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_by_entry,
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):
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
142
merger = Merger(tree.branch, other_tree, other_tree.basis_tree(), tree,
144
merger.base_rev_id = merger.base_tree.get_revision_id()
145
merger.other_rev_id = None
146
merger.other_basis = merger.base_rev_id
150
def from_mergeable(klass, tree, mergeable, pb):
151
"""Return a Merger for a bundle or merge directive.
153
:param tree: The tree to merge changes into
154
:param mergeable: A merge directive or bundle
155
:param pb: A progress indicator
157
mergeable.install_revisions(tree.branch.repository)
158
base_revision_id, other_revision_id, verified =\
159
mergeable.get_merge_request(tree.branch.repository)
160
revision_graph = tree.branch.repository.get_graph()
161
if base_revision_id is not None:
162
if (base_revision_id != _mod_revision.NULL_REVISION and
163
revision_graph.is_ancestor(
164
base_revision_id, tree.branch.last_revision())):
165
base_revision_id = None
167
warning('Performing cherrypick')
168
merger = klass.from_revision_ids(pb, tree, other_revision_id,
169
base_revision_id, revision_graph=
171
return merger, verified
174
def from_revision_ids(pb, tree, other, base=None, other_branch=None,
175
base_branch=None, revision_graph=None,
177
"""Return a Merger for revision-ids.
179
:param tree: The tree to merge changes into
180
:param other: The revision-id to use as OTHER
181
:param base: The revision-id to use as BASE. If not specified, will
183
:param other_branch: A branch containing the other revision-id. If
184
not supplied, tree.branch is used.
185
:param base_branch: A branch containing the base revision-id. If
186
not supplied, other_branch or tree.branch will be used.
187
:param revision_graph: If you have a revision_graph precomputed, pass
188
it in, otherwise it will be created for you.
189
:param pb: A progress indicator
191
if tree_branch is None:
192
tree_branch = tree.branch
193
merger = Merger(tree_branch, this_tree=tree, pb=pb,
194
revision_graph=revision_graph)
195
if other_branch is None:
196
other_branch = tree.branch
197
merger.set_other_revision(other, other_branch)
201
if base_branch is None:
202
base_branch = other_branch
203
merger.set_base_revision(base, base_branch)
206
def revision_tree(self, revision_id, branch=None):
207
if revision_id not in self._cached_trees:
209
branch = self.this_branch
211
tree = self.this_tree.revision_tree(revision_id)
212
except errors.NoSuchRevisionInTree:
213
tree = branch.repository.revision_tree(revision_id)
214
self._cached_trees[revision_id] = tree
215
return self._cached_trees[revision_id]
217
def _get_tree(self, treespec, possible_transports=None):
218
from bzrlib import workingtree
219
location, revno = treespec
221
tree = workingtree.WorkingTree.open_containing(location)[0]
222
return tree.branch, tree
223
branch = Branch.open_containing(location, possible_transports)[0]
225
revision_id = branch.last_revision()
227
revision_id = branch.get_rev_id(revno)
228
revision_id = ensure_null(revision_id)
229
return branch, self.revision_tree(revision_id, branch)
231
def ensure_revision_trees(self):
232
if self.this_revision_tree is None:
233
self.this_basis_tree = self.revision_tree(self.this_basis)
234
if self.this_basis == self.this_rev_id:
235
self.this_revision_tree = self.this_basis_tree
237
if self.other_rev_id is None:
238
other_basis_tree = self.revision_tree(self.other_basis)
239
changes = other_basis_tree.changes_from(self.other_tree)
240
if changes.has_changed():
241
raise WorkingTreeNotRevision(self.this_tree)
242
other_rev_id = self.other_basis
243
self.other_tree = other_basis_tree
245
def file_revisions(self, file_id):
246
self.ensure_revision_trees()
247
def get_id(tree, file_id):
248
revision_id = tree.inventory[file_id].revision
250
if self.this_rev_id is None:
251
if self.this_basis_tree.get_file_sha1(file_id) != \
252
self.this_tree.get_file_sha1(file_id):
253
raise WorkingTreeNotRevision(self.this_tree)
255
trees = (self.this_basis_tree, self.other_tree)
256
return [get_id(tree, file_id) for tree in trees]
258
def check_basis(self, check_clean, require_commits=True):
259
if self.this_basis is None and require_commits is True:
260
raise BzrCommandError("This branch has no commits."
261
" (perhaps you would prefer 'bzr pull')")
264
if self.this_basis != self.this_rev_id:
265
raise errors.UncommittedChanges(self.this_tree)
267
def compare_basis(self):
269
basis_tree = self.revision_tree(self.this_tree.last_revision())
270
except errors.NoSuchRevision:
271
basis_tree = self.this_tree.basis_tree()
272
changes = self.this_tree.changes_from(basis_tree)
273
if not changes.has_changed():
274
self.this_rev_id = self.this_basis
276
def set_interesting_files(self, file_list):
277
self.interesting_files = file_list
279
def set_pending(self):
280
if not self.base_is_ancestor or not self.base_is_other_ancestor or self.other_rev_id is None:
284
def _add_parent(self):
285
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
286
new_parent_trees = []
287
for revision_id in new_parents:
289
tree = self.revision_tree(revision_id)
290
except errors.NoSuchRevision:
294
new_parent_trees.append((revision_id, tree))
296
self.this_tree.set_parent_trees(new_parent_trees,
297
allow_leftmost_as_ghost=True)
299
for _revision_id, tree in new_parent_trees:
303
def set_other(self, other_revision, possible_transports=None):
304
"""Set the revision and tree to merge from.
306
This sets the other_tree, other_rev_id, other_basis attributes.
308
:param other_revision: The [path, revision] list to merge from.
