« back to all changes in this revision
Viewing changes to bzrlib/repository.py
- browse files at revision 6613.2.1
- compare with another revision
- download diff
- download tarball
- view history from revision 6613.2.1
- Committer: Richard Wilbur
- Date: 2016-02-04 19:07:28 UTC
- mto: This revision was merged to the branch mainline in revision 6618.
- Revision ID: richard.wilbur@gmail.com-20160204190728-p0zvfii6zase0fw7
Update COPYING.txt from the original http://www.gnu.org/licenses/gpl-2.0.txt (Only differences were in whitespace.) Thanks to Petr Stodulka for pointing out the discrepancy.
- files added:
- bzrlib/_termcolor.py
- bzrlib/branchfmt
- bzrlib/plugins/grep
- bzrlib/plugins/po_merge
- bzrlib/plugins/po_merge/tests
- po
- files removed:
- bzrlib/deprecated_graph.py
- bzrlib/doc_generate/builders
- bzrlib/doc_generate/writers
- bzrlib/tests/doc_generate
- bzrlib/tests/doc_generate/builders
- bzrlib/tests/doc_generate/writers
- bzrlib/util/effbot
- bzrlib/util/effbot/org
- bzrlib/util/elementtree
- files renamed:
- bzrlib/sign_my_commits.py => bzrlib/commit_signature_commands.py
- bzrlib/tests/per_repository/test_is_write_locked.py => bzrlib/tests/per_repository/test_locking.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_find_text_key_references.py => bzrlib/tests/per_repository_vf/test_find_text_key_references.py
- bzrlib/tests/per_repository/test_merge_directive.py => bzrlib/tests/per_repository_vf/test_merge_directive.py
- files modified:
- .bzrignore
added
removed
bzrlib/repository.py
14
14
# along with this program; if not, write to the Free Software
15
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
16
16
17
from __future__ import absolute_import
18
17
19
from bzrlib.lazy_import import lazy_import
18
20
lazy_import(globals(), """
21
import itertools
19
22
import time
20
23
21
24
from bzrlib import (
22
bzrdir,
23
check,
24
25
config,
25
26
controldir,
26
27
debug,
27
fetch as _mod_fetch,
28
fifo_cache,
29
28
generate_ids,
30
gpg,
31
29
graph,
32
inventory_delta,
33
30
lockable_files,
34
31
lockdir,
35
lru_cache,
36
32
osutils,
37
33
revision as _mod_revision,
38
static_tuple,
34
testament as _mod_testament,
39
35
tsort,
40
versionedfile,
36
gpg,
41
37
)
42
38
from bzrlib.bundle import serializer
43
from bzrlib.recordcounter import RecordCounter
44
from bzrlib.revisiontree import InventoryRevisionTree
45
from bzrlib.store.versioned import VersionedFileStore
46
from bzrlib.testament import Testament
39
from bzrlib.i18n import gettext
47
40
""")
48
41
49
42
from bzrlib import (
43
bzrdir,
50
44
errors,
51
45
registry,
52
46
symbol_versioning,
54
48
)
55
49
from bzrlib.decorators import needs_read_lock, needs_write_lock, only_raises
56
50
from bzrlib.inter import InterObject
57
from bzrlib.inventory import (
58
Inventory,
59
InventoryDirectory,
60
ROOT_ID,
61
entry_factory,
62
)
63
51
from bzrlib.lock import _RelockDebugMixin, LogicalLockResult
64
52
from bzrlib.trace import (
65
53
log_exception_quietly, note, mutter, mutter_callsite, warning)
86
74
87
75
# all clients should supply tree roots.
88
76
record_root_entry = True
89
# the default CommitBuilder does not manage trees whose root is versioned.
90
_versioned_root = False
91
# this commit builder supports the record_entry_contents interface
92
supports_record_entry_contents = True
77
# whether this commit builder supports the record_entry_contents interface
78
supports_record_entry_contents = False
79
# whether this commit builder will automatically update the branch that is
80
# being committed to
81
updates_branch = False
93
82
94
def __init__(self, repository, parents, config, timestamp=None,
83
def __init__(self, repository, parents, config_stack, timestamp=None,
95
84
timezone=None, committer=None, revprops=None,
96
85
revision_id=None, lossy=False):
97
86
"""Initiate a CommitBuilder.
106
95
:param lossy: Whether to discard data that can not be natively
107
96
represented, when pushing to a foreign VCS
108
97
"""
109
self._config = config
98
self._config_stack = config_stack
110
99
self._lossy = lossy
111
100
112
101
if committer is None:
113
self._committer = self._config.username()
102
self._committer = self._config_stack.get('email')
114
103
elif not isinstance(committer, unicode):
115
104
self._committer = committer.decode() # throw if non-ascii
116
105
else:
117
106
self._committer = committer
118
107
119
self.new_inventory = Inventory(None)
120
108
self._new_revision_id = revision_id
121
109
self.parents = parents
122
110
self.repository = repository
137
125
self._timezone = int(timezone)
138
126
139
127
self._generate_revision_if_needed()
140
self.__heads = graph.HeadsCache(repository.get_graph()).heads
141
self._basis_delta = []
142
# API compatibility, older code that used CommitBuilder did not call
143
# .record_delete(), which means the delta that is computed would not be
144
# valid. Callers that will call record_delete() should call
145
# .will_record_deletes() to indicate that.
146
self._recording_deletes = False
147
# memo'd check for no-op commits.
148
self._any_changes = False
149
128
150
129
def any_changes(self):
151
130
"""Return True if any entries were changed.
152
131
153
132
This includes merge-only changes. It is the core for the --unchanged
154
133
detection in commit.
155
134
156
135
:return: True if any changes have occured.
157
136
"""
158
return self._any_changes
137
raise NotImplementedError(self.any_changes)
159
138
160
139
def _validate_unicode_text(self, text, context):
161
140
"""Verify things like commit messages don't have bogus characters."""
172
151
self._validate_unicode_text(value,
173
152
'revision property (%s)' % (key,))
174
153
175
def _ensure_fallback_inventories(self):
176
"""Ensure that appropriate inventories are available.
177
178
This only applies to repositories that are stacked, and is about
179
enusring the stacking invariants. Namely, that for any revision that is
180
present, we either have all of the file content, or we have the parent
181
inventory and the delta file content.
182
"""
183
if not self.repository._fallback_repositories:
184
return
185
if not self.repository._format.supports_chks:
186
raise errors.BzrError("Cannot commit directly to a stacked branch"
187
" in pre-2a formats. See "
188
"https://bugs.launchpad.net/bzr/+bug/375013 for details.")
189
# This is a stacked repo, we need to make sure we have the parent
190
# inventories for the parents.
191
parent_keys = [(p,) for p in self.parents]
192
parent_map = self.repository.inventories._index.get_parent_map(parent_keys)
193
missing_parent_keys = set([pk for pk in parent_keys
194
if pk not in parent_map])
195
fallback_repos = list(reversed(self.repository._fallback_repositories))
196
missing_keys = [('inventories', pk[0])
197
for pk in missing_parent_keys]
198
resume_tokens = []
199
while missing_keys and fallback_repos:
200
fallback_repo = fallback_repos.pop()
201
source = fallback_repo._get_source(self.repository._format)
202
sink = self.repository._get_sink()
203
stream = source.get_stream_for_missing_keys(missing_keys)
204
missing_keys = sink.insert_stream_without_locking(stream,
205
self.repository._format)
206
if missing_keys:
207
raise errors.BzrError('Unable to fill in parent inventories for a'
208
' stacked branch')
209
210
154
def commit(self, message):
211
155
"""Make the actual commit.
212
156
213
157
:return: The revision id of the recorded revision.
214
158
"""
215
self._validate_unicode_text(message, 'commit message')
216
rev = _mod_revision.Revision(
217
timestamp=self._timestamp,
218
timezone=self._timezone,
219
committer=self._committer,
220
message=message,
221
inventory_sha1=self.inv_sha1,
222
revision_id=self._new_revision_id,
223
properties=self._revprops)
224
rev.parent_ids = self.parents
225
self.repository.add_revision(self._new_revision_id, rev,
226
self.new_inventory, self._config)
227
self._ensure_fallback_inventories()
228
self.repository.commit_write_group()
229
return self._new_revision_id
159
raise NotImplementedError(self.commit)
230
160
231
161
def abort(self):
232
162
"""Abort the commit that is being built.
233
163
"""
234
self.repository.abort_write_group()
164
raise NotImplementedError(self.abort)
235
165
236
166
def revision_tree(self):
237
167
"""Return the tree that was just committed.
238
168
239
169
After calling commit() this can be called to get a
240
InventoryRevisionTree representing the newly committed tree. This is
170
RevisionTree representing the newly committed tree. This is
241
171
preferred to calling Repository.revision_tree() because that may
242
172
require deserializing the inventory, while we already have a copy in
243
173
memory.
244
174
"""
245
if self.new_inventory is None:
246
self.new_inventory = self.repository.get_inventory(
247
self._new_revision_id)
248
return InventoryRevisionTree(self.repository, self.new_inventory,
249
self._new_revision_id)
175
raise NotImplementedError(self.revision_tree)
250
176
251
177
def finish_inventory(self):
252
178
"""Tell the builder that the inventory is finished.
254
180
:return: The inventory id in the repository, which can be used with
255
181
repository.get_inventory.
256
182
"""
257
if self.new_inventory is None:
258
# an inventory delta was accumulated without creating a new
259
# inventory.
260
basis_id = self.basis_delta_revision
261
# We ignore the 'inventory' returned by add_inventory_by_delta
262
# because self.new_inventory is used to hint to the rest of the
263
# system what code path was taken
264
self.inv_sha1, _ = self.repository.add_inventory_by_delta(
265
basis_id, self._basis_delta, self._new_revision_id,
266
self.parents)
267
else:
268
if self.new_inventory.root is None:
269
raise AssertionError('Root entry should be supplied to'
270
' record_entry_contents, as of bzr 0.10.')
271
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
272
self.new_inventory.revision_id = self._new_revision_id
273
self.inv_sha1 = self.repository.add_inventory(
274
self._new_revision_id,
275
self.new_inventory,
276
self.parents
277
)
278
return self._new_revision_id
183
raise NotImplementedError(self.finish_inventory)
279
184
280
185
def _gen_revision_id(self):
281
186
"""Return new revision-id."""
296
201
else:
297
202
self.random_revid = False
298
203
299
def _heads(self, file_id, revision_ids):
300
"""Calculate the graph heads for revision_ids in the graph of file_id.
301
302
This can use either a per-file graph or a global revision graph as we
303
have an identity relationship between the two graphs.
304
"""
305
return self.__heads(revision_ids)
306
307
def _check_root(self, ie, parent_invs, tree):
308
"""Helper for record_entry_contents.
309
310
:param ie: An entry being added.
311
:param parent_invs: The inventories of the parent revisions of the
312
commit.
313
:param tree: The tree that is being committed.
314
"""
315
# In this revision format, root entries have no knit or weave When
316
# serializing out to disk and back in root.revision is always
317
# _new_revision_id
318
ie.revision = self._new_revision_id
319
320
def _require_root_change(self, tree):
321
"""Enforce an appropriate root object change.
322
323
This is called once when record_iter_changes is called, if and only if
324
the root was not in the delta calculated by record_iter_changes.
325
326
:param tree: The tree which is being committed.
327
"""
328
if len(self.parents) == 0:
329
raise errors.RootMissing()
330
entry = entry_factory['directory'](tree.path2id(''), '',
331
None)
332
entry.revision = self._new_revision_id
333
self._basis_delta.append(('', '', entry.file_id, entry))
334
335
def _get_delta(self, ie, basis_inv, path):
336
"""Get a delta against the basis inventory for ie."""
337
if ie.file_id not in basis_inv:
338
# add
339
result = (None, path, ie.file_id, ie)
340
self._basis_delta.append(result)
341
return result
342
elif ie != basis_inv[ie.file_id]:
343
# common but altered
344
# TODO: avoid tis id2path call.
345
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
346
self._basis_delta.append(result)
347
return result
348
else:
349
# common, unaltered
350
return None
351
352
def get_basis_delta(self):
353
"""Return the complete inventory delta versus the basis inventory.
354
355
This has been built up with the calls to record_delete and
356
record_entry_contents. The client must have already called
357
will_record_deletes() to indicate that they will be generating a
358
complete delta.
359
360
:return: An inventory delta, suitable for use with apply_delta, or
361
Repository.add_inventory_by_delta, etc.
362
"""
363
if not self._recording_deletes:
364
raise AssertionError("recording deletes not activated.")
365
return self._basis_delta
366
367
def record_delete(self, path, file_id):
368
"""Record that a delete occured against a basis tree.
369
370
This is an optional API - when used it adds items to the basis_delta
371
being accumulated by the commit builder. It cannot be called unless the
372
method will_record_deletes() has been called to inform the builder that
373
a delta is being supplied.
374
375
:param path: The path of the thing deleted.
376
:param file_id: The file id that was deleted.
377
"""
378
if not self._recording_deletes:
379
raise AssertionError("recording deletes not activated.")
380
delta = (path, None, file_id, None)
381
self._basis_delta.append(delta)
382
self._any_changes = True
383
return delta
384
385
204
def will_record_deletes(self):
386
205
"""Tell the commit builder that deletes are being notified.
387
206
389
208
commit to be valid, deletes against the basis MUST be recorded via
390
209
builder.record_delete().
391
210
"""
392
self._recording_deletes = True
393
try:
394
basis_id = self.parents[0]
395
except IndexError:
396
basis_id = _mod_revision.NULL_REVISION
397
self.basis_delta_revision = basis_id
398
399
def record_entry_contents(self, ie, parent_invs, path, tree,
400
content_summary):
401
"""Record the content of ie from tree into the commit if needed.
402
403
Side effect: sets ie.revision when unchanged
404
405
:param ie: An inventory entry present in the commit.
406
:param parent_invs: The inventories of the parent revisions of the
407
commit.
408
:param path: The path the entry is at in the tree.
409
:param tree: The tree which contains this entry and should be used to
410
obtain content.
411
:param content_summary: Summary data from the tree about the paths
412
content - stat, length, exec, sha/link target. This is only
413
accessed when the entry has a revision of None - that is when it is
414
a candidate to commit.
415
:return: A tuple (change_delta, version_recorded, fs_hash).
416
change_delta is an inventory_delta change for this entry against
417
the basis tree of the commit, or None if no change occured against
418
the basis tree.
419
version_recorded is True if a new version of the entry has been
420
recorded. For instance, committing a merge where a file was only
421
changed on the other side will return (delta, False).
422
fs_hash is either None, or the hash details for the path (currently
423
a tuple of the contents sha1 and the statvalue returned by
424
tree.get_file_with_stat()).
425
"""
426
if self.new_inventory.root is None:
427
if ie.parent_id is not None:
428
raise errors.RootMissing()
429
self._check_root(ie, parent_invs, tree)
430
if ie.revision is None:
431
kind = content_summary[0]
432
else:
433
# ie is carried over from a prior commit
434
kind = ie.kind
435
# XXX: repository specific check for nested tree support goes here - if
436
# the repo doesn't want nested trees we skip it ?
437
if (kind == 'tree-reference' and
438
not self.repository._format.supports_tree_reference):
439
# mismatch between commit builder logic and repository:
440
# this needs the entry creation pushed down into the builder.
441
raise NotImplementedError('Missing repository subtree support.')
442
self.new_inventory.add(ie)
443
444
# TODO: slow, take it out of the inner loop.
445
try:
446
basis_inv = parent_invs[0]
447
except IndexError:
448
basis_inv = Inventory(root_id=None)
449
450
# ie.revision is always None if the InventoryEntry is considered
451
# for committing. We may record the previous parents revision if the
452
# content is actually unchanged against a sole head.
453
if ie.revision is not None:
454
if not self._versioned_root and path == '':
455
# repositories that do not version the root set the root's
456
# revision to the new commit even when no change occurs (more
457
# specifically, they do not record a revision on the root; and
458
# the rev id is assigned to the root during deserialisation -
459
# this masks when a change may have occurred against the basis.
460
# To match this we always issue a delta, because the revision
461
# of the root will always be changing.
462
if ie.file_id in basis_inv:
463
delta = (basis_inv.id2path(ie.file_id), path,
464
ie.file_id, ie)
465
else:
466
# add
467
delta = (None, path, ie.file_id, ie)
468
self._basis_delta.append(delta)
469
return delta, False, None
470
else:
471
# we don't need to commit this, because the caller already
472
# determined that an existing revision of this file is
473
# appropriate. If it's not being considered for committing then
474
# it and all its parents to the root must be unaltered so
475
# no-change against the basis.
476
if ie.revision == self._new_revision_id:
477
raise AssertionError("Impossible situation, a skipped "
478
"inventory entry (%r) claims to be modified in this "
479
"commit (%r).", (ie, self._new_revision_id))
480
return None, False, None
481
# XXX: Friction: parent_candidates should return a list not a dict
482
# so that we don't have to walk the inventories again.
483
parent_candiate_entries = ie.parent_candidates(parent_invs)
484
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
485
heads = []
486
for inv in parent_invs:
487
if ie.file_id in inv:
488
old_rev = inv[ie.file_id].revision
489
if old_rev in head_set:
490
heads.append(inv[ie.file_id].revision)
491
head_set.remove(inv[ie.file_id].revision)
492
493
store = False
494
# now we check to see if we need to write a new record to the
495
# file-graph.
496
# We write a new entry unless there is one head to the ancestors, and
497
# the kind-derived content is unchanged.
498
499
# Cheapest check first: no ancestors, or more the one head in the
500
# ancestors, we write a new node.
501
if len(heads) != 1:
502
store = True
503
if not store:
504
# There is a single head, look it up for comparison
505
parent_entry = parent_candiate_entries[heads[0]]
506
# if the non-content specific data has changed, we'll be writing a
507
# node:
508
if (parent_entry.parent_id != ie.parent_id or
509
parent_entry.name != ie.name):
510
store = True
511
# now we need to do content specific checks:
512
if not store:
513
# if the kind changed the content obviously has
514
if kind != parent_entry.kind:
515
store = True
516
# Stat cache fingerprint feedback for the caller - None as we usually
517
# don't generate one.
518
fingerprint = None
519
if kind == 'file':
520
if content_summary[2] is None:
521
raise ValueError("Files must not have executable = None")
522
if not store:
523
# We can't trust a check of the file length because of content
524
# filtering...
525
if (# if the exec bit has changed we have to store:
526
parent_entry.executable != content_summary[2]):
527
store = True
528
elif parent_entry.text_sha1 == content_summary[3]:
529
# all meta and content is unchanged (using a hash cache
530
# hit to check the sha)
531
ie.revision = parent_entry.revision
532
ie.text_size = parent_entry.text_size
533
ie.text_sha1 = parent_entry.text_sha1
534
ie.executable = parent_entry.executable
535
return self._get_delta(ie, basis_inv, path), False, None
536
else:
537
# Either there is only a hash change(no hash cache entry,
538
# or same size content change), or there is no change on
539
# this file at all.
540
# Provide the parent's hash to the store layer, so that the
541
# content is unchanged we will not store a new node.
542
nostore_sha = parent_entry.text_sha1
543
if store:
544
# We want to record a new node regardless of the presence or
545
# absence of a content change in the file.
546
nostore_sha = None
547
ie.executable = content_summary[2]
548
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
549
try:
550
text = file_obj.read()
551
finally:
552
file_obj.close()
553
try:
554
ie.text_sha1, ie.text_size = self._add_text_to_weave(
555
ie.file_id, text, heads, nostore_sha)
556
# Let the caller know we generated a stat fingerprint.
557
fingerprint = (ie.text_sha1, stat_value)
558
except errors.ExistingContent:
559
# Turns out that the file content was unchanged, and we were
560
# only going to store a new node if it was changed. Carry over
561
# the entry.
562
ie.revision = parent_entry.revision
563
ie.text_size = parent_entry.text_size
564
ie.text_sha1 = parent_entry.text_sha1
565
ie.executable = parent_entry.executable
566
return self._get_delta(ie, basis_inv, path), False, None
567
elif kind == 'directory':
568
if not store:
569
# all data is meta here, nothing specific to directory, so
570
# carry over:
571
ie.revision = parent_entry.revision
572
return self._get_delta(ie, basis_inv, path), False, None
573
self._add_text_to_weave(ie.file_id, '', heads, None)
574
elif kind == 'symlink':
575
current_link_target = content_summary[3]
576
if not store:
577
# symlink target is not generic metadata, check if it has
578
# changed.
579
if current_link_target != parent_entry.symlink_target:
580
store = True
581
if not store:
582
# unchanged, carry over.
583
ie.revision = parent_entry.revision
584
ie.symlink_target = parent_entry.symlink_target
585
return self._get_delta(ie, basis_inv, path), False, None
586
ie.symlink_target = current_link_target
587
self._add_text_to_weave(ie.file_id, '', heads, None)
588
elif kind == 'tree-reference':
589
if not store:
590
if content_summary[3] != parent_entry.reference_revision:
591
store = True
592
if not store:
593
# unchanged, carry over.
594
ie.reference_revision = parent_entry.reference_revision
595
ie.revision = parent_entry.revision
596
return self._get_delta(ie, basis_inv, path), False, None
597
ie.reference_revision = content_summary[3]
598
if ie.reference_revision is None:
599
raise AssertionError("invalid content_summary for nested tree: %r"
600
% (content_summary,))
601
self._add_text_to_weave(ie.file_id, '', heads, None)
602
else:
603
raise NotImplementedError('unknown kind')
604
ie.revision = self._new_revision_id
605
self._any_changes = True
606
return self._get_delta(ie, basis_inv, path), True, fingerprint
607
608
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
609
_entry_factory=entry_factory):
211
raise NotImplementedError(self.will_record_deletes)
212
213
def record_iter_changes(self, tree, basis_revision_id, iter_changes):
610
214
"""Record a new tree via iter_changes.
