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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 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
merge_inner(from_tree.branch, to_tree, from_tree, ignore_zero=True,
68
interesting_ids=interesting_ids, this_tree=from_tree)
72
def __init__(self, this_branch, other_tree=None, base_tree=None,
73
this_tree=None, pb=DummyProgress(), change_reporter=None,
74
recurse='down', revision_graph=None):
76
self.this_branch = this_branch
77
self.this_basis = _mod_revision.ensure_null(
78
this_branch.last_revision())
79
self.this_rev_id = None
80
self.this_tree = this_tree
81
self.this_revision_tree = None
82
self.this_basis_tree = None
83
self.other_tree = other_tree
84
self.other_branch = None
85
self.base_tree = base_tree
86
self.ignore_zero = False
87
self.backup_files = False
88
self.interesting_ids = None
89
self.interesting_files = None
90
self.show_base = False
91
self.reprocess = False
94
self.recurse = recurse
95
self.change_reporter = change_reporter
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)
236
def ensure_revision_trees(self):
237
if self.this_revision_tree is None:
238
self.this_basis_tree = self.revision_tree(self.this_basis)
239
if self.this_basis == self.this_rev_id:
240
self.this_revision_tree = self.this_basis_tree
242
if self.other_rev_id is None:
243
other_basis_tree = self.revision_tree(self.other_basis)
244
changes = other_basis_tree.changes_from(self.other_tree)
245
if changes.has_changed():
246
raise WorkingTreeNotRevision(self.this_tree)
247
other_rev_id = self.other_basis
248
self.other_tree = other_basis_tree
250
def file_revisions(self, file_id):
251
self.ensure_revision_trees()
252
def get_id(tree, file_id):
253
revision_id = tree.inventory[file_id].revision
255
if self.this_rev_id is None:
256
if self.this_basis_tree.get_file_sha1(file_id) != \
257
self.this_tree.get_file_sha1(file_id):
258
raise WorkingTreeNotRevision(self.this_tree)
260
trees = (self.this_basis_tree, self.other_tree)
261
return [get_id(tree, file_id) for tree in trees]
263
def check_basis(self, check_clean, require_commits=True):
264
if self.this_basis is None and require_commits is True:
265
raise BzrCommandError("This branch has no commits."
266
" (perhaps you would prefer 'bzr pull')")
269
if self.this_basis != self.this_rev_id:
270
raise errors.UncommittedChanges(self.this_tree)
272
def compare_basis(self):
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)
278
if not changes.has_changed():
279
self.this_rev_id = self.this_basis
281
def set_interesting_files(self, file_list):
282
self.interesting_files = file_list
284
def set_pending(self):
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):
309
"""Set the revision and tree to merge from.
311
This sets the other_tree, other_rev_id, other_basis attributes.
313
:param other_revision: The [path, revision] list to merge from.
315
self.other_branch, self.other_tree = self._get_tree(other_revision,
317
if other_revision[1] == -1:
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)
322
self.other_basis = self.other_rev_id
323
elif other_revision[1] is not None:
324
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
325
self.other_basis = self.other_rev_id
327
self.other_rev_id = None
328
self.other_basis = self.other_branch.last_revision()
329
if self.other_basis is None:
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)
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)
411
def set_base(self, base_revision):
412
"""Set the base revision to use for the merge.
414
:param base_revision: A 2-list containing a path and revision number.
416
mutter("doing merge() with no base_revision specified")
417
if base_revision == [None, None]:
420
base_branch, self.base_tree = self._get_tree(base_revision)
421
if base_revision[1] == -1:
422
self.base_rev_id = base_branch.last_revision()
423
elif base_revision[1] is None:
424
self.base_rev_id = _mod_revision.NULL_REVISION
426
self.base_rev_id = _mod_revision.ensure_null(
427
base_branch.get_rev_id(base_revision[1]))
428
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
430
def make_merger(self):
431
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
432
'other_tree': self.other_tree,
433
'interesting_ids': self.interesting_ids,
434
'interesting_files': self.interesting_files,
437
if self.merge_type.requires_base:
438
kwargs['base_tree'] = self.base_tree
439
if self.merge_type.supports_reprocess:
440
kwargs['reprocess'] = self.reprocess
442
raise BzrError("Conflict reduction is not supported for merge"
443
" type %s." % self.merge_type)
444
if self.merge_type.supports_show_base:
445
kwargs['show_base'] = self.show_base
447
raise BzrError("Showing base is not supported for this"
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
482
self.this_tree.lock_tree_write()
484
if self.base_tree is not None:
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()
499
self.this_tree.unlock()
500
if len(merge.cooked_conflicts) == 0:
501
if not self.ignore_zero and not is_quiet():
502
note("All changes applied successfully.")
