89
71
class Merger(object):
90
72
def __init__(self, this_branch, other_tree=None, base_tree=None,
91
this_tree=None, pb=DummyProgress(), change_reporter=None):
73
this_tree=None, pb=DummyProgress(), change_reporter=None,
74
recurse='down', revision_graph=None):
92
75
object.__init__(self)
93
assert this_tree is not None, "this_tree is required"
94
76
self.this_branch = this_branch
95
self.this_basis = this_branch.last_revision()
77
self.this_basis = _mod_revision.ensure_null(
78
this_branch.last_revision())
96
79
self.this_rev_id = None
97
80
self.this_tree = this_tree
98
81
self.this_revision_tree = None
99
82
self.this_basis_tree = None
100
83
self.other_tree = other_tree
84
self.other_branch = None
101
85
self.base_tree = base_tree
102
86
self.ignore_zero = False
103
87
self.backup_files = False
104
88
self.interesting_ids = None
89
self.interesting_files = None
105
90
self.show_base = False
106
91
self.reprocess = False
94
self.recurse = recurse
109
95
self.change_reporter = change_reporter
111
def revision_tree(self, revision_id):
112
return self.this_branch.repository.revision_tree(revision_id)
96
self._cached_trees = {}
97
self._revision_graph = revision_graph
98
self._base_is_ancestor = None
99
self._base_is_other_ancestor = None
100
self._is_criss_cross = None
101
self._lca_trees = None
104
def revision_graph(self):
105
if self._revision_graph is None:
106
self._revision_graph = self.this_branch.repository.get_graph()
107
return self._revision_graph
109
def _set_base_is_ancestor(self, value):
110
self._base_is_ancestor = value
112
def _get_base_is_ancestor(self):
113
if self._base_is_ancestor is None:
114
self._base_is_ancestor = self.revision_graph.is_ancestor(
115
self.base_rev_id, self.this_basis)
116
return self._base_is_ancestor
118
base_is_ancestor = property(_get_base_is_ancestor, _set_base_is_ancestor)
120
def _set_base_is_other_ancestor(self, value):
121
self._base_is_other_ancestor = value
123
def _get_base_is_other_ancestor(self):
124
if self._base_is_other_ancestor is None:
125
if self.other_basis is None:
127
self._base_is_other_ancestor = self.revision_graph.is_ancestor(
128
self.base_rev_id, self.other_basis)
129
return self._base_is_other_ancestor
131
base_is_other_ancestor = property(_get_base_is_other_ancestor,
132
_set_base_is_other_ancestor)
135
def from_uncommitted(tree, other_tree, pb, base_tree=None):
136
"""Return a Merger for uncommitted changes in other_tree.
138
:param tree: The tree to merge into
139
:param other_tree: The tree to get uncommitted changes from
140
:param pb: A progress indicator
141
:param base_tree: The basis to use for the merge. If unspecified,
142
other_tree.basis_tree() will be used.
144
if base_tree is None:
145
base_tree = other_tree.basis_tree()
146
merger = Merger(tree.branch, other_tree, base_tree, tree, pb)
147
merger.base_rev_id = merger.base_tree.get_revision_id()
148
merger.other_rev_id = None
149
merger.other_basis = merger.base_rev_id
153
def from_mergeable(klass, tree, mergeable, pb):
154
"""Return a Merger for a bundle or merge directive.
156
:param tree: The tree to merge changes into
157
:param mergeable: A merge directive or bundle
158
:param pb: A progress indicator
160
mergeable.install_revisions(tree.branch.repository)
161
base_revision_id, other_revision_id, verified =\
162
mergeable.get_merge_request(tree.branch.repository)
163
revision_graph = tree.branch.repository.get_graph()
164
if base_revision_id is not None:
165
if (base_revision_id != _mod_revision.NULL_REVISION and
166
revision_graph.is_ancestor(
167
base_revision_id, tree.branch.last_revision())):
168
base_revision_id = None
170
warning('Performing cherrypick')
171
merger = klass.from_revision_ids(pb, tree, other_revision_id,
172
base_revision_id, revision_graph=
174
return merger, verified
177
def from_revision_ids(pb, tree, other, base=None, other_branch=None,
178
base_branch=None, revision_graph=None,
180
"""Return a Merger for revision-ids.
182
:param pb: A progress indicator
183
:param tree: The tree to merge changes into
184
:param other: The revision-id to use as OTHER
185
:param base: The revision-id to use as BASE. If not specified, will
187
:param other_branch: A branch containing the other revision-id. If
188
not supplied, tree.branch is used.
189
:param base_branch: A branch containing the base revision-id. If
190
not supplied, other_branch or tree.branch will be used.
191
:param revision_graph: If you have a revision_graph precomputed, pass
192
it in, otherwise it will be created for you.
193
:param tree_branch: The branch associated with tree. If not supplied,
194
tree.branch will be used.
196
if tree_branch is None:
197
tree_branch = tree.branch
198
merger = Merger(tree_branch, this_tree=tree, pb=pb,
199
revision_graph=revision_graph)
200
if other_branch is None:
201
other_branch = tree.branch
202
merger.set_other_revision(other, other_branch)
206
if base_branch is None:
207
base_branch = other_branch
208
merger.set_base_revision(base, base_branch)
211
def revision_tree(self, revision_id, branch=None):
212
if revision_id not in self._cached_trees:
214
branch = self.this_branch
216
tree = self.this_tree.revision_tree(revision_id)
217
except errors.NoSuchRevisionInTree:
218
tree = branch.repository.revision_tree(revision_id)
219
self._cached_trees[revision_id] = tree
220
return self._cached_trees[revision_id]
222
def _get_tree(self, treespec, possible_transports=None):
223
from bzrlib import workingtree
224
location, revno = treespec
226
tree = workingtree.WorkingTree.open_containing(location)[0]
227
return tree.branch, tree
228
branch = Branch.open_containing(location, possible_transports)[0]
230
revision_id = branch.last_revision()
232
revision_id = branch.get_rev_id(revno)
233
revision_id = ensure_null(revision_id)
234
return branch, self.revision_tree(revision_id, branch)
114
236
def ensure_revision_trees(self):
115
237
if self.this_revision_tree is None:
116
self.this_basis_tree = self.this_branch.repository.revision_tree(
238
self.this_basis_tree = self.revision_tree(self.this_basis)
118
239
if self.this_basis == self.this_rev_id:
119
240
self.this_revision_tree = self.this_basis_tree
148
268
self.compare_basis()
149
269
if self.this_basis != self.this_rev_id:
150
raise BzrCommandError("Working tree has uncommitted changes.")
