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
interesting_revision_ids = [self.base_rev_id]
394
interesting_revision_ids.extend(lcas)
395
interesting_trees = dict((t.get_revision_id(), t)
396
for t in self.this_branch.repository.revision_trees(
397
interesting_revision_ids))
398
self._cached_trees.update(interesting_trees)
399
self.base_tree = interesting_trees.pop(self.base_rev_id)
400
sorted_lca_keys = self.revision_graph.find_merge_order(
402
self._lca_trees = [interesting_trees[key]
403
for key in sorted_lca_keys]
405
self.base_tree = self.revision_tree(self.base_rev_id)
406
self.base_is_ancestor = True
407
self.base_is_other_ancestor = True
409
def set_base(self, base_revision):
410
"""Set the base revision to use for the merge.
412
:param base_revision: A 2-list containing a path and revision number.
414
mutter("doing merge() with no base_revision specified")
415
if base_revision == [None, None]:
418
base_branch, self.base_tree = self._get_tree(base_revision)
419
if base_revision[1] == -1:
420
self.base_rev_id = base_branch.last_revision()
421
elif base_revision[1] is None:
422
self.base_rev_id = _mod_revision.NULL_REVISION
424
self.base_rev_id = _mod_revision.ensure_null(
425
base_branch.get_rev_id(base_revision[1]))
426
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
428
def make_merger(self):
429
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
430
'other_tree': self.other_tree,
431
'interesting_ids': self.interesting_ids,
432
'interesting_files': self.interesting_files,
435
if self.merge_type.requires_base:
436
kwargs['base_tree'] = self.base_tree
437
if self.merge_type.supports_reprocess:
438
kwargs['reprocess'] = self.reprocess
440
raise BzrError("Conflict reduction is not supported for merge"
441
" type %s." % self.merge_type)
442
if self.merge_type.supports_show_base:
443
kwargs['show_base'] = self.show_base
445
raise BzrError("Showing base is not supported for this"
446
" merge type. %s" % self.merge_type)
447
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
448
and not self.base_is_other_ancestor):
449
raise errors.CannotReverseCherrypick()
450
if self.merge_type.supports_cherrypick:
451
kwargs['cherrypick'] = (not self.base_is_ancestor or
452
not self.base_is_other_ancestor)
453
if self._is_criss_cross and getattr(self.merge_type,
454
'supports_lca_trees', False):
455
kwargs['lca_trees'] = self._lca_trees
456
return self.merge_type(pb=self._pb,
457
change_reporter=self.change_reporter,
460
def _do_merge_to(self, merge):
462
if self.recurse == 'down':
463
for relpath, file_id in self.this_tree.iter_references():
464
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
465
other_revision = self.other_tree.get_reference_revision(
467
if other_revision == sub_tree.last_revision():
469
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
470
sub_merge.merge_type = self.merge_type
471
other_branch = self.other_branch.reference_parent(file_id, relpath)
472
sub_merge.set_other_revision(other_revision, other_branch)
473
base_revision = self.base_tree.get_reference_revision(file_id)
474
sub_merge.base_tree = \
475
sub_tree.branch.repository.revision_tree(base_revision)
476
sub_merge.base_rev_id = base_revision
480
self.this_tree.lock_tree_write()
482
if self.base_tree is not None:
483
self.base_tree.lock_read()
485
if self.other_tree is not None:
486
self.other_tree.lock_read()
488
merge = self.make_merger()
489
self._do_merge_to(merge)
491
if self.other_tree is not None:
492
self.other_tree.unlock()
494
if self.base_tree is not None:
495
self.base_tree.unlock()
497
self.this_tree.unlock()
498
if len(merge.cooked_conflicts) == 0:
499
if not self.ignore_zero and not is_quiet():
500
note("All changes applied successfully.")
502
note("%d conflicts encountered." % len(merge.cooked_conflicts))
504
return len(merge.cooked_conflicts)
507
class _InventoryNoneEntry(object):
508
"""This represents an inventory entry which *isn't there*.
510
It simplifies the merging logic if we always have an InventoryEntry, even
511
if it isn't actually present
518
symlink_target = None
521
_none_entry = _InventoryNoneEntry()
524
class Merge3Merger(object):
525
"""Three-way merger that uses the merge3 text merger"""
527
supports_reprocess = True
528
supports_show_base = True
529
history_based = False
530
supports_cherrypick = True
531
supports_reverse_cherrypick = True
532
winner_idx = {"this": 2, "other": 1, "conflict": 1}
533
supports_lca_trees = True
535
def __init__(self, working_tree, this_tree, base_tree, other_tree,
536
interesting_ids=None, reprocess=False, show_base=False,
537
pb=DummyProgress(), pp=None, change_reporter=None,
538
interesting_files=None, do_merge=True,
539
cherrypick=False, lca_trees=None):
540
"""Initialize the merger object and perform the merge.
542
:param working_tree: The working tree to apply the merge to
543
:param this_tree: The local tree in the merge operation
544
:param base_tree: The common tree in the merge operation
545
:param other_tree: The other other tree to merge changes from
546
:param interesting_ids: The file_ids of files that should be
547
participate in the merge. May not be combined with
549
:param: reprocess If True, perform conflict-reduction processing.
550
:param show_base: If True, show the base revision in text conflicts.
551
(incompatible with reprocess)
552
:param pb: A Progress bar
553
:param pp: A ProgressPhase object
554
:param change_reporter: An object that should report changes made
555
:param interesting_files: The tree-relative paths of files that should
556
participate in the merge. If these paths refer to directories,
557
the contents of those directories will also be included. May not
558
be combined with interesting_ids. If neither interesting_files nor
559
interesting_ids is specified, all files may participate in the
561
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
562
if the ancestry was found to include a criss-cross merge.
563
Otherwise should be None.
