« back to all changes in this revision
Viewing changes to bzrlib/repository.py
- browse files at revision 2598.1.3
- compare with another revision
- download diff
- download tarball
- view history from revision 2598.1.3
- Committer: Martin Pool
- Date: 2007-07-11 01:55:33 UTC
- mto: This revision was merged to the branch mainline in revision 2599.
- Revision ID: mbp@sourcefrog.net-20070711015533-dzcxkjg0ujh8yuhl
Option help improvements (thanks jamesw)
- files added:
- bzrlib/bundle/common.py
- bzrlib/bundle/old
- bzrlib/tests/interversionedfile_implementations
- bzrlib/tests/revisionstore_implementations
- bzrlib/util/configobj/docs
- files removed:
- bzrlib/_annotator_py.py
- bzrlib/filters
- bzrlib/help_topics
- bzrlib/help_topics/en
- bzrlib/plugins/netrc_credential_store
- bzrlib/plugins/netrc_credential_store/tests
- bzrlib/tests/commands
- bzrlib/tests/ftp_server
- bzrlib/tests/inventory_implementations
- bzrlib/tests/per_interbranch
- bzrlib/tests/per_repository_chk
- bzrlib/tests/per_repository_reference
- bzrlib/tests/ssl_certs
- bzrlib/transport/ftp
- bzrlib/util/tests
- doc/en
- doc/en/admin-guide
- doc/en/developer-guide
- doc/en/mini-tutorial
- doc/en/quick-reference
- doc/en/release-notes
- doc/en/tutorials
- doc/en/user-guide
- doc/en/user-guide/images
- doc/en/user-reference
- doc/es
- doc/es/developer-guide
- doc/es/mini-tutorial
- doc/es/quick-reference
- doc/es/release-notes
- doc/es/user-guide
- doc/es/user-reference
- doc/ru
- doc/ru/mini-tutorial
- doc/ru/quick-reference
- doc/ru/tutorials
- doc/ru/user-guide
- doc/ru/user-guide/images
- tools/packaging
- files renamed:
- bzrlib/help_topics/__init__.py => bzrlib/help_topics.py
- bzrlib/_patiencediff_py.py => bzrlib/patiencediff.py
- bzrlib/plugins/launchpad/lp_directory.py => bzrlib/plugins/launchpad/lp_indirect.py
- bzrlib/plugins/launchpad/test_lp_directory.py => bzrlib/plugins/launchpad/test_lp_indirect.py
- bzrlib/tests/http_utils.py => bzrlib/tests/HTTPTestUtil.py
- bzrlib/tests/http_server.py => bzrlib/tests/HttpServer.py
- bzrlib/tests/blackbox/test_send.py => bzrlib/tests/blackbox/test_bundle.py
- bzrlib/tests/per_branch/ => bzrlib/tests/branch_implementations/
- bzrlib/tests/per_bzrdir/ => bzrlib/tests/bzrdir_implementations/
- bzrlib/tests/per_interrepository/ => bzrlib/tests/interrepository_implementations/
- bzrlib/tests/per_repository/ => bzrlib/tests/repository_implementations/
- bzrlib/tests/test_revisionspec.py => bzrlib/tests/test_revisionnamespaces.py
- bzrlib/transport/ftp/__init__.py => bzrlib/transport/ftp.py
- bzrlib/util/_bencode_py.py => bzrlib/util/bencode.py
- bzrlib/xml8.py => bzrlib/xml5.py
- doc/en/tutorials/centralized_workflow.txt => doc/centralized_workflow.txt
- bzrlib/help_topics/en/configuration.txt => doc/configuration.txt
- doc/en/developer-guide/HACKING.txt => doc/developers/HACKING
- doc/en/user-guide/http_smart_server.txt => doc/http_smart_server.txt
- doc/en/user-guide/index.txt => doc/index.txt
- doc/en/user-guide/plugins.txt => doc/plugins.txt
- doc/en/user-guide/server.txt => doc/server.txt
- doc/en/user-guide/setting_up_email.txt => doc/setting_up_email.txt
- doc/en/user-guide/shared_repository_layouts.txt => doc/shared_repository_layouts.txt
- doc/en/user-guide/specifying_revisions.txt => doc/specifying_revisions.txt
- doc/en/tutorials/tutorial.txt => doc/tutorial.txt
- doc/en/user-guide/using_aliases.txt => doc/using_aliases.txt
- doc/en/user-guide/version_info.txt => doc/version_info.txt
- tools/generate_docs.py => generate_docs.py
- bzrlib/doc_generate/ => tools/doc_generate/
- files modified:
- .bzrignore
added
removed
bzrlib/repository.py
1
# Copyright (C) 2005, 2006, 2007, 2008, 2009 Canonical Ltd
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
16
17
from cStringIO import StringIO
16
18
17
19
from bzrlib.lazy_import import lazy_import
18
20
lazy_import(globals(), """
19
import cStringIO
20
21
import re
21
22
import time
22
23
23
24
from bzrlib import (
24
25
bzrdir,
25
26
check,
26
chk_map,
27
debug,
27
deprecated_graph,
28
28
errors,
29
fifo_cache,
30
29
generate_ids,
31
30
gpg,
32
31
graph,
33
inventory,
34
32
lazy_regex,
35
33
lockable_files,
36
34
lockdir,
37
lru_cache,
38
35
osutils,
36
registry,
37
remote,
39
38
revision as _mod_revision,
40
39
symbol_versioning,
41
tsort,
40
transactions,
42
41
ui,
43
versionedfile,
44
42
)
45
from bzrlib.bundle import serializer
46
43
from bzrlib.revisiontree import RevisionTree
47
44
from bzrlib.store.versioned import VersionedFileStore
45
from bzrlib.store.text import TextStore
48
46
from bzrlib.testament import Testament
47
49
48
""")
50
49
51
50
from bzrlib.decorators import needs_read_lock, needs_write_lock
52
51
from bzrlib.inter import InterObject
53
from bzrlib.inventory import (
54
Inventory,
55
InventoryDirectory,
56
ROOT_ID,
57
entry_factory,
58
)
59
from bzrlib import registry
60
from bzrlib.trace import (
61
log_exception_quietly, note, mutter, mutter_callsite, warning)
52
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
53
from bzrlib.symbol_versioning import (
54
deprecated_method,
55
zero_nine,
56
)
57
from bzrlib.trace import mutter, note, warning
62
58
63
59
64
60
# Old formats display a warning, but only once
65
61
_deprecation_warning_done = False
66
62
67
63
68
class CommitBuilder(object):
69
"""Provides an interface to build up a commit.
70
71
This allows describing a tree to be committed without needing to
72
know the internals of the format of the repository.
73
"""
74
75
# all clients should supply tree roots.
76
record_root_entry = True
77
# the default CommitBuilder does not manage trees whose root is versioned.
78
_versioned_root = False
79
80
def __init__(self, repository, parents, config, timestamp=None,
81
timezone=None, committer=None, revprops=None,
82
revision_id=None):
83
"""Initiate a CommitBuilder.
84
85
:param repository: Repository to commit to.
86
:param parents: Revision ids of the parents of the new revision.
87
:param config: Configuration to use.
88
:param timestamp: Optional timestamp recorded for commit.
89
:param timezone: Optional timezone for timestamp.
90
:param committer: Optional committer to set for commit.
91
:param revprops: Optional dictionary of revision properties.
92
:param revision_id: Optional revision id.
93
"""
94
self._config = config
95
96
if committer is None:
97
self._committer = self._config.username()
98
else:
99
self._committer = committer
100
101
self.new_inventory = Inventory(None)
102
self._new_revision_id = revision_id
103
self.parents = parents
104
self.repository = repository
105
106
self._revprops = {}
107
if revprops is not None:
108
self._validate_revprops(revprops)
109
self._revprops.update(revprops)
110
111
if timestamp is None:
112
timestamp = time.time()
113
# Restrict resolution to 1ms
114
self._timestamp = round(timestamp, 3)
115
116
if timezone is None:
117
self._timezone = osutils.local_time_offset()
118
else:
119
self._timezone = int(timezone)
120
121
self._generate_revision_if_needed()
122
self.__heads = graph.HeadsCache(repository.get_graph()).heads
123
self._basis_delta = []
124
# API compatibility, older code that used CommitBuilder did not call
125
# .record_delete(), which means the delta that is computed would not be
126
# valid. Callers that will call record_delete() should call
127
# .will_record_deletes() to indicate that.
128
self._recording_deletes = False
129
# memo'd check for no-op commits.
130
self._any_changes = False
131
132
def any_changes(self):
133
"""Return True if any entries were changed.
134
135
This includes merge-only changes. It is the core for the --unchanged
136
detection in commit.
137
138
:return: True if any changes have occured.
139
"""
140
return self._any_changes
141
142
def _validate_unicode_text(self, text, context):
143
"""Verify things like commit messages don't have bogus characters."""
144
if '\r' in text:
145
raise ValueError('Invalid value for %s: %r' % (context, text))
146
147
def _validate_revprops(self, revprops):
148
for key, value in revprops.iteritems():
149
# We know that the XML serializers do not round trip '\r'
150
# correctly, so refuse to accept them
151
if not isinstance(value, basestring):
152
raise ValueError('revision property (%s) is not a valid'
153
' (unicode) string: %r' % (key, value))
154
self._validate_unicode_text(value,
155
'revision property (%s)' % (key,))
156
157
def commit(self, message):
158
"""Make the actual commit.
159
160
:return: The revision id of the recorded revision.
161
"""
162
self._validate_unicode_text(message, 'commit message')
163
rev = _mod_revision.Revision(
164
timestamp=self._timestamp,
165
timezone=self._timezone,
166
committer=self._committer,
167
message=message,
168
inventory_sha1=self.inv_sha1,
169
revision_id=self._new_revision_id,
170
properties=self._revprops)
171
rev.parent_ids = self.parents
172
self.repository.add_revision(self._new_revision_id, rev,
173
self.new_inventory, self._config)
174
self.repository.commit_write_group()
175
return self._new_revision_id
176
177
def abort(self):
178
"""Abort the commit that is being built.
179
"""
180
self.repository.abort_write_group()
181
182
def revision_tree(self):
183
"""Return the tree that was just committed.
184
185
After calling commit() this can be called to get a RevisionTree
186
representing the newly committed tree. This is preferred to
187
calling Repository.revision_tree() because that may require
188
deserializing the inventory, while we already have a copy in
189
memory.
190
"""
191
if self.new_inventory is None:
192
self.new_inventory = self.repository.get_inventory(
193
self._new_revision_id)
194
return RevisionTree(self.repository, self.new_inventory,
195
self._new_revision_id)
196
197
def finish_inventory(self):
198
"""Tell the builder that the inventory is finished.
199
200
:return: The inventory id in the repository, which can be used with
201
repository.get_inventory.
202
"""
203
if self.new_inventory is None:
204
# an inventory delta was accumulated without creating a new
205
# inventory.
206
basis_id = self.basis_delta_revision
207
self.inv_sha1 = self.repository.add_inventory_by_delta(
208
basis_id, self._basis_delta, self._new_revision_id,
209
self.parents)
210
else:
211
if self.new_inventory.root is None:
212
raise AssertionError('Root entry should be supplied to'
213
' record_entry_contents, as of bzr 0.10.')
214
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
215
self.new_inventory.revision_id = self._new_revision_id
216
self.inv_sha1 = self.repository.add_inventory(
217
self._new_revision_id,
218
self.new_inventory,
219
self.parents
220
)
221
return self._new_revision_id
222
223
def _gen_revision_id(self):
224
"""Return new revision-id."""
225
return generate_ids.gen_revision_id(self._config.username(),
226
self._timestamp)
227
228
def _generate_revision_if_needed(self):
229
"""Create a revision id if None was supplied.
230
231
If the repository can not support user-specified revision ids
232
they should override this function and raise CannotSetRevisionId
233
if _new_revision_id is not None.
234
235
:raises: CannotSetRevisionId
236
"""
237
if self._new_revision_id is None:
238
self._new_revision_id = self._gen_revision_id()
239
self.random_revid = True
240
else:
241
self.random_revid = False
242
243
def _heads(self, file_id, revision_ids):
244
"""Calculate the graph heads for revision_ids in the graph of file_id.
245
246
This can use either a per-file graph or a global revision graph as we
247
have an identity relationship between the two graphs.
248
"""
249
return self.__heads(revision_ids)
250
251
def _check_root(self, ie, parent_invs, tree):
252
"""Helper for record_entry_contents.
253
254
:param ie: An entry being added.
255
:param parent_invs: The inventories of the parent revisions of the
256
commit.
257
:param tree: The tree that is being committed.
258
"""
259
# In this revision format, root entries have no knit or weave When
260
# serializing out to disk and back in root.revision is always
261
# _new_revision_id
262
ie.revision = self._new_revision_id
263
264
def _require_root_change(self, tree):
265
"""Enforce an appropriate root object change.
266
267
This is called once when record_iter_changes is called, if and only if
268
the root was not in the delta calculated by record_iter_changes.
269
270
:param tree: The tree which is being committed.
271
"""
272
# NB: if there are no parents then this method is not called, so no
273
# need to guard on parents having length.
274
entry = entry_factory['directory'](tree.path2id(''), '',
275
None)
276
entry.revision = self._new_revision_id
277
self._basis_delta.append(('', '', entry.file_id, entry))
278
279
def _get_delta(self, ie, basis_inv, path):
280
"""Get a delta against the basis inventory for ie."""
281
if ie.file_id not in basis_inv:
282
# add
283
result = (None, path, ie.file_id, ie)
284
self._basis_delta.append(result)
285
return result
286
elif ie != basis_inv[ie.file_id]:
287
# common but altered
288
# TODO: avoid tis id2path call.
289
result = (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
290
self._basis_delta.append(result)
291
return result
292
else:
293
# common, unaltered
294
return None
295
296
def get_basis_delta(self):
297
"""Return the complete inventory delta versus the basis inventory.
298
299
This has been built up with the calls to record_delete and
300
record_entry_contents. The client must have already called
301
will_record_deletes() to indicate that they will be generating a
302
complete delta.
303
304
:return: An inventory delta, suitable for use with apply_delta, or
305
Repository.add_inventory_by_delta, etc.
306
"""
307
if not self._recording_deletes:
308
raise AssertionError("recording deletes not activated.")
309
return self._basis_delta
310
311
def record_delete(self, path, file_id):
312
"""Record that a delete occured against a basis tree.
313
314
This is an optional API - when used it adds items to the basis_delta
315
being accumulated by the commit builder. It cannot be called unless the
316
method will_record_deletes() has been called to inform the builder that
317
a delta is being supplied.
318
319
:param path: The path of the thing deleted.
320
:param file_id: The file id that was deleted.
321
"""
322
if not self._recording_deletes:
323
raise AssertionError("recording deletes not activated.")
324
delta = (path, None, file_id, None)
325
self._basis_delta.append(delta)
326
self._any_changes = True
327
return delta
328
329
def will_record_deletes(self):
330
"""Tell the commit builder that deletes are being notified.
331
332
This enables the accumulation of an inventory delta; for the resulting
333
commit to be valid, deletes against the basis MUST be recorded via
334
builder.record_delete().
335
"""
336
self._recording_deletes = True
337
try:
338
basis_id = self.parents[0]
339
except IndexError:
340
basis_id = _mod_revision.NULL_REVISION
341
self.basis_delta_revision = basis_id
342
343
def record_entry_contents(self, ie, parent_invs, path, tree,
344
content_summary):
345
"""Record the content of ie from tree into the commit if needed.
346
347
Side effect: sets ie.revision when unchanged
348
349
:param ie: An inventory entry present in the commit.
350
:param parent_invs: The inventories of the parent revisions of the
351
commit.
352
:param path: The path the entry is at in the tree.
353
:param tree: The tree which contains this entry and should be used to
354
obtain content.
355
:param content_summary: Summary data from the tree about the paths
356
content - stat, length, exec, sha/link target. This is only
357
accessed when the entry has a revision of None - that is when it is
358
a candidate to commit.
359
:return: A tuple (change_delta, version_recorded, fs_hash).
360
change_delta is an inventory_delta change for this entry against
361
the basis tree of the commit, or None if no change occured against
362
the basis tree.
363
version_recorded is True if a new version of the entry has been
364
recorded. For instance, committing a merge where a file was only
365
changed on the other side will return (delta, False).
366
fs_hash is either None, or the hash details for the path (currently
367
a tuple of the contents sha1 and the statvalue returned by
368
tree.get_file_with_stat()).
369
"""
370
if self.new_inventory.root is None:
371
if ie.parent_id is not None:
372
raise errors.RootMissing()
373
self._check_root(ie, parent_invs, tree)
374
if ie.revision is None:
375
kind = content_summary[0]
376
else:
377
# ie is carried over from a prior commit
378
kind = ie.kind
379
# XXX: repository specific check for nested tree support goes here - if
380
# the repo doesn't want nested trees we skip it ?
381
if (kind == 'tree-reference' and
382
not self.repository._format.supports_tree_reference):
383
# mismatch between commit builder logic and repository:
384
# this needs the entry creation pushed down into the builder.
385
raise NotImplementedError('Missing repository subtree support.')
386
self.new_inventory.add(ie)
387
388
# TODO: slow, take it out of the inner loop.
389
try:
390
basis_inv = parent_invs[0]
391
except IndexError:
392
basis_inv = Inventory(root_id=None)
393
394
# ie.revision is always None if the InventoryEntry is considered
395
# for committing. We may record the previous parents revision if the
396
# content is actually unchanged against a sole head.
397
if ie.revision is not None:
398
if not self._versioned_root and path == '':
399
# repositories that do not version the root set the root's
400
# revision to the new commit even when no change occurs (more
401
# specifically, they do not record a revision on the root; and
402
# the rev id is assigned to the root during deserialisation -
403
# this masks when a change may have occurred against the basis.
404
# To match this we always issue a delta, because the revision
405
# of the root will always be changing.
406
if ie.file_id in basis_inv:
407
delta = (basis_inv.id2path(ie.file_id), path,
408
ie.file_id, ie)
409
else:
410
# add
411
delta = (None, path, ie.file_id, ie)
412
self._basis_delta.append(delta)
413
return delta, False, None
414
else:
415
# we don't need to commit this, because the caller already
416
# determined that an existing revision of this file is
417
# appropriate. If its not being considered for committing then
418
# it and all its parents to the root must be unaltered so
419
# no-change against the basis.
420
if ie.revision == self._new_revision_id:
421
raise AssertionError("Impossible situation, a skipped "
422
"inventory entry (%r) claims to be modified in this "
423
"commit (%r).", (ie, self._new_revision_id))
424
return None, False, None
425
# XXX: Friction: parent_candidates should return a list not a dict
426
# so that we don't have to walk the inventories again.
427
parent_candiate_entries = ie.parent_candidates(parent_invs)
428
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
429
heads = []
430
for inv in parent_invs:
431
if ie.file_id in inv:
432
old_rev = inv[ie.file_id].revision
433
if old_rev in head_set:
434
heads.append(inv[ie.file_id].revision)
435
head_set.remove(inv[ie.file_id].revision)
436
437
store = False
438
# now we check to see if we need to write a new record to the
439
# file-graph.
440
# We write a new entry unless there is one head to the ancestors, and
441
# the kind-derived content is unchanged.
442
443
# Cheapest check first: no ancestors, or more the one head in the
444
# ancestors, we write a new node.
445
if len(heads) != 1:
446
store = True
447
if not store:
448
# There is a single head, look it up for comparison
449
parent_entry = parent_candiate_entries[heads[0]]
450
# if the non-content specific data has changed, we'll be writing a
451
# node:
452
if (parent_entry.parent_id != ie.parent_id or
453
parent_entry.name != ie.name):
454
store = True
455
# now we need to do content specific checks:
456
if not store:
457
# if the kind changed the content obviously has
458
if kind != parent_entry.kind:
459
store = True
460
# Stat cache fingerprint feedback for the caller - None as we usually
461
# don't generate one.
462
fingerprint = None
463
if kind == 'file':
464
if content_summary[2] is None:
465
raise ValueError("Files must not have executable = None")
466
if not store:
467
if (# if the file length changed we have to store:
468
parent_entry.text_size != content_summary[1] or
469
# if the exec bit has changed we have to store:
470
parent_entry.executable != content_summary[2]):
471
store = True
472
elif parent_entry.text_sha1 == content_summary[3]:
473
# all meta and content is unchanged (using a hash cache
474
# hit to check the sha)
475
ie.revision = parent_entry.revision
476
ie.text_size = parent_entry.text_size
477
ie.text_sha1 = parent_entry.text_sha1
478
ie.executable = parent_entry.executable
479
return self._get_delta(ie, basis_inv, path), False, None
480
else:
481
# Either there is only a hash change(no hash cache entry,
482
# or same size content change), or there is no change on
483
# this file at all.
484
# Provide the parent's hash to the store layer, so that the
485
# content is unchanged we will not store a new node.
486
nostore_sha = parent_entry.text_sha1
487
if store:
488
# We want to record a new node regardless of the presence or
489
# absence of a content change in the file.
490
nostore_sha = None
491
ie.executable = content_summary[2]
492
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
493
try:
494
text = file_obj.read()
495
finally:
496
file_obj.close()
497
try:
498
ie.text_sha1, ie.text_size = self._add_text_to_weave(
499
ie.file_id, text, heads, nostore_sha)
500
# Let the caller know we generated a stat fingerprint.
501
fingerprint = (ie.text_sha1, stat_value)
502
except errors.ExistingContent:
503
# Turns out that the file content was unchanged, and we were
504
# only going to store a new node if it was changed. Carry over
505
# the entry.
506
ie.revision = parent_entry.revision
507
ie.text_size = parent_entry.text_size
508
ie.text_sha1 = parent_entry.text_sha1
509
ie.executable = parent_entry.executable
510
return self._get_delta(ie, basis_inv, path), False, None
511
elif kind == 'directory':
512
if not store:
513
# all data is meta here, nothing specific to directory, so
514
# carry over:
515
ie.revision = parent_entry.revision
516
return self._get_delta(ie, basis_inv, path), False, None
517
self._add_text_to_weave(ie.file_id, '', heads, None)
518
elif kind == 'symlink':
519
current_link_target = content_summary[3]
520
if not store:
521
# symlink target is not generic metadata, check if it has
522
# changed.
523
if current_link_target != parent_entry.symlink_target:
524
store = True
525
if not store:
526
# unchanged, carry over.
527
ie.revision = parent_entry.revision
528
ie.symlink_target = parent_entry.symlink_target
529
return self._get_delta(ie, basis_inv, path), False, None
530
ie.symlink_target = current_link_target
531
self._add_text_to_weave(ie.file_id, '', heads, None)
532
elif kind == 'tree-reference':
533
if not store:
534
if content_summary[3] != parent_entry.reference_revision:
535
store = True
536
if not store:
537
# unchanged, carry over.
538
ie.reference_revision = parent_entry.reference_revision
539
ie.revision = parent_entry.revision
540
return self._get_delta(ie, basis_inv, path), False, None
541
ie.reference_revision = content_summary[3]
542
self._add_text_to_weave(ie.file_id, '', heads, None)
543
else:
544
raise NotImplementedError('unknown kind')
545
ie.revision = self._new_revision_id
546
self._any_changes = True
547
return self._get_delta(ie, basis_inv, path), True, fingerprint
548
549
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
550
_entry_factory=entry_factory):
551
"""Record a new tree via iter_changes.
552
553
:param tree: The tree to obtain text contents from for changed objects.
554
:param basis_revision_id: The revision id of the tree the iter_changes
555
has been generated against. Currently assumed to be the same
556
as self.parents[0] - if it is not, errors may occur.
557
:param iter_changes: An iter_changes iterator with the changes to apply
558
to basis_revision_id. The iterator must not include any items with
559
a current kind of None - missing items must be either filtered out
560
or errored-on beefore record_iter_changes sees the item.
561
:param _entry_factory: Private method to bind entry_factory locally for
562
performance.
563
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
564
tree._observed_sha1.