310
self.other_branch, self.other_tree = self._get_tree(other_revision,
312
if other_revision[1] == -1:
313
self.other_rev_id = _mod_revision.ensure_null(
314
self.other_branch.last_revision())
315
if _mod_revision.is_null(self.other_rev_id):
316
raise NoCommits(self.other_branch)
317
self.other_basis = self.other_rev_id
318
elif other_revision[1] is not None:
319
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
320
self.other_basis = self.other_rev_id
322
self.other_rev_id = None
323
self.other_basis = self.other_branch.last_revision()
324
if self.other_basis is None:
325
raise NoCommits(self.other_branch)
326
if self.other_rev_id is not None:
327
self._cached_trees[self.other_rev_id] = self.other_tree
328
self._maybe_fetch(self.other_branch,self.this_branch, self.other_basis)
330
def set_other_revision(self, revision_id, other_branch):
331
"""Set 'other' based on a branch and revision id
333
:param revision_id: The revision to use for a tree
334
:param other_branch: The branch containing this tree
336
self.other_rev_id = revision_id
337
self.other_branch = other_branch
338
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
339
self.other_tree = self.revision_tree(revision_id)
340
self.other_basis = revision_id
342
def set_base_revision(self, revision_id, branch):
343
"""Set 'base' based on a branch and revision id
345
:param revision_id: The revision to use for a tree
346
:param branch: The branch containing this tree
348
self.base_rev_id = revision_id
349
self.base_branch = branch
350
self._maybe_fetch(branch, self.this_branch, revision_id)
351
self.base_tree = self.revision_tree(revision_id)
353
def _maybe_fetch(self, source, target, revision_id):
354
if not source.repository.has_same_location(target.repository):
355
target.fetch(source, revision_id)
358
revisions = [ensure_null(self.this_basis),
359
ensure_null(self.other_basis)]
360
if NULL_REVISION in revisions:
361
self.base_rev_id = NULL_REVISION
362
self.base_tree = self.revision_tree(self.base_rev_id)
363
self._is_criss_cross = False
365
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
366
self._is_criss_cross = False
368
self.base_rev_id = NULL_REVISION
370
self.base_rev_id = list(lcas)[0]
371
else: # len(lcas) > 1
373
# find_unique_lca can only handle 2 nodes, so we have to
374
# start back at the beginning. It is a shame to traverse
375
# the graph again, but better than re-implementing
377
self.base_rev_id = self.revision_graph.find_unique_lca(
378
revisions[0], revisions[1])
380
self.base_rev_id = self.revision_graph.find_unique_lca(
382
self._is_criss_cross = True
383
if self.base_rev_id == NULL_REVISION:
384
raise UnrelatedBranches()
385
if self._is_criss_cross:
386
warning('Warning: criss-cross merge encountered. See bzr'
387
' help criss-cross.')
388
interesting_revision_ids = [self.base_rev_id]
389
interesting_revision_ids.extend(lcas)
390
interesting_trees = dict((t.get_revision_id(), t)
391
for t in self.this_branch.repository.revision_trees(
392
interesting_revision_ids))
393
self._cached_trees.update(interesting_trees)
394
self.base_tree = interesting_trees.pop(self.base_rev_id)
395
sorted_lca_keys = self.revision_graph.find_merge_order(
397
self._lca_trees = [interesting_trees[key]
398
for key in sorted_lca_keys]
400
self.base_tree = self.revision_tree(self.base_rev_id)
401
self.base_is_ancestor = True
402
self.base_is_other_ancestor = True
404
def set_base(self, base_revision):
405
"""Set the base revision to use for the merge.
407
:param base_revision: A 2-list containing a path and revision number.
409
mutter("doing merge() with no base_revision specified")
410
if base_revision == [None, None]:
413
base_branch, self.base_tree = self._get_tree(base_revision)
414
if base_revision[1] == -1:
415
self.base_rev_id = base_branch.last_revision()
416
elif base_revision[1] is None:
417
self.base_rev_id = _mod_revision.NULL_REVISION
419
self.base_rev_id = _mod_revision.ensure_null(
420
base_branch.get_rev_id(base_revision[1]))
421
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
423
def make_merger(self):
424
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
425
'other_tree': self.other_tree,
426
'interesting_ids': self.interesting_ids,
427
'interesting_files': self.interesting_files,
430
if self.merge_type.requires_base:
431
kwargs['base_tree'] = self.base_tree
432
if self.merge_type.supports_reprocess:
433
kwargs['reprocess'] = self.reprocess
435
raise BzrError("Conflict reduction is not supported for merge"
436
" type %s." % self.merge_type)
437
if self.merge_type.supports_show_base:
438
kwargs['show_base'] = self.show_base
440
raise BzrError("Showing base is not supported for this"
441
" merge type. %s" % self.merge_type)
442
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
443
and not self.base_is_other_ancestor):
444
raise errors.CannotReverseCherrypick()
445
if self.merge_type.supports_cherrypick:
446
kwargs['cherrypick'] = (not self.base_is_ancestor or
447
not self.base_is_other_ancestor)
448
if self._is_criss_cross and getattr(self.merge_type,
449
'supports_lca_trees', False):
450
kwargs['lca_trees'] = self._lca_trees
451
return self.merge_type(pb=self._pb,
452
change_reporter=self.change_reporter,
455
def _do_merge_to(self, merge):
457
if self.recurse == 'down':
458
for relpath, file_id in self.this_tree.iter_references():
459
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
460
other_revision = self.other_tree.get_reference_revision(
462
if other_revision == sub_tree.last_revision():
464
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
465
sub_merge.merge_type = self.merge_type
466
other_branch = self.other_branch.reference_parent(file_id, relpath)
467
sub_merge.set_other_revision(other_revision, other_branch)
468
base_revision = self.base_tree.get_reference_revision(file_id)
469
sub_merge.base_tree = \
470
sub_tree.branch.repository.revision_tree(base_revision)
471
sub_merge.base_rev_id = base_revision
475
self.this_tree.lock_tree_write()
477
if self.base_tree is not None:
478
self.base_tree.lock_read()
480
if self.other_tree is not None:
481
self.other_tree.lock_read()
483
merge = self.make_merger()
484
self._do_merge_to(merge)
486
if self.other_tree is not None:
487
self.other_tree.unlock()
489
if self.base_tree is not None:
490
self.base_tree.unlock()
492
self.this_tree.unlock()
493
if len(merge.cooked_conflicts) == 0:
494
if not self.ignore_zero and not is_quiet():
495
note("All changes applied successfully.")
497
note("%d conflicts encountered." % len(merge.cooked_conflicts))
499
return len(merge.cooked_conflicts)
502
class _InventoryNoneEntry(object):
503
"""This represents an inventory entry which *isn't there*.
505
It simplifies the merging logic if we always have an InventoryEntry, even
506
if it isn't actually present
513
symlink_target = None
516
_none_entry = _InventoryNoneEntry()
519
class Merge3Merger(object):
520
"""Three-way merger that uses the merge3 text merger"""
522
supports_reprocess = True
523
supports_show_base = True
524
history_based = False
525
supports_cherrypick = True
526
supports_reverse_cherrypick = True
527
winner_idx = {"this": 2, "other": 1, "conflict": 1}
528
supports_lca_trees = True
530
def __init__(self, working_tree, this_tree, base_tree, other_tree,
531
interesting_ids=None, reprocess=False, show_base=False,
532
pb=DummyProgress(), pp=None, change_reporter=None,
533
interesting_files=None, do_merge=True,
534
cherrypick=False, lca_trees=None):
535
"""Initialize the merger object and perform the merge.
