« back to all changes in this revision
Viewing changes to bzrlib/repository.py
- browse files at revision 6379.8.2
- compare with another revision
- download diff
- download tarball
- view history from revision 6379.8.2
- Committer: Jelmer Vernooij
- Date: 2011-12-16 16:40:10 UTC
- mto: This revision was merged to the branch mainline in revision 6391.
- Revision ID: jelmer@samba.org-20111216164010-z3hy00xrnclnkf7a
Update tests.
- files added:
- bzrlib/bzr_distutils.py
- bzrlib/plugins/changelog_merge
- bzrlib/plugins/changelog_merge/tests
- bzrlib/plugins/po_merge
- bzrlib/plugins/po_merge/tests
- bzrlib/plugins/weave_fmt
- bzrlib/plugins/weave_fmt/tests
- bzrlib/tests/per_repository_vf
- po
- files removed:
- bzrlib/deprecated_graph.py
- files renamed:
- bzrlib/sign_my_commits.py => bzrlib/commit_signature_commands.py
- bzrlib/repofmt/weaverepo.py => bzrlib/plugins/weave_fmt/repository.py
- bzrlib/xml4.py => bzrlib/plugins/weave_fmt/xml4.py
- bzrlib/tests/per_repository/test_is_write_locked.py => bzrlib/tests/per_repository/test_locking.py
- bzrlib/tests/per_repository/helpers.py => bzrlib/tests/per_repository_vf/helpers.py
- bzrlib/tests/per_repository/test__generate_text_key_index.py => bzrlib/tests/per_repository_vf/test__generate_text_key_index.py
- bzrlib/tests/per_repository/test_add_inventory_by_delta.py => bzrlib/tests/per_repository_vf/test_add_inventory_by_delta.py
- bzrlib/tests/per_repository/test_check_reconcile.py => bzrlib/tests/per_repository_vf/test_check_reconcile.py
- bzrlib/tests/per_repository/test_find_text_key_references.py => bzrlib/tests/per_repository_vf/test_find_text_key_references.py
- bzrlib/tests/per_repository/test_merge_directive.py => bzrlib/tests/per_repository_vf/test_merge_directive.py
- files modified:
- .testr.conf
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
19
import itertools
21
20
import time
22
21
23
22
from bzrlib import (
24
23
bzrdir,
25
check,
26
chk_map,
27
24
config,
28
25
controldir,
29
26
debug,
30
fetch as _mod_fetch,
31
fifo_cache,
32
27
generate_ids,
33
gpg,
34
28
graph,
35
inventory,
36
inventory_delta,
37
lazy_regex,
38
29
lockable_files,
39
30
lockdir,
40
lru_cache,
41
31
osutils,
42
pyutils,
43
32
revision as _mod_revision,
44
static_tuple,
45
symbol_versioning,
46
trace,
33
testament as _mod_testament,
47
34
tsort,
48
versionedfile,
35
gpg,
49
36
)
50
37
from bzrlib.bundle import serializer
51
from bzrlib.revisiontree import RevisionTree
52
from bzrlib.store.versioned import VersionedFileStore
53
from bzrlib.testament import Testament
38
from bzrlib.i18n import gettext
54
39
""")
55
40
56
import sys
57
41
from bzrlib import (
58
42
errors,
59
43
registry,
44
symbol_versioning,
60
45
ui,
61
46
)
62
47
from bzrlib.decorators import needs_read_lock, needs_write_lock, only_raises
63
48
from bzrlib.inter import InterObject
64
from bzrlib.inventory import (
65
Inventory,
66
InventoryDirectory,
67
ROOT_ID,
68
entry_factory,
69
)
70
from bzrlib.recordcounter import RecordCounter
71
49
from bzrlib.lock import _RelockDebugMixin, LogicalLockResult
72
50
from bzrlib.trace import (
73
51
log_exception_quietly, note, mutter, mutter_callsite, warning)
94
72
95
73
# all clients should supply tree roots.
96
74
record_root_entry = True
97
# the default CommitBuilder does not manage trees whose root is versioned.
98
_versioned_root = False
75
# whether this commit builder supports the record_entry_contents interface
76
supports_record_entry_contents = False
77
# whether this commit builder will automatically update the branch that is
78
# being committed to
79
updates_branch = False
99
80
100
def __init__(self, repository, parents, config, timestamp=None,
81
def __init__(self, repository, parents, config_stack, timestamp=None,
101
82
timezone=None, committer=None, revprops=None,
102
revision_id=None):
83
revision_id=None, lossy=False):
103
84
"""Initiate a CommitBuilder.
104
85
105
86
:param repository: Repository to commit to.
106
87
:param parents: Revision ids of the parents of the new revision.
107
:param config: Configuration to use.
108
88
:param timestamp: Optional timestamp recorded for commit.
109
89
:param timezone: Optional timezone for timestamp.
110
90
:param committer: Optional committer to set for commit.
111
91
:param revprops: Optional dictionary of revision properties.
112
92
:param revision_id: Optional revision id.
93
:param lossy: Whether to discard data that can not be natively
94
represented, when pushing to a foreign VCS
113
95
"""
114
self._config = config
96
self._config_stack = config_stack
97
self._lossy = lossy
115
98
116
99
if committer is None:
117
self._committer = self._config.username()
100
self._committer = self._config_stack.get('email')
118
101
elif not isinstance(committer, unicode):
119
102
self._committer = committer.decode() # throw if non-ascii
120
103
else:
121
104
self._committer = committer
122
105
123
self.new_inventory = Inventory(None)
124
106
self._new_revision_id = revision_id
125
107
self.parents = parents
126
108
self.repository = repository
141
123
self._timezone = int(timezone)
142
124
143
125
self._generate_revision_if_needed()
144
self.__heads = graph.HeadsCache(repository.get_graph()).heads
145
self._basis_delta = []
146
# API compatibility, older code that used CommitBuilder did not call
147
# .record_delete(), which means the delta that is computed would not be
148
# valid. Callers that will call record_delete() should call
149
# .will_record_deletes() to indicate that.
150
self._recording_deletes = False
151
# memo'd check for no-op commits.
152
self._any_changes = False
153
126
154
127
def any_changes(self):
155
128
"""Return True if any entries were changed.
156
129
157
130
This includes merge-only changes. It is the core for the --unchanged
158
131
detection in commit.
159
132
160
133
:return: True if any changes have occured.
161
134
"""
162
return self._any_changes
135
raise NotImplementedError(self.any_changes)
163
136
164
137
def _validate_unicode_text(self, text, context):
165
138
"""Verify things like commit messages don't have bogus characters."""
181
154
182
155
:return: The revision id of the recorded revision.
183
156
"""
184
self._validate_unicode_text(message, 'commit message')
185
rev = _mod_revision.Revision(
186
timestamp=self._timestamp,
187
timezone=self._timezone,
188
committer=self._committer,
189
message=message,
190
inventory_sha1=self.inv_sha1,
191
revision_id=self._new_revision_id,
192
properties=self._revprops)
193
rev.parent_ids = self.parents
194
self.repository.add_revision(self._new_revision_id, rev,
195
self.new_inventory, self._config)
196
self.repository.commit_write_group()
197
return self._new_revision_id
157
raise NotImplementedError(self.commit)
198
158
199
159
def abort(self):
200
160
"""Abort the commit that is being built.
201
161
"""
202
self.repository.abort_write_group()
162
raise NotImplementedError(self.abort)
203
163
204
164
def revision_tree(self):
205
165
"""Return the tree that was just committed.
206
166
207
After calling commit() this can be called to get a RevisionTree
208
representing the newly committed tree. This is preferred to
209
calling Repository.revision_tree() because that may require
210
deserializing the inventory, while we already have a copy in
167
After calling commit() this can be called to get a
168
RevisionTree representing the newly committed tree. This is
169
preferred to calling Repository.revision_tree() because that may
170
require deserializing the inventory, while we already have a copy in
211
171
memory.
212
172
"""
213
if self.new_inventory is None:
214
self.new_inventory = self.repository.get_inventory(
215
self._new_revision_id)
216
return RevisionTree(self.repository, self.new_inventory,
217
self._new_revision_id)
173
raise NotImplementedError(self.revision_tree)
218
174
219
175
def finish_inventory(self):
220
176
"""Tell the builder that the inventory is finished.
222
178
:return: The inventory id in the repository, which can be used with
223
179
repository.get_inventory.
224
180
"""
225
if self.new_inventory is None:
226
# an inventory delta was accumulated without creating a new
227
# inventory.
228
basis_id = self.basis_delta_revision
229
# We ignore the 'inventory' returned by add_inventory_by_delta
230
# because self.new_inventory is used to hint to the rest of the
231
# system what code path was taken
232
self.inv_sha1, _ = self.repository.add_inventory_by_delta(
233
basis_id, self._basis_delta, self._new_revision_id,
234
self.parents)
235
else:
236
if self.new_inventory.root is None:
237
raise AssertionError('Root entry should be supplied to'
238
' record_entry_contents, as of bzr 0.10.')
239
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
240
self.new_inventory.revision_id = self._new_revision_id
241
self.inv_sha1 = self.repository.add_inventory(
242
self._new_revision_id,
243
self.new_inventory,
244
self.parents
245
)
246
return self._new_revision_id
181
raise NotImplementedError(self.finish_inventory)
247
182
248
183
def _gen_revision_id(self):
249
184
"""Return new revision-id."""
264
199
else:
265
200
self.random_revid = False
266
201
267
def _heads(self, file_id, revision_ids):
268
"""Calculate the graph heads for revision_ids in the graph of file_id.
269
270
This can use either a per-file graph or a global revision graph as we
271
have an identity relationship between the two graphs.
272
"""
273
return self.__heads(revision_ids)
274
275
def _check_root(self, ie, parent_invs, tree):
276
"""Helper for record_entry_contents.
277
278
:param ie: An entry being added.
279
:param parent_invs: The inventories of the parent revisions of the
280
commit.
281
:param tree: The tree that is being committed.
282
"""
283
# In this revision format, root entries have no knit or weave When
284
# serializing out to disk and back in root.revision is always
285
# _new_revision_id
286
ie.revision = self._new_revision_id
287
288
def _require_root_change(self, tree):
289
"""Enforce an appropriate root object change.
290
291
This is called once when record_iter_changes is called, if and only if
292
the root was not in the delta calculated by record_iter_changes.
293
294
:param tree: The tree which is being committed.
295
"""
296
if len(self.parents) == 0:
297
raise errors.RootMissing()
298
entry = entry_factory['directory'](tree.path2id(''), '',
299
None)
300
entry.revision = self._new_revision_id
301
self._basis_delta.append(('', '', entry.file_id, entry))
302
303
def _get_delta(self, ie, basis_inv, path):
304
"""Get a delta against the basis inventory for ie."""
305
if ie.file_id not in basis_inv:
306
# add
307
result = (None, path, ie.file_id, ie)
308
self._basis_delta.append(result)
309
return result
310
elif ie != basis_inv[ie.file_id]:
311
# common but altered
312
# TODO: avoid tis id2path call.
313
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
314
self._basis_delta.append(result)
315
return result
316
else:
317
# common, unaltered
318
return None
319
320
def get_basis_delta(self):
321
"""Return the complete inventory delta versus the basis inventory.
322
323
This has been built up with the calls to record_delete and
324
record_entry_contents. The client must have already called
325
will_record_deletes() to indicate that they will be generating a
326
complete delta.
327
328
:return: An inventory delta, suitable for use with apply_delta, or
329
Repository.add_inventory_by_delta, etc.
330
"""
331
if not self._recording_deletes:
332
raise AssertionError("recording deletes not activated.")
333
return self._basis_delta
334
335
def record_delete(self, path, file_id):
336
"""Record that a delete occured against a basis tree.
337
338
This is an optional API - when used it adds items to the basis_delta
339
being accumulated by the commit builder. It cannot be called unless the
340
method will_record_deletes() has been called to inform the builder that
341
a delta is being supplied.
342
343
:param path: The path of the thing deleted.
344
:param file_id: The file id that was deleted.
345
"""
346
if not self._recording_deletes:
347
raise AssertionError("recording deletes not activated.")
348
delta = (path, None, file_id, None)
349
self._basis_delta.append(delta)
350
self._any_changes = True
351
return delta
352
353
202
def will_record_deletes(self):
354
203
"""Tell the commit builder that deletes are being notified.
355
204
357
206
commit to be valid, deletes against the basis MUST be recorded via
358
207
builder.record_delete().
359
208
"""
360
self._recording_deletes = True
361
try:
362
basis_id = self.parents[0]
363
except IndexError:
364
basis_id = _mod_revision.NULL_REVISION
365
self.basis_delta_revision = basis_id
366
367
def record_entry_contents(self, ie, parent_invs, path, tree,
368
content_summary):
369
"""Record the content of ie from tree into the commit if needed.
370
371
Side effect: sets ie.revision when unchanged
372
373
:param ie: An inventory entry present in the commit.
374
:param parent_invs: The inventories of the parent revisions of the
375
commit.
376
:param path: The path the entry is at in the tree.
377
:param tree: The tree which contains this entry and should be used to
378
obtain content.
379
:param content_summary: Summary data from the tree about the paths
380
content - stat, length, exec, sha/link target. This is only
381
accessed when the entry has a revision of None - that is when it is
382
a candidate to commit.
383
:return: A tuple (change_delta, version_recorded, fs_hash).
384
change_delta is an inventory_delta change for this entry against
385
the basis tree of the commit, or None if no change occured against
386
the basis tree.
387
version_recorded is True if a new version of the entry has been
388
recorded. For instance, committing a merge where a file was only
389
changed on the other side will return (delta, False).
390
fs_hash is either None, or the hash details for the path (currently
391
a tuple of the contents sha1 and the statvalue returned by
392
tree.get_file_with_stat()).
393
"""
394
if self.new_inventory.root is None:
395
if ie.parent_id is not None:
396
raise errors.RootMissing()
397
self._check_root(ie, parent_invs, tree)
398
if ie.revision is None:
399
kind = content_summary[0]
400
else:
401
# ie is carried over from a prior commit
402
kind = ie.kind
403
# XXX: repository specific check for nested tree support goes here - if
404
# the repo doesn't want nested trees we skip it ?
405
if (kind == 'tree-reference' and
406
not self.repository._format.supports_tree_reference):
407
# mismatch between commit builder logic and repository:
408
# this needs the entry creation pushed down into the builder.
409
raise NotImplementedError('Missing repository subtree support.')
410
self.new_inventory.add(ie)
411
412
# TODO: slow, take it out of the inner loop.
413
try:
414
basis_inv = parent_invs[0]
415
except IndexError:
416
basis_inv = Inventory(root_id=None)
417
418
# ie.revision is always None if the InventoryEntry is considered
419
# for committing. We may record the previous parents revision if the
420
# content is actually unchanged against a sole head.
421
if ie.revision is not None:
422
if not self._versioned_root and path == '':
423
# repositories that do not version the root set the root's
424
# revision to the new commit even when no change occurs (more
425
# specifically, they do not record a revision on the root; and
426
# the rev id is assigned to the root during deserialisation -
427
# this masks when a change may have occurred against the basis.
428
# To match this we always issue a delta, because the revision
429
# of the root will always be changing.
430
if ie.file_id in basis_inv:
431
delta = (basis_inv.id2path(ie.file_id), path,
432
ie.file_id, ie)
433
else:
434
# add
435
delta = (None, path, ie.file_id, ie)
436
self._basis_delta.append(delta)
437
return delta, False, None
438
else:
439
# we don't need to commit this, because the caller already
440
# determined that an existing revision of this file is
441
# appropriate. If it's not being considered for committing then
442
# it and all its parents to the root must be unaltered so
443
# no-change against the basis.
444
if ie.revision == self._new_revision_id:
445
raise AssertionError("Impossible situation, a skipped "
446
"inventory entry (%r) claims to be modified in this "
447
"commit (%r).", (ie, self._new_revision_id))
448
return None, False, None
449
# XXX: Friction: parent_candidates should return a list not a dict
450
# so that we don't have to walk the inventories again.
451
parent_candiate_entries = ie.parent_candidates(parent_invs)
452
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
453
heads = []
454
for inv in parent_invs:
455
if ie.file_id in inv:
456
old_rev = inv[ie.file_id].revision
457
if old_rev in head_set:
458
heads.append(inv[ie.file_id].revision)
459
head_set.remove(inv[ie.file_id].revision)
460
461
store = False
462
# now we check to see if we need to write a new record to the
463
# file-graph.
464
# We write a new entry unless there is one head to the ancestors, and
465
# the kind-derived content is unchanged.
466
467
# Cheapest check first: no ancestors, or more the one head in the
468
# ancestors, we write a new node.
469
if len(heads) != 1:
470
store = True
471
if not store:
472
# There is a single head, look it up for comparison
473
parent_entry = parent_candiate_entries[heads[0]]
474
# if the non-content specific data has changed, we'll be writing a
475
# node:
476
if (parent_entry.parent_id != ie.parent_id or
477
parent_entry.name != ie.name):
478
store = True
479
# now we need to do content specific checks:
480
if not store:
481
# if the kind changed the content obviously has
482
if kind != parent_entry.kind:
483
store = True
484
# Stat cache fingerprint feedback for the caller - None as we usually
485
# don't generate one.
486
fingerprint = None
487
if kind == 'file':
488
if content_summary[2] is None:
489
raise ValueError("Files must not have executable = None")
490
if not store:
491
# We can't trust a check of the file length because of content
492
# filtering...
493
if (# if the exec bit has changed we have to store:
494
parent_entry.executable != content_summary[2]):
495
store = True
496
elif parent_entry.text_sha1 == content_summary[3]:
497
# all meta and content is unchanged (using a hash cache
498
# hit to check the sha)
499
ie.revision = parent_entry.revision
500
ie.text_size = parent_entry.text_size
501
ie.text_sha1 = parent_entry.text_sha1
502
ie.executable = parent_entry.executable
503
return self._get_delta(ie, basis_inv, path), False, None
504
else:
505
# Either there is only a hash change(no hash cache entry,
506
# or same size content change), or there is no change on
507
# this file at all.
508
# Provide the parent's hash to the store layer, so that the
509
# content is unchanged we will not store a new node.
510
nostore_sha = parent_entry.text_sha1
511
if store:
512
# We want to record a new node regardless of the presence or
513
# absence of a content change in the file.
514
nostore_sha = None
515
ie.executable = content_summary[2]
516
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
517
try:
518
text = file_obj.read()
519
finally:
520
file_obj.close()
521
try:
522
ie.text_sha1, ie.text_size = self._add_text_to_weave(
523
ie.file_id, text, heads, nostore_sha)
524
# Let the caller know we generated a stat fingerprint.
525
fingerprint = (ie.text_sha1, stat_value)
526
except errors.ExistingContent:
527
# Turns out that the file content was unchanged, and we were
528
# only going to store a new node if it was changed. Carry over
529
# the entry.
530
ie.revision = parent_entry.revision
531
ie.text_size = parent_entry.text_size
532
ie.text_sha1 = parent_entry.text_sha1
533
ie.executable = parent_entry.executable
534
return self._get_delta(ie, basis_inv, path), False, None
535
elif kind == 'directory':
536
if not store:
537
# all data is meta here, nothing specific to directory, so
538
# carry over:
539
ie.revision = parent_entry.revision
540
return self._get_delta(ie, basis_inv, path), False, None
541
self._add_text_to_weave(ie.file_id, '', heads, None)
542
elif kind == 'symlink':
543
current_link_target = content_summary[3]
544
if not store:
545
# symlink target is not generic metadata, check if it has
546
# changed.
547
if current_link_target != parent_entry.symlink_target:
548
store = True
549
if not store:
550
# unchanged, carry over.
551
ie.revision = parent_entry.revision
552
ie.symlink_target = parent_entry.symlink_target
553
return self._get_delta(ie, basis_inv, path), False, None
554
ie.symlink_target = current_link_target
555
self._add_text_to_weave(ie.file_id, '', heads, None)
556
elif kind == 'tree-reference':
557
if not store:
558
if content_summary[3] != parent_entry.reference_revision:
559
store = True
560
if not store:
561
# unchanged, carry over.
562
ie.reference_revision = parent_entry.reference_revision
563
ie.revision = parent_entry.revision
564
return self._get_delta(ie, basis_inv, path), False, None
565
ie.reference_revision = content_summary[3]
566
if ie.reference_revision is None:
567
raise AssertionError("invalid content_summary for nested tree: %r"
568
% (content_summary,))
569
self._add_text_to_weave(ie.file_id, '', heads, None)
570
else:
571
raise NotImplementedError('unknown kind')
572
ie.revision = self._new_revision_id
573
self._any_changes = True
574
return self._get_delta(ie, basis_inv, path), True, fingerprint
575
576
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
577
_entry_factory=entry_factory):
209
raise NotImplementedError(self.will_record_deletes)
210
211
def record_iter_changes(self, tree, basis_revision_id, iter_changes):
578
212
"""Record a new tree via iter_changes.
