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 PerFileMerger(AbstractPerFileMerger):
97
"""Merge individual files when self.file_matches returns True.
99
This class is intended to be subclassed. The file_matches and
100
merge_matching methods should be overridden with concrete implementations.
103
def file_matches(self, params):
104
"""Return True if merge_matching should be called on this file.
106
Only called with merges of plain files with no clear winner.
108
Subclasses must override this.
110
raise NotImplementedError(self.file_matches)
112
def get_filename(self, params, tree):
113
"""Lookup the filename (i.e. basename, not path), given a Tree (e.g.
114
self.merger.this_tree) and a MergeHookParams.
116
return osutils.basename(tree.id2path(params.file_id))
118
def get_filepath(self, params, tree):
119
"""Calculate the path to the file in a tree.
121
:param params: A MergeHookParams describing the file to merge
122
:param tree: a Tree, e.g. self.merger.this_tree.
124
return tree.id2path(params.file_id)
126
def merge_contents(self, params):
127
"""Merge the contents of a single file."""
128
# Check whether this custom merge logic should be used.
130
# OTHER is a straight winner, rely on default merge.
131
params.winner == 'other' or
132
# THIS and OTHER aren't both files.
133
not params.is_file_merge() or
134
# The filename doesn't match *.xml
135
not self.file_matches(params)):
136
return 'not_applicable', None
137
return self.merge_matching(params)
139
def merge_matching(self, params):
140
"""Merge the contents of a single file that has matched the criteria
141
in PerFileMerger.merge_contents (is a conflict, is a file,
142
self.file_matches is True).
144
Subclasses must override this.
146
raise NotImplementedError(self.merge_matching)
149
class ConfigurableFileMerger(PerFileMerger):
150
"""Merge individual files when configured via a .conf file.
152
This is a base class for concrete custom file merging logic. Concrete
153
classes should implement ``merge_text``.
155
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
157
:ivar affected_files: The configured file paths to merge.
159
:cvar name_prefix: The prefix to use when looking up configuration
160
details. <name_prefix>_merge_files describes the files targeted by the
163
:cvar default_files: The default file paths to merge when no configuration
170
def __init__(self, merger):
171
super(ConfigurableFileMerger, self).__init__(merger)
172
self.affected_files = None
173
self.default_files = self.__class__.default_files or []
174
self.name_prefix = self.__class__.name_prefix
175
if self.name_prefix is None:
176
raise ValueError("name_prefix must be set.")
178
def file_matches(self, params):
179
"""Check whether the file should call the merge hook.
181
<name_prefix>_merge_files configuration variable is a list of files
182
that should use the hook.
184
affected_files = self.affected_files
185
if affected_files is None:
186
config = self.merger.this_branch.get_config()
187
# Until bzr provides a better policy for caching the config, we
188
# just add the part we're interested in to the params to avoid
189
# reading the config files repeatedly (bazaar.conf, location.conf,
191
config_key = self.name_prefix + '_merge_files'
192
affected_files = config.get_user_option_as_list(config_key)
193
if affected_files is None:
194
# If nothing was specified in the config, use the default.
195
affected_files = self.default_files
196
self.affected_files = affected_files
198
filepath = self.get_filepath(params, self.merger.this_tree)
199
if filepath in affected_files:
203
def merge_matching(self, params):
204
return self.merge_text(params)
206
def merge_text(self, params):
207
"""Merge the byte contents of a single file.
209
This is called after checking that the merge should be performed in
210
merge_contents, and it should behave as per
211
``bzrlib.merge.AbstractPerFileMerger.merge_contents``.
213
raise NotImplementedError(self.merge_text)
216
class MergeHookParams(object):
217
"""Object holding parameters passed to merge_file_content hooks.
219
There are some fields hooks can access:
221
:ivar file_id: the file ID of the file being merged
222
:ivar trans_id: the transform ID for the merge of this file
223
:ivar this_kind: kind of file_id in 'this' tree
224
:ivar other_kind: kind of file_id in 'other' tree
225
:ivar winner: one of 'this', 'other', 'conflict'
228
def __init__(self, merger, file_id, trans_id, this_kind, other_kind,
230
self._merger = merger
231
self.file_id = file_id
232
self.trans_id = trans_id
233
self.this_kind = this_kind
234
self.other_kind = other_kind
237
def is_file_merge(self):
238
"""True if this_kind and other_kind are both 'file'."""
239
return self.this_kind == 'file' and self.other_kind == 'file'
241
@decorators.cachedproperty
242
def base_lines(self):
243
"""The lines of the 'base' version of the file."""
244
return self._merger.get_lines(self._merger.base_tree, self.file_id)
246
@decorators.cachedproperty
247
def this_lines(self):
248
"""The lines of the 'this' version of the file."""
249
return self._merger.get_lines(self._merger.this_tree, self.file_id)
251
@decorators.cachedproperty
252
def other_lines(self):
253
"""The lines of the 'other' version of the file."""
254
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)
257
64
class Merger(object):
261
65
def __init__(self, this_branch, other_tree=None, base_tree=None,
262
this_tree=None, pb=None, change_reporter=None,
263
recurse='down', revision_graph=None):
66
this_tree=None, pb=DummyProgress(), change_reporter=None,
264
68
object.__init__(self)
69
assert this_tree is not None, "this_tree is required"
265
70
self.this_branch = this_branch
266
71
self.this_basis = _mod_revision.ensure_null(
267
72
this_branch.last_revision())
445
183
if self.other_rev_id is None:
446
184
other_basis_tree = self.revision_tree(self.other_basis)
447
if other_basis_tree.has_changes(self.other_tree):
448
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)
449
188
other_rev_id = self.other_basis
450
189
self.other_tree = other_basis_tree
452
@deprecated_method(deprecated_in((2, 1, 0)))
453
191
def file_revisions(self, file_id):
454
192
self.ensure_revision_trees()
455
193
def get_id(tree, file_id):
456
194
revision_id = tree.inventory[file_id].revision
195
assert revision_id is not None
457
196
return revision_id
458
197
if self.this_rev_id is None:
459
198
if self.this_basis_tree.get_file_sha1(file_id) != \
460
199
self.this_tree.get_file_sha1(file_id):
461
raise errors.WorkingTreeNotRevision(self.this_tree)
200
raise WorkingTreeNotRevision(self.this_tree)
463
202
trees = (self.this_basis_tree, self.other_tree)
464
203
return [get_id(tree, file_id) for tree in trees]
466
@deprecated_method(deprecated_in((2, 1, 0)))
467
205
def check_basis(self, check_clean, require_commits=True):
468
206
if self.this_basis is None and require_commits is True:
469
raise errors.BzrCommandError(
470
"This branch has no commits."
