_id))
2927
lines = revision_tree.get_file(ie.file_id).readlines()
2928
repository.texts.add_lines(text_key, text_parents, lines)
2929
try:
2930
# install the inventory
2931
if repository._format._commit_inv_deltas and len(rev.parent_ids):
2932
# Cache this inventory
2933
inventory_cache[rev.revision_id] = inv
2934
try:
2935
basis_inv = inventory_cache[rev.parent_ids[0]]
2936
except KeyError:
2937
repository.add_inventory(rev.revision_id, inv, present_parents)
2938
else:
2939
delta = inv._make_delta(basis_inv)
2940
repository.add_inventory_by_delta(rev.parent_ids[0], delta,
2941
rev.revision_id, present_parents)
2942
else:
2943
repository.add_inventory(rev.revision_id, inv, present_parents)
2944
except errors.RevisionAlreadyPresent:
2945
pass
2946
if signature is not None:
2947
repository.add_signature_text(rev.revision_id, signature)
2948
repository.add_revision(rev.revision_id, rev, inv)
2949
2950
1292
2951
1293
class MetaDirRepository(Repository):
2952
1294
"""Repositories in the new meta-dir layout.
2987
1329
return not self._transport.has('no-working-trees')
2988
1330
2989
1331
2990
class MetaDirVersionedFileRepository(MetaDirRepository):
2991
"""Repositories in a meta-dir, that work via versioned file objects."""
1332
class RepositoryFormatRegistry(controldir.ControlComponentFormatRegistry):
1333
"""Repository format registry."""
2992
1334
2993
def __init__(self, _format, a_bzrdir, control_files):
2994
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2995
control_files)
1335
def get_default(self):
1336
"""Return the current default format."""
1337
from bzrlib import bzrdir
1338
return bzrdir.format_registry.make_bzrdir('default').repository_format
2996
1339
2997
1340
2998
1341
network_format_registry = registry.FormatRegistry()
3004
1347
"""
3005
1348
3006
1349
3007
format_registry = registry.FormatRegistry(network_format_registry)
1350
format_registry = RepositoryFormatRegistry(network_format_registry)
3008
1351
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
3009
1352
3010
1353
This can contain either format instances themselves, or classes/factories that
3015
1358
#####################################################################
3016
1359
# Repository Formats
3017
1360
3018
class RepositoryFormat(object):
1361
class RepositoryFormat(controldir.ControlComponentFormat):
3019
1362
"""A repository format.
3020
1363
3021
1364
Formats provide four things:
3056
1399
# Does this format support CHK bytestring lookups. Set to True or False in
3057
1400
# derived classes.
3058
1401
supports_chks = None
3059
# Should commit add an inventory, or an inventory delta to the repository.
3060
_commit_inv_deltas = True
3061
# What order should fetch operations request streams in?
3062
# The default is unordered as that is the cheapest for an origin to
3063
# provide.
3064
_fetch_order = 'unordered'
3065
# Does this repository format use deltas that can be fetched as-deltas ?
3066
# (E.g. knits, where the knit deltas can be transplanted intact.
3067
# We default to False, which will ensure that enough data to get
3068
# a full text out of any fetch stream will be grabbed.
3069
_fetch_uses_deltas = False
3070
1402
# Should fetch trigger a reconcile after the fetch? Only needed for
3071
1403
# some repository formats that can suffer internal inconsistencies.
3072
1404
_fetch_reconcile = False
3078
1410
# help), and for fetching when data won't have come from the same
3079
1411
# compressor.
3080
1412
pack_compresses = False
3081
# Does the repository inventory storage understand references to trees?
1413
# Does the repository storage understand references to trees?
3082
1414
supports_tree_reference = None
3083
1415
# Is the format experimental ?
3084
1416
experimental = False
3085
# Does this repository format escape funky characters, or does it create files with
3086
# similar names as the versioned files in its contents on disk ?
3087
supports_funky_characters = True
1417
# Does this repository format escape funky characters, or does it create
1418
# files with similar names as the versioned files in its contents on disk
1419
# ?
1420
supports_funky_characters = None
1421
# Does this repository format support leaving locks?
1422
supports_leaving_lock = None
1423
# Does this format support the full VersionedFiles interface?
1424
supports_full_versioned_files = None
1425
# Does this format support signing revision signatures?
1426
supports_revision_signatures = True
1427
# Can the revision graph have incorrect parents?
1428
revision_graph_can_have_wrong_parents = None
1429
# Does this format support rich root data?
1430
rich_root_data = None
1431
# Does this format support explicitly versioned directories?
1432
supports_versioned_directories = None
3088
1433
3089
1434
def __repr__(self):
3090
1435
return "%s()" % self.__class__.__name__
3115
1460
kind='repository')
3116
1461
3117
1462
@classmethod
1463
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3118
1464
def register_format(klass, format):
3119
format_registry.register(format.get_format_string(), format)
1465
format_registry.register(format)
3120
1466
3121
1467
@classmethod
1468
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3122
1469
def unregister_format(klass, format):
3123
format_registry.remove(format.get_format_string())
1470
format_registry.remove(format)
3124
1471
3125
1472
@classmethod
1473
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3126
1474
def get_default_format(klass):
3127
1475
"""Return the current default format."""
3128
from bzrlib import bzrdir
3129
return bzrdir.format_registry.make_bzrdir('default').repository_format
1476
return format_registry.get_default()
3130
1477
3131
1478
def get_format_string(self):
3132
1479
"""Return the ASCII format string that identifies this format.
3140
1487
"""Return the short description for this format."""
3141
1488
raise NotImplementedError(self.get_format_description)
3142
1489
3143
# TODO: this shouldn't be in the base class, it's specific to things that
3144
# use weaves or knits -- mbp 20070207
3145
def _get_versioned_file_store(self,
3146
name,
3147
transport,
3148
control_files,
3149
prefixed=True,
3150
versionedfile_class=None,
3151
versionedfile_kwargs={},
3152
escaped=False):
3153
if versionedfile_class is None:
3154
versionedfile_class = self._versionedfile_class
3155
weave_transport = control_files._transport.clone(name)
3156
dir_mode = control_files._dir_mode
3157
file_mode = control_files._file_mode
3158
return VersionedFileStore(weave_transport, prefixed=prefixed,
3159
dir_mode=dir_mode,
3160
file_mode=file_mode,
3161
versionedfile_class=versionedfile_class,
3162
versionedfile_kwargs=versionedfile_kwargs,
3163
escaped=escaped)
3164
3165
1490
def initialize(self, a_bzrdir, shared=False):
3166
1491
"""Initialize a repository of this format in a_bzrdir.
3167
1492
3183
1508
"""
3184
1509
return True
3185
1510
1511
def is_deprecated(self):
1512
"""Is this format deprecated?
1513
1514
Deprecated formats may trigger a user-visible warning recommending
1515
the user to upgrade. They are still fully supported.
1516
"""
1517
return False
1518
3186
1519
def network_name(self):
3187
1520
"""A simple byte string uniquely identifying this format for RPC calls.