270
raise errors.UncommittedChanges(self.this_tree)
152
272
def compare_basis(self):
153
changes = self.this_tree.changes_from(self.this_tree.basis_tree())
274
basis_tree = self.revision_tree(self.this_tree.last_revision())
275
except errors.NoSuchRevision:
276
basis_tree = self.this_tree.basis_tree()
277
changes = self.this_tree.changes_from(basis_tree)
154
278
if not changes.has_changed():
155
279
self.this_rev_id = self.this_basis
157
281
def set_interesting_files(self, file_list):
159
self._set_interesting_files(file_list)
160
except NotVersionedError, e:
161
raise BzrCommandError("%s is not a source file in any"
164
def _set_interesting_files(self, file_list):
165
"""Set the list of interesting ids from a list of files."""
166
if file_list is None:
167
self.interesting_ids = None
170
interesting_ids = set()
171
for path in file_list:
173
# TODO: jam 20070226 The trees are not locked at this time,
174
# wouldn't it make merge faster if it locks everything in the
175
# beginning? It locks at do_merge time, but this happens
177
for tree in (self.this_tree, self.base_tree, self.other_tree):
178
file_id = tree.path2id(path)
179
if file_id is not None:
180
interesting_ids.add(file_id)
183
raise NotVersionedError(path=path)
184
self.interesting_ids = interesting_ids
282
self.interesting_files = file_list
186
284
def set_pending(self):
187
if not self.base_is_ancestor:
189
if self.other_rev_id is None:
191
ancestry = self.this_branch.repository.get_ancestry(self.this_basis)
192
if self.other_rev_id in ancestry:
194
self.this_tree.add_parent_tree((self.other_rev_id, self.other_tree))
196
def set_other(self, other_revision):
285
if not self.base_is_ancestor or not self.base_is_other_ancestor or self.other_rev_id is None:
289
def _add_parent(self):
290
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
291
new_parent_trees = []
292
for revision_id in new_parents:
294
tree = self.revision_tree(revision_id)
295
except errors.NoSuchRevision:
299
new_parent_trees.append((revision_id, tree))
301
self.this_tree.set_parent_trees(new_parent_trees,
302
allow_leftmost_as_ghost=True)
304
for _revision_id, tree in new_parent_trees:
308
def set_other(self, other_revision, possible_transports=None):
197
309
"""Set the revision and tree to merge from.
199
311
This sets the other_tree, other_rev_id, other_basis attributes.
201
313
:param other_revision: The [path, revision] list to merge from.
203
other_branch, self.other_tree = _get_tree(other_revision,
315
self.other_branch, self.other_tree = self._get_tree(other_revision,
205
317
if other_revision[1] == -1:
206
self.other_rev_id = other_branch.last_revision()
207
if self.other_rev_id is None:
208
raise NoCommits(other_branch)
318
self.other_rev_id = _mod_revision.ensure_null(
319
self.other_branch.last_revision())
320
if _mod_revision.is_null(self.other_rev_id):
321
raise NoCommits(self.other_branch)
209
322
self.other_basis = self.other_rev_id
210
323
elif other_revision[1] is not None:
211
self.other_rev_id = other_branch.get_rev_id(other_revision[1])
324
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
212
325
self.other_basis = self.other_rev_id
214
327
self.other_rev_id = None
215
self.other_basis = other_branch.last_revision()
328
self.other_basis = self.other_branch.last_revision()
216
329
if self.other_basis is None:
217
raise NoCommits(other_branch)
218
if other_branch.base != self.this_branch.base:
219
self.this_branch.fetch(other_branch, last_revision=self.other_basis)
330
raise NoCommits(self.other_branch)
331
if self.other_rev_id is not None:
332
self._cached_trees[self.other_rev_id] = self.other_tree
333
self._maybe_fetch(self.other_branch,self.this_branch, self.other_basis)
335
def set_other_revision(self, revision_id, other_branch):
336
"""Set 'other' based on a branch and revision id
338
:param revision_id: The revision to use for a tree
339
:param other_branch: The branch containing this tree
341
self.other_rev_id = revision_id
342
self.other_branch = other_branch
343
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
344
self.other_tree = self.revision_tree(revision_id)
345
self.other_basis = revision_id
347
def set_base_revision(self, revision_id, branch):
348
"""Set 'base' based on a branch and revision id
350
:param revision_id: The revision to use for a tree
351
:param branch: The branch containing this tree
353
self.base_rev_id = revision_id
354
self.base_branch = branch
355
self._maybe_fetch(branch, self.this_branch, revision_id)
356
self.base_tree = self.revision_tree(revision_id)
358
def _maybe_fetch(self, source, target, revision_id):
359
if not source.repository.has_same_location(target.repository):
360
target.fetch(source, revision_id)
221
362
def find_base(self):
222
self.set_base([None, None])
363
revisions = [ensure_null(self.this_basis),
364
ensure_null(self.other_basis)]
365
if NULL_REVISION in revisions:
366
self.base_rev_id = NULL_REVISION
367
self.base_tree = self.revision_tree(self.base_rev_id)
368
self._is_criss_cross = False
370
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
371
self._is_criss_cross = False
373
self.base_rev_id = NULL_REVISION
375
self.base_rev_id = list(lcas)[0]
376
else: # len(lcas) > 1
378
# find_unique_lca can only handle 2 nodes, so we have to
379
# start back at the beginning. It is a shame to traverse
380
# the graph again, but better than re-implementing
382
self.base_rev_id = self.revision_graph.find_unique_lca(
383
revisions[0], revisions[1])
385
self.base_rev_id = self.revision_graph.find_unique_lca(
387
self._is_criss_cross = True
388
if self.base_rev_id == NULL_REVISION:
389
raise UnrelatedBranches()
390
if self._is_criss_cross:
391
warning('Warning: criss-cross merge encountered. See bzr'
392
' help criss-cross.')