579
213
580
214
:param tree: The tree to obtain text contents from for changed objects.
585
219
to basis_revision_id. The iterator must not include any items with
586
220
a current kind of None - missing items must be either filtered out
587
221
or errored-on beefore record_iter_changes sees the item.
588
:param _entry_factory: Private method to bind entry_factory locally for
589
performance.
590
222
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
591
223
tree._observed_sha1.
592
224
"""
593
# Create an inventory delta based on deltas between all the parents and
594
# deltas between all the parent inventories. We use inventory delta's
595
# between the inventory objects because iter_changes masks
596
# last-changed-field only changes.
597
# Working data:
598
# file_id -> change map, change is fileid, paths, changed, versioneds,
599
# parents, names, kinds, executables
600
merged_ids = {}
601
# {file_id -> revision_id -> inventory entry, for entries in parent
602
# trees that are not parents[0]
603
parent_entries = {}
604
ghost_basis = False
605
try:
606
revtrees = list(self.repository.revision_trees(self.parents))
607
except errors.NoSuchRevision:
608
# one or more ghosts, slow path.
609
revtrees = []
610
for revision_id in self.parents:
611
try:
612
revtrees.append(self.repository.revision_tree(revision_id))
613
except errors.NoSuchRevision:
614
if not revtrees:
615
basis_revision_id = _mod_revision.NULL_REVISION
616
ghost_basis = True
617
revtrees.append(self.repository.revision_tree(
618
_mod_revision.NULL_REVISION))
619
# The basis inventory from a repository
620
if revtrees:
621
basis_inv = revtrees[0].inventory
622
else:
623
basis_inv = self.repository.revision_tree(
624
_mod_revision.NULL_REVISION).inventory
625
if len(self.parents) > 0:
626
if basis_revision_id != self.parents[0] and not ghost_basis:
627
raise Exception(
628
"arbitrary basis parents not yet supported with merges")
629
for revtree in revtrees[1:]:
630
for change in revtree.inventory._make_delta(basis_inv):
631
if change[1] is None:
632
# Not present in this parent.
633
continue
634
if change[2] not in merged_ids:
635
if change[0] is not None:
636
basis_entry = basis_inv[change[2]]
637
merged_ids[change[2]] = [
638
# basis revid
639
basis_entry.revision,
640
# new tree revid
641
change[3].revision]
642
parent_entries[change[2]] = {
643
# basis parent
644
basis_entry.revision:basis_entry,
645
# this parent
646
change[3].revision:change[3],
647
}
648
else:
649
merged_ids[change[2]] = [change[3].revision]
650
parent_entries[change[2]] = {change[3].revision:change[3]}
651
else:
652
merged_ids[change[2]].append(change[3].revision)
653
parent_entries[change[2]][change[3].revision] = change[3]
654
else:
655
merged_ids = {}
656
# Setup the changes from the tree:
657
# changes maps file_id -> (change, [parent revision_ids])
658
changes= {}
659
for change in iter_changes:
660
# This probably looks up in basis_inv way to much.
661
if change[1][0] is not None:
662
head_candidate = [basis_inv[change[0]].revision]
663
else:
664
head_candidate = []
665
changes[change[0]] = change, merged_ids.get(change[0],
666
head_candidate)
667
unchanged_merged = set(merged_ids) - set(changes)
668
# Extend the changes dict with synthetic changes to record merges of
669
# texts.
670
for file_id in unchanged_merged:
671
# Record a merged version of these items that did not change vs the
672
# basis. This can be either identical parallel changes, or a revert
673
# of a specific file after a merge. The recorded content will be
674
# that of the current tree (which is the same as the basis), but
675
# the per-file graph will reflect a merge.
676
# NB:XXX: We are reconstructing path information we had, this
677
# should be preserved instead.
678
# inv delta change: (file_id, (path_in_source, path_in_target),
679
# changed_content, versioned, parent, name, kind,
680
# executable)
681
try:
682
basis_entry = basis_inv[file_id]
683
except errors.NoSuchId:
684
# a change from basis->some_parents but file_id isn't in basis
685
# so was new in the merge, which means it must have changed
686
# from basis -> current, and as it hasn't the add was reverted
687
# by the user. So we discard this change.
688
pass
689
else:
690
change = (file_id,
691
(basis_inv.id2path(file_id), tree.id2path(file_id)),
692
False, (True, True),
693
(basis_entry.parent_id, basis_entry.parent_id),
694
(basis_entry.name, basis_entry.name),
695
(basis_entry.kind, basis_entry.kind),
696
(basis_entry.executable, basis_entry.executable))
697
changes[file_id] = (change, merged_ids[file_id])
698
# changes contains tuples with the change and a set of inventory
699
# candidates for the file.
700
# inv delta is:
701
# old_path, new_path, file_id, new_inventory_entry
702
seen_root = False # Is the root in the basis delta?
703
inv_delta = self._basis_delta
704
modified_rev = self._new_revision_id
705
for change, head_candidates in changes.values():
706
if change[3][1]: # versioned in target.
707
# Several things may be happening here:
708
# We may have a fork in the per-file graph
709
# - record a change with the content from tree
710
# We may have a change against < all trees
711
# - carry over the tree that hasn't changed
712
# We may have a change against all trees
713
# - record the change with the content from tree
714
kind = change[6][1]
715
file_id = change[0]
716
entry = _entry_factory[kind](file_id, change[5][1],
717
change[4][1])
718
head_set = self._heads(change[0], set(head_candidates))
719
heads = []
720
# Preserve ordering.
721
for head_candidate in head_candidates:
722
if head_candidate in head_set:
723
heads.append(head_candidate)
724
head_set.remove(head_candidate)
725
carried_over = False
726
if len(heads) == 1:
727
# Could be a carry-over situation:
728
parent_entry_revs = parent_entries.get(file_id, None)
729
if parent_entry_revs:
730
parent_entry = parent_entry_revs.get(heads[0], None)
731
else:
732
parent_entry = None
733
if parent_entry is None:
734
# The parent iter_changes was called against is the one
735
# that is the per-file head, so any change is relevant
736
# iter_changes is valid.
737
carry_over_possible = False
738
else:
739
# could be a carry over situation
740
# A change against the basis may just indicate a merge,
741
# we need to check the content against the source of the
742
# merge to determine if it was changed after the merge
743
# or carried over.
744
if (parent_entry.kind != entry.kind or
745
parent_entry.parent_id != entry.parent_id or
746
parent_entry.name != entry.name):
747
# Metadata common to all entries has changed
748
# against per-file parent
749
carry_over_possible = False
750
else:
751
carry_over_possible = True
752
# per-type checks for changes against the parent_entry
753
# are done below.
754
else:
755
# Cannot be a carry-over situation
756
carry_over_possible = False
757
# Populate the entry in the delta
758
if kind == 'file':
759
# XXX: There is still a small race here: If someone reverts the content of a file
760
# after iter_changes examines and decides it has changed,
761
# we will unconditionally record a new version even if some
762
# other process reverts it while commit is running (with
763
# the revert happening after iter_changes did its
764
# examination).
765
if change[7][1]:
766
entry.executable = True
767
else:
768
entry.executable = False
769
if (carry_over_possible and
770
parent_entry.executable == entry.executable):
771
# Check the file length, content hash after reading
772
# the file.
773
nostore_sha = parent_entry.text_sha1
774
else:
775
nostore_sha = None
776
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
777
try:
778
text = file_obj.read()
779
finally:
780
file_obj.close()
781
try:
782
entry.text_sha1, entry.text_size = self._add_text_to_weave(
783
file_id, text, heads, nostore_sha)
784
yield file_id, change[1][1], (entry.text_sha1, stat_value)
785
except errors.ExistingContent:
786
# No content change against a carry_over parent
787
# Perhaps this should also yield a fs hash update?
788
carried_over = True
789
entry.text_size = parent_entry.text_size
790
entry.text_sha1 = parent_entry.text_sha1
791
elif kind == 'symlink':
792
# Wants a path hint?
793
entry.symlink_target = tree.get_symlink_target(file_id)
794
if (carry_over_possible and
795
parent_entry.symlink_target == entry.symlink_target):
796
carried_over = True
797
else:
798
self._add_text_to_weave(change[0], '', heads, None)
799
elif kind == 'directory':
800
if carry_over_possible:
801
carried_over = True
802
else:
803
# Nothing to set on the entry.
804
# XXX: split into the Root and nonRoot versions.
805
if change[1][1] != '' or self.repository.supports_rich_root():
806
self._add_text_to_weave(change[0], '', heads, None)
807
elif kind == 'tree-reference':
808
if not self.repository._format.supports_tree_reference:
809
# This isn't quite sane as an error, but we shouldn't
810
# ever see this code path in practice: tree's don't
811
# permit references when the repo doesn't support tree
812
# references.
813
raise errors.UnsupportedOperation(tree.add_reference,
814
self.repository)
815
reference_revision = tree.get_reference_revision(change[0])
816
entry.reference_revision = reference_revision
817
if (carry_over_possible and
818
parent_entry.reference_revision == reference_revision):
819
carried_over = True
820
else:
821
self._add_text_to_weave(change[0], '', heads, None)
822
else:
823
raise AssertionError('unknown kind %r' % kind)
824
if not carried_over:
825
entry.revision = modified_rev
826
else:
827
entry.revision = parent_entry.revision
828
else:
829
entry = None
830
new_path = change[1][1]
831
inv_delta.append((change[1][0], new_path, change[0], entry))
832
if new_path == '':
833
seen_root = True
834
self.new_inventory = None
835
if len(inv_delta):
836
# This should perhaps be guarded by a check that the basis we
837
# commit against is the basis for the commit and if not do a delta
838
# against the basis.
839
self._any_changes = True
840
if not seen_root:
841
# housekeeping root entry changes do not affect no-change commits.
842
self._require_root_change(tree)
843
self.basis_delta_revision = basis_revision_id
844
845
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
846
parent_keys = tuple([(file_id, parent) for parent in parents])
847
return self.repository.texts._add_text(
848
(file_id, self._new_revision_id), parent_keys, new_text,
849
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
850
851
852
class RootCommitBuilder(CommitBuilder):
853
"""This commitbuilder actually records the root id"""
854
855
# the root entry gets versioned properly by this builder.
856
_versioned_root = True
857
858
def _check_root(self, ie, parent_invs, tree):
859
"""Helper for record_entry_contents.
860
861
:param ie: An entry being added.
862
:param parent_invs: The inventories of the parent revisions of the
863
commit.
864
:param tree: The tree that is being committed.
865
"""
866
867
def _require_root_change(self, tree):
868
"""Enforce an appropriate root object change.
869
870
This is called once when record_iter_changes is called, if and only if
871
the root was not in the delta calculated by record_iter_changes.
872
873
:param tree: The tree which is being committed.
874
"""
875
# versioned roots do not change unless the tree found a change.
225
raise NotImplementedError(self.record_iter_changes)
876
226
877
227
878
228
class RepositoryWriteLockResult(LogicalLockResult):
903
253
revisions and file history. It's normally accessed only by the Branch,
904
254
which views a particular line of development through that history.
905
255
906
The Repository builds on top of some byte storage facilies (the revisions,
907
signatures, inventories, texts and chk_bytes attributes) and a Transport,
908
which respectively provide byte storage and a means to access the (possibly
909
remote) disk.
910
911
The byte storage facilities are addressed via tuples, which we refer to
912
as 'keys' throughout the code base. Revision_keys, inventory_keys and
913
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
914
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
915
byte string made up of a hash identifier and a hash value.
916
We use this interface because it allows low friction with the underlying
917
code that implements disk indices, network encoding and other parts of
918
bzrlib.
919
920
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
921
the serialised revisions for the repository. This can be used to obtain
922
revision graph information or to access raw serialised revisions.
923
The result of trying to insert data into the repository via this store
924
is undefined: it should be considered read-only except for implementors
925
of repositories.
926
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
927
the serialised signatures for the repository. This can be used to
928
obtain access to raw serialised signatures. The result of trying to
929
insert data into the repository via this store is undefined: it should
930
be considered read-only except for implementors of repositories.
931
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
932
the serialised inventories for the repository. This can be used to
933
obtain unserialised inventories. The result of trying to insert data
934
into the repository via this store is undefined: it should be
935
considered read-only except for implementors of repositories.
936
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
937
texts of files and directories for the repository. This can be used to
938
obtain file texts or file graphs. Note that Repository.iter_file_bytes
939
is usually a better interface for accessing file texts.
940
The result of trying to insert data into the repository via this store
941
is undefined: it should be considered read-only except for implementors
942
of repositories.
943
:ivar chk_bytes: A bzrlib.versionedfile.VersionedFiles instance containing
944
any data the repository chooses to store or have indexed by its hash.
945
The result of trying to insert data into the repository via this store
946
is undefined: it should be considered read-only except for implementors
947
of repositories.
948
:ivar _transport: Transport for file access to repository, typically
949
pointing to .bzr/repository.
256
See VersionedFileRepository in bzrlib.vf_repository for the
257
base class for most Bazaar repositories.
950
258
"""
951
259
952
# What class to use for a CommitBuilder. Often it's simpler to change this
953
# in a Repository class subclass rather than to override
954
# get_commit_builder.
955
_commit_builder_class = CommitBuilder
956
# The search regex used by xml based repositories to determine what things
957
# where changed in a single commit.
958
_file_ids_altered_regex = lazy_regex.lazy_compile(
959
r'file_id="(?P<file_id>[^"]+)"'
960
r'.* revision="(?P<revision_id>[^"]+)"'
961
)
962
963
260
def abort_write_group(self, suppress_errors=False):
964
261
"""Commit the contents accrued within the current write group.
965
262
987
284
raise
988
285
mutter('abort_write_group failed')
989
286
log_exception_quietly()
990
note('bzr: ERROR (ignored): %s', exc)
287
note(gettext('bzr: ERROR (ignored): %s'), exc)
991
288
self._write_group = None
992
289
993
290
def _abort_write_group(self):
1008
305
1009
306
:param repository: A repository.
1010
307
"""
1011
if not self._format.supports_external_lookups:
1012
raise errors.UnstackableRepositoryFormat(self._format, self.base)
1013
if self.is_locked():
1014
# This repository will call fallback.unlock() when we transition to
1015
# the unlocked state, so we make sure to increment the lock count
1016
repository.lock_read()
1017
self._check_fallback_repository(repository)
1018
self._fallback_repositories.append(repository)
1019
self.texts.add_fallback_versioned_files(repository.texts)
1020
self.inventories.add_fallback_versioned_files(repository.inventories)
1021
self.revisions.add_fallback_versioned_files(repository.revisions)
1022
self.signatures.add_fallback_versioned_files(repository.signatures)
1023
if self.chk_bytes is not None:
1024
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
308
raise NotImplementedError(self.add_fallback_repository)
1025
309
1026
310
def _check_fallback_repository(self, repository):
1027
311
"""Check that this repository can fallback to repository safely.
1032
316
"""
1033
317
return InterRepository._assert_same_model(self, repository)
1034
318
1035
def add_inventory(self, revision_id, inv, parents):
1036
"""Add the inventory inv to the repository as revision_id.
1037
1038
:param parents: The revision ids of the parents that revision_id
1039
is known to have and are in the repository already.
1040
1041
:returns: The validator(which is a sha1 digest, though what is sha'd is
1042
repository format specific) of the serialized inventory.
1043
"""
1044
if not self.is_in_write_group():
1045
raise AssertionError("%r not in write group" % (self,))
1046
_mod_revision.check_not_reserved_id(revision_id)
1047
if not (inv.revision_id is None or inv.revision_id == revision_id):
1048
raise AssertionError(
1049
"Mismatch between inventory revision"
1050
" id and insertion revid (%r, %r)"
1051
% (inv.revision_id, revision_id))
1052
if inv.root is None:
1053
raise errors.RootMissing()
1054
return self._add_inventory_checked(revision_id, inv, parents)
1055
1056
def _add_inventory_checked(self, revision_id, inv, parents):
1057
"""Add inv to the repository after checking the inputs.
1058
1059
This function can be overridden to allow different inventory styles.
1060
1061
:seealso: add_inventory, for the contract.
1062
"""
1063
inv_lines = self._serializer.write_inventory_to_lines(inv)
1064
return self._inventory_add_lines(revision_id, parents,
1065
inv_lines, check_content=False)
1066
1067
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1068
parents, basis_inv=None, propagate_caches=False):
1069
"""Add a new inventory expressed as a delta against another revision.
1070
1071
See the inventory developers documentation for the theory behind
1072
inventory deltas.
1073
1074
:param basis_revision_id: The inventory id the delta was created
1075
against. (This does not have to be a direct parent.)
1076
:param delta: The inventory delta (see Inventory.apply_delta for
1077
details).
1078
:param new_revision_id: The revision id that the inventory is being
1079
added for.
1080
:param parents: The revision ids of the parents that revision_id is
1081
known to have and are in the repository already. These are supplied
1082
for repositories that depend on the inventory graph for revision
1083
graph access, as well as for those that pun ancestry with delta
1084
compression.
1085
:param basis_inv: The basis inventory if it is already known,
1086
otherwise None.
1087
:param propagate_caches: If True, the caches for this inventory are
1088
copied to and updated for the result if possible.
1089
1090
:returns: (validator, new_inv)
1091
The validator(which is a sha1 digest, though what is sha'd is
1092
repository format specific) of the serialized inventory, and the
1093
resulting inventory.
1094
"""
1095
if not self.is_in_write_group():
1096
raise AssertionError("%r not in write group" % (self,))
1097
_mod_revision.check_not_reserved_id(new_revision_id)
1098
basis_tree = self.revision_tree(basis_revision_id)
1099
basis_tree.lock_read()
1100
try:
1101
# Note that this mutates the inventory of basis_tree, which not all
1102
# inventory implementations may support: A better idiom would be to
1103
# return a new inventory, but as there is no revision tree cache in
1104
# repository this is safe for now - RBC 20081013
1105
if basis_inv is None:
1106
basis_inv = basis_tree.inventory
1107
basis_inv.apply_delta(delta)
1108
basis_inv.revision_id = new_revision_id
1109
return (self.add_inventory(new_revision_id, basis_inv, parents),
1110
basis_inv)
1111
finally:
1112
basis_tree.unlock()
1113
1114
def _inventory_add_lines(self, revision_id, parents, lines,
1115
check_content=True):
1116
"""Store lines in inv_vf and return the sha1 of the inventory."""
1117
parents = [(parent,) for parent in parents]
1118
result = self.inventories.add_lines((revision_id,), parents, lines,
1119
check_content=check_content)[0]
1120
self.inventories._access.flush()
1121
return result
1122
1123
def add_revision(self, revision_id, rev, inv=None, config=None):
1124
"""Add rev to the revision store as revision_id.
1125
1126
:param revision_id: the revision id to use.
1127
:param rev: The revision object.
1128
:param inv: The inventory for the revision. if None, it will be looked
1129
up in the inventory storer
1130
:param config: If None no digital signature will be created.
1131
If supplied its signature_needed method will be used
1132
to determine if a signature should be made.
1133
"""
1134
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
1135
# rev.parent_ids?
1136
_mod_revision.check_not_reserved_id(revision_id)
1137
if config is not None and config.signature_needed():
1138
if inv is None:
1139
inv = self.get_inventory(revision_id)
1140
plaintext = Testament(rev, inv).as_short_text()
1141
self.store_revision_signature(
1142
gpg.GPGStrategy(config), plaintext, revision_id)
1143
# check inventory present
1144
if not self.inventories.get_parent_map([(revision_id,)]):
1145
if inv is None:
1146
raise errors.WeaveRevisionNotPresent(revision_id,
1147
self.inventories)
1148
else:
1149
# yes, this is not suitable for adding with ghosts.
1150
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
1151
rev.parent_ids)
1152
else:
1153
key = (revision_id,)
1154
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
1155
self._add_revision(rev)
1156
1157
def _add_revision(self, revision):
1158
text = self._serializer.write_revision_to_string(revision)
1159
key = (revision.revision_id,)
1160
parents = tuple((parent,) for parent in revision.parent_ids)
1161
self.revisions.add_lines(key, parents, osutils.split_lines(text))
1162
1163
319
def all_revision_ids(self):
1164
320
"""Returns a list of all the revision ids in the repository.