504
note("%d conflicts encountered." % len(merge.cooked_conflicts))
506
return len(merge.cooked_conflicts)
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()
526
class Merge3Merger(object):
527
"""Three-way merger that uses the merge3 text merger"""
529
supports_reprocess = True
530
supports_show_base = True
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
537
def __init__(self, working_tree, this_tree, base_tree, other_tree,
538
interesting_ids=None, reprocess=False, show_base=False,
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.
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
573
self.this_tree = working_tree
574
self.base_tree = base_tree
575
self.other_tree = other_tree
576
self._raw_conflicts = []
577
self.cooked_conflicts = []
578
self.reprocess = reprocess
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]
588
self.change_reporter = change_reporter
589
self.cherrypick = cherrypick
591
self.pp = ProgressPhase("Merge phase", 3, 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)
602
self._compute_transform()
604
results = self.tt.apply(no_conflicts=True)
605
self.write_modified(results)
607
self.this_tree.add_conflicts(self.cooked_conflicts)
608
except UnsupportedOperation:
612
self.other_tree.unlock()
613
self.base_tree.unlock()
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)
873
self.tt.final_kind(self.tt.root)
875
self.tt.cancel_deletion(self.tt.root)
876
if self.tt.final_file_id(self.tt.root) is None:
877
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
879
other_root_file_id = self.other_tree.get_root_id()
880
if other_root_file_id is None:
882
other_root = self.tt.trans_id_file_id(other_root_file_id)
883
if other_root == self.tt.root:
886
self.tt.final_kind(other_root)
889
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
890
# the other tree's root is a non-root in the current tree
892
self.reparent_children(self.other_tree.inventory.root, self.tt.root)
893
self.tt.cancel_creation(other_root)
894
self.tt.cancel_versioning(other_root)
896
def reparent_children(self, ie, target):
897
for thing, child in ie.children.iteritems():
898
trans_id = self.tt.trans_id_file_id(child.file_id)
899
self.tt.adjust_path(self.tt.final_name(trans_id), target, trans_id)
901
def write_modified(self, results):
903
for path in results.modified_paths:
904
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
907
hash = self.this_tree.get_file_sha1(file_id)
910
modified_hashes[file_id] = hash
911
self.this_tree.set_merge_modified(modified_hashes)
914
def parent(entry, file_id):
915
"""Determine the parent for a file_id (used as a key method)"""
918
return entry.parent_id
921
def name(entry, file_id):
922
"""Determine the name for a file_id (used as a key method)"""
928
def contents_sha1(tree, file_id):
929
"""Determine the sha1 of the file contents (used as a key method)."""
930
if file_id not in tree:
932
return tree.get_file_sha1(file_id)
935
def executable(tree, file_id):
936
"""Determine the executability of a file-id (used as a key method)."""
937
if file_id not in tree:
939
if tree.kind(file_id) != "file":
941
return tree.is_executable(file_id)
944
def kind(tree, file_id):
945
"""Determine the kind of a file-id (used as a key method)."""
946
if file_id not in tree:
948
return tree.kind(file_id)
951
def _three_way(base, other, this):
952
#if base == other, either they all agree, or only THIS has changed.
955
elif this not in (base, other):
957
# "Ambiguous clean merge" -- both sides have made the same change.
960
# this == base: only other has changed.
965
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
966
"""Consider LCAs when determining whether a change has occurred.
968
If LCAS are all identical, this is the same as a _three_way comparison.
970
:param bases: value in (BASE, [LCAS])
971
:param other: value in OTHER
972
:param this: value in THIS
973
:param allow_overriding_lca: If there is more than one unique lca
974
value, allow OTHER to override THIS if it has a new value, and
975
THIS only has an lca value, or vice versa. This is appropriate for
976
truly scalar values, not as much for non-scalars.