565
"""
566
# Create an inventory delta based on deltas between all the parents and
567
# deltas between all the parent inventories. We use inventory delta's
568
# between the inventory objects because iter_changes masks
569
# last-changed-field only changes.
570
# Working data:
571
# file_id -> change map, change is fileid, paths, changed, versioneds,
572
# parents, names, kinds, executables
573
merged_ids = {}
574
# {file_id -> revision_id -> inventory entry, for entries in parent
575
# trees that are not parents[0]
576
parent_entries = {}
577
ghost_basis = False
578
try:
579
revtrees = list(self.repository.revision_trees(self.parents))
580
except errors.NoSuchRevision:
581
# one or more ghosts, slow path.
582
revtrees = []
583
for revision_id in self.parents:
584
try:
585
revtrees.append(self.repository.revision_tree(revision_id))
586
except errors.NoSuchRevision:
587
if not revtrees:
588
basis_revision_id = _mod_revision.NULL_REVISION
589
ghost_basis = True
590
revtrees.append(self.repository.revision_tree(
591
_mod_revision.NULL_REVISION))
592
# The basis inventory from a repository
593
if revtrees:
594
basis_inv = revtrees[0].inventory
595
else:
596
basis_inv = self.repository.revision_tree(
597
_mod_revision.NULL_REVISION).inventory
598
if len(self.parents) > 0:
599
if basis_revision_id != self.parents[0] and not ghost_basis:
600
raise Exception(
601
"arbitrary basis parents not yet supported with merges")
602
for revtree in revtrees[1:]:
603
for change in revtree.inventory._make_delta(basis_inv):
604
if change[1] is None:
605
# Not present in this parent.
606
continue
607
if change[2] not in merged_ids:
608
if change[0] is not None:
609
basis_entry = basis_inv[change[2]]
610
merged_ids[change[2]] = [
611
# basis revid
612
basis_entry.revision,
613
# new tree revid
614
change[3].revision]
615
parent_entries[change[2]] = {
616
# basis parent
617
basis_entry.revision:basis_entry,
618
# this parent
619
change[3].revision:change[3],
620
}
621
else:
622
merged_ids[change[2]] = [change[3].revision]
623
parent_entries[change[2]] = {change[3].revision:change[3]}
624
else:
625
merged_ids[change[2]].append(change[3].revision)
626
parent_entries[change[2]][change[3].revision] = change[3]
627
else:
628
merged_ids = {}
629
# Setup the changes from the tree:
630
# changes maps file_id -> (change, [parent revision_ids])
631
changes= {}
632
for change in iter_changes:
633
# This probably looks up in basis_inv way to much.
634
if change[1][0] is not None:
635
head_candidate = [basis_inv[change[0]].revision]
636
else:
637
head_candidate = []
638
changes[change[0]] = change, merged_ids.get(change[0],
639
head_candidate)
640
unchanged_merged = set(merged_ids) - set(changes)
641
# Extend the changes dict with synthetic changes to record merges of
642
# texts.
643
for file_id in unchanged_merged:
644
# Record a merged version of these items that did not change vs the
645
# basis. This can be either identical parallel changes, or a revert
646
# of a specific file after a merge. The recorded content will be
647
# that of the current tree (which is the same as the basis), but
648
# the per-file graph will reflect a merge.
649
# NB:XXX: We are reconstructing path information we had, this
650
# should be preserved instead.
651
# inv delta change: (file_id, (path_in_source, path_in_target),
652
# changed_content, versioned, parent, name, kind,
653
# executable)
654
try:
655
basis_entry = basis_inv[file_id]
656
except errors.NoSuchId:
657
# a change from basis->some_parents but file_id isn't in basis
658
# so was new in the merge, which means it must have changed
659
# from basis -> current, and as it hasn't the add was reverted
660
# by the user. So we discard this change.
661
pass
662
else:
663
change = (file_id,
664
(basis_inv.id2path(file_id), tree.id2path(file_id)),
665
False, (True, True),
666
(basis_entry.parent_id, basis_entry.parent_id),
667
(basis_entry.name, basis_entry.name),
668
(basis_entry.kind, basis_entry.kind),
669
(basis_entry.executable, basis_entry.executable))
670
changes[file_id] = (change, merged_ids[file_id])
671
# changes contains tuples with the change and a set of inventory
672
# candidates for the file.
673
# inv delta is:
674
# old_path, new_path, file_id, new_inventory_entry
675
seen_root = False # Is the root in the basis delta?
676
inv_delta = self._basis_delta
677
modified_rev = self._new_revision_id
678
for change, head_candidates in changes.values():
679
if change[3][1]: # versioned in target.
680
# Several things may be happening here:
681
# We may have a fork in the per-file graph
682
# - record a change with the content from tree
683
# We may have a change against < all trees
684
# - carry over the tree that hasn't changed
685
# We may have a change against all trees
686
# - record the change with the content from tree
687
kind = change[6][1]
688
file_id = change[0]
689
entry = _entry_factory[kind](file_id, change[5][1],
690
change[4][1])
691
head_set = self._heads(change[0], set(head_candidates))
692
heads = []
693
# Preserve ordering.
694
for head_candidate in head_candidates:
695
if head_candidate in head_set:
696
heads.append(head_candidate)
697
head_set.remove(head_candidate)
698
carried_over = False
699
if len(heads) == 1:
700
# Could be a carry-over situation:
701
parent_entry_revs = parent_entries.get(file_id, None)
702
if parent_entry_revs:
703
parent_entry = parent_entry_revs.get(heads[0], None)
704
else:
705
parent_entry = None
706
if parent_entry is None:
707
# The parent iter_changes was called against is the one
708
# that is the per-file head, so any change is relevant
709
# iter_changes is valid.
710
carry_over_possible = False
711
else:
712
# could be a carry over situation
713
# A change against the basis may just indicate a merge,
714
# we need to check the content against the source of the
715
# merge to determine if it was changed after the merge
716
# or carried over.
717
if (parent_entry.kind != entry.kind or
718
parent_entry.parent_id != entry.parent_id or
719
parent_entry.name != entry.name):
720
# Metadata common to all entries has changed
721
# against per-file parent
722
carry_over_possible = False
723
else:
724
carry_over_possible = True
725
# per-type checks for changes against the parent_entry
726
# are done below.
727
else:
728
# Cannot be a carry-over situation
729
carry_over_possible = False
730
# Populate the entry in the delta
731
if kind == 'file':
732
# XXX: There is still a small race here: If someone reverts the content of a file
733
# after iter_changes examines and decides it has changed,
734
# we will unconditionally record a new version even if some
735
# other process reverts it while commit is running (with
736
# the revert happening after iter_changes did it's
737
# examination).
738
if change[7][1]:
739
entry.executable = True
740
else:
741
entry.executable = False
742
if (carry_over_possible and
743
parent_entry.executable == entry.executable):
744
# Check the file length, content hash after reading
745
# the file.
746
nostore_sha = parent_entry.text_sha1
747
else:
748
nostore_sha = None
749
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
750
try:
751
text = file_obj.read()
752
finally:
753
file_obj.close()
754
try:
755
entry.text_sha1, entry.text_size = self._add_text_to_weave(
756
file_id, text, heads, nostore_sha)
757
yield file_id, change[1][1], (entry.text_sha1, stat_value)
758
except errors.ExistingContent:
759
# No content change against a carry_over parent
760
# Perhaps this should also yield a fs hash update?
761
carried_over = True
762
entry.text_size = parent_entry.text_size
763
entry.text_sha1 = parent_entry.text_sha1
764
elif kind == 'symlink':
765
# Wants a path hint?
766
entry.symlink_target = tree.get_symlink_target(file_id)
767
if (carry_over_possible and
768
parent_entry.symlink_target == entry.symlink_target):
769
carried_over = True
770
else:
771
self._add_text_to_weave(change[0], '', heads, None)
772
elif kind == 'directory':
773
if carry_over_possible:
774
carried_over = True
775
else:
776
# Nothing to set on the entry.
777
# XXX: split into the Root and nonRoot versions.
778
if change[1][1] != '' or self.repository.supports_rich_root():
779
self._add_text_to_weave(change[0], '', heads, None)
780
elif kind == 'tree-reference':
781
if not self.repository._format.supports_tree_reference:
782
# This isn't quite sane as an error, but we shouldn't
783
# ever see this code path in practice: tree's don't
784
# permit references when the repo doesn't support tree
785
# references.
786
raise errors.UnsupportedOperation(tree.add_reference,
787
self.repository)
788
reference_revision = tree.get_reference_revision(change[0])
789
entry.reference_revision = reference_revision
790
if (carry_over_possible and
791
parent_entry.reference_revision == reference_revision):
792
carried_over = True
793
else:
794
self._add_text_to_weave(change[0], '', heads, None)
795
else:
796
raise AssertionError('unknown kind %r' % kind)
797
if not carried_over:
798
entry.revision = modified_rev
799
else:
800
entry.revision = parent_entry.revision
801
else:
802
entry = None
803
new_path = change[1][1]
804
inv_delta.append((change[1][0], new_path, change[0], entry))
805
if new_path == '':
806
seen_root = True
807
self.new_inventory = None
808
if len(inv_delta):
809
self._any_changes = True
810
if not seen_root:
811
# housekeeping root entry changes do not affect no-change commits.
812
self._require_root_change(tree)
813
self.basis_delta_revision = basis_revision_id
814
815
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
816
parent_keys = tuple([(file_id, parent) for parent in parents])
817
return self.repository.texts._add_text(
818
(file_id, self._new_revision_id), parent_keys, new_text,
819
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
820
821
822
class RootCommitBuilder(CommitBuilder):
823
"""This commitbuilder actually records the root id"""
824
825
# the root entry gets versioned properly by this builder.
826
_versioned_root = True
827
828
def _check_root(self, ie, parent_invs, tree):
829
"""Helper for record_entry_contents.
830
831
:param ie: An entry being added.
832
:param parent_invs: The inventories of the parent revisions of the
833
commit.
834
:param tree: The tree that is being committed.
835
"""
836
837
def _require_root_change(self, tree):
838
"""Enforce an appropriate root object change.
839
840
This is called once when record_iter_changes is called, if and only if
841
the root was not in the delta calculated by record_iter_changes.
842
843
:param tree: The tree which is being committed.
844
"""
845
# versioned roots do not change unless the tree found a change.
846
847
848
64
######################################################################
849
65
# Repositories
850
66
855
71
revisions and file history. It's normally accessed only by the Branch,
856
72
which views a particular line of development through that history.
857
73
858
The Repository builds on top of some byte storage facilies (the revisions,
859
signatures, inventories, texts and chk_bytes attributes) and a Transport,
860
which respectively provide byte storage and a means to access the (possibly
74
The Repository builds on top of Stores and a Transport, which respectively
75
describe the disk data format and the way of accessing the (possibly
861
76
remote) disk.
862
863
The byte storage facilities are addressed via tuples, which we refer to
864
as 'keys' throughout the code base. Revision_keys, inventory_keys and
865
signature_keys are all 1-tuples: (revision_id,). text_keys are two-tuples:
866
(file_id, revision_id). chk_bytes uses CHK keys - a 1-tuple with a single
867
byte string made up of a hash identifier and a hash value.
868
We use this interface because it allows low friction with the underlying
869
code that implements disk indices, network encoding and other parts of
870
bzrlib.
871
872
:ivar revisions: A bzrlib.versionedfile.VersionedFiles instance containing
873
the serialised revisions for the repository. This can be used to obtain
874
revision graph information or to access raw serialised revisions.
875
The result of trying to insert data into the repository via this store
876
is undefined: it should be considered read-only except for implementors
877
of repositories.
878
:ivar signatures: A bzrlib.versionedfile.VersionedFiles instance containing
879
the serialised signatures for the repository. This can be used to
880
obtain access to raw serialised signatures. The result of trying to
881
insert data into the repository via this store is undefined: it should
882
be considered read-only except for implementors of repositories.
883
:ivar inventories: A bzrlib.versionedfile.VersionedFiles instance containing
884
the serialised inventories for the repository. This can be used to
885
obtain unserialised inventories. The result of trying to insert data
886
into the repository via this store is undefined: it should be
887
considered read-only except for implementors of repositories.
888
:ivar texts: A bzrlib.versionedfile.VersionedFiles instance containing the
889
texts of files and directories for the repository. This can be used to
890
obtain file texts or file graphs. Note that Repository.iter_file_bytes
891
is usually a better interface for accessing file texts.
892
The result of trying to insert data into the repository via this store
893
is undefined: it should be considered read-only except for implementors
894
of repositories.
895
:ivar chk_bytes: A bzrlib.versionedfile.VersionedFiles instance containing
896
any data the repository chooses to store or have indexed by its hash.
897
The result of trying to insert data into the repository via this store
898
is undefined: it should be considered read-only except for implementors
899
of repositories.
900
:ivar _transport: Transport for file access to repository, typically
901
pointing to .bzr/repository.
902
77
"""
903
78
904
# What class to use for a CommitBuilder. Often its simpler to change this
905
# in a Repository class subclass rather than to override
906
# get_commit_builder.
907
_commit_builder_class = CommitBuilder
908
# The search regex used by xml based repositories to determine what things
909
# where changed in a single commit.
910
79
_file_ids_altered_regex = lazy_regex.lazy_compile(
911
80
r'file_id="(?P<file_id>[^"]+)"'
912
r'.* revision="(?P<revision_id>[^"]+)"'
81
r'.*revision="(?P<revision_id>[^"]+)"'
913
82
)
914
83
915
def abort_write_group(self, suppress_errors=False):
916
"""Commit the contents accrued within the current write group.
917
918
:param suppress_errors: if true, abort_write_group will catch and log
919
unexpected errors that happen during the abort, rather than
920
allowing them to propagate. Defaults to False.
921
922
:seealso: start_write_group.
923
"""
924
if self._write_group is not self.get_transaction():
925
# has an unlock or relock occured ?
926
raise errors.BzrError(
927
'mismatched lock context and write group. %r, %r' %
928
(self._write_group, self.get_transaction()))
929
try:
930
self._abort_write_group()
931
except Exception, exc:
932
self._write_group = None
933
if not suppress_errors:
934
raise
935
mutter('abort_write_group failed')
936
log_exception_quietly()
937
note('bzr: ERROR (ignored): %s', exc)
938
self._write_group = None
939
940
def _abort_write_group(self):
941
"""Template method for per-repository write group cleanup.
942
943
This is called during abort before the write group is considered to be
944
finished and should cleanup any internal state accrued during the write
945
group. There is no requirement that data handed to the repository be
946
*not* made available - this is not a rollback - but neither should any
947
attempt be made to ensure that data added is fully commited. Abort is
948
invoked when an error has occured so futher disk or network operations
949
may not be possible or may error and if possible should not be
950
attempted.
951
"""
952
953
def add_fallback_repository(self, repository):
954
"""Add a repository to use for looking up data not held locally.
955
956
:param repository: A repository.
957
"""
958
if not self._format.supports_external_lookups:
959
raise errors.UnstackableRepositoryFormat(self._format, self.base)
960
if self.is_locked():
961
# This repository will call fallback.unlock() when we transition to
962
# the unlocked state, so we make sure to increment the lock count
963
repository.lock_read()
964
self._check_fallback_repository(repository)
965
self._fallback_repositories.append(repository)
966
self.texts.add_fallback_versioned_files(repository.texts)
967
self.inventories.add_fallback_versioned_files(repository.inventories)
968
self.revisions.add_fallback_versioned_files(repository.revisions)
969
self.signatures.add_fallback_versioned_files(repository.signatures)
970
if self.chk_bytes is not None:
971
self.chk_bytes.add_fallback_versioned_files(repository.chk_bytes)
972
973
def _check_fallback_repository(self, repository):
974
"""Check that this repository can fallback to repository safely.
975
976
Raise an error if not.
977
978
:param repository: A repository to fallback to.
979
"""
980
return InterRepository._assert_same_model(self, repository)
981
84
@needs_write_lock
982
85
def add_inventory(self, revision_id, inv, parents):
983
86
"""Add the inventory inv to the repository as revision_id.
984
87
985
88
:param parents: The revision ids of the parents that revision_id
986
89
is known to have and are in the repository already.
987
90
988
:returns: The validator(which is a sha1 digest, though what is sha'd is
989
repository format specific) of the serialized inventory.
91
returns the sha1 of the serialized inventory.
990
92
"""
991
if not self.is_in_write_group():
992
raise AssertionError("%r not in write group" % (self,))
93
revision_id = osutils.safe_revision_id(revision_id)
993
94
_mod_revision.check_not_reserved_id(revision_id)
994
if not (inv.revision_id is None or inv.revision_id == revision_id):
995
raise AssertionError(
996
"Mismatch between inventory revision"
997
" id and insertion revid (%r, %r)"
998
% (inv.revision_id, revision_id))
999
if inv.root is None:
1000
raise AssertionError()
1001
return self._add_inventory_checked(revision_id, inv, parents)
1002
1003
def _add_inventory_checked(self, revision_id, inv, parents):
1004
"""Add inv to the repository after checking the inputs.
1005
1006
This function can be overridden to allow different inventory styles.
1007
1008
:seealso: add_inventory, for the contract.
1009
"""
1010
inv_lines = self._serialise_inventory_to_lines(inv)
1011
return self._inventory_add_lines(revision_id, parents,
1012
inv_lines, check_content=False)
1013
1014
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1015
parents, basis_inv=None, propagate_caches=False):
1016
"""Add a new inventory expressed as a delta against another revision.
1017
1018
See the inventory developers documentation for the theory behind
1019
inventory deltas.
1020
1021
:param basis_revision_id: The inventory id the delta was created
1022
against. (This does not have to be a direct parent.)
1023
:param delta: The inventory delta (see Inventory.apply_delta for
1024
details).
1025
:param new_revision_id: The revision id that the inventory is being
1026
added for.
1027
:param parents: The revision ids of the parents that revision_id is
1028
known to have and are in the repository already. These are supplied
1029
for repositories that depend on the inventory graph for revision
1030
graph access, as well as for those that pun ancestry with delta
1031
compression.
1032
:param basis_inv: The basis inventory if it is already known,
1033
otherwise None.
1034
:param propagate_caches: If True, the caches for this inventory are
1035
copied to and updated for the result if possible.
1036
1037
:returns: (validator, new_inv)
1038
The validator(which is a sha1 digest, though what is sha'd is
1039
repository format specific) of the serialized inventory, and the
1040
resulting inventory.
1041
"""
1042
if not self.is_in_write_group():
1043
raise AssertionError("%r not in write group" % (self,))
1044
_mod_revision.check_not_reserved_id(new_revision_id)
1045
basis_tree = self.revision_tree(basis_revision_id)
1046
basis_tree.lock_read()
1047
try:
1048
# Note that this mutates the inventory of basis_tree, which not all
1049
# inventory implementations may support: A better idiom would be to
1050
# return a new inventory, but as there is no revision tree cache in
1051
# repository this is safe for now - RBC 20081013
1052
if basis_inv is None:
1053
basis_inv = basis_tree.inventory
1054
basis_inv.apply_delta(delta)
1055
basis_inv.revision_id = new_revision_id
1056
return (self.add_inventory(new_revision_id, basis_inv, parents),
1057
basis_inv)
1058
finally:
1059
basis_tree.unlock()
1060
1061
def _inventory_add_lines(self, revision_id, parents, lines,
1062
check_content=True):
1063
"""Store lines in inv_vf and return the sha1 of the inventory."""
1064
parents = [(parent,) for parent in parents]
1065
return self.inventories.add_lines((revision_id,), parents, lines,
1066
check_content=check_content)[0]
1067
95
assert inv.revision_id is None or inv.revision_id == revision_id, \
96
"Mismatch between inventory revision" \
97
" id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
98
assert inv.root is not None
99
inv_text = self.serialise_inventory(inv)
100
inv_sha1 = osutils.sha_string(inv_text)
101
inv_vf = self.control_weaves.get_weave('inventory',
102
self.get_transaction())
103
self._inventory_add_lines(inv_vf, revision_id, parents,
104
osutils.split_lines(inv_text))
105
return inv_sha1
106
107
def _inventory_add_lines(self, inv_vf, revision_id, parents, lines):
108
final_parents = []
109
for parent in parents:
110
if parent in inv_vf:
111
final_parents.append(parent)
112
113
inv_vf.add_lines(revision_id, final_parents, lines)
114
115
@needs_write_lock
1068
116
def add_revision(self, revision_id, rev, inv=None, config=None):
1069
117
"""Add rev to the revision store as revision_id.
1070
118
1076
124
If supplied its signature_needed method will be used
1077
125
to determine if a signature should be made.
1078
126
"""
127
revision_id = osutils.safe_revision_id(revision_id)
1079
128
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
1080
129
# rev.parent_ids?
1081
130
_mod_revision.check_not_reserved_id(revision_id)
1085
134
plaintext = Testament(rev, inv).as_short_text()
1086
135
self.store_revision_signature(
1087
136
gpg.GPGStrategy(config), plaintext, revision_id)
1088
# check inventory present
1089
if not self.inventories.get_parent_map([(revision_id,)]):
137
if not revision_id in self.get_inventory_weave():
1090
138
if inv is None:
1091
139
raise errors.WeaveRevisionNotPresent(revision_id,
1092
self.inventories)
140
self.get_inventory_weave())
1093
141
else:
1094
142
# yes, this is not suitable for adding with ghosts.
1095
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
1096
rev.parent_ids)
1097
else:
1098
key = (revision_id,)
1099
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
1100
self._add_revision(rev)
143
self.add_inventory(revision_id, inv, rev.parent_ids)
144
self._revision_store.add_revision(rev, self.get_transaction())
1101
145
1102
def _add_revision(self, revision):
1103
text = self._serializer.write_revision_to_string(revision)
1104
key = (revision.revision_id,)
1105
parents = tuple((parent,) for parent in revision.parent_ids)
1106
self.revisions.add_lines(key, parents, osutils.split_lines(text))
146
@needs_read_lock
147
def _all_possible_ids(self):
148
"""Return all the possible revisions that we could find."""
149
return self.get_inventory_weave().versions()
1107
150
1108
151
def all_revision_ids(self):
1109
"""Returns a list of all the revision ids in the repository.
152
"""Returns a list of all the revision ids in the repository.
1110
153
1111
This is conceptually deprecated because code should generally work on
1112
the graph reachable from a particular revision, and ignore any other
1113
revisions that might be present. There is no direct replacement
1114
method.
154
This is deprecated because code should generally work on the graph
155
reachable from a particular revision, and ignore any other revisions
156
that might be present. There is no direct replacement method.
1115
157
"""
1116
if 'evil' in debug.debug_flags:
1117
mutter_callsite(2, "all_revision_ids is linear with history.")
1118
158
return self._all_revision_ids()
1119
159
160
@needs_read_lock
1120
161
def _all_revision_ids(self):
1121
"""Returns a list of all the revision ids in the repository.
162
"""Returns a list of all the revision ids in the repository.
1122
163
1123
These are in as much topological order as the underlying store can
1124
present.
164
These are in as much topological order as the underlying store can
165
present: for weaves ghosts may lead to a lack of correctness until
166
the reweave updates the parents list.
1125
167
"""
1126
raise NotImplementedError(self._all_revision_ids)
168
if self._revision_store.text_store.listable():
169
return self._revision_store.all_revision_ids(self.get_transaction())
170
result = self._all_possible_ids()
171
# TODO: jam 20070210 Ensure that _all_possible_ids returns non-unicode
172
# ids. (It should, since _revision_store's API should change to
173
# return utf8 revision_ids)
174
return self._eliminate_revisions_not_present(result)
1127
175
1128
176
def break_lock(self):
1129
177
"""Break a lock if one is present from another instance.
1140
188
Returns a set of the present revisions.
1141
189
"""
1142
190
result = []
1143
graph = self.get_graph()
1144
parent_map = graph.get_parent_map(revision_ids)
1145
# The old API returned a list, should this actually be a set?
1146
return parent_map.keys()
191
for id in revision_ids:
192
if self.has_revision(id):
193
result.append(id)
194
return result
1147
195
1148
196
@staticmethod
1149
197
def create(a_bzrdir):
1150
198
"""Construct the current default format repository in a_bzrdir."""
1151
199
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
1152
200
1153
def __init__(self, _format, a_bzrdir, control_files):
201
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1154
202
"""instantiate a Repository.
1155
203
1156
204
:param _format: The format of the repository on disk.
1165
213
# the following are part of the public API for Repository:
1166
214
self.bzrdir = a_bzrdir
1167
215
self.control_files = control_files
1168
self._transport = control_files._transport
1169
self.base = self._transport.base
1170
# for tests
1171
self._reconcile_does_inventory_gc = True
1172
self._reconcile_fixes_text_parents = False
1173
self._reconcile_backsup_inventory = True
1174
# not right yet - should be more semantically clear ?
1175
#
216
self._revision_store = _revision_store
217
self.text_store = text_store
218
# backwards compatibility
219
self.weave_store = text_store
220
# not right yet - should be more semantically clear ?
221
#
222
self.control_store = control_store
223
self.control_weaves = control_store
1176
224
# TODO: make sure to construct the right store classes, etc, depending
1177
225
# on whether escaping is required.
1178
226
self._warn_if_deprecated()
1179
self._write_group = None
1180
# Additional places to query for data.
1181
self._fallback_repositories = []
1182
# An InventoryEntry cache, used during deserialization
1183
self._inventory_entry_cache = fifo_cache.FIFOCache(10*1024)
1184
227
1185
228
def __repr__(self):
1186
return '%s(%r)' % (self.__class__.__name__,
1187
self.base)
1188
1189
def has_same_location(self, other):
1190
"""Returns a boolean indicating if this repository is at the same
1191
location as another repository.
1192
1193
This might return False even when two repository objects are accessing
1194
the same physical repository via different URLs.
1195
"""
1196
if self.__class__ is not other.__class__:
1197
return False
1198
return (self._transport.base == other._transport.base)
1199
1200
def is_in_write_group(self):
1201
"""Return True if there is an open write group.