1165
321
1188
344
"""
1189
345
self.control_files.break_lock()
1190
346
1191
@needs_read_lock
1192
def _eliminate_revisions_not_present(self, revision_ids):
1193
"""Check every revision id in revision_ids to see if we have it.
1194
1195
Returns a set of the present revisions.
1196
"""
1197
result = []
1198
graph = self.get_graph()
1199
parent_map = graph.get_parent_map(revision_ids)
1200
# The old API returned a list, should this actually be a set?
1201
return parent_map.keys()
1202
1203
def _check_inventories(self, checker):
1204
"""Check the inventories found from the revision scan.
1205
1206
This is responsible for verifying the sha1 of inventories and
1207
creating a pending_keys set that covers data referenced by inventories.
1208
"""
1209
bar = ui.ui_factory.nested_progress_bar()
1210
try:
1211
self._do_check_inventories(checker, bar)
1212
finally:
1213
bar.finished()
1214
1215
def _do_check_inventories(self, checker, bar):
1216
"""Helper for _check_inventories."""
1217
revno = 0
1218
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1219
kinds = ['chk_bytes', 'texts']
1220
count = len(checker.pending_keys)
1221
bar.update("inventories", 0, 2)
1222
current_keys = checker.pending_keys
1223
checker.pending_keys = {}
1224
# Accumulate current checks.
1225
for key in current_keys:
1226
if key[0] != 'inventories' and key[0] not in kinds:
1227
checker._report_items.append('unknown key type %r' % (key,))
1228
keys[key[0]].add(key[1:])
1229
if keys['inventories']:
1230
# NB: output order *should* be roughly sorted - topo or
1231
# inverse topo depending on repository - either way decent
1232
# to just delta against. However, pre-CHK formats didn't
1233
# try to optimise inventory layout on disk. As such the
1234
# pre-CHK code path does not use inventory deltas.
1235
last_object = None
1236
for record in self.inventories.check(keys=keys['inventories']):
1237
if record.storage_kind == 'absent':
1238
checker._report_items.append(
1239
'Missing inventory {%s}' % (record.key,))
1240
else:
1241
last_object = self._check_record('inventories', record,
1242
checker, last_object,
1243
current_keys[('inventories',) + record.key])
1244
del keys['inventories']
1245
else:
1246
return
1247
bar.update("texts", 1)
1248
while (checker.pending_keys or keys['chk_bytes']
1249
or keys['texts']):
1250
# Something to check.
1251
current_keys = checker.pending_keys
1252
checker.pending_keys = {}
1253
# Accumulate current checks.
1254
for key in current_keys:
1255
if key[0] not in kinds:
1256
checker._report_items.append('unknown key type %r' % (key,))
1257
keys[key[0]].add(key[1:])
1258
# Check the outermost kind only - inventories || chk_bytes || texts
1259
for kind in kinds:
1260
if keys[kind]:
1261
last_object = None
1262
for record in getattr(self, kind).check(keys=keys[kind]):
1263
if record.storage_kind == 'absent':
1264
checker._report_items.append(
1265
'Missing %s {%s}' % (kind, record.key,))
1266
else:
1267
last_object = self._check_record(kind, record,
1268
checker, last_object, current_keys[(kind,) + record.key])
1269
keys[kind] = set()
1270
break
1271
1272
def _check_record(self, kind, record, checker, last_object, item_data):
1273
"""Check a single text from this repository."""
1274
if kind == 'inventories':
1275
rev_id = record.key[0]
1276
inv = self._deserialise_inventory(rev_id,
1277
record.get_bytes_as('fulltext'))
1278
if last_object is not None:
1279
delta = inv._make_delta(last_object)
1280
for old_path, path, file_id, ie in delta:
1281
if ie is None:
1282
continue
1283
ie.check(checker, rev_id, inv)
1284
else:
1285
for path, ie in inv.iter_entries():
1286
ie.check(checker, rev_id, inv)
1287
if self._format.fast_deltas:
1288
return inv
1289
elif kind == 'chk_bytes':
1290
# No code written to check chk_bytes for this repo format.
1291
checker._report_items.append(
1292
'unsupported key type chk_bytes for %s' % (record.key,))
1293
elif kind == 'texts':
1294
self._check_text(record, checker, item_data)
1295
else:
1296
checker._report_items.append(
1297
'unknown key type %s for %s' % (kind, record.key))
1298
1299
def _check_text(self, record, checker, item_data):
1300
"""Check a single text."""
1301
# Check it is extractable.
1302
# TODO: check length.
1303
if record.storage_kind == 'chunked':
1304
chunks = record.get_bytes_as(record.storage_kind)
1305
sha1 = osutils.sha_strings(chunks)
1306
length = sum(map(len, chunks))
1307
else:
1308
content = record.get_bytes_as('fulltext')
1309
sha1 = osutils.sha_string(content)
1310
length = len(content)
1311
if item_data and sha1 != item_data[1]:
1312
checker._report_items.append(
1313
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1314
(record.key, sha1, item_data[1], item_data[2]))
1315
1316
347
@staticmethod
1317
def create(a_bzrdir):
1318
"""Construct the current default format repository in a_bzrdir."""
1319
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
348
def create(controldir):
349
"""Construct the current default format repository in controldir."""
350
return RepositoryFormat.get_default_format().initialize(controldir)
1320
351
1321
def __init__(self, _format, a_bzrdir, control_files):
352
def __init__(self, _format, controldir, control_files):
1322
353
"""instantiate a Repository.
1323
354
1324
355
:param _format: The format of the repository on disk.
1325
:param a_bzrdir: The BzrDir of the repository.
356
:param controldir: The ControlDir of the repository.
357
:param control_files: Control files to use for locking, etc.
1326
358
"""
1327
359
# In the future we will have a single api for all stores for
1328
360
# getting file texts, inventories and revisions, then
1330
362
super(Repository, self).__init__()
1331
363
self._format = _format
1332
364
# the following are part of the public API for Repository:
1333
self.bzrdir = a_bzrdir
365
self.bzrdir = controldir
1334
366
self.control_files = control_files
1335
self._transport = control_files._transport
1336
self.base = self._transport.base
1337
367
# for tests
1338
self._reconcile_does_inventory_gc = True
1339
self._reconcile_fixes_text_parents = False
1340
self._reconcile_backsup_inventory = True
1341
368
self._write_group = None
1342
369
# Additional places to query for data.
1343
370
self._fallback_repositories = []
1344
# An InventoryEntry cache, used during deserialization
1345
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1346
# Is it safe to return inventory entries directly from the entry cache,
1347
# rather copying them?
1348
self._safe_to_return_from_cache = False
1349
371
1350
372
@property
1351
373
def user_transport(self):
1385
407
"""
1386
408
if self.__class__ is not other.__class__:
1387
409
return False
1388
return (self._transport.base == other._transport.base)
410
return (self.control_url == other.control_url)
1389
411
1390
412
def is_in_write_group(self):
1391
413
"""Return True if there is an open write group.
1491
513
if revid and committers:
1492
514
result['committers'] = 0
1493
515
if revid and revid != _mod_revision.NULL_REVISION:
516
graph = self.get_graph()
1494
517
if committers:
1495
518
all_committers = set()
1496
revisions = self.get_ancestry(revid)
1497
# pop the leading None
1498
revisions.pop(0)
1499
first_revision = None
519
revisions = [r for (r, p) in graph.iter_ancestry([revid])
520
if r != _mod_revision.NULL_REVISION]
521
last_revision = None
1500
522
if not committers:
1501
523
# ignore the revisions in the middle - just grab first and last
1502
524
revisions = revisions[0], revisions[-1]
1503
525
for revision in self.get_revisions(revisions):
1504
if not first_revision:
1505
first_revision = revision
526
if not last_revision:
527
last_revision = revision
1506
528
if committers:
1507
529
all_committers.add(revision.committer)
1508
last_revision = revision
530
first_revision = revision
1509
531
if committers:
1510
532
result['committers'] = len(all_committers)
1511
533
result['firstrev'] = (first_revision.timestamp,
1512
534
first_revision.timezone)
1513
535
result['latestrev'] = (last_revision.timestamp,
1514
536
last_revision.timezone)
1515
1516
# now gather global repository information
1517
# XXX: This is available for many repos regardless of listability.
1518
if self.user_transport.listable():
1519
# XXX: do we want to __define len__() ?
1520
# Maybe the versionedfiles object should provide a different
1521
# method to get the number of keys.
1522
result['revisions'] = len(self.revisions.keys())
1523
# result['size'] = t
1524
537
return result
1525
538
1526
539
def find_branches(self, using=False):
1537
550
def __init__(self):
1538
551
self.first_call = True
1539
552
1540
def __call__(self, bzrdir):
1541
# On the first call, the parameter is always the bzrdir
553
def __call__(self, controldir):
554
# On the first call, the parameter is always the controldir
1542
555
# containing the current repo.
1543
556
if not self.first_call:
1544
557
try:
1545
repository = bzrdir.open_repository()
558
repository = controldir.open_repository()
1546
559
except errors.NoRepositoryPresent:
1547
560
pass
1548
561
else:
1549
562
return False, ([], repository)
1550
563
self.first_call = False
1551
value = (bzrdir.list_branches(), None)
564
value = (controldir.list_branches(), None)
1552
565
return True, value
1553
566
1554
567
ret = []
1555
for branches, repository in bzrdir.BzrDir.find_bzrdirs(
568
for branches, repository in controldir.ControlDir.find_bzrdirs(
1556
569
self.user_transport, evaluate=Evaluator()):
1557
570
if branches is not None:
1558
571
ret.extend(branches)
1561
574
return ret
1562
575
1563
576
@needs_read_lock
1564
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
577
def search_missing_revision_ids(self, other,
578
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
579
find_ghosts=True, revision_ids=None, if_present_ids=None,
580
limit=None):
1565
581
"""Return the revision ids that other has that this does not.
1566
582
1567
583
These are returned in topological order.
1568
584
1569
585
revision_id: only return revision ids included by revision_id.
1570
586
"""
587
if symbol_versioning.deprecated_passed(revision_id):
588
symbol_versioning.warn(
589
'search_missing_revision_ids(revision_id=...) was '
590
'deprecated in 2.4. Use revision_ids=[...] instead.',
591
DeprecationWarning, stacklevel=3)
592
if revision_ids is not None:
593
raise AssertionError(
594
'revision_ids is mutually exclusive with revision_id')
595
if revision_id is not None:
596
revision_ids = [revision_id]
1571
597
return InterRepository.get(other, self).search_missing_revision_ids(
1572
revision_id, find_ghosts)
598
find_ghosts=find_ghosts, revision_ids=revision_ids,
599
if_present_ids=if_present_ids, limit=limit)
1573
600
1574
601
@staticmethod
1575
602
def open(base):
1578
605
For instance, if the repository is at URL/.bzr/repository,
1579
606
Repository.open(URL) -> a Repository instance.
1580
607
"""
1581
control = bzrdir.BzrDir.open(base)
608
control = controldir.ControlDir.open(base)
1582
609
return control.open_repository()
1583
610
1584
611
def copy_content_into(self, destination, revision_id=None):
1615
642
"""
1616
643
1617
644
def suspend_write_group(self):
645
"""Suspend a write group.
646
647
:raise UnsuspendableWriteGroup: If the write group can not be
648
suspended.
649
:return: List of tokens
650
"""
1618
651
raise errors.UnsuspendableWriteGroup(self)
1619
652
1620
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1621
"""Return the keys of missing inventory parents for revisions added in
1622
this write group.
1623
1624
A revision is not complete if the inventory delta for that revision
1625
cannot be calculated. Therefore if the parent inventories of a
1626
revision are not present, the revision is incomplete, and e.g. cannot
1627
be streamed by a smart server. This method finds missing inventory
1628
parents for revisions added in this write group.
1629
"""
1630
if not self._format.supports_external_lookups:
1631
# This is only an issue for stacked repositories
1632
return set()
1633
if not self.is_in_write_group():
1634
raise AssertionError('not in a write group')
1635
1636
# XXX: We assume that every added revision already has its
1637
# corresponding inventory, so we only check for parent inventories that
1638
# might be missing, rather than all inventories.
1639
parents = set(self.revisions._index.get_missing_parents())
1640
parents.discard(_mod_revision.NULL_REVISION)
1641
unstacked_inventories = self.inventories._index
1642
present_inventories = unstacked_inventories.get_parent_map(
1643
key[-1:] for key in parents)
1644
parents.difference_update(present_inventories)
1645
if len(parents) == 0:
1646
# No missing parent inventories.
1647
return set()
1648
if not check_for_missing_texts:
1649
return set(('inventories', rev_id) for (rev_id,) in parents)
1650
# Ok, now we have a list of missing inventories. But these only matter
1651
# if the inventories that reference them are missing some texts they
1652
# appear to introduce.
1653
# XXX: Texts referenced by all added inventories need to be present,
1654
# but at the moment we're only checking for texts referenced by
1655
# inventories at the graph's edge.
1656
key_deps = self.revisions._index._key_dependencies
1657
key_deps.satisfy_refs_for_keys(present_inventories)
1658
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1659
file_ids = self.fileids_altered_by_revision_ids(referrers)
1660
missing_texts = set()
1661
for file_id, version_ids in file_ids.iteritems():
1662
missing_texts.update(
1663
(file_id, version_id) for version_id in version_ids)
1664
present_texts = self.texts.get_parent_map(missing_texts)
1665
missing_texts.difference_update(present_texts)
1666
if not missing_texts:
1667
# No texts are missing, so all revisions and their deltas are
1668
# reconstructable.
1669
return set()
1670
# Alternatively the text versions could be returned as the missing
1671
# keys, but this is likely to be less data.
1672
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1673
return missing_keys
1674
1675
653
def refresh_data(self):
1676
654
"""Re-read any data needed to synchronise with disk.
1677
655
1697
675
def _resume_write_group(self, tokens):
1698
676
raise errors.UnsuspendableWriteGroup(self)
1699
677
1700
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
1701
fetch_spec=None):
678
def fetch(self, source, revision_id=None, find_ghosts=False):
1702
679
"""Fetch the content required to construct revision_id from source.
1703
680
1704
If revision_id is None and fetch_spec is None, then all content is
1705
copied.
681
If revision_id is None, then all content is copied.
1706
682
1707
683
fetch() may not be used when the repository is in a write group -
1708
684
either finish the current write group before using fetch, or use
1714
690
:param revision_id: If specified, all the content needed for this
1715
691
revision ID will be copied to the target. Fetch will determine for
1716
692
itself which content needs to be copied.
1717
:param fetch_spec: If specified, a SearchResult or
1718
PendingAncestryResult that describes which revisions to copy. This
1719
allows copying multiple heads at once. Mutually exclusive with
1720
revision_id.
1721
693
"""
1722
if fetch_spec is not None and revision_id is not None:
1723
raise AssertionError(
1724
"fetch_spec and revision_id are mutually exclusive.")
1725
694
if self.is_in_write_group():
1726
695
raise errors.InternalBzrError(
1727
696
"May not fetch while in a write group.")
1729
698
# TODO: lift out to somewhere common with RemoteRepository
1730
699
# <https://bugs.launchpad.net/bzr/+bug/401646>
1731
700
if (self.has_same_location(source)
1732
and fetch_spec is None
1733
701
and self._has_same_fallbacks(source)):
1734
702
# check that last_revision is in 'from' and then return a
1735
703
# no-operation.
1737
705
not _mod_revision.is_null(revision_id)):
1738
706
self.get_revision(revision_id)
1739
707
return 0, []
1740
# if there is no specific appropriate InterRepository, this will get
1741
# the InterRepository base class, which raises an
1742
# IncompatibleRepositories when asked to fetch.
1743
708
inter = InterRepository.get(source, self)
1744
return inter.fetch(revision_id=revision_id, pb=pb,
1745
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
709
return inter.fetch(revision_id=revision_id, find_ghosts=find_ghosts)
1746
710
1747
711
def create_bundle(self, target, base, fileobj, format=None):
1748
712
return serializer.write_bundle(self, target, base, fileobj, format)
1749
713
1750
def get_commit_builder(self, branch, parents, config, timestamp=None,
714
def get_commit_builder(self, branch, parents, config_stack, timestamp=None,
1751
715
timezone=None, committer=None, revprops=None,
1752
revision_id=None):
716
revision_id=None, lossy=False):
1753
717
"""Obtain a CommitBuilder for this repository.
1754
718
1755
719
:param branch: Branch to commit to.
1756
720
:param parents: Revision ids of the parents of the new revision.
1757
:param config: Configuration to use.
721
:param config_stack: Configuration stack to use.
1758
722
:param timestamp: Optional timestamp recorded for commit.
1759
723
:param timezone: Optional timezone for timestamp.
1760
724
:param committer: Optional committer to set for commit.
1761
725
:param revprops: Optional dictionary of revision properties.
1762
726
:param revision_id: Optional revision id.
727
:param lossy: Whether to discard data that can not be natively
728
represented, when pushing to a foreign VCS
1763
729
"""
1764
if self._fallback_repositories:
1765
raise errors.BzrError("Cannot commit from a lightweight checkout "
1766
"to a stacked branch. See "
1767
"https://bugs.launchpad.net/bzr/+bug/375013 for details.")
1768
result = self._commit_builder_class(self, parents, config,
1769
timestamp, timezone, committer, revprops, revision_id)
1770
self.start_write_group()
1771
return result
730
raise NotImplementedError(self.get_commit_builder)
1772
731
1773
732
@only_raises(errors.LockNotHeld, errors.LockBroken)
1774
733
def unlock(self):
1781
740
'Must end write groups before releasing write locks.')
1782
741
self.control_files.unlock()
1783
742
if self.control_files._lock_count == 0:
1784
self._inventory_entry_cache.clear()
1785
743
for repo in self._fallback_repositories:
1786
744
repo.unlock()
1787
745
1788
746
@needs_read_lock
1789
def clone(self, a_bzrdir, revision_id=None):
1790
"""Clone this repository into a_bzrdir using the current format.
747
def clone(self, controldir, revision_id=None):
748
"""Clone this repository into controldir using the current format.
1791
749
1792
750
Currently no check is made that the format of this repository and
1793
751
the bzrdir format are compatible. FIXME RBC 20060201.
1796
754
"""
1797
755
# TODO: deprecate after 0.16; cloning this with all its settings is
1798
756
# probably not very useful -- mbp 20070423
1799
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
757
dest_repo = self._create_sprouting_repo(
758
controldir, shared=self.is_shared())
1800
759
self.copy_content_into(dest_repo, revision_id)
1801
760
return dest_repo
1802
761
1852
811
dest_repo = a_bzrdir.open_repository()
1853
812
return dest_repo
1854
813
1855
def _get_sink(self):
1856
"""Return a sink for streaming into this repository."""
1857
return StreamSink(self)
1858
1859
def _get_source(self, to_format):
1860
"""Return a source for streaming from this repository."""
1861
return StreamSource(self, to_format)
1862
1863
814
@needs_read_lock
1864
815
def has_revision(self, revision_id):
1865
816
"""True if this repository has a copy of the revision."""
1872
823
:param revision_ids: An iterable of revision_ids.
1873
824
:return: A set of the revision_ids that were present.
1874
825
"""
1875
parent_map = self.revisions.get_parent_map(
1876
[(rev_id,) for rev_id in revision_ids])
1877
result = set()
1878
if _mod_revision.NULL_REVISION in revision_ids:
1879
result.add(_mod_revision.NULL_REVISION)
1880
result.update([key[0] for key in parent_map])
1881
return result
826
raise NotImplementedError(self.has_revisions)
1882
827
1883
828
@needs_read_lock
1884
829
def get_revision(self, revision_id):
1885
830
"""Return the Revision object for a named revision."""
1886
831
return self.get_revisions([revision_id])[0]
1887
832
1888
@needs_read_lock
1889
833
def get_revision_reconcile(self, revision_id):
1890
834
"""'reconcile' helper routine that allows access to a revision always.
1891
835
1894
838
be used by reconcile, or reconcile-alike commands that are correcting
1895
839
or testing the revision graph.
1896
840
"""
1897
return self._get_revisions([revision_id])[0]
841
raise NotImplementedError(self.get_revision_reconcile)
1898
842
1899
@needs_read_lock
1900
843
def get_revisions(self, revision_ids):
1901
844
"""Get many revisions at once.
