« back to all changes in this revision
Viewing changes to bzrlib/repository.py
- browse files at revision 5848.2.1
- compare with another revision
- download diff
- download tarball
- view history from revision 5848.2.1
- Committer: John Arbash Meinel
- Date: 2011-05-11 11:35:28 UTC
- mto: This revision was merged to the branch mainline in revision 5851.
- Revision ID: john@arbash-meinel.com-20110511113528-qepibuwxicjrbb2h
Break compatibility with python <2.6.
This includes auditing the code for places where we were doing
explicit 'sys.version' checks and removing them as appropriate.
This includes auditing the code for places where we were doing
explicit 'sys.version' checks and removing them as appropriate.
- files added:
- NEWS
- bzrlib/plugins/changelog_merge
- bzrlib/plugins/changelog_merge/tests
- bzrlib/plugins/weave_fmt
- bzrlib/plugins/weave_fmt/tests
- bzrlib/tests/per_bzrdir
- bzrlib/tests/per_repository_vf
- files removed:
- bzrlib/deprecated_graph.py
- files renamed:
- bzrlib/repofmt/weaverepo.py => bzrlib/plugins/weave_fmt/repository.py
- bzrlib/xml4.py => bzrlib/plugins/weave_fmt/xml4.py
- bzrlib/tests/per_repository/helpers.py => bzrlib/tests/per_repository_vf/helpers.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
- NEWS => doc/en/release-notes/bzr-2.3.txt
- NEWS-template.txt => doc/en/release-notes/release-template.txt
- files modified:
- .bzrignore
added
removed
bzrlib/repository.py
1
# Copyright (C) 2005-2010 Canonical Ltd
1
# Copyright (C) 2005-2011 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
16
16
17
17
from bzrlib.lazy_import import lazy_import
18
18
lazy_import(globals(), """
19
import cStringIO
20
import re
21
19
import time
22
20
23
21
from bzrlib import (
24
22
bzrdir,
25
23
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
32
revision as _mod_revision,
43
static_tuple,
44
symbol_versioning,
45
trace,
33
testament as _mod_testament,
46
34
tsort,
47
versionedfile,
48
35
)
49
36
from bzrlib.bundle import serializer
50
from bzrlib.revisiontree import RevisionTree
51
from bzrlib.store.versioned import VersionedFileStore
52
from bzrlib.testament import Testament
53
37
""")
54
38
55
39
from bzrlib import (
56
40
errors,
57
41
registry,
42
symbol_versioning,
58
43
ui,
59
44
)
60
45
from bzrlib.decorators import needs_read_lock, needs_write_lock, only_raises
61
46
from bzrlib.inter import InterObject
62
from bzrlib.inventory import (
63
Inventory,
64
InventoryDirectory,
65
ROOT_ID,
66
entry_factory,
67
)
68
from bzrlib.recordcounter import RecordCounter
69
47
from bzrlib.lock import _RelockDebugMixin, LogicalLockResult
70
48
from bzrlib.trace import (
71
49
log_exception_quietly, note, mutter, mutter_callsite, warning)
92
70
93
71
# all clients should supply tree roots.
94
72
record_root_entry = True
95
# the default CommitBuilder does not manage trees whose root is versioned.
96
_versioned_root = False
73
# whether this commit builder supports the record_entry_contents interface
74
supports_record_entry_contents = False
97
75
98
76
def __init__(self, repository, parents, config, timestamp=None,
99
77
timezone=None, committer=None, revprops=None,
100
revision_id=None):
78
revision_id=None, lossy=False):
101
79
"""Initiate a CommitBuilder.
102
80
103
81
:param repository: Repository to commit to.
104
82
:param parents: Revision ids of the parents of the new revision.
105
:param config: Configuration to use.
106
83
:param timestamp: Optional timestamp recorded for commit.
107
84
:param timezone: Optional timezone for timestamp.
108
85
:param committer: Optional committer to set for commit.
109
86
:param revprops: Optional dictionary of revision properties.
110
87
:param revision_id: Optional revision id.
88
:param lossy: Whether to discard data that can not be natively
89
represented, when pushing to a foreign VCS
111
90
"""
112
91
self._config = config
92
self._lossy = lossy
113
93
114
94
if committer is None:
115
95
self._committer = self._config.username()
96
elif not isinstance(committer, unicode):
97
self._committer = committer.decode() # throw if non-ascii
116
98
else:
117
99
self._committer = committer
118
100
119
self.new_inventory = Inventory(None)
120
101
self._new_revision_id = revision_id
121
102
self.parents = parents
122
103
self.repository = repository
137
118
self._timezone = int(timezone)
138
119
139
120
self._generate_revision_if_needed()
140
self.__heads = graph.HeadsCache(repository.get_graph()).heads
141
self._basis_delta = []
142
# API compatibility, older code that used CommitBuilder did not call
143
# .record_delete(), which means the delta that is computed would not be
144
# valid. Callers that will call record_delete() should call
145
# .will_record_deletes() to indicate that.
146
self._recording_deletes = False
147
# memo'd check for no-op commits.
148
self._any_changes = False
149
121
150
122
def any_changes(self):
151
123
"""Return True if any entries were changed.
152
124
153
125
This includes merge-only changes. It is the core for the --unchanged
154
126
detection in commit.
155
127
156
128
:return: True if any changes have occured.
157
129
"""
158
return self._any_changes
130
raise NotImplementedError(self.any_changes)
159
131
160
132
def _validate_unicode_text(self, text, context):
161
133
"""Verify things like commit messages don't have bogus characters."""
177
149
178
150
:return: The revision id of the recorded revision.
179
151
"""
180
self._validate_unicode_text(message, 'commit message')
181
rev = _mod_revision.Revision(
182
timestamp=self._timestamp,
183
timezone=self._timezone,
184
committer=self._committer,
185
message=message,
186
inventory_sha1=self.inv_sha1,
187
revision_id=self._new_revision_id,
188
properties=self._revprops)
189
rev.parent_ids = self.parents
190
self.repository.add_revision(self._new_revision_id, rev,
191
self.new_inventory, self._config)
192
self.repository.commit_write_group()
193
return self._new_revision_id
152
raise NotImplementedError(self.commit)
194
153
195
154
def abort(self):
196
155
"""Abort the commit that is being built.
197
156
"""
198
self.repository.abort_write_group()
157
raise NotImplementedError(self.abort)
199
158
200
159
def revision_tree(self):
201
160
"""Return the tree that was just committed.
202
161
203
After calling commit() this can be called to get a RevisionTree
204
representing the newly committed tree. This is preferred to
205
calling Repository.revision_tree() because that may require
206
deserializing the inventory, while we already have a copy in
162
After calling commit() this can be called to get a
163
RevisionTree representing the newly committed tree. This is
164
preferred to calling Repository.revision_tree() because that may
165
require deserializing the inventory, while we already have a copy in
207
166
memory.
208
167
"""
209
if self.new_inventory is None:
210
self.new_inventory = self.repository.get_inventory(
211
self._new_revision_id)
212
return RevisionTree(self.repository, self.new_inventory,
213
self._new_revision_id)
168
raise NotImplementedError(self.revision_tree)
214
169
215
170
def finish_inventory(self):
216
171
"""Tell the builder that the inventory is finished.
218
173
:return: The inventory id in the repository, which can be used with
219
174
repository.get_inventory.
220
175
"""
221
if self.new_inventory is None:
222
# an inventory delta was accumulated without creating a new
223
# inventory.
224
basis_id = self.basis_delta_revision
225
# We ignore the 'inventory' returned by add_inventory_by_delta
226
# because self.new_inventory is used to hint to the rest of the
227
# system what code path was taken
228
self.inv_sha1, _ = self.repository.add_inventory_by_delta(
229
basis_id, self._basis_delta, self._new_revision_id,
230
self.parents)
231
else:
232
if self.new_inventory.root is None:
233
raise AssertionError('Root entry should be supplied to'
234
' record_entry_contents, as of bzr 0.10.')
235
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
236
self.new_inventory.revision_id = self._new_revision_id
237
self.inv_sha1 = self.repository.add_inventory(
238
self._new_revision_id,
239
self.new_inventory,
240
self.parents
241
)
242
return self._new_revision_id
176
raise NotImplementedError(self.finish_inventory)
243
177
244
178
def _gen_revision_id(self):
245
179
"""Return new revision-id."""
260
194
else:
261
195
self.random_revid = False
262
196
263
def _heads(self, file_id, revision_ids):
264
"""Calculate the graph heads for revision_ids in the graph of file_id.
265
266
This can use either a per-file graph or a global revision graph as we
267
have an identity relationship between the two graphs.
268
"""
269
return self.__heads(revision_ids)
270
271
def _check_root(self, ie, parent_invs, tree):
272
"""Helper for record_entry_contents.
273
274
:param ie: An entry being added.
275
:param parent_invs: The inventories of the parent revisions of the
276
commit.
277
:param tree: The tree that is being committed.
278
"""
279
# In this revision format, root entries have no knit or weave When
280
# serializing out to disk and back in root.revision is always
281
# _new_revision_id
282
ie.revision = self._new_revision_id
283
284
def _require_root_change(self, tree):
285
"""Enforce an appropriate root object change.
286
287
This is called once when record_iter_changes is called, if and only if
288
the root was not in the delta calculated by record_iter_changes.
289
290
:param tree: The tree which is being committed.
291
"""
292
if len(self.parents) == 0:
293
raise errors.RootMissing()
294
entry = entry_factory['directory'](tree.path2id(''), '',
295
None)
296
entry.revision = self._new_revision_id
297
self._basis_delta.append(('', '', entry.file_id, entry))
298
299
def _get_delta(self, ie, basis_inv, path):
300
"""Get a delta against the basis inventory for ie."""
301
if ie.file_id not in basis_inv:
302
# add
303
result = (None, path, ie.file_id, ie)
304
self._basis_delta.append(result)
305
return result
306
elif ie != basis_inv[ie.file_id]:
307
# common but altered
308
# TODO: avoid tis id2path call.
309
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
310
self._basis_delta.append(result)
311
return result
312
else:
313
# common, unaltered
314
return None
315
316
def get_basis_delta(self):
317
"""Return the complete inventory delta versus the basis inventory.
318
319
This has been built up with the calls to record_delete and
320
record_entry_contents. The client must have already called
321
will_record_deletes() to indicate that they will be generating a
322
complete delta.
323
324
:return: An inventory delta, suitable for use with apply_delta, or
325
Repository.add_inventory_by_delta, etc.
326
"""
327
if not self._recording_deletes:
328
raise AssertionError("recording deletes not activated.")
329
return self._basis_delta
330
331
def record_delete(self, path, file_id):
332
"""Record that a delete occured against a basis tree.
333
334
This is an optional API - when used it adds items to the basis_delta
335
being accumulated by the commit builder. It cannot be called unless the
336
method will_record_deletes() has been called to inform the builder that
337
a delta is being supplied.
338
339
:param path: The path of the thing deleted.
340
:param file_id: The file id that was deleted.
341
"""
342
if not self._recording_deletes:
343
raise AssertionError("recording deletes not activated.")
344
delta = (path, None, file_id, None)
345
self._basis_delta.append(delta)
346
self._any_changes = True
347
return delta
348
349
197
def will_record_deletes(self):
350
198
"""Tell the commit builder that deletes are being notified.
351
199
353
201
commit to be valid, deletes against the basis MUST be recorded via
354
202
builder.record_delete().
355
203
"""
356
self._recording_deletes = True
357
try:
358
basis_id = self.parents[0]
359
except IndexError:
360
basis_id = _mod_revision.NULL_REVISION
361
self.basis_delta_revision = basis_id
362
363
def record_entry_contents(self, ie, parent_invs, path, tree,
364
content_summary):
365
"""Record the content of ie from tree into the commit if needed.
366
367
Side effect: sets ie.revision when unchanged
368
369
:param ie: An inventory entry present in the commit.
370
:param parent_invs: The inventories of the parent revisions of the
371
commit.
372
:param path: The path the entry is at in the tree.
373
:param tree: The tree which contains this entry and should be used to
374
obtain content.
375
:param content_summary: Summary data from the tree about the paths
376
content - stat, length, exec, sha/link target. This is only
377
accessed when the entry has a revision of None - that is when it is
378
a candidate to commit.
379
:return: A tuple (change_delta, version_recorded, fs_hash).
380
change_delta is an inventory_delta change for this entry against
381
the basis tree of the commit, or None if no change occured against
382
the basis tree.
383
version_recorded is True if a new version of the entry has been
384
recorded. For instance, committing a merge where a file was only
385
changed on the other side will return (delta, False).
386
fs_hash is either None, or the hash details for the path (currently
387
a tuple of the contents sha1 and the statvalue returned by
388
tree.get_file_with_stat()).
389
"""
390
if self.new_inventory.root is None:
391
if ie.parent_id is not None:
392
raise errors.RootMissing()
393
self._check_root(ie, parent_invs, tree)
394
if ie.revision is None:
395
kind = content_summary[0]
396
else:
397
# ie is carried over from a prior commit
398
kind = ie.kind
399
# XXX: repository specific check for nested tree support goes here - if
400
# the repo doesn't want nested trees we skip it ?
401
if (kind == 'tree-reference' and
402
not self.repository._format.supports_tree_reference):
403
# mismatch between commit builder logic and repository:
404
# this needs the entry creation pushed down into the builder.
405
raise NotImplementedError('Missing repository subtree support.')
406
self.new_inventory.add(ie)
407
408
# TODO: slow, take it out of the inner loop.
409
try:
410
basis_inv = parent_invs[0]
411
except IndexError:
412
basis_inv = Inventory(root_id=None)
413
414
# ie.revision is always None if the InventoryEntry is considered
415
# for committing. We may record the previous parents revision if the
416
# content is actually unchanged against a sole head.
417
if ie.revision is not None:
418
if not self._versioned_root and path == '':
419
# repositories that do not version the root set the root's
420
# revision to the new commit even when no change occurs (more
421
# specifically, they do not record a revision on the root; and
422
# the rev id is assigned to the root during deserialisation -
423
# this masks when a change may have occurred against the basis.
424
# To match this we always issue a delta, because the revision
425
# of the root will always be changing.
426
if ie.file_id in basis_inv:
427
delta = (basis_inv.id2path(ie.file_id), path,
428
ie.file_id, ie)
429
else:
430
# add
431
delta = (None, path, ie.file_id, ie)
432
self._basis_delta.append(delta)
433
return delta, False, None
434
else:
435
# we don't need to commit this, because the caller already
436
# determined that an existing revision of this file is
437
# appropriate. If its not being considered for committing then
438
# it and all its parents to the root must be unaltered so
439
# no-change against the basis.
440
if ie.revision == self._new_revision_id:
441
raise AssertionError("Impossible situation, a skipped "
442
"inventory entry (%r) claims to be modified in this "
443
"commit (%r).", (ie, self._new_revision_id))
444
return None, False, None
445
# XXX: Friction: parent_candidates should return a list not a dict
446
# so that we don't have to walk the inventories again.
447
parent_candiate_entries = ie.parent_candidates(parent_invs)
448
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
449
heads = []
450
for inv in parent_invs:
451
if ie.file_id in inv:
452
old_rev = inv[ie.file_id].revision
453
if old_rev in head_set:
454
heads.append(inv[ie.file_id].revision)
455
head_set.remove(inv[ie.file_id].revision)
456
457
store = False
458
# now we check to see if we need to write a new record to the
459
# file-graph.
460
# We write a new entry unless there is one head to the ancestors, and
461
# the kind-derived content is unchanged.
462
463
# Cheapest check first: no ancestors, or more the one head in the
464
# ancestors, we write a new node.
465
if len(heads) != 1:
466
store = True
467
if not store:
468
# There is a single head, look it up for comparison
469
parent_entry = parent_candiate_entries[heads[0]]
470
# if the non-content specific data has changed, we'll be writing a
471
# node:
472
if (parent_entry.parent_id != ie.parent_id or
473
parent_entry.name != ie.name):
474
store = True
475
# now we need to do content specific checks:
476
if not store:
477
# if the kind changed the content obviously has
478
if kind != parent_entry.kind:
479
store = True
480
# Stat cache fingerprint feedback for the caller - None as we usually
481
# don't generate one.
482
fingerprint = None
483
if kind == 'file':
484
if content_summary[2] is None:
485
raise ValueError("Files must not have executable = None")
486
if not store:
487
# We can't trust a check of the file length because of content
488
# filtering...
489
if (# if the exec bit has changed we have to store:
490
parent_entry.executable != content_summary[2]):
491
store = True
492
elif parent_entry.text_sha1 == content_summary[3]:
493
# all meta and content is unchanged (using a hash cache
494
# hit to check the sha)
495
ie.revision = parent_entry.revision
496
ie.text_size = parent_entry.text_size
497
ie.text_sha1 = parent_entry.text_sha1
498
ie.executable = parent_entry.executable
499
return self._get_delta(ie, basis_inv, path), False, None
500
else:
501
# Either there is only a hash change(no hash cache entry,
502
# or same size content change), or there is no change on
503
# this file at all.
504
# Provide the parent's hash to the store layer, so that the
505
# content is unchanged we will not store a new node.
506
nostore_sha = parent_entry.text_sha1
507
if store:
508
# We want to record a new node regardless of the presence or
509
# absence of a content change in the file.
510
nostore_sha = None
511
ie.executable = content_summary[2]
512
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
513
try:
514
text = file_obj.read()
515
finally:
516
file_obj.close()
517
try:
518
ie.text_sha1, ie.text_size = self._add_text_to_weave(
519
ie.file_id, text, heads, nostore_sha)
520
# Let the caller know we generated a stat fingerprint.
521
fingerprint = (ie.text_sha1, stat_value)
522
except errors.ExistingContent:
523
# Turns out that the file content was unchanged, and we were
524
# only going to store a new node if it was changed. Carry over
525
# the entry.
526
ie.revision = parent_entry.revision
527
ie.text_size = parent_entry.text_size
528
ie.text_sha1 = parent_entry.text_sha1
529
ie.executable = parent_entry.executable
530
return self._get_delta(ie, basis_inv, path), False, None
531
elif kind == 'directory':
532
if not store:
533
# all data is meta here, nothing specific to directory, so
534
# carry over:
535
ie.revision = parent_entry.revision
536
return self._get_delta(ie, basis_inv, path), False, None
537
self._add_text_to_weave(ie.file_id, '', heads, None)
538
elif kind == 'symlink':
539
current_link_target = content_summary[3]
540
if not store:
541
# symlink target is not generic metadata, check if it has
542
# changed.
543
if current_link_target != parent_entry.symlink_target:
544
store = True
545
if not store:
546
# unchanged, carry over.
547
ie.revision = parent_entry.revision
548
ie.symlink_target = parent_entry.symlink_target
549
return self._get_delta(ie, basis_inv, path), False, None
550
ie.symlink_target = current_link_target
551
self._add_text_to_weave(ie.file_id, '', heads, None)
552
elif kind == 'tree-reference':
553
if not store:
554
if content_summary[3] != parent_entry.reference_revision:
555
store = True
556
if not store:
557
# unchanged, carry over.
558
ie.reference_revision = parent_entry.reference_revision
559
ie.revision = parent_entry.revision
560
return self._get_delta(ie, basis_inv, path), False, None
561
ie.reference_revision = content_summary[3]
562
if ie.reference_revision is None:
563
raise AssertionError("invalid content_summary for nested tree: %r"
564
% (content_summary,))
565
self._add_text_to_weave(ie.file_id, '', heads, None)
566
else:
567
raise NotImplementedError('unknown kind')
568
ie.revision = self._new_revision_id
569
self._any_changes = True
570
return self._get_delta(ie, basis_inv, path), True, fingerprint
571
572
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
573
_entry_factory=entry_factory):
204
raise NotImplementedError(self.will_record_deletes)
205
206
def record_iter_changes(self, tree, basis_revision_id, iter_changes):
574
207
"""Record a new tree via iter_changes.
