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
19
22
from bzrlib import (
20
branch as _mod_branch,
21
conflicts as _mod_conflicts,
30
27
revision as _mod_revision,
39
from bzrlib.cleanup import OperationWithCleanups
40
from bzrlib.symbol_versioning import (
29
from bzrlib.branch import Branch
30
from bzrlib.conflicts import ConflictList, Conflict
31
from bzrlib.errors import (BzrCommandError,
41
WorkingTreeNotRevision,
44
from bzrlib.merge3 import Merge3
45
from bzrlib.osutils import rename, pathjoin
46
from progress import DummyProgress, ProgressPhase
47
from bzrlib.revision import (is_ancestor, NULL_REVISION, ensure_null)
48
from bzrlib.textfile import check_text_lines
49
from bzrlib.trace import mutter, warning, note
50
from bzrlib.transform import (TreeTransform, resolve_conflicts, cook_conflicts,
51
conflict_pass, FinalPaths, create_by_entry,
52
unique_add, ROOT_PARENT)
53
from bzrlib.versionedfile import PlanWeaveMerge
44
56
# TODO: Report back as changes are merged in
47
59
def transform_tree(from_tree, to_tree, interesting_ids=None):
48
from_tree.lock_tree_write()
49
operation = OperationWithCleanups(merge_inner)
50
operation.add_cleanup(from_tree.unlock)
51
operation.run_simple(from_tree.branch, to_tree, from_tree,
52
ignore_zero=True, interesting_ids=interesting_ids, this_tree=from_tree)
55
class MergeHooks(hooks.Hooks):
58
hooks.Hooks.__init__(self)
59
self.create_hook(hooks.HookPoint('merge_file_content',
60
"Called with a bzrlib.merge.Merger object to create a per file "
61
"merge object when starting a merge. "
62
"Should return either None or a subclass of "
63
"``bzrlib.merge.AbstractPerFileMerger``. "
64
"Such objects will then be called per file "
65
"that needs to be merged (including when one "
66
"side has deleted the file and the other has changed it). "
67
"See the AbstractPerFileMerger API docs for details on how it is "
72
class AbstractPerFileMerger(object):
73
"""PerFileMerger objects are used by plugins extending merge for bzrlib.
75
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
77
:ivar merger: The Merge3Merger performing the merge.
80
def __init__(self, merger):
81
"""Create a PerFileMerger for use with merger."""
84
def merge_contents(self, merge_params):
85
"""Attempt to merge the contents of a single file.
87
:param merge_params: A bzrlib.merge.MergeHookParams
88
:return : A tuple of (status, chunks), where status is one of
89
'not_applicable', 'success', 'conflicted', or 'delete'. If status
90
is 'success' or 'conflicted', then chunks should be an iterable of
91
strings for the new file contents.
93
return ('not applicable', None)
96
class ConfigurableFileMerger(AbstractPerFileMerger):
97
"""Merge individual files when configured via a .conf file.
99
This is a base class for concrete custom file merging logic. Concrete
100
classes should implement ``merge_text``.
102
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
104
:ivar affected_files: The configured file paths to merge.
106
:cvar name_prefix: The prefix to use when looking up configuration
107
details. <name_prefix>_merge_files describes the files targeted by the
110
:cvar default_files: The default file paths to merge when no configuration
117
def __init__(self, merger):
118
super(ConfigurableFileMerger, self).__init__(merger)
119
self.affected_files = None
120
self.default_files = self.__class__.default_files or []
121
self.name_prefix = self.__class__.name_prefix
122
if self.name_prefix is None:
123
raise ValueError("name_prefix must be set.")
125
def filename_matches_config(self, params):
126
"""Check whether the file should call the merge hook.
128
<name_prefix>_merge_files configuration variable is a list of files
129
that should use the hook.
131
affected_files = self.affected_files
132
if affected_files is None:
133
config = self.merger.this_branch.get_config()
134
# Until bzr provides a better policy for caching the config, we
135
# just add the part we're interested in to the params to avoid
136
# reading the config files repeatedly (bazaar.conf, location.conf,
138
config_key = self.name_prefix + '_merge_files'
139
affected_files = config.get_user_option_as_list(config_key)
140
if affected_files is None:
141
# If nothing was specified in the config, use the default.
142
affected_files = self.default_files
143
self.affected_files = affected_files
145
filename = self.merger.this_tree.id2path(params.file_id)
146
if filename in affected_files:
150
def merge_contents(self, params):
151
"""Merge the contents of a single file."""
152
# First, check whether this custom merge logic should be used. We
153
# expect most files should not be merged by this handler.
155
# OTHER is a straight winner, rely on default merge.
156
params.winner == 'other' or
157
# THIS and OTHER aren't both files.
158
not params.is_file_merge() or
159
# The filename isn't listed in the 'NAME_merge_files' config
161
not self.filename_matches_config(params)):
162
return 'not_applicable', None
163
return self.merge_text(params)
165
def merge_text(self, params):
166
"""Merge the byte contents of a single file.
168
This is called after checking that the merge should be performed in
169
merge_contents, and it should behave as per
170
``bzrlib.merge.AbstractPerFileMerger.merge_contents``.
172
raise NotImplementedError(self.merge_text)
175
class MergeHookParams(object):
176
"""Object holding parameters passed to merge_file_content hooks.
178
There are some fields hooks can access:
180
:ivar file_id: the file ID of the file being merged
181
:ivar trans_id: the transform ID for the merge of this file
182
:ivar this_kind: kind of file_id in 'this' tree
183
:ivar other_kind: kind of file_id in 'other' tree
184
:ivar winner: one of 'this', 'other', 'conflict'
187
def __init__(self, merger, file_id, trans_id, this_kind, other_kind,
189
self._merger = merger
190
self.file_id = file_id
191
self.trans_id = trans_id
192
self.this_kind = this_kind
193
self.other_kind = other_kind
196
def is_file_merge(self):
197
"""True if this_kind and other_kind are both 'file'."""
198
return self.this_kind == 'file' and self.other_kind == 'file'
200
@decorators.cachedproperty
201
def base_lines(self):
202
"""The lines of the 'base' version of the file."""
203
return self._merger.get_lines(self._merger.base_tree, self.file_id)
205
@decorators.cachedproperty
206
def this_lines(self):
207
"""The lines of the 'this' version of the file."""
208
return self._merger.get_lines(self._merger.this_tree, self.file_id)
210
@decorators.cachedproperty
211
def other_lines(self):
212
"""The lines of the 'other' version of the file."""
