136
174
working inventory.
138
176
def __init__(self,
140
if reporter is not None:
141
self.reporter = reporter
143
self.reporter = NullCommitReporter()
179
"""Create a Commit object.
181
:param reporter: the default reporter to use or None to decide later
183
self.reporter = reporter
151
191
specific_files=None,
153
193
allow_pointless=True,
201
message_callback=None,
204
possible_master_transports=None):
155
205
"""Commit working copy as a new revision.
157
timestamp -- if not None, seconds-since-epoch for a
158
postdated/predated commit.
160
specific_files -- If true, commit only those files.
162
rev_id -- If set, use this as the new revision id.
207
:param message: the commit message (it or message_callback is required)
208
:param message_callback: A callback: message = message_callback(cmt_obj)
210
:param timestamp: if not None, seconds-since-epoch for a
211
postdated/predated commit.
213
:param specific_files: If not None, commit only those files. An empty
214
list means 'commit no files'.
216
:param rev_id: If set, use this as the new revision id.
163
217
Useful for test or import commands that need to tightly
164
218
control what revisions are assigned. If you duplicate
165
219
a revision id that exists elsewhere it is your own fault.
166
220
If null (default), a time/random revision id is generated.
168
allow_pointless -- If true (default), commit even if nothing
222
:param allow_pointless: If true (default), commit even if nothing
169
223
has changed and no merges are recorded.
225
:param strict: If true, don't allow a commit if the working tree
226
contains unknown files.
228
:param revprops: Properties for new revision
229
:param local: Perform a local only commit.
230
:param reporter: the reporter to use or None for the default
231
:param verbose: if True and the reporter is not None, report everything
232
:param recursive: If set to 'down', commit in any subtrees that have
233
pending changes of any sort during this commit.
234
:param exclude: None or a list of relative paths to exclude from the
235
commit. Pending changes to excluded files will be ignored by the
238
operation = OperationWithCleanups(self._commit)
239
return operation.run(
244
specific_files=specific_files,
246
allow_pointless=allow_pointless,
250
working_tree=working_tree,
254
message_callback=message_callback,
257
possible_master_transports=possible_master_transports)
259
def _commit(self, operation, message, timestamp, timezone, committer,
260
specific_files, rev_id, allow_pointless, strict, verbose, revprops,
261
working_tree, local, reporter, config, message_callback, recursive,
262
exclude, possible_master_transports):
171
263
mutter('preparing to commit')
174
self.weave_store = branch.weave_store
176
self.specific_files = specific_files
265
if working_tree is None:
266
raise BzrError("working_tree must be passed into commit().")
268
self.work_tree = working_tree
269
self.branch = self.work_tree.branch
270
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
271
if not self.branch.repository.supports_rich_root():
272
raise errors.RootNotRich()
273
if message_callback is None:
274
if message is not None:
275
if isinstance(message, str):
276
message = message.decode(get_user_encoding())
277
message_callback = lambda x: message
279
raise BzrError("The message or message_callback keyword"
280
" parameter is required for commit().")
282
self.bound_branch = None
283
self.any_entries_deleted = False
284
if exclude is not None:
285
self.exclude = sorted(
286
minimum_path_selection(exclude))
290
self.master_branch = None
291
self.recursive = recursive
293
# self.specific_files is None to indicate no filter, or any iterable to
294
# indicate a filter - [] means no files at all, as per iter_changes.
295
if specific_files is not None:
296
self.specific_files = sorted(
297
minimum_path_selection(specific_files))
299
self.specific_files = None
177
301
self.allow_pointless = allow_pointless
179
if timestamp is None:
180
self.timestamp = time.time()
182
self.timestamp = long(timestamp)
185
self.rev_id = _gen_revision_id(self.branch, self.timestamp)
189
if committer is None:
190
self.committer = username(self.branch)
192
assert isinstance(committer, basestring), type(committer)
193
self.committer = committer
196
self.timezone = local_time_offset()
198
self.timezone = int(timezone)
200
assert isinstance(message, basestring), type(message)
201
self.message = message
202
self._escape_commit_message()
204
self.branch.lock_write()
206
self.work_tree = self.branch.working_tree()
207
self.work_inv = self.work_tree.inventory
208
self.basis_tree = self.branch.basis_tree()
209
self.basis_inv = self.basis_tree.inventory
211
self._gather_parents()
212
if len(self.parents) > 1 and self.specific_files:
213
raise NotImplementedError('selected-file commit of merges is not supported yet')
214
self._check_parents_present()
216
self._remove_deleted()
217
self.new_inv = Inventory()
218
self._store_entries()
219
self._report_deletes()
220
self._set_name_versions()
222
if not (self.allow_pointless
223
or len(self.parents) > 1
224
or self.new_inv != self.basis_inv):
225
raise PointlessCommit()
227
self._record_inventory()
228
self._record_ancestry()
229
self._make_revision()
230
note('committed r%d {%s}', (self.branch.revno() + 1),
232
self.branch.append_revision(self.rev_id)
233
self.branch.set_pending_merges([])
239
def _record_inventory(self):
240
"""Store the inventory for the new revision."""