393
mutter('Criss-cross lcas: %r' % lcas)
394
interesting_revision_ids = [self.base_rev_id]
395
interesting_revision_ids.extend(lcas)
396
interesting_trees = dict((t.get_revision_id(), t)
397
for t in self.this_branch.repository.revision_trees(
398
interesting_revision_ids))
399
self._cached_trees.update(interesting_trees)
400
self.base_tree = interesting_trees.pop(self.base_rev_id)
401
sorted_lca_keys = self.revision_graph.find_merge_order(
403
self._lca_trees = [interesting_trees[key]
404
for key in sorted_lca_keys]
406
self.base_tree = self.revision_tree(self.base_rev_id)
407
self.base_is_ancestor = True
408
self.base_is_other_ancestor = True
409
mutter('Base revid: %r' % self.base_rev_id)
224
411
def set_base(self, base_revision):
225
412
"""Set the base revision to use for the merge.
272
445
kwargs['show_base'] = self.show_base
273
446
elif self.show_base:
274
447
raise BzrError("Showing base is not supported for this"
275
" merge type. %s" % self.merge_type)
448
" merge type. %s" % self.merge_type)
449
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
450
and not self.base_is_other_ancestor):
451
raise errors.CannotReverseCherrypick()
452
if self.merge_type.supports_cherrypick:
453
kwargs['cherrypick'] = (not self.base_is_ancestor or
454
not self.base_is_other_ancestor)
455
if self._is_criss_cross and getattr(self.merge_type,
456
'supports_lca_trees', False):
457
kwargs['lca_trees'] = self._lca_trees
458
return self.merge_type(pb=self._pb,
459
change_reporter=self.change_reporter,
462
def _do_merge_to(self, merge):
464
if self.recurse == 'down':
465
for relpath, file_id in self.this_tree.iter_references():
466
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
467
other_revision = self.other_tree.get_reference_revision(
469
if other_revision == sub_tree.last_revision():
471
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
472
sub_merge.merge_type = self.merge_type
473
other_branch = self.other_branch.reference_parent(file_id, relpath)
474
sub_merge.set_other_revision(other_revision, other_branch)
475
base_revision = self.base_tree.get_reference_revision(file_id)
476
sub_merge.base_tree = \
477
sub_tree.branch.repository.revision_tree(base_revision)
478
sub_merge.base_rev_id = base_revision
276
482
self.this_tree.lock_tree_write()
277
if self.base_tree is not None:
278
self.base_tree.lock_read()
279
if self.other_tree is not None:
280
self.other_tree.lock_read()
282
merge = self.merge_type(pb=self._pb,
283
change_reporter=self.change_reporter,
286
if self.other_tree is not None:
287
self.other_tree.unlock()
288
484
if self.base_tree is not None:
289
self.base_tree.unlock()
485
self.base_tree.lock_read()
487
if self.other_tree is not None:
488
self.other_tree.lock_read()
490
merge = self.make_merger()
491
self._do_merge_to(merge)
493
if self.other_tree is not None:
494
self.other_tree.unlock()
496
if self.base_tree is not None:
497
self.base_tree.unlock()
290
499
self.this_tree.unlock()
291
500
if len(merge.cooked_conflicts) == 0:
292
if not self.ignore_zero:
501
if not self.ignore_zero and not is_quiet():
293
502
note("All changes applied successfully.")
295
504
note("%d conflicts encountered." % len(merge.cooked_conflicts))
297
506
return len(merge.cooked_conflicts)
299
def regen_inventory(self, new_entries):
300
old_entries = self.this_tree.read_working_inventory()
304
for path, file_id in new_entries:
307
new_entries_map[file_id] = path
309
def id2path(file_id):
310
path = new_entries_map.get(file_id)
313
entry = old_entries[file_id]
314
if entry.parent_id is None:
316
return pathjoin(id2path(entry.parent_id), entry.name)
318
for file_id in old_entries:
319
entry = old_entries[file_id]
320
path = id2path(file_id)
321
if file_id in self.base_tree.inventory:
322
executable = getattr(self.base_tree.inventory[file_id], 'executable', False)
324
executable = getattr(entry, 'executable', False)
325
new_inventory[file_id] = (path, file_id, entry.parent_id,
326
entry.kind, executable)
328
by_path[path] = file_id
333
for path, file_id in new_entries:
335
del new_inventory[file_id]
338
new_path_list.append((path, file_id))
339
if file_id not in old_entries:
341
# Ensure no file is added before its parent
343
for path, file_id in new_path_list:
347
parent = by_path[os.path.dirname(path)]
348
abspath = pathjoin(self.this_tree.basedir, path)
349
kind = osutils.file_kind(abspath)
350
if file_id in self.base_tree.inventory:
351
executable = getattr(self.base_tree.inventory[file_id], 'executable', False)
354
new_inventory[file_id] = (path, file_id, parent, kind, executable)
355
by_path[path] = file_id
357
# Get a list in insertion order
358
new_inventory_list = new_inventory.values()
359
mutter ("""Inventory regeneration:
360
old length: %i insertions: %i deletions: %i new_length: %i"""\
361
% (len(old_entries), insertions, deletions,
362
len(new_inventory_list)))
363
assert len(new_inventory_list) == len(old_entries) + insertions\
365
new_inventory_list.sort()
366
return new_inventory_list
509
class _InventoryNoneEntry(object):
510
"""This represents an inventory entry which *isn't there*.