565
object.__init__(self)
566
if interesting_files is not None and interesting_ids is not None:
568
'specify either interesting_ids or interesting_files')
569
self.interesting_ids = interesting_ids
570
self.interesting_files = interesting_files
571
self.this_tree = working_tree
572
self.base_tree = base_tree
573
self.other_tree = other_tree
574
self._raw_conflicts = []
575
self.cooked_conflicts = []
576
self.reprocess = reprocess
577
self.show_base = show_base
578
self._lca_trees = lca_trees
579
# Uncommenting this will change the default algorithm to always use
580
# _entries_lca. This can be useful for running the test suite and
581
# making sure we haven't missed any corner cases.
582
# if lca_trees is None:
583
# self._lca_trees = [self.base_tree]
586
self.change_reporter = change_reporter
587
self.cherrypick = cherrypick
589
self.pp = ProgressPhase("Merge phase", 3, self.pb)
594
self.this_tree.lock_tree_write()
595
self.base_tree.lock_read()
596
self.other_tree.lock_read()
597
self.tt = TreeTransform(self.this_tree, self.pb)
600
self._compute_transform()
602
results = self.tt.apply(no_conflicts=True)
603
self.write_modified(results)
605
self.this_tree.add_conflicts(self.cooked_conflicts)
606
except UnsupportedOperation:
610
self.other_tree.unlock()
611
self.base_tree.unlock()
612
self.this_tree.unlock()
615
def make_preview_transform(self):
616
self.base_tree.lock_read()
617
self.other_tree.lock_read()
618
self.tt = TransformPreview(self.this_tree)
621
self._compute_transform()
624
self.other_tree.unlock()
625
self.base_tree.unlock()
629
def _compute_transform(self):
630
if self._lca_trees is None:
631
entries = self._entries3()
632
resolver = self._three_way
634
entries = self._entries_lca()
635
resolver = self._lca_multi_way
636
child_pb = ui.ui_factory.nested_progress_bar()
638
for num, (file_id, changed, parents3, names3,
639
executable3) in enumerate(entries):
640
child_pb.update('Preparing file merge', num, len(entries))
641
self._merge_names(file_id, parents3, names3, resolver=resolver)
643
file_status = self.merge_contents(file_id)
645
file_status = 'unmodified'
646
self._merge_executable(file_id,
647
executable3, file_status, resolver=resolver)
652
child_pb = ui.ui_factory.nested_progress_bar()
654
fs_conflicts = resolve_conflicts(self.tt, child_pb,
655
lambda t, c: conflict_pass(t, c, self.other_tree))
658
if self.change_reporter is not None:
659
from bzrlib import delta
660
delta.report_changes(
661
self.tt.iter_changes(), self.change_reporter)
662
self.cook_conflicts(fs_conflicts)
663
for conflict in self.cooked_conflicts:
667
"""Gather data about files modified between three trees.
669
Return a list of tuples of file_id, changed, parents3, names3,
670
executable3. changed is a boolean indicating whether the file contents
671
or kind were changed. parents3 is a tuple of parent ids for base,
672
other and this. names3 is a tuple of names for base, other and this.
673
executable3 is a tuple of execute-bit values for base, other and this.
676
iterator = self.other_tree.iter_changes(self.base_tree,
677
include_unchanged=True, specific_files=self.interesting_files,
678
extra_trees=[self.this_tree])
679
this_entries = dict((e.file_id, e) for p, e in
680
self.this_tree.iter_entries_by_dir(
681
self.interesting_ids))
682
for (file_id, paths, changed, versioned, parents, names, kind,
683
executable) in iterator:
684
if (self.interesting_ids is not None and
685
file_id not in self.interesting_ids):
687
entry = this_entries.get(file_id)
688
if entry is not None:
689
this_name = entry.name
690
this_parent = entry.parent_id
691
this_executable = entry.executable
695
this_executable = None
696
parents3 = parents + (this_parent,)
697
names3 = names + (this_name,)
698
executable3 = executable + (this_executable,)
699
result.append((file_id, changed, parents3, names3, executable3))
702
def _entries_lca(self):
703
"""Gather data about files modified between multiple trees.
705
This compares OTHER versus all LCA trees, and for interesting entries,
706
it then compares with THIS and BASE.
708
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
709
:return: [(file_id, changed, parents, names, executable)]
710
file_id Simple file_id of the entry
711
changed Boolean, True if the kind or contents changed
713
parents ((base, [parent_id, in, lcas]), parent_id_other,
715
names ((base, [name, in, lcas]), name_in_other, name_in_this)
716
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
718
if self.interesting_files is not None:
719
lookup_trees = [self.this_tree, self.base_tree]
720
lookup_trees.extend(self._lca_trees)
721
# I think we should include the lca trees as well
722
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
725
interesting_ids = self.interesting_ids
727
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
729
base_inventory = self.base_tree.inventory
730
this_inventory = self.this_tree.inventory
731
for path, file_id, other_ie, lca_values in walker.iter_all():
732
# Is this modified at all from any of the other trees?
734
other_ie = _none_entry
735
if interesting_ids is not None and file_id not in interesting_ids:
738
# If other_revision is found in any of the lcas, that means this
739
# node is uninteresting. This is because when merging, if there are
740
# multiple heads(), we have to create a new node. So if we didn't,
741
# we know that the ancestry is linear, and that OTHER did not
743
# See doc/developers/lca_merge_resolution.txt for details
744
other_revision = other_ie.revision
745
if other_revision is not None:
746
# We can't use this shortcut when other_revision is None,
747
# because it may be None because things are WorkingTrees, and
748
# not because it is *actually* None.
749
is_unmodified = False
750
for lca_path, ie in lca_values:
751
if ie is not None and ie.revision == other_revision:
758
for lca_path, lca_ie in lca_values:
760
lca_entries.append(_none_entry)
762
lca_entries.append(lca_ie)
764
if file_id in base_inventory:
765
base_ie = base_inventory[file_id]
767
base_ie = _none_entry
769
if file_id in this_inventory:
770
this_ie = this_inventory[file_id]
772
this_ie = _none_entry
778
for lca_ie in lca_entries:
779
lca_kinds.append(lca_ie.kind)
780
lca_parent_ids.append(lca_ie.parent_id)
781
lca_names.append(lca_ie.name)
782
lca_executable.append(lca_ie.executable)
784
kind_winner = self._lca_multi_way(
785
(base_ie.kind, lca_kinds),
786
other_ie.kind, this_ie.kind)
787
parent_id_winner = self._lca_multi_way(
788
(base_ie.parent_id, lca_parent_ids),
789
other_ie.parent_id, this_ie.parent_id)
790
name_winner = self._lca_multi_way(
791
(base_ie.name, lca_names),
792
other_ie.name, this_ie.name)
794
content_changed = True
795
if kind_winner == 'this':
796
# No kind change in OTHER, see if there are *any* changes
797
if other_ie.kind == None:
798
# No content and 'this' wins the kind, so skip this?
801
elif other_ie.kind == 'directory':
802
if parent_id_winner == 'this' and name_winner == 'this':
803
# No change for this directory in OTHER, skip
805
content_changed = False
806
elif other_ie.kind == 'file':
807
def get_sha1(ie, tree):
808
if ie.kind != 'file':
810
return tree.get_file_sha1(file_id)
811
base_sha1 = get_sha1(base_ie, self.base_tree)
812
lca_sha1s = [get_sha1(ie, tree) for ie, tree
813
in zip(lca_entries, self._lca_trees)]
814
this_sha1 = get_sha1(this_ie, self.this_tree)
815
other_sha1 = get_sha1(other_ie, self.other_tree)
816
sha1_winner = self._lca_multi_way(
817
(base_sha1, lca_sha1s), other_sha1, this_sha1,
818
allow_overriding_lca=False)
819
exec_winner = self._lca_multi_way(
820
(base_ie.executable, lca_executable),
821
other_ie.executable, this_ie.executable)
822
if (parent_id_winner == 'this' and name_winner == 'this'
823
and sha1_winner == 'this' and exec_winner == 'this'):
824
# No kind, parent, name, exec, or content change for
825
# OTHER, so this node is not considered interesting
827
if sha1_winner == 'this':
828
content_changed = False
829
elif other_ie.kind == 'symlink':
830
def get_target(ie, tree):
831
if ie.kind != 'symlink':
833
return tree.get_symlink_target(file_id)
834
base_target = get_target(base_ie, self.base_tree)
835
lca_targets = [get_target(ie, tree) for ie, tree
836
in zip(lca_entries, self._lca_trees)]
837
this_target = get_target(this_ie, self.this_tree)
838
other_target = get_target(other_ie, self.other_tree)
839
target_winner = self._lca_multi_way(
840
(base_target, lca_targets),
841
other_target, this_target)
842
if (parent_id_winner == 'this' and name_winner == 'this'
843
and target_winner == 'this'):
844
# No kind, parent, name, or symlink target change
847
if target_winner == 'this':
848
content_changed = False
849
elif other_ie.kind == 'tree-reference':
850
# The 'changed' information seems to be handled at a higher
851
# level. At least, _entries3 returns False for content
852
# changed, even when at a new revision_id.
853
content_changed = False
854
if (parent_id_winner == 'this' and name_winner == 'this'):
855
# Nothing interesting
858
raise AssertionError('unhandled kind: %s' % other_ie.kind)
859
# XXX: We need to handle kind == 'symlink'
861
# If we have gotten this far, that means something has changed
862
result.append((file_id, content_changed,
863
((base_ie.parent_id, lca_parent_ids),
864
other_ie.parent_id, this_ie.parent_id),
865
((base_ie.name, lca_names),
866
other_ie.name, this_ie.name),
867
((base_ie.executable, lca_executable),
868
other_ie.executable, this_ie.executable)
875
self.tt.final_kind(self.tt.root)
877
self.tt.cancel_deletion(self.tt.root)
878
if self.tt.final_file_id(self.tt.root) is None:
879
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
881
other_root_file_id = self.other_tree.get_root_id()
882
if other_root_file_id is None:
884
other_root = self.tt.trans_id_file_id(other_root_file_id)
885
if other_root == self.tt.root:
888
self.tt.final_kind(other_root)
891
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
892
# the other tree's root is a non-root in the current tree
894
self.reparent_children(self.other_tree.inventory.root, self.tt.root)
895
self.tt.cancel_creation(other_root)
896
self.tt.cancel_versioning(other_root)
898
def reparent_children(self, ie, target):
899
for thing, child in ie.children.iteritems():
900
trans_id = self.tt.trans_id_file_id(child.file_id)
901
self.tt.adjust_path(self.tt.final_name(trans_id), target, trans_id)
903
def write_modified(self, results):
905
for path in results.modified_paths:
906
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
909
hash = self.this_tree.get_file_sha1(file_id)
912
modified_hashes[file_id] = hash
913
self.this_tree.set_merge_modified(modified_hashes)
916
def parent(entry, file_id):
917
"""Determine the parent for a file_id (used as a key method)"""
920
return entry.parent_id
923
def name(entry, file_id):
924
"""Determine the name for a file_id (used as a key method)"""
930
def contents_sha1(tree, file_id):
931
"""Determine the sha1 of the file contents (used as a key method)."""
932
if file_id not in tree:
934
return tree.get_file_sha1(file_id)
937
def executable(tree, file_id):
938
"""Determine the executability of a file-id (used as a key method)."""
939
if file_id not in tree:
941
if tree.kind(file_id) != "file":
943
return tree.is_executable(file_id)
946
def kind(tree, file_id):
947
"""Determine the kind of a file-id (used as a key method)."""
948
if file_id not in tree:
950
return tree.kind(file_id)
953
def _three_way(base, other, this):
954
#if base == other, either they all agree, or only THIS has changed.
957
elif this not in (base, other):
959
# "Ambiguous clean merge" -- both sides have made the same change.
962
# this == base: only other has changed.
967
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
968
"""Consider LCAs when determining whether a change has occurred.
970
If LCAS are all identical, this is the same as a _three_way comparison.
972
:param bases: value in (BASE, [LCAS])
973
:param other: value in OTHER
974
:param this: value in THIS
975
:param allow_overriding_lca: If there is more than one unique lca
976
value, allow OTHER to override THIS if it has a new value, and
977
THIS only has an lca value, or vice versa. This is appropriate for
978
truly scalar values, not as much for non-scalars.
979
:return: 'this', 'other', or 'conflict' depending on whether an entry
982
# See doc/developers/lca_merge_resolution.txt for details about this
985
# Either Ambiguously clean, or nothing was actually changed. We
988
base_val, lca_vals = bases
989
# Remove 'base_val' from the lca_vals, because it is not interesting
990
filtered_lca_vals = [lca_val for lca_val in lca_vals
991
if lca_val != base_val]
992
if len(filtered_lca_vals) == 0:
993
return Merge3Merger._three_way(base_val, other, this)
995
unique_lca_vals = set(filtered_lca_vals)
996
if len(unique_lca_vals) == 1:
997
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
999
if allow_overriding_lca:
1000
if other in unique_lca_vals:
1001
if this in unique_lca_vals:
1002
# Each side picked a different lca, conflict
1005
# This has a value which supersedes both lca values, and
1006
# other only has an lca value
1008
elif this in unique_lca_vals:
1009
# OTHER has a value which supersedes both lca values, and this
1010
# only has an lca value
1013
# At this point, the lcas disagree, and the tips disagree
1017
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1018
"""Do a three-way test on a scalar.
1019
Return "this", "other" or "conflict", depending whether a value wins.
1021
key_base = key(base_tree, file_id)
1022
key_other = key(other_tree, file_id)
1023
#if base == other, either they all agree, or only THIS has changed.