241
inv_text = serializer_v5.write_inventory_to_string(self.new_inv)
242
self.inv_sha1 = sha_string(inv_text)
243
s = self.branch.control_weaves
244
s.add_text('inventory', self.rev_id,
245
split_lines(inv_text), self.parents)
247
def _escape_commit_message(self):
248
"""Replace xml-incompatible control characters."""
249
# Python strings can include characters that can't be
250
# represented in well-formed XML; escape characters that
251
# aren't listed in the XML specification
252
# (http://www.w3.org/TR/REC-xml/#NT-Char).
253
if isinstance(self.message, unicode):
254
char_pattern = u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]'
256
# Use a regular 'str' as pattern to avoid having re.subn
257
# return 'unicode' results.
258
char_pattern = '[^x09\x0A\x0D\x20-\xFF]'
259
self.message, escape_count = re.subn(
261
lambda match: match.group(0).encode('unicode_escape'),
264
note("replaced %d control characters in message", escape_count)
266
def _record_ancestry(self):
267
"""Append merged revision ancestry to the ancestry file.
269
This should be the merged ancestry of all parents, plus the
271
s = self.branch.control_weaves
272
w = s.get_weave_or_empty('ancestry')
273
lines = self._make_ancestry(w)
274
w.add(self.rev_id, self.parents, lines)
275
s.put_weave('ancestry', w)
278
def _make_ancestry(self, ancestry_weave):
279
"""Return merged ancestry lines.
281
The lines are revision-ids followed by newlines."""
282
parent_ancestries = [ancestry_weave.get(p) for p in self.parents]
283
new_lines = merge_ancestry_lines(self.rev_id, parent_ancestries)
284
mutter('merged ancestry of {%s}:\n%s', self.rev_id, ''.join(new_lines))
302
self.revprops = revprops
303
self.message_callback = message_callback
304
self.timestamp = timestamp
305
self.timezone = timezone
306
self.committer = committer
308
self.verbose = verbose
310
self.work_tree.lock_write()
311
operation.add_cleanup(self.work_tree.unlock)
312
self.parents = self.work_tree.get_parent_ids()
313
# We can use record_iter_changes IFF iter_changes is compatible with
314
# the command line parameters, and the repository has fast delta
315
# generation. See bug 347649.
316
self.use_record_iter_changes = (
318
not self.branch.repository._format.supports_tree_reference and
319
(self.branch.repository._format.fast_deltas or
320
len(self.parents) < 2))
321
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
322
operation.add_cleanup(self.pb.finished)
323
self.basis_revid = self.work_tree.last_revision()
324
self.basis_tree = self.work_tree.basis_tree()
325
self.basis_tree.lock_read()
326
operation.add_cleanup(self.basis_tree.unlock)
327
# Cannot commit with conflicts present.
328
if len(self.work_tree.conflicts()) > 0:
329
raise ConflictsInTree
331
# Setup the bound branch variables as needed.
332
self._check_bound_branch(operation, possible_master_transports)
334
# Check that the working tree is up to date
335
old_revno, new_revno = self._check_out_of_date_tree()
337
# Complete configuration setup
338
if reporter is not None:
339
self.reporter = reporter
340
elif self.reporter is None:
341
self.reporter = self._select_reporter()
342
if self.config is None:
343
self.config = self.branch.get_config()
345
self._set_specific_file_ids()
347
# Setup the progress bar. As the number of files that need to be
348
# committed in unknown, progress is reported as stages.
349
# We keep track of entries separately though and include that
350
# information in the progress bar during the relevant stages.
351
self.pb_stage_name = ""
352
self.pb_stage_count = 0
353
self.pb_stage_total = 5
354
if self.bound_branch:
355
self.pb_stage_total += 1
356
self.pb.show_pct = False
357
self.pb.show_spinner = False
358
self.pb.show_eta = False
359
self.pb.show_count = True
360
self.pb.show_bar = True
362
self._gather_parents()
363
# After a merge, a selected file commit is not supported.
364
# See 'bzr help merge' for an explanation as to why.
365
if len(self.parents) > 1 and self.specific_files is not None:
366
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
367
# Excludes are a form of selected file commit.
368
if len(self.parents) > 1 and self.exclude:
369
raise errors.CannotCommitSelectedFileMerge(self.exclude)
371
# Collect the changes
372
self._set_progress_stage("Collecting changes", counter=True)
373
self.builder = self.branch.get_commit_builder(self.parents,
374
self.config, timestamp, timezone, committer, revprops, rev_id)
377
self.builder.will_record_deletes()
378
# find the location being committed to
379
if self.bound_branch:
380
master_location = self.master_branch.base
382
master_location = self.branch.base
384
# report the start of the commit
385
self.reporter.started(new_revno, self.rev_id, master_location)
387
self._update_builder_with_changes()
388
self._check_pointless()
390
# TODO: Now the new inventory is known, check for conflicts.
391
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
392
# weave lines, because nothing should be recorded until it is known
393
# that commit will succeed.
394
self._set_progress_stage("Saving data locally")
395
self.builder.finish_inventory()
397
# Prompt the user for a commit message if none provided
398
message = message_callback(self)
399
self.message = message
401
# Add revision data to the local branch
402
self.rev_id = self.builder.commit(self.message)
405
mutter("aborting commit write group because of exception:")
406
trace.log_exception_quietly()
407
note("aborting commit write group: %r" % (e,))
411
self._process_pre_hooks(old_revno, new_revno)
413
# Upload revision data to the master.
414
# this will propagate merged revisions too if needed.
415
if self.bound_branch:
416
self._set_progress_stage("Uploading data to master branch")
417
# 'commit' to the master first so a timeout here causes the
418
# local branch to be out of date
419
self.master_branch.import_last_revision_info(
420
self.branch.repository, new_revno, self.rev_id)
422
# and now do the commit locally.
423
self.branch.set_last_revision_info(new_revno, self.rev_id)
425
# Make the working tree be up to date with the branch. This
426
# includes automatic changes scheduled to be made to the tree, such
427
# as updating its basis and unversioning paths that were missing.
428
self.work_tree.unversion(self.deleted_ids)
429
self._set_progress_stage("Updating the working tree")
430
self.work_tree.update_basis_by_delta(self.rev_id,
431
self.builder.get_basis_delta())
432
self.reporter.completed(new_revno, self.rev_id)
433
self._process_post_hooks(old_revno, new_revno)
436
def _select_reporter(self):
437
"""Select the CommitReporter to use."""
439
return NullCommitReporter()
440
return ReportCommitToLog()
442
def _check_pointless(self):
443
if self.allow_pointless:
445
# A merge with no effect on files
446
if len(self.parents) > 1:
448
# TODO: we could simplify this by using self.builder.basis_delta.
450
# The initial commit adds a root directory, but this in itself is not
451
# a worthwhile commit.
452
if (self.basis_revid == revision.NULL_REVISION and
453
((self.builder.new_inventory is not None and
454
len(self.builder.new_inventory) == 1) or
455
len(self.builder._basis_delta) == 1)):
456
raise PointlessCommit()
457
if self.builder.any_changes():
459
raise PointlessCommit()
461
def _check_bound_branch(self, operation, possible_master_transports=None):
462
"""Check to see if the local branch is bound.
464
If it is bound, then most of the commit will actually be
465
done using the remote branch as the target branch.
466
Only at the end will the local branch be updated.
468
if self.local and not self.branch.get_bound_location():
469
raise errors.LocalRequiresBoundBranch()
472
self.master_branch = self.branch.get_master_branch(
473
possible_master_transports)
475
if not self.master_branch:
476
# make this branch the reference branch for out of date checks.
477
self.master_branch = self.branch
480
# If the master branch is bound, we must fail
481
master_bound_location = self.master_branch.get_bound_location()
482
if master_bound_location:
483
raise errors.CommitToDoubleBoundBranch(self.branch,
484
self.master_branch, master_bound_location)
486
# TODO: jam 20051230 We could automatically push local
487
# commits to the remote branch if they would fit.
488
# But for now, just require remote to be identical
491
# Make sure the local branch is identical to the master
492
master_info = self.master_branch.last_revision_info()
493
local_info = self.branch.last_revision_info()
494
if local_info != master_info:
495
raise errors.BoundBranchOutOfDate(self.branch,
498
# Now things are ready to change the master branch
500
self.bound_branch = self.branch
501
self.master_branch.lock_write()
502
operation.add_cleanup(self.master_branch.unlock)
504
def _check_out_of_date_tree(self):
505
"""Check that the working tree is up to date.