213
return self._merger.get_lines(self._merger.other_tree, self.file_id)
60
merge_inner(from_tree.branch, to_tree, from_tree, ignore_zero=True,
61
interesting_ids=interesting_ids, this_tree=from_tree)
216
64
class Merger(object):
220
65
def __init__(self, this_branch, other_tree=None, base_tree=None,
221
this_tree=None, pb=None, change_reporter=None,
222
recurse='down', revision_graph=None):
66
this_tree=None, pb=DummyProgress(), change_reporter=None,
223
68
object.__init__(self)
69
assert this_tree is not None, "this_tree is required"
224
70
self.this_branch = this_branch
225
71
self.this_basis = _mod_revision.ensure_null(
226
72
this_branch.last_revision())
404
183
if self.other_rev_id is None:
405
184
other_basis_tree = self.revision_tree(self.other_basis)
406
if other_basis_tree.has_changes(self.other_tree):
407
raise errors.WorkingTreeNotRevision(self.this_tree)
185
changes = other_basis_tree.changes_from(self.other_tree)
186
if changes.has_changed():
187
raise WorkingTreeNotRevision(self.this_tree)
408
188
other_rev_id = self.other_basis
409
189
self.other_tree = other_basis_tree
411
@deprecated_method(deprecated_in((2, 1, 0)))
412
191
def file_revisions(self, file_id):
413
192
self.ensure_revision_trees()
414
193
def get_id(tree, file_id):
415
194
revision_id = tree.inventory[file_id].revision
195
assert revision_id is not None
416
196
return revision_id
417
197
if self.this_rev_id is None:
418
198
if self.this_basis_tree.get_file_sha1(file_id) != \
419
199
self.this_tree.get_file_sha1(file_id):
420
raise errors.WorkingTreeNotRevision(self.this_tree)
200
raise WorkingTreeNotRevision(self.this_tree)
422
202
trees = (self.this_basis_tree, self.other_tree)
423
203
return [get_id(tree, file_id) for tree in trees]
425
@deprecated_method(deprecated_in((2, 1, 0)))
426
205
def check_basis(self, check_clean, require_commits=True):
427
206
if self.this_basis is None and require_commits is True:
428
raise errors.BzrCommandError(
429
"This branch has no commits."
430
" (perhaps you would prefer 'bzr pull')")
207
raise BzrCommandError("This branch has no commits."
208
" (perhaps you would prefer 'bzr pull')")
432
210
self.compare_basis()
433
211
if self.this_basis != self.this_rev_id:
434
212
raise errors.UncommittedChanges(self.this_tree)
436
@deprecated_method(deprecated_in((2, 1, 0)))
437
214
def compare_basis(self):
439
216
basis_tree = self.revision_tree(self.this_tree.last_revision())
440
except errors.NoSuchRevision:
217
except errors.RevisionNotPresent:
441
218
basis_tree = self.this_tree.basis_tree()
442
if not self.this_tree.has_changes(basis_tree):
219
changes = self.this_tree.changes_from(basis_tree)
220
if not changes.has_changed():
443
221
self.this_rev_id = self.this_basis
445
223
def set_interesting_files(self, file_list):
446
224
self.interesting_files = file_list
448
226
def set_pending(self):
449
if (not self.base_is_ancestor or not self.base_is_other_ancestor
450
or self.other_rev_id is None):
227
if not self.base_is_ancestor or not self.base_is_other_ancestor or self.other_rev_id is None:
452
229
self._add_parent()
454
231
def _add_parent(self):
455
232
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
456
233
new_parent_trees = []
457
operation = OperationWithCleanups(self.this_tree.set_parent_trees)
458
234
for revision_id in new_parents:
460
236
tree = self.revision_tree(revision_id)
461
except errors.NoSuchRevision:
237
except errors.RevisionNotPresent:
465
operation.add_cleanup(tree.unlock)
466
241
new_parent_trees.append((revision_id, tree))
467
operation.run_simple(new_parent_trees, allow_leftmost_as_ghost=True)
243
self.this_tree.set_parent_trees(new_parent_trees,
244
allow_leftmost_as_ghost=True)
246
for _revision_id, tree in new_parent_trees:
469
250
def set_other(self, other_revision, possible_transports=None):
470
251
"""Set the revision and tree to merge from.
587
340
self.base_rev_id = _mod_revision.ensure_null(
588
341
base_branch.get_rev_id(base_revision[1]))
589
342
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
343
self.base_is_ancestor = is_ancestor(self.this_basis,
346
self.base_is_other_ancestor = is_ancestor(self.other_basis,
591
def make_merger(self):
592
351
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
593
352
'other_tree': self.other_tree,
594
353
'interesting_ids': self.interesting_ids,
595
354
'interesting_files': self.interesting_files,
596
'this_branch': self.this_branch,
598
356
if self.merge_type.requires_base:
599
357
kwargs['base_tree'] = self.base_tree
600
358
if self.merge_type.supports_reprocess:
601
359
kwargs['reprocess'] = self.reprocess
602
360
elif self.reprocess:
603
raise errors.BzrError(
604
"Conflict reduction is not supported for merge"
605
" type %s." % self.merge_type)
361
raise BzrError("Conflict reduction is not supported for merge"
362
" type %s." % self.merge_type)
606
363
if self.merge_type.supports_show_base:
607
364
kwargs['show_base'] = self.show_base
608
365
elif self.show_base:
609
raise errors.BzrError("Showing base is not supported for this"
366
raise BzrError("Showing base is not supported for this"
610
367
" merge type. %s" % self.merge_type)
611
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
612
and not self.base_is_other_ancestor):
613
raise errors.CannotReverseCherrypick()
614
if self.merge_type.supports_cherrypick:
615
kwargs['cherrypick'] = (not self.base_is_ancestor or
616
not self.base_is_other_ancestor)
617
if self._is_criss_cross and getattr(self.merge_type,
618
'supports_lca_trees', False):
619
kwargs['lca_trees'] = self._lca_trees
620
return self.merge_type(pb=None,
621
change_reporter=self.change_reporter,
624
def _do_merge_to(self):
625
merge = self.make_merger()
626
if self.other_branch is not None:
627
self.other_branch.update_references(self.this_branch)
629
if self.recurse == 'down':
630
for relpath, file_id in self.this_tree.iter_references():
631
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
632
other_revision = self.other_tree.get_reference_revision(
634
if other_revision == sub_tree.last_revision():
636
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
637
sub_merge.merge_type = self.merge_type
638
other_branch = self.other_branch.reference_parent(file_id, relpath)
639
sub_merge.set_other_revision(other_revision, other_branch)
640
base_revision = self.base_tree.get_reference_revision(file_id)
641
sub_merge.base_tree = \
642
sub_tree.branch.repository.revision_tree(base_revision)
643
sub_merge.base_rev_id = base_revision
648
operation = OperationWithCleanups(self._do_merge_to)
649
368
self.this_tree.lock_tree_write()
650
operation.add_cleanup(self.this_tree.unlock)
651
369
if self.base_tree is not None:
652
370
self.base_tree.lock_read()
653
operation.add_cleanup(self.base_tree.unlock)
654
371
if self.other_tree is not None:
655
372
self.other_tree.lock_read()
656
operation.add_cleanup(self.other_tree.unlock)
657
merge = operation.run_simple()
374
merge = self.merge_type(pb=self._pb,
375
change_reporter=self.change_reporter,
377
if self.recurse == 'down':
378
for path, file_id in self.this_tree.iter_references():
379
sub_tree = self.this_tree.get_nested_tree(file_id, path)
380
other_revision = self.other_tree.get_reference_revision(
382
if other_revision == sub_tree.last_revision():
384
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
385
sub_merge.merge_type = self.merge_type
386
relpath = self.this_tree.relpath(path)
387
other_branch = self.other_branch.reference_parent(file_id, relpath)
388
sub_merge.set_other_revision(other_revision, other_branch)
389
base_revision = self.base_tree.get_reference_revision(file_id)
390
sub_merge.base_tree = \
391
sub_tree.branch.repository.revision_tree(base_revision)
395
if self.other_tree is not None:
396
self.other_tree.unlock()
397
if self.base_tree is not None:
398
self.base_tree.unlock()
399
self.this_tree.unlock()
658
400
if len(merge.cooked_conflicts) == 0:
659
if not self.ignore_zero and not trace.is_quiet():
660
trace.note("All changes applied successfully.")