1024
if key_base == key_other:
1026
key_this = key(this_tree, file_id)
1027
# "Ambiguous clean merge"
1028
if key_this == key_other:
1030
elif key_this == key_base:
1035
def merge_names(self, file_id):
1036
def get_entry(tree):
1037
if file_id in tree.inventory:
1038
return tree.inventory[file_id]
1041
this_entry = get_entry(self.this_tree)
1042
other_entry = get_entry(self.other_tree)
1043
base_entry = get_entry(self.base_tree)
1044
entries = (base_entry, other_entry, this_entry)
1047
for entry in entries:
1050
parents.append(None)
1052
names.append(entry.name)
1053
parents.append(entry.parent_id)
1054
return self._merge_names(file_id, parents, names,
1055
resolver=self._three_way)
1057
def _merge_names(self, file_id, parents, names, resolver):
1058
"""Perform a merge on file_id names and parents"""
1059
base_name, other_name, this_name = names
1060
base_parent, other_parent, this_parent = parents
1062
name_winner = resolver(*names)
1064
parent_id_winner = resolver(*parents)
1065
if this_name is None:
1066
if name_winner == "this":
1067
name_winner = "other"
1068
if parent_id_winner == "this":
1069
parent_id_winner = "other"
1070
if name_winner == "this" and parent_id_winner == "this":
1072
if name_winner == "conflict":
1073
trans_id = self.tt.trans_id_file_id(file_id)
1074
self._raw_conflicts.append(('name conflict', trans_id,
1075
this_name, other_name))
1076
if parent_id_winner == "conflict":
1077
trans_id = self.tt.trans_id_file_id(file_id)
1078
self._raw_conflicts.append(('parent conflict', trans_id,
1079
this_parent, other_parent))
1080
if other_name is None:
1081
# it doesn't matter whether the result was 'other' or
1082
# 'conflict'-- if there's no 'other', we leave it alone.
1084
# if we get here, name_winner and parent_winner are set to safe values.
1085
trans_id = self.tt.trans_id_file_id(file_id)
1086
parent_id = parents[self.winner_idx[parent_id_winner]]
1087
if parent_id is not None:
1088
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1089
self.tt.adjust_path(names[self.winner_idx[name_winner]],
1090
parent_trans_id, trans_id)
1092
def merge_contents(self, file_id):
1093
"""Performa a merge on file_id contents."""
1094
def contents_pair(tree):
1095
if file_id not in tree:
1097
kind = tree.kind(file_id)
1099
contents = tree.get_file_sha1(file_id)
1100
elif kind == "symlink":
1101
contents = tree.get_symlink_target(file_id)
1104
return kind, contents
1106
def contents_conflict():
1107
trans_id = self.tt.trans_id_file_id(file_id)
1108
name = self.tt.final_name(trans_id)
1109
parent_id = self.tt.final_parent(trans_id)
1110
if file_id in self.this_tree.inventory:
1111
self.tt.unversion_file(trans_id)
1112
if file_id in self.this_tree:
1113
self.tt.delete_contents(trans_id)
1114
file_group = self._dump_conflicts(name, parent_id, file_id,
1116
self._raw_conflicts.append(('contents conflict', file_group))
1118
# See SPOT run. run, SPOT, run.
1119
# So we're not QUITE repeating ourselves; we do tricky things with
1121
base_pair = contents_pair(self.base_tree)
1122
other_pair = contents_pair(self.other_tree)
1123
if base_pair == other_pair:
1124
# OTHER introduced no changes
1126
this_pair = contents_pair(self.this_tree)
1127
if this_pair == other_pair:
1128
# THIS and OTHER introduced the same changes
1131
trans_id = self.tt.trans_id_file_id(file_id)
1132
if this_pair == base_pair:
1133
# only OTHER introduced changes
1134
if file_id in self.this_tree:
1135
# Remove any existing contents
1136
self.tt.delete_contents(trans_id)
1137
if file_id in self.other_tree:
1138
# OTHER changed the file
1139
create_from_tree(self.tt, trans_id,
1140
self.other_tree, file_id)
1141
if file_id not in self.this_tree:
1142
self.tt.version_file(file_id, trans_id)
1144
elif file_id in self.this_tree.inventory:
1145
# OTHER deleted the file
1146
self.tt.unversion_file(trans_id)
1148
#BOTH THIS and OTHER introduced changes; scalar conflict
1149
elif this_pair[0] == "file" and other_pair[0] == "file":
1150
# THIS and OTHER are both files, so text merge. Either
1151
# BASE is a file, or both converted to files, so at least we
1152
# have agreement that output should be a file.
1154
self.text_merge(file_id, trans_id)
1156
return contents_conflict()
1157
if file_id not in self.this_tree:
1158
self.tt.version_file(file_id, trans_id)
1160
self.tt.tree_kind(trans_id)
1161
self.tt.delete_contents(trans_id)
1166
# Scalar conflict, can't text merge. Dump conflicts
1167
return contents_conflict()
1169
def get_lines(self, tree, file_id):
1170
"""Return the lines in a file, or an empty list."""
1172
return tree.get_file(file_id).readlines()
1176
def text_merge(self, file_id, trans_id):
1177
"""Perform a three-way text merge on a file_id"""
1178
# it's possible that we got here with base as a different type.
1179
# if so, we just want two-way text conflicts.
1180
if file_id in self.base_tree and \
1181
self.base_tree.kind(file_id) == "file":
1182
base_lines = self.get_lines(self.base_tree, file_id)
1185
other_lines = self.get_lines(self.other_tree, file_id)
1186
this_lines = self.get_lines(self.this_tree, file_id)
1187
m3 = Merge3(base_lines, this_lines, other_lines,
1188
is_cherrypick=self.cherrypick)
1189
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1190
if self.show_base is True:
1191
base_marker = '|' * 7
1195
def iter_merge3(retval):
1196
retval["text_conflicts"] = False
1197
for line in m3.merge_lines(name_a = "TREE",
1198
name_b = "MERGE-SOURCE",
1199
name_base = "BASE-REVISION",
1200
start_marker=start_marker,
1201
base_marker=base_marker,
1202
reprocess=self.reprocess):
1203
if line.startswith(start_marker):
1204
retval["text_conflicts"] = True
1205
yield line.replace(start_marker, '<' * 7)
1209
merge3_iterator = iter_merge3(retval)
1210
self.tt.create_file(merge3_iterator, trans_id)
1211
if retval["text_conflicts"] is True:
1212
self._raw_conflicts.append(('text conflict', trans_id))
1213
name = self.tt.final_name(trans_id)
1214
parent_id = self.tt.final_parent(trans_id)
1215
file_group = self._dump_conflicts(name, parent_id, file_id,
1216
this_lines, base_lines,
1218
file_group.append(trans_id)
1220
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1221
base_lines=None, other_lines=None, set_version=False,
1223
"""Emit conflict files.
