1
# Copyright (C) 2005, 2006, 2008 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
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
19
from itertools import chain
30
revision as _mod_revision,
34
from bzrlib.branch import Branch
35
from bzrlib.conflicts import ConflictList, Conflict
36
from bzrlib.errors import (BzrCommandError,
46
WorkingTreeNotRevision,
49
from bzrlib.graph import Graph
50
from bzrlib.merge3 import Merge3
51
from bzrlib.osutils import rename, pathjoin
52
from progress import DummyProgress, ProgressPhase
53
from bzrlib.revision import (NULL_REVISION, ensure_null)
54
from bzrlib.textfile import check_text_lines
55
from bzrlib.trace import mutter, warning, note, is_quiet
56
from bzrlib.transform import (TransformPreview, TreeTransform,
57
resolve_conflicts, cook_conflicts,
58
conflict_pass, FinalPaths, create_from_tree,
59
unique_add, ROOT_PARENT)
60
from bzrlib.versionedfile import PlanWeaveMerge
63
# TODO: Report back as changes are merged in
66
def transform_tree(from_tree, to_tree, interesting_ids=None):
67
from_tree.lock_tree_write()
69
merge_inner(from_tree.branch, to_tree, from_tree, ignore_zero=True,
70
interesting_ids=interesting_ids, this_tree=from_tree)
76
def __init__(self, this_branch, other_tree=None, base_tree=None,
77
this_tree=None, pb=None, change_reporter=None,
78
recurse='down', revision_graph=None):
80
self.this_branch = this_branch
81
self.this_basis = _mod_revision.ensure_null(
82
this_branch.last_revision())
83
self.this_rev_id = None
84
self.this_tree = this_tree
85
self.this_revision_tree = None
86
self.this_basis_tree = None
87
self.other_tree = other_tree
88
self.other_branch = None
89
self.base_tree = base_tree
90
self.ignore_zero = False
91
self.backup_files = False
92
self.interesting_ids = None
93
self.interesting_files = None
94
self.show_base = False
95
self.reprocess = False
100
self.recurse = recurse
101
self.change_reporter = change_reporter
102
self._cached_trees = {}
103
self._revision_graph = revision_graph
104
self._base_is_ancestor = None
105
self._base_is_other_ancestor = None
106
self._is_criss_cross = None
107
self._lca_trees = None
109
def cache_trees_with_revision_ids(self, trees):
110
"""Cache any tree in trees if it has a revision_id."""
111
for maybe_tree in trees:
112
if maybe_tree is None:
115
rev_id = maybe_tree.get_revision_id()
116
except AttributeError:
118
self._cached_trees[rev_id] = maybe_tree
121
def revision_graph(self):
122
if self._revision_graph is None:
123
self._revision_graph = self.this_branch.repository.get_graph()
124
return self._revision_graph
126
def _set_base_is_ancestor(self, value):
127
self._base_is_ancestor = value
129
def _get_base_is_ancestor(self):
130
if self._base_is_ancestor is None:
131
self._base_is_ancestor = self.revision_graph.is_ancestor(
132
self.base_rev_id, self.this_basis)
133
return self._base_is_ancestor
135
base_is_ancestor = property(_get_base_is_ancestor, _set_base_is_ancestor)
137
def _set_base_is_other_ancestor(self, value):
138
self._base_is_other_ancestor = value
140
def _get_base_is_other_ancestor(self):
141
if self._base_is_other_ancestor is None:
142
if self.other_basis is None:
144
self._base_is_other_ancestor = self.revision_graph.is_ancestor(
145
self.base_rev_id, self.other_basis)
146
return self._base_is_other_ancestor
148
base_is_other_ancestor = property(_get_base_is_other_ancestor,
149
_set_base_is_other_ancestor)
152
def from_uncommitted(tree, other_tree, pb=None, base_tree=None):
153
"""Return a Merger for uncommitted changes in other_tree.
155
:param tree: The tree to merge into
156
:param other_tree: The tree to get uncommitted changes from
157
:param pb: A progress indicator
158
:param base_tree: The basis to use for the merge. If unspecified,
159
other_tree.basis_tree() will be used.
161
if base_tree is None:
162
base_tree = other_tree.basis_tree()
163
merger = Merger(tree.branch, other_tree, base_tree, tree, pb)
164
merger.base_rev_id = merger.base_tree.get_revision_id()
165
merger.other_rev_id = None
166
merger.other_basis = merger.base_rev_id
170
def from_mergeable(klass, tree, mergeable, pb):
171
"""Return a Merger for a bundle or merge directive.
173
:param tree: The tree to merge changes into
174
:param mergeable: A merge directive or bundle
175
:param pb: A progress indicator
177
mergeable.install_revisions(tree.branch.repository)
178
base_revision_id, other_revision_id, verified =\
179
mergeable.get_merge_request(tree.branch.repository)
180
revision_graph = tree.branch.repository.get_graph()
181
if base_revision_id is not None:
182
if (base_revision_id != _mod_revision.NULL_REVISION and
183
revision_graph.is_ancestor(
184
base_revision_id, tree.branch.last_revision())):
185
base_revision_id = None
187
warning('Performing cherrypick')
188
merger = klass.from_revision_ids(pb, tree, other_revision_id,
189
base_revision_id, revision_graph=
191
return merger, verified
194
def from_revision_ids(pb, tree, other, base=None, other_branch=None,
195
base_branch=None, revision_graph=None,
197
"""Return a Merger for revision-ids.
199
:param pb: A progress indicator
200
:param tree: The tree to merge changes into
201
:param other: The revision-id to use as OTHER
202
:param base: The revision-id to use as BASE. If not specified, will
204
:param other_branch: A branch containing the other revision-id. If
205
not supplied, tree.branch is used.
206
:param base_branch: A branch containing the base revision-id. If
207
not supplied, other_branch or tree.branch will be used.
208
:param revision_graph: If you have a revision_graph precomputed, pass
209
it in, otherwise it will be created for you.
210
:param tree_branch: The branch associated with tree. If not supplied,
211
tree.branch will be used.
213
if tree_branch is None:
214
tree_branch = tree.branch
215
merger = Merger(tree_branch, this_tree=tree, pb=pb,
216
revision_graph=revision_graph)
217
if other_branch is None:
218
other_branch = tree.branch
219
merger.set_other_revision(other, other_branch)
223
if base_branch is None:
224
base_branch = other_branch
225
merger.set_base_revision(base, base_branch)
228
def revision_tree(self, revision_id, branch=None):
229
if revision_id not in self._cached_trees:
231
branch = self.this_branch
233
tree = self.this_tree.revision_tree(revision_id)
234
except errors.NoSuchRevisionInTree:
235
tree = branch.repository.revision_tree(revision_id)
236
self._cached_trees[revision_id] = tree
237
return self._cached_trees[revision_id]
239
def _get_tree(self, treespec, possible_transports=None):
240
from bzrlib import workingtree
241
location, revno = treespec
243
tree = workingtree.WorkingTree.open_containing(location)[0]
244
return tree.branch, tree
245
branch = Branch.open_containing(location, possible_transports)[0]
247
revision_id = branch.last_revision()
249
revision_id = branch.get_rev_id(revno)
250
revision_id = ensure_null(revision_id)
251
return branch, self.revision_tree(revision_id, branch)
253
def ensure_revision_trees(self):
254
if self.this_revision_tree is None:
255
self.this_basis_tree = self.revision_tree(self.this_basis)
256
if self.this_basis == self.this_rev_id:
257
self.this_revision_tree = self.this_basis_tree
259
if self.other_rev_id is None:
260
other_basis_tree = self.revision_tree(self.other_basis)
261
if other_basis_tree.has_changes(self.other_tree):
262
raise WorkingTreeNotRevision(self.this_tree)
263
other_rev_id = self.other_basis
264
self.other_tree = other_basis_tree
266
def file_revisions(self, file_id):
267
self.ensure_revision_trees()
268
def get_id(tree, file_id):
269
revision_id = tree.inventory[file_id].revision
271
if self.this_rev_id is None:
272
if self.this_basis_tree.get_file_sha1(file_id) != \
273
self.this_tree.get_file_sha1(file_id):
274
raise WorkingTreeNotRevision(self.this_tree)
276
trees = (self.this_basis_tree, self.other_tree)
277
return [get_id(tree, file_id) for tree in trees]
279
def check_basis(self, check_clean, require_commits=True):
280
if self.this_basis is None and require_commits is True:
281
raise BzrCommandError("This branch has no commits."