471
" (perhaps you would prefer 'bzr pull')")
207
raise BzrCommandError("This branch has no commits."
208
" (perhaps you would prefer 'bzr pull')")
473
210
self.compare_basis()
474
211
if self.this_basis != self.this_rev_id:
475
212
raise errors.UncommittedChanges(self.this_tree)
477
@deprecated_method(deprecated_in((2, 1, 0)))
478
214
def compare_basis(self):
480
216
basis_tree = self.revision_tree(self.this_tree.last_revision())
481
except errors.NoSuchRevision:
217
except errors.RevisionNotPresent:
482
218
basis_tree = self.this_tree.basis_tree()
483
if not self.this_tree.has_changes(basis_tree):
219
changes = self.this_tree.changes_from(basis_tree)
220
if not changes.has_changed():
484
221
self.this_rev_id = self.this_basis
486
223
def set_interesting_files(self, file_list):
487
224
self.interesting_files = file_list
489
226
def set_pending(self):
490
if (not self.base_is_ancestor or not self.base_is_other_ancestor
491
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:
493
229
self._add_parent()
495
231
def _add_parent(self):
496
232
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
497
233
new_parent_trees = []
498
operation = OperationWithCleanups(self.this_tree.set_parent_trees)
499
234
for revision_id in new_parents:
501
236
tree = self.revision_tree(revision_id)
502
except errors.NoSuchRevision:
237
except errors.RevisionNotPresent:
506
operation.add_cleanup(tree.unlock)
507
241
new_parent_trees.append((revision_id, tree))
508
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:
510
250
def set_other(self, other_revision, possible_transports=None):
511
251
"""Set the revision and tree to merge from.
628
340
self.base_rev_id = _mod_revision.ensure_null(
629
341
base_branch.get_rev_id(base_revision[1]))
630
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,
632
def make_merger(self):
633
351
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
634
352
'other_tree': self.other_tree,
635
353
'interesting_ids': self.interesting_ids,
636
354
'interesting_files': self.interesting_files,
637
'this_branch': self.this_branch,
639
356
if self.merge_type.requires_base:
640
357
kwargs['base_tree'] = self.base_tree
641
358
if self.merge_type.supports_reprocess:
642
359
kwargs['reprocess'] = self.reprocess
643
360
elif self.reprocess:
644
raise errors.BzrError(
645
"Conflict reduction is not supported for merge"
646
" type %s." % self.merge_type)
361
raise BzrError("Conflict reduction is not supported for merge"
362
" type %s." % self.merge_type)
647
363
if self.merge_type.supports_show_base:
648
364
kwargs['show_base'] = self.show_base
649
365
elif self.show_base:
650
raise errors.BzrError("Showing base is not supported for this"
366
raise BzrError("Showing base is not supported for this"
651
367
" merge type. %s" % self.merge_type)
652
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
653
and not self.base_is_other_ancestor):
654
raise errors.CannotReverseCherrypick()
655
if self.merge_type.supports_cherrypick:
656
kwargs['cherrypick'] = (not self.base_is_ancestor or
657
not self.base_is_other_ancestor)
658
if self._is_criss_cross and getattr(self.merge_type,
659
'supports_lca_trees', False):
660
kwargs['lca_trees'] = self._lca_trees
661
return self.merge_type(pb=None,
662
change_reporter=self.change_reporter,
665
def _do_merge_to(self):
666
merge = self.make_merger()
667
if self.other_branch is not None:
668
self.other_branch.update_references(self.this_branch)
670
if self.recurse == 'down':
671
for relpath, file_id in self.this_tree.iter_references():
672
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
673
other_revision = self.other_tree.get_reference_revision(
675
if other_revision == sub_tree.last_revision():
677
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
678
sub_merge.merge_type = self.merge_type
679
other_branch = self.other_branch.reference_parent(file_id, relpath)
680
sub_merge.set_other_revision(other_revision, other_branch)
681
base_revision = self.base_tree.get_reference_revision(file_id)
682
sub_merge.base_tree = \
683
sub_tree.branch.repository.revision_tree(base_revision)
684
sub_merge.base_rev_id = base_revision
689
operation = OperationWithCleanups(self._do_merge_to)
690
368
self.this_tree.lock_tree_write()
691
operation.add_cleanup(self.this_tree.unlock)
692
369
if self.base_tree is not None:
693
370
self.base_tree.lock_read()
694
operation.add_cleanup(self.base_tree.unlock)
695
371
if self.other_tree is not None:
696
372
self.other_tree.lock_read()
697
operation.add_cleanup(self.other_tree.unlock)
698
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()
699
400
if len(merge.cooked_conflicts) == 0:
700
if not self.ignore_zero and not trace.is_quiet():
701
trace.note("All changes applied successfully.")
401
if not self.ignore_zero:
402
note("All changes applied successfully.")
703
trace.note("%d conflicts encountered."
704
% len(merge.cooked_conflicts))
404
note("%d conflicts encountered." % len(merge.cooked_conflicts))
706
406
return len(merge.cooked_conflicts)
709
class _InventoryNoneEntry(object):
710
"""This represents an inventory entry which *isn't there*.
712
It simplifies the merging logic if we always have an InventoryEntry, even
713
if it isn't actually present
720
symlink_target = None
723
_none_entry = _InventoryNoneEntry()
726
409
class Merge3Merger(object):
727
410
"""Three-way merger that uses the merge3 text merger"""
728
411
requires_base = True
729
412
supports_reprocess = True
730
413
supports_show_base = True
731
414
history_based = False
732
supports_cherrypick = True
733
supports_reverse_cherrypick = True
734
415
winner_idx = {"this": 2, "other": 1, "conflict": 1}
735
supports_lca_trees = True
737
def __init__(self, working_tree, this_tree, base_tree, other_tree,
417
def __init__(self, working_tree, this_tree, base_tree, other_tree,
738
418
interesting_ids=None, reprocess=False, show_base=False,
739
pb=None, pp=None, change_reporter=None,
740
interesting_files=None, do_merge=True,
741
cherrypick=False, lca_trees=None, this_branch=None):
419
pb=DummyProgress(), pp=None, change_reporter=None,
420
interesting_files=None):
742
421
"""Initialize the merger object and perform the merge.
744
423
:param working_tree: The working tree to apply the merge to
745
424
:param this_tree: The local tree in the merge operation
746
425
:param base_tree: The common tree in the merge operation
747
:param other_tree: The other tree to merge changes from
748
:param this_branch: The branch associated with this_tree. Defaults to
749
this_tree.branch if not supplied.
