« back to all changes in this revision
Viewing changes to bzrlib/repository.py
- browse files at revision 6351.4.2
- compare with another revision
- download diff
- download tarball
- view history from revision 6351.4.2
- Committer: Jelmer Vernooij
- Date: 2011-12-14 00:54:18 UTC
- mfrom: (6362.2.4 path_from_environ_832028)
- mto: (6351.3.14 commit-uses-config-stacks)
- mto: This revision was merged to the branch mainline in revision 6370.
- Revision ID: jelmer@samba.org-20111214005418-q58vle04mp6futbv
Merge lp:~gz/bzr/path_from_environ_832028
- files added:
- NEWS
- 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
- NEWS => doc/en/release-notes/bzr-2.3.txt
- NEWS-template.txt => doc/en/release-notes/release-template.txt
- files modified:
- .bzrignore
added
removed
bzrlib/repository.py
1
# Copyright (C) 2005-2010 Canonical Ltd
1
# Copyright (C) 2005-2011 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
16
16
17
17
from bzrlib.lazy_import import lazy_import
18
18
lazy_import(globals(), """
19
import cStringIO
20
import re
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
32
revision as _mod_revision,
43
static_tuple,
44
symbol_versioning,
45
trace,
33
testament as _mod_testament,
46
34
tsort,
47
versionedfile,
35
gpg,
48
36
)
49
37
from bzrlib.bundle import serializer
50
from bzrlib.revisiontree import RevisionTree
51
from bzrlib.store.versioned import VersionedFileStore
52
from bzrlib.testament import Testament
38
from bzrlib.i18n import gettext
53
39
""")
54
40
55
41
from bzrlib import (
56
42
errors,
57
43
registry,
44
symbol_versioning,
58
45
ui,
59
46
)
60
47
from bzrlib.decorators import needs_read_lock, needs_write_lock, only_raises
61
48
from bzrlib.inter import InterObject
62
from bzrlib.inventory import (
63
Inventory,
64
InventoryDirectory,
65
ROOT_ID,
66
entry_factory,
67
)
68
from bzrlib.recordcounter import RecordCounter
69
49
from bzrlib.lock import _RelockDebugMixin, LogicalLockResult
70
50
from bzrlib.trace import (
71
51
log_exception_quietly, note, mutter, mutter_callsite, warning)
92
72
93
73
# all clients should supply tree roots.
94
74
record_root_entry = True
95
# the default CommitBuilder does not manage trees whose root is versioned.
96
_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
97
80
98
def __init__(self, repository, parents, config, timestamp=None,
81
def __init__(self, repository, parents, config_stack, timestamp=None,
99
82
timezone=None, committer=None, revprops=None,
100
revision_id=None):
83
revision_id=None, lossy=False):
101
84
"""Initiate a CommitBuilder.
102
85
103
86
:param repository: Repository to commit to.
104
87
:param parents: Revision ids of the parents of the new revision.
105
:param config: Configuration to use.
106
88
:param timestamp: Optional timestamp recorded for commit.
107
89
:param timezone: Optional timezone for timestamp.
108
90
:param committer: Optional committer to set for commit.
109
91
:param revprops: Optional dictionary of revision properties.
110
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
111
95
"""
112
self._config = config
96
self._config_stack = config_stack
97
self._lossy = lossy
113
98
114
99
if committer is None:
115
self._committer = self._config.username()
100
self._committer = self._config_stack.get('email')
101
elif not isinstance(committer, unicode):
102
self._committer = committer.decode() # throw if non-ascii
116
103
else:
117
104
self._committer = committer
118
105
119
self.new_inventory = Inventory(None)
120
106
self._new_revision_id = revision_id
121
107
self.parents = parents
122
108
self.repository = repository
137
123
self._timezone = int(timezone)
138
124
139
125
self._generate_revision_if_needed()
140
self.__heads = graph.HeadsCache(repository.get_graph()).heads
141
self._basis_delta = []
142
# API compatibility, older code that used CommitBuilder did not call
143
# .record_delete(), which means the delta that is computed would not be
144
# valid. Callers that will call record_delete() should call
145
# .will_record_deletes() to indicate that.
146
self._recording_deletes = False
147
# memo'd check for no-op commits.
148
self._any_changes = False
149
126
150
127
def any_changes(self):
151
128
"""Return True if any entries were changed.
152
129
153
130
This includes merge-only changes. It is the core for the --unchanged
154
131
detection in commit.
155
132
156
133
:return: True if any changes have occured.
157
134
"""
158
return self._any_changes
135
raise NotImplementedError(self.any_changes)
159
136
160
137
def _validate_unicode_text(self, text, context):
161
138
"""Verify things like commit messages don't have bogus characters."""
177
154
178
155
:return: The revision id of the recorded revision.
179
156
"""
180
self._validate_unicode_text(message, 'commit message')
181
rev = _mod_revision.Revision(
182
timestamp=self._timestamp,
183
timezone=self._timezone,
184
committer=self._committer,
185
message=message,
186
inventory_sha1=self.inv_sha1,
187
revision_id=self._new_revision_id,
188
properties=self._revprops)
189
rev.parent_ids = self.parents
190
self.repository.add_revision(self._new_revision_id, rev,
191
self.new_inventory, self._config)
192
self.repository.commit_write_group()
193
return self._new_revision_id
157
raise NotImplementedError(self.commit)
194
158
195
159
def abort(self):
196
160
"""Abort the commit that is being built.
197
161
"""
198
self.repository.abort_write_group()
162
raise NotImplementedError(self.abort)
199
163
200
164
def revision_tree(self):
201
165
"""Return the tree that was just committed.
202
166
203
After calling commit() this can be called to get a RevisionTree
204
representing the newly committed tree. This is preferred to
205
calling Repository.revision_tree() because that may require
206
deserializing the inventory, while we already have a copy in
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
207
171
memory.
208
172
"""
209
if self.new_inventory is None:
210
self.new_inventory = self.repository.get_inventory(
211
self._new_revision_id)
212
return RevisionTree(self.repository, self.new_inventory,
213
self._new_revision_id)
173
raise NotImplementedError(self.revision_tree)
214
174
215
175
def finish_inventory(self):
216
176
"""Tell the builder that the inventory is finished.
218
178
:return: The inventory id in the repository, which can be used with
219
179
repository.get_inventory.
220
180
"""
221
if self.new_inventory is None:
222
# an inventory delta was accumulated without creating a new
223
# inventory.
224
basis_id = self.basis_delta_revision
225
# We ignore the 'inventory' returned by add_inventory_by_delta
226
# because self.new_inventory is used to hint to the rest of the
227
# system what code path was taken
228
self.inv_sha1, _ = self.repository.add_inventory_by_delta(
229
basis_id, self._basis_delta, self._new_revision_id,
230
self.parents)
231
else:
232
if self.new_inventory.root is None:
233
raise AssertionError('Root entry should be supplied to'
234
' record_entry_contents, as of bzr 0.10.')
235
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
236
self.new_inventory.revision_id = self._new_revision_id
237
self.inv_sha1 = self.repository.add_inventory(
238
self._new_revision_id,
239
self.new_inventory,
240
self.parents
241
)
242
return self._new_revision_id
181
raise NotImplementedError(self.finish_inventory)
243
182
244
183
def _gen_revision_id(self):
245
184
"""Return new revision-id."""
260
199
else:
261
200
self.random_revid = False
262
201
263
def _heads(self, file_id, revision_ids):
264
"""Calculate the graph heads for revision_ids in the graph of file_id.
265
266
This can use either a per-file graph or a global revision graph as we
267
have an identity relationship between the two graphs.
268
"""
269
return self.__heads(revision_ids)
270
271
def _check_root(self, ie, parent_invs, tree):
272
"""Helper for record_entry_contents.
273
274
:param ie: An entry being added.
275
:param parent_invs: The inventories of the parent revisions of the
276
commit.
277
:param tree: The tree that is being committed.
278
"""
279
# In this revision format, root entries have no knit or weave When
280
# serializing out to disk and back in root.revision is always
281
# _new_revision_id
282
ie.revision = self._new_revision_id
283
284
def _require_root_change(self, tree):
285
"""Enforce an appropriate root object change.
286
287
This is called once when record_iter_changes is called, if and only if
288
the root was not in the delta calculated by record_iter_changes.
289
290
:param tree: The tree which is being committed.
291
"""
292
if len(self.parents) == 0:
293
raise errors.RootMissing()
294
entry = entry_factory['directory'](tree.path2id(''), '',
295
None)
296
entry.revision = self._new_revision_id
297
self._basis_delta.append(('', '', entry.file_id, entry))
298
299
def _get_delta(self, ie, basis_inv, path):
300
"""Get a delta against the basis inventory for ie."""
301
if ie.file_id not in basis_inv:
302
# add
303
result = (None, path, ie.file_id, ie)
304
self._basis_delta.append(result)
305
return result
306
elif ie != basis_inv[ie.file_id]:
307
# common but altered
308
# TODO: avoid tis id2path call.
309
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
310
self._basis_delta.append(result)
311
return result
312
else:
313
# common, unaltered
314
return None
315
316
def get_basis_delta(self):
317
"""Return the complete inventory delta versus the basis inventory.
318
319
This has been built up with the calls to record_delete and
320
record_entry_contents. The client must have already called
321
will_record_deletes() to indicate that they will be generating a
322
complete delta.
323
324
:return: An inventory delta, suitable for use with apply_delta, or
325
Repository.add_inventory_by_delta, etc.
326
"""
327
if not self._recording_deletes:
328
raise AssertionError("recording deletes not activated.")
329
return self._basis_delta
330
331
def record_delete(self, path, file_id):
332
"""Record that a delete occured against a basis tree.
333
334
This is an optional API - when used it adds items to the basis_delta
335
being accumulated by the commit builder. It cannot be called unless the
336
method will_record_deletes() has been called to inform the builder that
337
a delta is being supplied.
338
339
:param path: The path of the thing deleted.
340
:param file_id: The file id that was deleted.
341
"""
342
if not self._recording_deletes:
343
raise AssertionError("recording deletes not activated.")
344
delta = (path, None, file_id, None)
345
self._basis_delta.append(delta)
346
self._any_changes = True
347
return delta
348
349
202
def will_record_deletes(self):
350
203
"""Tell the commit builder that deletes are being notified.
351
204
353
206
commit to be valid, deletes against the basis MUST be recorded via
354
207
builder.record_delete().
355
208
"""
356
self._recording_deletes = True
357
try:
358
basis_id = self.parents[0]
359
except IndexError:
360
basis_id = _mod_revision.NULL_REVISION
361
self.basis_delta_revision = basis_id
362
363
def record_entry_contents(self, ie, parent_invs, path, tree,
364
content_summary):
365
"""Record the content of ie from tree into the commit if needed.
366
367
Side effect: sets ie.revision when unchanged
368
369
:param ie: An inventory entry present in the commit.
370
:param parent_invs: The inventories of the parent revisions of the
371
commit.
372
:param path: The path the entry is at in the tree.
373
:param tree: The tree which contains this entry and should be used to
374
obtain content.
375
:param content_summary: Summary data from the tree about the paths
376
content - stat, length, exec, sha/link target. This is only
377
accessed when the entry has a revision of None - that is when it is
378
a candidate to commit.
379
:return: A tuple (change_delta, version_recorded, fs_hash).
380
change_delta is an inventory_delta change for this entry against
381
the basis tree of the commit, or None if no change occured against
382
the basis tree.
383
version_recorded is True if a new version of the entry has been
384
recorded. For instance, committing a merge where a file was only
385
changed on the other side will return (delta, False).
386
fs_hash is either None, or the hash details for the path (currently
387
a tuple of the contents sha1 and the statvalue returned by
388
tree.get_file_with_stat()).
389
"""
390
if self.new_inventory.root is None:
391
if ie.parent_id is not None:
392
raise errors.RootMissing()
393
self._check_root(ie, parent_invs, tree)
394
if ie.revision is None:
395
kind = content_summary[0]
396
else:
397
# ie is carried over from a prior commit
398
kind = ie.kind
399
# XXX: repository specific check for nested tree support goes here - if
400
# the repo doesn't want nested trees we skip it ?
401
if (kind == 'tree-reference' and
402
not self.repository._format.supports_tree_reference):
403
# mismatch between commit builder logic and repository:
404
# this needs the entry creation pushed down into the builder.
405
raise NotImplementedError('Missing repository subtree support.')
406
self.new_inventory.add(ie)
407
408
# TODO: slow, take it out of the inner loop.
409
try:
410
basis_inv = parent_invs[0]
411
except IndexError:
412
basis_inv = Inventory(root_id=None)
413
414
# ie.revision is always None if the InventoryEntry is considered
415
# for committing. We may record the previous parents revision if the
416
# content is actually unchanged against a sole head.
417
if ie.revision is not None:
418
if not self._versioned_root and path == '':
419
# repositories that do not version the root set the root's
420
# revision to the new commit even when no change occurs (more
421
# specifically, they do not record a revision on the root; and
422
# the rev id is assigned to the root during deserialisation -
423
# this masks when a change may have occurred against the basis.
424
# To match this we always issue a delta, because the revision
425
# of the root will always be changing.
426
if ie.file_id in basis_inv:
427
delta = (basis_inv.id2path(ie.file_id), path,
428
ie.file_id, ie)
429
else:
430
# add
431
delta = (None, path, ie.file_id, ie)
432
self._basis_delta.append(delta)
433
return delta, False, None
434
else:
435
# we don't need to commit this, because the caller already
436
# determined that an existing revision of this file is
437
# appropriate. If its not being considered for committing then
438
# it and all its parents to the root must be unaltered so
439
# no-change against the basis.
440
if ie.revision == self._new_revision_id:
441
raise AssertionError("Impossible situation, a skipped "
442
"inventory entry (%r) claims to be modified in this "
443
"commit (%r).", (ie, self._new_revision_id))
444
return None, False, None
445
# XXX: Friction: parent_candidates should return a list not a dict
446
# so that we don't have to walk the inventories again.
447
parent_candiate_entries = ie.parent_candidates(parent_invs)
448
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
449
heads = []
450
for inv in parent_invs:
451
if ie.file_id in inv:
452
old_rev = inv[ie.file_id].revision
453
if old_rev in head_set:
454
heads.append(inv[ie.file_id].revision)
455
head_set.remove(inv[ie.file_id].revision)
456
457
store = False
458
# now we check to see if we need to write a new record to the
459
# file-graph.
460
# We write a new entry unless there is one head to the ancestors, and
461
# the kind-derived content is unchanged.
462
463
# Cheapest check first: no ancestors, or more the one head in the
464
# ancestors, we write a new node.
465
if len(heads) != 1:
466
store = True
467
if not store:
468
# There is a single head, look it up for comparison
469
parent_entry = parent_candiate_entries[heads[0]]
470
# if the non-content specific data has changed, we'll be writing a
471
# node:
472
if (parent_entry.parent_id != ie.parent_id or
473
parent_entry.name != ie.name):
474
store = True
475
# now we need to do content specific checks:
476
if not store:
477
# if the kind changed the content obviously has
478
if kind != parent_entry.kind:
479
store = True
480
# Stat cache fingerprint feedback for the caller - None as we usually
481
# don't generate one.
482
fingerprint = None
483
if kind == 'file':
484
if content_summary[2] is None:
485
raise ValueError("Files must not have executable = None")
486
if not store:
487
# We can't trust a check of the file length because of content
488
# filtering...
489
if (# if the exec bit has changed we have to store:
490
parent_entry.executable != content_summary[2]):
491
store = True
492
elif parent_entry.text_sha1 == content_summary[3]:
493
# all meta and content is unchanged (using a hash cache
494
# hit to check the sha)
495
ie.revision = parent_entry.revision
496
ie.text_size = parent_entry.text_size
497
ie.text_sha1 = parent_entry.text_sha1
498
ie.executable = parent_entry.executable
499
return self._get_delta(ie, basis_inv, path), False, None
500
else:
501
# Either there is only a hash change(no hash cache entry,
502
# or same size content change), or there is no change on
503
# this file at all.
504
# Provide the parent's hash to the store layer, so that the
505
# content is unchanged we will not store a new node.
506
nostore_sha = parent_entry.text_sha1
507
if store:
508
# We want to record a new node regardless of the presence or
509
# absence of a content change in the file.
510
nostore_sha = None
511
ie.executable = content_summary[2]
512
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
513
try:
514
text = file_obj.read()
515
finally:
516
file_obj.close()
517
try:
518
ie.text_sha1, ie.text_size = self._add_text_to_weave(
519
ie.file_id, text, heads, nostore_sha)
520
# Let the caller know we generated a stat fingerprint.
521
fingerprint = (ie.text_sha1, stat_value)
522
except errors.ExistingContent:
523
# Turns out that the file content was unchanged, and we were
524
# only going to store a new node if it was changed. Carry over
525
# the entry.
526
ie.revision = parent_entry.revision
527
ie.text_size = parent_entry.text_size
528
ie.text_sha1 = parent_entry.text_sha1
529
ie.executable = parent_entry.executable
530
return self._get_delta(ie, basis_inv, path), False, None
531
elif kind == 'directory':
532
if not store:
533
# all data is meta here, nothing specific to directory, so
534
# carry over:
535
ie.revision = parent_entry.revision
536
return self._get_delta(ie, basis_inv, path), False, None
537
self._add_text_to_weave(ie.file_id, '', heads, None)
538
elif kind == 'symlink':
539
current_link_target = content_summary[3]
540
if not store:
541
# symlink target is not generic metadata, check if it has
542
# changed.
543
if current_link_target != parent_entry.symlink_target:
544
store = True
545
if not store:
546
# unchanged, carry over.
547
ie.revision = parent_entry.revision
548
ie.symlink_target = parent_entry.symlink_target
549
return self._get_delta(ie, basis_inv, path), False, None
550
ie.symlink_target = current_link_target
551
self._add_text_to_weave(ie.file_id, '', heads, None)
552
elif kind == 'tree-reference':
553
if not store:
554
if content_summary[3] != parent_entry.reference_revision:
555
store = True
556
if not store:
557
# unchanged, carry over.
558
ie.reference_revision = parent_entry.reference_revision
559
ie.revision = parent_entry.revision
560
return self._get_delta(ie, basis_inv, path), False, None
561
ie.reference_revision = content_summary[3]
562
if ie.reference_revision is None:
563
raise AssertionError("invalid content_summary for nested tree: %r"
564
% (content_summary,))
565
self._add_text_to_weave(ie.file_id, '', heads, None)
566
else:
567
raise NotImplementedError('unknown kind')
568
ie.revision = self._new_revision_id
569
self._any_changes = True
570
return self._get_delta(ie, basis_inv, path), True, fingerprint
571
572
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
573
_entry_factory=entry_factory):
209
raise NotImplementedError(self.will_record_deletes)
210
211
def record_iter_changes(self, tree, basis_revision_id, iter_changes):
574
212
"""Record a new tree via iter_changes.
575
213
576
214
:param tree: The tree to obtain text contents from for changed objects.
581
219
to basis_revision_id. The iterator must not include any items with
582
220
a current kind of None - missing items must be either filtered out
583
221
or errored-on beefore record_iter_changes sees the item.
584
:param _entry_factory: Private method to bind entry_factory locally for
585
performance.
586
222
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
587
223
tree._observed_sha1.