282
" (perhaps you would prefer 'bzr pull')")
285
if self.this_basis != self.this_rev_id:
286
raise errors.UncommittedChanges(self.this_tree)
288
def compare_basis(self):
290
basis_tree = self.revision_tree(self.this_tree.last_revision())
291
except errors.NoSuchRevision:
292
basis_tree = self.this_tree.basis_tree()
293
if not self.this_tree.has_changes(basis_tree):
294
self.this_rev_id = self.this_basis
296
def set_interesting_files(self, file_list):
297
self.interesting_files = file_list
299
def set_pending(self):
300
if not self.base_is_ancestor or not self.base_is_other_ancestor or self.other_rev_id is None:
304
def _add_parent(self):
305
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
306
new_parent_trees = []
307
for revision_id in new_parents:
309
tree = self.revision_tree(revision_id)
310
except errors.NoSuchRevision:
314
new_parent_trees.append((revision_id, tree))
316
self.this_tree.set_parent_trees(new_parent_trees,
317
allow_leftmost_as_ghost=True)
319
for _revision_id, tree in new_parent_trees:
323
def set_other(self, other_revision, possible_transports=None):
324
"""Set the revision and tree to merge from.
326
This sets the other_tree, other_rev_id, other_basis attributes.
328
:param other_revision: The [path, revision] list to merge from.
330
self.other_branch, self.other_tree = self._get_tree(other_revision,
332
if other_revision[1] == -1:
333
self.other_rev_id = _mod_revision.ensure_null(
334
self.other_branch.last_revision())
335
if _mod_revision.is_null(self.other_rev_id):
336
raise NoCommits(self.other_branch)
337
self.other_basis = self.other_rev_id
338
elif other_revision[1] is not None:
339
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
340
self.other_basis = self.other_rev_id
342
self.other_rev_id = None
343
self.other_basis = self.other_branch.last_revision()
344
if self.other_basis is None:
345
raise NoCommits(self.other_branch)
346
if self.other_rev_id is not None:
347
self._cached_trees[self.other_rev_id] = self.other_tree
348
self._maybe_fetch(self.other_branch,self.this_branch, self.other_basis)
350
def set_other_revision(self, revision_id, other_branch):
351
"""Set 'other' based on a branch and revision id
353
:param revision_id: The revision to use for a tree
354
:param other_branch: The branch containing this tree
356
self.other_rev_id = revision_id
357
self.other_branch = other_branch
358
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
359
self.other_tree = self.revision_tree(revision_id)
360
self.other_basis = revision_id
362
def set_base_revision(self, revision_id, branch):
363
"""Set 'base' based on a branch and revision id
365
:param revision_id: The revision to use for a tree
366
:param branch: The branch containing this tree
368
self.base_rev_id = revision_id
369
self.base_branch = branch
370
self._maybe_fetch(branch, self.this_branch, revision_id)
371
self.base_tree = self.revision_tree(revision_id)
373
def _maybe_fetch(self, source, target, revision_id):
374
if not source.repository.has_same_location(target.repository):
375
target.fetch(source, revision_id)
378
revisions = [ensure_null(self.this_basis),
379
ensure_null(self.other_basis)]
380
if NULL_REVISION in revisions:
381
self.base_rev_id = NULL_REVISION
382
self.base_tree = self.revision_tree(self.base_rev_id)
383
self._is_criss_cross = False
385
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
386
self._is_criss_cross = False
388
self.base_rev_id = NULL_REVISION
390
self.base_rev_id = list(lcas)[0]
391
else: # len(lcas) > 1
393
# find_unique_lca can only handle 2 nodes, so we have to
394
# start back at the beginning. It is a shame to traverse
395
# the graph again, but better than re-implementing
397
self.base_rev_id = self.revision_graph.find_unique_lca(
398
revisions[0], revisions[1])
400
self.base_rev_id = self.revision_graph.find_unique_lca(
402
self._is_criss_cross = True
403
if self.base_rev_id == NULL_REVISION:
404
raise UnrelatedBranches()
405
if self._is_criss_cross:
406
warning('Warning: criss-cross merge encountered. See bzr'
407
' help criss-cross.')
408
mutter('Criss-cross lcas: %r' % lcas)
409
interesting_revision_ids = [self.base_rev_id]
410
interesting_revision_ids.extend(lcas)
411
interesting_trees = dict((t.get_revision_id(), t)
412
for t in self.this_branch.repository.revision_trees(
413
interesting_revision_ids))
414
self._cached_trees.update(interesting_trees)
415
self.base_tree = interesting_trees.pop(self.base_rev_id)
416
sorted_lca_keys = self.revision_graph.find_merge_order(
418
self._lca_trees = [interesting_trees[key]
419
for key in sorted_lca_keys]
421
self.base_tree = self.revision_tree(self.base_rev_id)
422
self.base_is_ancestor = True
423
self.base_is_other_ancestor = True
424
mutter('Base revid: %r' % self.base_rev_id)
426
def set_base(self, base_revision):
427
"""Set the base revision to use for the merge.
429
:param base_revision: A 2-list containing a path and revision number.
431
mutter("doing merge() with no base_revision specified")
432
if base_revision == [None, None]:
435
base_branch, self.base_tree = self._get_tree(base_revision)
436
if base_revision[1] == -1:
437
self.base_rev_id = base_branch.last_revision()
438
elif base_revision[1] is None:
439
self.base_rev_id = _mod_revision.NULL_REVISION
441
self.base_rev_id = _mod_revision.ensure_null(
442
base_branch.get_rev_id(base_revision[1]))
443
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
445
def make_merger(self):
446
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
447
'other_tree': self.other_tree,
448
'interesting_ids': self.interesting_ids,
449
'interesting_files': self.interesting_files,
452
if self.merge_type.requires_base:
453
kwargs['base_tree'] = self.base_tree
454
if self.merge_type.supports_reprocess:
455
kwargs['reprocess'] = self.reprocess
457
raise BzrError("Conflict reduction is not supported for merge"
458
" type %s." % self.merge_type)
459
if self.merge_type.supports_show_base:
460
kwargs['show_base'] = self.show_base
462
raise BzrError("Showing base is not supported for this"
463
" merge type. %s" % self.merge_type)
464
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
465
and not self.base_is_other_ancestor):
466
raise errors.CannotReverseCherrypick()
467
if self.merge_type.supports_cherrypick:
468
kwargs['cherrypick'] = (not self.base_is_ancestor or
469
not self.base_is_other_ancestor)
470
if self._is_criss_cross and getattr(self.merge_type,
471
'supports_lca_trees', False):
472
kwargs['lca_trees'] = self._lca_trees
473
return self.merge_type(pb=self._pb,
474
change_reporter=self.change_reporter,
477
def _do_merge_to(self, merge):
478
if self.other_branch is not None:
479
self.other_branch.update_references(self.this_branch)
481
if self.recurse == 'down':
482
for relpath, file_id in self.this_tree.iter_references():
483
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
484
other_revision = self.other_tree.get_reference_revision(
486
if other_revision == sub_tree.last_revision():
488
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
489
sub_merge.merge_type = self.merge_type
490
other_branch = self.other_branch.reference_parent(file_id, relpath)
491
sub_merge.set_other_revision(other_revision, other_branch)
492
base_revision = self.base_tree.get_reference_revision(file_id)
493
sub_merge.base_tree = \
494
sub_tree.branch.repository.revision_tree(base_revision)
495
sub_merge.base_rev_id = base_revision
499
self.this_tree.lock_tree_write()
501
if self.base_tree is not None:
502
self.base_tree.lock_read()
504
if self.other_tree is not None:
505
self.other_tree.lock_read()
507
merge = self.make_merger()
508
self._do_merge_to(merge)
510
if self.other_tree is not None:
511
self.other_tree.unlock()
513
if self.base_tree is not None:
514
self.base_tree.unlock()
516
self.this_tree.unlock()
517
if len(merge.cooked_conflicts) == 0:
518
if not self.ignore_zero and not is_quiet():
519
note("All changes applied successfully.")