611
215
612
216
:param tree: The tree to obtain text contents from for changed objects.
617
221
to basis_revision_id. The iterator must not include any items with
618
222
a current kind of None - missing items must be either filtered out
619
223
or errored-on beefore record_iter_changes sees the item.
620
:param _entry_factory: Private method to bind entry_factory locally for
621
performance.
622
224
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
623
225
tree._observed_sha1.
624
226
"""
625
# Create an inventory delta based on deltas between all the parents and
626
# deltas between all the parent inventories. We use inventory delta's
627
# between the inventory objects because iter_changes masks
628
# last-changed-field only changes.
629
# Working data:
630
# file_id -> change map, change is fileid, paths, changed, versioneds,
631
# parents, names, kinds, executables
632
merged_ids = {}
633
# {file_id -> revision_id -> inventory entry, for entries in parent
634
# trees that are not parents[0]
635
parent_entries = {}
636
ghost_basis = False
637
try:
638
revtrees = list(self.repository.revision_trees(self.parents))
639
except errors.NoSuchRevision:
640
# one or more ghosts, slow path.
641
revtrees = []
642
for revision_id in self.parents:
643
try:
644
revtrees.append(self.repository.revision_tree(revision_id))
645
except errors.NoSuchRevision:
646
if not revtrees:
647
basis_revision_id = _mod_revision.NULL_REVISION
648
ghost_basis = True
649
revtrees.append(self.repository.revision_tree(
650
_mod_revision.NULL_REVISION))
651
# The basis inventory from a repository
652
if revtrees:
653
basis_inv = revtrees[0].inventory
654
else:
655
basis_inv = self.repository.revision_tree(
656
_mod_revision.NULL_REVISION).inventory
657
if len(self.parents) > 0:
658
if basis_revision_id != self.parents[0] and not ghost_basis:
659
raise Exception(
660
"arbitrary basis parents not yet supported with merges")
661
for revtree in revtrees[1:]:
662
for change in revtree.inventory._make_delta(basis_inv):
663
if change[1] is None:
664
# Not present in this parent.
665
continue
666
if change[2] not in merged_ids:
667
if change[0] is not None:
668
basis_entry = basis_inv[change[2]]
669
merged_ids[change[2]] = [
670
# basis revid
671
basis_entry.revision,
672
# new tree revid
673
change[3].revision]
674
parent_entries[change[2]] = {
675
# basis parent
676
basis_entry.revision:basis_entry,
677
# this parent
678
change[3].revision:change[3],
679
}
680
else:
681
merged_ids[change[2]] = [change[3].revision]
682
parent_entries[change[2]] = {change[3].revision:change[3]}
683
else:
684
merged_ids[change[2]].append(change[3].revision)
685
parent_entries[change[2]][change[3].revision] = change[3]
686
else:
687
merged_ids = {}
688
# Setup the changes from the tree:
689
# changes maps file_id -> (change, [parent revision_ids])
690
changes= {}
691
for change in iter_changes:
692
# This probably looks up in basis_inv way to much.
693
if change[1][0] is not None:
694
head_candidate = [basis_inv[change[0]].revision]
695
else:
696
head_candidate = []
697
changes[change[0]] = change, merged_ids.get(change[0],
698
head_candidate)
699
unchanged_merged = set(merged_ids) - set(changes)
700
# Extend the changes dict with synthetic changes to record merges of
701
# texts.
702
for file_id in unchanged_merged:
703
# Record a merged version of these items that did not change vs the
704
# basis. This can be either identical parallel changes, or a revert
705
# of a specific file after a merge. The recorded content will be
706
# that of the current tree (which is the same as the basis), but
707
# the per-file graph will reflect a merge.
708
# NB:XXX: We are reconstructing path information we had, this
709
# should be preserved instead.
710
# inv delta change: (file_id, (path_in_source, path_in_target),
711
# changed_content, versioned, parent, name, kind,
712
# executable)
713
try:
714
basis_entry = basis_inv[file_id]
715
except errors.NoSuchId:
716
# a change from basis->some_parents but file_id isn't in basis
717
# so was new in the merge, which means it must have changed
718
# from basis -> current, and as it hasn't the add was reverted
719
# by the user. So we discard this change.
720
pass
721
else:
722
change = (file_id,
723
(basis_inv.id2path(file_id), tree.id2path(file_id)),
724
False, (True, True),
725
(basis_entry.parent_id, basis_entry.parent_id),
726
(basis_entry.name, basis_entry.name),
727
(basis_entry.kind, basis_entry.kind),
728
(basis_entry.executable, basis_entry.executable))
729
changes[file_id] = (change, merged_ids[file_id])
730
# changes contains tuples with the change and a set of inventory
731
# candidates for the file.
732
# inv delta is:
733
# old_path, new_path, file_id, new_inventory_entry
734
seen_root = False # Is the root in the basis delta?
735
inv_delta = self._basis_delta
736
modified_rev = self._new_revision_id
737
for change, head_candidates in changes.values():
738
if change[3][1]: # versioned in target.
739
# Several things may be happening here:
740
# We may have a fork in the per-file graph
741
# - record a change with the content from tree
742
# We may have a change against < all trees
743
# - carry over the tree that hasn't changed
744
# We may have a change against all trees
745
# - record the change with the content from tree
746
kind = change[6][1]
747
file_id = change[0]
748
entry = _entry_factory[kind](file_id, change[5][1],
749
change[4][1])
750
head_set = self._heads(change[0], set(head_candidates))
751
heads = []
752
# Preserve ordering.
753
for head_candidate in head_candidates:
754
if head_candidate in head_set:
755
heads.append(head_candidate)
756
head_set.remove(head_candidate)
757
carried_over = False
758
if len(heads) == 1:
759
# Could be a carry-over situation:
760
parent_entry_revs = parent_entries.get(file_id, None)
761
if parent_entry_revs:
762
parent_entry = parent_entry_revs.get(heads[0], None)
763
else:
764
parent_entry = None
765
if parent_entry is None:
766
# The parent iter_changes was called against is the one
767
# that is the per-file head, so any change is relevant
768
# iter_changes is valid.
769
carry_over_possible = False
770
else:
771
# could be a carry over situation
772
# A change against the basis may just indicate a merge,
773
# we need to check the content against the source of the
774
# merge to determine if it was changed after the merge
775
# or carried over.
776
if (parent_entry.kind != entry.kind or
777
parent_entry.parent_id != entry.parent_id or
778
parent_entry.name != entry.name):
779
# Metadata common to all entries has changed
780
# against per-file parent
781
carry_over_possible = False
782
else:
783
carry_over_possible = True
784
# per-type checks for changes against the parent_entry
785
# are done below.
786
else:
787
# Cannot be a carry-over situation
788
carry_over_possible = False
789
# Populate the entry in the delta
790
if kind == 'file':
791
# XXX: There is still a small race here: If someone reverts the content of a file
792
# after iter_changes examines and decides it has changed,
793
# we will unconditionally record a new version even if some
794
# other process reverts it while commit is running (with
795
# the revert happening after iter_changes did its
796
# examination).
797
if change[7][1]:
798
entry.executable = True
799
else:
800
entry.executable = False
801
if (carry_over_possible and
802
parent_entry.executable == entry.executable):
803
# Check the file length, content hash after reading
804
# the file.
805
nostore_sha = parent_entry.text_sha1
806
else:
807
nostore_sha = None
808
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
809
try:
810
text = file_obj.read()
811
finally:
812
file_obj.close()
813
try:
814
entry.text_sha1, entry.text_size = self._add_text_to_weave(
815
file_id, text, heads, nostore_sha)
816
yield file_id, change[1][1], (entry.text_sha1, stat_value)
817
except errors.ExistingContent:
818
# No content change against a carry_over parent
819
# Perhaps this should also yield a fs hash update?
820
carried_over = True
821
entry.text_size = parent_entry.text_size
822
entry.text_sha1 = parent_entry.text_sha1
823
elif kind == 'symlink':
824
# Wants a path hint?
825
entry.symlink_target = tree.get_symlink_target(file_id)
826
if (carry_over_possible and
827
parent_entry.symlink_target == entry.symlink_target):
828
carried_over = True
829
else:
830
self._add_text_to_weave(change[0], '', heads, None)
831
elif kind == 'directory':
832
if carry_over_possible:
833
carried_over = True
834
else:
835
# Nothing to set on the entry.
836
# XXX: split into the Root and nonRoot versions.
837
if change[1][1] != '' or self.repository.supports_rich_root():
838
self._add_text_to_weave(change[0], '', heads, None)
839
elif kind == 'tree-reference':
840
if not self.repository._format.supports_tree_reference:
841
# This isn't quite sane as an error, but we shouldn't
842
# ever see this code path in practice: tree's don't
843
# permit references when the repo doesn't support tree
844
# references.
845
raise errors.UnsupportedOperation(tree.add_reference,
846
self.repository)
847
reference_revision = tree.get_reference_revision(change[0])
848
entry.reference_revision = reference_revision
849
if (carry_over_possible and
850
parent_entry.reference_revision == reference_revision):
851
carried_over = True
852
else:
853
self._add_text_to_weave(change[0], '', heads, None)
854
else:
855
raise AssertionError('unknown kind %r' % kind)
856
if not carried_over:
857
entry.revision = modified_rev
858
else:
859
entry.revision = parent_entry.revision
860
else:
861
entry = None
862
new_path = change[1][1]
863
inv_delta.append((change[1][0], new_path, change[0], entry))
864
if new_path == '':
865
seen_root = True
866
self.new_inventory = None
867
if len(inv_delta):
868
# This should perhaps be guarded by a check that the basis we
869
# commit against is the basis for the commit and if not do a delta
870
# against the basis.
871
self._any_changes = True
872
if not seen_root:
873
# housekeeping root entry changes do not affect no-change commits.
874
self._require_root_change(tree)
875
self.basis_delta_revision = basis_revision_id
876
877
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
878
parent_keys = tuple([(file_id, parent) for parent in parents])
879
return self.repository.texts._add_text(
880
(file_id, self._new_revision_id), parent_keys, new_text,
881
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
882
883
884
class RootCommitBuilder(CommitBuilder):
885
"""This commitbuilder actually records the root id"""
886
887
# the root entry gets versioned properly by this builder.
888
_versioned_root = True
889
890
def _check_root(self, ie, parent_invs, tree):
891
"""Helper for record_entry_contents.
892
893
:param ie: An entry being added.
894
:param parent_invs: The inventories of the parent revisions of the
895
commit.
896
:param tree: The tree that is being committed.
897
"""
898
899
def _require_root_change(self, tree):
900
"""Enforce an appropriate root object change.
901
902
This is called once when record_iter_changes is called, if and only if
903
the root was not in the delta calculated by record_iter_changes.
904
905
:param tree: The tree which is being committed.
906
"""
907
# versioned roots do not change unless the tree found a change.
227
raise NotImplementedError(self.record_iter_changes)
908
228
909
229
910
230
class RepositoryWriteLockResult(LogicalLockResult):
935
255
revisions and file history. It's normally accessed only by the Branch,
936
256
which views a particular line of development through that history.
937
257
938
The Repository builds on top of some byte storage facilies (the revisions,
939
signatures, inventories, texts and chk_bytes attributes) and a Transport,
940
which respectively provide byte storage and a means to access the (possibly
941
remote) disk.
942
943
The byte storage facilities are addressed via tuples, which we refer to
944
as 'keys' throughout the code base. Revision_keys, inventory_keys and
945
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
946
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
947
byte string made up of a hash identifier and a hash value.
948
We use this interface because it allows low friction with the underlying
949
code that implements disk indices, network encoding and other parts of
950
bzrlib.
951
952
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
953
the serialised revisions for the repository. This can be used to obtain
954
revision graph information or to access raw serialised revisions.
955
The result of trying to insert data into the repository via this store
956
is undefined: it should be considered read-only except for implementors
957
of repositories.
958
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
959
the serialised signatures for the repository. This can be used to
960
obtain access to raw serialised signatures. The result of trying to
961
insert data into the repository via this store is undefined: it should
962
be considered read-only except for implementors of repositories.
963
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
964
the serialised inventories for the repository. This can be used to
965
obtain unserialised inventories. The result of trying to insert data
966
into the repository via this store is undefined: it should be
967
considered read-only except for implementors of repositories.
968
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
969
texts of files and directories for the repository. This can be used to
970
obtain file texts or file graphs. Note that Repository.iter_file_bytes
971
is usually a better interface for accessing file texts.
972
The result of trying to insert data into the repository via this store
973
is undefined: it should be considered read-only except for implementors
974
of repositories.
975
:ivar chk_bytes: A bzrlib.versionedfile.VersionedFiles instance containing
976
any data the repository chooses to store or have indexed by its hash.
977
The result of trying to insert data into the repository via this store
978
is undefined: it should be considered read-only except for implementors
979
of repositories.
980
:ivar _transport: Transport for file access to repository, typically
981
pointing to .bzr/repository.
258
See VersionedFileRepository in bzrlib.vf_repository for the
259
base class for most Bazaar repositories.
982
260
"""
983
261
984
# What class to use for a CommitBuilder. Often it's simpler to change this
985
# in a Repository class subclass rather than to override
986
# get_commit_builder.
987
_commit_builder_class = CommitBuilder
988
989
262
def abort_write_group(self, suppress_errors=False):
990
263
"""Commit the contents accrued within the current write group.
991
264
1013
286
raise
1014
287
mutter('abort_write_group failed')
1015
288
log_exception_quietly()
1016
note('bzr: ERROR (ignored): %s', exc)
289
note(gettext('bzr: ERROR (ignored): %s'), exc)
1017
290
self._write_group = None
1018
291
1019
292
def _abort_write_group(self):
1034
307
1035
308
:param repository: A repository.
1036
309
"""
1037
if not self._format.supports_external_lookups:
1038
raise errors.UnstackableRepositoryFormat(self._format, self.base)
1039
if self.is_locked():
1040
# This repository will call fallback.unlock() when we transition to
1041
# the unlocked state, so we make sure to increment the lock count
1042
repository.lock_read()
1043
self._check_fallback_repository(repository)
1044
self._fallback_repositories.append(repository)
1045
self.texts.add_fallback_versioned_files(repository.texts)
1046
self.inventories.add_fallback_versioned_files(repository.inventories)
1047
self.revisions.add_fallback_versioned_files(repository.revisions)
1048
self.signatures.add_fallback_versioned_files(repository.signatures)
1049
if self.chk_bytes is not None:
1050
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
310
raise NotImplementedError(self.add_fallback_repository)
1051
311
1052
312
def _check_fallback_repository(self, repository):
1053
313
"""Check that this repository can fallback to repository safely.
1058
318
"""
1059
319
return InterRepository._assert_same_model(self, repository)
1060
320
1061
def add_inventory(self, revision_id, inv, parents):
1062
"""Add the inventory inv to the repository as revision_id.
1063
1064
:param parents: The revision ids of the parents that revision_id
1065
is known to have and are in the repository already.
1066
1067
:returns: The validator(which is a sha1 digest, though what is sha'd is
1068
repository format specific) of the serialized inventory.
1069
"""
1070
if not self.is_in_write_group():
1071
raise AssertionError("%r not in write group" % (self,))
1072
_mod_revision.check_not_reserved_id(revision_id)
1073
if not (inv.revision_id is None or inv.revision_id == revision_id):
1074
raise AssertionError(
1075
"Mismatch between inventory revision"
1076
" id and insertion revid (%r, %r)"
1077
% (inv.revision_id, revision_id))
1078
if inv.root is None:
1079
raise errors.RootMissing()
1080
return self._add_inventory_checked(revision_id, inv, parents)
1081
1082
def _add_inventory_checked(self, revision_id, inv, parents):
1083
"""Add inv to the repository after checking the inputs.
1084
1085
This function can be overridden to allow different inventory styles.
1086
1087
:seealso: add_inventory, for the contract.
1088
"""
1089
inv_lines = self._serializer.write_inventory_to_lines(inv)
1090
return self._inventory_add_lines(revision_id, parents,
1091
inv_lines, check_content=False)
1092
1093
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1094
parents, basis_inv=None, propagate_caches=False):
1095
"""Add a new inventory expressed as a delta against another revision.
1096
1097
See the inventory developers documentation for the theory behind
1098
inventory deltas.
1099
1100
:param basis_revision_id: The inventory id the delta was created
1101
against. (This does not have to be a direct parent.)
