« back to all changes in this revision
Viewing changes to bzrlib/repository.py
- browse files at revision 5436.3.7
- compare with another revision
- download diff
- download tarball
- view history from revision 5436.3.7
- Committer: Jelmer Vernooij
- Date: 2011-12-14 19:44:40 UTC
- mfrom: (6368 +trunk)
- mto: (6366.1.2 count-transport-connections)
- mto: This revision was merged to the branch mainline in revision 6405.
- Revision ID: jelmer@samba.org-20111214194440-uh68nvn5gzihn2xc
Merge bzr.dev, update to use new hooks.
- 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
pyutils,
43
32
revision as _mod_revision,
44
static_tuple,
45
symbol_versioning,
46
trace,
33
testament as _mod_testament,
47
34
tsort,
48
versionedfile,
35
gpg,
49
36
)
50
37
from bzrlib.bundle import serializer
51
from bzrlib.revisiontree import RevisionTree
52
from bzrlib.store.versioned import VersionedFileStore
53
from bzrlib.testament import Testament
38
from bzrlib.i18n import gettext
54
39
""")
55
40
56
import sys
57
41
from bzrlib import (
58
42
errors,
59
43
registry,
44
symbol_versioning,
60
45
ui,
61
46
)
62
47
from bzrlib.decorators import needs_read_lock, needs_write_lock, only_raises
63
48
from bzrlib.inter import InterObject
64
from bzrlib.inventory import (
65
Inventory,
66
InventoryDirectory,
67
ROOT_ID,
68
entry_factory,
69
)
70
from bzrlib.recordcounter import RecordCounter
71
49
from bzrlib.lock import _RelockDebugMixin, LogicalLockResult
72
50
from bzrlib.trace import (
73
51
log_exception_quietly, note, mutter, mutter_callsite, warning)
94
72
95
73
# all clients should supply tree roots.
96
74
record_root_entry = True
97
# the default CommitBuilder does not manage trees whose root is versioned.
98
_versioned_root = False
75
# whether this commit builder supports the record_entry_contents interface
76
supports_record_entry_contents = False
77
# whether this commit builder will automatically update the branch that is
78
# being committed to
79
updates_branch = False
99
80
100
81
def __init__(self, repository, parents, config, timestamp=None,
101
82
timezone=None, committer=None, revprops=None,
102
revision_id=None):
83
revision_id=None, lossy=False):
103
84
"""Initiate a CommitBuilder.
104
85
105
86
:param repository: Repository to commit to.
106
87
:param parents: Revision ids of the parents of the new revision.
107
:param config: Configuration to use.
108
88
:param timestamp: Optional timestamp recorded for commit.
109
89
:param timezone: Optional timezone for timestamp.
110
90
:param committer: Optional committer to set for commit.
111
91
:param revprops: Optional dictionary of revision properties.
112
92
:param revision_id: Optional revision id.
93
:param lossy: Whether to discard data that can not be natively
94
represented, when pushing to a foreign VCS
113
95
"""
114
96
self._config = config
97
self._lossy = lossy
115
98
116
99
if committer is None:
117
100
self._committer = self._config.username()
101
elif not isinstance(committer, unicode):
102
self._committer = committer.decode() # throw if non-ascii
118
103
else:
119
104
self._committer = committer
120
105
121
self.new_inventory = Inventory(None)
122
106
self._new_revision_id = revision_id
123
107
self.parents = parents
124
108
self.repository = repository
139
123
self._timezone = int(timezone)
140
124
141
125
self._generate_revision_if_needed()
142
self.__heads = graph.HeadsCache(repository.get_graph()).heads
143
self._basis_delta = []
144
# API compatibility, older code that used CommitBuilder did not call
145
# .record_delete(), which means the delta that is computed would not be
146
# valid. Callers that will call record_delete() should call
147
# .will_record_deletes() to indicate that.
148
self._recording_deletes = False
149
# memo'd check for no-op commits.
150
self._any_changes = False
151
126
152
127
def any_changes(self):
153
128
"""Return True if any entries were changed.
154
129
155
130
This includes merge-only changes. It is the core for the --unchanged
156
131
detection in commit.
157
132
158
133
:return: True if any changes have occured.
159
134
"""
160
return self._any_changes
135
raise NotImplementedError(self.any_changes)
161
136
162
137
def _validate_unicode_text(self, text, context):
163
138
"""Verify things like commit messages don't have bogus characters."""
179
154
180
155
:return: The revision id of the recorded revision.
181
156
"""
182
self._validate_unicode_text(message, 'commit message')
183
rev = _mod_revision.Revision(
184
timestamp=self._timestamp,
185
timezone=self._timezone,
186
committer=self._committer,
187
message=message,
188
inventory_sha1=self.inv_sha1,
189
revision_id=self._new_revision_id,
190
properties=self._revprops)
191
rev.parent_ids = self.parents
192
self.repository.add_revision(self._new_revision_id, rev,
193
self.new_inventory, self._config)
194
self.repository.commit_write_group()
195
return self._new_revision_id
157
raise NotImplementedError(self.commit)
196
158
197
159
def abort(self):
198
160
"""Abort the commit that is being built.
199
161
"""
200
self.repository.abort_write_group()
162
raise NotImplementedError(self.abort)
201
163
202
164
def revision_tree(self):
203
165
"""Return the tree that was just committed.
204
166
205
After calling commit() this can be called to get a RevisionTree
206
representing the newly committed tree. This is preferred to
207
calling Repository.revision_tree() because that may require
208
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
209
171
memory.
210
172
"""
211
if self.new_inventory is None:
212
self.new_inventory = self.repository.get_inventory(
213
self._new_revision_id)
214
return RevisionTree(self.repository, self.new_inventory,
215
self._new_revision_id)
173
raise NotImplementedError(self.revision_tree)
216
174
217
175
def finish_inventory(self):
218
176
"""Tell the builder that the inventory is finished.
220
178
:return: The inventory id in the repository, which can be used with
221
179
repository.get_inventory.
222
180
"""
223
if self.new_inventory is None:
224
# an inventory delta was accumulated without creating a new
225
# inventory.
226
basis_id = self.basis_delta_revision
227
# We ignore the 'inventory' returned by add_inventory_by_delta
228
# because self.new_inventory is used to hint to the rest of the
229
# system what code path was taken
230
self.inv_sha1, _ = self.repository.add_inventory_by_delta(
231
basis_id, self._basis_delta, self._new_revision_id,
232
self.parents)
233
else:
234
if self.new_inventory.root is None:
235
raise AssertionError('Root entry should be supplied to'
236
' record_entry_contents, as of bzr 0.10.')
237
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
238
self.new_inventory.revision_id = self._new_revision_id
239
self.inv_sha1 = self.repository.add_inventory(
240
self._new_revision_id,
241
self.new_inventory,
242
self.parents
243
)
244
return self._new_revision_id
181
raise NotImplementedError(self.finish_inventory)
245
182
246
183
def _gen_revision_id(self):
247
184
"""Return new revision-id."""
262
199
else:
263
200
self.random_revid = False
264
201
265
def _heads(self, file_id, revision_ids):
266
"""Calculate the graph heads for revision_ids in the graph of file_id.
267
268
This can use either a per-file graph or a global revision graph as we
269
have an identity relationship between the two graphs.
270
"""
271
return self.__heads(revision_ids)
272
273
def _check_root(self, ie, parent_invs, tree):
274
"""Helper for record_entry_contents.
275
276
:param ie: An entry being added.
277
:param parent_invs: The inventories of the parent revisions of the
278
commit.
279
:param tree: The tree that is being committed.
280
"""
281
# In this revision format, root entries have no knit or weave When
282
# serializing out to disk and back in root.revision is always
283
# _new_revision_id
284
ie.revision = self._new_revision_id
285
286
def _require_root_change(self, tree):
287
"""Enforce an appropriate root object change.
288
289
This is called once when record_iter_changes is called, if and only if
290
the root was not in the delta calculated by record_iter_changes.
291
292
:param tree: The tree which is being committed.
293
"""
294
if len(self.parents) == 0:
295
raise errors.RootMissing()
296
entry = entry_factory['directory'](tree.path2id(''), '',
297
None)
298
entry.revision = self._new_revision_id
299
self._basis_delta.append(('', '', entry.file_id, entry))
300
301
def _get_delta(self, ie, basis_inv, path):
302
"""Get a delta against the basis inventory for ie."""
303
if ie.file_id not in basis_inv:
304
# add
305
result = (None, path, ie.file_id, ie)
306
self._basis_delta.append(result)
307
return result
308
elif ie != basis_inv[ie.file_id]:
309
# common but altered
310
# TODO: avoid tis id2path call.
311
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
312
self._basis_delta.append(result)
313
return result
314
else:
315
# common, unaltered
316
return None
317
318
def get_basis_delta(self):
319
"""Return the complete inventory delta versus the basis inventory.
320
321
This has been built up with the calls to record_delete and
322
record_entry_contents. The client must have already called
323
will_record_deletes() to indicate that they will be generating a
324
complete delta.
325
326
:return: An inventory delta, suitable for use with apply_delta, or
327
Repository.add_inventory_by_delta, etc.
328
"""
329
if not self._recording_deletes:
330
raise AssertionError("recording deletes not activated.")
331
return self._basis_delta
332
333
def record_delete(self, path, file_id):
334
"""Record that a delete occured against a basis tree.
335
336
This is an optional API - when used it adds items to the basis_delta
337
being accumulated by the commit builder. It cannot be called unless the
338
method will_record_deletes() has been called to inform the builder that
339
a delta is being supplied.
340
341
:param path: The path of the thing deleted.
342
:param file_id: The file id that was deleted.
343
"""
344
if not self._recording_deletes:
345
raise AssertionError("recording deletes not activated.")
346
delta = (path, None, file_id, None)
347
self._basis_delta.append(delta)
348
self._any_changes = True
349
return delta
350
351
202
def will_record_deletes(self):
352
203
"""Tell the commit builder that deletes are being notified.
353
204
355
206
commit to be valid, deletes against the basis MUST be recorded via
356
207
builder.record_delete().
357
208
"""
358
self._recording_deletes = True
359
try:
360
basis_id = self.parents[0]
361
except IndexError:
362
basis_id = _mod_revision.NULL_REVISION
363
self.basis_delta_revision = basis_id
364
365
def record_entry_contents(self, ie, parent_invs, path, tree,
366
content_summary):
367
"""Record the content of ie from tree into the commit if needed.
368
369
Side effect: sets ie.revision when unchanged
370
371
:param ie: An inventory entry present in the commit.
372
:param parent_invs: The inventories of the parent revisions of the
373
commit.
374
:param path: The path the entry is at in the tree.
375
:param tree: The tree which contains this entry and should be used to
376
obtain content.
377
:param content_summary: Summary data from the tree about the paths
378
content - stat, length, exec, sha/link target. This is only
379
accessed when the entry has a revision of None - that is when it is
380
a candidate to commit.
381
:return: A tuple (change_delta, version_recorded, fs_hash).
382
change_delta is an inventory_delta change for this entry against
383
the basis tree of the commit, or None if no change occured against
384
the basis tree.
385
version_recorded is True if a new version of the entry has been
386
recorded. For instance, committing a merge where a file was only
387
changed on the other side will return (delta, False).
388
fs_hash is either None, or the hash details for the path (currently
389
a tuple of the contents sha1 and the statvalue returned by
390
tree.get_file_with_stat()).
391
"""
392
if self.new_inventory.root is None:
393
if ie.parent_id is not None:
394
raise errors.RootMissing()
395
self._check_root(ie, parent_invs, tree)
396
if ie.revision is None:
397
kind = content_summary[0]
398
else:
399
# ie is carried over from a prior commit
400
kind = ie.kind
401
# XXX: repository specific check for nested tree support goes here - if
402
# the repo doesn't want nested trees we skip it ?
403
if (kind == 'tree-reference' and
404
not self.repository._format.supports_tree_reference):
405
# mismatch between commit builder logic and repository:
406
# this needs the entry creation pushed down into the builder.
407
raise NotImplementedError('Missing repository subtree support.')
408
self.new_inventory.add(ie)
409
410
# TODO: slow, take it out of the inner loop.
411
try:
412
basis_inv = parent_invs[0]
413
except IndexError:
414
basis_inv = Inventory(root_id=None)
415
416
# ie.revision is always None if the InventoryEntry is considered
417
# for committing. We may record the previous parents revision if the
418
# content is actually unchanged against a sole head.
419
if ie.revision is not None:
420
if not self._versioned_root and path == '':
421
# repositories that do not version the root set the root's
422
# revision to the new commit even when no change occurs (more
423
# specifically, they do not record a revision on the root; and
424
# the rev id is assigned to the root during deserialisation -
425
# this masks when a change may have occurred against the basis.
426
# To match this we always issue a delta, because the revision
427
# of the root will always be changing.
428
if ie.file_id in basis_inv:
429
delta = (basis_inv.id2path(ie.file_id), path,
430
ie.file_id, ie)
431
else:
432
# add
433
delta = (None, path, ie.file_id, ie)
434
self._basis_delta.append(delta)
435
return delta, False, None
436
else:
437
# we don't need to commit this, because the caller already
438
# determined that an existing revision of this file is
439
# appropriate. If its not being considered for committing then
440
# it and all its parents to the root must be unaltered so
441
# no-change against the basis.
442
if ie.revision == self._new_revision_id:
443
raise AssertionError("Impossible situation, a skipped "
444
"inventory entry (%r) claims to be modified in this "
445
"commit (%r).", (ie, self._new_revision_id))
446
return None, False, None
447
# XXX: Friction: parent_candidates should return a list not a dict
448
# so that we don't have to walk the inventories again.
449
parent_candiate_entries = ie.parent_candidates(parent_invs)
450
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
451
heads = []
452
for inv in parent_invs:
453
if ie.file_id in inv:
454
old_rev = inv[ie.file_id].revision
455
if old_rev in head_set:
456
heads.append(inv[ie.file_id].revision)
457
head_set.remove(inv[ie.file_id].revision)
458
459
store = False
460
# now we check to see if we need to write a new record to the
461
# file-graph.
462
# We write a new entry unless there is one head to the ancestors, and
463
# the kind-derived content is unchanged.
464
465
# Cheapest check first: no ancestors, or more the one head in the
466
# ancestors, we write a new node.
467
if len(heads) != 1:
468
store = True
469
if not store:
470
# There is a single head, look it up for comparison
471
parent_entry = parent_candiate_entries[heads[0]]
472
# if the non-content specific data has changed, we'll be writing a
473
# node:
474
if (parent_entry.parent_id != ie.parent_id or
475
parent_entry.name != ie.name):
476
store = True
477
# now we need to do content specific checks:
478
if not store:
479
# if the kind changed the content obviously has
480
if kind != parent_entry.kind:
481
store = True
482
# Stat cache fingerprint feedback for the caller - None as we usually
483
# don't generate one.
484
fingerprint = None
485
if kind == 'file':
486
if content_summary[2] is None:
487
raise ValueError("Files must not have executable = None")
488
if not store:
489
# We can't trust a check of the file length because of content
490
# filtering...
491
if (# if the exec bit has changed we have to store:
492
parent_entry.executable != content_summary[2]):
493
store = True
494
elif parent_entry.text_sha1 == content_summary[3]:
495
# all meta and content is unchanged (using a hash cache
496
# hit to check the sha)
497
ie.revision = parent_entry.revision
498
ie.text_size = parent_entry.text_size
499
ie.text_sha1 = parent_entry.text_sha1
500
ie.executable = parent_entry.executable
501
return self._get_delta(ie, basis_inv, path), False, None
502
else:
503
# Either there is only a hash change(no hash cache entry,
504
# or same size content change), or there is no change on
505
# this file at all.
506
# Provide the parent's hash to the store layer, so that the
507
# content is unchanged we will not store a new node.
508
nostore_sha = parent_entry.text_sha1
509
if store:
510
# We want to record a new node regardless of the presence or
511
# absence of a content change in the file.
512
nostore_sha = None
513
ie.executable = content_summary[2]
514
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
515
try:
516
text = file_obj.read()
517
finally:
518
file_obj.close()
519
try:
520
ie.text_sha1, ie.text_size = self._add_text_to_weave(
521
ie.file_id, text, heads, nostore_sha)
522
# Let the caller know we generated a stat fingerprint.
523
fingerprint = (ie.text_sha1, stat_value)
524
except errors.ExistingContent:
525
# Turns out that the file content was unchanged, and we were
526
# only going to store a new node if it was changed. Carry over
527
# the entry.
528
ie.revision = parent_entry.revision
529
ie.text_size = parent_entry.text_size
530
ie.text_sha1 = parent_entry.text_sha1
531
ie.executable = parent_entry.executable
532
return self._get_delta(ie, basis_inv, path), False, None
533
elif kind == 'directory':
534
if not store:
535
# all data is meta here, nothing specific to directory, so
536
# carry over:
537
ie.revision = parent_entry.revision
538
return self._get_delta(ie, basis_inv, path), False, None
539
self._add_text_to_weave(ie.file_id, '', heads, None)
540
elif kind == 'symlink':
541
current_link_target = content_summary[3]
542
if not store:
543
# symlink target is not generic metadata, check if it has
544
# changed.
545
if current_link_target != parent_entry.symlink_target:
546
store = True
547
if not store:
548
# unchanged, carry over.
549
ie.revision = parent_entry.revision
550
ie.symlink_target = parent_entry.symlink_target
551
return self._get_delta(ie, basis_inv, path), False, None
552
ie.symlink_target = current_link_target
553
self._add_text_to_weave(ie.file_id, '', heads, None)
554
elif kind == 'tree-reference':
555
if not store:
556
if content_summary[3] != parent_entry.reference_revision:
557
store = True
558
if not store:
559
# unchanged, carry over.
560
ie.reference_revision = parent_entry.reference_revision
561
ie.revision = parent_entry.revision
562
return self._get_delta(ie, basis_inv, path), False, None
563
ie.reference_revision = content_summary[3]
564
if ie.reference_revision is None:
565
raise AssertionError("invalid content_summary for nested tree: %r"
566
% (content_summary,))
567
self._add_text_to_weave(ie.file_id, '', heads, None)
568
else:
569
raise NotImplementedError('unknown kind')
570
ie.revision = self._new_revision_id
571
self._any_changes = True
572
return self._get_delta(ie, basis_inv, path), True, fingerprint
573
574
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
575
_entry_factory=entry_factory):
209
raise NotImplementedError(self.will_record_deletes)
210
211
def record_iter_changes(self, tree, basis_revision_id, iter_changes):
576
212
"""Record a new tree via iter_changes.
577
213
578
214
:param tree: The tree to obtain text contents from for changed objects.
583
219
to basis_revision_id. The iterator must not include any items with
584
220
a current kind of None - missing items must be either filtered out
585
221
or errored-on beefore record_iter_changes sees the item.
586
:param _entry_factory: Private method to bind entry_factory locally for
587
performance.
588
222
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
589
223
tree._observed_sha1.