521
note("%d conflicts encountered." % len(merge.cooked_conflicts))
523
return len(merge.cooked_conflicts)
526
class _InventoryNoneEntry(object):
527
"""This represents an inventory entry which *isn't there*.
529
It simplifies the merging logic if we always have an InventoryEntry, even
530
if it isn't actually present
537
symlink_target = None
540
_none_entry = _InventoryNoneEntry()
543
class Merge3Merger(object):
544
"""Three-way merger that uses the merge3 text merger"""
546
supports_reprocess = True
547
supports_show_base = True
548
history_based = False
549
supports_cherrypick = True
550
supports_reverse_cherrypick = True
551
winner_idx = {"this": 2, "other": 1, "conflict": 1}
552
supports_lca_trees = True
554
def __init__(self, working_tree, this_tree, base_tree, other_tree,
555
interesting_ids=None, reprocess=False, show_base=False,
556
pb=DummyProgress(), pp=None, change_reporter=None,
557
interesting_files=None, do_merge=True,
558
cherrypick=False, lca_trees=None):
559
"""Initialize the merger object and perform the merge.
561
:param working_tree: The working tree to apply the merge to
562
:param this_tree: The local tree in the merge operation
563
:param base_tree: The common tree in the merge operation
564
:param other_tree: The other tree to merge changes from
565
:param interesting_ids: The file_ids of files that should be
566
participate in the merge. May not be combined with
568
:param: reprocess If True, perform conflict-reduction processing.
569
:param show_base: If True, show the base revision in text conflicts.
570
(incompatible with reprocess)
571
:param pb: A Progress bar
572
:param pp: A ProgressPhase object
573
:param change_reporter: An object that should report changes made
574
:param interesting_files: The tree-relative paths of files that should
575
participate in the merge. If these paths refer to directories,
576
the contents of those directories will also be included. May not
577
be combined with interesting_ids. If neither interesting_files nor
578
interesting_ids is specified, all files may participate in the
580
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
581
if the ancestry was found to include a criss-cross merge.
582
Otherwise should be None.
584
object.__init__(self)
585
if interesting_files is not None and interesting_ids is not None:
587
'specify either interesting_ids or interesting_files')
588
self.interesting_ids = interesting_ids
589
self.interesting_files = interesting_files
590
self.this_tree = working_tree
591
self.base_tree = base_tree
592
self.other_tree = other_tree
593
self._raw_conflicts = []
594
self.cooked_conflicts = []
595
self.reprocess = reprocess
596
self.show_base = show_base
597
self._lca_trees = lca_trees
598
# Uncommenting this will change the default algorithm to always use
599
# _entries_lca. This can be useful for running the test suite and
600
# making sure we haven't missed any corner cases.
601
# if lca_trees is None:
602
# self._lca_trees = [self.base_tree]
605
self.change_reporter = change_reporter
606
self.cherrypick = cherrypick
608
self.pp = ProgressPhase("Merge phase", 3, self.pb)
613
self.this_tree.lock_tree_write()
614
self.base_tree.lock_read()
615
self.other_tree.lock_read()
617
self.tt = TreeTransform(self.this_tree, self.pb)
620
self._compute_transform()
622
results = self.tt.apply(no_conflicts=True)
623
self.write_modified(results)
625
self.this_tree.add_conflicts(self.cooked_conflicts)
626
except UnsupportedOperation:
631
self.other_tree.unlock()
632
self.base_tree.unlock()
633
self.this_tree.unlock()
636
def make_preview_transform(self):
637
self.base_tree.lock_read()
638
self.other_tree.lock_read()
639
self.tt = TransformPreview(self.this_tree)
642
self._compute_transform()
645
self.other_tree.unlock()
646
self.base_tree.unlock()
650
def _compute_transform(self):
651
if self._lca_trees is None:
652
entries = self._entries3()
653
resolver = self._three_way
655
entries = self._entries_lca()
656
resolver = self._lca_multi_way
657
child_pb = ui.ui_factory.nested_progress_bar()
659
for num, (file_id, changed, parents3, names3,
660
executable3) in enumerate(entries):
661
child_pb.update('Preparing file merge', num, len(entries))
662
self._merge_names(file_id, parents3, names3, resolver=resolver)
664
file_status = self.merge_contents(file_id)
666
file_status = 'unmodified'
667
self._merge_executable(file_id,
668
executable3, file_status, resolver=resolver)
673
child_pb = ui.ui_factory.nested_progress_bar()
675
fs_conflicts = resolve_conflicts(self.tt, child_pb,
676
lambda t, c: conflict_pass(t, c, self.other_tree))
679
if self.change_reporter is not None:
680
from bzrlib import delta
681
delta.report_changes(
682
self.tt.iter_changes(), self.change_reporter)
683
self.cook_conflicts(fs_conflicts)
684
for conflict in self.cooked_conflicts:
688
"""Gather data about files modified between three trees.
690
Return a list of tuples of file_id, changed, parents3, names3,
691
executable3. changed is a boolean indicating whether the file contents
692
or kind were changed. parents3 is a tuple of parent ids for base,
693
other and this. names3 is a tuple of names for base, other and this.
694
executable3 is a tuple of execute-bit values for base, other and this.
697
iterator = self.other_tree.iter_changes(self.base_tree,
698
include_unchanged=True, specific_files=self.interesting_files,
699
extra_trees=[self.this_tree])
700
this_entries = dict((e.file_id, e) for p, e in
701
self.this_tree.iter_entries_by_dir(
702
self.interesting_ids))
703
for (file_id, paths, changed, versioned, parents, names, kind,
704
executable) in iterator:
705
if (self.interesting_ids is not None and
706
file_id not in self.interesting_ids):
708
entry = this_entries.get(file_id)
709
if entry is not None:
710
this_name = entry.name
711
this_parent = entry.parent_id
712
this_executable = entry.executable
716
this_executable = None
717
parents3 = parents + (this_parent,)
718
names3 = names + (this_name,)
719
executable3 = executable + (this_executable,)
720
result.append((file_id, changed, parents3, names3, executable3))
723
def _entries_lca(self):
724
"""Gather data about files modified between multiple trees.
726
This compares OTHER versus all LCA trees, and for interesting entries,
727
it then compares with THIS and BASE.
729
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
730
:return: [(file_id, changed, parents, names, executable)]
731
file_id Simple file_id of the entry
732
changed Boolean, True if the kind or contents changed
734
parents ((base, [parent_id, in, lcas]), parent_id_other,
736
names ((base, [name, in, lcas]), name_in_other, name_in_this)
737
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
739
if self.interesting_files is not None:
740
lookup_trees = [self.this_tree, self.base_tree]
741
lookup_trees.extend(self._lca_trees)
742
# I think we should include the lca trees as well
743
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
746
interesting_ids = self.interesting_ids
748
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
750
base_inventory = self.base_tree.inventory
751
this_inventory = self.this_tree.inventory
752
for path, file_id, other_ie, lca_values in walker.iter_all():
753
# Is this modified at all from any of the other trees?
755
other_ie = _none_entry
756
if interesting_ids is not None and file_id not in interesting_ids:
759
# If other_revision is found in any of the lcas, that means this
760
# node is uninteresting. This is because when merging, if there are
761
# multiple heads(), we have to create a new node. So if we didn't,
762
# we know that the ancestry is linear, and that OTHER did not
764
# See doc/developers/lca_merge_resolution.txt for details
765
other_revision = other_ie.revision
766
if other_revision is not None:
767
# We can't use this shortcut when other_revision is None,
768
# because it may be None because things are WorkingTrees, and
769
# not because it is *actually* None.
770
is_unmodified = False
771
for lca_path, ie in lca_values:
772
if ie is not None and ie.revision == other_revision:
779
for lca_path, lca_ie in lca_values:
781
lca_entries.append(_none_entry)
783
lca_entries.append(lca_ie)
785
if file_id in base_inventory:
786
base_ie = base_inventory[file_id]
788
base_ie = _none_entry
790
if file_id in this_inventory:
791
this_ie = this_inventory[file_id]
793
this_ie = _none_entry
799
for lca_ie in lca_entries:
800
lca_kinds.append(lca_ie.kind)
801
lca_parent_ids.append(lca_ie.parent_id)
802
lca_names.append(lca_ie.name)
803
lca_executable.append(lca_ie.executable)
805
kind_winner = self._lca_multi_way(
806
(base_ie.kind, lca_kinds),
807
other_ie.kind, this_ie.kind)
808
parent_id_winner = self._lca_multi_way(
809
(base_ie.parent_id, lca_parent_ids),
810
other_ie.parent_id, this_ie.parent_id)
811
name_winner = self._lca_multi_way(
812
(base_ie.name, lca_names),
813
other_ie.name, this_ie.name)
815
content_changed = True
816
if kind_winner == 'this':
817
# No kind change in OTHER, see if there are *any* changes
818
if other_ie.kind == 'directory':
819
if parent_id_winner == 'this' and name_winner == 'this':
820
# No change for this directory in OTHER, skip
822
content_changed = False
823
elif other_ie.kind is None or other_ie.kind == 'file':
824
def get_sha1(ie, tree):
825
if ie.kind != 'file':
827
return tree.get_file_sha1(file_id)
828
base_sha1 = get_sha1(base_ie, self.base_tree)
829
lca_sha1s = [get_sha1(ie, tree) for ie, tree
830
in zip(lca_entries, self._lca_trees)]
831
this_sha1 = get_sha1(this_ie, self.this_tree)
832
other_sha1 = get_sha1(other_ie, self.other_tree)
833
sha1_winner = self._lca_multi_way(
834
(base_sha1, lca_sha1s), other_sha1, this_sha1,
835
allow_overriding_lca=False)
836
exec_winner = self._lca_multi_way(
837
(base_ie.executable, lca_executable),
838
other_ie.executable, this_ie.executable)
839
if (parent_id_winner == 'this' and name_winner == 'this'
840
and sha1_winner == 'this' and exec_winner == 'this'):
841
# No kind, parent, name, exec, or content change for
842
# OTHER, so this node is not considered interesting
844
if sha1_winner == 'this':
845
content_changed = False
846
elif other_ie.kind == 'symlink':
847
def get_target(ie, tree):
848
if ie.kind != 'symlink':
850
return tree.get_symlink_target(file_id)
851
base_target = get_target(base_ie, self.base_tree)
852
lca_targets = [get_target(ie, tree) for ie, tree
853
in zip(lca_entries, self._lca_trees)]
854
this_target = get_target(this_ie, self.this_tree)
855
other_target = get_target(other_ie, self.other_tree)
856
target_winner = self._lca_multi_way(
857
(base_target, lca_targets),
858
other_target, this_target)
859
if (parent_id_winner == 'this' and name_winner == 'this'
860
and target_winner == 'this'):
861
# No kind, parent, name, or symlink target change
864
if target_winner == 'this':
865
content_changed = False
866
elif other_ie.kind == 'tree-reference':
867
# The 'changed' information seems to be handled at a higher
868
# level. At least, _entries3 returns False for content
869
# changed, even when at a new revision_id.
870
content_changed = False
871
if (parent_id_winner == 'this' and name_winner == 'this'):
872
# Nothing interesting
875
raise AssertionError('unhandled kind: %s' % other_ie.kind)
876
# XXX: We need to handle kind == 'symlink'
878
# If we have gotten this far, that means something has changed
879
result.append((file_id, content_changed,
880
((base_ie.parent_id, lca_parent_ids),
881
other_ie.parent_id, this_ie.parent_id),
882
((base_ie.name, lca_names),
883
other_ie.name, this_ie.name),
884
((base_ie.executable, lca_executable),
885
other_ie.executable, this_ie.executable)
892
self.tt.final_kind(self.tt.root)
894
self.tt.cancel_deletion(self.tt.root)
895
if self.tt.final_file_id(self.tt.root) is None:
896
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
898
other_root_file_id = self.other_tree.get_root_id()
899
if other_root_file_id is None:
901
other_root = self.tt.trans_id_file_id(other_root_file_id)
902
if other_root == self.tt.root:
905
self.tt.final_kind(other_root)
908
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
909
# the other tree's root is a non-root in the current tree
911
self.reparent_children(self.other_tree.inventory.root, self.tt.root)
912
self.tt.cancel_creation(other_root)
913
self.tt.cancel_versioning(other_root)
915
def reparent_children(self, ie, target):
916
for thing, child in ie.children.iteritems():
917
trans_id = self.tt.trans_id_file_id(child.file_id)
918
self.tt.adjust_path(self.tt.final_name(trans_id), target, trans_id)
920
def write_modified(self, results):
922
for path in results.modified_paths:
923
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
926
hash = self.this_tree.get_file_sha1(file_id)
929
modified_hashes[file_id] = hash
930
self.this_tree.set_merge_modified(modified_hashes)
933
def parent(entry, file_id):
934
"""Determine the parent for a file_id (used as a key method)"""
937
return entry.parent_id
940
def name(entry, file_id):
941
"""Determine the name for a file_id (used as a key method)"""
947
def contents_sha1(tree, file_id):
948
"""Determine the sha1 of the file contents (used as a key method)."""
949
if file_id not in tree:
951
return tree.get_file_sha1(file_id)
954
def executable(tree, file_id):
955
"""Determine the executability of a file-id (used as a key method)."""
956
if file_id not in tree:
958
if tree.kind(file_id) != "file":
960
return tree.is_executable(file_id)
963
def kind(tree, file_id):
964
"""Determine the kind of a file-id (used as a key method)."""
965
if file_id not in tree:
967
return tree.kind(file_id)
970
def _three_way(base, other, this):
971
#if base == other, either they all agree, or only THIS has changed.
974
elif this not in (base, other):
976
# "Ambiguous clean merge" -- both sides have made the same change.
979
# this == base: only other has changed.
984
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
985
"""Consider LCAs when determining whether a change has occurred.
987
If LCAS are all identical, this is the same as a _three_way comparison.
989
:param bases: value in (BASE, [LCAS])
990
:param other: value in OTHER
991
:param this: value in THIS
992
:param allow_overriding_lca: If there is more than one unique lca
993
value, allow OTHER to override THIS if it has a new value, and
994
THIS only has an lca value, or vice versa. This is appropriate for
995
truly scalar values, not as much for non-scalars.
996
:return: 'this', 'other', or 'conflict' depending on whether an entry
999
# See doc/developers/lca_tree_merging.txt for details about this
1002
# Either Ambiguously clean, or nothing was actually changed. We
1005
base_val, lca_vals = bases
1006
# Remove 'base_val' from the lca_vals, because it is not interesting
1007
filtered_lca_vals = [lca_val for lca_val in lca_vals
1008
if lca_val != base_val]
1009
if len(filtered_lca_vals) == 0:
1010
return Merge3Merger._three_way(base_val, other, this)
1012
unique_lca_vals = set(filtered_lca_vals)
1013
if len(unique_lca_vals) == 1:
1014
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1016
if allow_overriding_lca:
1017
if other in unique_lca_vals:
1018
if this in unique_lca_vals:
1019
# Each side picked a different lca, conflict
1022
# This has a value which supersedes both lca values, and
1023
# other only has an lca value
1025
elif this in unique_lca_vals:
1026
# OTHER has a value which supersedes both lca values, and this
1027
# only has an lca value
1030
# At this point, the lcas disagree, and the tips disagree
1034
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1035
"""Do a three-way test on a scalar.
1036
Return "this", "other" or "conflict", depending whether a value wins.
1038
key_base = key(base_tree, file_id)
1039
key_other = key(other_tree, file_id)
1040
#if base == other, either they all agree, or only THIS has changed.