507
:return: old_revision_number,new_revision_number tuple
510
first_tree_parent = self.work_tree.get_parent_ids()[0]
512
# if there are no parents, treat our parent as 'None'
513
# this is so that we still consider the master branch
514
# - in a checkout scenario the tree may have no
515
# parents but the branch may do.
516
first_tree_parent = bzrlib.revision.NULL_REVISION
517
old_revno, master_last = self.master_branch.last_revision_info()
518
if master_last != first_tree_parent:
519
if master_last != bzrlib.revision.NULL_REVISION:
520
raise errors.OutOfDateTree(self.work_tree)
521
if self.branch.repository.has_revision(first_tree_parent):
522
new_revno = old_revno + 1
524
# ghost parents never appear in revision history.
526
return old_revno,new_revno
528
def _process_pre_hooks(self, old_revno, new_revno):
529
"""Process any registered pre commit hooks."""
530
self._set_progress_stage("Running pre_commit hooks")
531
self._process_hooks("pre_commit", old_revno, new_revno)
533
def _process_post_hooks(self, old_revno, new_revno):
534
"""Process any registered post commit hooks."""
535
# Process the post commit hooks, if any
536
self._set_progress_stage("Running post_commit hooks")
537
# old style commit hooks - should be deprecated ? (obsoleted in
539
if self.config.post_commit() is not None:
540
hooks = self.config.post_commit().split(' ')
541
# this would be nicer with twisted.python.reflect.namedAny
543
result = eval(hook + '(branch, rev_id)',
544
{'branch':self.branch,
546
'rev_id':self.rev_id})
547
# process new style post commit hooks
548
self._process_hooks("post_commit", old_revno, new_revno)
550
def _process_hooks(self, hook_name, old_revno, new_revno):
551
if not Branch.hooks[hook_name]:
554
# new style commit hooks:
555
if not self.bound_branch:
556
hook_master = self.branch
559
hook_master = self.master_branch
560
hook_local = self.branch
561
# With bound branches, when the master is behind the local branch,
562
# the 'old_revno' and old_revid values here are incorrect.
563
# XXX: FIXME ^. RBC 20060206
565
old_revid = self.parents[0]
567
old_revid = bzrlib.revision.NULL_REVISION
569
if hook_name == "pre_commit":
570
future_tree = self.builder.revision_tree()
571
tree_delta = future_tree.changes_from(self.basis_tree,
574
for hook in Branch.hooks[hook_name]:
575
# show the running hook in the progress bar. As hooks may
576
# end up doing nothing (e.g. because they are not configured by
577
# the user) this is still showing progress, not showing overall
578
# actions - its up to each plugin to show a UI if it want's to
579
# (such as 'Emailing diff to foo@example.com').
580
self.pb_stage_name = "Running %s hooks [%s]" % \
581
(hook_name, Branch.hooks.get_hook_name(hook))
582
self._emit_progress()
583
if 'hooks' in debug.debug_flags:
584
mutter("Invoking commit hook: %r", hook)
585
if hook_name == "post_commit":
586
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
588
elif hook_name == "pre_commit":
589
hook(hook_local, hook_master,
590
old_revno, old_revid, new_revno, self.rev_id,
591
tree_delta, future_tree)
288
593
def _gather_parents(self):
289
pending_merges = self.branch.pending_merges()
291
self.parent_trees = []
292
precursor_id = self.branch.last_revision()
294
self.parents.append(precursor_id)
295
self.parent_trees.append(self.basis_tree)
296
self.parents += pending_merges
297
self.parent_trees.extend(map(self.branch.revision_tree, pending_merges))
300
def _check_parents_present(self):
301
for parent_id in self.parents:
302
mutter('commit parent revision {%s}', parent_id)
303
if not self.branch.has_revision(parent_id):
304
warning("can't commit a merge from an absent parent")
305
raise HistoryMissing(self.branch, 'revision', parent_id)
308
def _make_revision(self):
309
"""Record a new revision object for this commit."""
310
self.rev = Revision(timestamp=self.timestamp,
311
timezone=self.timezone,
312
committer=self.committer,
313
message=self.message,
314
inventory_sha1=self.inv_sha1,
315
revision_id=self.rev_id)
316
self.rev.parent_ids = self.parents
318
serializer_v5.write_revision(self.rev, rev_tmp)
320
self.branch.revision_store.add(rev_tmp, self.rev_id, compressed=False)
321
mutter('new revision_id is {%s}', self.rev_id)
324
def _remove_deleted(self):
325
"""Remove deleted files from the working inventories.
327
This is done prior to taking the working inventory as the
328
basis for the new committed inventory.
330
This returns true if any files
331
*that existed in the basis inventory* were deleted.
332
Files that were added and deleted
333
in the working copy don't matter.