1102
:param delta: The inventory delta (see Inventory.apply_delta for
1103
details).
1104
:param new_revision_id: The revision id that the inventory is being
1105
added for.
1106
:param parents: The revision ids of the parents that revision_id is
1107
known to have and are in the repository already. These are supplied
1108
for repositories that depend on the inventory graph for revision
1109
graph access, as well as for those that pun ancestry with delta
1110
compression.
1111
:param basis_inv: The basis inventory if it is already known,
1112
otherwise None.
1113
:param propagate_caches: If True, the caches for this inventory are
1114
copied to and updated for the result if possible.
1115
1116
:returns: (validator, new_inv)
1117
The validator(which is a sha1 digest, though what is sha'd is
1118
repository format specific) of the serialized inventory, and the
1119
resulting inventory.
1120
"""
1121
if not self.is_in_write_group():
1122
raise AssertionError("%r not in write group" % (self,))
1123
_mod_revision.check_not_reserved_id(new_revision_id)
1124
basis_tree = self.revision_tree(basis_revision_id)
1125
basis_tree.lock_read()
1126
try:
1127
# Note that this mutates the inventory of basis_tree, which not all
1128
# inventory implementations may support: A better idiom would be to
1129
# return a new inventory, but as there is no revision tree cache in
1130
# repository this is safe for now - RBC 20081013
1131
if basis_inv is None:
1132
basis_inv = basis_tree.inventory
1133
basis_inv.apply_delta(delta)
1134
basis_inv.revision_id = new_revision_id
1135
return (self.add_inventory(new_revision_id, basis_inv, parents),
1136
basis_inv)
1137
finally:
1138
basis_tree.unlock()
1139
1140
def _inventory_add_lines(self, revision_id, parents, lines,
1141
check_content=True):
1142
"""Store lines in inv_vf and return the sha1 of the inventory."""
1143
parents = [(parent,) for parent in parents]
1144
result = self.inventories.add_lines((revision_id,), parents, lines,
1145
check_content=check_content)[0]
1146
self.inventories._access.flush()
1147
return result
1148
1149
def add_revision(self, revision_id, rev, inv=None, config=None):
1150
"""Add rev to the revision store as revision_id.
1151
1152
:param revision_id: the revision id to use.
1153
:param rev: The revision object.
1154
:param inv: The inventory for the revision. if None, it will be looked
1155
up in the inventory storer
1156
:param config: If None no digital signature will be created.
1157
If supplied its signature_needed method will be used
1158
to determine if a signature should be made.
1159
"""
1160
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
1161
# rev.parent_ids?
1162
_mod_revision.check_not_reserved_id(revision_id)
1163
if config is not None and config.signature_needed():
1164
if inv is None:
1165
inv = self.get_inventory(revision_id)
1166
tree = InventoryRevisionTree(self, inv, revision_id)
1167
testament = Testament(rev, tree)
1168
plaintext = testament.as_short_text()
1169
self.store_revision_signature(
1170
gpg.GPGStrategy(config), plaintext, revision_id)
1171
# check inventory present
1172
if not self.inventories.get_parent_map([(revision_id,)]):
1173
if inv is None:
1174
raise errors.WeaveRevisionNotPresent(revision_id,
1175
self.inventories)
1176
else:
1177
# yes, this is not suitable for adding with ghosts.
1178
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
1179
rev.parent_ids)
1180
else:
1181
key = (revision_id,)
1182
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
1183
self._add_revision(rev)
1184
1185
def _add_revision(self, revision):
1186
text = self._serializer.write_revision_to_string(revision)
1187
key = (revision.revision_id,)
1188
parents = tuple((parent,) for parent in revision.parent_ids)
1189
self.revisions.add_lines(key, parents, osutils.split_lines(text))
1190
1191
321
def all_revision_ids(self):
1192
322
"""Returns a list of all the revision ids in the repository.
1193
323
1216
346
"""
1217
347
self.control_files.break_lock()
1218
348
1219
@needs_read_lock
1220
def _eliminate_revisions_not_present(self, revision_ids):
1221
"""Check every revision id in revision_ids to see if we have it.
1222
1223
Returns a set of the present revisions.
1224
"""
1225
result = []
1226
graph = self.get_graph()
1227
parent_map = graph.get_parent_map(revision_ids)
1228
# The old API returned a list, should this actually be a set?
1229
return parent_map.keys()
1230
1231
def _check_inventories(self, checker):
1232
"""Check the inventories found from the revision scan.
1233
1234
This is responsible for verifying the sha1 of inventories and
1235
creating a pending_keys set that covers data referenced by inventories.
1236
"""
1237
bar = ui.ui_factory.nested_progress_bar()
1238
try:
1239
self._do_check_inventories(checker, bar)
1240
finally:
1241
bar.finished()
1242
1243
def _do_check_inventories(self, checker, bar):
1244
"""Helper for _check_inventories."""
1245
revno = 0
1246
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1247
kinds = ['chk_bytes', 'texts']
1248
count = len(checker.pending_keys)
1249
bar.update("inventories", 0, 2)
1250
current_keys = checker.pending_keys
1251
checker.pending_keys = {}
1252
# Accumulate current checks.
1253
for key in current_keys:
1254
if key[0] != 'inventories' and key[0] not in kinds:
1255
checker._report_items.append('unknown key type %r' % (key,))
1256
keys[key[0]].add(key[1:])
1257
if keys['inventories']:
1258
# NB: output order *should* be roughly sorted - topo or
1259
# inverse topo depending on repository - either way decent
1260
# to just delta against. However, pre-CHK formats didn't
1261
# try to optimise inventory layout on disk. As such the
1262
# pre-CHK code path does not use inventory deltas.
1263
last_object = None
1264
for record in self.inventories.check(keys=keys['inventories']):
1265
if record.storage_kind == 'absent':
1266
checker._report_items.append(
1267
'Missing inventory {%s}' % (record.key,))
1268
else:
1269
last_object = self._check_record('inventories', record,
1270
checker, last_object,
1271
current_keys[('inventories',) + record.key])
1272
del keys['inventories']
1273
else:
1274
return
1275
bar.update("texts", 1)
1276
while (checker.pending_keys or keys['chk_bytes']
1277
or keys['texts']):
1278
# Something to check.
1279
current_keys = checker.pending_keys
1280
checker.pending_keys = {}
1281
# Accumulate current checks.
1282
for key in current_keys:
1283
if key[0] not in kinds:
1284
checker._report_items.append('unknown key type %r' % (key,))
1285
keys[key[0]].add(key[1:])
1286
# Check the outermost kind only - inventories || chk_bytes || texts
1287
for kind in kinds:
1288
if keys[kind]:
1289
last_object = None
1290
for record in getattr(self, kind).check(keys=keys[kind]):
1291
if record.storage_kind == 'absent':
1292
checker._report_items.append(
1293
'Missing %s {%s}' % (kind, record.key,))
1294
else:
1295
last_object = self._check_record(kind, record,
1296
checker, last_object, current_keys[(kind,) + record.key])
1297
keys[kind] = set()
1298
break
1299
1300
def _check_record(self, kind, record, checker, last_object, item_data):
1301
"""Check a single text from this repository."""
1302
if kind == 'inventories':
1303
rev_id = record.key[0]
1304
inv = self._deserialise_inventory(rev_id,
1305
record.get_bytes_as('fulltext'))
1306
if last_object is not None:
1307
delta = inv._make_delta(last_object)
1308
for old_path, path, file_id, ie in delta:
1309
if ie is None:
1310
continue
1311
ie.check(checker, rev_id, inv)
1312
else:
1313
for path, ie in inv.iter_entries():
1314
ie.check(checker, rev_id, inv)
1315
if self._format.fast_deltas:
1316
return inv
1317
elif kind == 'chk_bytes':
1318
# No code written to check chk_bytes for this repo format.
1319
checker._report_items.append(
1320
'unsupported key type chk_bytes for %s' % (record.key,))
1321
elif kind == 'texts':
1322
self._check_text(record, checker, item_data)
1323
else:
1324
checker._report_items.append(
1325
'unknown key type %s for %s' % (kind, record.key))
1326
1327
def _check_text(self, record, checker, item_data):
1328
"""Check a single text."""
1329
# Check it is extractable.
1330
# TODO: check length.
1331
if record.storage_kind == 'chunked':
1332
chunks = record.get_bytes_as(record.storage_kind)
1333
sha1 = osutils.sha_strings(chunks)
1334
length = sum(map(len, chunks))
1335
else:
1336
content = record.get_bytes_as('fulltext')
1337
sha1 = osutils.sha_string(content)
1338
length = len(content)
1339
if item_data and sha1 != item_data[1]:
1340
checker._report_items.append(
1341
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1342
(record.key, sha1, item_data[1], item_data[2]))
1343
1344
349
@staticmethod
1345
def create(a_bzrdir):
1346
"""Construct the current default format repository in a_bzrdir."""
1347
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
350
def create(controldir):
351
"""Construct the current default format repository in controldir."""
352
return RepositoryFormat.get_default_format().initialize(controldir)
1348
353
1349
def __init__(self, _format, a_bzrdir, control_files):
354
def __init__(self, _format, controldir, control_files):
1350
355
"""instantiate a Repository.
1351
356
1352
357
:param _format: The format of the repository on disk.
1353
:param a_bzrdir: The BzrDir of the repository.
358
:param controldir: The ControlDir of the repository.
359
:param control_files: Control files to use for locking, etc.
1354
360
"""
1355
361
# In the future we will have a single api for all stores for
1356
362
# getting file texts, inventories and revisions, then
1358
364
super(Repository, self).__init__()
1359
365
self._format = _format
1360
366
# the following are part of the public API for Repository:
1361
self.bzrdir = a_bzrdir
367
self.bzrdir = controldir
1362
368
self.control_files = control_files
1363
self._transport = control_files._transport
1364
self.base = self._transport.base
1365
369
# for tests
1366
self._reconcile_does_inventory_gc = True
1367
self._reconcile_fixes_text_parents = False
1368
self._reconcile_backsup_inventory = True
1369
370
self._write_group = None
1370
371
# Additional places to query for data.
1371
372
self._fallback_repositories = []
1372
# An InventoryEntry cache, used during deserialization
1373
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1374
# Is it safe to return inventory entries directly from the entry cache,
1375
# rather copying them?
1376
self._safe_to_return_from_cache = False
1377
373
1378
374
@property
1379
375
def user_transport(self):
1413
409
"""
1414
410
if self.__class__ is not other.__class__:
1415
411
return False
1416
return (self._transport.base == other._transport.base)
412
return (self.control_url == other.control_url)
1417
413
1418
414
def is_in_write_group(self):
1419
415
"""Return True if there is an open write group.
1519
515
if revid and committers:
1520
516
result['committers'] = 0
1521
517
if revid and revid != _mod_revision.NULL_REVISION:
518
graph = self.get_graph()
1522
519
if committers:
1523
520
all_committers = set()
1524
revisions = self.get_ancestry(revid)
1525
# pop the leading None
1526
revisions.pop(0)
1527
first_revision = None
521
revisions = [r for (r, p) in graph.iter_ancestry([revid])
522
if r != _mod_revision.NULL_REVISION]
523
last_revision = None
1528
524
if not committers:
1529
525
# ignore the revisions in the middle - just grab first and last
1530
526
revisions = revisions[0], revisions[-1]
1531
527
for revision in self.get_revisions(revisions):
1532
if not first_revision:
1533
first_revision = revision
528
if not last_revision:
529
last_revision = revision
1534
530
if committers:
1535
531
all_committers.add(revision.committer)
1536
last_revision = revision
532
first_revision = revision
1537
533
if committers:
1538
534
result['committers'] = len(all_committers)
1539
535
result['firstrev'] = (first_revision.timestamp,
1540
536
first_revision.timezone)
1541
537
result['latestrev'] = (last_revision.timestamp,
1542
538
last_revision.timezone)
1543
1544
# now gather global repository information
1545
# XXX: This is available for many repos regardless of listability.
1546
if self.user_transport.listable():
1547
# XXX: do we want to __define len__() ?
1548
# Maybe the versionedfiles object should provide a different
1549
# method to get the number of keys.
1550
result['revisions'] = len(self.revisions.keys())
1551
# result['size'] = t
1552
539
return result
1553
540
1554
541
def find_branches(self, using=False):
1565
552
def __init__(self):
1566
553
self.first_call = True
1567
554
1568
def __call__(self, bzrdir):
1569
# On the first call, the parameter is always the bzrdir
555
def __call__(self, controldir):
556
# On the first call, the parameter is always the controldir
1570
557
# containing the current repo.
1571
558
if not self.first_call:
1572
559
try:
1573
repository = bzrdir.open_repository()
560
repository = controldir.open_repository()
1574
561
except errors.NoRepositoryPresent:
1575
562
pass
1576
563
else:
1577
564
return False, ([], repository)
1578
565
self.first_call = False
1579
value = (bzrdir.list_branches(), None)
566
value = (controldir.list_branches(), None)
1580
567
return True, value
1581
568
1582
569
ret = []
1583
for branches, repository in bzrdir.BzrDir.find_bzrdirs(
570
for branches, repository in controldir.ControlDir.find_bzrdirs(
1584
571
self.user_transport, evaluate=Evaluator()):
1585
572
if branches is not None:
1586
573
ret.extend(branches)
1591
578
@needs_read_lock
1592
579
def search_missing_revision_ids(self, other,
1593
580
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
1594
find_ghosts=True, revision_ids=None, if_present_ids=None):
581
find_ghosts=True, revision_ids=None, if_present_ids=None,
582
limit=None):
1595
583
"""Return the revision ids that other has that this does not.
1596
584
1597
585
These are returned in topological order.
1610
598
revision_ids = [revision_id]
1611
599
return InterRepository.get(other, self).search_missing_revision_ids(
1612
600
find_ghosts=find_ghosts, revision_ids=revision_ids,
1613
if_present_ids=if_present_ids)
601
if_present_ids=if_present_ids, limit=limit)
1614
602
1615
603
@staticmethod
1616
604
def open(base):
1619
607
For instance, if the repository is at URL/.bzr/repository,
1620
608
Repository.open(URL) -> a Repository instance.
1621
609
"""
1622
control = bzrdir.BzrDir.open(base)
610
control = controldir.ControlDir.open(base)
1623
611
return control.open_repository()
1624
612
1625
613
def copy_content_into(self, destination, revision_id=None):
1656
644
"""
1657
645
1658
646
def suspend_write_group(self):
647
"""Suspend a write group.
648
649
:raise UnsuspendableWriteGroup: If the write group can not be
650
suspended.
651
:return: List of tokens
652
"""
1659
653
raise errors.UnsuspendableWriteGroup(self)
1660
654
1661
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1662
"""Return the keys of missing inventory parents for revisions added in
1663
this write group.
1664
1665
A revision is not complete if the inventory delta for that revision
1666
cannot be calculated. Therefore if the parent inventories of a
1667
revision are not present, the revision is incomplete, and e.g. cannot
1668
be streamed by a smart server. This method finds missing inventory
1669
parents for revisions added in this write group.
1670
"""
1671
if not self._format.supports_external_lookups:
1672
# This is only an issue for stacked repositories
1673
return set()
1674
if not self.is_in_write_group():
1675
raise AssertionError('not in a write group')
1676
1677
# XXX: We assume that every added revision already has its
1678
# corresponding inventory, so we only check for parent inventories that
1679
# might be missing, rather than all inventories.
1680
parents = set(self.revisions._index.get_missing_parents())
1681
parents.discard(_mod_revision.NULL_REVISION)
1682
unstacked_inventories = self.inventories._index
1683
present_inventories = unstacked_inventories.get_parent_map(
1684
key[-1:] for key in parents)
1685
parents.difference_update(present_inventories)
1686
if len(parents) == 0:
1687
# No missing parent inventories.
1688
return set()
1689
if not check_for_missing_texts:
1690
return set(('inventories', rev_id) for (rev_id,) in parents)
1691
# Ok, now we have a list of missing inventories. But these only matter
1692
# if the inventories that reference them are missing some texts they
1693
# appear to introduce.
1694
# XXX: Texts referenced by all added inventories need to be present,
1695
# but at the moment we're only checking for texts referenced by
1696
# inventories at the graph's edge.
1697
key_deps = self.revisions._index._key_dependencies
1698
key_deps.satisfy_refs_for_keys(present_inventories)
1699
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1700
file_ids = self.fileids_altered_by_revision_ids(referrers)
1701
missing_texts = set()
1702
for file_id, version_ids in file_ids.iteritems():
1703
missing_texts.update(
1704
(file_id, version_id) for version_id in version_ids)
1705
present_texts = self.texts.get_parent_map(missing_texts)
1706
missing_texts.difference_update(present_texts)
1707
if not missing_texts:
1708
# No texts are missing, so all revisions and their deltas are
1709
# reconstructable.
1710
return set()
1711
# Alternatively the text versions could be returned as the missing
1712
# keys, but this is likely to be less data.
1713
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1714
return missing_keys
1715
1716
655
def refresh_data(self):
1717
656
"""Re-read any data needed to synchronise with disk.
1718
657
1738
677
def _resume_write_group(self, tokens):
1739
678
raise errors.UnsuspendableWriteGroup(self)
1740
679
1741
def fetch(self, source, revision_id=None, find_ghosts=False,
1742
fetch_spec=None):
680
def fetch(self, source, revision_id=None, find_ghosts=False):
1743
681
"""Fetch the content required to construct revision_id from source.
1744
682
1745
If revision_id is None and fetch_spec is None, then all content is
1746
copied.
683
If revision_id is None, then all content is copied.
1747
684
1748
685
fetch() may not be used when the repository is in a write group -
1749
686
either finish the current write group before using fetch, or use
1755
692
:param revision_id: If specified, all the content needed for this
1756
693
revision ID will be copied to the target. Fetch will determine for
1757
694
itself which content needs to be copied.
1758
:param fetch_spec: If specified, a SearchResult or
1759
PendingAncestryResult that describes which revisions to copy. This
1760
allows copying multiple heads at once. Mutually exclusive with
1761
revision_id.
1762
695
"""
1763
if fetch_spec is not None and revision_id is not None:
1764
raise AssertionError(
1765
"fetch_spec and revision_id are mutually exclusive.")
1766
696
if self.is_in_write_group():
1767
697
raise errors.InternalBzrError(
1768
698
"May not fetch while in a write group.")
1770
700
# TODO: lift out to somewhere common with RemoteRepository
1771
701
# <https://bugs.launchpad.net/bzr/+bug/401646>
1772
702
if (self.has_same_location(source)
1773
and fetch_spec is None
1774
703
and self._has_same_fallbacks(source)):
1775
704
# check that last_revision is in 'from' and then return a
1776
705
# no-operation.
1779
708
self.get_revision(revision_id)
1780
709
return 0, []
1781
710
inter = InterRepository.get(source, self)
1782
return inter.fetch(revision_id=revision_id,
1783
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
711
return inter.fetch(revision_id=revision_id, find_ghosts=find_ghosts)
1784
712
1785
713
def create_bundle(self, target, base, fileobj, format=None):
1786
714
return serializer.write_bundle(self, target, base, fileobj, format)
1787
715
1788
def get_commit_builder(self, branch, parents, config, timestamp=None,
716
def get_commit_builder(self, branch, parents, config_stack, timestamp=None,
1789
717
timezone=None, committer=None, revprops=None,
1790
718
revision_id=None, lossy=False):
1791
719
"""Obtain a CommitBuilder for this repository.