537
:param working_tree: The working tree to apply the merge to
538
:param this_tree: The local tree in the merge operation
539
:param base_tree: The common tree in the merge operation
540
:param other_tree: The other other tree to merge changes from
541
:param interesting_ids: The file_ids of files that should be
542
participate in the merge. May not be combined with
544
:param: reprocess If True, perform conflict-reduction processing.
545
:param show_base: If True, show the base revision in text conflicts.
546
(incompatible with reprocess)
547
:param pb: A Progress bar
548
:param pp: A ProgressPhase object
549
:param change_reporter: An object that should report changes made
550
:param interesting_files: The tree-relative paths of files that should
551
participate in the merge. If these paths refer to directories,
552
the contents of those directories will also be included. May not
553
be combined with interesting_ids. If neither interesting_files nor
554
interesting_ids is specified, all files may participate in the
556
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
557
if the ancestry was found to include a criss-cross merge.
558
Otherwise should be None.
560
object.__init__(self)
561
if interesting_files is not None and interesting_ids is not None:
563
'specify either interesting_ids or interesting_files')
564
self.interesting_ids = interesting_ids
565
self.interesting_files = interesting_files
566
self.this_tree = working_tree
567
self.base_tree = base_tree
568
self.other_tree = other_tree
569
self._raw_conflicts = []
570
self.cooked_conflicts = []
571
self.reprocess = reprocess
572
self.show_base = show_base
573
self._lca_trees = lca_trees
574
# Uncommenting this will change the default algorithm to always use
575
# _entries_lca. This can be useful for running the test suite and
576
# making sure we haven't missed any corner cases.
577
# if lca_trees is None:
578
# self._lca_trees = [self.base_tree]
581
self.change_reporter = change_reporter
582
self.cherrypick = cherrypick
584
self.pp = ProgressPhase("Merge phase", 3, self.pb)
589
self.this_tree.lock_tree_write()
590
self.base_tree.lock_read()
591
self.other_tree.lock_read()
592
self.tt = TreeTransform(self.this_tree, self.pb)
595
self._compute_transform()
597
results = self.tt.apply(no_conflicts=True)
598
self.write_modified(results)
600
self.this_tree.add_conflicts(self.cooked_conflicts)
601
except UnsupportedOperation:
605
self.other_tree.unlock()
606
self.base_tree.unlock()
607
self.this_tree.unlock()
610
def make_preview_transform(self):
611
self.base_tree.lock_read()
612
self.other_tree.lock_read()
613
self.tt = TransformPreview(self.this_tree)
616
self._compute_transform()
619
self.other_tree.unlock()
620
self.base_tree.unlock()
624
def _compute_transform(self):
625
if self._lca_trees is None:
626
entries = self._entries3()
627
resolver = self._three_way
629
entries = self._entries_lca()
630
resolver = self._lca_multi_way
631
child_pb = ui.ui_factory.nested_progress_bar()
633
for num, (file_id, changed, parents3, names3,
634
executable3) in enumerate(entries):
635
child_pb.update('Preparing file merge', num, len(entries))
636
self._merge_names(file_id, parents3, names3, resolver=resolver)
638
file_status = self.merge_contents(file_id)
640
file_status = 'unmodified'
641
self._merge_executable(file_id,
642
executable3, file_status, resolver=resolver)
647
child_pb = ui.ui_factory.nested_progress_bar()
649
fs_conflicts = resolve_conflicts(self.tt, child_pb,
650
lambda t, c: conflict_pass(t, c, self.other_tree))
653
if self.change_reporter is not None:
654
from bzrlib import delta
655
delta.report_changes(
656
self.tt.iter_changes(), self.change_reporter)
657
self.cook_conflicts(fs_conflicts)
658
for conflict in self.cooked_conflicts:
662
"""Gather data about files modified between three trees.
664
Return a list of tuples of file_id, changed, parents3, names3,
665
executable3. changed is a boolean indicating whether the file contents
666
or kind were changed. parents3 is a tuple of parent ids for base,
667
other and this. names3 is a tuple of names for base, other and this.
668
executable3 is a tuple of execute-bit values for base, other and this.
671
iterator = self.other_tree.iter_changes(self.base_tree,
672
include_unchanged=True, specific_files=self.interesting_files,
673
extra_trees=[self.this_tree])
674
this_entries = dict((e.file_id, e) for p, e in
675
self.this_tree.iter_entries_by_dir(
676
self.interesting_ids))
677
for (file_id, paths, changed, versioned, parents, names, kind,
678
executable) in iterator:
679
if (self.interesting_ids is not None and
680
file_id not in self.interesting_ids):
682
entry = this_entries.get(file_id)
683
if entry is not None:
684
this_name = entry.name
685
this_parent = entry.parent_id
686
this_executable = entry.executable
690
this_executable = None
691
parents3 = parents + (this_parent,)
692
names3 = names + (this_name,)
693
executable3 = executable + (this_executable,)
694
result.append((file_id, changed, parents3, names3, executable3))
697
def _entries_lca(self):
698
"""Gather data about files modified between multiple trees.
700
This compares OTHER versus all LCA trees, and for interesting entries,
701
it then compares with THIS and BASE.
703
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
704
:return: [(file_id, changed, parents, names, executable)]
705
file_id Simple file_id of the entry
706
changed Boolean, True if the kind or contents changed
708
parents ((base, [parent_id, in, lcas]), parent_id_other,
710
names ((base, [name, in, lcas]), name_in_other, name_in_this)
711
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
713
if self.interesting_files is not None:
714
lookup_trees = [self.this_tree, self.base_tree]
715
lookup_trees.extend(self._lca_trees)
716
# I think we should include the lca trees as well
717
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
720
interesting_ids = self.interesting_ids
722
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
724
base_inventory = self.base_tree.inventory
725
this_inventory = self.this_tree.inventory
726
for path, file_id, other_ie, lca_values in walker.iter_all():
727
# Is this modified at all from any of the other trees?
729
other_ie = _none_entry
730
if interesting_ids is not None and file_id not in interesting_ids:
733
# If other_revision is found in any of the lcas, that means this
734
# node is uninteresting. This is because when merging, if there are
735
# multiple heads(), we have to create a new node. So if we didn't,
736
# we know that the ancestry is linear, and that OTHER did not
738
# See doc/developers/lca_merge_resolution.txt for details
739
other_revision = other_ie.revision
740
if other_revision is not None:
741
# We can't use this shortcut when other_revision is None,
742
# because it may be None because things are WorkingTrees, and
743
# not because it is *actually* None.
744
is_unmodified = False
745
for lca_path, ie in lca_values:
746
if ie is not None and ie.revision == other_revision:
753
for lca_path, lca_ie in lca_values:
755
lca_entries.append(_none_entry)
757
lca_entries.append(lca_ie)
759
if file_id in base_inventory:
760
base_ie = base_inventory[file_id]
762
base_ie = _none_entry
764
if file_id in this_inventory:
765
this_ie = this_inventory[file_id]
767
this_ie = _none_entry
773
for lca_ie in lca_entries:
774
lca_kinds.append(lca_ie.kind)
775
lca_parent_ids.append(lca_ie.parent_id)
776
lca_names.append(lca_ie.name)
777
lca_executable.append(lca_ie.executable)
779
kind_winner = self._lca_multi_way(
780
(base_ie.kind, lca_kinds),
781
other_ie.kind, this_ie.kind)
782
parent_id_winner = self._lca_multi_way(
783
(base_ie.parent_id, lca_parent_ids),
784
other_ie.parent_id, this_ie.parent_id)
785
name_winner = self._lca_multi_way(
786
(base_ie.name, lca_names),
787
other_ie.name, this_ie.name)
789
content_changed = True
790
if kind_winner == 'this':
791
# No kind change in OTHER, see if there are *any* changes
792
if other_ie.kind == None:
793
# No content and 'this' wins the kind, so skip this?
796
elif other_ie.kind == 'directory':
797
if parent_id_winner == 'this' and name_winner == 'this':
798
# No change for this directory in OTHER, skip
800
content_changed = False
801
elif other_ie.kind == 'file':
802
def get_sha1(ie, tree):
803
if ie.kind != 'file':
805
return tree.get_file_sha1(file_id)
806
base_sha1 = get_sha1(base_ie, self.base_tree)
807
lca_sha1s = [get_sha1(ie, tree) for ie, tree
808
in zip(lca_entries, self._lca_trees)]
809
this_sha1 = get_sha1(this_ie, self.this_tree)
810
other_sha1 = get_sha1(other_ie, self.other_tree)
811
sha1_winner = self._lca_multi_way(
812
(base_sha1, lca_sha1s), other_sha1, this_sha1,
813
allow_overriding_lca=False)
814
exec_winner = self._lca_multi_way(
815
(base_ie.executable, lca_executable),
816
other_ie.executable, this_ie.executable)
817
if (parent_id_winner == 'this' and name_winner == 'this'
818
and sha1_winner == 'this' and exec_winner == 'this'):
819
# No kind, parent, name, exec, or content change for
820
# OTHER, so this node is not considered interesting
822
if sha1_winner == 'this':
823
content_changed = False
824
elif other_ie.kind == 'symlink':
825
def get_target(ie, tree):
826
if ie.kind != 'symlink':
828
return tree.get_symlink_target(file_id)
829
base_target = get_target(base_ie, self.base_tree)
830
lca_targets = [get_target(ie, tree) for ie, tree
831
in zip(lca_entries, self._lca_trees)]
832
this_target = get_target(this_ie, self.this_tree)
833
other_target = get_target(other_ie, self.other_tree)
834
target_winner = self._lca_multi_way(
835
(base_target, lca_targets),
836
other_target, this_target)
837
if (parent_id_winner == 'this' and name_winner == 'this'
838
and target_winner == 'this'):
839
# No kind, parent, name, or symlink target change
842
if target_winner == 'this':
843
content_changed = False
844
elif other_ie.kind == 'tree-reference':
845
# The 'changed' information seems to be handled at a higher
846
# level. At least, _entries3 returns False for content
847
# changed, even when at a new revision_id.
848
content_changed = False
849
if (parent_id_winner == 'this' and name_winner == 'this'):
850
# Nothing interesting
853
raise AssertionError('unhandled kind: %s' % other_ie.kind)
854
# XXX: We need to handle kind == 'symlink'
856
# If we have gotten this far, that means something has changed
857
result.append((file_id, content_changed,
858
((base_ie.parent_id, lca_parent_ids),
859
other_ie.parent_id, this_ie.parent_id),
860
((base_ie.name, lca_names),
861
other_ie.name, this_ie.name),
862
((base_ie.executable, lca_executable),
863
other_ie.executable, this_ie.executable)
870
self.tt.final_kind(self.tt.root)
872
self.tt.cancel_deletion(self.tt.root)
873
if self.tt.final_file_id(self.tt.root) is None:
874
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
876
if self.other_tree.inventory.root is None:
878
other_root_file_id = self.other_tree.get_root_id()
879
other_root = self.tt.trans_id_file_id(other_root_file_id)
880
if other_root == self.tt.root:
883
self.tt.final_kind(other_root)
886
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
887
# the other tree's root is a non-root in the current tree
889
self.reparent_children(self.other_tree.inventory.root, self.tt.root)
890
self.tt.cancel_creation(other_root)
891
self.tt.cancel_versioning(other_root)
893
def reparent_children(self, ie, target):
894
for thing, child in ie.children.iteritems():
895
trans_id = self.tt.trans_id_file_id(child.file_id)
896
self.tt.adjust_path(self.tt.final_name(trans_id), target, trans_id)
898
def write_modified(self, results):
900
for path in results.modified_paths:
901
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
904
hash = self.this_tree.get_file_sha1(file_id)
907
modified_hashes[file_id] = hash
908
self.this_tree.set_merge_modified(modified_hashes)
911
def parent(entry, file_id):
912
"""Determine the parent for a file_id (used as a key method)"""
915
return entry.parent_id
918
def name(entry, file_id):
919
"""Determine the name for a file_id (used as a key method)"""
925
def contents_sha1(tree, file_id):
926
"""Determine the sha1 of the file contents (used as a key method)."""
927
if file_id not in tree:
929
return tree.get_file_sha1(file_id)
932
def executable(tree, file_id):
933
"""Determine the executability of a file-id (used as a key method)."""
934
if file_id not in tree:
936
if tree.kind(file_id) != "file":
938
return tree.is_executable(file_id)
941
def kind(tree, file_id):
942
"""Determine the kind of a file-id (used as a key method)."""
943
if file_id not in tree:
945
return tree.kind(file_id)
948
def _three_way(base, other, this):
949
#if base == other, either they all agree, or only THIS has changed.
952
elif this not in (base, other):
954
# "Ambiguous clean merge" -- both sides have made the same change.
957
# this == base: only other has changed.
962
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
963
"""Consider LCAs when determining whether a change has occurred.
965
If LCAS are all identical, this is the same as a _three_way comparison.