977
:return: 'this', 'other', or 'conflict' depending on whether an entry
980
# See doc/developers/lca_merge_resolution.txt for details about this
983
# Either Ambiguously clean, or nothing was actually changed. We
986
base_val, lca_vals = bases
987
# Remove 'base_val' from the lca_vals, because it is not interesting
988
filtered_lca_vals = [lca_val for lca_val in lca_vals
989
if lca_val != base_val]
990
if len(filtered_lca_vals) == 0:
991
return Merge3Merger._three_way(base_val, other, this)
993
unique_lca_vals = set(filtered_lca_vals)
994
if len(unique_lca_vals) == 1:
995
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
997
if allow_overriding_lca:
998
if other in unique_lca_vals:
999
if this in unique_lca_vals:
1000
# Each side picked a different lca, conflict
1003
# This has a value which supersedes both lca values, and
1004
# other only has an lca value
1006
elif this in unique_lca_vals:
1007
# OTHER has a value which supersedes both lca values, and this
1008
# only has an lca value
1011
# At this point, the lcas disagree, and the tips disagree
1015
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1016
"""Do a three-way test on a scalar.
1017
Return "this", "other" or "conflict", depending whether a value wins.
1019
key_base = key(base_tree, file_id)
1020
key_other = key(other_tree, file_id)
1021
#if base == other, either they all agree, or only THIS has changed.
1022
if key_base == key_other:
1024
key_this = key(this_tree, file_id)
1025
# "Ambiguous clean merge"
1026
if key_this == key_other:
1028
elif key_this == key_base:
1033
def merge_names(self, file_id):
1034
def get_entry(tree):
1035
if file_id in tree.inventory:
1036
return tree.inventory[file_id]
1039
this_entry = get_entry(self.this_tree)
1040
other_entry = get_entry(self.other_tree)
1041
base_entry = get_entry(self.base_tree)
1042
entries = (base_entry, other_entry, this_entry)
1045
for entry in entries:
1048
parents.append(None)
1050
names.append(entry.name)
1051
parents.append(entry.parent_id)
1052
return self._merge_names(file_id, parents, names,
1053
resolver=self._three_way)
1055
def _merge_names(self, file_id, parents, names, resolver):
1056
"""Perform a merge on file_id names and parents"""
1057
base_name, other_name, this_name = names
1058
base_parent, other_parent, this_parent = parents
1060
name_winner = resolver(*names)
1062
parent_id_winner = resolver(*parents)
1063
if this_name is None:
1064
if name_winner == "this":
1065
name_winner = "other"
1066
if parent_id_winner == "this":
1067
parent_id_winner = "other"
1068
if name_winner == "this" and parent_id_winner == "this":
1070
if name_winner == "conflict":
1071
trans_id = self.tt.trans_id_file_id(file_id)
1072
self._raw_conflicts.append(('name conflict', trans_id,
1073
this_name, other_name))
1074
if parent_id_winner == "conflict":
1075
trans_id = self.tt.trans_id_file_id(file_id)
1076
self._raw_conflicts.append(('parent conflict', trans_id,
1077
this_parent, other_parent))
1078
if other_name is None:
1079
# it doesn't matter whether the result was 'other' or
1080
# 'conflict'-- if there's no 'other', we leave it alone.
1082
# if we get here, name_winner and parent_winner are set to safe values.
1083
trans_id = self.tt.trans_id_file_id(file_id)
1084
parent_id = parents[self.winner_idx[parent_id_winner]]
1085
if parent_id is not None:
1086
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1087
self.tt.adjust_path(names[self.winner_idx[name_winner]],
1088
parent_trans_id, trans_id)
1090
def merge_contents(self, file_id):
1091
"""Performs a merge on file_id contents."""
1092
def contents_pair(tree):
1093
if file_id not in tree:
1095
kind = tree.kind(file_id)
1097
contents = tree.get_file_sha1(file_id)
1098
elif kind == "symlink":
1099
contents = tree.get_symlink_target(file_id)
1102
return kind, contents
1104
def contents_conflict():
1105
trans_id = self.tt.trans_id_file_id(file_id)
1106
name = self.tt.final_name(trans_id)
1107
parent_id = self.tt.final_parent(trans_id)
1108
if file_id in self.this_tree.inventory:
1109
self.tt.unversion_file(trans_id)
1110
if file_id in self.this_tree:
1111
self.tt.delete_contents(trans_id)
1112
file_group = self._dump_conflicts(name, parent_id, file_id,
1114
self._raw_conflicts.append(('contents conflict', file_group))