426
:param other_tree: The other other tree to merge changes from
750
427
:param interesting_ids: The file_ids of files that should be
751
428
participate in the merge. May not be combined with
752
429
interesting_files.
753
430
:param: reprocess If True, perform conflict-reduction processing.
754
431
:param show_base: If True, show the base revision in text conflicts.
755
432
(incompatible with reprocess)
433
:param pb: A Progress bar
757
434
:param pp: A ProgressPhase object
758
435
:param change_reporter: An object that should report changes made
759
436
:param interesting_files: The tree-relative paths of files that should
762
439
be combined with interesting_ids. If neither interesting_files nor
763
440
interesting_ids is specified, all files may participate in the
765
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
766
if the ancestry was found to include a criss-cross merge.
767
Otherwise should be None.
769
443
object.__init__(self)
770
if interesting_files is not None and interesting_ids is not None:
772
'specify either interesting_ids or interesting_files')
773
if this_branch is None:
774
this_branch = this_tree.branch
444
if interesting_files is not None:
445
assert interesting_ids is None
775
446
self.interesting_ids = interesting_ids
776
447
self.interesting_files = interesting_files
777
448
self.this_tree = working_tree
449
self.this_tree.lock_tree_write()
778
450
self.base_tree = base_tree
451
self.base_tree.lock_read()
779
452
self.other_tree = other_tree
780
self.this_branch = this_branch
453
self.other_tree.lock_read()
781
454
self._raw_conflicts = []
782
455
self.cooked_conflicts = []
783
456
self.reprocess = reprocess
784
457
self.show_base = show_base
785
self._lca_trees = lca_trees
786
# Uncommenting this will change the default algorithm to always use
787
# _entries_lca. This can be useful for running the test suite and
788
# making sure we haven't missed any corner cases.
789
# if lca_trees is None:
790
# self._lca_trees = [self.base_tree]
791
460
self.change_reporter = change_reporter
792
self.cherrypick = cherrypick
796
warnings.warn("pp argument to Merge3Merger is deprecated")
798
warnings.warn("pb argument to Merge3Merger is deprecated")
801
operation = OperationWithCleanups(self._do_merge)
802
self.this_tree.lock_tree_write()
803
operation.add_cleanup(self.this_tree.unlock)
804
self.base_tree.lock_read()
805
operation.add_cleanup(self.base_tree.unlock)
806
self.other_tree.lock_read()
807
operation.add_cleanup(self.other_tree.unlock)
810
def _do_merge(self, operation):
811
self.tt = transform.TreeTransform(self.this_tree, None)
812
operation.add_cleanup(self.tt.finalize)
813
self._compute_transform()
814
results = self.tt.apply(no_conflicts=True)
815
self.write_modified(results)
462
self.pp = ProgressPhase("Merge phase", 3, self.pb)
464
self.tt = TreeTransform(working_tree, self.pb)
817
self.this_tree.add_conflicts(self.cooked_conflicts)
818
except errors.UnsupportedOperation:
821
def make_preview_transform(self):
822
operation = OperationWithCleanups(self._make_preview_transform)
823
self.base_tree.lock_read()
824
operation.add_cleanup(self.base_tree.unlock)
825
self.other_tree.lock_read()
826
operation.add_cleanup(self.other_tree.unlock)
827
return operation.run_simple()
829
def _make_preview_transform(self):
830
self.tt = transform.TransformPreview(self.this_tree)
831
self._compute_transform()
834
def _compute_transform(self):
835
if self._lca_trees is None:
836
467
entries = self._entries3()
837
resolver = self._three_way
839
entries = self._entries_lca()
840
resolver = self._lca_multi_way
841
child_pb = ui.ui_factory.nested_progress_bar()
843
factories = Merger.hooks['merge_file_content']
844
hooks = [factory(self) for factory in factories] + [self]
845
self.active_hooks = [hook for hook in hooks if hook is not None]
846
for num, (file_id, changed, parents3, names3,
847
executable3) in enumerate(entries):
848
child_pb.update('Preparing file merge', num, len(entries))
849
self._merge_names(file_id, parents3, names3, resolver=resolver)
851
file_status = self._do_merge_contents(file_id)
853
file_status = 'unmodified'
854
self._merge_executable(file_id,
855
executable3, file_status, resolver=resolver)
859
child_pb = ui.ui_factory.nested_progress_bar()
861
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
862
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
865
if self.change_reporter is not None:
866
from bzrlib import delta
867
delta.report_changes(
868
self.tt.iter_changes(), self.change_reporter)
869
self.cook_conflicts(fs_conflicts)
870
for conflict in self.cooked_conflicts:
871
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()
873
510
def _entries3(self):
874
511
"""Gather data about files modified between three trees.
906
540
result.append((file_id, changed, parents3, names3, executable3))
909
def _entries_lca(self):
910
"""Gather data about files modified between multiple trees.
912
This compares OTHER versus all LCA trees, and for interesting entries,
913
it then compares with THIS and BASE.
915
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
916
:return: [(file_id, changed, parents, names, executable)]
917
file_id Simple file_id of the entry
918
changed Boolean, True if the kind or contents changed
920
parents ((base, [parent_id, in, lcas]), parent_id_other,
922
names ((base, [name, in, lcas]), name_in_other, name_in_this)
923
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
925
if self.interesting_files is not None:
926
lookup_trees = [self.this_tree, self.base_tree]
927
lookup_trees.extend(self._lca_trees)
928
# I think we should include the lca trees as well
929
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
932
interesting_ids = self.interesting_ids
934
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
936
base_inventory = self.base_tree.inventory
937
this_inventory = self.this_tree.inventory
938
for path, file_id, other_ie, lca_values in walker.iter_all():
939
# Is this modified at all from any of the other trees?
941
other_ie = _none_entry
942
if interesting_ids is not None and file_id not in interesting_ids:
945
# If other_revision is found in any of the lcas, that means this
946
# node is uninteresting. This is because when merging, if there are
947
# multiple heads(), we have to create a new node. So if we didn't,
948
# we know that the ancestry is linear, and that OTHER did not
950
# See doc/developers/lca_merge_resolution.txt for details
951
other_revision = other_ie.revision
952
if other_revision is not None:
953
# We can't use this shortcut when other_revision is None,
954
# because it may be None because things are WorkingTrees, and
955
# not because it is *actually* None.
956
is_unmodified = False
957
for lca_path, ie in lca_values:
958
if ie is not None and ie.revision == other_revision:
965
for lca_path, lca_ie in lca_values:
967
lca_entries.append(_none_entry)
969
lca_entries.append(lca_ie)
971
if file_id in base_inventory:
972
base_ie = base_inventory[file_id]
974
base_ie = _none_entry
976
if file_id in this_inventory:
977
this_ie = this_inventory[file_id]
979
this_ie = _none_entry
985
for lca_ie in lca_entries:
986
lca_kinds.append(lca_ie.kind)
987
lca_parent_ids.append(lca_ie.parent_id)
988
lca_names.append(lca_ie.name)
989
lca_executable.append(lca_ie.executable)
991
kind_winner = self._lca_multi_way(
992
(base_ie.kind, lca_kinds),
993
other_ie.kind, this_ie.kind)
994
parent_id_winner = self._lca_multi_way(
995
(base_ie.parent_id, lca_parent_ids),
996
other_ie.parent_id, this_ie.parent_id)
997
name_winner = self._lca_multi_way(
998
(base_ie.name, lca_names),
999
other_ie.name, this_ie.name)
1001
content_changed = True
1002
if kind_winner == 'this':
1003
# No kind change in OTHER, see if there are *any* changes
1004
if other_ie.kind == 'directory':
1005
if parent_id_winner == 'this' and name_winner == 'this':
1006
# No change for this directory in OTHER, skip
1008
content_changed = False
1009
elif other_ie.kind is None or other_ie.kind == 'file':
1010
def get_sha1(ie, tree):
1011
if ie.kind != 'file':
1013
return tree.get_file_sha1(file_id)
1014
base_sha1 = get_sha1(base_ie, self.base_tree)
1015
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1016
in zip(lca_entries, self._lca_trees)]
1017
this_sha1 = get_sha1(this_ie, self.this_tree)
1018
other_sha1 = get_sha1(other_ie, self.other_tree)
1019
sha1_winner = self._lca_multi_way(
1020
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1021
allow_overriding_lca=False)
1022
exec_winner = self._lca_multi_way(
1023
(base_ie.executable, lca_executable),
1024
other_ie.executable, this_ie.executable)
1025
if (parent_id_winner == 'this' and name_winner == 'this'
1026
and sha1_winner == 'this' and exec_winner == 'this'):
1027
# No kind, parent, name, exec, or content change for
1028
# OTHER, so this node is not considered interesting
1030
if sha1_winner == 'this':
1031
content_changed = False
1032
elif other_ie.kind == 'symlink':
1033
def get_target(ie, tree):
1034
if ie.kind != 'symlink':
1036
return tree.get_symlink_target(file_id)
1037
base_target = get_target(base_ie, self.base_tree)
1038
lca_targets = [get_target(ie, tree) for ie, tree
1039
in zip(lca_entries, self._lca_trees)]
1040
this_target = get_target(this_ie, self.this_tree)
1041
other_target = get_target(other_ie, self.other_tree)
1042
target_winner = self._lca_multi_way(
1043
(base_target, lca_targets),
1044
other_target, this_target)
1045
if (parent_id_winner == 'this' and name_winner == 'this'
1046
and target_winner == 'this'):
1047
# No kind, parent, name, or symlink target change
1050
if target_winner == 'this':
1051
content_changed = False
1052
elif other_ie.kind == 'tree-reference':
1053
# The 'changed' information seems to be handled at a higher
1054
# level. At least, _entries3 returns False for content
1055
# changed, even when at a new revision_id.
1056
content_changed = False
1057
if (parent_id_winner == 'this' and name_winner == 'this'):
1058
# Nothing interesting
1061
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1063
# If we have gotten this far, that means something has changed
1064
result.append((file_id, content_changed,
1065
((base_ie.parent_id, lca_parent_ids),
1066
other_ie.parent_id, this_ie.parent_id),
1067
((base_ie.name, lca_names),
1068
other_ie.name, this_ie.name),
1069
((base_ie.executable, lca_executable),
1070
other_ie.executable, this_ie.executable)
1075
543
def fix_root(self):
1077
545
self.tt.final_kind(self.tt.root)
1078
except errors.NoSuchFile:
1079
547
self.tt.cancel_deletion(self.tt.root)
1080
548
if self.tt.final_file_id(self.tt.root) is None:
1081
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:
1083
553
other_root_file_id = self.other_tree.get_root_id()
1084
if other_root_file_id is None:
1086
554
other_root = self.tt.trans_id_file_id(other_root_file_id)
1087
555
if other_root == self.tt.root:
1089
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
1090
# the other tree's root is a non-root in the current tree (as when
1091
# a previously unrelated branch is merged into another)
1094
558
self.tt.final_kind(other_root)
1095
other_root_is_present = True
1096
except errors.NoSuchFile:
1097
# other_root doesn't have a physical representation. We still need
1098
# to move any references to the actual root of the tree.
1099
other_root_is_present = False
1100
# 'other_tree.inventory.root' is not present in this tree. We are
1101
# calling adjust_path for children which *want* to be present with a
1102
# correct place to go.
1103
for thing, child in self.other_tree.inventory.root.children.iteritems():
561
self.reparent_children(self.other_tree.inventory.root, self.tt.root)
562
self.tt.cancel_creation(other_root)
563
self.tt.cancel_versioning(other_root)
565
def reparent_children(self, ie, target):
566
for thing, child in ie.children.iteritems():
1104
567
trans_id = self.tt.trans_id_file_id(child.file_id)
1105
if not other_root_is_present:
1106
# FIXME: Make final_kind returns None instead of raising
1107
# NoSuchFile to avoid the ugly construct below -- vila 20100402
1109
self.tt.final_kind(trans_id)
1110
# The item exist in the final tree and has a defined place
1113
except errors.NoSuchFile, e:
1115
# Move the item into the root
1116
self.tt.adjust_path(self.tt.final_name(trans_id),
1117
self.tt.root, trans_id)
1118
if other_root_is_present:
1119
self.tt.cancel_creation(other_root)
1120
self.tt.cancel_versioning(other_root)
568
self.tt.adjust_path(self.tt.final_name(trans_id), target, trans_id)
1122
570
def write_modified(self, results):
1123
571
modified_hashes = {}
1326
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))