967
:param bases: value in (BASE, [LCAS])
968
:param other: value in OTHER
969
:param this: value in THIS
970
:param allow_overriding_lca: If there is more than one unique lca
971
value, allow OTHER to override THIS if it has a new value, and
972
THIS only has an lca value, or vice versa. This is appropriate for
973
truly scalar values, not as much for non-scalars.
974
:return: 'this', 'other', or 'conflict' depending on whether an entry
977
# See doc/developers/lca_merge_resolution.txt for details about this
980
# Either Ambiguously clean, or nothing was actually changed. We
983
base_val, lca_vals = bases
984
# Remove 'base_val' from the lca_vals, because it is not interesting
985
filtered_lca_vals = [lca_val for lca_val in lca_vals
986
if lca_val != base_val]
987
if len(filtered_lca_vals) == 0:
988
return Merge3Merger._three_way(base_val, other, this)
990
unique_lca_vals = set(filtered_lca_vals)
991
if len(unique_lca_vals) == 1:
992
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
994
if allow_overriding_lca:
995
if other in unique_lca_vals:
996
if this in unique_lca_vals:
997
# Each side picked a different lca, conflict
1000
# This has a value which supersedes both lca values, and
1001
# other only has an lca value
1003
elif this in unique_lca_vals:
1004
# OTHER has a value which supersedes both lca values, and this
1005
# only has an lca value
1008
# At this point, the lcas disagree, and the tips disagree
1012
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1013
"""Do a three-way test on a scalar.
1014
Return "this", "other" or "conflict", depending whether a value wins.
1016
key_base = key(base_tree, file_id)
1017
key_other = key(other_tree, file_id)
1018
#if base == other, either they all agree, or only THIS has changed.
1019
if key_base == key_other:
1021
key_this = key(this_tree, file_id)
1022
# "Ambiguous clean merge"
1023
if key_this == key_other:
1025
elif key_this == key_base:
1030
def merge_names(self, file_id):
1031
def get_entry(tree):
1032
if file_id in tree.inventory:
1033
return tree.inventory[file_id]
1036
this_entry = get_entry(self.this_tree)
1037
other_entry = get_entry(self.other_tree)
1038
base_entry = get_entry(self.base_tree)
1039
entries = (base_entry, other_entry, this_entry)
1042
for entry in entries:
1045
parents.append(None)
1047
names.append(entry.name)
1048
parents.append(entry.parent_id)
1049
return self._merge_names(file_id, parents, names,
1050
resolver=self._three_way)
1052
def _merge_names(self, file_id, parents, names, resolver):
1053
"""Perform a merge on file_id names and parents"""
1054
base_name, other_name, this_name = names
1055
base_parent, other_parent, this_parent = parents
1057
name_winner = resolver(*names)
1059
parent_id_winner = resolver(*parents)
1060
if this_name is None:
1061
if name_winner == "this":
1062
name_winner = "other"
1063
if parent_id_winner == "this":
1064
parent_id_winner = "other"
1065
if name_winner == "this" and parent_id_winner == "this":
1067
if name_winner == "conflict":
1068
trans_id = self.tt.trans_id_file_id(file_id)
1069
self._raw_conflicts.append(('name conflict', trans_id,
1070
this_name, other_name))
1071
if parent_id_winner == "conflict":
1072
trans_id = self.tt.trans_id_file_id(file_id)
1073
self._raw_conflicts.append(('parent conflict', trans_id,
1074
this_parent, other_parent))
1075
if other_name is None:
1076
# it doesn't matter whether the result was 'other' or
1077
# 'conflict'-- if there's no 'other', we leave it alone.
1079
# if we get here, name_winner and parent_winner are set to safe values.
1080
trans_id = self.tt.trans_id_file_id(file_id)
1081
parent_id = parents[self.winner_idx[parent_id_winner]]
1082
if parent_id is not None:
1083
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1084
self.tt.adjust_path(names[self.winner_idx[name_winner]],
1085
parent_trans_id, trans_id)
1087
def merge_contents(self, file_id):
1088
"""Performa a merge on file_id contents."""
1089
def contents_pair(tree):
1090
if file_id not in tree:
1092
kind = tree.kind(file_id)
1094
contents = tree.get_file_sha1(file_id)
1095
elif kind == "symlink":
1096
contents = tree.get_symlink_target(file_id)
1099
return kind, contents
1101
def contents_conflict():
1102
trans_id = self.tt.trans_id_file_id(file_id)
1103
name = self.tt.final_name(trans_id)
1104
parent_id = self.tt.final_parent(trans_id)
1105
if file_id in self.this_tree.inventory:
1106
self.tt.unversion_file(trans_id)
1107
if file_id in self.this_tree:
1108
self.tt.delete_contents(trans_id)
1109
file_group = self._dump_conflicts(name, parent_id, file_id,
1111
self._raw_conflicts.append(('contents conflict', file_group))
1113
# See SPOT run. run, SPOT, run.
1114
# So we're not QUITE repeating ourselves; we do tricky things with
1116
base_pair = contents_pair(self.base_tree)
1117
other_pair = contents_pair(self.other_tree)
1118
if base_pair == other_pair:
1119
# OTHER introduced no changes
1121
this_pair = contents_pair(self.this_tree)
1122
if this_pair == other_pair:
1123
# THIS and OTHER introduced the same changes
1126
trans_id = self.tt.trans_id_file_id(file_id)
1127
if this_pair == base_pair:
1128
# only OTHER introduced changes
1129
if file_id in self.this_tree:
1130
# Remove any existing contents
1131
self.tt.delete_contents(trans_id)
1132
if file_id in self.other_tree:
1133
# OTHER changed the file
1134
create_by_entry(self.tt,
1135
self.other_tree.inventory[file_id],
1136
self.other_tree, trans_id)
1137
if file_id not in self.this_tree:
1138
self.tt.version_file(file_id, trans_id)
1140
elif file_id in self.this_tree.inventory:
1141
# OTHER deleted the file
1142
self.tt.unversion_file(trans_id)
1144
#BOTH THIS and OTHER introduced changes; scalar conflict
1145
elif this_pair[0] == "file" and other_pair[0] == "file":
1146
# THIS and OTHER are both files, so text merge. Either
1147
# BASE is a file, or both converted to files, so at least we
1148
# have agreement that output should be a file.
1150
self.text_merge(file_id, trans_id)
1152
return contents_conflict()
1153
if file_id not in self.this_tree:
1154
self.tt.version_file(file_id, trans_id)
1156
self.tt.tree_kind(trans_id)
1157
self.tt.delete_contents(trans_id)
1162
# Scalar conflict, can't text merge. Dump conflicts
1163
return contents_conflict()
1165
def get_lines(self, tree, file_id):
1166
"""Return the lines in a file, or an empty list."""