1041
if key_base == key_other:
1043
key_this = key(this_tree, file_id)
1044
# "Ambiguous clean merge"
1045
if key_this == key_other:
1047
elif key_this == key_base:
1052
def merge_names(self, file_id):
1053
def get_entry(tree):
1054
if file_id in tree.inventory:
1055
return tree.inventory[file_id]
1058
this_entry = get_entry(self.this_tree)
1059
other_entry = get_entry(self.other_tree)
1060
base_entry = get_entry(self.base_tree)
1061
entries = (base_entry, other_entry, this_entry)
1064
for entry in entries:
1067
parents.append(None)
1069
names.append(entry.name)
1070
parents.append(entry.parent_id)
1071
return self._merge_names(file_id, parents, names,
1072
resolver=self._three_way)
1074
def _merge_names(self, file_id, parents, names, resolver):
1075
"""Perform a merge on file_id names and parents"""
1076
base_name, other_name, this_name = names
1077
base_parent, other_parent, this_parent = parents
1079
name_winner = resolver(*names)
1081
parent_id_winner = resolver(*parents)
1082
if this_name is None:
1083
if name_winner == "this":
1084
name_winner = "other"
1085
if parent_id_winner == "this":
1086
parent_id_winner = "other"
1087
if name_winner == "this" and parent_id_winner == "this":
1089
if name_winner == "conflict":
1090
trans_id = self.tt.trans_id_file_id(file_id)
1091
self._raw_conflicts.append(('name conflict', trans_id,
1092
this_name, other_name))
1093
if parent_id_winner == "conflict":
1094
trans_id = self.tt.trans_id_file_id(file_id)
1095
self._raw_conflicts.append(('parent conflict', trans_id,
1096
this_parent, other_parent))
1097
if other_name is None:
1098
# it doesn't matter whether the result was 'other' or
1099
# 'conflict'-- if there's no 'other', we leave it alone.
1101
# if we get here, name_winner and parent_winner are set to safe values.
1102
trans_id = self.tt.trans_id_file_id(file_id)
1103
parent_id = parents[self.winner_idx[parent_id_winner]]
1104
if parent_id is not None:
1105
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1106
self.tt.adjust_path(names[self.winner_idx[name_winner]],
1107
parent_trans_id, trans_id)
1109
def merge_contents(self, file_id):
1110
"""Performs a merge on file_id contents."""
1111
def contents_pair(tree):
1112
if file_id not in tree:
1114
kind = tree.kind(file_id)
1116
contents = tree.get_file_sha1(file_id)
1117
elif kind == "symlink":
1118
contents = tree.get_symlink_target(file_id)
1121
return kind, contents
1123
def contents_conflict():
1124
trans_id = self.tt.trans_id_file_id(file_id)
1125
name = self.tt.final_name(trans_id)
1126
parent_id = self.tt.final_parent(trans_id)
1127
if file_id in self.this_tree.inventory:
1128
self.tt.unversion_file(trans_id)
1129
if file_id in self.this_tree:
1130
self.tt.delete_contents(trans_id)
1131
file_group = self._dump_conflicts(name, parent_id, file_id,
1133
self._raw_conflicts.append(('contents conflict', file_group))
1135
# See SPOT run. run, SPOT, run.
1136
# So we're not QUITE repeating ourselves; we do tricky things with
1138
base_pair = contents_pair(self.base_tree)
1139
other_pair = contents_pair(self.other_tree)
1141
this_pair = contents_pair(self.this_tree)
1142
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1143
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1144
this_pair, allow_overriding_lca=False)
1146
if base_pair == other_pair:
1149
# We delayed evaluating this_pair as long as we can to avoid
1150
# unnecessary sha1 calculation
1151
this_pair = contents_pair(self.this_tree)
1152
winner = self._three_way(base_pair, other_pair, this_pair)
1153
if winner == 'this':
1154
# No interesting changes introduced by OTHER
1156
trans_id = self.tt.trans_id_file_id(file_id)
1157
if winner == 'other':
1158
# OTHER is a straight winner, so replace this contents with other
1159
file_in_this = file_id in self.this_tree
1161
# Remove any existing contents
1162
self.tt.delete_contents(trans_id)
1163
if file_id in self.other_tree:
1164
# OTHER changed the file
1165
create_from_tree(self.tt, trans_id,
1166
self.other_tree, file_id)
1167
if not file_in_this:
1168
self.tt.version_file(file_id, trans_id)
1171
# OTHER deleted the file
1172
self.tt.unversion_file(trans_id)
1175
# We have a hypothetical conflict, but if we have files, then we
1176
# can try to merge the content
1177
if this_pair[0] == 'file' and other_pair[0] == 'file':
1178
# THIS and OTHER are both files, so text merge. Either
1179
# BASE is a file, or both converted to files, so at least we
1180
# have agreement that output should be a file.
1182
self.text_merge(file_id, trans_id)
1184
return contents_conflict()
1185
if file_id not in self.this_tree:
1186
self.tt.version_file(file_id, trans_id)
1188
self.tt.tree_kind(trans_id)
1189
self.tt.delete_contents(trans_id)
1194
return contents_conflict()
1196
def get_lines(self, tree, file_id):
1197
"""Return the lines in a file, or an empty list."""
1199
return tree.get_file(file_id).readlines()
1203
def text_merge(self, file_id, trans_id):
1204
"""Perform a three-way text merge on a file_id"""
1205
# it's possible that we got here with base as a different type.
1206
# if so, we just want two-way text conflicts.
1207
if file_id in self.base_tree and \
1208
self.base_tree.kind(file_id) == "file":
1209
base_lines = self.get_lines(self.base_tree, file_id)
1212
other_lines = self.get_lines(self.other_tree, file_id)
1213
this_lines = self.get_lines(self.this_tree, file_id)
1214
m3 = Merge3(base_lines, this_lines, other_lines,
1215
is_cherrypick=self.cherrypick)
1216
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1217
if self.show_base is True:
1218
base_marker = '|' * 7
1222
def iter_merge3(retval):
1223
retval["text_conflicts"] = False
1224
for line in m3.merge_lines(name_a = "TREE",
1225
name_b = "MERGE-SOURCE",
1226
name_base = "BASE-REVISION",
1227
start_marker=start_marker,
1228
base_marker=base_marker,
1229
reprocess=self.reprocess):
1230
if line.startswith(start_marker):
1231
retval["text_conflicts"] = True
1232
yield line.replace(start_marker, '<' * 7)
1236
merge3_iterator = iter_merge3(retval)
1237
self.tt.create_file(merge3_iterator, trans_id)
1238
if retval["text_conflicts"] is True:
1239
self._raw_conflicts.append(('text conflict', trans_id))
1240
name = self.tt.final_name(trans_id)
1241
parent_id = self.tt.final_parent(trans_id)
1242
file_group = self._dump_conflicts(name, parent_id, file_id,
1243
this_lines, base_lines,
1245
file_group.append(trans_id)
1247
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1248
base_lines=None, other_lines=None, set_version=False,
1250
"""Emit conflict files.
1251
If this_lines, base_lines, or other_lines are omitted, they will be
1252
determined automatically. If set_version is true, the .OTHER, .THIS
1253
or .BASE (in that order) will be created as versioned files.
1255
data = [('OTHER', self.other_tree, other_lines),
1256
('THIS', self.this_tree, this_lines)]
1258
data.append(('BASE', self.base_tree, base_lines))
1261
for suffix, tree, lines in data:
1263
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1265
file_group.append(trans_id)
1266
if set_version and not versioned:
1267
self.tt.version_file(file_id, trans_id)
1271
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1273
"""Emit a single conflict file."""
1274
name = name + '.' + suffix
1275
trans_id = self.tt.create_path(name, parent_id)
1276
create_from_tree(self.tt, trans_id, tree, file_id, lines)
1279
def merge_executable(self, file_id, file_status):
1280
"""Perform a merge on the execute bit."""
1281
executable = [self.executable(t, file_id) for t in (self.base_tree,
1282
self.other_tree, self.this_tree)]
1283
self._merge_executable(file_id, executable, file_status,
1284
resolver=self._three_way)
1286
def _merge_executable(self, file_id, executable, file_status,
1288
"""Perform a merge on the execute bit."""