3188
1521
3227
1560
rich_root_data = False
3228
1561
supports_tree_reference = False
3229
1562
supports_external_lookups = False
1563
supports_leaving_lock = True
3230
1564
3231
1565
@property
3232
1566
def _matchingbzrdir(self):
3270
1604
return self.get_format_string()
3271
1605
3272
1606
3273
# Pre-0.8 formats that don't have a disk format string (because they are
3274
# versioned by the matching control directory). We use the control directories
3275
# disk format string as a key for the network_name because they meet the
3276
# constraints (simple string, unique, immutable).
3277
network_format_registry.register_lazy(
3278
"Bazaar-NG branch, format 5\n",
3279
'bzrlib.repofmt.weaverepo',
3280
'RepositoryFormat5',
3281
)
3282
network_format_registry.register_lazy(
3283
"Bazaar-NG branch, format 6\n",
3284
'bzrlib.repofmt.weaverepo',
3285
'RepositoryFormat6',
3286
)
3287
3288
1607
# formats which have no format string are not discoverable or independently
3289
1608
# creatable on disk, so are not registered in format_registry. They're
3290
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1609
# all in bzrlib.repofmt.knitreponow. When an instance of one of these is
3291
1610
# needed, it's constructed directly by the BzrDir. Non-native formats where
3292
1611
# the repository is not separately opened are similar.
3293
1612
3294
1613
format_registry.register_lazy(
3295
'Bazaar-NG Repository format 7',
3296
'bzrlib.repofmt.weaverepo',
3297
'RepositoryFormat7'
3298
)
3299
3300
format_registry.register_lazy(
3301
1614
'Bazaar-NG Knit Repository Format 1',
3302
1615
'bzrlib.repofmt.knitrepo',
3303
1616
'RepositoryFormatKnit1',
3320
1633
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
3321
1634
format_registry.register_lazy(
3322
1635
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
3323
'bzrlib.repofmt.pack_repo',
1636
'bzrlib.repofmt.knitpack_repo',
3324
1637
'RepositoryFormatKnitPack1',
3325
1638
)
3326
1639
format_registry.register_lazy(
3327
1640
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
3328
'bzrlib.repofmt.pack_repo',
1641
'bzrlib.repofmt.knitpack_repo',
3329
1642
'RepositoryFormatKnitPack3',
3330
1643
)
3331
1644
format_registry.register_lazy(
3332
1645
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
3333
'bzrlib.repofmt.pack_repo',
1646
'bzrlib.repofmt.knitpack_repo',
3334
1647
'RepositoryFormatKnitPack4',
3335
1648
)
3336
1649
format_registry.register_lazy(
3337
1650
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
3338
'bzrlib.repofmt.pack_repo',
1651
'bzrlib.repofmt.knitpack_repo',
3339
1652
'RepositoryFormatKnitPack5',
3340
1653
)
3341
1654
format_registry.register_lazy(
3342
1655
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
3343
'bzrlib.repofmt.pack_repo',
1656
'bzrlib.repofmt.knitpack_repo',
3344
1657
'RepositoryFormatKnitPack5RichRoot',
3345
1658
)
3346
1659
format_registry.register_lazy(
3347
1660
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
3348
'bzrlib.repofmt.pack_repo',
1661
'bzrlib.repofmt.knitpack_repo',
3349
1662
'RepositoryFormatKnitPack5RichRootBroken',
3350
1663
)
3351
1664
format_registry.register_lazy(
3352
1665
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
3353
'bzrlib.repofmt.pack_repo',
1666
'bzrlib.repofmt.knitpack_repo',
3354
1667
'RepositoryFormatKnitPack6',
3355
1668
)
3356
1669
format_registry.register_lazy(
3357
1670
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
3358
'bzrlib.repofmt.pack_repo',
1671
'bzrlib.repofmt.knitpack_repo',
3359
1672
'RepositoryFormatKnitPack6RichRoot',
3360
1673
)
3361
1674
format_registry.register_lazy(
3369
1682
format_registry.register_lazy(
3370
1683
("Bazaar development format 2 with subtree support "
3371
1684
"(needs bzr.dev from before 1.8)\n"),
3372
'bzrlib.repofmt.pack_repo',
1685
'bzrlib.repofmt.knitpack_repo',
3373
1686
'RepositoryFormatPackDevelopment2Subtree',
3374
1687
)
3375
1688
format_registry.register_lazy(
3391
1704
InterRepository.get(other).method_name(parameters).
3392
1705
"""
3393
1706
3394
_walk_to_common_revisions_batch_size = 50
3395
1707
_optimisers = []
3396
1708
"""The available optimised InterRepository types."""
3397
1709
3412
1724
self.target.fetch(self.source, revision_id=revision_id)
3413
1725
3414
1726
@needs_write_lock
3415
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
1727
def fetch(self, revision_id=None, find_ghosts=False,
3416
1728
fetch_spec=None):
3417
1729
"""Fetch the content required to construct revision_id.
3418
1730
3420
1732
3421
1733
:param revision_id: if None all content is copied, if NULL_REVISION no
3422
1734
content is copied.
3423
:param pb: ignored.
3424
1735
:return: None.
3425
1736
"""
3426
ui.ui_factory.warn_experimental_format_fetch(self)
3427
from bzrlib.fetch import RepoFetcher
3428
# See <https://launchpad.net/bugs/456077> asking for a warning here
3429
if self.source._format.network_name() != self.target._format.network_name():
3430
ui.ui_factory.show_user_warning('cross_format_fetch',
3431
from_format=self.source._format,
3432
to_format=self.target._format)
3433
f = RepoFetcher(to_repository=self.target,
3434
from_repository=self.source,
3435
last_revision=revision_id,
3436
fetch_spec=fetch_spec,
3437
find_ghosts=find_ghosts)
3438
3439
def _walk_to_common_revisions(self, revision_ids):
3440
"""Walk out from revision_ids in source to revisions target has.
3441
3442
:param revision_ids: The start point for the search.
3443
:return: A set of revision ids.
3444
"""
3445
target_graph = self.target.get_graph()
3446
revision_ids = frozenset(revision_ids)
3447
missing_revs = set()
3448
source_graph = self.source.get_graph()
3449
# ensure we don't pay silly lookup costs.
3450
searcher = source_graph._make_breadth_first_searcher(revision_ids)
3451
null_set = frozenset([_mod_revision.NULL_REVISION])
3452
searcher_exhausted = False
3453
while True:
3454
next_revs = set()
3455
ghosts = set()
3456
# Iterate the searcher until we have enough next_revs
3457
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3458
try:
3459
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3460
next_revs.update(next_revs_part)
3461
ghosts.update(ghosts_part)
3462
except StopIteration:
3463
searcher_exhausted = True
3464
break
3465
# If there are ghosts in the source graph, and the caller asked for
3466
# them, make sure that they are present in the target.
3467
# We don't care about other ghosts as we can't fetch them and
3468
# haven't been asked to.
3469
ghosts_to_check = set(revision_ids.intersection(ghosts))
3470
revs_to_get = set(next_revs).union(ghosts_to_check)
3471
if revs_to_get:
3472
have_revs = set(target_graph.get_parent_map(revs_to_get))
3473
# we always have NULL_REVISION present.
3474
have_revs = have_revs.union(null_set)
3475
# Check if the target is missing any ghosts we need.
3476
ghosts_to_check.difference_update(have_revs)
3477
if ghosts_to_check:
3478
# One of the caller's revision_ids is a ghost in both the
3479
# source and the target.
3480
raise errors.NoSuchRevision(
3481
self.source, ghosts_to_check.pop())
3482
missing_revs.update(next_revs - have_revs)
3483
# Because we may have walked past the original stop point, make
3484
# sure everything is stopped
3485
stop_revs = searcher.find_seen_ancestors(have_revs)
3486
searcher.stop_searching_any(stop_revs)
3487
if searcher_exhausted:
3488
break
3489
return searcher.get_result()
1737
raise NotImplementedError(self.fetch)
3490
1738
3491
1739
@needs_read_lock
3492
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
1740
def search_missing_revision_ids(self,
1741
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
1742
find_ghosts=True, revision_ids=None, if_present_ids=None,
1743
limit=None):
3493
1744
"""Return the revision ids that source has that target does not.