590
224
"""
591
# Create an inventory delta based on deltas between all the parents and
592
# deltas between all the parent inventories. We use inventory delta's
593
# between the inventory objects because iter_changes masks
594
# last-changed-field only changes.
595
# Working data:
596
# file_id -> change map, change is fileid, paths, changed, versioneds,
597
# parents, names, kinds, executables
598
merged_ids = {}
599
# {file_id -> revision_id -> inventory entry, for entries in parent
600
# trees that are not parents[0]
601
parent_entries = {}
602
ghost_basis = False
603
try:
604
revtrees = list(self.repository.revision_trees(self.parents))
605
except errors.NoSuchRevision:
606
# one or more ghosts, slow path.
607
revtrees = []
608
for revision_id in self.parents:
609
try:
610
revtrees.append(self.repository.revision_tree(revision_id))
611
except errors.NoSuchRevision:
612
if not revtrees:
613
basis_revision_id = _mod_revision.NULL_REVISION
614
ghost_basis = True
615
revtrees.append(self.repository.revision_tree(
616
_mod_revision.NULL_REVISION))
617
# The basis inventory from a repository
618
if revtrees:
619
basis_inv = revtrees[0].inventory
620
else:
621
basis_inv = self.repository.revision_tree(
622
_mod_revision.NULL_REVISION).inventory
623
if len(self.parents) > 0:
624
if basis_revision_id != self.parents[0] and not ghost_basis:
625
raise Exception(
626
"arbitrary basis parents not yet supported with merges")
627
for revtree in revtrees[1:]:
628
for change in revtree.inventory._make_delta(basis_inv):
629
if change[1] is None:
630
# Not present in this parent.
631
continue
632
if change[2] not in merged_ids:
633
if change[0] is not None:
634
basis_entry = basis_inv[change[2]]
635
merged_ids[change[2]] = [
636
# basis revid
637
basis_entry.revision,
638
# new tree revid
639
change[3].revision]
640
parent_entries[change[2]] = {
641
# basis parent
642
basis_entry.revision:basis_entry,
643
# this parent
644
change[3].revision:change[3],
645
}
646
else:
647
merged_ids[change[2]] = [change[3].revision]
648
parent_entries[change[2]] = {change[3].revision:change[3]}
649
else:
650
merged_ids[change[2]].append(change[3].revision)
651
parent_entries[change[2]][change[3].revision] = change[3]
652
else:
653
merged_ids = {}
654
# Setup the changes from the tree:
655
# changes maps file_id -> (change, [parent revision_ids])
656
changes= {}
657
for change in iter_changes:
658
# This probably looks up in basis_inv way to much.
659
if change[1][0] is not None:
660
head_candidate = [basis_inv[change[0]].revision]
661
else:
662
head_candidate = []
663
changes[change[0]] = change, merged_ids.get(change[0],
664
head_candidate)
665
unchanged_merged = set(merged_ids) - set(changes)
666
# Extend the changes dict with synthetic changes to record merges of
667
# texts.
668
for file_id in unchanged_merged:
669
# Record a merged version of these items that did not change vs the
670
# basis. This can be either identical parallel changes, or a revert
671
# of a specific file after a merge. The recorded content will be
672
# that of the current tree (which is the same as the basis), but
673
# the per-file graph will reflect a merge.
674
# NB:XXX: We are reconstructing path information we had, this
675
# should be preserved instead.
676
# inv delta change: (file_id, (path_in_source, path_in_target),
677
# changed_content, versioned, parent, name, kind,
678
# executable)
679
try:
680
basis_entry = basis_inv[file_id]
681
except errors.NoSuchId:
682
# a change from basis->some_parents but file_id isn't in basis
683
# so was new in the merge, which means it must have changed
684
# from basis -> current, and as it hasn't the add was reverted
685
# by the user. So we discard this change.
686
pass
687
else:
688
change = (file_id,
689
(basis_inv.id2path(file_id), tree.id2path(file_id)),
690
False, (True, True),
691
(basis_entry.parent_id, basis_entry.parent_id),
692
(basis_entry.name, basis_entry.name),
693
(basis_entry.kind, basis_entry.kind),
694
(basis_entry.executable, basis_entry.executable))
695
changes[file_id] = (change, merged_ids[file_id])
696
# changes contains tuples with the change and a set of inventory
697
# candidates for the file.
698
# inv delta is:
699
# old_path, new_path, file_id, new_inventory_entry
700
seen_root = False # Is the root in the basis delta?
701
inv_delta = self._basis_delta
702
modified_rev = self._new_revision_id
703
for change, head_candidates in changes.values():
704
if change[3][1]: # versioned in target.
705
# Several things may be happening here:
706
# We may have a fork in the per-file graph
707
# - record a change with the content from tree
708
# We may have a change against < all trees
709
# - carry over the tree that hasn't changed
710
# We may have a change against all trees
711
# - record the change with the content from tree
712
kind = change[6][1]
713
file_id = change[0]
714
entry = _entry_factory[kind](file_id, change[5][1],
715
change[4][1])
716
head_set = self._heads(change[0], set(head_candidates))
717
heads = []
718
# Preserve ordering.
719
for head_candidate in head_candidates:
720
if head_candidate in head_set:
721
heads.append(head_candidate)
722
head_set.remove(head_candidate)
723
carried_over = False
724
if len(heads) == 1:
725
# Could be a carry-over situation:
726
parent_entry_revs = parent_entries.get(file_id, None)
727
if parent_entry_revs:
728
parent_entry = parent_entry_revs.get(heads[0], None)
729
else:
730
parent_entry = None
731
if parent_entry is None:
732
# The parent iter_changes was called against is the one
733
# that is the per-file head, so any change is relevant
734
# iter_changes is valid.
735
carry_over_possible = False
736
else:
737
# could be a carry over situation
738
# A change against the basis may just indicate a merge,
739
# we need to check the content against the source of the
740
# merge to determine if it was changed after the merge
741
# or carried over.
742
if (parent_entry.kind != entry.kind or
743
parent_entry.parent_id != entry.parent_id or
744
parent_entry.name != entry.name):
745
# Metadata common to all entries has changed
746
# against per-file parent
747
carry_over_possible = False
748
else:
749
carry_over_possible = True
750
# per-type checks for changes against the parent_entry
751
# are done below.
752
else:
753
# Cannot be a carry-over situation
754
carry_over_possible = False
755
# Populate the entry in the delta
756
if kind == 'file':
757
# XXX: There is still a small race here: If someone reverts the content of a file
758
# after iter_changes examines and decides it has changed,
759
# we will unconditionally record a new version even if some
760
# other process reverts it while commit is running (with
761
# the revert happening after iter_changes did it's
762
# examination).
763
if change[7][1]:
764
entry.executable = True
765
else:
766
entry.executable = False
767
if (carry_over_possible and
768
parent_entry.executable == entry.executable):
769
# Check the file length, content hash after reading
770
# the file.
771
nostore_sha = parent_entry.text_sha1
772
else:
773
nostore_sha = None
774
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
775
try:
776
text = file_obj.read()
777
finally:
778
file_obj.close()
779
try:
780
entry.text_sha1, entry.text_size = self._add_text_to_weave(
781
file_id, text, heads, nostore_sha)
782
yield file_id, change[1][1], (entry.text_sha1, stat_value)
783
except errors.ExistingContent:
784
# No content change against a carry_over parent
785
# Perhaps this should also yield a fs hash update?
786
carried_over = True
787
entry.text_size = parent_entry.text_size
788
entry.text_sha1 = parent_entry.text_sha1
789
elif kind == 'symlink':
790
# Wants a path hint?
791
entry.symlink_target = tree.get_symlink_target(file_id)
792
if (carry_over_possible and
793
parent_entry.symlink_target == entry.symlink_target):
794
carried_over = True
795
else:
796
self._add_text_to_weave(change[0], '', heads, None)
797
elif kind == 'directory':
798
if carry_over_possible:
799
carried_over = True
800
else:
801
# Nothing to set on the entry.
802
# XXX: split into the Root and nonRoot versions.
803
if change[1][1] != '' or self.repository.supports_rich_root():
804
self._add_text_to_weave(change[0], '', heads, None)
805
elif kind == 'tree-reference':
806
if not self.repository._format.supports_tree_reference:
807
# This isn't quite sane as an error, but we shouldn't
808
# ever see this code path in practice: tree's don't
809
# permit references when the repo doesn't support tree
810
# references.
811
raise errors.UnsupportedOperation(tree.add_reference,
812
self.repository)
813
reference_revision = tree.get_reference_revision(change[0])
814
entry.reference_revision = reference_revision
815
if (carry_over_possible and
816
parent_entry.reference_revision == reference_revision):
817
carried_over = True
818
else:
819
self._add_text_to_weave(change[0], '', heads, None)
820
else:
821
raise AssertionError('unknown kind %r' % kind)
822
if not carried_over:
823
entry.revision = modified_rev
824
else:
825
entry.revision = parent_entry.revision
826
else:
827
entry = None
828
new_path = change[1][1]
829
inv_delta.append((change[1][0], new_path, change[0], entry))
830
if new_path == '':
831
seen_root = True
832
self.new_inventory = None
833
if len(inv_delta):
834
# This should perhaps be guarded by a check that the basis we
835
# commit against is the basis for the commit and if not do a delta
836
# against the basis.
837
self._any_changes = True
838
if not seen_root:
839
# housekeeping root entry changes do not affect no-change commits.
840
self._require_root_change(tree)
841
self.basis_delta_revision = basis_revision_id
842
843
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
844
parent_keys = tuple([(file_id, parent) for parent in parents])
845
return self.repository.texts._add_text(
846
(file_id, self._new_revision_id), parent_keys, new_text,
847
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
848
849
850
class RootCommitBuilder(CommitBuilder):
851
"""This commitbuilder actually records the root id"""
852
853
# the root entry gets versioned properly by this builder.
854
_versioned_root = True
855
856
def _check_root(self, ie, parent_invs, tree):
857
"""Helper for record_entry_contents.
858
859
:param ie: An entry being added.
860
:param parent_invs: The inventories of the parent revisions of the
861
commit.
862
:param tree: The tree that is being committed.
863
"""
864
865
def _require_root_change(self, tree):
866
"""Enforce an appropriate root object change.
867
868
This is called once when record_iter_changes is called, if and only if
869
the root was not in the delta calculated by record_iter_changes.
870
871
:param tree: The tree which is being committed.
872
"""
873
# versioned roots do not change unless the tree found a change.
225
raise NotImplementedError(self.record_iter_changes)
874
226
875
227
876
228
class RepositoryWriteLockResult(LogicalLockResult):
901
253
revisions and file history. It's normally accessed only by the Branch,
902
254
which views a particular line of development through that history.
903
255
904
The Repository builds on top of some byte storage facilies (the revisions,
905
signatures, inventories, texts and chk_bytes attributes) and a Transport,
906
which respectively provide byte storage and a means to access the (possibly
907
remote) disk.
908
909
The byte storage facilities are addressed via tuples, which we refer to
910
as 'keys' throughout the code base. Revision_keys, inventory_keys and
911
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
912
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
913
byte string made up of a hash identifier and a hash value.
914
We use this interface because it allows low friction with the underlying
915
code that implements disk indices, network encoding and other parts of
916
bzrlib.
917
918
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
919
the serialised revisions for the repository. This can be used to obtain
920
revision graph information or to access raw serialised revisions.
921
The result of trying to insert data into the repository via this store
922
is undefined: it should be considered read-only except for implementors
923
of repositories.
924
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
925
the serialised signatures for the repository. This can be used to
926
obtain access to raw serialised signatures. The result of trying to
927
insert data into the repository via this store is undefined: it should
928
be considered read-only except for implementors of repositories.
929
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
930
the serialised inventories for the repository. This can be used to
931
obtain unserialised inventories. The result of trying to insert data
932
into the repository via this store is undefined: it should be
933
considered read-only except for implementors of repositories.
934
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
935
texts of files and directories for the repository. This can be used to
936
obtain file texts or file graphs. Note that Repository.iter_file_bytes
937
is usually a better interface for accessing file texts.
938
The result of trying to insert data into the repository via this store
939
is undefined: it should be considered read-only except for implementors
940
of repositories.
941
:ivar chk_bytes: A bzrlib.versionedfile.VersionedFiles instance containing
942
any data the repository chooses to store or have indexed by its hash.
943
The result of trying to insert data into the repository via this store
944
is undefined: it should be considered read-only except for implementors
945
of repositories.
946
:ivar _transport: Transport for file access to repository, typically
947
pointing to .bzr/repository.
256
See VersionedFileRepository in bzrlib.vf_repository for the
257
base class for most Bazaar repositories.
948
258
"""
949
259
950
# What class to use for a CommitBuilder. Often its simpler to change this
951
# in a Repository class subclass rather than to override
952
# get_commit_builder.
953
_commit_builder_class = CommitBuilder
954
# The search regex used by xml based repositories to determine what things
955
# where changed in a single commit.
956
_file_ids_altered_regex = lazy_regex.lazy_compile(
957
r'file_id="(?P<file_id>[^"]+)"'
958
r'.* revision="(?P<revision_id>[^"]+)"'
959
)
960
961
260
def abort_write_group(self, suppress_errors=False):
962
261
"""Commit the contents accrued within the current write group.
963
262
985
284
raise
986
285
mutter('abort_write_group failed')
987
286
log_exception_quietly()
988
note('bzr: ERROR (ignored): %s', exc)
287
note(gettext('bzr: ERROR (ignored): %s'), exc)
989
288
self._write_group = None
990
289
991
290
def _abort_write_group(self):
1006
305
1007
306
:param repository: A repository.
1008
307
"""
1009
if not self._format.supports_external_lookups:
1010
raise errors.UnstackableRepositoryFormat(self._format, self.base)
1011
if self.is_locked():
1012
# This repository will call fallback.unlock() when we transition to
1013
# the unlocked state, so we make sure to increment the lock count
1014
repository.lock_read()
1015
self._check_fallback_repository(repository)
1016
self._fallback_repositories.append(repository)
1017
self.texts.add_fallback_versioned_files(repository.texts)
1018
self.inventories.add_fallback_versioned_files(repository.inventories)
1019
self.revisions.add_fallback_versioned_files(repository.revisions)
1020
self.signatures.add_fallback_versioned_files(repository.signatures)
1021
if self.chk_bytes is not None:
1022
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
308
raise NotImplementedError(self.add_fallback_repository)
1023
309
1024
310
def _check_fallback_repository(self, repository):
1025
311
"""Check that this repository can fallback to repository safely.
1030
316
"""
1031
317
return InterRepository._assert_same_model(self, repository)
1032
318
1033
def add_inventory(self, revision_id, inv, parents):
1034
"""Add the inventory inv to the repository as revision_id.
1035
1036
:param parents: The revision ids of the parents that revision_id
1037
is known to have and are in the repository already.
1038
1039
:returns: The validator(which is a sha1 digest, though what is sha'd is
1040
repository format specific) of the serialized inventory.
1041
"""
1042
if not self.is_in_write_group():
1043
raise AssertionError("%r not in write group" % (self,))
1044
_mod_revision.check_not_reserved_id(revision_id)
1045
if not (inv.revision_id is None or inv.revision_id == revision_id):
1046
raise AssertionError(
1047
"Mismatch between inventory revision"
1048
" id and insertion revid (%r, %r)"
1049
% (inv.revision_id, revision_id))
1050
if inv.root is None:
1051
raise errors.RootMissing()
1052
return self._add_inventory_checked(revision_id, inv, parents)
1053
1054
def _add_inventory_checked(self, revision_id, inv, parents):
1055
"""Add inv to the repository after checking the inputs.
1056
1057
This function can be overridden to allow different inventory styles.
1058
1059
:seealso: add_inventory, for the contract.
1060
"""
1061
inv_lines = self._serializer.write_inventory_to_lines(inv)
1062
return self._inventory_add_lines(revision_id, parents,
1063
inv_lines, check_content=False)
1064
1065
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1066
parents, basis_inv=None, propagate_caches=False):
1067
"""Add a new inventory expressed as a delta against another revision.
1068
1069
See the inventory developers documentation for the theory behind
1070
inventory deltas.
1071
1072
:param basis_revision_id: The inventory id the delta was created
1073
against. (This does not have to be a direct parent.)
1074
:param delta: The inventory delta (see Inventory.apply_delta for
1075
details).
1076
:param new_revision_id: The revision id that the inventory is being
1077
added for.
1078
:param parents: The revision ids of the parents that revision_id is
1079
known to have and are in the repository already. These are supplied
1080
for repositories that depend on the inventory graph for revision
1081
graph access, as well as for those that pun ancestry with delta
1082
compression.
1083
:param basis_inv: The basis inventory if it is already known,
1084
otherwise None.
1085
:param propagate_caches: If True, the caches for this inventory are
1086
copied to and updated for the result if possible.
1087
1088
:returns: (validator, new_inv)
1089
The validator(which is a sha1 digest, though what is sha'd is
1090
repository format specific) of the serialized inventory, and the
1091
resulting inventory.
1092
"""
1093
if not self.is_in_write_group():
1094
raise AssertionError("%r not in write group" % (self,))
1095
_mod_revision.check_not_reserved_id(new_revision_id)
1096
basis_tree = self.revision_tree(basis_revision_id)
1097
basis_tree.lock_read()
1098
try:
1099
# Note that this mutates the inventory of basis_tree, which not all
1100
# inventory implementations may support: A better idiom would be to
1101
# return a new inventory, but as there is no revision tree cache in
1102
# repository this is safe for now - RBC 20081013
1103
if basis_inv is None:
1104
basis_inv = basis_tree.inventory
1105
basis_inv.apply_delta(delta)
1106
basis_inv.revision_id = new_revision_id
1107
return (self.add_inventory(new_revision_id, basis_inv, parents),
1108
basis_inv)
1109
finally:
1110
basis_tree.unlock()
1111
1112
def _inventory_add_lines(self, revision_id, parents, lines,
1113
check_content=True):
1114
"""Store lines in inv_vf and return the sha1 of the inventory."""
1115
parents = [(parent,) for parent in parents]
1116
result = self.inventories.add_lines((revision_id,), parents, lines,
1117
check_content=check_content)[0]
1118
self.inventories._access.flush()
1119
return result
1120
1121
def add_revision(self, revision_id, rev, inv=None, config=None):
1122
"""Add rev to the revision store as revision_id.
1123
1124
:param revision_id: the revision id to use.
1125
:param rev: The revision object.
1126
:param inv: The inventory for the revision. if None, it will be looked
1127
up in the inventory storer
1128
:param config: If None no digital signature will be created.
1129
If supplied its signature_needed method will be used
1130
to determine if a signature should be made.
1131
"""
1132
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
1133
# rev.parent_ids?
1134
_mod_revision.check_not_reserved_id(revision_id)
1135
if config is not None and config.signature_needed():
1136
if inv is None:
1137
inv = self.get_inventory(revision_id)
1138
plaintext = Testament(rev, inv).as_short_text()
1139
self.store_revision_signature(
1140
gpg.GPGStrategy(config), plaintext, revision_id)
1141
# check inventory present
1142
if not self.inventories.get_parent_map([(revision_id,)]):
1143
if inv is None:
1144
raise errors.WeaveRevisionNotPresent(revision_id,
1145
self.inventories)
1146
else:
1147
# yes, this is not suitable for adding with ghosts.
1148
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
1149
rev.parent_ids)
1150
else:
1151
key = (revision_id,)
1152
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
1153
self._add_revision(rev)
1154
1155
def _add_revision(self, revision):
1156
text = self._serializer.write_revision_to_string(revision)
1157
key = (revision.revision_id,)
1158
parents = tuple((parent,) for parent in revision.parent_ids)
1159
self.revisions.add_lines(key, parents, osutils.split_lines(text))
1160
1161
319
def all_revision_ids(self):
1162
320
"""Returns a list of all the revision ids in the repository.