1328
735
# See SPOT run. run, SPOT, run.
1329
736
# So we're not QUITE repeating ourselves; we do tricky things with
1331
738
base_pair = contents_pair(self.base_tree)
1332
739
other_pair = contents_pair(self.other_tree)
1334
this_pair = contents_pair(self.this_tree)
1335
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1336
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1337
this_pair, allow_overriding_lca=False)
1339
if base_pair == other_pair:
1342
# We delayed evaluating this_pair as long as we can to avoid
1343
# unnecessary sha1 calculation
1344
this_pair = contents_pair(self.this_tree)
1345
winner = self._three_way(base_pair, other_pair, this_pair)
1346
if winner == 'this':
1347
# No interesting changes introduced by OTHER
1349
# We have a hypothetical conflict, but if we have files, then we
1350
# can try to merge the content
1351
trans_id = self.tt.trans_id_file_id(file_id)
1352
params = MergeHookParams(self, file_id, trans_id, this_pair[0],
1353
other_pair[0], winner)
1354
hooks = self.active_hooks
1355
hook_status = 'not_applicable'
1357
hook_status, lines = hook.merge_contents(params)
1358
if hook_status != 'not_applicable':
1359
# Don't try any more hooks, this one applies.
1362
if hook_status == 'not_applicable':
1363
# This is a contents conflict, because none of the available
1364
# functions could merge it.
1366
name = self.tt.final_name(trans_id)
1367
parent_id = self.tt.final_parent(trans_id)
1368
if self.this_tree.has_id(file_id):
1369
self.tt.unversion_file(trans_id)
1370
file_group = self._dump_conflicts(name, parent_id, file_id,
1372
self._raw_conflicts.append(('contents conflict', file_group))
1373
elif hook_status == 'success':
1374
self.tt.create_file(lines, trans_id)
1375
elif hook_status == 'conflicted':
1376
# XXX: perhaps the hook should be able to provide
1377
# the BASE/THIS/OTHER files?
1378
self.tt.create_file(lines, trans_id)
1379
self._raw_conflicts.append(('text conflict', trans_id))
1380
name = self.tt.final_name(trans_id)
1381
parent_id = self.tt.final_parent(trans_id)
1382
self._dump_conflicts(name, parent_id, file_id)
1383
elif hook_status == 'delete':
1384
self.tt.unversion_file(trans_id)
1386
elif hook_status == 'done':
1387
# The hook function did whatever it needs to do directly, no
1388
# further action needed here.
1391
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1392
if not self.this_tree.has_id(file_id) and result == "modified":
1393
self.tt.version_file(file_id, trans_id)
1394
# The merge has been performed, so the old contents should not be
1397
self.tt.delete_contents(trans_id)
1398
except errors.NoSuchFile:
1402
def _default_other_winner_merge(self, merge_hook_params):
1403
"""Replace this contents with other."""
1404
file_id = merge_hook_params.file_id
1405
trans_id = merge_hook_params.trans_id
1406
file_in_this = self.this_tree.has_id(file_id)
1407
if self.other_tree.has_id(file_id):
1408
# OTHER changed the file
1410
if wt.supports_content_filtering():
1411
# We get the path from the working tree if it exists.
1412
# That fails though when OTHER is adding a file, so
1413
# we fall back to the other tree to find the path if
1414
# it doesn't exist locally.
1416
filter_tree_path = wt.id2path(file_id)
1417
except errors.NoSuchId:
1418
filter_tree_path = self.other_tree.id2path(file_id)
1420
# Skip the id2path lookup for older formats
1421
filter_tree_path = None
1422
transform.create_from_tree(self.tt, trans_id,
1423
self.other_tree, file_id,
1424
filter_tree_path=filter_tree_path)
1427
# OTHER deleted the file
1428
return 'delete', None
1430
raise AssertionError(
1431
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1434
def merge_contents(self, merge_hook_params):
1435
"""Fallback merge logic after user installed hooks."""
1436
# This function is used in merge hooks as the fallback instance.
1437
# Perhaps making this function and the functions it calls be a
1438
# a separate class would be better.
1439
if merge_hook_params.winner == 'other':
1440
# OTHER is a straight winner, so replace this contents with other
1441
return self._default_other_winner_merge(merge_hook_params)
1442
elif merge_hook_params.is_file_merge():
1443
# THIS and OTHER are both files, so text merge. Either
1444
# BASE is a file, or both converted to files, so at least we
1445
# have agreement that output should be a file.
1447
self.text_merge(merge_hook_params.file_id,
1448
merge_hook_params.trans_id)
1449
except errors.BinaryFile:
1450
return 'not_applicable', None
1453
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()
1455
787
def get_lines(self, tree, file_id):
1456
788
"""Return the lines in a file, or an empty list."""
1457
if tree.has_id(file_id):
1458
790
return tree.get_file(file_id).readlines()
1511
842
determined automatically. If set_version is true, the .OTHER, .THIS
1512
843
or .BASE (in that order) will be created as versioned files.
1514
data = [('OTHER', self.other_tree, other_lines),
845
data = [('OTHER', self.other_tree, other_lines),
1515
846
('THIS', self.this_tree, this_lines)]
1517
848
data.append(('BASE', self.base_tree, base_lines))
1519
# We need to use the actual path in the working tree of the file here,
1520
# ignoring the conflict suffixes
1522
if wt.supports_content_filtering():
1524
filter_tree_path = wt.id2path(file_id)
1525
except errors.NoSuchId:
1526
# file has been deleted
1527
filter_tree_path = None
1529
# Skip the id2path lookup for older formats
1530
filter_tree_path = None
1532
849
versioned = False
1534
851
for suffix, tree, lines in data:
1535
if tree.has_id(file_id):
1536
853
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1537
suffix, lines, filter_tree_path)
1538
855
file_group.append(trans_id)
1539
856
if set_version and not versioned:
1540
857
self.tt.version_file(file_id, trans_id)
1541
858
versioned = True
1542
859
return file_group
1544
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1545
lines=None, filter_tree_path=None):
861
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1546
863
"""Emit a single conflict file."""
1547
864
name = name + '.' + suffix
1548
865
trans_id = self.tt.create_path(name, parent_id)
1549
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)
1553
870
def merge_executable(self, file_id, file_status):
1554
871
"""Perform a merge on the execute bit."""
1555
872
executable = [self.executable(t, file_id) for t in (self.base_tree,
1556
873
self.other_tree, self.this_tree)]
1557
self._merge_executable(file_id, executable, file_status,
1558
resolver=self._three_way)
874
self._merge_executable(file_id, executable, file_status)
1560
def _merge_executable(self, file_id, executable, file_status,
876
def _merge_executable(self, file_id, executable, file_status):
1562
877
"""Perform a merge on the execute bit."""