575
208
576
209
:param tree: The tree to obtain text contents from for changed objects.
581
214
to basis_revision_id. The iterator must not include any items with
582
215
a current kind of None - missing items must be either filtered out
583
216
or errored-on beefore record_iter_changes sees the item.
584
:param _entry_factory: Private method to bind entry_factory locally for
585
performance.
586
217
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
587
218
tree._observed_sha1.
588
219
"""
589
# Create an inventory delta based on deltas between all the parents and
590
# deltas between all the parent inventories. We use inventory delta's
591
# between the inventory objects because iter_changes masks
592
# last-changed-field only changes.
593
# Working data:
594
# file_id -> change map, change is fileid, paths, changed, versioneds,
595
# parents, names, kinds, executables
596
merged_ids = {}
597
# {file_id -> revision_id -> inventory entry, for entries in parent
598
# trees that are not parents[0]
599
parent_entries = {}
600
ghost_basis = False
601
try:
602
revtrees = list(self.repository.revision_trees(self.parents))
603
except errors.NoSuchRevision:
604
# one or more ghosts, slow path.
605
revtrees = []
606
for revision_id in self.parents:
607
try:
608
revtrees.append(self.repository.revision_tree(revision_id))
609
except errors.NoSuchRevision:
610
if not revtrees:
611
basis_revision_id = _mod_revision.NULL_REVISION
612
ghost_basis = True
613
revtrees.append(self.repository.revision_tree(
614
_mod_revision.NULL_REVISION))
615
# The basis inventory from a repository
616
if revtrees:
617
basis_inv = revtrees[0].inventory
618
else:
619
basis_inv = self.repository.revision_tree(
620
_mod_revision.NULL_REVISION).inventory
621
if len(self.parents) > 0:
622
if basis_revision_id != self.parents[0] and not ghost_basis:
623
raise Exception(
624
"arbitrary basis parents not yet supported with merges")
625
for revtree in revtrees[1:]:
626
for change in revtree.inventory._make_delta(basis_inv):
627
if change[1] is None:
628
# Not present in this parent.
629
continue
630
if change[2] not in merged_ids:
631
if change[0] is not None:
632
basis_entry = basis_inv[change[2]]
633
merged_ids[change[2]] = [
634
# basis revid
635
basis_entry.revision,
636
# new tree revid
637
change[3].revision]
638
parent_entries[change[2]] = {
639
# basis parent
640
basis_entry.revision:basis_entry,
641
# this parent
642
change[3].revision:change[3],
643
}
644
else:
645
merged_ids[change[2]] = [change[3].revision]
646
parent_entries[change[2]] = {change[3].revision:change[3]}
647
else:
648
merged_ids[change[2]].append(change[3].revision)
649
parent_entries[change[2]][change[3].revision] = change[3]
650
else:
651
merged_ids = {}
652
# Setup the changes from the tree:
653
# changes maps file_id -> (change, [parent revision_ids])
654
changes= {}
655
for change in iter_changes:
656
# This probably looks up in basis_inv way to much.
657
if change[1][0] is not None:
658
head_candidate = [basis_inv[change[0]].revision]
659
else:
660
head_candidate = []
661
changes[change[0]] = change, merged_ids.get(change[0],
662
head_candidate)
663
unchanged_merged = set(merged_ids) - set(changes)
664
# Extend the changes dict with synthetic changes to record merges of
665
# texts.
666
for file_id in unchanged_merged:
667
# Record a merged version of these items that did not change vs the
668
# basis. This can be either identical parallel changes, or a revert
669
# of a specific file after a merge. The recorded content will be
670
# that of the current tree (which is the same as the basis), but
671
# the per-file graph will reflect a merge.
672
# NB:XXX: We are reconstructing path information we had, this
673
# should be preserved instead.
674
# inv delta change: (file_id, (path_in_source, path_in_target),
675
# changed_content, versioned, parent, name, kind,
676
# executable)
677
try:
678
basis_entry = basis_inv[file_id]
679
except errors.NoSuchId:
680
# a change from basis->some_parents but file_id isn't in basis
681
# so was new in the merge, which means it must have changed
682
# from basis -> current, and as it hasn't the add was reverted
683
# by the user. So we discard this change.
684
pass
685
else:
686
change = (file_id,
687
(basis_inv.id2path(file_id), tree.id2path(file_id)),
688
False, (True, True),
689
(basis_entry.parent_id, basis_entry.parent_id),
690
(basis_entry.name, basis_entry.name),
691
(basis_entry.kind, basis_entry.kind),
692
(basis_entry.executable, basis_entry.executable))
693
changes[file_id] = (change, merged_ids[file_id])
694
# changes contains tuples with the change and a set of inventory
695
# candidates for the file.
696
# inv delta is:
697
# old_path, new_path, file_id, new_inventory_entry
698
seen_root = False # Is the root in the basis delta?
699
inv_delta = self._basis_delta
700
modified_rev = self._new_revision_id
701
for change, head_candidates in changes.values():
702
if change[3][1]: # versioned in target.
703
# Several things may be happening here:
704
# We may have a fork in the per-file graph
705
# - record a change with the content from tree
706
# We may have a change against < all trees
707
# - carry over the tree that hasn't changed
708
# We may have a change against all trees
709
# - record the change with the content from tree
710
kind = change[6][1]
711
file_id = change[0]
712
entry = _entry_factory[kind](file_id, change[5][1],
713
change[4][1])
714
head_set = self._heads(change[0], set(head_candidates))
715
heads = []
716
# Preserve ordering.
717
for head_candidate in head_candidates:
718
if head_candidate in head_set:
719
heads.append(head_candidate)
720
head_set.remove(head_candidate)
721
carried_over = False
722
if len(heads) == 1:
723
# Could be a carry-over situation:
724
parent_entry_revs = parent_entries.get(file_id, None)
725
if parent_entry_revs:
726
parent_entry = parent_entry_revs.get(heads[0], None)
727
else:
728
parent_entry = None
729
if parent_entry is None:
730
# The parent iter_changes was called against is the one
731
# that is the per-file head, so any change is relevant
732
# iter_changes is valid.
733
carry_over_possible = False
734
else:
735
# could be a carry over situation
736
# A change against the basis may just indicate a merge,
737
# we need to check the content against the source of the
738
# merge to determine if it was changed after the merge
739
# or carried over.
740
if (parent_entry.kind != entry.kind or
741
parent_entry.parent_id != entry.parent_id or
742
parent_entry.name != entry.name):
743
# Metadata common to all entries has changed
744
# against per-file parent
745
carry_over_possible = False
746
else:
747
carry_over_possible = True
748
# per-type checks for changes against the parent_entry
749
# are done below.
750
else:
751
# Cannot be a carry-over situation
752
carry_over_possible = False
753
# Populate the entry in the delta
754
if kind == 'file':
755
# XXX: There is still a small race here: If someone reverts the content of a file
756
# after iter_changes examines and decides it has changed,
757
# we will unconditionally record a new version even if some
758
# other process reverts it while commit is running (with
759
# the revert happening after iter_changes did it's
760
# examination).
761
if change[7][1]:
762
entry.executable = True
763
else:
764
entry.executable = False
765
if (carry_over_possible and
766
parent_entry.executable == entry.executable):
767
# Check the file length, content hash after reading
768
# the file.
769
nostore_sha = parent_entry.text_sha1
770
else:
771
nostore_sha = None
772
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
773
try:
774
text = file_obj.read()
775
finally:
776
file_obj.close()
777
try:
778
entry.text_sha1, entry.text_size = self._add_text_to_weave(
779
file_id, text, heads, nostore_sha)
780
yield file_id, change[1][1], (entry.text_sha1, stat_value)
781
except errors.ExistingContent:
782
# No content change against a carry_over parent
783
# Perhaps this should also yield a fs hash update?
784
carried_over = True
785
entry.text_size = parent_entry.text_size
786
entry.text_sha1 = parent_entry.text_sha1
787
elif kind == 'symlink':
788
# Wants a path hint?
789
entry.symlink_target = tree.get_symlink_target(file_id)
790
if (carry_over_possible and
791
parent_entry.symlink_target == entry.symlink_target):
792
carried_over = True
793
else:
794
self._add_text_to_weave(change[0], '', heads, None)
795
elif kind == 'directory':
796
if carry_over_possible:
797
carried_over = True
798
else:
799
# Nothing to set on the entry.
800
# XXX: split into the Root and nonRoot versions.
801
if change[1][1] != '' or self.repository.supports_rich_root():
802
self._add_text_to_weave(change[0], '', heads, None)
803
elif kind == 'tree-reference':
804
if not self.repository._format.supports_tree_reference:
805
# This isn't quite sane as an error, but we shouldn't
806
# ever see this code path in practice: tree's don't
807
# permit references when the repo doesn't support tree
808
# references.
809
raise errors.UnsupportedOperation(tree.add_reference,
810
self.repository)
811
reference_revision = tree.get_reference_revision(change[0])
812
entry.reference_revision = reference_revision
813
if (carry_over_possible and
814
parent_entry.reference_revision == reference_revision):
815
carried_over = True
816
else:
817
self._add_text_to_weave(change[0], '', heads, None)
818
else:
819
raise AssertionError('unknown kind %r' % kind)
820
if not carried_over:
821
entry.revision = modified_rev
822
else:
823
entry.revision = parent_entry.revision
824
else:
825
entry = None
826
new_path = change[1][1]
827
inv_delta.append((change[1][0], new_path, change[0], entry))
828
if new_path == '':
829
seen_root = True
830
self.new_inventory = None
831
if len(inv_delta):
832
# This should perhaps be guarded by a check that the basis we
833
# commit against is the basis for the commit and if not do a delta
834
# against the basis.
835
self._any_changes = True
836
if not seen_root:
837
# housekeeping root entry changes do not affect no-change commits.
838
self._require_root_change(tree)
839
self.basis_delta_revision = basis_revision_id
840
841
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
842
parent_keys = tuple([(file_id, parent) for parent in parents])
843
return self.repository.texts._add_text(
844
(file_id, self._new_revision_id), parent_keys, new_text,
845
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
846
847
848
class RootCommitBuilder(CommitBuilder):
849
"""This commitbuilder actually records the root id"""
850
851
# the root entry gets versioned properly by this builder.
852
_versioned_root = True
853
854
def _check_root(self, ie, parent_invs, tree):
855
"""Helper for record_entry_contents.
856
857
:param ie: An entry being added.
858
:param parent_invs: The inventories of the parent revisions of the
859
commit.
860
:param tree: The tree that is being committed.
861
"""
862
863
def _require_root_change(self, tree):
864
"""Enforce an appropriate root object change.
865
866
This is called once when record_iter_changes is called, if and only if
867
the root was not in the delta calculated by record_iter_changes.
868
869
:param tree: The tree which is being committed.
870
"""
871
# versioned roots do not change unless the tree found a change.
220
raise NotImplementedError(self.record_iter_changes)
872
221
873
222
874
223
class RepositoryWriteLockResult(LogicalLockResult):
899
248
revisions and file history. It's normally accessed only by the Branch,
900
249
which views a particular line of development through that history.
901
250
902
The Repository builds on top of some byte storage facilies (the revisions,
903
signatures, inventories, texts and chk_bytes attributes) and a Transport,
904
which respectively provide byte storage and a means to access the (possibly
905
remote) disk.
906
907
The byte storage facilities are addressed via tuples, which we refer to
908
as 'keys' throughout the code base. Revision_keys, inventory_keys and
909
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
910
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
911
byte string made up of a hash identifier and a hash value.
912
We use this interface because it allows low friction with the underlying
913
code that implements disk indices, network encoding and other parts of
914
bzrlib.
915
916
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
917
the serialised revisions for the repository. This can be used to obtain
918
revision graph information or to access raw serialised revisions.
919
The result of trying to insert data into the repository via this store
920
is undefined: it should be considered read-only except for implementors
921
of repositories.
922
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
923
the serialised signatures for the repository. This can be used to
924
obtain access to raw serialised signatures. The result of trying to
925
insert data into the repository via this store is undefined: it should
926
be considered read-only except for implementors of repositories.
927
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
928
the serialised inventories for the repository. This can be used to
929
obtain unserialised inventories. The result of trying to insert data
930
into the repository via this store is undefined: it should be
931
considered read-only except for implementors of repositories.
932
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
933
texts of files and directories for the repository. This can be used to
934
obtain file texts or file graphs. Note that Repository.iter_file_bytes
935
is usually a better interface for accessing file texts.
936
The result of trying to insert data into the repository via this store
937
is undefined: it should be considered read-only except for implementors
938
of repositories.
939
:ivar chk_bytes: A bzrlib.versionedfile.VersionedFiles instance containing
940
any data the repository chooses to store or have indexed by its hash.
941
The result of trying to insert data into the repository via this store
942
is undefined: it should be considered read-only except for implementors
943
of repositories.
944
:ivar _transport: Transport for file access to repository, typically
945
pointing to .bzr/repository.
251
See VersionedFileRepository in bzrlib.vf_repository for the
252
base class for most Bazaar repositories.
946
253
"""
947
254
948
# What class to use for a CommitBuilder. Often its simpler to change this
949
# in a Repository class subclass rather than to override
950
# get_commit_builder.
951
_commit_builder_class = CommitBuilder
952
# The search regex used by xml based repositories to determine what things
953
# where changed in a single commit.
954
_file_ids_altered_regex = lazy_regex.lazy_compile(
955
r'file_id="(?P<file_id>[^"]+)"'
956
r'.* revision="(?P<revision_id>[^"]+)"'
957
)
958
959
255
def abort_write_group(self, suppress_errors=False):
960
256
"""Commit the contents accrued within the current write group.
961
257
1004
300
1005
301
:param repository: A repository.
1006
302
"""
1007
if not self._format.supports_external_lookups:
1008
raise errors.UnstackableRepositoryFormat(self._format, self.base)
1009
if self.is_locked():
1010
# This repository will call fallback.unlock() when we transition to
1011
# the unlocked state, so we make sure to increment the lock count
1012
repository.lock_read()
1013
self._check_fallback_repository(repository)
1014
self._fallback_repositories.append(repository)
1015
self.texts.add_fallback_versioned_files(repository.texts)
1016
self.inventories.add_fallback_versioned_files(repository.inventories)
1017
self.revisions.add_fallback_versioned_files(repository.revisions)
1018
self.signatures.add_fallback_versioned_files(repository.signatures)
1019
if self.chk_bytes is not None:
1020
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
303
raise NotImplementedError(self.add_fallback_repository)
1021
304
1022
305
def _check_fallback_repository(self, repository):
1023
306
"""Check that this repository can fallback to repository safely.
1028
311
"""
1029
312
return InterRepository._assert_same_model(self, repository)
1030
313
1031
def add_inventory(self, revision_id, inv, parents):
1032
"""Add the inventory inv to the repository as revision_id.
1033
1034
:param parents: The revision ids of the parents that revision_id
1035
is known to have and are in the repository already.
1036
1037
:returns: The validator(which is a sha1 digest, though what is sha'd is
1038
repository format specific) of the serialized inventory.
1039
"""
1040
if not self.is_in_write_group():
1041
raise AssertionError("%r not in write group" % (self,))
1042
_mod_revision.check_not_reserved_id(revision_id)
1043
if not (inv.revision_id is None or inv.revision_id == revision_id):
1044
raise AssertionError(
1045
"Mismatch between inventory revision"
1046
" id and insertion revid (%r, %r)"
1047
% (inv.revision_id, revision_id))
1048
if inv.root is None:
1049
raise errors.RootMissing()
1050
return self._add_inventory_checked(revision_id, inv, parents)
1051
1052
def _add_inventory_checked(self, revision_id, inv, parents):
1053
"""Add inv to the repository after checking the inputs.
1054
1055
This function can be overridden to allow different inventory styles.
1056
1057
:seealso: add_inventory, for the contract.
1058
"""
1059
inv_lines = self._serializer.write_inventory_to_lines(inv)
1060
return self._inventory_add_lines(revision_id, parents,
1061
inv_lines, check_content=False)
1062
1063
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1064
parents, basis_inv=None, propagate_caches=False):
1065
"""Add a new inventory expressed as a delta against another revision.
1066
1067
See the inventory developers documentation for the theory behind
1068
inventory deltas.
1069
1070
:param basis_revision_id: The inventory id the delta was created
1071
against. (This does not have to be a direct parent.)
1072
:param delta: The inventory delta (see Inventory.apply_delta for
1073
details).
1074
:param new_revision_id: The revision id that the inventory is being
1075
added for.
1076
:param parents: The revision ids of the parents that revision_id is
1077
known to have and are in the repository already. These are supplied
1078
for repositories that depend on the inventory graph for revision
1079
graph access, as well as for those that pun ancestry with delta
1080
compression.
1081
:param basis_inv: The basis inventory if it is already known,
1082
otherwise None.
1083
:param propagate_caches: If True, the caches for this inventory are
1084
copied to and updated for the result if possible.
1085
1086
:returns: (validator, new_inv)
1087
The validator(which is a sha1 digest, though what is sha'd is
1088
repository format specific) of the serialized inventory, and the
1089
resulting inventory.
