283
281
self.committer = committer
284
282
self.strict = strict
285
283
self.verbose = verbose
286
# accumulates an inventory delta to the basis entry, so we can make
287
# just the necessary updates to the workingtree's cached basis.
288
self._basis_delta = []
290
285
self.work_tree.lock_write()
286
self.parents = self.work_tree.get_parent_ids()
287
# We can use record_iter_changes IFF iter_changes is compatible with
288
# the command line parameters, and the repository has fast delta
289
# generation. See bug 347649.
290
self.use_record_iter_changes = (
292
not self.branch.repository._format.supports_tree_reference and
293
(self.branch.repository._format.fast_deltas or
294
len(self.parents) < 2))
291
295
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
292
296
self.basis_revid = self.work_tree.last_revision()
293
297
self.basis_tree = self.work_tree.basis_tree()
311
315
if self.config is None:
312
316
self.config = self.branch.get_config()
314
# If provided, ensure the specified files are versioned
315
if self.specific_files is not None:
316
# Note: This routine is being called because it raises
317
# PathNotVersionedError as a side effect of finding the IDs. We
318
# later use the ids we found as input to the working tree
319
# inventory iterator, so we only consider those ids rather than
320
# examining the whole tree again.
321
# XXX: Dont we have filter_unversioned to do this more
323
self.specific_file_ids = tree.find_ids_across_trees(
324
specific_files, [self.basis_tree, self.work_tree])
318
self._set_specific_file_ids()
326
320
# Setup the progress bar. As the number of files that need to be
327
321
# committed in unknown, progress is reported as stages.
338
332
self.pb.show_count = True
339
333
self.pb.show_bar = True
341
self.basis_inv = self.basis_tree.inventory
342
335
self._gather_parents()
343
336
# After a merge, a selected file commit is not supported.
344
337
# See 'bzr help merge' for an explanation as to why.
345
if len(self.parents) > 1 and self.specific_files:
338
if len(self.parents) > 1 and self.specific_files is not None:
346
339
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
347
340
# Excludes are a form of selected file commit.
348
341
if len(self.parents) > 1 and self.exclude:
349
342
raise errors.CannotCommitSelectedFileMerge(self.exclude)
351
344
# Collect the changes
352
self._set_progress_stage("Collecting changes",
353
entries_title="Directory")
345
self._set_progress_stage("Collecting changes", counter=True)
354
346
self.builder = self.branch.get_commit_builder(self.parents,
355
347
self.config, timestamp, timezone, committer, revprops, rev_id)
350
self.builder.will_record_deletes()
358
351
# find the location being committed to
359
352
if self.bound_branch:
360
353
master_location = self.master_branch.base
392
386
# Upload revision data to the master.
393
387
# this will propagate merged revisions too if needed.
394
388
if self.bound_branch:
395
if not self.master_branch.repository.has_same_location(
396
self.branch.repository):
397
self._set_progress_stage("Uploading data to master branch")
398
self.master_branch.repository.fetch(self.branch.repository,
399
revision_id=self.rev_id)
400
# now the master has the revision data
389
self._set_progress_stage("Uploading data to master branch")
401
390
# 'commit' to the master first so a timeout here causes the
402
391
# local branch to be out of date
403
self.master_branch.set_last_revision_info(new_revno,
392
self.master_branch.import_last_revision_info(
393
self.branch.repository, new_revno, self.rev_id)
406
395
# and now do the commit locally.
407
396
self.branch.set_last_revision_info(new_revno, self.rev_id)
409
# Make the working tree up to date with the branch
398
# Make the working tree be up to date with the branch. This
399
# includes automatic changes scheduled to be made to the tree, such
400
# as updating its basis and unversioning paths that were missing.
401
self.work_tree.unversion(self.deleted_ids)
410
402
self._set_progress_stage("Updating the working tree")
411
403
self.work_tree.update_basis_by_delta(self.rev_id,
404
self.builder.get_basis_delta())
413
405
self.reporter.completed(new_revno, self.rev_id)
414
406
self._process_post_hooks(old_revno, new_revno)
428
420
# A merge with no effect on files
429
421
if len(self.parents) > 1:
431
# TODO: we could simplify this by using self._basis_delta.
423
# TODO: we could simplify this by using self.builder.basis_delta.
433
425
# The initial commit adds a root directory, but this in itself is not
434
426
# a worthwhile commit.
435
427
if (self.basis_revid == revision.NULL_REVISION and
436
len(self.builder.new_inventory) == 1):
428
((self.builder.new_inventory is not None and
429
len(self.builder.new_inventory) == 1) or
430
len(self.builder._basis_delta) == 1)):
437
431
raise PointlessCommit()
438
# If length == 1, then we only have the root entry. Which means
439
# that there is no real difference (only the root could be different)
440
# unless deletes occured, in which case the length is irrelevant.
441
if (self.any_entries_deleted or
442
(len(self.builder.new_inventory) != 1 and
443
self.any_entries_changed)):
432
if self.builder.any_changes():
445
434
raise PointlessCommit()
447
def _check_bound_branch(self):
436
def _check_bound_branch(self, possible_master_transports=None):
448
437
"""Check to see if the local branch is bound.
450
439
If it is bound, then most of the commit will actually be
611
601
if self.master_locked:
612
602
self.master_branch.unlock()
614
def _escape_commit_message(self):
615
"""Replace xml-incompatible control characters."""
616
# FIXME: RBC 20060419 this should be done by the revision
617
# serialiser not by commit. Then we can also add an unescaper
618
# in the deserializer and start roundtripping revision messages
619
# precisely. See repository_implementations/test_repository.py
621
# Python strings can include characters that can't be
622
# represented in well-formed XML; escape characters that
623
# aren't listed in the XML specification
624
# (http://www.w3.org/TR/REC-xml/#NT-Char).
625
self.message, escape_count = re.subn(
626
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
627
lambda match: match.group(0).encode('unicode_escape'),
630
self.reporter.escaped(escape_count, self.message)
632
604
def _gather_parents(self):
633
605
"""Record the parents of a merge for merge detection."""
634
# TODO: Make sure that this list doesn't contain duplicate
606
# TODO: Make sure that this list doesn't contain duplicate
635
607
# entries and the order is preserved when doing this.
636
self.parents = self.work_tree.get_parent_ids()
608
if self.use_record_iter_changes:
610
self.basis_inv = self.basis_tree.inventory
637
611
self.parent_invs = [self.basis_inv]
638
612
for revision in self.parents[1:]:
639
613
if self.branch.repository.has_revision(revision):
646
620
def _update_builder_with_changes(self):
647
621
"""Update the commit builder with the data about what has changed.
649
# Build the revision inventory.
651
# This starts by creating a new empty inventory. Depending on
652
# which files are selected for commit, and what is present in the
653
# current tree, the new inventory is populated. inventory entries
654
# which are candidates for modification have their revision set to
655
# None; inventory entries that are carried over untouched have their
656
# revision set to their prior value.
658
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
659
# results to create a new inventory at the same time, which results
660
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
663
623
exclude = self.exclude
664
specific_files = self.specific_files or []
624
specific_files = self.specific_files
665
625
mutter("Selecting files for commit with filter %s", specific_files)
667
# Build the new inventory
668
self._populate_from_inventory()
628
if self.use_record_iter_changes:
629
iter_changes = self.work_tree.iter_changes(self.basis_tree,
630
specific_files=specific_files)
631
iter_changes = self._filter_iter_changes(iter_changes)
632
for file_id, path, fs_hash in self.builder.record_iter_changes(
633
self.work_tree, self.basis_revid, iter_changes):
634
self.work_tree._observed_sha1(file_id, path, fs_hash)
636
# Build the new inventory
637
self._populate_from_inventory()
638
self._record_unselected()
639
self._report_and_accumulate_deletes()
641
def _filter_iter_changes(self, iter_changes):
642
"""Process iter_changes.
644
This method reports on the changes in iter_changes to the user, and
645
converts 'missing' entries in the iter_changes iterator to 'deleted'
646
entries. 'missing' entries have their
648
:param iter_changes: An iter_changes to process.
649
:return: A generator of changes.