1116
# See SPOT run. run, SPOT, run.
1117
# So we're not QUITE repeating ourselves; we do tricky things with
1119
base_pair = contents_pair(self.base_tree)
1120
other_pair = contents_pair(self.other_tree)
1121
if base_pair == other_pair:
1122
# OTHER introduced no changes
1124
this_pair = contents_pair(self.this_tree)
1125
if this_pair == other_pair:
1126
# THIS and OTHER introduced the same changes
1129
trans_id = self.tt.trans_id_file_id(file_id)
1130
if this_pair == base_pair:
1131
# only OTHER introduced changes
1132
if file_id in self.this_tree:
1133
# Remove any existing contents
1134
self.tt.delete_contents(trans_id)
1135
if file_id in self.other_tree:
1136
# OTHER changed the file
1137
create_from_tree(self.tt, trans_id,
1138
self.other_tree, file_id)
1139
if file_id not in self.this_tree:
1140
self.tt.version_file(file_id, trans_id)
1142
elif file_id in self.this_tree.inventory:
1143
# OTHER deleted the file
1144
self.tt.unversion_file(trans_id)
1146
#BOTH THIS and OTHER introduced changes; scalar conflict
1147
elif this_pair[0] == "file" and other_pair[0] == "file":
1148
# THIS and OTHER are both files, so text merge. Either
1149
# BASE is a file, or both converted to files, so at least we
1150
# have agreement that output should be a file.
1152
self.text_merge(file_id, trans_id)
1154
return contents_conflict()
1155
if file_id not in self.this_tree:
1156
self.tt.version_file(file_id, trans_id)
1158
self.tt.tree_kind(trans_id)
1159
self.tt.delete_contents(trans_id)
1164
# Scalar conflict, can't text merge. Dump conflicts
1165
return contents_conflict()
1167
def get_lines(self, tree, file_id):
1168
"""Return the lines in a file, or an empty list."""
1170
return tree.get_file(file_id).readlines()
1174
def text_merge(self, file_id, trans_id):
1175
"""Perform a three-way text merge on a file_id"""
1176
# it's possible that we got here with base as a different type.
1177
# if so, we just want two-way text conflicts.
1178
if file_id in self.base_tree and \
1179
self.base_tree.kind(file_id) == "file":
1180
base_lines = self.get_lines(self.base_tree, file_id)
1183
other_lines = self.get_lines(self.other_tree, file_id)
1184
this_lines = self.get_lines(self.this_tree, file_id)
1185
m3 = Merge3(base_lines, this_lines, other_lines,
1186
is_cherrypick=self.cherrypick)
1187
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1188
if self.show_base is True:
1189
base_marker = '|' * 7
1193
def iter_merge3(retval):
1194
retval["text_conflicts"] = False
1195
for line in m3.merge_lines(name_a = "TREE",
1196
name_b = "MERGE-SOURCE",
1197
name_base = "BASE-REVISION",
1198
start_marker=start_marker,
1199
base_marker=base_marker,
1200
reprocess=self.reprocess):
1201
if line.startswith(start_marker):
1202
retval["text_conflicts"] = True
1203
yield line.replace(start_marker, '<' * 7)
1207
merge3_iterator = iter_merge3(retval)
1208
self.tt.create_file(merge3_iterator, trans_id)
1209
if retval["text_conflicts"] is True:
1210
self._raw_conflicts.append(('text conflict', trans_id))
1211
name = self.tt.final_name(trans_id)
1212
parent_id = self.tt.final_parent(trans_id)
1213
file_group = self._dump_conflicts(name, parent_id, file_id,
1214
this_lines, base_lines,
1216
file_group.append(trans_id)
1218
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1219
base_lines=None, other_lines=None, set_version=False,
1221
"""Emit conflict files.
1222
If this_lines, base_lines, or other_lines are omitted, they will be
1223
determined automatically. If set_version is true, the .OTHER, .THIS
1224
or .BASE (in that order) will be created as versioned files.
1226
data = [('OTHER', self.other_tree, other_lines),
1227
('THIS', self.this_tree, this_lines)]
1229
data.append(('BASE', self.base_tree, base_lines))
1232
for suffix, tree, lines in data:
1234
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1236
file_group.append(trans_id)
1237
if set_version and not versioned:
1238
self.tt.version_file(file_id, trans_id)
1242
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1244
"""Emit a single conflict file."""
1245
name = name + '.' + suffix
1246
trans_id = self.tt.create_path(name, parent_id)
1247
create_from_tree(self.tt, trans_id, tree, file_id, lines)
1250
def merge_executable(self, file_id, file_status):
1251
"""Perform a merge on the execute bit."""