1090
"""
1091
if not self.is_in_write_group():
1092
raise AssertionError("%r not in write group" % (self,))
1093
_mod_revision.check_not_reserved_id(new_revision_id)
1094
basis_tree = self.revision_tree(basis_revision_id)
1095
basis_tree.lock_read()
1096
try:
1097
# Note that this mutates the inventory of basis_tree, which not all
1098
# inventory implementations may support: A better idiom would be to
1099
# return a new inventory, but as there is no revision tree cache in
1100
# repository this is safe for now - RBC 20081013
1101
if basis_inv is None:
1102
basis_inv = basis_tree.inventory
1103
basis_inv.apply_delta(delta)
1104
basis_inv.revision_id = new_revision_id
1105
return (self.add_inventory(new_revision_id, basis_inv, parents),
1106
basis_inv)
1107
finally:
1108
basis_tree.unlock()
1109
1110
def _inventory_add_lines(self, revision_id, parents, lines,
1111
check_content=True):
1112
"""Store lines in inv_vf and return the sha1 of the inventory."""
1113
parents = [(parent,) for parent in parents]
1114
result = self.inventories.add_lines((revision_id,), parents, lines,
1115
check_content=check_content)[0]
1116
self.inventories._access.flush()
1117
return result
1118
1119
def add_revision(self, revision_id, rev, inv=None, config=None):
1120
"""Add rev to the revision store as revision_id.
1121
1122
:param revision_id: the revision id to use.
1123
:param rev: The revision object.
1124
:param inv: The inventory for the revision. if None, it will be looked
1125
up in the inventory storer
1126
:param config: If None no digital signature will be created.
1127
If supplied its signature_needed method will be used
1128
to determine if a signature should be made.
1129
"""
1130
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
1131
# rev.parent_ids?
1132
_mod_revision.check_not_reserved_id(revision_id)
1133
if config is not None and config.signature_needed():
1134
if inv is None:
1135
inv = self.get_inventory(revision_id)
1136
plaintext = Testament(rev, inv).as_short_text()
1137
self.store_revision_signature(
1138
gpg.GPGStrategy(config), plaintext, revision_id)
1139
# check inventory present
1140
if not self.inventories.get_parent_map([(revision_id,)]):
1141
if inv is None:
1142
raise errors.WeaveRevisionNotPresent(revision_id,
1143
self.inventories)
1144
else:
1145
# yes, this is not suitable for adding with ghosts.
1146
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
1147
rev.parent_ids)
1148
else:
1149
key = (revision_id,)
1150
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
1151
self._add_revision(rev)
1152
1153
def _add_revision(self, revision):
1154
text = self._serializer.write_revision_to_string(revision)
1155
key = (revision.revision_id,)
1156
parents = tuple((parent,) for parent in revision.parent_ids)
1157
self.revisions.add_lines(key, parents, osutils.split_lines(text))
1158
1159
314
def all_revision_ids(self):
1160
315
"""Returns a list of all the revision ids in the repository.
1161
316
1196
351
# The old API returned a list, should this actually be a set?
1197
352
return parent_map.keys()
1198
353
1199
def _check_inventories(self, checker):
1200
"""Check the inventories found from the revision scan.
1201
1202
This is responsible for verifying the sha1 of inventories and
1203
creating a pending_keys set that covers data referenced by inventories.
1204
"""
1205
bar = ui.ui_factory.nested_progress_bar()
1206
try:
1207
self._do_check_inventories(checker, bar)
1208
finally:
1209
bar.finished()
1210
1211
def _do_check_inventories(self, checker, bar):
1212
"""Helper for _check_inventories."""
1213
revno = 0
1214
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1215
kinds = ['chk_bytes', 'texts']
1216
count = len(checker.pending_keys)
1217
bar.update("inventories", 0, 2)
1218
current_keys = checker.pending_keys
1219
checker.pending_keys = {}
1220
# Accumulate current checks.
1221
for key in current_keys:
1222
if key[0] != 'inventories' and key[0] not in kinds:
1223
checker._report_items.append('unknown key type %r' % (key,))
1224
keys[key[0]].add(key[1:])
1225
if keys['inventories']:
1226
# NB: output order *should* be roughly sorted - topo or
1227
# inverse topo depending on repository - either way decent
1228
# to just delta against. However, pre-CHK formats didn't
1229
# try to optimise inventory layout on disk. As such the
1230
# pre-CHK code path does not use inventory deltas.
1231
last_object = None
1232
for record in self.inventories.check(keys=keys['inventories']):
1233
if record.storage_kind == 'absent':
1234
checker._report_items.append(
1235
'Missing inventory {%s}' % (record.key,))
1236
else:
1237
last_object = self._check_record('inventories', record,
1238
checker, last_object,
1239
current_keys[('inventories',) + record.key])
1240
del keys['inventories']
1241
else:
1242
return
1243
bar.update("texts", 1)
1244
while (checker.pending_keys or keys['chk_bytes']
1245
or keys['texts']):
1246
# Something to check.
1247
current_keys = checker.pending_keys
1248
checker.pending_keys = {}
1249
# Accumulate current checks.
1250
for key in current_keys:
1251
if key[0] not in kinds:
1252
checker._report_items.append('unknown key type %r' % (key,))
1253
keys[key[0]].add(key[1:])
1254
# Check the outermost kind only - inventories || chk_bytes || texts
1255
for kind in kinds:
1256
if keys[kind]:
1257
last_object = None
1258
for record in getattr(self, kind).check(keys=keys[kind]):
1259
if record.storage_kind == 'absent':
1260
checker._report_items.append(
1261
'Missing %s {%s}' % (kind, record.key,))
1262
else:
1263
last_object = self._check_record(kind, record,
1264
checker, last_object, current_keys[(kind,) + record.key])
1265
keys[kind] = set()
1266
break
1267
1268
def _check_record(self, kind, record, checker, last_object, item_data):
1269
"""Check a single text from this repository."""
1270
if kind == 'inventories':
1271
rev_id = record.key[0]
1272
inv = self._deserialise_inventory(rev_id,
1273
record.get_bytes_as('fulltext'))
1274
if last_object is not None:
1275
delta = inv._make_delta(last_object)
1276
for old_path, path, file_id, ie in delta:
1277
if ie is None:
1278
continue
1279
ie.check(checker, rev_id, inv)
1280
else:
1281
for path, ie in inv.iter_entries():
1282
ie.check(checker, rev_id, inv)
1283
if self._format.fast_deltas:
1284
return inv
1285
elif kind == 'chk_bytes':
1286
# No code written to check chk_bytes for this repo format.
1287
checker._report_items.append(
1288
'unsupported key type chk_bytes for %s' % (record.key,))
1289
elif kind == 'texts':
1290
self._check_text(record, checker, item_data)
1291
else:
1292
checker._report_items.append(
1293
'unknown key type %s for %s' % (kind, record.key))
1294
1295
def _check_text(self, record, checker, item_data):
1296
"""Check a single text."""
1297
# Check it is extractable.
1298
# TODO: check length.
1299
if record.storage_kind == 'chunked':
1300
chunks = record.get_bytes_as(record.storage_kind)
1301
sha1 = osutils.sha_strings(chunks)
1302
length = sum(map(len, chunks))
1303
else:
1304
content = record.get_bytes_as('fulltext')
1305
sha1 = osutils.sha_string(content)
1306
length = len(content)
1307
if item_data and sha1 != item_data[1]:
1308
checker._report_items.append(
1309
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1310
(record.key, sha1, item_data[1], item_data[2]))
1311
1312
354
@staticmethod
1313
355
def create(a_bzrdir):
1314
356
"""Construct the current default format repository in a_bzrdir."""
1319
361
1320
362
:param _format: The format of the repository on disk.
1321
363
:param a_bzrdir: The BzrDir of the repository.
364
:param control_files: Control files to use for locking, etc.
1322
365
"""
1323
366
# In the future we will have a single api for all stores for
1324
367
# getting file texts, inventories and revisions, then
1331
374
self._transport = control_files._transport
1332
375
self.base = self._transport.base
1333
376
# for tests
1334
self._reconcile_does_inventory_gc = True
1335
self._reconcile_fixes_text_parents = False
1336
self._reconcile_backsup_inventory = True
1337
377
self._write_group = None
1338
378
# Additional places to query for data.
1339
379
self._fallback_repositories = []
1340
# An InventoryEntry cache, used during deserialization
1341
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1342
# Is it safe to return inventory entries directly from the entry cache,
1343
# rather copying them?
1344
self._safe_to_return_from_cache = False
1345
380
1346
381
@property
1347
382
def user_transport(self):
1508
543
first_revision.timezone)
1509
544
result['latestrev'] = (last_revision.timestamp,
1510
545
last_revision.timezone)
1511
1512
# now gather global repository information
1513
# XXX: This is available for many repos regardless of listability.
1514
if self.user_transport.listable():
1515
# XXX: do we want to __define len__() ?
1516
# Maybe the versionedfiles object should provide a different
1517
# method to get the number of keys.
1518
result['revisions'] = len(self.revisions.keys())
1519
# result['size'] = t
1520
546
return result
1521
547
1522
548
def find_branches(self, using=False):
1557
583
return ret
1558
584
1559
585
@needs_read_lock
1560
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
586
def search_missing_revision_ids(self, other,
587
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
588
find_ghosts=True, revision_ids=None, if_present_ids=None):
1561
589
"""Return the revision ids that other has that this does not.
1562
590
1563
591
These are returned in topological order.
1564
592
1565
593
revision_id: only return revision ids included by revision_id.
1566
594
"""
595
if symbol_versioning.deprecated_passed(revision_id):
596
symbol_versioning.warn(
597
'search_missing_revision_ids(revision_id=...) was '
598
'deprecated in 2.4. Use revision_ids=[...] instead.',
599
DeprecationWarning, stacklevel=3)
600
if revision_ids is not None:
601
raise AssertionError(
602
'revision_ids is mutually exclusive with revision_id')
603
if revision_id is not None:
604
revision_ids = [revision_id]
1567
605
return InterRepository.get(other, self).search_missing_revision_ids(
1568
revision_id, find_ghosts)
606
find_ghosts=find_ghosts, revision_ids=revision_ids,
607
if_present_ids=if_present_ids)
1569
608
1570
609
@staticmethod
1571
610
def open(base):
1613
652
def suspend_write_group(self):
1614
653
raise errors.UnsuspendableWriteGroup(self)
1615
654
1616
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1617
"""Return the keys of missing inventory parents for revisions added in
1618
this write group.
1619
1620
A revision is not complete if the inventory delta for that revision
1621
cannot be calculated. Therefore if the parent inventories of a
1622
revision are not present, the revision is incomplete, and e.g. cannot
1623
be streamed by a smart server. This method finds missing inventory
1624
parents for revisions added in this write group.
1625
"""
1626
if not self._format.supports_external_lookups:
1627
# This is only an issue for stacked repositories
1628
return set()
1629
if not self.is_in_write_group():
1630
raise AssertionError('not in a write group')
1631
1632
# XXX: We assume that every added revision already has its
1633
# corresponding inventory, so we only check for parent inventories that
1634
# might be missing, rather than all inventories.
1635
parents = set(self.revisions._index.get_missing_parents())
1636
parents.discard(_mod_revision.NULL_REVISION)
1637
unstacked_inventories = self.inventories._index
1638
present_inventories = unstacked_inventories.get_parent_map(
1639
key[-1:] for key in parents)
1640
parents.difference_update(present_inventories)
1641
if len(parents) == 0:
1642
# No missing parent inventories.
1643
return set()
1644
if not check_for_missing_texts:
1645
return set(('inventories', rev_id) for (rev_id,) in parents)
1646
# Ok, now we have a list of missing inventories. But these only matter
1647
# if the inventories that reference them are missing some texts they
1648
# appear to introduce.
1649
# XXX: Texts referenced by all added inventories need to be present,
1650
# but at the moment we're only checking for texts referenced by
1651
# inventories at the graph's edge.
1652
key_deps = self.revisions._index._key_dependencies
1653
key_deps.satisfy_refs_for_keys(present_inventories)
1654
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1655
file_ids = self.fileids_altered_by_revision_ids(referrers)
1656
missing_texts = set()
1657
for file_id, version_ids in file_ids.iteritems():
1658
missing_texts.update(
1659
(file_id, version_id) for version_id in version_ids)
1660
present_texts = self.texts.get_parent_map(missing_texts)
1661
missing_texts.difference_update(present_texts)
1662
if not missing_texts:
1663
# No texts are missing, so all revisions and their deltas are
1664
# reconstructable.
1665
return set()
1666
# Alternatively the text versions could be returned as the missing
1667
# keys, but this is likely to be less data.
1668
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1669
return missing_keys
1670
1671
655
def refresh_data(self):
1672
656
"""Re-read any data needed to synchronise with disk.
1673
657
1693
677
def _resume_write_group(self, tokens):
1694
678
raise errors.UnsuspendableWriteGroup(self)
1695
679
1696
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
680
def fetch(self, source, revision_id=None, find_ghosts=False,
1697
681
fetch_spec=None):
1698
682
"""Fetch the content required to construct revision_id from source.
1699
683
1733
717
not _mod_revision.is_null(revision_id)):
1734
718
self.get_revision(revision_id)
1735
719
return 0, []
1736
# if there is no specific appropriate InterRepository, this will get
1737
# the InterRepository base class, which raises an
1738
# IncompatibleRepositories when asked to fetch.
1739
720
inter = InterRepository.get(source, self)
1740
return inter.fetch(revision_id=revision_id, pb=pb,
721
return inter.fetch(revision_id=revision_id,
1741
722
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
1742
723
1743
724
def create_bundle(self, target, base, fileobj, format=None):
1745
726
1746
727
def get_commit_builder(self, branch, parents, config, timestamp=None,
1747
728
timezone=None, committer=None, revprops=None,
1748
revision_id=None):
729
revision_id=None, lossy=False):
1749
730
"""Obtain a CommitBuilder for this repository.
1750
731
1751
732
:param branch: Branch to commit to.
1756
737
:param committer: Optional committer to set for commit.
1757
738
:param revprops: Optional dictionary of revision properties.
1758
739
:param revision_id: Optional revision id.
740
:param lossy: Whether to discard data that can not be natively
741
represented, when pushing to a foreign VCS
1759
742
"""
1760
if self._fallback_repositories:
1761
raise errors.BzrError("Cannot commit from a lightweight checkout "
1762
"to a stacked branch. See "
1763
"https://bugs.launchpad.net/bzr/+bug/375013 for details.")
1764
result = self._commit_builder_class(self, parents, config,
1765
timestamp, timezone, committer, revprops, revision_id)
1766
self.start_write_group()
1767
return result
743
raise NotImplementedError(self.get_commit_builder)
1768
744
1769
745
@only_raises(errors.LockNotHeld, errors.LockBroken)
1770
746
def unlock(self):
1777
753
'Must end write groups before releasing write locks.')
1778
754
self.control_files.unlock()
1779
755
if self.control_files._lock_count == 0:
1780
self._inventory_entry_cache.clear()
1781
756
for repo in self._fallback_repositories:
1782
757
repo.unlock()
1783
758
1848
823
dest_repo = a_bzrdir.open_repository()
1849
824
return dest_repo
1850
825
1851
def _get_sink(self):
1852
"""Return a sink for streaming into this repository."""
1853
return StreamSink(self)
1854
1855
def _get_source(self, to_format):
1856
"""Return a source for streaming from this repository."""
1857
return StreamSource(self, to_format)
1858
1859
826
@needs_read_lock
1860
827
def has_revision(self, revision_id):
1861
828
"""True if this repository has a copy of the revision."""
1868
835
:param revision_ids: An iterable of revision_ids.
1869
836
:return: A set of the revision_ids that were present.
1870
837
"""
1871
parent_map = self.revisions.get_parent_map(
1872
[(rev_id,) for rev_id in revision_ids])
1873
result = set()
1874
if _mod_revision.NULL_REVISION in revision_ids:
1875
result.add(_mod_revision.NULL_REVISION)
1876
result.update([key[0] for key in parent_map])
1877
return result
838
raise NotImplementedError(self.has_revisions)
1878
839
1879
840
@needs_read_lock
1880
841
def get_revision(self, revision_id):
1881
842
"""Return the Revision object for a named revision."""
1882
843
return self.get_revisions([revision_id])[0]
1883
844
1884
@needs_read_lock
1885
845
def get_revision_reconcile(self, revision_id):
1886
846
"""'reconcile' helper routine that allows access to a revision always.
1887
847
1890
850
be used by reconcile, or reconcile-alike commands that are correcting
1891
851
or testing the revision graph.
1892
852
"""
1893
return self._get_revisions([revision_id])[0]
853
raise NotImplementedError(self.get_revision_reconcile)
1894
854
1895
@needs_read_lock
1896
855
def get_revisions(self, revision_ids):
1897
856
"""Get many revisions at once.
1898
857
1899
858
Repositories that need to check data on every revision read should
1900
859
subclass this method.
1901
860
"""
1902
return self._get_revisions(revision_ids)
1903
1904
@needs_read_lock
1905
def _get_revisions(self, revision_ids):
1906
"""Core work logic to get many revisions without sanity checks."""
1907
revs = {}
1908
for revid, rev in self._iter_revisions(revision_ids):
1909
if rev is None:
1910
raise errors.NoSuchRevision(self, revid)
1911
revs[revid] = rev
1912
return [revs[revid] for revid in revision_ids]
1913
1914
def _iter_revisions(self, revision_ids):
1915
"""Iterate over revision objects.
1916
1917
:param revision_ids: An iterable of revisions to examine. None may be
1918
passed to request all revisions known to the repository. Note that
1919
not all repositories can find unreferenced revisions; for those
1920
repositories only referenced ones will be returned.
1921
:return: An iterator of (revid, revision) tuples. Absent revisions (
1922
those asked for but not available) are returned as (revid, None).
1923
"""
1924
if revision_ids is None:
1925
revision_ids = self.all_revision_ids()
1926
else:
1927
for rev_id in revision_ids:
1928
if not rev_id or not isinstance(rev_id, basestring):
1929
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1930
keys = [(key,) for key in revision_ids]
1931
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1932
for record in stream:
1933
revid = record.key[0]
1934
if record.storage_kind == 'absent':
1935
yield (revid, None)
1936
else:
1937
text = record.get_bytes_as('fulltext')
1938
rev = self._serializer.read_revision_from_string(text)
1939
yield (revid, rev)
861
raise NotImplementedError(self.get_revisions)
1940
862
1941
863
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1942
864
"""Produce a generator of revision deltas.
1997
919
signature = gpg_strategy.sign(plaintext)
1998
920
self.add_signature_text(revision_id, signature)
1999
921
2000
@needs_write_lock
2001
922
def add_signature_text(self, revision_id, signature):
2002
self.signatures.add_lines((revision_id,), (),
2003
osutils.split_lines(signature))
923
"""Store a signature text for a revision.