1202
1203
:seealso: start_write_group.
1204
"""
1205
return self._write_group is not None
229
return '%s(%r)' % (self.__class__.__name__,
230
self.bzrdir.transport.base)
1206
231
1207
232
def is_locked(self):
1208
233
return self.control_files.is_locked()
1209
234
1210
def is_write_locked(self):
1211
"""Return True if this object is write locked."""
1212
return self.is_locked() and self.control_files._lock_mode == 'w'
1213
1214
235
def lock_write(self, token=None):
1215
236
"""Lock this repository for writing.
1216
1217
This causes caching within the repository obejct to start accumlating
1218
data during reads, and allows a 'write_group' to be obtained. Write
1219
groups must be used for actual data insertion.
1220
237
1221
238
:param token: if this is already locked, then lock_write will fail
1222
239
unless the token matches the existing lock.
1223
240
:returns: a token if this instance supports tokens, otherwise None.
1225
242
instance doesn't support using token locks.
1226
243
:raises MismatchedToken: if the specified token doesn't match the token
1227
244
of the existing lock.
1228
:seealso: start_write_group.
1229
245
1230
246
A token should be passed in if you know that you have locked the object
1231
247
some other way, and need to synchronise this object's state with that
1233
249
1234
250
XXX: this docstring is duplicated in many places, e.g. lockable_files.py
1235
251
"""
1236
locked = self.is_locked()
1237
result = self.control_files.lock_write(token=token)
1238
if not locked:
1239
for repo in self._fallback_repositories:
1240
# Writes don't affect fallback repos
1241
repo.lock_read()
1242
self._refresh_data()
1243
return result
252
return self.control_files.lock_write(token=token)
1244
253
1245
254
def lock_read(self):
1246
locked = self.is_locked()
1247
255
self.control_files.lock_read()
1248
if not locked:
1249
for repo in self._fallback_repositories:
1250
repo.lock_read()
1251
self._refresh_data()
1252
256
1253
257
def get_physical_lock_status(self):
1254
258
return self.control_files.get_physical_lock_status()
1256
260
def leave_lock_in_place(self):
1257
261
"""Tell this repository not to release the physical lock when this
1258
262
object is unlocked.
1259
263
1260
264
If lock_write doesn't return a token, then this method is not supported.
1261
265
"""
1262
266
self.control_files.leave_in_place()
1313
317
last_revision.timezone)
1314
318
1315
319
# now gather global repository information
1316
# XXX: This is available for many repos regardless of listability.
1317
320
if self.bzrdir.root_transport.listable():
1318
# XXX: do we want to __define len__() ?
1319
# Maybe the versionedfiles object should provide a different
1320
# method to get the number of keys.
1321
result['revisions'] = len(self.revisions.keys())
1322
# result['size'] = t
321
c, t = self._revision_store.total_size(self.get_transaction())
322
result['revisions'] = c
323
result['size'] = t
1323
324
return result
1324
325
1325
def find_branches(self, using=False):
1326
"""Find branches underneath this repository.
1327
1328
This will include branches inside other branches.
1329
1330
:param using: If True, list only branches using this repository.
1331
"""
1332
if using and not self.is_shared():
1333
try:
1334
return [self.bzrdir.open_branch()]
1335
except errors.NotBranchError:
1336
return []
1337
class Evaluator(object):
1338
1339
def __init__(self):
1340
self.first_call = True
1341
1342
def __call__(self, bzrdir):
1343
# On the first call, the parameter is always the bzrdir
1344
# containing the current repo.
1345
if not self.first_call:
1346
try:
1347
repository = bzrdir.open_repository()
1348
except errors.NoRepositoryPresent:
1349
pass
1350
else:
1351
return False, (None, repository)
1352
self.first_call = False
1353
try:
1354
value = (bzrdir.open_branch(), None)
1355
except errors.NotBranchError:
1356
value = (None, None)
1357
return True, value
1358
1359
branches = []
1360
for branch, repository in bzrdir.BzrDir.find_bzrdirs(
1361
self.bzrdir.root_transport, evaluate=Evaluator()):
1362
if branch is not None:
1363
branches.append(branch)
1364
if not using and repository is not None:
1365
branches.extend(repository.find_branches())
1366
return branches
1367
1368
326
@needs_read_lock
1369
def search_missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
327
def missing_revision_ids(self, other, revision_id=None):
1370
328
"""Return the revision ids that other has that this does not.
1371
329
1372
330
These are returned in topological order.
1373
331
1374
332
revision_id: only return revision ids included by revision_id.
1375
333
"""
1376
return InterRepository.get(other, self).search_missing_revision_ids(
1377
revision_id, find_ghosts)
334
revision_id = osutils.safe_revision_id(revision_id)
335
return InterRepository.get(other, self).missing_revision_ids(revision_id)
1378
336
1379
337
@staticmethod
1380
338
def open(base):
1388
346
1389
347
def copy_content_into(self, destination, revision_id=None):
1390
348
"""Make a complete copy of the content in self into destination.
1391
1392
This is a destructive operation! Do not use it on existing
349
350
This is a destructive operation! Do not use it on existing
1393
351
repositories.
1394
352
"""
353
revision_id = osutils.safe_revision_id(revision_id)
1395
354
return InterRepository.get(self, destination).copy_content(revision_id)
1396
355
1397
def commit_write_group(self):
1398
"""Commit the contents accrued within the current write group.
1399
1400
:seealso: start_write_group.
1401
"""
1402
if self._write_group is not self.get_transaction():
1403
# has an unlock or relock occured ?
1404
raise errors.BzrError('mismatched lock context %r and '
1405
'write group %r.' %
1406
(self.get_transaction(), self._write_group))
1407
result = self._commit_write_group()
1408
self._write_group = None
1409
return result
1410
1411
def _commit_write_group(self):
1412
"""Template method for per-repository write group cleanup.
1413
1414
This is called before the write group is considered to be
1415
finished and should ensure that all data handed to the repository
1416
for writing during the write group is safely committed (to the
1417
extent possible considering file system caching etc).
1418
"""
1419
1420
def suspend_write_group(self):
1421
raise errors.UnsuspendableWriteGroup(self)
1422
1423
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1424
"""Return the keys of missing inventory parents for revisions added in
1425
this write group.
1426
1427
A revision is not complete if the inventory delta for that revision
1428
cannot be calculated. Therefore if the parent inventories of a
1429
revision are not present, the revision is incomplete, and e.g. cannot
1430
be streamed by a smart server. This method finds missing inventory
1431
parents for revisions added in this write group.
1432
"""
1433
if not self._format.supports_external_lookups:
1434
# This is only an issue for stacked repositories
1435
return set()
1436
if not self.is_in_write_group():
1437
raise AssertionError('not in a write group')
1438
1439
# XXX: We assume that every added revision already has its
1440
# corresponding inventory, so we only check for parent inventories that
1441
# might be missing, rather than all inventories.
1442
parents = set(self.revisions._index.get_missing_parents())
1443
parents.discard(_mod_revision.NULL_REVISION)
1444
unstacked_inventories = self.inventories._index
1445
present_inventories = unstacked_inventories.get_parent_map(
1446
key[-1:] for key in parents)
1447
parents.difference_update(present_inventories)
1448
if len(parents) == 0:
1449
# No missing parent inventories.
1450
return set()
1451
if not check_for_missing_texts:
1452
return set(('inventories', rev_id) for (rev_id,) in parents)
1453
# Ok, now we have a list of missing inventories. But these only matter
1454
# if the inventories that reference them are missing some texts they
1455
# appear to introduce.
1456
# XXX: Texts referenced by all added inventories need to be present,
1457
# but at the moment we're only checking for texts referenced by
1458
# inventories at the graph's edge.
1459
key_deps = self.revisions._index._key_dependencies
1460
key_deps.add_keys(present_inventories)
1461
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1462
file_ids = self.fileids_altered_by_revision_ids(referrers)
1463
missing_texts = set()
1464
for file_id, version_ids in file_ids.iteritems():
1465
missing_texts.update(
1466
(file_id, version_id) for version_id in version_ids)
1467
present_texts = self.texts.get_parent_map(missing_texts)
1468
missing_texts.difference_update(present_texts)
1469
if not missing_texts:
1470
# No texts are missing, so all revisions and their deltas are
1471
# reconstructable.
1472
return set()
1473
# Alternatively the text versions could be returned as the missing
1474
# keys, but this is likely to be less data.
1475
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1476
return missing_keys
1477
1478
def refresh_data(self):
1479
"""Re-read any data needed to to synchronise with disk.
1480
1481
This method is intended to be called after another repository instance
1482
(such as one used by a smart server) has inserted data into the
1483
repository. It may not be called during a write group, but may be
1484
called at any other time.
1485
"""
1486
if self.is_in_write_group():
1487
raise errors.InternalBzrError(
1488
"May not refresh_data while in a write group.")
1489
self._refresh_data()
1490
1491
def resume_write_group(self, tokens):
1492
if not self.is_write_locked():
1493
raise errors.NotWriteLocked(self)
1494
if self._write_group:
1495
raise errors.BzrError('already in a write group')
1496
self._resume_write_group(tokens)
1497
# so we can detect unlock/relock - the write group is now entered.
1498
self._write_group = self.get_transaction()
1499
1500
def _resume_write_group(self, tokens):
1501
raise errors.UnsuspendableWriteGroup(self)
1502
1503
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
1504
fetch_spec=None):
356
def fetch(self, source, revision_id=None, pb=None):
1505
357
"""Fetch the content required to construct revision_id from source.
1506
358
1507
If revision_id is None and fetch_spec is None, then all content is
1508
copied.
1509
1510
fetch() may not be used when the repository is in a write group -
1511
either finish the current write group before using fetch, or use
1512
fetch before starting the write group.
1513
1514
:param find_ghosts: Find and copy revisions in the source that are
1515
ghosts in the target (and not reachable directly by walking out to
1516
the first-present revision in target from revision_id).
1517
:param revision_id: If specified, all the content needed for this
1518
revision ID will be copied to the target. Fetch will determine for
1519
itself which content needs to be copied.
1520
:param fetch_spec: If specified, a SearchResult or
1521
PendingAncestryResult that describes which revisions to copy. This
1522
allows copying multiple heads at once. Mutually exclusive with
1523
revision_id.
359
If revision_id is None all content is copied.
1524
360
"""
1525
if fetch_spec is not None and revision_id is not None:
1526
raise AssertionError(
1527
"fetch_spec and revision_id are mutually exclusive.")
1528
if self.is_in_write_group():
1529
raise errors.InternalBzrError(
1530
"May not fetch while in a write group.")
1531
# fast path same-url fetch operations
1532
if self.has_same_location(source) and fetch_spec is None:
1533
# check that last_revision is in 'from' and then return a
1534
# no-operation.
1535
if (revision_id is not None and
1536
not _mod_revision.is_null(revision_id)):
1537
self.get_revision(revision_id)
1538
return 0, []
1539
# if there is no specific appropriate InterRepository, this will get
1540
# the InterRepository base class, which raises an
1541
# IncompatibleRepositories when asked to fetch.
361
revision_id = osutils.safe_revision_id(revision_id)
1542
362
inter = InterRepository.get(source, self)
1543
return inter.fetch(revision_id=revision_id, pb=pb,
1544
find_ghosts=find_ghosts, fetch_spec=fetch_spec)
1545
1546
def create_bundle(self, target, base, fileobj, format=None):
1547
return serializer.write_bundle(self, target, base, fileobj, format)
1548
1549
def get_commit_builder(self, branch, parents, config, timestamp=None,
1550
timezone=None, committer=None, revprops=None,
363
try:
364
return inter.fetch(revision_id=revision_id, pb=pb)
365
except NotImplementedError:
366
raise errors.IncompatibleRepositories(source, self)
367
368
def get_commit_builder(self, branch, parents, config, timestamp=None,
369
timezone=None, committer=None, revprops=None,
1551
370
revision_id=None):
1552
371
"""Obtain a CommitBuilder for this repository.
1553
372
1554
373
:param branch: Branch to commit to.
1555
374
:param parents: Revision ids of the parents of the new revision.
1556
375
:param config: Configuration to use.
1560
379
:param revprops: Optional dictionary of revision properties.
1561
380
:param revision_id: Optional revision id.
1562
381
"""
1563
result = self._commit_builder_class(self, parents, config,
1564
timestamp, timezone, committer, revprops, revision_id)
1565
self.start_write_group()
1566
return result
382
revision_id = osutils.safe_revision_id(revision_id)
383
return _CommitBuilder(self, parents, config, timestamp, timezone,
384
committer, revprops, revision_id)
1567
385
1568
386
def unlock(self):
1569
if (self.control_files._lock_count == 1 and
1570
self.control_files._lock_mode == 'w'):
1571
if self._write_group is not None:
1572
self.abort_write_group()
1573
self.control_files.unlock()
1574
raise errors.BzrError(
1575
'Must end write groups before releasing write locks.')
1576
387
self.control_files.unlock()
1577
if self.control_files._lock_count == 0:
1578
self._inventory_entry_cache.clear()
1579
for repo in self._fallback_repositories:
1580
repo.unlock()
1581
388
1582
389
@needs_read_lock
1583
390
def clone(self, a_bzrdir, revision_id=None):
1594
401
self.copy_content_into(dest_repo, revision_id)
1595
402
return dest_repo
1596
403
1597
def start_write_group(self):
1598
"""Start a write group in the repository.
1599
1600
Write groups are used by repositories which do not have a 1:1 mapping
1601
between file ids and backend store to manage the insertion of data from
1602
both fetch and commit operations.
1603
1604
A write lock is required around the start_write_group/commit_write_group
1605
for the support of lock-requiring repository formats.
1606
1607
One can only insert data into a repository inside a write group.
1608
1609
:return: None.
1610
"""
1611
if not self.is_write_locked():
1612
raise errors.NotWriteLocked(self)
1613
if self._write_group:
1614
raise errors.BzrError('already in a write group')
1615
self._start_write_group()
1616
# so we can detect unlock/relock - the write group is now entered.
1617
self._write_group = self.get_transaction()
1618
1619
def _start_write_group(self):
1620
"""Template method for per-repository write group startup.
1621
1622
This is called before the write group is considered to be
1623
entered.
1624
"""
1625
1626
404
@needs_read_lock
1627
405
def sprout(self, to_bzrdir, revision_id=None):
1628
406
"""Create a descendent repository for new development.
1646
424
dest_repo = a_bzrdir.open_repository()
1647
425
return dest_repo
1648
426
1649
def _get_sink(self):
1650
"""Return a sink for streaming into this repository."""
1651
return StreamSink(self)
1652
1653
def _get_source(self, to_format):
1654
"""Return a source for streaming from this repository."""
1655
return StreamSource(self, to_format)
1656
1657
427
@needs_read_lock
1658
428
def has_revision(self, revision_id):
1659
429
"""True if this repository has a copy of the revision."""
1660
return revision_id in self.has_revisions((revision_id,))
1661
1662
@needs_read_lock
1663
def has_revisions(self, revision_ids):
1664
"""Probe to find out the presence of multiple revisions.
1665
1666
:param revision_ids: An iterable of revision_ids.
1667
:return: A set of the revision_ids that were present.
1668
"""
1669
parent_map = self.revisions.get_parent_map(
1670
[(rev_id,) for rev_id in revision_ids])
1671
result = set()
1672
if _mod_revision.NULL_REVISION in revision_ids:
1673
result.add(_mod_revision.NULL_REVISION)
1674
result.update([key[0] for key in parent_map])
1675
return result
1676
1677
@needs_read_lock
1678
def get_revision(self, revision_id):
1679
"""Return the Revision object for a named revision."""
1680
return self.get_revisions([revision_id])[0]
430
revision_id = osutils.safe_revision_id(revision_id)
431
return self._revision_store.has_revision_id(revision_id,
432
self.get_transaction())
1681
433
1682
434
@needs_read_lock
1683
435
def get_revision_reconcile(self, revision_id):
1684
436
"""'reconcile' helper routine that allows access to a revision always.
1685
437
1686
438
This variant of get_revision does not cross check the weave graph
1687
439
against the revision one as get_revision does: but it should only
1688
440
be used by reconcile, or reconcile-alike commands that are correcting
1689
441
or testing the revision graph.
1690
442
"""
1691
return self._get_revisions([revision_id])[0]
443
if not revision_id or not isinstance(revision_id, basestring):
444
raise errors.InvalidRevisionId(revision_id=revision_id,
445
branch=self)
446
return self.get_revisions([revision_id])[0]
1692
447
1693
448
@needs_read_lock
1694
449
def get_revisions(self, revision_ids):
1695
"""Get many revisions at once."""
1696
return self._get_revisions(revision_ids)
1697
1698
@needs_read_lock
1699
def _get_revisions(self, revision_ids):
1700
"""Core work logic to get many revisions without sanity checks."""
1701
for rev_id in revision_ids:
1702
if not rev_id or not isinstance(rev_id, basestring):
1703
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1704
keys = [(key,) for key in revision_ids]
1705
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1706
revs = {}
1707
for record in stream:
1708
if record.storage_kind == 'absent':
1709
raise errors.NoSuchRevision(self, record.key[0])
1710
text = record.get_bytes_as('fulltext')
1711
rev = self._serializer.read_revision_from_string(text)
1712
revs[record.key[0]] = rev
1713
return [revs[revid] for revid in revision_ids]
450
revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
451
revs = self._revision_store.get_revisions(revision_ids,
452
self.get_transaction())
453
for rev in revs:
454
assert not isinstance(rev.revision_id, unicode)
455
for parent_id in rev.parent_ids:
456
assert not isinstance(parent_id, unicode)
457
return revs
1714
458
1715
459
@needs_read_lock
1716
460
def get_revision_xml(self, revision_id):
1717
461
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1718
462
# would have already do it.
1719
463
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1720
# TODO: this can't just be replaced by:
1721
# return self._serializer.write_revision_to_string(
1722
# self.get_revision(revision_id))
1723
# as cStringIO preservers the encoding unlike write_revision_to_string
1724
# or some other call down the path.
464
revision_id = osutils.safe_revision_id(revision_id)
1725
465
rev = self.get_revision(revision_id)
1726
rev_tmp = cStringIO.StringIO()
466
rev_tmp = StringIO()
1727
467
# the current serializer..
1728
self._serializer.write_revision(rev, rev_tmp)
468
self._revision_store._serializer.write_revision(rev, rev_tmp)
1729
469
rev_tmp.seek(0)
1730
470
return rev_tmp.getvalue()
1731
471
1732
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
472
@needs_read_lock
473
def get_revision(self, revision_id):
474
"""Return the Revision object for a named revision"""
475
# TODO: jam 20070210 get_revision_reconcile should do this for us
476
revision_id = osutils.safe_revision_id(revision_id)
477
r = self.get_revision_reconcile(revision_id)
478
# weave corruption can lead to absent revision markers that should be
479
# present.
480
# the following test is reasonably cheap (it needs a single weave read)
481
# and the weave is cached in read transactions. In write transactions
482
# it is not cached but typically we only read a small number of
483
# revisions. For knits when they are introduced we will probably want
484
# to ensure that caching write transactions are in use.
485
inv = self.get_inventory_weave()
486
self._check_revision_parents(r, inv)
487
return r
488
489
@needs_read_lock
490
def get_deltas_for_revisions(self, revisions):
1733
491
"""Produce a generator of revision deltas.
1734
492
1735
493
Note that the input is a sequence of REVISIONS, not revision_ids.
1736
494
Trees will be held in memory until the generator exits.
1737
495
Each delta is relative to the revision's lefthand predecessor.
1738
1739
:param specific_fileids: if not None, the result is filtered
1740
so that only those file-ids, their parents and their
1741
children are included.
1742
496
"""
1743
# Get the revision-ids of interest
1744
497
required_trees = set()
1745
498
for revision in revisions:
1746
499
required_trees.add(revision.revision_id)
1747
500
required_trees.update(revision.parent_ids[:1])
1748
1749
# Get the matching filtered trees. Note that it's more
1750
# efficient to pass filtered trees to changes_from() rather
1751
# than doing the filtering afterwards. changes_from() could
1752
# arguably do the filtering itself but it's path-based, not
1753
# file-id based, so filtering before or afterwards is
1754
# currently easier.
1755
if specific_fileids is None:
1756
trees = dict((t.get_revision_id(), t) for
1757
t in self.revision_trees(required_trees))
1758
else:
1759
trees = dict((t.get_revision_id(), t) for
1760
t in self._filtered_revision_trees(required_trees,
1761
specific_fileids))
1762
1763
# Calculate the deltas
501
trees = dict((t.get_revision_id(), t) for
502
t in self.revision_trees(required_trees))
1764
503
for revision in revisions:
1765
504
if not revision.parent_ids:
1766
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
505
old_tree = self.revision_tree(None)
1767
506
else:
1768
507
old_tree = trees[revision.parent_ids[0]]
1769
508
yield trees[revision.revision_id].changes_from(old_tree)
1770
509
1771
510
@needs_read_lock
1772
def get_revision_delta(self, revision_id, specific_fileids=None):
511
def get_revision_delta(self, revision_id):
1773
512
"""Return the delta for one revision.
1774
513
1775
514
The delta is relative to the left-hand predecessor of the
1776
515
revision.
1777
1778
:param specific_fileids: if not None, the result is filtered
1779
so that only those file-ids, their parents and their
1780
children are included.
1781
516
"""
1782
517
r = self.get_revision(revision_id)
1783
return list(self.get_deltas_for_revisions([r],
1784
specific_fileids=specific_fileids))[0]
518
return list(self.get_deltas_for_revisions([r]))[0]
519
520
def _check_revision_parents(self, revision, inventory):
521
"""Private to Repository and Fetch.
522
523
This checks the parentage of revision in an inventory weave for
524
consistency and is only applicable to inventory-weave-for-ancestry
525
using repository formats & fetchers.
526
"""
527
weave_parents = inventory.get_parents(revision.revision_id)
528
weave_names = inventory.versions()
529
for parent_id in revision.parent_ids:
530
if parent_id in weave_names:
531
# this parent must not be a ghost.
532
if not parent_id in weave_parents:
533
# but it is a ghost
534
raise errors.CorruptRepository(self)
1785
535
1786
536
@needs_write_lock
1787
537
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
538
revision_id = osutils.safe_revision_id(revision_id)
1788
539
signature = gpg_strategy.sign(plaintext)
1789
self.add_signature_text(revision_id, signature)
1790
1791
@needs_write_lock
1792
def add_signature_text(self, revision_id, signature):
1793
self.signatures.add_lines((revision_id,), (),
1794
osutils.split_lines(signature))
1795
1796
def find_text_key_references(self):
1797
"""Find the text key references within the repository.
1798
1799
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1800
to whether they were referred to by the inventory of the
1801
revision_id that they contain. The inventory texts from all present
1802
revision ids are assessed to generate this report.
1803
"""
1804
revision_keys = self.revisions.keys()
1805
w = self.inventories
1806
pb = ui.ui_factory.nested_progress_bar()
1807
try:
1808
return self._find_text_key_references_from_xml_inventory_lines(
1809
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
1810
finally:
1811
pb.finished()
1812
1813
def _find_text_key_references_from_xml_inventory_lines(self,
1814
line_iterator):
1815
"""Core routine for extracting references to texts from inventories.
1816
1817
This performs the translation of xml lines to revision ids.
1818
1819
:param line_iterator: An iterator of lines, origin_version_id
1820
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1821
to whether they were referred to by the inventory of the
1822
revision_id that they contain. Note that if that revision_id was
1823
not part of the line_iterator's output then False will be given -
1824
even though it may actually refer to that key.
1825
"""
1826
if not self._serializer.support_altered_by_hack:
1827
raise AssertionError(
1828
"_find_text_key_references_from_xml_inventory_lines only "
1829
"supported for branches which store inventory as unnested xml"
1830
", not on %r" % self)
540
self._revision_store.add_revision_signature_text(revision_id,
541
signature,
542
self.get_transaction())
543
544
def fileids_altered_by_revision_ids(self, revision_ids):
545
"""Find the file ids and versions affected by revisions.
546
547
:param revisions: an iterable containing revision ids.
548
:return: a dictionary mapping altered file-ids to an iterable of
549
revision_ids. Each altered file-ids has the exact revision_ids that
550
altered it listed explicitly.