1902
845
1903
846
Repositories that need to check data on every revision read should
1904
847
subclass this method.
1905
848
"""
1906
return self._get_revisions(revision_ids)
1907
1908
@needs_read_lock
1909
def _get_revisions(self, revision_ids):
1910
"""Core work logic to get many revisions without sanity checks."""
1911
revs = {}
1912
for revid, rev in self._iter_revisions(revision_ids):
1913
if rev is None:
1914
raise errors.NoSuchRevision(self, revid)
1915
revs[revid] = rev
1916
return [revs[revid] for revid in revision_ids]
1917
1918
def _iter_revisions(self, revision_ids):
1919
"""Iterate over revision objects.
1920
1921
:param revision_ids: An iterable of revisions to examine. None may be
1922
passed to request all revisions known to the repository. Note that
1923
not all repositories can find unreferenced revisions; for those
1924
repositories only referenced ones will be returned.
1925
:return: An iterator of (revid, revision) tuples. Absent revisions (
1926
those asked for but not available) are returned as (revid, None).
1927
"""
1928
if revision_ids is None:
1929
revision_ids = self.all_revision_ids()
1930
else:
1931
for rev_id in revision_ids:
1932
if not rev_id or not isinstance(rev_id, basestring):
1933
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1934
keys = [(key,) for key in revision_ids]
1935
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1936
for record in stream:
1937
revid = record.key[0]
1938
if record.storage_kind == 'absent':
1939
yield (revid, None)
1940
else:
1941
text = record.get_bytes_as('fulltext')
1942
rev = self._serializer.read_revision_from_string(text)
1943
yield (revid, rev)
849
raise NotImplementedError(self.get_revisions)
1944
850
1945
851
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1946
852
"""Produce a generator of revision deltas.
2001
907
signature = gpg_strategy.sign(plaintext)
2002
908
self.add_signature_text(revision_id, signature)
2003
909
2004
@needs_write_lock
2005
910
def add_signature_text(self, revision_id, signature):
2006
self.signatures.add_lines((revision_id,), (),
2007
osutils.split_lines(signature))
2008
2009
def find_text_key_references(self):
2010
"""Find the text key references within the repository.
2011
2012
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2013
to whether they were referred to by the inventory of the
2014
revision_id that they contain. The inventory texts from all present
2015
revision ids are assessed to generate this report.
2016
"""
2017
revision_keys = self.revisions.keys()
2018
w = self.inventories
2019
pb = ui.ui_factory.nested_progress_bar()
2020
try:
2021
return self._find_text_key_references_from_xml_inventory_lines(
2022
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
2023
finally:
2024
pb.finished()
2025
2026
def _find_text_key_references_from_xml_inventory_lines(self,
2027
line_iterator):
2028
"""Core routine for extracting references to texts from inventories.
2029
2030
This performs the translation of xml lines to revision ids.
2031
2032
:param line_iterator: An iterator of lines, origin_version_id
2033
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2034
to whether they were referred to by the inventory of the
2035
revision_id that they contain. Note that if that revision_id was
2036
not part of the line_iterator's output then False will be given -
2037
even though it may actually refer to that key.
2038
"""
2039
if not self._serializer.support_altered_by_hack:
2040
raise AssertionError(
2041
"_find_text_key_references_from_xml_inventory_lines only "
2042
"supported for branches which store inventory as unnested xml"
2043
", not on %r" % self)
2044
result = {}
2045
2046
# this code needs to read every new line in every inventory for the
2047
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
2048
# not present in one of those inventories is unnecessary but not
2049
# harmful because we are filtering by the revision id marker in the
2050
# inventory lines : we only select file ids altered in one of those
2051
# revisions. We don't need to see all lines in the inventory because
2052
# only those added in an inventory in rev X can contain a revision=X
2053
# line.
2054
unescape_revid_cache = {}
2055
unescape_fileid_cache = {}
2056
2057
# jam 20061218 In a big fetch, this handles hundreds of thousands
2058
# of lines, so it has had a lot of inlining and optimizing done.
2059
# Sorry that it is a little bit messy.
2060
# Move several functions to be local variables, since this is a long
2061
# running loop.
2062
search = self._file_ids_altered_regex.search
2063
unescape = _unescape_xml
2064
setdefault = result.setdefault
2065
for line, line_key in line_iterator:
2066
match = search(line)
2067
if match is None:
2068
continue
2069
# One call to match.group() returning multiple items is quite a
2070
# bit faster than 2 calls to match.group() each returning 1
2071
file_id, revision_id = match.group('file_id', 'revision_id')
2072
2073
# Inlining the cache lookups helps a lot when you make 170,000
2074
# lines and 350k ids, versus 8.4 unique ids.
2075
# Using a cache helps in 2 ways:
2076
# 1) Avoids unnecessary decoding calls
2077
# 2) Re-uses cached strings, which helps in future set and
2078
# equality checks.
2079
# (2) is enough that removing encoding entirely along with
2080
# the cache (so we are using plain strings) results in no
2081
# performance improvement.
2082
try:
2083
revision_id = unescape_revid_cache[revision_id]
2084
except KeyError:
2085
unescaped = unescape(revision_id)
2086
unescape_revid_cache[revision_id] = unescaped
2087
revision_id = unescaped
2088
2089
# Note that unconditionally unescaping means that we deserialise
2090
# every fileid, which for general 'pull' is not great, but we don't
2091
# really want to have some many fulltexts that this matters anyway.
2092
# RBC 20071114.
2093
try:
2094
file_id = unescape_fileid_cache[file_id]
2095
except KeyError:
2096
unescaped = unescape(file_id)
2097
unescape_fileid_cache[file_id] = unescaped
2098
file_id = unescaped
2099
2100
key = (file_id, revision_id)
2101
setdefault(key, False)
2102
if revision_id == line_key[-1]:
2103
result[key] = True
2104
return result
2105
2106
def _inventory_xml_lines_for_keys(self, keys):
2107
"""Get a line iterator of the sort needed for findind references.
2108
2109
Not relevant for non-xml inventory repositories.
2110
2111
Ghosts in revision_keys are ignored.
2112
2113
:param revision_keys: The revision keys for the inventories to inspect.
2114
:return: An iterator over (inventory line, revid) for the fulltexts of
2115
all of the xml inventories specified by revision_keys.
2116
"""
2117
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2118
for record in stream:
2119
if record.storage_kind != 'absent':
2120
chunks = record.get_bytes_as('chunked')
2121
revid = record.key[-1]
2122
lines = osutils.chunks_to_lines(chunks)
2123
for line in lines:
2124
yield line, revid
2125
2126
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2127
revision_keys):
2128
"""Helper routine for fileids_altered_by_revision_ids.
2129
2130
This performs the translation of xml lines to revision ids.
2131
2132
:param line_iterator: An iterator of lines, origin_version_id
2133
:param revision_keys: The revision ids to filter for. This should be a
2134
set or other type which supports efficient __contains__ lookups, as
2135
the revision key from each parsed line will be looked up in the
2136
revision_keys filter.
2137
:return: a dictionary mapping altered file-ids to an iterable of
2138
revision_ids. Each altered file-ids has the exact revision_ids that
2139
altered it listed explicitly.
2140
"""
2141
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2142
line_iterator).iterkeys())
2143
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2144
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2145
self._inventory_xml_lines_for_keys(parent_keys)))
2146
new_keys = seen - parent_seen
2147
result = {}
2148
setdefault = result.setdefault
2149
for key in new_keys:
2150
setdefault(key[0], set()).add(key[-1])
2151
return result
911
"""Store a signature text for a revision.
912
913
:param revision_id: Revision id of the revision
914
:param signature: Signature text.
915
"""
916
raise NotImplementedError(self.add_signature_text)
2152
917
2153
918
def _find_parent_ids_of_revisions(self, revision_ids):
2154
919
"""Find all parent ids that are mentioned in the revision graph.
2163
928
parent_ids.discard(_mod_revision.NULL_REVISION)
2164
929
return parent_ids
2165
930
2166
def _find_parent_keys_of_revisions(self, revision_keys):
2167
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2168
2169
:param revision_keys: An iterable of revision_keys.
2170
:return: The parents of all revision_keys that are not already in
2171
revision_keys
2172
"""
2173
parent_map = self.revisions.get_parent_map(revision_keys)
2174
parent_keys = set()
2175
map(parent_keys.update, parent_map.itervalues())
2176
parent_keys.difference_update(revision_keys)
2177
parent_keys.discard(_mod_revision.NULL_REVISION)
2178
return parent_keys
2179
2180
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
2181
"""Find the file ids and versions affected by revisions.
2182
2183
:param revisions: an iterable containing revision ids.
2184
:param _inv_weave: The inventory weave from this repository or None.
2185
If None, the inventory weave will be opened automatically.
2186
:return: a dictionary mapping altered file-ids to an iterable of
2187
revision_ids. Each altered file-ids has the exact revision_ids that
2188
altered it listed explicitly.
2189
"""
2190
selected_keys = set((revid,) for revid in revision_ids)
2191
w = _inv_weave or self.inventories
2192
return self._find_file_ids_from_xml_inventory_lines(
2193
w.iter_lines_added_or_present_in_keys(
2194
selected_keys, pb=None),
2195
selected_keys)
2196
2197
931
def iter_files_bytes(self, desired_files):
2198
932
"""Iterate through file versions.
2199
933
2205
939
uniquely identify the file version in the caller's context. (Examples:
2206
940
an index number or a TreeTransform trans_id.)
2207
941
2208
bytes_iterator is an iterable of bytestrings for the file. The
2209
kind of iterable and length of the bytestrings are unspecified, but for
2210
this implementation, it is a list of bytes produced by
2211
VersionedFile.get_record_stream().
2212
2213
942
:param desired_files: a list of (file_id, revision_id, identifier)
2214
943
triples
2215
944
"""
2216
text_keys = {}
2217
for file_id, revision_id, callable_data in desired_files:
2218
text_keys[(file_id, revision_id)] = callable_data
2219
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2220
if record.storage_kind == 'absent':
2221
raise errors.RevisionNotPresent(record.key, self)
2222
yield text_keys[record.key], record.get_bytes_as('chunked')
2223
2224
def _generate_text_key_index(self, text_key_references=None,
2225
ancestors=None):
2226
"""Generate a new text key index for the repository.
2227
2228
This is an expensive function that will take considerable time to run.
2229
2230
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2231
list of parents, also text keys. When a given key has no parents,
2232
the parents list will be [NULL_REVISION].
2233
"""
2234
# All revisions, to find inventory parents.
2235
if ancestors is None:
2236
graph = self.get_graph()
2237
ancestors = graph.get_parent_map(self.all_revision_ids())
2238
if text_key_references is None:
2239
text_key_references = self.find_text_key_references()
2240
pb = ui.ui_factory.nested_progress_bar()
2241
try:
2242
return self._do_generate_text_key_index(ancestors,
2243
text_key_references, pb)
2244
finally:
2245
pb.finished()
2246
2247
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2248
"""Helper for _generate_text_key_index to avoid deep nesting."""
2249
revision_order = tsort.topo_sort(ancestors)
2250
invalid_keys = set()
2251
revision_keys = {}
2252
for revision_id in revision_order:
2253
revision_keys[revision_id] = set()
2254
text_count = len(text_key_references)
2255
# a cache of the text keys to allow reuse; costs a dict of all the
2256
# keys, but saves a 2-tuple for every child of a given key.
2257
text_key_cache = {}
2258
for text_key, valid in text_key_references.iteritems():
2259
if not valid:
2260
invalid_keys.add(text_key)
2261
else:
2262
revision_keys[text_key[1]].add(text_key)
2263
text_key_cache[text_key] = text_key
2264
del text_key_references
2265
text_index = {}
2266
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2267
NULL_REVISION = _mod_revision.NULL_REVISION
2268
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2269
# too small for large or very branchy trees. However, for 55K path
2270
# trees, it would be easy to use too much memory trivially. Ideally we
2271
# could gauge this by looking at available real memory etc, but this is
2272
# always a tricky proposition.
2273
inventory_cache = lru_cache.LRUCache(10)
2274
batch_size = 10 # should be ~150MB on a 55K path tree
2275
batch_count = len(revision_order) / batch_size + 1
2276
processed_texts = 0
2277
pb.update("Calculating text parents", processed_texts, text_count)
2278
for offset in xrange(batch_count):
2279
to_query = revision_order[offset * batch_size:(offset + 1) *
2280
batch_size]
2281
if not to_query:
2282
break
2283
for revision_id in to_query:
2284
parent_ids = ancestors[revision_id]
2285
for text_key in revision_keys[revision_id]:
2286
pb.update("Calculating text parents", processed_texts)
2287
processed_texts += 1
2288
candidate_parents = []
2289
for parent_id in parent_ids:
2290
parent_text_key = (text_key[0], parent_id)
2291
try:
2292
check_parent = parent_text_key not in \
2293
revision_keys[parent_id]
2294
except KeyError:
2295
# the parent parent_id is a ghost:
2296
check_parent = False
2297
# truncate the derived graph against this ghost.
2298
parent_text_key = None
2299
if check_parent:
2300
# look at the parent commit details inventories to
2301
# determine possible candidates in the per file graph.
2302
# TODO: cache here.
2303
try:
2304
inv = inventory_cache[parent_id]
2305
except KeyError:
2306
inv = self.revision_tree(parent_id).inventory
2307
inventory_cache[parent_id] = inv
2308
try:
2309
parent_entry = inv[text_key[0]]
2310
except (KeyError, errors.NoSuchId):
2311
parent_entry = None
2312
if parent_entry is not None:
2313
parent_text_key = (
2314
text_key[0], parent_entry.revision)
2315
else:
2316
parent_text_key = None
2317
if parent_text_key is not None:
2318
candidate_parents.append(
2319
text_key_cache[parent_text_key])
2320
parent_heads = text_graph.heads(candidate_parents)
2321
new_parents = list(parent_heads)
2322
new_parents.sort(key=lambda x:candidate_parents.index(x))
2323
if new_parents == []:
2324
new_parents = [NULL_REVISION]
2325
text_index[text_key] = new_parents
2326
2327
for text_key in invalid_keys:
2328
text_index[text_key] = [NULL_REVISION]
2329
return text_index
2330
2331
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2332
"""Get an iterable listing the keys of all the data introduced by a set
2333
of revision IDs.
2334
2335
The keys will be ordered so that the corresponding items can be safely
2336
fetched and inserted in that order.
2337
2338
:returns: An iterable producing tuples of (knit-kind, file-id,
2339
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2340
'revisions'. file-id is None unless knit-kind is 'file'.
2341
"""
2342
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2343
yield result
2344
del _files_pb
2345
for result in self._find_non_file_keys_to_fetch(revision_ids):
2346
yield result
2347
2348
def _find_file_keys_to_fetch(self, revision_ids, pb):
2349
# XXX: it's a bit weird to control the inventory weave caching in this
2350
# generator. Ideally the caching would be done in fetch.py I think. Or
2351
# maybe this generator should explicitly have the contract that it
2352
# should not be iterated until the previously yielded item has been
2353
# processed?
2354
inv_w = self.inventories
2355
2356
# file ids that changed
2357
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2358
count = 0
2359
num_file_ids = len(file_ids)
2360
for file_id, altered_versions in file_ids.iteritems():
2361
if pb is not None:
2362
pb.update("Fetch texts", count, num_file_ids)
2363
count += 1
2364
yield ("file", file_id, altered_versions)
2365
2366
def _find_non_file_keys_to_fetch(self, revision_ids):
2367
# inventory
2368
yield ("inventory", None, revision_ids)
2369
2370
# signatures
2371
# XXX: Note ATM no callers actually pay attention to this return
2372
# instead they just use the list of revision ids and ignore
2373
# missing sigs. Consider removing this work entirely
2374
revisions_with_signatures = set(self.signatures.get_parent_map(
2375
[(r,) for r in revision_ids]))
2376
revisions_with_signatures = set(
2377
[r for (r,) in revisions_with_signatures])
2378
revisions_with_signatures.intersection_update(revision_ids)
2379
yield ("signatures", None, revisions_with_signatures)
2380
2381
# revisions
2382
yield ("revisions", None, revision_ids)
2383
2384
@needs_read_lock
2385
def get_inventory(self, revision_id):
2386
"""Get Inventory object by revision id."""
2387
return self.iter_inventories([revision_id]).next()
2388
2389
def iter_inventories(self, revision_ids, ordering=None):
2390
"""Get many inventories by revision_ids.
2391
2392
This will buffer some or all of the texts used in constructing the
2393
inventories in memory, but will only parse a single inventory at a
2394
time.
2395
2396
:param revision_ids: The expected revision ids of the inventories.
2397
:param ordering: optional ordering, e.g. 'topological'. If not
2398
specified, the order of revision_ids will be preserved (by
2399
buffering if necessary).
2400
:return: An iterator of inventories.
2401
"""
2402
if ((None in revision_ids)
2403
or (_mod_revision.NULL_REVISION in revision_ids)):
2404
raise ValueError('cannot get null revision inventory')
2405
return self._iter_inventories(revision_ids, ordering)
2406
2407
def _iter_inventories(self, revision_ids, ordering):
2408
"""single-document based inventory iteration."""
2409
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2410
for text, revision_id in inv_xmls:
2411
yield self._deserialise_inventory(revision_id, text)
2412
2413
def _iter_inventory_xmls(self, revision_ids, ordering):
2414
if ordering is None:
2415
order_as_requested = True
2416
ordering = 'unordered'
2417
else:
2418
order_as_requested = False
2419
keys = [(revision_id,) for revision_id in revision_ids]
2420
if not keys:
2421
return
2422
if order_as_requested:
2423
key_iter = iter(keys)
2424
next_key = key_iter.next()
2425
stream = self.inventories.get_record_stream(keys, ordering, True)
2426
text_chunks = {}
2427
for record in stream:
2428
if record.storage_kind != 'absent':
2429
chunks = record.get_bytes_as('chunked')
2430
if order_as_requested:
2431
text_chunks[record.key] = chunks
2432
else:
2433
yield ''.join(chunks), record.key[-1]
2434
else:
2435
raise errors.NoSuchRevision(self, record.key)
2436
if order_as_requested:
2437
# Yield as many results as we can while preserving order.
2438
while next_key in text_chunks:
2439
chunks = text_chunks.pop(next_key)
2440
yield ''.join(chunks), next_key[-1]
2441
try:
2442
next_key = key_iter.next()
2443
except StopIteration:
2444
# We still want to fully consume the get_record_stream,
2445
# just in case it is not actually finished at this point
2446
next_key = None
2447
break
2448
2449
def _deserialise_inventory(self, revision_id, xml):
2450
"""Transform the xml into an inventory object.
2451
2452
:param revision_id: The expected revision id of the inventory.
2453
:param xml: A serialised inventory.
2454
"""
2455
result = self._serializer.read_inventory_from_string(xml, revision_id,
2456
entry_cache=self._inventory_entry_cache,
2457
return_from_cache=self._safe_to_return_from_cache)
2458
if result.revision_id != revision_id:
2459
raise AssertionError('revision id mismatch %s != %s' % (
2460
result.revision_id, revision_id))
2461
return result
2462
2463
def get_serializer_format(self):
2464
return self._serializer.format_num
2465
2466
@needs_read_lock
2467
def _get_inventory_xml(self, revision_id):
2468
"""Get serialized inventory as a string."""
2469
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2470
try:
2471
text, revision_id = texts.next()
2472
except StopIteration:
2473
raise errors.HistoryMissing(self, 'inventory', revision_id)
2474
return text
945
raise NotImplementedError(self.iter_files_bytes)
2475
946
2476
947
def get_rev_id_for_revno(self, revno, known_pair):
2477
948
"""Return the revision id of a revno, given a later (revno, revid)
2508
979
raise AssertionError('_iter_for_revno returned too much history')
2509
980
return (True, partial_history[-1])
2510
981
982
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
2511
983
def iter_reverse_revision_history(self, revision_id):
2512
984
"""Iterate backwards through revision ids in the lefthand history
2513
985
2550
1022
2551
1023
`revision_id` may be NULL_REVISION for the empty tree revision.