924
925
:param revision_id: Revision id of the revision
926
:param signature: Signature text.
927
"""
928
raise NotImplementedError(self.add_signature_text)
2004
929
2005
930
def find_text_key_references(self):
2006
931
"""Find the text key references within the repository.
2007
932
2008
933
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2009
to whether they were referred to by the inventory of the
2010
revision_id that they contain. The inventory texts from all present
2011
revision ids are assessed to generate this report.
2012
"""
2013
revision_keys = self.revisions.keys()
2014
w = self.inventories
2015
pb = ui.ui_factory.nested_progress_bar()
2016
try:
2017
return self._find_text_key_references_from_xml_inventory_lines(
2018
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
2019
finally:
2020
pb.finished()
2021
2022
def _find_text_key_references_from_xml_inventory_lines(self,
2023
line_iterator):
2024
"""Core routine for extracting references to texts from inventories.
2025
2026
This performs the translation of xml lines to revision ids.
2027
2028
:param line_iterator: An iterator of lines, origin_version_id
2029
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2030
to whether they were referred to by the inventory of the
2031
revision_id that they contain. Note that if that revision_id was
2032
not part of the line_iterator's output then False will be given -
2033
even though it may actually refer to that key.
2034
"""
2035
if not self._serializer.support_altered_by_hack:
2036
raise AssertionError(
2037
"_find_text_key_references_from_xml_inventory_lines only "
2038
"supported for branches which store inventory as unnested xml"
2039
", not on %r" % self)
2040
result = {}
2041
2042
# this code needs to read every new line in every inventory for the
2043
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
2044
# not present in one of those inventories is unnecessary but not
2045
# harmful because we are filtering by the revision id marker in the
2046
# inventory lines : we only select file ids altered in one of those
2047
# revisions. We don't need to see all lines in the inventory because
2048
# only those added in an inventory in rev X can contain a revision=X
2049
# line.
2050
unescape_revid_cache = {}
2051
unescape_fileid_cache = {}
2052
2053
# jam 20061218 In a big fetch, this handles hundreds of thousands
2054
# of lines, so it has had a lot of inlining and optimizing done.
2055
# Sorry that it is a little bit messy.
2056
# Move several functions to be local variables, since this is a long
2057
# running loop.
2058
search = self._file_ids_altered_regex.search
2059
unescape = _unescape_xml
2060
setdefault = result.setdefault
2061
for line, line_key in line_iterator:
2062
match = search(line)
2063
if match is None:
2064
continue
2065
# One call to match.group() returning multiple items is quite a
2066
# bit faster than 2 calls to match.group() each returning 1
2067
file_id, revision_id = match.group('file_id', 'revision_id')
2068
2069
# Inlining the cache lookups helps a lot when you make 170,000
2070
# lines and 350k ids, versus 8.4 unique ids.
2071
# Using a cache helps in 2 ways:
2072
# 1) Avoids unnecessary decoding calls
2073
# 2) Re-uses cached strings, which helps in future set and
2074
# equality checks.
2075
# (2) is enough that removing encoding entirely along with
2076
# the cache (so we are using plain strings) results in no
2077
# performance improvement.
2078
try:
2079
revision_id = unescape_revid_cache[revision_id]
2080
except KeyError:
2081
unescaped = unescape(revision_id)
2082
unescape_revid_cache[revision_id] = unescaped
2083
revision_id = unescaped
2084
2085
# Note that unconditionally unescaping means that we deserialise
2086
# every fileid, which for general 'pull' is not great, but we don't
2087
# really want to have some many fulltexts that this matters anyway.
2088
# RBC 20071114.
2089
try:
2090
file_id = unescape_fileid_cache[file_id]
2091
except KeyError:
2092
unescaped = unescape(file_id)
2093
unescape_fileid_cache[file_id] = unescaped
2094
file_id = unescaped
2095
2096
key = (file_id, revision_id)
2097
setdefault(key, False)
2098
if revision_id == line_key[-1]:
2099
result[key] = True
2100
return result
2101
2102
def _inventory_xml_lines_for_keys(self, keys):
2103
"""Get a line iterator of the sort needed for findind references.
2104
2105
Not relevant for non-xml inventory repositories.
2106
2107
Ghosts in revision_keys are ignored.
2108
2109
:param revision_keys: The revision keys for the inventories to inspect.
2110
:return: An iterator over (inventory line, revid) for the fulltexts of
2111
all of the xml inventories specified by revision_keys.
2112
"""
2113
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2114
for record in stream:
2115
if record.storage_kind != 'absent':
2116
chunks = record.get_bytes_as('chunked')
2117
revid = record.key[-1]
2118
lines = osutils.chunks_to_lines(chunks)
2119
for line in lines:
2120
yield line, revid
2121
2122
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2123
revision_keys):
2124
"""Helper routine for fileids_altered_by_revision_ids.
2125
2126
This performs the translation of xml lines to revision ids.
2127
2128
:param line_iterator: An iterator of lines, origin_version_id
2129
:param revision_keys: The revision ids to filter for. This should be a
2130
set or other type which supports efficient __contains__ lookups, as
2131
the revision key from each parsed line will be looked up in the
2132
revision_keys filter.
2133
:return: a dictionary mapping altered file-ids to an iterable of
2134
revision_ids. Each altered file-ids has the exact revision_ids that
2135
altered it listed explicitly.
2136
"""
2137
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2138
line_iterator).iterkeys())
2139
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2140
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2141
self._inventory_xml_lines_for_keys(parent_keys)))
2142
new_keys = seen - parent_seen
2143
result = {}
2144
setdefault = result.setdefault
2145
for key in new_keys:
2146
setdefault(key[0], set()).add(key[-1])
2147
return result
934
to whether they were referred to by the tree of the
935
revision_id that they contain.
936
"""
937
raise NotImplementedError(self.find_text_key_references)
2148
938
2149
939
def _find_parent_ids_of_revisions(self, revision_ids):
2150
940
"""Find all parent ids that are mentioned in the revision graph.
2159
949
parent_ids.discard(_mod_revision.NULL_REVISION)
2160
950
return parent_ids
2161
951
2162
def _find_parent_keys_of_revisions(self, revision_keys):
2163
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2164
2165
:param revision_keys: An iterable of revision_keys.
2166
:return: The parents of all revision_keys that are not already in
2167
revision_keys
2168
"""
2169
parent_map = self.revisions.get_parent_map(revision_keys)
2170
parent_keys = set()
2171
map(parent_keys.update, parent_map.itervalues())
2172
parent_keys.difference_update(revision_keys)
2173
parent_keys.discard(_mod_revision.NULL_REVISION)
2174
return parent_keys
2175
2176
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
952
def fileids_altered_by_revision_ids(self, revision_ids):
2177
953
"""Find the file ids and versions affected by revisions.
2178
954
2179
955
:param revisions: an iterable containing revision ids.
2180
:param _inv_weave: The inventory weave from this repository or None.
2181
If None, the inventory weave will be opened automatically.
2182
956
:return: a dictionary mapping altered file-ids to an iterable of
2183
revision_ids. Each altered file-ids has the exact revision_ids that
2184
altered it listed explicitly.
957
revision_ids. Each altered file-ids has the exact revision_ids
958
that altered it listed explicitly.
2185
959
"""
2186
selected_keys = set((revid,) for revid in revision_ids)
2187
w = _inv_weave or self.inventories
2188
return self._find_file_ids_from_xml_inventory_lines(
2189
w.iter_lines_added_or_present_in_keys(
2190
selected_keys, pb=None),
2191
selected_keys)
960
raise NotImplementedError(self.fileids_altered_by_revision_ids)
2192
961
2193
962
def iter_files_bytes(self, desired_files):
2194
963
"""Iterate through file versions.
2201
970
uniquely identify the file version in the caller's context. (Examples:
2202
971
an index number or a TreeTransform trans_id.)
2203
972
2204
bytes_iterator is an iterable of bytestrings for the file. The
2205
kind of iterable and length of the bytestrings are unspecified, but for
2206
this implementation, it is a list of bytes produced by
2207
VersionedFile.get_record_stream().
2208
2209
973
:param desired_files: a list of (file_id, revision_id, identifier)
2210
974
triples
2211
975
"""
2212
text_keys = {}
2213
for file_id, revision_id, callable_data in desired_files:
2214
text_keys[(file_id, revision_id)] = callable_data
2215
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2216
if record.storage_kind == 'absent':
2217
raise errors.RevisionNotPresent(record.key, self)
2218
yield text_keys[record.key], record.get_bytes_as('chunked')
2219
2220
def _generate_text_key_index(self, text_key_references=None,
2221
ancestors=None):
2222
"""Generate a new text key index for the repository.
2223
2224
This is an expensive function that will take considerable time to run.
2225
2226
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2227
list of parents, also text keys. When a given key has no parents,
2228
the parents list will be [NULL_REVISION].
2229
"""
2230
# All revisions, to find inventory parents.
2231
if ancestors is None:
2232
graph = self.get_graph()
2233
ancestors = graph.get_parent_map(self.all_revision_ids())
2234
if text_key_references is None:
2235
text_key_references = self.find_text_key_references()
2236
pb = ui.ui_factory.nested_progress_bar()
2237
try:
2238
return self._do_generate_text_key_index(ancestors,
2239
text_key_references, pb)
2240
finally:
2241
pb.finished()
2242
2243
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2244
"""Helper for _generate_text_key_index to avoid deep nesting."""
2245
revision_order = tsort.topo_sort(ancestors)
2246
invalid_keys = set()
2247
revision_keys = {}
2248
for revision_id in revision_order:
2249
revision_keys[revision_id] = set()
2250
text_count = len(text_key_references)
2251
# a cache of the text keys to allow reuse; costs a dict of all the
2252
# keys, but saves a 2-tuple for every child of a given key.
2253
text_key_cache = {}
2254
for text_key, valid in text_key_references.iteritems():
2255
if not valid:
2256
invalid_keys.add(text_key)
2257
else:
2258
revision_keys[text_key[1]].add(text_key)
2259
text_key_cache[text_key] = text_key
2260
del text_key_references
2261
text_index = {}
2262
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2263
NULL_REVISION = _mod_revision.NULL_REVISION
2264
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2265
# too small for large or very branchy trees. However, for 55K path
2266
# trees, it would be easy to use too much memory trivially. Ideally we
2267
# could gauge this by looking at available real memory etc, but this is
2268
# always a tricky proposition.
2269
inventory_cache = lru_cache.LRUCache(10)
2270
batch_size = 10 # should be ~150MB on a 55K path tree
2271
batch_count = len(revision_order) / batch_size + 1
2272
processed_texts = 0
2273
pb.update("Calculating text parents", processed_texts, text_count)
2274
for offset in xrange(batch_count):
2275
to_query = revision_order[offset * batch_size:(offset + 1) *
2276
batch_size]
2277
if not to_query:
2278
break
2279
for revision_id in to_query:
2280
parent_ids = ancestors[revision_id]
2281
for text_key in revision_keys[revision_id]:
2282
pb.update("Calculating text parents", processed_texts)
2283
processed_texts += 1
2284
candidate_parents = []
2285
for parent_id in parent_ids:
2286
parent_text_key = (text_key[0], parent_id)
2287
try:
2288
check_parent = parent_text_key not in \
2289
revision_keys[parent_id]
2290
except KeyError:
2291
# the parent parent_id is a ghost:
2292
check_parent = False
2293
# truncate the derived graph against this ghost.
2294
parent_text_key = None
2295
if check_parent:
2296
# look at the parent commit details inventories to
2297
# determine possible candidates in the per file graph.
2298
# TODO: cache here.
2299
try:
2300
inv = inventory_cache[parent_id]
2301
except KeyError:
2302
inv = self.revision_tree(parent_id).inventory
2303
inventory_cache[parent_id] = inv
2304
try:
2305
parent_entry = inv[text_key[0]]
2306
except (KeyError, errors.NoSuchId):
2307
parent_entry = None
2308
if parent_entry is not None:
2309
parent_text_key = (
2310
text_key[0], parent_entry.revision)
2311
else:
2312
parent_text_key = None
2313
if parent_text_key is not None:
2314
candidate_parents.append(
2315
text_key_cache[parent_text_key])
2316
parent_heads = text_graph.heads(candidate_parents)
2317
new_parents = list(parent_heads)
2318
new_parents.sort(key=lambda x:candidate_parents.index(x))
2319
if new_parents == []:
2320
new_parents = [NULL_REVISION]
2321
text_index[text_key] = new_parents
2322
2323
for text_key in invalid_keys:
2324
text_index[text_key] = [NULL_REVISION]
2325
return text_index
2326
2327
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2328
"""Get an iterable listing the keys of all the data introduced by a set
2329
of revision IDs.
2330
2331
The keys will be ordered so that the corresponding items can be safely
2332
fetched and inserted in that order.
2333
2334
:returns: An iterable producing tuples of (knit-kind, file-id,
2335
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2336
'revisions'. file-id is None unless knit-kind is 'file'.
2337
"""
2338
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2339
yield result
2340
del _files_pb
2341
for result in self._find_non_file_keys_to_fetch(revision_ids):
2342
yield result
2343
2344
def _find_file_keys_to_fetch(self, revision_ids, pb):
2345
# XXX: it's a bit weird to control the inventory weave caching in this
2346
# generator. Ideally the caching would be done in fetch.py I think. Or
2347
# maybe this generator should explicitly have the contract that it
2348
# should not be iterated until the previously yielded item has been
2349
# processed?
2350
inv_w = self.inventories
2351
2352
# file ids that changed
2353
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2354
count = 0
2355
num_file_ids = len(file_ids)
2356
for file_id, altered_versions in file_ids.iteritems():
2357
if pb is not None:
2358
pb.update("Fetch texts", count, num_file_ids)
2359
count += 1
2360
yield ("file", file_id, altered_versions)
2361
2362
def _find_non_file_keys_to_fetch(self, revision_ids):
2363
# inventory
2364
yield ("inventory", None, revision_ids)
2365
2366
# signatures
2367
# XXX: Note ATM no callers actually pay attention to this return
2368
# instead they just use the list of revision ids and ignore
2369
# missing sigs. Consider removing this work entirely
2370
revisions_with_signatures = set(self.signatures.get_parent_map(
2371
[(r,) for r in revision_ids]))
2372
revisions_with_signatures = set(
2373
[r for (r,) in revisions_with_signatures])
2374
revisions_with_signatures.intersection_update(revision_ids)
2375
yield ("signatures", None, revisions_with_signatures)
2376
2377
# revisions
2378
yield ("revisions", None, revision_ids)
2379
2380
@needs_read_lock
2381
def get_inventory(self, revision_id):
2382
"""Get Inventory object by revision id."""
2383
return self.iter_inventories([revision_id]).next()
2384
2385
def iter_inventories(self, revision_ids, ordering=None):
2386
"""Get many inventories by revision_ids.
2387
2388
This will buffer some or all of the texts used in constructing the
2389
inventories in memory, but will only parse a single inventory at a
2390
time.
2391
2392
:param revision_ids: The expected revision ids of the inventories.
2393
:param ordering: optional ordering, e.g. 'topological'. If not
2394
specified, the order of revision_ids will be preserved (by
2395
buffering if necessary).
2396
:return: An iterator of inventories.
2397
"""
2398
if ((None in revision_ids)
2399
or (_mod_revision.NULL_REVISION in revision_ids)):
2400
raise ValueError('cannot get null revision inventory')
2401
return self._iter_inventories(revision_ids, ordering)
2402
2403
def _iter_inventories(self, revision_ids, ordering):
2404
"""single-document based inventory iteration."""
2405
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2406
for text, revision_id in inv_xmls:
2407
yield self._deserialise_inventory(revision_id, text)
2408
2409
def _iter_inventory_xmls(self, revision_ids, ordering):
2410
if ordering is None:
2411
order_as_requested = True
2412
ordering = 'unordered'
2413
else:
2414
order_as_requested = False
2415
keys = [(revision_id,) for revision_id in revision_ids]
2416
if not keys:
2417
return
2418
if order_as_requested:
2419
key_iter = iter(keys)
2420
next_key = key_iter.next()
2421
stream = self.inventories.get_record_stream(keys, ordering, True)
2422
text_chunks = {}
2423
for record in stream:
2424
if record.storage_kind != 'absent':
2425
chunks = record.get_bytes_as('chunked')
2426
if order_as_requested:
2427
text_chunks[record.key] = chunks
2428
else:
2429
yield ''.join(chunks), record.key[-1]
2430
else:
2431
raise errors.NoSuchRevision(self, record.key)
2432
if order_as_requested:
2433
# Yield as many results as we can while preserving order.
2434
while next_key in text_chunks:
2435
chunks = text_chunks.pop(next_key)
2436
yield ''.join(chunks), next_key[-1]
2437
try:
2438
next_key = key_iter.next()
2439
except StopIteration:
2440
# We still want to fully consume the get_record_stream,
2441
# just in case it is not actually finished at this point
2442
next_key = None
2443
break
2444
2445
def _deserialise_inventory(self, revision_id, xml):
2446
"""Transform the xml into an inventory object.
2447
2448
:param revision_id: The expected revision id of the inventory.
2449
:param xml: A serialised inventory.
2450
"""
2451
result = self._serializer.read_inventory_from_string(xml, revision_id,
2452
entry_cache=self._inventory_entry_cache,
2453
return_from_cache=self._safe_to_return_from_cache)
2454
if result.revision_id != revision_id:
2455
raise AssertionError('revision id mismatch %s != %s' % (
2456
result.revision_id, revision_id))
2457
return result
2458
2459
def get_serializer_format(self):
2460
return self._serializer.format_num
2461
2462
@needs_read_lock
2463
def _get_inventory_xml(self, revision_id):
2464
"""Get serialized inventory as a string."""
2465
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2466
try:
2467
text, revision_id = texts.next()
2468
except StopIteration:
2469
raise errors.HistoryMissing(self, 'inventory', revision_id)
2470
return text
976
raise NotImplementedError(self.iter_files_bytes)
2471
977
2472
978
def get_rev_id_for_revno(self, revno, known_pair):
2473
979
"""Return the revision id of a revno, given a later (revno, revid)
2511
1017
ancestors will be traversed.
2512
1018
"""
2513
1019
graph = self.get_graph()
2514
next_id = revision_id
2515
while True:
2516
if next_id in (None, _mod_revision.NULL_REVISION):
2517
return
2518
try:
2519
parents = graph.get_parent_map([next_id])[next_id]
2520
except KeyError:
2521
raise errors.RevisionNotPresent(next_id, self)
2522
yield next_id
2523
if len(parents) == 0:
2524
return
2525
else:
2526
next_id = parents[0]
1020
stop_revisions = (None, _mod_revision.NULL_REVISION)
1021
return graph.iter_lefthand_ancestry(revision_id, stop_revisions)
2527
1022
2528
1023
def is_shared(self):
2529
1024
"""Return True if this repository is flagged as a shared repository."""