3494
1745
3495
1746
:param revision_id: only return revision ids included by this
3496
revision_id.
1747
revision_id.
1748
:param revision_ids: return revision ids included by these
1749
revision_ids. NoSuchRevision will be raised if any of these
1750
revisions are not present.
1751
:param if_present_ids: like revision_ids, but will not cause
1752
NoSuchRevision if any of these are absent, instead they will simply
1753
not be in the result. This is useful for e.g. finding revisions
1754
to fetch for tags, which may reference absent revisions.
3497
1755
:param find_ghosts: If True find missing revisions in deep history
3498
1756
rather than just finding the surface difference.
1757
:param limit: Maximum number of revisions to return, topologically
1758
ordered
3499
1759
:return: A bzrlib.graph.SearchResult.
3500
1760
"""
3501
# stop searching at found target revisions.
3502
if not find_ghosts and revision_id is not None:
3503
return self._walk_to_common_revisions([revision_id])
3504
# generic, possibly worst case, slow code path.
3505
target_ids = set(self.target.all_revision_ids())
3506
if revision_id is not None:
3507
source_ids = self.source.get_ancestry(revision_id)
3508
if source_ids[0] is not None:
3509
raise AssertionError()
3510
source_ids.pop(0)
3511
else:
3512
source_ids = self.source.all_revision_ids()
3513
result_set = set(source_ids).difference(target_ids)
3514
return self.source.revision_ids_to_search_result(result_set)
1761
raise NotImplementedError(self.search_missing_revision_ids)
3515
1762
3516
1763
@staticmethod
3517
1764
def _same_model(source, target):
3538
1785
"different serializers")
3539
1786
3540
1787
3541
class InterSameDataRepository(InterRepository):
3542
"""Code for converting between repositories that represent the same data.
3543
3544
Data format and model must match for this to work.
3545
"""
3546
3547
@classmethod
3548
def _get_repo_format_to_test(self):
3549
"""Repository format for testing with.
3550
3551
InterSameData can pull from subtree to subtree and from non-subtree to
3552
non-subtree, so we test this with the richest repository format.
3553
"""
3554
from bzrlib.repofmt import knitrepo
3555
return knitrepo.RepositoryFormatKnit3()
3556
3557
@staticmethod
3558
def is_compatible(source, target):
3559
return InterRepository._same_model(source, target)
3560
3561
3562
class InterDifferingSerializer(InterRepository):
3563
3564
@classmethod
3565
def _get_repo_format_to_test(self):
3566
return None
3567
3568
@staticmethod
3569
def is_compatible(source, target):
3570
"""Be compatible with Knit2 source and Knit3 target"""
3571
# This is redundant with format.check_conversion_target(), however that
3572
# raises an exception, and we just want to say "False" as in we won't
3573
# support converting between these formats.
3574
if 'IDS_never' in debug.debug_flags:
3575
return False
3576
if source.supports_rich_root() and not target.supports_rich_root():
3577
return False
3578
if (source._format.supports_tree_reference
3579
and not target._format.supports_tree_reference):
3580
return False
3581
if target._fallback_repositories and target._format.supports_chks:
3582
# IDS doesn't know how to copy CHKs for the parent inventories it
3583
# adds to stacked repos.
3584
return False
3585
if 'IDS_always' in debug.debug_flags:
3586
return True
3587
# Only use this code path for local source and target. IDS does far
3588
# too much IO (both bandwidth and roundtrips) over a network.
3589
if not source.bzrdir.transport.base.startswith('file:///'):
3590
return False
3591
if not target.bzrdir.transport.base.startswith('file:///'):
3592
return False
3593
return True
3594
3595
def _get_trees(self, revision_ids, cache):
3596
possible_trees = []
3597
for rev_id in revision_ids:
3598
if rev_id in cache:
3599
possible_trees.append((rev_id, cache[rev_id]))
3600
else:
3601
# Not cached, but inventory might be present anyway.
3602
try:
3603
tree = self.source.revision_tree(rev_id)
3604
except errors.NoSuchRevision:
3605
# Nope, parent is ghost.
3606
pass
3607
else:
3608
cache[rev_id] = tree
3609
possible_trees.append((rev_id, tree))
3610
return possible_trees
3611
3612
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3613
"""Get the best delta and base for this revision.
3614
3615
:return: (basis_id, delta)
3616
"""
3617
deltas = []
3618
# Generate deltas against each tree, to find the shortest.
3619
texts_possibly_new_in_tree = set()
3620
for basis_id, basis_tree in possible_trees:
3621
delta = tree.inventory._make_delta(basis_tree.inventory)
3622
for old_path, new_path, file_id, new_entry in delta:
3623
if new_path is None:
3624
# This file_id isn't present in the new rev, so we don't
3625
# care about it.
3626
continue
3627
if not new_path:
3628
# Rich roots are handled elsewhere...
3629
continue
3630
kind = new_entry.kind
3631
if kind != 'directory' and kind != 'file':
3632
# No text record associated with this inventory entry.
3633
continue
3634
# This is a directory or file that has changed somehow.
3635
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3636
deltas.append((len(delta), basis_id, delta))
3637
deltas.sort()
3638
return deltas[0][1:]
3639
3640
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3641
"""Find all parent revisions that are absent, but for which the
3642
inventory is present, and copy those inventories.
3643
3644
This is necessary to preserve correctness when the source is stacked
3645
without fallbacks configured. (Note that in cases like upgrade the
3646
source may be not have _fallback_repositories even though it is
3647
stacked.)
3648
"""
3649
parent_revs = set()
3650
for parents in parent_map.values():
3651
parent_revs.update(parents)
3652
present_parents = self.source.get_parent_map(parent_revs)
3653
absent_parents = set(parent_revs).difference(present_parents)
3654
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3655
(rev_id,) for rev_id in absent_parents)
3656
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3657
for parent_tree in self.source.revision_trees(parent_inv_ids):
3658
current_revision_id = parent_tree.get_revision_id()
3659
parents_parents_keys = parent_invs_keys_for_stacking[
3660
(current_revision_id,)]
3661
parents_parents = [key[-1] for key in parents_parents_keys]
3662
basis_id = _mod_revision.NULL_REVISION
3663
basis_tree = self.source.revision_tree(basis_id)
3664
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3665
self.target.add_inventory_by_delta(
3666
basis_id, delta, current_revision_id, parents_parents)
3667
cache[current_revision_id] = parent_tree
3668
3669
def _fetch_batch(self, revision_ids, basis_id, cache):
3670
"""Fetch across a few revisions.