1168
return tree.get_file(file_id).readlines()
1172
def text_merge(self, file_id, trans_id):
1173
"""Perform a three-way text merge on a file_id"""
1174
# it's possible that we got here with base as a different type.
1175
# if so, we just want two-way text conflicts.
1176
if file_id in self.base_tree and \
1177
self.base_tree.kind(file_id) == "file":
1178
base_lines = self.get_lines(self.base_tree, file_id)
1181
other_lines = self.get_lines(self.other_tree, file_id)
1182
this_lines = self.get_lines(self.this_tree, file_id)
1183
m3 = Merge3(base_lines, this_lines, other_lines,
1184
is_cherrypick=self.cherrypick)
1185
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1186
if self.show_base is True:
1187
base_marker = '|' * 7
1191
def iter_merge3(retval):
1192
retval["text_conflicts"] = False
1193
for line in m3.merge_lines(name_a = "TREE",
1194
name_b = "MERGE-SOURCE",
1195
name_base = "BASE-REVISION",
1196
start_marker=start_marker,
1197
base_marker=base_marker,
1198
reprocess=self.reprocess):
1199
if line.startswith(start_marker):
1200
retval["text_conflicts"] = True
1201
yield line.replace(start_marker, '<' * 7)
1205
merge3_iterator = iter_merge3(retval)
1206
self.tt.create_file(merge3_iterator, trans_id)
1207
if retval["text_conflicts"] is True:
1208
self._raw_conflicts.append(('text conflict', trans_id))
1209
name = self.tt.final_name(trans_id)
1210
parent_id = self.tt.final_parent(trans_id)
1211
file_group = self._dump_conflicts(name, parent_id, file_id,
1212
this_lines, base_lines,
1214
file_group.append(trans_id)
1216
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1217
base_lines=None, other_lines=None, set_version=False,
1219
"""Emit conflict files.
1220
If this_lines, base_lines, or other_lines are omitted, they will be
1221
determined automatically. If set_version is true, the .OTHER, .THIS
1222
or .BASE (in that order) will be created as versioned files.
1224
data = [('OTHER', self.other_tree, other_lines),
1225
('THIS', self.this_tree, this_lines)]
1227
data.append(('BASE', self.base_tree, base_lines))
1230
for suffix, tree, lines in data:
1232
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1234
file_group.append(trans_id)
1235
if set_version and not versioned:
1236
self.tt.version_file(file_id, trans_id)
1240
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1242
"""Emit a single conflict file."""
1243
name = name + '.' + suffix
1244
trans_id = self.tt.create_path(name, parent_id)
1245
entry = tree.inventory[file_id]
1246
create_by_entry(self.tt, entry, tree, trans_id, lines)
1249
def merge_executable(self, file_id, file_status):
1250
"""Perform a merge on the execute bit."""
1251
executable = [self.executable(t, file_id) for t in (self.base_tree,
1252
self.other_tree, self.this_tree)]
1253
self._merge_executable(file_id, executable, file_status,
1254
resolver=self._three_way)
1256
def _merge_executable(self, file_id, executable, file_status,
1258
"""Perform a merge on the execute bit."""
1259
base_executable, other_executable, this_executable = executable
1260
if file_status == "deleted":
1262
winner = resolver(*executable)
1263
if winner == "conflict":
1264
# There must be a None in here, if we have a conflict, but we
1265
# need executability since file status was not deleted.
1266
if self.executable(self.other_tree, file_id) is None:
1270
if winner == 'this' and file_status != "modified":
1272
trans_id = self.tt.trans_id_file_id(file_id)
1274
if self.tt.final_kind(trans_id) != "file":
1278
if winner == "this":
1279
executability = this_executable
1281
if file_id in self.other_tree:
1282
executability = other_executable
1283
elif file_id in self.this_tree:
1284
executability = this_executable
1285
elif file_id in self.base_tree:
1286
executability = base_executable
1287
if executability is not None:
1288
trans_id = self.tt.trans_id_file_id(file_id)
1289
self.tt.set_executability(executability, trans_id)
1291
def cook_conflicts(self, fs_conflicts):
1292
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1293
from conflicts import Conflict
1295
self.cooked_conflicts.extend(cook_conflicts(fs_conflicts, self.tt))
1296
fp = FinalPaths(self.tt)
1297
for conflict in self._raw_conflicts:
1298
conflict_type = conflict[0]
1299
if conflict_type in ('name conflict', 'parent conflict'):
1300
trans_id = conflict[1]
1301
conflict_args = conflict[2:]
1302
if trans_id not in name_conflicts:
1303
name_conflicts[trans_id] = {}
1304
unique_add(name_conflicts[trans_id], conflict_type,
1306
if conflict_type == 'contents conflict':
1307
for trans_id in conflict[1]:
1308
file_id = self.tt.final_file_id(trans_id)
1309
if file_id is not None:
1311
path = fp.get_path(trans_id)
1312
for suffix in ('.BASE', '.THIS', '.OTHER'):
1313
if path.endswith(suffix):
1314
path = path[:-len(suffix)]
1316
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
1317
self.cooked_conflicts.append(c)
1318
if conflict_type == 'text conflict':
1319
trans_id = conflict[1]
1320
path = fp.get_path(trans_id)
1321
file_id = self.tt.final_file_id(trans_id)
1322
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
1323
self.cooked_conflicts.append(c)
1325
for trans_id, conflicts in name_conflicts.iteritems():
1327
this_parent, other_parent = conflicts['parent conflict']
1328
if this_parent == other_parent:
1329
raise AssertionError()
1331
this_parent = other_parent = \
1332
self.tt.final_file_id(self.tt.final_parent(trans_id))
1334
this_name, other_name = conflicts['name conflict']
1335
if this_name == other_name:
1336
raise AssertionError()
1338
this_name = other_name = self.tt.final_name(trans_id)
1339
other_path = fp.get_path(trans_id)
1340
if this_parent is not None and this_name is not None:
1341
this_parent_path = \
1342
fp.get_path(self.tt.trans_id_file_id(this_parent))
1343
this_path = pathjoin(this_parent_path, this_name)
1345
this_path = "<deleted>"
1346
file_id = self.tt.final_file_id(trans_id)
1347
c = Conflict.factory('path conflict', path=this_path,
1348
conflict_path=other_path, file_id=file_id)
1349
self.cooked_conflicts.append(c)
1350
self.cooked_conflicts.sort(key=Conflict.sort_key)
1353
class WeaveMerger(Merge3Merger):
1354
"""Three-way tree merger, text weave merger."""
1355
supports_reprocess = True
1356
supports_show_base = False
1357
supports_reverse_cherrypick = False
1358
history_based = True
1360
def _merged_lines(self, file_id):
1361
"""Generate the merged lines.