2557
1052
2558
1053
`revision_id` may be NULL_REVISION for the empty tree revision.
2559
1054
"""
2560
revision_id = _mod_revision.ensure_null(revision_id)
2561
# TODO: refactor this to use an existing revision object
2562
# so we don't need to read it in twice.
2563
if revision_id == _mod_revision.NULL_REVISION:
2564
return RevisionTree(self, Inventory(root_id=None),
2565
_mod_revision.NULL_REVISION)
2566
else:
2567
inv = self.get_inventory(revision_id)
2568
return RevisionTree(self, inv, revision_id)
1055
raise NotImplementedError(self.revision_tree)
2569
1056
2570
1057
def revision_trees(self, revision_ids):
2571
1058
"""Return Trees for revisions in this repository.
2573
1060
:param revision_ids: a sequence of revision-ids;
2574
1061
a revision-id may not be None or 'null:'
2575
1062
"""
2576
inventories = self.iter_inventories(revision_ids)
2577
for inv in inventories:
2578
yield RevisionTree(self, inv, inv.revision_id)
2579
2580
def _filtered_revision_trees(self, revision_ids, file_ids):
2581
"""Return Tree for a revision on this branch with only some files.
2582
2583
:param revision_ids: a sequence of revision-ids;
2584
a revision-id may not be None or 'null:'
2585
:param file_ids: if not None, the result is filtered
2586
so that only those file-ids, their parents and their
2587
children are included.
2588
"""
2589
inventories = self.iter_inventories(revision_ids)
2590
for inv in inventories:
2591
# Should we introduce a FilteredRevisionTree class rather
2592
# than pre-filter the inventory here?
2593
filtered_inv = inv.filter(file_ids)
2594
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
1063
raise NotImplementedError(self.revision_trees)
2595
1064
2596
1065
@needs_read_lock
2597
1066
def get_ancestry(self, revision_id, topo_sorted=True):
2630
1099
types it should be a no-op that just returns.
2631
1100
2632
1101
This stub method does not require a lock, but subclasses should use
2633
@needs_write_lock as this is a long running call its reasonable to
1102
@needs_write_lock as this is a long running call it's reasonable to
2634
1103
implicitly lock for the user.
2635
1104
2636
1105
:param hint: If not supplied, the whole repository is packed.
2649
1118
2650
1119
def get_parent_map(self, revision_ids):
2651
1120
"""See graph.StackedParentsProvider.get_parent_map"""
1121
raise NotImplementedError(self.get_parent_map)
1122
1123
def _get_parent_map_no_fallbacks(self, revision_ids):
1124
"""Same as Repository.get_parent_map except doesn't query fallbacks."""
2652
1125
# revisions index works in keys; this just works in revisions
2653
1126
# therefore wrap and unwrap
2654
1127
query_keys = []
2660
1133
raise ValueError('get_parent_map(None) is not valid')
2661
1134
else:
2662
1135
query_keys.append((revision_id ,))
1136
vf = self.revisions.without_fallbacks()
2663
1137
for ((revision_id,), parent_keys) in \
2664
self.revisions.get_parent_map(query_keys).iteritems():
1138
vf.get_parent_map(query_keys).iteritems():
2665
1139
if parent_keys:
2666
1140
result[revision_id] = tuple([parent_revid
2667
1141
for (parent_revid,) in parent_keys])
2670
1144
return result
2671
1145
2672
1146
def _make_parents_provider(self):
2673
return self
1147
if not self._format.supports_external_lookups:
1148
return self
1149
return graph.StackedParentsProvider(_LazyListJoin(
1150
[self._make_parents_provider_unstacked()],
1151
self._fallback_repositories))
1152
1153
def _make_parents_provider_unstacked(self):
1154
return graph.CallableToParentsProviderAdapter(
1155
self._get_parent_map_no_fallbacks)
2674
1156
2675
1157
@needs_read_lock
2676
1158
def get_known_graph_ancestry(self, revision_ids):
2677
1159
"""Return the known graph for a set of revision ids and their ancestors.
2678
1160
"""
2679
st = static_tuple.StaticTuple
2680
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2681
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2682
return graph.GraphThunkIdsToKeys(known_graph)
1161
raise NotImplementedError(self.get_known_graph_ancestry)
2683
1162
2684
1163
def get_graph(self, other_repository=None):
2685
1164
"""Return the graph walker for this repository format"""
2690
1169
[parents_provider, other_repository._make_parents_provider()])
2691
1170
return graph.Graph(parents_provider)
2692
1171
2693
def _get_versioned_file_checker(self, text_key_references=None,
2694
ancestors=None):
2695
"""Return an object suitable for checking versioned files.
2696
2697
:param text_key_references: if non-None, an already built
2698
dictionary mapping text keys ((fileid, revision_id) tuples)
2699
to whether they were referred to by the inventory of the
2700
revision_id that they contain. If None, this will be
2701
calculated.
2702
:param ancestors: Optional result from
2703
self.get_graph().get_parent_map(self.all_revision_ids()) if already
2704
available.
2705
"""
2706
return _VersionedFileChecker(self,
2707
text_key_references=text_key_references, ancestors=ancestors)
2708
2709
1172
def revision_ids_to_search_result(self, result_set):
2710
1173
"""Convert a set of revision ids to a graph SearchResult."""
2711
1174
result_parents = set()
2737
1200
2738
1201
@needs_write_lock
2739
1202
def sign_revision(self, revision_id, gpg_strategy):
2740
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1203
testament = _mod_testament.Testament.from_revision(self, revision_id)
1204
plaintext = testament.as_short_text()
2741
1205
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2742
1206
2743
@needs_read_lock
2744
1207
def has_signature_for_revision_id(self, revision_id):
2745
1208
"""Query for a revision signature for revision_id in the repository."""
2746
if not self.has_revision(revision_id):
2747
raise errors.NoSuchRevision(self, revision_id)
2748
sig_present = (1 == len(
2749
self.signatures.get_parent_map([(revision_id,)])))
2750
return sig_present
1209
raise NotImplementedError(self.has_signature_for_revision_id)
2751
1210
2752
@needs_read_lock
2753
1211
def get_signature_text(self, revision_id):
2754
1212
"""Return the text for a signature."""
2755
stream = self.signatures.get_record_stream([(revision_id,)],
2756
'unordered', True)
2757
record = stream.next()
2758
if record.storage_kind == 'absent':
2759
raise errors.NoSuchRevision(self, revision_id)
2760
return record.get_bytes_as('fulltext')
1213
raise NotImplementedError(self.get_signature_text)
2761
1214
2762
1215
@needs_read_lock
2763
1216
def check(self, revision_ids=None, callback_refs=None, check_repo=True):
2782
1235
return result
2783
1236
2784
1237
def _warn_if_deprecated(self, branch=None):
1238
if not self._format.is_deprecated():
1239
return
2785
1240
global _deprecation_warning_done
2786
1241
if _deprecation_warning_done:
2787
1242
return
2817
1272
except UnicodeDecodeError:
2818
1273
raise errors.NonAsciiRevisionId(method, self)
2819
1274
2820
def revision_graph_can_have_wrong_parents(self):
2821
"""Is it possible for this repository to have a revision graph with
2822
incorrect parents?
2823
2824
If True, then this repository must also implement
2825
_find_inconsistent_revision_parents so that check and reconcile can
2826
check for inconsistencies before proceeding with other checks that may
2827
depend on the revision index being consistent.
2828
"""
2829
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2830
2831
2832
# remove these delegates a while after bzr 0.15
2833
def __make_delegated(name, from_module):
2834
def _deprecated_repository_forwarder():
2835
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2836
% (name, from_module),
2837
DeprecationWarning,
2838
stacklevel=2)
2839
m = __import__(from_module, globals(), locals(), [name])
2840
try:
2841
return getattr(m, name)
2842
except AttributeError:
2843
raise AttributeError('module %s has no name %s'
2844
% (m, name))
2845
globals()[name] = _deprecated_repository_forwarder
2846
2847
for _name in [
2848
'AllInOneRepository',
2849
'WeaveMetaDirRepository',
2850
'PreSplitOutRepositoryFormat',
2851
'RepositoryFormat4',
2852
'RepositoryFormat5',
2853
'RepositoryFormat6',
2854
'RepositoryFormat7',
2855
]:
2856
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2857
2858
for _name in [
2859
'KnitRepository',
2860
'RepositoryFormatKnit',
2861
'RepositoryFormatKnit1',
2862
]:
2863
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2864
2865
2866
def install_revision(repository, rev, revision_tree):
2867
"""Install all revision data into a repository."""
2868
install_revisions(repository, [(rev, revision_tree, None)])
2869
2870
2871
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2872
"""Install all revision data into a repository.
2873
2874
Accepts an iterable of revision, tree, signature tuples. The signature
2875
may be None.
2876
"""
2877
repository.start_write_group()
2878
try:
2879
inventory_cache = lru_cache.LRUCache(10)
2880
for n, (revision, revision_tree, signature) in enumerate(iterable):
2881
_install_revision(repository, revision, revision_tree, signature,
2882
inventory_cache)
2883
if pb is not None:
2884
pb.update('Transferring revisions', n + 1, num_revisions)
2885
except:
2886
repository.abort_write_group()
2887
raise
2888
else:
2889
repository.commit_write_group()
2890
2891
2892
def _install_revision(repository, rev, revision_tree, signature,
2893
inventory_cache):
2894
"""Install all revision data into a repository."""
2895
present_parents = []
2896
parent_trees = {}
2897
for p_id in rev.parent_ids:
2898
if repository.has_revision(p_id):
2899
present_parents.append(p_id)
2900
parent_trees[p_id] = repository.revision_tree(p_id)
2901
else:
2902
parent_trees[p_id] = repository.revision_tree(
2903
_mod_revision.NULL_REVISION)
2904
2905
inv = revision_tree.inventory
2906
entries = inv.iter_entries()
2907
# backwards compatibility hack: skip the root id.
2908
if not repository.supports_rich_root():
2909
path, root = entries.next()
2910
if root.revision != rev.revision_id:
2911
raise errors.IncompatibleRevision(repr(repository))
2912
text_keys = {}
2913
for path, ie in entries:
2914
text_keys[(ie.file_id, ie.revision)] = ie
2915
text_parent_map = repository.texts.get_parent_map(text_keys)
2916
missing_texts = set(text_keys) - set(text_parent_map)
2917
# Add the texts that are not already present
2918
for text_key in missing_texts:
2919
ie = text_keys[text_key]
2920
text_parents = []
2921
# FIXME: TODO: The following loop overlaps/duplicates that done by
2922
# commit to determine parents. There is a latent/real bug here where
2923
# the parents inserted are not those commit would do - in particular
2924
# they are not filtered by heads(). RBC, AB
2925
for revision, tree in parent_trees.iteritems():
2926
if ie.file_id not in tree:
2927
continue
2928
parent_id = tree.inventory[ie.file_id].revision
2929
if parent_id in text_parents:
2930
continue
2931
text_parents.append((ie.file_id, parent_id))
2932
lines = revision_tree.get_file(ie.file_id).readlines()
2933
repository.texts.add_lines(text_key, text_parents, lines)
2934
try:
2935
# install the inventory
2936
if repository._format._commit_inv_deltas and len(rev.parent_ids):
2937
# Cache this inventory
2938
inventory_cache[rev.revision_id] = inv
2939
try:
2940
basis_inv = inventory_cache[rev.parent_ids[0]]
2941
except KeyError:
2942
repository.add_inventory(rev.revision_id, inv, present_parents)
2943
else:
2944
delta = inv._make_delta(basis_inv)
2945
repository.add_inventory_by_delta(rev.parent_ids[0], delta,
2946
rev.revision_id, present_parents)
2947
else:
2948
repository.add_inventory(rev.revision_id, inv, present_parents)
2949
except errors.RevisionAlreadyPresent:
2950
pass
2951
if signature is not None:
2952
repository.add_signature_text(rev.revision_id, signature)
2953
repository.add_revision(rev.revision_id, rev, inv)
2954
2955
1275
2956
1276
class MetaDirRepository(Repository):
2957
1277
"""Repositories in the new meta-dir layout.
2992
1312
return not self._transport.has('no-working-trees')
2993
1313
2994
1314
2995
class MetaDirVersionedFileRepository(MetaDirRepository):
2996
"""Repositories in a meta-dir, that work via versioned file objects."""
1315
class RepositoryFormatRegistry(controldir.ControlComponentFormatRegistry):
1316
"""Repository format registry."""
2997
1317
2998
def __init__(self, _format, a_bzrdir, control_files):
2999
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
3000
control_files)
1318
def get_default(self):
1319
"""Return the current default format."""
1320
from bzrlib import bzrdir
1321
return bzrdir.format_registry.make_bzrdir('default').repository_format
3001
1322
3002
1323
3003
1324
network_format_registry = registry.FormatRegistry()
3009
1330
"""
3010
1331
3011
1332
3012
format_registry = registry.FormatRegistry(network_format_registry)
1333
format_registry = RepositoryFormatRegistry(network_format_registry)
3013
1334
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
3014
1335
3015
1336
This can contain either format instances themselves, or classes/factories that
3020
1341
#####################################################################
3021
1342
# Repository Formats
3022
1343
3023
class RepositoryFormat(object):
1344
class RepositoryFormat(controldir.ControlComponentFormat):
3024
1345
"""A repository format.
3025
1346
3026
1347
Formats provide four things:
3061
1382
# Does this format support CHK bytestring lookups. Set to True or False in
3062
1383
# derived classes.
3063
1384
supports_chks = None
3064
# Should commit add an inventory, or an inventory delta to the repository.
3065
_commit_inv_deltas = True
3066
# What order should fetch operations request streams in?
3067
# The default is unordered as that is the cheapest for an origin to
3068
# provide.
3069
_fetch_order = 'unordered'
3070
# Does this repository format use deltas that can be fetched as-deltas ?
3071
# (E.g. knits, where the knit deltas can be transplanted intact.
3072
# We default to False, which will ensure that enough data to get
3073
# a full text out of any fetch stream will be grabbed.
3074
_fetch_uses_deltas = False
3075
1385
# Should fetch trigger a reconcile after the fetch? Only needed for
3076
1386
# some repository formats that can suffer internal inconsistencies.
3077
1387
_fetch_reconcile = False
3083
1393
# help), and for fetching when data won't have come from the same
3084
1394
# compressor.
3085
1395
pack_compresses = False
3086
# Does the repository inventory storage understand references to trees?
1396
# Does the repository storage understand references to trees?
3087
1397
supports_tree_reference = None
3088
1398
# Is the format experimental ?
3089
1399
experimental = False
1400
# Does this repository format escape funky characters, or does it create
1401
# files with similar names as the versioned files in its contents on disk
1402
# ?
1403
supports_funky_characters = None
1404
# Does this repository format support leaving locks?
1405
supports_leaving_lock = None
1406
# Does this format support the full VersionedFiles interface?
1407
supports_full_versioned_files = None
1408
# Does this format support signing revision signatures?
1409
supports_revision_signatures = True
1410
# Can the revision graph have incorrect parents?
1411
revision_graph_can_have_wrong_parents = None
1412
# Does this format support rich root data?
1413
rich_root_data = None
3090
1414
3091
1415
def __repr__(self):
3092
1416
return "%s()" % self.__class__.__name__
3117
1441
kind='repository')
3118
1442
3119
1443
@classmethod
1444
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3120
1445
def register_format(klass, format):
3121
format_registry.register(format.get_format_string(), format)
1446
format_registry.register(format)
3122
1447
3123
1448
@classmethod
1449
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3124
1450
def unregister_format(klass, format):
3125
format_registry.remove(format.get_format_string())
1451
format_registry.remove(format)
3126
1452
3127
1453
@classmethod
1454
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3128
1455
def get_default_format(klass):
3129
1456
"""Return the current default format."""
3130
from bzrlib import bzrdir
3131
return bzrdir.format_registry.make_bzrdir('default').repository_format
1457
return format_registry.get_default()
3132
1458
3133
1459
def get_format_string(self):
3134
1460
"""Return the ASCII format string that identifies this format.
3142
1468
"""Return the short description for this format."""
3143
1469
raise NotImplementedError(self.get_format_description)
3144
1470
3145
# TODO: this shouldn't be in the base class, it's specific to things that
3146
# use weaves or knits -- mbp 20070207
3147
def _get_versioned_file_store(self,
3148
name,
3149
transport,
3150
control_files,
3151
prefixed=True,
3152
versionedfile_class=None,
3153
versionedfile_kwargs={},
3154
escaped=False):
3155
if versionedfile_class is None:
3156
versionedfile_class = self._versionedfile_class
3157
weave_transport = control_files._transport.clone(name)
3158
dir_mode = control_files._dir_mode
3159
file_mode = control_files._file_mode
3160
return VersionedFileStore(weave_transport, prefixed=prefixed,
3161
dir_mode=dir_mode,
3162
file_mode=file_mode,
3163
versionedfile_class=versionedfile_class,
3164
versionedfile_kwargs=versionedfile_kwargs,
3165
escaped=escaped)
3166
3167
1471
def initialize(self, a_bzrdir, shared=False):
3168
1472
"""Initialize a repository of this format in a_bzrdir.
3169
1473
3185
1489
"""
3186
1490
return True
3187
1491
1492
def is_deprecated(self):
1493
"""Is this format deprecated?
1494
1495
Deprecated formats may trigger a user-visible warning recommending
1496
the user to upgrade. They are still fully supported.
1497
"""
1498
return False
1499
3188
1500
def network_name(self):
3189
1501
"""A simple byte string uniquely identifying this format for RPC calls.