1224
If this_lines, base_lines, or other_lines are omitted, they will be
1225
determined automatically. If set_version is true, the .OTHER, .THIS
1226
or .BASE (in that order) will be created as versioned files.
1228
data = [('OTHER', self.other_tree, other_lines),
1229
('THIS', self.this_tree, this_lines)]
1231
data.append(('BASE', self.base_tree, base_lines))
1234
for suffix, tree, lines in data:
1236
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1238
file_group.append(trans_id)
1239
if set_version and not versioned:
1240
self.tt.version_file(file_id, trans_id)
1244
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1246
"""Emit a single conflict file."""
1247
name = name + '.' + suffix
1248
trans_id = self.tt.create_path(name, parent_id)
1249
create_from_tree(self.tt, trans_id, tree, file_id, lines)
1252
def merge_executable(self, file_id, file_status):
1253
"""Perform a merge on the execute bit."""
1254
executable = [self.executable(t, file_id) for t in (self.base_tree,
1255
self.other_tree, self.this_tree)]
1256
self._merge_executable(file_id, executable, file_status,
1257
resolver=self._three_way)
1259
def _merge_executable(self, file_id, executable, file_status,
1261
"""Perform a merge on the execute bit."""
1262
base_executable, other_executable, this_executable = executable
1263
if file_status == "deleted":
1265
winner = resolver(*executable)
1266
if winner == "conflict":
1267
# There must be a None in here, if we have a conflict, but we
1268
# need executability since file status was not deleted.
1269
if self.executable(self.other_tree, file_id) is None:
1273
if winner == 'this' and file_status != "modified":
1275
trans_id = self.tt.trans_id_file_id(file_id)
1277
if self.tt.final_kind(trans_id) != "file":
1281
if winner == "this":
1282
executability = this_executable
1284
if file_id in self.other_tree:
1285
executability = other_executable
1286
elif file_id in self.this_tree:
1287
executability = this_executable
1288
elif file_id in self.base_tree:
1289
executability = base_executable
1290
if executability is not None:
1291
trans_id = self.tt.trans_id_file_id(file_id)
1292
self.tt.set_executability(executability, trans_id)
1294
def cook_conflicts(self, fs_conflicts):
1295
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1296
from conflicts import Conflict
1298
self.cooked_conflicts.extend(cook_conflicts(fs_conflicts, self.tt))
1299
fp = FinalPaths(self.tt)
1300
for conflict in self._raw_conflicts:
1301
conflict_type = conflict[0]
1302
if conflict_type in ('name conflict', 'parent conflict'):
1303
trans_id = conflict[1]
1304
conflict_args = conflict[2:]
1305
if trans_id not in name_conflicts:
1306
name_conflicts[trans_id] = {}
1307
unique_add(name_conflicts[trans_id], conflict_type,
1309
if conflict_type == 'contents conflict':
1310
for trans_id in conflict[1]:
1311
file_id = self.tt.final_file_id(trans_id)
1312
if file_id is not None:
1314
path = fp.get_path(trans_id)
1315
for suffix in ('.BASE', '.THIS', '.OTHER'):
1316
if path.endswith(suffix):
1317
path = path[:-len(suffix)]
1319
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
1320
self.cooked_conflicts.append(c)
1321
if conflict_type == 'text conflict':
1322
trans_id = conflict[1]
1323
path = fp.get_path(trans_id)
1324
file_id = self.tt.final_file_id(trans_id)
1325
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
1326
self.cooked_conflicts.append(c)
1328
for trans_id, conflicts in name_conflicts.iteritems():
1330
this_parent, other_parent = conflicts['parent conflict']
1331
if this_parent == other_parent:
1332
raise AssertionError()
1334
this_parent = other_parent = \
1335
self.tt.final_file_id(self.tt.final_parent(trans_id))
1337
this_name, other_name = conflicts['name conflict']
1338
if this_name == other_name:
1339
raise AssertionError()
1341
this_name = other_name = self.tt.final_name(trans_id)
1342
other_path = fp.get_path(trans_id)
1343
if this_parent is not None and this_name is not None:
1344
this_parent_path = \
1345
fp.get_path(self.tt.trans_id_file_id(this_parent))
1346
this_path = pathjoin(this_parent_path, this_name)
1348
this_path = "<deleted>"
1349
file_id = self.tt.final_file_id(trans_id)
1350
c = Conflict.factory('path conflict', path=this_path,
1351
conflict_path=other_path, file_id=file_id)
1352
self.cooked_conflicts.append(c)
1353
self.cooked_conflicts.sort(key=Conflict.sort_key)
1356
class WeaveMerger(Merge3Merger):
1357
"""Three-way tree merger, text weave merger."""
1358
supports_reprocess = True
1359
supports_show_base = False
1360
supports_reverse_cherrypick = False
1361
history_based = True
1363
def _merged_lines(self, file_id):
1364
"""Generate the merged lines.
1365
There is no distinction between lines that are meant to contain <<<<<<<
1369
base = self.base_tree
1372
plan = self.this_tree.plan_file_merge(file_id, self.other_tree,
1374
if 'merge' in debug.debug_flags:
1376
trans_id = self.tt.trans_id_file_id(file_id)
1377
name = self.tt.final_name(trans_id) + '.plan'
1378
contents = ('%10s|%s' % l for l in plan)
1379
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1380
textmerge = PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1381
'>>>>>>> MERGE-SOURCE\n')
1382
return textmerge.merge_lines(self.reprocess)
1384
def text_merge(self, file_id, trans_id):
1385
"""Perform a (weave) text merge for a given file and file-id.
1386
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1387
and a conflict will be noted.
1389
lines, conflicts = self._merged_lines(file_id)
1391
# Note we're checking whether the OUTPUT is binary in this case,
1392
# because we don't want to get into weave merge guts.
1393
check_text_lines(lines)
1394
self.tt.create_file(lines, trans_id)
1396
self._raw_conflicts.append(('text conflict', trans_id))
1397
name = self.tt.final_name(trans_id)
1398
parent_id = self.tt.final_parent(trans_id)
1399
file_group = self._dump_conflicts(name, parent_id, file_id,
1401
file_group.append(trans_id)
1404
class LCAMerger(WeaveMerger):
1406
def _merged_lines(self, file_id):
1407
"""Generate the merged lines.