512
It simplifies the merging logic if we always have an InventoryEntry, even
513
if it isn't actually present
520
symlink_target = None
523
_none_entry = _InventoryNoneEntry()
369
526
class Merge3Merger(object):
372
529
supports_reprocess = True
373
530
supports_show_base = True
374
531
history_based = False
532
supports_cherrypick = True
533
supports_reverse_cherrypick = True
534
winner_idx = {"this": 2, "other": 1, "conflict": 1}
535
supports_lca_trees = True
376
537
def __init__(self, working_tree, this_tree, base_tree, other_tree,
377
538
interesting_ids=None, reprocess=False, show_base=False,
378
pb=DummyProgress(), pp=None, change_reporter=None):
379
"""Initialize the merger object and perform the merge."""
539
pb=DummyProgress(), pp=None, change_reporter=None,
540
interesting_files=None, do_merge=True,
541
cherrypick=False, lca_trees=None):
542
"""Initialize the merger object and perform the merge.
544
:param working_tree: The working tree to apply the merge to
545
:param this_tree: The local tree in the merge operation
546
:param base_tree: The common tree in the merge operation
547
:param other_tree: The other other tree to merge changes from
548
:param interesting_ids: The file_ids of files that should be
549
participate in the merge. May not be combined with
551
:param: reprocess If True, perform conflict-reduction processing.
552
:param show_base: If True, show the base revision in text conflicts.
553
(incompatible with reprocess)
554
:param pb: A Progress bar
555
:param pp: A ProgressPhase object
556
:param change_reporter: An object that should report changes made
557
:param interesting_files: The tree-relative paths of files that should
558
participate in the merge. If these paths refer to directories,
559
the contents of those directories will also be included. May not
560
be combined with interesting_ids. If neither interesting_files nor
561
interesting_ids is specified, all files may participate in the
563
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
564
if the ancestry was found to include a criss-cross merge.
565
Otherwise should be None.
380
567
object.__init__(self)
568
if interesting_files is not None and interesting_ids is not None:
570
'specify either interesting_ids or interesting_files')
571
self.interesting_ids = interesting_ids
572
self.interesting_files = interesting_files
381
573
self.this_tree = working_tree
382
self.this_tree.lock_tree_write()
383
574
self.base_tree = base_tree
384
self.base_tree.lock_read()
385
575
self.other_tree = other_tree
386
self.other_tree.lock_read()
387
576
self._raw_conflicts = []
388
577
self.cooked_conflicts = []
389
578
self.reprocess = reprocess
390
579
self.show_base = show_base
580
self._lca_trees = lca_trees
581
# Uncommenting this will change the default algorithm to always use
582
# _entries_lca. This can be useful for running the test suite and
583
# making sure we haven't missed any corner cases.
584
# if lca_trees is None:
585
# self._lca_trees = [self.base_tree]
393
588
self.change_reporter = change_reporter
589
self.cherrypick = cherrypick
394
590
if self.pp is None:
395
591
self.pp = ProgressPhase("Merge phase", 3, self.pb)
397
if interesting_ids is not None:
398
all_ids = interesting_ids
400
all_ids = set(base_tree)
401
all_ids.update(other_tree)
402
self.tt = TreeTransform(working_tree, self.pb)
596
self.this_tree.lock_tree_write()
597
self.base_tree.lock_read()
598
self.other_tree.lock_read()
599
self.tt = TreeTransform(self.this_tree, self.pb)
404
601
self.pp.next_phase()
405
child_pb = ui.ui_factory.nested_progress_bar()
407
for num, file_id in enumerate(all_ids):
408
child_pb.update('Preparing file merge', num, len(all_ids))
409
self.merge_names(file_id)
410
file_status = self.merge_contents(file_id)
411
self.merge_executable(file_id, file_status)
416
child_pb = ui.ui_factory.nested_progress_bar()
418
fs_conflicts = resolve_conflicts(self.tt, child_pb)
421
if change_reporter is not None:
422
from bzrlib import delta
423
delta.report_changes(self.tt._iter_changes(), change_reporter)
424
self.cook_conflicts(fs_conflicts)
425
for conflict in self.cooked_conflicts:
428
results = self.tt.apply()
602
self._compute_transform()
604
results = self.tt.apply(no_conflicts=True)
429
605
self.write_modified(results)
431
working_tree.add_conflicts(self.cooked_conflicts)
607
self.this_tree.add_conflicts(self.cooked_conflicts)
432
608
except UnsupportedOperation:
438
614
self.this_tree.unlock()
617
def make_preview_transform(self):
618
self.base_tree.lock_read()
619
self.other_tree.lock_read()
620
self.tt = TransformPreview(self.this_tree)
623
self._compute_transform()
626
self.other_tree.unlock()
627
self.base_tree.unlock()
631
def _compute_transform(self):
632
if self._lca_trees is None:
633
entries = self._entries3()
634
resolver = self._three_way
636
entries = self._entries_lca()
637
resolver = self._lca_multi_way
638
child_pb = ui.ui_factory.nested_progress_bar()
640
for num, (file_id, changed, parents3, names3,
641
executable3) in enumerate(entries):
642
child_pb.update('Preparing file merge', num, len(entries))
643
self._merge_names(file_id, parents3, names3, resolver=resolver)
645
file_status = self.merge_contents(file_id)
647
file_status = 'unmodified'
648
self._merge_executable(file_id,
649
executable3, file_status, resolver=resolver)
654
child_pb = ui.ui_factory.nested_progress_bar()
656
fs_conflicts = resolve_conflicts(self.tt, child_pb,
657
lambda t, c: conflict_pass(t, c, self.other_tree))
660
if self.change_reporter is not None:
661
from bzrlib import delta
662
delta.report_changes(
663
self.tt.iter_changes(), self.change_reporter)
664
self.cook_conflicts(fs_conflicts)
665
for conflict in self.cooked_conflicts:
669
"""Gather data about files modified between three trees.