551
"""
552
assert self._serializer.support_altered_by_hack, \
553
("fileids_altered_by_revision_ids only supported for branches "
554
"which store inventory as unnested xml, not on %r" % self)
555
selected_revision_ids = set(osutils.safe_revision_id(r)
556
for r in revision_ids)
557
w = self.get_inventory_weave()
1831
558
result = {}
1832
559
1833
560
# this code needs to read every new line in every inventory for the
1834
561
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1835
# not present in one of those inventories is unnecessary but not
562
# not present in one of those inventories is unnecessary but not
1836
563
# harmful because we are filtering by the revision id marker in the
1837
# inventory lines : we only select file ids altered in one of those
564
# inventory lines : we only select file ids altered in one of those
1838
565
# revisions. We don't need to see all lines in the inventory because
1839
566
# only those added in an inventory in rev X can contain a revision=X
1840
567
# line.
1849
576
search = self._file_ids_altered_regex.search
1850
577
unescape = _unescape_xml
1851
578
setdefault = result.setdefault
1852
for line, line_key in line_iterator:
1853
match = search(line)
1854
if match is None:
1855
continue
1856
# One call to match.group() returning multiple items is quite a
1857
# bit faster than 2 calls to match.group() each returning 1
1858
file_id, revision_id = match.group('file_id', 'revision_id')
1859
1860
# Inlining the cache lookups helps a lot when you make 170,000
1861
# lines and 350k ids, versus 8.4 unique ids.
1862
# Using a cache helps in 2 ways:
1863
# 1) Avoids unnecessary decoding calls
1864
# 2) Re-uses cached strings, which helps in future set and
1865
# equality checks.
1866
# (2) is enough that removing encoding entirely along with
1867
# the cache (so we are using plain strings) results in no
1868
# performance improvement.
1869
try:
1870
revision_id = unescape_revid_cache[revision_id]
1871
except KeyError:
1872
unescaped = unescape(revision_id)
1873
unescape_revid_cache[revision_id] = unescaped
1874
revision_id = unescaped
1875
1876
# Note that unconditionally unescaping means that we deserialise
1877
# every fileid, which for general 'pull' is not great, but we don't
1878
# really want to have some many fulltexts that this matters anyway.
1879
# RBC 20071114.
1880
try:
1881
file_id = unescape_fileid_cache[file_id]
1882
except KeyError:
1883
unescaped = unescape(file_id)
1884
unescape_fileid_cache[file_id] = unescaped
1885
file_id = unescaped
1886
1887
key = (file_id, revision_id)
1888
setdefault(key, False)
1889
if revision_id == line_key[-1]:
1890
result[key] = True
1891
return result
1892
1893
def _inventory_xml_lines_for_keys(self, keys):
1894
"""Get a line iterator of the sort needed for findind references.
1895
1896
Not relevant for non-xml inventory repositories.
1897
1898
Ghosts in revision_keys are ignored.
1899
1900
:param revision_keys: The revision keys for the inventories to inspect.
1901
:return: An iterator over (inventory line, revid) for the fulltexts of
1902
all of the xml inventories specified by revision_keys.
1903
"""
1904
stream = self.inventories.get_record_stream(keys, 'unordered', True)
1905
for record in stream:
1906
if record.storage_kind != 'absent':
1907
chunks = record.get_bytes_as('chunked')
1908
revid = record.key[-1]
1909
lines = osutils.chunks_to_lines(chunks)
1910
for line in lines:
1911
yield line, revid
1912
1913
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1914
revision_keys):
1915
"""Helper routine for fileids_altered_by_revision_ids.
1916
1917
This performs the translation of xml lines to revision ids.
1918
1919
:param line_iterator: An iterator of lines, origin_version_id
1920
:param revision_keys: The revision ids to filter for. This should be a
1921
set or other type which supports efficient __contains__ lookups, as
1922
the revision key from each parsed line will be looked up in the
1923
revision_keys filter.
1924
:return: a dictionary mapping altered file-ids to an iterable of
1925
revision_ids. Each altered file-ids has the exact revision_ids that
1926
altered it listed explicitly.
1927
"""
1928
seen = set(self._find_text_key_references_from_xml_inventory_lines(
1929
line_iterator).iterkeys())
1930
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
1931
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
1932
self._inventory_xml_lines_for_keys(parent_keys)))
1933
new_keys = seen - parent_seen
1934
result = {}
1935
setdefault = result.setdefault
1936
for key in new_keys:
1937
setdefault(key[0], set()).add(key[-1])
1938
return result
1939
1940
def _find_parent_ids_of_revisions(self, revision_ids):
1941
"""Find all parent ids that are mentioned in the revision graph.
1942
1943
:return: set of revisions that are parents of revision_ids which are
1944
not part of revision_ids themselves
1945
"""
1946
parent_map = self.get_parent_map(revision_ids)
1947
parent_ids = set()
1948
map(parent_ids.update, parent_map.itervalues())
1949
parent_ids.difference_update(revision_ids)
1950
parent_ids.discard(_mod_revision.NULL_REVISION)
1951
return parent_ids
1952
1953
def _find_parent_keys_of_revisions(self, revision_keys):
1954
"""Similar to _find_parent_ids_of_revisions, but used with keys.
1955
1956
:param revision_keys: An iterable of revision_keys.
1957
:return: The parents of all revision_keys that are not already in
1958
revision_keys
1959
"""
1960
parent_map = self.revisions.get_parent_map(revision_keys)
1961
parent_keys = set()
1962
map(parent_keys.update, parent_map.itervalues())
1963
parent_keys.difference_update(revision_keys)
1964
parent_keys.discard(_mod_revision.NULL_REVISION)
1965
return parent_keys
1966
1967
def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1968
"""Find the file ids and versions affected by revisions.
1969
1970
:param revisions: an iterable containing revision ids.
1971
:param _inv_weave: The inventory weave from this repository or None.
1972
If None, the inventory weave will be opened automatically.
1973
:return: a dictionary mapping altered file-ids to an iterable of
1974
revision_ids. Each altered file-ids has the exact revision_ids that
1975
altered it listed explicitly.
1976
"""
1977
selected_keys = set((revid,) for revid in revision_ids)
1978
w = _inv_weave or self.inventories
1979
pb = ui.ui_factory.nested_progress_bar()
1980
try:
1981
return self._find_file_ids_from_xml_inventory_lines(
1982
w.iter_lines_added_or_present_in_keys(
1983
selected_keys, pb=pb),
1984
selected_keys)
1985
finally:
1986
pb.finished()
1987
1988
def iter_files_bytes(self, desired_files):
1989
"""Iterate through file versions.
1990
1991
Files will not necessarily be returned in the order they occur in
1992
desired_files. No specific order is guaranteed.
1993
1994
Yields pairs of identifier, bytes_iterator. identifier is an opaque
1995
value supplied by the caller as part of desired_files. It should
1996
uniquely identify the file version in the caller's context. (Examples:
1997
an index number or a TreeTransform trans_id.)
1998
1999
bytes_iterator is an iterable of bytestrings for the file. The
2000
kind of iterable and length of the bytestrings are unspecified, but for
2001
this implementation, it is a list of bytes produced by
2002
VersionedFile.get_record_stream().
2003
2004
:param desired_files: a list of (file_id, revision_id, identifier)
2005
triples
2006
"""
2007
text_keys = {}
2008
for file_id, revision_id, callable_data in desired_files:
2009
text_keys[(file_id, revision_id)] = callable_data
2010
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2011
if record.storage_kind == 'absent':
2012
raise errors.RevisionNotPresent(record.key, self)
2013
yield text_keys[record.key], record.get_bytes_as('chunked')
2014
2015
def _generate_text_key_index(self, text_key_references=None,
2016
ancestors=None):
2017
"""Generate a new text key index for the repository.
2018
2019
This is an expensive function that will take considerable time to run.
2020
2021
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2022
list of parents, also text keys. When a given key has no parents,
2023
the parents list will be [NULL_REVISION].
2024
"""
2025
# All revisions, to find inventory parents.
2026
if ancestors is None:
2027
graph = self.get_graph()
2028
ancestors = graph.get_parent_map(self.all_revision_ids())
2029
if text_key_references is None:
2030
text_key_references = self.find_text_key_references()
2031
pb = ui.ui_factory.nested_progress_bar()
2032
try:
2033
return self._do_generate_text_key_index(ancestors,
2034
text_key_references, pb)
2035
finally:
2036
pb.finished()
2037
2038
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2039
"""Helper for _generate_text_key_index to avoid deep nesting."""
2040
revision_order = tsort.topo_sort(ancestors)
2041
invalid_keys = set()
2042
revision_keys = {}
2043
for revision_id in revision_order:
2044
revision_keys[revision_id] = set()
2045
text_count = len(text_key_references)
2046
# a cache of the text keys to allow reuse; costs a dict of all the
2047
# keys, but saves a 2-tuple for every child of a given key.
2048
text_key_cache = {}
2049
for text_key, valid in text_key_references.iteritems():
2050
if not valid:
2051
invalid_keys.add(text_key)
2052
else:
2053
revision_keys[text_key[1]].add(text_key)
2054
text_key_cache[text_key] = text_key
2055
del text_key_references
2056
text_index = {}
2057
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2058
NULL_REVISION = _mod_revision.NULL_REVISION
2059
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2060
# too small for large or very branchy trees. However, for 55K path
2061
# trees, it would be easy to use too much memory trivially. Ideally we
2062
# could gauge this by looking at available real memory etc, but this is
2063
# always a tricky proposition.
2064
inventory_cache = lru_cache.LRUCache(10)
2065
batch_size = 10 # should be ~150MB on a 55K path tree
2066
batch_count = len(revision_order) / batch_size + 1
2067
processed_texts = 0
2068
pb.update("Calculating text parents", processed_texts, text_count)
2069
for offset in xrange(batch_count):
2070
to_query = revision_order[offset * batch_size:(offset + 1) *
2071
batch_size]
2072
if not to_query:
2073
break
2074
for rev_tree in self.revision_trees(to_query):
2075
revision_id = rev_tree.get_revision_id()
2076
parent_ids = ancestors[revision_id]
2077
for text_key in revision_keys[revision_id]:
2078
pb.update("Calculating text parents", processed_texts)
2079
processed_texts += 1
2080
candidate_parents = []
2081
for parent_id in parent_ids:
2082
parent_text_key = (text_key[0], parent_id)
2083
try:
2084
check_parent = parent_text_key not in \
2085
revision_keys[parent_id]
2086
except KeyError:
2087
# the parent parent_id is a ghost:
2088
check_parent = False
2089
# truncate the derived graph against this ghost.
2090
parent_text_key = None
2091
if check_parent:
2092
# look at the parent commit details inventories to
2093
# determine possible candidates in the per file graph.
2094
# TODO: cache here.
2095
try:
2096
inv = inventory_cache[parent_id]
2097
except KeyError:
2098
inv = self.revision_tree(parent_id).inventory
2099
inventory_cache[parent_id] = inv
2100
try:
2101
parent_entry = inv[text_key[0]]
2102
except (KeyError, errors.NoSuchId):
2103
parent_entry = None
2104
if parent_entry is not None:
2105
parent_text_key = (
2106
text_key[0], parent_entry.revision)
2107
else:
2108
parent_text_key = None
2109
if parent_text_key is not None:
2110
candidate_parents.append(
2111
text_key_cache[parent_text_key])
2112
parent_heads = text_graph.heads(candidate_parents)
2113
new_parents = list(parent_heads)
2114
new_parents.sort(key=lambda x:candidate_parents.index(x))
2115
if new_parents == []:
2116
new_parents = [NULL_REVISION]
2117
text_index[text_key] = new_parents
2118
2119
for text_key in invalid_keys:
2120
text_index[text_key] = [NULL_REVISION]
2121
return text_index
2122
2123
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2124
"""Get an iterable listing the keys of all the data introduced by a set
2125
of revision IDs.
2126
2127
The keys will be ordered so that the corresponding items can be safely
2128
fetched and inserted in that order.
2129
2130
:returns: An iterable producing tuples of (knit-kind, file-id,
2131
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2132
'revisions'. file-id is None unless knit-kind is 'file'.
2133
"""
2134
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2135
yield result
2136
del _files_pb
2137
for result in self._find_non_file_keys_to_fetch(revision_ids):
2138
yield result
2139
2140
def _find_file_keys_to_fetch(self, revision_ids, pb):
2141
# XXX: it's a bit weird to control the inventory weave caching in this
2142
# generator. Ideally the caching would be done in fetch.py I think. Or
2143
# maybe this generator should explicitly have the contract that it
2144
# should not be iterated until the previously yielded item has been
2145
# processed?
2146
inv_w = self.inventories
2147
2148
# file ids that changed
2149
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2150
count = 0
2151
num_file_ids = len(file_ids)
2152
for file_id, altered_versions in file_ids.iteritems():
2153
if pb is not None:
2154
pb.update("fetch texts", count, num_file_ids)
2155
count += 1
2156
yield ("file", file_id, altered_versions)
2157
2158
def _find_non_file_keys_to_fetch(self, revision_ids):
2159
# inventory
2160
yield ("inventory", None, revision_ids)
2161
2162
# signatures
2163
# XXX: Note ATM no callers actually pay attention to this return
2164
# instead they just use the list of revision ids and ignore
2165
# missing sigs. Consider removing this work entirely
2166
revisions_with_signatures = set(self.signatures.get_parent_map(
2167
[(r,) for r in revision_ids]))
2168
revisions_with_signatures = set(
2169
[r for (r,) in revisions_with_signatures])
2170
revisions_with_signatures.intersection_update(revision_ids)
2171
yield ("signatures", None, revisions_with_signatures)
2172
2173
# revisions
2174
yield ("revisions", None, revision_ids)
579
pb = ui.ui_factory.nested_progress_bar()
580
try:
581
for line in w.iter_lines_added_or_present_in_versions(
582
selected_revision_ids, pb=pb):
583
match = search(line)
584
if match is None:
585
continue
586
# One call to match.group() returning multiple items is quite a
587
# bit faster than 2 calls to match.group() each returning 1
588
file_id, revision_id = match.group('file_id', 'revision_id')
589
590
# Inlining the cache lookups helps a lot when you make 170,000
591
# lines and 350k ids, versus 8.4 unique ids.
592
# Using a cache helps in 2 ways:
593
# 1) Avoids unnecessary decoding calls
594
# 2) Re-uses cached strings, which helps in future set and
595
# equality checks.
596
# (2) is enough that removing encoding entirely along with
597
# the cache (so we are using plain strings) results in no
598
# performance improvement.
599
try:
600
revision_id = unescape_revid_cache[revision_id]
601
except KeyError:
602
unescaped = unescape(revision_id)
603
unescape_revid_cache[revision_id] = unescaped
604
revision_id = unescaped
605
606
if revision_id in selected_revision_ids:
607
try:
608
file_id = unescape_fileid_cache[file_id]
609
except KeyError:
610
unescaped = unescape(file_id)
611
unescape_fileid_cache[file_id] = unescaped
612
file_id = unescaped
613
setdefault(file_id, set()).add(revision_id)
614
finally:
615
pb.finished()
616
return result
617
618
@needs_read_lock
619
def get_inventory_weave(self):
620
return self.control_weaves.get_weave('inventory',
621
self.get_transaction())
2175
622
2176
623
@needs_read_lock
2177
624
def get_inventory(self, revision_id):
2178
"""Get Inventory object by revision id."""
2179
return self.iter_inventories([revision_id]).next()
2180
2181
def iter_inventories(self, revision_ids):
2182
"""Get many inventories by revision_ids.
2183
2184
This will buffer some or all of the texts used in constructing the
2185
inventories in memory, but will only parse a single inventory at a
2186
time.
2187
2188
:param revision_ids: The expected revision ids of the inventories.
2189
:return: An iterator of inventories.
2190
"""
2191
if ((None in revision_ids)
2192
or (_mod_revision.NULL_REVISION in revision_ids)):
2193
raise ValueError('cannot get null revision inventory')
2194
return self._iter_inventories(revision_ids)
2195
2196
def _iter_inventories(self, revision_ids):
2197
"""single-document based inventory iteration."""
2198
for text, revision_id in self._iter_inventory_xmls(revision_ids):
2199
yield self.deserialise_inventory(revision_id, text)
2200
2201
def _iter_inventory_xmls(self, revision_ids):
2202
keys = [(revision_id,) for revision_id in revision_ids]
2203
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2204
text_chunks = {}
2205
for record in stream:
2206
if record.storage_kind != 'absent':
2207
text_chunks[record.key] = record.get_bytes_as('chunked')
2208
else:
2209
raise errors.NoSuchRevision(self, record.key)
2210
for key in keys:
2211
chunks = text_chunks.pop(key)
2212
yield ''.join(chunks), key[-1]
625
"""Get Inventory object by hash."""
626
# TODO: jam 20070210 Technically we don't need to sanitize, since all
627
# called functions must sanitize.
628
revision_id = osutils.safe_revision_id(revision_id)
629
return self.deserialise_inventory(
630
revision_id, self.get_inventory_xml(revision_id))
2213
631
2214
632
def deserialise_inventory(self, revision_id, xml):
2215
"""Transform the xml into an inventory object.
633
"""Transform the xml into an inventory object.
2216
634
2217
635
:param revision_id: The expected revision id of the inventory.
2218
636
:param xml: A serialised inventory.
2219
637
"""
2220
result = self._serializer.read_inventory_from_string(xml, revision_id,
2221
entry_cache=self._inventory_entry_cache)
2222
if result.revision_id != revision_id:
2223
raise AssertionError('revision id mismatch %s != %s' % (
2224
result.revision_id, revision_id))
638
revision_id = osutils.safe_revision_id(revision_id)
639
result = self._serializer.read_inventory_from_string(xml)
640
result.root.revision = revision_id
2225
641
return result
2226
642
2227
643
def serialise_inventory(self, inv):
2228
644
return self._serializer.write_inventory_to_string(inv)
2229
645
2230
def _serialise_inventory_to_lines(self, inv):
2231
return self._serializer.write_inventory_to_lines(inv)
2232
2233
def get_serializer_format(self):
2234
return self._serializer.format_num
2235
2236
646
@needs_read_lock
2237
647
def get_inventory_xml(self, revision_id):
2238
648
"""Get inventory XML as a file object."""
2239
texts = self._iter_inventory_xmls([revision_id])
649
revision_id = osutils.safe_revision_id(revision_id)
2240
650
try:
2241
text, revision_id = texts.next()
2242
except StopIteration:
651
assert isinstance(revision_id, str), type(revision_id)
652
iw = self.get_inventory_weave()
653
return iw.get_text(revision_id)
654
except IndexError:
2243
655
raise errors.HistoryMissing(self, 'inventory', revision_id)
2244
return text
2245
656
2246
657
@needs_read_lock
2247
658
def get_inventory_sha1(self, revision_id):
2248
659
"""Return the sha1 hash of the inventory entry
2249
660
"""
661
# TODO: jam 20070210 Shouldn't this be deprecated / removed?
662
revision_id = osutils.safe_revision_id(revision_id)
2250
663
return self.get_revision(revision_id).inventory_sha1
2251
664
2252
def get_rev_id_for_revno(self, revno, known_pair):
2253
"""Return the revision id of a revno, given a later (revno, revid)
2254
pair in the same history.
2255
2256
:return: if found (True, revid). If the available history ran out
2257
before reaching the revno, then this returns
2258
(False, (closest_revno, closest_revid)).
2259
"""
2260
known_revno, known_revid = known_pair
2261
partial_history = [known_revid]
2262
distance_from_known = known_revno - revno
2263
if distance_from_known < 0:
2264
raise ValueError(
2265
'requested revno (%d) is later than given known revno (%d)'
2266
% (revno, known_revno))
2267
try:
2268
_iter_for_revno(
2269
self, partial_history, stop_index=distance_from_known)
2270
except errors.RevisionNotPresent, err:
2271
if err.revision_id == known_revid:
2272
# The start revision (known_revid) wasn't found.
2273
raise
2274
# This is a stacked repository with no fallbacks, or a there's a
2275
# left-hand ghost. Either way, even though the revision named in
2276
# the error isn't in this repo, we know it's the next step in this
2277
# left-hand history.
2278
partial_history.append(err.revision_id)
2279
if len(partial_history) <= distance_from_known:
2280
# Didn't find enough history to get a revid for the revno.
2281
earliest_revno = known_revno - len(partial_history) + 1
2282
return (False, (earliest_revno, partial_history[-1]))
2283
if len(partial_history) - 1 > distance_from_known:
2284
raise AssertionError('_iter_for_revno returned too much history')
2285
return (True, partial_history[-1])
665
@needs_read_lock
666
def get_revision_graph(self, revision_id=None):
667
"""Return a dictionary containing the revision graph.
668
669
:param revision_id: The revision_id to get a graph from. If None, then
670
the entire revision graph is returned. This is a deprecated mode of
671
operation and will be removed in the future.
672
:return: a dictionary of revision_id->revision_parents_list.
673
"""
674
# special case NULL_REVISION
675
if revision_id == _mod_revision.NULL_REVISION:
676
return {}
677
revision_id = osutils.safe_revision_id(revision_id)
678
a_weave = self.get_inventory_weave()
679
all_revisions = self._eliminate_revisions_not_present(
680
a_weave.versions())
681
entire_graph = dict([(node, a_weave.get_parents(node)) for
682
node in all_revisions])
683
if revision_id is None:
684
return entire_graph
685
elif revision_id not in entire_graph:
686
raise errors.NoSuchRevision(self, revision_id)
687
else:
688
# add what can be reached from revision_id
689
result = {}
690
pending = set([revision_id])
691
while len(pending) > 0:
692
node = pending.pop()
693
result[node] = entire_graph[node]
694
for revision_id in result[node]:
695
if revision_id not in result:
696
pending.add(revision_id)
697
return result
698
699
@needs_read_lock
700
def get_revision_graph_with_ghosts(self, revision_ids=None):
701
"""Return a graph of the revisions with ghosts marked as applicable.
702
703
:param revision_ids: an iterable of revisions to graph or None for all.
704
:return: a Graph object with the graph reachable from revision_ids.
705
"""
706
result = deprecated_graph.Graph()
707
if not revision_ids:
708
pending = set(self.all_revision_ids())
709
required = set([])
710
else:
711
pending = set(osutils.safe_revision_id(r) for r in revision_ids)
712
# special case NULL_REVISION
713
if _mod_revision.NULL_REVISION in pending:
714
pending.remove(_mod_revision.NULL_REVISION)
715
required = set(pending)
716
done = set([])
717
while len(pending):
718
revision_id = pending.pop()
719
try:
720
rev = self.get_revision(revision_id)
721
except errors.NoSuchRevision:
722
if revision_id in required:
723
raise
724
# a ghost
725
result.add_ghost(revision_id)
726
continue
727
for parent_id in rev.parent_ids:
728
# is this queued or done ?
729
if (parent_id not in pending and
730
parent_id not in done):
731
# no, queue it.
732
pending.add(parent_id)
733
result.add_node(revision_id, rev.parent_ids)
734
done.add(revision_id)
735
return result
736
737
def _get_history_vf(self):
738
"""Get a versionedfile whose history graph reflects all revisions.
739
740
For weave repositories, this is the inventory weave.
741
"""
742
return self.get_inventory_weave()
2286
743
2287
744
def iter_reverse_revision_history(self, revision_id):
2288
745
"""Iterate backwards through revision ids in the lefthand history
2290
747
:param revision_id: The revision id to start with. All its lefthand
2291
748
ancestors will be traversed.
2292
749
"""
2293
graph = self.get_graph()
750
revision_id = osutils.safe_revision_id(revision_id)
751
if revision_id in (None, _mod_revision.NULL_REVISION):
752
return
2294
753
next_id = revision_id
754
versionedfile = self._get_history_vf()
2295
755
while True:
2296
if next_id in (None, _mod_revision.NULL_REVISION):
2297
return
2298
try:
2299
parents = graph.get_parent_map([next_id])[next_id]
2300
except KeyError:
2301
raise errors.RevisionNotPresent(next_id, self)
2302
756
yield next_id
757
parents = versionedfile.get_parents(next_id)
2303
758
if len(parents) == 0:
2304
759
return
2305
760
else:
2321
776
else:
2322
777
return self.get_inventory(revision_id)
2323
778
779
@needs_read_lock
2324
780
def is_shared(self):
2325
781
"""Return True if this repository is flagged as a shared repository."""
2326
782
raise NotImplementedError(self.is_shared)
2333
789
reconciler.reconcile()
2334
790
return reconciler
2335
791
2336
def _refresh_data(self):
2337
"""Helper called from lock_* to ensure coherency with disk.
2338
2339
The default implementation does nothing; it is however possible
2340
for repositories to maintain loaded indices across multiple locks
2341
by checking inside their implementation of this method to see
2342
whether their indices are still valid. This depends of course on
2343
the disk format being validatable in this manner. This method is
2344
also called by the refresh_data() public interface to cause a refresh
2345
to occur while in a write lock so that data inserted by a smart server
2346
push operation is visible on the client's instance of the physical
2347
repository.
2348
"""
2349
2350
792
@needs_read_lock
2351
793
def revision_tree(self, revision_id):
2352
794
"""Return Tree for a revision on this branch.
2353
795
2354
`revision_id` may be NULL_REVISION for the empty tree revision.
796
`revision_id` may be None for the empty tree revision.