401
if not self.ignore_zero:
402
note("All changes applied successfully.")
662
trace.note("%d conflicts encountered."
663
% len(merge.cooked_conflicts))
404
note("%d conflicts encountered." % len(merge.cooked_conflicts))
665
406
return len(merge.cooked_conflicts)
668
class _InventoryNoneEntry(object):
669
"""This represents an inventory entry which *isn't there*.
671
It simplifies the merging logic if we always have an InventoryEntry, even
672
if it isn't actually present
679
symlink_target = None
682
_none_entry = _InventoryNoneEntry()
685
409
class Merge3Merger(object):
686
410
"""Three-way merger that uses the merge3 text merger"""
687
411
requires_base = True
688
412
supports_reprocess = True
689
413
supports_show_base = True
690
414
history_based = False
691
supports_cherrypick = True
692
supports_reverse_cherrypick = True
693
415
winner_idx = {"this": 2, "other": 1, "conflict": 1}
694
supports_lca_trees = True
696
def __init__(self, working_tree, this_tree, base_tree, other_tree,
417
def __init__(self, working_tree, this_tree, base_tree, other_tree,
697
418
interesting_ids=None, reprocess=False, show_base=False,
698
pb=None, pp=None, change_reporter=None,
699
interesting_files=None, do_merge=True,
700
cherrypick=False, lca_trees=None, this_branch=None):
419
pb=DummyProgress(), pp=None, change_reporter=None,
420
interesting_files=None):
701
421
"""Initialize the merger object and perform the merge.
703
423
:param working_tree: The working tree to apply the merge to
704
424
:param this_tree: The local tree in the merge operation
705
425
:param base_tree: The common tree in the merge operation
706
:param other_tree: The other tree to merge changes from
707
:param this_branch: The branch associated with this_tree
426
:param other_tree: The other other tree to merge changes from
708
427
:param interesting_ids: The file_ids of files that should be
709
428
participate in the merge. May not be combined with
710
429
interesting_files.
711
430
:param: reprocess If True, perform conflict-reduction processing.
712
431
:param show_base: If True, show the base revision in text conflicts.
713
432
(incompatible with reprocess)
433
:param pb: A Progress bar
715
434
:param pp: A ProgressPhase object
716
435
:param change_reporter: An object that should report changes made
717
436
:param interesting_files: The tree-relative paths of files that should
720
439
be combined with interesting_ids. If neither interesting_files nor
721
440
interesting_ids is specified, all files may participate in the
723
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
724
if the ancestry was found to include a criss-cross merge.
725
Otherwise should be None.
727
443
object.__init__(self)
728
if interesting_files is not None and interesting_ids is not None:
730
'specify either interesting_ids or interesting_files')
444
if interesting_files is not None:
445
assert interesting_ids is None
731
446
self.interesting_ids = interesting_ids
732
447
self.interesting_files = interesting_files
733
448
self.this_tree = working_tree
449
self.this_tree.lock_tree_write()
734
450
self.base_tree = base_tree
451
self.base_tree.lock_read()
735
452
self.other_tree = other_tree
736
self.this_branch = this_branch
453
self.other_tree.lock_read()
737
454
self._raw_conflicts = []
738
455
self.cooked_conflicts = []
739
456
self.reprocess = reprocess
740
457
self.show_base = show_base
741
self._lca_trees = lca_trees
742
# Uncommenting this will change the default algorithm to always use
743
# _entries_lca. This can be useful for running the test suite and
744
# making sure we haven't missed any corner cases.
745
# if lca_trees is None:
746
# self._lca_trees = [self.base_tree]
747
460
self.change_reporter = change_reporter
748
self.cherrypick = cherrypick
752
warnings.warn("pp argument to Merge3Merger is deprecated")
754
warnings.warn("pb argument to Merge3Merger is deprecated")
757
operation = OperationWithCleanups(self._do_merge)
758
self.this_tree.lock_tree_write()
759
operation.add_cleanup(self.this_tree.unlock)
760
self.base_tree.lock_read()
761
operation.add_cleanup(self.base_tree.unlock)
762
self.other_tree.lock_read()
763
operation.add_cleanup(self.other_tree.unlock)
766
def _do_merge(self, operation):
767
self.tt = transform.TreeTransform(self.this_tree, None)
768
operation.add_cleanup(self.tt.finalize)
769
self._compute_transform()
770
results = self.tt.apply(no_conflicts=True)
771
self.write_modified(results)
462
self.pp = ProgressPhase("Merge phase", 3, self.pb)
464
self.tt = TreeTransform(working_tree, self.pb)
773
self.this_tree.add_conflicts(self.cooked_conflicts)
774
except errors.UnsupportedOperation:
777
def make_preview_transform(self):
778
operation = OperationWithCleanups(self._make_preview_transform)
779
self.base_tree.lock_read()
780
operation.add_cleanup(self.base_tree.unlock)
781
self.other_tree.lock_read()
782
operation.add_cleanup(self.other_tree.unlock)
783
return operation.run_simple()
785
def _make_preview_transform(self):
786
self.tt = transform.TransformPreview(self.this_tree)
787
self._compute_transform()
790
def _compute_transform(self):
791
if self._lca_trees is None:
792
467
entries = self._entries3()
793
resolver = self._three_way
795
entries = self._entries_lca()
796
resolver = self._lca_multi_way
797
child_pb = ui.ui_factory.nested_progress_bar()
799
factories = Merger.hooks['merge_file_content']
800
hooks = [factory(self) for factory in factories] + [self]
801
self.active_hooks = [hook for hook in hooks if hook is not None]
802
for num, (file_id, changed, parents3, names3,
803
executable3) in enumerate(entries):
804
child_pb.update('Preparing file merge', num, len(entries))
805
self._merge_names(file_id, parents3, names3, resolver=resolver)
807
file_status = self._do_merge_contents(file_id)
809
file_status = 'unmodified'
810
self._merge_executable(file_id,
811
executable3, file_status, resolver=resolver)
815
child_pb = ui.ui_factory.nested_progress_bar()
817
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
818
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
821
if self.change_reporter is not None:
822
from bzrlib import delta
823
delta.report_changes(
824
self.tt.iter_changes(), self.change_reporter)
825
self.cook_conflicts(fs_conflicts)
826
for conflict in self.cooked_conflicts:
827
trace.warning(conflict)
468
child_pb = ui.ui_factory.nested_progress_bar()
470
for num, (file_id, changed, parents3, names3,
471
executable3) in enumerate(entries):
472
child_pb.update('Preparing file merge', num, len(entries))
473
self._merge_names(file_id, parents3, names3)
475
file_status = self.merge_contents(file_id)
477
file_status = 'unmodified'
478
self._merge_executable(file_id,
479
executable3, file_status)
484
child_pb = ui.ui_factory.nested_progress_bar()
486
fs_conflicts = resolve_conflicts(self.tt, child_pb,
487
lambda t, c: conflict_pass(t, c, self.other_tree))
490
if change_reporter is not None:
491
from bzrlib import delta
492
delta.report_changes(self.tt._iter_changes(), change_reporter)
493
self.cook_conflicts(fs_conflicts)
494
for conflict in self.cooked_conflicts:
497
results = self.tt.apply(no_conflicts=True)
498
self.write_modified(results)
500
working_tree.add_conflicts(self.cooked_conflicts)
501
except UnsupportedOperation:
505
self.other_tree.unlock()
506
self.base_tree.unlock()
507
self.this_tree.unlock()
829
510
def _entries3(self):
830
511
"""Gather data about files modified between three trees.