1792
720
1793
721
:param branch: Branch to commit to.
1794
722
:param parents: Revision ids of the parents of the new revision.
1795
:param config: Configuration to use.
723
:param config_stack: Configuration stack to use.
1796
724
:param timestamp: Optional timestamp recorded for commit.
1797
725
:param timezone: Optional timezone for timestamp.
1798
726
:param committer: Optional committer to set for commit.
1801
729
:param lossy: Whether to discard data that can not be natively
1802
730
represented, when pushing to a foreign VCS
1803
731
"""
1804
if self._fallback_repositories and not self._format.supports_chks:
1805
raise errors.BzrError("Cannot commit directly to a stacked branch"
1806
" in pre-2a formats. See "
1807
"https://bugs.launchpad.net/bzr/+bug/375013 for details.")
1808
result = self._commit_builder_class(self, parents, config,
1809
timestamp, timezone, committer, revprops, revision_id,
1810
lossy)
1811
self.start_write_group()
1812
return result
732
raise NotImplementedError(self.get_commit_builder)
1813
733
1814
734
@only_raises(errors.LockNotHeld, errors.LockBroken)
1815
735
def unlock(self):
1822
742
'Must end write groups before releasing write locks.')
1823
743
self.control_files.unlock()
1824
744
if self.control_files._lock_count == 0:
1825
self._inventory_entry_cache.clear()
1826
745
for repo in self._fallback_repositories:
1827
746
repo.unlock()
1828
747
1829
748
@needs_read_lock
1830
def clone(self, a_bzrdir, revision_id=None):
1831
"""Clone this repository into a_bzrdir using the current format.
749
def clone(self, controldir, revision_id=None):
750
"""Clone this repository into controldir using the current format.
1832
751
1833
752
Currently no check is made that the format of this repository and
1834
753
the bzrdir format are compatible. FIXME RBC 20060201.
1837
756
"""
1838
757
# TODO: deprecate after 0.16; cloning this with all its settings is
1839
758
# probably not very useful -- mbp 20070423
1840
dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
759
dest_repo = self._create_sprouting_repo(
760
controldir, shared=self.is_shared())
1841
761
self.copy_content_into(dest_repo, revision_id)
1842
762
return dest_repo
1843
763
1893
813
dest_repo = a_bzrdir.open_repository()
1894
814
return dest_repo
1895
815
1896
def _get_sink(self):
1897
"""Return a sink for streaming into this repository."""
1898
return StreamSink(self)
1899
1900
def _get_source(self, to_format):
1901
"""Return a source for streaming from this repository."""
1902
return StreamSource(self, to_format)
1903
1904
816
@needs_read_lock
1905
817
def has_revision(self, revision_id):
1906
818
"""True if this repository has a copy of the revision."""
1913
825
:param revision_ids: An iterable of revision_ids.
1914
826
:return: A set of the revision_ids that were present.
1915
827
"""
1916
parent_map = self.revisions.get_parent_map(
1917
[(rev_id,) for rev_id in revision_ids])
1918
result = set()
1919
if _mod_revision.NULL_REVISION in revision_ids:
1920
result.add(_mod_revision.NULL_REVISION)
1921
result.update([key[0] for key in parent_map])
1922
return result
828
raise NotImplementedError(self.has_revisions)
1923
829
1924
830
@needs_read_lock
1925
831
def get_revision(self, revision_id):
1926
832
"""Return the Revision object for a named revision."""
1927
833
return self.get_revisions([revision_id])[0]
1928
834
1929
@needs_read_lock
1930
835
def get_revision_reconcile(self, revision_id):
1931
836
"""'reconcile' helper routine that allows access to a revision always.
1932
837
1935
840
be used by reconcile, or reconcile-alike commands that are correcting
1936
841
or testing the revision graph.
1937
842
"""
1938
return self._get_revisions([revision_id])[0]
843
raise NotImplementedError(self.get_revision_reconcile)
1939
844
1940
@needs_read_lock
1941
845
def get_revisions(self, revision_ids):
1942
846
"""Get many revisions at once.
1943
847
1944
848
Repositories that need to check data on every revision read should
1945
849
subclass this method.
1946
850
"""
1947
return self._get_revisions(revision_ids)
1948
1949
@needs_read_lock
1950
def _get_revisions(self, revision_ids):
1951
"""Core work logic to get many revisions without sanity checks."""
1952
revs = {}
1953
for revid, rev in self._iter_revisions(revision_ids):
1954
if rev is None:
1955
raise errors.NoSuchRevision(self, revid)
1956
revs[revid] = rev
1957
return [revs[revid] for revid in revision_ids]
1958
1959
def _iter_revisions(self, revision_ids):
1960
"""Iterate over revision objects.
1961
1962
:param revision_ids: An iterable of revisions to examine. None may be
1963
passed to request all revisions known to the repository. Note that
1964
not all repositories can find unreferenced revisions; for those
1965
repositories only referenced ones will be returned.
1966
:return: An iterator of (revid, revision) tuples. Absent revisions (
1967
those asked for but not available) are returned as (revid, None).
1968
"""
1969
if revision_ids is None:
1970
revision_ids = self.all_revision_ids()
1971
else:
1972
for rev_id in revision_ids:
1973
if not rev_id or not isinstance(rev_id, basestring):
1974
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1975
keys = [(key,) for key in revision_ids]
1976
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1977
for record in stream:
1978
revid = record.key[0]
1979
if record.storage_kind == 'absent':
1980
yield (revid, None)
1981
else:
1982
text = record.get_bytes_as('fulltext')
1983
rev = self._serializer.read_revision_from_string(text)
1984
yield (revid, rev)
851
raise NotImplementedError(self.get_revisions)
1985
852
1986
853
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1987
854
"""Produce a generator of revision deltas.
2042
909
signature = gpg_strategy.sign(plaintext)
2043
910
self.add_signature_text(revision_id, signature)
2044
911
2045
@needs_write_lock
2046
912
def add_signature_text(self, revision_id, signature):
2047
self.signatures.add_lines((revision_id,), (),
2048
osutils.split_lines(signature))
2049
2050
def find_text_key_references(self):
2051
"""Find the text key references within the repository.
2052
2053
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2054
to whether they were referred to by the inventory of the
2055
revision_id that they contain. The inventory texts from all present
2056
revision ids are assessed to generate this report.
2057
"""
2058
revision_keys = self.revisions.keys()
2059
w = self.inventories
2060
pb = ui.ui_factory.nested_progress_bar()
2061
try:
2062
return self._serializer._find_text_key_references(
2063
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
2064
finally:
2065
pb.finished()
2066
2067
def _inventory_xml_lines_for_keys(self, keys):
2068
"""Get a line iterator of the sort needed for findind references.
2069
2070
Not relevant for non-xml inventory repositories.
2071
2072
Ghosts in revision_keys are ignored.
2073
2074
:param revision_keys: The revision keys for the inventories to inspect.
2075
:return: An iterator over (inventory line, revid) for the fulltexts of
2076
all of the xml inventories specified by revision_keys.
2077
"""
2078
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2079
for record in stream:
2080
if record.storage_kind != 'absent':
2081
chunks = record.get_bytes_as('chunked')
2082
revid = record.key[-1]
2083
lines = osutils.chunks_to_lines(chunks)
2084
for line in lines:
2085
yield line, revid
2086
2087
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2088
revision_keys):
2089
"""Helper routine for fileids_altered_by_revision_ids.
2090
2091
This performs the translation of xml lines to revision ids.
2092
2093
:param line_iterator: An iterator of lines, origin_version_id
2094
:param revision_keys: The revision ids to filter for. This should be a
2095
set or other type which supports efficient __contains__ lookups, as
2096
the revision key from each parsed line will be looked up in the
2097
revision_keys filter.
2098
:return: a dictionary mapping altered file-ids to an iterable of
2099
revision_ids. Each altered file-ids has the exact revision_ids that
2100
altered it listed explicitly.
2101
"""
2102
seen = set(self._serializer._find_text_key_references(
2103
line_iterator).iterkeys())
2104
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2105
parent_seen = set(self._serializer._find_text_key_references(
2106
self._inventory_xml_lines_for_keys(parent_keys)))
2107
new_keys = seen - parent_seen
2108
result = {}
2109
setdefault = result.setdefault
2110
for key in new_keys:
2111
setdefault(key[0], set()).add(key[-1])
2112
return result
913
"""Store a signature text for a revision.
914
915
:param revision_id: Revision id of the revision
916
:param signature: Signature text.
917
"""
918
raise NotImplementedError(self.add_signature_text)
2113
919
2114
920
def _find_parent_ids_of_revisions(self, revision_ids):
2115
921
"""Find all parent ids that are mentioned in the revision graph.
2124
930
parent_ids.discard(_mod_revision.NULL_REVISION)
2125
931
return parent_ids
2126
932
2127
def _find_parent_keys_of_revisions(self, revision_keys):
2128
"""Similar to _find_parent_ids_of_revisions, but used with keys.
2129
2130
:param revision_keys: An iterable of revision_keys.
2131
:return: The parents of all revision_keys that are not already in
2132
revision_keys
2133
"""
2134
parent_map = self.revisions.get_parent_map(revision_keys)
2135
parent_keys = set()
2136
map(parent_keys.update, parent_map.itervalues())
2137
parent_keys.difference_update(revision_keys)
2138
parent_keys.discard(_mod_revision.NULL_REVISION)
2139
return parent_keys
2140
2141
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
2142
"""Find the file ids and versions affected by revisions.
2143
2144
:param revisions: an iterable containing revision ids.
2145
:param _inv_weave: The inventory weave from this repository or None.
2146
If None, the inventory weave will be opened automatically.
2147
:return: a dictionary mapping altered file-ids to an iterable of
2148
revision_ids. Each altered file-ids has the exact revision_ids that
2149
altered it listed explicitly.
2150
"""
2151
selected_keys = set((revid,) for revid in revision_ids)
2152
w = _inv_weave or self.inventories
2153
return self._find_file_ids_from_xml_inventory_lines(
2154
w.iter_lines_added_or_present_in_keys(
2155
selected_keys, pb=None),
2156
selected_keys)
2157
2158
933
def iter_files_bytes(self, desired_files):
2159
934
"""Iterate through file versions.
2160
935
2166
941
uniquely identify the file version in the caller's context. (Examples:
2167
942
an index number or a TreeTransform trans_id.)
2168
943
2169
bytes_iterator is an iterable of bytestrings for the file. The
2170
kind of iterable and length of the bytestrings are unspecified, but for
2171
this implementation, it is a list of bytes produced by
2172
VersionedFile.get_record_stream().
2173
2174
944
:param desired_files: a list of (file_id, revision_id, identifier)
2175
945
triples
2176
946
"""
2177
text_keys = {}
2178
for file_id, revision_id, callable_data in desired_files:
2179
text_keys[(file_id, revision_id)] = callable_data
2180
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2181
if record.storage_kind == 'absent':
2182
raise errors.RevisionNotPresent(record.key, self)
2183
yield text_keys[record.key], record.get_bytes_as('chunked')
2184
2185
def _generate_text_key_index(self, text_key_references=None,
2186
ancestors=None):
2187
"""Generate a new text key index for the repository.
2188
2189
This is an expensive function that will take considerable time to run.
2190
2191
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2192
list of parents, also text keys. When a given key has no parents,
2193
the parents list will be [NULL_REVISION].
2194
"""
2195
# All revisions, to find inventory parents.
2196
if ancestors is None:
2197
graph = self.get_graph()
2198
ancestors = graph.get_parent_map(self.all_revision_ids())
2199
if text_key_references is None:
2200
text_key_references = self.find_text_key_references()
2201
pb = ui.ui_factory.nested_progress_bar()
2202
try:
2203
return self._do_generate_text_key_index(ancestors,
2204
text_key_references, pb)
2205
finally:
2206
pb.finished()
2207
2208
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2209
"""Helper for _generate_text_key_index to avoid deep nesting."""
2210
revision_order = tsort.topo_sort(ancestors)
2211
invalid_keys = set()
2212
revision_keys = {}
2213
for revision_id in revision_order:
2214
revision_keys[revision_id] = set()
2215
text_count = len(text_key_references)
2216
# a cache of the text keys to allow reuse; costs a dict of all the
2217
# keys, but saves a 2-tuple for every child of a given key.
2218
text_key_cache = {}
2219
for text_key, valid in text_key_references.iteritems():
2220
if not valid:
2221
invalid_keys.add(text_key)
2222
else:
2223
revision_keys[text_key[1]].add(text_key)
2224
text_key_cache[text_key] = text_key
2225
del text_key_references
2226
text_index = {}
2227
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2228
NULL_REVISION = _mod_revision.NULL_REVISION
2229
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2230
# too small for large or very branchy trees. However, for 55K path
2231
# trees, it would be easy to use too much memory trivially. Ideally we
2232
# could gauge this by looking at available real memory etc, but this is
2233
# always a tricky proposition.
2234
inventory_cache = lru_cache.LRUCache(10)
2235
batch_size = 10 # should be ~150MB on a 55K path tree
2236
batch_count = len(revision_order) / batch_size + 1
2237
processed_texts = 0
2238
pb.update("Calculating text parents", processed_texts, text_count)
2239
for offset in xrange(batch_count):
2240
to_query = revision_order[offset * batch_size:(offset + 1) *
2241
batch_size]
2242
if not to_query:
2243
break
2244
for revision_id in to_query:
2245
parent_ids = ancestors[revision_id]
2246
for text_key in revision_keys[revision_id]:
2247
pb.update("Calculating text parents", processed_texts)
2248
processed_texts += 1
2249
candidate_parents = []
2250
for parent_id in parent_ids:
2251
parent_text_key = (text_key[0], parent_id)
2252
try:
2253
check_parent = parent_text_key not in \
2254
revision_keys[parent_id]
2255
except KeyError:
2256
# the parent parent_id is a ghost:
2257
check_parent = False
2258
# truncate the derived graph against this ghost.
2259
parent_text_key = None
2260
if check_parent:
2261
# look at the parent commit details inventories to
2262
# determine possible candidates in the per file graph.
2263
# TODO: cache here.
2264
try:
2265
inv = inventory_cache[parent_id]
2266
except KeyError:
2267
inv = self.revision_tree(parent_id).inventory
2268
inventory_cache[parent_id] = inv
2269
try:
2270
parent_entry = inv[text_key[0]]
2271
except (KeyError, errors.NoSuchId):
2272
parent_entry = None
2273
if parent_entry is not None:
2274
parent_text_key = (
2275
text_key[0], parent_entry.revision)
2276
else:
2277
parent_text_key = None
2278
if parent_text_key is not None:
2279
candidate_parents.append(
2280
text_key_cache[parent_text_key])
2281
parent_heads = text_graph.heads(candidate_parents)
2282
new_parents = list(parent_heads)
2283
new_parents.sort(key=lambda x:candidate_parents.index(x))
2284
if new_parents == []:
2285
new_parents = [NULL_REVISION]
2286
text_index[text_key] = new_parents
2287
2288
for text_key in invalid_keys:
2289
text_index[text_key] = [NULL_REVISION]
2290
return text_index
2291
2292
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2293
"""Get an iterable listing the keys of all the data introduced by a set
2294
of revision IDs.
2295
2296
The keys will be ordered so that the corresponding items can be safely
2297
fetched and inserted in that order.
2298
2299
:returns: An iterable producing tuples of (knit-kind, file-id,
2300
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2301
'revisions'. file-id is None unless knit-kind is 'file'.
2302
"""
2303
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2304
yield result
2305
del _files_pb
2306
for result in self._find_non_file_keys_to_fetch(revision_ids):
2307
yield result
2308
2309
def _find_file_keys_to_fetch(self, revision_ids, pb):
2310
# XXX: it's a bit weird to control the inventory weave caching in this
2311
# generator. Ideally the caching would be done in fetch.py I think. Or
2312
# maybe this generator should explicitly have the contract that it
2313
# should not be iterated until the previously yielded item has been
2314
# processed?
2315
inv_w = self.inventories
2316
2317
# file ids that changed
2318
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2319
count = 0
2320
num_file_ids = len(file_ids)
2321
for file_id, altered_versions in file_ids.iteritems():
2322
if pb is not None:
2323
pb.update("Fetch texts", count, num_file_ids)
2324
count += 1
2325
yield ("file", file_id, altered_versions)
2326
2327
def _find_non_file_keys_to_fetch(self, revision_ids):
2328
# inventory
2329
yield ("inventory", None, revision_ids)
2330
2331
# signatures
2332
# XXX: Note ATM no callers actually pay attention to this return
2333
# instead they just use the list of revision ids and ignore
2334
# missing sigs. Consider removing this work entirely
2335
revisions_with_signatures = set(self.signatures.get_parent_map(
2336
[(r,) for r in revision_ids]))
2337
revisions_with_signatures = set(
2338
[r for (r,) in revisions_with_signatures])
2339
revisions_with_signatures.intersection_update(revision_ids)
2340
yield ("signatures", None, revisions_with_signatures)
2341
2342
# revisions
2343
yield ("revisions", None, revision_ids)
2344
2345
@needs_read_lock
2346
def get_inventory(self, revision_id):
2347
"""Get Inventory object by revision id."""
2348
return self.iter_inventories([revision_id]).next()
2349
2350
def iter_inventories(self, revision_ids, ordering=None):
2351
"""Get many inventories by revision_ids.
2352
2353
This will buffer some or all of the texts used in constructing the
2354
inventories in memory, but will only parse a single inventory at a
2355
time.
2356
2357
:param revision_ids: The expected revision ids of the inventories.
2358
:param ordering: optional ordering, e.g. 'topological'. If not
2359
specified, the order of revision_ids will be preserved (by
2360
buffering if necessary).
2361
:return: An iterator of inventories.
2362
"""
2363
if ((None in revision_ids)
2364
or (_mod_revision.NULL_REVISION in revision_ids)):
2365
raise ValueError('cannot get null revision inventory')
2366
return self._iter_inventories(revision_ids, ordering)
2367
2368
def _iter_inventories(self, revision_ids, ordering):
2369
"""single-document based inventory iteration."""