335
specific = self.specific_files
594
"""Record the parents of a merge for merge detection."""
595
# TODO: Make sure that this list doesn't contain duplicate
596
# entries and the order is preserved when doing this.
597
if self.use_record_iter_changes:
599
self.basis_inv = self.basis_tree.inventory
600
self.parent_invs = [self.basis_inv]
601
for revision in self.parents[1:]:
602
if self.branch.repository.has_revision(revision):
603
mutter('commit parent revision {%s}', revision)
604
inventory = self.branch.repository.get_inventory(revision)
605
self.parent_invs.append(inventory)
607
mutter('commit parent ghost revision {%s}', revision)
609
def _update_builder_with_changes(self):
610
"""Update the commit builder with the data about what has changed.
612
exclude = self.exclude
613
specific_files = self.specific_files
614
mutter("Selecting files for commit with filter %s", specific_files)
617
if self.use_record_iter_changes:
618
iter_changes = self.work_tree.iter_changes(self.basis_tree,
619
specific_files=specific_files)
620
iter_changes = self._filter_iter_changes(iter_changes)
621
for file_id, path, fs_hash in self.builder.record_iter_changes(
622
self.work_tree, self.basis_revid, iter_changes):
623
self.work_tree._observed_sha1(file_id, path, fs_hash)
625
# Build the new inventory
626
self._populate_from_inventory()
627
self._record_unselected()
628
self._report_and_accumulate_deletes()
630
def _filter_iter_changes(self, iter_changes):
631
"""Process iter_changes.
633
This method reports on the changes in iter_changes to the user, and
634
converts 'missing' entries in the iter_changes iterator to 'deleted'
635
entries. 'missing' entries have their
637
:param iter_changes: An iter_changes to process.
638
:return: A generator of changes.
640
reporter = self.reporter
641
report_changes = reporter.is_verbose()
337
for path, ie in self.work_inv.iter_entries():
338
if specific and not is_inside_any(specific, path):
340
if not self.work_tree.has_filename(path):
341
note('missing %s', path)
342
deleted_ids.append(ie.file_id)
643
for change in iter_changes:
645
old_path = change[1][0]
646
new_path = change[1][1]
647
versioned = change[3][1]
649
versioned = change[3][1]
650
if kind is None and versioned:
653
reporter.missing(new_path)
654
deleted_ids.append(change[0])
655
# Reset the new path (None) and new versioned flag (False)
656
change = (change[0], (change[1][0], None), change[2],
657
(change[3][0], False)) + change[4:]
658
elif kind == 'tree-reference':
659
if self.recursive == 'down':
660
self._commit_nested_tree(change[0], change[1][1])
661
if change[3][0] or change[3][1]:
665
reporter.deleted(old_path)
666
elif old_path is None:
667
reporter.snapshot_change('added', new_path)
668
elif old_path != new_path:
669
reporter.renamed('renamed', old_path, new_path)
672
self.work_tree.branch.repository._format.rich_root_data):
673
# Don't report on changes to '' in non rich root
675
reporter.snapshot_change('modified', new_path)
676
self._next_progress_entry()
677
# Unversion IDs that were found to be deleted
678
self.deleted_ids = deleted_ids
680
def _record_unselected(self):
681
# If specific files are selected, then all un-selected files must be
682
# recorded in their previous state. For more details, see
683
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
684
if self.specific_files or self.exclude:
685
specific_files = self.specific_files or []
686
for path, old_ie in self.basis_inv.iter_entries():
687
if old_ie.file_id in self.builder.new_inventory:
688
# already added - skip.
690
if (is_inside_any(specific_files, path)
691
and not is_inside_any(self.exclude, path)):
692
# was inside the selected path, and not excluded - if not
693
# present it has been deleted so skip.
695
# From here down it was either not selected, or was excluded:
696
# We preserve the entry unaltered.
698
# Note: specific file commits after a merge are currently
699
# prohibited. This test is for sanity/safety in case it's
700
# required after that changes.
701
if len(self.parents) > 1:
703
self.builder.record_entry_contents(ie, self.parent_invs, path,
704
self.basis_tree, None)
706
def _report_and_accumulate_deletes(self):
707
if (isinstance(self.basis_inv, Inventory)
708
and isinstance(self.builder.new_inventory, Inventory)):
709
# the older Inventory classes provide a _byid dict, and building a
710
# set from the keys of this dict is substantially faster than even
711
# getting a set of ids from the inventory
713
# <lifeless> set(dict) is roughly the same speed as
714
# set(iter(dict)) and both are significantly slower than
716
deleted_ids = set(self.basis_inv._byid.keys()) - \
717
set(self.builder.new_inventory._byid.keys())
719
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
344
for file_id in deleted_ids:
345
del self.work_inv[file_id]
346
self.branch._write_inventory(self.work_inv)
349
def _find_file_parents(self, file_id):
350
"""Return the text versions and hashes for all file parents.
