283
277
self.committer = committer
284
278
self.strict = strict
285
279
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
281
self.work_tree.lock_write()
282
self.parents = self.work_tree.get_parent_ids()
283
# We can use record_iter_changes IFF iter_changes is compatible with
284
# the command line parameters, and the repository has fast delta
285
# generation. See bug 347649.
286
self.use_record_iter_changes = (
287
not self.specific_files and
289
not self.branch.repository._format.supports_tree_reference and
290
(self.branch.repository._format.fast_deltas or
291
len(self.parents) < 2))
291
292
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
292
293
self.basis_revid = self.work_tree.last_revision()
293
294
self.basis_tree = self.work_tree.basis_tree()
311
312
if self.config is None:
312
313
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])
315
self._set_specific_file_ids()
326
317
# Setup the progress bar. As the number of files that need to be
327
318
# committed in unknown, progress is reported as stages.
392
383
# Upload revision data to the master.
393
384
# this will propagate merged revisions too if needed.
394
385
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
386
self._set_progress_stage("Uploading data to master branch")
401
387
# 'commit' to the master first so a timeout here causes the
402
388
# local branch to be out of date
403
self.master_branch.set_last_revision_info(new_revno,
389
self.master_branch.import_last_revision_info(
390
self.branch.repository, new_revno, self.rev_id)
406
392
# and now do the commit locally.
407
393
self.branch.set_last_revision_info(new_revno, self.rev_id)
428
414
# A merge with no effect on files
429
415
if len(self.parents) > 1:
431
# TODO: we could simplify this by using self._basis_delta.
417
# TODO: we could simplify this by using self.builder.basis_delta.
433
419
# The initial commit adds a root directory, but this in itself is not
434
420
# a worthwhile commit.
435
421
if (self.basis_revid == revision.NULL_REVISION and
436
len(self.builder.new_inventory) == 1):
422
((self.builder.new_inventory is not None and
423
len(self.builder.new_inventory) == 1) or
424
len(self.builder._basis_delta) == 1)):
437
425
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)):
426
if self.builder.any_changes():
445
428
raise PointlessCommit()
447
def _check_bound_branch(self):
430
def _check_bound_branch(self, possible_master_transports=None):
448
431
"""Check to see if the local branch is bound.
450
433
If it is bound, then most of the commit will actually be
611
595
if self.master_locked:
612
596
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
598
def _gather_parents(self):
633
599
"""Record the parents of a merge for merge detection."""
634
# TODO: Make sure that this list doesn't contain duplicate
600
# TODO: Make sure that this list doesn't contain duplicate
635
601
# entries and the order is preserved when doing this.
636
self.parents = self.work_tree.get_parent_ids()
602
if self.use_record_iter_changes:
604
self.basis_inv = self.basis_tree.inventory
637
605
self.parent_invs = [self.basis_inv]
638
606
for revision in self.parents[1:]:
639
607
if self.branch.repository.has_revision(revision):
646
614
def _update_builder_with_changes(self):
647
615
"""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
617
exclude = self.exclude
664
618
specific_files = self.specific_files or []
665
619
mutter("Selecting files for commit with filter %s", specific_files)
667
# Build the new inventory
668
self._populate_from_inventory()
622
if self.use_record_iter_changes:
623
iter_changes = self.work_tree.iter_changes(self.basis_tree)
624
iter_changes = self._filter_iter_changes(iter_changes)
625
for file_id, path, fs_hash in self.builder.record_iter_changes(
626
self.work_tree, self.basis_revid, iter_changes):
627
self.work_tree._observed_sha1(file_id, path, fs_hash)
629
# Build the new inventory
630
self._populate_from_inventory()
631
self._record_unselected()
632
self._report_and_accumulate_deletes()
634
def _filter_iter_changes(self, iter_changes):
635
"""Process iter_changes.
637
This method reports on the changes in iter_changes to the user, and
638
converts 'missing' entries in the iter_changes iterator to 'deleted'
639
entries. 'missing' entries have their
641
:param iter_changes: An iter_changes to process.
642
:return: A generator of changes.
644
reporter = self.reporter
645
report_changes = reporter.is_verbose()
647
for change in iter_changes:
649
old_path = change[1][0]
650
new_path = change[1][1]
651
versioned = change[3][1]
653
versioned = change[3][1]
654
if kind is None and versioned:
657
reporter.missing(new_path)
658
deleted_ids.append(change[0])
659
# Reset the new path (None) and new versioned flag (False)
660
change = (change[0], (change[1][0], None), change[2],
661
(change[3][0], False)) + change[4:]
662
elif kind == 'tree-reference':
663
if self.recursive == 'down':
664
self._commit_nested_tree(change[0], change[1][1])
665
if change[3][0] or change[3][1]:
669
reporter.deleted(old_path)
670
elif old_path is None:
671
reporter.snapshot_change('added', new_path)
672
elif old_path != new_path:
673
reporter.renamed('renamed', old_path, new_path)
676
self.work_tree.branch.repository._format.rich_root_data):
677
# Don't report on changes to '' in non rich root
679
reporter.snapshot_change('modified', new_path)
680
self._next_progress_entry()
681
# Unversion IDs that were found to be deleted
682
self.work_tree.unversion(deleted_ids)
684
def _record_unselected(self):
670
685
# If specific files are selected, then all un-selected files must be
671
686
# recorded in their previous state. For more details, see
672
687
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
673
if specific_files or exclude:
688
if self.specific_files or self.exclude:
689
specific_files = self.specific_files or []
674
690
for path, old_ie in self.basis_inv.iter_entries():
675
691
if old_ie.file_id in self.builder.new_inventory:
676
692
# already added - skip.
678
694
if (is_inside_any(specific_files, path)
679
and not is_inside_any(exclude, path)):
695
and not is_inside_any(self.exclude, path)):
680
696
# was inside the selected path, and not excluded - if not
681
697
# present it has been deleted so skip.
683
699
# From here down it was either not selected, or was excluded:
684
if old_ie.kind == 'directory':
685
self._next_progress_entry()
686
700
# We preserve the entry unaltered.
687
701
ie = old_ie.copy()
688
702
# Note: specific file commits after a merge are currently
690
704
# required after that changes.
691
705
if len(self.parents) > 1:
692
706
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)
707
self.builder.record_entry_contents(ie, self.parent_invs, path,
708
self.basis_tree, None)
699
710
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())
711
if (isinstance(self.basis_inv, Inventory)
712
and isinstance(self.builder.new_inventory, Inventory)):
713
# the older Inventory classes provide a _byid dict, and building a
714
# set from the keys of this dict is substantially faster than even
715
# getting a set of ids from the inventory
717
# <lifeless> set(dict) is roughly the same speed as
718
# set(iter(dict)) and both are significantly slower than
720
deleted_ids = set(self.basis_inv._byid.keys()) - \
721
set(self.builder.new_inventory._byid.keys())
723
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
705
725
self.any_entries_deleted = True
706
726
deleted = [(self.basis_tree.id2path(file_id), file_id)
709
729
# XXX: this is not quite directory-order sorting
710
730
for path, file_id in deleted:
711
self._basis_delta.append((path, None, file_id, None))
731
self.builder.record_delete(path, file_id)
712
732
self.reporter.deleted(path)
714
def _populate_from_inventory(self):
715
"""Populate the CommitBuilder by walking the working tree inventory."""
734
def _check_strict(self):
735
# XXX: when we use iter_changes this would likely be faster if
736
# iter_changes would check for us (even in the presence of
717
739
# raise an exception as soon as we find a single unknown.
718
740
for unknown in self.work_tree.unknowns():
719
741
raise StrictCommitFailed()
743
def _populate_from_inventory(self):
744
"""Populate the CommitBuilder by walking the working tree inventory."""
745
# Build the revision inventory.
747
# This starts by creating a new empty inventory. Depending on
748
# which files are selected for commit, and what is present in the
749
# current tree, the new inventory is populated. inventory entries
750
# which are candidates for modification have their revision set to
751
# None; inventory entries that are carried over untouched have their
752
# revision set to their prior value.
754
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
755
# results to create a new inventory at the same time, which results
756
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
721
759
specific_files = self.specific_files
722
760
exclude = self.exclude
723
761
report_changes = self.reporter.is_verbose()
844
881
ie = existing_ie.copy()
845
882
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
883
# For carried over entries we don't care about the fs hash - the repo
884
# isn't generating a sha, so we're not saving computation time.
885
_, _, fs_hash = self.builder.record_entry_contents(
886
ie, self.parent_invs, path, self.work_tree, content_summary)
852
887
if report_changes:
853
888
self._report_change(ie, path)
890
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
856
893
def _report_change(self, ie, path):
866
903
change = ie.describe_change(basis_ie, ie)
867
if change in (InventoryEntry.RENAMED,
904
if change in (InventoryEntry.RENAMED,
868
905
InventoryEntry.MODIFIED_AND_RENAMED):
869
906
old_path = self.basis_inv.id2path(ie.file_id)
870
907
self.reporter.renamed(change, old_path, path)
908
self._next_progress_entry()
910
if change == 'unchanged':
872
912
self.reporter.snapshot_change(change, path)
913
self._next_progress_entry()
874
def _set_progress_stage(self, name, entries_title=None):
915
def _set_progress_stage(self, name, counter=False):
875
916
"""Set the progress stage and emit an update to the progress bar."""
876
917
self.pb_stage_name = name
877
918
self.pb_stage_count += 1
878
self.pb_entries_title = entries_title
879
if entries_title is not None:
880
920
self.pb_entries_count = 0
881
self.pb_entries_total = '?'
922
self.pb_entries_count = None
882
923
self._emit_progress()
884
925
def _next_progress_entry(self):
887
928
self._emit_progress()
889
930
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))
931
if self.pb_entries_count is not None:
932
text = "%s [%d] - Stage" % (self.pb_stage_name,
933
self.pb_entries_count)
899
text = "%s - Stage" % (self.pb_stage_name)
935
text = "%s - Stage" % (self.pb_stage_name, )
900
936
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
938
def _set_specific_file_ids(self):
939
"""populate self.specific_file_ids if we will use it."""
940
if not self.use_record_iter_changes:
941
# If provided, ensure the specified files are versioned
942
if self.specific_files is not None:
943
# Note: This routine is being called because it raises
944
# PathNotVersionedError as a side effect of finding the IDs. We
945
# later use the ids we found as input to the working tree
946
# inventory iterator, so we only consider those ids rather than
947
# examining the whole tree again.
948
# XXX: Dont we have filter_unversioned to do this more
950
self.specific_file_ids = tree.find_ids_across_trees(
951
self.specific_files, [self.basis_tree, self.work_tree])
953
self.specific_file_ids = None
added
removed
bzrlib/commit.py