1163
321
1186
344
"""
1187
345
self.control_files.break_lock()
1188
346
1189
@needs_read_lock
1190
def _eliminate_revisions_not_present(self, revision_ids):
1191
"""Check every revision id in revision_ids to see if we have it.
1192
1193
Returns a set of the present revisions.
1194
"""
1195
result = []
1196
graph = self.get_graph()
1197
parent_map = graph.get_parent_map(revision_ids)
1198
# The old API returned a list, should this actually be a set?
1199
return parent_map.keys()
1200
1201
def _check_inventories(self, checker):
1202
"""Check the inventories found from the revision scan.
1203
1204
This is responsible for verifying the sha1 of inventories and
1205
creating a pending_keys set that covers data referenced by inventories.
1206
"""
1207
bar = ui.ui_factory.nested_progress_bar()
1208
try:
1209
self._do_check_inventories(checker, bar)
1210
finally:
1211
bar.finished()
1212
1213
def _do_check_inventories(self, checker, bar):
1214
"""Helper for _check_inventories."""
1215
revno = 0
1216
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1217
kinds = ['chk_bytes', 'texts']
1218
count = len(checker.pending_keys)
1219
bar.update("inventories", 0, 2)
1220
current_keys = checker.pending_keys
1221
checker.pending_keys = {}
1222
# Accumulate current checks.
1223
for key in current_keys:
1224
if key[0] != 'inventories' and key[0] not in kinds:
1225
checker._report_items.append('unknown key type %r' % (key,))
1226
keys[key[0]].add(key[1:])
1227
if keys['inventories']:
1228
# NB: output order *should* be roughly sorted - topo or
1229
# inverse topo depending on repository - either way decent
1230
# to just delta against. However, pre-CHK formats didn't
1231
# try to optimise inventory layout on disk. As such the
1232
# pre-CHK code path does not use inventory deltas.
1233
last_object = None
1234
for record in self.inventories.check(keys=keys['inventories']):
1235
if record.storage_kind == 'absent':
1236
checker._report_items.append(
1237
'Missing inventory {%s}' % (record.key,))
1238
else:
1239
last_object = self._check_record('inventories', record,
1240
checker, last_object,
1241
current_keys[('inventories',) + record.key])
1242
del keys['inventories']
1243
else:
1244
return
1245
bar.update("texts", 1)
1246
while (checker.pending_keys or keys['chk_bytes']
1247
or keys['texts']):
1248
# Something to check.
1249
current_keys = checker.pending_keys
1250
checker.pending_keys = {}
1251
# Accumulate current checks.
1252
for key in current_keys:
1253
if key[0] not in kinds:
1254
checker._report_items.append('unknown key type %r' % (key,))
1255
keys[key[0]].add(key[1:])
1256
# Check the outermost kind only - inventories || chk_bytes || texts
1257
for kind in kinds:
1258
if keys[kind]:
1259
last_object = None
1260
for record in getattr(self, kind).check(keys=keys[kind]):
1261
if record.storage_kind == 'absent':
1262
checker._report_items.append(
1263
'Missing %s {%s}' % (kind, record.key,))
1264
else:
1265
last_object = self._check_record(kind, record,
1266
checker, last_object, current_keys[(kind,) + record.key])
1267
keys[kind] = set()
1268
break
1269
1270
def _check_record(self, kind, record, checker, last_object, item_data):
1271
"""Check a single text from this repository."""
1272
if kind == 'inventories':
1273
rev_id = record.key[0]
1274
inv = self._deserialise_inventory(rev_id,
1275
record.get_bytes_as('fulltext'))
1276
if last_object is not None:
1277
delta = inv._make_delta(last_object)
1278
for old_path, path, file_id, ie in delta:
1279
if ie is None:
1280
continue
1281
ie.check(checker, rev_id, inv)
1282
else:
1283
for path, ie in inv.iter_entries():
1284
ie.check(checker, rev_id, inv)
1285
if self._format.fast_deltas:
1286
return inv
1287
elif kind == 'chk_bytes':
1288
# No code written to check chk_bytes for this repo format.
1289
checker._report_items.append(
1290
'unsupported key type chk_bytes for %s' % (record.key,))
1291
elif kind == 'texts':
1292
self._check_text(record, checker, item_data)
1293
else:
1294
checker._report_items.append(
1295
'unknown key type %s for %s' % (kind, record.key))
1296
1297
def _check_text(self, record, checker, item_data):
1298
"""Check a single text."""
1299
# Check it is extractable.
1300
# TODO: check length.
1301
if record.storage_kind == 'chunked':
1302
chunks = record.get_bytes_as(record.storage_kind)
1303
sha1 = osutils.sha_strings(chunks)
1304
length = sum(map(len, chunks))
1305
else:
1306
content = record.get_bytes_as('fulltext')
1307
sha1 = osutils.sha_string(content)
1308
length = len(content)
1309
if item_data and sha1 != item_data[1]:
1310
checker._report_items.append(
1311
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1312
(record.key, sha1, item_data[1], item_data[2]))
1313
1314
347
@staticmethod
1315
def create(a_bzrdir):
1316
"""Construct the current default format repository in a_bzrdir."""
1317
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)
1318
351
1319
def __init__(self, _format, a_bzrdir, control_files):
352
def __init__(self, _format, controldir, control_files):
1320
353
"""instantiate a Repository.
1321
354
1322
355
:param _format: The format of the repository on disk.
1323
: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.
1324
358
"""
1325
359
# In the future we will have a single api for all stores for
1326
360
# getting file texts, inventories and revisions, then
1328
362
super(Repository, self).__init__()
1329
363
self._format = _format
1330
364
# the following are part of the public API for Repository:
1331
self.bzrdir = a_bzrdir
365
self.bzrdir = controldir
1332
366
self.control_files = control_files
1333
self._transport = control_files._transport
1334
self.base = self._transport.base
1335
367
# for tests
1336
self._reconcile_does_inventory_gc = True
1337
self._reconcile_fixes_text_parents = False
1338
self._reconcile_backsup_inventory = True
1339
368
self._write_group = None
1340
369
# Additional places to query for data.
1341
370
self._fallback_repositories = []
1342
# An InventoryEntry cache, used during deserialization
1343
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1344
# Is it safe to return inventory entries directly from the entry cache,
1345
# rather copying them?
1346
self._safe_to_return_from_cache = False
1347
371
1348
372
@property
1349
373
def user_transport(self):
1383
407
"""
1384
408
if self.__class__ is not other.__class__:
1385
409
return False
1386
return (self._transport.base == other._transport.base)
410
return (self.control_url == other.control_url)
1387
411
1388
412
def is_in_write_group(self):
1389
413
"""Return True if there is an open write group.
1489
513
if revid and committers:
1490
514
result['committers'] = 0
1491
515
if revid and revid != _mod_revision.NULL_REVISION:
516
graph = self.get_graph()
1492
517
if committers:
1493
518
all_committers = set()
1494
revisions = self.get_ancestry(revid)
1495
# pop the leading None
1496
revisions.pop(0)
1497
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
1498
522
if not committers:
1499
523
# ignore the revisions in the middle - just grab first and last
1500
524
revisions = revisions[0], revisions[-1]
1501
525
for revision in self.get_revisions(revisions):
1502
if not first_revision:
1503
first_revision = revision
526
if not last_revision:
527
last_revision = revision
1504
528
if committers:
1505
529
all_committers.add(revision.committer)
1506
last_revision = revision
530
first_revision = revision
1507
531
if committers:
1508
532
result['committers'] = len(all_committers)
1509
533
result['firstrev'] = (first_revision.timestamp,
1510
534
first_revision.timezone)
1511
535
result['latestrev'] = (last_revision.timestamp,
1512
536
last_revision.timezone)
1513
1514
# now gather global repository information
1515
# XXX: This is available for many repos regardless of listability.
1516
if self.user_transport.listable():
1517
# XXX: do we want to __define len__() ?
1518
# Maybe the versionedfiles object should provide a different
1519
# method to get the number of keys.
1520
result['revisions'] = len(self.revisions.keys())
1521
# result['size'] = t
1522
537
return result
1523
538
1524
539
def find_branches(self, using=False):
1535
550
def __init__(self):
1536
551
self.first_call = True
1537
552
1538
def __call__(self, bzrdir):
1539
# 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
1540
555
# containing the current repo.
1541
556
if not self.first_call:
1542
557
try:
1543
repository = bzrdir.open_repository()
558
repository = controldir.open_repository()
1544
559
except errors.NoRepositoryPresent:
1545
560
pass
1546
561
else:
1547
562
return False, ([], repository)
1548
563
self.first_call = False
1549
value = (bzrdir.list_branches(), None)
564
value = (controldir.list_branches(), None)
1550
565
return True, value
1551
566
1552
567
ret = []
1553
for branches, repository in bzrdir.BzrDir.find_bzrdirs(
568
for branches, repository in controldir.ControlDir.find_bzrdirs(
1554
569
self.user_transport, evaluate=Evaluator()):
1555
570
if branches is not None:
1556
571
ret.extend(branches)
1559
574
return ret
1560
575
1561
576
@needs_read_lock
1562
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):
1563
581
"""Return the revision ids that other has that this does not.
1564
582
1565
583
These are returned in topological order.
1566
584
1567
585
revision_id: only return revision ids included by revision_id.
1568
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]
1569
597
return InterRepository.get(other, self).search_missing_revision_ids(
1570
revision_id, find_ghosts)
598
find_ghosts=find_ghosts, revision_ids=revision_ids,
599
if_present_ids=if_present_ids, limit=limit)
1571
600
1572
601
@staticmethod
1573
602
def open(base):
1576
605
For instance, if the repository is at URL/.bzr/repository,
1577
606
Repository.open(URL) -> a Repository instance.
1578
607
"""
1579
control = bzrdir.BzrDir.open(base)
608
control = controldir.ControlDir.open(base)
1580
609
return control.open_repository()
1581
610
1582
611
def copy_content_into(self, destination, revision_id=None):
1613
642
"""
1614
643
1615
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
"""
1616
651
raise errors.UnsuspendableWriteGroup(self)
1617
652
1618
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1619
"""Return the keys of missing inventory parents for revisions added in
1620
this write group.
1621
1622
A revision is not complete if the inventory delta for that revision
1623
cannot be calculated. Therefore if the parent inventories of a
1624
revision are not present, the revision is incomplete, and e.g. cannot
1625
be streamed by a smart server. This method finds missing inventory
1626
parents for revisions added in this write group.
1627
"""
1628
if not self._format.supports_external_lookups:
1629
# This is only an issue for stacked repositories
1630
return set()
1631
if not self.is_in_write_group():
1632
raise AssertionError('not in a write group')
1633
1634
# XXX: We assume that every added revision already has its
1635
# corresponding inventory, so we only check for parent inventories that
1636
# might be missing, rather than all inventories.
1637
parents = set(self.revisions._index.get_missing_parents())
1638
parents.discard(_mod_revision.NULL_REVISION)
1639
unstacked_inventories = self.inventories._index
1640
present_inventories = unstacked_inventories.get_parent_map(
1641
key[-1:] for key in parents)
1642
parents.difference_update(present_inventories)
1643
if len(parents) == 0:
1644
# No missing parent inventories.
1645
return set()
1646
if not check_for_missing_texts:
1647
return set(('inventories', rev_id) for (rev_id,) in parents)
1648
# Ok, now we have a list of missing inventories. But these only matter
1649
# if the inventories that reference them are missing some texts they
1650
# appear to introduce.
1651
# XXX: Texts referenced by all added inventories need to be present,
1652
# but at the moment we're only checking for texts referenced by
1653
# inventories at the graph's edge.
1654
key_deps = self.revisions._index._key_dependencies
1655
key_deps.satisfy_refs_for_keys(present_inventories)
1656
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1657
file_ids = self.fileids_altered_by_revision_ids(referrers)
1658
missing_texts = set()
1659
for file_id, version_ids in file_ids.iteritems():
1660
missing_texts.update(
1661
(file_id, version_id) for version_id in version_ids)
1662
present_texts = self.texts.get_parent_map(missing_texts)
1663
missing_texts.difference_update(present_texts)
1664
if not missing_texts:
1665
# No texts are missing, so all revisions and their deltas are
1666
# reconstructable.
1667
return set()
1668
# Alternatively the text versions could be returned as the missing
1669
# keys, but this is likely to be less data.
1670
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1671
return missing_keys
1672
1673
653
def refresh_data(self):
1674
654
"""Re-read any data needed to synchronise with disk.
1675
655
1695
675
def _resume_write_group(self, tokens):
1696
676
raise errors.UnsuspendableWriteGroup(self)
1697
677
1698
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
1699
fetch_spec=None):
678
def fetch(self, source, revision_id=None, find_ghosts=False):
1700
679
"""Fetch the content required to construct revision_id from source.
1701
680
1702
If revision_id is None and fetch_spec is None, then all content is
1703
copied.
681
If revision_id is None, then all content is copied.
1704
682
1705
683
fetch() may not be used when the repository is in a write group -
1706
684
either finish the current write group before using fetch, or use
1712
690
:param revision_id: If specified, all the content needed for this
1713
691
revision ID will be copied to the target. Fetch will determine for
1714
692
itself which content needs to be copied.
1715
:param fetch_spec: If specified, a SearchResult or
1716
PendingAncestryResult that describes which revisions to copy. This
1717
allows copying multiple heads at once. Mutually exclusive with
1718
revision_id.
1719
693
"""
1720
if fetch_spec is not None and revision_id is not None:
1721
raise AssertionError(
1722
"fetch_spec and revision_id are mutually exclusive.")
1723
694
if self.is_in_write_group():
1724
695
raise errors.InternalBzrError(
1725
696
"May not fetch while in a write group.")
1727
698
# TODO: lift out to somewhere common with RemoteRepository
1728
699
# <https://bugs.launchpad.net/bzr/+bug/401646>
1729
700
if (self.has_same_location(source)
1730
and fetch_spec is None
1731
701
and self._has_same_fallbacks(source)):
1732
702
# check that last_revision is in 'from' and then return a
1733
703
# no-operation.
1735
705
not _mod_revision.is_null(revision_id)):
1736
706
self.get_revision(revision_id)
1737
707
return 0, []
1738
# if there is no specific appropriate InterRepository, this will get
1739
# the InterRepository base class, which raises an
1740
# IncompatibleRepositories when asked to fetch.
1741
708
inter = InterRepository.get(source, self)
1742
return inter.fetch(revision_id=revision_id, pb=pb,
1743
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
709
return inter.fetch(revision_id=revision_id, find_ghosts=find_ghosts)
1744
710
1745
711
def create_bundle(self, target, base, fileobj, format=None):
1746
712
return serializer.write_bundle(self, target, base, fileobj, format)
1747
713
1748
714
def get_commit_builder(self, branch, parents, config, timestamp=None,
1749
715
timezone=None, committer=None, revprops=None,
1750
revision_id=None):
716
revision_id=None, lossy=False):
1751
717
"""Obtain a CommitBuilder for this repository.
1752
718
1753
719
:param branch: Branch to commit to.
1758
724
:param committer: Optional committer to set for commit.
1759
725
:param revprops: Optional dictionary of revision properties.
1760
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
1761
729
"""
1762
if self._fallback_repositories:
1763
raise errors.BzrError("Cannot commit from a lightweight checkout "
1764
"to a stacked branch. See "
1765
"https://bugs.launchpad.net/bzr/+bug/375013 for details.")
1766
result = self._commit_builder_class(self, parents, config,
1767
timestamp, timezone, committer, revprops, revision_id)
1768
self.start_write_group()
1769
return result
730
raise NotImplementedError(self.get_commit_builder)
1770
731
1771
732
@only_raises(errors.LockNotHeld, errors.LockBroken)
1772
733
def unlock(self):
1779
740
'Must end write groups before releasing write locks.')
1780
741
self.control_files.unlock()
1781
742
if self.control_files._lock_count == 0:
1782
self._inventory_entry_cache.clear()
1783
743
for repo in self._fallback_repositories:
1784
744
repo.unlock()
1785
745
1786
746
@needs_read_lock
1787
def clone(self, a_bzrdir, revision_id=None):
1788
"""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.
1789
749
1790
750
Currently no check is made that the format of this repository and
1791
751
the bzrdir format are compatible. FIXME RBC 20060201.
1794
754
"""
1795
755
# TODO: deprecate after 0.16; cloning this with all its settings is
1796
756
# probably not very useful -- mbp 20070423
1797
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())
1798
759
self.copy_content_into(dest_repo, revision_id)
1799
760
return dest_repo
1800
761
1850
811
dest_repo = a_bzrdir.open_repository()
1851
812
return dest_repo
1852
813
1853
def _get_sink(self):
1854
"""Return a sink for streaming into this repository."""
1855
return StreamSink(self)
1856
1857
def _get_source(self, to_format):
1858
"""Return a source for streaming from this repository."""
1859
return StreamSource(self, to_format)
1860
1861
814
@needs_read_lock
1862
815
def has_revision(self, revision_id):
1863
816
"""True if this repository has a copy of the revision."""
1870
823
:param revision_ids: An iterable of revision_ids.
1871
824
:return: A set of the revision_ids that were present.
1872
825
"""
1873
parent_map = self.revisions.get_parent_map(
1874
[(rev_id,) for rev_id in revision_ids])
1875
result = set()
1876
if _mod_revision.NULL_REVISION in revision_ids:
1877
result.add(_mod_revision.NULL_REVISION)
1878
result.update([key[0] for key in parent_map])
1879
return result
826
raise NotImplementedError(self.has_revisions)
1880
827
1881
828
@needs_read_lock
1882
829
def get_revision(self, revision_id):
1883
830
"""Return the Revision object for a named revision."""
1884
831
return self.get_revisions([revision_id])[0]
1885
832
1886
@needs_read_lock
1887
833
def get_revision_reconcile(self, revision_id):
1888
834
"""'reconcile' helper routine that allows access to a revision always.
1889
835
1892
838
be used by reconcile, or reconcile-alike commands that are correcting
1893
839
or testing the revision graph.
1894
840
"""
1895
return self._get_revisions([revision_id])[0]
841
raise NotImplementedError(self.get_revision_reconcile)
1896
842
1897
@needs_read_lock
1898
843
def get_revisions(self, revision_ids):
1899
844
"""Get many revisions at once.
1900
845
1901
846
Repositories that need to check data on every revision read should
1902
847
subclass this method.
1903
848
"""
1904
return self._get_revisions(revision_ids)
1905
1906
@needs_read_lock
1907
def _get_revisions(self, revision_ids):
1908
"""Core work logic to get many revisions without sanity checks."""