352
Returned as a map from text version to inventory entry.
354
This is a set containing the file versions in all parents
355
revisions containing the file. If the file is new, the set
358
for tree in self.parent_trees:
359
if file_id in tree.inventory:
360
ie = tree.inventory[file_id]
361
assert ie.kind == 'file'
362
assert ie.file_id == file_id
363
if ie.text_version in r:
364
assert r[ie.text_version] == ie
366
r[ie.text_version] = ie
370
def _set_name_versions(self):
371
"""Pass over inventory and mark new entry version as needed.
373
Files get a new name version when they are new, have a
374
different parent, or a different name from in the
375
basis inventory, or if the file is in a different place
376
to any of the parents."""
377
# XXX: Need to think more here about when the user has
378
# made a specific decision on a particular value -- c.f.
380
for path, ie in self.new_inv.iter_entries():
383
for parent_tree in self.parent_trees:
384
parent_inv = parent_tree.inventory
385
if file_id not in parent_inv:
387
parent_ie = parent_inv[file_id]
388
if parent_ie.parent_id != ie.parent_id:
391
elif parent_ie.name != ie.name:
394
elif old_version is None:
395
old_version = parent_ie.name_version
396
elif old_version != parent_ie.name_version:
400
pass # so far so good
401
if old_version is None:
402
mutter('new name_version for {%s}', file_id)
403
ie.name_version = self.rev_id
405
mutter('name_version for {%s} inherited as {%s}',
406
file_id, old_version)
407
ie.name_version = old_version
410
def _store_entries(self):
411
"""Build revision inventory and store modified files.
413
This is called with new_inv a new empty inventory. Depending on
414
which files are selected for commit, and which ones have
415
been modified or merged, new inventory entries are built
416
based on the working and parent inventories.
418
As a side-effect this stores new text versions for committed
419
files with text changes or merges.
421
Each entry can have one of several things happen:
423
carry_file -- carried from the previous version (if not
426
commit_nonfile -- no text to worry about
428
commit_old_text -- same text, may have moved
430
commit_file -- new text version
721
self.any_entries_deleted = True
722
deleted = [(self.basis_tree.id2path(file_id), file_id)
723
for file_id in deleted_ids]
725
# XXX: this is not quite directory-order sorting
726
for path, file_id in deleted:
727
self.builder.record_delete(path, file_id)
728
self.reporter.deleted(path)
730
def _check_strict(self):
731
# XXX: when we use iter_changes this would likely be faster if
732
# iter_changes would check for us (even in the presence of
735
# raise an exception as soon as we find a single unknown.
736
for unknown in self.work_tree.unknowns():
737
raise StrictCommitFailed()
739
def _populate_from_inventory(self):
740
"""Populate the CommitBuilder by walking the working tree inventory."""
741
# Build the revision inventory.
743
# This starts by creating a new empty inventory. Depending on
744
# which files are selected for commit, and what is present in the
745
# current tree, the new inventory is populated. inventory entries
746
# which are candidates for modification have their revision set to
747
# None; inventory entries that are carried over untouched have their
748
# revision set to their prior value.
750
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
751
# results to create a new inventory at the same time, which results
752
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
755
specific_files = self.specific_files
756
exclude = self.exclude
757
report_changes = self.reporter.is_verbose()
759
# A tree of paths that have been deleted. E.g. if foo/bar has been
760
# deleted, then we have {'foo':{'bar':{}}}
762
# XXX: Note that entries may have the wrong kind because the entry does
763
# not reflect the status on disk.
764
work_inv = self.work_tree.inventory
765
# NB: entries will include entries within the excluded ids/paths
766
# because iter_entries_by_dir has no 'exclude' facility today.
767
entries = work_inv.iter_entries_by_dir(
768
specific_file_ids=self.specific_file_ids, yield_parents=True)
769
for path, existing_ie in entries:
770
file_id = existing_ie.file_id
771
name = existing_ie.name
772
parent_id = existing_ie.parent_id
773
kind = existing_ie.kind
774
# Skip files that have been deleted from the working tree.
775
# The deleted path ids are also recorded so they can be explicitly
778
path_segments = splitpath(path)
779
deleted_dict = deleted_paths
780
for segment in path_segments:
781
deleted_dict = deleted_dict.get(segment, None)