862
540
result.append((file_id, changed, parents3, names3, executable3))
865
def _entries_lca(self):
866
"""Gather data about files modified between multiple trees.
868
This compares OTHER versus all LCA trees, and for interesting entries,
869
it then compares with THIS and BASE.
871
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
872
:return: [(file_id, changed, parents, names, executable)]
873
file_id Simple file_id of the entry
874
changed Boolean, True if the kind or contents changed
876
parents ((base, [parent_id, in, lcas]), parent_id_other,
878
names ((base, [name, in, lcas]), name_in_other, name_in_this)
879
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
881
if self.interesting_files is not None:
882
lookup_trees = [self.this_tree, self.base_tree]
883
lookup_trees.extend(self._lca_trees)
884
# I think we should include the lca trees as well
885
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
888
interesting_ids = self.interesting_ids
890
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
892
base_inventory = self.base_tree.inventory
893
this_inventory = self.this_tree.inventory
894
for path, file_id, other_ie, lca_values in walker.iter_all():
895
# Is this modified at all from any of the other trees?
897
other_ie = _none_entry
898
if interesting_ids is not None and file_id not in interesting_ids:
901
# If other_revision is found in any of the lcas, that means this
902
# node is uninteresting. This is because when merging, if there are
903
# multiple heads(), we have to create a new node. So if we didn't,
904
# we know that the ancestry is linear, and that OTHER did not
906
# See doc/developers/lca_merge_resolution.txt for details
907
other_revision = other_ie.revision
908
if other_revision is not None:
909
# We can't use this shortcut when other_revision is None,
910
# because it may be None because things are WorkingTrees, and
911
# not because it is *actually* None.
912
is_unmodified = False
913
for lca_path, ie in lca_values:
914
if ie is not None and ie.revision == other_revision:
921
for lca_path, lca_ie in lca_values:
923
lca_entries.append(_none_entry)
925
lca_entries.append(lca_ie)
927
if file_id in base_inventory:
928
base_ie = base_inventory[file_id]
930
base_ie = _none_entry
932
if file_id in this_inventory:
933
this_ie = this_inventory[file_id]
935
this_ie = _none_entry
941
for lca_ie in lca_entries:
942
lca_kinds.append(lca_ie.kind)
943
lca_parent_ids.append(lca_ie.parent_id)
944
lca_names.append(lca_ie.name)
945
lca_executable.append(lca_ie.executable)
947
kind_winner = self._lca_multi_way(
948
(base_ie.kind, lca_kinds),
949
other_ie.kind, this_ie.kind)
950
parent_id_winner = self._lca_multi_way(
951
(base_ie.parent_id, lca_parent_ids),
952
other_ie.parent_id, this_ie.parent_id)
953
name_winner = self._lca_multi_way(
954
(base_ie.name, lca_names),
955
other_ie.name, this_ie.name)
957
content_changed = True
958
if kind_winner == 'this':
959
# No kind change in OTHER, see if there are *any* changes
960
if other_ie.kind == 'directory':
961
if parent_id_winner == 'this' and name_winner == 'this':
962
# No change for this directory in OTHER, skip
964
content_changed = False
965
elif other_ie.kind is None or other_ie.kind == 'file':
966
def get_sha1(ie, tree):
967
if ie.kind != 'file':
969
return tree.get_file_sha1(file_id)
970
base_sha1 = get_sha1(base_ie, self.base_tree)
971
lca_sha1s = [get_sha1(ie, tree) for ie, tree
972
in zip(lca_entries, self._lca_trees)]
973
this_sha1 = get_sha1(this_ie, self.this_tree)
974
other_sha1 = get_sha1(other_ie, self.other_tree)
975
sha1_winner = self._lca_multi_way(
976
(base_sha1, lca_sha1s), other_sha1, this_sha1,
977
allow_overriding_lca=False)
978
exec_winner = self._lca_multi_way(
979
(base_ie.executable, lca_executable),
980
other_ie.executable, this_ie.executable)
981
if (parent_id_winner == 'this' and name_winner == 'this'
982
and sha1_winner == 'this' and exec_winner == 'this'):
983
# No kind, parent, name, exec, or content change for
984
# OTHER, so this node is not considered interesting
986
if sha1_winner == 'this':
987
content_changed = False
988
elif other_ie.kind == 'symlink':
989
def get_target(ie, tree):
990
if ie.kind != 'symlink':
992
return tree.get_symlink_target(file_id)
993
base_target = get_target(base_ie, self.base_tree)
994
lca_targets = [get_target(ie, tree) for ie, tree
995
in zip(lca_entries, self._lca_trees)]
996
this_target = get_target(this_ie, self.this_tree)
997
other_target = get_target(other_ie, self.other_tree)
998
target_winner = self._lca_multi_way(
999
(base_target, lca_targets),
1000
other_target, this_target)
1001
if (parent_id_winner == 'this' and name_winner == 'this'
1002
and target_winner == 'this'):
1003
# No kind, parent, name, or symlink target change
1006
if target_winner == 'this':
1007
content_changed = False
1008
elif other_ie.kind == 'tree-reference':
1009
# The 'changed' information seems to be handled at a higher
1010
# level. At least, _entries3 returns False for content
1011
# changed, even when at a new revision_id.
1012
content_changed = False
1013
if (parent_id_winner == 'this' and name_winner == 'this'):
1014
# Nothing interesting
1017
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1018
# XXX: We need to handle kind == 'symlink'
1020
# If we have gotten this far, that means something has changed
1021
result.append((file_id, content_changed,
1022
((base_ie.parent_id, lca_parent_ids),
1023
other_ie.parent_id, this_ie.parent_id),
1024
((base_ie.name, lca_names),
1025
other_ie.name, this_ie.name),
1026
((base_ie.executable, lca_executable),
1027
other_ie.executable, this_ie.executable)
1032
543
def fix_root(self):
1034
545
self.tt.final_kind(self.tt.root)
1035
except errors.NoSuchFile:
1036
547
self.tt.cancel_deletion(self.tt.root)
1037
548
if self.tt.final_file_id(self.tt.root) is None:
1038
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
549
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
551
if self.other_tree.inventory.root is None:
1040
553
other_root_file_id = self.other_tree.get_root_id()
1041
if other_root_file_id is None:
1043
554
other_root = self.tt.trans_id_file_id(other_root_file_id)
1044
555
if other_root == self.tt.root:
1047
558
self.tt.final_kind(other_root)
1048
except errors.NoSuchFile:
1050
if self.this_tree.has_id(self.other_tree.inventory.root.file_id):
1051
# the other tree's root is a non-root in the current tree
1053
561
self.reparent_children(self.other_tree.inventory.root, self.tt.root)
1054
562
self.tt.cancel_creation(other_root)
1264
721
return kind, contents
723
def contents_conflict():
724
trans_id = self.tt.trans_id_file_id(file_id)
725
name = self.tt.final_name(trans_id)
726
parent_id = self.tt.final_parent(trans_id)
727
if file_id in self.this_tree.inventory:
728
self.tt.unversion_file(trans_id)
729
if file_id in self.this_tree:
730
self.tt.delete_contents(trans_id)
731
file_group = self._dump_conflicts(name, parent_id, file_id,
733
self._raw_conflicts.append(('contents conflict', file_group))