1563
878
base_executable, other_executable, this_executable = executable
1564
879
if file_status == "deleted":
1566
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)
1567
888
if winner == "conflict":
1568
889
# There must be a None in here, if we have a conflict, but we
1569
890
# need executability since file status was not deleted.
1629
935
if path.endswith(suffix):
1630
936
path = path[:-len(suffix)]
1632
c = _mod_conflicts.Conflict.factory(conflict_type,
1633
path=path, file_id=file_id)
1634
elif conflict_type == 'text conflict':
938
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
939
self.cooked_conflicts.append(c)
940
if conflict_type == 'text conflict':
1635
941
trans_id = conflict[1]
1636
942
path = fp.get_path(trans_id)
1637
943
file_id = self.tt.final_file_id(trans_id)
1638
c = _mod_conflicts.Conflict.factory(conflict_type,
1639
path=path, file_id=file_id)
944
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
945
self.cooked_conflicts.append(c)
947
for trans_id, conflicts in name_conflicts.iteritems():
949
this_parent, other_parent = conflicts['parent conflict']
950
assert this_parent != other_parent
952
this_parent = other_parent = \
953
self.tt.final_file_id(self.tt.final_parent(trans_id))
955
this_name, other_name = conflicts['name conflict']
956
assert this_name != other_name
958
this_name = other_name = self.tt.final_name(trans_id)
959
other_path = fp.get_path(trans_id)
960
if this_parent is not None and this_name is not None:
962
fp.get_path(self.tt.trans_id_file_id(this_parent))
963
this_path = pathjoin(this_parent_path, this_name)
1641
raise AssertionError('bad conflict type: %r' % (conflict,))
965
this_path = "<deleted>"
966
file_id = self.tt.final_file_id(trans_id)
967
c = Conflict.factory('path conflict', path=this_path,
968
conflict_path=other_path, file_id=file_id)
1642
969
self.cooked_conflicts.append(c)
1643
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
970
self.cooked_conflicts.sort(key=Conflict.sort_key)
1646
973
class WeaveMerger(Merge3Merger):
1647
974
"""Three-way tree merger, text weave merger."""
1648
975
supports_reprocess = True
1649
976
supports_show_base = False
1650
supports_reverse_cherrypick = False
1651
history_based = True
1653
def _generate_merge_plan(self, file_id, base):
1654
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)
1657
988
def _merged_lines(self, file_id):
1658
989
"""Generate the merged lines.
1659
990
There is no distinction between lines that are meant to contain <<<<<<<
1663
base = self.base_tree
1666
plan = self._generate_merge_plan(file_id, base)
1667
if 'merge' in debug.debug_flags:
1669
trans_id = self.tt.trans_id_file_id(file_id)
1670
name = self.tt.final_name(trans_id) + '.plan'
1671
contents = ('%11s|%s' % l for l in plan)
1672
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1673
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1674
'>>>>>>> MERGE-SOURCE\n')
1675
lines, conflicts = textmerge.merge_lines(self.reprocess)
1677
base_lines = textmerge.base_from_plan()
1680
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)
1682
998
def text_merge(self, file_id, trans_id):
1683
999
"""Perform a (weave) text merge for a given file and file-id.
1684
1000
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1685
1001
and a conflict will be noted.
1687
lines, base_lines = self._merged_lines(file_id)
1003
lines, conflicts = self._merged_lines(file_id)
1688
1004
lines = list(lines)
1689
# 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,
1690
1006
# because we don't want to get into weave merge guts.
1691
textfile.check_text_lines(lines)
1007
check_text_lines(lines)
1692
1008
self.tt.create_file(lines, trans_id)
1693
if base_lines is not None:
1695
1010
self._raw_conflicts.append(('text conflict', trans_id))
1696
1011
name = self.tt.final_name(trans_id)
1697
1012
parent_id = self.tt.final_parent(trans_id)
1698
file_group = self._dump_conflicts(name, parent_id, file_id,
1700
base_lines=base_lines)
1013
file_group = self._dump_conflicts(name, parent_id, file_id,
1701
1015
file_group.append(trans_id)
1704
class LCAMerger(WeaveMerger):
1706
def _generate_merge_plan(self, file_id, base):
1707
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1710
1018
class Diff3Merger(Merge3Merger):
1711
1019
"""Three-way merger using external diff3 for text merging"""
1713
1021
def dump_file(self, temp_dir, name, tree, file_id):
1714
out_path = osutils.pathjoin(temp_dir, name)
1022
out_path = pathjoin(temp_dir, name)
1715
1023
out_file = open(out_path, "wb")
1717
1025
in_file = tree.get_file(file_id)
1828
1130
yield status_a(revision, text)
1829
1131
for revision, text in annotated_b[b_cur:bi]:
1830
1132
yield status_b(revision, text)
1831
1134
# and now the matched section
1834
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
1835
1139
yield "unchanged", text_a
1838
class _PlanMergeBase(object):
1840
def __init__(self, a_rev, b_rev, vf, key_prefix):
1843
:param a_rev: Revision-id of one revision to merge
1844
:param b_rev: Revision-id of the other revision to merge
1845
:param vf: A VersionedFiles containing both revisions
1846
:param key_prefix: A prefix for accessing keys in vf, typically
1852
self._last_lines = None
1853
self._last_lines_revision_id = None
1854
self._cached_matching_blocks = {}
1855
self._key_prefix = key_prefix
1856
self._precache_tip_lines()
1858
def _precache_tip_lines(self):
1859
lines = self.get_lines([self.a_rev, self.b_rev])
1860
self.lines_a = lines[self.a_rev]
1861
self.lines_b = lines[self.b_rev]
1863
def get_lines(self, revisions):
1864
"""Get lines for revisions from the backing VersionedFiles.
1866
:raises RevisionNotPresent: on absent texts.
1868
keys = [(self._key_prefix + (rev,)) for rev in revisions]
1870
for record in self.vf.get_record_stream(keys, 'unordered', True):
1871
if record.storage_kind == 'absent':
1872
raise errors.RevisionNotPresent(record.key, self.vf)
1873
result[record.key[-1]] = osutils.chunks_to_lines(
1874
record.get_bytes_as('chunked'))
1877
def plan_merge(self):
1878
"""Generate a 'plan' for merging the two revisions.