651
reporter = self.reporter
652
report_changes = reporter.is_verbose()
654
for change in iter_changes:
656
old_path = change[1][0]
657
new_path = change[1][1]
658
versioned = change[3][1]
660
versioned = change[3][1]
661
if kind is None and versioned:
664
reporter.missing(new_path)
665
deleted_ids.append(change[0])
666
# Reset the new path (None) and new versioned flag (False)
667
change = (change[0], (change[1][0], None), change[2],
668
(change[3][0], False)) + change[4:]
669
elif kind == 'tree-reference':
670
if self.recursive == 'down':
671
self._commit_nested_tree(change[0], change[1][1])
672
if change[3][0] or change[3][1]:
676
reporter.deleted(old_path)
677
elif old_path is None:
678
reporter.snapshot_change('added', new_path)
679
elif old_path != new_path:
680
reporter.renamed('renamed', old_path, new_path)
683
self.work_tree.branch.repository._format.rich_root_data):
684
# Don't report on changes to '' in non rich root
686
reporter.snapshot_change('modified', new_path)
687
self._next_progress_entry()
688
# Unversion IDs that were found to be deleted
689
self.deleted_ids = deleted_ids
691
def _record_unselected(self):
670
692
# If specific files are selected, then all un-selected files must be
671
693
# recorded in their previous state. For more details, see
672
694
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
673
if specific_files or exclude:
695
if self.specific_files or self.exclude:
696
specific_files = self.specific_files or []
674
697
for path, old_ie in self.basis_inv.iter_entries():
675
698
if old_ie.file_id in self.builder.new_inventory:
676
699
# already added - skip.
678
701
if (is_inside_any(specific_files, path)
679
and not is_inside_any(exclude, path)):
702
and not is_inside_any(self.exclude, path)):
680
703
# was inside the selected path, and not excluded - if not
681
704
# present it has been deleted so skip.
683
706
# From here down it was either not selected, or was excluded:
684
if old_ie.kind == 'directory':
685
self._next_progress_entry()
686
707
# We preserve the entry unaltered.
687
708
ie = old_ie.copy()
688
709
# Note: specific file commits after a merge are currently
690
711
# required after that changes.
691
712
if len(self.parents) > 1:
692
713
ie.revision = None
693
delta, version_recorded = self.builder.record_entry_contents(
694
ie, self.parent_invs, path, self.basis_tree, None)
696
self.any_entries_changed = True
697
if delta: self._basis_delta.append(delta)
714
self.builder.record_entry_contents(ie, self.parent_invs, path,
715
self.basis_tree, None)
699
717
def _report_and_accumulate_deletes(self):
700
# XXX: Could the list of deleted paths and ids be instead taken from
701
# _populate_from_inventory?
702
deleted_ids = set(self.basis_inv._byid.keys()) - \
703
set(self.builder.new_inventory._byid.keys())
718
if (isinstance(self.basis_inv, Inventory)
719
and isinstance(self.builder.new_inventory, Inventory)):
720
# the older Inventory classes provide a _byid dict, and building a
721
# set from the keys of this dict is substantially faster than even
722
# getting a set of ids from the inventory
724
# <lifeless> set(dict) is roughly the same speed as
725
# set(iter(dict)) and both are significantly slower than
727
deleted_ids = set(self.basis_inv._byid.keys()) - \
728
set(self.builder.new_inventory._byid.keys())
730
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
705
732
self.any_entries_deleted = True
706
733
deleted = [(self.basis_tree.id2path(file_id), file_id)
709
736
# XXX: this is not quite directory-order sorting
710
737
for path, file_id in deleted:
711
self._basis_delta.append((path, None, file_id, None))
738
self.builder.record_delete(path, file_id)
712
739
self.reporter.deleted(path)
714
def _populate_from_inventory(self):
715
"""Populate the CommitBuilder by walking the working tree inventory."""
741
def _check_strict(self):
742
# XXX: when we use iter_changes this would likely be faster if
743
# iter_changes would check for us (even in the presence of
717
746
# raise an exception as soon as we find a single unknown.
718
747
for unknown in self.work_tree.unknowns():
719
748
raise StrictCommitFailed()
750
def _populate_from_inventory(self):
751
"""Populate the CommitBuilder by walking the working tree inventory."""
752
# Build the revision inventory.
754
# This starts by creating a new empty inventory. Depending on
755
# which files are selected for commit, and what is present in the
756
# current tree, the new inventory is populated. inventory entries
757
# which are candidates for modification have their revision set to
758
# None; inventory entries that are carried over untouched have their
759
# revision set to their prior value.
761
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
762
# results to create a new inventory at the same time, which results
763
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
721
766
specific_files = self.specific_files
722
767
exclude = self.exclude
723
768
report_changes = self.reporter.is_verbose()
773
817
for segment in path_segments:
774
818
deleted_dict = deleted_dict.setdefault(segment, {})
775
819
self.reporter.missing(path)
820
self._next_progress_entry()
776
821
deleted_ids.append(file_id)