1408
There is no distinction between lines that are meant to contain <<<<<<<
1412
base = self.base_tree
1415
plan = self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1417
if 'merge' in debug.debug_flags:
1419
trans_id = self.tt.trans_id_file_id(file_id)
1420
name = self.tt.final_name(trans_id) + '.plan'
1421
contents = ('%10s|%s' % l for l in plan)
1422
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1423
textmerge = PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1424
'>>>>>>> MERGE-SOURCE\n')
1425
return textmerge.merge_lines(self.reprocess)
1428
class Diff3Merger(Merge3Merger):
1429
"""Three-way merger using external diff3 for text merging"""
1431
def dump_file(self, temp_dir, name, tree, file_id):
1432
out_path = pathjoin(temp_dir, name)
1433
out_file = open(out_path, "wb")
1435
in_file = tree.get_file(file_id)
1436
for line in in_file:
1437
out_file.write(line)
1442
def text_merge(self, file_id, trans_id):
1443
"""Perform a diff3 merge using a specified file-id and trans-id.
1444
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1445
will be dumped, and a will be conflict noted.
1448
temp_dir = osutils.mkdtemp(prefix="bzr-")
1450
new_file = pathjoin(temp_dir, "new")
1451
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1452
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1453
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1454
status = bzrlib.patch.diff3(new_file, this, base, other)
1455
if status not in (0, 1):
1456
raise BzrError("Unhandled diff3 exit code")
1457
f = open(new_file, 'rb')
1459
self.tt.create_file(f, trans_id)
1463
name = self.tt.final_name(trans_id)
1464
parent_id = self.tt.final_parent(trans_id)
1465
self._dump_conflicts(name, parent_id, file_id)
1466
self._raw_conflicts.append(('text conflict', trans_id))
1468
osutils.rmtree(temp_dir)
1471
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1473
merge_type=Merge3Merger,
1474
interesting_ids=None,
1478
interesting_files=None,
1481
change_reporter=None):
1482
"""Primary interface for merging.
1484
typical use is probably
1485
'merge_inner(branch, branch.get_revision_tree(other_revision),
1486
branch.get_revision_tree(base_revision))'
1488
if this_tree is None:
1489
raise BzrError("bzrlib.merge.merge_inner requires a this_tree "
1490
"parameter as of bzrlib version 0.8.")
1491
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1492
pb=pb, change_reporter=change_reporter)
1493
merger.backup_files = backup_files
1494
merger.merge_type = merge_type
1495
merger.interesting_ids = interesting_ids
1496
merger.ignore_zero = ignore_zero
1497
if interesting_files:
1499
raise ValueError('Only supply interesting_ids'
1500
' or interesting_files')
1501
merger.interesting_files = interesting_files
1502
merger.show_base = show_base
1503
merger.reprocess = reprocess
1504
merger.other_rev_id = other_rev_id
1505
merger.other_basis = other_rev_id
1506
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1507
if get_revision_id is None:
1508
get_revision_id = base_tree.last_revision
1509
merger.set_base_revision(get_revision_id(), this_branch)
1510
return merger.do_merge()
1512
def get_merge_type_registry():
1513
"""Merge type registry is in bzrlib.option to avoid circular imports.
1515
This method provides a sanctioned way to retrieve it.
1517
from bzrlib import option
1518
return option._merge_type_registry
1521
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1522
def status_a(revision, text):
1523
if revision in ancestors_b:
1524
return 'killed-b', text
1526
return 'new-a', text
1528
def status_b(revision, text):
1529
if revision in ancestors_a:
1530
return 'killed-a', text
1532
return 'new-b', text
1534
plain_a = [t for (a, t) in annotated_a]
1535
plain_b = [t for (a, t) in annotated_b]
1536
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1537
blocks = matcher.get_matching_blocks()
1540
for ai, bi, l in blocks:
1541
# process all mismatched sections
1542
# (last mismatched section is handled because blocks always
1543
# includes a 0-length last block)
1544
for revision, text in annotated_a[a_cur:ai]:
1545
yield status_a(revision, text)
1546
for revision, text in annotated_b[b_cur:bi]:
1547
yield status_b(revision, text)
1548
# and now the matched section
1551
for text_a in plain_a[ai:a_cur]:
1552
yield "unchanged", text_a
1555
class _PlanMergeBase(object):
1557
def __init__(self, a_rev, b_rev, vf, key_prefix):
1560
:param a_rev: Revision-id of one revision to merge
1561
:param b_rev: Revision-id of the other revision to merge
1562
:param vf: A VersionedFiles containing both revisions
1563
:param key_prefix: A prefix for accessing keys in vf, typically
1569
self._last_lines = None
1570
self._last_lines_revision_id = None
1571
self._cached_matching_blocks = {}
1572
self._key_prefix = key_prefix
1573
self._precache_tip_lines()
1575
def _precache_tip_lines(self):
1576
lines = self.get_lines([self.a_rev, self.b_rev])
1577
self.lines_a = lines[self.a_rev]
1578
self.lines_b = lines[self.b_rev]
1580
def get_lines(self, revisions):
1581
"""Get lines for revisions from the backing VersionedFiles.
1583
:raises RevisionNotPresent: on absent texts.
1585
keys = [(self._key_prefix + (rev,)) for rev in revisions]
1587
for record in self.vf.get_record_stream(keys, 'unordered', True):
1588
if record.storage_kind == 'absent':
1589
raise errors.RevisionNotPresent(record.key, self.vf)
1590
result[record.key[-1]] = osutils.split_lines(
1591
record.get_bytes_as('fulltext'))
1594
def plan_merge(self):
1595
"""Generate a 'plan' for merging the two revisions.
1597
This involves comparing their texts and determining the cause of
1598
differences. If text A has a line and text B does not, then either the
1599
line was added to text A, or it was deleted from B. Once the causes
1600
are combined, they are written out in the format described in
1601
VersionedFile.plan_merge
1603
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
1604
unique_a, unique_b = self._unique_lines(blocks)
1605
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
1606
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
1607
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
1609
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
1612
for i, j, n in blocks:
1613
for a_index in range(last_i, i):
1614
if a_index in new_a:
1615
if a_index in killed_b:
1616
yield 'conflicted-a', self.lines_a[a_index]
1618
yield 'new-a', self.lines_a[a_index]
1620
yield 'killed-b', self.lines_a[a_index]
1621
for b_index in range(last_j, j):
1622
if b_index in new_b:
1623
if b_index in killed_a:
1624
yield 'conflicted-b', self.lines_b[b_index]
1626
yield 'new-b', self.lines_b[b_index]
1628
yield 'killed-a', self.lines_b[b_index]
1629
# handle common lines
1630
for a_index in range(i, i+n):
1631
yield 'unchanged', self.lines_a[a_index]
1635
def _get_matching_blocks(self, left_revision, right_revision):
1636
"""Return a description of which sections of two revisions match.