3671
3672
:param revision_ids: The revisions to copy
3673
:param basis_id: The revision_id of a tree that must be in cache, used
3674
as a basis for delta when no other base is available
3675
:param cache: A cache of RevisionTrees that we can use.
3676
:return: The revision_id of the last converted tree. The RevisionTree
3677
for it will be in cache
3678
"""
3679
# Walk though all revisions; get inventory deltas, copy referenced
3680
# texts that delta references, insert the delta, revision and
3681
# signature.
3682
root_keys_to_create = set()
3683
text_keys = set()
3684
pending_deltas = []
3685
pending_revisions = []
3686
parent_map = self.source.get_parent_map(revision_ids)
3687
self._fetch_parent_invs_for_stacking(parent_map, cache)
3688
self.source._safe_to_return_from_cache = True
3689
for tree in self.source.revision_trees(revision_ids):
3690
# Find a inventory delta for this revision.
3691
# Find text entries that need to be copied, too.
3692
current_revision_id = tree.get_revision_id()
3693
parent_ids = parent_map.get(current_revision_id, ())
3694
parent_trees = self._get_trees(parent_ids, cache)
3695
possible_trees = list(parent_trees)
3696
if len(possible_trees) == 0:
3697
# There either aren't any parents, or the parents are ghosts,
3698
# so just use the last converted tree.
3699
possible_trees.append((basis_id, cache[basis_id]))
3700
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3701
possible_trees)
3702
revision = self.source.get_revision(current_revision_id)
3703
pending_deltas.append((basis_id, delta,
3704
current_revision_id, revision.parent_ids))
3705
if self._converting_to_rich_root:
3706
self._revision_id_to_root_id[current_revision_id] = \
3707
tree.get_root_id()
3708
# Determine which texts are in present in this revision but not in
3709
# any of the available parents.
3710
texts_possibly_new_in_tree = set()
3711
for old_path, new_path, file_id, entry in delta:
3712
if new_path is None:
3713
# This file_id isn't present in the new rev
3714
continue
3715
if not new_path:
3716
# This is the root
3717
if not self.target.supports_rich_root():
3718
# The target doesn't support rich root, so we don't
3719
# copy
3720
continue
3721
if self._converting_to_rich_root:
3722
# This can't be copied normally, we have to insert
3723
# it specially
3724
root_keys_to_create.add((file_id, entry.revision))
3725
continue
3726
kind = entry.kind
3727
texts_possibly_new_in_tree.add((file_id, entry.revision))
3728
for basis_id, basis_tree in possible_trees:
3729
basis_inv = basis_tree.inventory
3730
for file_key in list(texts_possibly_new_in_tree):
3731
file_id, file_revision = file_key
3732
try:
3733
entry = basis_inv[file_id]
3734
except errors.NoSuchId:
3735
continue
3736
if entry.revision == file_revision:
3737
texts_possibly_new_in_tree.remove(file_key)
3738
text_keys.update(texts_possibly_new_in_tree)
3739
pending_revisions.append(revision)
3740
cache[current_revision_id] = tree
3741
basis_id = current_revision_id
3742
self.source._safe_to_return_from_cache = False
3743
# Copy file texts
3744
from_texts = self.source.texts
3745
to_texts = self.target.texts
3746
if root_keys_to_create:
3747
root_stream = _mod_fetch._new_root_data_stream(
3748
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3749
self.source)
3750
to_texts.insert_record_stream(root_stream)
3751
to_texts.insert_record_stream(from_texts.get_record_stream(
3752
text_keys, self.target._format._fetch_order,
3753
not self.target._format._fetch_uses_deltas))
3754
# insert inventory deltas
3755
for delta in pending_deltas:
3756
self.target.add_inventory_by_delta(*delta)
3757
if self.target._fallback_repositories:
3758
# Make sure this stacked repository has all the parent inventories
3759
# for the new revisions that we are about to insert. We do this
3760
# before adding the revisions so that no revision is added until
3761
# all the inventories it may depend on are added.
3762
# Note that this is overzealous, as we may have fetched these in an
3763
# earlier batch.
3764
parent_ids = set()
3765
revision_ids = set()
3766
for revision in pending_revisions:
3767
revision_ids.add(revision.revision_id)
3768
parent_ids.update(revision.parent_ids)
3769
parent_ids.difference_update(revision_ids)
3770
parent_ids.discard(_mod_revision.NULL_REVISION)
3771
parent_map = self.source.get_parent_map(parent_ids)
3772
# we iterate over parent_map and not parent_ids because we don't
3773
# want to try copying any revision which is a ghost
3774
for parent_tree in self.source.revision_trees(parent_map):
3775
current_revision_id = parent_tree.get_revision_id()
3776
parents_parents = parent_map[current_revision_id]
3777
possible_trees = self._get_trees(parents_parents, cache)
3778
if len(possible_trees) == 0:
3779
# There either aren't any parents, or the parents are
3780
# ghosts, so just use the last converted tree.
3781
possible_trees.append((basis_id, cache[basis_id]))
3782
basis_id, delta = self._get_delta_for_revision(parent_tree,
3783
parents_parents, possible_trees)
3784
self.target.add_inventory_by_delta(
3785
basis_id, delta, current_revision_id, parents_parents)
3786
# insert signatures and revisions
3787
for revision in pending_revisions:
3788
try:
3789
signature = self.source.get_signature_text(
3790
revision.revision_id)
3791
self.target.add_signature_text(revision.revision_id,
3792
signature)
3793
except errors.NoSuchRevision:
3794
pass
3795
self.target.add_revision(revision.revision_id, revision)
3796
return basis_id
3797
3798
def _fetch_all_revisions(self, revision_ids, pb):
3799
"""Fetch everything for the list of revisions.
3800
3801
:param revision_ids: The list of revisions to fetch. Must be in
3802
topological order.
3803
:param pb: A ProgressTask
3804
:return: None
3805
"""
3806
basis_id, basis_tree = self._get_basis(revision_ids[0])
3807
batch_size = 100
3808
cache = lru_cache.LRUCache(100)
3809
cache[basis_id] = basis_tree
3810
del basis_tree # We don't want to hang on to it here
3811
hints = []
3812
a_graph = None
3813
3814
for offset in range(0, len(revision_ids), batch_size):
3815
self.target.start_write_group()
3816
try:
3817
pb.update('Transferring revisions', offset,
3818
len(revision_ids))
3819
batch = revision_ids[offset:offset+batch_size]
3820
basis_id = self._fetch_batch(batch, basis_id, cache)
3821
except:
3822
self.source._safe_to_return_from_cache = False
3823
self.target.abort_write_group()
3824
raise
3825
else:
3826
hint = self.target.commit_write_group()
3827
if hint:
3828
hints.extend(hint)
3829
if hints and self.target._format.pack_compresses:
3830
self.target.pack(hint=hints)
3831
pb.update('Transferring revisions', len(revision_ids),
3832
len(revision_ids))
3833
3834
@needs_write_lock
3835
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3836
fetch_spec=None):
3837
"""See InterRepository.fetch()."""