671
Return a list of tuples of file_id, changed, parents3, names3,
672
executable3. changed is a boolean indicating whether the file contents
673
or kind were changed. parents3 is a tuple of parent ids for base,
674
other and this. names3 is a tuple of names for base, other and this.
675
executable3 is a tuple of execute-bit values for base, other and this.
678
iterator = self.other_tree.iter_changes(self.base_tree,
679
include_unchanged=True, specific_files=self.interesting_files,
680
extra_trees=[self.this_tree])
681
this_entries = dict((e.file_id, e) for p, e in
682
self.this_tree.iter_entries_by_dir(
683
self.interesting_ids))
684
for (file_id, paths, changed, versioned, parents, names, kind,
685
executable) in iterator:
686
if (self.interesting_ids is not None and
687
file_id not in self.interesting_ids):
689
entry = this_entries.get(file_id)
690
if entry is not None:
691
this_name = entry.name
692
this_parent = entry.parent_id
693
this_executable = entry.executable
697
this_executable = None
698
parents3 = parents + (this_parent,)
699
names3 = names + (this_name,)
700
executable3 = executable + (this_executable,)
701
result.append((file_id, changed, parents3, names3, executable3))
704
def _entries_lca(self):
705
"""Gather data about files modified between multiple trees.
707
This compares OTHER versus all LCA trees, and for interesting entries,
708
it then compares with THIS and BASE.
710
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
711
:return: [(file_id, changed, parents, names, executable)]
712
file_id Simple file_id of the entry
713
changed Boolean, True if the kind or contents changed
715
parents ((base, [parent_id, in, lcas]), parent_id_other,
717
names ((base, [name, in, lcas]), name_in_other, name_in_this)
718
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
720
if self.interesting_files is not None:
721
lookup_trees = [self.this_tree, self.base_tree]
722
lookup_trees.extend(self._lca_trees)
723
# I think we should include the lca trees as well
724
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
727
interesting_ids = self.interesting_ids
729
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
731
base_inventory = self.base_tree.inventory
732
this_inventory = self.this_tree.inventory
733
for path, file_id, other_ie, lca_values in walker.iter_all():
734
# Is this modified at all from any of the other trees?
736
other_ie = _none_entry
737
if interesting_ids is not None and file_id not in interesting_ids:
740
# If other_revision is found in any of the lcas, that means this
741
# node is uninteresting. This is because when merging, if there are
742
# multiple heads(), we have to create a new node. So if we didn't,
743
# we know that the ancestry is linear, and that OTHER did not
745
# See doc/developers/lca_merge_resolution.txt for details
746
other_revision = other_ie.revision
747
if other_revision is not None:
748
# We can't use this shortcut when other_revision is None,
749
# because it may be None because things are WorkingTrees, and
750
# not because it is *actually* None.
751
is_unmodified = False
752
for lca_path, ie in lca_values:
753
if ie is not None and ie.revision == other_revision:
760
for lca_path, lca_ie in lca_values:
762
lca_entries.append(_none_entry)
764
lca_entries.append(lca_ie)
766
if file_id in base_inventory:
767
base_ie = base_inventory[file_id]
769
base_ie = _none_entry
771
if file_id in this_inventory:
772
this_ie = this_inventory[file_id]
774
this_ie = _none_entry
780
for lca_ie in lca_entries:
781
lca_kinds.append(lca_ie.kind)
782
lca_parent_ids.append(lca_ie.parent_id)
783
lca_names.append(lca_ie.name)
784
lca_executable.append(lca_ie.executable)
786
kind_winner = self._lca_multi_way(
787
(base_ie.kind, lca_kinds),
788
other_ie.kind, this_ie.kind)
789
parent_id_winner = self._lca_multi_way(
790
(base_ie.parent_id, lca_parent_ids),
791
other_ie.parent_id, this_ie.parent_id)
792
name_winner = self._lca_multi_way(
793
(base_ie.name, lca_names),
794
other_ie.name, this_ie.name)
796
content_changed = True
797
if kind_winner == 'this':
798
# No kind change in OTHER, see if there are *any* changes
799
if other_ie.kind == 'directory':
800
if parent_id_winner == 'this' and name_winner == 'this':
801
# No change for this directory in OTHER, skip
803
content_changed = False
804
elif other_ie.kind is None or other_ie.kind == 'file':
805
def get_sha1(ie, tree):
806
if ie.kind != 'file':
808
return tree.get_file_sha1(file_id)
809
base_sha1 = get_sha1(base_ie, self.base_tree)
810
lca_sha1s = [get_sha1(ie, tree) for ie, tree
811
in zip(lca_entries, self._lca_trees)]
812
this_sha1 = get_sha1(this_ie, self.this_tree)
813
other_sha1 = get_sha1(other_ie, self.other_tree)
814
sha1_winner = self._lca_multi_way(
815
(base_sha1, lca_sha1s), other_sha1, this_sha1,
816
allow_overriding_lca=False)
817
exec_winner = self._lca_multi_way(
818
(base_ie.executable, lca_executable),
819
other_ie.executable, this_ie.executable)
820
if (parent_id_winner == 'this' and name_winner == 'this'
821
and sha1_winner == 'this' and exec_winner == 'this'):
822
# No kind, parent, name, exec, or content change for
823
# OTHER, so this node is not considered interesting
825
if sha1_winner == 'this':
826
content_changed = False
827
elif other_ie.kind == 'symlink':
828
def get_target(ie, tree):
829
if ie.kind != 'symlink':
831
return tree.get_symlink_target(file_id)
832
base_target = get_target(base_ie, self.base_tree)
833
lca_targets = [get_target(ie, tree) for ie, tree
834
in zip(lca_entries, self._lca_trees)]
835
this_target = get_target(this_ie, self.this_tree)
836
other_target = get_target(other_ie, self.other_tree)
837
target_winner = self._lca_multi_way(
838
(base_target, lca_targets),
839
other_target, this_target)
840
if (parent_id_winner == 'this' and name_winner == 'this'
841
and target_winner == 'this'):
842
# No kind, parent, name, or symlink target change
845
if target_winner == 'this':
846
content_changed = False
847
elif other_ie.kind == 'tree-reference':