588
224
"""
589
# Create an inventory delta based on deltas between all the parents and
590
# deltas between all the parent inventories. We use inventory delta's
591
# between the inventory objects because iter_changes masks
592
# last-changed-field only changes.
593
# Working data:
594
# file_id -> change map, change is fileid, paths, changed, versioneds,
595
# parents, names, kinds, executables
596
merged_ids = {}
597
# {file_id -> revision_id -> inventory entry, for entries in parent
598
# trees that are not parents[0]
599
parent_entries = {}
600
ghost_basis = False
601
try:
602
revtrees = list(self.repository.revision_trees(self.parents))
603
except errors.NoSuchRevision:
604
# one or more ghosts, slow path.
605
revtrees = []
606
for revision_id in self.parents:
607
try:
608
revtrees.append(self.repository.revision_tree(revision_id))
609
except errors.NoSuchRevision:
610
if not revtrees:
611
basis_revision_id = _mod_revision.NULL_REVISION
612
ghost_basis = True
613
revtrees.append(self.repository.revision_tree(
614
_mod_revision.NULL_REVISION))
615
# The basis inventory from a repository
616
if revtrees:
617
basis_inv = revtrees[0].inventory
618
else:
619
basis_inv = self.repository.revision_tree(
620
_mod_revision.NULL_REVISION).inventory
621
if len(self.parents) > 0:
622
if basis_revision_id != self.parents[0] and not ghost_basis:
623
raise Exception(
624
"arbitrary basis parents not yet supported with merges")
625
for revtree in revtrees[1:]:
626
for change in revtree.inventory._make_delta(basis_inv):
627
if change[1] is None:
628
# Not present in this parent.
629
continue
630
if change[2] not in merged_ids:
631
if change[0] is not None:
632
basis_entry = basis_inv[change[2]]
633
merged_ids[change[2]] = [
634
# basis revid
635
basis_entry.revision,
636
# new tree revid
637
change[3].revision]
638
parent_entries[change[2]] = {
639
# basis parent
640
basis_entry.revision:basis_entry,
641
# this parent
642
change[3].revision:change[3],
643
}
644
else:
645
merged_ids[change[2]] = [change[3].revision]
646
parent_entries[change[2]] = {change[3].revision:change[3]}
647
else:
648
merged_ids[change[2]].append(change[3].revision)
649
parent_entries[change[2]][change[3].revision] = change[3]
650
else:
651
merged_ids = {}
652
# Setup the changes from the tree:
653
# changes maps file_id -> (change, [parent revision_ids])
654
changes= {}
655
for change in iter_changes:
656
# This probably looks up in basis_inv way to much.
657
if change[1][0] is not None:
658
head_candidate = [basis_inv[change[0]].revision]
659
else:
660
head_candidate = []
661
changes[change[0]] = change, merged_ids.get(change[0],
662
head_candidate)
663
unchanged_merged = set(merged_ids) - set(changes)
664
# Extend the changes dict with synthetic changes to record merges of
665
# texts.
666
for file_id in unchanged_merged:
667
# Record a merged version of these items that did not change vs the
668
# basis. This can be either identical parallel changes, or a revert
669
# of a specific file after a merge. The recorded content will be
670
# that of the current tree (which is the same as the basis), but
671
# the per-file graph will reflect a merge.
672
# NB:XXX: We are reconstructing path information we had, this
673
# should be preserved instead.
674
# inv delta change: (file_id, (path_in_source, path_in_target),
675
# changed_content, versioned, parent, name, kind,
676
# executable)
677
try:
678
basis_entry = basis_inv[file_id]
679
except errors.NoSuchId:
680
# a change from basis->some_parents but file_id isn't in basis
681
# so was new in the merge, which means it must have changed
682
# from basis -> current, and as it hasn't the add was reverted
683
# by the user. So we discard this change.
684
pass
685
else:
686
change = (file_id,
687
(basis_inv.id2path(file_id), tree.id2path(file_id)),
688
False, (True, True),
689
(basis_entry.parent_id, basis_entry.parent_id),
690
(basis_entry.name, basis_entry.name),
691
(basis_entry.kind, basis_entry.kind),
692
(basis_entry.executable, basis_entry.executable))
693
changes[file_id] = (change, merged_ids[file_id])
694
# changes contains tuples with the change and a set of inventory
695
# candidates for the file.
696
# inv delta is:
697
# old_path, new_path, file_id, new_inventory_entry
698
seen_root = False # Is the root in the basis delta?
699
inv_delta = self._basis_delta
700
modified_rev = self._new_revision_id
701
for change, head_candidates in changes.values():
702
if change[3][1]: # versioned in target.
703
# Several things may be happening here:
704
# We may have a fork in the per-file graph
705
# - record a change with the content from tree
706
# We may have a change against < all trees
707
# - carry over the tree that hasn't changed
708
# We may have a change against all trees
709
# - record the change with the content from tree
710
kind = change[6][1]
711
file_id = change[0]
712
entry = _entry_factory[kind](file_id, change[5][1],
713
change[4][1])
714
head_set = self._heads(change[0], set(head_candidates))
715
heads = []
716
# Preserve ordering.
717
for head_candidate in head_candidates:
718
if head_candidate in head_set:
719
heads.append(head_candidate)
720
head_set.remove(head_candidate)
721
carried_over = False
722
if len(heads) == 1:
723
# Could be a carry-over situation:
724
parent_entry_revs = parent_entries.get(file_id, None)
725
if parent_entry_revs:
726
parent_entry = parent_entry_revs.get(heads[0], None)
727
else:
728
parent_entry = None
729
if parent_entry is None:
730
# The parent iter_changes was called against is the one
731
# that is the per-file head, so any change is relevant
732
# iter_changes is valid.
733
carry_over_possible = False
734
else:
735
# could be a carry over situation
736
# A change against the basis may just indicate a merge,
737
# we need to check the content against the source of the
738
# merge to determine if it was changed after the merge
739
# or carried over.
740
if (parent_entry.kind != entry.kind or
741
parent_entry.parent_id != entry.parent_id or
742
parent_entry.name != entry.name):
743
# Metadata common to all entries has changed
744
# against per-file parent
745
carry_over_possible = False
746
else:
747
carry_over_possible = True
748
# per-type checks for changes against the parent_entry
749
# are done below.
750
else:
751
# Cannot be a carry-over situation
752
carry_over_possible = False
753
# Populate the entry in the delta
754
if kind == 'file':
755
# XXX: There is still a small race here: If someone reverts the content of a file
756
# after iter_changes examines and decides it has changed,
757
# we will unconditionally record a new version even if some
758
# other process reverts it while commit is running (with
759
# the revert happening after iter_changes did it's
760
# examination).
761
if change[7][1]:
762
entry.executable = True
763
else:
764
entry.executable = False
765
if (carry_over_possible and
766
parent_entry.executable == entry.executable):
767
# Check the file length, content hash after reading
768
# the file.
769
nostore_sha = parent_entry.text_sha1
770
else:
771
nostore_sha = None
772
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
773
try:
774
text = file_obj.read()
775
finally:
776
file_obj.close()
777
try:
778
entry.text_sha1, entry.text_size = self._add_text_to_weave(
779
file_id, text, heads, nostore_sha)
780
yield file_id, change[1][1], (entry.text_sha1, stat_value)
781
except errors.ExistingContent:
782
# No content change against a carry_over parent
783
# Perhaps this should also yield a fs hash update?
784
carried_over = True
785
entry.text_size = parent_entry.text_size
786
entry.text_sha1 = parent_entry.text_sha1
787
elif kind == 'symlink':
788
# Wants a path hint?
789
entry.symlink_target = tree.get_symlink_target(file_id)
790
if (carry_over_possible and
791
parent_entry.symlink_target == entry.symlink_target):
792
carried_over = True
793
else:
794
self._add_text_to_weave(change[0], '', heads, None)
795
elif kind == 'directory':
796
if carry_over_possible:
797
carried_over = True
798
else:
799
# Nothing to set on the entry.
800
# XXX: split into the Root and nonRoot versions.
801
if change[1][1] != '' or self.repository.supports_rich_root():
802
self._add_text_to_weave(change[0], '', heads, None)
803
elif kind == 'tree-reference':
804
if not self.repository._format.supports_tree_reference:
805
# This isn't quite sane as an error, but we shouldn't
806
# ever see this code path in practice: tree's don't
807
# permit references when the repo doesn't support tree
808
# references.
809
raise errors.UnsupportedOperation(tree.add_reference,
810
self.repository)
811
reference_revision = tree.get_reference_revision(change[0])
812
entry.reference_revision = reference_revision
813
if (carry_over_possible and
814
parent_entry.reference_revision == reference_revision):
815
carried_over = True
816
else:
817
self._add_text_to_weave(change[0], '', heads, None)
818
else:
819
raise AssertionError('unknown kind %r' % kind)
820
if not carried_over:
821
entry.revision = modified_rev
822
else:
823
entry.revision = parent_entry.revision
824
else:
825
entry = None
826
new_path = change[1][1]
827
inv_delta.append((change[1][0], new_path, change[0], entry))
828
if new_path == '':
829
seen_root = True
830
self.new_inventory = None
831
if len(inv_delta):
832
# This should perhaps be guarded by a check that the basis we
833
# commit against is the basis for the commit and if not do a delta
834
# against the basis.
835
self._any_changes = True
836
if not seen_root:
837
# housekeeping root entry changes do not affect no-change commits.
838
self._require_root_change(tree)
839
self.basis_delta_revision = basis_revision_id
840
841
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
842
parent_keys = tuple([(file_id, parent) for parent in parents])
843
return self.repository.texts._add_text(
844
(file_id, self._new_revision_id), parent_keys, new_text,
845
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
846
847
848
class RootCommitBuilder(CommitBuilder):
849
"""This commitbuilder actually records the root id"""
850
851
# the root entry gets versioned properly by this builder.
852
_versioned_root = True
853
854
def _check_root(self, ie, parent_invs, tree):
855
"""Helper for record_entry_contents.
856
857
:param ie: An entry being added.
858
:param parent_invs: The inventories of the parent revisions of the
859
commit.
860
:param tree: The tree that is being committed.
861
"""
862
863
def _require_root_change(self, tree):
864
"""Enforce an appropriate root object change.
865
866
This is called once when record_iter_changes is called, if and only if
867
the root was not in the delta calculated by record_iter_changes.
868
869
:param tree: The tree which is being committed.
870
"""
871
# versioned roots do not change unless the tree found a change.
225
raise NotImplementedError(self.record_iter_changes)
872
226
873
227
874
228
class RepositoryWriteLockResult(LogicalLockResult):
899
253
revisions and file history. It's normally accessed only by the Branch,
900
254
which views a particular line of development through that history.
901
255
902
The Repository builds on top of some byte storage facilies (the revisions,
903
signatures, inventories, texts and chk_bytes attributes) and a Transport,
904
which respectively provide byte storage and a means to access the (possibly
905
remote) disk.
906
907
The byte storage facilities are addressed via tuples, which we refer to
908
as 'keys' throughout the code base. Revision_keys, inventory_keys and
909
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
910
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
911
byte string made up of a hash identifier and a hash value.
912
We use this interface because it allows low friction with the underlying
913
code that implements disk indices, network encoding and other parts of
914
bzrlib.
915
916
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
917
the serialised revisions for the repository. This can be used to obtain
918
revision graph information or to access raw serialised revisions.
919
The result of trying to insert data into the repository via this store
920
is undefined: it should be considered read-only except for implementors
921
of repositories.
922
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
923
the serialised signatures for the repository. This can be used to
924
obtain access to raw serialised signatures. The result of trying to
925
insert data into the repository via this store is undefined: it should
926
be considered read-only except for implementors of repositories.
927
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
928
the serialised inventories for the repository. This can be used to
929
obtain unserialised inventories. The result of trying to insert data
930
into the repository via this store is undefined: it should be
931
considered read-only except for implementors of repositories.
932
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
933
texts of files and directories for the repository. This can be used to
934
obtain file texts or file graphs. Note that Repository.iter_file_bytes
935
is usually a better interface for accessing file texts.
936
The result of trying to insert data into the repository via this store
937
is undefined: it should be considered read-only except for implementors
938
of repositories.
939
:ivar chk_bytes: A bzrlib.versionedfile.VersionedFiles instance containing
940
any data the repository chooses to store or have indexed by its hash.
941
The result of trying to insert data into the repository via this store
942
is undefined: it should be considered read-only except for implementors
943
of repositories.
944
:ivar _transport: Transport for file access to repository, typically
945
pointing to .bzr/repository.
256
See VersionedFileRepository in bzrlib.vf_repository for the
257
base class for most Bazaar repositories.
946
258
"""
947
259
948
# What class to use for a CommitBuilder. Often its simpler to change this
949
# in a Repository class subclass rather than to override
950
# get_commit_builder.
951
_commit_builder_class = CommitBuilder
952
# The search regex used by xml based repositories to determine what things
953
# where changed in a single commit.
954
_file_ids_altered_regex = lazy_regex.lazy_compile(
955
r'file_id="(?P<file_id>[^"]+)"'
956
r'.* revision="(?P<revision_id>[^"]+)"'
957
)
958
959
260
def abort_write_group(self, suppress_errors=False):
960
261
"""Commit the contents accrued within the current write group.
961
262
983
284
raise
984
285
mutter('abort_write_group failed')
985
286
log_exception_quietly()
986
note('bzr: ERROR (ignored): %s', exc)
287
note(gettext('bzr: ERROR (ignored): %s'), exc)
987
288
self._write_group = None
988
289
989
290
def _abort_write_group(self):
1004
305
1005
306
:param repository: A repository.
1006
307
"""
1007
if not self._format.supports_external_lookups:
1008
raise errors.UnstackableRepositoryFormat(self._format, self.base)
1009
if self.is_locked():
1010
# This repository will call fallback.unlock() when we transition to
1011
# the unlocked state, so we make sure to increment the lock count
1012
repository.lock_read()
1013
self._check_fallback_repository(repository)
1014
self._fallback_repositories.append(repository)
1015
self.texts.add_fallback_versioned_files(repository.texts)
1016
self.inventories.add_fallback_versioned_files(repository.inventories)
1017
self.revisions.add_fallback_versioned_files(repository.revisions)
1018
self.signatures.add_fallback_versioned_files(repository.signatures)
1019
if self.chk_bytes is not None:
1020
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
308
raise NotImplementedError(self.add_fallback_repository)
1021
309
1022
310
def _check_fallback_repository(self, repository):
1023
311
"""Check that this repository can fallback to repository safely.
1028
316
"""
1029
317
return InterRepository._assert_same_model(self, repository)
1030
318
1031
def add_inventory(self, revision_id, inv, parents):
1032
"""Add the inventory inv to the repository as revision_id.
1033
1034
:param parents: The revision ids of the parents that revision_id
1035
is known to have and are in the repository already.
1036
1037
:returns: The validator(which is a sha1 digest, though what is sha'd is
1038
repository format specific) of the serialized inventory.
1039
"""
1040
if not self.is_in_write_group():
1041
raise AssertionError("%r not in write group" % (self,))
1042
_mod_revision.check_not_reserved_id(revision_id)
1043
if not (inv.revision_id is None or inv.revision_id == revision_id):
1044
raise AssertionError(
1045
"Mismatch between inventory revision"
1046
" id and insertion revid (%r, %r)"
1047
% (inv.revision_id, revision_id))
1048
if inv.root is None:
1049
raise errors.RootMissing()
1050
return self._add_inventory_checked(revision_id, inv, parents)
1051
1052
def _add_inventory_checked(self, revision_id, inv, parents):
1053
"""Add inv to the repository after checking the inputs.
1054
1055
This function can be overridden to allow different inventory styles.
1056
1057
:seealso: add_inventory, for the contract.
1058
"""
1059
inv_lines = self._serializer.write_inventory_to_lines(inv)
1060
return self._inventory_add_lines(revision_id, parents,
1061
inv_lines, check_content=False)
1062
1063
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1064
parents, basis_inv=None, propagate_caches=False):
1065
"""Add a new inventory expressed as a delta against another revision.
1066
1067
See the inventory developers documentation for the theory behind
1068
inventory deltas.
1069
1070
:param basis_revision_id: The inventory id the delta was created
1071
against. (This does not have to be a direct parent.)
1072
:param delta: The inventory delta (see Inventory.apply_delta for
1073
details).
1074
:param new_revision_id: The revision id that the inventory is being
1075
added for.
1076
:param parents: The revision ids of the parents that revision_id is
1077
known to have and are in the repository already. These are supplied
1078
for repositories that depend on the inventory graph for revision
1079
graph access, as well as for those that pun ancestry with delta
1080
compression.
1081
:param basis_inv: The basis inventory if it is already known,
1082
otherwise None.
1083
:param propagate_caches: If True, the caches for this inventory are
1084
copied to and updated for the result if possible.
1085
1086
:returns: (validator, new_inv)
1087
The validator(which is a sha1 digest, though what is sha'd is
1088
repository format specific) of the serialized inventory, and the
1089
resulting inventory.
1090
"""
1091
if not self.is_in_write_group():
1092
raise AssertionError("%r not in write group" % (self,))
1093
_mod_revision.check_not_reserved_id(new_revision_id)
1094
basis_tree = self.revision_tree(basis_revision_id)
1095
basis_tree.lock_read()
1096
try:
1097
# Note that this mutates the inventory of basis_tree, which not all
1098
# inventory implementations may support: A better idiom would be to
1099
# return a new inventory, but as there is no revision tree cache in
1100
# repository this is safe for now - RBC 20081013
1101
if basis_inv is None:
1102
basis_inv = basis_tree.inventory
1103
basis_inv.apply_delta(delta)
1104
basis_inv.revision_id = new_revision_id
1105
return (self.add_inventory(new_revision_id, basis_inv, parents),
1106
basis_inv)
1107
finally:
1108
basis_tree.unlock()
1109
1110
def _inventory_add_lines(self, revision_id, parents, lines,
1111
check_content=True):
1112
"""Store lines in inv_vf and return the sha1 of the inventory."""
1113
parents = [(parent,) for parent in parents]
1114
result = self.inventories.add_lines((revision_id,), parents, lines,
1115
check_content=check_content)[0]
1116
self.inventories._access.flush()
1117
return result
1118
1119
def add_revision(self, revision_id, rev, inv=None, config=None):
1120
"""Add rev to the revision store as revision_id.
1121
1122
:param revision_id: the revision id to use.
1123
:param rev: The revision object.
1124
:param inv: The inventory for the revision. if None, it will be looked
1125
up in the inventory storer
1126
:param config: If None no digital signature will be created.
1127
If supplied its signature_needed method will be used
1128
to determine if a signature should be made.
1129
"""
1130
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
1131
# rev.parent_ids?
1132
_mod_revision.check_not_reserved_id(revision_id)
1133
if config is not None and config.signature_needed():
1134
if inv is None:
1135
inv = self.get_inventory(revision_id)
1136
plaintext = Testament(rev, inv).as_short_text()
1137
self.store_revision_signature(
1138
gpg.GPGStrategy(config), plaintext, revision_id)
1139
# check inventory present
1140
if not self.inventories.get_parent_map([(revision_id,)]):
1141
if inv is None:
1142
raise errors.WeaveRevisionNotPresent(revision_id,
1143
self.inventories)
1144
else:
1145
# yes, this is not suitable for adding with ghosts.
1146
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
1147
rev.parent_ids)
1148
else:
1149
key = (revision_id,)
1150
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
1151
self._add_revision(rev)
1152
1153
def _add_revision(self, revision):
1154
text = self._serializer.write_revision_to_string(revision)
1155
key = (revision.revision_id,)
1156
parents = tuple((parent,) for parent in revision.parent_ids)
1157
self.revisions.add_lines(key, parents, osutils.split_lines(text))
1158
1159
319
def all_revision_ids(self):
1160
320
"""Returns a list of all the revision ids in the repository.