783
# We either took a path not present in the dict
784
# (deleted_dict was None), or we've reached an empty
785
# child dir in the dict, so are now a sub-path.
789
if deleted_dict is not None:
790
# the path has a deleted parent, do not add it.
792
if exclude and is_inside_any(exclude, path):
793
# Skip excluded paths. Excluded paths are processed by
794
# _update_builder_with_changes.
796
content_summary = self.work_tree.path_content_summary(path)
797
kind = content_summary[0]
798
# Note that when a filter of specific files is given, we must only
799
# skip/record deleted files matching that filter.
800
if not specific_files or is_inside_any(specific_files, path):
801
if kind == 'missing':
802
if not deleted_paths:
803
# path won't have been split yet.
804
path_segments = splitpath(path)
805
deleted_dict = deleted_paths
806
for segment in path_segments:
807
deleted_dict = deleted_dict.setdefault(segment, {})
808
self.reporter.missing(path)
809
self._next_progress_entry()
810
deleted_ids.append(file_id)
812
# TODO: have the builder do the nested commit just-in-time IF and
814
if kind == 'tree-reference':
815
# enforce repository nested tree policy.
816
if (not self.work_tree.supports_tree_reference() or
817
# repository does not support it either.
818
not self.branch.repository._format.supports_tree_reference):
820
content_summary = (kind, None, None, None)
821
elif self.recursive == 'down':
822
nested_revision_id = self._commit_nested_tree(
824
content_summary = (kind, None, None, nested_revision_id)
826
nested_revision_id = self.work_tree.get_reference_revision(file_id)
827
content_summary = (kind, None, None, nested_revision_id)
829
# Record an entry for this item
830
# Note: I don't particularly want to have the existing_ie
831
# parameter but the test suite currently (28-Jun-07) breaks
832
# without it thanks to a unicode normalisation issue. :-(
833
definitely_changed = kind != existing_ie.kind
834
self._record_entry(path, file_id, specific_files, kind, name,
835
parent_id, definitely_changed, existing_ie, report_changes,
838
# Unversion IDs that were found to be deleted
839
self.deleted_ids = deleted_ids
841
def _commit_nested_tree(self, file_id, path):
842
"Commit a nested tree."
843
sub_tree = self.work_tree.get_nested_tree(file_id, path)
844
# FIXME: be more comprehensive here:
845
# this works when both trees are in --trees repository,
846
# but when both are bound to a different repository,
847
# it fails; a better way of approaching this is to
848
# finally implement the explicit-caches approach design
849
# a while back - RBC 20070306.
850
if sub_tree.branch.repository.has_same_location(
851
self.work_tree.branch.repository):
852
sub_tree.branch.repository = \
853
self.work_tree.branch.repository
855
return sub_tree.commit(message=None, revprops=self.revprops,
856
recursive=self.recursive,
857
message_callback=self.message_callback,
858
timestamp=self.timestamp, timezone=self.timezone,
859
committer=self.committer,
860
allow_pointless=self.allow_pointless,
861
strict=self.strict, verbose=self.verbose,
862
local=self.local, reporter=self.reporter)
863
except errors.PointlessCommit:
864
return self.work_tree.get_reference_revision(file_id)
866
def _record_entry(self, path, file_id, specific_files, kind, name,
867
parent_id, definitely_changed, existing_ie, report_changes,
869
"Record the new inventory entry for a path if any."
870
# mutter('check %s {%s}', path, file_id)
871
# mutter('%s selected for commit', path)
872
if definitely_changed or existing_ie is None:
873
ie = make_entry(kind, name, parent_id, file_id)
875
ie = existing_ie.copy()
877
# For carried over entries we don't care about the fs hash - the repo
878
# isn't generating a sha, so we're not saving computation time.
879
_, _, fs_hash = self.builder.record_entry_contents(
880
ie, self.parent_invs, path, self.work_tree, content_summary)
882
self._report_change(ie, path)
884
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
887
def _report_change(self, ie, path):
888
"""Report a change to the user.