2370
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2371
for text, revision_id in inv_xmls:
2372
yield self._deserialise_inventory(revision_id, text)
2373
2374
def _iter_inventory_xmls(self, revision_ids, ordering):
2375
if ordering is None:
2376
order_as_requested = True
2377
ordering = 'unordered'
2378
else:
2379
order_as_requested = False
2380
keys = [(revision_id,) for revision_id in revision_ids]
2381
if not keys:
2382
return
2383
if order_as_requested:
2384
key_iter = iter(keys)
2385
next_key = key_iter.next()
2386
stream = self.inventories.get_record_stream(keys, ordering, True)
2387
text_chunks = {}
2388
for record in stream:
2389
if record.storage_kind != 'absent':
2390
chunks = record.get_bytes_as('chunked')
2391
if order_as_requested:
2392
text_chunks[record.key] = chunks
2393
else:
2394
yield ''.join(chunks), record.key[-1]
2395
else:
2396
raise errors.NoSuchRevision(self, record.key)
2397
if order_as_requested:
2398
# Yield as many results as we can while preserving order.
2399
while next_key in text_chunks:
2400
chunks = text_chunks.pop(next_key)
2401
yield ''.join(chunks), next_key[-1]
2402
try:
2403
next_key = key_iter.next()
2404
except StopIteration:
2405
# We still want to fully consume the get_record_stream,
2406
# just in case it is not actually finished at this point
2407
next_key = None
2408
break
2409
2410
def _deserialise_inventory(self, revision_id, xml):
2411
"""Transform the xml into an inventory object.
2412
2413
:param revision_id: The expected revision id of the inventory.
2414
:param xml: A serialised inventory.
2415
"""
2416
result = self._serializer.read_inventory_from_string(xml, revision_id,
2417
entry_cache=self._inventory_entry_cache,
2418
return_from_cache=self._safe_to_return_from_cache)
2419
if result.revision_id != revision_id:
2420
raise AssertionError('revision id mismatch %s != %s' % (
2421
result.revision_id, revision_id))
2422
return result
2423
2424
def get_serializer_format(self):
2425
return self._serializer.format_num
2426
2427
@needs_read_lock
2428
def _get_inventory_xml(self, revision_id):
2429
"""Get serialized inventory as a string."""
2430
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2431
try:
2432
text, revision_id = texts.next()
2433
except StopIteration:
2434
raise errors.HistoryMissing(self, 'inventory', revision_id)
2435
return text
947
raise NotImplementedError(self.iter_files_bytes)
2436
948
2437
949
def get_rev_id_for_revno(self, revno, known_pair):
2438
950
"""Return the revision id of a revno, given a later (revno, revid)
2469
981
raise AssertionError('_iter_for_revno returned too much history')
2470
982
return (True, partial_history[-1])
2471
983
2472
def iter_reverse_revision_history(self, revision_id):
2473
"""Iterate backwards through revision ids in the lefthand history
2474
2475
:param revision_id: The revision id to start with. All its lefthand
2476
ancestors will be traversed.
2477
"""
2478
graph = self.get_graph()
2479
stop_revisions = (None, _mod_revision.NULL_REVISION)
2480
return graph.iter_lefthand_ancestry(revision_id, stop_revisions)
2481
2482
984
def is_shared(self):
2483
985
"""Return True if this repository is flagged as a shared repository."""
2484
986
raise NotImplementedError(self.is_shared)
2511
1013
2512
1014
`revision_id` may be NULL_REVISION for the empty tree revision.
2513
1015
"""
2514
revision_id = _mod_revision.ensure_null(revision_id)
2515
# TODO: refactor this to use an existing revision object
2516
# so we don't need to read it in twice.
2517
if revision_id == _mod_revision.NULL_REVISION:
2518
return InventoryRevisionTree(self,
2519
Inventory(root_id=None), _mod_revision.NULL_REVISION)
2520
else:
2521
inv = self.get_inventory(revision_id)
2522
return InventoryRevisionTree(self, inv, revision_id)
1016
raise NotImplementedError(self.revision_tree)
2523
1017
2524
1018
def revision_trees(self, revision_ids):
2525
1019
"""Return Trees for revisions in this repository.
2527
1021
:param revision_ids: a sequence of revision-ids;
2528
1022
a revision-id may not be None or 'null:'
2529
1023
"""
2530
inventories = self.iter_inventories(revision_ids)
2531
for inv in inventories:
2532
yield InventoryRevisionTree(self, inv, inv.revision_id)
2533
2534
def _filtered_revision_trees(self, revision_ids, file_ids):
2535
"""Return Tree for a revision on this branch with only some files.
2536
2537
:param revision_ids: a sequence of revision-ids;
2538
a revision-id may not be None or 'null:'
2539
:param file_ids: if not None, the result is filtered
2540
so that only those file-ids, their parents and their
2541
children are included.
2542
"""
2543
inventories = self.iter_inventories(revision_ids)
2544
for inv in inventories:
2545
# Should we introduce a FilteredRevisionTree class rather
2546
# than pre-filter the inventory here?
2547
filtered_inv = inv.filter(file_ids)
2548
yield InventoryRevisionTree(self, filtered_inv, filtered_inv.revision_id)
2549
2550
@needs_read_lock
2551
def get_ancestry(self, revision_id, topo_sorted=True):
2552
"""Return a list of revision-ids integrated by a revision.
2553
2554
The first element of the list is always None, indicating the origin
2555
revision. This might change when we have history horizons, or
2556
perhaps we should have a new API.
2557
2558
This is topologically sorted.
2559
"""
2560
if _mod_revision.is_null(revision_id):
2561
return [None]
2562
if not self.has_revision(revision_id):
2563
raise errors.NoSuchRevision(self, revision_id)
2564
graph = self.get_graph()
2565
keys = set()
2566
search = graph._make_breadth_first_searcher([revision_id])
2567
while True:
2568
try:
2569
found, ghosts = search.next_with_ghosts()
2570
except StopIteration:
2571
break
2572
keys.update(found)
2573
if _mod_revision.NULL_REVISION in keys:
2574
keys.remove(_mod_revision.NULL_REVISION)
2575
if topo_sorted:
2576
parent_map = graph.get_parent_map(keys)
2577
keys = tsort.topo_sort(parent_map)
2578
return [None] + list(keys)
1024
raise NotImplementedError(self.revision_trees)
2579
1025
2580
1026
def pack(self, hint=None, clean_obsolete_packs=False):
2581
1027
"""Compress the data within the repository.
2603
1049
2604
1050
def get_parent_map(self, revision_ids):
2605
1051
"""See graph.StackedParentsProvider.get_parent_map"""
1052
raise NotImplementedError(self.get_parent_map)
1053
1054
def _get_parent_map_no_fallbacks(self, revision_ids):
1055
"""Same as Repository.get_parent_map except doesn't query fallbacks."""
2606
1056
# revisions index works in keys; this just works in revisions
2607
1057
# therefore wrap and unwrap
2608
1058
query_keys = []
2614
1064
raise ValueError('get_parent_map(None) is not valid')
2615
1065
else:
2616
1066
query_keys.append((revision_id ,))
1067
vf = self.revisions.without_fallbacks()
2617
1068
for ((revision_id,), parent_keys) in \
2618
self.revisions.get_parent_map(query_keys).iteritems():
1069
vf.get_parent_map(query_keys).iteritems():
2619
1070
if parent_keys:
2620
1071
result[revision_id] = tuple([parent_revid
2621
1072
for (parent_revid,) in parent_keys])
2624
1075
return result
2625
1076
2626
1077
def _make_parents_provider(self):
2627
return self
1078
if not self._format.supports_external_lookups:
1079
return self
1080
return graph.StackedParentsProvider(_LazyListJoin(
1081
[self._make_parents_provider_unstacked()],
1082
self._fallback_repositories))
1083
1084
def _make_parents_provider_unstacked(self):
1085
return graph.CallableToParentsProviderAdapter(
1086
self._get_parent_map_no_fallbacks)
2628
1087
2629
1088
@needs_read_lock
2630
1089
def get_known_graph_ancestry(self, revision_ids):
2631
1090
"""Return the known graph for a set of revision ids and their ancestors.
2632
1091
"""
2633
st = static_tuple.StaticTuple
2634
revision_keys = [st(r_id).intern() for r_id in revision_ids]
2635
known_graph = self.revisions.get_known_graph_ancestry(revision_keys)
2636
return graph.GraphThunkIdsToKeys(known_graph)
1092
raise NotImplementedError(self.get_known_graph_ancestry)
1093
1094
def get_file_graph(self):
1095
"""Return the graph walker for files."""
1096
raise NotImplementedError(self.get_file_graph)
2637
1097
2638
1098
def get_graph(self, other_repository=None):
2639
1099
"""Return the graph walker for this repository format"""
2644
1104
[parents_provider, other_repository._make_parents_provider()])
2645
1105
return graph.Graph(parents_provider)
2646
1106
2647
def _get_versioned_file_checker(self, text_key_references=None,
2648
ancestors=None):
2649
"""Return an object suitable for checking versioned files.
2650
2651
:param text_key_references: if non-None, an already built
2652
dictionary mapping text keys ((fileid, revision_id) tuples)
2653
to whether they were referred to by the inventory of the
2654
revision_id that they contain. If None, this will be
2655
calculated.
2656
:param ancestors: Optional result from
2657
self.get_graph().get_parent_map(self.all_revision_ids()) if already
2658
available.
2659
"""
2660
return _VersionedFileChecker(self,
2661
text_key_references=text_key_references, ancestors=ancestors)
2662
2663
def revision_ids_to_search_result(self, result_set):
2664
"""Convert a set of revision ids to a graph SearchResult."""
2665
result_parents = set()
2666
for parents in self.get_graph().get_parent_map(
2667
result_set).itervalues():
2668
result_parents.update(parents)
2669
included_keys = result_set.intersection(result_parents)
2670
start_keys = result_set.difference(included_keys)
2671
exclude_keys = result_parents.difference(result_set)
2672
result = graph.SearchResult(start_keys, exclude_keys,
2673
len(result_set), result_set)
2674
return result
2675
2676
1107
@needs_write_lock
2677
1108
def set_make_working_trees(self, new_value):
2678
1109
"""Set the policy flag for making working trees when creating branches.
2691
1122
2692
1123
@needs_write_lock
2693
1124
def sign_revision(self, revision_id, gpg_strategy):
2694
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1125
testament = _mod_testament.Testament.from_revision(self, revision_id)
1126
plaintext = testament.as_short_text()
2695
1127
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2696
1128
2697
1129
@needs_read_lock
1130
def verify_revision_signature(self, revision_id, gpg_strategy):
1131
"""Verify the signature on a revision.
1132
1133
:param revision_id: the revision to verify
1134
:gpg_strategy: the GPGStrategy object to used
1135
1136
:return: gpg.SIGNATURE_VALID or a failed SIGNATURE_ value
1137
"""
1138
if not self.has_signature_for_revision_id(revision_id):
1139
return gpg.SIGNATURE_NOT_SIGNED, None
1140
signature = self.get_signature_text(revision_id)
1141
1142
testament = _mod_testament.Testament.from_revision(self, revision_id)
1143
plaintext = testament.as_short_text()
1144
1145
return gpg_strategy.verify(signature, plaintext)
1146
1147
@needs_read_lock
1148
def verify_revision_signatures(self, revision_ids, gpg_strategy):
1149
"""Verify revision signatures for a number of revisions.
1150
1151
:param revision_id: the revision to verify
1152
:gpg_strategy: the GPGStrategy object to used
1153
:return: Iterator over tuples with revision id, result and keys
1154
"""
1155
for revid in revision_ids:
1156
(result, key) = self.verify_revision_signature(revid, gpg_strategy)
1157
yield revid, result, key
1158
2698
1159
def has_signature_for_revision_id(self, revision_id):
2699
1160
"""Query for a revision signature for revision_id in the repository."""
2700
if not self.has_revision(revision_id):
2701
raise errors.NoSuchRevision(self, revision_id)
2702
sig_present = (1 == len(
2703
self.signatures.get_parent_map([(revision_id,)])))
2704
return sig_present
1161
raise NotImplementedError(self.has_signature_for_revision_id)
2705
1162
2706
@needs_read_lock
2707
1163
def get_signature_text(self, revision_id):
2708
1164
"""Return the text for a signature."""
2709
stream = self.signatures.get_record_stream([(revision_id,)],
2710
'unordered', True)
2711
record = stream.next()
2712
if record.storage_kind == 'absent':
2713
raise errors.NoSuchRevision(self, revision_id)
2714
return record.get_bytes_as('fulltext')
1165
raise NotImplementedError(self.get_signature_text)
2715
1166
2716
@needs_read_lock
2717
1167
def check(self, revision_ids=None, callback_refs=None, check_repo=True):
2718
1168
"""Check consistency of all history of given revision_ids.
2719
1169
2727
1177
:param check_repo: If False do not check the repository contents, just
2728
1178
calculate the data callback_refs requires and call them back.
2729
1179
"""
2730
return self._check(revision_ids, callback_refs=callback_refs,
1180
return self._check(revision_ids=revision_ids, callback_refs=callback_refs,
2731
1181
check_repo=check_repo)
2732
1182
2733
def _check(self, revision_ids, callback_refs, check_repo):
2734
result = check.Check(self, check_repo=check_repo)
2735
result.check(callback_refs)
2736
return result
1183
def _check(self, revision_ids=None, callback_refs=None, check_repo=True):
1184
raise NotImplementedError(self.check)
2737
1185
2738
1186
def _warn_if_deprecated(self, branch=None):
2739
1187
if not self._format.is_deprecated():
2743
1191
return
2744
1192
try:
2745
1193
if branch is None:
2746
conf = config.GlobalConfig()
1194
conf = config.GlobalStack()
2747
1195
else:
2748
conf = branch.get_config()
2749
if conf.suppress_warning('format_deprecation'):
1196
conf = branch.get_config_stack()
1197
if 'format_deprecation' in conf.get('suppress_warnings'):
2750
1198
return
2751
1199
warning("Format %s for %s is deprecated -"
2752
1200
" please use 'bzr upgrade' to get better performance"
2773
1221
except UnicodeDecodeError:
2774
1222
raise errors.NonAsciiRevisionId(method, self)
2775
1223
2776
def _find_inconsistent_revision_parents(self, revisions_iterator=None):
2777
"""Find revisions with different parent lists in the revision object
2778
and in the index graph.
2779
2780
:param revisions_iterator: None, or an iterator of (revid,
2781
Revision-or-None). This iterator controls the revisions checked.
2782
:returns: an iterator yielding tuples of (revison-id, parents-in-index,
2783
parents-in-revision).
2784
"""
2785
if not self.is_locked():
2786
raise AssertionError()
2787
vf = self.revisions
2788
if revisions_iterator is None:
2789
revisions_iterator = self._iter_revisions(None)
2790
for revid, revision in revisions_iterator:
2791
if revision is None:
2792
pass
2793
parent_map = vf.get_parent_map([(revid,)])
2794
parents_according_to_index = tuple(parent[-1] for parent in
2795
parent_map[(revid,)])
2796
parents_according_to_revision = tuple(revision.parent_ids)
2797
if parents_according_to_index != parents_according_to_revision:
2798
yield (revid, parents_according_to_index,
2799
parents_according_to_revision)
2800
2801
def _check_for_inconsistent_revision_parents(self):
2802
inconsistencies = list(self._find_inconsistent_revision_parents())
2803
if inconsistencies:
2804
raise errors.BzrCheckError(
2805
"Revision knit has inconsistent parents.")
2806
2807
2808
def install_revision(repository, rev, revision_tree):
2809
"""Install all revision data into a repository."""
2810
install_revisions(repository, [(rev, revision_tree, None)])
2811
2812
2813
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2814
"""Install all revision data into a repository.
2815
2816
Accepts an iterable of revision, tree, signature tuples. The signature
2817
may be None.
2818
"""
2819
repository.start_write_group()
2820
try:
2821
inventory_cache = lru_cache.LRUCache(10)
2822
for n, (revision, revision_tree, signature) in enumerate(iterable):
2823
_install_revision(repository, revision, revision_tree, signature,
2824
inventory_cache)
2825
if pb is not None:
2826
pb.update('Transferring revisions', n + 1, num_revisions)
2827
except:
2828
repository.abort_write_group()
2829
raise
2830
else:
2831
repository.commit_write_group()
2832
2833
2834
def _install_revision(repository, rev, revision_tree, signature,
2835
inventory_cache):
2836
"""Install all revision data into a repository."""
2837
present_parents = []
2838
parent_trees = {}
2839
for p_id in rev.parent_ids:
2840
if repository.has_revision(p_id):
2841
present_parents.append(p_id)
2842
parent_trees[p_id] = repository.revision_tree(p_id)
2843
else:
2844
parent_trees[p_id] = repository.revision_tree(
2845
_mod_revision.NULL_REVISION)
2846
2847
inv = revision_tree.inventory
2848
entries = inv.iter_entries()
2849
# backwards compatibility hack: skip the root id.
2850
if not repository.supports_rich_root():
2851
path, root = entries.next()
2852
if root.revision != rev.revision_id:
2853
raise errors.IncompatibleRevision(repr(repository))
2854
text_keys = {}
2855
for path, ie in entries:
2856
text_keys[(ie.file_id, ie.revision)] = ie
2857
text_parent_map = repository.texts.get_parent_map(text_keys)
2858
missing_texts = set(text_keys) - set(text_parent_map)
2859
# Add the texts that are not already present
2860
for text_key in missing_texts:
2861
ie = text_keys[text_key]
2862
text_parents = []
2863
# FIXME: TODO: The following loop overlaps/duplicates that done by
2864
# commit to determine parents. There is a latent/real bug here where
2865
# the parents inserted are not those commit would do - in particular
2866
# they are not filtered by heads(). RBC, AB
2867
for revision, tree in parent_trees.iteritems():
2868
if ie.file_id not in tree:
2869
continue
2870
parent_id = tree.get_file_revision(ie.file_id)
2871
if parent_id in text_parents:
2872
continue
2873
text_parents.append((ie.file_id, parent_id))
2874
lines = revision_tree.get_file(ie.file_id).readlines()
2875
repository.texts.add_lines(text_key, text_parents, lines)
2876
try:
2877
# install the inventory
2878
if repository._format._commit_inv_deltas and len(rev.parent_ids):
2879
# Cache this inventory
2880
inventory_cache[rev.revision_id] = inv
2881
try:
2882
basis_inv = inventory_cache[rev.parent_ids[0]]
2883
except KeyError:
2884
repository.add_inventory(rev.revision_id, inv, present_parents)
2885
else:
2886
delta = inv._make_delta(basis_inv)
2887
repository.add_inventory_by_delta(rev.parent_ids[0], delta,
2888
rev.revision_id, present_parents)
2889
else:
2890
repository.add_inventory(rev.revision_id, inv, present_parents)
2891
except errors.RevisionAlreadyPresent:
2892
pass
2893
if signature is not None:
2894
repository.add_signature_text(rev.revision_id, signature)
2895
repository.add_revision(rev.revision_id, rev, inv)
2896
2897
1224
2898
1225
class MetaDirRepository(Repository):
2899
1226
"""Repositories in the new meta-dir layout.
2933
1260
"""Returns the policy for making working trees on new branches."""
2934
1261
return not self._transport.has('no-working-trees')
2935
1262
2936
2937
class MetaDirVersionedFileRepository(MetaDirRepository):
2938
"""Repositories in a meta-dir, that work via versioned file objects."""
2939
2940
def __init__(self, _format, a_bzrdir, control_files):
2941
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2942
control_files)
1263
@needs_write_lock
1264
def update_feature_flags(self, updated_flags):
1265
"""Update the feature flags for this branch.