1161
321
1184
344
"""
1185
345
self.control_files.break_lock()
1186
346
1187
@needs_read_lock
1188
def _eliminate_revisions_not_present(self, revision_ids):
1189
"""Check every revision id in revision_ids to see if we have it.
1190
1191
Returns a set of the present revisions.
1192
"""
1193
result = []
1194
graph = self.get_graph()
1195
parent_map = graph.get_parent_map(revision_ids)
1196
# The old API returned a list, should this actually be a set?
1197
return parent_map.keys()
1198
1199
def _check_inventories(self, checker):
1200
"""Check the inventories found from the revision scan.
1201
1202
This is responsible for verifying the sha1 of inventories and
1203
creating a pending_keys set that covers data referenced by inventories.
1204
"""
1205
bar = ui.ui_factory.nested_progress_bar()
1206
try:
1207
self._do_check_inventories(checker, bar)
1208
finally:
1209
bar.finished()
1210
1211
def _do_check_inventories(self, checker, bar):
1212
"""Helper for _check_inventories."""
1213
revno = 0
1214
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1215
kinds = ['chk_bytes', 'texts']
1216
count = len(checker.pending_keys)
1217
bar.update("inventories", 0, 2)
1218
current_keys = checker.pending_keys
1219
checker.pending_keys = {}
1220
# Accumulate current checks.
1221
for key in current_keys:
1222
if key[0] != 'inventories' and key[0] not in kinds:
1223
checker._report_items.append('unknown key type %r' % (key,))
1224
keys[key[0]].add(key[1:])
1225
if keys['inventories']:
1226
# NB: output order *should* be roughly sorted - topo or
1227
# inverse topo depending on repository - either way decent
1228
# to just delta against. However, pre-CHK formats didn't
1229
# try to optimise inventory layout on disk. As such the
1230
# pre-CHK code path does not use inventory deltas.
1231
last_object = None
1232
for record in self.inventories.check(keys=keys['inventories']):
1233
if record.storage_kind == 'absent':
1234
checker._report_items.append(
1235
'Missing inventory {%s}' % (record.key,))
1236
else:
1237
last_object = self._check_record('inventories', record,
1238
checker, last_object,
1239
current_keys[('inventories',) + record.key])
1240
del keys['inventories']
1241
else:
1242
return
1243
bar.update("texts", 1)
1244
while (checker.pending_keys or keys['chk_bytes']
1245
or keys['texts']):
1246
# Something to check.
1247
current_keys = checker.pending_keys
1248
checker.pending_keys = {}
1249
# Accumulate current checks.
1250
for key in current_keys:
1251
if key[0] not in kinds:
1252
checker._report_items.append('unknown key type %r' % (key,))
1253
keys[key[0]].add(key[1:])
1254
# Check the outermost kind only - inventories || chk_bytes || texts
1255
for kind in kinds:
1256
if keys[kind]:
1257
last_object = None
1258
for record in getattr(self, kind).check(keys=keys[kind]):
1259
if record.storage_kind == 'absent':
1260
checker._report_items.append(
1261
'Missing %s {%s}' % (kind, record.key,))
1262
else:
1263
last_object = self._check_record(kind, record,
1264
checker, last_object, current_keys[(kind,) + record.key])
1265
keys[kind] = set()
1266
break
1267
1268
def _check_record(self, kind, record, checker, last_object, item_data):
1269
"""Check a single text from this repository."""
1270
if kind == 'inventories':
1271
rev_id = record.key[0]
1272
inv = self._deserialise_inventory(rev_id,
1273
record.get_bytes_as('fulltext'))
1274
if last_object is not None:
1275
delta = inv._make_delta(last_object)
1276
for old_path, path, file_id, ie in delta:
1277
if ie is None:
1278
continue
1279
ie.check(checker, rev_id, inv)
1280
else:
1281
for path, ie in inv.iter_entries():
1282
ie.check(checker, rev_id, inv)
1283
if self._format.fast_deltas:
1284
return inv
1285
elif kind == 'chk_bytes':
1286
# No code written to check chk_bytes for this repo format.
1287
checker._report_items.append(
1288
'unsupported key type chk_bytes for %s' % (record.key,))
1289
elif kind == 'texts':
1290
self._check_text(record, checker, item_data)
1291
else:
1292
checker._report_items.append(
1293
'unknown key type %s for %s' % (kind, record.key))
1294
1295
def _check_text(self, record, checker, item_data):
1296
"""Check a single text."""
1297
# Check it is extractable.
1298
# TODO: check length.
1299
if record.storage_kind == 'chunked':
1300
chunks = record.get_bytes_as(record.storage_kind)
1301
sha1 = osutils.sha_strings(chunks)
1302
length = sum(map(len, chunks))
1303
else:
1304
content = record.get_bytes_as('fulltext')
1305
sha1 = osutils.sha_string(content)
1306
length = len(content)
1307
if item_data and sha1 != item_data[1]:
1308
checker._report_items.append(
1309
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1310
(record.key, sha1, item_data[1], item_data[2]))
1311
1312
347
@staticmethod
1313
def create(a_bzrdir):
1314
"""Construct the current default format repository in a_bzrdir."""
1315
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)
1316
351
1317
def __init__(self, _format, a_bzrdir, control_files):
352
def __init__(self, _format, controldir, control_files):
1318
353
"""instantiate a Repository.
1319
354
1320
355
:param _format: The format of the repository on disk.
1321
: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.
1322
358
"""
1323
359
# In the future we will have a single api for all stores for
1324
360
# getting file texts, inventories and revisions, then
1326
362
super(Repository, self).__init__()
1327
363
self._format = _format
1328
364
# the following are part of the public API for Repository:
1329
self.bzrdir = a_bzrdir
365
self.bzrdir = controldir
1330
366
self.control_files = control_files
1331
self._transport = control_files._transport
1332
self.base = self._transport.base
1333
367
# for tests
1334
self._reconcile_does_inventory_gc = True
1335
self._reconcile_fixes_text_parents = False
1336
self._reconcile_backsup_inventory = True
1337
368
self._write_group = None
1338
369
# Additional places to query for data.
1339
370
self._fallback_repositories = []
1340
# An InventoryEntry cache, used during deserialization
1341
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1342
# Is it safe to return inventory entries directly from the entry cache,
1343
# rather copying them?
1344
self._safe_to_return_from_cache = False
1345
371
1346
372
@property
1347
373
def user_transport(self):
1381
407
"""
1382
408
if self.__class__ is not other.__class__:
1383
409
return False
1384
return (self._transport.base == other._transport.base)
410
return (self.control_url == other.control_url)
1385
411
1386
412
def is_in_write_group(self):
1387
413
"""Return True if there is an open write group.
1487
513
if revid and committers:
1488
514
result['committers'] = 0
1489
515
if revid and revid != _mod_revision.NULL_REVISION:
516
graph = self.get_graph()
1490
517
if committers:
1491
518
all_committers = set()
1492
revisions = self.get_ancestry(revid)
1493
# pop the leading None
1494
revisions.pop(0)
1495
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
1496
522
if not committers:
1497
523
# ignore the revisions in the middle - just grab first and last
1498
524
revisions = revisions[0], revisions[-1]
1499
525
for revision in self.get_revisions(revisions):
1500
if not first_revision:
1501
first_revision = revision
526
if not last_revision:
527
last_revision = revision
1502
528
if committers:
1503
529
all_committers.add(revision.committer)
1504
last_revision = revision
530
first_revision = revision
1505
531
if committers:
1506
532
result['committers'] = len(all_committers)
1507
533
result['firstrev'] = (first_revision.timestamp,
1508
534
first_revision.timezone)
1509
535
result['latestrev'] = (last_revision.timestamp,
1510
536
last_revision.timezone)
1511
1512
# now gather global repository information
1513
# XXX: This is available for many repos regardless of listability.
1514
if self.user_transport.listable():
1515
# XXX: do we want to __define len__() ?
1516
# Maybe the versionedfiles object should provide a different
1517
# method to get the number of keys.
1518
result['revisions'] = len(self.revisions.keys())
1519
# result['size'] = t
1520
537
return result
1521
538
1522
539
def find_branches(self, using=False):
1533
550
def __init__(self):
1534
551
self.first_call = True
1535
552
1536
def __call__(self, bzrdir):
1537
# 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
1538
555
# containing the current repo.
1539
556
if not self.first_call:
1540
557
try:
1541
repository = bzrdir.open_repository()
558
repository = controldir.open_repository()
1542
559
except errors.NoRepositoryPresent:
1543
560
pass
1544
561
else:
1545
562
return False, ([], repository)
1546
563
self.first_call = False
1547
value = (bzrdir.list_branches(), None)
564
value = (controldir.list_branches(), None)
1548
565
return True, value
1549
566
1550
567
ret = []
1551
for branches, repository in bzrdir.BzrDir.find_bzrdirs(
568
for branches, repository in controldir.ControlDir.find_bzrdirs(
1552
569
self.user_transport, evaluate=Evaluator()):
1553
570
if branches is not None:
1554
571
ret.extend(branches)
1557
574
return ret
1558
575
1559
576
@needs_read_lock
1560
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):
1561
581
"""Return the revision ids that other has that this does not.
1562
582
1563
583
These are returned in topological order.
1564
584
1565
585
revision_id: only return revision ids included by revision_id.
1566
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]
1567
597
return InterRepository.get(other, self).search_missing_revision_ids(
1568
revision_id, find_ghosts)
598
find_ghosts=find_ghosts, revision_ids=revision_ids,
599
if_present_ids=if_present_ids, limit=limit)
1569
600
1570
601
@staticmethod
1571
602
def open(base):
1574
605
For instance, if the repository is at URL/.bzr/repository,
1575
606
Repository.open(URL) -> a Repository instance.
1576
607
"""
1577
control = bzrdir.BzrDir.open(base)
608
control = controldir.ControlDir.open(base)
1578
609
return control.open_repository()
1579
610
1580
611
def copy_content_into(self, destination, revision_id=None):
1611
642
"""
1612
643
1613
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
"""
1614
651
raise errors.UnsuspendableWriteGroup(self)
1615
652
1616
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1617
"""Return the keys of missing inventory parents for revisions added in
1618
this write group.
1619
1620
A revision is not complete if the inventory delta for that revision
1621
cannot be calculated. Therefore if the parent inventories of a
1622
revision are not present, the revision is incomplete, and e.g. cannot
1623
be streamed by a smart server. This method finds missing inventory
1624
parents for revisions added in this write group.
1625
"""
1626
if not self._format.supports_external_lookups:
1627
# This is only an issue for stacked repositories
1628
return set()
1629
if not self.is_in_write_group():
1630
raise AssertionError('not in a write group')
1631
1632
# XXX: We assume that every added revision already has its
1633
# corresponding inventory, so we only check for parent inventories that
1634
# might be missing, rather than all inventories.
1635
parents = set(self.revisions._index.get_missing_parents())
1636
parents.discard(_mod_revision.NULL_REVISION)
1637
unstacked_inventories = self.inventories._index
1638
present_inventories = unstacked_inventories.get_parent_map(
1639
key[-1:] for key in parents)
1640
parents.difference_update(present_inventories)
1641
if len(parents) == 0:
1642
# No missing parent inventories.
1643
return set()
1644
if not check_for_missing_texts:
1645
return set(('inventories', rev_id) for (rev_id,) in parents)
1646
# Ok, now we have a list of missing inventories. But these only matter
1647
# if the inventories that reference them are missing some texts they
1648
# appear to introduce.
1649
# XXX: Texts referenced by all added inventories need to be present,
1650
# but at the moment we're only checking for texts referenced by
1651
# inventories at the graph's edge.
1652
key_deps = self.revisions._index._key_dependencies
1653
key_deps.satisfy_refs_for_keys(present_inventories)
1654
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1655
file_ids = self.fileids_altered_by_revision_ids(referrers)
1656
missing_texts = set()
1657
for file_id, version_ids in file_ids.iteritems():
1658
missing_texts.update(
1659
(file_id, version_id) for version_id in version_ids)
1660
present_texts = self.texts.get_parent_map(missing_texts)
1661
missing_texts.difference_update(present_texts)
1662
if not missing_texts:
1663
# No texts are missing, so all revisions and their deltas are
1664
# reconstructable.
1665
return set()
1666
# Alternatively the text versions could be returned as the missing
1667
# keys, but this is likely to be less data.
1668
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1669
return missing_keys
1670
1671
653
def refresh_data(self):
1672
654
"""Re-read any data needed to synchronise with disk.
1673
655
1693
675
def _resume_write_group(self, tokens):
1694
676
raise errors.UnsuspendableWriteGroup(self)
1695
677
1696
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
1697
fetch_spec=None):
678
def fetch(self, source, revision_id=None, find_ghosts=False):
1698
679
"""Fetch the content required to construct revision_id from source.
1699
680
1700
If revision_id is None and fetch_spec is None, then all content is
1701
copied.
681
If revision_id is None, then all content is copied.
1702
682
1703
683
fetch() may not be used when the repository is in a write group -
1704
684
either finish the current write group before using fetch, or use
1710
690
:param revision_id: If specified, all the content needed for this
1711
691
revision ID will be copied to the target. Fetch will determine for
1712
692
itself which content needs to be copied.
1713
:param fetch_spec: If specified, a SearchResult or
1714
PendingAncestryResult that describes which revisions to copy. This
1715
allows copying multiple heads at once. Mutually exclusive with
1716
revision_id.
1717
693
"""
1718
if fetch_spec is not None and revision_id is not None:
1719
raise AssertionError(
1720
"fetch_spec and revision_id are mutually exclusive.")
1721
694
if self.is_in_write_group():
1722
695
raise errors.InternalBzrError(
1723
696
"May not fetch while in a write group.")
1725
698
# TODO: lift out to somewhere common with RemoteRepository
1726
699
# <https://bugs.launchpad.net/bzr/+bug/401646>
1727
700
if (self.has_same_location(source)
1728
and fetch_spec is None
1729
701
and self._has_same_fallbacks(source)):
1730
702
# check that last_revision is in 'from' and then return a
1731
703
# no-operation.
1733
705
not _mod_revision.is_null(revision_id)):
1734
706
self.get_revision(revision_id)
1735
707
return 0, []
1736
# if there is no specific appropriate InterRepository, this will get
1737
# the InterRepository base class, which raises an
1738
# IncompatibleRepositories when asked to fetch.
1739
708
inter = InterRepository.get(source, self)
1740
return inter.fetch(revision_id=revision_id, pb=pb,
1741
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
709
return inter.fetch(revision_id=revision_id, find_ghosts=find_ghosts)
1742
710
1743
711
def create_bundle(self, target, base, fileobj, format=None):
1744
712
return serializer.write_bundle(self, target, base, fileobj, format)
1745
713
1746
def get_commit_builder(self, branch, parents, config, timestamp=None,
714
def get_commit_builder(self, branch, parents, config_stack, timestamp=None,
1747
715
timezone=None, committer=None, revprops=None,
1748
revision_id=None):
716
revision_id=None, lossy=False):
1749
717
"""Obtain a CommitBuilder for this repository.
1750
718
1751
719
:param branch: Branch to commit to.
1752
720
:param parents: Revision ids of the parents of the new revision.
1753
:param config: Configuration to use.
721
:param config_stack: Configuration stack to use.
1754
722
:param timestamp: Optional timestamp recorded for commit.
1755
723
:param timezone: Optional timezone for timestamp.
1756
724
:param committer: Optional committer to set for commit.
1757
725
:param revprops: Optional dictionary of revision properties.
1758
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
1759
729
"""
1760
if self._fallback_repositories:
1761
raise errors.BzrError("Cannot commit from a lightweight checkout "
1762
"to a stacked branch. See "
1763
"https://bugs.launchpad.net/bzr/+bug/375013 for details.")
1764
result = self._commit_builder_class(self, parents, config,
1765
timestamp, timezone, committer, revprops, revision_id)
1766
self.start_write_group()
1767
return result
730
raise NotImplementedError(self.get_commit_builder)
1768
731
1769
732
@only_raises(errors.LockNotHeld, errors.LockBroken)
1770
733
def unlock(self):
1777
740
'Must end write groups before releasing write locks.')
1778
741
self.control_files.unlock()
1779
742
if self.control_files._lock_count == 0:
1780
self._inventory_entry_cache.clear()
1781
743
for repo in self._fallback_repositories:
1782
744
repo.unlock()
1783
745
1784
746
@needs_read_lock
1785
def clone(self, a_bzrdir, revision_id=None):
1786
"""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.
1787
749
1788
750
Currently no check is made that the format of this repository and
1789
751
the bzrdir format are compatible. FIXME RBC 20060201.
1792
754
"""
1793
755
# TODO: deprecate after 0.16; cloning this with all its settings is
1794
756
# probably not very useful -- mbp 20070423
1795
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())
1796
759
self.copy_content_into(dest_repo, revision_id)
1797
760
return dest_repo
1798
761
1848
811
dest_repo = a_bzrdir.open_repository()
1849
812
return dest_repo
1850
813
1851
def _get_sink(self):
1852
"""Return a sink for streaming into this repository."""
1853
return StreamSink(self)
1854
1855
def _get_source(self, to_format):
1856
"""Return a source for streaming from this repository."""
1857
return StreamSource(self, to_format)
1858
1859
814
@needs_read_lock
1860
815
def has_revision(self, revision_id):
1861
816
"""True if this repository has a copy of the revision."""
1868
823
:param revision_ids: An iterable of revision_ids.
1869
824
:return: A set of the revision_ids that were present.
1870
825
"""
1871
parent_map = self.revisions.get_parent_map(
1872
[(rev_id,) for rev_id in revision_ids])
1873
result = set()
1874
if _mod_revision.NULL_REVISION in revision_ids:
1875
result.add(_mod_revision.NULL_REVISION)
1876
result.update([key[0] for key in parent_map])
1877
return result
826
raise NotImplementedError(self.has_revisions)
1878
827
1879
828
@needs_read_lock
1880
829
def get_revision(self, revision_id):
1881
830
"""Return the Revision object for a named revision."""
1882
831
return self.get_revisions([revision_id])[0]
1883
832
1884
@needs_read_lock
1885
833
def get_revision_reconcile(self, revision_id):
1886
834
"""'reconcile' helper routine that allows access to a revision always.
1887
835
1890
838
be used by reconcile, or reconcile-alike commands that are correcting
1891
839
or testing the revision graph.
1892
840
"""
1893
return self._get_revisions([revision_id])[0]
841
raise NotImplementedError(self.get_revision_reconcile)
1894
842
1895
@needs_read_lock
1896
843
def get_revisions(self, revision_ids):
1897
844
"""Get many revisions at once.
1898
845
1899
846
Repositories that need to check data on every revision read should
1900
847
subclass this method.
1901
848
"""
1902
return self._get_revisions(revision_ids)
1903
1904
@needs_read_lock
1905
def _get_revisions(self, revision_ids):
1906
"""Core work logic to get many revisions without sanity checks."""