2552
1024
"""
2553
revision_id = _mod_revision.ensure_null(revision_id)
2554
# TODO: refactor this to use an existing revision object
2555
# so we don't need to read it in twice.
2556
if revision_id == _mod_revision.NULL_REVISION:
2557
return RevisionTree(self, Inventory(root_id=None),
2558
_mod_revision.NULL_REVISION)
2559
else:
2560
inv = self.get_inventory(revision_id)
2561
return RevisionTree(self, inv, revision_id)
1025
raise NotImplementedError(self.revision_tree)
2562
1026
2563
1027
def revision_trees(self, revision_ids):
2564
1028
"""Return Trees for revisions in this repository.
2566
1030
:param revision_ids: a sequence of revision-ids;
2567
1031
a revision-id may not be None or 'null:'
2568
1032
"""
2569
inventories = self.iter_inventories(revision_ids)
2570
for inv in inventories:
2571
yield RevisionTree(self, inv, inv.revision_id)
2572
2573
def _filtered_revision_trees(self, revision_ids, file_ids):
2574
"""Return Tree for a revision on this branch with only some files.
2575
2576
:param revision_ids: a sequence of revision-ids;
2577
a revision-id may not be None or 'null:'
2578
:param file_ids: if not None, the result is filtered
2579
so that only those file-ids, their parents and their
2580
children are included.
2581
"""
2582
inventories = self.iter_inventories(revision_ids)
2583
for inv in inventories:
2584
# Should we introduce a FilteredRevisionTree class rather
2585
# than pre-filter the inventory here?
2586
filtered_inv = inv.filter(file_ids)
2587
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
1033
raise NotImplementedError(self.revision_trees)
2588
1034
2589
1035
@needs_read_lock
1036
@symbol_versioning.deprecated_method(
1037
symbol_versioning.deprecated_in((2, 4, 0)))
2590
1038
def get_ancestry(self, revision_id, topo_sorted=True):
2591
1039
"""Return a list of revision-ids integrated by a revision.
2592
1040
2596
1044
2597
1045
This is topologically sorted.
2598
1046
"""
1047
if 'evil' in debug.debug_flags:
1048
mutter_callsite(2, "get_ancestry is linear with history.")
2599
1049
if _mod_revision.is_null(revision_id):
2600
1050
return [None]
2601
1051
if not self.has_revision(revision_id):
2642
1092
2643
1093
def get_parent_map(self, revision_ids):
2644
1094
"""See graph.StackedParentsProvider.get_parent_map"""
1095
raise NotImplementedError(self.get_parent_map)
1096
1097
def _get_parent_map_no_fallbacks(self, revision_ids):
1098
"""Same as Repository.get_parent_map except doesn't query fallbacks."""
2645
1099
# revisions index works in keys; this just works in revisions
2646
1100
# therefore wrap and unwrap
2647
1101
query_keys = []
2653
1107
raise ValueError('get_parent_map(None) is not valid')
2654
1108
else:
2655
1109
query_keys.append((revision_id ,))
1110
vf = self.revisions.without_fallbacks()
2656
1111
for ((revision_id,), parent_keys) in \
2657
self.revisions.get_parent_map(query_keys).iteritems():
1112
vf.get_parent_map(query_keys).iteritems():
2658
1113
if parent_keys:
2659
1114
result[revision_id] = tuple([parent_revid
2660
1115
for (parent_revid,) in parent_keys])
2663
1118
return result
2664
1119
2665
1120
def _make_parents_provider(self):
2666
return self
1121
if not self._format.supports_external_lookups:
1122
return self
1123
return graph.StackedParentsProvider(_LazyListJoin(
1124
[self._make_parents_provider_unstacked()],
1125
self._fallback_repositories))
1126
1127
def _make_parents_provider_unstacked(self):
1128
return graph.CallableToParentsProviderAdapter(
1129
self._get_parent_map_no_fallbacks)
2667
1130
2668
1131
@needs_read_lock
2669
1132
def get_known_graph_ancestry(self, revision_ids):
2670
1133
"""Return the known graph for a set of revision ids and their ancestors.
2671
1134
"""
2672
st = static_tuple.StaticTuple
2673
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2674
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2675
return graph.GraphThunkIdsToKeys(known_graph)
1135
raise NotImplementedError(self.get_known_graph_ancestry)
1136
1137
def get_file_graph(self):
1138
"""Return the graph walker for files."""
1139
raise NotImplementedError(self.get_file_graph)
2676
1140
2677
1141
def get_graph(self, other_repository=None):
2678
1142
"""Return the graph walker for this repository format"""
2683
1147
[parents_provider, other_repository._make_parents_provider()])
2684
1148
return graph.Graph(parents_provider)
2685
1149
2686
def _get_versioned_file_checker(self, text_key_references=None,
2687
ancestors=None):
2688
"""Return an object suitable for checking versioned files.
2689
2690
:param text_key_references: if non-None, an already built
2691
dictionary mapping text keys ((fileid, revision_id) tuples)
2692
to whether they were referred to by the inventory of the
2693
revision_id that they contain. If None, this will be
2694
calculated.
2695
:param ancestors: Optional result from
2696
self.get_graph().get_parent_map(self.all_revision_ids()) if already
2697
available.
2698
"""
2699
return _VersionedFileChecker(self,
2700
text_key_references=text_key_references, ancestors=ancestors)
2701
2702
def revision_ids_to_search_result(self, result_set):
2703
"""Convert a set of revision ids to a graph SearchResult."""
2704
result_parents = set()
2705
for parents in self.get_graph().get_parent_map(
2706
result_set).itervalues():
2707
result_parents.update(parents)
2708
included_keys = result_set.intersection(result_parents)
2709
start_keys = result_set.difference(included_keys)
2710
exclude_keys = result_parents.difference(result_set)
2711
result = graph.SearchResult(start_keys, exclude_keys,
2712
len(result_set), result_set)
2713
return result
2714
2715
1150
@needs_write_lock
2716
1151
def set_make_working_trees(self, new_value):
2717
1152
"""Set the policy flag for making working trees when creating branches.
2730
1165
2731
1166
@needs_write_lock
2732
1167
def sign_revision(self, revision_id, gpg_strategy):
2733
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1168
testament = _mod_testament.Testament.from_revision(self, revision_id)
1169
plaintext = testament.as_short_text()
2734
1170
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2735
1171
2736
1172
@needs_read_lock
1173
def verify_revision_signature(self, revision_id, gpg_strategy):
1174
"""Verify the signature on a revision.
1175
1176
:param revision_id: the revision to verify
1177
:gpg_strategy: the GPGStrategy object to used
1178
1179
:return: gpg.SIGNATURE_VALID or a failed SIGNATURE_ value
1180
"""
1181
if not self.has_signature_for_revision_id(revision_id):
1182
return gpg.SIGNATURE_NOT_SIGNED, None
1183
signature = self.get_signature_text(revision_id)
1184
1185
testament = _mod_testament.Testament.from_revision(self, revision_id)
1186
plaintext = testament.as_short_text()
1187
1188
return gpg_strategy.verify(signature, plaintext)
1189
2737
1190
def has_signature_for_revision_id(self, revision_id):
2738
1191
"""Query for a revision signature for revision_id in the repository."""
2739
if not self.has_revision(revision_id):
2740
raise errors.NoSuchRevision(self, revision_id)
2741
sig_present = (1 == len(
2742
self.signatures.get_parent_map([(revision_id,)])))
2743
return sig_present
1192
raise NotImplementedError(self.has_signature_for_revision_id)
2744
1193
2745
@needs_read_lock
2746
1194
def get_signature_text(self, revision_id):
2747
1195
"""Return the text for a signature."""
2748
stream = self.signatures.get_record_stream([(revision_id,)],
2749
'unordered', True)
2750
record = stream.next()
2751
if record.storage_kind == 'absent':
2752
raise errors.NoSuchRevision(self, revision_id)
2753
return record.get_bytes_as('fulltext')
1196
raise NotImplementedError(self.get_signature_text)
2754
1197
2755
@needs_read_lock
2756
1198
def check(self, revision_ids=None, callback_refs=None, check_repo=True):
2757
1199
"""Check consistency of all history of given revision_ids.
2758
1200
2766
1208
:param check_repo: If False do not check the repository contents, just
2767
1209
calculate the data callback_refs requires and call them back.
2768
1210
"""
2769
return self._check(revision_ids, callback_refs=callback_refs,
1211
return self._check(revision_ids=revision_ids, callback_refs=callback_refs,
2770
1212
check_repo=check_repo)
2771
1213
2772
def _check(self, revision_ids, callback_refs, check_repo):
2773
result = check.Check(self, check_repo=check_repo)
2774
result.check(callback_refs)
2775
return result
1214
def _check(self, revision_ids=None, callback_refs=None, check_repo=True):
1215
raise NotImplementedError(self.check)
2776
1216
2777
1217
def _warn_if_deprecated(self, branch=None):
1218
if not self._format.is_deprecated():
1219
return
2778
1220
global _deprecation_warning_done
2779
1221
if _deprecation_warning_done:
2780
1222
return
2810
1252
except UnicodeDecodeError:
2811
1253
raise errors.NonAsciiRevisionId(method, self)
2812
1254
2813
def revision_graph_can_have_wrong_parents(self):
2814
"""Is it possible for this repository to have a revision graph with
2815
incorrect parents?
2816
2817
If True, then this repository must also implement
2818
_find_inconsistent_revision_parents so that check and reconcile can
2819
check for inconsistencies before proceeding with other checks that may
2820
depend on the revision index being consistent.
2821
"""
2822
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2823
2824
2825
# remove these delegates a while after bzr 0.15
2826
def __make_delegated(name, from_module):
2827
def _deprecated_repository_forwarder():
2828
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2829
% (name, from_module),
2830
DeprecationWarning,
2831
stacklevel=2)
2832
try:
2833
return pyutils.get_named_object(from_module, name)
2834
except AttributeError:
2835
raise AttributeError('module %s has no name %s'
2836
% (sys.modules[from_module], name))
2837
globals()[name] = _deprecated_repository_forwarder
2838
2839
for _name in [
2840
'AllInOneRepository',
2841
'WeaveMetaDirRepository',
2842
'PreSplitOutRepositoryFormat',
2843
'RepositoryFormat4',
2844
'RepositoryFormat5',
2845
'RepositoryFormat6',
2846
'RepositoryFormat7',
2847
]:
2848
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2849
2850
for _name in [
2851
'KnitRepository',
2852
'RepositoryFormatKnit',
2853
'RepositoryFormatKnit1',
2854
]:
2855
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2856
2857
2858
def install_revision(repository, rev, revision_tree):
2859
"""Install all revision data into a repository."""
2860
install_revisions(repository, [(rev, revision_tree, None)])
2861
2862
2863
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2864
"""Install all revision data into a repository.
2865
2866
Accepts an iterable of revision, tree, signature tuples. The signature
2867
may be None.
2868
"""
2869
repository.start_write_group()
2870
try:
2871
inventory_cache = lru_cache.LRUCache(10)
2872
for n, (revision, revision_tree, signature) in enumerate(iterable):
2873
_install_revision(repository, revision, revision_tree, signature,
2874
inventory_cache)
2875
if pb is not None:
2876
pb.update('Transferring revisions', n + 1, num_revisions)
2877
except:
2878
repository.abort_write_group()
2879
raise
2880
else:
2881
repository.commit_write_group()
2882
2883
2884
def _install_revision(repository, rev, revision_tree, signature,
2885
inventory_cache):
2886
"""Install all revision data into a repository."""
2887
present_parents = []
2888
parent_trees = {}
2889
for p_id in rev.parent_ids:
2890
if repository.has_revision(p_id):
2891
present_parents.append(p_id)
2892
parent_trees[p_id] = repository.revision_tree(p_id)
2893
else:
2894
parent_trees[p_id] = repository.revision_tree(
2895
_mod_revision.NULL_REVISION)
2896
2897
inv = revision_tree.inventory
2898
entries = inv.iter_entries()
2899
# backwards compatibility hack: skip the root id.
2900
if not repository.supports_rich_root():
2901
path, root = entries.next()
2902
if root.revision != rev.revision_id:
2903
raise errors.IncompatibleRevision(repr(repository))
2904
text_keys = {}
2905
for path, ie in entries:
2906
text_keys[(ie.file_id, ie.revision)] = ie
2907
text_parent_map = repository.texts.get_parent_map(text_keys)
2908
missing_texts = set(text_keys) - set(text_parent_map)
2909
# Add the texts that are not already present
2910
for text_key in missing_texts:
2911
ie = text_keys[text_key]
2912
text_parents = []
2913
# FIXME: TODO: The following loop overlaps/duplicates that done by
2914
# commit to determine parents. There is a latent/real bug here where
2915
# the parents inserted are not those commit would do - in particular
2916
# they are not filtered by heads(). RBC, AB
2917
for revision, tree in parent_trees.iteritems():
2918
if ie.file_id not in tree:
2919
continue
2920
parent_id = tree.inventory[ie.file_id].revision
2921
if parent_id in text_parents:
2922
continue
2923
text_parents.append((ie.file_id, parent_id))
2924
lines = revision_tree.get_file(ie.file_id).readlines()
2925
repository.texts.add_lines(text_key, text_parents, lines)
2926
try:
2927
# install the inventory
2928
if repository._format._commit_inv_deltas and len(rev.parent_ids):
2929
# Cache this inventory
2930
inventory_cache[rev.revision_id] = inv
2931
try:
2932
basis_inv = inventory_cache[rev.parent_ids[0]]
2933
except KeyError:
2934
repository.add_inventory(rev.revision_id, inv, present_parents)
2935
else:
2936
delta = inv._make_delta(basis_inv)
2937
repository.add_inventory_by_delta(rev.parent_ids[0], delta,
2938
rev.revision_id, present_parents)
2939
else:
2940
repository.add_inventory(rev.revision_id, inv, present_parents)
2941
except errors.RevisionAlreadyPresent:
2942
pass
2943
if signature is not None:
2944
repository.add_signature_text(rev.revision_id, signature)
2945
repository.add_revision(rev.revision_id, rev, inv)
2946
2947
1255
2948
1256
class MetaDirRepository(Repository):
2949
1257
"""Repositories in the new meta-dir layout.
2984
1292
return not self._transport.has('no-working-trees')
2985
1293
2986
1294
2987
class MetaDirVersionedFileRepository(MetaDirRepository):
2988
"""Repositories in a meta-dir, that work via versioned file objects."""
1295
class RepositoryFormatRegistry(controldir.ControlComponentFormatRegistry):
1296
"""Repository format registry."""
2989
1297
2990
def __init__(self, _format, a_bzrdir, control_files):
2991
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2992
control_files)
1298
def get_default(self):
1299
"""Return the current default format."""
1300
return controldir.format_registry.make_bzrdir('default').repository_format
2993
1301
2994
1302
2995
1303
network_format_registry = registry.FormatRegistry()
3001
1309
"""
3002
1310
3003
1311
3004
format_registry = registry.FormatRegistry(network_format_registry)
1312
format_registry = RepositoryFormatRegistry(network_format_registry)
3005
1313
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
3006
1314
3007
1315
This can contain either format instances themselves, or classes/factories that
3012
1320
#####################################################################
3013
1321
# Repository Formats
3014
1322
3015
class RepositoryFormat(object):
1323
class RepositoryFormat(controldir.ControlComponentFormat):
3016
1324
"""A repository format.
3017
1325
3018
1326
Formats provide four things:
3038
1346
created.
3039
1347
3040
1348
Common instance attributes:
3041
_matchingbzrdir - the bzrdir format that the repository format was
1349
_matchingbzrdir - the controldir format that the repository format was
3042
1350
originally written to work with. This can be used if manually
3043
1351
constructing a bzrdir and repository, or more commonly for test suite
3044
1352
parameterization.
3053
1361
# Does this format support CHK bytestring lookups. Set to True or False in
3054
1362
# derived classes.
3055
1363
supports_chks = None
3056
# Should commit add an inventory, or an inventory delta to the repository.
3057
_commit_inv_deltas = True
3058
# What order should fetch operations request streams in?
3059
# The default is unordered as that is the cheapest for an origin to
3060
# provide.
3061
_fetch_order = 'unordered'
3062
# Does this repository format use deltas that can be fetched as-deltas ?
3063
# (E.g. knits, where the knit deltas can be transplanted intact.
3064
# We default to False, which will ensure that enough data to get
3065
# a full text out of any fetch stream will be grabbed.
3066
_fetch_uses_deltas = False
3067
1364
# Should fetch trigger a reconcile after the fetch? Only needed for
3068
1365
# some repository formats that can suffer internal inconsistencies.
3069
1366
_fetch_reconcile = False
3075
1372
# help), and for fetching when data won't have come from the same
3076
1373
# compressor.
3077
1374
pack_compresses = False
3078
# Does the repository inventory storage understand references to trees?
1375
# Does the repository storage understand references to trees?
3079
1376
supports_tree_reference = None
3080
1377
# Is the format experimental ?
3081
1378
experimental = False
1379
# Does this repository format escape funky characters, or does it create
1380
# files with similar names as the versioned files in its contents on disk
1381
# ?
1382
supports_funky_characters = None
1383
# Does this repository format support leaving locks?
1384
supports_leaving_lock = None
1385
# Does this format support the full VersionedFiles interface?
1386
supports_full_versioned_files = None
1387
# Does this format support signing revision signatures?
1388
supports_revision_signatures = True
1389
# Can the revision graph have incorrect parents?
1390
revision_graph_can_have_wrong_parents = None
1391
# Does this format support rich root data?
1392
rich_root_data = None
1393
# Does this format support explicitly versioned directories?
1394
supports_versioned_directories = None
1395
# Can other repositories be nested into one of this format?
1396
supports_nesting_repositories = None
1397
# Is it possible for revisions to be present without being referenced
1398
# somewhere ?
1399
supports_unreferenced_revisions = None
3082
1400
3083
1401
def __repr__(self):
3084
1402
return "%s()" % self.__class__.__name__
3091
1409
return not self == other
3092
1410
3093
1411
@classmethod
3094
def find_format(klass, a_bzrdir):
3095
"""Return the format for the repository object in a_bzrdir.
3096
3097
This is used by bzr native formats that have a "format" file in
3098
the repository. Other methods may be used by different types of
3099
control directory.
3100
"""
3101
try:
3102
transport = a_bzrdir.get_repository_transport(None)
3103
format_string = transport.get_bytes("format")
3104
return format_registry.get(format_string)
3105
except errors.NoSuchFile:
3106
raise errors.NoRepositoryPresent(a_bzrdir)
3107
except KeyError:
3108
raise errors.UnknownFormatError(format=format_string,
3109
kind='repository')
3110
3111
@classmethod
1412
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3112
1413
def register_format(klass, format):
3113
format_registry.register(format.get_format_string(), format)
1414
format_registry.register(format)
3114
1415
3115
1416
@classmethod
1417
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3116
1418
def unregister_format(klass, format):
3117
format_registry.remove(format.get_format_string())
1419
format_registry.remove(format)
3118
1420
3119
1421
@classmethod
1422
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3120
1423
def get_default_format(klass):
3121
1424
"""Return the current default format."""
3122
from bzrlib import bzrdir
3123
return bzrdir.format_registry.make_bzrdir('default').repository_format
3124
3125
def get_format_string(self):
3126
"""Return the ASCII format string that identifies this format.
3127
3128
Note that in pre format ?? repositories the format string is
3129
not permitted nor written to disk.
3130
"""
3131
raise NotImplementedError(self.get_format_string)
1425
return format_registry.get_default()
3132
1426
3133
1427
def get_format_description(self):
3134
1428
"""Return the short description for this format."""
3135
1429
raise NotImplementedError(self.get_format_description)
3136
1430
3137
# TODO: this shouldn't be in the base class, it's specific to things that
3138
# use weaves or knits -- mbp 20070207
3139
def _get_versioned_file_store(self,
3140
name,
3141
transport,
3142
control_files,
3143
prefixed=True,
3144
versionedfile_class=None,
3145
versionedfile_kwargs={},
3146
escaped=False):
3147
if versionedfile_class is None:
3148
versionedfile_class = self._versionedfile_class
3149
weave_transport = control_files._transport.clone(name)
3150
dir_mode = control_files._dir_mode
3151
file_mode = control_files._file_mode
3152
return VersionedFileStore(weave_transport, prefixed=prefixed,
3153
dir_mode=dir_mode,
3154
file_mode=file_mode,
3155
versionedfile_class=versionedfile_class,
3156
versionedfile_kwargs=versionedfile_kwargs,
3157
escaped=escaped)
3158
3159
def initialize(self, a_bzrdir, shared=False):
3160
"""Initialize a repository of this format in a_bzrdir.
3161
3162
:param a_bzrdir: The bzrdir to put the new repository in it.