1909
revs = {}
1910
for revid, rev in self._iter_revisions(revision_ids):
1911
if rev is None:
1912
raise errors.NoSuchRevision(self, revid)
1913
revs[revid] = rev
1914
return [revs[revid] for revid in revision_ids]
1915
1916
def _iter_revisions(self, revision_ids):
1917
"""Iterate over revision objects.
1918
1919
:param revision_ids: An iterable of revisions to examine. None may be
1920
passed to request all revisions known to the repository. Note that
1921
not all repositories can find unreferenced revisions; for those
1922
repositories only referenced ones will be returned.
1923
:return: An iterator of (revid, revision) tuples. Absent revisions (
1924
those asked for but not available) are returned as (revid, None).
1925
"""
1926
if revision_ids is None:
1927
revision_ids = self.all_revision_ids()
1928
else:
1929
for rev_id in revision_ids:
1930
if not rev_id or not isinstance(rev_id, basestring):
1931
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1932
keys = [(key,) for key in revision_ids]
1933
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1934
for record in stream:
1935
revid = record.key[0]
1936
if record.storage_kind == 'absent':
1937
yield (revid, None)
1938
else:
1939
text = record.get_bytes_as('fulltext')
1940
rev = self._serializer.read_revision_from_string(text)
1941
yield (revid, rev)
849
raise NotImplementedError(self.get_revisions)
1942
850
1943
851
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1944
852
"""Produce a generator of revision deltas.
1999
907
signature = gpg_strategy.sign(plaintext)
2000
908
self.add_signature_text(revision_id, signature)
2001
909
2002
@needs_write_lock
2003
910
def add_signature_text(self, revision_id, signature):
2004
self.signatures.add_lines((revision_id,), (),
2005
osutils.split_lines(signature))
2006
2007
def find_text_key_references(self):
2008
"""Find the text key references within the repository.
2009
2010
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2011
to whether they were referred to by the inventory of the
2012
revision_id that they contain. The inventory texts from all present
2013
revision ids are assessed to generate this report.
2014
"""
2015
revision_keys = self.revisions.keys()
2016
w = self.inventories
2017
pb = ui.ui_factory.nested_progress_bar()
2018
try:
2019
return self._find_text_key_references_from_xml_inventory_lines(
2020
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
2021
finally:
2022
pb.finished()
2023
2024
def _find_text_key_references_from_xml_inventory_lines(self,
2025
line_iterator):
2026
"""Core routine for extracting references to texts from inventories.
2027
2028
This performs the translation of xml lines to revision ids.
2029
2030
:param line_iterator: An iterator of lines, origin_version_id
2031
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2032
to whether they were referred to by the inventory of the
2033
revision_id that they contain. Note that if that revision_id was
2034
not part of the line_iterator's output then False will be given -
2035
even though it may actually refer to that key.
2036
"""
2037
if not self._serializer.support_altered_by_hack:
2038
raise AssertionError(
2039
"_find_text_key_references_from_xml_inventory_lines only "
2040
"supported for branches which store inventory as unnested xml"
2041
", not on %r" % self)
2042
result = {}
2043
2044
# this code needs to read every new line in every inventory for the
2045
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
2046
# not present in one of those inventories is unnecessary but not
2047
# harmful because we are filtering by the revision id marker in the
2048
# inventory lines : we only select file ids altered in one of those
2049
# revisions. We don't need to see all lines in the inventory because
2050
# only those added in an inventory in rev X can contain a revision=X
2051
# line.
2052
unescape_revid_cache = {}
2053
unescape_fileid_cache = {}
2054
2055
# jam 20061218 In a big fetch, this handles hundreds of thousands
2056
# of lines, so it has had a lot of inlining and optimizing done.
2057
# Sorry that it is a little bit messy.
2058
# Move several functions to be local variables, since this is a long
2059
# running loop.
2060
search = self._file_ids_altered_regex.search
2061
unescape = _unescape_xml
2062
setdefault = result.setdefault
2063
for line, line_key in line_iterator:
2064
match = search(line)
2065
if match is None:
2066
continue
2067
# One call to match.group() returning multiple items is quite a
2068
# bit faster than 2 calls to match.group() each returning 1
2069
file_id, revision_id = match.group('file_id', 'revision_id')
2070
2071
# Inlining the cache lookups helps a lot when you make 170,000
2072
# lines and 350k ids, versus 8.4 unique ids.
2073
# Using a cache helps in 2 ways:
2074
# 1) Avoids unnecessary decoding calls
2075
# 2) Re-uses cached strings, which helps in future set and
2076
# equality checks.
2077
# (2) is enough that removing encoding entirely along with
2078
# the cache (so we are using plain strings) results in no
2079
# performance improvement.
2080
try:
2081
revision_id = unescape_revid_cache[revision_id]
2082
except KeyError:
2083
unescaped = unescape(revision_id)
2084
unescape_revid_cache[revision_id] = unescaped
2085
revision_id = unescaped
2086
2087
# Note that unconditionally unescaping means that we deserialise
2088
# every fileid, which for general 'pull' is not great, but we don't
2089
# really want to have some many fulltexts that this matters anyway.
2090
# RBC 20071114.
2091
try:
2092
file_id = unescape_fileid_cache[file_id]
2093
except KeyError:
2094
unescaped = unescape(file_id)
2095
unescape_fileid_cache[file_id] = unescaped
2096
file_id = unescaped
2097
2098
key = (file_id, revision_id)
2099
setdefault(key, False)
2100
if revision_id == line_key[-1]:
2101
result[key] = True
2102
return result
2103
2104
def _inventory_xml_lines_for_keys(self, keys):
2105
"""Get a line iterator of the sort needed for findind references.
2106
2107
Not relevant for non-xml inventory repositories.
2108
2109
Ghosts in revision_keys are ignored.
2110
2111
:param revision_keys: The revision keys for the inventories to inspect.
2112
:return: An iterator over (inventory line, revid) for the fulltexts of
2113
all of the xml inventories specified by revision_keys.
2114
"""
2115
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2116
for record in stream:
2117
if record.storage_kind != 'absent':
2118
chunks = record.get_bytes_as('chunked')
2119
revid = record.key[-1]
2120
lines = osutils.chunks_to_lines(chunks)
2121
for line in lines:
2122
yield line, revid
2123
2124
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2125
revision_keys):
2126
"""Helper routine for fileids_altered_by_revision_ids.
2127
2128
This performs the translation of xml lines to revision ids.
2129
2130
:param line_iterator: An iterator of lines, origin_version_id
2131
:param revision_keys: The revision ids to filter for. This should be a
2132
set or other type which supports efficient __contains__ lookups, as
2133
the revision key from each parsed line will be looked up in the
2134
revision_keys filter.
2135
:return: a dictionary mapping altered file-ids to an iterable of
2136
revision_ids. Each altered file-ids has the exact revision_ids that
2137
altered it listed explicitly.
2138
"""
2139
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2140
line_iterator).iterkeys())
2141
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2142
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2143
self._inventory_xml_lines_for_keys(parent_keys)))
2144
new_keys = seen - parent_seen
2145
result = {}
2146
setdefault = result.setdefault
2147
for key in new_keys:
2148
setdefault(key[0], set()).add(key[-1])
2149
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)
2150
917
2151
918
def _find_parent_ids_of_revisions(self, revision_ids):
2152
919
"""Find all parent ids that are mentioned in the revision graph.
2161
928
parent_ids.discard(_mod_revision.NULL_REVISION)
2162
929
return parent_ids
2163
930
2164
def _find_parent_keys_of_revisions(self, revision_keys):
2165
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2166
2167
:param revision_keys: An iterable of revision_keys.
2168
:return: The parents of all revision_keys that are not already in
2169
revision_keys
2170
"""
2171
parent_map = self.revisions.get_parent_map(revision_keys)
2172
parent_keys = set()
2173
map(parent_keys.update, parent_map.itervalues())
2174
parent_keys.difference_update(revision_keys)
2175
parent_keys.discard(_mod_revision.NULL_REVISION)
2176
return parent_keys
2177
2178
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
2179
"""Find the file ids and versions affected by revisions.
2180
2181
:param revisions: an iterable containing revision ids.
2182
:param _inv_weave: The inventory weave from this repository or None.
2183
If None, the inventory weave will be opened automatically.
2184
:return: a dictionary mapping altered file-ids to an iterable of
2185
revision_ids. Each altered file-ids has the exact revision_ids that
2186
altered it listed explicitly.
2187
"""
2188
selected_keys = set((revid,) for revid in revision_ids)
2189
w = _inv_weave or self.inventories
2190
return self._find_file_ids_from_xml_inventory_lines(
2191
w.iter_lines_added_or_present_in_keys(
2192
selected_keys, pb=None),
2193
selected_keys)
2194
2195
931
def iter_files_bytes(self, desired_files):
2196
932
"""Iterate through file versions.
2197
933
2203
939
uniquely identify the file version in the caller's context. (Examples:
2204
940
an index number or a TreeTransform trans_id.)
2205
941
2206
bytes_iterator is an iterable of bytestrings for the file. The
2207
kind of iterable and length of the bytestrings are unspecified, but for
2208
this implementation, it is a list of bytes produced by
2209
VersionedFile.get_record_stream().
2210
2211
942
:param desired_files: a list of (file_id, revision_id, identifier)
2212
943
triples
2213
944
"""
2214
text_keys = {}
2215
for file_id, revision_id, callable_data in desired_files:
2216
text_keys[(file_id, revision_id)] = callable_data
2217
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2218
if record.storage_kind == 'absent':
2219
raise errors.RevisionNotPresent(record.key, self)
2220
yield text_keys[record.key], record.get_bytes_as('chunked')
2221
2222
def _generate_text_key_index(self, text_key_references=None,
2223
ancestors=None):
2224
"""Generate a new text key index for the repository.
2225
2226
This is an expensive function that will take considerable time to run.
2227
2228
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2229
list of parents, also text keys. When a given key has no parents,
2230
the parents list will be [NULL_REVISION].
2231
"""
2232
# All revisions, to find inventory parents.
2233
if ancestors is None:
2234
graph = self.get_graph()
2235
ancestors = graph.get_parent_map(self.all_revision_ids())
2236
if text_key_references is None:
2237
text_key_references = self.find_text_key_references()
2238
pb = ui.ui_factory.nested_progress_bar()
2239
try:
2240
return self._do_generate_text_key_index(ancestors,
2241
text_key_references, pb)
2242
finally:
2243
pb.finished()
2244
2245
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2246
"""Helper for _generate_text_key_index to avoid deep nesting."""
2247
revision_order = tsort.topo_sort(ancestors)
2248
invalid_keys = set()
2249
revision_keys = {}
2250
for revision_id in revision_order:
2251
revision_keys[revision_id] = set()
2252
text_count = len(text_key_references)
2253
# a cache of the text keys to allow reuse; costs a dict of all the
2254
# keys, but saves a 2-tuple for every child of a given key.
2255
text_key_cache = {}
2256
for text_key, valid in text_key_references.iteritems():
2257
if not valid:
2258
invalid_keys.add(text_key)
2259
else:
2260
revision_keys[text_key[1]].add(text_key)
2261
text_key_cache[text_key] = text_key
2262
del text_key_references
2263
text_index = {}
2264
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2265
NULL_REVISION = _mod_revision.NULL_REVISION
2266
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2267
# too small for large or very branchy trees. However, for 55K path
2268
# trees, it would be easy to use too much memory trivially. Ideally we
2269
# could gauge this by looking at available real memory etc, but this is
2270
# always a tricky proposition.
2271
inventory_cache = lru_cache.LRUCache(10)
2272
batch_size = 10 # should be ~150MB on a 55K path tree
2273
batch_count = len(revision_order) / batch_size + 1
2274
processed_texts = 0
2275
pb.update("Calculating text parents", processed_texts, text_count)
2276
for offset in xrange(batch_count):
2277
to_query = revision_order[offset * batch_size:(offset + 1) *
2278
batch_size]
2279
if not to_query:
2280
break
2281
for revision_id in to_query:
2282
parent_ids = ancestors[revision_id]
2283
for text_key in revision_keys[revision_id]:
2284
pb.update("Calculating text parents", processed_texts)
2285
processed_texts += 1
2286
candidate_parents = []
2287
for parent_id in parent_ids:
2288
parent_text_key = (text_key[0], parent_id)
2289
try:
2290
check_parent = parent_text_key not in \
2291
revision_keys[parent_id]
2292
except KeyError:
2293
# the parent parent_id is a ghost:
2294
check_parent = False
2295
# truncate the derived graph against this ghost.
2296
parent_text_key = None
2297
if check_parent:
2298
# look at the parent commit details inventories to
2299
# determine possible candidates in the per file graph.
2300
# TODO: cache here.
2301
try:
2302
inv = inventory_cache[parent_id]
2303
except KeyError:
2304
inv = self.revision_tree(parent_id).inventory
2305
inventory_cache[parent_id] = inv
2306
try:
2307
parent_entry = inv[text_key[0]]
2308
except (KeyError, errors.NoSuchId):
2309
parent_entry = None
2310
if parent_entry is not None:
2311
parent_text_key = (
2312
text_key[0], parent_entry.revision)
2313
else:
2314
parent_text_key = None
2315
if parent_text_key is not None:
2316
candidate_parents.append(
2317
text_key_cache[parent_text_key])
2318
parent_heads = text_graph.heads(candidate_parents)
2319
new_parents = list(parent_heads)
2320
new_parents.sort(key=lambda x:candidate_parents.index(x))
2321
if new_parents == []:
2322
new_parents = [NULL_REVISION]
2323
text_index[text_key] = new_parents
2324
2325
for text_key in invalid_keys:
2326
text_index[text_key] = [NULL_REVISION]
2327
return text_index
2328
2329
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2330
"""Get an iterable listing the keys of all the data introduced by a set
2331
of revision IDs.
2332
2333
The keys will be ordered so that the corresponding items can be safely
2334
fetched and inserted in that order.
2335
2336
:returns: An iterable producing tuples of (knit-kind, file-id,
2337
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2338
'revisions'. file-id is None unless knit-kind is 'file'.
2339
"""
2340
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2341
yield result
2342
del _files_pb
2343
for result in self._find_non_file_keys_to_fetch(revision_ids):
2344
yield result
2345
2346
def _find_file_keys_to_fetch(self, revision_ids, pb):
2347
# XXX: it's a bit weird to control the inventory weave caching in this
2348
# generator. Ideally the caching would be done in fetch.py I think. Or
2349
# maybe this generator should explicitly have the contract that it
2350
# should not be iterated until the previously yielded item has been
2351
# processed?
2352
inv_w = self.inventories
2353
2354
# file ids that changed
2355
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2356
count = 0
2357
num_file_ids = len(file_ids)
2358
for file_id, altered_versions in file_ids.iteritems():
2359
if pb is not None:
2360
pb.update("Fetch texts", count, num_file_ids)
2361
count += 1
2362
yield ("file", file_id, altered_versions)
2363
2364
def _find_non_file_keys_to_fetch(self, revision_ids):
2365
# inventory
2366
yield ("inventory", None, revision_ids)
2367
2368
# signatures
2369
# XXX: Note ATM no callers actually pay attention to this return
2370
# instead they just use the list of revision ids and ignore
2371
# missing sigs. Consider removing this work entirely
2372
revisions_with_signatures = set(self.signatures.get_parent_map(
2373
[(r,) for r in revision_ids]))
2374
revisions_with_signatures = set(
2375
[r for (r,) in revisions_with_signatures])
2376
revisions_with_signatures.intersection_update(revision_ids)
2377
yield ("signatures", None, revisions_with_signatures)
2378
2379
# revisions
2380
yield ("revisions", None, revision_ids)
2381
2382
@needs_read_lock
2383
def get_inventory(self, revision_id):
2384
"""Get Inventory object by revision id."""
2385
return self.iter_inventories([revision_id]).next()
2386
2387
def iter_inventories(self, revision_ids, ordering=None):
2388
"""Get many inventories by revision_ids.
2389
2390
This will buffer some or all of the texts used in constructing the
2391
inventories in memory, but will only parse a single inventory at a
2392
time.
2393
2394
:param revision_ids: The expected revision ids of the inventories.
2395
:param ordering: optional ordering, e.g. 'topological'. If not
2396
specified, the order of revision_ids will be preserved (by
2397
buffering if necessary).
2398
:return: An iterator of inventories.
2399
"""
2400
if ((None in revision_ids)
2401
or (_mod_revision.NULL_REVISION in revision_ids)):
2402
raise ValueError('cannot get null revision inventory')
2403
return self._iter_inventories(revision_ids, ordering)
2404
2405
def _iter_inventories(self, revision_ids, ordering):
2406
"""single-document based inventory iteration."""
2407
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2408
for text, revision_id in inv_xmls:
2409
yield self._deserialise_inventory(revision_id, text)
2410
2411
def _iter_inventory_xmls(self, revision_ids, ordering):
2412
if ordering is None:
2413
order_as_requested = True
2414
ordering = 'unordered'
2415
else:
2416
order_as_requested = False
2417
keys = [(revision_id,) for revision_id in revision_ids]
2418
if not keys:
2419
return
2420
if order_as_requested:
2421
key_iter = iter(keys)
2422
next_key = key_iter.next()
2423
stream = self.inventories.get_record_stream(keys, ordering, True)
2424
text_chunks = {}
2425
for record in stream:
2426
if record.storage_kind != 'absent':
2427
chunks = record.get_bytes_as('chunked')
2428
if order_as_requested:
2429
text_chunks[record.key] = chunks
2430
else:
2431
yield ''.join(chunks), record.key[-1]
2432
else:
2433
raise errors.NoSuchRevision(self, record.key)
2434
if order_as_requested:
2435
# Yield as many results as we can while preserving order.
2436
while next_key in text_chunks:
2437
chunks = text_chunks.pop(next_key)
2438
yield ''.join(chunks), next_key[-1]
2439
try:
2440
next_key = key_iter.next()
2441
except StopIteration:
2442
# We still want to fully consume the get_record_stream,
2443
# just in case it is not actually finished at this point
2444
next_key = None
2445
break
2446
2447
def _deserialise_inventory(self, revision_id, xml):
2448
"""Transform the xml into an inventory object.
2449
2450
:param revision_id: The expected revision id of the inventory.
2451
:param xml: A serialised inventory.
2452
"""
2453
result = self._serializer.read_inventory_from_string(xml, revision_id,
2454
entry_cache=self._inventory_entry_cache,
2455
return_from_cache=self._safe_to_return_from_cache)
2456
if result.revision_id != revision_id:
2457
raise AssertionError('revision id mismatch %s != %s' % (
2458
result.revision_id, revision_id))
2459
return result
2460
2461
def get_serializer_format(self):
2462
return self._serializer.format_num
2463
2464
@needs_read_lock
2465
def _get_inventory_xml(self, revision_id):
2466
"""Get serialized inventory as a string."""
2467
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2468
try:
2469
text, revision_id = texts.next()
2470
except StopIteration:
2471
raise errors.HistoryMissing(self, 'inventory', revision_id)
2472
return text
945
raise NotImplementedError(self.iter_files_bytes)
2473
946
2474
947
def get_rev_id_for_revno(self, revno, known_pair):
2475
948
"""Return the revision id of a revno, given a later (revno, revid)
2506
979
raise AssertionError('_iter_for_revno returned too much history')
2507
980
return (True, partial_history[-1])
2508
981
982
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
2509
983
def iter_reverse_revision_history(self, revision_id):
2510
984
"""Iterate backwards through revision ids in the lefthand history
2511
985
2548
1022
2549
1023
`revision_id` may be NULL_REVISION for the empty tree revision.