1638
See SequenceMatcher.get_matching_blocks
1640
cached = self._cached_matching_blocks.get((left_revision,
1642
if cached is not None:
1644
if self._last_lines_revision_id == left_revision:
1645
left_lines = self._last_lines
1646
right_lines = self.get_lines([right_revision])[right_revision]
1648
lines = self.get_lines([left_revision, right_revision])
1649
left_lines = lines[left_revision]
1650
right_lines = lines[right_revision]
1651
self._last_lines = right_lines
1652
self._last_lines_revision_id = right_revision
1653
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
1655
return matcher.get_matching_blocks()
1657
def _unique_lines(self, matching_blocks):
1658
"""Analyse matching_blocks to determine which lines are unique
1660
:return: a tuple of (unique_left, unique_right), where the values are
1661
sets of line numbers of unique lines.
1667
for i, j, n in matching_blocks:
1668
unique_left.extend(range(last_i, i))
1669
unique_right.extend(range(last_j, j))
1672
return unique_left, unique_right
1675
def _subtract_plans(old_plan, new_plan):
1676
"""Remove changes from new_plan that came from old_plan.
1678
It is assumed that the difference between the old_plan and new_plan
1679
is their choice of 'b' text.
1681
All lines from new_plan that differ from old_plan are emitted
1682
verbatim. All lines from new_plan that match old_plan but are
1683
not about the 'b' revision are emitted verbatim.
1685
Lines that match and are about the 'b' revision are the lines we
1686
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
1687
is skipped entirely.
1689
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
1692
for i, j, n in matcher.get_matching_blocks():
1693
for jj in range(last_j, j):
1695
for jj in range(j, j+n):
1696
plan_line = new_plan[jj]
1697
if plan_line[0] == 'new-b':
1699
elif plan_line[0] == 'killed-b':
1700
yield 'unchanged', plan_line[1]
1706
class _PlanMerge(_PlanMergeBase):
1707
"""Plan an annotate merge using on-the-fly annotation"""
1709
def __init__(self, a_rev, b_rev, vf, key_prefix):
1710
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
1711
self.a_key = self._key_prefix + (self.a_rev,)
1712
self.b_key = self._key_prefix + (self.b_rev,)
1713
self.graph = Graph(self.vf)
1714
heads = self.graph.heads((self.a_key, self.b_key))
1716
# one side dominates, so we can just return its values, yay for
1718
# Ideally we would know that before we get this far
1719
self._head_key = heads.pop()
1720
if self._head_key == self.a_key:
1724
mutter('found dominating revision for %s\n%s > %s', self.vf,
1725
self._head_key[-1], other)
1728
self._head_key = None
1731
def _precache_tip_lines(self):
1732
# Turn this into a no-op, because we will do this later
1735
def _find_recursive_lcas(self):
1736
"""Find all the ancestors back to a unique lca"""
1737
cur_ancestors = (self.a_key, self.b_key)
1738
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
1739
# rather than a key tuple. We will just map that directly to no common
1743
next_lcas = self.graph.find_lca(*cur_ancestors)
1744
# Map a plain NULL_REVISION to a simple no-ancestors
1745
if next_lcas == set([NULL_REVISION]):
1747
# Order the lca's based on when they were merged into the tip
1748
# While the actual merge portion of weave merge uses a set() of
1749
# active revisions, the order of insertion *does* effect the
1750
# implicit ordering of the texts.
1751
for rev_key in cur_ancestors:
1752
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
1754
parent_map[rev_key] = ordered_parents
1755
if len(next_lcas) == 0:
1757
elif len(next_lcas) == 1:
1758
parent_map[list(next_lcas)[0]] = ()
1760
elif len(next_lcas) > 2:
1761
# More than 2 lca's, fall back to grabbing all nodes between
1762
# this and the unique lca.
1763
mutter('More than 2 LCAs, falling back to all nodes for:'
1764
' %s, %s\n=> %s', self.a_key, self.b_key, cur_ancestors)
1765
cur_lcas = next_lcas
1766
while len(cur_lcas) > 1:
1767
cur_lcas = self.graph.find_lca(*cur_lcas)
1768
if len(cur_lcas) == 0:
1769
# No common base to find, use the full ancestry
1772
unique_lca = list(cur_lcas)[0]
1773
if unique_lca == NULL_REVISION:
1774
# find_lca will return a plain 'NULL_REVISION' rather
1775
# than a key tuple when there is no common ancestor, we
1776
# prefer to just use None, because it doesn't confuse
1777
# _get_interesting_texts()
1779
parent_map.update(self._find_unique_parents(next_lcas,
1782
cur_ancestors = next_lcas
1785
def _find_unique_parents(self, tip_keys, base_key):
1786
"""Find ancestors of tip that aren't ancestors of base.
1788
:param tip_keys: Nodes that are interesting
1789
:param base_key: Cull all ancestors of this node
1790
:return: The parent map for all revisions between tip_keys and
1791
base_key. base_key will be included. References to nodes outside of
1792
the ancestor set will also be removed.
1794
# TODO: this would be simpler if find_unique_ancestors took a list
1795
# instead of a single tip, internally it supports it, but it
1796
# isn't a "backwards compatible" api change.
1797
if base_key is None:
1798
parent_map = dict(self.graph.iter_ancestry(tip_keys))
1799
# We remove NULL_REVISION because it isn't a proper tuple key, and
1800
# thus confuses things like _get_interesting_texts, and our logic
1801
# to add the texts into the memory weave.
1802
if NULL_REVISION in parent_map:
1803
parent_map.pop(NULL_REVISION)
1806
for tip in tip_keys:
1808
self.graph.find_unique_ancestors(tip, [base_key]))
1809
parent_map = self.graph.get_parent_map(interesting)
1810
parent_map[base_key] = ()
1811
culled_parent_map, child_map, tails = self._remove_external_references(
1813
# Remove all the tails but base_key
1814
if base_key is not None:
1815
tails.remove(base_key)
1816
self._prune_tails(culled_parent_map, child_map, tails)
1817
# Now remove all the uninteresting 'linear' regions
1818
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
1822
def _remove_external_references(parent_map):
1823
"""Remove references that go outside of the parent map.