1907
revs = {}
1908
for revid, rev in self._iter_revisions(revision_ids):
1909
if rev is None:
1910
raise errors.NoSuchRevision(self, revid)
1911
revs[revid] = rev
1912
return [revs[revid] for revid in revision_ids]
1913
1914
def _iter_revisions(self, revision_ids):
1915
"""Iterate over revision objects.
1916
1917
:param revision_ids: An iterable of revisions to examine. None may be
1918
passed to request all revisions known to the repository. Note that
1919
not all repositories can find unreferenced revisions; for those
1920
repositories only referenced ones will be returned.
1921
:return: An iterator of (revid, revision) tuples. Absent revisions (
1922
those asked for but not available) are returned as (revid, None).
1923
"""
1924
if revision_ids is None:
1925
revision_ids = self.all_revision_ids()
1926
else:
1927
for rev_id in revision_ids:
1928
if not rev_id or not isinstance(rev_id, basestring):
1929
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1930
keys = [(key,) for key in revision_ids]
1931
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1932
for record in stream:
1933
revid = record.key[0]
1934
if record.storage_kind == 'absent':
1935
yield (revid, None)
1936
else:
1937
text = record.get_bytes_as('fulltext')
1938
rev = self._serializer.read_revision_from_string(text)
1939
yield (revid, rev)
849
raise NotImplementedError(self.get_revisions)
1940
850
1941
851
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1942
852
"""Produce a generator of revision deltas.
1997
907
signature = gpg_strategy.sign(plaintext)
1998
908
self.add_signature_text(revision_id, signature)
1999
909
2000
@needs_write_lock
2001
910
def add_signature_text(self, revision_id, signature):
2002
self.signatures.add_lines((revision_id,), (),
2003
osutils.split_lines(signature))
2004
2005
def find_text_key_references(self):
2006
"""Find the text key references within the repository.
2007
2008
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2009
to whether they were referred to by the inventory of the
2010
revision_id that they contain. The inventory texts from all present
2011
revision ids are assessed to generate this report.
2012
"""
2013
revision_keys = self.revisions.keys()
2014
w = self.inventories
2015
pb = ui.ui_factory.nested_progress_bar()
2016
try:
2017
return self._find_text_key_references_from_xml_inventory_lines(
2018
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
2019
finally:
2020
pb.finished()
2021
2022
def _find_text_key_references_from_xml_inventory_lines(self,
2023
line_iterator):
2024
"""Core routine for extracting references to texts from inventories.
2025
2026
This performs the translation of xml lines to revision ids.
2027
2028
:param line_iterator: An iterator of lines, origin_version_id
2029
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2030
to whether they were referred to by the inventory of the
2031
revision_id that they contain. Note that if that revision_id was
2032
not part of the line_iterator's output then False will be given -
2033
even though it may actually refer to that key.
2034
"""
2035
if not self._serializer.support_altered_by_hack:
2036
raise AssertionError(
2037
"_find_text_key_references_from_xml_inventory_lines only "
2038
"supported for branches which store inventory as unnested xml"
2039
", not on %r" % self)
2040
result = {}
2041
2042
# this code needs to read every new line in every inventory for the
2043
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
2044
# not present in one of those inventories is unnecessary but not
2045
# harmful because we are filtering by the revision id marker in the
2046
# inventory lines : we only select file ids altered in one of those
2047
# revisions. We don't need to see all lines in the inventory because
2048
# only those added in an inventory in rev X can contain a revision=X
2049
# line.
2050
unescape_revid_cache = {}
2051
unescape_fileid_cache = {}
2052
2053
# jam 20061218 In a big fetch, this handles hundreds of thousands
2054
# of lines, so it has had a lot of inlining and optimizing done.
2055
# Sorry that it is a little bit messy.
2056
# Move several functions to be local variables, since this is a long
2057
# running loop.
2058
search = self._file_ids_altered_regex.search
2059
unescape = _unescape_xml
2060
setdefault = result.setdefault
2061
for line, line_key in line_iterator:
2062
match = search(line)
2063
if match is None:
2064
continue
2065
# One call to match.group() returning multiple items is quite a
2066
# bit faster than 2 calls to match.group() each returning 1
2067
file_id, revision_id = match.group('file_id', 'revision_id')
2068
2069
# Inlining the cache lookups helps a lot when you make 170,000
2070
# lines and 350k ids, versus 8.4 unique ids.
2071
# Using a cache helps in 2 ways:
2072
# 1) Avoids unnecessary decoding calls
2073
# 2) Re-uses cached strings, which helps in future set and
2074
# equality checks.
2075
# (2) is enough that removing encoding entirely along with
2076
# the cache (so we are using plain strings) results in no
2077
# performance improvement.
2078
try:
2079
revision_id = unescape_revid_cache[revision_id]
2080
except KeyError:
2081
unescaped = unescape(revision_id)
2082
unescape_revid_cache[revision_id] = unescaped
2083
revision_id = unescaped
2084
2085
# Note that unconditionally unescaping means that we deserialise
2086
# every fileid, which for general 'pull' is not great, but we don't
2087
# really want to have some many fulltexts that this matters anyway.
2088
# RBC 20071114.
2089
try:
2090
file_id = unescape_fileid_cache[file_id]
2091
except KeyError:
2092
unescaped = unescape(file_id)
2093
unescape_fileid_cache[file_id] = unescaped
2094
file_id = unescaped
2095
2096
key = (file_id, revision_id)
2097
setdefault(key, False)
2098
if revision_id == line_key[-1]:
2099
result[key] = True
2100
return result
2101
2102
def _inventory_xml_lines_for_keys(self, keys):
2103
"""Get a line iterator of the sort needed for findind references.
2104
2105
Not relevant for non-xml inventory repositories.
2106
2107
Ghosts in revision_keys are ignored.
2108
2109
:param revision_keys: The revision keys for the inventories to inspect.
2110
:return: An iterator over (inventory line, revid) for the fulltexts of
2111
all of the xml inventories specified by revision_keys.
2112
"""
2113
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2114
for record in stream:
2115
if record.storage_kind != 'absent':
2116
chunks = record.get_bytes_as('chunked')
2117
revid = record.key[-1]
2118
lines = osutils.chunks_to_lines(chunks)
2119
for line in lines:
2120
yield line, revid
2121
2122
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2123
revision_keys):
2124
"""Helper routine for fileids_altered_by_revision_ids.
2125
2126
This performs the translation of xml lines to revision ids.
2127
2128
:param line_iterator: An iterator of lines, origin_version_id
2129
:param revision_keys: The revision ids to filter for. This should be a
2130
set or other type which supports efficient __contains__ lookups, as
2131
the revision key from each parsed line will be looked up in the
2132
revision_keys filter.
2133
:return: a dictionary mapping altered file-ids to an iterable of
2134
revision_ids. Each altered file-ids has the exact revision_ids that
2135
altered it listed explicitly.
2136
"""
2137
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2138
line_iterator).iterkeys())
2139
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2140
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2141
self._inventory_xml_lines_for_keys(parent_keys)))
2142
new_keys = seen - parent_seen
2143
result = {}
2144
setdefault = result.setdefault
2145
for key in new_keys:
2146
setdefault(key[0], set()).add(key[-1])
2147
return result
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)
2148
917
2149
918
def _find_parent_ids_of_revisions(self, revision_ids):
2150
919
"""Find all parent ids that are mentioned in the revision graph.
2159
928
parent_ids.discard(_mod_revision.NULL_REVISION)
2160
929
return parent_ids
2161
930
2162
def _find_parent_keys_of_revisions(self, revision_keys):
2163
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2164
2165
:param revision_keys: An iterable of revision_keys.
2166
:return: The parents of all revision_keys that are not already in
2167
revision_keys
2168
"""
2169
parent_map = self.revisions.get_parent_map(revision_keys)
2170
parent_keys = set()
2171
map(parent_keys.update, parent_map.itervalues())
2172
parent_keys.difference_update(revision_keys)
2173
parent_keys.discard(_mod_revision.NULL_REVISION)
2174
return parent_keys
2175
2176
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
2177
"""Find the file ids and versions affected by revisions.
2178
2179
:param revisions: an iterable containing revision ids.
2180
:param _inv_weave: The inventory weave from this repository or None.
2181
If None, the inventory weave will be opened automatically.
2182
:return: a dictionary mapping altered file-ids to an iterable of
2183
revision_ids. Each altered file-ids has the exact revision_ids that
2184
altered it listed explicitly.
2185
"""
2186
selected_keys = set((revid,) for revid in revision_ids)
2187
w = _inv_weave or self.inventories
2188
return self._find_file_ids_from_xml_inventory_lines(
2189
w.iter_lines_added_or_present_in_keys(
2190
selected_keys, pb=None),
2191
selected_keys)
2192
2193
931
def iter_files_bytes(self, desired_files):
2194
932
"""Iterate through file versions.
2195
933
2201
939
uniquely identify the file version in the caller's context. (Examples:
2202
940
an index number or a TreeTransform trans_id.)
2203
941
2204
bytes_iterator is an iterable of bytestrings for the file. The
2205
kind of iterable and length of the bytestrings are unspecified, but for
2206
this implementation, it is a list of bytes produced by
2207
VersionedFile.get_record_stream().
2208
2209
942
:param desired_files: a list of (file_id, revision_id, identifier)
2210
943
triples
2211
944
"""
2212
text_keys = {}
2213
for file_id, revision_id, callable_data in desired_files:
2214
text_keys[(file_id, revision_id)] = callable_data
2215
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2216
if record.storage_kind == 'absent':
2217
raise errors.RevisionNotPresent(record.key, self)
2218
yield text_keys[record.key], record.get_bytes_as('chunked')
2219
2220
def _generate_text_key_index(self, text_key_references=None,
2221
ancestors=None):
2222
"""Generate a new text key index for the repository.
2223
2224
This is an expensive function that will take considerable time to run.
2225
2226
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2227
list of parents, also text keys. When a given key has no parents,
2228
the parents list will be [NULL_REVISION].
2229
"""
2230
# All revisions, to find inventory parents.
2231
if ancestors is None:
2232
graph = self.get_graph()
2233
ancestors = graph.get_parent_map(self.all_revision_ids())
2234
if text_key_references is None:
2235
text_key_references = self.find_text_key_references()
2236
pb = ui.ui_factory.nested_progress_bar()
2237
try:
2238
return self._do_generate_text_key_index(ancestors,
2239
text_key_references, pb)
2240
finally:
2241
pb.finished()
2242
2243
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2244
"""Helper for _generate_text_key_index to avoid deep nesting."""
2245
revision_order = tsort.topo_sort(ancestors)
2246
invalid_keys = set()
2247
revision_keys = {}
2248
for revision_id in revision_order:
2249
revision_keys[revision_id] = set()
2250
text_count = len(text_key_references)
2251
# a cache of the text keys to allow reuse; costs a dict of all the
2252
# keys, but saves a 2-tuple for every child of a given key.
2253
text_key_cache = {}
2254
for text_key, valid in text_key_references.iteritems():
2255
if not valid:
2256
invalid_keys.add(text_key)
2257
else:
2258
revision_keys[text_key[1]].add(text_key)
2259
text_key_cache[text_key] = text_key
2260
del text_key_references
2261
text_index = {}
2262
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2263
NULL_REVISION = _mod_revision.NULL_REVISION
2264
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2265
# too small for large or very branchy trees. However, for 55K path
2266
# trees, it would be easy to use too much memory trivially. Ideally we
2267
# could gauge this by looking at available real memory etc, but this is
2268
# always a tricky proposition.
2269
inventory_cache = lru_cache.LRUCache(10)
2270
batch_size = 10 # should be ~150MB on a 55K path tree
2271
batch_count = len(revision_order) / batch_size + 1
2272
processed_texts = 0
2273
pb.update("Calculating text parents", processed_texts, text_count)
2274
for offset in xrange(batch_count):
2275
to_query = revision_order[offset * batch_size:(offset + 1) *
2276
batch_size]
2277
if not to_query:
2278
break
2279
for revision_id in to_query:
2280
parent_ids = ancestors[revision_id]
2281
for text_key in revision_keys[revision_id]:
2282
pb.update("Calculating text parents", processed_texts)
2283
processed_texts += 1
2284
candidate_parents = []
2285
for parent_id in parent_ids:
2286
parent_text_key = (text_key[0], parent_id)
2287
try:
2288
check_parent = parent_text_key not in \
2289
revision_keys[parent_id]
2290
except KeyError:
2291
# the parent parent_id is a ghost:
2292
check_parent = False
2293
# truncate the derived graph against this ghost.
2294
parent_text_key = None
2295
if check_parent:
2296
# look at the parent commit details inventories to
2297
# determine possible candidates in the per file graph.
2298
# TODO: cache here.
2299
try:
2300
inv = inventory_cache[parent_id]
2301
except KeyError:
2302
inv = self.revision_tree(parent_id).inventory
2303
inventory_cache[parent_id] = inv
2304
try:
2305
parent_entry = inv[text_key[0]]
2306
except (KeyError, errors.NoSuchId):
2307
parent_entry = None
2308
if parent_entry is not None:
2309
parent_text_key = (
2310
text_key[0], parent_entry.revision)
2311
else:
2312
parent_text_key = None
2313
if parent_text_key is not None:
2314
candidate_parents.append(
2315
text_key_cache[parent_text_key])
2316
parent_heads = text_graph.heads(candidate_parents)
2317
new_parents = list(parent_heads)
2318
new_parents.sort(key=lambda x:candidate_parents.index(x))
2319
if new_parents == []:
2320
new_parents = [NULL_REVISION]
2321
text_index[text_key] = new_parents
2322
2323
for text_key in invalid_keys:
2324
text_index[text_key] = [NULL_REVISION]
2325
return text_index
2326
2327
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2328
"""Get an iterable listing the keys of all the data introduced by a set
2329
of revision IDs.
2330
2331
The keys will be ordered so that the corresponding items can be safely
2332
fetched and inserted in that order.
2333
2334
:returns: An iterable producing tuples of (knit-kind, file-id,
2335
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2336
'revisions'. file-id is None unless knit-kind is 'file'.
2337
"""
2338
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2339
yield result
2340
del _files_pb
2341
for result in self._find_non_file_keys_to_fetch(revision_ids):
2342
yield result
2343
2344
def _find_file_keys_to_fetch(self, revision_ids, pb):
2345
# XXX: it's a bit weird to control the inventory weave caching in this
2346
# generator. Ideally the caching would be done in fetch.py I think. Or
2347
# maybe this generator should explicitly have the contract that it
2348
# should not be iterated until the previously yielded item has been
2349
# processed?
2350
inv_w = self.inventories
2351
2352
# file ids that changed
2353
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2354
count = 0
2355
num_file_ids = len(file_ids)
2356
for file_id, altered_versions in file_ids.iteritems():
2357
if pb is not None:
2358
pb.update("Fetch texts", count, num_file_ids)
2359
count += 1
2360
yield ("file", file_id, altered_versions)
2361
2362
def _find_non_file_keys_to_fetch(self, revision_ids):
2363
# inventory
2364
yield ("inventory", None, revision_ids)
2365
2366
# signatures
2367
# XXX: Note ATM no callers actually pay attention to this return
2368
# instead they just use the list of revision ids and ignore
2369
# missing sigs. Consider removing this work entirely
2370
revisions_with_signatures = set(self.signatures.get_parent_map(
2371
[(r,) for r in revision_ids]))
2372
revisions_with_signatures = set(
2373
[r for (r,) in revisions_with_signatures])
2374
revisions_with_signatures.intersection_update(revision_ids)
2375
yield ("signatures", None, revisions_with_signatures)
2376
2377
# revisions
2378
yield ("revisions", None, revision_ids)
2379
2380
@needs_read_lock
2381
def get_inventory(self, revision_id):
2382
"""Get Inventory object by revision id."""
2383
return self.iter_inventories([revision_id]).next()
2384
2385
def iter_inventories(self, revision_ids, ordering=None):
2386
"""Get many inventories by revision_ids.
2387
2388
This will buffer some or all of the texts used in constructing the
2389
inventories in memory, but will only parse a single inventory at a
2390
time.
2391
2392
:param revision_ids: The expected revision ids of the inventories.
2393
:param ordering: optional ordering, e.g. 'topological'. If not
2394
specified, the order of revision_ids will be preserved (by
2395
buffering if necessary).
2396
:return: An iterator of inventories.
2397
"""
2398
if ((None in revision_ids)
2399
or (_mod_revision.NULL_REVISION in revision_ids)):
2400
raise ValueError('cannot get null revision inventory')
2401
return self._iter_inventories(revision_ids, ordering)
2402
2403
def _iter_inventories(self, revision_ids, ordering):
2404
"""single-document based inventory iteration."""
2405
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2406
for text, revision_id in inv_xmls:
2407
yield self._deserialise_inventory(revision_id, text)
2408
2409
def _iter_inventory_xmls(self, revision_ids, ordering):
2410
if ordering is None:
2411
order_as_requested = True
2412
ordering = 'unordered'
2413
else:
2414
order_as_requested = False
2415
keys = [(revision_id,) for revision_id in revision_ids]
2416
if not keys:
2417
return
2418
if order_as_requested:
2419
key_iter = iter(keys)
2420
next_key = key_iter.next()
2421
stream = self.inventories.get_record_stream(keys, ordering, True)
2422
text_chunks = {}
2423
for record in stream:
2424
if record.storage_kind != 'absent':
2425
chunks = record.get_bytes_as('chunked')
2426
if order_as_requested:
2427
text_chunks[record.key] = chunks
2428
else:
2429
yield ''.join(chunks), record.key[-1]
2430
else:
2431
raise errors.NoSuchRevision(self, record.key)
2432
if order_as_requested:
2433
# Yield as many results as we can while preserving order.
2434
while next_key in text_chunks:
2435
chunks = text_chunks.pop(next_key)
2436
yield ''.join(chunks), next_key[-1]
2437
try:
2438
next_key = key_iter.next()
2439
except StopIteration:
2440
# We still want to fully consume the get_record_stream,
2441
# just in case it is not actually finished at this point
2442
next_key = None
2443
break
2444
2445
def _deserialise_inventory(self, revision_id, xml):
2446
"""Transform the xml into an inventory object.
2447
2448
:param revision_id: The expected revision id of the inventory.
2449
:param xml: A serialised inventory.
2450
"""
2451
result = self._serializer.read_inventory_from_string(xml, revision_id,
2452
entry_cache=self._inventory_entry_cache,
2453
return_from_cache=self._safe_to_return_from_cache)
2454
if result.revision_id != revision_id:
2455
raise AssertionError('revision id mismatch %s != %s' % (
2456
result.revision_id, revision_id))
2457
return result
2458
2459
def get_serializer_format(self):
2460
return self._serializer.format_num
2461
2462
@needs_read_lock
2463
def _get_inventory_xml(self, revision_id):
2464
"""Get serialized inventory as a string."""
2465
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2466
try:
2467
text, revision_id = texts.next()
2468
except StopIteration:
2469
raise errors.HistoryMissing(self, 'inventory', revision_id)
2470
return text
945
raise NotImplementedError(self.iter_files_bytes)
2471
946
2472
947
def get_rev_id_for_revno(self, revno, known_pair):
2473
948
"""Return the revision id of a revno, given a later (revno, revid)
2504
979
raise AssertionError('_iter_for_revno returned too much history')
2505
980
return (True, partial_history[-1])
2506
981
982
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
2507
983
def iter_reverse_revision_history(self, revision_id):
2508
984
"""Iterate backwards through revision ids in the lefthand history
2509
985
2511
987
ancestors will be traversed.