1431
def initialize(self, controldir, shared=False):
1432
"""Initialize a repository of this format in controldir.
1433
1434
:param controldir: The controldir to put the new repository in it.
3163
1435
:param shared: The repository should be initialized as a sharable one.
3164
1436
:returns: The new repository object.
3165
1437
3166
1438
This may raise UninitializableFormat if shared repository are not
3167
compatible the a_bzrdir.
1439
compatible the controldir.
3168
1440
"""
3169
1441
raise NotImplementedError(self.initialize)
3170
1442
3177
1449
"""
3178
1450
return True
3179
1451
1452
def is_deprecated(self):
1453
"""Is this format deprecated?
1454
1455
Deprecated formats may trigger a user-visible warning recommending
1456
the user to upgrade. They are still fully supported.
1457
"""
1458
return False
1459
3180
1460
def network_name(self):
3181
1461
"""A simple byte string uniquely identifying this format for RPC calls.
3182
1462
3198
1478
'Does not support nested trees', target_format,
3199
1479
from_format=self)
3200
1480
3201
def open(self, a_bzrdir, _found=False):
3202
"""Return an instance of this format for the bzrdir a_bzrdir.
1481
def open(self, controldir, _found=False):
1482
"""Return an instance of this format for a controldir.
3203
1483
3204
1484
_found is a private parameter, do not use it.
3205
1485
"""
3206
1486
raise NotImplementedError(self.open)
3207
1487
3208
def _run_post_repo_init_hooks(self, repository, a_bzrdir, shared):
3209
from bzrlib.bzrdir import BzrDir, RepoInitHookParams
3210
hooks = BzrDir.hooks['post_repo_init']
1488
def _run_post_repo_init_hooks(self, repository, controldir, shared):
1489
from bzrlib.controldir import ControlDir, RepoInitHookParams
1490
hooks = ControlDir.hooks['post_repo_init']
3211
1491
if not hooks:
3212
1492
return
3213
params = RepoInitHookParams(repository, self, a_bzrdir, shared)
1493
params = RepoInitHookParams(repository, self, controldir, shared)
3214
1494
for hook in hooks:
3215
1495
hook(params)
3216
1496
3217
1497
3218
class MetaDirRepositoryFormat(RepositoryFormat):
1498
class RepositoryFormatMetaDir(bzrdir.BzrDirMetaComponentFormat, RepositoryFormat):
3219
1499
"""Common base class for the new repositories using the metadir layout."""
3220
1500
3221
1501
rich_root_data = False
3222
1502
supports_tree_reference = False
3223
1503
supports_external_lookups = False
1504
supports_leaving_lock = True
1505
supports_nesting_repositories = True
3224
1506
3225
1507
@property
3226
1508
def _matchingbzrdir(self):
3229
1511
return matching
3230
1512
3231
1513
def __init__(self):
3232
super(MetaDirRepositoryFormat, self).__init__()
1514
RepositoryFormat.__init__(self)
1515
bzrdir.BzrDirMetaComponentFormat.__init__(self)
3233
1516
3234
1517
def _create_control_files(self, a_bzrdir):
3235
1518
"""Create the required files and the initial control_files object."""
3259
1542
finally:
3260
1543
control_files.unlock()
3261
1544
3262
def network_name(self):
3263
"""Metadir formats have matching disk and network format strings."""
3264
return self.get_format_string()
3265
3266
3267
# Pre-0.8 formats that don't have a disk format string (because they are
3268
# versioned by the matching control directory). We use the control directories
3269
# disk format string as a key for the network_name because they meet the
3270
# constraints (simple string, unique, immutable).
3271
network_format_registry.register_lazy(
3272
"Bazaar-NG branch, format 5\n",
3273
'bzrlib.repofmt.weaverepo',
3274
'RepositoryFormat5',
3275
)
3276
network_format_registry.register_lazy(
3277
"Bazaar-NG branch, format 6\n",
3278
'bzrlib.repofmt.weaverepo',
3279
'RepositoryFormat6',
3280
)
1545
@classmethod
1546
def find_format(klass, a_bzrdir):
1547
"""Return the format for the repository object in a_bzrdir.
1548
1549
This is used by bzr native formats that have a "format" file in
1550
the repository. Other methods may be used by different types of
1551
control directory.
1552
"""
1553
try:
1554
transport = a_bzrdir.get_repository_transport(None)
1555
format_string = transport.get_bytes("format")
1556
except errors.NoSuchFile:
1557
raise errors.NoRepositoryPresent(a_bzrdir)
1558
return klass._find_format(format_registry, 'repository', format_string)
1559
3281
1560
3282
1561
# formats which have no format string are not discoverable or independently
3283
1562
# creatable on disk, so are not registered in format_registry. They're
3284
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
3285
# needed, it's constructed directly by the BzrDir. Non-native formats where
1563
# all in bzrlib.repofmt.knitreponow. When an instance of one of these is
1564
# needed, it's constructed directly by the ControlDir. Non-native formats where
3286
1565
# the repository is not separately opened are similar.
3287
1566
3288
1567
format_registry.register_lazy(
3289
'Bazaar-NG Repository format 7',
3290
'bzrlib.repofmt.weaverepo',
3291
'RepositoryFormat7'
3292
)
3293
3294
format_registry.register_lazy(
3295
1568
'Bazaar-NG Knit Repository Format 1',
3296
1569
'bzrlib.repofmt.knitrepo',
3297
1570
'RepositoryFormatKnit1',
3314
1587
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
3315
1588
format_registry.register_lazy(
3316
1589
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
3317
'bzrlib.repofmt.pack_repo',
1590
'bzrlib.repofmt.knitpack_repo',
3318
1591
'RepositoryFormatKnitPack1',
3319
1592
)
3320
1593
format_registry.register_lazy(
3321
1594
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
3322
'bzrlib.repofmt.pack_repo',
1595
'bzrlib.repofmt.knitpack_repo',
3323
1596
'RepositoryFormatKnitPack3',
3324
1597
)
3325
1598
format_registry.register_lazy(
3326
1599
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
3327
'bzrlib.repofmt.pack_repo',
1600
'bzrlib.repofmt.knitpack_repo',
3328
1601
'RepositoryFormatKnitPack4',
3329
1602
)
3330
1603
format_registry.register_lazy(
3331
1604
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
3332
'bzrlib.repofmt.pack_repo',
1605
'bzrlib.repofmt.knitpack_repo',
3333
1606
'RepositoryFormatKnitPack5',
3334
1607
)
3335
1608
format_registry.register_lazy(
3336
1609
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
3337
'bzrlib.repofmt.pack_repo',
1610
'bzrlib.repofmt.knitpack_repo',
3338
1611
'RepositoryFormatKnitPack5RichRoot',
3339
1612
)
3340
1613
format_registry.register_lazy(
3341
1614
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
3342
'bzrlib.repofmt.pack_repo',
1615
'bzrlib.repofmt.knitpack_repo',
3343
1616
'RepositoryFormatKnitPack5RichRootBroken',
3344
1617
)
3345
1618
format_registry.register_lazy(
3346
1619
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
3347
'bzrlib.repofmt.pack_repo',
1620
'bzrlib.repofmt.knitpack_repo',
3348
1621
'RepositoryFormatKnitPack6',
3349
1622
)
3350
1623
format_registry.register_lazy(
3351
1624
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
3352
'bzrlib.repofmt.pack_repo',
1625
'bzrlib.repofmt.knitpack_repo',
3353
1626
'RepositoryFormatKnitPack6RichRoot',
3354
1627
)
1628
format_registry.register_lazy(
1629
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
1630
'bzrlib.repofmt.groupcompress_repo',
1631
'RepositoryFormat2a',
1632
)
3355
1633
3356
1634
# Development formats.
3357
# Obsolete but kept pending a CHK based subtree format.
1635
# Check their docstrings to see if/when they are obsolete.
3358
1636
format_registry.register_lazy(
3359
1637
("Bazaar development format 2 with subtree support "
3360
1638
"(needs bzr.dev from before 1.8)\n"),
3361
'bzrlib.repofmt.pack_repo',
1639
'bzrlib.repofmt.knitpack_repo',
3362
1640
'RepositoryFormatPackDevelopment2Subtree',
3363
1641
)
3364
3365
# 1.14->1.16 go below here
3366
format_registry.register_lazy(
3367
'Bazaar development format - group compression and chk inventory'
3368
' (needs bzr.dev from 1.14)\n',
3369
'bzrlib.repofmt.groupcompress_repo',
3370
'RepositoryFormatCHK1',
3371
)
3372
3373
format_registry.register_lazy(
3374
'Bazaar development format - chk repository with bencode revision '
3375
'serialization (needs bzr.dev from 1.16)\n',
3376
'bzrlib.repofmt.groupcompress_repo',
3377
'RepositoryFormatCHK2',
3378
)
3379
format_registry.register_lazy(
3380
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3381
'bzrlib.repofmt.groupcompress_repo',
3382
'RepositoryFormat2a',
3383
)
3384
1642
format_registry.register_lazy(
3385
1643
'Bazaar development format 8\n',
3386
1644
'bzrlib.repofmt.groupcompress_repo',
3400
1658
InterRepository.get(other).method_name(parameters).
3401
1659
"""
3402
1660
3403
_walk_to_common_revisions_batch_size = 50
3404
1661
_optimisers = []
3405
1662
"""The available optimised InterRepository types."""
3406
1663
3421
1678
self.target.fetch(self.source, revision_id=revision_id)
3422
1679
3423
1680
@needs_write_lock
3424
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3425
fetch_spec=None):
1681
def fetch(self, revision_id=None, find_ghosts=False):
3426
1682
"""Fetch the content required to construct revision_id.
3427
1683
3428
1684
The content is copied from self.source to self.target.
3429
1685
3430
1686
:param revision_id: if None all content is copied, if NULL_REVISION no
3431
1687
content is copied.
3432
:param pb: ignored.
3433
1688
:return: None.
3434
1689
"""
3435
ui.ui_factory.warn_experimental_format_fetch(self)
3436
from bzrlib.fetch import RepoFetcher
3437
# See <https://launchpad.net/bugs/456077> asking for a warning here
3438
if self.source._format.network_name() != self.target._format.network_name():
3439
ui.ui_factory.show_user_warning('cross_format_fetch',
3440
from_format=self.source._format,
3441
to_format=self.target._format)
3442
f = RepoFetcher(to_repository=self.target,
3443
from_repository=self.source,
3444
last_revision=revision_id,
3445
fetch_spec=fetch_spec,
3446
find_ghosts=find_ghosts)
3447
3448
def _walk_to_common_revisions(self, revision_ids):
3449
"""Walk out from revision_ids in source to revisions target has.
3450
3451
:param revision_ids: The start point for the search.
3452
:return: A set of revision ids.
3453
"""
3454
target_graph = self.target.get_graph()
3455
revision_ids = frozenset(revision_ids)
3456
missing_revs = set()
3457
source_graph = self.source.get_graph()
3458
# ensure we don't pay silly lookup costs.
3459
searcher = source_graph._make_breadth_first_searcher(revision_ids)
3460
null_set = frozenset([_mod_revision.NULL_REVISION])
3461
searcher_exhausted = False
3462
while True:
3463
next_revs = set()
3464
ghosts = set()
3465
# Iterate the searcher until we have enough next_revs
3466
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3467
try:
3468
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3469
next_revs.update(next_revs_part)
3470
ghosts.update(ghosts_part)
3471
except StopIteration:
3472
searcher_exhausted = True
3473
break
3474
# If there are ghosts in the source graph, and the caller asked for
3475
# them, make sure that they are present in the target.
3476
# We don't care about other ghosts as we can't fetch them and
3477
# haven't been asked to.
3478
ghosts_to_check = set(revision_ids.intersection(ghosts))
3479
revs_to_get = set(next_revs).union(ghosts_to_check)
3480
if revs_to_get:
3481
have_revs = set(target_graph.get_parent_map(revs_to_get))
3482
# we always have NULL_REVISION present.
3483
have_revs = have_revs.union(null_set)
3484
# Check if the target is missing any ghosts we need.
3485
ghosts_to_check.difference_update(have_revs)
3486
if ghosts_to_check:
3487
# One of the caller's revision_ids is a ghost in both the
3488
# source and the target.
3489
raise errors.NoSuchRevision(
3490
self.source, ghosts_to_check.pop())
3491
missing_revs.update(next_revs - have_revs)
3492
# Because we may have walked past the original stop point, make
3493
# sure everything is stopped
3494
stop_revs = searcher.find_seen_ancestors(have_revs)
3495
searcher.stop_searching_any(stop_revs)
3496
if searcher_exhausted:
3497
break
3498
return searcher.get_result()
1690
raise NotImplementedError(self.fetch)
3499
1691
3500
1692
@needs_read_lock
3501
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
1693
def search_missing_revision_ids(self,
1694
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
1695
find_ghosts=True, revision_ids=None, if_present_ids=None,
1696
limit=None):
3502
1697
"""Return the revision ids that source has that target does not.
3503
1698
3504
1699
:param revision_id: only return revision ids included by this
3505
revision_id.
1700
revision_id.
1701
:param revision_ids: return revision ids included by these
1702
revision_ids. NoSuchRevision will be raised if any of these
1703
revisions are not present.
1704
:param if_present_ids: like revision_ids, but will not cause
1705
NoSuchRevision if any of these are absent, instead they will simply
1706
not be in the result. This is useful for e.g. finding revisions
1707
to fetch for tags, which may reference absent revisions.
3506
1708
:param find_ghosts: If True find missing revisions in deep history
3507
1709
rather than just finding the surface difference.
1710
:param limit: Maximum number of revisions to return, topologically
1711
ordered
3508
1712
:return: A bzrlib.graph.SearchResult.
3509
1713
"""
3510
# stop searching at found target revisions.
3511
if not find_ghosts and revision_id is not None:
3512
return self._walk_to_common_revisions([revision_id])
3513
# generic, possibly worst case, slow code path.
3514
target_ids = set(self.target.all_revision_ids())
3515
if revision_id is not None:
3516
source_ids = self.source.get_ancestry(revision_id)
3517
if source_ids[0] is not None:
3518
raise AssertionError()
3519
source_ids.pop(0)
3520
else:
3521
source_ids = self.source.all_revision_ids()
3522
result_set = set(source_ids).difference(target_ids)
3523
return self.source.revision_ids_to_search_result(result_set)
1714
raise NotImplementedError(self.search_missing_revision_ids)
3524
1715
3525
1716
@staticmethod
3526
1717
def _same_model(source, target):
3547
1738
"different serializers")
3548
1739
3549
1740
3550
class InterSameDataRepository(InterRepository):
3551
"""Code for converting between repositories that represent the same data.
3552
3553
Data format and model must match for this to work.
3554
"""
3555
3556
@classmethod
3557
def _get_repo_format_to_test(self):
3558
"""Repository format for testing with.
3559
3560
InterSameData can pull from subtree to subtree and from non-subtree to
3561
non-subtree, so we test this with the richest repository format.
3562
"""
3563
from bzrlib.repofmt import knitrepo
3564
return knitrepo.RepositoryFormatKnit3()
3565
3566
@staticmethod
3567
def is_compatible(source, target):
3568
return InterRepository._same_model(source, target)
3569
3570
3571
class InterWeaveRepo(InterSameDataRepository):
3572
"""Optimised code paths between Weave based repositories.
3573
3574
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
3575
implemented lazy inter-object optimisation.
3576
"""
3577
3578
@classmethod
3579
def _get_repo_format_to_test(self):
3580
from bzrlib.repofmt import weaverepo
3581
return weaverepo.RepositoryFormat7()
3582
3583
@staticmethod
3584
def is_compatible(source, target):
3585
"""Be compatible with known Weave formats.
3586
3587
We don't test for the stores being of specific types because that
3588
could lead to confusing results, and there is no need to be
3589
overly general.
3590
"""
3591
from bzrlib.repofmt.weaverepo import (
3592
RepositoryFormat5,
3593
RepositoryFormat6,
3594
RepositoryFormat7,
3595
)
3596
try:
3597
return (isinstance(source._format, (RepositoryFormat5,
3598
RepositoryFormat6,
3599
RepositoryFormat7)) and
3600
isinstance(target._format, (RepositoryFormat5,
3601
RepositoryFormat6,
3602
RepositoryFormat7)))
3603
except AttributeError:
3604
return False
3605
3606
@needs_write_lock
3607
def copy_content(self, revision_id=None):
3608
"""See InterRepository.copy_content()."""
3609
# weave specific optimised path:
3610
try:
3611
self.target.set_make_working_trees(self.source.make_working_trees())
3612
except (errors.RepositoryUpgradeRequired, NotImplemented):
3613
pass
3614
# FIXME do not peek!
3615
if self.source._transport.listable():
3616
pb = ui.ui_factory.nested_progress_bar()
3617
try:
3618
self.target.texts.insert_record_stream(
3619
self.source.texts.get_record_stream(
3620
self.source.texts.keys(), 'topological', False))
3621
pb.update('Copying inventory', 0, 1)
3622
self.target.inventories.insert_record_stream(
3623
self.source.inventories.get_record_stream(
3624
self.source.inventories.keys(), 'topological', False))
3625
self.target.signatures.insert_record_stream(
3626
self.source.signatures.get_record_stream(
3627
self.source.signatures.keys(),
3628
'unordered', True))
3629
self.target.revisions.insert_record_stream(
3630
self.source.revisions.get_record_stream(
3631
self.source.revisions.keys(),
3632
'topological', True))
3633
finally:
3634
pb.finished()
3635
else:
3636
self.target.fetch(self.source, revision_id=revision_id)
3637
3638
@needs_read_lock
3639
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3640
"""See InterRepository.missing_revision_ids()."""
3641
# we want all revisions to satisfy revision_id in source.
3642
# but we don't want to stat every file here and there.
3643
# we want then, all revisions other needs to satisfy revision_id
3644
# checked, but not those that we have locally.
3645
# so the first thing is to get a subset of the revisions to
3646
# satisfy revision_id in source, and then eliminate those that
3647
# we do already have.
3648
# this is slow on high latency connection to self, but as this
3649
# disk format scales terribly for push anyway due to rewriting
3650
# inventory.weave, this is considered acceptable.
3651
# - RBC 20060209
3652
if revision_id is not None:
3653
source_ids = self.source.get_ancestry(revision_id)
3654
if source_ids[0] is not None:
3655
raise AssertionError()
3656
source_ids.pop(0)
3657
else:
3658
source_ids = self.source._all_possible_ids()
3659
source_ids_set = set(source_ids)
3660
# source_ids is the worst possible case we may need to pull.
3661
# now we want to filter source_ids against what we actually
3662
# have in target, but don't try to check for existence where we know
3663
# we do not have a revision as that would be pointless.
3664
target_ids = set(self.target._all_possible_ids())
3665
possibly_present_revisions = target_ids.intersection(source_ids_set)
3666
actually_present_revisions = set(
3667
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3668
required_revisions = source_ids_set.difference(actually_present_revisions)
3669
if revision_id is not None:
3670
# we used get_ancestry to determine source_ids then we are assured all
3671
# revisions referenced are present as they are installed in topological order.
3672
# and the tip revision was validated by get_ancestry.
3673
result_set = required_revisions
3674
else:
3675
# if we just grabbed the possibly available ids, then
3676
# we only have an estimate of whats available and need to validate
3677
# that against the revision records.
3678
result_set = set(
3679
self.source._eliminate_revisions_not_present(required_revisions))
3680
return self.source.revision_ids_to_search_result(result_set)
3681
3682
3683
class InterKnitRepo(InterSameDataRepository):
3684
"""Optimised code paths between Knit based repositories."""
3685
3686
@classmethod
3687
def _get_repo_format_to_test(self):
3688
from bzrlib.repofmt import knitrepo
3689
return knitrepo.RepositoryFormatKnit1()
3690
3691
@staticmethod
3692
def is_compatible(source, target):
3693
"""Be compatible with known Knit formats.
3694
3695
We don't test for the stores being of specific types because that
3696
could lead to confusing results, and there is no need to be
3697
overly general.
3698
"""
3699
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
3700
try:
3701
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
3702
isinstance(target._format, RepositoryFormatKnit))
3703
except AttributeError:
3704
return False
3705
return are_knits and InterRepository._same_model(source, target)
3706
3707
@needs_read_lock
3708
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3709
"""See InterRepository.missing_revision_ids()."""