890
The change that has occurred is described relative to the basis
432
for path, new_ie in self.work_inv.iter_entries():
433
file_id = new_ie.file_id
434
mutter('check %s {%s}', path, new_ie.file_id)
435
if self.specific_files:
436
if not is_inside_any(self.specific_files, path):
437
mutter('%s not selected for commit', path)
438
self._carry_file(file_id)
440
if new_ie.kind != 'file':
441
self._commit_nonfile(file_id)
444
file_parents = self._find_file_parents(file_id)
445
mutter('parents of %s are %r', path, file_parents)
446
if len(file_parents) == 1:
447
parent_ie = file_parents.values()[0]
448
wc_sha1 = self.work_tree.get_file_sha1(file_id)
449
if parent_ie.text_sha1 == wc_sha1:
450
# text not changed or merged
451
self._commit_old_text(file_id, parent_ie)
453
# file is either new, or a file merge; need to record
455
if len(file_parents) > 1:
456
note('merged %s', path)
457
elif len(file_parents) == 0:
458
note('added %s', path)
893
if (self.basis_inv.has_id(ie.file_id)):
894
basis_ie = self.basis_inv[ie.file_id]
897
change = ie.describe_change(basis_ie, ie)
898
if change in (InventoryEntry.RENAMED,
899
InventoryEntry.MODIFIED_AND_RENAMED):
900
old_path = self.basis_inv.id2path(ie.file_id)
901
self.reporter.renamed(change, old_path, path)
902
self._next_progress_entry()
904
if change == 'unchanged':
906
self.reporter.snapshot_change(change, path)
907
self._next_progress_entry()
909
def _set_progress_stage(self, name, counter=False):
910
"""Set the progress stage and emit an update to the progress bar."""
911
self.pb_stage_name = name
912
self.pb_stage_count += 1
914
self.pb_entries_count = 0
916
self.pb_entries_count = None
917
self._emit_progress()
919
def _next_progress_entry(self):
920
"""Emit an update to the progress bar and increment the entry count."""
921
self.pb_entries_count += 1
922
self._emit_progress()
924
def _emit_progress(self):
925
if self.pb_entries_count is not None:
926
text = "%s [%d] - Stage" % (self.pb_stage_name,
927
self.pb_entries_count)
929
text = "%s - Stage" % (self.pb_stage_name, )
930
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
932
def _set_specific_file_ids(self):
933
"""populate self.specific_file_ids if we will use it."""
934
if not self.use_record_iter_changes:
935
# If provided, ensure the specified files are versioned
936
if self.specific_files is not None:
937
# Note: This routine is being called because it raises
938
# PathNotVersionedError as a side effect of finding the IDs. We
939
# later use the ids we found as input to the working tree
940
# inventory iterator, so we only consider those ids rather than
941
# examining the whole tree again.
942
# XXX: Dont we have filter_unversioned to do this more
944
self.specific_file_ids = tree.find_ids_across_trees(
945
self.specific_files, [self.basis_tree, self.work_tree])
460
note('modified %s', path)
461
self._commit_file(new_ie, file_id, file_parents)
464
def _commit_nonfile(self, file_id):
465
self.new_inv.add(self.work_inv[file_id].copy())
468
def _carry_file(self, file_id):
469
"""Carry the file unchanged from the basis revision."""
470
if self.basis_inv.has_id(file_id):
471
self.new_inv.add(self.basis_inv[file_id].copy())
474
def _commit_old_text(self, file_id, parent_ie):
475
"""Keep the same text as last time, but possibly a different name."""
476
ie = self.work_inv[file_id].copy()
477
ie.text_version = parent_ie.text_version
478
ie.text_size = parent_ie.text_size
479
ie.text_sha1 = parent_ie.text_sha1
483
def _report_deletes(self):
484
for file_id in self.basis_inv:
485
if file_id not in self.new_inv:
486
note('deleted %s', self.basis_inv.id2path(file_id))
489
def _commit_file(self, new_ie, file_id, file_parents):
490
mutter('store new text for {%s} in revision {%s}',
491
file_id, self.rev_id)
492
new_lines = self.work_tree.get_file(file_id).readlines()
493
self._add_text_to_weave(file_id, new_lines, file_parents)
494
new_ie.text_version = self.rev_id
495
new_ie.text_sha1 = sha_strings(new_lines)
496
new_ie.text_size = sum(map(len, new_lines))
497
self.new_inv.add(new_ie)
500
def _add_text_to_weave(self, file_id, new_lines, parents):
501
self.weave_store.add_text(file_id, self.rev_id, new_lines, parents)
504
def _gen_revision_id(branch, when):
505
"""Return new revision-id."""
506
s = '%s-%s-' % (user_email(branch), compact_date(when))
507
s += hexlify(rand_bytes(8))
513
def merge_ancestry_lines(rev_id, ancestries):
514
"""Return merged ancestry lines.
516
rev_id -- id of the new revision
518
ancestries -- a sequence of ancestries for parent revisions,
519
as newline-terminated line lists.
521
if len(ancestries) == 0:
522
return [rev_id + '\n']
523
seen = set(ancestries[0])
524
ancs = ancestries[0][:]
525
for parent_ancestry in ancestries[1:]:
526
for line in parent_ancestry:
527
assert line[-1] == '\n'
947
self.specific_file_ids = None
added
removed
bzrlib/commit.py