2512
988
"""
2513
989
graph = self.get_graph()
2514
next_id = revision_id
2515
while True:
2516
if next_id in (None, _mod_revision.NULL_REVISION):
2517
return
2518
try:
2519
parents = graph.get_parent_map([next_id])[next_id]
2520
except KeyError:
2521
raise errors.RevisionNotPresent(next_id, self)
2522
yield next_id
2523
if len(parents) == 0:
2524
return
2525
else:
2526
next_id = parents[0]
990
stop_revisions = (None, _mod_revision.NULL_REVISION)
991
return graph.iter_lefthand_ancestry(revision_id, stop_revisions)
2527
992
2528
993
def is_shared(self):
2529
994
"""Return True if this repository is flagged as a shared repository."""
2557
1022
2558
1023
`revision_id` may be NULL_REVISION for the empty tree revision.
2559
1024
"""
2560
revision_id = _mod_revision.ensure_null(revision_id)
2561
# TODO: refactor this to use an existing revision object
2562
# so we don't need to read it in twice.
2563
if revision_id == _mod_revision.NULL_REVISION:
2564
return RevisionTree(self, Inventory(root_id=None),
2565
_mod_revision.NULL_REVISION)
2566
else:
2567
inv = self.get_inventory(revision_id)
2568
return RevisionTree(self, inv, revision_id)
1025
raise NotImplementedError(self.revision_tree)
2569
1026
2570
1027
def revision_trees(self, revision_ids):
2571
1028
"""Return Trees for revisions in this repository.
2573
1030
:param revision_ids: a sequence of revision-ids;
2574
1031
a revision-id may not be None or 'null:'
2575
1032
"""
2576
inventories = self.iter_inventories(revision_ids)
2577
for inv in inventories:
2578
yield RevisionTree(self, inv, inv.revision_id)
2579
2580
def _filtered_revision_trees(self, revision_ids, file_ids):
2581
"""Return Tree for a revision on this branch with only some files.
2582
2583
:param revision_ids: a sequence of revision-ids;
2584
a revision-id may not be None or 'null:'
2585
:param file_ids: if not None, the result is filtered
2586
so that only those file-ids, their parents and their
2587
children are included.
2588
"""
2589
inventories = self.iter_inventories(revision_ids)
2590
for inv in inventories:
2591
# Should we introduce a FilteredRevisionTree class rather
2592
# than pre-filter the inventory here?
2593
filtered_inv = inv.filter(file_ids)
2594
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
1033
raise NotImplementedError(self.revision_trees)
2595
1034
2596
1035
@needs_read_lock
1036
@symbol_versioning.deprecated_method(
1037
symbol_versioning.deprecated_in((2, 4, 0)))
2597
1038
def get_ancestry(self, revision_id, topo_sorted=True):
2598
1039
"""Return a list of revision-ids integrated by a revision.
2599
1040
2603
1044
2604
1045
This is topologically sorted.
2605
1046
"""
1047
if 'evil' in debug.debug_flags:
1048
mutter_callsite(2, "get_ancestry is linear with history.")
2606
1049
if _mod_revision.is_null(revision_id):
2607
1050
return [None]
2608
1051
if not self.has_revision(revision_id):
2630
1073
types it should be a no-op that just returns.
2631
1074
2632
1075
This stub method does not require a lock, but subclasses should use
2633
@needs_write_lock as this is a long running call its reasonable to
1076
@needs_write_lock as this is a long running call it's reasonable to
2634
1077
implicitly lock for the user.
2635
1078
2636
1079
:param hint: If not supplied, the whole repository is packed.
2649
1092
2650
1093
def get_parent_map(self, revision_ids):
2651
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."""
2652
1099
# revisions index works in keys; this just works in revisions
2653
1100
# therefore wrap and unwrap
2654
1101
query_keys = []
2660
1107
raise ValueError('get_parent_map(None) is not valid')
2661
1108
else:
2662
1109
query_keys.append((revision_id ,))
1110
vf = self.revisions.without_fallbacks()
2663
1111
for ((revision_id,), parent_keys) in \
2664
self.revisions.get_parent_map(query_keys).iteritems():
1112
vf.get_parent_map(query_keys).iteritems():
2665
1113
if parent_keys:
2666
1114
result[revision_id] = tuple([parent_revid
2667
1115
for (parent_revid,) in parent_keys])
2670
1118
return result
2671
1119
2672
1120
def _make_parents_provider(self):
2673
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)
2674
1130
2675
1131
@needs_read_lock
2676
1132
def get_known_graph_ancestry(self, revision_ids):
2677
1133
"""Return the known graph for a set of revision ids and their ancestors.
2678
1134
"""
2679
st = static_tuple.StaticTuple
2680
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2681
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2682
return graph.GraphThunkIdsToKeys(known_graph)
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)
2683
1140
2684
1141
def get_graph(self, other_repository=None):
2685
1142
"""Return the graph walker for this repository format"""
2690
1147
[parents_provider, other_repository._make_parents_provider()])
2691
1148
return graph.Graph(parents_provider)
2692
1149
2693
def _get_versioned_file_checker(self, text_key_references=None,
2694
ancestors=None):
2695
"""Return an object suitable for checking versioned files.
2696
2697
:param text_key_references: if non-None, an already built
2698
dictionary mapping text keys ((fileid, revision_id) tuples)
2699
to whether they were referred to by the inventory of the
2700
revision_id that they contain. If None, this will be
2701
calculated.
2702
:param ancestors: Optional result from
2703
self.get_graph().get_parent_map(self.all_revision_ids()) if already
2704
available.
2705
"""
2706
return _VersionedFileChecker(self,
2707
text_key_references=text_key_references, ancestors=ancestors)
2708
2709
def revision_ids_to_search_result(self, result_set):
2710
"""Convert a set of revision ids to a graph SearchResult."""
2711
result_parents = set()
2712
for parents in self.get_graph().get_parent_map(
2713
result_set).itervalues():
2714
result_parents.update(parents)
2715
included_keys = result_set.intersection(result_parents)
2716
start_keys = result_set.difference(included_keys)
2717
exclude_keys = result_parents.difference(result_set)
2718
result = graph.SearchResult(start_keys, exclude_keys,
2719
len(result_set), result_set)
2720
return result
2721
2722
1150
@needs_write_lock
2723
1151
def set_make_working_trees(self, new_value):
2724
1152
"""Set the policy flag for making working trees when creating branches.
2737
1165
2738
1166
@needs_write_lock
2739
1167
def sign_revision(self, revision_id, gpg_strategy):
2740
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()
2741
1170
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2742
1171
2743
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
2744
1190
def has_signature_for_revision_id(self, revision_id):
2745
1191
"""Query for a revision signature for revision_id in the repository."""
2746
if not self.has_revision(revision_id):
2747
raise errors.NoSuchRevision(self, revision_id)
2748
sig_present = (1 == len(
2749
self.signatures.get_parent_map([(revision_id,)])))
2750
return sig_present
1192
raise NotImplementedError(self.has_signature_for_revision_id)
2751
1193
2752
@needs_read_lock
2753
1194
def get_signature_text(self, revision_id):
2754
1195
"""Return the text for a signature."""
2755
stream = self.signatures.get_record_stream([(revision_id,)],
2756
'unordered', True)
2757
record = stream.next()
2758
if record.storage_kind == 'absent':
2759
raise errors.NoSuchRevision(self, revision_id)
2760
return record.get_bytes_as('fulltext')
1196
raise NotImplementedError(self.get_signature_text)
2761
1197
2762
@needs_read_lock
2763
1198
def check(self, revision_ids=None, callback_refs=None, check_repo=True):
2764
1199
"""Check consistency of all history of given revision_ids.
2765
1200
2773
1208
:param check_repo: If False do not check the repository contents, just
2774
1209
calculate the data callback_refs requires and call them back.
2775
1210
"""
2776
return self._check(revision_ids, callback_refs=callback_refs,
1211
return self._check(revision_ids=revision_ids, callback_refs=callback_refs,
2777
1212
check_repo=check_repo)
2778
1213
2779
def _check(self, revision_ids, callback_refs, check_repo):
2780
result = check.Check(self, check_repo=check_repo)
2781
result.check(callback_refs)
2782
return result
1214
def _check(self, revision_ids=None, callback_refs=None, check_repo=True):
1215
raise NotImplementedError(self.check)
2783
1216
2784
1217
def _warn_if_deprecated(self, branch=None):
1218
if not self._format.is_deprecated():
1219
return
2785
1220
global _deprecation_warning_done
2786
1221
if _deprecation_warning_done:
2787
1222
return
2817
1252
except UnicodeDecodeError:
2818
1253
raise errors.NonAsciiRevisionId(method, self)
2819
1254
2820
def revision_graph_can_have_wrong_parents(self):
2821
"""Is it possible for this repository to have a revision graph with
2822
incorrect parents?
2823
2824
If True, then this repository must also implement
2825
_find_inconsistent_revision_parents so that check and reconcile can
2826
check for inconsistencies before proceeding with other checks that may
2827
depend on the revision index being consistent.
2828
"""
2829
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2830
2831
2832
# remove these delegates a while after bzr 0.15
2833
def __make_delegated(name, from_module):
2834
def _deprecated_repository_forwarder():
2835
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2836
% (name, from_module),
2837
DeprecationWarning,
2838
stacklevel=2)
2839
m = __import__(from_module, globals(), locals(), [name])
2840
try:
2841
return getattr(m, name)
2842
except AttributeError:
2843
raise AttributeError('module %s has no name %s'
2844
% (m, name))
2845
globals()[name] = _deprecated_repository_forwarder
2846
2847
for _name in [
2848
'AllInOneRepository',
2849
'WeaveMetaDirRepository',
2850
'PreSplitOutRepositoryFormat',
2851
'RepositoryFormat4',
2852
'RepositoryFormat5',
2853
'RepositoryFormat6',
2854
'RepositoryFormat7',
2855
]:
2856
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2857
2858
for _name in [
2859
'KnitRepository',
2860
'RepositoryFormatKnit',
2861
'RepositoryFormatKnit1',
2862
]:
2863
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2864
2865
2866
def install_revision(repository, rev, revision_tree):
2867
"""Install all revision data into a repository."""
2868
install_revisions(repository, [(rev, revision_tree, None)])
2869
2870
2871
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2872
"""Install all revision data into a repository.
2873
2874
Accepts an iterable of revision, tree, signature tuples. The signature
2875
may be None.
2876
"""
2877
repository.start_write_group()
2878
try:
2879
inventory_cache = lru_cache.LRUCache(10)
2880
for n, (revision, revision_tree, signature) in enumerate(iterable):
2881
_install_revision(repository, revision, revision_tree, signature,
2882
inventory_cache)
2883
if pb is not None:
2884
pb.update('Transferring revisions', n + 1, num_revisions)
2885
except:
2886
repository.abort_write_group()
2887
raise
2888
else:
2889
repository.commit_write_group()
2890
2891
2892
def _install_revision(repository, rev, revision_tree, signature,
2893
inventory_cache):
2894
"""Install all revision data into a repository."""
2895
present_parents = []
2896
parent_trees = {}
2897
for p_id in rev.parent_ids:
2898
if repository.has_revision(p_id):
2899
present_parents.append(p_id)
2900
parent_trees[p_id] = repository.revision_tree(p_id)
2901
else:
2902
parent_trees[p_id] = repository.revision_tree(
2903
_mod_revision.NULL_REVISION)
2904
2905
inv = revision_tree.inventory
2906
entries = inv.iter_entries()
2907
# backwards compatibility hack: skip the root id.
2908
if not repository.supports_rich_root():
2909
path, root = entries.next()
2910
if root.revision != rev.revision_id:
2911
raise errors.IncompatibleRevision(repr(repository))
2912
text_keys = {}
2913
for path, ie in entries:
2914
text_keys[(ie.file_id, ie.revision)] = ie
2915
text_parent_map = repository.texts.get_parent_map(text_keys)
2916
missing_texts = set(text_keys) - set(text_parent_map)
2917
# Add the texts that are not already present
2918
for text_key in missing_texts:
2919
ie = text_keys[text_key]
2920
text_parents = []
2921
# FIXME: TODO: The following loop overlaps/duplicates that done by
2922
# commit to determine parents. There is a latent/real bug here where
2923
# the parents inserted are not those commit would do - in particular
2924
# they are not filtered by heads(). RBC, AB
2925
for revision, tree in parent_trees.iteritems():
2926
if ie.file_id not in tree:
2927
continue
2928
parent_id = tree.inventory[ie.file_id].revision
2929
if parent_id in text_parents:
2930
continue
2931
text_parents.append((ie.file_id, parent_id))
2932
lines = revision_tree.get_file(ie.file_id).readlines()
2933
repository.texts.add_lines(text_key, text_parents, lines)
2934
try:
2935
# install the inventory
2936
if repository._format._commit_inv_deltas and len(rev.parent_ids):
2937
# Cache this inventory
2938
inventory_cache[rev.revision_id] = inv
2939
try:
2940
basis_inv = inventory_cache[rev.parent_ids[0]]
2941
except KeyError:
2942
repository.add_inventory(rev.revision_id, inv, present_parents)
2943
else:
2944
delta = inv._make_delta(basis_inv)
2945
repository.add_inventory_by_delta(rev.parent_ids[0], delta,
2946
rev.revision_id, present_parents)
2947
else:
2948
repository.add_inventory(rev.revision_id, inv, present_parents)
2949
except errors.RevisionAlreadyPresent:
2950
pass
2951
if signature is not None:
2952
repository.add_signature_text(rev.revision_id, signature)
2953
repository.add_revision(rev.revision_id, rev, inv)
2954
2955
1255
2956
1256
class MetaDirRepository(Repository):
2957
1257
"""Repositories in the new meta-dir layout.
2992
1292
return not self._transport.has('no-working-trees')
2993
1293
2994
1294
2995
class MetaDirVersionedFileRepository(MetaDirRepository):
2996
"""Repositories in a meta-dir, that work via versioned file objects."""
1295
class RepositoryFormatRegistry(controldir.ControlComponentFormatRegistry):
1296
"""Repository format registry."""
2997
1297
2998
def __init__(self, _format, a_bzrdir, control_files):
2999
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
3000
control_files)
1298
def get_default(self):
1299
"""Return the current default format."""
1300
return controldir.format_registry.make_bzrdir('default').repository_format
3001
1301
3002
1302
3003
1303
network_format_registry = registry.FormatRegistry()
3009
1309
"""
3010
1310
3011
1311
3012
format_registry = registry.FormatRegistry(network_format_registry)
1312
format_registry = RepositoryFormatRegistry(network_format_registry)
3013
1313
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
3014
1314
3015
1315
This can contain either format instances themselves, or classes/factories that
3020
1320
#####################################################################
3021
1321
# Repository Formats
3022
1322
3023
class RepositoryFormat(object):
1323
class RepositoryFormat(controldir.ControlComponentFormat):
3024
1324
"""A repository format.
3025
1325
3026
1326
Formats provide four things:
3046
1346
created.
3047
1347
3048
1348
Common instance attributes:
3049
_matchingbzrdir - the bzrdir format that the repository format was
1349
_matchingbzrdir - the controldir format that the repository format was
3050
1350
originally written to work with. This can be used if manually
3051
1351
constructing a bzrdir and repository, or more commonly for test suite
3052
1352
parameterization.
3061
1361
# Does this format support CHK bytestring lookups. Set to True or False in
3062
1362
# derived classes.
3063
1363
supports_chks = None
3064
# Should commit add an inventory, or an inventory delta to the repository.
3065
_commit_inv_deltas = True
3066
# What order should fetch operations request streams in?
3067
# The default is unordered as that is the cheapest for an origin to
3068
# provide.
3069
_fetch_order = 'unordered'
3070
# Does this repository format use deltas that can be fetched as-deltas ?
3071
# (E.g. knits, where the knit deltas can be transplanted intact.
3072
# We default to False, which will ensure that enough data to get
3073
# a full text out of any fetch stream will be grabbed.
3074
_fetch_uses_deltas = False
3075
1364
# Should fetch trigger a reconcile after the fetch? Only needed for
3076
1365
# some repository formats that can suffer internal inconsistencies.
3077
1366
_fetch_reconcile = False
3083
1372
# help), and for fetching when data won't have come from the same
3084
1373
# compressor.
3085
1374
pack_compresses = False
3086
# Does the repository inventory storage understand references to trees?
1375
# Does the repository storage understand references to trees?
3087
1376
supports_tree_reference = None
3088
1377
# Is the format experimental ?
3089
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
3090
1400
3091
1401
def __repr__(self):
3092
1402
return "%s()" % self.__class__.__name__
3099
1409
return not self == other
3100
1410
3101
1411
@classmethod
3102
def find_format(klass, a_bzrdir):
3103
"""Return the format for the repository object in a_bzrdir.
3104
3105
This is used by bzr native formats that have a "format" file in
3106
the repository. Other methods may be used by different types of
3107
control directory.
3108
"""
3109
try:
3110
transport = a_bzrdir.get_repository_transport(None)
3111
format_string = transport.get_bytes("format")
3112
return format_registry.get(format_string)
3113
except errors.NoSuchFile:
3114
raise errors.NoRepositoryPresent(a_bzrdir)
3115
except KeyError:
3116
raise errors.UnknownFormatError(format=format_string,
3117
kind='repository')
3118
3119
@classmethod
1412
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3120
1413
def register_format(klass, format):
3121
format_registry.register(format.get_format_string(), format)
1414
format_registry.register(format)
3122
1415
3123
1416
@classmethod
1417
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3124
1418
def unregister_format(klass, format):
3125
format_registry.remove(format.get_format_string())
1419
format_registry.remove(format)
3126
1420
3127
1421
@classmethod
1422
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3128
1423
def get_default_format(klass):
3129
1424
"""Return the current default format."""
3130
from bzrlib import bzrdir
3131
return bzrdir.format_registry.make_bzrdir('default').repository_format
3132
3133
def get_format_string(self):
3134
"""Return the ASCII format string that identifies this format.
3135
3136
Note that in pre format ?? repositories the format string is
3137
not permitted nor written to disk.
3138
"""
3139
raise NotImplementedError(self.get_format_string)
1425
return format_registry.get_default()
3140
1426
3141
1427
def get_format_description(self):
3142
1428
"""Return the short description for this format."""
3143
1429
raise NotImplementedError(self.get_format_description)
3144
1430
3145
# TODO: this shouldn't be in the base class, it's specific to things that
3146
# use weaves or knits -- mbp 20070207
3147
def _get_versioned_file_store(self,
3148
name,
3149
transport,
3150
control_files,
3151
prefixed=True,
3152
versionedfile_class=None,
3153
versionedfile_kwargs={},
3154
escaped=False):
3155
if versionedfile_class is None:
3156
versionedfile_class = self._versionedfile_class
3157
weave_transport = control_files._transport.clone(name)
3158
dir_mode = control_files._dir_mode
3159
file_mode = control_files._file_mode
3160
return VersionedFileStore(weave_transport, prefixed=prefixed,
3161
dir_mode=dir_mode,
3162
file_mode=file_mode,
3163
versionedfile_class=versionedfile_class,
3164
versionedfile_kwargs=versionedfile_kwargs,
3165
escaped=escaped)
3166
3167
def initialize(self, a_bzrdir, shared=False):
3168
"""Initialize a repository of this format in a_bzrdir.