848
# The 'changed' information seems to be handled at a higher
849
# level. At least, _entries3 returns False for content
850
# changed, even when at a new revision_id.
851
content_changed = False
852
if (parent_id_winner == 'this' and name_winner == 'this'):
853
# Nothing interesting
856
raise AssertionError('unhandled kind: %s' % other_ie.kind)
857
# XXX: We need to handle kind == 'symlink'
859
# If we have gotten this far, that means something has changed
860
result.append((file_id, content_changed,
861
((base_ie.parent_id, lca_parent_ids),
862
other_ie.parent_id, this_ie.parent_id),
863
((base_ie.name, lca_names),
864
other_ie.name, this_ie.name),
865
((base_ie.executable, lca_executable),
866
other_ie.executable, this_ie.executable)
441
871
def fix_root(self):
443
873
self.tt.final_kind(self.tt.root)
998
1515
from bzrlib import option
999
1516
return option._merge_type_registry
1519
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1520
def status_a(revision, text):
1521
if revision in ancestors_b:
1522
return 'killed-b', text
1524
return 'new-a', text
1526
def status_b(revision, text):
1527
if revision in ancestors_a:
1528
return 'killed-a', text
1530
return 'new-b', text
1532
plain_a = [t for (a, t) in annotated_a]
1533
plain_b = [t for (a, t) in annotated_b]
1534
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1535
blocks = matcher.get_matching_blocks()
1538
for ai, bi, l in blocks:
1539
# process all mismatched sections
1540
# (last mismatched section is handled because blocks always
1541
# includes a 0-length last block)
1542
for revision, text in annotated_a[a_cur:ai]:
1543
yield status_a(revision, text)
1544
for revision, text in annotated_b[b_cur:bi]:
1545
yield status_b(revision, text)
1546
# and now the matched section
1549
for text_a in plain_a[ai:a_cur]:
1550
yield "unchanged", text_a
1553
class _PlanMergeBase(object):
1555
def __init__(self, a_rev, b_rev, vf, key_prefix):
1558
:param a_rev: Revision-id of one revision to merge
1559
:param b_rev: Revision-id of the other revision to merge
1560
:param vf: A VersionedFiles containing both revisions
1561
:param key_prefix: A prefix for accessing keys in vf, typically
1567
self._last_lines = None
1568
self._last_lines_revision_id = None
1569
self._cached_matching_blocks = {}
1570
self._key_prefix = key_prefix
1571
self._precache_tip_lines()
1573
def _precache_tip_lines(self):
1574
lines = self.get_lines([self.a_rev, self.b_rev])
1575
self.lines_a = lines[self.a_rev]
1576
self.lines_b = lines[self.b_rev]
1578
def get_lines(self, revisions):
1579
"""Get lines for revisions from the backing VersionedFiles.
1581
:raises RevisionNotPresent: on absent texts.
1583
keys = [(self._key_prefix + (rev,)) for rev in revisions]
1585
for record in self.vf.get_record_stream(keys, 'unordered', True):
1586
if record.storage_kind == 'absent':
1587
raise errors.RevisionNotPresent(record.key, self.vf)
1588
result[record.key[-1]] = osutils.chunks_to_lines(
1589
record.get_bytes_as('chunked'))
1592
def plan_merge(self):
1593
"""Generate a 'plan' for merging the two revisions.
1595
This involves comparing their texts and determining the cause of
1596
differences. If text A has a line and text B does not, then either the
1597
line was added to text A, or it was deleted from B. Once the causes
1598
are combined, they are written out in the format described in
1599
VersionedFile.plan_merge
1601
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
1602
unique_a, unique_b = self._unique_lines(blocks)
1603
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
1604
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
1605
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
1607
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
1610
for i, j, n in blocks:
1611
for a_index in range(last_i, i):
1612
if a_index in new_a:
1613
if a_index in killed_b:
1614
yield 'conflicted-a', self.lines_a[a_index]
1616
yield 'new-a', self.lines_a[a_index]
1618
yield 'killed-b', self.lines_a[a_index]
1619
for b_index in range(last_j, j):
1620
if b_index in new_b:
1621
if b_index in killed_a:
1622
yield 'conflicted-b', self.lines_b[b_index]
1624
yield 'new-b', self.lines_b[b_index]
1626
yield 'killed-a', self.lines_b[b_index]
1627
# handle common lines
1628
for a_index in range(i, i+n):
1629
yield 'unchanged', self.lines_a[a_index]
1633
def _get_matching_blocks(self, left_revision, right_revision):
1634
"""Return a description of which sections of two revisions match.
1636
See SequenceMatcher.get_matching_blocks
1638
cached = self._cached_matching_blocks.get((left_revision,
1640
if cached is not None:
1642
if self._last_lines_revision_id == left_revision:
1643
left_lines = self._last_lines
1644
right_lines = self.get_lines([right_revision])[right_revision]
1646
lines = self.get_lines([left_revision, right_revision])
1647
left_lines = lines[left_revision]
1648
right_lines = lines[right_revision]
1649
self._last_lines = right_lines
1650
self._last_lines_revision_id = right_revision
1651
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
1653
return matcher.get_matching_blocks()
1655
def _unique_lines(self, matching_blocks):
1656
"""Analyse matching_blocks to determine which lines are unique
1658
:return: a tuple of (unique_left, unique_right), where the values are
1659
sets of line numbers of unique lines.