2550
1024
"""
2551
revision_id = _mod_revision.ensure_null(revision_id)
2552
# TODO: refactor this to use an existing revision object
2553
# so we don't need to read it in twice.
2554
if revision_id == _mod_revision.NULL_REVISION:
2555
return RevisionTree(self, Inventory(root_id=None),
2556
_mod_revision.NULL_REVISION)
2557
else:
2558
inv = self.get_inventory(revision_id)
2559
return RevisionTree(self, inv, revision_id)
1025
raise NotImplementedError(self.revision_tree)
2560
1026
2561
1027
def revision_trees(self, revision_ids):
2562
1028
"""Return Trees for revisions in this repository.
2564
1030
:param revision_ids: a sequence of revision-ids;
2565
1031
a revision-id may not be None or 'null:'
2566
1032
"""
2567
inventories = self.iter_inventories(revision_ids)
2568
for inv in inventories:
2569
yield RevisionTree(self, inv, inv.revision_id)
2570
2571
def _filtered_revision_trees(self, revision_ids, file_ids):
2572
"""Return Tree for a revision on this branch with only some files.
2573
2574
:param revision_ids: a sequence of revision-ids;
2575
a revision-id may not be None or 'null:'
2576
:param file_ids: if not None, the result is filtered
2577
so that only those file-ids, their parents and their
2578
children are included.
2579
"""
2580
inventories = self.iter_inventories(revision_ids)
2581
for inv in inventories:
2582
# Should we introduce a FilteredRevisionTree class rather
2583
# than pre-filter the inventory here?
2584
filtered_inv = inv.filter(file_ids)
2585
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
1033
raise NotImplementedError(self.revision_trees)
2586
1034
2587
1035
@needs_read_lock
1036
@symbol_versioning.deprecated_method(
1037
symbol_versioning.deprecated_in((2, 4, 0)))
2588
1038
def get_ancestry(self, revision_id, topo_sorted=True):
2589
1039
"""Return a list of revision-ids integrated by a revision.
2590
1040
2594
1044
2595
1045
This is topologically sorted.
2596
1046
"""
1047
if 'evil' in debug.debug_flags:
1048
mutter_callsite(2, "get_ancestry is linear with history.")
2597
1049
if _mod_revision.is_null(revision_id):
2598
1050
return [None]
2599
1051
if not self.has_revision(revision_id):
2621
1073
types it should be a no-op that just returns.
2622
1074
2623
1075
This stub method does not require a lock, but subclasses should use
2624
@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
2625
1077
implicitly lock for the user.
2626
1078
2627
1079
:param hint: If not supplied, the whole repository is packed.
2640
1092
2641
1093
def get_parent_map(self, revision_ids):
2642
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."""
2643
1099
# revisions index works in keys; this just works in revisions
2644
1100
# therefore wrap and unwrap
2645
1101
query_keys = []
2651
1107
raise ValueError('get_parent_map(None) is not valid')
2652
1108
else:
2653
1109
query_keys.append((revision_id ,))
1110
vf = self.revisions.without_fallbacks()
2654
1111
for ((revision_id,), parent_keys) in \
2655
self.revisions.get_parent_map(query_keys).iteritems():
1112
vf.get_parent_map(query_keys).iteritems():
2656
1113
if parent_keys:
2657
1114
result[revision_id] = tuple([parent_revid
2658
1115
for (parent_revid,) in parent_keys])
2661
1118
return result
2662
1119
2663
1120
def _make_parents_provider(self):
2664
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)
2665
1130
2666
1131
@needs_read_lock
2667
1132
def get_known_graph_ancestry(self, revision_ids):
2668
1133
"""Return the known graph for a set of revision ids and their ancestors.
2669
1134
"""
2670
st = static_tuple.StaticTuple
2671
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2672
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2673
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)
2674
1140
2675
1141
def get_graph(self, other_repository=None):
2676
1142
"""Return the graph walker for this repository format"""
2681
1147
[parents_provider, other_repository._make_parents_provider()])
2682
1148
return graph.Graph(parents_provider)
2683
1149
2684
def _get_versioned_file_checker(self, text_key_references=None,
2685
ancestors=None):
2686
"""Return an object suitable for checking versioned files.
2687
2688
:param text_key_references: if non-None, an already built
2689
dictionary mapping text keys ((fileid, revision_id) tuples)
2690
to whether they were referred to by the inventory of the
2691
revision_id that they contain. If None, this will be
2692
calculated.
2693
:param ancestors: Optional result from
2694
self.get_graph().get_parent_map(self.all_revision_ids()) if already
2695
available.
2696
"""
2697
return _VersionedFileChecker(self,
2698
text_key_references=text_key_references, ancestors=ancestors)
2699
2700
def revision_ids_to_search_result(self, result_set):
2701
"""Convert a set of revision ids to a graph SearchResult."""
2702
result_parents = set()
2703
for parents in self.get_graph().get_parent_map(
2704
result_set).itervalues():
2705
result_parents.update(parents)
2706
included_keys = result_set.intersection(result_parents)
2707
start_keys = result_set.difference(included_keys)
2708
exclude_keys = result_parents.difference(result_set)
2709
result = graph.SearchResult(start_keys, exclude_keys,
2710
len(result_set), result_set)
2711
return result
2712
2713
1150
@needs_write_lock
2714
1151
def set_make_working_trees(self, new_value):
2715
1152
"""Set the policy flag for making working trees when creating branches.
2728
1165
2729
1166
@needs_write_lock
2730
1167
def sign_revision(self, revision_id, gpg_strategy):
2731
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()
2732
1170
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2733
1171
2734
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
2735
1190
def has_signature_for_revision_id(self, revision_id):
2736
1191
"""Query for a revision signature for revision_id in the repository."""
2737
if not self.has_revision(revision_id):
2738
raise errors.NoSuchRevision(self, revision_id)
2739
sig_present = (1 == len(
2740
self.signatures.get_parent_map([(revision_id,)])))
2741
return sig_present
1192
raise NotImplementedError(self.has_signature_for_revision_id)
2742
1193
2743
@needs_read_lock
2744
1194
def get_signature_text(self, revision_id):
2745
1195
"""Return the text for a signature."""
2746
stream = self.signatures.get_record_stream([(revision_id,)],
2747
'unordered', True)
2748
record = stream.next()
2749
if record.storage_kind == 'absent':
2750
raise errors.NoSuchRevision(self, revision_id)
2751
return record.get_bytes_as('fulltext')
1196
raise NotImplementedError(self.get_signature_text)
2752
1197
2753
@needs_read_lock
2754
1198
def check(self, revision_ids=None, callback_refs=None, check_repo=True):
2755
1199
"""Check consistency of all history of given revision_ids.
2756
1200
2764
1208
:param check_repo: If False do not check the repository contents, just
2765
1209
calculate the data callback_refs requires and call them back.
2766
1210
"""
2767
return self._check(revision_ids, callback_refs=callback_refs,
1211
return self._check(revision_ids=revision_ids, callback_refs=callback_refs,
2768
1212
check_repo=check_repo)
2769
1213
2770
def _check(self, revision_ids, callback_refs, check_repo):
2771
result = check.Check(self, check_repo=check_repo)
2772
result.check(callback_refs)
2773
return result
1214
def _check(self, revision_ids=None, callback_refs=None, check_repo=True):
1215
raise NotImplementedError(self.check)
2774
1216
2775
1217
def _warn_if_deprecated(self, branch=None):
1218
if not self._format.is_deprecated():
1219
return
2776
1220
global _deprecation_warning_done
2777
1221
if _deprecation_warning_done:
2778
1222
return
2808
1252
except UnicodeDecodeError:
2809
1253
raise errors.NonAsciiRevisionId(method, self)
2810
1254
2811
def revision_graph_can_have_wrong_parents(self):
2812
"""Is it possible for this repository to have a revision graph with
2813
incorrect parents?
2814
2815
If True, then this repository must also implement
2816
_find_inconsistent_revision_parents so that check and reconcile can
2817
check for inconsistencies before proceeding with other checks that may
2818
depend on the revision index being consistent.
2819
"""
2820
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2821
2822
2823
# remove these delegates a while after bzr 0.15
2824
def __make_delegated(name, from_module):
2825
def _deprecated_repository_forwarder():
2826
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2827
% (name, from_module),
2828
DeprecationWarning,
2829
stacklevel=2)
2830
try:
2831
return pyutils.get_named_object(from_module, name)
2832
except AttributeError:
2833
raise AttributeError('module %s has no name %s'
2834
% (sys.modules[from_module], name))
2835
globals()[name] = _deprecated_repository_forwarder
2836
2837
for _name in [
2838
'AllInOneRepository',
2839
'WeaveMetaDirRepository',
2840
'PreSplitOutRepositoryFormat',
2841
'RepositoryFormat4',
2842
'RepositoryFormat5',
2843
'RepositoryFormat6',
2844
'RepositoryFormat7',
2845
]:
2846
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2847
2848
for _name in [
2849
'KnitRepository',
2850
'RepositoryFormatKnit',
2851
'RepositoryFormatKnit1',
2852
]:
2853
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2854
2855
2856
def install_revision(repository, rev, revision_tree):
2857
"""Install all revision data into a repository."""
2858
install_revisions(repository, [(rev, revision_tree, None)])
2859
2860
2861
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2862
"""Install all revision data into a repository.
2863
2864
Accepts an iterable of revision, tree, signature tuples. The signature
2865
may be None.
2866
"""
2867
repository.start_write_group()
2868
try:
2869
inventory_cache = lru_cache.LRUCache(10)
2870
for n, (revision, revision_tree, signature) in enumerate(iterable):
2871
_install_revision(repository, revision, revision_tree, signature,
2872
inventory_cache)
2873
if pb is not None:
2874
pb.update('Transferring revisions', n + 1, num_revisions)
2875
except:
2876
repository.abort_write_group()
2877
raise
2878
else:
2879
repository.commit_write_group()
2880
2881
2882
def _install_revision(repository, rev, revision_tree, signature,
2883
inventory_cache):
2884
"""Install all revision data into a repository."""
2885
present_parents = []
2886
parent_trees = {}
2887
for p_id in rev.parent_ids:
2888
if repository.has_revision(p_id):
2889
present_parents.append(p_id)
2890
parent_trees[p_id] = repository.revision_tree(p_id)
2891
else:
2892
parent_trees[p_id] = repository.revision_tree(
2893
_mod_revision.NULL_REVISION)
2894
2895
inv = revision_tree.inventory
2896
entries = inv.iter_entries()
2897
# backwards compatibility hack: skip the root id.
2898
if not repository.supports_rich_root():
2899
path, root = entries.next()
2900
if root.revision != rev.revision_id:
2901
raise errors.IncompatibleRevision(repr(repository))
2902
text_keys = {}
2903
for path, ie in entries:
2904
text_keys[(ie.file_id, ie.revision)] = ie
2905
text_parent_map = repository.texts.get_parent_map(text_keys)
2906
missing_texts = set(text_keys) - set(text_parent_map)
2907
# Add the texts that are not already present
2908
for text_key in missing_texts:
2909
ie = text_keys[text_key]
2910
text_parents = []
2911
# FIXME: TODO: The following loop overlaps/duplicates that done by
2912
# commit to determine parents. There is a latent/real bug here where
2913
# the parents inserted are not those commit would do - in particular
2914
# they are not filtered by heads(). RBC, AB
2915
for revision, tree in parent_trees.iteritems():
2916
if ie.file_id not in tree:
2917
continue
2918
parent_id = tree.inventory[ie.file_id].revision
2919
if parent_id in text_parents:
2920
continue
2921
text_parents.append((ie.file_id, parent_id))
2922
lines = revision_tree.get_file(ie.file_id).readlines()
2923
repository.texts.add_lines(text_key, text_parents, lines)
2924
try:
2925
# install the inventory
2926
if repository._format._commit_inv_deltas and len(rev.parent_ids):
2927
# Cache this inventory
2928
inventory_cache[rev.revision_id] = inv
2929
try:
2930
basis_inv = inventory_cache[rev.parent_ids[0]]
2931
except KeyError:
2932
repository.add_inventory(rev.revision_id, inv, present_parents)
2933
else:
2934
delta = inv._make_delta(basis_inv)
2935
repository.add_inventory_by_delta(rev.parent_ids[0], delta,
2936
rev.revision_id, present_parents)
2937
else:
2938
repository.add_inventory(rev.revision_id, inv, present_parents)
2939
except errors.RevisionAlreadyPresent:
2940
pass
2941
if signature is not None:
2942
repository.add_signature_text(rev.revision_id, signature)
2943
repository.add_revision(rev.revision_id, rev, inv)
2944
2945
1255
2946
1256
class MetaDirRepository(Repository):
2947
1257
"""Repositories in the new meta-dir layout.
2982
1292
return not self._transport.has('no-working-trees')
2983
1293
2984
1294
2985
class MetaDirVersionedFileRepository(MetaDirRepository):
2986
"""Repositories in a meta-dir, that work via versioned file objects."""
1295
class RepositoryFormatRegistry(controldir.ControlComponentFormatRegistry):
1296
"""Repository format registry."""
2987
1297
2988
def __init__(self, _format, a_bzrdir, control_files):
2989
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2990
control_files)
1298
def get_default(self):
1299
"""Return the current default format."""
1300
return controldir.format_registry.make_bzrdir('default').repository_format
2991
1301
2992
1302
2993
1303
network_format_registry = registry.FormatRegistry()
2999
1309
"""
3000
1310
3001
1311
3002
format_registry = registry.FormatRegistry(network_format_registry)
1312
format_registry = RepositoryFormatRegistry(network_format_registry)
3003
1313
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
3004
1314
3005
1315
This can contain either format instances themselves, or classes/factories that
3010
1320
#####################################################################
3011
1321
# Repository Formats
3012
1322
3013
class RepositoryFormat(object):
1323
class RepositoryFormat(controldir.ControlComponentFormat):
3014
1324
"""A repository format.
3015
1325
3016
1326
Formats provide four things:
3036
1346
created.
3037
1347
3038
1348
Common instance attributes:
3039
_matchingbzrdir - the bzrdir format that the repository format was
1349
_matchingbzrdir - the controldir format that the repository format was
3040
1350
originally written to work with. This can be used if manually
3041
1351
constructing a bzrdir and repository, or more commonly for test suite
3042
1352
parameterization.
3051
1361
# Does this format support CHK bytestring lookups. Set to True or False in
3052
1362
# derived classes.
3053
1363
supports_chks = None
3054
# Should commit add an inventory, or an inventory delta to the repository.
3055
_commit_inv_deltas = True
3056
# What order should fetch operations request streams in?
3057
# The default is unordered as that is the cheapest for an origin to
3058
# provide.
3059
_fetch_order = 'unordered'
3060
# Does this repository format use deltas that can be fetched as-deltas ?
3061
# (E.g. knits, where the knit deltas can be transplanted intact.
3062
# We default to False, which will ensure that enough data to get
3063
# a full text out of any fetch stream will be grabbed.
3064
_fetch_uses_deltas = False
3065
1364
# Should fetch trigger a reconcile after the fetch? Only needed for
3066
1365
# some repository formats that can suffer internal inconsistencies.
3067
1366
_fetch_reconcile = False
3073
1372
# help), and for fetching when data won't have come from the same
3074
1373
# compressor.
3075
1374
pack_compresses = False
3076
# Does the repository inventory storage understand references to trees?
1375
# Does the repository storage understand references to trees?
3077
1376
supports_tree_reference = None
3078
1377
# Is the format experimental ?
3079
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
3080
1400
3081
1401
def __repr__(self):
3082
1402
return "%s()" % self.__class__.__name__
3089
1409
return not self == other
3090
1410
3091
1411
@classmethod
3092
def find_format(klass, a_bzrdir):
3093
"""Return the format for the repository object in a_bzrdir.
3094
3095
This is used by bzr native formats that have a "format" file in
3096
the repository. Other methods may be used by different types of
3097
control directory.
3098
"""
3099
try:
3100
transport = a_bzrdir.get_repository_transport(None)
3101
format_string = transport.get_bytes("format")
3102
return format_registry.get(format_string)
3103
except errors.NoSuchFile:
3104
raise errors.NoRepositoryPresent(a_bzrdir)
3105
except KeyError:
3106
raise errors.UnknownFormatError(format=format_string,
3107
kind='repository')
3108
3109
@classmethod
1412
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3110
1413
def register_format(klass, format):
3111
format_registry.register(format.get_format_string(), format)
1414
format_registry.register(format)
3112
1415
3113
1416
@classmethod
1417
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3114
1418
def unregister_format(klass, format):
3115
format_registry.remove(format.get_format_string())
1419
format_registry.remove(format)
3116
1420
3117
1421
@classmethod
1422
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3118
1423
def get_default_format(klass):
3119
1424
"""Return the current default format."""
3120
from bzrlib import bzrdir
3121
return bzrdir.format_registry.make_bzrdir('default').repository_format
3122
3123
def get_format_string(self):
3124
"""Return the ASCII format string that identifies this format.
3125
3126
Note that in pre format ?? repositories the format string is
3127
not permitted nor written to disk.
3128
"""
3129
raise NotImplementedError(self.get_format_string)
1425
return format_registry.get_default()
3130
1426
3131
1427
def get_format_description(self):
3132
1428
"""Return the short description for this format."""
3133
1429
raise NotImplementedError(self.get_format_description)
3134
1430
3135
# TODO: this shouldn't be in the base class, it's specific to things that
3136
# use weaves or knits -- mbp 20070207
3137
def _get_versioned_file_store(self,
3138
name,
3139
transport,
3140
control_files,
3141
prefixed=True,
3142
versionedfile_class=None,
3143
versionedfile_kwargs={},
3144
escaped=False):
3145
if versionedfile_class is None:
3146
versionedfile_class = self._versionedfile_class
3147
weave_transport = control_files._transport.clone(name)
3148
dir_mode = control_files._dir_mode
3149
file_mode = control_files._file_mode
3150
return VersionedFileStore(weave_transport, prefixed=prefixed,
3151
dir_mode=dir_mode,
3152
file_mode=file_mode,
3153
versionedfile_class=versionedfile_class,
3154
versionedfile_kwargs=versionedfile_kwargs,
3155
escaped=escaped)
3156
3157
def initialize(self, a_bzrdir, shared=False):
3158
"""Initialize a repository of this format in a_bzrdir.
3159
3160
: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.
3161
1435
:param shared: The repository should be initialized as a sharable one.
3162
1436
:returns: The new repository object.
3163
1437
3164
1438
This may raise UninitializableFormat if shared repository are not
3165
compatible the a_bzrdir.
1439
compatible the controldir.
3166
1440
"""
3167
1441
raise NotImplementedError(self.initialize)
3168
1442
3175
1449
"""
3176
1450
return True
3177
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
3178
1460
def network_name(self):
3179
1461
"""A simple byte string uniquely identifying this format for RPC calls.