3169
3170
: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.
3171
1435
:param shared: The repository should be initialized as a sharable one.
3172
1436
:returns: The new repository object.
3173
1437
3174
1438
This may raise UninitializableFormat if shared repository are not
3175
compatible the a_bzrdir.
1439
compatible the controldir.
3176
1440
"""
3177
1441
raise NotImplementedError(self.initialize)
3178
1442
3185
1449
"""
3186
1450
return True
3187
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
3188
1460
def network_name(self):
3189
1461
"""A simple byte string uniquely identifying this format for RPC calls.
3190
1462
3206
1478
'Does not support nested trees', target_format,
3207
1479
from_format=self)
3208
1480
3209
def open(self, a_bzrdir, _found=False):
3210
"""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.
3211
1483
3212
1484
_found is a private parameter, do not use it.
3213
1485
"""
3214
1486
raise NotImplementedError(self.open)
3215
1487
3216
def _run_post_repo_init_hooks(self, repository, a_bzrdir, shared):
3217
from bzrlib.bzrdir import BzrDir, RepoInitHookParams
3218
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']
3219
1491
if not hooks:
3220
1492
return
3221
params = RepoInitHookParams(repository, self, a_bzrdir, shared)
1493
params = RepoInitHookParams(repository, self, controldir, shared)
3222
1494
for hook in hooks:
3223
1495
hook(params)
3224
1496
3225
1497
3226
class MetaDirRepositoryFormat(RepositoryFormat):
1498
class RepositoryFormatMetaDir(bzrdir.BzrDirMetaComponentFormat, RepositoryFormat):
3227
1499
"""Common base class for the new repositories using the metadir layout."""
3228
1500
3229
1501
rich_root_data = False
3230
1502
supports_tree_reference = False
3231
1503
supports_external_lookups = False
1504
supports_leaving_lock = True
1505
supports_nesting_repositories = True
3232
1506
3233
1507
@property
3234
1508
def _matchingbzrdir(self):
3237
1511
return matching
3238
1512
3239
1513
def __init__(self):
3240
super(MetaDirRepositoryFormat, self).__init__()
1514
RepositoryFormat.__init__(self)
1515
bzrdir.BzrDirMetaComponentFormat.__init__(self)
3241
1516
3242
1517
def _create_control_files(self, a_bzrdir):
3243
1518
"""Create the required files and the initial control_files object."""
3267
1542
finally:
3268
1543
control_files.unlock()
3269
1544
3270
def network_name(self):
3271
"""Metadir formats have matching disk and network format strings."""
3272
return self.get_format_string()
3273
3274
3275
# Pre-0.8 formats that don't have a disk format string (because they are
3276
# versioned by the matching control directory). We use the control directories
3277
# disk format string as a key for the network_name because they meet the
3278
# constraints (simple string, unique, immutable).
3279
network_format_registry.register_lazy(
3280
"Bazaar-NG branch, format 5\n",
3281
'bzrlib.repofmt.weaverepo',
3282
'RepositoryFormat5',
3283
)
3284
network_format_registry.register_lazy(
3285
"Bazaar-NG branch, format 6\n",
3286
'bzrlib.repofmt.weaverepo',
3287
'RepositoryFormat6',
3288
)
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
3289
1560
3290
1561
# formats which have no format string are not discoverable or independently
3291
1562
# creatable on disk, so are not registered in format_registry. They're
3292
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
3293
# 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
3294
1565
# the repository is not separately opened are similar.
3295
1566
3296
1567
format_registry.register_lazy(
3297
'Bazaar-NG Repository format 7',
3298
'bzrlib.repofmt.weaverepo',
3299
'RepositoryFormat7'
3300
)
3301
3302
format_registry.register_lazy(
3303
1568
'Bazaar-NG Knit Repository Format 1',
3304
1569
'bzrlib.repofmt.knitrepo',
3305
1570
'RepositoryFormatKnit1',
3322
1587
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
3323
1588
format_registry.register_lazy(
3324
1589
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
3325
'bzrlib.repofmt.pack_repo',
1590
'bzrlib.repofmt.knitpack_repo',
3326
1591
'RepositoryFormatKnitPack1',
3327
1592
)
3328
1593
format_registry.register_lazy(
3329
1594
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
3330
'bzrlib.repofmt.pack_repo',
1595
'bzrlib.repofmt.knitpack_repo',
3331
1596
'RepositoryFormatKnitPack3',
3332
1597
)
3333
1598
format_registry.register_lazy(
3334
1599
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
3335
'bzrlib.repofmt.pack_repo',
1600
'bzrlib.repofmt.knitpack_repo',
3336
1601
'RepositoryFormatKnitPack4',
3337
1602
)
3338
1603
format_registry.register_lazy(
3339
1604
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
3340
'bzrlib.repofmt.pack_repo',
1605
'bzrlib.repofmt.knitpack_repo',
3341
1606
'RepositoryFormatKnitPack5',
3342
1607
)
3343
1608
format_registry.register_lazy(
3344
1609
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
3345
'bzrlib.repofmt.pack_repo',
1610
'bzrlib.repofmt.knitpack_repo',
3346
1611
'RepositoryFormatKnitPack5RichRoot',
3347
1612
)
3348
1613
format_registry.register_lazy(
3349
1614
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
3350
'bzrlib.repofmt.pack_repo',
1615
'bzrlib.repofmt.knitpack_repo',
3351
1616
'RepositoryFormatKnitPack5RichRootBroken',
3352
1617
)
3353
1618
format_registry.register_lazy(
3354
1619
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
3355
'bzrlib.repofmt.pack_repo',
1620
'bzrlib.repofmt.knitpack_repo',
3356
1621
'RepositoryFormatKnitPack6',
3357
1622
)
3358
1623
format_registry.register_lazy(
3359
1624
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
3360
'bzrlib.repofmt.pack_repo',
1625
'bzrlib.repofmt.knitpack_repo',
3361
1626
'RepositoryFormatKnitPack6RichRoot',
3362
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
)
3363
1633
3364
1634
# Development formats.
3365
# Obsolete but kept pending a CHK based subtree format.
1635
# Check their docstrings to see if/when they are obsolete.
3366
1636
format_registry.register_lazy(
3367
1637
("Bazaar development format 2 with subtree support "
3368
1638
"(needs bzr.dev from before 1.8)\n"),
3369
'bzrlib.repofmt.pack_repo',
1639
'bzrlib.repofmt.knitpack_repo',
3370
1640
'RepositoryFormatPackDevelopment2Subtree',
3371
1641
)
3372
3373
# 1.14->1.16 go below here
3374
format_registry.register_lazy(
3375
'Bazaar development format - group compression and chk inventory'
3376
' (needs bzr.dev from 1.14)\n',
3377
'bzrlib.repofmt.groupcompress_repo',
3378
'RepositoryFormatCHK1',
3379
)
3380
3381
format_registry.register_lazy(
3382
'Bazaar development format - chk repository with bencode revision '
3383
'serialization (needs bzr.dev from 1.16)\n',
3384
'bzrlib.repofmt.groupcompress_repo',
3385
'RepositoryFormatCHK2',
3386
)
3387
format_registry.register_lazy(
3388
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3389
'bzrlib.repofmt.groupcompress_repo',
3390
'RepositoryFormat2a',
3391
)
3392
1642
format_registry.register_lazy(
3393
1643
'Bazaar development format 8\n',
3394
1644
'bzrlib.repofmt.groupcompress_repo',
3408
1658
InterRepository.get(other).method_name(parameters).
3409
1659
"""
3410
1660
3411
_walk_to_common_revisions_batch_size = 50
3412
1661
_optimisers = []
3413
1662
"""The available optimised InterRepository types."""
3414
1663
3429
1678
self.target.fetch(self.source, revision_id=revision_id)
3430
1679
3431
1680
@needs_write_lock
3432
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3433
fetch_spec=None):
1681
def fetch(self, revision_id=None, find_ghosts=False):
3434
1682
"""Fetch the content required to construct revision_id.
3435
1683
3436
1684
The content is copied from self.source to self.target.
3437
1685
3438
1686
:param revision_id: if None all content is copied, if NULL_REVISION no
3439
1687
content is copied.
3440
:param pb: ignored.
3441
1688
:return: None.
3442
1689
"""
3443
ui.ui_factory.warn_experimental_format_fetch(self)
3444
from bzrlib.fetch import RepoFetcher
3445
# See <https://launchpad.net/bugs/456077> asking for a warning here
3446
if self.source._format.network_name() != self.target._format.network_name():
3447
ui.ui_factory.show_user_warning('cross_format_fetch',
3448
from_format=self.source._format,
3449
to_format=self.target._format)
3450
f = RepoFetcher(to_repository=self.target,
3451
from_repository=self.source,
3452
last_revision=revision_id,
3453
fetch_spec=fetch_spec,
3454
find_ghosts=find_ghosts)
3455
3456
def _walk_to_common_revisions(self, revision_ids):
3457
"""Walk out from revision_ids in source to revisions target has.
3458
3459
:param revision_ids: The start point for the search.
3460
:return: A set of revision ids.
3461
"""
3462
target_graph = self.target.get_graph()
3463
revision_ids = frozenset(revision_ids)
3464
missing_revs = set()
3465
source_graph = self.source.get_graph()
3466
# ensure we don't pay silly lookup costs.
3467
searcher = source_graph._make_breadth_first_searcher(revision_ids)
3468
null_set = frozenset([_mod_revision.NULL_REVISION])
3469
searcher_exhausted = False
3470
while True:
3471
next_revs = set()
3472
ghosts = set()
3473
# Iterate the searcher until we have enough next_revs
3474
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3475
try:
3476
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3477
next_revs.update(next_revs_part)
3478
ghosts.update(ghosts_part)
3479
except StopIteration:
3480
searcher_exhausted = True
3481
break
3482
# If there are ghosts in the source graph, and the caller asked for
3483
# them, make sure that they are present in the target.
3484
# We don't care about other ghosts as we can't fetch them and
3485
# haven't been asked to.
3486
ghosts_to_check = set(revision_ids.intersection(ghosts))
3487
revs_to_get = set(next_revs).union(ghosts_to_check)
3488
if revs_to_get:
3489
have_revs = set(target_graph.get_parent_map(revs_to_get))
3490
# we always have NULL_REVISION present.
3491
have_revs = have_revs.union(null_set)
3492
# Check if the target is missing any ghosts we need.
3493
ghosts_to_check.difference_update(have_revs)
3494
if ghosts_to_check:
3495
# One of the caller's revision_ids is a ghost in both the
3496
# source and the target.
3497
raise errors.NoSuchRevision(
3498
self.source, ghosts_to_check.pop())
3499
missing_revs.update(next_revs - have_revs)
3500
# Because we may have walked past the original stop point, make
3501
# sure everything is stopped
3502
stop_revs = searcher.find_seen_ancestors(have_revs)
3503
searcher.stop_searching_any(stop_revs)
3504
if searcher_exhausted:
3505
break
3506
return searcher.get_result()
1690
raise NotImplementedError(self.fetch)
3507
1691
3508
1692
@needs_read_lock
3509
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):
3510
1697
"""Return the revision ids that source has that target does not.
3511
1698
3512
1699
:param revision_id: only return revision ids included by this
3513
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.
3514
1708
:param find_ghosts: If True find missing revisions in deep history
3515
1709
rather than just finding the surface difference.
1710
:param limit: Maximum number of revisions to return, topologically
1711
ordered
3516
1712
:return: A bzrlib.graph.SearchResult.
3517
1713
"""
3518
# stop searching at found target revisions.
3519
if not find_ghosts and revision_id is not None:
3520
return self._walk_to_common_revisions([revision_id])
3521
# generic, possibly worst case, slow code path.
3522
target_ids = set(self.target.all_revision_ids())
3523
if revision_id is not None:
3524
source_ids = self.source.get_ancestry(revision_id)
3525
if source_ids[0] is not None:
3526
raise AssertionError()
3527
source_ids.pop(0)
3528
else:
3529
source_ids = self.source.all_revision_ids()
3530
result_set = set(source_ids).difference(target_ids)
3531
return self.source.revision_ids_to_search_result(result_set)
1714
raise NotImplementedError(self.search_missing_revision_ids)
3532
1715
3533
1716
@staticmethod
3534
1717
def _same_model(source, target):
3555
1738
"different serializers")
3556
1739
3557
1740
3558
class InterSameDataRepository(InterRepository):
3559
"""Code for converting between repositories that represent the same data.
3560
3561
Data format and model must match for this to work.
3562
"""
3563
3564
@classmethod
3565
def _get_repo_format_to_test(self):
3566
"""Repository format for testing with.
3567
3568
InterSameData can pull from subtree to subtree and from non-subtree to
3569
non-subtree, so we test this with the richest repository format.
3570
"""
3571
from bzrlib.repofmt import knitrepo
3572
return knitrepo.RepositoryFormatKnit3()
3573
3574
@staticmethod
3575
def is_compatible(source, target):
3576
return InterRepository._same_model(source, target)
3577
3578
3579
class InterWeaveRepo(InterSameDataRepository):
3580
"""Optimised code paths between Weave based repositories.
3581
3582
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
3583
implemented lazy inter-object optimisation.
3584
"""
3585
3586
@classmethod
3587
def _get_repo_format_to_test(self):
3588
from bzrlib.repofmt import weaverepo
3589
return weaverepo.RepositoryFormat7()
3590
3591
@staticmethod
3592
def is_compatible(source, target):
3593
"""Be compatible with known Weave formats.
3594
3595
We don't test for the stores being of specific types because that
3596
could lead to confusing results, and there is no need to be
3597
overly general.
3598
"""
3599
from bzrlib.repofmt.weaverepo import (
3600
RepositoryFormat5,
3601
RepositoryFormat6,
3602
RepositoryFormat7,
3603
)
3604
try:
3605
return (isinstance(source._format, (RepositoryFormat5,
3606
RepositoryFormat6,
3607
RepositoryFormat7)) and
3608
isinstance(target._format, (RepositoryFormat5,
3609
RepositoryFormat6,
3610
RepositoryFormat7)))
3611
except AttributeError:
3612
return False
3613
3614
@needs_write_lock
3615
def copy_content(self, revision_id=None):
3616
"""See InterRepository.copy_content()."""
3617
# weave specific optimised path:
3618
try:
3619
self.target.set_make_working_trees(self.source.make_working_trees())
3620
except (errors.RepositoryUpgradeRequired, NotImplemented):
3621
pass
3622
# FIXME do not peek!
3623
if self.source._transport.listable():
3624
pb = ui.ui_factory.nested_progress_bar()
3625
try:
3626
self.target.texts.insert_record_stream(
3627
self.source.texts.get_record_stream(
3628
self.source.texts.keys(), 'topological', False))
3629
pb.update('Copying inventory', 0, 1)
3630
self.target.inventories.insert_record_stream(
3631
self.source.inventories.get_record_stream(
3632
self.source.inventories.keys(), 'topological', False))
3633
self.target.signatures.insert_record_stream(
3634
self.source.signatures.get_record_stream(
3635
self.source.signatures.keys(),
3636
'unordered', True))
3637
self.target.revisions.insert_record_stream(
3638
self.source.revisions.get_record_stream(
3639
self.source.revisions.keys(),
3640
'topological', True))
3641
finally:
3642
pb.finished()
3643
else:
3644
self.target.fetch(self.source, revision_id=revision_id)
3645
3646
@needs_read_lock
3647
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3648
"""See InterRepository.missing_revision_ids()."""
3649
# we want all revisions to satisfy revision_id in source.
3650
# but we don't want to stat every file here and there.
3651
# we want then, all revisions other needs to satisfy revision_id
3652
# checked, but not those that we have locally.
3653
# so the first thing is to get a subset of the revisions to
3654
# satisfy revision_id in source, and then eliminate those that
3655
# we do already have.
3656
# this is slow on high latency connection to self, but as this
3657
# disk format scales terribly for push anyway due to rewriting
3658
# inventory.weave, this is considered acceptable.
3659
# - RBC 20060209
3660
if revision_id is not None:
3661
source_ids = self.source.get_ancestry(revision_id)
3662
if source_ids[0] is not None:
3663
raise AssertionError()
3664
source_ids.pop(0)
3665
else:
3666
source_ids = self.source._all_possible_ids()
3667
source_ids_set = set(source_ids)
3668
# source_ids is the worst possible case we may need to pull.
3669
# now we want to filter source_ids against what we actually
3670
# have in target, but don't try to check for existence where we know
3671
# we do not have a revision as that would be pointless.
3672
target_ids = set(self.target._all_possible_ids())
3673
possibly_present_revisions = target_ids.intersection(source_ids_set)
3674
actually_present_revisions = set(
3675
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3676
required_revisions = source_ids_set.difference(actually_present_revisions)
3677
if revision_id is not None:
3678
# we used get_ancestry to determine source_ids then we are assured all
3679
# revisions referenced are present as they are installed in topological order.
3680
# and the tip revision was validated by get_ancestry.
3681
result_set = required_revisions
3682
else:
3683
# if we just grabbed the possibly available ids, then
3684
# we only have an estimate of whats available and need to validate
3685
# that against the revision records.
3686
result_set = set(
3687
self.source._eliminate_revisions_not_present(required_revisions))
3688
return self.source.revision_ids_to_search_result(result_set)
3689
3690
3691
class InterKnitRepo(InterSameDataRepository):
3692
"""Optimised code paths between Knit based repositories."""
3693
3694
@classmethod
3695
def _get_repo_format_to_test(self):
3696
from bzrlib.repofmt import knitrepo
3697
return knitrepo.RepositoryFormatKnit1()
3698
3699
@staticmethod
3700
def is_compatible(source, target):
3701
"""Be compatible with known Knit formats.
3702
3703
We don't test for the stores being of specific types because that
3704
could lead to confusing results, and there is no need to be
3705
overly general.
3706
"""
3707
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
3708
try:
3709
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
3710
isinstance(target._format, RepositoryFormatKnit))
3711
except AttributeError:
3712
return False
3713
return are_knits and InterRepository._same_model(source, target)
3714
3715
@needs_read_lock
3716
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3717
"""See InterRepository.missing_revision_ids()."""