3838
if fetch_spec is not None:
3839
raise AssertionError("Not implemented yet...")
3840
ui.ui_factory.warn_experimental_format_fetch(self)
3841
if (not self.source.supports_rich_root()
3842
and self.target.supports_rich_root()):
3843
self._converting_to_rich_root = True
3844
self._revision_id_to_root_id = {}
3845
else:
3846
self._converting_to_rich_root = False
3847
# See <https://launchpad.net/bugs/456077> asking for a warning here
3848
if self.source._format.network_name() != self.target._format.network_name():
3849
ui.ui_factory.show_user_warning('cross_format_fetch',
3850
from_format=self.source._format,
3851
to_format=self.target._format)
3852
revision_ids = self.target.search_missing_revision_ids(self.source,
3853
revision_id, find_ghosts=find_ghosts).get_keys()
3854
if not revision_ids:
3855
return 0, 0
3856
revision_ids = tsort.topo_sort(
3857
self.source.get_graph().get_parent_map(revision_ids))
3858
if not revision_ids:
3859
return 0, 0
3860
# Walk though all revisions; get inventory deltas, copy referenced
3861
# texts that delta references, insert the delta, revision and
3862
# signature.
3863
if pb is None:
3864
my_pb = ui.ui_factory.nested_progress_bar()
3865
pb = my_pb
3866
else:
3867
symbol_versioning.warn(
3868
symbol_versioning.deprecated_in((1, 14, 0))
3869
% "pb parameter to fetch()")
3870
my_pb = None
3871
try:
3872
self._fetch_all_revisions(revision_ids, pb)
3873
finally:
3874
if my_pb is not None:
3875
my_pb.finished()
3876
return len(revision_ids), 0
3877
3878
def _get_basis(self, first_revision_id):
3879
"""Get a revision and tree which exists in the target.
3880
3881
This assumes that first_revision_id is selected for transmission
3882
because all other ancestors are already present. If we can't find an
3883
ancestor we fall back to NULL_REVISION since we know that is safe.
3884
3885
:return: (basis_id, basis_tree)
3886
"""
3887
first_rev = self.source.get_revision(first_revision_id)
3888
try:
3889
basis_id = first_rev.parent_ids[0]
3890
# only valid as a basis if the target has it
3891
self.target.get_revision(basis_id)
3892
# Try to get a basis tree - if it's a ghost it will hit the
3893
# NoSuchRevision case.
3894
basis_tree = self.source.revision_tree(basis_id)
3895
except (IndexError, errors.NoSuchRevision):
3896
basis_id = _mod_revision.NULL_REVISION
3897
basis_tree = self.source.revision_tree(basis_id)
3898
return basis_id, basis_tree
3899
3900
3901
InterRepository.register_optimiser(InterDifferingSerializer)
3902
InterRepository.register_optimiser(InterSameDataRepository)
3903
3904
3905
1788
class CopyConverter(object):
3906
1789
"""A repository conversion tool which just performs a copy of the content.
3907
1790
3950
1833
pb.finished()
3951
1834
3952
1835
3953
_unescape_map = {
3954
'apos':"'",
3955
'quot':'"',
3956
'amp':'&',
3957
'lt':'<',
3958
'gt':'>'
3959
}
3960
3961
3962
def _unescaper(match, _map=_unescape_map):
3963
code = match.group(1)
3964
try:
3965
return _map[code]
3966
except KeyError:
3967
if not code.startswith('#'):
3968
raise
3969
return unichr(int(code[1:])).encode('utf8')
3970
3971
3972
_unescape_re = None
3973
3974
3975
def _unescape_xml(data):
3976
"""Unescape predefined XML entities in a string of data."""
3977
global _unescape_re
3978
if _unescape_re is None:
3979
_unescape_re = re.compile('\&([^;]*);')
3980
return _unescape_re.sub(_unescaper, data)
3981
3982
3983
class _VersionedFileChecker(object):
3984
3985
def __init__(self, repository, text_key_references=None, ancestors=None):
3986
self.repository = repository
3987
self.text_index = self.repository._generate_text_key_index(
3988
text_key_references=text_key_references, ancestors=ancestors)
3989
3990
def calculate_file_version_parents(self, text_key):
3991
"""Calculate the correct parents for a file version according to
3992
the inventories.
3993
"""
3994
parent_keys = self.text_index[text_key]
3995
if parent_keys == [_mod_revision.NULL_REVISION]:
3996
return ()
3997
return tuple(parent_keys)
3998
3999
def check_file_version_parents(self, texts, progress_bar=None):
4000
"""Check the parents stored in a versioned file are correct.
4001
4002
It also detects file versions that are not referenced by their
4003
corresponding revision's inventory.
4004
4005
:returns: A tuple of (wrong_parents, dangling_file_versions).
4006
wrong_parents is a dict mapping {revision_id: (stored_parents,
4007
correct_parents)} for each revision_id where the stored parents
4008
are not correct. dangling_file_versions is a set of (file_id,
4009
revision_id) tuples for versions that are present in this versioned
4010
file, but not used by the corresponding inventory.
4011
"""
4012
local_progress = None
4013
if progress_bar is None:
4014
local_progress = ui.ui_factory.nested_progress_bar()
4015
progress_bar = local_progress
4016
try:
4017
return self._check_file_version_parents(texts, progress_bar)
4018
finally:
4019
if local_progress:
4020
local_progress.finished()
4021
4022
def _check_file_version_parents(self, texts, progress_bar):
4023
"""See check_file_version_parents."""
4024
wrong_parents = {}
4025
self.file_ids = set([file_id for file_id, _ in
4026
self.text_index.iterkeys()])
4027
# text keys is now grouped by file_id
4028
n_versions = len(self.text_index)
4029
progress_bar.update('loading text store', 0, n_versions)
4030
parent_map = self.repository.texts.get_parent_map(self.text_index)
4031
# On unlistable transports this could well be empty/error...
4032
text_keys = self.repository.texts.keys()
4033
unused_keys = frozenset(text_keys) - set(self.text_index)
4034
for num, key in enumerate(self.text_index.iterkeys()):
4035
progress_bar.update('checking text graph', num, n_versions)
4036
correct_parents = self.calculate_file_version_parents(key)
4037
try:
4038
knit_parents = parent_map[key]
4039
except errors.RevisionNotPresent:
4040
# Missing text!
4041
knit_parents = None
4042
if correct_parents != knit_parents:
4043
wrong_parents[key] = (knit_parents, correct_parents)
4044
return wrong_parents, unused_keys
4045
4046
4047
def _old_get_graph(repository, revision_id):
4048
"""DO NOT USE. That is all. I'm serious."""
4049
graph = repository.get_graph()
4050
revision_graph = dict(((key, value) for key, value in
4051
graph.iter_ancestry([revision_id]) if value is not None))
4052
return _strip_NULL_ghosts(revision_graph)
4053
4054
4055
1836
def _strip_NULL_ghosts(revision_graph):
4056
1837
"""Also don't use this. more compatibility code for unmigrated clients."""