3190
1502
3229
1541
rich_root_data = False
3230
1542
supports_tree_reference = False
3231
1543
supports_external_lookups = False
1544
supports_leaving_lock = True
3232
1545
3233
1546
@property
3234
1547
def _matchingbzrdir(self):
3272
1585
return self.get_format_string()
3273
1586
3274
1587
3275
# Pre-0.8 formats that don't have a disk format string (because they are
3276
# versioned by the matching control directory). We use the control directories
3277
# disk format string as a key for the network_name because they meet the
3278
# constraints (simple string, unique, immutable).
3279
network_format_registry.register_lazy(
3280
"Bazaar-NG branch, format 5\n",
3281
'bzrlib.repofmt.weaverepo',
3282
'RepositoryFormat5',
3283
)
3284
network_format_registry.register_lazy(
3285
"Bazaar-NG branch, format 6\n",
3286
'bzrlib.repofmt.weaverepo',
3287
'RepositoryFormat6',
3288
)
3289
3290
1588
# formats which have no format string are not discoverable or independently
3291
1589
# creatable on disk, so are not registered in format_registry. They're
3292
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
1590
# all in bzrlib.repofmt.knitreponow. When an instance of one of these is
3293
1591
# needed, it's constructed directly by the BzrDir. Non-native formats where
3294
1592
# the repository is not separately opened are similar.
3295
1593
3296
1594
format_registry.register_lazy(
3297
'Bazaar-NG Repository format 7',
3298
'bzrlib.repofmt.weaverepo',
3299
'RepositoryFormat7'
3300
)
3301
3302
format_registry.register_lazy(
3303
1595
'Bazaar-NG Knit Repository Format 1',
3304
1596
'bzrlib.repofmt.knitrepo',
3305
1597
'RepositoryFormatKnit1',
3322
1614
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
3323
1615
format_registry.register_lazy(
3324
1616
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
3325
'bzrlib.repofmt.pack_repo',
1617
'bzrlib.repofmt.knitpack_repo',
3326
1618
'RepositoryFormatKnitPack1',
3327
1619
)
3328
1620
format_registry.register_lazy(
3329
1621
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
3330
'bzrlib.repofmt.pack_repo',
1622
'bzrlib.repofmt.knitpack_repo',
3331
1623
'RepositoryFormatKnitPack3',
3332
1624
)
3333
1625
format_registry.register_lazy(
3334
1626
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
3335
'bzrlib.repofmt.pack_repo',
1627
'bzrlib.repofmt.knitpack_repo',
3336
1628
'RepositoryFormatKnitPack4',
3337
1629
)
3338
1630
format_registry.register_lazy(
3339
1631
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
3340
'bzrlib.repofmt.pack_repo',
1632
'bzrlib.repofmt.knitpack_repo',
3341
1633
'RepositoryFormatKnitPack5',
3342
1634
)
3343
1635
format_registry.register_lazy(
3344
1636
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
3345
'bzrlib.repofmt.pack_repo',
1637
'bzrlib.repofmt.knitpack_repo',
3346
1638
'RepositoryFormatKnitPack5RichRoot',
3347
1639
)
3348
1640
format_registry.register_lazy(
3349
1641
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
3350
'bzrlib.repofmt.pack_repo',
1642
'bzrlib.repofmt.knitpack_repo',
3351
1643
'RepositoryFormatKnitPack5RichRootBroken',
3352
1644
)
3353
1645
format_registry.register_lazy(
3354
1646
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
3355
'bzrlib.repofmt.pack_repo',
1647
'bzrlib.repofmt.knitpack_repo',
3356
1648
'RepositoryFormatKnitPack6',
3357
1649
)
3358
1650
format_registry.register_lazy(
3359
1651
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
3360
'bzrlib.repofmt.pack_repo',
1652
'bzrlib.repofmt.knitpack_repo',
3361
1653
'RepositoryFormatKnitPack6RichRoot',
3362
1654
)
1655
format_registry.register_lazy(
1656
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
1657
'bzrlib.repofmt.groupcompress_repo',
1658
'RepositoryFormat2a',
1659
)
3363
1660
3364
1661
# Development formats.
3365
# Obsolete but kept pending a CHK based subtree format.
1662
# Check their docstrings to see if/when they are obsolete.
3366
1663
format_registry.register_lazy(
3367
1664
("Bazaar development format 2 with subtree support "
3368
1665
"(needs bzr.dev from before 1.8)\n"),
3369
'bzrlib.repofmt.pack_repo',
1666
'bzrlib.repofmt.knitpack_repo',
3370
1667
'RepositoryFormatPackDevelopment2Subtree',
3371
1668
)
3372
3373
# 1.14->1.16 go below here
3374
format_registry.register_lazy(
3375
'Bazaar development format - group compression and chk inventory'
3376
' (needs bzr.dev from 1.14)\n',
3377
'bzrlib.repofmt.groupcompress_repo',
3378
'RepositoryFormatCHK1',
3379
)
3380
3381
format_registry.register_lazy(
3382
'Bazaar development format - chk repository with bencode revision '
3383
'serialization (needs bzr.dev from 1.16)\n',
3384
'bzrlib.repofmt.groupcompress_repo',
3385
'RepositoryFormatCHK2',
3386
)
3387
format_registry.register_lazy(
3388
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3389
'bzrlib.repofmt.groupcompress_repo',
3390
'RepositoryFormat2a',
1669
format_registry.register_lazy(
1670
'Bazaar development format 8\n',
1671
'bzrlib.repofmt.groupcompress_repo',
1672
'RepositoryFormat2aSubtree',
3391
1673
)
3392
1674
3393
1675
3424
1706
self.target.fetch(self.source, revision_id=revision_id)
3425
1707
3426
1708
@needs_write_lock
3427
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
1709
def fetch(self, revision_id=None, find_ghosts=False,
3428
1710
fetch_spec=None):
3429
1711
"""Fetch the content required to construct revision_id.
3430
1712
3432
1714
3433
1715
:param revision_id: if None all content is copied, if NULL_REVISION no
3434
1716
content is copied.
3435
:param pb: ignored.
3436
1717
:return: None.
3437
1718
"""
3438
1719
ui.ui_factory.warn_experimental_format_fetch(self)
3448
1729
fetch_spec=fetch_spec,
3449
1730
find_ghosts=find_ghosts)
3450
1731
3451
def _walk_to_common_revisions(self, revision_ids):
1732
def _walk_to_common_revisions(self, revision_ids, if_present_ids=None):
3452
1733
"""Walk out from revision_ids in source to revisions target has.
3453
1734
3454
1735
:param revision_ids: The start point for the search.
3456
1737
"""
3457
1738
target_graph = self.target.get_graph()
3458
1739
revision_ids = frozenset(revision_ids)
1740
if if_present_ids:
1741
all_wanted_revs = revision_ids.union(if_present_ids)
1742
else:
1743
all_wanted_revs = revision_ids
3459
1744
missing_revs = set()
3460
1745
source_graph = self.source.get_graph()
3461
1746
# ensure we don't pay silly lookup costs.
3462
searcher = source_graph._make_breadth_first_searcher(revision_ids)
1747
searcher = source_graph._make_breadth_first_searcher(all_wanted_revs)
3463
1748
null_set = frozenset([_mod_revision.NULL_REVISION])
3464
1749
searcher_exhausted = False
3465
1750
while True:
3501
1786
return searcher.get_result()
3502
1787
3503
1788
@needs_read_lock
3504
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
1789
def search_missing_revision_ids(self,
1790
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
1791
find_ghosts=True, revision_ids=None, if_present_ids=None):
3505
1792
"""Return the revision ids that source has that target does not.
3506
1793
3507
1794
:param revision_id: only return revision ids included by this
3508
revision_id.
1795
revision_id.
1796
:param revision_ids: return revision ids included by these
1797
revision_ids. NoSuchRevision will be raised if any of these
1798
revisions are not present.
1799
:param if_present_ids: like revision_ids, but will not cause
1800
NoSuchRevision if any of these are absent, instead they will simply
1801
not be in the result. This is useful for e.g. finding revisions
1802
to fetch for tags, which may reference absent revisions.
3509
1803
:param find_ghosts: If True find missing revisions in deep history
3510
1804
rather than just finding the surface difference.
3511
1805
:return: A bzrlib.graph.SearchResult.
3512
1806
"""
1807
if symbol_versioning.deprecated_passed(revision_id):
1808
symbol_versioning.warn(
1809
'search_missing_revision_ids(revision_id=...) was '
1810
'deprecated in 2.4. Use revision_ids=[...] instead.',
1811
DeprecationWarning, stacklevel=2)
1812
if revision_ids is not None:
1813
raise AssertionError(
1814
'revision_ids is mutually exclusive with revision_id')
1815
if revision_id is not None:
1816
revision_ids = [revision_id]
1817
del revision_id
3513
1818
# stop searching at found target revisions.
3514
if not find_ghosts and revision_id is not None:
3515
return self._walk_to_common_revisions([revision_id])
1819
if not find_ghosts and (revision_ids is not None or if_present_ids is
1820
not None):
1821
return self._walk_to_common_revisions(revision_ids,
1822
if_present_ids=if_present_ids)
3516
1823
# generic, possibly worst case, slow code path.
3517
1824
target_ids = set(self.target.all_revision_ids())
3518
if revision_id is not None:
3519
source_ids = self.source.get_ancestry(revision_id)
3520
if source_ids[0] is not None:
3521
raise AssertionError()
3522
source_ids.pop(0)
3523
else:
3524
source_ids = self.source.all_revision_ids()
1825
source_ids = self._present_source_revisions_for(
1826
revision_ids, if_present_ids)
3525
1827
result_set = set(source_ids).difference(target_ids)
3526
1828
return self.source.revision_ids_to_search_result(result_set)
3527
1829
1830
def _present_source_revisions_for(self, revision_ids, if_present_ids=None):
1831
"""Returns set of all revisions in ancestry of revision_ids present in
1832
the source repo.
1833
1834
:param revision_ids: if None, all revisions in source are returned.
1835
:param if_present_ids: like revision_ids, but if any/all of these are
1836
absent no error is raised.
1837
"""
1838
if revision_ids is not None or if_present_ids is not None:
1839
# First, ensure all specified revisions exist. Callers expect
1840
# NoSuchRevision when they pass absent revision_ids here.
1841
if revision_ids is None:
1842
revision_ids = set()
1843
if if_present_ids is None:
1844
if_present_ids = set()
1845
revision_ids = set(revision_ids)
1846
if_present_ids = set(if_present_ids)
1847
all_wanted_ids = revision_ids.union(if_present_ids)
1848
graph = self.source.get_graph()
1849
present_revs = set(graph.get_parent_map(all_wanted_ids))
1850
missing = revision_ids.difference(present_revs)
1851
if missing:
1852
raise errors.NoSuchRevision(self.source, missing.pop())
1853
found_ids = all_wanted_ids.intersection(present_revs)
1854
source_ids = [rev_id for (rev_id, parents) in
1855
graph.iter_ancestry(found_ids)
1856
if rev_id != _mod_revision.NULL_REVISION
1857
and parents is not None]
1858
else:
1859
source_ids = self.source.all_revision_ids()
1860
return set(source_ids)
1861
3528
1862
@staticmethod
3529
1863
def _same_model(source, target):
3530
1864
"""True if source and target have the same data representation.
3549
1883
raise errors.IncompatibleRepositories(source, target,
3550
1884
"different serializers")
3551
1885
3552
3553
class InterSameDataRepository(InterRepository):
3554
"""Code for converting between repositories that represent the same data.
3555
3556
Data format and model must match for this to work.
3557
"""
3558
3559
@classmethod
3560
def _get_repo_format_to_test(self):
3561
"""Repository format for testing with.
3562
3563
InterSameData can pull from subtree to subtree and from non-subtree to
3564
non-subtree, so we test this with the richest repository format.
3565
"""
3566
from bzrlib.repofmt import knitrepo
3567
return knitrepo.RepositoryFormatKnit3()
3568
3569
@staticmethod
3570
def is_compatible(source, target):
3571
return InterRepository._same_model(source, target)
3572
3573
3574
class InterWeaveRepo(InterSameDataRepository):
3575
"""Optimised code paths between Weave based repositories.
3576
3577
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
3578
implemented lazy inter-object optimisation.
3579
"""
3580
3581
@classmethod
3582
def _get_repo_format_to_test(self):
3583
from bzrlib.repofmt import weaverepo
3584
return weaverepo.RepositoryFormat7()
3585
3586
@staticmethod
3587
def is_compatible(source, target):
3588
"""Be compatible with known Weave formats.
3589
3590
We don't test for the stores being of specific types because that
3591
could lead to confusing results, and there is no need to be
3592
overly general.
3593
"""
3594
from bzrlib.repofmt.weaverepo import (
3595
RepositoryFormat5,
3596
RepositoryFormat6,
3597
RepositoryFormat7,
3598
)
3599
try:
3600
return (isinstance(source._format, (RepositoryFormat5,
3601
RepositoryFormat6,
3602
RepositoryFormat7)) and
3603
isinstance(target._format, (RepositoryFormat5,
3604
RepositoryFormat6,
3605
RepositoryFormat7)))
3606
except AttributeError:
3607
return False
3608
3609
@needs_write_lock
3610
def copy_content(self, revision_id=None):
3611
"""See InterRepository.copy_content()."""
3612
# weave specific optimised path:
3613
try:
3614
self.target.set_make_working_trees(self.source.make_working_trees())
3615
except (errors.RepositoryUpgradeRequired, NotImplemented):
3616
pass
3617
# FIXME do not peek!
3618
if self.source._transport.listable():
3619
pb = ui.ui_factory.nested_progress_bar()
3620
try:
3621
self.target.texts.insert_record_stream(
3622
self.source.texts.get_record_stream(
3623
self.source.texts.keys(), 'topological', False))
3624
pb.update('Copying inventory', 0, 1)
3625
self.target.inventories.insert_record_stream(
3626
self.source.inventories.get_record_stream(
3627
self.source.inventories.keys(), 'topological', False))
3628
self.target.signatures.insert_record_stream(
3629
self.source.signatures.get_record_stream(
3630
self.source.signatures.keys(),
3631
'unordered', True))
3632
self.target.revisions.insert_record_stream(
3633
self.source.revisions.get_record_stream(
3634
self.source.revisions.keys(),
3635
'topological', True))
3636
finally:
3637
pb.finished()
3638
else:
3639
self.target.fetch(self.source, revision_id=revision_id)
3640
3641
@needs_read_lock
3642
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3643
"""See InterRepository.missing_revision_ids()."""
3644
# we want all revisions to satisfy revision_id in source.
3645
# but we don't want to stat every file here and there.
3646
# we want then, all revisions other needs to satisfy revision_id
3647
# checked, but not those that we have locally.
3648
# so the first thing is to get a subset of the revisions to
3649
# satisfy revision_id in source, and then eliminate those that
3650
# we do already have.
3651
# this is slow on high latency connection to self, but as this
3652
# disk format scales terribly for push anyway due to rewriting
3653
# inventory.weave, this is considered acceptable.
3654
# - RBC 20060209
3655
if revision_id is not None:
3656
source_ids = self.source.get_ancestry(revision_id)
3657
if source_ids[0] is not None:
3658
raise AssertionError()
3659
source_ids.pop(0)
3660
else:
3661
source_ids = self.source._all_possible_ids()
3662
source_ids_set = set(source_ids)
3663
# source_ids is the worst possible case we may need to pull.
3664
# now we want to filter source_ids against what we actually
3665
# have in target, but don't try to check for existence where we know
3666
# we do not have a revision as that would be pointless.
3667
target_ids = set(self.target._all_possible_ids())
3668
possibly_present_revisions = target_ids.intersection(source_ids_set)
3669
actually_present_revisions = set(
3670
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3671
required_revisions = source_ids_set.difference(actually_present_revisions)
3672
if revision_id is not None:
3673
# we used get_ancestry to determine source_ids then we are assured all
3674
# revisions referenced are present as they are installed in topological order.
3675
# and the tip revision was validated by get_ancestry.
3676
result_set = required_revisions
3677
else:
3678
# if we just grabbed the possibly available ids, then
3679
# we only have an estimate of whats available and need to validate
3680
# that against the revision records.
3681
result_set = set(
3682
self.source._eliminate_revisions_not_present(required_revisions))
3683
return self.source.revision_ids_to_search_result(result_set)
3684
3685
3686
class InterKnitRepo(InterSameDataRepository):
3687
"""Optimised code paths between Knit based repositories."""
3688
3689
@classmethod
3690
def _get_repo_format_to_test(self):
3691
from bzrlib.repofmt import knitrepo
3692
return knitrepo.RepositoryFormatKnit1()
3693
3694
@staticmethod
3695
def is_compatible(source, target):
3696
"""Be compatible with known Knit formats.
3697
3698
We don't test for the stores being of specific types because that
3699
could lead to confusing results, and there is no need to be
3700
overly general.
3701
"""
3702
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
3703
try:
3704
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
3705
isinstance(target._format, RepositoryFormatKnit))
3706
except AttributeError:
3707
return False
3708
return are_knits and InterRepository._same_model(source, target)
3709
3710
@needs_read_lock
3711
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3712
"""See InterRepository.missing_revision_ids()."""
3713
if revision_id is not None:
3714
source_ids = self.source.get_ancestry(revision_id)
3715
if source_ids[0] is not None:
3716
raise AssertionError()
3717
source_ids.pop(0)
3718
else:
3719
source_ids = self.source.all_revision_ids()
3720
source_ids_set = set(source_ids)
3721
# source_ids is the worst possible case we may need to pull.
3722
# now we want to filter source_ids against what we actually
3723
# have in target, but don't try to check for existence where we know
3724
# we do not have a revision as that would be pointless.
3725
target_ids = set(self.target.all_revision_ids())
3726
possibly_present_revisions = target_ids.intersection(source_ids_set)
3727
actually_present_revisions = set(
3728
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3729
required_revisions = source_ids_set.difference(actually_present_revisions)
3730
if revision_id is not None:
3731
# we used get_ancestry to determine source_ids then we are assured all
3732
# revisions referenced are present as they are installed in topological order.
3733
# and the tip revision was validated by get_ancestry.
3734
result_set = required_revisions
3735
else:
3736
# if we just grabbed the possibly available ids, then
3737
# we only have an estimate of whats available and need to validate
3738
# that against the revision records.
3739
result_set = set(
3740
self.source._eliminate_revisions_not_present(required_revisions))
3741
return self.source.revision_ids_to_search_result(result_set)
3742
3743
3744
class InterDifferingSerializer(InterRepository):
3745
3746
1886
@classmethod
3747
1887
def _get_repo_format_to_test(self):
3748
1888
return None
3749
1889
3750
@staticmethod
3751
def is_compatible(source, target):
3752
"""Be compatible with Knit2 source and Knit3 target"""
3753
# This is redundant with format.check_conversion_target(), however that
3754
# raises an exception, and we just want to say "False" as in we won't
3755
# support converting between these formats.
3756
if 'IDS_never' in debug.debug_flags:
3757
return False
3758
if source.supports_rich_root() and not target.supports_rich_root():
3759
return False
3760
if (source._format.supports_tree_reference
3761
and not target._format.supports_tree_reference):
3762
return False
3763
if target._fallback_repositories and target._format.supports_chks:
3764
# IDS doesn't know how to copy CHKs for the parent inventories it
3765
# adds to stacked repos.
3766
return False
3767
if 'IDS_always' in debug.debug_flags:
3768
return True
3769
# Only use this code path for local source and target. IDS does far
3770
# too much IO (both bandwidth and roundtrips) over a network.
3771
if not source.bzrdir.transport.base.startswith('file:///'):
3772
return False
3773
if not target.bzrdir.transport.base.startswith('file:///'):
3774
return False
1890
@classmethod
1891
def is_compatible(cls, source, target):
1892
# The default implementation is compatible with everything
3775
1893
return True
3776
1894
3777
def _get_trees(self, revision_ids, cache):
3778
possible_trees = []
3779
for rev_id in revision_ids:
3780
if rev_id in cache:
3781
possible_trees.append((rev_id, cache[rev_id]))
3782
else:
3783
# Not cached, but inventory might be present anyway.
3784
try:
3785
tree = self.source.revision_tree(rev_id)
3786
except errors.NoSuchRevision:
3787
# Nope, parent is ghost.
3788
pass
3789
else:
3790
cache[rev_id] = tree
3791
possible_trees.append((rev_id, tree))
3792
return possible_trees
3793
3794
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3795
"""Get the best delta and base for this revision.