2355
797
"""
2356
revision_id = _mod_revision.ensure_null(revision_id)
2357
798
# TODO: refactor this to use an existing revision object
2358
799
# so we don't need to read it in twice.
2359
if revision_id == _mod_revision.NULL_REVISION:
2360
return RevisionTree(self, Inventory(root_id=None),
800
if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
801
return RevisionTree(self, Inventory(root_id=None),
2361
802
_mod_revision.NULL_REVISION)
2362
803
else:
804
revision_id = osutils.safe_revision_id(revision_id)
2363
805
inv = self.get_revision_inventory(revision_id)
2364
806
return RevisionTree(self, inv, revision_id)
2365
807
808
@needs_read_lock
2366
809
def revision_trees(self, revision_ids):
2367
"""Return Trees for revisions in this repository.
2368
2369
:param revision_ids: a sequence of revision-ids;
2370
a revision-id may not be None or 'null:'
2371
"""
2372
inventories = self.iter_inventories(revision_ids)
2373
for inv in inventories:
2374
yield RevisionTree(self, inv, inv.revision_id)
2375
2376
def _filtered_revision_trees(self, revision_ids, file_ids):
2377
"""Return Tree for a revision on this branch with only some files.
2378
2379
:param revision_ids: a sequence of revision-ids;
2380
a revision-id may not be None or 'null:'
2381
:param file_ids: if not None, the result is filtered
2382
so that only those file-ids, their parents and their
2383
children are included.
2384
"""
2385
inventories = self.iter_inventories(revision_ids)
2386
for inv in inventories:
2387
# Should we introduce a FilteredRevisionTree class rather
2388
# than pre-filter the inventory here?
2389
filtered_inv = inv.filter(file_ids)
2390
yield RevisionTree(self, filtered_inv, filtered_inv.revision_id)
810
"""Return Tree for a revision on this branch.
811
812
`revision_id` may not be None or 'null:'"""
813
assert None not in revision_ids
814
assert _mod_revision.NULL_REVISION not in revision_ids
815
texts = self.get_inventory_weave().get_texts(revision_ids)
816
for text, revision_id in zip(texts, revision_ids):
817
inv = self.deserialise_inventory(revision_id, text)
818
yield RevisionTree(self, inv, revision_id)
2391
819
2392
820
@needs_read_lock
2393
821
def get_ancestry(self, revision_id, topo_sorted=True):
2394
822
"""Return a list of revision-ids integrated by a revision.
2395
823
2396
The first element of the list is always None, indicating the origin
2397
revision. This might change when we have history horizons, or
824
The first element of the list is always None, indicating the origin
825
revision. This might change when we have history horizons, or
2398
826
perhaps we should have a new API.
2399
827
2400
828
This is topologically sorted.
2401
829
"""
2402
if _mod_revision.is_null(revision_id):
830
if revision_id is None:
2403
831
return [None]
832
revision_id = osutils.safe_revision_id(revision_id)
2404
833
if not self.has_revision(revision_id):
2405
834
raise errors.NoSuchRevision(self, revision_id)
2406
graph = self.get_graph()
2407
keys = set()
2408
search = graph._make_breadth_first_searcher([revision_id])
2409
while True:
2410
try:
2411
found, ghosts = search.next_with_ghosts()
2412
except StopIteration:
2413
break
2414
keys.update(found)
2415
if _mod_revision.NULL_REVISION in keys:
2416
keys.remove(_mod_revision.NULL_REVISION)
2417
if topo_sorted:
2418
parent_map = graph.get_parent_map(keys)
2419
keys = tsort.topo_sort(parent_map)
2420
return [None] + list(keys)
2421
2422
def pack(self, hint=None):
2423
"""Compress the data within the repository.
2424
2425
This operation only makes sense for some repository types. For other
2426
types it should be a no-op that just returns.
2427
2428
This stub method does not require a lock, but subclasses should use
2429
@needs_write_lock as this is a long running call its reasonable to
2430
implicitly lock for the user.
2431
2432
:param hint: If not supplied, the whole repository is packed.
2433
If supplied, the repository may use the hint parameter as a
2434
hint for the parts of the repository to pack. A hint can be
2435
obtained from the result of commit_write_group(). Out of
2436
date hints are simply ignored, because concurrent operations
2437
can obsolete them rapidly.
835
w = self.get_inventory_weave()
836
candidates = w.get_ancestry(revision_id, topo_sorted)
837
return [None] + candidates # self._eliminate_revisions_not_present(candidates)
838
839
@needs_read_lock
840
def print_file(self, file, revision_id):
841
"""Print `file` to stdout.
842
843
FIXME RBC 20060125 as John Meinel points out this is a bad api
844
- it writes to stdout, it assumes that that is valid etc. Fix
845
by creating a new more flexible convenience function.
2438
846
"""
847
revision_id = osutils.safe_revision_id(revision_id)
848
tree = self.revision_tree(revision_id)
849
# use inventory as it was in that revision
850
file_id = tree.inventory.path2id(file)
851
if not file_id:
852
# TODO: jam 20060427 Write a test for this code path
853
# it had a bug in it, and was raising the wrong
854
# exception.
855
raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
856
tree.print_file(file_id)
2439
857
2440
858
def get_transaction(self):
2441
859
return self.control_files.get_transaction()
2442
860
2443
def get_parent_map(self, revision_ids):
2444
"""See graph.StackedParentsProvider.get_parent_map"""
2445
# revisions index works in keys; this just works in revisions
2446
# therefore wrap and unwrap
2447
query_keys = []
2448
result = {}
861
def revision_parents(self, revision_id):
862
revision_id = osutils.safe_revision_id(revision_id)
863
return self.get_inventory_weave().parent_names(revision_id)
864
865
def get_parents(self, revision_ids):
866
"""See StackedParentsProvider.get_parents"""
867
parents_list = []
2449
868
for revision_id in revision_ids:
2450
869
if revision_id == _mod_revision.NULL_REVISION:
2451
result[revision_id] = ()
2452
elif revision_id is None:
2453
raise ValueError('get_parent_map(None) is not valid')
2454
else:
2455
query_keys.append((revision_id ,))
2456
for ((revision_id,), parent_keys) in \
2457
self.revisions.get_parent_map(query_keys).iteritems():
2458
if parent_keys:
2459
result[revision_id] = tuple(parent_revid
2460
for (parent_revid,) in parent_keys)
2461
else:
2462
result[revision_id] = (_mod_revision.NULL_REVISION,)
2463
return result
870
parents = []
871
else:
872
try:
873
parents = self.get_revision(revision_id).parent_ids
874
except errors.NoSuchRevision:
875
parents = None
876
else:
877
if len(parents) == 0:
878
parents = [_mod_revision.NULL_REVISION]
879
parents_list.append(parents)
880
return parents_list
2464
881
2465
882
def _make_parents_provider(self):
2466
883
return self
2469
886
"""Return the graph walker for this repository format"""
2470
887
parents_provider = self._make_parents_provider()
2471
888
if (other_repository is not None and
2472
not self.has_same_location(other_repository)):
2473
parents_provider = graph.StackedParentsProvider(
889
other_repository.bzrdir.transport.base !=
890
self.bzrdir.transport.base):
891
parents_provider = graph._StackedParentsProvider(
2474
892
[parents_provider, other_repository._make_parents_provider()])
2475
893
return graph.Graph(parents_provider)
2476
894
2477
def _get_versioned_file_checker(self, text_key_references=None):
2478
"""Return an object suitable for checking versioned files.
2479
2480
:param text_key_references: if non-None, an already built
2481
dictionary mapping text keys ((fileid, revision_id) tuples)
2482
to whether they were referred to by the inventory of the
2483
revision_id that they contain. If None, this will be
2484
calculated.
2485
"""
2486
return _VersionedFileChecker(self,
2487
text_key_references=text_key_references)
2488
2489
def revision_ids_to_search_result(self, result_set):
2490
"""Convert a set of revision ids to a graph SearchResult."""
2491
result_parents = set()
2492
for parents in self.get_graph().get_parent_map(
2493
result_set).itervalues():
2494
result_parents.update(parents)
2495
included_keys = result_set.intersection(result_parents)
2496
start_keys = result_set.difference(included_keys)
2497
exclude_keys = result_parents.difference(result_set)
2498
result = graph.SearchResult(start_keys, exclude_keys,
2499
len(result_set), result_set)
2500
return result
2501
2502
895
@needs_write_lock
2503
896
def set_make_working_trees(self, new_value):
2504
897
"""Set the policy flag for making working trees when creating branches.
2510
903
working trees.
2511
904
"""
2512
905
raise NotImplementedError(self.set_make_working_trees)
2513
906
2514
907
def make_working_trees(self):
2515
908
"""Returns the policy for making working trees on new branches."""
2516
909
raise NotImplementedError(self.make_working_trees)
2517
910
2518
911
@needs_write_lock
2519
912
def sign_revision(self, revision_id, gpg_strategy):
913
revision_id = osutils.safe_revision_id(revision_id)
2520
914
plaintext = Testament.from_revision(self, revision_id).as_short_text()
2521
915
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
2522
916
2523
917
@needs_read_lock
2524
918
def has_signature_for_revision_id(self, revision_id):
2525
919
"""Query for a revision signature for revision_id in the repository."""
2526
if not self.has_revision(revision_id):
2527
raise errors.NoSuchRevision(self, revision_id)
2528
sig_present = (1 == len(
2529
self.signatures.get_parent_map([(revision_id,)])))
2530
return sig_present
920
revision_id = osutils.safe_revision_id(revision_id)
921
return self._revision_store.has_signature(revision_id,
922
self.get_transaction())
2531
923
2532
924
@needs_read_lock
2533
925
def get_signature_text(self, revision_id):
2534
926
"""Return the text for a signature."""
2535
stream = self.signatures.get_record_stream([(revision_id,)],
2536
'unordered', True)
2537
record = stream.next()
2538
if record.storage_kind == 'absent':
2539
raise errors.NoSuchRevision(self, revision_id)
2540
return record.get_bytes_as('fulltext')
927
revision_id = osutils.safe_revision_id(revision_id)
928
return self._revision_store.get_signature_text(revision_id,
929
self.get_transaction())
2541
930
2542
931
@needs_read_lock
2543
def check(self, revision_ids=None):
932
def check(self, revision_ids):
2544
933
"""Check consistency of all history of given revision_ids.
2545
934
2546
935
Different repository implementations should override _check().
2548
937
:param revision_ids: A non-empty list of revision_ids whose ancestry
2549
938
will be checked. Typically the last revision_id of a branch.
2550
939
"""
940
if not revision_ids:
941
raise ValueError("revision_ids must be non-empty in %s.check"
942
% (self,))
943
revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
2551
944
return self._check(revision_ids)
2552
945
2553
946
def _check(self, revision_ids):
2582
975
except UnicodeDecodeError:
2583
976
raise errors.NonAsciiRevisionId(method, self)
2584
977
2585
def revision_graph_can_have_wrong_parents(self):
2586
"""Is it possible for this repository to have a revision graph with
2587
incorrect parents?
2588
2589
If True, then this repository must also implement
2590
_find_inconsistent_revision_parents so that check and reconcile can
2591
check for inconsistencies before proceeding with other checks that may
2592
depend on the revision index being consistent.
2593
"""
2594
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2595
978
2596
979
2597
980
# remove these delegates a while after bzr 0.15
2630
1013
2631
1014
def install_revision(repository, rev, revision_tree):
2632
1015
"""Install all revision data into a repository."""
2633
install_revisions(repository, [(rev, revision_tree, None)])
2634
2635
2636
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2637
"""Install all revision data into a repository.
2638
2639
Accepts an iterable of revision, tree, signature tuples. The signature
2640
may be None.
2641
"""
2642
repository.start_write_group()
2643
try:
2644
inventory_cache = lru_cache.LRUCache(10)
2645
for n, (revision, revision_tree, signature) in enumerate(iterable):
2646
_install_revision(repository, revision, revision_tree, signature,
2647
inventory_cache)
2648
if pb is not None:
2649
pb.update('Transferring revisions', n + 1, num_revisions)
2650
except:
2651
repository.abort_write_group()
2652
raise
2653
else:
2654
repository.commit_write_group()
2655
2656
2657
def _install_revision(repository, rev, revision_tree, signature,
2658
inventory_cache):
2659
"""Install all revision data into a repository."""
2660
1016
present_parents = []
2661
1017
parent_trees = {}
2662
1018
for p_id in rev.parent_ids:
2664
1020
present_parents.append(p_id)
2665
1021
parent_trees[p_id] = repository.revision_tree(p_id)
2666
1022
else:
2667
parent_trees[p_id] = repository.revision_tree(
2668
_mod_revision.NULL_REVISION)
1023
parent_trees[p_id] = repository.revision_tree(None)
2669
1024
2670
1025
inv = revision_tree.inventory
2671
1026
entries = inv.iter_entries()
2672
# backwards compatibility hack: skip the root id.
1027
# backwards compatability hack: skip the root id.
2673
1028
if not repository.supports_rich_root():
2674
1029
path, root = entries.next()
2675
1030
if root.revision != rev.revision_id:
2676
1031
raise errors.IncompatibleRevision(repr(repository))
2677
text_keys = {}
1032
# Add the texts that are not already present
2678
1033
for path, ie in entries:
2679
text_keys[(ie.file_id, ie.revision)] = ie
2680
text_parent_map = repository.texts.get_parent_map(text_keys)
2681
missing_texts = set(text_keys) - set(text_parent_map)
2682
# Add the texts that are not already present
2683
for text_key in missing_texts:
2684
ie = text_keys[text_key]
2685
text_parents = []
2686
# FIXME: TODO: The following loop overlaps/duplicates that done by
2687
# commit to determine parents. There is a latent/real bug here where
2688
# the parents inserted are not those commit would do - in particular
2689
# they are not filtered by heads(). RBC, AB
2690
for revision, tree in parent_trees.iteritems():
2691
if ie.file_id not in tree:
2692
continue
2693
parent_id = tree.inventory[ie.file_id].revision
2694
if parent_id in text_parents:
2695
continue
2696
text_parents.append((ie.file_id, parent_id))
2697
lines = revision_tree.get_file(ie.file_id).readlines()
2698
repository.texts.add_lines(text_key, text_parents, lines)
1034
w = repository.weave_store.get_weave_or_empty(ie.file_id,
1035
repository.get_transaction())
1036
if ie.revision not in w:
1037
text_parents = []
1038
# FIXME: TODO: The following loop *may* be overlapping/duplicate
1039
# with InventoryEntry.find_previous_heads(). if it is, then there
1040
# is a latent bug here where the parents may have ancestors of each
1041
# other. RBC, AB
1042
for revision, tree in parent_trees.iteritems():
1043
if ie.file_id not in tree:
1044
continue
1045
parent_id = tree.inventory[ie.file_id].revision
1046
if parent_id in text_parents:
1047
continue
1048
text_parents.append(parent_id)
1049
1050
vfile = repository.weave_store.get_weave_or_empty(ie.file_id,
1051
repository.get_transaction())
1052
lines = revision_tree.get_file(ie.file_id).readlines()
1053
vfile.add_lines(rev.revision_id, text_parents, lines)
2699
1054
try:
2700
1055
# install the inventory
2701
if repository._format._commit_inv_deltas and len(rev.parent_ids):
2702
# Cache this inventory
2703
inventory_cache[rev.revision_id] = inv
2704
try:
2705
basis_inv = inventory_cache[rev.parent_ids[0]]
2706
except KeyError:
2707
repository.add_inventory(rev.revision_id, inv, present_parents)
2708
else:
2709
delta = inv._make_delta(basis_inv)
2710
repository.add_inventory_by_delta(rev.parent_ids[0], delta,
2711
rev.revision_id, present_parents)
2712
else:
2713
repository.add_inventory(rev.revision_id, inv, present_parents)
1056
repository.add_inventory(rev.revision_id, inv, present_parents)
2714
1057
except errors.RevisionAlreadyPresent:
2715
1058
pass
2716
if signature is not None:
2717
repository.add_signature_text(rev.revision_id, signature)
2718
1059
repository.add_revision(rev.revision_id, rev, inv)
2719
1060
2720
1061
2721
1062
class MetaDirRepository(Repository):
2722
"""Repositories in the new meta-dir layout.
2723
2724
:ivar _transport: Transport for access to repository control files,
2725
typically pointing to .bzr/repository.
2726
"""
2727
2728
def __init__(self, _format, a_bzrdir, control_files):
2729
super(MetaDirRepository, self).__init__(_format, a_bzrdir, control_files)
2730
self._transport = control_files._transport
2731
1063
"""Repositories in the new meta-dir layout."""
1064
1065
def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1066
super(MetaDirRepository, self).__init__(_format,
1067
a_bzrdir,
1068
control_files,
1069
_revision_store,
1070
control_store,
1071
text_store)
1072
dir_mode = self.control_files._dir_mode
1073
file_mode = self.control_files._file_mode
1074
1075
@needs_read_lock
2732
1076
def is_shared(self):
2733
1077
"""Return True if this repository is flagged as a shared repository."""
2734
return self._transport.has('shared-storage')
1078
return self.control_files._transport.has('shared-storage')
2735
1079
2736
1080
@needs_write_lock
2737
1081
def set_make_working_trees(self, new_value):
2745
1089
"""
2746
1090
if new_value:
2747
1091
try:
2748
self._transport.delete('no-working-trees')
1092
self.control_files._transport.delete('no-working-trees')
2749
1093
except errors.NoSuchFile:
2750
1094
pass
2751
1095
else:
2752
self._transport.put_bytes('no-working-trees', '',
2753
mode=self.bzrdir._get_file_mode())
2754
1096
self.control_files.put_utf8('no-working-trees', '')
1097
2755
1098
def make_working_trees(self):
2756
1099
"""Returns the policy for making working trees on new branches."""
2757
return not self._transport.has('no-working-trees')
2758
2759
2760
class MetaDirVersionedFileRepository(MetaDirRepository):
2761
"""Repositories in a meta-dir, that work via versioned file objects."""
2762
2763
def __init__(self, _format, a_bzrdir, control_files):
2764
super(MetaDirVersionedFileRepository, self).__init__(_format, a_bzrdir,
2765
control_files)
2766
2767
2768
network_format_registry = registry.FormatRegistry()
2769
"""Registry of formats indexed by their network name.
2770
2771
The network name for a repository format is an identifier that can be used when
2772
referring to formats with smart server operations. See
2773
RepositoryFormat.network_name() for more detail.
2774
"""
2775
2776
2777
format_registry = registry.FormatRegistry(network_format_registry)
2778
"""Registry of formats, indexed by their BzrDirMetaFormat format string.
1100
return not self.control_files._transport.has('no-working-trees')
1101
1102
1103
class RepositoryFormatRegistry(registry.Registry):
1104
"""Registry of RepositoryFormats.
1105
"""
1106
1107
def get(self, format_string):
1108
r = registry.Registry.get(self, format_string)
1109
if callable(r):
1110
r = r()
1111
return r
1112
1113
1114
format_registry = RepositoryFormatRegistry()
1115
"""Registry of formats, indexed by their identifying format string.
2779
1116
2780
1117
This can contain either format instances themselves, or classes/factories that
2781
1118
can be called to obtain one.
2788
1125
class RepositoryFormat(object):
2789
1126
"""A repository format.
2790
1127
2791
Formats provide four things:
1128
Formats provide three things:
2792
1129
* An initialization routine to construct repository data on disk.
2793
* a optional format string which is used when the BzrDir supports
2794
versioned children.
1130
* a format string which is used when the BzrDir supports versioned
1131
children.
2795
1132
* an open routine which returns a Repository instance.
2796
* A network name for referring to the format in smart server RPC
2797
methods.
2798
2799
There is one and only one Format subclass for each on-disk format. But
2800
there can be one Repository subclass that is used for several different
2801
formats. The _format attribute on a Repository instance can be used to
2802
determine the disk format.
2803
2804
Formats are placed in a registry by their format string for reference
2805
during opening. These should be subclasses of RepositoryFormat for
2806
consistency.
1133
1134
Formats are placed in an dict by their format string for reference
1135
during opening. These should be subclasses of RepositoryFormat
1136
for consistency.
2807
1137
2808
1138
Once a format is deprecated, just deprecate the initialize and open
2809
methods on the format class. Do not deprecate the object, as the
2810
object may be created even when a repository instance hasn't been
2811
created.
1139
methods on the format class. Do not deprecate the object, as the
1140
object will be created every system load.
2812
1141
2813
1142
Common instance attributes:
2814
1143
_matchingbzrdir - the bzrdir format that the repository format was
2815
1144
originally written to work with. This can be used if manually
2816
1145
constructing a bzrdir and repository, or more commonly for test suite
2817
parameterization.
1146
parameterisation.
2818
1147
"""
2819
1148
2820
# Set to True or False in derived classes. True indicates that the format
2821
# supports ghosts gracefully.
2822
supports_ghosts = None
2823
# Can this repository be given external locations to lookup additional
2824
# data. Set to True or False in derived classes.
2825
supports_external_lookups = None
2826
# Does this format support CHK bytestring lookups. Set to True or False in
2827
# derived classes.
2828
supports_chks = None
2829
# Should commit add an inventory, or an inventory delta to the repository.
2830
_commit_inv_deltas = True
2831
# What order should fetch operations request streams in?
2832
# The default is unordered as that is the cheapest for an origin to
2833
# provide.
2834
_fetch_order = 'unordered'
2835
# Does this repository format use deltas that can be fetched as-deltas ?
2836
# (E.g. knits, where the knit deltas can be transplanted intact.
2837
# We default to False, which will ensure that enough data to get
2838
# a full text out of any fetch stream will be grabbed.
2839
_fetch_uses_deltas = False
2840
# Should fetch trigger a reconcile after the fetch? Only needed for
2841
# some repository formats that can suffer internal inconsistencies.
2842
_fetch_reconcile = False
2843
# Does this format have < O(tree_size) delta generation. Used to hint what
2844
# code path for commit, amongst other things.
2845
fast_deltas = None
2846
# Does doing a pack operation compress data? Useful for the pack UI command
2847
# (so if there is one pack, the operation can still proceed because it may
2848
# help), and for fetching when data won't have come from the same
2849
# compressor.
2850
pack_compresses = False
2851
2852
1149
def __str__(self):
2853
1150
return "<%s>" % self.__class__.__name__
2854
1151
2862
1159
@classmethod
2863
1160
def find_format(klass, a_bzrdir):
2864
1161
"""Return the format for the repository object in a_bzrdir.
2865
1162
2866
1163
This is used by bzr native formats that have a "format" file in
2867
the repository. Other methods may be used by different types of
1164
the repository. Other methods may be used by different types of
2868
1165
control directory.
2869
1166
"""
2870
1167
try:
2874
1171
except errors.NoSuchFile:
2875
1172
raise errors.NoRepositoryPresent(a_bzrdir)
2876
1173
except KeyError:
2877
raise errors.UnknownFormatError(format=format_string,
2878
kind='repository')
1174
raise errors.UnknownFormatError(format=format_string)
2879
1175
2880
1176
@classmethod
2881
1177
def register_format(klass, format):
2884
1180
@classmethod
2885
1181
def unregister_format(klass, format):
2886
1182
format_registry.remove(format.get_format_string())
2887
1183
2888
1184
@classmethod
2889
1185
def get_default_format(klass):
2890
1186
"""Return the current default format."""
2891
1187
from bzrlib import bzrdir
2892
1188
return bzrdir.format_registry.make_bzrdir('default').repository_format
2893
1189
1190
def _get_control_store(self, repo_transport, control_files):
1191
"""Return the control store for this repository."""
1192
raise NotImplementedError(self._get_control_store)
1193
2894
1194
def get_format_string(self):
2895
1195
"""Return the ASCII format string that identifies this format.
2896
2897
Note that in pre format ?? repositories the format string is
1196
1197
Note that in pre format ?? repositories the format string is
2898
1198
not permitted nor written to disk.
2899
1199
"""
2900
1200
raise NotImplementedError(self.get_format_string)
2903
1203
"""Return the short description for this format."""
2904
1204
raise NotImplementedError(self.get_format_description)
2905
1205
1206
def _get_revision_store(self, repo_transport, control_files):
1207
"""Return the revision store object for this a_bzrdir."""
1208
raise NotImplementedError(self._get_revision_store)
1209
1210
def _get_text_rev_store(self,
1211
transport,
1212
control_files,
1213
name,
1214
compressed=True,
1215
prefixed=False,
1216
serializer=None):
1217
"""Common logic for getting a revision store for a repository.
1218
1219
see self._get_revision_store for the subclass-overridable method to
1220
get the store for a repository.
1221
"""
1222
from bzrlib.store.revision.text import TextRevisionStore
1223
dir_mode = control_files._dir_mode
1224
file_mode = control_files._file_mode
1225
text_store = TextStore(transport.clone(name),
1226
prefixed=prefixed,
1227
compressed=compressed,
1228
dir_mode=dir_mode,
1229
file_mode=file_mode)
1230
_revision_store = TextRevisionStore(text_store, serializer)
1231
return _revision_store
1232
2906
1233
# TODO: this shouldn't be in the base class, it's specific to things that
2907
1234
# use weaves or knits -- mbp 20070207
2908
1235
def _get_versioned_file_store(self,
2931
1258
:param a_bzrdir: The bzrdir to put the new repository in it.