1266
735
# See SPOT run. run, SPOT, run.
1267
736
# So we're not QUITE repeating ourselves; we do tricky things with
1269
738
base_pair = contents_pair(self.base_tree)
1270
739
other_pair = contents_pair(self.other_tree)
1272
this_pair = contents_pair(self.this_tree)
1273
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1274
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1275
this_pair, allow_overriding_lca=False)
1277
if base_pair == other_pair:
1280
# We delayed evaluating this_pair as long as we can to avoid
1281
# unnecessary sha1 calculation
1282
this_pair = contents_pair(self.this_tree)
1283
winner = self._three_way(base_pair, other_pair, this_pair)
1284
if winner == 'this':
1285
# No interesting changes introduced by OTHER
1287
# We have a hypothetical conflict, but if we have files, then we
1288
# can try to merge the content
1289
trans_id = self.tt.trans_id_file_id(file_id)
1290
params = MergeHookParams(self, file_id, trans_id, this_pair[0],
1291
other_pair[0], winner)
1292
hooks = self.active_hooks
1293
hook_status = 'not_applicable'
1295
hook_status, lines = hook.merge_contents(params)
1296
if hook_status != 'not_applicable':
1297
# Don't try any more hooks, this one applies.
1300
if hook_status == 'not_applicable':
1301
# This is a contents conflict, because none of the available
1302
# functions could merge it.
1304
name = self.tt.final_name(trans_id)
1305
parent_id = self.tt.final_parent(trans_id)
1306
if self.this_tree.has_id(file_id):
1307
self.tt.unversion_file(trans_id)
1308
file_group = self._dump_conflicts(name, parent_id, file_id,
1310
self._raw_conflicts.append(('contents conflict', file_group))
1311
elif hook_status == 'success':
1312
self.tt.create_file(lines, trans_id)
1313
elif hook_status == 'conflicted':
1314
# XXX: perhaps the hook should be able to provide
1315
# the BASE/THIS/OTHER files?
1316
self.tt.create_file(lines, trans_id)
1317
self._raw_conflicts.append(('text conflict', trans_id))
1318
name = self.tt.final_name(trans_id)
1319
parent_id = self.tt.final_parent(trans_id)
1320
self._dump_conflicts(name, parent_id, file_id)
1321
elif hook_status == 'delete':
1322
self.tt.unversion_file(trans_id)
1324
elif hook_status == 'done':
1325
# The hook function did whatever it needs to do directly, no
1326
# further action needed here.
1329
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1330
if not self.this_tree.has_id(file_id) and result == "modified":
1331
self.tt.version_file(file_id, trans_id)
1332
# The merge has been performed, so the old contents should not be
1335
self.tt.delete_contents(trans_id)
1336
except errors.NoSuchFile:
1340
def _default_other_winner_merge(self, merge_hook_params):
1341
"""Replace this contents with other."""
1342
file_id = merge_hook_params.file_id
1343
trans_id = merge_hook_params.trans_id
1344
file_in_this = self.this_tree.has_id(file_id)
1345
if self.other_tree.has_id(file_id):
1346
# OTHER changed the file
1348
if wt.supports_content_filtering():
1349
# We get the path from the working tree if it exists.
1350
# That fails though when OTHER is adding a file, so
1351
# we fall back to the other tree to find the path if
1352
# it doesn't exist locally.
1354
filter_tree_path = wt.id2path(file_id)
1355
except errors.NoSuchId:
1356
filter_tree_path = self.other_tree.id2path(file_id)
1358
# Skip the id2path lookup for older formats
1359
filter_tree_path = None
1360
transform.create_from_tree(self.tt, trans_id,
1361
self.other_tree, file_id,
1362
filter_tree_path=filter_tree_path)
1365
# OTHER deleted the file
1366
return 'delete', None
1368
raise AssertionError(
1369
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1372
def merge_contents(self, merge_hook_params):
1373
"""Fallback merge logic after user installed hooks."""
1374
# This function is used in merge hooks as the fallback instance.
1375
# Perhaps making this function and the functions it calls be a
1376
# a separate class would be better.
1377
if merge_hook_params.winner == 'other':
1378
# OTHER is a straight winner, so replace this contents with other
1379
return self._default_other_winner_merge(merge_hook_params)
1380
elif merge_hook_params.is_file_merge():
1381
# THIS and OTHER are both files, so text merge. Either
1382
# BASE is a file, or both converted to files, so at least we
1383
# have agreement that output should be a file.
1385
self.text_merge(merge_hook_params.file_id,
1386
merge_hook_params.trans_id)
1387
except errors.BinaryFile:
1388
return 'not_applicable', None
1391
return 'not_applicable', None
740
if base_pair == other_pair:
741
# OTHER introduced no changes
743
this_pair = contents_pair(self.this_tree)
744
if this_pair == other_pair:
745
# THIS and OTHER introduced the same changes
748
trans_id = self.tt.trans_id_file_id(file_id)
749
if this_pair == base_pair:
750
# only OTHER introduced changes
751
if file_id in self.this_tree:
752
# Remove any existing contents
753
self.tt.delete_contents(trans_id)
754
if file_id in self.other_tree:
755
# OTHER changed the file
756
create_by_entry(self.tt,
757
self.other_tree.inventory[file_id],
758
self.other_tree, trans_id)
759
if file_id not in self.this_tree.inventory:
760
self.tt.version_file(file_id, trans_id)
762
elif file_id in self.this_tree.inventory:
763
# OTHER deleted the file
764
self.tt.unversion_file(trans_id)
766
#BOTH THIS and OTHER introduced changes; scalar conflict
767
elif this_pair[0] == "file" and other_pair[0] == "file":
768
# THIS and OTHER are both files, so text merge. Either
769
# BASE is a file, or both converted to files, so at least we
770
# have agreement that output should be a file.
772
self.text_merge(file_id, trans_id)
774
return contents_conflict()
775
if file_id not in self.this_tree.inventory:
776
self.tt.version_file(file_id, trans_id)
778
self.tt.tree_kind(trans_id)
779
self.tt.delete_contents(trans_id)
784
# Scalar conflict, can't text merge. Dump conflicts
785
return contents_conflict()
1393
787
def get_lines(self, tree, file_id):
1394
788
"""Return the lines in a file, or an empty list."""
1395
if tree.has_id(file_id):
1396
790
return tree.get_file(file_id).readlines()
1449
842
determined automatically. If set_version is true, the .OTHER, .THIS
1450
843
or .BASE (in that order) will be created as versioned files.
1452
data = [('OTHER', self.other_tree, other_lines),
845
data = [('OTHER', self.other_tree, other_lines),
1453
846
('THIS', self.this_tree, this_lines)]
1455
848
data.append(('BASE', self.base_tree, base_lines))
1457
# We need to use the actual path in the working tree of the file here,
1458
# ignoring the conflict suffixes
1460
if wt.supports_content_filtering():
1462
filter_tree_path = wt.id2path(file_id)
1463
except errors.NoSuchId:
1464
# file has been deleted
1465
filter_tree_path = None
1467
# Skip the id2path lookup for older formats
1468
filter_tree_path = None
1470
849
versioned = False
1472
851
for suffix, tree, lines in data:
1473
if tree.has_id(file_id):
1474
853
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1475
suffix, lines, filter_tree_path)
1476
855
file_group.append(trans_id)
1477
856
if set_version and not versioned:
1478
857
self.tt.version_file(file_id, trans_id)
1479
858
versioned = True
1480
859
return file_group
1482
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1483
lines=None, filter_tree_path=None):
861
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1484
863
"""Emit a single conflict file."""