3710
if revision_id is not None:
3711
source_ids = self.source.get_ancestry(revision_id)
3712
if source_ids[0] is not None:
3713
raise AssertionError()
3714
source_ids.pop(0)
3715
else:
3716
source_ids = self.source.all_revision_ids()
3717
source_ids_set = set(source_ids)
3718
# source_ids is the worst possible case we may need to pull.
3719
# now we want to filter source_ids against what we actually
3720
# have in target, but don't try to check for existence where we know
3721
# we do not have a revision as that would be pointless.
3722
target_ids = set(self.target.all_revision_ids())
3723
possibly_present_revisions = target_ids.intersection(source_ids_set)
3724
actually_present_revisions = set(
3725
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3726
required_revisions = source_ids_set.difference(actually_present_revisions)
3727
if revision_id is not None:
3728
# we used get_ancestry to determine source_ids then we are assured all
3729
# revisions referenced are present as they are installed in topological order.
3730
# and the tip revision was validated by get_ancestry.
3731
result_set = required_revisions
3732
else:
3733
# if we just grabbed the possibly available ids, then
3734
# we only have an estimate of whats available and need to validate
3735
# that against the revision records.
3736
result_set = set(
3737
self.source._eliminate_revisions_not_present(required_revisions))
3738
return self.source.revision_ids_to_search_result(result_set)
3739
3740
3741
class InterDifferingSerializer(InterRepository):
3742
3743
@classmethod
3744
def _get_repo_format_to_test(self):
3745
return None
3746
3747
@staticmethod
3748
def is_compatible(source, target):
3749
"""Be compatible with Knit2 source and Knit3 target"""
3750
# This is redundant with format.check_conversion_target(), however that
3751
# raises an exception, and we just want to say "False" as in we won't
3752
# support converting between these formats.
3753
if 'IDS_never' in debug.debug_flags:
3754
return False
3755
if source.supports_rich_root() and not target.supports_rich_root():
3756
return False
3757
if (source._format.supports_tree_reference
3758
and not target._format.supports_tree_reference):
3759
return False
3760
if target._fallback_repositories and target._format.supports_chks:
3761
# IDS doesn't know how to copy CHKs for the parent inventories it
3762
# adds to stacked repos.
3763
return False
3764
if 'IDS_always' in debug.debug_flags:
3765
return True
3766
# Only use this code path for local source and target. IDS does far
3767
# too much IO (both bandwidth and roundtrips) over a network.
3768
if not source.bzrdir.transport.base.startswith('file:///'):
3769
return False
3770
if not target.bzrdir.transport.base.startswith('file:///'):
3771
return False
3772
return True
3773
3774
def _get_trees(self, revision_ids, cache):
3775
possible_trees = []
3776
for rev_id in revision_ids:
3777
if rev_id in cache:
3778
possible_trees.append((rev_id, cache[rev_id]))
3779
else:
3780
# Not cached, but inventory might be present anyway.
3781
try:
3782
tree = self.source.revision_tree(rev_id)
3783
except errors.NoSuchRevision:
3784
# Nope, parent is ghost.
3785
pass
3786
else:
3787
cache[rev_id] = tree
3788
possible_trees.append((rev_id, tree))
3789
return possible_trees
3790
3791
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3792
"""Get the best delta and base for this revision.
3793
3794
:return: (basis_id, delta)
3795
"""
3796
deltas = []
3797
# Generate deltas against each tree, to find the shortest.
3798
texts_possibly_new_in_tree = set()
3799
for basis_id, basis_tree in possible_trees:
3800
delta = tree.inventory._make_delta(basis_tree.inventory)
3801
for old_path, new_path, file_id, new_entry in delta:
3802
if new_path is None:
3803
# This file_id isn't present in the new rev, so we don't
3804
# care about it.
3805
continue
3806
if not new_path:
3807
# Rich roots are handled elsewhere...
3808
continue
3809
kind = new_entry.kind
3810
if kind != 'directory' and kind != 'file':
3811
# No text record associated with this inventory entry.
3812
continue
3813
# This is a directory or file that has changed somehow.
3814
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3815
deltas.append((len(delta), basis_id, delta))
3816
deltas.sort()
3817
return deltas[0][1:]
3818
3819
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3820
"""Find all parent revisions that are absent, but for which the
3821
inventory is present, and copy those inventories.
3822
3823
This is necessary to preserve correctness when the source is stacked
3824
without fallbacks configured. (Note that in cases like upgrade the
3825
source may be not have _fallback_repositories even though it is
3826
stacked.)
3827
"""
3828
parent_revs = set()
3829
for parents in parent_map.values():
3830
parent_revs.update(parents)
3831
present_parents = self.source.get_parent_map(parent_revs)
3832
absent_parents = set(parent_revs).difference(present_parents)
3833
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3834
(rev_id,) for rev_id in absent_parents)
3835
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3836
for parent_tree in self.source.revision_trees(parent_inv_ids):
3837
current_revision_id = parent_tree.get_revision_id()
3838
parents_parents_keys = parent_invs_keys_for_stacking[
3839
(current_revision_id,)]
3840
parents_parents = [key[-1] for key in parents_parents_keys]
3841
basis_id = _mod_revision.NULL_REVISION
3842
basis_tree = self.source.revision_tree(basis_id)
3843
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3844
self.target.add_inventory_by_delta(
3845
basis_id, delta, current_revision_id, parents_parents)
3846
cache[current_revision_id] = parent_tree
3847
3848
def _fetch_batch(self, revision_ids, basis_id, cache):
3849
"""Fetch across a few revisions.
3850
3851
:param revision_ids: The revisions to copy
3852
:param basis_id: The revision_id of a tree that must be in cache, used
3853
as a basis for delta when no other base is available
3854
:param cache: A cache of RevisionTrees that we can use.
3855
:return: The revision_id of the last converted tree. The RevisionTree
3856
for it will be in cache
3857
"""
3858
# Walk though all revisions; get inventory deltas, copy referenced
3859
# texts that delta references, insert the delta, revision and
3860
# signature.
3861
root_keys_to_create = set()
3862
text_keys = set()
3863
pending_deltas = []
3864
pending_revisions = []
3865
parent_map = self.source.get_parent_map(revision_ids)
3866
self._fetch_parent_invs_for_stacking(parent_map, cache)
3867
self.source._safe_to_return_from_cache = True
3868
for tree in self.source.revision_trees(revision_ids):
3869
# Find a inventory delta for this revision.
3870
# Find text entries that need to be copied, too.
3871
current_revision_id = tree.get_revision_id()
3872
parent_ids = parent_map.get(current_revision_id, ())
3873
parent_trees = self._get_trees(parent_ids, cache)
3874
possible_trees = list(parent_trees)
3875
if len(possible_trees) == 0:
3876
# There either aren't any parents, or the parents are ghosts,
3877
# so just use the last converted tree.
3878
possible_trees.append((basis_id, cache[basis_id]))
3879
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3880
possible_trees)
3881
revision = self.source.get_revision(current_revision_id)
3882
pending_deltas.append((basis_id, delta,
3883
current_revision_id, revision.parent_ids))
3884
if self._converting_to_rich_root:
3885
self._revision_id_to_root_id[current_revision_id] = \
3886
tree.get_root_id()
3887
# Determine which texts are in present in this revision but not in
3888
# any of the available parents.
3889
texts_possibly_new_in_tree = set()
3890
for old_path, new_path, file_id, entry in delta:
3891
if new_path is None:
3892
# This file_id isn't present in the new rev
3893
continue
3894
if not new_path:
3895
# This is the root
3896
if not self.target.supports_rich_root():
3897
# The target doesn't support rich root, so we don't
3898
# copy
3899
continue
3900
if self._converting_to_rich_root:
3901
# This can't be copied normally, we have to insert
3902
# it specially
3903
root_keys_to_create.add((file_id, entry.revision))
3904
continue
3905
kind = entry.kind
3906
texts_possibly_new_in_tree.add((file_id, entry.revision))
3907
for basis_id, basis_tree in possible_trees:
3908
basis_inv = basis_tree.inventory
3909
for file_key in list(texts_possibly_new_in_tree):
3910
file_id, file_revision = file_key
3911
try:
3912
entry = basis_inv[file_id]
3913
except errors.NoSuchId:
3914
continue
3915
if entry.revision == file_revision:
3916
texts_possibly_new_in_tree.remove(file_key)
3917
text_keys.update(texts_possibly_new_in_tree)
3918
pending_revisions.append(revision)
3919
cache[current_revision_id] = tree
3920
basis_id = current_revision_id
3921
self.source._safe_to_return_from_cache = False
3922
# Copy file texts
3923
from_texts = self.source.texts
3924
to_texts = self.target.texts
3925
if root_keys_to_create:
3926
root_stream = _mod_fetch._new_root_data_stream(
3927
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3928
self.source)
3929
to_texts.insert_record_stream(root_stream)
3930
to_texts.insert_record_stream(from_texts.get_record_stream(
3931
text_keys, self.target._format._fetch_order,
3932
not self.target._format._fetch_uses_deltas))
3933
# insert inventory deltas
3934
for delta in pending_deltas:
3935
self.target.add_inventory_by_delta(*delta)
3936
if self.target._fallback_repositories:
3937
# Make sure this stacked repository has all the parent inventories
3938
# for the new revisions that we are about to insert. We do this
3939
# before adding the revisions so that no revision is added until
3940
# all the inventories it may depend on are added.
3941
# Note that this is overzealous, as we may have fetched these in an
3942
# earlier batch.
3943
parent_ids = set()
3944
revision_ids = set()
3945
for revision in pending_revisions:
3946
revision_ids.add(revision.revision_id)
3947
parent_ids.update(revision.parent_ids)
3948
parent_ids.difference_update(revision_ids)
3949
parent_ids.discard(_mod_revision.NULL_REVISION)
3950
parent_map = self.source.get_parent_map(parent_ids)
3951
# we iterate over parent_map and not parent_ids because we don't
3952
# want to try copying any revision which is a ghost
3953
for parent_tree in self.source.revision_trees(parent_map):
3954
current_revision_id = parent_tree.get_revision_id()
3955
parents_parents = parent_map[current_revision_id]
3956
possible_trees = self._get_trees(parents_parents, cache)
3957
if len(possible_trees) == 0:
3958
# There either aren't any parents, or the parents are
3959
# ghosts, so just use the last converted tree.
3960
possible_trees.append((basis_id, cache[basis_id]))
3961
basis_id, delta = self._get_delta_for_revision(parent_tree,
3962
parents_parents, possible_trees)
3963
self.target.add_inventory_by_delta(
3964
basis_id, delta, current_revision_id, parents_parents)
3965
# insert signatures and revisions
3966
for revision in pending_revisions:
3967
try:
3968
signature = self.source.get_signature_text(
3969
revision.revision_id)
3970
self.target.add_signature_text(revision.revision_id,
3971
signature)
3972
except errors.NoSuchRevision:
3973
pass
3974
self.target.add_revision(revision.revision_id, revision)
3975
return basis_id
3976
3977
def _fetch_all_revisions(self, revision_ids, pb):
3978
"""Fetch everything for the list of revisions.
3979
3980
:param revision_ids: The list of revisions to fetch. Must be in
3981
topological order.
3982
:param pb: A ProgressTask
3983
:return: None
3984
"""
3985
basis_id, basis_tree = self._get_basis(revision_ids[0])
3986
batch_size = 100
3987
cache = lru_cache.LRUCache(100)
3988
cache[basis_id] = basis_tree
3989
del basis_tree # We don't want to hang on to it here
3990
hints = []
3991
a_graph = None
3992
3993
for offset in range(0, len(revision_ids), batch_size):
3994
self.target.start_write_group()
3995
try:
3996
pb.update('Transferring revisions', offset,
3997
len(revision_ids))
3998
batch = revision_ids[offset:offset+batch_size]
3999
basis_id = self._fetch_batch(batch, basis_id, cache)
4000
except:
4001
self.source._safe_to_return_from_cache = False
4002
self.target.abort_write_group()
4003
raise
4004
else:
4005
hint = self.target.commit_write_group()
4006
if hint:
4007
hints.extend(hint)
4008
if hints and self.target._format.pack_compresses:
4009
self.target.pack(hint=hints)
4010
pb.update('Transferring revisions', len(revision_ids),
4011
len(revision_ids))
4012
4013
@needs_write_lock
4014
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
4015
fetch_spec=None):
4016
"""See InterRepository.fetch()."""
4017
if fetch_spec is not None:
4018
raise AssertionError("Not implemented yet...")
4019
ui.ui_factory.warn_experimental_format_fetch(self)
4020
if (not self.source.supports_rich_root()
4021
and self.target.supports_rich_root()):
4022
self._converting_to_rich_root = True
4023
self._revision_id_to_root_id = {}
4024
else:
4025
self._converting_to_rich_root = False
4026
# See <https://launchpad.net/bugs/456077> asking for a warning here
4027
if self.source._format.network_name() != self.target._format.network_name():
4028
ui.ui_factory.show_user_warning('cross_format_fetch',
4029
from_format=self.source._format,
4030
to_format=self.target._format)
4031
revision_ids = self.target.search_missing_revision_ids(self.source,
4032
revision_id, find_ghosts=find_ghosts).get_keys()
4033
if not revision_ids:
4034
return 0, 0
4035
revision_ids = tsort.topo_sort(
4036
self.source.get_graph().get_parent_map(revision_ids))
4037
if not revision_ids:
4038
return 0, 0
4039
# Walk though all revisions; get inventory deltas, copy referenced
4040
# texts that delta references, insert the delta, revision and
4041
# signature.
4042
if pb is None:
4043
my_pb = ui.ui_factory.nested_progress_bar()
4044
pb = my_pb
4045
else:
4046
symbol_versioning.warn(
4047
symbol_versioning.deprecated_in((1, 14, 0))
4048
% "pb parameter to fetch()")
4049
my_pb = None
4050
try:
4051
self._fetch_all_revisions(revision_ids, pb)
4052
finally:
4053
if my_pb is not None:
4054
my_pb.finished()
4055
return len(revision_ids), 0
4056
4057
def _get_basis(self, first_revision_id):
4058
"""Get a revision and tree which exists in the target.
4059
4060
This assumes that first_revision_id is selected for transmission
4061
because all other ancestors are already present. If we can't find an
4062
ancestor we fall back to NULL_REVISION since we know that is safe.
4063
4064
:return: (basis_id, basis_tree)
4065
"""
4066
first_rev = self.source.get_revision(first_revision_id)
4067
try:
4068
basis_id = first_rev.parent_ids[0]
4069
# only valid as a basis if the target has it
4070
self.target.get_revision(basis_id)
4071
# Try to get a basis tree - if it's a ghost it will hit the
4072
# NoSuchRevision case.
4073
basis_tree = self.source.revision_tree(basis_id)
4074
except (IndexError, errors.NoSuchRevision):
4075
basis_id = _mod_revision.NULL_REVISION
4076
basis_tree = self.source.revision_tree(basis_id)
4077
return basis_id, basis_tree
4078
4079
4080
InterRepository.register_optimiser(InterDifferingSerializer)
4081
InterRepository.register_optimiser(InterSameDataRepository)
4082
InterRepository.register_optimiser(InterWeaveRepo)
4083
InterRepository.register_optimiser(InterKnitRepo)
4084
4085
4086
1741
class CopyConverter(object):
4087
1742
"""A repository conversion tool which just performs a copy of the content.
4088
1743
4109
1764
# trigger an assertion if not such
4110
1765
repo._format.get_format_string()
4111
1766
self.repo_dir = repo.bzrdir
4112
pb.update('Moving repository to repository.backup')
1767
pb.update(gettext('Moving repository to repository.backup'))
4113
1768
self.repo_dir.transport.move('repository', 'repository.backup')
4114
1769
backup_transport = self.repo_dir.transport.clone('repository.backup')
4115
1770
repo._format.check_conversion_target(self.target_format)
4116
1771
self.source_repo = repo._format.open(self.repo_dir,
4117
1772
_found=True,
4118
1773
_override_transport=backup_transport)
4119
pb.update('Creating new repository')
1774
pb.update(gettext('Creating new repository'))
4120
1775
converted = self.target_format.initialize(self.repo_dir,
4121
1776
self.source_repo.is_shared())
4122
1777
converted.lock_write()
4123
1778
try:
4124
pb.update('Copying content')
1779
pb.update(gettext('Copying content'))
4125
1780
self.source_repo.copy_content_into(converted)
4126
1781
finally:
4127
1782
converted.unlock()
4128
pb.update('Deleting old repository content')
1783
pb.update(gettext('Deleting old repository content'))
4129
1784
self.repo_dir.transport.delete_tree('repository.backup')
4130
ui.ui_factory.note('repository converted')
1785
ui.ui_factory.note(gettext('repository converted'))
4131
1786
pb.finished()
4132
1787
4133
1788
4134
_unescape_map = {
4135
'apos':"'",
4136
'quot':'"',
4137
'amp':'&',
4138
'lt':'<',
4139
'gt':'>'
4140
}
4141
4142
4143
def _unescaper(match, _map=_unescape_map):
4144
code = match.group(1)
4145
try:
4146
return _map[code]
4147
except KeyError:
4148
if not code.startswith('#'):
4149
raise
4150
return unichr(int(code[1:])).encode('utf8')
4151
4152
4153
_unescape_re = None
4154
4155
4156
def _unescape_xml(data):
4157
"""Unescape predefined XML entities in a string of data."""
4158
global _unescape_re
4159
if _unescape_re is None:
4160
_unescape_re = re.compile('\&([^;]*);')
4161
return _unescape_re.sub(_unescaper, data)
4162
4163
4164
class _VersionedFileChecker(object):
4165
4166
def __init__(self, repository, text_key_references=None, ancestors=None):
4167
self.repository = repository
4168
self.text_index = self.repository._generate_text_key_index(
4169
text_key_references=text_key_references, ancestors=ancestors)
4170
4171
def calculate_file_version_parents(self, text_key):
4172
"""Calculate the correct parents for a file version according to
4173
the inventories.
4174
"""
4175
parent_keys = self.text_index[text_key]
4176
if parent_keys == [_mod_revision.NULL_REVISION]:
4177
return ()
4178
return tuple(parent_keys)
4179
4180
def check_file_version_parents(self, texts, progress_bar=None):
4181
"""Check the parents stored in a versioned file are correct.
4182
4183
It also detects file versions that are not referenced by their
4184
corresponding revision's inventory.
4185
4186
:returns: A tuple of (wrong_parents, dangling_file_versions).
4187
wrong_parents is a dict mapping {revision_id: (stored_parents,
4188
correct_parents)} for each revision_id where the stored parents
4189
are not correct. dangling_file_versions is a set of (file_id,
4190
revision_id) tuples for versions that are present in this versioned
4191
file, but not used by the corresponding inventory.
4192
"""
4193
local_progress = None
4194
if progress_bar is None:
4195
local_progress = ui.ui_factory.nested_progress_bar()
4196
progress_bar = local_progress
4197
try:
4198
return self._check_file_version_parents(texts, progress_bar)
4199
finally:
4200
if local_progress:
4201
local_progress.finished()
4202
4203
def _check_file_version_parents(self, texts, progress_bar):
4204
"""See check_file_version_parents."""
4205
wrong_parents = {}
4206
self.file_ids = set([file_id for file_id, _ in
4207
self.text_index.iterkeys()])
4208
# text keys is now grouped by file_id
4209
n_versions = len(self.text_index)
4210
progress_bar.update('loading text store', 0, n_versions)
4211
parent_map = self.repository.texts.get_parent_map(self.text_index)
4212
# On unlistable transports this could well be empty/error...
4213
text_keys = self.repository.texts.keys()
4214
unused_keys = frozenset(text_keys) - set(self.text_index)
4215
for num, key in enumerate(self.text_index.iterkeys()):
4216
progress_bar.update('checking text graph', num, n_versions)
4217
correct_parents = self.calculate_file_version_parents(key)
4218
try:
4219
knit_parents = parent_map[key]
4220
except errors.RevisionNotPresent:
4221
# Missing text!
4222
knit_parents = None
4223
if correct_parents != knit_parents:
4224
wrong_parents[key] = (knit_parents, correct_parents)
4225
return wrong_parents, unused_keys
4226
4227
4228
def _old_get_graph(repository, revision_id):
4229
"""DO NOT USE. That is all. I'm serious."""
4230
graph = repository.get_graph()
4231
revision_graph = dict(((key, value) for key, value in
4232
graph.iter_ancestry([revision_id]) if value is not None))
4233
return _strip_NULL_ghosts(revision_graph)
4234
4235
4236
1789
def _strip_NULL_ghosts(revision_graph):
4237
1790
"""Also don't use this. more compatibility code for unmigrated clients."""