2932
1259
:param shared: The repository should be initialized as a sharable one.
2933
1260
:returns: The new repository object.
2934
1261
2935
1262
This may raise UninitializableFormat if shared repository are not
2936
1263
compatible the a_bzrdir.
2937
1264
"""
2941
1268
"""Is this format supported?
2942
1269
2943
1270
Supported formats must be initializable and openable.
2944
Unsupported formats may not support initialization or committing or
1271
Unsupported formats may not support initialization or committing or
2945
1272
some other features depending on the reason for not being supported.
2946
1273
"""
2947
1274
return True
2948
1275
2949
def network_name(self):
2950
"""A simple byte string uniquely identifying this format for RPC calls.
2951
2952
MetaDir repository formats use their disk format string to identify the
2953
repository over the wire. All in one formats such as bzr < 0.8, and
2954
foreign formats like svn/git and hg should use some marker which is
2955
unique and immutable.
2956
"""
2957
raise NotImplementedError(self.network_name)
2958
2959
1276
def check_conversion_target(self, target_format):
2960
1277
raise NotImplementedError(self.check_conversion_target)
2961
1278
2962
1279
def open(self, a_bzrdir, _found=False):
2963
1280
"""Return an instance of this format for the bzrdir a_bzrdir.
2964
1281
2965
1282
_found is a private parameter, do not use it.
2966
1283
"""
2967
1284
raise NotImplementedError(self.open)
2972
1289
2973
1290
rich_root_data = False
2974
1291
supports_tree_reference = False
2975
supports_external_lookups = False
2976
2977
@property
2978
def _matchingbzrdir(self):
2979
matching = bzrdir.BzrDirMetaFormat1()
2980
matching.repository_format = self
2981
return matching
1292
_matchingbzrdir = bzrdir.BzrDirMetaFormat1()
2982
1293
2983
1294
def __init__(self):
2984
1295
super(MetaDirRepositoryFormat, self).__init__()
2997
1308
"""Upload the initial blank content."""
2998
1309
control_files = self._create_control_files(a_bzrdir)
2999
1310
control_files.lock_write()
3000
transport = control_files._transport
3001
if shared == True:
3002
utf8_files += [('shared-storage', '')]
3003
1311
try:
3004
transport.mkdir_multi(dirs, mode=a_bzrdir._get_dir_mode())
3005
for (filename, content_stream) in files:
3006
transport.put_file(filename, content_stream,
3007
mode=a_bzrdir._get_file_mode())
3008
for (filename, content_bytes) in utf8_files:
3009
transport.put_bytes_non_atomic(filename, content_bytes,
3010
mode=a_bzrdir._get_file_mode())
1312
control_files._transport.mkdir_multi(dirs,
1313
mode=control_files._dir_mode)
1314
for file, content in files:
1315
control_files.put(file, content)
1316
for file, content in utf8_files:
1317
control_files.put_utf8(file, content)
1318
if shared == True:
1319
control_files.put_utf8('shared-storage', '')
3011
1320
finally:
3012
1321
control_files.unlock()
3013
1322
3014
def network_name(self):
3015
"""Metadir formats have matching disk and network format strings."""
3016
return self.get_format_string()
3017
3018
3019
# Pre-0.8 formats that don't have a disk format string (because they are
3020
# versioned by the matching control directory). We use the control directories
3021
# disk format string as a key for the network_name because they meet the
3022
# constraints (simple string, unique, immutable).
3023
network_format_registry.register_lazy(
3024
"Bazaar-NG branch, format 5\n",
3025
'bzrlib.repofmt.weaverepo',
3026
'RepositoryFormat5',
3027
)
3028
network_format_registry.register_lazy(
3029
"Bazaar-NG branch, format 6\n",
3030
'bzrlib.repofmt.weaverepo',
3031
'RepositoryFormat6',
3032
)
3033
3034
# formats which have no format string are not discoverable or independently
3035
# creatable on disk, so are not registered in format_registry. They're
1323
1324
# formats which have no format string are not discoverable
1325
# and not independently creatable, so are not registered. They're
3036
1326
# all in bzrlib.repofmt.weaverepo now. When an instance of one of these is
3037
1327
# needed, it's constructed directly by the BzrDir. Non-native formats where
3038
1328
# the repository is not separately opened are similar.
3042
1332
'bzrlib.repofmt.weaverepo',
3043
1333
'RepositoryFormat7'
3044
1334
)
1335
# KEEP in sync with bzrdir.format_registry default, which controls the overall
1336
# default control directory format
3045
1337
3046
1338
format_registry.register_lazy(
3047
1339
'Bazaar-NG Knit Repository Format 1',
3048
1340
'bzrlib.repofmt.knitrepo',
3049
1341
'RepositoryFormatKnit1',
3050
1342
)
1343
format_registry.default_key = 'Bazaar-NG Knit Repository Format 1'
3051
1344
3052
1345
format_registry.register_lazy(
3053
1346
'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
3055
1348
'RepositoryFormatKnit3',
3056
1349
)
3057
1350
3058
format_registry.register_lazy(
3059
'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
3060
'bzrlib.repofmt.knitrepo',
3061
'RepositoryFormatKnit4',
3062
)
3063
3064
# Pack-based formats. There is one format for pre-subtrees, and one for
3065
# post-subtrees to allow ease of testing.
3066
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
3067
format_registry.register_lazy(
3068
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
3069
'bzrlib.repofmt.pack_repo',
3070
'RepositoryFormatKnitPack1',
3071
)
3072
format_registry.register_lazy(
3073
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
3074
'bzrlib.repofmt.pack_repo',
3075
'RepositoryFormatKnitPack3',
3076
)
3077
format_registry.register_lazy(
3078
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
3079
'bzrlib.repofmt.pack_repo',
3080
'RepositoryFormatKnitPack4',
3081
)
3082
format_registry.register_lazy(
3083
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
3084
'bzrlib.repofmt.pack_repo',
3085
'RepositoryFormatKnitPack5',
3086
)
3087
format_registry.register_lazy(
3088
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
3089
'bzrlib.repofmt.pack_repo',
3090
'RepositoryFormatKnitPack5RichRoot',
3091
)
3092
format_registry.register_lazy(
3093
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
3094
'bzrlib.repofmt.pack_repo',
3095
'RepositoryFormatKnitPack5RichRootBroken',
3096
)
3097
format_registry.register_lazy(
3098
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
3099
'bzrlib.repofmt.pack_repo',
3100
'RepositoryFormatKnitPack6',
3101
)
3102
format_registry.register_lazy(
3103
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
3104
'bzrlib.repofmt.pack_repo',
3105
'RepositoryFormatKnitPack6RichRoot',
3106
)
3107
3108
# Development formats.
3109
# Obsolete but kept pending a CHK based subtree format.
3110
format_registry.register_lazy(
3111
("Bazaar development format 2 with subtree support "
3112
"(needs bzr.dev from before 1.8)\n"),
3113
'bzrlib.repofmt.pack_repo',
3114
'RepositoryFormatPackDevelopment2Subtree',
3115
)
3116
3117
# 1.14->1.16 go below here
3118
format_registry.register_lazy(
3119
'Bazaar development format - group compression and chk inventory'
3120
' (needs bzr.dev from 1.14)\n',
3121
'bzrlib.repofmt.groupcompress_repo',
3122
'RepositoryFormatCHK1',
3123
)
3124
3125
format_registry.register_lazy(
3126
'Bazaar development format - chk repository with bencode revision '
3127
'serialization (needs bzr.dev from 1.16)\n',
3128
'bzrlib.repofmt.groupcompress_repo',
3129
'RepositoryFormatCHK2',
3130
)
3131
format_registry.register_lazy(
3132
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3133
'bzrlib.repofmt.groupcompress_repo',
3134
'RepositoryFormat2a',
3135
)
3136
3137
1351
3138
1352
class InterRepository(InterObject):
3139
1353
"""This class represents operations taking place between two repositories.
3140
1354
3141
1355
Its instances have methods like copy_content and fetch, and contain
3142
references to the source and target repositories these operations can be
1356
references to the source and target repositories these operations can be
3143
1357
carried out on.
3144
1358
3145
1359
Often we will provide convenience methods on 'repository' which carry out
3147
1361
InterRepository.get(other).method_name(parameters).
3148
1362
"""
3149
1363
3150
_walk_to_common_revisions_batch_size = 50
3151
1364
_optimisers = []
3152
1365
"""The available optimised InterRepository types."""
3153
1366
3154
@needs_write_lock
3155
1367
def copy_content(self, revision_id=None):
3156
"""Make a complete copy of the content in self into destination.
3157
3158
This is a destructive operation! Do not use it on existing
3159
repositories.
3160
3161
:param revision_id: Only copy the content needed to construct
3162
revision_id and its parents.
3163
"""
3164
try:
3165
self.target.set_make_working_trees(self.source.make_working_trees())
3166
except NotImplementedError:
3167
pass
3168
self.target.fetch(self.source, revision_id=revision_id)
3169
3170
@needs_write_lock
3171
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3172
fetch_spec=None):
1368
raise NotImplementedError(self.copy_content)
1369
1370
def fetch(self, revision_id=None, pb=None):
3173
1371
"""Fetch the content required to construct revision_id.
3174
1372
3175
1373
The content is copied from self.source to self.target.
3178
1376
content is copied.
3179
1377
:param pb: optional progress bar to use for progress reports. If not
3180
1378
provided a default one will be created.
3181
:return: None.
3182
"""
3183
from bzrlib.fetch import RepoFetcher
3184
f = RepoFetcher(to_repository=self.target,
3185
from_repository=self.source,
3186
last_revision=revision_id,
3187
fetch_spec=fetch_spec,
3188
pb=pb, find_ghosts=find_ghosts)
3189
3190
def _walk_to_common_revisions(self, revision_ids):
3191
"""Walk out from revision_ids in source to revisions target has.
3192
3193
:param revision_ids: The start point for the search.
3194
:return: A set of revision ids.
3195
"""
3196
target_graph = self.target.get_graph()
3197
revision_ids = frozenset(revision_ids)
3198
missing_revs = set()
3199
source_graph = self.source.get_graph()
3200
# ensure we don't pay silly lookup costs.
3201
searcher = source_graph._make_breadth_first_searcher(revision_ids)
3202
null_set = frozenset([_mod_revision.NULL_REVISION])
3203
searcher_exhausted = False
3204
while True:
3205
next_revs = set()
3206
ghosts = set()
3207
# Iterate the searcher until we have enough next_revs
3208
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3209
try:
3210
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3211
next_revs.update(next_revs_part)
3212
ghosts.update(ghosts_part)
3213
except StopIteration:
3214
searcher_exhausted = True
3215
break
3216
# If there are ghosts in the source graph, and the caller asked for
3217
# them, make sure that they are present in the target.
3218
# We don't care about other ghosts as we can't fetch them and
3219
# haven't been asked to.
3220
ghosts_to_check = set(revision_ids.intersection(ghosts))
3221
revs_to_get = set(next_revs).union(ghosts_to_check)
3222
if revs_to_get:
3223
have_revs = set(target_graph.get_parent_map(revs_to_get))
3224
# we always have NULL_REVISION present.
3225
have_revs = have_revs.union(null_set)
3226
# Check if the target is missing any ghosts we need.
3227
ghosts_to_check.difference_update(have_revs)
3228
if ghosts_to_check:
3229
# One of the caller's revision_ids is a ghost in both the
3230
# source and the target.
3231
raise errors.NoSuchRevision(
3232
self.source, ghosts_to_check.pop())
3233
missing_revs.update(next_revs - have_revs)
3234
# Because we may have walked past the original stop point, make
3235
# sure everything is stopped
3236
stop_revs = searcher.find_seen_ancestors(have_revs)
3237
searcher.stop_searching_any(stop_revs)
3238
if searcher_exhausted:
3239
break
3240
return searcher.get_result()
3241
1379
1380
Returns the copied revision count and the failed revisions in a tuple:
1381
(copied, failures).
1382
"""
1383
raise NotImplementedError(self.fetch)
1384
3242
1385
@needs_read_lock
3243
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
1386
def missing_revision_ids(self, revision_id=None):
3244
1387
"""Return the revision ids that source has that target does not.
1388
1389
These are returned in topological order.
3245
1390
3246
1391
:param revision_id: only return revision ids included by this
3247
1392
revision_id.
3248
:param find_ghosts: If True find missing revisions in deep history
3249
rather than just finding the surface difference.
3250
:return: A bzrlib.graph.SearchResult.
3251
1393
"""
3252
# stop searching at found target revisions.
3253
if not find_ghosts and revision_id is not None:
3254
return self._walk_to_common_revisions([revision_id])
3255
1394
# generic, possibly worst case, slow code path.
3256
1395
target_ids = set(self.target.all_revision_ids())
3257
1396
if revision_id is not None:
1397
# TODO: jam 20070210 InterRepository is internal enough that it
1398
# should assume revision_ids are already utf-8
1399
revision_id = osutils.safe_revision_id(revision_id)
3258
1400
source_ids = self.source.get_ancestry(revision_id)
3259
if source_ids[0] is not None:
3260
raise AssertionError()
1401
assert source_ids[0] is None
3261
1402
source_ids.pop(0)
3262
1403
else:
3263
1404
source_ids = self.source.all_revision_ids()
3264
1405
result_set = set(source_ids).difference(target_ids)
3265
return self.source.revision_ids_to_search_result(result_set)
3266
3267
@staticmethod
3268
def _same_model(source, target):
3269
"""True if source and target have the same data representation.
3270
3271
Note: this is always called on the base class; overriding it in a
3272
subclass will have no effect.
3273
"""
3274
try:
3275
InterRepository._assert_same_model(source, target)
3276
return True
3277
except errors.IncompatibleRepositories, e:
3278
return False
3279
3280
@staticmethod
3281
def _assert_same_model(source, target):
3282
"""Raise an exception if two repositories do not use the same model.
3283
"""
3284
if source.supports_rich_root() != target.supports_rich_root():
3285
raise errors.IncompatibleRepositories(source, target,
3286
"different rich-root support")
3287
if source._serializer != target._serializer:
3288
raise errors.IncompatibleRepositories(source, target,
3289
"different serializers")
1406
# this may look like a no-op: its not. It preserves the ordering
1407
# other_ids had while only returning the members from other_ids
1408
# that we've decided we need.
1409
return [rev_id for rev_id in source_ids if rev_id in result_set]
3290
1410
3291
1411
3292
1412
class InterSameDataRepository(InterRepository):
3293
1413
"""Code for converting between repositories that represent the same data.
3294
1414
3295
1415
Data format and model must match for this to work.
3296
1416
"""
3297
1417
3298
1418
@classmethod
3299
1419
def _get_repo_format_to_test(self):
3300
"""Repository format for testing with.
3301
3302
InterSameData can pull from subtree to subtree and from non-subtree to
3303
non-subtree, so we test this with the richest repository format.
3304
"""
3305
from bzrlib.repofmt import knitrepo
3306
return knitrepo.RepositoryFormatKnit3()
1420
"""Repository format for testing with."""
1421
return RepositoryFormat.get_default_format()
3307
1422
3308
1423
@staticmethod
3309
1424
def is_compatible(source, target):
3310
return InterRepository._same_model(source, target)
1425
if source.supports_rich_root() != target.supports_rich_root():
1426
return False
1427
if source._serializer != target._serializer:
1428
return False
1429
return True
1430
1431
@needs_write_lock
1432
def copy_content(self, revision_id=None):
1433
"""Make a complete copy of the content in self into destination.
1434
1435
This copies both the repository's revision data, and configuration information
1436
such as the make_working_trees setting.
1437
1438
This is a destructive operation! Do not use it on existing
1439
repositories.
1440
1441
:param revision_id: Only copy the content needed to construct
1442
revision_id and its parents.
1443
"""
1444
try:
1445
self.target.set_make_working_trees(self.source.make_working_trees())
1446
except NotImplementedError:
1447
pass
1448
# TODO: jam 20070210 This is fairly internal, so we should probably
1449
# just assert that revision_id is not unicode.
1450
revision_id = osutils.safe_revision_id(revision_id)
1451
# but don't bother fetching if we have the needed data now.
1452
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
1453
self.target.has_revision(revision_id)):
1454
return
1455
self.target.fetch(self.source, revision_id=revision_id)
1456
1457
@needs_write_lock
1458
def fetch(self, revision_id=None, pb=None):
1459
"""See InterRepository.fetch()."""
1460
from bzrlib.fetch import GenericRepoFetcher
1461
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1462
self.source, self.source._format, self.target,
1463
self.target._format)
1464
# TODO: jam 20070210 This should be an assert, not a translate
1465
revision_id = osutils.safe_revision_id(revision_id)
1466
f = GenericRepoFetcher(to_repository=self.target,
1467
from_repository=self.source,
1468
last_revision=revision_id,
1469
pb=pb)
1470
return f.count_copied, f.failed_revisions
3311
1471
3312
1472
3313
1473
class InterWeaveRepo(InterSameDataRepository):
3314
"""Optimised code paths between Weave based repositories.
3315
3316
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
3317
implemented lazy inter-object optimisation.
3318
"""
1474
"""Optimised code paths between Weave based repositories."""
3319
1475
3320
1476
@classmethod
3321
1477
def _get_repo_format_to_test(self):
3325
1481
@staticmethod
3326
1482
def is_compatible(source, target):
3327
1483
"""Be compatible with known Weave formats.
3328
1484
3329
1485
We don't test for the stores being of specific types because that
3330
could lead to confusing results, and there is no need to be
1486
could lead to confusing results, and there is no need to be
3331
1487
overly general.
3332
1488
"""
3333
1489
from bzrlib.repofmt.weaverepo import (
3344
1500
RepositoryFormat7)))
3345
1501
except AttributeError:
3346
1502
return False
3347
1503
3348
1504
@needs_write_lock
3349
1505
def copy_content(self, revision_id=None):
3350
1506
"""See InterRepository.copy_content()."""
3351
1507
# weave specific optimised path:
1508
# TODO: jam 20070210 Internal, should be an assert, not translate
1509
revision_id = osutils.safe_revision_id(revision_id)
3352
1510
try:
3353
1511
self.target.set_make_working_trees(self.source.make_working_trees())
3354
except (errors.RepositoryUpgradeRequired, NotImplemented):
1512
except NotImplementedError:
3355
1513
pass
3356
1514
# FIXME do not peek!
3357
if self.source._transport.listable():
1515
if self.source.control_files._transport.listable():
3358
1516
pb = ui.ui_factory.nested_progress_bar()
3359
1517
try:
3360
self.target.texts.insert_record_stream(
3361
self.source.texts.get_record_stream(
3362
self.source.texts.keys(), 'topological', False))
1518
self.target.weave_store.copy_all_ids(
1519
self.source.weave_store,
1520
pb=pb,
1521
from_transaction=self.source.get_transaction(),
1522
to_transaction=self.target.get_transaction())
3363
1523
pb.update('copying inventory', 0, 1)
3364
self.target.inventories.insert_record_stream(
3365
self.source.inventories.get_record_stream(
3366
self.source.inventories.keys(), 'topological', False))
3367
self.target.signatures.insert_record_stream(
3368
self.source.signatures.get_record_stream(
3369
self.source.signatures.keys(),
3370
'unordered', True))
3371
self.target.revisions.insert_record_stream(
3372
self.source.revisions.get_record_stream(
3373
self.source.revisions.keys(),
3374
'topological', True))
1524
self.target.control_weaves.copy_multi(
1525
self.source.control_weaves, ['inventory'],
1526
from_transaction=self.source.get_transaction(),
1527
to_transaction=self.target.get_transaction())
1528
self.target._revision_store.text_store.copy_all_ids(
1529
self.source._revision_store.text_store,
1530
pb=pb)
3375
1531
finally:
3376
1532
pb.finished()
3377
1533
else:
3378
1534
self.target.fetch(self.source, revision_id=revision_id)
3379
1535
1536
@needs_write_lock
1537
def fetch(self, revision_id=None, pb=None):
1538
"""See InterRepository.fetch()."""
1539
from bzrlib.fetch import GenericRepoFetcher
1540
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1541
self.source, self.source._format, self.target, self.target._format)
1542
# TODO: jam 20070210 This should be an assert, not a translate
1543
revision_id = osutils.safe_revision_id(revision_id)
1544
f = GenericRepoFetcher(to_repository=self.target,
1545
from_repository=self.source,
1546
last_revision=revision_id,
1547
pb=pb)
1548
return f.count_copied, f.failed_revisions
1549
3380
1550
@needs_read_lock
3381
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
1551
def missing_revision_ids(self, revision_id=None):
3382
1552
"""See InterRepository.missing_revision_ids()."""
3383
1553
# we want all revisions to satisfy revision_id in source.
3384
1554
# but we don't want to stat every file here and there.
3385
# we want then, all revisions other needs to satisfy revision_id
1555
# we want then, all revisions other needs to satisfy revision_id
3386
1556
# checked, but not those that we have locally.
3387
# so the first thing is to get a subset of the revisions to
1557
# so the first thing is to get a subset of the revisions to
3388
1558
# satisfy revision_id in source, and then eliminate those that
3389
# we do already have.
3390
# this is slow on high latency connection to self, but as this
3391
# disk format scales terribly for push anyway due to rewriting
1559
# we do already have.
1560
# this is slow on high latency connection to self, but as as this
1561
# disk format scales terribly for push anyway due to rewriting
3392
1562
# inventory.weave, this is considered acceptable.
3393
1563
# - RBC 20060209
3394
1564
if revision_id is not None:
3395
1565
source_ids = self.source.get_ancestry(revision_id)
3396
if source_ids[0] is not None:
3397
raise AssertionError()
1566
assert source_ids[0] is None
3398
1567
source_ids.pop(0)
3399
1568
else:
3400
1569
source_ids = self.source._all_possible_ids()
3405
1574
# we do not have a revision as that would be pointless.
3406
1575
target_ids = set(self.target._all_possible_ids())
3407
1576
possibly_present_revisions = target_ids.intersection(source_ids_set)
3408
actually_present_revisions = set(
3409
self.target._eliminate_revisions_not_present(possibly_present_revisions))
1577
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
3410
1578
required_revisions = source_ids_set.difference(actually_present_revisions)
1579
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
3411
1580
if revision_id is not None:
3412
1581
# we used get_ancestry to determine source_ids then we are assured all
3413
1582
# revisions referenced are present as they are installed in topological order.
3414
1583
# and the tip revision was validated by get_ancestry.
3415
result_set = required_revisions
1584
return required_topo_revisions
3416
1585
else:
3417
# if we just grabbed the possibly available ids, then
1586
# if we just grabbed the possibly available ids, then
3418
1587
# we only have an estimate of whats available and need to validate
3419
1588
# that against the revision records.
3420
result_set = set(
3421
self.source._eliminate_revisions_not_present(required_revisions))
3422
return self.source.revision_ids_to_search_result(result_set)
1589
return self.source._eliminate_revisions_not_present(required_topo_revisions)
3423
1590
3424
1591
3425
1592
class InterKnitRepo(InterSameDataRepository):
3433
1600
@staticmethod
3434
1601
def is_compatible(source, target):
3435
1602
"""Be compatible with known Knit formats.
3436
1603
3437
1604
We don't test for the stores being of specific types because that
3438
could lead to confusing results, and there is no need to be
1605
could lead to confusing results, and there is no need to be
3439
1606
overly general.
3440
1607
"""
3441
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
1608
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit1
3442
1609
try:
3443
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
3444
isinstance(target._format, RepositoryFormatKnit))
1610
return (isinstance(source._format, (RepositoryFormatKnit1)) and
1611
isinstance(target._format, (RepositoryFormatKnit1)))
3445
1612
except AttributeError:
3446
1613
return False
3447
return are_knits and InterRepository._same_model(source, target)
1614
1615
@needs_write_lock
1616
def fetch(self, revision_id=None, pb=None):
1617
"""See InterRepository.fetch()."""
1618
from bzrlib.fetch import KnitRepoFetcher
1619
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1620
self.source, self.source._format, self.target, self.target._format)
1621
# TODO: jam 20070210 This should be an assert, not a translate
1622
revision_id = osutils.safe_revision_id(revision_id)
1623
f = KnitRepoFetcher(to_repository=self.target,
1624
from_repository=self.source,
1625
last_revision=revision_id,
1626
pb=pb)
1627
return f.count_copied, f.failed_revisions
3448
1628
3449
1629
@needs_read_lock
3450
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
1630
def missing_revision_ids(self, revision_id=None):
3451
1631
"""See InterRepository.missing_revision_ids()."""