1362
There is no distinction between lines that are meant to contain <<<<<<<
1366
base = self.base_tree
1369
plan = self.this_tree.plan_file_merge(file_id, self.other_tree,
1371
if 'merge' in debug.debug_flags:
1373
trans_id = self.tt.trans_id_file_id(file_id)
1374
name = self.tt.final_name(trans_id) + '.plan'
1375
contents = ('%10s|%s' % l for l in plan)
1376
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1377
textmerge = PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1378
'>>>>>>> MERGE-SOURCE\n')
1379
return textmerge.merge_lines(self.reprocess)
1381
def text_merge(self, file_id, trans_id):
1382
"""Perform a (weave) text merge for a given file and file-id.
1383
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1384
and a conflict will be noted.
1386
lines, conflicts = self._merged_lines(file_id)
1388
# Note we're checking whether the OUTPUT is binary in this case,
1389
# because we don't want to get into weave merge guts.
1390
check_text_lines(lines)
1391
self.tt.create_file(lines, trans_id)
1393
self._raw_conflicts.append(('text conflict', trans_id))
1394
name = self.tt.final_name(trans_id)
1395
parent_id = self.tt.final_parent(trans_id)
1396
file_group = self._dump_conflicts(name, parent_id, file_id,
1398
file_group.append(trans_id)
1401
class LCAMerger(WeaveMerger):
1403
def _merged_lines(self, file_id):
1404
"""Generate the merged lines.
1405
There is no distinction between lines that are meant to contain <<<<<<<
1409
base = self.base_tree
1412
plan = self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1414
if 'merge' in debug.debug_flags:
1416
trans_id = self.tt.trans_id_file_id(file_id)
1417
name = self.tt.final_name(trans_id) + '.plan'
1418
contents = ('%10s|%s' % l for l in plan)
1419
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1420
textmerge = PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1421
'>>>>>>> MERGE-SOURCE\n')
1422
return textmerge.merge_lines(self.reprocess)
1425
class Diff3Merger(Merge3Merger):
1426
"""Three-way merger using external diff3 for text merging"""
1428
def dump_file(self, temp_dir, name, tree, file_id):
1429
out_path = pathjoin(temp_dir, name)
1430
out_file = open(out_path, "wb")
1432
in_file = tree.get_file(file_id)
1433
for line in in_file:
1434
out_file.write(line)
1439
def text_merge(self, file_id, trans_id):
1440
"""Perform a diff3 merge using a specified file-id and trans-id.
1441
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1442
will be dumped, and a will be conflict noted.
1445
temp_dir = osutils.mkdtemp(prefix="bzr-")
1447
new_file = pathjoin(temp_dir, "new")
1448
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1449
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1450
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1451
status = bzrlib.patch.diff3(new_file, this, base, other)
1452
if status not in (0, 1):
1453
raise BzrError("Unhandled diff3 exit code")
1454
f = open(new_file, 'rb')
1456
self.tt.create_file(f, trans_id)
1460
name = self.tt.final_name(trans_id)
1461
parent_id = self.tt.final_parent(trans_id)
1462
self._dump_conflicts(name, parent_id, file_id)
1463
self._raw_conflicts.append(('text conflict', trans_id))
1465
osutils.rmtree(temp_dir)
1468
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1470
merge_type=Merge3Merger,
1471
interesting_ids=None,
1475
interesting_files=None,
1478
change_reporter=None):
1479
"""Primary interface for merging.
1481
typical use is probably
1482
'merge_inner(branch, branch.get_revision_tree(other_revision),
1483
branch.get_revision_tree(base_revision))'
1485
if this_tree is None:
1486
raise BzrError("bzrlib.merge.merge_inner requires a this_tree "
1487
"parameter as of bzrlib version 0.8.")
1488
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1489
pb=pb, change_reporter=change_reporter)
1490
merger.backup_files = backup_files
1491
merger.merge_type = merge_type
1492
merger.interesting_ids = interesting_ids
1493
merger.ignore_zero = ignore_zero
1494
if interesting_files:
1496
raise ValueError('Only supply interesting_ids'
1497
' or interesting_files')
1498
merger.interesting_files = interesting_files
1499
merger.show_base = show_base
1500
merger.reprocess = reprocess
1501
merger.other_rev_id = other_rev_id
1502
merger.other_basis = other_rev_id
1503
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1504
if get_revision_id is None:
1505
get_revision_id = base_tree.last_revision
1506
merger.set_base_revision(get_revision_id(), this_branch)
1507
return merger.do_merge()
1509
def get_merge_type_registry():
1510
"""Merge type registry is in bzrlib.option to avoid circular imports.
1512
This method provides a sanctioned way to retrieve it.
1514
from bzrlib import option
1515
return option._merge_type_registry
1518
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1519
def status_a(revision, text):
1520
if revision in ancestors_b:
1521
return 'killed-b', text
1523
return 'new-a', text
1525
def status_b(revision, text):
1526
if revision in ancestors_a:
1527
return 'killed-a', text
1529
return 'new-b', text
1531
plain_a = [t for (a, t) in annotated_a]
1532
plain_b = [t for (a, t) in annotated_b]
1533
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1534
blocks = matcher.get_matching_blocks()
1537
for ai, bi, l in blocks:
1538
# process all mismatched sections
1539
# (last mismatched section is handled because blocks always
1540
# includes a 0-length last block)
1541
for revision, text in annotated_a[a_cur:ai]:
1542
yield status_a(revision, text)
1543
for revision, text in annotated_b[b_cur:bi]:
1544
yield status_b(revision, text)
1545
# and now the matched section
1548
for text_a in plain_a[ai:a_cur]:
1549
yield "unchanged", text_a
1552
class _PlanMergeBase(object):
1554
def __init__(self, a_rev, b_rev, vf, key_prefix):
1557
:param a_rev: Revision-id of one revision to merge
1558
:param b_rev: Revision-id of the other revision to merge
1559
:param vf: A VersionedFiles containing both revisions
1560
:param key_prefix: A prefix for accessing keys in vf, typically
1566
self._last_lines = None
1567
self._last_lines_revision_id = None
1568
self._cached_matching_blocks = {}
1569
self._key_prefix = key_prefix
1570
self._precache_tip_lines()
1572
def _precache_tip_lines(self):
1573
lines = self.get_lines([self.a_rev, self.b_rev])
1574
self.lines_a = lines[self.a_rev]
1575
self.lines_b = lines[self.b_rev]
1577
def get_lines(self, revisions):
1578
"""Get lines for revisions from the backing VersionedFiles.
1580
:raises RevisionNotPresent: on absent texts.