4057
1838
# Filter ghosts, and null:
4063
1844
return revision_graph
4064
1845
4065
1846
4066
class StreamSink(object):
4067
"""An object that can insert a stream into a repository.
4068
4069
This interface handles the complexity of reserialising inventories and
4070
revisions from different formats, and allows unidirectional insertion into
4071
stacked repositories without looking for the missing basis parents
4072
beforehand.
4073
"""
4074
4075
def __init__(self, target_repo):
4076
self.target_repo = target_repo
4077
4078
def insert_stream(self, stream, src_format, resume_tokens):
4079
"""Insert a stream's content into the target repository.
4080
4081
:param src_format: a bzr repository format.
4082
4083
:return: a list of resume tokens and an iterable of keys additional
4084
items required before the insertion can be completed.
4085
"""
4086
self.target_repo.lock_write()
4087
try:
4088
if resume_tokens:
4089
self.target_repo.resume_write_group(resume_tokens)
4090
is_resume = True
4091
else:
4092
self.target_repo.start_write_group()
4093
is_resume = False
4094
try:
4095
# locked_insert_stream performs a commit|suspend.
4096
missing_keys = self.insert_stream_without_locking(stream,
4097
src_format, is_resume)
4098
if missing_keys:
4099
# suspend the write group and tell the caller what we is
4100
# missing. We know we can suspend or else we would not have
4101
# entered this code path. (All repositories that can handle
4102
# missing keys can handle suspending a write group).
4103
write_group_tokens = self.target_repo.suspend_write_group()
4104
return write_group_tokens, missing_keys
4105
hint = self.target_repo.commit_write_group()
4106
to_serializer = self.target_repo._format._serializer
4107
src_serializer = src_format._serializer
4108
if (to_serializer != src_serializer and
4109
self.target_repo._format.pack_compresses):
4110
self.target_repo.pack(hint=hint)
4111
return [], set()
4112
except:
4113
self.target_repo.abort_write_group(suppress_errors=True)
4114
raise
4115
finally:
4116
self.target_repo.unlock()
4117
4118
def insert_stream_without_locking(self, stream, src_format,
4119
is_resume=False):
4120
"""Insert a stream's content into the target repository.
4121
4122
This assumes that you already have a locked repository and an active
4123
write group.
4124
4125
:param src_format: a bzr repository format.
4126
:param is_resume: Passed down to get_missing_parent_inventories to
4127
indicate if we should be checking for missing texts at the same
4128
time.
4129
4130
:return: A set of keys that are missing.
4131
"""
4132
if not self.target_repo.is_write_locked():
4133
raise errors.ObjectNotLocked(self)
4134
if not self.target_repo.is_in_write_group():
4135
raise errors.BzrError('you must already be in a write group')
4136
to_serializer = self.target_repo._format._serializer
4137
src_serializer = src_format._serializer
4138
new_pack = None
4139
if to_serializer == src_serializer:
4140
# If serializers match and the target is a pack repository, set the
4141
# write cache size on the new pack. This avoids poor performance
4142
# on transports where append is unbuffered (such as
4143
# RemoteTransport). This is safe to do because nothing should read
4144
# back from the target repository while a stream with matching
4145
# serialization is being inserted.
4146
# The exception is that a delta record from the source that should
4147
# be a fulltext may need to be expanded by the target (see
4148
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4149
# explicitly flush any buffered writes first in that rare case.
4150
try:
4151
new_pack = self.target_repo._pack_collection._new_pack
4152
except AttributeError:
4153
# Not a pack repository
4154
pass
4155
else:
4156
new_pack.set_write_cache_size(1024*1024)
4157
for substream_type, substream in stream:
4158
if 'stream' in debug.debug_flags:
4159
mutter('inserting substream: %s', substream_type)
4160
if substream_type == 'texts':
4161
self.target_repo.texts.insert_record_stream(substream)
4162
elif substream_type == 'inventories':
4163
if src_serializer == to_serializer:
4164
self.target_repo.inventories.insert_record_stream(
4165
substream)
4166
else:
4167
self._extract_and_insert_inventories(
4168
substream, src_serializer)
4169
elif substream_type == 'inventory-deltas':
4170
self._extract_and_insert_inventory_deltas(
4171
substream, src_serializer)
4172
elif substream_type == 'chk_bytes':
4173
# XXX: This doesn't support conversions, as it assumes the
4174
# conversion was done in the fetch code.
4175
self.target_repo.chk_bytes.insert_record_stream(substream)
4176
elif substream_type == 'revisions':
4177
# This may fallback to extract-and-insert more often than
4178
# required if the serializers are different only in terms of
4179
# the inventory.
4180
if src_serializer == to_serializer:
4181
self.target_repo.revisions.insert_record_stream(substream)
4182
else:
4183
self._extract_and_insert_revisions(substream,
4184
src_serializer)
4185
elif substream_type == 'signatures':
4186
self.target_repo.signatures.insert_record_stream(substream)
4187
else:
4188
raise AssertionError('kaboom! %s' % (substream_type,))
4189
# Done inserting data, and the missing_keys calculations will try to
4190
# read back from the inserted data, so flush the writes to the new pack
4191
# (if this is pack format).
4192
if new_pack is not None:
4193
new_pack._write_data('', flush=True)
4194
# Find all the new revisions (including ones from resume_tokens)
4195
missing_keys = self.target_repo.get_missing_parent_inventories(
4196
check_for_missing_texts=is_resume)
4197
try:
4198
for prefix, versioned_file in (
4199
('texts', self.target_repo.texts),
4200
('inventories', self.target_repo.inventories),
4201
('revisions', self.target_repo.revisions),
4202
('signatures', self.target_repo.signatures),
4203
('chk_bytes', self.target_repo.chk_bytes),
4204
):
4205
if versioned_file is None:
4206
continue
4207
# TODO: key is often going to be a StaticTuple object
4208
# I don't believe we can define a method by which
4209
# (prefix,) + StaticTuple will work, though we could
4210
# define a StaticTuple.sq_concat that would allow you to
4211
# pass in either a tuple or a StaticTuple as the second
4212
# object, so instead we could have:
4213
# StaticTuple(prefix) + key here...
4214
missing_keys.update((prefix,) + key for key in
4215
versioned_file.get_missing_compression_parent_keys())
4216
except NotImplementedError:
4217
# cannot even attempt suspending, and missing would have failed
4218
# during stream insertion.
4219
missing_keys = set()
4220
return missing_keys
4221
4222
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4223
target_rich_root = self.target_repo._format.rich_root_data
4224
target_tree_refs = self.target_repo._format.supports_tree_reference
4225
for record in substream:
4226
# Insert the delta directly
4227
inventory_delta_bytes = record.get_bytes_as('fulltext')
4228
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4229
try:
4230
parse_result = deserialiser.parse_text_bytes(
4231
inventory_delta_bytes)
4232
except inventory_delta.IncompatibleInventoryDelta, err:
4233
trace.mutter("Incompatible delta: %s", err.msg)
4234
raise errors.IncompatibleRevision(self.target_repo._format)
4235
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4236
revision_id = new_id
4237
parents = [key[0] for key in record.parents]
4238
self.target_repo.add_inventory_by_delta(
4239
basis_id, inv_delta, revision_id, parents)
4240
4241
def _extract_and_insert_inventories(self, substream, serializer,
4242
parse_delta=None):
4243
"""Generate a new inventory versionedfile in target, converting data.