3452
1632
if revision_id is not None:
3453
1633
source_ids = self.source.get_ancestry(revision_id)
3454
if source_ids[0] is not None:
3455
raise AssertionError()
1634
assert source_ids[0] is None
3456
1635
source_ids.pop(0)
3457
1636
else:
3458
source_ids = self.source.all_revision_ids()
1637
source_ids = self.source._all_possible_ids()
3459
1638
source_ids_set = set(source_ids)
3460
1639
# source_ids is the worst possible case we may need to pull.
3461
1640
# now we want to filter source_ids against what we actually
3462
1641
# have in target, but don't try to check for existence where we know
3463
1642
# we do not have a revision as that would be pointless.
3464
target_ids = set(self.target.all_revision_ids())
1643
target_ids = set(self.target._all_possible_ids())
3465
1644
possibly_present_revisions = target_ids.intersection(source_ids_set)
3466
actually_present_revisions = set(
3467
self.target._eliminate_revisions_not_present(possibly_present_revisions))
1645
actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
3468
1646
required_revisions = source_ids_set.difference(actually_present_revisions)
1647
required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
3469
1648
if revision_id is not None:
3470
1649
# we used get_ancestry to determine source_ids then we are assured all
3471
1650
# revisions referenced are present as they are installed in topological order.
3472
1651
# and the tip revision was validated by get_ancestry.
3473
result_set = required_revisions
1652
return required_topo_revisions
3474
1653
else:
3475
# if we just grabbed the possibly available ids, then
1654
# if we just grabbed the possibly available ids, then
3476
1655
# we only have an estimate of whats available and need to validate
3477
1656
# that against the revision records.
3478
result_set = set(
3479
self.source._eliminate_revisions_not_present(required_revisions))
3480
return self.source.revision_ids_to_search_result(result_set)
3481
3482
3483
class InterDifferingSerializer(InterRepository):
3484
3485
@classmethod
3486
def _get_repo_format_to_test(self):
3487
return None
3488
3489
@staticmethod
3490
def is_compatible(source, target):
3491
"""Be compatible with Knit2 source and Knit3 target"""
3492
# This is redundant with format.check_conversion_target(), however that
3493
# raises an exception, and we just want to say "False" as in we won't
3494
# support converting between these formats.
3495
if source.supports_rich_root() and not target.supports_rich_root():
3496
return False
3497
if (source._format.supports_tree_reference
3498
and not target._format.supports_tree_reference):
3499
return False
3500
return True
3501
3502
def _get_delta_for_revision(self, tree, parent_ids, basis_id, cache):
3503
"""Get the best delta and base for this revision.
3504
3505
:return: (basis_id, delta)
3506
"""
3507
possible_trees = [(parent_id, cache[parent_id])
3508
for parent_id in parent_ids
3509
if parent_id in cache]
3510
if len(possible_trees) == 0:
3511
# There either aren't any parents, or the parents aren't in the
3512
# cache, so just use the last converted tree
3513
possible_trees.append((basis_id, cache[basis_id]))
3514
deltas = []
3515
for basis_id, basis_tree in possible_trees:
3516
delta = tree.inventory._make_delta(basis_tree.inventory)
3517
deltas.append((len(delta), basis_id, delta))
3518
deltas.sort()
3519
return deltas[0][1:]
3520
3521
def _get_parent_keys(self, root_key, parent_map):
3522
"""Get the parent keys for a given root id."""
3523
root_id, rev_id = root_key
3524
# Include direct parents of the revision, but only if they used
3525
# the same root_id and are heads.
3526
parent_keys = []
3527
for parent_id in parent_map[rev_id]:
3528
if parent_id == _mod_revision.NULL_REVISION:
3529
continue
3530
if parent_id not in self._revision_id_to_root_id:
3531
# We probably didn't read this revision, go spend the
3532
# extra effort to actually check
3533
try:
3534
tree = self.source.revision_tree(parent_id)
3535
except errors.NoSuchRevision:
3536
# Ghost, fill out _revision_id_to_root_id in case we
3537
# encounter this again.
3538
# But set parent_root_id to None since we don't really know
3539
parent_root_id = None
3540
else:
3541
parent_root_id = tree.get_root_id()
3542
self._revision_id_to_root_id[parent_id] = None
3543
else:
3544
parent_root_id = self._revision_id_to_root_id[parent_id]
3545
if root_id == parent_root_id:
3546
# With stacking we _might_ want to refer to a non-local
3547
# revision, but this code path only applies when we have the
3548
# full content available, so ghosts really are ghosts, not just
3549
# the edge of local data.
3550
parent_keys.append((parent_id,))
3551
else:
3552
# root_id may be in the parent anyway.
3553
try:
3554
tree = self.source.revision_tree(parent_id)
3555
except errors.NoSuchRevision:
3556
# ghost, can't refer to it.
3557
pass
3558
else:
3559
try:
3560
parent_keys.append((tree.inventory[root_id].revision,))
3561
except errors.NoSuchId:
3562
# not in the tree
3563
pass
3564
g = graph.Graph(self.source.revisions)
3565
heads = g.heads(parent_keys)
3566
selected_keys = []
3567
for key in parent_keys:
3568
if key in heads and key not in selected_keys:
3569
selected_keys.append(key)
3570
return tuple([(root_id,)+ key for key in selected_keys])
3571
3572
def _new_root_data_stream(self, root_keys_to_create, parent_map):
3573
for root_key in root_keys_to_create:
3574
parent_keys = self._get_parent_keys(root_key, parent_map)
3575
yield versionedfile.FulltextContentFactory(root_key,
3576
parent_keys, None, '')
3577
3578
def _fetch_batch(self, revision_ids, basis_id, cache):
3579
"""Fetch across a few revisions.
3580
3581
:param revision_ids: The revisions to copy
3582
:param basis_id: The revision_id of a tree that must be in cache, used
3583
as a basis for delta when no other base is available
3584
:param cache: A cache of RevisionTrees that we can use.
3585
:return: The revision_id of the last converted tree. The RevisionTree
3586
for it will be in cache
3587
"""
3588
# Walk though all revisions; get inventory deltas, copy referenced
3589
# texts that delta references, insert the delta, revision and
3590
# signature.
3591
root_keys_to_create = set()
3592
text_keys = set()
3593
pending_deltas = []
3594
pending_revisions = []
3595
parent_map = self.source.get_parent_map(revision_ids)
3596
for tree in self.source.revision_trees(revision_ids):
3597
current_revision_id = tree.get_revision_id()
3598
parent_ids = parent_map.get(current_revision_id, ())
3599
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3600
basis_id, cache)
3601
if self._converting_to_rich_root:
3602
self._revision_id_to_root_id[current_revision_id] = \
3603
tree.get_root_id()
3604
# Find text entries that need to be copied
3605
for old_path, new_path, file_id, entry in delta:
3606
if new_path is not None:
3607
if not new_path:
3608
# This is the root
3609
if not self.target.supports_rich_root():
3610
# The target doesn't support rich root, so we don't
3611
# copy
3612
continue
3613
if self._converting_to_rich_root:
3614
# This can't be copied normally, we have to insert
3615
# it specially
3616
root_keys_to_create.add((file_id, entry.revision))
3617
continue
3618
text_keys.add((file_id, entry.revision))
3619
revision = self.source.get_revision(current_revision_id)
3620
pending_deltas.append((basis_id, delta,
3621
current_revision_id, revision.parent_ids))
3622
pending_revisions.append(revision)
3623
cache[current_revision_id] = tree
3624
basis_id = current_revision_id
3625
# Copy file texts
3626
from_texts = self.source.texts
3627
to_texts = self.target.texts
3628
if root_keys_to_create:
3629
root_stream = self._new_root_data_stream(root_keys_to_create,
3630
parent_map)
3631
to_texts.insert_record_stream(root_stream)
3632
to_texts.insert_record_stream(from_texts.get_record_stream(
3633
text_keys, self.target._format._fetch_order,
3634
not self.target._format._fetch_uses_deltas))
3635
# insert inventory deltas
3636
for delta in pending_deltas:
3637
self.target.add_inventory_by_delta(*delta)
3638
if self.target._fallback_repositories:
3639
# Make sure this stacked repository has all the parent inventories
3640
# for the new revisions that we are about to insert. We do this
3641
# before adding the revisions so that no revision is added until
3642
# all the inventories it may depend on are added.
3643
parent_ids = set()
3644
revision_ids = set()
3645
for revision in pending_revisions:
3646
revision_ids.add(revision.revision_id)
3647
parent_ids.update(revision.parent_ids)
3648
parent_ids.difference_update(revision_ids)
3649
parent_ids.discard(_mod_revision.NULL_REVISION)
3650
parent_map = self.source.get_parent_map(parent_ids)
3651
for parent_tree in self.source.revision_trees(parent_ids):
3652
basis_id, delta = self._get_delta_for_revision(tree, parent_ids, basis_id, cache)
3653
current_revision_id = parent_tree.get_revision_id()
3654
parents_parents = parent_map[current_revision_id]
3655
self.target.add_inventory_by_delta(
3656
basis_id, delta, current_revision_id, parents_parents)
3657
# insert signatures and revisions
3658
for revision in pending_revisions:
3659
try:
3660
signature = self.source.get_signature_text(
3661
revision.revision_id)
3662
self.target.add_signature_text(revision.revision_id,
3663
signature)
3664
except errors.NoSuchRevision:
3665
pass
3666
self.target.add_revision(revision.revision_id, revision)
3667
return basis_id
3668
3669
def _fetch_all_revisions(self, revision_ids, pb):
3670
"""Fetch everything for the list of revisions.
3671
3672
:param revision_ids: The list of revisions to fetch. Must be in
3673
topological order.
3674
:param pb: A ProgressBar
3675
:return: None
3676
"""
3677
basis_id, basis_tree = self._get_basis(revision_ids[0])
3678
batch_size = 100
3679
cache = lru_cache.LRUCache(100)
3680
cache[basis_id] = basis_tree
3681
del basis_tree # We don't want to hang on to it here
3682
hints = []
3683
for offset in range(0, len(revision_ids), batch_size):
3684
self.target.start_write_group()
3685
try:
3686
pb.update('Transferring revisions', offset,
3687
len(revision_ids))
3688
batch = revision_ids[offset:offset+batch_size]
3689
basis_id = self._fetch_batch(batch, basis_id, cache)
3690
except:
3691
self.target.abort_write_group()
3692
raise
3693
else:
3694
hint = self.target.commit_write_group()
3695
if hint:
3696
hints.extend(hint)
3697
if hints and self.target._format.pack_compresses:
3698
self.target.pack(hint=hints)
3699
pb.update('Transferring revisions', len(revision_ids),
3700
len(revision_ids))
3701
3702
@needs_write_lock
3703
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
3704
fetch_spec=None):
3705
"""See InterRepository.fetch()."""
3706
if fetch_spec is not None:
3707
raise AssertionError("Not implemented yet...")
3708
if (not self.source.supports_rich_root()
3709
and self.target.supports_rich_root()):
3710
self._converting_to_rich_root = True
3711
self._revision_id_to_root_id = {}
3712
else:
3713
self._converting_to_rich_root = False
3714
revision_ids = self.target.search_missing_revision_ids(self.source,
3715
revision_id, find_ghosts=find_ghosts).get_keys()
3716
if not revision_ids:
3717
return 0, 0
3718
revision_ids = tsort.topo_sort(
3719
self.source.get_graph().get_parent_map(revision_ids))
3720
if not revision_ids:
3721
return 0, 0
3722
# Walk though all revisions; get inventory deltas, copy referenced
3723
# texts that delta references, insert the delta, revision and
3724
# signature.
3725
first_rev = self.source.get_revision(revision_ids[0])
3726
if pb is None:
3727
my_pb = ui.ui_factory.nested_progress_bar()
3728
pb = my_pb
3729
else:
3730
symbol_versioning.warn(
3731
symbol_versioning.deprecated_in((1, 14, 0))
3732
% "pb parameter to fetch()")
3733
my_pb = None
3734
try:
3735
self._fetch_all_revisions(revision_ids, pb)
3736
finally:
3737
if my_pb is not None:
3738
my_pb.finished()
3739
return len(revision_ids), 0
3740
3741
def _get_basis(self, first_revision_id):
3742
"""Get a revision and tree which exists in the target.
3743
3744
This assumes that first_revision_id is selected for transmission
3745
because all other ancestors are already present. If we can't find an
3746
ancestor we fall back to NULL_REVISION since we know that is safe.
3747
3748
:return: (basis_id, basis_tree)
3749
"""
3750
first_rev = self.source.get_revision(first_revision_id)
3751
try:
3752
basis_id = first_rev.parent_ids[0]
3753
# only valid as a basis if the target has it
3754
self.target.get_revision(basis_id)
3755
# Try to get a basis tree - if its a ghost it will hit the
3756
# NoSuchRevision case.
3757
basis_tree = self.source.revision_tree(basis_id)
3758
except (IndexError, errors.NoSuchRevision):
3759
basis_id = _mod_revision.NULL_REVISION
3760
basis_tree = self.source.revision_tree(basis_id)
3761
return basis_id, basis_tree
3762
3763
3764
InterRepository.register_optimiser(InterDifferingSerializer)
1657
return self.source._eliminate_revisions_not_present(required_topo_revisions)
1658
1659
1660
class InterModel1and2(InterRepository):
1661
1662
@classmethod
1663
def _get_repo_format_to_test(self):
1664
return None
1665
1666
@staticmethod
1667
def is_compatible(source, target):
1668
if not source.supports_rich_root() and target.supports_rich_root():
1669
return True
1670
else:
1671
return False
1672
1673
@needs_write_lock
1674
def fetch(self, revision_id=None, pb=None):
1675
"""See InterRepository.fetch()."""
1676
from bzrlib.fetch import Model1toKnit2Fetcher
1677
# TODO: jam 20070210 This should be an assert, not a translate
1678
revision_id = osutils.safe_revision_id(revision_id)
1679
f = Model1toKnit2Fetcher(to_repository=self.target,
1680
from_repository=self.source,
1681
last_revision=revision_id,
1682
pb=pb)
1683
return f.count_copied, f.failed_revisions
1684
1685
@needs_write_lock
1686
def copy_content(self, revision_id=None):
1687
"""Make a complete copy of the content in self into destination.
1688
1689
This is a destructive operation! Do not use it on existing
1690
repositories.
1691
1692
:param revision_id: Only copy the content needed to construct
1693
revision_id and its parents.
1694
"""
1695
try:
1696
self.target.set_make_working_trees(self.source.make_working_trees())
1697
except NotImplementedError:
1698
pass
1699
# TODO: jam 20070210 Internal, assert, don't translate
1700
revision_id = osutils.safe_revision_id(revision_id)
1701
# but don't bother fetching if we have the needed data now.
1702
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
1703
self.target.has_revision(revision_id)):
1704
return
1705
self.target.fetch(self.source, revision_id=revision_id)
1706
1707
1708
class InterKnit1and2(InterKnitRepo):
1709
1710
@classmethod
1711
def _get_repo_format_to_test(self):
1712
return None
1713
1714
@staticmethod
1715
def is_compatible(source, target):
1716
"""Be compatible with Knit1 source and Knit3 target"""
1717
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
1718
try:
1719
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit1, \
1720
RepositoryFormatKnit3
1721
return (isinstance(source._format, (RepositoryFormatKnit1)) and
1722
isinstance(target._format, (RepositoryFormatKnit3)))
1723
except AttributeError:
1724
return False
1725
1726
@needs_write_lock
1727
def fetch(self, revision_id=None, pb=None):
1728
"""See InterRepository.fetch()."""
1729
from bzrlib.fetch import Knit1to2Fetcher
1730
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
1731
self.source, self.source._format, self.target,
1732
self.target._format)
1733
# TODO: jam 20070210 This should be an assert, not a translate
1734
revision_id = osutils.safe_revision_id(revision_id)
1735
f = Knit1to2Fetcher(to_repository=self.target,
1736
from_repository=self.source,
1737
last_revision=revision_id,
1738
pb=pb)
1739
return f.count_copied, f.failed_revisions
1740
1741
1742
class InterRemoteRepository(InterRepository):
1743
"""Code for converting between RemoteRepository objects.
1744
1745
This just gets an non-remote repository from the RemoteRepository, and calls
1746
InterRepository.get again.
1747
"""
1748
1749
def __init__(self, source, target):
1750
if isinstance(source, remote.RemoteRepository):
1751
source._ensure_real()
1752
real_source = source._real_repository
1753
else:
1754
real_source = source
1755
if isinstance(target, remote.RemoteRepository):
1756
target._ensure_real()
1757
real_target = target._real_repository
1758
else:
1759
real_target = target
1760
self.real_inter = InterRepository.get(real_source, real_target)
1761
1762
@staticmethod
1763
def is_compatible(source, target):
1764
if isinstance(source, remote.RemoteRepository):
1765
return True
1766
if isinstance(target, remote.RemoteRepository):
1767
return True
1768
return False
1769
1770
def copy_content(self, revision_id=None):
1771
self.real_inter.copy_content(revision_id=revision_id)
1772
1773
def fetch(self, revision_id=None, pb=None):
1774
self.real_inter.fetch(revision_id=revision_id, pb=pb)
1775
1776
@classmethod
1777
def _get_repo_format_to_test(self):
1778
return None
1779
1780
3765
1781
InterRepository.register_optimiser(InterSameDataRepository)
3766
1782
InterRepository.register_optimiser(InterWeaveRepo)
3767
1783
InterRepository.register_optimiser(InterKnitRepo)
1784
InterRepository.register_optimiser(InterModel1and2)
1785
InterRepository.register_optimiser(InterKnit1and2)
1786
InterRepository.register_optimiser(InterRemoteRepository)
3768
1787
3769
1788
3770
1789
class CopyConverter(object):
3771
1790
"""A repository conversion tool which just performs a copy of the content.
3772
1791
3773
1792
This is slow but quite reliable.
3774
1793
"""
3775
1794
3779
1798
:param target_format: The format the resulting repository should be.
3780
1799
"""
3781
1800
self.target_format = target_format
3782
1801
3783
1802
def convert(self, repo, pb):
3784
1803
"""Perform the conversion of to_convert, giving feedback via pb.
3785
1804
3819
1838
self.pb.update(message, self.count, self.total)
3820
1839
3821
1840
1841
class CommitBuilder(object):
1842
"""Provides an interface to build up a commit.
1843
1844
This allows describing a tree to be committed without needing to
1845
know the internals of the format of the repository.
1846
"""
1847
1848
record_root_entry = False
1849
def __init__(self, repository, parents, config, timestamp=None,
1850
timezone=None, committer=None, revprops=None,
1851
revision_id=None):
1852
"""Initiate a CommitBuilder.
1853
1854
:param repository: Repository to commit to.
1855
:param parents: Revision ids of the parents of the new revision.
1856
:param config: Configuration to use.
1857
:param timestamp: Optional timestamp recorded for commit.
1858
:param timezone: Optional timezone for timestamp.
1859
:param committer: Optional committer to set for commit.
1860
:param revprops: Optional dictionary of revision properties.
1861
:param revision_id: Optional revision id.
1862
"""
1863
self._config = config
1864
1865
if committer is None:
1866
self._committer = self._config.username()
1867
else:
1868
assert isinstance(committer, basestring), type(committer)
1869
self._committer = committer
1870
1871
self.new_inventory = Inventory(None)
1872
self._new_revision_id = osutils.safe_revision_id(revision_id)
1873
self.parents = parents
1874
self.repository = repository
1875
1876
self._revprops = {}
1877
if revprops is not None:
1878
self._revprops.update(revprops)
1879
1880
if timestamp is None:
1881
timestamp = time.time()
1882
# Restrict resolution to 1ms
1883
self._timestamp = round(timestamp, 3)
1884
1885
if timezone is None:
1886
self._timezone = osutils.local_time_offset()
1887
else:
1888
self._timezone = int(timezone)
1889
1890
self._generate_revision_if_needed()
1891
1892
def commit(self, message):
1893
"""Make the actual commit.
1894
1895
:return: The revision id of the recorded revision.
1896
"""
1897
rev = _mod_revision.Revision(
1898
timestamp=self._timestamp,
1899
timezone=self._timezone,
1900
committer=self._committer,
1901
message=message,
1902
inventory_sha1=self.inv_sha1,
1903
revision_id=self._new_revision_id,
1904
properties=self._revprops)
1905
rev.parent_ids = self.parents
1906
self.repository.add_revision(self._new_revision_id, rev,
1907
self.new_inventory, self._config)
1908
return self._new_revision_id
1909
1910
def revision_tree(self):
1911
"""Return the tree that was just committed.
1912
1913
After calling commit() this can be called to get a RevisionTree
1914
representing the newly committed tree. This is preferred to
1915
calling Repository.revision_tree() because that may require
1916
deserializing the inventory, while we already have a copy in
1917
memory.
1918
"""
1919
return RevisionTree(self.repository, self.new_inventory,
1920
self._new_revision_id)
1921
1922
def finish_inventory(self):
1923
"""Tell the builder that the inventory is finished."""
1924
if self.new_inventory.root is None:
1925
symbol_versioning.warn('Root entry should be supplied to'
1926
' record_entry_contents, as of bzr 0.10.',
1927
DeprecationWarning, stacklevel=2)
1928
self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
1929
self.new_inventory.revision_id = self._new_revision_id
1930
self.inv_sha1 = self.repository.add_inventory(
1931
self._new_revision_id,
1932
self.new_inventory,
1933
self.parents
1934
)
1935
1936
def _gen_revision_id(self):
1937
"""Return new revision-id."""
1938
return generate_ids.gen_revision_id(self._config.username(),
1939
self._timestamp)
1940
1941
def _generate_revision_if_needed(self):
1942
"""Create a revision id if None was supplied.
1943
1944
If the repository can not support user-specified revision ids
1945
they should override this function and raise CannotSetRevisionId
1946
if _new_revision_id is not None.
1947
1948
:raises: CannotSetRevisionId
1949
"""
1950
if self._new_revision_id is None:
1951
self._new_revision_id = self._gen_revision_id()
1952
1953
def record_entry_contents(self, ie, parent_invs, path, tree):
1954
"""Record the content of ie from tree into the commit if needed.
1955
1956
Side effect: sets ie.revision when unchanged
1957
1958
:param ie: An inventory entry present in the commit.
1959
:param parent_invs: The inventories of the parent revisions of the
1960
commit.
1961
:param path: The path the entry is at in the tree.
1962
:param tree: The tree which contains this entry and should be used to
1963
obtain content.
1964
"""
1965
if self.new_inventory.root is None and ie.parent_id is not None:
1966
symbol_versioning.warn('Root entry should be supplied to'
1967
' record_entry_contents, as of bzr 0.10.',
1968
DeprecationWarning, stacklevel=2)
1969
self.record_entry_contents(tree.inventory.root.copy(), parent_invs,
1970
'', tree)
1971
self.new_inventory.add(ie)
1972
1973
# ie.revision is always None if the InventoryEntry is considered
1974
# for committing. ie.snapshot will record the correct revision
1975
# which may be the sole parent if it is untouched.
1976
if ie.revision is not None:
1977
return
1978
1979
# In this revision format, root entries have no knit or weave
1980
if ie is self.new_inventory.root:
1981
# When serializing out to disk and back in
1982
# root.revision is always _new_revision_id
1983
ie.revision = self._new_revision_id
1984
return
1985
previous_entries = ie.find_previous_heads(
1986
parent_invs,
1987
self.repository.weave_store,
1988
self.repository.get_transaction())
1989
# we are creating a new revision for ie in the history store
1990
# and inventory.
1991
ie.snapshot(self._new_revision_id, path, previous_entries, tree, self)
1992
1993
def modified_directory(self, file_id, file_parents):
1994
"""Record the presence of a symbolic link.
1995
1996
:param file_id: The file_id of the link to record.
1997
:param file_parents: The per-file parent revision ids.
1998
"""
1999
self._add_text_to_weave(file_id, [], file_parents.keys())
2000
2001
def modified_reference(self, file_id, file_parents):
2002
"""Record the modification of a reference.
2003
2004
:param file_id: The file_id of the link to record.
2005
:param file_parents: The per-file parent revision ids.
2006
"""
2007
self._add_text_to_weave(file_id, [], file_parents.keys())
2008
2009
def modified_file_text(self, file_id, file_parents,
2010
get_content_byte_lines, text_sha1=None,
2011
text_size=None):
2012
"""Record the text of file file_id
2013
2014
:param file_id: The file_id of the file to record the text of.
2015
:param file_parents: The per-file parent revision ids.
2016
:param get_content_byte_lines: A callable which will return the byte
2017
lines for the file.
2018
:param text_sha1: Optional SHA1 of the file contents.
2019
:param text_size: Optional size of the file contents.
2020
"""
2021
# mutter('storing text of file {%s} in revision {%s} into %r',
2022
# file_id, self._new_revision_id, self.repository.weave_store)
2023
# special case to avoid diffing on renames or
2024
# reparenting
2025
if (len(file_parents) == 1
2026
and text_sha1 == file_parents.values()[0].text_sha1
2027
and text_size == file_parents.values()[0].text_size):
2028
previous_ie = file_parents.values()[0]
2029
versionedfile = self.repository.weave_store.get_weave(file_id,
2030
self.repository.get_transaction())
2031
versionedfile.clone_text(self._new_revision_id,
2032
previous_ie.revision, file_parents.keys())
2033
return text_sha1, text_size
2034
else:
2035
new_lines = get_content_byte_lines()
2036
# TODO: Rather than invoking sha_strings here, _add_text_to_weave
2037
# should return the SHA1 and size
2038
self._add_text_to_weave(file_id, new_lines, file_parents.keys())
2039
return osutils.sha_strings(new_lines), \
2040
sum(map(len, new_lines))
2041
2042
def modified_link(self, file_id, file_parents, link_target):
2043
"""Record the presence of a symbolic link.