1266
1267
:param updated_flags: Dictionary mapping feature names to necessities
1268
A necessity can be None to indicate the feature should be removed
1269
"""
1270
self._format._update_feature_flags(updated_flags)
1271
self.control_transport.put_bytes('format', self._format.as_string())
2943
1272
2944
1273
2945
1274
class RepositoryFormatRegistry(controldir.ControlComponentFormatRegistry):
2947
1276
2948
1277
def get_default(self):
2949
1278
"""Return the current default format."""
2950
from bzrlib import bzrdir
2951
return bzrdir.format_registry.make_bzrdir('default').repository_format
1279
return controldir.format_registry.make_bzrdir('default').repository_format
2952
1280
2953
1281
2954
1282
network_format_registry = registry.FormatRegistry()
2997
1325
created.
2998
1326
2999
1327
Common instance attributes:
3000
_matchingbzrdir - the bzrdir format that the repository format was
1328
_matchingbzrdir - the controldir format that the repository format was
3001
1329
originally written to work with. This can be used if manually
3002
1330
constructing a bzrdir and repository, or more commonly for test suite
3003
1331
parameterization.
3012
1340
# Does this format support CHK bytestring lookups. Set to True or False in
3013
1341
# derived classes.
3014
1342
supports_chks = None
3015
# Should commit add an inventory, or an inventory delta to the repository.
3016
_commit_inv_deltas = True
3017
# What order should fetch operations request streams in?
3018
# The default is unordered as that is the cheapest for an origin to
3019
# provide.
3020
_fetch_order = 'unordered'
3021
# Does this repository format use deltas that can be fetched as-deltas ?
3022
# (E.g. knits, where the knit deltas can be transplanted intact.
3023
# We default to False, which will ensure that enough data to get
3024
# a full text out of any fetch stream will be grabbed.
3025
_fetch_uses_deltas = False
3026
1343
# Should fetch trigger a reconcile after the fetch? Only needed for
3027
1344
# some repository formats that can suffer internal inconsistencies.
3028
1345
_fetch_reconcile = False
3034
1351
# help), and for fetching when data won't have come from the same
3035
1352
# compressor.
3036
1353
pack_compresses = False
3037
# Does the repository inventory storage understand references to trees?
1354
# Does the repository storage understand references to trees?
3038
1355
supports_tree_reference = None
3039
1356
# Is the format experimental ?
3040
1357
experimental = False
3041
# Does this repository format escape funky characters, or does it create files with
3042
# similar names as the versioned files in its contents on disk ?
1358
# Does this repository format escape funky characters, or does it create
1359
# files with similar names as the versioned files in its contents on disk
1360
# ?
3043
1361
supports_funky_characters = None
3044
1362
# Does this repository format support leaving locks?
3045
1363
supports_leaving_lock = None
3049
1367
supports_revision_signatures = True
3050
1368
# Can the revision graph have incorrect parents?
3051
1369
revision_graph_can_have_wrong_parents = None
1370
# Does this format support rich root data?
1371
rich_root_data = None
1372
# Does this format support explicitly versioned directories?
1373
supports_versioned_directories = None
1374
# Can other repositories be nested into one of this format?
1375
supports_nesting_repositories = None
1376
# Is it possible for revisions to be present without being referenced
1377
# somewhere ?
1378
supports_unreferenced_revisions = None
3052
1379
3053
1380
def __repr__(self):
3054
1381
return "%s()" % self.__class__.__name__
3060
1387
def __ne__(self, other):
3061
1388
return not self == other
3062
1389
3063
@classmethod
3064
def find_format(klass, a_bzrdir):
3065
"""Return the format for the repository object in a_bzrdir.
3066
3067
This is used by bzr native formats that have a "format" file in
3068
the repository. Other methods may be used by different types of
3069
control directory.
3070
"""
3071
try:
3072
transport = a_bzrdir.get_repository_transport(None)
3073
format_string = transport.get_bytes("format")
3074
return format_registry.get(format_string)
3075
except errors.NoSuchFile:
3076
raise errors.NoRepositoryPresent(a_bzrdir)
3077
except KeyError:
3078
raise errors.UnknownFormatError(format=format_string,
3079
kind='repository')
3080
3081
@classmethod
3082
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3083
def register_format(klass, format):
3084
format_registry.register(format)
3085
3086
@classmethod
3087
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3088
def unregister_format(klass, format):
3089
format_registry.remove(format)
3090
3091
@classmethod
3092
@symbol_versioning.deprecated_method(symbol_versioning.deprecated_in((2, 4, 0)))
3093
def get_default_format(klass):
3094
"""Return the current default format."""
3095
return format_registry.get_default()
3096
3097
def get_format_string(self):
3098
"""Return the ASCII format string that identifies this format.
3099
3100
Note that in pre format ?? repositories the format string is
3101
not permitted nor written to disk.
3102
"""
3103
raise NotImplementedError(self.get_format_string)
3104
3105
1390
def get_format_description(self):
3106
1391
"""Return the short description for this format."""
3107
1392
raise NotImplementedError(self.get_format_description)
3108
1393
3109
# TODO: this shouldn't be in the base class, it's specific to things that
3110
# use weaves or knits -- mbp 20070207
3111
def _get_versioned_file_store(self,
3112
name,
3113
transport,
3114
control_files,
3115
prefixed=True,
3116
versionedfile_class=None,
3117
versionedfile_kwargs={},
3118
escaped=False):
3119
if versionedfile_class is None:
3120
versionedfile_class = self._versionedfile_class
3121
weave_transport = control_files._transport.clone(name)
3122
dir_mode = control_files._dir_mode
3123
file_mode = control_files._file_mode
3124
return VersionedFileStore(weave_transport, prefixed=prefixed,
3125
dir_mode=dir_mode,
3126
file_mode=file_mode,
3127
versionedfile_class=versionedfile_class,
3128
versionedfile_kwargs=versionedfile_kwargs,
3129
escaped=escaped)
3130
3131
def initialize(self, a_bzrdir, shared=False):
3132
"""Initialize a repository of this format in a_bzrdir.
3133
3134
:param a_bzrdir: The bzrdir to put the new repository in it.
1394
def initialize(self, controldir, shared=False):
1395
"""Initialize a repository of this format in controldir.
1396
1397
:param controldir: The controldir to put the new repository in it.
3135
1398
:param shared: The repository should be initialized as a sharable one.
3136
1399
:returns: The new repository object.
3137
1400
3138
1401
This may raise UninitializableFormat if shared repository are not
3139
compatible the a_bzrdir.
1402
compatible the controldir.
3140
1403
"""
3141
1404
raise NotImplementedError(self.initialize)
3142
1405
3178
1441
'Does not support nested trees', target_format,
3179
1442
from_format=self)
3180
1443
3181
def open(self, a_bzrdir, _found=False):
3182
"""Return an instance of this format for the bzrdir a_bzrdir.
1444
def open(self, controldir, _found=False):
1445
"""Return an instance of this format for a controldir.
3183
1446
3184
1447
_found is a private parameter, do not use it.
3185
1448
"""
3186
1449
raise NotImplementedError(self.open)
3187
1450
3188
def _run_post_repo_init_hooks(self, repository, a_bzrdir, shared):
3189
from bzrlib.bzrdir import BzrDir, RepoInitHookParams
3190
hooks = BzrDir.hooks['post_repo_init']
1451
def _run_post_repo_init_hooks(self, repository, controldir, shared):
1452
from bzrlib.controldir import ControlDir, RepoInitHookParams
1453
hooks = ControlDir.hooks['post_repo_init']
3191
1454
if not hooks:
3192
1455
return
3193
params = RepoInitHookParams(repository, self, a_bzrdir, shared)
1456
params = RepoInitHookParams(repository, self, controldir, shared)
3194
1457
for hook in hooks:
3195
1458
hook(params)
3196
1459
3197
1460
3198
class MetaDirRepositoryFormat(RepositoryFormat):
1461
class RepositoryFormatMetaDir(bzrdir.BzrFormat, RepositoryFormat):
3199
1462
"""Common base class for the new repositories using the metadir layout."""
3200
1463
3201
1464
rich_root_data = False
3202
1465
supports_tree_reference = False
3203
1466
supports_external_lookups = False
3204
1467
supports_leaving_lock = True
1468
supports_nesting_repositories = True
3205
1469
3206
1470
@property
3207
1471
def _matchingbzrdir(self):
3210
1474
return matching
3211
1475
3212
1476
def __init__(self):
3213
super(MetaDirRepositoryFormat, self).__init__()
1477
RepositoryFormat.__init__(self)
1478
bzrdir.BzrFormat.__init__(self)
3214
1479
3215
1480
def _create_control_files(self, a_bzrdir):
3216
1481
"""Create the required files and the initial control_files object."""
3240
1505
finally:
3241
1506
control_files.unlock()
3242
1507
3243
def network_name(self):
3244
"""Metadir formats have matching disk and network format strings."""
3245
return self.get_format_string()
1508
@classmethod
1509
def find_format(klass, a_bzrdir):
1510
"""Return the format for the repository object in a_bzrdir.
1511
1512
This is used by bzr native formats that have a "format" file in
1513
the repository. Other methods may be used by different types of
1514
control directory.
1515
"""
1516
try:
1517
transport = a_bzrdir.get_repository_transport(None)
1518
format_string = transport.get_bytes("format")
1519
except errors.NoSuchFile:
1520
raise errors.NoRepositoryPresent(a_bzrdir)
1521
return klass._find_format(format_registry, 'repository', format_string)
1522
1523
def check_support_status(self, allow_unsupported, recommend_upgrade=True,
1524
basedir=None):
1525
RepositoryFormat.check_support_status(self,
1526
allow_unsupported=allow_unsupported, recommend_upgrade=recommend_upgrade,
1527
basedir=basedir)
1528
bzrdir.BzrFormat.check_support_status(self, allow_unsupported=allow_unsupported,
1529
recommend_upgrade=recommend_upgrade, basedir=basedir)
3246
1530
3247
1531
3248
1532
# formats which have no format string are not discoverable or independently
3249
1533
# creatable on disk, so are not registered in format_registry. They're
3250
1534
# all in bzrlib.repofmt.knitreponow. When an instance of one of these is
3251
# needed, it's constructed directly by the BzrDir. Non-native formats where
1535
# needed, it's constructed directly by the ControlDir. Non-native formats where
3252
1536
# the repository is not separately opened are similar.
3253
1537
3254
1538
format_registry.register_lazy(
3345
1629
InterRepository.get(other).method_name(parameters).
3346
1630
"""
3347
1631
3348
_walk_to_common_revisions_batch_size = 50
3349
1632
_optimisers = []
3350
1633
"""The available optimised InterRepository types."""
3351
1634
3366
1649
self.target.fetch(self.source, revision_id=revision_id)
3367
1650
3368
1651
@needs_write_lock
3369
def fetch(self, revision_id=None, find_ghosts=False,
3370
fetch_spec=None):
1652
def fetch(self, revision_id=None, find_ghosts=False):
3371
1653
"""Fetch the content required to construct revision_id.
3372
1654
3373
1655
The content is copied from self.source to self.target.
3376
1658
content is copied.
3377
1659
:return: None.
3378
1660
"""
3379
ui.ui_factory.warn_experimental_format_fetch(self)
3380
from bzrlib.fetch import RepoFetcher
3381
# See <https://launchpad.net/bugs/456077> asking for a warning here
3382
if self.source._format.network_name() != self.target._format.network_name():
3383
ui.ui_factory.show_user_warning('cross_format_fetch',
3384
from_format=self.source._format,
3385
to_format=self.target._format)
3386
f = RepoFetcher(to_repository=self.target,
3387
from_repository=self.source,
3388
last_revision=revision_id,
3389
fetch_spec=fetch_spec,
3390
find_ghosts=find_ghosts)
3391
3392
def _walk_to_common_revisions(self, revision_ids, if_present_ids=None):
3393
"""Walk out from revision_ids in source to revisions target has.
3394
3395
:param revision_ids: The start point for the search.
3396
:return: A set of revision ids.
3397
"""
3398
target_graph = self.target.get_graph()
3399
revision_ids = frozenset(revision_ids)
3400
if if_present_ids:
3401
all_wanted_revs = revision_ids.union(if_present_ids)
3402
else:
3403
all_wanted_revs = revision_ids
3404
missing_revs = set()
3405
source_graph = self.source.get_graph()
3406
# ensure we don't pay silly lookup costs.
3407
searcher = source_graph._make_breadth_first_searcher(all_wanted_revs)
3408
null_set = frozenset([_mod_revision.NULL_REVISION])
3409
searcher_exhausted = False
3410
while True:
3411
next_revs = set()
3412
ghosts = set()
3413
# Iterate the searcher until we have enough next_revs
3414
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3415
try:
3416
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3417
next_revs.update(next_revs_part)
3418
ghosts.update(ghosts_part)
3419
except StopIteration:
3420
searcher_exhausted = True
3421
break
3422
# If there are ghosts in the source graph, and the caller asked for
3423
# them, make sure that they are present in the target.
3424
# We don't care about other ghosts as we can't fetch them and
3425
# haven't been asked to.
3426
ghosts_to_check = set(revision_ids.intersection(ghosts))
3427
revs_to_get = set(next_revs).union(ghosts_to_check)
3428
if revs_to_get:
3429
have_revs = set(target_graph.get_parent_map(revs_to_get))
3430
# we always have NULL_REVISION present.
3431
have_revs = have_revs.union(null_set)
3432
# Check if the target is missing any ghosts we need.
3433
ghosts_to_check.difference_update(have_revs)
3434
if ghosts_to_check:
3435
# One of the caller's revision_ids is a ghost in both the
3436
# source and the target.
3437
raise errors.NoSuchRevision(
3438
self.source, ghosts_to_check.pop())
3439
missing_revs.update(next_revs - have_revs)
3440
# Because we may have walked past the original stop point, make
3441
# sure everything is stopped
3442
stop_revs = searcher.find_seen_ancestors(have_revs)
3443
searcher.stop_searching_any(stop_revs)
3444
if searcher_exhausted:
3445
break
3446
return searcher.get_result()
1661
raise NotImplementedError(self.fetch)
3447
1662
3448
1663
@needs_read_lock
3449
1664
def search_missing_revision_ids(self,
3450
1665
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
3451
find_ghosts=True, revision_ids=None, if_present_ids=None):
1666
find_ghosts=True, revision_ids=None, if_present_ids=None,
1667
limit=None):
3452
1668
"""Return the revision ids that source has that target does not.
3453
1669
3454
1670
:param revision_id: only return revision ids included by this
3462
1678
to fetch for tags, which may reference absent revisions.
3463
1679
:param find_ghosts: If True find missing revisions in deep history
3464
1680
rather than just finding the surface difference.
1681
:param limit: Maximum number of revisions to return, topologically
1682
ordered
3465
1683
:return: A bzrlib.graph.SearchResult.
3466
1684
"""
3467
if symbol_versioning.deprecated_passed(revision_id):
3468
symbol_versioning.warn(
3469
'search_missing_revision_ids(revision_id=...) was '
3470
'deprecated in 2.4. Use revision_ids=[...] instead.',
3471
DeprecationWarning, stacklevel=2)
3472
if revision_ids is not None:
3473
raise AssertionError(
3474
'revision_ids is mutually exclusive with revision_id')
3475
if revision_id is not None:
3476
revision_ids = [revision_id]
3477
del revision_id
3478
# stop searching at found target revisions.
3479
if not find_ghosts and (revision_ids is not None or if_present_ids is
3480
not None):
3481
return self._walk_to_common_revisions(revision_ids,
3482
if_present_ids=if_present_ids)
3483
# generic, possibly worst case, slow code path.
3484
target_ids = set(self.target.all_revision_ids())
3485
source_ids = self._present_source_revisions_for(
3486
revision_ids, if_present_ids)
3487
result_set = set(source_ids).difference(target_ids)
3488
return self.source.revision_ids_to_search_result(result_set)
3489
3490
def _present_source_revisions_for(self, revision_ids, if_present_ids=None):
3491
"""Returns set of all revisions in ancestry of revision_ids present in
3492
the source repo.
3493
3494
:param revision_ids: if None, all revisions in source are returned.
3495
:param if_present_ids: like revision_ids, but if any/all of these are
3496
absent no error is raised.
3497
"""
3498
if revision_ids is not None or if_present_ids is not None:
3499
# First, ensure all specified revisions exist. Callers expect
3500
# NoSuchRevision when they pass absent revision_ids here.
3501
if revision_ids is None:
3502
revision_ids = set()
3503
if if_present_ids is None:
3504
if_present_ids = set()
3505
revision_ids = set(revision_ids)
3506
if_present_ids = set(if_present_ids)
3507
all_wanted_ids = revision_ids.union(if_present_ids)
3508
graph = self.source.get_graph()
3509
present_revs = set(graph.get_parent_map(all_wanted_ids))
3510
missing = revision_ids.difference(present_revs)
3511
if missing:
3512
raise errors.NoSuchRevision(self.source, missing.pop())
3513
found_ids = all_wanted_ids.intersection(present_revs)
3514
source_ids = [rev_id for (rev_id, parents) in
3515
graph.iter_ancestry(found_ids)
3516
if rev_id != _mod_revision.NULL_REVISION
3517
and parents is not None]
3518
else:
3519
source_ids = self.source.all_revision_ids()
3520
return set(source_ids)
1685
raise NotImplementedError(self.search_missing_revision_ids)
3521
1686
3522
1687
@staticmethod
3523
1688
def _same_model(source, target):
3544
1709
"different serializers")
3545
1710
3546
1711
3547
class InterSameDataRepository(InterRepository):
3548
"""Code for converting between repositories that represent the same data.
3549
3550
Data format and model must match for this to work.
3551
"""
3552
3553
@classmethod
3554
def _get_repo_format_to_test(self):
3555
"""Repository format for testing with.
3556
3557
InterSameData can pull from subtree to subtree and from non-subtree to
3558
non-subtree, so we test this with the richest repository format.
3559
"""
3560
from bzrlib.repofmt import knitrepo
3561
return knitrepo.RepositoryFormatKnit3()
3562
3563
@staticmethod
3564
def is_compatible(source, target):
3565
return InterRepository._same_model(source, target)
3566
3567
3568
class InterDifferingSerializer(InterRepository):
3569
3570
@classmethod
3571
def _get_repo_format_to_test(self):
3572
return None
3573
3574
@staticmethod
3575
def is_compatible(source, target):
3576
"""Be compatible with Knit2 source and Knit3 target"""
3577
# This is redundant with format.check_conversion_target(), however that
3578
# raises an exception, and we just want to say "False" as in we won't
3579
# support converting between these formats.
3580
if 'IDS_never' in debug.debug_flags:
3581
return False
3582
if source.supports_rich_root() and not target.supports_rich_root():
3583
return False
3584
if (source._format.supports_tree_reference
3585
and not target._format.supports_tree_reference):
3586
return False
3587
if target._fallback_repositories and target._format.supports_chks:
3588
# IDS doesn't know how to copy CHKs for the parent inventories it
3589
# adds to stacked repos.
3590
return False
3591
if 'IDS_always' in debug.debug_flags:
3592
return True
3593
# Only use this code path for local source and target. IDS does far
3594
# too much IO (both bandwidth and roundtrips) over a network.
3595
if not source.bzrdir.transport.base.startswith('file:///'):
3596
return False
3597
if not target.bzrdir.transport.base.startswith('file:///'):
3598
return False
3599
return True
3600
3601
def _get_trees(self, revision_ids, cache):
3602
possible_trees = []
3603
for rev_id in revision_ids:
3604
if rev_id in cache:
3605
possible_trees.append((rev_id, cache[rev_id]))
3606
else:
3607
# Not cached, but inventory might be present anyway.
3608
try:
3609
tree = self.source.revision_tree(rev_id)
3610
except errors.NoSuchRevision:
3611
# Nope, parent is ghost.
3612
pass
3613
else:
3614
cache[rev_id] = tree
3615
possible_trees.append((rev_id, tree))
3616
return possible_trees
3617
3618
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3619
"""Get the best delta and base for this revision.