4238
1791
# Filter ghosts, and null:
4244
1797
return revision_graph
4245
1798
4246
1799
4247
class StreamSink(object):
4248
"""An object that can insert a stream into a repository.
4249
4250
This interface handles the complexity of reserialising inventories and
4251
revisions from different formats, and allows unidirectional insertion into
4252
stacked repositories without looking for the missing basis parents
4253
beforehand.
4254
"""
4255
4256
def __init__(self, target_repo):
4257
self.target_repo = target_repo
4258
4259
def insert_stream(self, stream, src_format, resume_tokens):
4260
"""Insert a stream's content into the target repository.
4261
4262
:param src_format: a bzr repository format.
4263
4264
:return: a list of resume tokens and an iterable of keys additional
4265
items required before the insertion can be completed.
4266
"""
4267
self.target_repo.lock_write()
4268
try:
4269
if resume_tokens:
4270
self.target_repo.resume_write_group(resume_tokens)
4271
is_resume = True
4272
else:
4273
self.target_repo.start_write_group()
4274
is_resume = False
4275
try:
4276
# locked_insert_stream performs a commit|suspend.
4277
return self._locked_insert_stream(stream, src_format,
4278
is_resume)
4279
except:
4280
self.target_repo.abort_write_group(suppress_errors=True)
4281
raise
4282
finally:
4283
self.target_repo.unlock()
4284
4285
def _locked_insert_stream(self, stream, src_format, is_resume):
4286
to_serializer = self.target_repo._format._serializer
4287
src_serializer = src_format._serializer
4288
new_pack = None
4289
if to_serializer == src_serializer:
4290
# If serializers match and the target is a pack repository, set the
4291
# write cache size on the new pack. This avoids poor performance
4292
# on transports where append is unbuffered (such as
4293
# RemoteTransport). This is safe to do because nothing should read
4294
# back from the target repository while a stream with matching
4295
# serialization is being inserted.
4296
# The exception is that a delta record from the source that should
4297
# be a fulltext may need to be expanded by the target (see
4298
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4299
# explicitly flush any buffered writes first in that rare case.
4300
try:
4301
new_pack = self.target_repo._pack_collection._new_pack
4302
except AttributeError:
4303
# Not a pack repository
4304
pass
4305
else:
4306
new_pack.set_write_cache_size(1024*1024)
4307
for substream_type, substream in stream:
4308
if 'stream' in debug.debug_flags:
4309
mutter('inserting substream: %s', substream_type)
4310
if substream_type == 'texts':
4311
self.target_repo.texts.insert_record_stream(substream)
4312
elif substream_type == 'inventories':
4313
if src_serializer == to_serializer:
4314
self.target_repo.inventories.insert_record_stream(
4315
substream)
4316
else:
4317
self._extract_and_insert_inventories(
4318
substream, src_serializer)
4319
elif substream_type == 'inventory-deltas':
4320
self._extract_and_insert_inventory_deltas(
4321
substream, src_serializer)
4322
elif substream_type == 'chk_bytes':
4323
# XXX: This doesn't support conversions, as it assumes the
4324
# conversion was done in the fetch code.
4325
self.target_repo.chk_bytes.insert_record_stream(substream)
4326
elif substream_type == 'revisions':
4327
# This may fallback to extract-and-insert more often than
4328
# required if the serializers are different only in terms of
4329
# the inventory.
4330
if src_serializer == to_serializer:
4331
self.target_repo.revisions.insert_record_stream(substream)
4332
else:
4333
self._extract_and_insert_revisions(substream,
4334
src_serializer)
4335
elif substream_type == 'signatures':
4336
self.target_repo.signatures.insert_record_stream(substream)
4337
else:
4338
raise AssertionError('kaboom! %s' % (substream_type,))
4339
# Done inserting data, and the missing_keys calculations will try to
4340
# read back from the inserted data, so flush the writes to the new pack
4341
# (if this is pack format).
4342
if new_pack is not None:
4343
new_pack._write_data('', flush=True)
4344
# Find all the new revisions (including ones from resume_tokens)
4345
missing_keys = self.target_repo.get_missing_parent_inventories(
4346
check_for_missing_texts=is_resume)
4347
try:
4348
for prefix, versioned_file in (
4349
('texts', self.target_repo.texts),
4350
('inventories', self.target_repo.inventories),
4351
('revisions', self.target_repo.revisions),
4352
('signatures', self.target_repo.signatures),
4353
('chk_bytes', self.target_repo.chk_bytes),
4354
):
4355
if versioned_file is None:
4356
continue
4357
# TODO: key is often going to be a StaticTuple object
4358
# I don't believe we can define a method by which
4359
# (prefix,) + StaticTuple will work, though we could
4360
# define a StaticTuple.sq_concat that would allow you to
4361
# pass in either a tuple or a StaticTuple as the second
4362
# object, so instead we could have:
4363
# StaticTuple(prefix) + key here...
4364
missing_keys.update((prefix,) + key for key in
4365
versioned_file.get_missing_compression_parent_keys())
4366
except NotImplementedError:
4367
# cannot even attempt suspending, and missing would have failed
4368
# during stream insertion.
4369
missing_keys = set()
4370
else:
4371
if missing_keys:
4372
# suspend the write group and tell the caller what we is
4373
# missing. We know we can suspend or else we would not have
4374
# entered this code path. (All repositories that can handle
4375
# missing keys can handle suspending a write group).
4376
write_group_tokens = self.target_repo.suspend_write_group()
4377
return write_group_tokens, missing_keys
4378
hint = self.target_repo.commit_write_group()
4379
if (to_serializer != src_serializer and
4380
self.target_repo._format.pack_compresses):
4381
self.target_repo.pack(hint=hint)
4382
return [], set()
4383
4384
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4385
target_rich_root = self.target_repo._format.rich_root_data
4386
target_tree_refs = self.target_repo._format.supports_tree_reference
4387
for record in substream:
4388
# Insert the delta directly
4389
inventory_delta_bytes = record.get_bytes_as('fulltext')
4390
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4391
try:
4392
parse_result = deserialiser.parse_text_bytes(
4393
inventory_delta_bytes)
4394
except inventory_delta.IncompatibleInventoryDelta, err:
4395
trace.mutter("Incompatible delta: %s", err.msg)
4396
raise errors.IncompatibleRevision(self.target_repo._format)
4397
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4398
revision_id = new_id
4399
parents = [key[0] for key in record.parents]
4400
self.target_repo.add_inventory_by_delta(
4401
basis_id, inv_delta, revision_id, parents)
4402
4403
def _extract_and_insert_inventories(self, substream, serializer,
4404
parse_delta=None):
4405
"""Generate a new inventory versionedfile in target, converting data.
4406
4407
The inventory is retrieved from the source, (deserializing it), and
4408
stored in the target (reserializing it in a different format).
4409
"""
4410
target_rich_root = self.target_repo._format.rich_root_data
4411
target_tree_refs = self.target_repo._format.supports_tree_reference
4412
for record in substream:
4413
# It's not a delta, so it must be a fulltext in the source
4414
# serializer's format.
4415
bytes = record.get_bytes_as('fulltext')
4416
revision_id = record.key[0]
4417
inv = serializer.read_inventory_from_string(bytes, revision_id)
4418
parents = [key[0] for key in record.parents]
4419
self.target_repo.add_inventory(revision_id, inv, parents)
4420
# No need to keep holding this full inv in memory when the rest of
4421
# the substream is likely to be all deltas.
4422
del inv
4423
4424
def _extract_and_insert_revisions(self, substream, serializer):
4425
for record in substream:
4426
bytes = record.get_bytes_as('fulltext')
4427
revision_id = record.key[0]
4428
rev = serializer.read_revision_from_string(bytes)
4429
if rev.revision_id != revision_id:
4430
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4431
self.target_repo.add_revision(revision_id, rev)
4432
4433
def finished(self):
4434
if self.target_repo._format._fetch_reconcile:
4435
self.target_repo.reconcile()
4436
4437
4438
class StreamSource(object):
4439
"""A source of a stream for fetching between repositories."""
4440
4441
def __init__(self, from_repository, to_format):
4442
"""Create a StreamSource streaming from from_repository."""
4443
self.from_repository = from_repository
4444
self.to_format = to_format
4445
self._record_counter = RecordCounter()
4446
4447
def delta_on_metadata(self):
4448
"""Return True if delta's are permitted on metadata streams.
4449
4450
That is on revisions and signatures.
4451
"""
4452
src_serializer = self.from_repository._format._serializer
4453
target_serializer = self.to_format._serializer
4454
return (self.to_format._fetch_uses_deltas and
4455
src_serializer == target_serializer)
4456
4457
def _fetch_revision_texts(self, revs):
4458
# fetch signatures first and then the revision texts
4459
# may need to be a InterRevisionStore call here.
4460
from_sf = self.from_repository.signatures
4461
# A missing signature is just skipped.
4462
keys = [(rev_id,) for rev_id in revs]
4463
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4464
keys,
4465
self.to_format._fetch_order,
4466
not self.to_format._fetch_uses_deltas))
4467
# If a revision has a delta, this is actually expanded inside the
4468
# insert_record_stream code now, which is an alternate fix for
4469
# bug #261339
4470
from_rf = self.from_repository.revisions
4471
revisions = from_rf.get_record_stream(
4472
keys,
4473
self.to_format._fetch_order,
4474
not self.delta_on_metadata())
4475
return [('signatures', signatures), ('revisions', revisions)]
4476
4477
def _generate_root_texts(self, revs):
4478
"""This will be called by get_stream between fetching weave texts and
4479
fetching the inventory weave.
4480
"""
4481
if self._rich_root_upgrade():
4482
return _mod_fetch.Inter1and2Helper(
4483
self.from_repository).generate_root_texts(revs)
4484
else:
4485
return []
4486
4487
def get_stream(self, search):
4488
phase = 'file'
4489
revs = search.get_keys()
4490
graph = self.from_repository.get_graph()
4491
revs = tsort.topo_sort(graph.get_parent_map(revs))
4492
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4493
text_keys = []
4494
for knit_kind, file_id, revisions in data_to_fetch:
4495
if knit_kind != phase:
4496
phase = knit_kind
4497
# Make a new progress bar for this phase
4498
if knit_kind == "file":
4499
# Accumulate file texts
4500
text_keys.extend([(file_id, revision) for revision in
4501
revisions])
4502
elif knit_kind == "inventory":
4503
# Now copy the file texts.
4504
from_texts = self.from_repository.texts
4505
yield ('texts', from_texts.get_record_stream(
4506
text_keys, self.to_format._fetch_order,
4507
not self.to_format._fetch_uses_deltas))
4508
# Cause an error if a text occurs after we have done the
4509
# copy.
4510
text_keys = None
4511
# Before we process the inventory we generate the root
4512
# texts (if necessary) so that the inventories references
4513
# will be valid.
4514
for _ in self._generate_root_texts(revs):
4515
yield _
4516
# we fetch only the referenced inventories because we do not
4517
# know for unselected inventories whether all their required
4518
# texts are present in the other repository - it could be
4519
# corrupt.
4520
for info in self._get_inventory_stream(revs):
4521
yield info
4522
elif knit_kind == "signatures":
4523
# Nothing to do here; this will be taken care of when
4524
# _fetch_revision_texts happens.
4525
pass
4526
elif knit_kind == "revisions":
4527
for record in self._fetch_revision_texts(revs):
4528
yield record
4529
else:
4530
raise AssertionError("Unknown knit kind %r" % knit_kind)
4531
4532
def get_stream_for_missing_keys(self, missing_keys):
4533
# missing keys can only occur when we are byte copying and not
4534
# translating (because translation means we don't send
4535
# unreconstructable deltas ever).
4536
keys = {}
4537
keys['texts'] = set()
4538
keys['revisions'] = set()
4539
keys['inventories'] = set()
4540
keys['chk_bytes'] = set()
4541
keys['signatures'] = set()
4542
for key in missing_keys:
4543
keys[key[0]].add(key[1:])
4544
if len(keys['revisions']):
4545
# If we allowed copying revisions at this point, we could end up
4546
# copying a revision without copying its required texts: a
4547
# violation of the requirements for repository integrity.
4548
raise AssertionError(
4549
'cannot copy revisions to fill in missing deltas %s' % (
4550
keys['revisions'],))
4551
for substream_kind, keys in keys.iteritems():
4552
vf = getattr(self.from_repository, substream_kind)
4553
if vf is None and keys:
4554
raise AssertionError(
4555
"cannot fill in keys for a versioned file we don't"
4556
" have: %s needs %s" % (substream_kind, keys))
4557
if not keys:
4558
# No need to stream something we don't have
4559
continue
4560
if substream_kind == 'inventories':
4561
# Some missing keys are genuinely ghosts, filter those out.
4562
present = self.from_repository.inventories.get_parent_map(keys)
4563
revs = [key[0] for key in present]
4564
# Get the inventory stream more-or-less as we do for the
4565
# original stream; there's no reason to assume that records
4566
# direct from the source will be suitable for the sink. (Think
4567
# e.g. 2a -> 1.9-rich-root).
4568
for info in self._get_inventory_stream(revs, missing=True):
4569
yield info
4570
continue
4571
4572
# Ask for full texts always so that we don't need more round trips
4573
# after this stream.
4574
# Some of the missing keys are genuinely ghosts, so filter absent
4575
# records. The Sink is responsible for doing another check to
4576
# ensure that ghosts don't introduce missing data for future
4577
# fetches.
4578
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4579
self.to_format._fetch_order, True))
4580
yield substream_kind, stream
4581
4582
def inventory_fetch_order(self):
4583
if self._rich_root_upgrade():
4584
return 'topological'
4585
else:
4586
return self.to_format._fetch_order
4587
4588
def _rich_root_upgrade(self):
4589
return (not self.from_repository._format.rich_root_data and
4590
self.to_format.rich_root_data)
4591
4592
def _get_inventory_stream(self, revision_ids, missing=False):
4593
from_format = self.from_repository._format
4594
if (from_format.supports_chks and self.to_format.supports_chks and
4595
from_format.network_name() == self.to_format.network_name()):
4596
raise AssertionError(
4597
"this case should be handled by GroupCHKStreamSource")
4598
elif 'forceinvdeltas' in debug.debug_flags:
4599
return self._get_convertable_inventory_stream(revision_ids,
4600
delta_versus_null=missing)
4601
elif from_format.network_name() == self.to_format.network_name():
4602
# Same format.
4603
return self._get_simple_inventory_stream(revision_ids,
4604
missing=missing)
4605
elif (not from_format.supports_chks and not self.to_format.supports_chks
4606
and from_format._serializer == self.to_format._serializer):
4607
# Essentially the same format.
4608
return self._get_simple_inventory_stream(revision_ids,
4609
missing=missing)
4610
else:
4611
# Any time we switch serializations, we want to use an
4612
# inventory-delta based approach.
4613
return self._get_convertable_inventory_stream(revision_ids,
4614
delta_versus_null=missing)
4615
4616
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4617
# NB: This currently reopens the inventory weave in source;
4618
# using a single stream interface instead would avoid this.
4619
from_weave = self.from_repository.inventories
4620
if missing:
4621
delta_closure = True
4622
else:
4623
delta_closure = not self.delta_on_metadata()
4624
yield ('inventories', from_weave.get_record_stream(
4625
[(rev_id,) for rev_id in revision_ids],
4626
self.inventory_fetch_order(), delta_closure))
4627
4628
def _get_convertable_inventory_stream(self, revision_ids,
4629
delta_versus_null=False):
4630
# The two formats are sufficiently different that there is no fast
4631
# path, so we need to send just inventorydeltas, which any
4632
# sufficiently modern client can insert into any repository.
4633
# The StreamSink code expects to be able to
4634
# convert on the target, so we need to put bytes-on-the-wire that can
4635
# be converted. That means inventory deltas (if the remote is <1.19,
4636
# RemoteStreamSink will fallback to VFS to insert the deltas).
4637
yield ('inventory-deltas',
4638
self._stream_invs_as_deltas(revision_ids,
4639
delta_versus_null=delta_versus_null))
4640
4641
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4642
"""Return a stream of inventory-deltas for the given rev ids.
4643
4644
:param revision_ids: The list of inventories to transmit
4645
:param delta_versus_null: Don't try to find a minimal delta for this
4646
entry, instead compute the delta versus the NULL_REVISION. This
4647
effectively streams a complete inventory. Used for stuff like
4648
filling in missing parents, etc.
4649
"""
4650
from_repo = self.from_repository
4651
revision_keys = [(rev_id,) for rev_id in revision_ids]
4652
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4653
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4654
# method...
4655
inventories = self.from_repository.iter_inventories(
4656
revision_ids, 'topological')
4657
format = from_repo._format
4658
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4659
inventory_cache = lru_cache.LRUCache(50)
4660
null_inventory = from_repo.revision_tree(
4661
_mod_revision.NULL_REVISION).inventory
4662
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4663
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4664
# repo back into a non-rich-root repo ought to be allowed)
4665
serializer = inventory_delta.InventoryDeltaSerializer(
4666
versioned_root=format.rich_root_data,
4667
tree_references=format.supports_tree_reference)
4668
for inv in inventories:
4669
key = (inv.revision_id,)
4670
parent_keys = parent_map.get(key, ())
4671
delta = None
4672
if not delta_versus_null and parent_keys:
4673
# The caller did not ask for complete inventories and we have
4674
# some parents that we can delta against. Make a delta against
4675
# each parent so that we can find the smallest.
4676
parent_ids = [parent_key[0] for parent_key in parent_keys]
4677
for parent_id in parent_ids:
4678
if parent_id not in invs_sent_so_far:
4679
# We don't know that the remote side has this basis, so
4680
# we can't use it.
4681
continue
4682
if parent_id == _mod_revision.NULL_REVISION:
4683
parent_inv = null_inventory
4684
else:
4685
parent_inv = inventory_cache.get(parent_id, None)
4686
if parent_inv is None:
4687
parent_inv = from_repo.get_inventory(parent_id)
4688
candidate_delta = inv._make_delta(parent_inv)
4689
if (delta is None or
4690
len(delta) > len(candidate_delta)):
4691
delta = candidate_delta
4692
basis_id = parent_id
4693
if delta is None:
4694
# Either none of the parents ended up being suitable, or we
4695
# were asked to delta against NULL
4696
basis_id = _mod_revision.NULL_REVISION
4697
delta = inv._make_delta(null_inventory)
4698
invs_sent_so_far.add(inv.revision_id)
4699
inventory_cache[inv.revision_id] = inv
4700
delta_serialized = ''.join(
4701
serializer.delta_to_lines(basis_id, key[-1], delta))
4702
yield versionedfile.FulltextContentFactory(
4703
key, parent_keys, None, delta_serialized)
4704
4705
4706
1800
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4707
1801
stop_revision=None):
4708
1802
"""Extend the partial history to include a given index
4718
1812
it is encountered, history extension will stop.
4719
1813
"""
4720
1814
start_revision = partial_history_cache[-1]
4721
iterator = repo.iter_reverse_revision_history(start_revision)
1815
graph = repo.get_graph()
1816
iterator = graph.iter_lefthand_ancestry(start_revision,
1817
(_mod_revision.NULL_REVISION,))
4722
1818
try:
4723
#skip the last revision in the list
1819
# skip the last revision in the list
4724
1820
iterator.next()
4725
1821
while True:
4726
1822
if (stop_index is not None and
4734
1830
# No more history
4735
1831
return
4736
1832
1833
1834
class _LazyListJoin(object):
1835
"""An iterable yielding the contents of many lists as one list.
1836
1837
Each iterator made from this will reflect the current contents of the lists
1838
at the time the iterator is made.
1839
1840
This is used by Repository's _make_parents_provider implementation so that
1841
it is safe to do::
1842
1843
pp = repo._make_parents_provider() # uses a list of fallback repos
1844
pp.add_fallback_repository(other_repo) # appends to that list
1845
result = pp.get_parent_map(...)
1846
# The result will include revs from other_repo
1847
"""
1848
1849
def __init__(self, *list_parts):
1850
self.list_parts = list_parts
1851
1852
def __iter__(self):
1853
full_list = []
1854
for list_part in self.list_parts:
1855
full_list.extend(list_part)
1856
return iter(full_list)
1857
1858
def __repr__(self):
1859
return "%s.%s(%s)" % (self.__module__, self.__class__.__name__,
1860
self.list_parts)
Loggerhead is a web-based interface for Breezy
Version: 2.0.1