3796
3797
:return: (basis_id, delta)
3798
"""
3799
deltas = []
3800
# Generate deltas against each tree, to find the shortest.
3801
texts_possibly_new_in_tree = set()
3802
for basis_id, basis_tree in possible_trees:
3803
delta = tree.inventory._make_delta(basis_tree.inventory)
3804
for old_path, new_path, file_id, new_entry in delta:
3805
if new_path is None:
3806
# This file_id isn't present in the new rev, so we don't
3807
# care about it.
3808
continue
3809
if not new_path:
3810
# Rich roots are handled elsewhere...
3811
continue
3812
kind = new_entry.kind
3813
if kind != 'directory' and kind != 'file':
3814
# No text record associated with this inventory entry.
3815
continue
3816
# This is a directory or file that has changed somehow.
3817
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3818
deltas.append((len(delta), basis_id, delta))
3819
deltas.sort()
3820
return deltas[0][1:]
3821
3822
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3823
"""Find all parent revisions that are absent, but for which the
3824
inventory is present, and copy those inventories.
3825
3826
This is necessary to preserve correctness when the source is stacked
3827
without fallbacks configured. (Note that in cases like upgrade the
3828
source may be not have _fallback_repositories even though it is
3829
stacked.)
3830
"""
3831
parent_revs = set()
3832
for parents in parent_map.values():
3833
parent_revs.update(parents)
3834
present_parents = self.source.get_parent_map(parent_revs)
3835
absent_parents = set(parent_revs).difference(present_parents)
3836
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3837
(rev_id,) for rev_id in absent_parents)
3838
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3839
for parent_tree in self.source.revision_trees(parent_inv_ids):
3840
current_revision_id = parent_tree.get_revision_id()
3841
parents_parents_keys = parent_invs_keys_for_stacking[
3842
(current_revision_id,)]
3843
parents_parents = [key[-1] for key in parents_parents_keys]
3844
basis_id = _mod_revision.NULL_REVISION
3845
basis_tree = self.source.revision_tree(basis_id)
3846
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3847
self.target.add_inventory_by_delta(
3848
basis_id, delta, current_revision_id, parents_parents)
3849
cache[current_revision_id] = parent_tree
3850
3851
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3852
"""Fetch across a few revisions.
3853
3854
:param revision_ids: The revisions to copy
3855
:param basis_id: The revision_id of a tree that must be in cache, used
3856
as a basis for delta when no other base is available
3857
:param cache: A cache of RevisionTrees that we can use.
3858
:param a_graph: A Graph object to determine the heads() of the
3859
rich-root data stream.
3860
:return: The revision_id of the last converted tree. The RevisionTree
3861
for it will be in cache
3862
"""
3863
# Walk though all revisions; get inventory deltas, copy referenced
3864
# texts that delta references, insert the delta, revision and
3865
# signature.
3866
root_keys_to_create = set()
3867
text_keys = set()
3868
pending_deltas = []
3869
pending_revisions = []
3870
parent_map = self.source.get_parent_map(revision_ids)
3871
self._fetch_parent_invs_for_stacking(parent_map, cache)
3872
self.source._safe_to_return_from_cache = True
3873
for tree in self.source.revision_trees(revision_ids):
3874
# Find a inventory delta for this revision.
3875
# Find text entries that need to be copied, too.
3876
current_revision_id = tree.get_revision_id()
3877
parent_ids = parent_map.get(current_revision_id, ())
3878
parent_trees = self._get_trees(parent_ids, cache)
3879
possible_trees = list(parent_trees)
3880
if len(possible_trees) == 0:
3881
# There either aren't any parents, or the parents are ghosts,
3882
# so just use the last converted tree.
3883
possible_trees.append((basis_id, cache[basis_id]))
3884
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3885
possible_trees)
3886
revision = self.source.get_revision(current_revision_id)
3887
pending_deltas.append((basis_id, delta,
3888
current_revision_id, revision.parent_ids))
3889
if self._converting_to_rich_root:
3890
self._revision_id_to_root_id[current_revision_id] = \
3891
tree.get_root_id()
3892
# Determine which texts are in present in this revision but not in
3893
# any of the available parents.
3894
texts_possibly_new_in_tree = set()
3895
for old_path, new_path, file_id, entry in delta:
3896
if new_path is None:
3897
# This file_id isn't present in the new rev
3898
continue
3899
if not new_path:
3900
# This is the root
3901
if not self.target.supports_rich_root():
3902
# The target doesn't support rich root, so we don't
3903
# copy
3904
continue
3905
if self._converting_to_rich_root:
3906
# This can't be copied normally, we have to insert
3907
# it specially
3908
root_keys_to_create.add((file_id, entry.revision))
3909
continue
3910
kind = entry.kind
3911
texts_possibly_new_in_tree.add((file_id, entry.revision))
3912
for basis_id, basis_tree in possible_trees:
3913
basis_inv = basis_tree.inventory
3914
for file_key in list(texts_possibly_new_in_tree):
3915
file_id, file_revision = file_key
3916
try:
3917
entry = basis_inv[file_id]
3918
except errors.NoSuchId:
3919
continue
3920
if entry.revision == file_revision:
3921
texts_possibly_new_in_tree.remove(file_key)
3922
text_keys.update(texts_possibly_new_in_tree)
3923
pending_revisions.append(revision)
3924
cache[current_revision_id] = tree
3925
basis_id = current_revision_id
3926
self.source._safe_to_return_from_cache = False
3927
# Copy file texts
3928
from_texts = self.source.texts
3929
to_texts = self.target.texts
3930
if root_keys_to_create:
3931
root_stream = _mod_fetch._new_root_data_stream(
3932
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3933
self.source, graph=a_graph)
3934
to_texts.insert_record_stream(root_stream)
3935
to_texts.insert_record_stream(from_texts.get_record_stream(
3936
text_keys, self.target._format._fetch_order,
3937
not self.target._format._fetch_uses_deltas))
3938
# insert inventory deltas
3939
for delta in pending_deltas:
3940
self.target.add_inventory_by_delta(*delta)
3941
if self.target._fallback_repositories:
3942
# Make sure this stacked repository has all the parent inventories
3943
# for the new revisions that we are about to insert. We do this
3944
# before adding the revisions so that no revision is added until
3945
# all the inventories it may depend on are added.
3946
# Note that this is overzealous, as we may have fetched these in an
3947
# earlier batch.
3948
parent_ids = set()
3949
revision_ids = set()
3950
for revision in pending_revisions:
3951
revision_ids.add(revision.revision_id)
3952
parent_ids.update(revision.parent_ids)
3953
parent_ids.difference_update(revision_ids)
3954
parent_ids.discard(_mod_revision.NULL_REVISION)
3955
parent_map = self.source.get_parent_map(parent_ids)
3956
# we iterate over parent_map and not parent_ids because we don't
3957
# want to try copying any revision which is a ghost
3958
for parent_tree in self.source.revision_trees(parent_map):
3959
current_revision_id = parent_tree.get_revision_id()
3960
parents_parents = parent_map[current_revision_id]
3961
possible_trees = self._get_trees(parents_parents, cache)
3962
if len(possible_trees) == 0:
3963
# There either aren't any parents, or the parents are
3964
# ghosts, so just use the last converted tree.
3965
possible_trees.append((basis_id, cache[basis_id]))
3966
basis_id, delta = self._get_delta_for_revision(parent_tree,
3967
parents_parents, possible_trees)
3968
self.target.add_inventory_by_delta(
3969
basis_id, delta, current_revision_id, parents_parents)
3970
# insert signatures and revisions
3971
for revision in pending_revisions:
3972
try:
3973
signature = self.source.get_signature_text(
3974
revision.revision_id)
3975
self.target.add_signature_text(revision.revision_id,
3976
signature)
3977
except errors.NoSuchRevision:
3978
pass
3979
self.target.add_revision(revision.revision_id, revision)
3980
return basis_id
3981
3982
def _fetch_all_revisions(self, revision_ids, pb):
3983
"""Fetch everything for the list of revisions.
3984
3985
:param revision_ids: The list of revisions to fetch. Must be in
3986
topological order.
3987
:param pb: A ProgressTask
3988
:return: None
3989
"""
3990
basis_id, basis_tree = self._get_basis(revision_ids[0])
3991
batch_size = 100
3992
cache = lru_cache.LRUCache(100)
3993
cache[basis_id] = basis_tree
3994
del basis_tree # We don't want to hang on to it here
3995
hints = []
3996
if self._converting_to_rich_root and len(revision_ids) > 100:
3997
a_graph = _mod_fetch._get_rich_root_heads_graph(self.source,
3998
revision_ids)
3999
else:
4000
a_graph = None
4001
4002
for offset in range(0, len(revision_ids), batch_size):
4003
self.target.start_write_group()
4004
try:
4005
pb.update('Transferring revisions', offset,
4006
len(revision_ids))
4007
batch = revision_ids[offset:offset+batch_size]
4008
basis_id = self._fetch_batch(batch, basis_id, cache,
4009
a_graph=a_graph)
4010
except:
4011
self.source._safe_to_return_from_cache = False
4012
self.target.abort_write_group()
4013
raise
4014
else:
4015
hint = self.target.commit_write_group()
4016
if hint:
4017
hints.extend(hint)
4018
if hints and self.target._format.pack_compresses:
4019
self.target.pack(hint=hints)
4020
pb.update('Transferring revisions', len(revision_ids),
4021
len(revision_ids))
4022
4023
@needs_write_lock
4024
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
4025
fetch_spec=None):
4026
"""See InterRepository.fetch()."""
4027
if fetch_spec is not None:
4028
raise AssertionError("Not implemented yet...")
4029
ui.ui_factory.warn_experimental_format_fetch(self)
4030
if (not self.source.supports_rich_root()
4031
and self.target.supports_rich_root()):
4032
self._converting_to_rich_root = True
4033
self._revision_id_to_root_id = {}
4034
else:
4035
self._converting_to_rich_root = False
4036
# See <https://launchpad.net/bugs/456077> asking for a warning here
4037
if self.source._format.network_name() != self.target._format.network_name():
4038
ui.ui_factory.show_user_warning('cross_format_fetch',
4039
from_format=self.source._format,
4040
to_format=self.target._format)
4041
revision_ids = self.target.search_missing_revision_ids(self.source,
4042
revision_id, find_ghosts=find_ghosts).get_keys()
4043
if not revision_ids:
4044
return 0, 0
4045
revision_ids = tsort.topo_sort(
4046
self.source.get_graph().get_parent_map(revision_ids))
4047
if not revision_ids:
4048
return 0, 0
4049
# Walk though all revisions; get inventory deltas, copy referenced
4050
# texts that delta references, insert the delta, revision and
4051
# signature.
4052
if pb is None:
4053
my_pb = ui.ui_factory.nested_progress_bar()
4054
pb = my_pb
4055
else:
4056
symbol_versioning.warn(
4057
symbol_versioning.deprecated_in((1, 14, 0))
4058
% "pb parameter to fetch()")
4059
my_pb = None
4060
try:
4061
self._fetch_all_revisions(revision_ids, pb)
4062
finally:
4063
if my_pb is not None:
4064
my_pb.finished()
4065
return len(revision_ids), 0
4066
4067
def _get_basis(self, first_revision_id):
4068
"""Get a revision and tree which exists in the target.
4069
4070
This assumes that first_revision_id is selected for transmission
4071
because all other ancestors are already present. If we can't find an
4072
ancestor we fall back to NULL_REVISION since we know that is safe.
4073
4074
:return: (basis_id, basis_tree)
4075
"""
4076
first_rev = self.source.get_revision(first_revision_id)
4077
try:
4078
basis_id = first_rev.parent_ids[0]
4079
# only valid as a basis if the target has it
4080
self.target.get_revision(basis_id)
4081
# Try to get a basis tree - if its a ghost it will hit the
4082
# NoSuchRevision case.
4083
basis_tree = self.source.revision_tree(basis_id)
4084
except (IndexError, errors.NoSuchRevision):
4085
basis_id = _mod_revision.NULL_REVISION
4086
basis_tree = self.source.revision_tree(basis_id)
4087
return basis_id, basis_tree
4088
4089
4090
InterRepository.register_optimiser(InterDifferingSerializer)
4091
InterRepository.register_optimiser(InterSameDataRepository)
4092
InterRepository.register_optimiser(InterWeaveRepo)
4093
InterRepository.register_optimiser(InterKnitRepo)
1895
1896
InterRepository.register_optimiser(InterRepository)
4094
1897
4095
1898
4096
1899
class CopyConverter(object):
4141
1944
pb.finished()
4142
1945
4143
1946
4144
_unescape_map = {
4145
'apos':"'",
4146
'quot':'"',
4147
'amp':'&',
4148
'lt':'<',
4149
'gt':'>'
4150
}
4151
4152
4153
def _unescaper(match, _map=_unescape_map):
4154
code = match.group(1)
4155
try:
4156
return _map[code]
4157
except KeyError:
4158
if not code.startswith('#'):
4159
raise
4160
return unichr(int(code[1:])).encode('utf8')
4161
4162
4163
_unescape_re = None
4164
4165
4166
def _unescape_xml(data):
4167
"""Unescape predefined XML entities in a string of data."""
4168
global _unescape_re
4169
if _unescape_re is None:
4170
_unescape_re = re.compile('\&([^;]*);')
4171
return _unescape_re.sub(_unescaper, data)
4172
4173
4174
class _VersionedFileChecker(object):
4175
4176
def __init__(self, repository, text_key_references=None, ancestors=None):
4177
self.repository = repository
4178
self.text_index = self.repository._generate_text_key_index(
4179
text_key_references=text_key_references, ancestors=ancestors)
4180
4181
def calculate_file_version_parents(self, text_key):
4182
"""Calculate the correct parents for a file version according to
4183
the inventories.
4184
"""
4185
parent_keys = self.text_index[text_key]
4186
if parent_keys == [_mod_revision.NULL_REVISION]:
4187
return ()
4188
return tuple(parent_keys)
4189
4190
def check_file_version_parents(self, texts, progress_bar=None):
4191
"""Check the parents stored in a versioned file are correct.
4192
4193
It also detects file versions that are not referenced by their
4194
corresponding revision's inventory.
4195
4196
:returns: A tuple of (wrong_parents, dangling_file_versions).
4197
wrong_parents is a dict mapping {revision_id: (stored_parents,
4198
correct_parents)} for each revision_id where the stored parents
4199
are not correct. dangling_file_versions is a set of (file_id,
4200
revision_id) tuples for versions that are present in this versioned
4201
file, but not used by the corresponding inventory.
4202
"""
4203
local_progress = None
4204
if progress_bar is None:
4205
local_progress = ui.ui_factory.nested_progress_bar()
4206
progress_bar = local_progress
4207
try:
4208
return self._check_file_version_parents(texts, progress_bar)
4209
finally:
4210
if local_progress:
4211
local_progress.finished()
4212
4213
def _check_file_version_parents(self, texts, progress_bar):
4214
"""See check_file_version_parents."""
4215
wrong_parents = {}
4216
self.file_ids = set([file_id for file_id, _ in
4217
self.text_index.iterkeys()])
4218
# text keys is now grouped by file_id
4219
n_versions = len(self.text_index)
4220
progress_bar.update('loading text store', 0, n_versions)
4221
parent_map = self.repository.texts.get_parent_map(self.text_index)
4222
# On unlistable transports this could well be empty/error...
4223
text_keys = self.repository.texts.keys()
4224
unused_keys = frozenset(text_keys) - set(self.text_index)
4225
for num, key in enumerate(self.text_index.iterkeys()):
4226
progress_bar.update('checking text graph', num, n_versions)
4227
correct_parents = self.calculate_file_version_parents(key)
4228
try:
4229
knit_parents = parent_map[key]
4230
except errors.RevisionNotPresent:
4231
# Missing text!
4232
knit_parents = None
4233
if correct_parents != knit_parents:
4234
wrong_parents[key] = (knit_parents, correct_parents)
4235
return wrong_parents, unused_keys
4236
4237
4238
def _old_get_graph(repository, revision_id):
4239
"""DO NOT USE. That is all. I'm serious."""
4240
graph = repository.get_graph()
4241
revision_graph = dict(((key, value) for key, value in
4242
graph.iter_ancestry([revision_id]) if value is not None))
4243
return _strip_NULL_ghosts(revision_graph)
4244
4245
4246
1947
def _strip_NULL_ghosts(revision_graph):
4247
1948
"""Also don't use this. more compatibility code for unmigrated clients."""
4248
1949
# Filter ghosts, and null:
4254
1955
return revision_graph
4255
1956
4256
1957
4257
class StreamSink(object):
4258
"""An object that can insert a stream into a repository.
4259
4260
This interface handles the complexity of reserialising inventories and
4261
revisions from different formats, and allows unidirectional insertion into
4262
stacked repositories without looking for the missing basis parents
4263
beforehand.
4264
"""
4265
4266
def __init__(self, target_repo):
4267
self.target_repo = target_repo
4268
4269
def insert_stream(self, stream, src_format, resume_tokens):
4270
"""Insert a stream's content into the target repository.
4271
4272
:param src_format: a bzr repository format.
4273
4274
:return: a list of resume tokens and an iterable of keys additional
4275
items required before the insertion can be completed.