1289
base_executable, other_executable, this_executable = executable
1290
if file_status == "deleted":
1292
winner = resolver(*executable)
1293
if winner == "conflict":
1294
# There must be a None in here, if we have a conflict, but we
1295
# need executability since file status was not deleted.
1296
if self.executable(self.other_tree, file_id) is None:
1300
if winner == 'this' and file_status != "modified":
1302
trans_id = self.tt.trans_id_file_id(file_id)
1304
if self.tt.final_kind(trans_id) != "file":
1308
if winner == "this":
1309
executability = this_executable
1311
if file_id in self.other_tree:
1312
executability = other_executable
1313
elif file_id in self.this_tree:
1314
executability = this_executable
1315
elif file_id in self.base_tree:
1316
executability = base_executable
1317
if executability is not None:
1318
trans_id = self.tt.trans_id_file_id(file_id)
1319
self.tt.set_executability(executability, trans_id)
1321
def cook_conflicts(self, fs_conflicts):
1322
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1323
from conflicts import Conflict
1325
self.cooked_conflicts.extend(cook_conflicts(fs_conflicts, self.tt))
1326
fp = FinalPaths(self.tt)
1327
for conflict in self._raw_conflicts:
1328
conflict_type = conflict[0]
1329
if conflict_type in ('name conflict', 'parent conflict'):
1330
trans_id = conflict[1]
1331
conflict_args = conflict[2:]
1332
if trans_id not in name_conflicts:
1333
name_conflicts[trans_id] = {}
1334
unique_add(name_conflicts[trans_id], conflict_type,
1336
if conflict_type == 'contents conflict':
1337
for trans_id in conflict[1]:
1338
file_id = self.tt.final_file_id(trans_id)
1339
if file_id is not None:
1341
path = fp.get_path(trans_id)
1342
for suffix in ('.BASE', '.THIS', '.OTHER'):
1343
if path.endswith(suffix):
1344
path = path[:-len(suffix)]
1346
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
1347
self.cooked_conflicts.append(c)
1348
if conflict_type == 'text conflict':
1349
trans_id = conflict[1]
1350
path = fp.get_path(trans_id)
1351
file_id = self.tt.final_file_id(trans_id)
1352
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
1353
self.cooked_conflicts.append(c)
1355
for trans_id, conflicts in name_conflicts.iteritems():
1357
this_parent, other_parent = conflicts['parent conflict']
1358
if this_parent == other_parent:
1359
raise AssertionError()
1361
this_parent = other_parent = \
1362
self.tt.final_file_id(self.tt.final_parent(trans_id))
1364
this_name, other_name = conflicts['name conflict']
1365
if this_name == other_name:
1366
raise AssertionError()
1368
this_name = other_name = self.tt.final_name(trans_id)
1369
other_path = fp.get_path(trans_id)
1370
if this_parent is not None and this_name is not None:
1371
this_parent_path = \
1372
fp.get_path(self.tt.trans_id_file_id(this_parent))
1373
this_path = pathjoin(this_parent_path, this_name)
1375
this_path = "<deleted>"
1376
file_id = self.tt.final_file_id(trans_id)
1377
c = Conflict.factory('path conflict', path=this_path,
1378
conflict_path=other_path, file_id=file_id)
1379
self.cooked_conflicts.append(c)
1380
self.cooked_conflicts.sort(key=Conflict.sort_key)
1383
class WeaveMerger(Merge3Merger):
1384
"""Three-way tree merger, text weave merger."""
1385
supports_reprocess = True
1386
supports_show_base = False
1387
supports_reverse_cherrypick = False
1388
history_based = True
1390
def _merged_lines(self, file_id):
1391
"""Generate the merged lines.
1392
There is no distinction between lines that are meant to contain <<<<<<<
1396
base = self.base_tree
1399
plan = self.this_tree.plan_file_merge(file_id, self.other_tree,
1401
if 'merge' in debug.debug_flags:
1403
trans_id = self.tt.trans_id_file_id(file_id)
1404
name = self.tt.final_name(trans_id) + '.plan'
1405
contents = ('%10s|%s' % l for l in plan)
1406
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1407
textmerge = PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1408
'>>>>>>> MERGE-SOURCE\n')
1409
return textmerge.merge_lines(self.reprocess)
1411
def text_merge(self, file_id, trans_id):
1412
"""Perform a (weave) text merge for a given file and file-id.
1413
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1414
and a conflict will be noted.
1416
lines, conflicts = self._merged_lines(file_id)
1418
# Note we're checking whether the OUTPUT is binary in this case,
1419
# because we don't want to get into weave merge guts.
1420
check_text_lines(lines)
1421
self.tt.create_file(lines, trans_id)
1423
self._raw_conflicts.append(('text conflict', trans_id))
1424
name = self.tt.final_name(trans_id)
1425
parent_id = self.tt.final_parent(trans_id)
1426
file_group = self._dump_conflicts(name, parent_id, file_id,
1428
file_group.append(trans_id)
1431
class LCAMerger(WeaveMerger):
1433
def _merged_lines(self, file_id):
1434
"""Generate the merged lines.
1435
There is no distinction between lines that are meant to contain <<<<<<<
1439
base = self.base_tree
1442
plan = self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1444
if 'merge' in debug.debug_flags:
1446
trans_id = self.tt.trans_id_file_id(file_id)
1447
name = self.tt.final_name(trans_id) + '.plan'
1448
contents = ('%10s|%s' % l for l in plan)
1449
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1450
textmerge = PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1451
'>>>>>>> MERGE-SOURCE\n')
1452
return textmerge.merge_lines(self.reprocess)
1455
class Diff3Merger(Merge3Merger):
1456
"""Three-way merger using external diff3 for text merging"""
1458
def dump_file(self, temp_dir, name, tree, file_id):
1459
out_path = pathjoin(temp_dir, name)
1460
out_file = open(out_path, "wb")
1462
in_file = tree.get_file(file_id)
1463
for line in in_file:
1464
out_file.write(line)
1469
def text_merge(self, file_id, trans_id):
1470
"""Perform a diff3 merge using a specified file-id and trans-id.
1471
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1472
will be dumped, and a will be conflict noted.
1475
temp_dir = osutils.mkdtemp(prefix="bzr-")
1477
new_file = pathjoin(temp_dir, "new")
1478
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1479
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1480
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1481
status = bzrlib.patch.diff3(new_file, this, base, other)
1482
if status not in (0, 1):
1483
raise BzrError("Unhandled diff3 exit code")
1484
f = open(new_file, 'rb')
1486
self.tt.create_file(f, trans_id)
1490
name = self.tt.final_name(trans_id)
1491
parent_id = self.tt.final_parent(trans_id)
1492
self._dump_conflicts(name, parent_id, file_id)
1493
self._raw_conflicts.append(('text conflict', trans_id))
1495
osutils.rmtree(temp_dir)
1498
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1500
merge_type=Merge3Merger,
1501
interesting_ids=None,
1505
interesting_files=None,
1508
change_reporter=None):
1509
"""Primary interface for merging.
1511
typical use is probably
1512
'merge_inner(branch, branch.get_revision_tree(other_revision),
1513
branch.get_revision_tree(base_revision))'
1515
if this_tree is None:
1516
raise BzrError("bzrlib.merge.merge_inner requires a this_tree "
1517
"parameter as of bzrlib version 0.8.")
1518
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1519
pb=pb, change_reporter=change_reporter)
1520
merger.backup_files = backup_files
1521
merger.merge_type = merge_type
1522
merger.interesting_ids = interesting_ids
1523
merger.ignore_zero = ignore_zero
1524
if interesting_files:
1526
raise ValueError('Only supply interesting_ids'
1527
' or interesting_files')
1528
merger.interesting_files = interesting_files
1529
merger.show_base = show_base
1530
merger.reprocess = reprocess
1531
merger.other_rev_id = other_rev_id
1532
merger.other_basis = other_rev_id
1533
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1534
if get_revision_id is None:
1535
get_revision_id = base_tree.last_revision
1536
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
1537
merger.set_base_revision(get_revision_id(), this_branch)
1538
return merger.do_merge()
1540
def get_merge_type_registry():
1541
"""Merge type registry is in bzrlib.option to avoid circular imports.