1665
for i, j, n in matching_blocks:
1666
unique_left.extend(range(last_i, i))
1667
unique_right.extend(range(last_j, j))
1670
return unique_left, unique_right
1673
def _subtract_plans(old_plan, new_plan):
1674
"""Remove changes from new_plan that came from old_plan.
1676
It is assumed that the difference between the old_plan and new_plan
1677
is their choice of 'b' text.
1679
All lines from new_plan that differ from old_plan are emitted
1680
verbatim. All lines from new_plan that match old_plan but are
1681
not about the 'b' revision are emitted verbatim.
1683
Lines that match and are about the 'b' revision are the lines we
1684
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
1685
is skipped entirely.
1687
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
1690
for i, j, n in matcher.get_matching_blocks():
1691
for jj in range(last_j, j):
1693
for jj in range(j, j+n):
1694
plan_line = new_plan[jj]
1695
if plan_line[0] == 'new-b':
1697
elif plan_line[0] == 'killed-b':
1698
yield 'unchanged', plan_line[1]
1704
class _PlanMerge(_PlanMergeBase):
1705
"""Plan an annotate merge using on-the-fly annotation"""
1707
def __init__(self, a_rev, b_rev, vf, key_prefix):
1708
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
1709
self.a_key = self._key_prefix + (self.a_rev,)
1710
self.b_key = self._key_prefix + (self.b_rev,)
1711
self.graph = Graph(self.vf)
1712
heads = self.graph.heads((self.a_key, self.b_key))
1714
# one side dominates, so we can just return its values, yay for
1716
# Ideally we would know that before we get this far
1717
self._head_key = heads.pop()
1718
if self._head_key == self.a_key:
1722
mutter('found dominating revision for %s\n%s > %s', self.vf,
1723
self._head_key[-1], other)
1726
self._head_key = None
1729
def _precache_tip_lines(self):
1730
# Turn this into a no-op, because we will do this later
1733
def _find_recursive_lcas(self):
1734
"""Find all the ancestors back to a unique lca"""
1735
cur_ancestors = (self.a_key, self.b_key)
1736
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
1737
# rather than a key tuple. We will just map that directly to no common
1741
next_lcas = self.graph.find_lca(*cur_ancestors)
1742
# Map a plain NULL_REVISION to a simple no-ancestors
1743
if next_lcas == set([NULL_REVISION]):
1745
# Order the lca's based on when they were merged into the tip
1746
# While the actual merge portion of weave merge uses a set() of
1747
# active revisions, the order of insertion *does* effect the
1748
# implicit ordering of the texts.
1749
for rev_key in cur_ancestors:
1750
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
1752
parent_map[rev_key] = ordered_parents
1753
if len(next_lcas) == 0:
1755
elif len(next_lcas) == 1:
1756
parent_map[list(next_lcas)[0]] = ()
1758
elif len(next_lcas) > 2:
1759
# More than 2 lca's, fall back to grabbing all nodes between
1760
# this and the unique lca.
1761
mutter('More than 2 LCAs, falling back to all nodes for:'
1762
' %s, %s\n=> %s', self.a_key, self.b_key, cur_ancestors)
1763
cur_lcas = next_lcas
1764
while len(cur_lcas) > 1:
1765
cur_lcas = self.graph.find_lca(*cur_lcas)
1766
if len(cur_lcas) == 0:
1767
# No common base to find, use the full ancestry
1770
unique_lca = list(cur_lcas)[0]
1771
if unique_lca == NULL_REVISION:
1772
# find_lca will return a plain 'NULL_REVISION' rather
1773
# than a key tuple when there is no common ancestor, we
1774
# prefer to just use None, because it doesn't confuse
1775
# _get_interesting_texts()
1777
parent_map.update(self._find_unique_parents(next_lcas,
1780
cur_ancestors = next_lcas
1783
def _find_unique_parents(self, tip_keys, base_key):
1784
"""Find ancestors of tip that aren't ancestors of base.
1786
:param tip_keys: Nodes that are interesting
1787
:param base_key: Cull all ancestors of this node
1788
:return: The parent map for all revisions between tip_keys and
1789
base_key. base_key will be included. References to nodes outside of
1790
the ancestor set will also be removed.
1792
# TODO: this would be simpler if find_unique_ancestors took a list
1793
# instead of a single tip, internally it supports it, but it
1794
# isn't a "backwards compatible" api change.
1795
if base_key is None:
1796
parent_map = dict(self.graph.iter_ancestry(tip_keys))
1797
# We remove NULL_REVISION because it isn't a proper tuple key, and
1798
# thus confuses things like _get_interesting_texts, and our logic
1799
# to add the texts into the memory weave.
1800
if NULL_REVISION in parent_map:
1801
parent_map.pop(NULL_REVISION)
1804
for tip in tip_keys:
1806
self.graph.find_unique_ancestors(tip, [base_key]))
1807
parent_map = self.graph.get_parent_map(interesting)
1808
parent_map[base_key] = ()
1809
culled_parent_map, child_map, tails = self._remove_external_references(
1811
# Remove all the tails but base_key
1812
if base_key is not None:
1813
tails.remove(base_key)
1814
self._prune_tails(culled_parent_map, child_map, tails)
1815
# Now remove all the uninteresting 'linear' regions
1816
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
1820
def _remove_external_references(parent_map):
1821
"""Remove references that go outside of the parent map.