1252
executable = [self.executable(t, file_id) for t in (self.base_tree,
1253
self.other_tree, self.this_tree)]
1254
self._merge_executable(file_id, executable, file_status,
1255
resolver=self._three_way)
1257
def _merge_executable(self, file_id, executable, file_status,
1259
"""Perform a merge on the execute bit."""
1260
base_executable, other_executable, this_executable = executable
1261
if file_status == "deleted":
1263
winner = resolver(*executable)
1264
if winner == "conflict":
1265
# There must be a None in here, if we have a conflict, but we
1266
# need executability since file status was not deleted.
1267
if self.executable(self.other_tree, file_id) is None:
1271
if winner == 'this' and file_status != "modified":
1273
trans_id = self.tt.trans_id_file_id(file_id)
1275
if self.tt.final_kind(trans_id) != "file":
1279
if winner == "this":
1280
executability = this_executable
1282
if file_id in self.other_tree:
1283
executability = other_executable
1284
elif file_id in self.this_tree:
1285
executability = this_executable
1286
elif file_id in self.base_tree:
1287
executability = base_executable
1288
if executability is not None:
1289
trans_id = self.tt.trans_id_file_id(file_id)
1290
self.tt.set_executability(executability, trans_id)
1292
def cook_conflicts(self, fs_conflicts):
1293
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1294
from conflicts import Conflict
1296
self.cooked_conflicts.extend(cook_conflicts(fs_conflicts, self.tt))
1297
fp = FinalPaths(self.tt)
1298
for conflict in self._raw_conflicts:
1299
conflict_type = conflict[0]
1300
if conflict_type in ('name conflict', 'parent conflict'):
1301
trans_id = conflict[1]
1302
conflict_args = conflict[2:]
1303
if trans_id not in name_conflicts:
1304
name_conflicts[trans_id] = {}
1305
unique_add(name_conflicts[trans_id], conflict_type,
1307
if conflict_type == 'contents conflict':
1308
for trans_id in conflict[1]:
1309
file_id = self.tt.final_file_id(trans_id)
1310
if file_id is not None:
1312
path = fp.get_path(trans_id)
1313
for suffix in ('.BASE', '.THIS', '.OTHER'):
1314
if path.endswith(suffix):
1315
path = path[:-len(suffix)]
1317
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
1318
self.cooked_conflicts.append(c)
1319
if conflict_type == 'text conflict':
1320
trans_id = conflict[1]
1321
path = fp.get_path(trans_id)
1322
file_id = self.tt.final_file_id(trans_id)
1323
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
1324
self.cooked_conflicts.append(c)
1326
for trans_id, conflicts in name_conflicts.iteritems():
1328
this_parent, other_parent = conflicts['parent conflict']
1329
if this_parent == other_parent:
1330
raise AssertionError()
1332
this_parent = other_parent = \
1333
self.tt.final_file_id(self.tt.final_parent(trans_id))
1335
this_name, other_name = conflicts['name conflict']
1336
if this_name == other_name:
1337
raise AssertionError()
1339
this_name = other_name = self.tt.final_name(trans_id)
1340
other_path = fp.get_path(trans_id)
1341
if this_parent is not None and this_name is not None:
1342
this_parent_path = \
1343
fp.get_path(self.tt.trans_id_file_id(this_parent))
1344
this_path = pathjoin(this_parent_path, this_name)
1346
this_path = "<deleted>"
1347
file_id = self.tt.final_file_id(trans_id)
1348
c = Conflict.factory('path conflict', path=this_path,
1349
conflict_path=other_path, file_id=file_id)
1350
self.cooked_conflicts.append(c)
1351
self.cooked_conflicts.sort(key=Conflict.sort_key)
1354
class WeaveMerger(Merge3Merger):
1355
"""Three-way tree merger, text weave merger."""
1356
supports_reprocess = True
1357
supports_show_base = False
1358
supports_reverse_cherrypick = False
1359
history_based = True
1361
def _merged_lines(self, file_id):
1362
"""Generate the merged lines.