4244
4245
The inventory is retrieved from the source, (deserializing it), and
4246
stored in the target (reserializing it in a different format).
4247
"""
4248
target_rich_root = self.target_repo._format.rich_root_data
4249
target_tree_refs = self.target_repo._format.supports_tree_reference
4250
for record in substream:
4251
# It's not a delta, so it must be a fulltext in the source
4252
# serializer's format.
4253
bytes = record.get_bytes_as('fulltext')
4254
revision_id = record.key[0]
4255
inv = serializer.read_inventory_from_string(bytes, revision_id)
4256
parents = [key[0] for key in record.parents]
4257
self.target_repo.add_inventory(revision_id, inv, parents)
4258
# No need to keep holding this full inv in memory when the rest of
4259
# the substream is likely to be all deltas.
4260
del inv
4261
4262
def _extract_and_insert_revisions(self, substream, serializer):
4263
for record in substream:
4264
bytes = record.get_bytes_as('fulltext')
4265
revision_id = record.key[0]
4266
rev = serializer.read_revision_from_string(bytes)
4267
if rev.revision_id != revision_id:
4268
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4269
self.target_repo.add_revision(revision_id, rev)
4270
4271
def finished(self):
4272
if self.target_repo._format._fetch_reconcile:
4273
self.target_repo.reconcile()
4274
4275
4276
class StreamSource(object):
4277
"""A source of a stream for fetching between repositories."""
4278
4279
def __init__(self, from_repository, to_format):
4280
"""Create a StreamSource streaming from from_repository."""
4281
self.from_repository = from_repository
4282
self.to_format = to_format
4283
self._record_counter = RecordCounter()
4284
4285
def delta_on_metadata(self):
4286
"""Return True if delta's are permitted on metadata streams.
4287
4288
That is on revisions and signatures.
4289
"""
4290
src_serializer = self.from_repository._format._serializer
4291
target_serializer = self.to_format._serializer
4292
return (self.to_format._fetch_uses_deltas and
4293
src_serializer == target_serializer)
4294
4295
def _fetch_revision_texts(self, revs):
4296
# fetch signatures first and then the revision texts
4297
# may need to be a InterRevisionStore call here.
4298
from_sf = self.from_repository.signatures
4299
# A missing signature is just skipped.
4300
keys = [(rev_id,) for rev_id in revs]
4301
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4302
keys,
4303
self.to_format._fetch_order,
4304
not self.to_format._fetch_uses_deltas))
4305
# If a revision has a delta, this is actually expanded inside the
4306
# insert_record_stream code now, which is an alternate fix for
4307
# bug #261339
4308
from_rf = self.from_repository.revisions
4309
revisions = from_rf.get_record_stream(
4310
keys,
4311
self.to_format._fetch_order,
4312
not self.delta_on_metadata())
4313
return [('signatures', signatures), ('revisions', revisions)]
4314
4315
def _generate_root_texts(self, revs):
4316
"""This will be called by get_stream between fetching weave texts and
4317
fetching the inventory weave.
4318
"""
4319
if self._rich_root_upgrade():
4320
return _mod_fetch.Inter1and2Helper(
4321
self.from_repository).generate_root_texts(revs)
4322
else:
4323
return []
4324
4325
def get_stream(self, search):
4326
phase = 'file'
4327
revs = search.get_keys()
4328
graph = self.from_repository.get_graph()
4329
revs = tsort.topo_sort(graph.get_parent_map(revs))
4330
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4331
text_keys = []
4332
for knit_kind, file_id, revisions in data_to_fetch:
4333
if knit_kind != phase:
4334
phase = knit_kind
4335
# Make a new progress bar for this phase
4336
if knit_kind == "file":
4337
# Accumulate file texts
4338
text_keys.extend([(file_id, revision) for revision in
4339
revisions])
4340
elif knit_kind == "inventory":
4341
# Now copy the file texts.
4342
from_texts = self.from_repository.texts
4343
yield ('texts', from_texts.get_record_stream(
4344
text_keys, self.to_format._fetch_order,
4345
not self.to_format._fetch_uses_deltas))
4346
# Cause an error if a text occurs after we have done the
4347
# copy.
4348
text_keys = None
4349
# Before we process the inventory we generate the root
4350
# texts (if necessary) so that the inventories references
4351
# will be valid.
4352
for _ in self._generate_root_texts(revs):
4353
yield _
4354
# we fetch only the referenced inventories because we do not
4355
# know for unselected inventories whether all their required
4356
# texts are present in the other repository - it could be
4357
# corrupt.
4358
for info in self._get_inventory_stream(revs):
4359
yield info
4360
elif knit_kind == "signatures":
4361
# Nothing to do here; this will be taken care of when
4362
# _fetch_revision_texts happens.
4363
pass
4364
elif knit_kind == "revisions":
4365
for record in self._fetch_revision_texts(revs):
4366
yield record
4367
else:
4368
raise AssertionError("Unknown knit kind %r" % knit_kind)
4369
4370
def get_stream_for_missing_keys(self, missing_keys):
4371
# missing keys can only occur when we are byte copying and not
4372
# translating (because translation means we don't send
4373
# unreconstructable deltas ever).
4374
keys = {}
4375
keys['texts'] = set()
4376
keys['revisions'] = set()
4377
keys['inventories'] = set()
4378
keys['chk_bytes'] = set()
4379
keys['signatures'] = set()
4380
for key in missing_keys:
4381
keys[key[0]].add(key[1:])
4382
if len(keys['revisions']):
4383
# If we allowed copying revisions at this point, we could end up
4384
# copying a revision without copying its required texts: a
4385
# violation of the requirements for repository integrity.
4386
raise AssertionError(
4387
'cannot copy revisions to fill in missing deltas %s' % (
4388
keys['revisions'],))
4389
for substream_kind, keys in keys.iteritems():
4390
vf = getattr(self.from_repository, substream_kind)
4391
if vf is None and keys:
4392
raise AssertionError(
4393
"cannot fill in keys for a versioned file we don't"
4394
" have: %s needs %s" % (substream_kind, keys))
4395
if not keys:
4396
# No need to stream something we don't have
4397
continue
4398
if substream_kind == 'inventories':
4399
# Some missing keys are genuinely ghosts, filter those out.
4400
present = self.from_repository.inventories.get_parent_map(keys)
4401
revs = [key[0] for key in present]
4402
# Get the inventory stream more-or-less as we do for the
4403
# original stream; there's no reason to assume that records
4404
# direct from the source will be suitable for the sink. (Think
4405
# e.g. 2a -> 1.9-rich-root).
4406
for info in self._get_inventory_stream(revs, missing=True):