3180
1462
3196
1478
'Does not support nested trees', target_format,
3197
1479
from_format=self)
3198
1480
3199
def open(self, a_bzrdir, _found=False):
3200
"""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.
3201
1483
3202
1484
_found is a private parameter, do not use it.
3203
1485
"""
3204
1486
raise NotImplementedError(self.open)
3205
1487
3206
def _run_post_repo_init_hooks(self, repository, a_bzrdir, shared):
3207
from bzrlib.bzrdir import BzrDir, RepoInitHookParams
3208
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']
3209
1491
if not hooks:
3210
1492
return
3211
params = RepoInitHookParams(repository, self, a_bzrdir, shared)
1493
params = RepoInitHookParams(repository, self, controldir, shared)
3212
1494
for hook in hooks:
3213
1495
hook(params)
3214
1496
3215
1497
3216
class MetaDirRepositoryFormat(RepositoryFormat):
1498
class RepositoryFormatMetaDir(bzrdir.BzrDirMetaComponentFormat, RepositoryFormat):
3217
1499
"""Common base class for the new repositories using the metadir layout."""
3218
1500
3219
1501
rich_root_data = False
3220
1502
supports_tree_reference = False
3221
1503
supports_external_lookups = False
1504
supports_leaving_lock = True
1505
supports_nesting_repositories = True
3222
1506
3223
1507
@property
3224
1508
def _matchingbzrdir(self):
3227
1511
return matching
3228
1512
3229
1513
def __init__(self):
3230
super(MetaDirRepositoryFormat, self).__init__()
1514
RepositoryFormat.__init__(self)
1515
bzrdir.BzrDirMetaComponentFormat.__init__(self)
3231
1516
3232
1517
def _create_control_files(self, a_bzrdir):
3233
1518
"""Create the required files and the initial control_files object."""
3257
1542
finally:
3258
1543
control_files.unlock()
3259
1544
3260
def network_name(self):
3261
"""Metadir formats have matching disk and network format strings."""
3262
return self.get_format_string()
3263
3264
3265
# Pre-0.8 formats that don't have a disk format string (because they are
3266
# versioned by the matching control directory). We use the control directories
3267
# disk format string as a key for the network_name because they meet the
3268
# constraints (simple string, unique, immutable).
3269
network_format_registry.register_lazy(
3270
"Bazaar-NG branch, format 5\n",
3271
'bzrlib.repofmt.weaverepo',
3272
'RepositoryFormat5',
3273
)
3274
network_format_registry.register_lazy(
3275
"Bazaar-NG branch, format 6\n",
3276
'bzrlib.repofmt.weaverepo',
3277
'RepositoryFormat6',
3278
)
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
3279
1560
3280
1561
# formats which have no format string are not discoverable or independently
3281
1562
# creatable on disk, so are not registered in format_registry. They're
3282
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
3283
# 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
3284
1565
# the repository is not separately opened are similar.
3285
1566
3286
1567
format_registry.register_lazy(
3287
'Bazaar-NG Repository format 7',
3288
'bzrlib.repofmt.weaverepo',
3289
'RepositoryFormat7'
3290
)
3291
3292
format_registry.register_lazy(
3293
1568
'Bazaar-NG Knit Repository Format 1',
3294
1569
'bzrlib.repofmt.knitrepo',
3295
1570
'RepositoryFormatKnit1',
3312
1587
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
3313
1588
format_registry.register_lazy(
3314
1589
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
3315
'bzrlib.repofmt.pack_repo',
1590
'bzrlib.repofmt.knitpack_repo',
3316
1591
'RepositoryFormatKnitPack1',
3317
1592
)
3318
1593
format_registry.register_lazy(
3319
1594
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
3320
'bzrlib.repofmt.pack_repo',
1595
'bzrlib.repofmt.knitpack_repo',
3321
1596
'RepositoryFormatKnitPack3',
3322
1597
)
3323
1598
format_registry.register_lazy(
3324
1599
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
3325
'bzrlib.repofmt.pack_repo',
1600
'bzrlib.repofmt.knitpack_repo',
3326
1601
'RepositoryFormatKnitPack4',
3327
1602
)
3328
1603
format_registry.register_lazy(
3329
1604
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
3330
'bzrlib.repofmt.pack_repo',
1605
'bzrlib.repofmt.knitpack_repo',
3331
1606
'RepositoryFormatKnitPack5',
3332
1607
)
3333
1608
format_registry.register_lazy(
3334
1609
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
3335
'bzrlib.repofmt.pack_repo',
1610
'bzrlib.repofmt.knitpack_repo',
3336
1611
'RepositoryFormatKnitPack5RichRoot',
3337
1612
)
3338
1613
format_registry.register_lazy(
3339
1614
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
3340
'bzrlib.repofmt.pack_repo',
1615
'bzrlib.repofmt.knitpack_repo',
3341
1616
'RepositoryFormatKnitPack5RichRootBroken',
3342
1617
)
3343
1618
format_registry.register_lazy(
3344
1619
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
3345
'bzrlib.repofmt.pack_repo',
1620
'bzrlib.repofmt.knitpack_repo',
3346
1621
'RepositoryFormatKnitPack6',
3347
1622
)
3348
1623
format_registry.register_lazy(
3349
1624
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
3350
'bzrlib.repofmt.pack_repo',
1625
'bzrlib.repofmt.knitpack_repo',
3351
1626
'RepositoryFormatKnitPack6RichRoot',
3352
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
)
3353
1633
3354
1634
# Development formats.
3355
# Obsolete but kept pending a CHK based subtree format.
1635
# Check their docstrings to see if/when they are obsolete.
3356
1636
format_registry.register_lazy(
3357
1637
("Bazaar development format 2 with subtree support "
3358
1638
"(needs bzr.dev from before 1.8)\n"),
3359
'bzrlib.repofmt.pack_repo',
1639
'bzrlib.repofmt.knitpack_repo',
3360
1640
'RepositoryFormatPackDevelopment2Subtree',
3361
1641
)
3362
3363
# 1.14->1.16 go below here
3364
format_registry.register_lazy(
3365
'Bazaar development format - group compression and chk inventory'
3366
' (needs bzr.dev from 1.14)\n',
3367
'bzrlib.repofmt.groupcompress_repo',
3368
'RepositoryFormatCHK1',
3369
)
3370
3371
format_registry.register_lazy(
3372
'Bazaar development format - chk repository with bencode revision '
3373
'serialization (needs bzr.dev from 1.16)\n',
3374
'bzrlib.repofmt.groupcompress_repo',
3375
'RepositoryFormatCHK2',
3376
)
3377
format_registry.register_lazy(
3378
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3379
'bzrlib.repofmt.groupcompress_repo',
3380
'RepositoryFormat2a',
3381
)
3382
1642
format_registry.register_lazy(
3383
1643
'Bazaar development format 8\n',
3384
1644
'bzrlib.repofmt.groupcompress_repo',
3398
1658
InterRepository.get(other).method_name(parameters).
3399
1659
"""
3400
1660
3401
_walk_to_common_revisions_batch_size = 50
3402
1661
_optimisers = []
3403
1662
"""The available optimised InterRepository types."""
3404
1663
3419
1678
self.target.fetch(self.source, revision_id=revision_id)
3420
1679
3421
1680
@needs_write_lock
3422
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3423
fetch_spec=None):
1681
def fetch(self, revision_id=None, find_ghosts=False):
3424
1682
"""Fetch the content required to construct revision_id.
3425
1683
3426
1684
The content is copied from self.source to self.target.
3427
1685
3428
1686
:param revision_id: if None all content is copied, if NULL_REVISION no
3429
1687
content is copied.
3430
:param pb: ignored.
3431
1688
:return: None.
3432
1689
"""
3433
ui.ui_factory.warn_experimental_format_fetch(self)
3434
from bzrlib.fetch import RepoFetcher
3435
# See <https://launchpad.net/bugs/456077> asking for a warning here
3436
if self.source._format.network_name() != self.target._format.network_name():
3437
ui.ui_factory.show_user_warning('cross_format_fetch',
3438
from_format=self.source._format,
3439
to_format=self.target._format)
3440
f = RepoFetcher(to_repository=self.target,
3441
from_repository=self.source,
3442
last_revision=revision_id,
3443
fetch_spec=fetch_spec,
3444
find_ghosts=find_ghosts)
3445
3446
def _walk_to_common_revisions(self, revision_ids):
3447
"""Walk out from revision_ids in source to revisions target has.
3448
3449
:param revision_ids: The start point for the search.
3450
:return: A set of revision ids.
3451
"""
3452
target_graph = self.target.get_graph()
3453
revision_ids = frozenset(revision_ids)
3454
missing_revs = set()
3455
source_graph = self.source.get_graph()
3456
# ensure we don't pay silly lookup costs.
3457
searcher = source_graph._make_breadth_first_searcher(revision_ids)
3458
null_set = frozenset([_mod_revision.NULL_REVISION])
3459
searcher_exhausted = False
3460
while True:
3461
next_revs = set()
3462
ghosts = set()
3463
# Iterate the searcher until we have enough next_revs
3464
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3465
try:
3466
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3467
next_revs.update(next_revs_part)
3468
ghosts.update(ghosts_part)
3469
except StopIteration:
3470
searcher_exhausted = True
3471
break
3472
# If there are ghosts in the source graph, and the caller asked for
3473
# them, make sure that they are present in the target.
3474
# We don't care about other ghosts as we can't fetch them and
3475
# haven't been asked to.
3476
ghosts_to_check = set(revision_ids.intersection(ghosts))
3477
revs_to_get = set(next_revs).union(ghosts_to_check)
3478
if revs_to_get:
3479
have_revs = set(target_graph.get_parent_map(revs_to_get))
3480
# we always have NULL_REVISION present.
3481
have_revs = have_revs.union(null_set)
3482
# Check if the target is missing any ghosts we need.
3483
ghosts_to_check.difference_update(have_revs)
3484
if ghosts_to_check:
3485
# One of the caller's revision_ids is a ghost in both the
3486
# source and the target.
3487
raise errors.NoSuchRevision(
3488
self.source, ghosts_to_check.pop())
3489
missing_revs.update(next_revs - have_revs)
3490
# Because we may have walked past the original stop point, make
3491
# sure everything is stopped
3492
stop_revs = searcher.find_seen_ancestors(have_revs)
3493
searcher.stop_searching_any(stop_revs)
3494
if searcher_exhausted:
3495
break
3496
return searcher.get_result()
1690
raise NotImplementedError(self.fetch)
3497
1691
3498
1692
@needs_read_lock
3499
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):
3500
1697
"""Return the revision ids that source has that target does not.
3501
1698
3502
1699
:param revision_id: only return revision ids included by this
3503
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.
3504
1708
:param find_ghosts: If True find missing revisions in deep history
3505
1709
rather than just finding the surface difference.
1710
:param limit: Maximum number of revisions to return, topologically
1711
ordered
3506
1712
:return: A bzrlib.graph.SearchResult.
3507
1713
"""
3508
# stop searching at found target revisions.
3509
if not find_ghosts and revision_id is not None:
3510
return self._walk_to_common_revisions([revision_id])
3511
# generic, possibly worst case, slow code path.
3512
target_ids = set(self.target.all_revision_ids())
3513
if revision_id is not None:
3514
source_ids = self.source.get_ancestry(revision_id)
3515
if source_ids[0] is not None:
3516
raise AssertionError()
3517
source_ids.pop(0)
3518
else:
3519
source_ids = self.source.all_revision_ids()
3520
result_set = set(source_ids).difference(target_ids)
3521
return self.source.revision_ids_to_search_result(result_set)
1714
raise NotImplementedError(self.search_missing_revision_ids)
3522
1715
3523
1716
@staticmethod
3524
1717
def _same_model(source, target):
3545
1738
"different serializers")
3546
1739
3547
1740
3548
class InterSameDataRepository(InterRepository):
3549
"""Code for converting between repositories that represent the same data.
3550
3551
Data format and model must match for this to work.
3552
"""
3553
3554
@classmethod
3555
def _get_repo_format_to_test(self):
3556
"""Repository format for testing with.
3557
3558
InterSameData can pull from subtree to subtree and from non-subtree to
3559
non-subtree, so we test this with the richest repository format.
3560
"""
3561
from bzrlib.repofmt import knitrepo
3562
return knitrepo.RepositoryFormatKnit3()
3563
3564
@staticmethod
3565
def is_compatible(source, target):
3566
return InterRepository._same_model(source, target)
3567
3568
3569
class InterWeaveRepo(InterSameDataRepository):
3570
"""Optimised code paths between Weave based repositories.
3571
3572
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
3573
implemented lazy inter-object optimisation.
3574
"""
3575
3576
@classmethod
3577
def _get_repo_format_to_test(self):
3578
from bzrlib.repofmt import weaverepo
3579
return weaverepo.RepositoryFormat7()
3580
3581
@staticmethod
3582
def is_compatible(source, target):
3583
"""Be compatible with known Weave formats.
3584
3585
We don't test for the stores being of specific types because that
3586
could lead to confusing results, and there is no need to be
3587
overly general.
3588
"""
3589
from bzrlib.repofmt.weaverepo import (
3590
RepositoryFormat5,
3591
RepositoryFormat6,
3592
RepositoryFormat7,
3593
)
3594
try:
3595
return (isinstance(source._format, (RepositoryFormat5,
3596
RepositoryFormat6,
3597
RepositoryFormat7)) and
3598
isinstance(target._format, (RepositoryFormat5,
3599
RepositoryFormat6,
3600
RepositoryFormat7)))
3601
except AttributeError:
3602
return False
3603
3604
@needs_write_lock
3605
def copy_content(self, revision_id=None):
3606
"""See InterRepository.copy_content()."""
3607
# weave specific optimised path:
3608
try:
3609
self.target.set_make_working_trees(self.source.make_working_trees())
3610
except (errors.RepositoryUpgradeRequired, NotImplemented):
3611
pass
3612
# FIXME do not peek!
3613
if self.source._transport.listable():
3614
pb = ui.ui_factory.nested_progress_bar()
3615
try:
3616
self.target.texts.insert_record_stream(
3617
self.source.texts.get_record_stream(
3618
self.source.texts.keys(), 'topological', False))
3619
pb.update('Copying inventory', 0, 1)
3620
self.target.inventories.insert_record_stream(
3621
self.source.inventories.get_record_stream(
3622
self.source.inventories.keys(), 'topological', False))
3623
self.target.signatures.insert_record_stream(
3624
self.source.signatures.get_record_stream(
3625
self.source.signatures.keys(),
3626
'unordered', True))
3627
self.target.revisions.insert_record_stream(
3628
self.source.revisions.get_record_stream(
3629
self.source.revisions.keys(),
3630
'topological', True))
3631
finally:
3632
pb.finished()
3633
else:
3634
self.target.fetch(self.source, revision_id=revision_id)
3635
3636
@needs_read_lock
3637
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3638
"""See InterRepository.missing_revision_ids()."""
3639
# we want all revisions to satisfy revision_id in source.
3640
# but we don't want to stat every file here and there.
3641
# we want then, all revisions other needs to satisfy revision_id
3642
# checked, but not those that we have locally.
3643
# so the first thing is to get a subset of the revisions to
3644
# satisfy revision_id in source, and then eliminate those that
3645
# we do already have.
3646
# this is slow on high latency connection to self, but as this
3647
# disk format scales terribly for push anyway due to rewriting
3648
# inventory.weave, this is considered acceptable.
3649
# - RBC 20060209
3650
if revision_id is not None:
3651
source_ids = self.source.get_ancestry(revision_id)
3652
if source_ids[0] is not None:
3653
raise AssertionError()
3654
source_ids.pop(0)
3655
else:
3656
source_ids = self.source._all_possible_ids()
3657
source_ids_set = set(source_ids)
3658
# source_ids is the worst possible case we may need to pull.
3659
# now we want to filter source_ids against what we actually
3660
# have in target, but don't try to check for existence where we know
3661
# we do not have a revision as that would be pointless.
3662
target_ids = set(self.target._all_possible_ids())
3663
possibly_present_revisions = target_ids.intersection(source_ids_set)
3664
actually_present_revisions = set(
3665
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3666
required_revisions = source_ids_set.difference(actually_present_revisions)
3667
if revision_id is not None:
3668
# we used get_ancestry to determine source_ids then we are assured all
3669
# revisions referenced are present as they are installed in topological order.
3670
# and the tip revision was validated by get_ancestry.
3671
result_set = required_revisions
3672
else:
3673
# if we just grabbed the possibly available ids, then
3674
# we only have an estimate of whats available and need to validate
3675
# that against the revision records.
3676
result_set = set(
3677
self.source._eliminate_revisions_not_present(required_revisions))
3678
return self.source.revision_ids_to_search_result(result_set)
3679
3680
3681
class InterKnitRepo(InterSameDataRepository):
3682
"""Optimised code paths between Knit based repositories."""
3683
3684
@classmethod
3685
def _get_repo_format_to_test(self):
3686
from bzrlib.repofmt import knitrepo
3687
return knitrepo.RepositoryFormatKnit1()
3688
3689
@staticmethod
3690
def is_compatible(source, target):
3691
"""Be compatible with known Knit formats.
3692
3693
We don't test for the stores being of specific types because that
3694
could lead to confusing results, and there is no need to be
3695
overly general.
3696
"""
3697
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
3698
try:
3699
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
3700
isinstance(target._format, RepositoryFormatKnit))
3701
except AttributeError:
3702
return False
3703
return are_knits and InterRepository._same_model(source, target)
3704
3705
@needs_read_lock
3706
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3707
"""See InterRepository.missing_revision_ids()."""