1363
There is no distinction between lines that are meant to contain <<<<<<<
1367
base = self.base_tree
1370
plan = self.this_tree.plan_file_merge(file_id, self.other_tree,
1372
if 'merge' in debug.debug_flags:
1374
trans_id = self.tt.trans_id_file_id(file_id)
1375
name = self.tt.final_name(trans_id) + '.plan'
1376
contents = ('%10s|%s' % l for l in plan)
1377
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1378
textmerge = PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1379
'>>>>>>> MERGE-SOURCE\n')
1380
return textmerge.merge_lines(self.reprocess)
1382
def text_merge(self, file_id, trans_id):
1383
"""Perform a (weave) text merge for a given file and file-id.
1384
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1385
and a conflict will be noted.
1387
lines, conflicts = self._merged_lines(file_id)
1389
# Note we're checking whether the OUTPUT is binary in this case,
1390
# because we don't want to get into weave merge guts.
1391
check_text_lines(lines)
1392
self.tt.create_file(lines, trans_id)
1394
self._raw_conflicts.append(('text conflict', trans_id))
1395
name = self.tt.final_name(trans_id)
1396
parent_id = self.tt.final_parent(trans_id)
1397
file_group = self._dump_conflicts(name, parent_id, file_id,
1399
file_group.append(trans_id)
1402
class LCAMerger(WeaveMerger):
1404
def _merged_lines(self, file_id):
1405
"""Generate the merged lines.
1406
There is no distinction between lines that are meant to contain <<<<<<<
1410
base = self.base_tree
1413
plan = self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1415
if 'merge' in debug.debug_flags:
1417
trans_id = self.tt.trans_id_file_id(file_id)
1418
name = self.tt.final_name(trans_id) + '.plan'
1419
contents = ('%10s|%s' % l for l in plan)
1420
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1421
textmerge = PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1422
'>>>>>>> MERGE-SOURCE\n')
1423
return textmerge.merge_lines(self.reprocess)
1426
class Diff3Merger(Merge3Merger):
1427
"""Three-way merger using external diff3 for text merging"""
1429
def dump_file(self, temp_dir, name, tree, file_id):
1430
out_path = pathjoin(temp_dir, name)
1431
out_file = open(out_path, "wb")
1433
in_file = tree.get_file(file_id)
1434
for line in in_file:
1435
out_file.write(line)
1440
def text_merge(self, file_id, trans_id):
1441
"""Perform a diff3 merge using a specified file-id and trans-id.
1442
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1443
will be dumped, and a will be conflict noted.
1446
temp_dir = osutils.mkdtemp(prefix="bzr-")
1448
new_file = pathjoin(temp_dir, "new")
1449
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1450
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1451
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1452
status = bzrlib.patch.diff3(new_file, this, base, other)
1453
if status not in (0, 1):
1454
raise BzrError("Unhandled diff3 exit code")
1455
f = open(new_file, 'rb')
1457
self.tt.create_file(f, trans_id)
1461
name = self.tt.final_name(trans_id)
1462
parent_id = self.tt.final_parent(trans_id)
1463
self._dump_conflicts(name, parent_id, file_id)
1464
self._raw_conflicts.append(('text conflict', trans_id))
1466
osutils.rmtree(temp_dir)
1469
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1471
merge_type=Merge3Merger,
1472
interesting_ids=None,
1476
interesting_files=None,
1479
change_reporter=None):
1480
"""Primary interface for merging.
1482
typical use is probably
1483
'merge_inner(branch, branch.get_revision_tree(other_revision),
1484
branch.get_revision_tree(base_revision))'
1486
if this_tree is None:
1487
raise BzrError("bzrlib.merge.merge_inner requires a this_tree "
1488
"parameter as of bzrlib version 0.8.")
1489
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1490
pb=pb, change_reporter=change_reporter)
1491
merger.backup_files = backup_files
1492
merger.merge_type = merge_type
1493
merger.interesting_ids = interesting_ids
1494
merger.ignore_zero = ignore_zero
1495
if interesting_files:
1497
raise ValueError('Only supply interesting_ids'
1498
' or interesting_files')
1499
merger.interesting_files = interesting_files
1500
merger.show_base = show_base
1501
merger.reprocess = reprocess
1502
merger.other_rev_id = other_rev_id
1503
merger.other_basis = other_rev_id
1504
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1505
if get_revision_id is None:
1506
get_revision_id = base_tree.last_revision
1507
merger.set_base_revision(get_revision_id(), this_branch)
1508
return merger.do_merge()
1510
def get_merge_type_registry():
1511
"""Merge type registry is in bzrlib.option to avoid circular imports.
1513
This method provides a sanctioned way to retrieve it.
1515
from bzrlib import option
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