3620
3621
:return: (basis_id, delta)
3622
"""
3623
deltas = []
3624
# Generate deltas against each tree, to find the shortest.
3625
texts_possibly_new_in_tree = set()
3626
for basis_id, basis_tree in possible_trees:
3627
delta = tree.inventory._make_delta(basis_tree.inventory)
3628
for old_path, new_path, file_id, new_entry in delta:
3629
if new_path is None:
3630
# This file_id isn't present in the new rev, so we don't
3631
# care about it.
3632
continue
3633
if not new_path:
3634
# Rich roots are handled elsewhere...
3635
continue
3636
kind = new_entry.kind
3637
if kind != 'directory' and kind != 'file':
3638
# No text record associated with this inventory entry.
3639
continue
3640
# This is a directory or file that has changed somehow.
3641
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3642
deltas.append((len(delta), basis_id, delta))
3643
deltas.sort()
3644
return deltas[0][1:]
3645
3646
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3647
"""Find all parent revisions that are absent, but for which the
3648
inventory is present, and copy those inventories.
3649
3650
This is necessary to preserve correctness when the source is stacked
3651
without fallbacks configured. (Note that in cases like upgrade the
3652
source may be not have _fallback_repositories even though it is
3653
stacked.)
3654
"""
3655
parent_revs = set()
3656
for parents in parent_map.values():
3657
parent_revs.update(parents)
3658
present_parents = self.source.get_parent_map(parent_revs)
3659
absent_parents = set(parent_revs).difference(present_parents)
3660
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3661
(rev_id,) for rev_id in absent_parents)
3662
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3663
for parent_tree in self.source.revision_trees(parent_inv_ids):
3664
current_revision_id = parent_tree.get_revision_id()
3665
parents_parents_keys = parent_invs_keys_for_stacking[
3666
(current_revision_id,)]
3667
parents_parents = [key[-1] for key in parents_parents_keys]
3668
basis_id = _mod_revision.NULL_REVISION
3669
basis_tree = self.source.revision_tree(basis_id)
3670
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3671
self.target.add_inventory_by_delta(
3672
basis_id, delta, current_revision_id, parents_parents)
3673
cache[current_revision_id] = parent_tree
3674
3675
def _fetch_batch(self, revision_ids, basis_id, cache):
3676
"""Fetch across a few revisions.
3677
3678
:param revision_ids: The revisions to copy
3679
:param basis_id: The revision_id of a tree that must be in cache, used
3680
as a basis for delta when no other base is available
3681
:param cache: A cache of RevisionTrees that we can use.
3682
:return: The revision_id of the last converted tree. The RevisionTree
3683
for it will be in cache
3684
"""
3685
# Walk though all revisions; get inventory deltas, copy referenced
3686
# texts that delta references, insert the delta, revision and
3687
# signature.
3688
root_keys_to_create = set()
3689
text_keys = set()
3690
pending_deltas = []
3691
pending_revisions = []
3692
parent_map = self.source.get_parent_map(revision_ids)
3693
self._fetch_parent_invs_for_stacking(parent_map, cache)
3694
self.source._safe_to_return_from_cache = True
3695
for tree in self.source.revision_trees(revision_ids):
3696
# Find a inventory delta for this revision.
3697
# Find text entries that need to be copied, too.
3698
current_revision_id = tree.get_revision_id()
3699
parent_ids = parent_map.get(current_revision_id, ())
3700
parent_trees = self._get_trees(parent_ids, cache)
3701
possible_trees = list(parent_trees)
3702
if len(possible_trees) == 0:
3703
# There either aren't any parents, or the parents are ghosts,
3704
# so just use the last converted tree.
3705
possible_trees.append((basis_id, cache[basis_id]))
3706
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3707
possible_trees)
3708
revision = self.source.get_revision(current_revision_id)
3709
pending_deltas.append((basis_id, delta,
3710
current_revision_id, revision.parent_ids))
3711
if self._converting_to_rich_root:
3712
self._revision_id_to_root_id[current_revision_id] = \
3713
tree.get_root_id()
3714
# Determine which texts are in present in this revision but not in
3715
# any of the available parents.
3716
texts_possibly_new_in_tree = set()
3717
for old_path, new_path, file_id, entry in delta:
3718
if new_path is None:
3719
# This file_id isn't present in the new rev
3720
continue
3721
if not new_path:
3722
# This is the root
3723
if not self.target.supports_rich_root():
3724
# The target doesn't support rich root, so we don't
3725
# copy
3726
continue
3727
if self._converting_to_rich_root:
3728
# This can't be copied normally, we have to insert
3729
# it specially
3730
root_keys_to_create.add((file_id, entry.revision))
3731
continue
3732
kind = entry.kind
3733
texts_possibly_new_in_tree.add((file_id, entry.revision))
3734
for basis_id, basis_tree in possible_trees:
3735
basis_inv = basis_tree.inventory
3736
for file_key in list(texts_possibly_new_in_tree):
3737
file_id, file_revision = file_key
3738
try:
3739
entry = basis_inv[file_id]
3740
except errors.NoSuchId:
3741
continue
3742
if entry.revision == file_revision:
3743
texts_possibly_new_in_tree.remove(file_key)
3744
text_keys.update(texts_possibly_new_in_tree)
3745
pending_revisions.append(revision)
3746
cache[current_revision_id] = tree
3747
basis_id = current_revision_id
3748
self.source._safe_to_return_from_cache = False
3749
# Copy file texts
3750
from_texts = self.source.texts
3751
to_texts = self.target.texts
3752
if root_keys_to_create:
3753
root_stream = _mod_fetch._new_root_data_stream(
3754
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3755
self.source)
3756
to_texts.insert_record_stream(root_stream)
3757
to_texts.insert_record_stream(from_texts.get_record_stream(
3758
text_keys, self.target._format._fetch_order,
3759
not self.target._format._fetch_uses_deltas))
3760
# insert inventory deltas
3761
for delta in pending_deltas:
3762
self.target.add_inventory_by_delta(*delta)
3763
if self.target._fallback_repositories:
3764
# Make sure this stacked repository has all the parent inventories
3765
# for the new revisions that we are about to insert. We do this
3766
# before adding the revisions so that no revision is added until
3767
# all the inventories it may depend on are added.
3768
# Note that this is overzealous, as we may have fetched these in an
3769
# earlier batch.
3770
parent_ids = set()
3771
revision_ids = set()
3772
for revision in pending_revisions:
3773
revision_ids.add(revision.revision_id)
3774
parent_ids.update(revision.parent_ids)
3775
parent_ids.difference_update(revision_ids)
3776
parent_ids.discard(_mod_revision.NULL_REVISION)
3777
parent_map = self.source.get_parent_map(parent_ids)
3778
# we iterate over parent_map and not parent_ids because we don't
3779
# want to try copying any revision which is a ghost
3780
for parent_tree in self.source.revision_trees(parent_map):
3781
current_revision_id = parent_tree.get_revision_id()
3782
parents_parents = parent_map[current_revision_id]
3783
possible_trees = self._get_trees(parents_parents, cache)
3784
if len(possible_trees) == 0:
3785
# There either aren't any parents, or the parents are
3786
# ghosts, so just use the last converted tree.
3787
possible_trees.append((basis_id, cache[basis_id]))
3788
basis_id, delta = self._get_delta_for_revision(parent_tree,
3789
parents_parents, possible_trees)
3790
self.target.add_inventory_by_delta(
3791
basis_id, delta, current_revision_id, parents_parents)
3792
# insert signatures and revisions
3793
for revision in pending_revisions:
3794
try:
3795
signature = self.source.get_signature_text(
3796
revision.revision_id)
3797
self.target.add_signature_text(revision.revision_id,
3798
signature)
3799
except errors.NoSuchRevision:
3800
pass
3801
self.target.add_revision(revision.revision_id, revision)
3802
return basis_id
3803
3804
def _fetch_all_revisions(self, revision_ids, pb):
3805
"""Fetch everything for the list of revisions.
3806
3807
:param revision_ids: The list of revisions to fetch. Must be in
3808
topological order.
3809
:param pb: A ProgressTask
3810
:return: None
3811
"""
3812
basis_id, basis_tree = self._get_basis(revision_ids[0])
3813
batch_size = 100
3814
cache = lru_cache.LRUCache(100)
3815
cache[basis_id] = basis_tree
3816
del basis_tree # We don't want to hang on to it here
3817
hints = []
3818
a_graph = None
3819
3820
for offset in range(0, len(revision_ids), batch_size):
3821
self.target.start_write_group()
3822
try:
3823
pb.update('Transferring revisions', offset,
3824
len(revision_ids))
3825
batch = revision_ids[offset:offset+batch_size]
3826
basis_id = self._fetch_batch(batch, basis_id, cache)
3827
except:
3828
self.source._safe_to_return_from_cache = False
3829
self.target.abort_write_group()
3830
raise
3831
else:
3832
hint = self.target.commit_write_group()
3833
if hint:
3834
hints.extend(hint)
3835
if hints and self.target._format.pack_compresses:
3836
self.target.pack(hint=hints)
3837
pb.update('Transferring revisions', len(revision_ids),
3838
len(revision_ids))
3839
3840
@needs_write_lock
3841
def fetch(self, revision_id=None, find_ghosts=False,
3842
fetch_spec=None):
3843
"""See InterRepository.fetch()."""
3844
if fetch_spec is not None:
3845
revision_ids = fetch_spec.get_keys()
3846
else:
3847
revision_ids = None
3848
ui.ui_factory.warn_experimental_format_fetch(self)
3849
if (not self.source.supports_rich_root()
3850
and self.target.supports_rich_root()):
3851
self._converting_to_rich_root = True
3852
self._revision_id_to_root_id = {}
3853
else:
3854
self._converting_to_rich_root = False
3855
# See <https://launchpad.net/bugs/456077> asking for a warning here
3856
if self.source._format.network_name() != self.target._format.network_name():
3857
ui.ui_factory.show_user_warning('cross_format_fetch',
3858
from_format=self.source._format,
3859
to_format=self.target._format)
3860
if revision_ids is None:
3861
if revision_id:
3862
search_revision_ids = [revision_id]
3863
else:
3864
search_revision_ids = None
3865
revision_ids = self.target.search_missing_revision_ids(self.source,
3866
revision_ids=search_revision_ids,
3867
find_ghosts=find_ghosts).get_keys()
3868
if not revision_ids:
3869
return 0, 0
3870
revision_ids = tsort.topo_sort(
3871
self.source.get_graph().get_parent_map(revision_ids))
3872
if not revision_ids:
3873
return 0, 0
3874
# Walk though all revisions; get inventory deltas, copy referenced
3875
# texts that delta references, insert the delta, revision and
3876
# signature.
3877
pb = ui.ui_factory.nested_progress_bar()
3878
try:
3879
self._fetch_all_revisions(revision_ids, pb)
3880
finally:
3881
pb.finished()
3882
return len(revision_ids), 0
3883
3884
def _get_basis(self, first_revision_id):
3885
"""Get a revision and tree which exists in the target.
3886
3887
This assumes that first_revision_id is selected for transmission
3888
because all other ancestors are already present. If we can't find an
3889
ancestor we fall back to NULL_REVISION since we know that is safe.
3890
3891
:return: (basis_id, basis_tree)
3892
"""
3893
first_rev = self.source.get_revision(first_revision_id)
3894
try:
3895
basis_id = first_rev.parent_ids[0]
3896
# only valid as a basis if the target has it
3897
self.target.get_revision(basis_id)
3898
# Try to get a basis tree - if it's a ghost it will hit the
3899
# NoSuchRevision case.
3900
basis_tree = self.source.revision_tree(basis_id)
3901
except (IndexError, errors.NoSuchRevision):
3902
basis_id = _mod_revision.NULL_REVISION
3903
basis_tree = self.source.revision_tree(basis_id)
3904
return basis_id, basis_tree
3905
3906
3907
InterRepository.register_optimiser(InterDifferingSerializer)
3908
InterRepository.register_optimiser(InterSameDataRepository)
3909
3910
3911
1712
class CopyConverter(object):
3912
1713
"""A repository conversion tool which just performs a copy of the content.
3913
1714
3934
1735
# trigger an assertion if not such
3935
1736
repo._format.get_format_string()
3936
1737
self.repo_dir = repo.bzrdir
3937
pb.update('Moving repository to repository.backup')
1738
pb.update(gettext('Moving repository to repository.backup'))
3938
1739
self.repo_dir.transport.move('repository', 'repository.backup')
3939
1740
backup_transport = self.repo_dir.transport.clone('repository.backup')
3940
1741
repo._format.check_conversion_target(self.target_format)
3941
1742
self.source_repo = repo._format.open(self.repo_dir,
3942
1743
_found=True,
3943
1744
_override_transport=backup_transport)
3944
pb.update('Creating new repository')
1745
pb.update(gettext('Creating new repository'))
3945
1746
converted = self.target_format.initialize(self.repo_dir,
3946
1747
self.source_repo.is_shared())
3947
1748
converted.lock_write()
3948
1749
try:
3949
pb.update('Copying content')
1750
pb.update(gettext('Copying content'))
3950
1751
self.source_repo.copy_content_into(converted)
3951
1752
finally:
3952
1753
converted.unlock()
3953
pb.update('Deleting old repository content')
1754
pb.update(gettext('Deleting old repository content'))
3954
1755
self.repo_dir.transport.delete_tree('repository.backup')
3955
ui.ui_factory.note('repository converted')
1756
ui.ui_factory.note(gettext('repository converted'))
3956
1757
pb.finished()
3957
1758
3958
1759
3959
class _VersionedFileChecker(object):
3960
3961
def __init__(self, repository, text_key_references=None, ancestors=None):
3962
self.repository = repository
3963
self.text_index = self.repository._generate_text_key_index(
3964
text_key_references=text_key_references, ancestors=ancestors)
3965
3966
def calculate_file_version_parents(self, text_key):
3967
"""Calculate the correct parents for a file version according to
3968
the inventories.
3969
"""
3970
parent_keys = self.text_index[text_key]
3971
if parent_keys == [_mod_revision.NULL_REVISION]:
3972
return ()
3973
return tuple(parent_keys)
3974
3975
def check_file_version_parents(self, texts, progress_bar=None):
3976
"""Check the parents stored in a versioned file are correct.
3977
3978
It also detects file versions that are not referenced by their
3979
corresponding revision's inventory.
3980
3981
:returns: A tuple of (wrong_parents, dangling_file_versions).
3982
wrong_parents is a dict mapping {revision_id: (stored_parents,
3983
correct_parents)} for each revision_id where the stored parents
3984
are not correct. dangling_file_versions is a set of (file_id,
3985
revision_id) tuples for versions that are present in this versioned
3986
file, but not used by the corresponding inventory.
3987
"""
3988
local_progress = None
3989
if progress_bar is None:
3990
local_progress = ui.ui_factory.nested_progress_bar()
3991
progress_bar = local_progress
3992
try:
3993
return self._check_file_version_parents(texts, progress_bar)
3994
finally:
3995
if local_progress:
3996
local_progress.finished()
3997
3998
def _check_file_version_parents(self, texts, progress_bar):
3999
"""See check_file_version_parents."""
4000
wrong_parents = {}
4001
self.file_ids = set([file_id for file_id, _ in
4002
self.text_index.iterkeys()])
4003
# text keys is now grouped by file_id
4004
n_versions = len(self.text_index)
4005
progress_bar.update('loading text store', 0, n_versions)
4006
parent_map = self.repository.texts.get_parent_map(self.text_index)
4007
# On unlistable transports this could well be empty/error...
4008
text_keys = self.repository.texts.keys()
4009
unused_keys = frozenset(text_keys) - set(self.text_index)
4010
for num, key in enumerate(self.text_index.iterkeys()):
4011
progress_bar.update('checking text graph', num, n_versions)
4012
correct_parents = self.calculate_file_version_parents(key)
4013
try:
4014
knit_parents = parent_map[key]
4015
except errors.RevisionNotPresent:
4016
# Missing text!
4017
knit_parents = None
4018
if correct_parents != knit_parents:
4019
wrong_parents[key] = (knit_parents, correct_parents)
4020
return wrong_parents, unused_keys
4021
4022
4023
1760
def _strip_NULL_ghosts(revision_graph):
4024
1761
"""Also don't use this. more compatibility code for unmigrated clients."""
4025
1762
# Filter ghosts, and null:
4031
1768
return revision_graph
4032
1769
4033
1770
4034
class StreamSink(object):
4035
"""An object that can insert a stream into a repository.
4036
4037
This interface handles the complexity of reserialising inventories and
4038
revisions from different formats, and allows unidirectional insertion into
4039
stacked repositories without looking for the missing basis parents
4040
beforehand.
4041
"""
4042
4043
def __init__(self, target_repo):
4044
self.target_repo = target_repo
4045
4046
def insert_stream(self, stream, src_format, resume_tokens):
4047
"""Insert a stream's content into the target repository.
4048
4049
:param src_format: a bzr repository format.
4050
4051
:return: a list of resume tokens and an iterable of keys additional
4052
items required before the insertion can be completed.
4053
"""
4054
self.target_repo.lock_write()
4055
try:
4056
if resume_tokens:
4057
self.target_repo.resume_write_group(resume_tokens)
4058
is_resume = True
4059
else:
4060
self.target_repo.start_write_group()
4061
is_resume = False
4062
try:
4063
# locked_insert_stream performs a commit|suspend.
4064
missing_keys = self.insert_stream_without_locking(stream,
4065
src_format, is_resume)
4066
if missing_keys:
4067
# suspend the write group and tell the caller what we is
4068
# missing. We know we can suspend or else we would not have
4069
# entered this code path. (All repositories that can handle
4070
# missing keys can handle suspending a write group).
4071
write_group_tokens = self.target_repo.suspend_write_group()
4072
return write_group_tokens, missing_keys
4073
hint = self.target_repo.commit_write_group()
4074
to_serializer = self.target_repo._format._serializer
4075
src_serializer = src_format._serializer
4076
if (to_serializer != src_serializer and
4077
self.target_repo._format.pack_compresses):
4078
self.target_repo.pack(hint=hint)
4079
return [], set()
4080
except:
4081
self.target_repo.abort_write_group(suppress_errors=True)
4082
raise
4083
finally:
4084
self.target_repo.unlock()
4085
4086
def insert_stream_without_locking(self, stream, src_format,
4087
is_resume=False):
4088
"""Insert a stream's content into the target repository.