3708
if revision_id is not None:
3709
source_ids = self.source.get_ancestry(revision_id)
3710
if source_ids[0] is not None:
3711
raise AssertionError()
3712
source_ids.pop(0)
3713
else:
3714
source_ids = self.source.all_revision_ids()
3715
source_ids_set = set(source_ids)
3716
# source_ids is the worst possible case we may need to pull.
3717
# now we want to filter source_ids against what we actually
3718
# have in target, but don't try to check for existence where we know
3719
# we do not have a revision as that would be pointless.
3720
target_ids = set(self.target.all_revision_ids())
3721
possibly_present_revisions = target_ids.intersection(source_ids_set)
3722
actually_present_revisions = set(
3723
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3724
required_revisions = source_ids_set.difference(actually_present_revisions)
3725
if revision_id is not None:
3726
# we used get_ancestry to determine source_ids then we are assured all
3727
# revisions referenced are present as they are installed in topological order.
3728
# and the tip revision was validated by get_ancestry.
3729
result_set = required_revisions
3730
else:
3731
# if we just grabbed the possibly available ids, then
3732
# we only have an estimate of whats available and need to validate
3733
# that against the revision records.
3734
result_set = set(
3735
self.source._eliminate_revisions_not_present(required_revisions))
3736
return self.source.revision_ids_to_search_result(result_set)
3737
3738
3739
class InterDifferingSerializer(InterRepository):
3740
3741
@classmethod
3742
def _get_repo_format_to_test(self):
3743
return None
3744
3745
@staticmethod
3746
def is_compatible(source, target):
3747
"""Be compatible with Knit2 source and Knit3 target"""
3748
# This is redundant with format.check_conversion_target(), however that
3749
# raises an exception, and we just want to say "False" as in we won't
3750
# support converting between these formats.
3751
if 'IDS_never' in debug.debug_flags:
3752
return False
3753
if source.supports_rich_root() and not target.supports_rich_root():
3754
return False
3755
if (source._format.supports_tree_reference
3756
and not target._format.supports_tree_reference):
3757
return False
3758
if target._fallback_repositories and target._format.supports_chks:
3759
# IDS doesn't know how to copy CHKs for the parent inventories it
3760
# adds to stacked repos.
3761
return False
3762
if 'IDS_always' in debug.debug_flags:
3763
return True
3764
# Only use this code path for local source and target. IDS does far
3765
# too much IO (both bandwidth and roundtrips) over a network.
3766
if not source.bzrdir.transport.base.startswith('file:///'):
3767
return False
3768
if not target.bzrdir.transport.base.startswith('file:///'):
3769
return False
3770
return True
3771
3772
def _get_trees(self, revision_ids, cache):
3773
possible_trees = []
3774
for rev_id in revision_ids:
3775
if rev_id in cache:
3776
possible_trees.append((rev_id, cache[rev_id]))
3777
else:
3778
# Not cached, but inventory might be present anyway.
3779
try:
3780
tree = self.source.revision_tree(rev_id)
3781
except errors.NoSuchRevision:
3782
# Nope, parent is ghost.
3783
pass
3784
else:
3785
cache[rev_id] = tree
3786
possible_trees.append((rev_id, tree))
3787
return possible_trees
3788
3789
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3790
"""Get the best delta and base for this revision.
3791
3792
:return: (basis_id, delta)
3793
"""
3794
deltas = []
3795
# Generate deltas against each tree, to find the shortest.
3796
texts_possibly_new_in_tree = set()
3797
for basis_id, basis_tree in possible_trees:
3798
delta = tree.inventory._make_delta(basis_tree.inventory)
3799
for old_path, new_path, file_id, new_entry in delta:
3800
if new_path is None:
3801
# This file_id isn't present in the new rev, so we don't
3802
# care about it.
3803
continue
3804
if not new_path:
3805
# Rich roots are handled elsewhere...
3806
continue
3807
kind = new_entry.kind
3808
if kind != 'directory' and kind != 'file':
3809
# No text record associated with this inventory entry.
3810
continue
3811
# This is a directory or file that has changed somehow.
3812
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3813
deltas.append((len(delta), basis_id, delta))
3814
deltas.sort()
3815
return deltas[0][1:]
3816
3817
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3818
"""Find all parent revisions that are absent, but for which the
3819
inventory is present, and copy those inventories.
3820
3821
This is necessary to preserve correctness when the source is stacked
3822
without fallbacks configured. (Note that in cases like upgrade the
3823
source may be not have _fallback_repositories even though it is
3824
stacked.)
3825
"""
3826
parent_revs = set()
3827
for parents in parent_map.values():
3828
parent_revs.update(parents)
3829
present_parents = self.source.get_parent_map(parent_revs)
3830
absent_parents = set(parent_revs).difference(present_parents)
3831
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3832
(rev_id,) for rev_id in absent_parents)
3833
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3834
for parent_tree in self.source.revision_trees(parent_inv_ids):
3835
current_revision_id = parent_tree.get_revision_id()
3836
parents_parents_keys = parent_invs_keys_for_stacking[
3837
(current_revision_id,)]
3838
parents_parents = [key[-1] for key in parents_parents_keys]
3839
basis_id = _mod_revision.NULL_REVISION
3840
basis_tree = self.source.revision_tree(basis_id)
3841
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3842
self.target.add_inventory_by_delta(
3843
basis_id, delta, current_revision_id, parents_parents)
3844
cache[current_revision_id] = parent_tree
3845
3846
def _fetch_batch(self, revision_ids, basis_id, cache):
3847
"""Fetch across a few revisions.
3848
3849
:param revision_ids: The revisions to copy
3850
:param basis_id: The revision_id of a tree that must be in cache, used
3851
as a basis for delta when no other base is available
3852
:param cache: A cache of RevisionTrees that we can use.
3853
:return: The revision_id of the last converted tree. The RevisionTree
3854
for it will be in cache
3855
"""
3856
# Walk though all revisions; get inventory deltas, copy referenced
3857
# texts that delta references, insert the delta, revision and
3858
# signature.
3859
root_keys_to_create = set()
3860
text_keys = set()
3861
pending_deltas = []
3862
pending_revisions = []
3863
parent_map = self.source.get_parent_map(revision_ids)
3864
self._fetch_parent_invs_for_stacking(parent_map, cache)
3865
self.source._safe_to_return_from_cache = True
3866
for tree in self.source.revision_trees(revision_ids):
3867
# Find a inventory delta for this revision.
3868
# Find text entries that need to be copied, too.
3869
current_revision_id = tree.get_revision_id()
3870
parent_ids = parent_map.get(current_revision_id, ())
3871
parent_trees = self._get_trees(parent_ids, cache)
3872
possible_trees = list(parent_trees)
3873
if len(possible_trees) == 0:
3874
# There either aren't any parents, or the parents are ghosts,
3875
# so just use the last converted tree.
3876
possible_trees.append((basis_id, cache[basis_id]))
3877
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3878
possible_trees)
3879
revision = self.source.get_revision(current_revision_id)
3880
pending_deltas.append((basis_id, delta,
3881
current_revision_id, revision.parent_ids))
3882
if self._converting_to_rich_root:
3883
self._revision_id_to_root_id[current_revision_id] = \
3884
tree.get_root_id()
3885
# Determine which texts are in present in this revision but not in
3886
# any of the available parents.
3887
texts_possibly_new_in_tree = set()
3888
for old_path, new_path, file_id, entry in delta:
3889
if new_path is None:
3890
# This file_id isn't present in the new rev
3891
continue
3892
if not new_path:
3893
# This is the root
3894
if not self.target.supports_rich_root():
3895
# The target doesn't support rich root, so we don't
3896
# copy
3897
continue
3898
if self._converting_to_rich_root:
3899
# This can't be copied normally, we have to insert
3900
# it specially
3901
root_keys_to_create.add((file_id, entry.revision))
3902
continue
3903
kind = entry.kind
3904
texts_possibly_new_in_tree.add((file_id, entry.revision))
3905
for basis_id, basis_tree in possible_trees:
3906
basis_inv = basis_tree.inventory
3907
for file_key in list(texts_possibly_new_in_tree):
3908
file_id, file_revision = file_key
3909
try:
3910
entry = basis_inv[file_id]
3911
except errors.NoSuchId:
3912
continue
3913
if entry.revision == file_revision:
3914
texts_possibly_new_in_tree.remove(file_key)
3915
text_keys.update(texts_possibly_new_in_tree)
3916
pending_revisions.append(revision)
3917
cache[current_revision_id] = tree
3918
basis_id = current_revision_id
3919
self.source._safe_to_return_from_cache = False
3920
# Copy file texts
3921
from_texts = self.source.texts
3922
to_texts = self.target.texts
3923
if root_keys_to_create:
3924
root_stream = _mod_fetch._new_root_data_stream(
3925
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3926
self.source)
3927
to_texts.insert_record_stream(root_stream)
3928
to_texts.insert_record_stream(from_texts.get_record_stream(
3929
text_keys, self.target._format._fetch_order,
3930
not self.target._format._fetch_uses_deltas))
3931
# insert inventory deltas
3932
for delta in pending_deltas:
3933
self.target.add_inventory_by_delta(*delta)
3934
if self.target._fallback_repositories:
3935
# Make sure this stacked repository has all the parent inventories
3936
# for the new revisions that we are about to insert. We do this
3937
# before adding the revisions so that no revision is added until
3938
# all the inventories it may depend on are added.
3939
# Note that this is overzealous, as we may have fetched these in an
3940
# earlier batch.
3941
parent_ids = set()
3942
revision_ids = set()
3943
for revision in pending_revisions:
3944
revision_ids.add(revision.revision_id)
3945
parent_ids.update(revision.parent_ids)
3946
parent_ids.difference_update(revision_ids)
3947
parent_ids.discard(_mod_revision.NULL_REVISION)
3948
parent_map = self.source.get_parent_map(parent_ids)
3949
# we iterate over parent_map and not parent_ids because we don't
3950
# want to try copying any revision which is a ghost
3951
for parent_tree in self.source.revision_trees(parent_map):
3952
current_revision_id = parent_tree.get_revision_id()
3953
parents_parents = parent_map[current_revision_id]
3954
possible_trees = self._get_trees(parents_parents, cache)
3955
if len(possible_trees) == 0:
3956
# There either aren't any parents, or the parents are
3957
# ghosts, so just use the last converted tree.
3958
possible_trees.append((basis_id, cache[basis_id]))
3959
basis_id, delta = self._get_delta_for_revision(parent_tree,
3960
parents_parents, possible_trees)
3961
self.target.add_inventory_by_delta(
3962
basis_id, delta, current_revision_id, parents_parents)
3963
# insert signatures and revisions
3964
for revision in pending_revisions:
3965
try:
3966
signature = self.source.get_signature_text(
3967
revision.revision_id)
3968
self.target.add_signature_text(revision.revision_id,
3969
signature)
3970
except errors.NoSuchRevision:
3971
pass
3972
self.target.add_revision(revision.revision_id, revision)
3973
return basis_id
3974
3975
def _fetch_all_revisions(self, revision_ids, pb):
3976
"""Fetch everything for the list of revisions.
3977
3978
:param revision_ids: The list of revisions to fetch. Must be in
3979
topological order.
3980
:param pb: A ProgressTask
3981
:return: None
3982
"""
3983
basis_id, basis_tree = self._get_basis(revision_ids[0])
3984
batch_size = 100
3985
cache = lru_cache.LRUCache(100)
3986
cache[basis_id] = basis_tree
3987
del basis_tree # We don't want to hang on to it here
3988
hints = []
3989
a_graph = None
3990
3991
for offset in range(0, len(revision_ids), batch_size):
3992
self.target.start_write_group()
3993
try:
3994
pb.update('Transferring revisions', offset,
3995
len(revision_ids))
3996
batch = revision_ids[offset:offset+batch_size]
3997
basis_id = self._fetch_batch(batch, basis_id, cache)
3998
except:
3999
self.source._safe_to_return_from_cache = False
4000
self.target.abort_write_group()
4001
raise
4002
else:
4003
hint = self.target.commit_write_group()
4004
if hint:
4005
hints.extend(hint)
4006
if hints and self.target._format.pack_compresses:
4007
self.target.pack(hint=hints)
4008
pb.update('Transferring revisions', len(revision_ids),
4009
len(revision_ids))
4010
4011
@needs_write_lock
4012
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
4013
fetch_spec=None):
4014
"""See InterRepository.fetch()."""
4015
if fetch_spec is not None:
4016
raise AssertionError("Not implemented yet...")
4017
ui.ui_factory.warn_experimental_format_fetch(self)
4018
if (not self.source.supports_rich_root()
4019
and self.target.supports_rich_root()):
4020
self._converting_to_rich_root = True
4021
self._revision_id_to_root_id = {}
4022
else:
4023
self._converting_to_rich_root = False
4024
# See <https://launchpad.net/bugs/456077> asking for a warning here
4025
if self.source._format.network_name() != self.target._format.network_name():
4026
ui.ui_factory.show_user_warning('cross_format_fetch',
4027
from_format=self.source._format,
4028
to_format=self.target._format)
4029
revision_ids = self.target.search_missing_revision_ids(self.source,
4030
revision_id, find_ghosts=find_ghosts).get_keys()
4031
if not revision_ids:
4032
return 0, 0
4033
revision_ids = tsort.topo_sort(
4034
self.source.get_graph().get_parent_map(revision_ids))
4035
if not revision_ids:
4036
return 0, 0
4037
# Walk though all revisions; get inventory deltas, copy referenced
4038
# texts that delta references, insert the delta, revision and
4039
# signature.
4040
if pb is None:
4041
my_pb = ui.ui_factory.nested_progress_bar()
4042
pb = my_pb
4043
else:
4044
symbol_versioning.warn(
4045
symbol_versioning.deprecated_in((1, 14, 0))
4046
% "pb parameter to fetch()")
4047
my_pb = None
4048
try:
4049
self._fetch_all_revisions(revision_ids, pb)
4050
finally:
4051
if my_pb is not None:
4052
my_pb.finished()
4053
return len(revision_ids), 0
4054
4055
def _get_basis(self, first_revision_id):
4056
"""Get a revision and tree which exists in the target.
4057
4058
This assumes that first_revision_id is selected for transmission
4059
because all other ancestors are already present. If we can't find an
4060
ancestor we fall back to NULL_REVISION since we know that is safe.
4061
4062
:return: (basis_id, basis_tree)
4063
"""
4064
first_rev = self.source.get_revision(first_revision_id)
4065
try:
4066
basis_id = first_rev.parent_ids[0]
4067
# only valid as a basis if the target has it
4068
self.target.get_revision(basis_id)
4069
# Try to get a basis tree - if its a ghost it will hit the
4070
# NoSuchRevision case.
4071
basis_tree = self.source.revision_tree(basis_id)
4072
except (IndexError, errors.NoSuchRevision):
4073
basis_id = _mod_revision.NULL_REVISION
4074
basis_tree = self.source.revision_tree(basis_id)
4075
return basis_id, basis_tree
4076
4077
4078
InterRepository.register_optimiser(InterDifferingSerializer)
4079
InterRepository.register_optimiser(InterSameDataRepository)
4080
InterRepository.register_optimiser(InterWeaveRepo)
4081
InterRepository.register_optimiser(InterKnitRepo)
4082
4083
4084
1741
class CopyConverter(object):
4085
1742
"""A repository conversion tool which just performs a copy of the content.
4086
1743
4107
1764
# trigger an assertion if not such
4108
1765
repo._format.get_format_string()
4109
1766
self.repo_dir = repo.bzrdir
4110
pb.update('Moving repository to repository.backup')
1767
pb.update(gettext('Moving repository to repository.backup'))
4111
1768
self.repo_dir.transport.move('repository', 'repository.backup')
4112
1769
backup_transport = self.repo_dir.transport.clone('repository.backup')
4113
1770
repo._format.check_conversion_target(self.target_format)
4114
1771
self.source_repo = repo._format.open(self.repo_dir,
4115
1772
_found=True,
4116
1773
_override_transport=backup_transport)
4117
pb.update('Creating new repository')
1774
pb.update(gettext('Creating new repository'))
4118
1775
converted = self.target_format.initialize(self.repo_dir,
4119
1776
self.source_repo.is_shared())
4120
1777
converted.lock_write()
4121
1778
try:
4122
pb.update('Copying content')
1779
pb.update(gettext('Copying content'))
4123
1780
self.source_repo.copy_content_into(converted)
4124
1781
finally:
4125
1782
converted.unlock()
4126
pb.update('Deleting old repository content')
1783
pb.update(gettext('Deleting old repository content'))
4127
1784
self.repo_dir.transport.delete_tree('repository.backup')
4128
ui.ui_factory.note('repository converted')
1785
ui.ui_factory.note(gettext('repository converted'))
4129
1786
pb.finished()
4130
1787
4131
1788
4132
_unescape_map = {
4133
'apos':"'",
4134
'quot':'"',
4135
'amp':'&',
4136
'lt':'<',
4137
'gt':'>'
4138
}
4139
4140
4141
def _unescaper(match, _map=_unescape_map):
4142
code = match.group(1)
4143
try:
4144
return _map[code]
4145
except KeyError:
4146
if not code.startswith('#'):
4147
raise
4148
return unichr(int(code[1:])).encode('utf8')
4149
4150
4151
_unescape_re = None
4152
4153
4154
def _unescape_xml(data):
4155
"""Unescape predefined XML entities in a string of data."""
4156
global _unescape_re
4157
if _unescape_re is None:
4158
_unescape_re = re.compile('\&([^;]*);')
4159
return _unescape_re.sub(_unescaper, data)
4160
4161
4162
class _VersionedFileChecker(object):
4163
4164
def __init__(self, repository, text_key_references=None, ancestors=None):
4165
self.repository = repository
4166
self.text_index = self.repository._generate_text_key_index(
4167
text_key_references=text_key_references, ancestors=ancestors)
4168
4169
def calculate_file_version_parents(self, text_key):
4170
"""Calculate the correct parents for a file version according to
4171
the inventories.
4172
"""
4173
parent_keys = self.text_index[text_key]
4174
if parent_keys == [_mod_revision.NULL_REVISION]:
4175
return ()
4176
return tuple(parent_keys)
4177
4178
def check_file_version_parents(self, texts, progress_bar=None):
4179
"""Check the parents stored in a versioned file are correct.
4180
4181
It also detects file versions that are not referenced by their
4182
corresponding revision's inventory.
4183
4184
:returns: A tuple of (wrong_parents, dangling_file_versions).
4185
wrong_parents is a dict mapping {revision_id: (stored_parents,
4186
correct_parents)} for each revision_id where the stored parents
4187
are not correct. dangling_file_versions is a set of (file_id,
4188
revision_id) tuples for versions that are present in this versioned
4189
file, but not used by the corresponding inventory.
4190
"""
4191
local_progress = None
4192
if progress_bar is None:
4193
local_progress = ui.ui_factory.nested_progress_bar()
4194
progress_bar = local_progress
4195
try:
4196
return self._check_file_version_parents(texts, progress_bar)
4197
finally:
4198
if local_progress:
4199
local_progress.finished()
4200
4201
def _check_file_version_parents(self, texts, progress_bar):
4202
"""See check_file_version_parents."""
4203
wrong_parents = {}
4204
self.file_ids = set([file_id for file_id, _ in
4205
self.text_index.iterkeys()])
4206
# text keys is now grouped by file_id
4207
n_versions = len(self.text_index)
4208
progress_bar.update('loading text store', 0, n_versions)
4209
parent_map = self.repository.texts.get_parent_map(self.text_index)
4210
# On unlistable transports this could well be empty/error...
4211
text_keys = self.repository.texts.keys()
4212
unused_keys = frozenset(text_keys) - set(self.text_index)
4213
for num, key in enumerate(self.text_index.iterkeys()):
4214
progress_bar.update('checking text graph', num, n_versions)
4215
correct_parents = self.calculate_file_version_parents(key)
4216
try:
4217
knit_parents = parent_map[key]
4218
except errors.RevisionNotPresent:
4219
# Missing text!
4220
knit_parents = None
4221
if correct_parents != knit_parents:
4222
wrong_parents[key] = (knit_parents, correct_parents)
4223
return wrong_parents, unused_keys
4224
4225
4226
def _old_get_graph(repository, revision_id):
4227
"""DO NOT USE. That is all. I'm serious."""
4228
graph = repository.get_graph()
4229
revision_graph = dict(((key, value) for key, value in
4230
graph.iter_ancestry([revision_id]) if value is not None))
4231
return _strip_NULL_ghosts(revision_graph)
4232
4233
4234
1789
def _strip_NULL_ghosts(revision_graph):
4235
1790
"""Also don't use this. more compatibility code for unmigrated clients."""