1543
This method provides a sanctioned way to retrieve it.
1545
from bzrlib import option
1546
return option._merge_type_registry
1549
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1550
def status_a(revision, text):
1551
if revision in ancestors_b:
1552
return 'killed-b', text
1554
return 'new-a', text
1556
def status_b(revision, text):
1557
if revision in ancestors_a:
1558
return 'killed-a', text
1560
return 'new-b', text
1562
plain_a = [t for (a, t) in annotated_a]
1563
plain_b = [t for (a, t) in annotated_b]
1564
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1565
blocks = matcher.get_matching_blocks()
1568
for ai, bi, l in blocks:
1569
# process all mismatched sections
1570
# (last mismatched section is handled because blocks always
1571
# includes a 0-length last block)
1572
for revision, text in annotated_a[a_cur:ai]:
1573
yield status_a(revision, text)
1574
for revision, text in annotated_b[b_cur:bi]:
1575
yield status_b(revision, text)
1576
# and now the matched section
1579
for text_a in plain_a[ai:a_cur]:
1580
yield "unchanged", text_a
1583
class _PlanMergeBase(object):
1585
def __init__(self, a_rev, b_rev, vf, key_prefix):
1588
:param a_rev: Revision-id of one revision to merge
1589
:param b_rev: Revision-id of the other revision to merge
1590
:param vf: A VersionedFiles containing both revisions
1591
:param key_prefix: A prefix for accessing keys in vf, typically
1597
self._last_lines = None
1598
self._last_lines_revision_id = None
1599
self._cached_matching_blocks = {}
1600
self._key_prefix = key_prefix
1601
self._precache_tip_lines()
1603
def _precache_tip_lines(self):
1604
lines = self.get_lines([self.a_rev, self.b_rev])
1605
self.lines_a = lines[self.a_rev]
1606
self.lines_b = lines[self.b_rev]
1608
def get_lines(self, revisions):
1609
"""Get lines for revisions from the backing VersionedFiles.
1611
:raises RevisionNotPresent: on absent texts.
1613
keys = [(self._key_prefix + (rev,)) for rev in revisions]
1615
for record in self.vf.get_record_stream(keys, 'unordered', True):
1616
if record.storage_kind == 'absent':
1617
raise errors.RevisionNotPresent(record.key, self.vf)
1618
result[record.key[-1]] = osutils.chunks_to_lines(
1619
record.get_bytes_as('chunked'))
1622
def plan_merge(self):
1623
"""Generate a 'plan' for merging the two revisions.
1625
This involves comparing their texts and determining the cause of
1626
differences. If text A has a line and text B does not, then either the
1627
line was added to text A, or it was deleted from B. Once the causes
1628
are combined, they are written out in the format described in
1629
VersionedFile.plan_merge
1631
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
1632
unique_a, unique_b = self._unique_lines(blocks)
1633
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
1634
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
1635
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
1637
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
1640
for i, j, n in blocks:
1641
for a_index in range(last_i, i):
1642
if a_index in new_a:
1643
if a_index in killed_b:
1644
yield 'conflicted-a', self.lines_a[a_index]
1646
yield 'new-a', self.lines_a[a_index]
1648
yield 'killed-b', self.lines_a[a_index]
1649
for b_index in range(last_j, j):
1650
if b_index in new_b:
1651
if b_index in killed_a:
1652
yield 'conflicted-b', self.lines_b[b_index]
1654
yield 'new-b', self.lines_b[b_index]
1656
yield 'killed-a', self.lines_b[b_index]
1657
# handle common lines
1658
for a_index in range(i, i+n):
1659
yield 'unchanged', self.lines_a[a_index]
1663
def _get_matching_blocks(self, left_revision, right_revision):
1664
"""Return a description of which sections of two revisions match.
1666
See SequenceMatcher.get_matching_blocks
1668
cached = self._cached_matching_blocks.get((left_revision,
1670
if cached is not None:
1672
if self._last_lines_revision_id == left_revision:
1673
left_lines = self._last_lines
1674
right_lines = self.get_lines([right_revision])[right_revision]
1676
lines = self.get_lines([left_revision, right_revision])
1677
left_lines = lines[left_revision]
1678
right_lines = lines[right_revision]
1679
self._last_lines = right_lines
1680
self._last_lines_revision_id = right_revision
1681
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
1683
return matcher.get_matching_blocks()
1685
def _unique_lines(self, matching_blocks):
1686
"""Analyse matching_blocks to determine which lines are unique
1688
:return: a tuple of (unique_left, unique_right), where the values are
1689
sets of line numbers of unique lines.
1695
for i, j, n in matching_blocks:
1696
unique_left.extend(range(last_i, i))
1697
unique_right.extend(range(last_j, j))
1700
return unique_left, unique_right
1703
def _subtract_plans(old_plan, new_plan):
1704
"""Remove changes from new_plan that came from old_plan.
1706
It is assumed that the difference between the old_plan and new_plan
1707
is their choice of 'b' text.
1709
All lines from new_plan that differ from old_plan are emitted
1710
verbatim. All lines from new_plan that match old_plan but are
1711
not about the 'b' revision are emitted verbatim.
1713
Lines that match and are about the 'b' revision are the lines we
1714
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
1715
is skipped entirely.
1717
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
1720
for i, j, n in matcher.get_matching_blocks():
1721
for jj in range(last_j, j):
1723
for jj in range(j, j+n):
1724
plan_line = new_plan[jj]
1725
if plan_line[0] == 'new-b':
1727
elif plan_line[0] == 'killed-b':
1728
yield 'unchanged', plan_line[1]
1734
class _PlanMerge(_PlanMergeBase):
1735
"""Plan an annotate merge using on-the-fly annotation"""
1737
def __init__(self, a_rev, b_rev, vf, key_prefix):
1738
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
1739
self.a_key = self._key_prefix + (self.a_rev,)
1740
self.b_key = self._key_prefix + (self.b_rev,)
1741
self.graph = Graph(self.vf)
1742
heads = self.graph.heads((self.a_key, self.b_key))
1744
# one side dominates, so we can just return its values, yay for
1746
# Ideally we would know that before we get this far
1747
self._head_key = heads.pop()
1748
if self._head_key == self.a_key:
1752
mutter('found dominating revision for %s\n%s > %s', self.vf,
1753
self._head_key[-1], other)
1756
self._head_key = None
1759
def _precache_tip_lines(self):
1760
# Turn this into a no-op, because we will do this later
1763
def _find_recursive_lcas(self):
1764
"""Find all the ancestors back to a unique lca"""
1765
cur_ancestors = (self.a_key, self.b_key)
1766
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
1767
# rather than a key tuple. We will just map that directly to no common
1771
next_lcas = self.graph.find_lca(*cur_ancestors)
1772
# Map a plain NULL_REVISION to a simple no-ancestors
1773
if next_lcas == set([NULL_REVISION]):
1775
# Order the lca's based on when they were merged into the tip
1776
# While the actual merge portion of weave merge uses a set() of
1777
# active revisions, the order of insertion *does* effect the
1778
# implicit ordering of the texts.
1779
for rev_key in cur_ancestors:
1780
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
1782
parent_map[rev_key] = ordered_parents
1783
if len(next_lcas) == 0:
1785
elif len(next_lcas) == 1:
1786
parent_map[list(next_lcas)[0]] = ()
1788
elif len(next_lcas) > 2:
1789
# More than 2 lca's, fall back to grabbing all nodes between
1790
# this and the unique lca.
1791
mutter('More than 2 LCAs, falling back to all nodes for:'
1792
' %s, %s\n=> %s', self.a_key, self.b_key, cur_ancestors)
1793
cur_lcas = next_lcas
1794
while len(cur_lcas) > 1:
1795
cur_lcas = self.graph.find_lca(*cur_lcas)
1796
if len(cur_lcas) == 0:
1797
# No common base to find, use the full ancestry
1800
unique_lca = list(cur_lcas)[0]
1801
if unique_lca == NULL_REVISION:
1802
# find_lca will return a plain 'NULL_REVISION' rather
1803
# than a key tuple when there is no common ancestor, we
1804
# prefer to just use None, because it doesn't confuse
1805
# _get_interesting_texts()
1807
parent_map.update(self._find_unique_parents(next_lcas,
1810
cur_ancestors = next_lcas
1813
def _find_unique_parents(self, tip_keys, base_key):
1814
"""Find ancestors of tip that aren't ancestors of base.