1880
This involves comparing their texts and determining the cause of
1881
differences. If text A has a line and text B does not, then either the
1882
line was added to text A, or it was deleted from B. Once the causes
1883
are combined, they are written out in the format described in
1884
VersionedFile.plan_merge
1886
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
1887
unique_a, unique_b = self._unique_lines(blocks)
1888
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
1889
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
1890
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
1892
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
1895
for i, j, n in blocks:
1896
for a_index in range(last_i, i):
1897
if a_index in new_a:
1898
if a_index in killed_b:
1899
yield 'conflicted-a', self.lines_a[a_index]
1901
yield 'new-a', self.lines_a[a_index]
1903
yield 'killed-b', self.lines_a[a_index]
1904
for b_index in range(last_j, j):
1905
if b_index in new_b:
1906
if b_index in killed_a:
1907
yield 'conflicted-b', self.lines_b[b_index]
1909
yield 'new-b', self.lines_b[b_index]
1911
yield 'killed-a', self.lines_b[b_index]
1912
# handle common lines
1913
for a_index in range(i, i+n):
1914
yield 'unchanged', self.lines_a[a_index]
1918
def _get_matching_blocks(self, left_revision, right_revision):
1919
"""Return a description of which sections of two revisions match.
1921
See SequenceMatcher.get_matching_blocks
1923
cached = self._cached_matching_blocks.get((left_revision,
1925
if cached is not None:
1927
if self._last_lines_revision_id == left_revision:
1928
left_lines = self._last_lines
1929
right_lines = self.get_lines([right_revision])[right_revision]
1931
lines = self.get_lines([left_revision, right_revision])
1932
left_lines = lines[left_revision]
1933
right_lines = lines[right_revision]
1934
self._last_lines = right_lines
1935
self._last_lines_revision_id = right_revision
1936
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
1938
return matcher.get_matching_blocks()
1940
def _unique_lines(self, matching_blocks):
1941
"""Analyse matching_blocks to determine which lines are unique
1943
:return: a tuple of (unique_left, unique_right), where the values are
1944
sets of line numbers of unique lines.
1950
for i, j, n in matching_blocks:
1951
unique_left.extend(range(last_i, i))
1952
unique_right.extend(range(last_j, j))
1955
return unique_left, unique_right
1958
def _subtract_plans(old_plan, new_plan):
1959
"""Remove changes from new_plan that came from old_plan.
1961
It is assumed that the difference between the old_plan and new_plan
1962
is their choice of 'b' text.
1964
All lines from new_plan that differ from old_plan are emitted
1965
verbatim. All lines from new_plan that match old_plan but are
1966
not about the 'b' revision are emitted verbatim.
1968
Lines that match and are about the 'b' revision are the lines we
1969
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
1970
is skipped entirely.
1972
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
1975
for i, j, n in matcher.get_matching_blocks():
1976
for jj in range(last_j, j):
1978
for jj in range(j, j+n):
1979
plan_line = new_plan[jj]
1980
if plan_line[0] == 'new-b':
1982
elif plan_line[0] == 'killed-b':
1983
yield 'unchanged', plan_line[1]
1989
class _PlanMerge(_PlanMergeBase):
1990
"""Plan an annotate merge using on-the-fly annotation"""
1992
def __init__(self, a_rev, b_rev, vf, key_prefix):
1993
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
1994
self.a_key = self._key_prefix + (self.a_rev,)
1995
self.b_key = self._key_prefix + (self.b_rev,)
1996
self.graph = _mod_graph.Graph(self.vf)
1997
heads = self.graph.heads((self.a_key, self.b_key))
1999
# one side dominates, so we can just return its values, yay for
2001
# Ideally we would know that before we get this far
2002
self._head_key = heads.pop()
2003
if self._head_key == self.a_key:
2007
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2008
self._head_key[-1], other)
2011
self._head_key = None
2014
def _precache_tip_lines(self):
2015
# Turn this into a no-op, because we will do this later
2018
def _find_recursive_lcas(self):
2019
"""Find all the ancestors back to a unique lca"""
2020
cur_ancestors = (self.a_key, self.b_key)
2021
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2022
# rather than a key tuple. We will just map that directly to no common
2026
next_lcas = self.graph.find_lca(*cur_ancestors)
2027
# Map a plain NULL_REVISION to a simple no-ancestors
2028
if next_lcas == set([_mod_revision.NULL_REVISION]):
2030
# Order the lca's based on when they were merged into the tip
2031
# While the actual merge portion of weave merge uses a set() of
2032
# active revisions, the order of insertion *does* effect the
2033
# implicit ordering of the texts.
2034
for rev_key in cur_ancestors:
2035
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2037
parent_map[rev_key] = ordered_parents
2038
if len(next_lcas) == 0:
2040
elif len(next_lcas) == 1:
2041
parent_map[list(next_lcas)[0]] = ()
2043
elif len(next_lcas) > 2:
2044
# More than 2 lca's, fall back to grabbing all nodes between
2045
# this and the unique lca.
2046
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2048
self.a_key, self.b_key, cur_ancestors)
2049
cur_lcas = next_lcas
2050
while len(cur_lcas) > 1:
2051
cur_lcas = self.graph.find_lca(*cur_lcas)
2052
if len(cur_lcas) == 0:
2053
# No common base to find, use the full ancestry
2056
unique_lca = list(cur_lcas)[0]
2057
if unique_lca == _mod_revision.NULL_REVISION:
2058
# find_lca will return a plain 'NULL_REVISION' rather
2059
# than a key tuple when there is no common ancestor, we
2060
# prefer to just use None, because it doesn't confuse
2061
# _get_interesting_texts()
2063
parent_map.update(self._find_unique_parents(next_lcas,
2066
cur_ancestors = next_lcas
2069
def _find_unique_parents(self, tip_keys, base_key):
2070
"""Find ancestors of tip that aren't ancestors of base.
2072
:param tip_keys: Nodes that are interesting
2073
:param base_key: Cull all ancestors of this node
2074
:return: The parent map for all revisions between tip_keys and
2075
base_key. base_key will be included. References to nodes outside of
2076
the ancestor set will also be removed.
2078
# TODO: this would be simpler if find_unique_ancestors took a list
2079
# instead of a single tip, internally it supports it, but it
2080
# isn't a "backwards compatible" api change.
2081
if base_key is None:
2082
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2083
# We remove NULL_REVISION because it isn't a proper tuple key, and
2084
# thus confuses things like _get_interesting_texts, and our logic
2085
# to add the texts into the memory weave.
2086
if _mod_revision.NULL_REVISION in parent_map:
2087
parent_map.pop(_mod_revision.NULL_REVISION)
2090
for tip in tip_keys:
2092
self.graph.find_unique_ancestors(tip, [base_key]))
2093
parent_map = self.graph.get_parent_map(interesting)
2094
parent_map[base_key] = ()
2095
culled_parent_map, child_map, tails = self._remove_external_references(
2097
# Remove all the tails but base_key
2098
if base_key is not None:
2099
tails.remove(base_key)
2100
self._prune_tails(culled_parent_map, child_map, tails)
2101
# Now remove all the uninteresting 'linear' regions
2102
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2106
def _remove_external_references(parent_map):
2107
"""Remove references that go outside of the parent map.