4236
1791
# Filter ghosts, and null:
4242
1797
return revision_graph
4243
1798
4244
1799
4245
class StreamSink(object):
4246
"""An object that can insert a stream into a repository.
4247
4248
This interface handles the complexity of reserialising inventories and
4249
revisions from different formats, and allows unidirectional insertion into
4250
stacked repositories without looking for the missing basis parents
4251
beforehand.
4252
"""
4253
4254
def __init__(self, target_repo):
4255
self.target_repo = target_repo
4256
4257
def insert_stream(self, stream, src_format, resume_tokens):
4258
"""Insert a stream's content into the target repository.
4259
4260
:param src_format: a bzr repository format.
4261
4262
:return: a list of resume tokens and an iterable of keys additional
4263
items required before the insertion can be completed.
4264
"""
4265
self.target_repo.lock_write()
4266
try:
4267
if resume_tokens:
4268
self.target_repo.resume_write_group(resume_tokens)
4269
is_resume = True
4270
else:
4271
self.target_repo.start_write_group()
4272
is_resume = False
4273
try:
4274
# locked_insert_stream performs a commit|suspend.
4275
return self._locked_insert_stream(stream, src_format,
4276
is_resume)
4277
except:
4278
self.target_repo.abort_write_group(suppress_errors=True)
4279
raise
4280
finally:
4281
self.target_repo.unlock()
4282
4283
def _locked_insert_stream(self, stream, src_format, is_resume):
4284
to_serializer = self.target_repo._format._serializer
4285
src_serializer = src_format._serializer
4286
new_pack = None
4287
if to_serializer == src_serializer:
4288
# If serializers match and the target is a pack repository, set the
4289
# write cache size on the new pack. This avoids poor performance
4290
# on transports where append is unbuffered (such as
4291
# RemoteTransport). This is safe to do because nothing should read
4292
# back from the target repository while a stream with matching
4293
# serialization is being inserted.
4294
# The exception is that a delta record from the source that should
4295
# be a fulltext may need to be expanded by the target (see
4296
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4297
# explicitly flush any buffered writes first in that rare case.
4298
try:
4299
new_pack = self.target_repo._pack_collection._new_pack
4300
except AttributeError:
4301
# Not a pack repository
4302
pass
4303
else:
4304
new_pack.set_write_cache_size(1024*1024)
4305
for substream_type, substream in stream:
4306
if 'stream' in debug.debug_flags:
4307
mutter('inserting substream: %s', substream_type)
4308
if substream_type == 'texts':
4309
self.target_repo.texts.insert_record_stream(substream)
4310
elif substream_type == 'inventories':
4311
if src_serializer == to_serializer:
4312
self.target_repo.inventories.insert_record_stream(
4313
substream)
4314
else:
4315
self._extract_and_insert_inventories(
4316
substream, src_serializer)
4317
elif substream_type == 'inventory-deltas':
4318
self._extract_and_insert_inventory_deltas(
4319
substream, src_serializer)
4320
elif substream_type == 'chk_bytes':
4321
# XXX: This doesn't support conversions, as it assumes the
4322
# conversion was done in the fetch code.
4323
self.target_repo.chk_bytes.insert_record_stream(substream)
4324
elif substream_type == 'revisions':
4325
# This may fallback to extract-and-insert more often than
4326
# required if the serializers are different only in terms of
4327
# the inventory.
4328
if src_serializer == to_serializer:
4329
self.target_repo.revisions.insert_record_stream(substream)
4330
else:
4331
self._extract_and_insert_revisions(substream,
4332
src_serializer)
4333
elif substream_type == 'signatures':
4334
self.target_repo.signatures.insert_record_stream(substream)
4335
else:
4336
raise AssertionError('kaboom! %s' % (substream_type,))
4337
# Done inserting data, and the missing_keys calculations will try to
4338
# read back from the inserted data, so flush the writes to the new pack
4339
# (if this is pack format).
4340
if new_pack is not None:
4341
new_pack._write_data('', flush=True)
4342
# Find all the new revisions (including ones from resume_tokens)
4343
missing_keys = self.target_repo.get_missing_parent_inventories(
4344
check_for_missing_texts=is_resume)
4345
try:
4346
for prefix, versioned_file in (
4347
('texts', self.target_repo.texts),
4348
('inventories', self.target_repo.inventories),
4349
('revisions', self.target_repo.revisions),
4350
('signatures', self.target_repo.signatures),
4351
('chk_bytes', self.target_repo.chk_bytes),
4352
):
4353
if versioned_file is None:
4354
continue
4355
# TODO: key is often going to be a StaticTuple object
4356
# I don't believe we can define a method by which
4357
# (prefix,) + StaticTuple will work, though we could
4358
# define a StaticTuple.sq_concat that would allow you to
4359
# pass in either a tuple or a StaticTuple as the second
4360
# object, so instead we could have:
4361
# StaticTuple(prefix) + key here...
4362
missing_keys.update((prefix,) + key for key in
4363
versioned_file.get_missing_compression_parent_keys())
4364
except NotImplementedError:
4365
# cannot even attempt suspending, and missing would have failed
4366
# during stream insertion.
4367
missing_keys = set()
4368
else:
4369
if missing_keys:
4370
# suspend the write group and tell the caller what we is
4371
# missing. We know we can suspend or else we would not have
4372
# entered this code path. (All repositories that can handle
4373
# missing keys can handle suspending a write group).
4374
write_group_tokens = self.target_repo.suspend_write_group()
4375
return write_group_tokens, missing_keys
4376
hint = self.target_repo.commit_write_group()
4377
if (to_serializer != src_serializer and
4378
self.target_repo._format.pack_compresses):
4379
self.target_repo.pack(hint=hint)
4380
return [], set()
4381
4382
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4383
target_rich_root = self.target_repo._format.rich_root_data
4384
target_tree_refs = self.target_repo._format.supports_tree_reference
4385
for record in substream:
4386
# Insert the delta directly
4387
inventory_delta_bytes = record.get_bytes_as('fulltext')
4388
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4389
try:
4390
parse_result = deserialiser.parse_text_bytes(
4391
inventory_delta_bytes)
4392
except inventory_delta.IncompatibleInventoryDelta, err:
4393
trace.mutter("Incompatible delta: %s", err.msg)
4394
raise errors.IncompatibleRevision(self.target_repo._format)
4395
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4396
revision_id = new_id
4397
parents = [key[0] for key in record.parents]
4398
self.target_repo.add_inventory_by_delta(
4399
basis_id, inv_delta, revision_id, parents)
4400
4401
def _extract_and_insert_inventories(self, substream, serializer,
4402
parse_delta=None):
4403
"""Generate a new inventory versionedfile in target, converting data.
4404
4405
The inventory is retrieved from the source, (deserializing it), and
4406
stored in the target (reserializing it in a different format).
4407
"""
4408
target_rich_root = self.target_repo._format.rich_root_data
4409
target_tree_refs = self.target_repo._format.supports_tree_reference
4410
for record in substream:
4411
# It's not a delta, so it must be a fulltext in the source
4412
# serializer's format.
4413
bytes = record.get_bytes_as('fulltext')
4414
revision_id = record.key[0]
4415
inv = serializer.read_inventory_from_string(bytes, revision_id)
4416
parents = [key[0] for key in record.parents]
4417
self.target_repo.add_inventory(revision_id, inv, parents)
4418
# No need to keep holding this full inv in memory when the rest of
4419
# the substream is likely to be all deltas.
4420
del inv
4421
4422
def _extract_and_insert_revisions(self, substream, serializer):
4423
for record in substream:
4424
bytes = record.get_bytes_as('fulltext')
4425
revision_id = record.key[0]
4426
rev = serializer.read_revision_from_string(bytes)
4427
if rev.revision_id != revision_id:
4428
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4429
self.target_repo.add_revision(revision_id, rev)
4430
4431
def finished(self):
4432
if self.target_repo._format._fetch_reconcile:
4433
self.target_repo.reconcile()
4434
4435
4436
class StreamSource(object):
4437
"""A source of a stream for fetching between repositories."""
4438
4439
def __init__(self, from_repository, to_format):
4440
"""Create a StreamSource streaming from from_repository."""
4441
self.from_repository = from_repository
4442
self.to_format = to_format
4443
self._record_counter = RecordCounter()
4444
4445
def delta_on_metadata(self):
4446
"""Return True if delta's are permitted on metadata streams.
4447
4448
That is on revisions and signatures.
4449
"""
4450
src_serializer = self.from_repository._format._serializer
4451
target_serializer = self.to_format._serializer
4452
return (self.to_format._fetch_uses_deltas and
4453
src_serializer == target_serializer)
4454
4455
def _fetch_revision_texts(self, revs):
4456
# fetch signatures first and then the revision texts
4457
# may need to be a InterRevisionStore call here.
4458
from_sf = self.from_repository.signatures
4459
# A missing signature is just skipped.
4460
keys = [(rev_id,) for rev_id in revs]
4461
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4462
keys,
4463
self.to_format._fetch_order,
4464
not self.to_format._fetch_uses_deltas))
4465
# If a revision has a delta, this is actually expanded inside the
4466
# insert_record_stream code now, which is an alternate fix for
4467
# bug #261339
4468
from_rf = self.from_repository.revisions
4469
revisions = from_rf.get_record_stream(
4470
keys,
4471
self.to_format._fetch_order,
4472
not self.delta_on_metadata())
4473
return [('signatures', signatures), ('revisions', revisions)]
4474
4475
def _generate_root_texts(self, revs):
4476
"""This will be called by get_stream between fetching weave texts and
4477
fetching the inventory weave.
4478
"""
4479
if self._rich_root_upgrade():
4480
return _mod_fetch.Inter1and2Helper(
4481
self.from_repository).generate_root_texts(revs)
4482
else:
4483
return []
4484
4485
def get_stream(self, search):
4486
phase = 'file'
4487
revs = search.get_keys()
4488
graph = self.from_repository.get_graph()
4489
revs = tsort.topo_sort(graph.get_parent_map(revs))
4490
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4491
text_keys = []
4492
for knit_kind, file_id, revisions in data_to_fetch:
4493
if knit_kind != phase:
4494
phase = knit_kind
4495
# Make a new progress bar for this phase
4496
if knit_kind == "file":
4497
# Accumulate file texts
4498
text_keys.extend([(file_id, revision) for revision in
4499
revisions])
4500
elif knit_kind == "inventory":
4501
# Now copy the file texts.
4502
from_texts = self.from_repository.texts
4503
yield ('texts', from_texts.get_record_stream(
4504
text_keys, self.to_format._fetch_order,
4505
not self.to_format._fetch_uses_deltas))
4506
# Cause an error if a text occurs after we have done the
4507
# copy.
4508
text_keys = None
4509
# Before we process the inventory we generate the root
4510
# texts (if necessary) so that the inventories references
4511
# will be valid.
4512
for _ in self._generate_root_texts(revs):
4513
yield _
4514
# we fetch only the referenced inventories because we do not
4515
# know for unselected inventories whether all their required
4516
# texts are present in the other repository - it could be
4517
# corrupt.
4518
for info in self._get_inventory_stream(revs):
4519
yield info
4520
elif knit_kind == "signatures":
4521
# Nothing to do here; this will be taken care of when
4522
# _fetch_revision_texts happens.
4523
pass
4524
elif knit_kind == "revisions":
4525
for record in self._fetch_revision_texts(revs):
4526
yield record
4527
else:
4528
raise AssertionError("Unknown knit kind %r" % knit_kind)
4529
4530
def get_stream_for_missing_keys(self, missing_keys):
4531
# missing keys can only occur when we are byte copying and not
4532
# translating (because translation means we don't send
4533
# unreconstructable deltas ever).
4534
keys = {}
4535
keys['texts'] = set()
4536
keys['revisions'] = set()
4537
keys['inventories'] = set()
4538
keys['chk_bytes'] = set()
4539
keys['signatures'] = set()
4540
for key in missing_keys:
4541
keys[key[0]].add(key[1:])
4542
if len(keys['revisions']):
4543
# If we allowed copying revisions at this point, we could end up
4544
# copying a revision without copying its required texts: a
4545
# violation of the requirements for repository integrity.
4546
raise AssertionError(
4547
'cannot copy revisions to fill in missing deltas %s' % (
4548
keys['revisions'],))
4549
for substream_kind, keys in keys.iteritems():
4550
vf = getattr(self.from_repository, substream_kind)
4551
if vf is None and keys:
4552
raise AssertionError(
4553
"cannot fill in keys for a versioned file we don't"
4554
" have: %s needs %s" % (substream_kind, keys))
4555
if not keys:
4556
# No need to stream something we don't have
4557
continue
4558
if substream_kind == 'inventories':
4559
# Some missing keys are genuinely ghosts, filter those out.
4560
present = self.from_repository.inventories.get_parent_map(keys)
4561
revs = [key[0] for key in present]
4562
# Get the inventory stream more-or-less as we do for the
4563
# original stream; there's no reason to assume that records
4564
# direct from the source will be suitable for the sink. (Think
4565
# e.g. 2a -> 1.9-rich-root).
4566
for info in self._get_inventory_stream(revs, missing=True):
4567
yield info
4568
continue
4569
4570
# Ask for full texts always so that we don't need more round trips
4571
# after this stream.
4572
# Some of the missing keys are genuinely ghosts, so filter absent
4573
# records. The Sink is responsible for doing another check to
4574
# ensure that ghosts don't introduce missing data for future
4575
# fetches.
4576
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4577
self.to_format._fetch_order, True))
4578
yield substream_kind, stream
4579
4580
def inventory_fetch_order(self):
4581
if self._rich_root_upgrade():
4582
return 'topological'
4583
else:
4584
return self.to_format._fetch_order
4585
4586
def _rich_root_upgrade(self):
4587
return (not self.from_repository._format.rich_root_data and
4588
self.to_format.rich_root_data)
4589
4590
def _get_inventory_stream(self, revision_ids, missing=False):
4591
from_format = self.from_repository._format
4592
if (from_format.supports_chks and self.to_format.supports_chks and
4593
from_format.network_name() == self.to_format.network_name()):
4594
raise AssertionError(
4595
"this case should be handled by GroupCHKStreamSource")
4596
elif 'forceinvdeltas' in debug.debug_flags:
4597
return self._get_convertable_inventory_stream(revision_ids,
4598
delta_versus_null=missing)
4599
elif from_format.network_name() == self.to_format.network_name():
4600
# Same format.
4601
return self._get_simple_inventory_stream(revision_ids,
4602
missing=missing)
4603
elif (not from_format.supports_chks and not self.to_format.supports_chks
4604
and from_format._serializer == self.to_format._serializer):
4605
# Essentially the same format.
4606
return self._get_simple_inventory_stream(revision_ids,
4607
missing=missing)
4608
else:
4609
# Any time we switch serializations, we want to use an
4610
# inventory-delta based approach.
4611
return self._get_convertable_inventory_stream(revision_ids,
4612
delta_versus_null=missing)
4613
4614
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4615
# NB: This currently reopens the inventory weave in source;
4616
# using a single stream interface instead would avoid this.
4617
from_weave = self.from_repository.inventories
4618
if missing:
4619
delta_closure = True
4620
else:
4621
delta_closure = not self.delta_on_metadata()
4622
yield ('inventories', from_weave.get_record_stream(
4623
[(rev_id,) for rev_id in revision_ids],
4624
self.inventory_fetch_order(), delta_closure))
4625
4626
def _get_convertable_inventory_stream(self, revision_ids,
4627
delta_versus_null=False):
4628
# The two formats are sufficiently different that there is no fast
4629
# path, so we need to send just inventorydeltas, which any
4630
# sufficiently modern client can insert into any repository.
4631
# The StreamSink code expects to be able to
4632
# convert on the target, so we need to put bytes-on-the-wire that can
4633
# be converted. That means inventory deltas (if the remote is <1.19,
4634
# RemoteStreamSink will fallback to VFS to insert the deltas).
4635
yield ('inventory-deltas',
4636
self._stream_invs_as_deltas(revision_ids,
4637
delta_versus_null=delta_versus_null))
4638
4639
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4640
"""Return a stream of inventory-deltas for the given rev ids.
4641
4642
:param revision_ids: The list of inventories to transmit
4643
:param delta_versus_null: Don't try to find a minimal delta for this
4644
entry, instead compute the delta versus the NULL_REVISION. This
4645
effectively streams a complete inventory. Used for stuff like
4646
filling in missing parents, etc.
4647
"""
4648
from_repo = self.from_repository
4649
revision_keys = [(rev_id,) for rev_id in revision_ids]
4650
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4651
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4652
# method...
4653
inventories = self.from_repository.iter_inventories(
4654
revision_ids, 'topological')
4655
format = from_repo._format
4656
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4657
inventory_cache = lru_cache.LRUCache(50)
4658
null_inventory = from_repo.revision_tree(
4659
_mod_revision.NULL_REVISION).inventory
4660
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4661
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4662
# repo back into a non-rich-root repo ought to be allowed)
4663
serializer = inventory_delta.InventoryDeltaSerializer(
4664
versioned_root=format.rich_root_data,
4665
tree_references=format.supports_tree_reference)
4666
for inv in inventories:
4667
key = (inv.revision_id,)
4668
parent_keys = parent_map.get(key, ())
4669
delta = None
4670
if not delta_versus_null and parent_keys:
4671
# The caller did not ask for complete inventories and we have
4672
# some parents that we can delta against. Make a delta against
4673
# each parent so that we can find the smallest.
4674
parent_ids = [parent_key[0] for parent_key in parent_keys]
4675
for parent_id in parent_ids:
4676
if parent_id not in invs_sent_so_far:
4677
# We don't know that the remote side has this basis, so
4678
# we can't use it.
4679
continue
4680
if parent_id == _mod_revision.NULL_REVISION:
4681
parent_inv = null_inventory
4682
else:
4683
parent_inv = inventory_cache.get(parent_id, None)
4684
if parent_inv is None:
4685
parent_inv = from_repo.get_inventory(parent_id)
4686
candidate_delta = inv._make_delta(parent_inv)
4687
if (delta is None or
4688
len(delta) > len(candidate_delta)):
4689
delta = candidate_delta
4690
basis_id = parent_id
4691
if delta is None:
4692
# Either none of the parents ended up being suitable, or we
4693
# were asked to delta against NULL
4694
basis_id = _mod_revision.NULL_REVISION
4695
delta = inv._make_delta(null_inventory)
4696
invs_sent_so_far.add(inv.revision_id)
4697
inventory_cache[inv.revision_id] = inv
4698
delta_serialized = ''.join(
4699
serializer.delta_to_lines(basis_id, key[-1], delta))
4700
yield versionedfile.FulltextContentFactory(
4701
key, parent_keys, None, delta_serialized)
4702
4703
4704
1800
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4705
1801
stop_revision=None):
4706
1802
"""Extend the partial history to include a given index
4716
1812
it is encountered, history extension will stop.
4717
1813
"""
4718
1814
start_revision = partial_history_cache[-1]
4719
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,))
4720
1818
try:
4721
#skip the last revision in the list
1819
# skip the last revision in the list
4722
1820
iterator.next()
4723
1821
while True:
4724
1822
if (stop_index is not None and
4732
1830
# No more history
4733
1831
return
4734
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