1823
:param parent_map: Something returned from Graph.get_parent_map(keys)
1824
:return: (filtered_parent_map, child_map, tails)
1825
filtered_parent_map is parent_map without external references
1826
child_map is the {parent_key: [child_keys]} mapping
1827
tails is a list of nodes that do not have any parents in the map
1829
# TODO: The basic effect of this function seems more generic than
1830
# _PlanMerge. But the specific details of building a child_map,
1831
# and computing tails seems very specific to _PlanMerge.
1832
# Still, should this be in Graph land?
1833
filtered_parent_map = {}
1836
for key, parent_keys in parent_map.iteritems():
1837
culled_parent_keys = [p for p in parent_keys if p in parent_map]
1838
if not culled_parent_keys:
1840
for parent_key in culled_parent_keys:
1841
child_map.setdefault(parent_key, []).append(key)
1842
# TODO: Do we want to do this, it adds overhead for every node,
1843
# just to say that the node has no children
1844
child_map.setdefault(key, [])
1845
filtered_parent_map[key] = culled_parent_keys
1846
return filtered_parent_map, child_map, tails
1849
def _prune_tails(parent_map, child_map, tails_to_remove):
1850
"""Remove tails from the parent map.
1852
This will remove the supplied revisions until no more children have 0
1855
:param parent_map: A dict of {child: [parents]}, this dictionary will
1856
be modified in place.
1857
:param tails_to_remove: A list of tips that should be removed,
1858
this list will be consumed
1859
:param child_map: The reverse dict of parent_map ({parent: [children]})
1860
this dict will be modified
1861
:return: None, parent_map will be modified in place.
1863
while tails_to_remove:
1864
next = tails_to_remove.pop()
1865
parent_map.pop(next)
1866
children = child_map.pop(next)
1867
for child in children:
1868
child_parents = parent_map[child]
1869
child_parents.remove(next)
1870
if len(child_parents) == 0:
1871
tails_to_remove.append(child)
1873
def _get_interesting_texts(self, parent_map):
1874
"""Return a dict of texts we are interested in.
1876
Note that the input is in key tuples, but the output is in plain
1879
:param parent_map: The output from _find_recursive_lcas
1880
:return: A dict of {'revision_id':lines} as returned by
1881
_PlanMergeBase.get_lines()
1883
all_revision_keys = set(parent_map)
1884
all_revision_keys.add(self.a_key)
1885
all_revision_keys.add(self.b_key)
1887
# Everything else is in 'keys' but get_lines is in 'revision_ids'
1888
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
1891
def _build_weave(self):
1892
from bzrlib import weave
1893
self._weave = weave.Weave(weave_name='in_memory_weave',
1894
allow_reserved=True)
1895
parent_map = self._find_recursive_lcas()
1897
all_texts = self._get_interesting_texts(parent_map)
1899
# Note: Unfortunately, the order given by topo_sort will effect the
1900
# ordering resolution in the output. Specifically, if you add A then B,
1901
# then in the output text A lines will show up before B lines. And, of
1902
# course, topo_sort doesn't guarantee any real ordering.
1903
# So we use merge_sort, and add a fake node on the tip.
1904
# This ensures that left-hand parents will always be inserted into the
1905
# weave before right-hand parents.
1906
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
1907
parent_map[tip_key] = (self.a_key, self.b_key)
1909
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
1913
# for key in tsort.topo_sort(parent_map):
1914
parent_keys = parent_map[key]
1915
revision_id = key[-1]
1916
parent_ids = [k[-1] for k in parent_keys]
1917
self._weave.add_lines(revision_id, parent_ids,
1918
all_texts[revision_id])
1920
def plan_merge(self):
1921
"""Generate a 'plan' for merging the two revisions.
1923
This involves comparing their texts and determining the cause of
1924
differences. If text A has a line and text B does not, then either the
1925
line was added to text A, or it was deleted from B. Once the causes
1926
are combined, they are written out in the format described in
1927
VersionedFile.plan_merge
1929
if self._head_key is not None: # There was a single head
1930
if self._head_key == self.a_key:
1933
if self._head_key != self.b_key:
1934
raise AssertionError('There was an invalid head: %s != %s'
1935
% (self.b_key, self._head_key))
1937
head_rev = self._head_key[-1]
1938
lines = self.get_lines([head_rev])[head_rev]
1939
return ((plan, line) for line in lines)
1940
return self._weave.plan_merge(self.a_rev, self.b_rev)
1943
class _PlanLCAMerge(_PlanMergeBase):
1945
This merge algorithm differs from _PlanMerge in that:
1946
1. comparisons are done against LCAs only
1947
2. cases where a contested line is new versus one LCA but old versus
1948
another are marked as conflicts, by emitting the line as conflicted-a
1951
This is faster, and hopefully produces more useful output.
1954
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
1955
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
1956
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
1959
if lca == NULL_REVISION:
1962
self.lcas.add(lca[-1])
1963
for lca in self.lcas:
1964
if _mod_revision.is_null(lca):
1967
lca_lines = self.get_lines([lca])[lca]
1968
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
1970
blocks = list(matcher.get_matching_blocks())
1971
self._cached_matching_blocks[(a_rev, lca)] = blocks
1972
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
1974
blocks = list(matcher.get_matching_blocks())
1975
self._cached_matching_blocks[(b_rev, lca)] = blocks
1977
def _determine_status(self, revision_id, unique_line_numbers):
1978
"""Determines the status unique lines versus all lcas.
1980
Basically, determines why the line is unique to this revision.
1982
A line may be determined new, killed, or both.
1984
If a line is determined new, that means it was not present in at least
1985
one LCA, and is not present in the other merge revision.
1987
If a line is determined killed, that means the line was present in
1990
If a line is killed and new, this indicates that the two merge
1991
revisions contain differing conflict resolutions.
1992
:param revision_id: The id of the revision in which the lines are
1994
:param unique_line_numbers: The line numbers of unique lines.
1995
:return a tuple of (new_this, killed_other):
1999
unique_line_numbers = set(unique_line_numbers)
2000
for lca in self.lcas:
2001
blocks = self._get_matching_blocks(revision_id, lca)
2002
unique_vs_lca, _ignored = self._unique_lines(blocks)
2003
new.update(unique_line_numbers.intersection(unique_vs_lca))
2004
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