2044
2045
:param file_id: The file_id of the link to record.
2046
:param file_parents: The per-file parent revision ids.
2047
:param link_target: Target location of this link.
2048
"""
2049
self._add_text_to_weave(file_id, [], file_parents.keys())
2050
2051
def _add_text_to_weave(self, file_id, new_lines, parents):
2052
versionedfile = self.repository.weave_store.get_weave_or_empty(
2053
file_id, self.repository.get_transaction())
2054
versionedfile.add_lines(self._new_revision_id, parents, new_lines)
2055
versionedfile.clear_cache()
2056
2057
2058
class _CommitBuilder(CommitBuilder):
2059
"""Temporary class so old CommitBuilders are detected properly
2060
2061
Note: CommitBuilder works whether or not root entry is recorded.
2062
"""
2063
2064
record_root_entry = True
2065
2066
2067
class RootCommitBuilder(CommitBuilder):
2068
"""This commitbuilder actually records the root id"""
2069
2070
record_root_entry = True
2071
2072
def record_entry_contents(self, ie, parent_invs, path, tree):
2073
"""Record the content of ie from tree into the commit if needed.
2074
2075
Side effect: sets ie.revision when unchanged
2076
2077
:param ie: An inventory entry present in the commit.
2078
:param parent_invs: The inventories of the parent revisions of the
2079
commit.
2080
:param path: The path the entry is at in the tree.
2081
:param tree: The tree which contains this entry and should be used to
2082
obtain content.
2083
"""
2084
assert self.new_inventory.root is not None or ie.parent_id is None
2085
self.new_inventory.add(ie)
2086
2087
# ie.revision is always None if the InventoryEntry is considered
2088
# for committing. ie.snapshot will record the correct revision
2089
# which may be the sole parent if it is untouched.
2090
if ie.revision is not None:
2091
return
2092
2093
previous_entries = ie.find_previous_heads(
2094
parent_invs,
2095
self.repository.weave_store,
2096
self.repository.get_transaction())
2097
# we are creating a new revision for ie in the history store
2098
# and inventory.
2099
ie.snapshot(self._new_revision_id, path, previous_entries, tree, self)
2100
2101
3822
2102
_unescape_map = {
3823
2103
'apos':"'",
3824
2104
'quot':'"',
3847
2127
if _unescape_re is None:
3848
2128
_unescape_re = re.compile('\&([^;]*);')
3849
2129
return _unescape_re.sub(_unescaper, data)
3850
3851
3852
class _VersionedFileChecker(object):
3853
3854
def __init__(self, repository, text_key_references=None):
3855
self.repository = repository
3856
self.text_index = self.repository._generate_text_key_index(
3857
text_key_references=text_key_references)
3858
3859
def calculate_file_version_parents(self, text_key):
3860
"""Calculate the correct parents for a file version according to
3861
the inventories.
3862
"""
3863
parent_keys = self.text_index[text_key]
3864
if parent_keys == [_mod_revision.NULL_REVISION]:
3865
return ()
3866
return tuple(parent_keys)
3867
3868
def check_file_version_parents(self, texts, progress_bar=None):
3869
"""Check the parents stored in a versioned file are correct.
3870
3871
It also detects file versions that are not referenced by their
3872
corresponding revision's inventory.
3873
3874
:returns: A tuple of (wrong_parents, dangling_file_versions).
3875
wrong_parents is a dict mapping {revision_id: (stored_parents,
3876
correct_parents)} for each revision_id where the stored parents
3877
are not correct. dangling_file_versions is a set of (file_id,
3878
revision_id) tuples for versions that are present in this versioned
3879
file, but not used by the corresponding inventory.
3880
"""
3881
wrong_parents = {}
3882
self.file_ids = set([file_id for file_id, _ in
3883
self.text_index.iterkeys()])
3884
# text keys is now grouped by file_id
3885
n_weaves = len(self.file_ids)
3886
files_in_revisions = {}
3887
revisions_of_files = {}
3888
n_versions = len(self.text_index)
3889
progress_bar.update('loading text store', 0, n_versions)
3890
parent_map = self.repository.texts.get_parent_map(self.text_index)
3891
# On unlistable transports this could well be empty/error...
3892
text_keys = self.repository.texts.keys()
3893
unused_keys = frozenset(text_keys) - set(self.text_index)
3894
for num, key in enumerate(self.text_index.iterkeys()):
3895
if progress_bar is not None:
3896
progress_bar.update('checking text graph', num, n_versions)
3897
correct_parents = self.calculate_file_version_parents(key)
3898
try:
3899
knit_parents = parent_map[key]
3900
except errors.RevisionNotPresent:
3901
# Missing text!
3902
knit_parents = None
3903
if correct_parents != knit_parents:
3904
wrong_parents[key] = (knit_parents, correct_parents)
3905
return wrong_parents, unused_keys
3906
3907
3908
def _old_get_graph(repository, revision_id):
3909
"""DO NOT USE. That is all. I'm serious."""
3910
graph = repository.get_graph()
3911
revision_graph = dict(((key, value) for key, value in
3912
graph.iter_ancestry([revision_id]) if value is not None))
3913
return _strip_NULL_ghosts(revision_graph)
3914
3915
3916
def _strip_NULL_ghosts(revision_graph):
3917
"""Also don't use this. more compatibility code for unmigrated clients."""
3918
# Filter ghosts, and null:
3919
if _mod_revision.NULL_REVISION in revision_graph:
3920
del revision_graph[_mod_revision.NULL_REVISION]
3921
for key, parents in revision_graph.items():
3922
revision_graph[key] = tuple(parent for parent in parents if parent
3923
in revision_graph)
3924
return revision_graph
3925
3926
3927
class StreamSink(object):
3928
"""An object that can insert a stream into a repository.
3929
3930
This interface handles the complexity of reserialising inventories and
3931
revisions from different formats, and allows unidirectional insertion into
3932
stacked repositories without looking for the missing basis parents
3933
beforehand.
3934
"""
3935
3936
def __init__(self, target_repo):
3937
self.target_repo = target_repo
3938
3939
def insert_stream(self, stream, src_format, resume_tokens):
3940
"""Insert a stream's content into the target repository.
3941
3942
:param src_format: a bzr repository format.
3943
3944
:return: a list of resume tokens and an iterable of keys additional
3945
items required before the insertion can be completed.
3946
"""
3947
self.target_repo.lock_write()
3948
try:
3949
if resume_tokens:
3950
self.target_repo.resume_write_group(resume_tokens)
3951
is_resume = True
3952
else:
3953
self.target_repo.start_write_group()
3954
is_resume = False
3955
try:
3956
# locked_insert_stream performs a commit|suspend.
3957
return self._locked_insert_stream(stream, src_format, is_resume)
3958
except:
3959
self.target_repo.abort_write_group(suppress_errors=True)
3960
raise
3961
finally:
3962
self.target_repo.unlock()
3963
3964
def _locked_insert_stream(self, stream, src_format, is_resume):
3965
to_serializer = self.target_repo._format._serializer
3966
src_serializer = src_format._serializer
3967
new_pack = None
3968
if to_serializer == src_serializer:
3969
# If serializers match and the target is a pack repository, set the
3970
# write cache size on the new pack. This avoids poor performance
3971
# on transports where append is unbuffered (such as
3972
# RemoteTransport). This is safe to do because nothing should read
3973
# back from the target repository while a stream with matching
3974
# serialization is being inserted.
3975
# The exception is that a delta record from the source that should
3976
# be a fulltext may need to be expanded by the target (see
3977
# test_fetch_revisions_with_deltas_into_pack); but we take care to
3978
# explicitly flush any buffered writes first in that rare case.
3979
try:
3980
new_pack = self.target_repo._pack_collection._new_pack
3981
except AttributeError:
3982
# Not a pack repository
3983
pass
3984
else:
3985
new_pack.set_write_cache_size(1024*1024)
3986
for substream_type, substream in stream:
3987
if substream_type == 'texts':
3988
self.target_repo.texts.insert_record_stream(substream)
3989
elif substream_type == 'inventories':
3990
if src_serializer == to_serializer:
3991
self.target_repo.inventories.insert_record_stream(
3992
substream)
3993
else:
3994
self._extract_and_insert_inventories(
3995
substream, src_serializer)
3996
elif substream_type == 'chk_bytes':
3997
# XXX: This doesn't support conversions, as it assumes the
3998
# conversion was done in the fetch code.
3999
self.target_repo.chk_bytes.insert_record_stream(substream)
4000
elif substream_type == 'revisions':
4001
# This may fallback to extract-and-insert more often than
4002
# required if the serializers are different only in terms of
4003
# the inventory.
4004
if src_serializer == to_serializer:
4005
self.target_repo.revisions.insert_record_stream(
4006
substream)
4007
else:
4008
self._extract_and_insert_revisions(substream,
4009
src_serializer)
4010
elif substream_type == 'signatures':
4011
self.target_repo.signatures.insert_record_stream(substream)
4012
else:
4013
raise AssertionError('kaboom! %s' % (substream_type,))
4014
# Done inserting data, and the missing_keys calculations will try to
4015
# read back from the inserted data, so flush the writes to the new pack
4016
# (if this is pack format).
4017
if new_pack is not None:
4018
new_pack._write_data('', flush=True)
4019
# Find all the new revisions (including ones from resume_tokens)
4020
missing_keys = self.target_repo.get_missing_parent_inventories(
4021
check_for_missing_texts=is_resume)
4022
try:
4023
for prefix, versioned_file in (
4024
('texts', self.target_repo.texts),
4025
('inventories', self.target_repo.inventories),
4026
('revisions', self.target_repo.revisions),
4027
('signatures', self.target_repo.signatures),
4028
('chk_bytes', self.target_repo.chk_bytes),
4029
):
4030
if versioned_file is None:
4031
continue
4032
missing_keys.update((prefix,) + key for key in
4033
versioned_file.get_missing_compression_parent_keys())
4034
except NotImplementedError:
4035
# cannot even attempt suspending, and missing would have failed
4036
# during stream insertion.
4037
missing_keys = set()
4038
else:
4039
if missing_keys:
4040
# suspend the write group and tell the caller what we is
4041
# missing. We know we can suspend or else we would not have
4042
# entered this code path. (All repositories that can handle
4043
# missing keys can handle suspending a write group).
4044
write_group_tokens = self.target_repo.suspend_write_group()
4045
return write_group_tokens, missing_keys
4046
hint = self.target_repo.commit_write_group()
4047
if (to_serializer != src_serializer and
4048
self.target_repo._format.pack_compresses):
4049
self.target_repo.pack(hint=hint)
4050
return [], set()
4051
4052
def _extract_and_insert_inventories(self, substream, serializer):
4053
"""Generate a new inventory versionedfile in target, converting data.
4054
4055
The inventory is retrieved from the source, (deserializing it), and
4056
stored in the target (reserializing it in a different format).
4057
"""
4058
for record in substream:
4059
bytes = record.get_bytes_as('fulltext')
4060
revision_id = record.key[0]
4061
inv = serializer.read_inventory_from_string(bytes, revision_id)
4062
parents = [key[0] for key in record.parents]
4063
self.target_repo.add_inventory(revision_id, inv, parents)
4064
4065
def _extract_and_insert_revisions(self, substream, serializer):
4066
for record in substream:
4067
bytes = record.get_bytes_as('fulltext')
4068
revision_id = record.key[0]
4069
rev = serializer.read_revision_from_string(bytes)
4070
if rev.revision_id != revision_id:
4071
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4072
self.target_repo.add_revision(revision_id, rev)
4073
4074
def finished(self):
4075
if self.target_repo._format._fetch_reconcile:
4076
self.target_repo.reconcile()
4077
4078
4079
class StreamSource(object):
4080
"""A source of a stream for fetching between repositories."""
4081
4082
def __init__(self, from_repository, to_format):
4083
"""Create a StreamSource streaming from from_repository."""
4084
self.from_repository = from_repository
4085
self.to_format = to_format
4086
4087
def delta_on_metadata(self):
4088
"""Return True if delta's are permitted on metadata streams.
4089
4090
That is on revisions and signatures.
4091
"""
4092
src_serializer = self.from_repository._format._serializer
4093
target_serializer = self.to_format._serializer
4094
return (self.to_format._fetch_uses_deltas and
4095
src_serializer == target_serializer)
4096
4097
def _fetch_revision_texts(self, revs):
4098
# fetch signatures first and then the revision texts
4099
# may need to be a InterRevisionStore call here.
4100
from_sf = self.from_repository.signatures
4101
# A missing signature is just skipped.
4102
keys = [(rev_id,) for rev_id in revs]
4103
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4104
keys,
4105
self.to_format._fetch_order,
4106
not self.to_format._fetch_uses_deltas))
4107
# If a revision has a delta, this is actually expanded inside the
4108
# insert_record_stream code now, which is an alternate fix for
4109
# bug #261339
4110
from_rf = self.from_repository.revisions
4111
revisions = from_rf.get_record_stream(
4112
keys,
4113
self.to_format._fetch_order,
4114
not self.delta_on_metadata())
4115
return [('signatures', signatures), ('revisions', revisions)]
4116
4117
def _generate_root_texts(self, revs):
4118
"""This will be called by __fetch between fetching weave texts and
4119
fetching the inventory weave.
4120
4121
Subclasses should override this if they need to generate root texts
4122
after fetching weave texts.
4123
"""
4124
if self._rich_root_upgrade():
4125
import bzrlib.fetch
4126
return bzrlib.fetch.Inter1and2Helper(
4127
self.from_repository).generate_root_texts(revs)
4128
else:
4129
return []
4130
4131
def get_stream(self, search):
4132
phase = 'file'
4133
revs = search.get_keys()
4134
graph = self.from_repository.get_graph()
4135
revs = list(graph.iter_topo_order(revs))
4136
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4137
text_keys = []
4138
for knit_kind, file_id, revisions in data_to_fetch:
4139
if knit_kind != phase:
4140
phase = knit_kind
4141
# Make a new progress bar for this phase
4142
if knit_kind == "file":
4143
# Accumulate file texts
4144
text_keys.extend([(file_id, revision) for revision in
4145
revisions])
4146
elif knit_kind == "inventory":
4147
# Now copy the file texts.
4148
from_texts = self.from_repository.texts
4149
yield ('texts', from_texts.get_record_stream(
4150
text_keys, self.to_format._fetch_order,
4151
not self.to_format._fetch_uses_deltas))
4152
# Cause an error if a text occurs after we have done the
4153
# copy.
4154
text_keys = None
4155
# Before we process the inventory we generate the root
4156
# texts (if necessary) so that the inventories references
4157
# will be valid.
4158
for _ in self._generate_root_texts(revs):
4159
yield _
4160
# NB: This currently reopens the inventory weave in source;
4161
# using a single stream interface instead would avoid this.
4162
from_weave = self.from_repository.inventories
4163
# we fetch only the referenced inventories because we do not
4164
# know for unselected inventories whether all their required
4165
# texts are present in the other repository - it could be
4166
# corrupt.
4167
for info in self._get_inventory_stream(revs):
4168
yield info
4169
elif knit_kind == "signatures":
4170
# Nothing to do here; this will be taken care of when
4171
# _fetch_revision_texts happens.
4172
pass
4173
elif knit_kind == "revisions":
4174
for record in self._fetch_revision_texts(revs):
4175
yield record
4176
else:
4177
raise AssertionError("Unknown knit kind %r" % knit_kind)
4178
4179
def get_stream_for_missing_keys(self, missing_keys):
4180
# missing keys can only occur when we are byte copying and not
4181
# translating (because translation means we don't send
4182
# unreconstructable deltas ever).
4183
keys = {}
4184
keys['texts'] = set()
4185
keys['revisions'] = set()
4186
keys['inventories'] = set()
4187
keys['chk_bytes'] = set()
4188
keys['signatures'] = set()
4189
for key in missing_keys:
4190
keys[key[0]].add(key[1:])
4191
if len(keys['revisions']):
4192
# If we allowed copying revisions at this point, we could end up
4193
# copying a revision without copying its required texts: a
4194
# violation of the requirements for repository integrity.
4195
raise AssertionError(
4196
'cannot copy revisions to fill in missing deltas %s' % (
4197
keys['revisions'],))
4198
for substream_kind, keys in keys.iteritems():
4199
vf = getattr(self.from_repository, substream_kind)
4200
if vf is None and keys:
4201
raise AssertionError(
4202
"cannot fill in keys for a versioned file we don't"
4203
" have: %s needs %s" % (substream_kind, keys))
4204
if not keys:
4205
# No need to stream something we don't have
4206
continue
4207
# Ask for full texts always so that we don't need more round trips
4208
# after this stream.
4209
# Some of the missing keys are genuinely ghosts, so filter absent
4210
# records. The Sink is responsible for doing another check to
4211
# ensure that ghosts don't introduce missing data for future
4212
# fetches.
4213
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4214
self.to_format._fetch_order, True))
4215
yield substream_kind, stream
4216
4217
def inventory_fetch_order(self):
4218
if self._rich_root_upgrade():
4219
return 'topological'
4220
else:
4221
return self.to_format._fetch_order
4222
4223
def _rich_root_upgrade(self):
4224
return (not self.from_repository._format.rich_root_data and
4225
self.to_format.rich_root_data)
4226
4227
def _get_inventory_stream(self, revision_ids):
4228
from_format = self.from_repository._format
4229
if (from_format.supports_chks and self.to_format.supports_chks
4230
and (from_format._serializer == self.to_format._serializer)):
4231
# Both sides support chks, and they use the same serializer, so it
4232
# is safe to transmit the chk pages and inventory pages across
4233
# as-is.
4234
return self._get_chk_inventory_stream(revision_ids)
4235
elif (not from_format.supports_chks):
4236
# Source repository doesn't support chks. So we can transmit the
4237
# inventories 'as-is' and either they are just accepted on the
4238
# target, or the Sink will properly convert it.
4239
return self._get_simple_inventory_stream(revision_ids)
4240
else:
4241
# XXX: Hack to make not-chk->chk fetch: copy the inventories as
4242
# inventories. Note that this should probably be done somehow
4243
# as part of bzrlib.repository.StreamSink. Except JAM couldn't
4244
# figure out how a non-chk repository could possibly handle
4245
# deserializing an inventory stream from a chk repo, as it
4246
# doesn't have a way to understand individual pages.
4247
return self._get_convertable_inventory_stream(revision_ids)
4248
4249
def _get_simple_inventory_stream(self, revision_ids):
4250
from_weave = self.from_repository.inventories
4251
yield ('inventories', from_weave.get_record_stream(
4252
[(rev_id,) for rev_id in revision_ids],
4253
self.inventory_fetch_order(),
4254
not self.delta_on_metadata()))
4255
4256
def _get_chk_inventory_stream(self, revision_ids):
4257
"""Fetch the inventory texts, along with the associated chk maps."""
4258
# We want an inventory outside of the search set, so that we can filter
4259
# out uninteresting chk pages. For now we use
4260
# _find_revision_outside_set, but if we had a Search with cut_revs, we
4261
# could use that instead.
4262
start_rev_id = self.from_repository._find_revision_outside_set(
4263
revision_ids)
4264
start_rev_key = (start_rev_id,)
4265
inv_keys_to_fetch = [(rev_id,) for rev_id in revision_ids]
4266
if start_rev_id != _mod_revision.NULL_REVISION:
4267
inv_keys_to_fetch.append((start_rev_id,))
4268
# Any repo that supports chk_bytes must also support out-of-order
4269
# insertion. At least, that is how we expect it to work
4270
# We use get_record_stream instead of iter_inventories because we want
4271
# to be able to insert the stream as well. We could instead fetch
4272
# allowing deltas, and then iter_inventories, but we don't know whether
4273
# source or target is more 'local' anway.
4274
inv_stream = self.from_repository.inventories.get_record_stream(
4275
inv_keys_to_fetch, 'unordered',
4276
True) # We need them as full-texts so we can find their references
4277
uninteresting_chk_roots = set()
4278
interesting_chk_roots = set()
4279
def filter_inv_stream(inv_stream):
4280
for idx, record in enumerate(inv_stream):
4281
### child_pb.update('fetch inv', idx, len(inv_keys_to_fetch))
4282
bytes = record.get_bytes_as('fulltext')
4283
chk_inv = inventory.CHKInventory.deserialise(
4284
self.from_repository.chk_bytes, bytes, record.key)
4285
if record.key == start_rev_key:
4286
uninteresting_chk_roots.add(chk_inv.id_to_entry.key())
4287
p_id_map = chk_inv.parent_id_basename_to_file_id
4288
if p_id_map is not None:
4289
uninteresting_chk_roots.add(p_id_map.key())
4290
else:
4291
yield record
4292
interesting_chk_roots.add(chk_inv.id_to_entry.key())
4293
p_id_map = chk_inv.parent_id_basename_to_file_id
4294
if p_id_map is not None:
4295
interesting_chk_roots.add(p_id_map.key())
4296
### pb.update('fetch inventory', 0, 2)
4297
yield ('inventories', filter_inv_stream(inv_stream))
4298
# Now that we have worked out all of the interesting root nodes, grab
4299
# all of the interesting pages and insert them
4300
### pb.update('fetch inventory', 1, 2)
4301
interesting = chk_map.iter_interesting_nodes(
4302
self.from_repository.chk_bytes, interesting_chk_roots,
4303
uninteresting_chk_roots)
4304
def to_stream_adapter():
4305
"""Adapt the iter_interesting_nodes result to a single stream.
4306
4307
iter_interesting_nodes returns records as it processes them, along
4308
with keys. However, we only want to return the records themselves.
4309
"""
4310
for record, items in interesting:
4311
if record is not None:
4312
yield record
4313
# XXX: We could instead call get_record_stream(records.keys())
4314
# ATM, this will always insert the records as fulltexts, and
4315
# requires that you can hang on to records once you have gone
4316
# on to the next one. Further, it causes the target to
4317
# recompress the data. Testing shows it to be faster than
4318
# requesting the records again, though.
4319
yield ('chk_bytes', to_stream_adapter())
4320
### pb.update('fetch inventory', 2, 2)
4321
4322
def _get_convertable_inventory_stream(self, revision_ids):
4323
# XXX: One of source or target is using chks, and they don't have
4324
# compatible serializations. The StreamSink code expects to be
4325
# able to convert on the target, so we need to put
4326
# bytes-on-the-wire that can be converted
4327
yield ('inventories', self._stream_invs_as_fulltexts(revision_ids))
4328
4329
def _stream_invs_as_fulltexts(self, revision_ids):
4330
from_repo = self.from_repository
4331
from_serializer = from_repo._format._serializer
4332
revision_keys = [(rev_id,) for rev_id in revision_ids]
4333
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4334
for inv in self.from_repository.iter_inventories(revision_ids):
4335
# XXX: This is a bit hackish, but it works. Basically,
4336
# CHKSerializer 'accidentally' supports
4337
# read/write_inventory_to_string, even though that is never
4338
# the format that is stored on disk. It *does* give us a
4339
# single string representation for an inventory, so live with
4340
# it for now.
4341
# This would be far better if we had a 'serialized inventory
4342
# delta' form. Then we could use 'inventory._make_delta', and
4343
# transmit that. This would both be faster to generate, and
4344
# result in fewer bytes-on-the-wire.
4345
as_bytes = from_serializer.write_inventory_to_string(inv)
4346
key = (inv.revision_id,)
4347
parent_keys = parent_map.get(key, ())
4348
yield versionedfile.FulltextContentFactory(
4349
key, parent_keys, None, as_bytes)
4350
4351
4352
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4353
stop_revision=None):
4354
"""Extend the partial history to include a given index
4355
4356
If a stop_index is supplied, stop when that index has been reached.
4357
If a stop_revision is supplied, stop when that revision is
4358
encountered. Otherwise, stop when the beginning of history is
4359
reached.
4360
4361
:param stop_index: The index which should be present. When it is
4362
present, history extension will stop.
4363
:param stop_revision: The revision id which should be present. When
4364
it is encountered, history extension will stop.
4365
"""
4366
start_revision = partial_history_cache[-1]
4367
iterator = repo.iter_reverse_revision_history(start_revision)
4368
try:
4369
#skip the last revision in the list
4370
iterator.next()
4371
while True:
4372
if (stop_index is not None and
4373
len(partial_history_cache) > stop_index):
4374
break
4375
if partial_history_cache[-1] == stop_revision:
4376
break
4377
revision_id = iterator.next()
4378
partial_history_cache.append(revision_id)
4379
except StopIteration:
4380
# No more history
4381
return
4382
Loggerhead is a web-based interface for Breezy
Version: 2.0.1