4089
4090
This assumes that you already have a locked repository and an active
4091
write group.
4092
4093
:param src_format: a bzr repository format.
4094
:param is_resume: Passed down to get_missing_parent_inventories to
4095
indicate if we should be checking for missing texts at the same
4096
time.
4097
4098
:return: A set of keys that are missing.
4099
"""
4100
if not self.target_repo.is_write_locked():
4101
raise errors.ObjectNotLocked(self)
4102
if not self.target_repo.is_in_write_group():
4103
raise errors.BzrError('you must already be in a write group')
4104
to_serializer = self.target_repo._format._serializer
4105
src_serializer = src_format._serializer
4106
new_pack = None
4107
if to_serializer == src_serializer:
4108
# If serializers match and the target is a pack repository, set the
4109
# write cache size on the new pack. This avoids poor performance
4110
# on transports where append is unbuffered (such as
4111
# RemoteTransport). This is safe to do because nothing should read
4112
# back from the target repository while a stream with matching
4113
# serialization is being inserted.
4114
# The exception is that a delta record from the source that should
4115
# be a fulltext may need to be expanded by the target (see
4116
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4117
# explicitly flush any buffered writes first in that rare case.
4118
try:
4119
new_pack = self.target_repo._pack_collection._new_pack
4120
except AttributeError:
4121
# Not a pack repository
4122
pass
4123
else:
4124
new_pack.set_write_cache_size(1024*1024)
4125
for substream_type, substream in stream:
4126
if 'stream' in debug.debug_flags:
4127
mutter('inserting substream: %s', substream_type)
4128
if substream_type == 'texts':
4129
self.target_repo.texts.insert_record_stream(substream)
4130
elif substream_type == 'inventories':
4131
if src_serializer == to_serializer:
4132
self.target_repo.inventories.insert_record_stream(
4133
substream)
4134
else:
4135
self._extract_and_insert_inventories(
4136
substream, src_serializer)
4137
elif substream_type == 'inventory-deltas':
4138
self._extract_and_insert_inventory_deltas(
4139
substream, src_serializer)
4140
elif substream_type == 'chk_bytes':
4141
# XXX: This doesn't support conversions, as it assumes the
4142
# conversion was done in the fetch code.
4143
self.target_repo.chk_bytes.insert_record_stream(substream)
4144
elif substream_type == 'revisions':
4145
# This may fallback to extract-and-insert more often than
4146
# required if the serializers are different only in terms of
4147
# the inventory.
4148
if src_serializer == to_serializer:
4149
self.target_repo.revisions.insert_record_stream(substream)
4150
else:
4151
self._extract_and_insert_revisions(substream,
4152
src_serializer)
4153
elif substream_type == 'signatures':
4154
self.target_repo.signatures.insert_record_stream(substream)
4155
else:
4156
raise AssertionError('kaboom! %s' % (substream_type,))
4157
# Done inserting data, and the missing_keys calculations will try to
4158
# read back from the inserted data, so flush the writes to the new pack
4159
# (if this is pack format).
4160
if new_pack is not None:
4161
new_pack._write_data('', flush=True)
4162
# Find all the new revisions (including ones from resume_tokens)
4163
missing_keys = self.target_repo.get_missing_parent_inventories(
4164
check_for_missing_texts=is_resume)
4165
try:
4166
for prefix, versioned_file in (
4167
('texts', self.target_repo.texts),
4168
('inventories', self.target_repo.inventories),
4169
('revisions', self.target_repo.revisions),
4170
('signatures', self.target_repo.signatures),
4171
('chk_bytes', self.target_repo.chk_bytes),
4172
):
4173
if versioned_file is None:
4174
continue
4175
# TODO: key is often going to be a StaticTuple object
4176
# I don't believe we can define a method by which
4177
# (prefix,) + StaticTuple will work, though we could
4178
# define a StaticTuple.sq_concat that would allow you to
4179
# pass in either a tuple or a StaticTuple as the second
4180
# object, so instead we could have:
4181
# StaticTuple(prefix) + key here...
4182
missing_keys.update((prefix,) + key for key in
4183
versioned_file.get_missing_compression_parent_keys())
4184
except NotImplementedError:
4185
# cannot even attempt suspending, and missing would have failed
4186
# during stream insertion.
4187
missing_keys = set()
4188
return missing_keys
4189
4190
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4191
target_rich_root = self.target_repo._format.rich_root_data
4192
target_tree_refs = self.target_repo._format.supports_tree_reference
4193
for record in substream:
4194
# Insert the delta directly
4195
inventory_delta_bytes = record.get_bytes_as('fulltext')
4196
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4197
try:
4198
parse_result = deserialiser.parse_text_bytes(
4199
inventory_delta_bytes)
4200
except inventory_delta.IncompatibleInventoryDelta, err:
4201
mutter("Incompatible delta: %s", err.msg)
4202
raise errors.IncompatibleRevision(self.target_repo._format)
4203
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4204
revision_id = new_id
4205
parents = [key[0] for key in record.parents]
4206
self.target_repo.add_inventory_by_delta(
4207
basis_id, inv_delta, revision_id, parents)
4208
4209
def _extract_and_insert_inventories(self, substream, serializer,
4210
parse_delta=None):
4211
"""Generate a new inventory versionedfile in target, converting data.
4212
4213
The inventory is retrieved from the source, (deserializing it), and
4214
stored in the target (reserializing it in a different format).
4215
"""
4216
target_rich_root = self.target_repo._format.rich_root_data
4217
target_tree_refs = self.target_repo._format.supports_tree_reference
4218
for record in substream:
4219
# It's not a delta, so it must be a fulltext in the source
4220
# serializer's format.
4221
bytes = record.get_bytes_as('fulltext')
4222
revision_id = record.key[0]
4223
inv = serializer.read_inventory_from_string(bytes, revision_id)
4224
parents = [key[0] for key in record.parents]
4225
self.target_repo.add_inventory(revision_id, inv, parents)
4226
# No need to keep holding this full inv in memory when the rest of
4227
# the substream is likely to be all deltas.
4228
del inv
4229
4230
def _extract_and_insert_revisions(self, substream, serializer):
4231
for record in substream:
4232
bytes = record.get_bytes_as('fulltext')
4233
revision_id = record.key[0]
4234
rev = serializer.read_revision_from_string(bytes)
4235
if rev.revision_id != revision_id:
4236
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4237
self.target_repo.add_revision(revision_id, rev)
4238
4239
def finished(self):
4240
if self.target_repo._format._fetch_reconcile:
4241
self.target_repo.reconcile()
4242
4243
4244
class StreamSource(object):
4245
"""A source of a stream for fetching between repositories."""
4246
4247
def __init__(self, from_repository, to_format):
4248
"""Create a StreamSource streaming from from_repository."""
4249
self.from_repository = from_repository
4250
self.to_format = to_format
4251
self._record_counter = RecordCounter()
4252
4253
def delta_on_metadata(self):
4254
"""Return True if delta's are permitted on metadata streams.
4255
4256
That is on revisions and signatures.
4257
"""
4258
src_serializer = self.from_repository._format._serializer
4259
target_serializer = self.to_format._serializer
4260
return (self.to_format._fetch_uses_deltas and
4261
src_serializer == target_serializer)
4262
4263
def _fetch_revision_texts(self, revs):
4264
# fetch signatures first and then the revision texts
4265
# may need to be a InterRevisionStore call here.
4266
from_sf = self.from_repository.signatures
4267
# A missing signature is just skipped.
4268
keys = [(rev_id,) for rev_id in revs]
4269
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4270
keys,
4271
self.to_format._fetch_order,
4272
not self.to_format._fetch_uses_deltas))
4273
# If a revision has a delta, this is actually expanded inside the
4274
# insert_record_stream code now, which is an alternate fix for
4275
# bug #261339
4276
from_rf = self.from_repository.revisions
4277
revisions = from_rf.get_record_stream(
4278
keys,
4279
self.to_format._fetch_order,
4280
not self.delta_on_metadata())
4281
return [('signatures', signatures), ('revisions', revisions)]
4282
4283
def _generate_root_texts(self, revs):
4284
"""This will be called by get_stream between fetching weave texts and
4285
fetching the inventory weave.
4286
"""
4287
if self._rich_root_upgrade():
4288
return _mod_fetch.Inter1and2Helper(
4289
self.from_repository).generate_root_texts(revs)
4290
else:
4291
return []
4292
4293
def get_stream(self, search):
4294
phase = 'file'
4295
revs = search.get_keys()
4296
graph = self.from_repository.get_graph()
4297
revs = tsort.topo_sort(graph.get_parent_map(revs))
4298
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4299
text_keys = []
4300
for knit_kind, file_id, revisions in data_to_fetch:
4301
if knit_kind != phase:
4302
phase = knit_kind
4303
# Make a new progress bar for this phase
4304
if knit_kind == "file":
4305
# Accumulate file texts
4306
text_keys.extend([(file_id, revision) for revision in
4307
revisions])
4308
elif knit_kind == "inventory":
4309
# Now copy the file texts.
4310
from_texts = self.from_repository.texts
4311
yield ('texts', from_texts.get_record_stream(
4312
text_keys, self.to_format._fetch_order,
4313
not self.to_format._fetch_uses_deltas))
4314
# Cause an error if a text occurs after we have done the
4315
# copy.
4316
text_keys = None
4317
# Before we process the inventory we generate the root
4318
# texts (if necessary) so that the inventories references
4319
# will be valid.
4320
for _ in self._generate_root_texts(revs):
4321
yield _
4322
# we fetch only the referenced inventories because we do not
4323
# know for unselected inventories whether all their required
4324
# texts are present in the other repository - it could be
4325
# corrupt.
4326
for info in self._get_inventory_stream(revs):
4327
yield info
4328
elif knit_kind == "signatures":
4329
# Nothing to do here; this will be taken care of when
4330
# _fetch_revision_texts happens.
4331
pass
4332
elif knit_kind == "revisions":
4333
for record in self._fetch_revision_texts(revs):
4334
yield record
4335
else:
4336
raise AssertionError("Unknown knit kind %r" % knit_kind)
4337
4338
def get_stream_for_missing_keys(self, missing_keys):
4339
# missing keys can only occur when we are byte copying and not
4340
# translating (because translation means we don't send
4341
# unreconstructable deltas ever).
4342
keys = {}
4343
keys['texts'] = set()
4344
keys['revisions'] = set()
4345
keys['inventories'] = set()
4346
keys['chk_bytes'] = set()
4347
keys['signatures'] = set()
4348
for key in missing_keys:
4349
keys[key[0]].add(key[1:])
4350
if len(keys['revisions']):
4351
# If we allowed copying revisions at this point, we could end up
4352
# copying a revision without copying its required texts: a
4353
# violation of the requirements for repository integrity.
4354
raise AssertionError(
4355
'cannot copy revisions to fill in missing deltas %s' % (
4356
keys['revisions'],))
4357
for substream_kind, keys in keys.iteritems():
4358
vf = getattr(self.from_repository, substream_kind)
4359
if vf is None and keys:
4360
raise AssertionError(
4361
"cannot fill in keys for a versioned file we don't"
4362
" have: %s needs %s" % (substream_kind, keys))
4363
if not keys:
4364
# No need to stream something we don't have
4365
continue
4366
if substream_kind == 'inventories':
4367
# Some missing keys are genuinely ghosts, filter those out.
4368
present = self.from_repository.inventories.get_parent_map(keys)
4369
revs = [key[0] for key in present]
4370
# Get the inventory stream more-or-less as we do for the
4371
# original stream; there's no reason to assume that records
4372
# direct from the source will be suitable for the sink. (Think
4373
# e.g. 2a -> 1.9-rich-root).
4374
for info in self._get_inventory_stream(revs, missing=True):
4375
yield info
4376
continue
4377
4378
# Ask for full texts always so that we don't need more round trips
4379
# after this stream.
4380
# Some of the missing keys are genuinely ghosts, so filter absent
4381
# records. The Sink is responsible for doing another check to
4382
# ensure that ghosts don't introduce missing data for future
4383
# fetches.
4384
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4385
self.to_format._fetch_order, True))
4386
yield substream_kind, stream
4387
4388
def inventory_fetch_order(self):
4389
if self._rich_root_upgrade():
4390
return 'topological'
4391
else:
4392
return self.to_format._fetch_order
4393
4394
def _rich_root_upgrade(self):
4395
return (not self.from_repository._format.rich_root_data and
4396
self.to_format.rich_root_data)
4397
4398
def _get_inventory_stream(self, revision_ids, missing=False):
4399
from_format = self.from_repository._format
4400
if (from_format.supports_chks and self.to_format.supports_chks and
4401
from_format.network_name() == self.to_format.network_name()):
4402
raise AssertionError(
4403
"this case should be handled by GroupCHKStreamSource")
4404
elif 'forceinvdeltas' in debug.debug_flags:
4405
return self._get_convertable_inventory_stream(revision_ids,
4406
delta_versus_null=missing)
4407
elif from_format.network_name() == self.to_format.network_name():
4408
# Same format.
4409
return self._get_simple_inventory_stream(revision_ids,
4410
missing=missing)
4411
elif (not from_format.supports_chks and not self.to_format.supports_chks
4412
and from_format._serializer == self.to_format._serializer):
4413
# Essentially the same format.
4414
return self._get_simple_inventory_stream(revision_ids,
4415
missing=missing)
4416
else:
4417
# Any time we switch serializations, we want to use an
4418
# inventory-delta based approach.
4419
return self._get_convertable_inventory_stream(revision_ids,
4420
delta_versus_null=missing)
4421
4422
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4423
# NB: This currently reopens the inventory weave in source;
4424
# using a single stream interface instead would avoid this.
4425
from_weave = self.from_repository.inventories
4426
if missing:
4427
delta_closure = True
4428
else:
4429
delta_closure = not self.delta_on_metadata()
4430
yield ('inventories', from_weave.get_record_stream(
4431
[(rev_id,) for rev_id in revision_ids],
4432
self.inventory_fetch_order(), delta_closure))
4433
4434
def _get_convertable_inventory_stream(self, revision_ids,
4435
delta_versus_null=False):
4436
# The two formats are sufficiently different that there is no fast
4437
# path, so we need to send just inventorydeltas, which any
4438
# sufficiently modern client can insert into any repository.
4439
# The StreamSink code expects to be able to
4440
# convert on the target, so we need to put bytes-on-the-wire that can
4441
# be converted. That means inventory deltas (if the remote is <1.19,
4442
# RemoteStreamSink will fallback to VFS to insert the deltas).
4443
yield ('inventory-deltas',
4444
self._stream_invs_as_deltas(revision_ids,
4445
delta_versus_null=delta_versus_null))
4446
4447
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4448
"""Return a stream of inventory-deltas for the given rev ids.
4449
4450
:param revision_ids: The list of inventories to transmit
4451
:param delta_versus_null: Don't try to find a minimal delta for this
4452
entry, instead compute the delta versus the NULL_REVISION. This
4453
effectively streams a complete inventory. Used for stuff like
4454
filling in missing parents, etc.
4455
"""
4456
from_repo = self.from_repository
4457
revision_keys = [(rev_id,) for rev_id in revision_ids]
4458
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4459
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4460
# method...
4461
inventories = self.from_repository.iter_inventories(
4462
revision_ids, 'topological')
4463
format = from_repo._format
4464
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4465
inventory_cache = lru_cache.LRUCache(50)
4466
null_inventory = from_repo.revision_tree(
4467
_mod_revision.NULL_REVISION).inventory
4468
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4469
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4470
# repo back into a non-rich-root repo ought to be allowed)
4471
serializer = inventory_delta.InventoryDeltaSerializer(
4472
versioned_root=format.rich_root_data,
4473
tree_references=format.supports_tree_reference)
4474
for inv in inventories:
4475
key = (inv.revision_id,)
4476
parent_keys = parent_map.get(key, ())
4477
delta = None
4478
if not delta_versus_null and parent_keys:
4479
# The caller did not ask for complete inventories and we have
4480
# some parents that we can delta against. Make a delta against
4481
# each parent so that we can find the smallest.
4482
parent_ids = [parent_key[0] for parent_key in parent_keys]
4483
for parent_id in parent_ids:
4484
if parent_id not in invs_sent_so_far:
4485
# We don't know that the remote side has this basis, so
4486
# we can't use it.
4487
continue
4488
if parent_id == _mod_revision.NULL_REVISION:
4489
parent_inv = null_inventory
4490
else:
4491
parent_inv = inventory_cache.get(parent_id, None)
4492
if parent_inv is None:
4493
parent_inv = from_repo.get_inventory(parent_id)
4494
candidate_delta = inv._make_delta(parent_inv)
4495
if (delta is None or
4496
len(delta) > len(candidate_delta)):
4497
delta = candidate_delta
4498
basis_id = parent_id
4499
if delta is None:
4500
# Either none of the parents ended up being suitable, or we
4501
# were asked to delta against NULL
4502
basis_id = _mod_revision.NULL_REVISION
4503
delta = inv._make_delta(null_inventory)
4504
invs_sent_so_far.add(inv.revision_id)
4505
inventory_cache[inv.revision_id] = inv
4506
delta_serialized = ''.join(
4507
serializer.delta_to_lines(basis_id, key[-1], delta))
4508
yield versionedfile.FulltextContentFactory(
4509
key, parent_keys, None, delta_serialized)
4510
4511
4512
1771
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4513
1772
stop_revision=None):
4514
1773
"""Extend the partial history to include a given index
4524
1783
it is encountered, history extension will stop.
4525
1784
"""
4526
1785
start_revision = partial_history_cache[-1]
4527
iterator = repo.iter_reverse_revision_history(start_revision)
1786
graph = repo.get_graph()
1787
iterator = graph.iter_lefthand_ancestry(start_revision,
1788
(_mod_revision.NULL_REVISION,))
4528
1789
try:
4529
#skip the last revision in the list
1790
# skip the last revision in the list
4530
1791
iterator.next()
4531
1792
while True:
4532
1793
if (stop_index is not None and
4539
1800
except StopIteration:
4540
1801
# No more history
4541
1802
return
1803
1804
1805
class _LazyListJoin(object):
1806
"""An iterable yielding the contents of many lists as one list.
1807
1808
Each iterator made from this will reflect the current contents of the lists
1809
at the time the iterator is made.
1810
1811
This is used by Repository's _make_parents_provider implementation so that
1812
it is safe to do::
1813
1814
pp = repo._make_parents_provider() # uses a list of fallback repos
1815
pp.add_fallback_repository(other_repo) # appends to that list
1816
result = pp.get_parent_map(...)
1817
# The result will include revs from other_repo
1818
"""
1819
1820
def __init__(self, *list_parts):
1821
self.list_parts = list_parts
1822
1823
def __iter__(self):
1824
full_list = []
1825
for list_part in self.list_parts:
1826
full_list.extend(list_part)
1827
return iter(full_list)
1828
1829
def __repr__(self):
1830
return "%s.%s(%s)" % (self.__module__, self.__class__.__name__,
1831
self.list_parts)
Loggerhead is a web-based interface for Breezy
Version: 2.0.1