3718
if revision_id is not None:
3719
source_ids = self.source.get_ancestry(revision_id)
3720
if source_ids[0] is not None:
3721
raise AssertionError()
3722
source_ids.pop(0)
3723
else:
3724
source_ids = self.source.all_revision_ids()
3725
source_ids_set = set(source_ids)
3726
# source_ids is the worst possible case we may need to pull.
3727
# now we want to filter source_ids against what we actually
3728
# have in target, but don't try to check for existence where we know
3729
# we do not have a revision as that would be pointless.
3730
target_ids = set(self.target.all_revision_ids())
3731
possibly_present_revisions = target_ids.intersection(source_ids_set)
3732
actually_present_revisions = set(
3733
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3734
required_revisions = source_ids_set.difference(actually_present_revisions)
3735
if revision_id is not None:
3736
# we used get_ancestry to determine source_ids then we are assured all
3737
# revisions referenced are present as they are installed in topological order.
3738
# and the tip revision was validated by get_ancestry.
3739
result_set = required_revisions
3740
else:
3741
# if we just grabbed the possibly available ids, then
3742
# we only have an estimate of whats available and need to validate
3743
# that against the revision records.
3744
result_set = set(
3745
self.source._eliminate_revisions_not_present(required_revisions))
3746
return self.source.revision_ids_to_search_result(result_set)
3747
3748
3749
class InterDifferingSerializer(InterRepository):
3750
3751
@classmethod
3752
def _get_repo_format_to_test(self):
3753
return None
3754
3755
@staticmethod
3756
def is_compatible(source, target):
3757
"""Be compatible with Knit2 source and Knit3 target"""
3758
# This is redundant with format.check_conversion_target(), however that
3759
# raises an exception, and we just want to say "False" as in we won't
3760
# support converting between these formats.
3761
if 'IDS_never' in debug.debug_flags:
3762
return False
3763
if source.supports_rich_root() and not target.supports_rich_root():
3764
return False
3765
if (source._format.supports_tree_reference
3766
and not target._format.supports_tree_reference):
3767
return False
3768
if target._fallback_repositories and target._format.supports_chks:
3769
# IDS doesn't know how to copy CHKs for the parent inventories it
3770
# adds to stacked repos.
3771
return False
3772
if 'IDS_always' in debug.debug_flags:
3773
return True
3774
# Only use this code path for local source and target. IDS does far
3775
# too much IO (both bandwidth and roundtrips) over a network.
3776
if not source.bzrdir.transport.base.startswith('file:///'):
3777
return False
3778
if not target.bzrdir.transport.base.startswith('file:///'):
3779
return False
3780
return True
3781
3782
def _get_trees(self, revision_ids, cache):
3783
possible_trees = []
3784
for rev_id in revision_ids:
3785
if rev_id in cache:
3786
possible_trees.append((rev_id, cache[rev_id]))
3787
else:
3788
# Not cached, but inventory might be present anyway.
3789
try:
3790
tree = self.source.revision_tree(rev_id)
3791
except errors.NoSuchRevision:
3792
# Nope, parent is ghost.
3793
pass
3794
else:
3795
cache[rev_id] = tree
3796
possible_trees.append((rev_id, tree))
3797
return possible_trees
3798
3799
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3800
"""Get the best delta and base for this revision.
3801
3802
:return: (basis_id, delta)
3803
"""
3804
deltas = []
3805
# Generate deltas against each tree, to find the shortest.
3806
texts_possibly_new_in_tree = set()
3807
for basis_id, basis_tree in possible_trees:
3808
delta = tree.inventory._make_delta(basis_tree.inventory)
3809
for old_path, new_path, file_id, new_entry in delta:
3810
if new_path is None:
3811
# This file_id isn't present in the new rev, so we don't
3812
# care about it.
3813
continue
3814
if not new_path:
3815
# Rich roots are handled elsewhere...
3816
continue
3817
kind = new_entry.kind
3818
if kind != 'directory' and kind != 'file':
3819
# No text record associated with this inventory entry.
3820
continue
3821
# This is a directory or file that has changed somehow.
3822
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3823
deltas.append((len(delta), basis_id, delta))
3824
deltas.sort()
3825
return deltas[0][1:]
3826
3827
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3828
"""Find all parent revisions that are absent, but for which the
3829
inventory is present, and copy those inventories.
3830
3831
This is necessary to preserve correctness when the source is stacked
3832
without fallbacks configured. (Note that in cases like upgrade the
3833
source may be not have _fallback_repositories even though it is
3834
stacked.)
3835
"""
3836
parent_revs = set()
3837
for parents in parent_map.values():
3838
parent_revs.update(parents)
3839
present_parents = self.source.get_parent_map(parent_revs)
3840
absent_parents = set(parent_revs).difference(present_parents)
3841
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3842
(rev_id,) for rev_id in absent_parents)
3843
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3844
for parent_tree in self.source.revision_trees(parent_inv_ids):
3845
current_revision_id = parent_tree.get_revision_id()
3846
parents_parents_keys = parent_invs_keys_for_stacking[
3847
(current_revision_id,)]
3848
parents_parents = [key[-1] for key in parents_parents_keys]
3849
basis_id = _mod_revision.NULL_REVISION
3850
basis_tree = self.source.revision_tree(basis_id)
3851
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3852
self.target.add_inventory_by_delta(
3853
basis_id, delta, current_revision_id, parents_parents)
3854
cache[current_revision_id] = parent_tree
3855
3856
def _fetch_batch(self, revision_ids, basis_id, cache):
3857
"""Fetch across a few revisions.
3858
3859
:param revision_ids: The revisions to copy
3860
:param basis_id: The revision_id of a tree that must be in cache, used
3861
as a basis for delta when no other base is available
3862
:param cache: A cache of RevisionTrees that we can use.
3863
:return: The revision_id of the last converted tree. The RevisionTree
3864
for it will be in cache
3865
"""
3866
# Walk though all revisions; get inventory deltas, copy referenced
3867
# texts that delta references, insert the delta, revision and
3868
# signature.
3869
root_keys_to_create = set()
3870
text_keys = set()
3871
pending_deltas = []
3872
pending_revisions = []
3873
parent_map = self.source.get_parent_map(revision_ids)
3874
self._fetch_parent_invs_for_stacking(parent_map, cache)
3875
self.source._safe_to_return_from_cache = True
3876
for tree in self.source.revision_trees(revision_ids):
3877
# Find a inventory delta for this revision.
3878
# Find text entries that need to be copied, too.
3879
current_revision_id = tree.get_revision_id()
3880
parent_ids = parent_map.get(current_revision_id, ())
3881
parent_trees = self._get_trees(parent_ids, cache)
3882
possible_trees = list(parent_trees)
3883
if len(possible_trees) == 0:
3884
# There either aren't any parents, or the parents are ghosts,
3885
# so just use the last converted tree.
3886
possible_trees.append((basis_id, cache[basis_id]))
3887
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3888
possible_trees)
3889
revision = self.source.get_revision(current_revision_id)
3890
pending_deltas.append((basis_id, delta,
3891
current_revision_id, revision.parent_ids))
3892
if self._converting_to_rich_root:
3893
self._revision_id_to_root_id[current_revision_id] = \
3894
tree.get_root_id()
3895
# Determine which texts are in present in this revision but not in
3896
# any of the available parents.
3897
texts_possibly_new_in_tree = set()
3898
for old_path, new_path, file_id, entry in delta:
3899
if new_path is None:
3900
# This file_id isn't present in the new rev
3901
continue
3902
if not new_path:
3903
# This is the root
3904
if not self.target.supports_rich_root():
3905
# The target doesn't support rich root, so we don't
3906
# copy
3907
continue
3908
if self._converting_to_rich_root:
3909
# This can't be copied normally, we have to insert
3910
# it specially
3911
root_keys_to_create.add((file_id, entry.revision))
3912
continue
3913
kind = entry.kind
3914
texts_possibly_new_in_tree.add((file_id, entry.revision))
3915
for basis_id, basis_tree in possible_trees:
3916
basis_inv = basis_tree.inventory
3917
for file_key in list(texts_possibly_new_in_tree):
3918
file_id, file_revision = file_key
3919
try:
3920
entry = basis_inv[file_id]
3921
except errors.NoSuchId:
3922
continue
3923
if entry.revision == file_revision:
3924
texts_possibly_new_in_tree.remove(file_key)
3925
text_keys.update(texts_possibly_new_in_tree)
3926
pending_revisions.append(revision)
3927
cache[current_revision_id] = tree
3928
basis_id = current_revision_id
3929
self.source._safe_to_return_from_cache = False
3930
# Copy file texts
3931
from_texts = self.source.texts
3932
to_texts = self.target.texts
3933
if root_keys_to_create:
3934
root_stream = _mod_fetch._new_root_data_stream(
3935
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3936
self.source)
3937
to_texts.insert_record_stream(root_stream)
3938
to_texts.insert_record_stream(from_texts.get_record_stream(
3939
text_keys, self.target._format._fetch_order,
3940
not self.target._format._fetch_uses_deltas))
3941
# insert inventory deltas
3942
for delta in pending_deltas:
3943
self.target.add_inventory_by_delta(*delta)
3944
if self.target._fallback_repositories:
3945
# Make sure this stacked repository has all the parent inventories
3946
# for the new revisions that we are about to insert. We do this
3947
# before adding the revisions so that no revision is added until
3948
# all the inventories it may depend on are added.
3949
# Note that this is overzealous, as we may have fetched these in an
3950
# earlier batch.
3951
parent_ids = set()
3952
revision_ids = set()
3953
for revision in pending_revisions:
3954
revision_ids.add(revision.revision_id)
3955
parent_ids.update(revision.parent_ids)
3956
parent_ids.difference_update(revision_ids)
3957
parent_ids.discard(_mod_revision.NULL_REVISION)
3958
parent_map = self.source.get_parent_map(parent_ids)
3959
# we iterate over parent_map and not parent_ids because we don't
3960
# want to try copying any revision which is a ghost
3961
for parent_tree in self.source.revision_trees(parent_map):
3962
current_revision_id = parent_tree.get_revision_id()
3963
parents_parents = parent_map[current_revision_id]
3964
possible_trees = self._get_trees(parents_parents, cache)
3965
if len(possible_trees) == 0:
3966
# There either aren't any parents, or the parents are
3967
# ghosts, so just use the last converted tree.
3968
possible_trees.append((basis_id, cache[basis_id]))
3969
basis_id, delta = self._get_delta_for_revision(parent_tree,
3970
parents_parents, possible_trees)
3971
self.target.add_inventory_by_delta(
3972
basis_id, delta, current_revision_id, parents_parents)
3973
# insert signatures and revisions
3974
for revision in pending_revisions:
3975
try:
3976
signature = self.source.get_signature_text(
3977
revision.revision_id)
3978
self.target.add_signature_text(revision.revision_id,
3979
signature)
3980
except errors.NoSuchRevision:
3981
pass
3982
self.target.add_revision(revision.revision_id, revision)
3983
return basis_id
3984
3985
def _fetch_all_revisions(self, revision_ids, pb):
3986
"""Fetch everything for the list of revisions.
3987
3988
:param revision_ids: The list of revisions to fetch. Must be in
3989
topological order.
3990
:param pb: A ProgressTask
3991
:return: None
3992
"""
3993
basis_id, basis_tree = self._get_basis(revision_ids[0])
3994
batch_size = 100
3995
cache = lru_cache.LRUCache(100)
3996
cache[basis_id] = basis_tree
3997
del basis_tree # We don't want to hang on to it here
3998
hints = []
3999
a_graph = None
4000
4001
for offset in range(0, len(revision_ids), batch_size):
4002
self.target.start_write_group()
4003
try:
4004
pb.update('Transferring revisions', offset,
4005
len(revision_ids))
4006
batch = revision_ids[offset:offset+batch_size]
4007
basis_id = self._fetch_batch(batch, basis_id, cache)
4008
except:
4009
self.source._safe_to_return_from_cache = False
4010
self.target.abort_write_group()
4011
raise
4012
else:
4013
hint = self.target.commit_write_group()
4014
if hint:
4015
hints.extend(hint)
4016
if hints and self.target._format.pack_compresses:
4017
self.target.pack(hint=hints)
4018
pb.update('Transferring revisions', len(revision_ids),
4019
len(revision_ids))
4020
4021
@needs_write_lock
4022
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
4023
fetch_spec=None):
4024
"""See InterRepository.fetch()."""
4025
if fetch_spec is not None:
4026
raise AssertionError("Not implemented yet...")
4027
ui.ui_factory.warn_experimental_format_fetch(self)
4028
if (not self.source.supports_rich_root()
4029
and self.target.supports_rich_root()):
4030
self._converting_to_rich_root = True
4031
self._revision_id_to_root_id = {}
4032
else:
4033
self._converting_to_rich_root = False
4034
# See <https://launchpad.net/bugs/456077> asking for a warning here
4035
if self.source._format.network_name() != self.target._format.network_name():
4036
ui.ui_factory.show_user_warning('cross_format_fetch',
4037
from_format=self.source._format,
4038
to_format=self.target._format)
4039
revision_ids = self.target.search_missing_revision_ids(self.source,
4040
revision_id, find_ghosts=find_ghosts).get_keys()
4041
if not revision_ids:
4042
return 0, 0
4043
revision_ids = tsort.topo_sort(
4044
self.source.get_graph().get_parent_map(revision_ids))
4045
if not revision_ids:
4046
return 0, 0
4047
# Walk though all revisions; get inventory deltas, copy referenced
4048
# texts that delta references, insert the delta, revision and
4049
# signature.
4050
if pb is None:
4051
my_pb = ui.ui_factory.nested_progress_bar()
4052
pb = my_pb
4053
else:
4054
symbol_versioning.warn(
4055
symbol_versioning.deprecated_in((1, 14, 0))
4056
% "pb parameter to fetch()")
4057
my_pb = None
4058
try:
4059
self._fetch_all_revisions(revision_ids, pb)
4060
finally:
4061
if my_pb is not None:
4062
my_pb.finished()
4063
return len(revision_ids), 0
4064
4065
def _get_basis(self, first_revision_id):
4066
"""Get a revision and tree which exists in the target.
4067
4068
This assumes that first_revision_id is selected for transmission
4069
because all other ancestors are already present. If we can't find an
4070
ancestor we fall back to NULL_REVISION since we know that is safe.
4071
4072
:return: (basis_id, basis_tree)
4073
"""
4074
first_rev = self.source.get_revision(first_revision_id)
4075
try:
4076
basis_id = first_rev.parent_ids[0]
4077
# only valid as a basis if the target has it
4078
self.target.get_revision(basis_id)
4079
# Try to get a basis tree - if its a ghost it will hit the
4080
# NoSuchRevision case.
4081
basis_tree = self.source.revision_tree(basis_id)
4082
except (IndexError, errors.NoSuchRevision):
4083
basis_id = _mod_revision.NULL_REVISION
4084
basis_tree = self.source.revision_tree(basis_id)
4085
return basis_id, basis_tree
4086
4087
4088
InterRepository.register_optimiser(InterDifferingSerializer)
4089
InterRepository.register_optimiser(InterSameDataRepository)
4090
InterRepository.register_optimiser(InterWeaveRepo)
4091
InterRepository.register_optimiser(InterKnitRepo)
4092
4093
4094
1741
class CopyConverter(object):
4095
1742
"""A repository conversion tool which just performs a copy of the content.
4096
1743
4117
1764
# trigger an assertion if not such
4118
1765
repo._format.get_format_string()
4119
1766
self.repo_dir = repo.bzrdir
4120
pb.update('Moving repository to repository.backup')
1767
pb.update(gettext('Moving repository to repository.backup'))
4121
1768
self.repo_dir.transport.move('repository', 'repository.backup')
4122
1769
backup_transport = self.repo_dir.transport.clone('repository.backup')
4123
1770
repo._format.check_conversion_target(self.target_format)
4124
1771
self.source_repo = repo._format.open(self.repo_dir,
4125
1772
_found=True,
4126
1773
_override_transport=backup_transport)
4127
pb.update('Creating new repository')
1774
pb.update(gettext('Creating new repository'))
4128
1775
converted = self.target_format.initialize(self.repo_dir,
4129
1776
self.source_repo.is_shared())
4130
1777
converted.lock_write()
4131
1778
try:
4132
pb.update('Copying content')
1779
pb.update(gettext('Copying content'))
4133
1780
self.source_repo.copy_content_into(converted)
4134
1781
finally:
4135
1782
converted.unlock()
4136
pb.update('Deleting old repository content')
1783
pb.update(gettext('Deleting old repository content'))
4137
1784
self.repo_dir.transport.delete_tree('repository.backup')
4138
ui.ui_factory.note('repository converted')
1785
ui.ui_factory.note(gettext('repository converted'))
4139
1786
pb.finished()
4140
1787
4141
1788
4142
_unescape_map = {
4143
'apos':"'",
4144
'quot':'"',
4145
'amp':'&',
4146
'lt':'<',
4147
'gt':'>'
4148
}
4149
4150
4151
def _unescaper(match, _map=_unescape_map):
4152
code = match.group(1)
4153
try:
4154
return _map[code]
4155
except KeyError:
4156
if not code.startswith('#'):
4157
raise
4158
return unichr(int(code[1:])).encode('utf8')
4159
4160
4161
_unescape_re = None
4162
4163
4164
def _unescape_xml(data):
4165
"""Unescape predefined XML entities in a string of data."""
4166
global _unescape_re
4167
if _unescape_re is None:
4168
_unescape_re = re.compile('\&([^;]*);')
4169
return _unescape_re.sub(_unescaper, data)
4170
4171
4172
class _VersionedFileChecker(object):
4173
4174
def __init__(self, repository, text_key_references=None, ancestors=None):
4175
self.repository = repository
4176
self.text_index = self.repository._generate_text_key_index(
4177
text_key_references=text_key_references, ancestors=ancestors)
4178
4179
def calculate_file_version_parents(self, text_key):
4180
"""Calculate the correct parents for a file version according to
4181
the inventories.
4182
"""
4183
parent_keys = self.text_index[text_key]
4184
if parent_keys == [_mod_revision.NULL_REVISION]:
4185
return ()
4186
return tuple(parent_keys)
4187
4188
def check_file_version_parents(self, texts, progress_bar=None):
4189
"""Check the parents stored in a versioned file are correct.
4190
4191
It also detects file versions that are not referenced by their
4192
corresponding revision's inventory.
4193
4194
:returns: A tuple of (wrong_parents, dangling_file_versions).
4195
wrong_parents is a dict mapping {revision_id: (stored_parents,
4196
correct_parents)} for each revision_id where the stored parents
4197
are not correct. dangling_file_versions is a set of (file_id,
4198
revision_id) tuples for versions that are present in this versioned
4199
file, but not used by the corresponding inventory.
4200
"""
4201
local_progress = None
4202
if progress_bar is None:
4203
local_progress = ui.ui_factory.nested_progress_bar()
4204
progress_bar = local_progress
4205
try:
4206
return self._check_file_version_parents(texts, progress_bar)
4207
finally:
4208
if local_progress:
4209
local_progress.finished()
4210
4211
def _check_file_version_parents(self, texts, progress_bar):
4212
"""See check_file_version_parents."""
4213
wrong_parents = {}
4214
self.file_ids = set([file_id for file_id, _ in
4215
self.text_index.iterkeys()])
4216
# text keys is now grouped by file_id
4217
n_versions = len(self.text_index)
4218
progress_bar.update('loading text store', 0, n_versions)
4219
parent_map = self.repository.texts.get_parent_map(self.text_index)
4220
# On unlistable transports this could well be empty/error...
4221
text_keys = self.repository.texts.keys()
4222
unused_keys = frozenset(text_keys) - set(self.text_index)
4223
for num, key in enumerate(self.text_index.iterkeys()):
4224
progress_bar.update('checking text graph', num, n_versions)
4225
correct_parents = self.calculate_file_version_parents(key)
4226
try:
4227
knit_parents = parent_map[key]
4228
except errors.RevisionNotPresent:
4229
# Missing text!
4230
knit_parents = None
4231
if correct_parents != knit_parents:
4232
wrong_parents[key] = (knit_parents, correct_parents)
4233
return wrong_parents, unused_keys
4234
4235
4236
def _old_get_graph(repository, revision_id):
4237
"""DO NOT USE. That is all. I'm serious."""
4238
graph = repository.get_graph()
4239
revision_graph = dict(((key, value) for key, value in
4240
graph.iter_ancestry([revision_id]) if value is not None))
4241
return _strip_NULL_ghosts(revision_graph)
4242
4243
4244
1789
def _strip_NULL_ghosts(revision_graph):
4245
1790
"""Also don't use this. more compatibility code for unmigrated clients."""