1485
864
name = name + '.' + suffix
1486
865
trans_id = self.tt.create_path(name, parent_id)
1487
transform.create_from_tree(self.tt, trans_id, tree, file_id, lines,
866
entry = tree.inventory[file_id]
867
create_by_entry(self.tt, entry, tree, trans_id, lines)
1491
870
def merge_executable(self, file_id, file_status):
1492
871
"""Perform a merge on the execute bit."""
1493
872
executable = [self.executable(t, file_id) for t in (self.base_tree,
1494
873
self.other_tree, self.this_tree)]
1495
self._merge_executable(file_id, executable, file_status,
1496
resolver=self._three_way)
874
self._merge_executable(file_id, executable, file_status)
1498
def _merge_executable(self, file_id, executable, file_status,
876
def _merge_executable(self, file_id, executable, file_status):
1500
877
"""Perform a merge on the execute bit."""
1501
878
base_executable, other_executable, this_executable = executable
1502
879
if file_status == "deleted":
1504
winner = resolver(*executable)
881
trans_id = self.tt.trans_id_file_id(file_id)
883
if self.tt.final_kind(trans_id) != "file":
887
winner = self._three_way(*executable)
1505
888
if winner == "conflict":
1506
889
# There must be a None in here, if we have a conflict, but we
1507
890
# need executability since file status was not deleted.
1555
935
if path.endswith(suffix):
1556
936
path = path[:-len(suffix)]
1558
c = _mod_conflicts.Conflict.factory(conflict_type,
1559
path=path, file_id=file_id)
938
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
1560
939
self.cooked_conflicts.append(c)
1561
940
if conflict_type == 'text conflict':
1562
941
trans_id = conflict[1]
1563
942
path = fp.get_path(trans_id)
1564
943
file_id = self.tt.final_file_id(trans_id)
1565
c = _mod_conflicts.Conflict.factory(conflict_type,
1566
path=path, file_id=file_id)
944
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
1567
945
self.cooked_conflicts.append(c)
1569
947
for trans_id, conflicts in name_conflicts.iteritems():
1571
949
this_parent, other_parent = conflicts['parent conflict']
1572
if this_parent == other_parent:
1573
raise AssertionError()
950
assert this_parent != other_parent
1574
951
except KeyError:
1575
952
this_parent = other_parent = \
1576
953
self.tt.final_file_id(self.tt.final_parent(trans_id))
1578
955
this_name, other_name = conflicts['name conflict']
1579
if this_name == other_name:
1580
raise AssertionError()
956
assert this_name != other_name
1581
957
except KeyError:
1582
958
this_name = other_name = self.tt.final_name(trans_id)
1583
959
other_path = fp.get_path(trans_id)
1584
960
if this_parent is not None and this_name is not None:
1585
961
this_parent_path = \
1586
962
fp.get_path(self.tt.trans_id_file_id(this_parent))
1587
this_path = osutils.pathjoin(this_parent_path, this_name)
963
this_path = pathjoin(this_parent_path, this_name)
1589
965
this_path = "<deleted>"
1590
966
file_id = self.tt.final_file_id(trans_id)
1591
c = _mod_conflicts.Conflict.factory('path conflict', path=this_path,
1592
conflict_path=other_path,
967
c = Conflict.factory('path conflict', path=this_path,
968
conflict_path=other_path, file_id=file_id)
1594
969
self.cooked_conflicts.append(c)
1595
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
970
self.cooked_conflicts.sort(key=Conflict.sort_key)
1598
973
class WeaveMerger(Merge3Merger):
1599
974
"""Three-way tree merger, text weave merger."""
1600
975
supports_reprocess = True
1601
976
supports_show_base = False
1602
supports_reverse_cherrypick = False
1603
history_based = True
1605
def _generate_merge_plan(self, file_id, base):
1606
return self.this_tree.plan_file_merge(file_id, self.other_tree,
978
def __init__(self, working_tree, this_tree, base_tree, other_tree,
979
interesting_ids=None, pb=DummyProgress(), pp=None,
980
reprocess=False, change_reporter=None,
981
interesting_files=None):
982
super(WeaveMerger, self).__init__(working_tree, this_tree,
983
base_tree, other_tree,
984
interesting_ids=interesting_ids,
985
pb=pb, pp=pp, reprocess=reprocess,
986
change_reporter=change_reporter)
1609
988
def _merged_lines(self, file_id):
1610
989
"""Generate the merged lines.
1611
990
There is no distinction between lines that are meant to contain <<<<<<<
1615
base = self.base_tree
1618
plan = self._generate_merge_plan(file_id, base)
1619
if 'merge' in debug.debug_flags:
1621
trans_id = self.tt.trans_id_file_id(file_id)
1622
name = self.tt.final_name(trans_id) + '.plan'
1623
contents = ('%11s|%s' % l for l in plan)
1624
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1625
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1626
'>>>>>>> MERGE-SOURCE\n')
1627
lines, conflicts = textmerge.merge_lines(self.reprocess)
1629
base_lines = textmerge.base_from_plan()
1632
return lines, base_lines
993
plan = self.this_tree.plan_file_merge(file_id, self.other_tree)
994
textmerge = PlanWeaveMerge(plan, '<<<<<<< TREE\n',
995
'>>>>>>> MERGE-SOURCE\n')
996
return textmerge.merge_lines(self.reprocess)
1634
998
def text_merge(self, file_id, trans_id):
1635
999
"""Perform a (weave) text merge for a given file and file-id.
1636
1000
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1637
1001
and a conflict will be noted.
1639
lines, base_lines = self._merged_lines(file_id)
1003
lines, conflicts = self._merged_lines(file_id)
1640
1004
lines = list(lines)
1641
# Note we're checking whether the OUTPUT is binary in this case,
1005
# Note we're checking whether the OUTPUT is binary in this case,
1642
1006
# because we don't want to get into weave merge guts.
1643
textfile.check_text_lines(lines)
1007
check_text_lines(lines)
1644
1008
self.tt.create_file(lines, trans_id)
1645
if base_lines is not None:
1647
1010
self._raw_conflicts.append(('text conflict', trans_id))
1648
1011
name = self.tt.final_name(trans_id)
1649
1012
parent_id = self.tt.final_parent(trans_id)
1650
file_group = self._dump_conflicts(name, parent_id, file_id,
1652
base_lines=base_lines)
1013
file_group = self._dump_conflicts(name, parent_id, file_id,
1653
1015
file_group.append(trans_id)
1656
class LCAMerger(WeaveMerger):
1658
def _generate_merge_plan(self, file_id, base):
1659
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1662
1018
class Diff3Merger(Merge3Merger):
1663
1019
"""Three-way merger using external diff3 for text merging"""
1665
1021
def dump_file(self, temp_dir, name, tree, file_id):
1666
out_path = osutils.pathjoin(temp_dir, name)
1022
out_path = pathjoin(temp_dir, name)
1667
1023
out_file = open(out_path, "wb")
1669
1025
in_file = tree.get_file(file_id)
1780
1130
yield status_a(revision, text)
1781
1131
for revision, text in annotated_b[b_cur:bi]:
1782
1132
yield status_b(revision, text)
1783
1134
# and now the matched section
1786
for text_a in plain_a[ai:a_cur]:
1137
for text_a, text_b in zip(plain_a[ai:a_cur], plain_b[bi:b_cur]):
1138
assert text_a == text_b
1787
1139
yield "unchanged", text_a
1790
class _PlanMergeBase(object):
1792
def __init__(self, a_rev, b_rev, vf, key_prefix):
1795
:param a_rev: Revision-id of one revision to merge
1796
:param b_rev: Revision-id of the other revision to merge
1797
:param vf: A VersionedFiles containing both revisions
1798
:param key_prefix: A prefix for accessing keys in vf, typically
1804
self._last_lines = None
1805
self._last_lines_revision_id = None
1806
self._cached_matching_blocks = {}
1807
self._key_prefix = key_prefix
1808
self._precache_tip_lines()
1810
def _precache_tip_lines(self):
1811
lines = self.get_lines([self.a_rev, self.b_rev])
1812
self.lines_a = lines[self.a_rev]
1813
self.lines_b = lines[self.b_rev]
1815
def get_lines(self, revisions):
1816
"""Get lines for revisions from the backing VersionedFiles.