1816
:param tip_keys: Nodes that are interesting
1817
:param base_key: Cull all ancestors of this node
1818
:return: The parent map for all revisions between tip_keys and
1819
base_key. base_key will be included. References to nodes outside of
1820
the ancestor set will also be removed.
1822
# TODO: this would be simpler if find_unique_ancestors took a list
1823
# instead of a single tip, internally it supports it, but it
1824
# isn't a "backwards compatible" api change.
1825
if base_key is None:
1826
parent_map = dict(self.graph.iter_ancestry(tip_keys))
1827
# We remove NULL_REVISION because it isn't a proper tuple key, and
1828
# thus confuses things like _get_interesting_texts, and our logic
1829
# to add the texts into the memory weave.
1830
if NULL_REVISION in parent_map:
1831
parent_map.pop(NULL_REVISION)
1834
for tip in tip_keys:
1836
self.graph.find_unique_ancestors(tip, [base_key]))
1837
parent_map = self.graph.get_parent_map(interesting)
1838
parent_map[base_key] = ()
1839
culled_parent_map, child_map, tails = self._remove_external_references(
1841
# Remove all the tails but base_key
1842
if base_key is not None:
1843
tails.remove(base_key)
1844
self._prune_tails(culled_parent_map, child_map, tails)
1845
# Now remove all the uninteresting 'linear' regions
1846
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
1850
def _remove_external_references(parent_map):
1851
"""Remove references that go outside of the parent map.
1853
:param parent_map: Something returned from Graph.get_parent_map(keys)
1854
:return: (filtered_parent_map, child_map, tails)
1855
filtered_parent_map is parent_map without external references
1856
child_map is the {parent_key: [child_keys]} mapping
1857
tails is a list of nodes that do not have any parents in the map
1859
# TODO: The basic effect of this function seems more generic than
1860
# _PlanMerge. But the specific details of building a child_map,
1861
# and computing tails seems very specific to _PlanMerge.
1862
# Still, should this be in Graph land?
1863
filtered_parent_map = {}
1866
for key, parent_keys in parent_map.iteritems():
1867
culled_parent_keys = [p for p in parent_keys if p in parent_map]
1868
if not culled_parent_keys:
1870
for parent_key in culled_parent_keys:
1871
child_map.setdefault(parent_key, []).append(key)
1872
# TODO: Do we want to do this, it adds overhead for every node,
1873
# just to say that the node has no children
1874
child_map.setdefault(key, [])
1875
filtered_parent_map[key] = culled_parent_keys
1876
return filtered_parent_map, child_map, tails
1879
def _prune_tails(parent_map, child_map, tails_to_remove):
1880
"""Remove tails from the parent map.
1882
This will remove the supplied revisions until no more children have 0
1885
:param parent_map: A dict of {child: [parents]}, this dictionary will
1886
be modified in place.
1887
:param tails_to_remove: A list of tips that should be removed,
1888
this list will be consumed
1889
:param child_map: The reverse dict of parent_map ({parent: [children]})
1890
this dict will be modified
1891
:return: None, parent_map will be modified in place.
1893
while tails_to_remove:
1894
next = tails_to_remove.pop()
1895
parent_map.pop(next)
1896
children = child_map.pop(next)
1897
for child in children:
1898
child_parents = parent_map[child]
1899
child_parents.remove(next)
1900
if len(child_parents) == 0:
1901
tails_to_remove.append(child)
1903
def _get_interesting_texts(self, parent_map):
1904
"""Return a dict of texts we are interested in.
1906
Note that the input is in key tuples, but the output is in plain
1909
:param parent_map: The output from _find_recursive_lcas
1910
:return: A dict of {'revision_id':lines} as returned by
1911
_PlanMergeBase.get_lines()
1913
all_revision_keys = set(parent_map)
1914
all_revision_keys.add(self.a_key)
1915
all_revision_keys.add(self.b_key)
1917
# Everything else is in 'keys' but get_lines is in 'revision_ids'
1918
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
1921
def _build_weave(self):
1922
from bzrlib import weave
1923
self._weave = weave.Weave(weave_name='in_memory_weave',
1924
allow_reserved=True)
1925
parent_map = self._find_recursive_lcas()
1927
all_texts = self._get_interesting_texts(parent_map)
1929
# Note: Unfortunately, the order given by topo_sort will effect the
1930
# ordering resolution in the output. Specifically, if you add A then B,
1931
# then in the output text A lines will show up before B lines. And, of
1932
# course, topo_sort doesn't guarantee any real ordering.
1933
# So we use merge_sort, and add a fake node on the tip.
1934
# This ensures that left-hand parents will always be inserted into the
1935
# weave before right-hand parents.
1936
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
1937
parent_map[tip_key] = (self.a_key, self.b_key)
1939
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
1943
# for key in tsort.topo_sort(parent_map):
1944
parent_keys = parent_map[key]
1945
revision_id = key[-1]
1946
parent_ids = [k[-1] for k in parent_keys]
1947
self._weave.add_lines(revision_id, parent_ids,
1948
all_texts[revision_id])
1950
def plan_merge(self):
1951
"""Generate a 'plan' for merging the two revisions.
1953
This involves comparing their texts and determining the cause of
1954
differences. If text A has a line and text B does not, then either the
1955
line was added to text A, or it was deleted from B. Once the causes
1956
are combined, they are written out in the format described in
1957
VersionedFile.plan_merge
1959
if self._head_key is not None: # There was a single head
1960
if self._head_key == self.a_key:
1963
if self._head_key != self.b_key:
1964
raise AssertionError('There was an invalid head: %s != %s'
1965
% (self.b_key, self._head_key))
1967
head_rev = self._head_key[-1]
1968
lines = self.get_lines([head_rev])[head_rev]
1969
return ((plan, line) for line in lines)
1970
return self._weave.plan_merge(self.a_rev, self.b_rev)
1973
class _PlanLCAMerge(_PlanMergeBase):
1975
This merge algorithm differs from _PlanMerge in that:
1976
1. comparisons are done against LCAs only
1977
2. cases where a contested line is new versus one LCA but old versus
1978
another are marked as conflicts, by emitting the line as conflicted-a
1981
This is faster, and hopefully produces more useful output.
1984
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
1985
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
1986
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
1989
if lca == NULL_REVISION:
1992
self.lcas.add(lca[-1])
1993
for lca in self.lcas:
1994
if _mod_revision.is_null(lca):
1997
lca_lines = self.get_lines([lca])[lca]
1998
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2000
blocks = list(matcher.get_matching_blocks())
2001
self._cached_matching_blocks[(a_rev, lca)] = blocks
2002
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2004
blocks = list(matcher.get_matching_blocks())
2005
self._cached_matching_blocks[(b_rev, lca)] = blocks
2007
def _determine_status(self, revision_id, unique_line_numbers):
2008
"""Determines the status unique lines versus all lcas.
2010
Basically, determines why the line is unique to this revision.
2012
A line may be determined new, killed, or both.
2014
If a line is determined new, that means it was not present in at least
2015
one LCA, and is not present in the other merge revision.
2017
If a line is determined killed, that means the line was present in
2020
If a line is killed and new, this indicates that the two merge
2021
revisions contain differing conflict resolutions.
2022
:param revision_id: The id of the revision in which the lines are
2024
:param unique_line_numbers: The line numbers of unique lines.
2025
:return a tuple of (new_this, killed_other):
2029
unique_line_numbers = set(unique_line_numbers)
2030
for lca in self.lcas:
2031
blocks = self._get_matching_blocks(revision_id, lca)
2032
unique_vs_lca, _ignored = self._unique_lines(blocks)
2033
new.update(unique_line_numbers.intersection(unique_vs_lca))
2034
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