4246
1791
# Filter ghosts, and null:
4252
1797
return revision_graph
4253
1798
4254
1799
4255
class StreamSink(object):
4256
"""An object that can insert a stream into a repository.
4257
4258
This interface handles the complexity of reserialising inventories and
4259
revisions from different formats, and allows unidirectional insertion into
4260
stacked repositories without looking for the missing basis parents
4261
beforehand.
4262
"""
4263
4264
def __init__(self, target_repo):
4265
self.target_repo = target_repo
4266
4267
def insert_stream(self, stream, src_format, resume_tokens):
4268
"""Insert a stream's content into the target repository.
4269
4270
:param src_format: a bzr repository format.
4271
4272
:return: a list of resume tokens and an iterable of keys additional
4273
items required before the insertion can be completed.
4274
"""
4275
self.target_repo.lock_write()
4276
try:
4277
if resume_tokens:
4278
self.target_repo.resume_write_group(resume_tokens)
4279
is_resume = True
4280
else:
4281
self.target_repo.start_write_group()
4282
is_resume = False
4283
try:
4284
# locked_insert_stream performs a commit|suspend.
4285
return self._locked_insert_stream(stream, src_format,
4286
is_resume)
4287
except:
4288
self.target_repo.abort_write_group(suppress_errors=True)
4289
raise
4290
finally:
4291
self.target_repo.unlock()
4292
4293
def _locked_insert_stream(self, stream, src_format, is_resume):
4294
to_serializer = self.target_repo._format._serializer
4295
src_serializer = src_format._serializer
4296
new_pack = None
4297
if to_serializer == src_serializer:
4298
# If serializers match and the target is a pack repository, set the
4299
# write cache size on the new pack. This avoids poor performance
4300
# on transports where append is unbuffered (such as
4301
# RemoteTransport). This is safe to do because nothing should read
4302
# back from the target repository while a stream with matching
4303
# serialization is being inserted.
4304
# The exception is that a delta record from the source that should
4305
# be a fulltext may need to be expanded by the target (see
4306
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4307
# explicitly flush any buffered writes first in that rare case.
4308
try:
4309
new_pack = self.target_repo._pack_collection._new_pack
4310
except AttributeError:
4311
# Not a pack repository
4312
pass
4313
else:
4314
new_pack.set_write_cache_size(1024*1024)
4315
for substream_type, substream in stream:
4316
if 'stream' in debug.debug_flags:
4317
mutter('inserting substream: %s', substream_type)
4318
if substream_type == 'texts':
4319
self.target_repo.texts.insert_record_stream(substream)
4320
elif substream_type == 'inventories':
4321
if src_serializer == to_serializer:
4322
self.target_repo.inventories.insert_record_stream(
4323
substream)
4324
else:
4325
self._extract_and_insert_inventories(
4326
substream, src_serializer)
4327
elif substream_type == 'inventory-deltas':
4328
self._extract_and_insert_inventory_deltas(
4329
substream, src_serializer)
4330
elif substream_type == 'chk_bytes':
4331
# XXX: This doesn't support conversions, as it assumes the
4332
# conversion was done in the fetch code.
4333
self.target_repo.chk_bytes.insert_record_stream(substream)
4334
elif substream_type == 'revisions':
4335
# This may fallback to extract-and-insert more often than
4336
# required if the serializers are different only in terms of
4337
# the inventory.
4338
if src_serializer == to_serializer:
4339
self.target_repo.revisions.insert_record_stream(substream)
4340
else:
4341
self._extract_and_insert_revisions(substream,
4342
src_serializer)
4343
elif substream_type == 'signatures':
4344
self.target_repo.signatures.insert_record_stream(substream)
4345
else:
4346
raise AssertionError('kaboom! %s' % (substream_type,))
4347
# Done inserting data, and the missing_keys calculations will try to
4348
# read back from the inserted data, so flush the writes to the new pack
4349
# (if this is pack format).
4350
if new_pack is not None:
4351
new_pack._write_data('', flush=True)
4352
# Find all the new revisions (including ones from resume_tokens)
4353
missing_keys = self.target_repo.get_missing_parent_inventories(
4354
check_for_missing_texts=is_resume)
4355
try:
4356
for prefix, versioned_file in (
4357
('texts', self.target_repo.texts),
4358
('inventories', self.target_repo.inventories),
4359
('revisions', self.target_repo.revisions),
4360
('signatures', self.target_repo.signatures),
4361
('chk_bytes', self.target_repo.chk_bytes),
4362
):
4363
if versioned_file is None:
4364
continue
4365
# TODO: key is often going to be a StaticTuple object
4366
# I don't believe we can define a method by which
4367
# (prefix,) + StaticTuple will work, though we could
4368
# define a StaticTuple.sq_concat that would allow you to
4369
# pass in either a tuple or a StaticTuple as the second
4370
# object, so instead we could have:
4371
# StaticTuple(prefix) + key here...
4372
missing_keys.update((prefix,) + key for key in
4373
versioned_file.get_missing_compression_parent_keys())
4374
except NotImplementedError:
4375
# cannot even attempt suspending, and missing would have failed
4376
# during stream insertion.
4377
missing_keys = set()
4378
else:
4379
if missing_keys:
4380
# suspend the write group and tell the caller what we is
4381
# missing. We know we can suspend or else we would not have
4382
# entered this code path. (All repositories that can handle
4383
# missing keys can handle suspending a write group).
4384
write_group_tokens = self.target_repo.suspend_write_group()
4385
return write_group_tokens, missing_keys
4386
hint = self.target_repo.commit_write_group()
4387
if (to_serializer != src_serializer and
4388
self.target_repo._format.pack_compresses):
4389
self.target_repo.pack(hint=hint)
4390
return [], set()
4391
4392
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4393
target_rich_root = self.target_repo._format.rich_root_data
4394
target_tree_refs = self.target_repo._format.supports_tree_reference
4395
for record in substream:
4396
# Insert the delta directly
4397
inventory_delta_bytes = record.get_bytes_as('fulltext')
4398
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4399
try:
4400
parse_result = deserialiser.parse_text_bytes(
4401
inventory_delta_bytes)
4402
except inventory_delta.IncompatibleInventoryDelta, err:
4403
trace.mutter("Incompatible delta: %s", err.msg)
4404
raise errors.IncompatibleRevision(self.target_repo._format)
4405
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4406
revision_id = new_id
4407
parents = [key[0] for key in record.parents]
4408
self.target_repo.add_inventory_by_delta(
4409
basis_id, inv_delta, revision_id, parents)
4410
4411
def _extract_and_insert_inventories(self, substream, serializer,
4412
parse_delta=None):
4413
"""Generate a new inventory versionedfile in target, converting data.
4414
4415
The inventory is retrieved from the source, (deserializing it), and
4416
stored in the target (reserializing it in a different format).
4417
"""
4418
target_rich_root = self.target_repo._format.rich_root_data
4419
target_tree_refs = self.target_repo._format.supports_tree_reference
4420
for record in substream:
4421
# It's not a delta, so it must be a fulltext in the source
4422
# serializer's format.
4423
bytes = record.get_bytes_as('fulltext')
4424
revision_id = record.key[0]
4425
inv = serializer.read_inventory_from_string(bytes, revision_id)
4426
parents = [key[0] for key in record.parents]
4427
self.target_repo.add_inventory(revision_id, inv, parents)
4428
# No need to keep holding this full inv in memory when the rest of
4429
# the substream is likely to be all deltas.
4430
del inv
4431
4432
def _extract_and_insert_revisions(self, substream, serializer):
4433
for record in substream:
4434
bytes = record.get_bytes_as('fulltext')
4435
revision_id = record.key[0]
4436
rev = serializer.read_revision_from_string(bytes)
4437
if rev.revision_id != revision_id:
4438
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4439
self.target_repo.add_revision(revision_id, rev)
4440
4441
def finished(self):
4442
if self.target_repo._format._fetch_reconcile:
4443
self.target_repo.reconcile()
4444
4445
4446
class StreamSource(object):
4447
"""A source of a stream for fetching between repositories."""
4448
4449
def __init__(self, from_repository, to_format):
4450
"""Create a StreamSource streaming from from_repository."""
4451
self.from_repository = from_repository
4452
self.to_format = to_format
4453
self._record_counter = RecordCounter()
4454
4455
def delta_on_metadata(self):
4456
"""Return True if delta's are permitted on metadata streams.
4457
4458
That is on revisions and signatures.
4459
"""
4460
src_serializer = self.from_repository._format._serializer
4461
target_serializer = self.to_format._serializer
4462
return (self.to_format._fetch_uses_deltas and
4463
src_serializer == target_serializer)
4464
4465
def _fetch_revision_texts(self, revs):
4466
# fetch signatures first and then the revision texts
4467
# may need to be a InterRevisionStore call here.
4468
from_sf = self.from_repository.signatures
4469
# A missing signature is just skipped.
4470
keys = [(rev_id,) for rev_id in revs]
4471
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4472
keys,
4473
self.to_format._fetch_order,
4474
not self.to_format._fetch_uses_deltas))
4475
# If a revision has a delta, this is actually expanded inside the
4476
# insert_record_stream code now, which is an alternate fix for
4477
# bug #261339
4478
from_rf = self.from_repository.revisions
4479
revisions = from_rf.get_record_stream(
4480
keys,
4481
self.to_format._fetch_order,
4482
not self.delta_on_metadata())
4483
return [('signatures', signatures), ('revisions', revisions)]
4484
4485
def _generate_root_texts(self, revs):
4486
"""This will be called by get_stream between fetching weave texts and
4487
fetching the inventory weave.
4488
"""
4489
if self._rich_root_upgrade():
4490
return _mod_fetch.Inter1and2Helper(
4491
self.from_repository).generate_root_texts(revs)
4492
else:
4493
return []
4494
4495
def get_stream(self, search):
4496
phase = 'file'
4497
revs = search.get_keys()
4498
graph = self.from_repository.get_graph()
4499
revs = tsort.topo_sort(graph.get_parent_map(revs))
4500
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4501
text_keys = []
4502
for knit_kind, file_id, revisions in data_to_fetch:
4503
if knit_kind != phase:
4504
phase = knit_kind
4505
# Make a new progress bar for this phase
4506
if knit_kind == "file":
4507
# Accumulate file texts
4508
text_keys.extend([(file_id, revision) for revision in
4509
revisions])
4510
elif knit_kind == "inventory":
4511
# Now copy the file texts.
4512
from_texts = self.from_repository.texts
4513
yield ('texts', from_texts.get_record_stream(
4514
text_keys, self.to_format._fetch_order,
4515
not self.to_format._fetch_uses_deltas))
4516
# Cause an error if a text occurs after we have done the
4517
# copy.
4518
text_keys = None
4519
# Before we process the inventory we generate the root
4520
# texts (if necessary) so that the inventories references
4521
# will be valid.
4522
for _ in self._generate_root_texts(revs):
4523
yield _
4524
# we fetch only the referenced inventories because we do not
4525
# know for unselected inventories whether all their required
4526
# texts are present in the other repository - it could be
4527
# corrupt.
4528
for info in self._get_inventory_stream(revs):
4529
yield info
4530
elif knit_kind == "signatures":
4531
# Nothing to do here; this will be taken care of when
4532
# _fetch_revision_texts happens.
4533
pass
4534
elif knit_kind == "revisions":
4535
for record in self._fetch_revision_texts(revs):
4536
yield record
4537
else:
4538
raise AssertionError("Unknown knit kind %r" % knit_kind)
4539
4540
def get_stream_for_missing_keys(self, missing_keys):
4541
# missing keys can only occur when we are byte copying and not
4542
# translating (because translation means we don't send
4543
# unreconstructable deltas ever).
4544
keys = {}
4545
keys['texts'] = set()
4546
keys['revisions'] = set()
4547
keys['inventories'] = set()
4548
keys['chk_bytes'] = set()
4549
keys['signatures'] = set()
4550
for key in missing_keys:
4551
keys[key[0]].add(key[1:])
4552
if len(keys['revisions']):
4553
# If we allowed copying revisions at this point, we could end up
4554
# copying a revision without copying its required texts: a
4555
# violation of the requirements for repository integrity.
4556
raise AssertionError(
4557
'cannot copy revisions to fill in missing deltas %s' % (
4558
keys['revisions'],))
4559
for substream_kind, keys in keys.iteritems():
4560
vf = getattr(self.from_repository, substream_kind)
4561
if vf is None and keys:
4562
raise AssertionError(
4563
"cannot fill in keys for a versioned file we don't"
4564
" have: %s needs %s" % (substream_kind, keys))
4565
if not keys:
4566
# No need to stream something we don't have
4567
continue
4568
if substream_kind == 'inventories':
4569
# Some missing keys are genuinely ghosts, filter those out.
4570
present = self.from_repository.inventories.get_parent_map(keys)
4571
revs = [key[0] for key in present]
4572
# Get the inventory stream more-or-less as we do for the
4573
# original stream; there's no reason to assume that records
4574
# direct from the source will be suitable for the sink. (Think
4575
# e.g. 2a -> 1.9-rich-root).
4576
for info in self._get_inventory_stream(revs, missing=True):
4577
yield info
4578
continue
4579
4580
# Ask for full texts always so that we don't need more round trips
4581
# after this stream.
4582
# Some of the missing keys are genuinely ghosts, so filter absent
4583
# records. The Sink is responsible for doing another check to
4584
# ensure that ghosts don't introduce missing data for future
4585
# fetches.
4586
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4587
self.to_format._fetch_order, True))
4588
yield substream_kind, stream
4589
4590
def inventory_fetch_order(self):
4591
if self._rich_root_upgrade():
4592
return 'topological'
4593
else:
4594
return self.to_format._fetch_order
4595
4596
def _rich_root_upgrade(self):
4597
return (not self.from_repository._format.rich_root_data and
4598
self.to_format.rich_root_data)
4599
4600
def _get_inventory_stream(self, revision_ids, missing=False):
4601
from_format = self.from_repository._format
4602
if (from_format.supports_chks and self.to_format.supports_chks and
4603
from_format.network_name() == self.to_format.network_name()):
4604
raise AssertionError(
4605
"this case should be handled by GroupCHKStreamSource")
4606
elif 'forceinvdeltas' in debug.debug_flags:
4607
return self._get_convertable_inventory_stream(revision_ids,
4608
delta_versus_null=missing)
4609
elif from_format.network_name() == self.to_format.network_name():
4610
# Same format.
4611
return self._get_simple_inventory_stream(revision_ids,
4612
missing=missing)
4613
elif (not from_format.supports_chks and not self.to_format.supports_chks
4614
and from_format._serializer == self.to_format._serializer):
4615
# Essentially the same format.
4616
return self._get_simple_inventory_stream(revision_ids,
4617
missing=missing)
4618
else:
4619
# Any time we switch serializations, we want to use an
4620
# inventory-delta based approach.
4621
return self._get_convertable_inventory_stream(revision_ids,
4622
delta_versus_null=missing)
4623
4624
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4625
# NB: This currently reopens the inventory weave in source;
4626
# using a single stream interface instead would avoid this.
4627
from_weave = self.from_repository.inventories
4628
if missing:
4629
delta_closure = True
4630
else:
4631
delta_closure = not self.delta_on_metadata()
4632
yield ('inventories', from_weave.get_record_stream(
4633
[(rev_id,) for rev_id in revision_ids],
4634
self.inventory_fetch_order(), delta_closure))
4635
4636
def _get_convertable_inventory_stream(self, revision_ids,
4637
delta_versus_null=False):
4638
# The two formats are sufficiently different that there is no fast
4639
# path, so we need to send just inventorydeltas, which any
4640
# sufficiently modern client can insert into any repository.
4641
# The StreamSink code expects to be able to
4642
# convert on the target, so we need to put bytes-on-the-wire that can
4643
# be converted. That means inventory deltas (if the remote is <1.19,
4644
# RemoteStreamSink will fallback to VFS to insert the deltas).
4645
yield ('inventory-deltas',
4646
self._stream_invs_as_deltas(revision_ids,
4647
delta_versus_null=delta_versus_null))
4648
4649
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4650
"""Return a stream of inventory-deltas for the given rev ids.
4651
4652
:param revision_ids: The list of inventories to transmit
4653
:param delta_versus_null: Don't try to find a minimal delta for this
4654
entry, instead compute the delta versus the NULL_REVISION. This
4655
effectively streams a complete inventory. Used for stuff like
4656
filling in missing parents, etc.
4657
"""
4658
from_repo = self.from_repository
4659
revision_keys = [(rev_id,) for rev_id in revision_ids]
4660
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4661
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4662
# method...
4663
inventories = self.from_repository.iter_inventories(
4664
revision_ids, 'topological')
4665
format = from_repo._format
4666
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4667
inventory_cache = lru_cache.LRUCache(50)
4668
null_inventory = from_repo.revision_tree(
4669
_mod_revision.NULL_REVISION).inventory
4670
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4671
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4672
# repo back into a non-rich-root repo ought to be allowed)
4673
serializer = inventory_delta.InventoryDeltaSerializer(
4674
versioned_root=format.rich_root_data,
4675
tree_references=format.supports_tree_reference)
4676
for inv in inventories:
4677
key = (inv.revision_id,)
4678
parent_keys = parent_map.get(key, ())
4679
delta = None
4680
if not delta_versus_null and parent_keys:
4681
# The caller did not ask for complete inventories and we have
4682
# some parents that we can delta against. Make a delta against
4683
# each parent so that we can find the smallest.
4684
parent_ids = [parent_key[0] for parent_key in parent_keys]
4685
for parent_id in parent_ids:
4686
if parent_id not in invs_sent_so_far:
4687
# We don't know that the remote side has this basis, so
4688
# we can't use it.
4689
continue
4690
if parent_id == _mod_revision.NULL_REVISION:
4691
parent_inv = null_inventory
4692
else:
4693
parent_inv = inventory_cache.get(parent_id, None)
4694
if parent_inv is None:
4695
parent_inv = from_repo.get_inventory(parent_id)
4696
candidate_delta = inv._make_delta(parent_inv)
4697
if (delta is None or
4698
len(delta) > len(candidate_delta)):
4699
delta = candidate_delta
4700
basis_id = parent_id
4701
if delta is None:
4702
# Either none of the parents ended up being suitable, or we
4703
# were asked to delta against NULL
4704
basis_id = _mod_revision.NULL_REVISION
4705
delta = inv._make_delta(null_inventory)
4706
invs_sent_so_far.add(inv.revision_id)
4707
inventory_cache[inv.revision_id] = inv
4708
delta_serialized = ''.join(
4709
serializer.delta_to_lines(basis_id, key[-1], delta))
4710
yield versionedfile.FulltextContentFactory(
4711
key, parent_keys, None, delta_serialized)
4712
4713
4714
1800
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4715
1801
stop_revision=None):
4716
1802
"""Extend the partial history to include a given index
4726
1812
it is encountered, history extension will stop.
4727
1813
"""
4728
1814
start_revision = partial_history_cache[-1]
4729
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,))
4730
1818
try:
4731
#skip the last revision in the list
1819
# skip the last revision in the list
4732
1820
iterator.next()
4733
1821
while True:
4734
1822
if (stop_index is not None and
4742
1830
# No more history
4743
1831
return
4744
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