1818
:raises RevisionNotPresent: on absent texts.
1820
keys = [(self._key_prefix + (rev,)) for rev in revisions]
1822
for record in self.vf.get_record_stream(keys, 'unordered', True):
1823
if record.storage_kind == 'absent':
1824
raise errors.RevisionNotPresent(record.key, self.vf)
1825
result[record.key[-1]] = osutils.chunks_to_lines(
1826
record.get_bytes_as('chunked'))
1829
def plan_merge(self):
1830
"""Generate a 'plan' for merging the two revisions.
1832
This involves comparing their texts and determining the cause of
1833
differences. If text A has a line and text B does not, then either the
1834
line was added to text A, or it was deleted from B. Once the causes
1835
are combined, they are written out in the format described in
1836
VersionedFile.plan_merge
1838
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
1839
unique_a, unique_b = self._unique_lines(blocks)
1840
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
1841
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
1842
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
1844
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
1847
for i, j, n in blocks:
1848
for a_index in range(last_i, i):
1849
if a_index in new_a:
1850
if a_index in killed_b:
1851
yield 'conflicted-a', self.lines_a[a_index]
1853
yield 'new-a', self.lines_a[a_index]
1855
yield 'killed-b', self.lines_a[a_index]
1856
for b_index in range(last_j, j):
1857
if b_index in new_b:
1858
if b_index in killed_a:
1859
yield 'conflicted-b', self.lines_b[b_index]
1861
yield 'new-b', self.lines_b[b_index]
1863
yield 'killed-a', self.lines_b[b_index]
1864
# handle common lines
1865
for a_index in range(i, i+n):
1866
yield 'unchanged', self.lines_a[a_index]
1870
def _get_matching_blocks(self, left_revision, right_revision):
1871
"""Return a description of which sections of two revisions match.
1873
See SequenceMatcher.get_matching_blocks
1875
cached = self._cached_matching_blocks.get((left_revision,
1877
if cached is not None:
1879
if self._last_lines_revision_id == left_revision:
1880
left_lines = self._last_lines
1881
right_lines = self.get_lines([right_revision])[right_revision]
1883
lines = self.get_lines([left_revision, right_revision])
1884
left_lines = lines[left_revision]
1885
right_lines = lines[right_revision]
1886
self._last_lines = right_lines
1887
self._last_lines_revision_id = right_revision
1888
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
1890
return matcher.get_matching_blocks()
1892
def _unique_lines(self, matching_blocks):
1893
"""Analyse matching_blocks to determine which lines are unique
1895
:return: a tuple of (unique_left, unique_right), where the values are
1896
sets of line numbers of unique lines.
1902
for i, j, n in matching_blocks:
1903
unique_left.extend(range(last_i, i))
1904
unique_right.extend(range(last_j, j))
1907
return unique_left, unique_right
1910
def _subtract_plans(old_plan, new_plan):
1911
"""Remove changes from new_plan that came from old_plan.
1913
It is assumed that the difference between the old_plan and new_plan
1914
is their choice of 'b' text.
1916
All lines from new_plan that differ from old_plan are emitted
1917
verbatim. All lines from new_plan that match old_plan but are
1918
not about the 'b' revision are emitted verbatim.
1920
Lines that match and are about the 'b' revision are the lines we
1921
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
1922
is skipped entirely.
1924
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
1927
for i, j, n in matcher.get_matching_blocks():
1928
for jj in range(last_j, j):
1930
for jj in range(j, j+n):
1931
plan_line = new_plan[jj]
1932
if plan_line[0] == 'new-b':
1934
elif plan_line[0] == 'killed-b':
1935
yield 'unchanged', plan_line[1]
1941
class _PlanMerge(_PlanMergeBase):
1942
"""Plan an annotate merge using on-the-fly annotation"""
1944
def __init__(self, a_rev, b_rev, vf, key_prefix):
1945
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
1946
self.a_key = self._key_prefix + (self.a_rev,)
1947
self.b_key = self._key_prefix + (self.b_rev,)
1948
self.graph = _mod_graph.Graph(self.vf)
1949
heads = self.graph.heads((self.a_key, self.b_key))
1951
# one side dominates, so we can just return its values, yay for
1953
# Ideally we would know that before we get this far
1954
self._head_key = heads.pop()
1955
if self._head_key == self.a_key:
1959
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
1960
self._head_key[-1], other)
1963
self._head_key = None
1966
def _precache_tip_lines(self):
1967
# Turn this into a no-op, because we will do this later
1970
def _find_recursive_lcas(self):
1971
"""Find all the ancestors back to a unique lca"""
1972
cur_ancestors = (self.a_key, self.b_key)
1973
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
1974
# rather than a key tuple. We will just map that directly to no common
1978
next_lcas = self.graph.find_lca(*cur_ancestors)
1979
# Map a plain NULL_REVISION to a simple no-ancestors
1980
if next_lcas == set([_mod_revision.NULL_REVISION]):
1982
# Order the lca's based on when they were merged into the tip
1983
# While the actual merge portion of weave merge uses a set() of
1984
# active revisions, the order of insertion *does* effect the
1985
# implicit ordering of the texts.
1986
for rev_key in cur_ancestors:
1987
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
1989
parent_map[rev_key] = ordered_parents
1990
if len(next_lcas) == 0:
1992
elif len(next_lcas) == 1:
1993
parent_map[list(next_lcas)[0]] = ()
1995
elif len(next_lcas) > 2:
1996
# More than 2 lca's, fall back to grabbing all nodes between
1997
# this and the unique lca.
1998
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2000
self.a_key, self.b_key, cur_ancestors)
2001
cur_lcas = next_lcas
2002
while len(cur_lcas) > 1:
2003
cur_lcas = self.graph.find_lca(*cur_lcas)
2004
if len(cur_lcas) == 0:
2005
# No common base to find, use the full ancestry
2008
unique_lca = list(cur_lcas)[0]
2009
if unique_lca == _mod_revision.NULL_REVISION:
2010
# find_lca will return a plain 'NULL_REVISION' rather
2011
# than a key tuple when there is no common ancestor, we
2012
# prefer to just use None, because it doesn't confuse
2013
# _get_interesting_texts()
2015
parent_map.update(self._find_unique_parents(next_lcas,
2018
cur_ancestors = next_lcas
2021
def _find_unique_parents(self, tip_keys, base_key):
2022
"""Find ancestors of tip that aren't ancestors of base.