2109
:param parent_map: Something returned from Graph.get_parent_map(keys)
2110
:return: (filtered_parent_map, child_map, tails)
2111
filtered_parent_map is parent_map without external references
2112
child_map is the {parent_key: [child_keys]} mapping
2113
tails is a list of nodes that do not have any parents in the map
2115
# TODO: The basic effect of this function seems more generic than
2116
# _PlanMerge. But the specific details of building a child_map,
2117
# and computing tails seems very specific to _PlanMerge.
2118
# Still, should this be in Graph land?
2119
filtered_parent_map = {}
2122
for key, parent_keys in parent_map.iteritems():
2123
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2124
if not culled_parent_keys:
2126
for parent_key in culled_parent_keys:
2127
child_map.setdefault(parent_key, []).append(key)
2128
# TODO: Do we want to do this, it adds overhead for every node,
2129
# just to say that the node has no children
2130
child_map.setdefault(key, [])
2131
filtered_parent_map[key] = culled_parent_keys
2132
return filtered_parent_map, child_map, tails
2135
def _prune_tails(parent_map, child_map, tails_to_remove):
2136
"""Remove tails from the parent map.
2138
This will remove the supplied revisions until no more children have 0
2141
:param parent_map: A dict of {child: [parents]}, this dictionary will
2142
be modified in place.
2143
:param tails_to_remove: A list of tips that should be removed,
2144
this list will be consumed
2145
:param child_map: The reverse dict of parent_map ({parent: [children]})
2146
this dict will be modified
2147
:return: None, parent_map will be modified in place.
2149
while tails_to_remove:
2150
next = tails_to_remove.pop()
2151
parent_map.pop(next)
2152
children = child_map.pop(next)
2153
for child in children:
2154
child_parents = parent_map[child]
2155
child_parents.remove(next)
2156
if len(child_parents) == 0:
2157
tails_to_remove.append(child)
2159
def _get_interesting_texts(self, parent_map):
2160
"""Return a dict of texts we are interested in.
2162
Note that the input is in key tuples, but the output is in plain
2165
:param parent_map: The output from _find_recursive_lcas
2166
:return: A dict of {'revision_id':lines} as returned by
2167
_PlanMergeBase.get_lines()
2169
all_revision_keys = set(parent_map)
2170
all_revision_keys.add(self.a_key)
2171
all_revision_keys.add(self.b_key)
2173
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2174
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2177
def _build_weave(self):
2178
from bzrlib import weave
2179
self._weave = weave.Weave(weave_name='in_memory_weave',
2180
allow_reserved=True)
2181
parent_map = self._find_recursive_lcas()
2183
all_texts = self._get_interesting_texts(parent_map)
2185
# Note: Unfortunately, the order given by topo_sort will effect the
2186
# ordering resolution in the output. Specifically, if you add A then B,
2187
# then in the output text A lines will show up before B lines. And, of
2188
# course, topo_sort doesn't guarantee any real ordering.
2189
# So we use merge_sort, and add a fake node on the tip.
2190
# This ensures that left-hand parents will always be inserted into the
2191
# weave before right-hand parents.
2192
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2193
parent_map[tip_key] = (self.a_key, self.b_key)
2195
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2199
# for key in tsort.topo_sort(parent_map):
2200
parent_keys = parent_map[key]
2201
revision_id = key[-1]
2202
parent_ids = [k[-1] for k in parent_keys]
2203
self._weave.add_lines(revision_id, parent_ids,
2204
all_texts[revision_id])
2206
def plan_merge(self):
2207
"""Generate a 'plan' for merging the two revisions.
2209
This involves comparing their texts and determining the cause of
2210
differences. If text A has a line and text B does not, then either the
2211
line was added to text A, or it was deleted from B. Once the causes
2212
are combined, they are written out in the format described in
2213
VersionedFile.plan_merge
2215
if self._head_key is not None: # There was a single head
2216
if self._head_key == self.a_key:
2219
if self._head_key != self.b_key:
2220
raise AssertionError('There was an invalid head: %s != %s'
2221
% (self.b_key, self._head_key))
2223
head_rev = self._head_key[-1]
2224
lines = self.get_lines([head_rev])[head_rev]
2225
return ((plan, line) for line in lines)
2226
return self._weave.plan_merge(self.a_rev, self.b_rev)
2229
class _PlanLCAMerge(_PlanMergeBase):
2231
This merge algorithm differs from _PlanMerge in that:
2232
1. comparisons are done against LCAs only
2233
2. cases where a contested line is new versus one LCA but old versus
2234
another are marked as conflicts, by emitting the line as conflicted-a
2237
This is faster, and hopefully produces more useful output.
2240
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2241
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2242
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2245
if lca == _mod_revision.NULL_REVISION:
2248
self.lcas.add(lca[-1])
2249
for lca in self.lcas:
2250
if _mod_revision.is_null(lca):
2253
lca_lines = self.get_lines([lca])[lca]
2254
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2256
blocks = list(matcher.get_matching_blocks())
2257
self._cached_matching_blocks[(a_rev, lca)] = blocks
2258
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2260
blocks = list(matcher.get_matching_blocks())
2261
self._cached_matching_blocks[(b_rev, lca)] = blocks
2263
def _determine_status(self, revision_id, unique_line_numbers):
2264
"""Determines the status unique lines versus all lcas.
2266
Basically, determines why the line is unique to this revision.
2268
A line may be determined new, killed, or both.
2270
If a line is determined new, that means it was not present in at least
2271
one LCA, and is not present in the other merge revision.
2273
If a line is determined killed, that means the line was present in
2276
If a line is killed and new, this indicates that the two merge
2277
revisions contain differing conflict resolutions.
2278
:param revision_id: The id of the revision in which the lines are
2280
:param unique_line_numbers: The line numbers of unique lines.
2281
:return a tuple of (new_this, killed_other):
2285
unique_line_numbers = set(unique_line_numbers)
2286
for lca in self.lcas:
2287
blocks = self._get_matching_blocks(revision_id, lca)
2288
unique_vs_lca, _ignored = self._unique_lines(blocks)
2289
new.update(unique_line_numbers.intersection(unique_vs_lca))
2290
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