4407
yield info
4408
continue
4409
4410
# Ask for full texts always so that we don't need more round trips
4411
# after this stream.
4412
# Some of the missing keys are genuinely ghosts, so filter absent
4413
# records. The Sink is responsible for doing another check to
4414
# ensure that ghosts don't introduce missing data for future
4415
# fetches.
4416
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4417
self.to_format._fetch_order, True))
4418
yield substream_kind, stream
4419
4420
def inventory_fetch_order(self):
4421
if self._rich_root_upgrade():
4422
return 'topological'
4423
else:
4424
return self.to_format._fetch_order
4425
4426
def _rich_root_upgrade(self):
4427
return (not self.from_repository._format.rich_root_data and
4428
self.to_format.rich_root_data)
4429
4430
def _get_inventory_stream(self, revision_ids, missing=False):
4431
from_format = self.from_repository._format
4432
if (from_format.supports_chks and self.to_format.supports_chks and
4433
from_format.network_name() == self.to_format.network_name()):
4434
raise AssertionError(
4435
"this case should be handled by GroupCHKStreamSource")
4436
elif 'forceinvdeltas' in debug.debug_flags:
4437
return self._get_convertable_inventory_stream(revision_ids,
4438
delta_versus_null=missing)
4439
elif from_format.network_name() == self.to_format.network_name():
4440
# Same format.
4441
return self._get_simple_inventory_stream(revision_ids,
4442
missing=missing)
4443
elif (not from_format.supports_chks and not self.to_format.supports_chks
4444
and from_format._serializer == self.to_format._serializer):
4445
# Essentially the same format.
4446
return self._get_simple_inventory_stream(revision_ids,
4447
missing=missing)
4448
else:
4449
# Any time we switch serializations, we want to use an
4450
# inventory-delta based approach.
4451
return self._get_convertable_inventory_stream(revision_ids,
4452
delta_versus_null=missing)
4453
4454
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4455
# NB: This currently reopens the inventory weave in source;
4456
# using a single stream interface instead would avoid this.
4457
from_weave = self.from_repository.inventories
4458
if missing:
4459
delta_closure = True
4460
else:
4461
delta_closure = not self.delta_on_metadata()
4462
yield ('inventories', from_weave.get_record_stream(
4463
[(rev_id,) for rev_id in revision_ids],
4464
self.inventory_fetch_order(), delta_closure))
4465
4466
def _get_convertable_inventory_stream(self, revision_ids,
4467
delta_versus_null=False):
4468
# The two formats are sufficiently different that there is no fast
4469
# path, so we need to send just inventorydeltas, which any
4470
# sufficiently modern client can insert into any repository.
4471
# The StreamSink code expects to be able to
4472
# convert on the target, so we need to put bytes-on-the-wire that can
4473
# be converted. That means inventory deltas (if the remote is <1.19,
4474
# RemoteStreamSink will fallback to VFS to insert the deltas).
4475
yield ('inventory-deltas',
4476
self._stream_invs_as_deltas(revision_ids,
4477
delta_versus_null=delta_versus_null))
4478
4479
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4480
"""Return a stream of inventory-deltas for the given rev ids.
4481
4482
:param revision_ids: The list of inventories to transmit
4483
:param delta_versus_null: Don't try to find a minimal delta for this
4484
entry, instead compute the delta versus the NULL_REVISION. This
4485
effectively streams a complete inventory. Used for stuff like
4486
filling in missing parents, etc.
4487
"""
4488
from_repo = self.from_repository
4489
revision_keys = [(rev_id,) for rev_id in revision_ids]
4490
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4491
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4492
# method...
4493
inventories = self.from_repository.iter_inventories(
4494
revision_ids, 'topological')
4495
format = from_repo._format
4496
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4497
inventory_cache = lru_cache.LRUCache(50)
4498
null_inventory = from_repo.revision_tree(
4499
_mod_revision.NULL_REVISION).inventory
4500
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4501
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4502
# repo back into a non-rich-root repo ought to be allowed)
4503
serializer = inventory_delta.InventoryDeltaSerializer(
4504
versioned_root=format.rich_root_data,
4505
tree_references=format.supports_tree_reference)
4506
for inv in inventories:
4507
key = (inv.revision_id,)
4508
parent_keys = parent_map.get(key, ())
4509
delta = None
4510
if not delta_versus_null and parent_keys:
4511
# The caller did not ask for complete inventories and we have
4512
# some parents that we can delta against. Make a delta against
4513
# each parent so that we can find the smallest.
4514
parent_ids = [parent_key[0] for parent_key in parent_keys]
4515
for parent_id in parent_ids:
4516
if parent_id not in invs_sent_so_far:
4517
# We don't know that the remote side has this basis, so
4518
# we can't use it.
4519
continue
4520
if parent_id == _mod_revision.NULL_REVISION:
4521
parent_inv = null_inventory
4522
else:
4523
parent_inv = inventory_cache.get(parent_id, None)
4524
if parent_inv is None:
4525
parent_inv = from_repo.get_inventory(parent_id)
4526
candidate_delta = inv._make_delta(parent_inv)
4527
if (delta is None or
4528
len(delta) > len(candidate_delta)):
4529
delta = candidate_delta
4530
basis_id = parent_id
4531
if delta is None:
4532
# Either none of the parents ended up being suitable, or we
4533
# were asked to delta against NULL
4534
basis_id = _mod_revision.NULL_REVISION
4535
delta = inv._make_delta(null_inventory)
4536
invs_sent_so_far.add(inv.revision_id)
4537
inventory_cache[inv.revision_id] = inv
4538
delta_serialized = ''.join(
4539
serializer.delta_to_lines(basis_id, key[-1], delta))
4540
yield versionedfile.FulltextContentFactory(
4541
key, parent_keys, None, delta_serialized)
4542
4543
4544
1847
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4545
1848
stop_revision=None):
4546
1849
"""Extend the partial history to include a given index
4556
1859
it is encountered, history extension will stop.
4557
1860
"""
4558
1861
start_revision = partial_history_cache[-1]
4559
iterator = repo.iter_reverse_revision_history(start_revision)
1862
graph = repo.get_graph()
1863
iterator = graph.iter_lefthand_ancestry(start_revision,
1864
(_mod_revision.NULL_REVISION,))
4560
1865
try:
4561
#skip the last revision in the list
1866
# skip the last revision in the list
4562
1867
iterator.next()
4563
1868
while True:
4564
1869
if (stop_index is not None and
4572
1877
# No more history
4573
1878
return
4574
1879
1880
1881
class _LazyListJoin(object):
1882
"""An iterable yielding the contents of many lists as one list.
1883
1884
Each iterator made from this will reflect the current contents of the lists
1885
at the time the iterator is made.
1886
1887
This is used by Repository's _make_parents_provider implementation so that
1888
it is safe to do::
1889
1890
pp = repo._make_parents_provider() # uses a list of fallback repos
1891
pp.add_fallback_repository(other_repo) # appends to that list
1892
result = pp.get_parent_map(...)
1893
# The result will include revs from other_repo
1894
"""
1895
1896
def __init__(self, *list_parts):
1897
self.list_parts = list_parts
1898
1899
def __iter__(self):
1900
full_list = []
1901
for list_part in self.list_parts:
1902
full_list.extend(list_part)
1903
return iter(full_list)
Loggerhead is a web-based interface for Breezy
Version: 2.0.1