4276
"""
4277
self.target_repo.lock_write()
4278
try:
4279
if resume_tokens:
4280
self.target_repo.resume_write_group(resume_tokens)
4281
is_resume = True
4282
else:
4283
self.target_repo.start_write_group()
4284
is_resume = False
4285
try:
4286
# locked_insert_stream performs a commit|suspend.
4287
return self._locked_insert_stream(stream, src_format,
4288
is_resume)
4289
except:
4290
self.target_repo.abort_write_group(suppress_errors=True)
4291
raise
4292
finally:
4293
self.target_repo.unlock()
4294
4295
def _locked_insert_stream(self, stream, src_format, is_resume):
4296
to_serializer = self.target_repo._format._serializer
4297
src_serializer = src_format._serializer
4298
new_pack = None
4299
if to_serializer == src_serializer:
4300
# If serializers match and the target is a pack repository, set the
4301
# write cache size on the new pack. This avoids poor performance
4302
# on transports where append is unbuffered (such as
4303
# RemoteTransport). This is safe to do because nothing should read
4304
# back from the target repository while a stream with matching
4305
# serialization is being inserted.
4306
# The exception is that a delta record from the source that should
4307
# be a fulltext may need to be expanded by the target (see
4308
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4309
# explicitly flush any buffered writes first in that rare case.
4310
try:
4311
new_pack = self.target_repo._pack_collection._new_pack
4312
except AttributeError:
4313
# Not a pack repository
4314
pass
4315
else:
4316
new_pack.set_write_cache_size(1024*1024)
4317
for substream_type, substream in stream:
4318
if 'stream' in debug.debug_flags:
4319
mutter('inserting substream: %s', substream_type)
4320
if substream_type == 'texts':
4321
self.target_repo.texts.insert_record_stream(substream)
4322
elif substream_type == 'inventories':
4323
if src_serializer == to_serializer:
4324
self.target_repo.inventories.insert_record_stream(
4325
substream)
4326
else:
4327
self._extract_and_insert_inventories(
4328
substream, src_serializer)
4329
elif substream_type == 'inventory-deltas':
4330
self._extract_and_insert_inventory_deltas(
4331
substream, src_serializer)
4332
elif substream_type == 'chk_bytes':
4333
# XXX: This doesn't support conversions, as it assumes the
4334
# conversion was done in the fetch code.
4335
self.target_repo.chk_bytes.insert_record_stream(substream)
4336
elif substream_type == 'revisions':
4337
# This may fallback to extract-and-insert more often than
4338
# required if the serializers are different only in terms of
4339
# the inventory.
4340
if src_serializer == to_serializer:
4341
self.target_repo.revisions.insert_record_stream(substream)
4342
else:
4343
self._extract_and_insert_revisions(substream,
4344
src_serializer)
4345
elif substream_type == 'signatures':
4346
self.target_repo.signatures.insert_record_stream(substream)
4347
else:
4348
raise AssertionError('kaboom! %s' % (substream_type,))
4349
# Done inserting data, and the missing_keys calculations will try to
4350
# read back from the inserted data, so flush the writes to the new pack
4351
# (if this is pack format).
4352
if new_pack is not None:
4353
new_pack._write_data('', flush=True)
4354
# Find all the new revisions (including ones from resume_tokens)
4355
missing_keys = self.target_repo.get_missing_parent_inventories(
4356
check_for_missing_texts=is_resume)
4357
try:
4358
for prefix, versioned_file in (
4359
('texts', self.target_repo.texts),
4360
('inventories', self.target_repo.inventories),
4361
('revisions', self.target_repo.revisions),
4362
('signatures', self.target_repo.signatures),
4363
('chk_bytes', self.target_repo.chk_bytes),
4364
):
4365
if versioned_file is None:
4366
continue
4367
# TODO: key is often going to be a StaticTuple object
4368
# I don't believe we can define a method by which
4369
# (prefix,) + StaticTuple will work, though we could
4370
# define a StaticTuple.sq_concat that would allow you to
4371
# pass in either a tuple or a StaticTuple as the second
4372
# object, so instead we could have:
4373
# StaticTuple(prefix) + key here...
4374
missing_keys.update((prefix,) + key for key in
4375
versioned_file.get_missing_compression_parent_keys())
4376
except NotImplementedError:
4377
# cannot even attempt suspending, and missing would have failed
4378
# during stream insertion.
4379
missing_keys = set()
4380
else:
4381
if missing_keys:
4382
# suspend the write group and tell the caller what we is
4383
# missing. We know we can suspend or else we would not have
4384
# entered this code path. (All repositories that can handle
4385
# missing keys can handle suspending a write group).
4386
write_group_tokens = self.target_repo.suspend_write_group()
4387
return write_group_tokens, missing_keys
4388
hint = self.target_repo.commit_write_group()
4389
if (to_serializer != src_serializer and
4390
self.target_repo._format.pack_compresses):
4391
self.target_repo.pack(hint=hint)
4392
return [], set()
4393
4394
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4395
target_rich_root = self.target_repo._format.rich_root_data
4396
target_tree_refs = self.target_repo._format.supports_tree_reference
4397
for record in substream:
4398
# Insert the delta directly
4399
inventory_delta_bytes = record.get_bytes_as('fulltext')
4400
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4401
try:
4402
parse_result = deserialiser.parse_text_bytes(
4403
inventory_delta_bytes)
4404
except inventory_delta.IncompatibleInventoryDelta, err:
4405
trace.mutter("Incompatible delta: %s", err.msg)
4406
raise errors.IncompatibleRevision(self.target_repo._format)
4407
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4408
revision_id = new_id
4409
parents = [key[0] for key in record.parents]
4410
self.target_repo.add_inventory_by_delta(
4411
basis_id, inv_delta, revision_id, parents)
4412
4413
def _extract_and_insert_inventories(self, substream, serializer,
4414
parse_delta=None):
4415
"""Generate a new inventory versionedfile in target, converting data.
4416
4417
The inventory is retrieved from the source, (deserializing it), and
4418
stored in the target (reserializing it in a different format).
4419
"""
4420
target_rich_root = self.target_repo._format.rich_root_data
4421
target_tree_refs = self.target_repo._format.supports_tree_reference
4422
for record in substream:
4423
# It's not a delta, so it must be a fulltext in the source
4424
# serializer's format.
4425
bytes = record.get_bytes_as('fulltext')
4426
revision_id = record.key[0]
4427
inv = serializer.read_inventory_from_string(bytes, revision_id)
4428
parents = [key[0] for key in record.parents]
4429
self.target_repo.add_inventory(revision_id, inv, parents)
4430
# No need to keep holding this full inv in memory when the rest of
4431
# the substream is likely to be all deltas.
4432
del inv
4433
4434
def _extract_and_insert_revisions(self, substream, serializer):
4435
for record in substream:
4436
bytes = record.get_bytes_as('fulltext')
4437
revision_id = record.key[0]
4438
rev = serializer.read_revision_from_string(bytes)
4439
if rev.revision_id != revision_id:
4440
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4441
self.target_repo.add_revision(revision_id, rev)
4442
4443
def finished(self):
4444
if self.target_repo._format._fetch_reconcile:
4445
self.target_repo.reconcile()
4446
4447
4448
class StreamSource(object):
4449
"""A source of a stream for fetching between repositories."""
4450
4451
def __init__(self, from_repository, to_format):
4452
"""Create a StreamSource streaming from from_repository."""
4453
self.from_repository = from_repository
4454
self.to_format = to_format
4455
self._record_counter = RecordCounter()
4456
4457
def delta_on_metadata(self):
4458
"""Return True if delta's are permitted on metadata streams.
4459
4460
That is on revisions and signatures.
4461
"""
4462
src_serializer = self.from_repository._format._serializer
4463
target_serializer = self.to_format._serializer
4464
return (self.to_format._fetch_uses_deltas and
4465
src_serializer == target_serializer)
4466
4467
def _fetch_revision_texts(self, revs):
4468
# fetch signatures first and then the revision texts
4469
# may need to be a InterRevisionStore call here.
4470
from_sf = self.from_repository.signatures
4471
# A missing signature is just skipped.
4472
keys = [(rev_id,) for rev_id in revs]
4473
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4474
keys,
4475
self.to_format._fetch_order,
4476
not self.to_format._fetch_uses_deltas))
4477
# If a revision has a delta, this is actually expanded inside the
4478
# insert_record_stream code now, which is an alternate fix for
4479
# bug #261339
4480
from_rf = self.from_repository.revisions
4481
revisions = from_rf.get_record_stream(
4482
keys,
4483
self.to_format._fetch_order,
4484
not self.delta_on_metadata())
4485
return [('signatures', signatures), ('revisions', revisions)]
4486
4487
def _generate_root_texts(self, revs):
4488
"""This will be called by get_stream between fetching weave texts and
4489
fetching the inventory weave.
4490
"""
4491
if self._rich_root_upgrade():
4492
return _mod_fetch.Inter1and2Helper(
4493
self.from_repository).generate_root_texts(revs)
4494
else:
4495
return []
4496
4497
def get_stream(self, search):
4498
phase = 'file'
4499
revs = search.get_keys()
4500
graph = self.from_repository.get_graph()
4501
revs = tsort.topo_sort(graph.get_parent_map(revs))
4502
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4503
text_keys = []
4504
for knit_kind, file_id, revisions in data_to_fetch:
4505
if knit_kind != phase:
4506
phase = knit_kind
4507
# Make a new progress bar for this phase
4508
if knit_kind == "file":
4509
# Accumulate file texts
4510
text_keys.extend([(file_id, revision) for revision in
4511
revisions])
4512
elif knit_kind == "inventory":
4513
# Now copy the file texts.
4514
from_texts = self.from_repository.texts
4515
yield ('texts', from_texts.get_record_stream(
4516
text_keys, self.to_format._fetch_order,
4517
not self.to_format._fetch_uses_deltas))
4518
# Cause an error if a text occurs after we have done the
4519
# copy.
4520
text_keys = None
4521
# Before we process the inventory we generate the root
4522
# texts (if necessary) so that the inventories references
4523
# will be valid.
4524
for _ in self._generate_root_texts(revs):
4525
yield _
4526
# we fetch only the referenced inventories because we do not
4527
# know for unselected inventories whether all their required
4528
# texts are present in the other repository - it could be
4529
# corrupt.
4530
for info in self._get_inventory_stream(revs):
4531
yield info
4532
elif knit_kind == "signatures":
4533
# Nothing to do here; this will be taken care of when
4534
# _fetch_revision_texts happens.
4535
pass
4536
elif knit_kind == "revisions":
4537
for record in self._fetch_revision_texts(revs):
4538
yield record
4539
else:
4540
raise AssertionError("Unknown knit kind %r" % knit_kind)
4541
4542
def get_stream_for_missing_keys(self, missing_keys):
4543
# missing keys can only occur when we are byte copying and not
4544
# translating (because translation means we don't send
4545
# unreconstructable deltas ever).
4546
keys = {}
4547
keys['texts'] = set()
4548
keys['revisions'] = set()
4549
keys['inventories'] = set()
4550
keys['chk_bytes'] = set()
4551
keys['signatures'] = set()
4552
for key in missing_keys:
4553
keys[key[0]].add(key[1:])
4554
if len(keys['revisions']):
4555
# If we allowed copying revisions at this point, we could end up
4556
# copying a revision without copying its required texts: a
4557
# violation of the requirements for repository integrity.
4558
raise AssertionError(
4559
'cannot copy revisions to fill in missing deltas %s' % (
4560
keys['revisions'],))
4561
for substream_kind, keys in keys.iteritems():
4562
vf = getattr(self.from_repository, substream_kind)
4563
if vf is None and keys:
4564
raise AssertionError(
4565
"cannot fill in keys for a versioned file we don't"
4566
" have: %s needs %s" % (substream_kind, keys))
4567
if not keys:
4568
# No need to stream something we don't have
4569
continue
4570
if substream_kind == 'inventories':
4571
# Some missing keys are genuinely ghosts, filter those out.
4572
present = self.from_repository.inventories.get_parent_map(keys)
4573
revs = [key[0] for key in present]
4574
# Get the inventory stream more-or-less as we do for the
4575
# original stream; there's no reason to assume that records
4576
# direct from the source will be suitable for the sink. (Think
4577
# e.g. 2a -> 1.9-rich-root).
4578
for info in self._get_inventory_stream(revs, missing=True):
4579
yield info
4580
continue
4581
4582
# Ask for full texts always so that we don't need more round trips
4583
# after this stream.
4584
# Some of the missing keys are genuinely ghosts, so filter absent
4585
# records. The Sink is responsible for doing another check to
4586
# ensure that ghosts don't introduce missing data for future
4587
# fetches.
4588
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4589
self.to_format._fetch_order, True))
4590
yield substream_kind, stream
4591
4592
def inventory_fetch_order(self):
4593
if self._rich_root_upgrade():
4594
return 'topological'
4595
else:
4596
return self.to_format._fetch_order
4597
4598
def _rich_root_upgrade(self):
4599
return (not self.from_repository._format.rich_root_data and
4600
self.to_format.rich_root_data)
4601
4602
def _get_inventory_stream(self, revision_ids, missing=False):
4603
from_format = self.from_repository._format
4604
if (from_format.supports_chks and self.to_format.supports_chks and
4605
from_format.network_name() == self.to_format.network_name()):
4606
raise AssertionError(
4607
"this case should be handled by GroupCHKStreamSource")
4608
elif 'forceinvdeltas' in debug.debug_flags:
4609
return self._get_convertable_inventory_stream(revision_ids,
4610
delta_versus_null=missing)
4611
elif from_format.network_name() == self.to_format.network_name():
4612
# Same format.
4613
return self._get_simple_inventory_stream(revision_ids,
4614
missing=missing)
4615
elif (not from_format.supports_chks and not self.to_format.supports_chks
4616
and from_format._serializer == self.to_format._serializer):
4617
# Essentially the same format.
4618
return self._get_simple_inventory_stream(revision_ids,
4619
missing=missing)
4620
else:
4621
# Any time we switch serializations, we want to use an
4622
# inventory-delta based approach.
4623
return self._get_convertable_inventory_stream(revision_ids,
4624
delta_versus_null=missing)
4625
4626
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4627
# NB: This currently reopens the inventory weave in source;
4628
# using a single stream interface instead would avoid this.
4629
from_weave = self.from_repository.inventories
4630
if missing:
4631
delta_closure = True
4632
else:
4633
delta_closure = not self.delta_on_metadata()
4634
yield ('inventories', from_weave.get_record_stream(
4635
[(rev_id,) for rev_id in revision_ids],
4636
self.inventory_fetch_order(), delta_closure))
4637
4638
def _get_convertable_inventory_stream(self, revision_ids,
4639
delta_versus_null=False):
4640
# The two formats are sufficiently different that there is no fast
4641
# path, so we need to send just inventorydeltas, which any
4642
# sufficiently modern client can insert into any repository.
4643
# The StreamSink code expects to be able to
4644
# convert on the target, so we need to put bytes-on-the-wire that can
4645
# be converted. That means inventory deltas (if the remote is <1.19,
4646
# RemoteStreamSink will fallback to VFS to insert the deltas).
4647
yield ('inventory-deltas',
4648
self._stream_invs_as_deltas(revision_ids,
4649
delta_versus_null=delta_versus_null))
4650
4651
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4652
"""Return a stream of inventory-deltas for the given rev ids.
4653
4654
:param revision_ids: The list of inventories to transmit
4655
:param delta_versus_null: Don't try to find a minimal delta for this
4656
entry, instead compute the delta versus the NULL_REVISION. This
4657
effectively streams a complete inventory. Used for stuff like
4658
filling in missing parents, etc.
4659
"""
4660
from_repo = self.from_repository
4661
revision_keys = [(rev_id,) for rev_id in revision_ids]
4662
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4663
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4664
# method...
4665
inventories = self.from_repository.iter_inventories(
4666
revision_ids, 'topological')
4667
format = from_repo._format
4668
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4669
inventory_cache = lru_cache.LRUCache(50)
4670
null_inventory = from_repo.revision_tree(
4671
_mod_revision.NULL_REVISION).inventory
4672
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4673
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4674
# repo back into a non-rich-root repo ought to be allowed)
4675
serializer = inventory_delta.InventoryDeltaSerializer(
4676
versioned_root=format.rich_root_data,
4677
tree_references=format.supports_tree_reference)
4678
for inv in inventories:
4679
key = (inv.revision_id,)
4680
parent_keys = parent_map.get(key, ())
4681
delta = None
4682
if not delta_versus_null and parent_keys:
4683
# The caller did not ask for complete inventories and we have
4684
# some parents that we can delta against. Make a delta against
4685
# each parent so that we can find the smallest.
4686
parent_ids = [parent_key[0] for parent_key in parent_keys]
4687
for parent_id in parent_ids:
4688
if parent_id not in invs_sent_so_far:
4689
# We don't know that the remote side has this basis, so
4690
# we can't use it.
4691
continue
4692
if parent_id == _mod_revision.NULL_REVISION:
4693
parent_inv = null_inventory
4694
else:
4695
parent_inv = inventory_cache.get(parent_id, None)
4696
if parent_inv is None:
4697
parent_inv = from_repo.get_inventory(parent_id)
4698
candidate_delta = inv._make_delta(parent_inv)
4699
if (delta is None or
4700
len(delta) > len(candidate_delta)):
4701
delta = candidate_delta
4702
basis_id = parent_id
4703
if delta is None:
4704
# Either none of the parents ended up being suitable, or we
4705
# were asked to delta against NULL
4706
basis_id = _mod_revision.NULL_REVISION
4707
delta = inv._make_delta(null_inventory)
4708
invs_sent_so_far.add(inv.revision_id)
4709
inventory_cache[inv.revision_id] = inv
4710
delta_serialized = ''.join(
4711
serializer.delta_to_lines(basis_id, key[-1], delta))
4712
yield versionedfile.FulltextContentFactory(
4713
key, parent_keys, None, delta_serialized)
4714
4715
4716
1958
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4717
1959
stop_revision=None):
4718
1960
"""Extend the partial history to include a given index
4744
1986
# No more history
4745
1987
return
4746
1988
1989
1990
class _LazyListJoin(object):
1991
"""An iterable yielding the contents of many lists as one list.
1992
1993
Each iterator made from this will reflect the current contents of the lists
1994
at the time the iterator is made.
1995
1996
This is used by Repository's _make_parents_provider implementation so that
1997
it is safe to do::
1998
1999
pp = repo._make_parents_provider() # uses a list of fallback repos
2000
pp.add_fallback_repository(other_repo) # appends to that list
2001
result = pp.get_parent_map(...)
2002
# The result will include revs from other_repo
2003
"""
2004
2005
def __init__(self, *list_parts):
2006
self.list_parts = list_parts
2007
2008
def __iter__(self):
2009
full_list = []
2010
for list_part in self.list_parts:
2011
full_list.extend(list_part)
2012
return iter(full_list)
Loggerhead is a web-based interface for Breezy
Version: 2.0.1