778
823
# TODO: have the builder do the nested commit just-in-time IF and
779
824
# only if needed.
780
if content_summary[0] == 'tree-reference':
825
if kind == 'tree-reference':
781
826
# enforce repository nested tree policy.
782
827
if (not self.work_tree.supports_tree_reference() or
783
828
# repository does not support it either.
784
829
not self.branch.repository._format.supports_tree_reference):
785
content_summary = ('directory',) + content_summary[1:]
786
kind = content_summary[0]
787
# TODO: specific_files filtering before nested tree processing
788
if kind == 'tree-reference':
789
if self.recursive == 'down':
831
content_summary = (kind, None, None, None)
832
elif self.recursive == 'down':
790
833
nested_revision_id = self._commit_nested_tree(
792
content_summary = content_summary[:3] + (
835
content_summary = (kind, None, None, nested_revision_id)
795
content_summary = content_summary[:3] + (
796
self.work_tree.get_reference_revision(file_id),)
837
nested_revision_id = self.work_tree.get_reference_revision(file_id)
838
content_summary = (kind, None, None, nested_revision_id)
798
840
# Record an entry for this item
799
841
# Note: I don't particularly want to have the existing_ie
844
886
ie = existing_ie.copy()
845
887
ie.revision = None
846
delta, version_recorded = self.builder.record_entry_contents(ie,
847
self.parent_invs, path, self.work_tree, content_summary)
849
self._basis_delta.append(delta)
851
self.any_entries_changed = True
888
# For carried over entries we don't care about the fs hash - the repo
889
# isn't generating a sha, so we're not saving computation time.
890
_, _, fs_hash = self.builder.record_entry_contents(
891
ie, self.parent_invs, path, self.work_tree, content_summary)
852
892
if report_changes:
853
893
self._report_change(ie, path)
895
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
856
898
def _report_change(self, ie, path):
866
908
change = ie.describe_change(basis_ie, ie)
867
if change in (InventoryEntry.RENAMED,
909
if change in (InventoryEntry.RENAMED,
868
910
InventoryEntry.MODIFIED_AND_RENAMED):
869
911
old_path = self.basis_inv.id2path(ie.file_id)
870
912
self.reporter.renamed(change, old_path, path)
913
self._next_progress_entry()
915
if change == 'unchanged':
872
917
self.reporter.snapshot_change(change, path)
918
self._next_progress_entry()
874
def _set_progress_stage(self, name, entries_title=None):
920
def _set_progress_stage(self, name, counter=False):
875
921
"""Set the progress stage and emit an update to the progress bar."""
876
922
self.pb_stage_name = name
877
923
self.pb_stage_count += 1
878
self.pb_entries_title = entries_title
879
if entries_title is not None:
880
925
self.pb_entries_count = 0
881
self.pb_entries_total = '?'
927
self.pb_entries_count = None
882
928
self._emit_progress()
884
930
def _next_progress_entry(self):
887
933
self._emit_progress()
889
935
def _emit_progress(self):
890
if self.pb_entries_title:
891
if self.pb_entries_total == '?':
892
text = "%s [%s %d] - Stage" % (self.pb_stage_name,
893
self.pb_entries_title, self.pb_entries_count)
895
text = "%s [%s %d/%s] - Stage" % (self.pb_stage_name,
896
self.pb_entries_title, self.pb_entries_count,
897
str(self.pb_entries_total))
936
if self.pb_entries_count is not None:
937
text = "%s [%d] - Stage" % (self.pb_stage_name,
938
self.pb_entries_count)
899
text = "%s - Stage" % (self.pb_stage_name)
940
text = "%s - Stage" % (self.pb_stage_name, )
900
941
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
943
def _set_specific_file_ids(self):
944
"""populate self.specific_file_ids if we will use it."""
945
if not self.use_record_iter_changes:
946
# If provided, ensure the specified files are versioned
947
if self.specific_files is not None:
948
# Note: This routine is being called because it raises
949
# PathNotVersionedError as a side effect of finding the IDs. We
950
# later use the ids we found as input to the working tree
951
# inventory iterator, so we only consider those ids rather than
952
# examining the whole tree again.
953
# XXX: Dont we have filter_unversioned to do this more
955
self.specific_file_ids = tree.find_ids_across_trees(
956
self.specific_files, [self.basis_tree, self.work_tree])
958
self.specific_file_ids = None
added
removed
bzrlib/commit.py