2024
:param tip_keys: Nodes that are interesting
2025
:param base_key: Cull all ancestors of this node
2026
:return: The parent map for all revisions between tip_keys and
2027
base_key. base_key will be included. References to nodes outside of
2028
the ancestor set will also be removed.
2030
# TODO: this would be simpler if find_unique_ancestors took a list
2031
# instead of a single tip, internally it supports it, but it
2032
# isn't a "backwards compatible" api change.
2033
if base_key is None:
2034
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2035
# We remove NULL_REVISION because it isn't a proper tuple key, and
2036
# thus confuses things like _get_interesting_texts, and our logic
2037
# to add the texts into the memory weave.
2038
if _mod_revision.NULL_REVISION in parent_map:
2039
parent_map.pop(_mod_revision.NULL_REVISION)
2042
for tip in tip_keys:
2044
self.graph.find_unique_ancestors(tip, [base_key]))
2045
parent_map = self.graph.get_parent_map(interesting)
2046
parent_map[base_key] = ()
2047
culled_parent_map, child_map, tails = self._remove_external_references(
2049
# Remove all the tails but base_key
2050
if base_key is not None:
2051
tails.remove(base_key)
2052
self._prune_tails(culled_parent_map, child_map, tails)
2053
# Now remove all the uninteresting 'linear' regions
2054
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2058
def _remove_external_references(parent_map):
2059
"""Remove references that go outside of the parent map.
2061
:param parent_map: Something returned from Graph.get_parent_map(keys)
2062
:return: (filtered_parent_map, child_map, tails)
2063
filtered_parent_map is parent_map without external references
2064
child_map is the {parent_key: [child_keys]} mapping
2065
tails is a list of nodes that do not have any parents in the map
2067
# TODO: The basic effect of this function seems more generic than
2068
# _PlanMerge. But the specific details of building a child_map,
2069
# and computing tails seems very specific to _PlanMerge.
2070
# Still, should this be in Graph land?
2071
filtered_parent_map = {}
2074
for key, parent_keys in parent_map.iteritems():
2075
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2076
if not culled_parent_keys:
2078
for parent_key in culled_parent_keys:
2079
child_map.setdefault(parent_key, []).append(key)
2080
# TODO: Do we want to do this, it adds overhead for every node,
2081
# just to say that the node has no children
2082
child_map.setdefault(key, [])
2083
filtered_parent_map[key] = culled_parent_keys
2084
return filtered_parent_map, child_map, tails
2087
def _prune_tails(parent_map, child_map, tails_to_remove):
2088
"""Remove tails from the parent map.
2090
This will remove the supplied revisions until no more children have 0
2093
:param parent_map: A dict of {child: [parents]}, this dictionary will
2094
be modified in place.
2095
:param tails_to_remove: A list of tips that should be removed,
2096
this list will be consumed
2097
:param child_map: The reverse dict of parent_map ({parent: [children]})
2098
this dict will be modified
2099
:return: None, parent_map will be modified in place.
2101
while tails_to_remove:
2102
next = tails_to_remove.pop()
2103
parent_map.pop(next)
2104
children = child_map.pop(next)
2105
for child in children:
2106
child_parents = parent_map[child]
2107
child_parents.remove(next)
2108
if len(child_parents) == 0:
2109
tails_to_remove.append(child)
2111
def _get_interesting_texts(self, parent_map):
2112
"""Return a dict of texts we are interested in.
2114
Note that the input is in key tuples, but the output is in plain
2117
:param parent_map: The output from _find_recursive_lcas
2118
:return: A dict of {'revision_id':lines} as returned by
2119
_PlanMergeBase.get_lines()
2121
all_revision_keys = set(parent_map)
2122
all_revision_keys.add(self.a_key)
2123
all_revision_keys.add(self.b_key)
2125
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2126
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2129
def _build_weave(self):
2130
from bzrlib import weave
2131
self._weave = weave.Weave(weave_name='in_memory_weave',
2132
allow_reserved=True)
2133
parent_map = self._find_recursive_lcas()
2135
all_texts = self._get_interesting_texts(parent_map)
2137
# Note: Unfortunately, the order given by topo_sort will effect the
2138
# ordering resolution in the output. Specifically, if you add A then B,
2139
# then in the output text A lines will show up before B lines. And, of
2140
# course, topo_sort doesn't guarantee any real ordering.
2141
# So we use merge_sort, and add a fake node on the tip.
2142
# This ensures that left-hand parents will always be inserted into the
2143
# weave before right-hand parents.
2144
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2145
parent_map[tip_key] = (self.a_key, self.b_key)
2147
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2151
# for key in tsort.topo_sort(parent_map):
2152
parent_keys = parent_map[key]
2153
revision_id = key[-1]
2154
parent_ids = [k[-1] for k in parent_keys]
2155
self._weave.add_lines(revision_id, parent_ids,
2156
all_texts[revision_id])
2158
def plan_merge(self):
2159
"""Generate a 'plan' for merging the two revisions.
2161
This involves comparing their texts and determining the cause of
2162
differences. If text A has a line and text B does not, then either the
2163
line was added to text A, or it was deleted from B. Once the causes
2164
are combined, they are written out in the format described in
2165
VersionedFile.plan_merge
2167
if self._head_key is not None: # There was a single head
2168
if self._head_key == self.a_key:
2171
if self._head_key != self.b_key:
2172
raise AssertionError('There was an invalid head: %s != %s'
2173
% (self.b_key, self._head_key))
2175
head_rev = self._head_key[-1]
2176
lines = self.get_lines([head_rev])[head_rev]
2177
return ((plan, line) for line in lines)
2178
return self._weave.plan_merge(self.a_rev, self.b_rev)
2181
class _PlanLCAMerge(_PlanMergeBase):
2183
This merge algorithm differs from _PlanMerge in that:
2184
1. comparisons are done against LCAs only
2185
2. cases where a contested line is new versus one LCA but old versus
2186
another are marked as conflicts, by emitting the line as conflicted-a
2189
This is faster, and hopefully produces more useful output.
2192
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2193
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2194
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2197
if lca == _mod_revision.NULL_REVISION:
2200
self.lcas.add(lca[-1])
2201
for lca in self.lcas:
2202
if _mod_revision.is_null(lca):
2205
lca_lines = self.get_lines([lca])[lca]
2206
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2208
blocks = list(matcher.get_matching_blocks())
2209
self._cached_matching_blocks[(a_rev, lca)] = blocks
2210
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2212
blocks = list(matcher.get_matching_blocks())
2213
self._cached_matching_blocks[(b_rev, lca)] = blocks
2215
def _determine_status(self, revision_id, unique_line_numbers):
2216
"""Determines the status unique lines versus all lcas.
2218
Basically, determines why the line is unique to this revision.
2220
A line may be determined new, killed, or both.
2222
If a line is determined new, that means it was not present in at least
2223
one LCA, and is not present in the other merge revision.
2225
If a line is determined killed, that means the line was present in
2228
If a line is killed and new, this indicates that the two merge
2229
revisions contain differing conflict resolutions.
2230
:param revision_id: The id of the revision in which the lines are
2232
:param unique_line_numbers: The line numbers of unique lines.
2233
:return a tuple of (new_this, killed_other):
2237
unique_line_numbers = set(unique_line_numbers)
2238
for lca in self.lcas:
2239
blocks = self._get_matching_blocks(revision_id, lca)
2240
unique_vs_lca, _ignored = self._unique_lines(blocks)
2241
new.update(unique_line_numbers.intersection(unique_vs_lca))
2242
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