1825
:param parent_map: Something returned from Graph.get_parent_map(keys)
1826
:return: (filtered_parent_map, child_map, tails)
1827
filtered_parent_map is parent_map without external references
1828
child_map is the {parent_key: [child_keys]} mapping
1829
tails is a list of nodes that do not have any parents in the map
1831
# TODO: The basic effect of this function seems more generic than
1832
# _PlanMerge. But the specific details of building a child_map,
1833
# and computing tails seems very specific to _PlanMerge.
1834
# Still, should this be in Graph land?
1835
filtered_parent_map = {}
1838
for key, parent_keys in parent_map.iteritems():
1839
culled_parent_keys = [p for p in parent_keys if p in parent_map]
1840
if not culled_parent_keys:
1842
for parent_key in culled_parent_keys:
1843
child_map.setdefault(parent_key, []).append(key)
1844
# TODO: Do we want to do this, it adds overhead for every node,
1845
# just to say that the node has no children
1846
child_map.setdefault(key, [])
1847
filtered_parent_map[key] = culled_parent_keys
1848
return filtered_parent_map, child_map, tails
1851
def _prune_tails(parent_map, child_map, tails_to_remove):
1852
"""Remove tails from the parent map.
1854
This will remove the supplied revisions until no more children have 0
1857
:param parent_map: A dict of {child: [parents]}, this dictionary will
1858
be modified in place.
1859
:param tails_to_remove: A list of tips that should be removed,
1860
this list will be consumed
1861
:param child_map: The reverse dict of parent_map ({parent: [children]})
1862
this dict will be modified
1863
:return: None, parent_map will be modified in place.
1865
while tails_to_remove:
1866
next = tails_to_remove.pop()
1867
parent_map.pop(next)
1868
children = child_map.pop(next)
1869
for child in children:
1870
child_parents = parent_map[child]
1871
child_parents.remove(next)
1872
if len(child_parents) == 0:
1873
tails_to_remove.append(child)
1875
def _get_interesting_texts(self, parent_map):
1876
"""Return a dict of texts we are interested in.
1878
Note that the input is in key tuples, but the output is in plain
1881
:param parent_map: The output from _find_recursive_lcas
1882
:return: A dict of {'revision_id':lines} as returned by
1883
_PlanMergeBase.get_lines()
1885
all_revision_keys = set(parent_map)
1886
all_revision_keys.add(self.a_key)
1887
all_revision_keys.add(self.b_key)
1889
# Everything else is in 'keys' but get_lines is in 'revision_ids'
1890
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
1893
def _build_weave(self):
1894
from bzrlib import weave
1895
self._weave = weave.Weave(weave_name='in_memory_weave',
1896
allow_reserved=True)
1897
parent_map = self._find_recursive_lcas()
1899
all_texts = self._get_interesting_texts(parent_map)
1901
# Note: Unfortunately, the order given by topo_sort will effect the
1902
# ordering resolution in the output. Specifically, if you add A then B,
1903
# then in the output text A lines will show up before B lines. And, of
1904
# course, topo_sort doesn't guarantee any real ordering.
1905
# So we use merge_sort, and add a fake node on the tip.
1906
# This ensures that left-hand parents will always be inserted into the
1907
# weave before right-hand parents.
1908
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
1909
parent_map[tip_key] = (self.a_key, self.b_key)
1911
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
1915
# for key in tsort.topo_sort(parent_map):
1916
parent_keys = parent_map[key]
1917
revision_id = key[-1]
1918
parent_ids = [k[-1] for k in parent_keys]
1919
self._weave.add_lines(revision_id, parent_ids,
1920
all_texts[revision_id])
1922
def plan_merge(self):
1923
"""Generate a 'plan' for merging the two revisions.
1925
This involves comparing their texts and determining the cause of
1926
differences. If text A has a line and text B does not, then either the
1927
line was added to text A, or it was deleted from B. Once the causes
1928
are combined, they are written out in the format described in
1929
VersionedFile.plan_merge
1931
if self._head_key is not None: # There was a single head
1932
if self._head_key == self.a_key:
1935
if self._head_key != self.b_key:
1936
raise AssertionError('There was an invalid head: %s != %s'
1937
% (self.b_key, self._head_key))
1939
head_rev = self._head_key[-1]
1940
lines = self.get_lines([head_rev])[head_rev]
1941
return ((plan, line) for line in lines)
1942
return self._weave.plan_merge(self.a_rev, self.b_rev)
1945
class _PlanLCAMerge(_PlanMergeBase):
1947
This merge algorithm differs from _PlanMerge in that:
1948
1. comparisons are done against LCAs only
1949
2. cases where a contested line is new versus one LCA but old versus
1950
another are marked as conflicts, by emitting the line as conflicted-a
1953
This is faster, and hopefully produces more useful output.
1956
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
1957
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
1958
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
1961
if lca == NULL_REVISION:
1964
self.lcas.add(lca[-1])
1965
for lca in self.lcas:
1966
if _mod_revision.is_null(lca):
1969
lca_lines = self.get_lines([lca])[lca]
1970
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
1972
blocks = list(matcher.get_matching_blocks())
1973
self._cached_matching_blocks[(a_rev, lca)] = blocks
1974
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
1976
blocks = list(matcher.get_matching_blocks())
1977
self._cached_matching_blocks[(b_rev, lca)] = blocks
1979
def _determine_status(self, revision_id, unique_line_numbers):
1980
"""Determines the status unique lines versus all lcas.
1982
Basically, determines why the line is unique to this revision.
1984
A line may be determined new, killed, or both.
1986
If a line is determined new, that means it was not present in at least
1987
one LCA, and is not present in the other merge revision.
1989
If a line is determined killed, that means the line was present in
1992
If a line is killed and new, this indicates that the two merge
1993
revisions contain differing conflict resolutions.
1994
:param revision_id: The id of the revision in which the lines are
1996
:param unique_line_numbers: The line numbers of unique lines.
1997
:return a tuple of (new_this, killed_other):
2001
unique_line_numbers = set(unique_line_numbers)
2002
for lca in self.lcas:
2003
blocks = self._get_matching_blocks(revision_id, lca)
2004
unique_vs_lca, _ignored = self._unique_lines(blocks)
2005
new.update(unique_line_numbers.intersection(unique_vs_lca))
2006
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