1582
keys = [(self._key_prefix + (rev,)) for rev in revisions]
1584
for record in self.vf.get_record_stream(keys, 'unordered', True):
1585
if record.storage_kind == 'absent':
1586
raise errors.RevisionNotPresent(record.key, self.vf)
1587
result[record.key[-1]] = osutils.split_lines(
1588
record.get_bytes_as('fulltext'))
1591
def plan_merge(self):
1592
"""Generate a 'plan' for merging the two revisions.
1594
This involves comparing their texts and determining the cause of
1595
differences. If text A has a line and text B does not, then either the
1596
line was added to text A, or it was deleted from B. Once the causes
1597
are combined, they are written out in the format described in
1598
VersionedFile.plan_merge
1600
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
1601
unique_a, unique_b = self._unique_lines(blocks)
1602
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
1603
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
1604
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
1606
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
1609
for i, j, n in blocks:
1610
for a_index in range(last_i, i):
1611
if a_index in new_a:
1612
if a_index in killed_b:
1613
yield 'conflicted-a', self.lines_a[a_index]
1615
yield 'new-a', self.lines_a[a_index]
1617
yield 'killed-b', self.lines_a[a_index]
1618
for b_index in range(last_j, j):
1619
if b_index in new_b:
1620
if b_index in killed_a:
1621
yield 'conflicted-b', self.lines_b[b_index]
1623
yield 'new-b', self.lines_b[b_index]
1625
yield 'killed-a', self.lines_b[b_index]
1626
# handle common lines
1627
for a_index in range(i, i+n):
1628
yield 'unchanged', self.lines_a[a_index]
1632
def _get_matching_blocks(self, left_revision, right_revision):
1633
"""Return a description of which sections of two revisions match.
1635
See SequenceMatcher.get_matching_blocks
1637
cached = self._cached_matching_blocks.get((left_revision,
1639
if cached is not None:
1641
if self._last_lines_revision_id == left_revision:
1642
left_lines = self._last_lines
1643
right_lines = self.get_lines([right_revision])[right_revision]
1645
lines = self.get_lines([left_revision, right_revision])
1646
left_lines = lines[left_revision]
1647
right_lines = lines[right_revision]
1648
self._last_lines = right_lines
1649
self._last_lines_revision_id = right_revision
1650
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
1652
return matcher.get_matching_blocks()
1654
def _unique_lines(self, matching_blocks):
1655
"""Analyse matching_blocks to determine which lines are unique
1657
:return: a tuple of (unique_left, unique_right), where the values are
1658
sets of line numbers of unique lines.
1664
for i, j, n in matching_blocks:
1665
unique_left.extend(range(last_i, i))
1666
unique_right.extend(range(last_j, j))
1669
return unique_left, unique_right
1672
def _subtract_plans(old_plan, new_plan):
1673
"""Remove changes from new_plan that came from old_plan.
1675
It is assumed that the difference between the old_plan and new_plan
1676
is their choice of 'b' text.
1678
All lines from new_plan that differ from old_plan are emitted
1679
verbatim. All lines from new_plan that match old_plan but are
1680
not about the 'b' revision are emitted verbatim.
1682
Lines that match and are about the 'b' revision are the lines we
1683
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
1684
is skipped entirely.
1686
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
1689
for i, j, n in matcher.get_matching_blocks():
1690
for jj in range(last_j, j):
1692
for jj in range(j, j+n):
1693
plan_line = new_plan[jj]
1694
if plan_line[0] == 'new-b':
1696
elif plan_line[0] == 'killed-b':
1697
yield 'unchanged', plan_line[1]
1703
class _PlanMerge(_PlanMergeBase):
1704
"""Plan an annotate merge using on-the-fly annotation"""
1706
def __init__(self, a_rev, b_rev, vf, key_prefix):
1707
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
1708
self.a_key = self._key_prefix + (self.a_rev,)
1709
self.b_key = self._key_prefix + (self.b_rev,)
1710
self.graph = Graph(self.vf)
1711
heads = self.graph.heads((self.a_key, self.b_key))
1713
# one side dominates, so we can just return its values, yay for
1715
# Ideally we would know that before we get this far
1716
self._head_key = heads.pop()
1717
if self._head_key == self.a_key:
1721
mutter('found dominating revision for %s\n%s > %s', self.vf,
1722
self._head_key[-1], other)
1725
self._head_key = None
1728
def _precache_tip_lines(self):
1729
# Turn this into a no-op, because we will do this later
1732
def _find_recursive_lcas(self):
1733
"""Find all the ancestors back to a unique lca"""
1734
cur_ancestors = (self.a_key, self.b_key)
1735
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
1736
# rather than a key tuple. We will just map that directly to no common
1740
next_lcas = self.graph.find_lca(*cur_ancestors)
1741
# Map a plain NULL_REVISION to a simple no-ancestors
1742
if next_lcas == set([NULL_REVISION]):
1744
# Order the lca's based on when they were merged into the tip
1745
# While the actual merge portion of weave merge uses a set() of
1746
# active revisions, the order of insertion *does* effect the
1747
# implicit ordering of the texts.
1748
for rev_key in cur_ancestors:
1749
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
1751
parent_map[rev_key] = ordered_parents
1752
if len(next_lcas) == 0:
1754
elif len(next_lcas) == 1:
1755
parent_map[list(next_lcas)[0]] = ()
1757
elif len(next_lcas) > 2:
1758
# More than 2 lca's, fall back to grabbing all nodes between
1759
# this and the unique lca.
1760
mutter('More than 2 LCAs, falling back to all nodes for:'
1761
' %s, %s\n=> %s', self.a_key, self.b_key, cur_ancestors)
1762
cur_lcas = next_lcas
1763
while len(cur_lcas) > 1:
1764
cur_lcas = self.graph.find_lca(*cur_lcas)
1765
if len(cur_lcas) == 0:
1766
# No common base to find, use the full ancestry
1769
unique_lca = list(cur_lcas)[0]
1770
if unique_lca == NULL_REVISION:
1771
# find_lca will return a plain 'NULL_REVISION' rather
1772
# than a key tuple when there is no common ancestor, we
1773
# prefer to just use None, because it doesn't confuse
1774
# _get_interesting_texts()
1776
parent_map.update(self._find_unique_parents(next_lcas,
1779
cur_ancestors = next_lcas
1782
def _find_unique_parents(self, tip_keys, base_key):
1783
"""Find ancestors of tip that aren't ancestors of base.
1785
:param tip_keys: Nodes that are interesting
1786
:param base_key: Cull all ancestors of this node
1787
:return: The parent map for all revisions between tip_keys and
1788
base_key. base_key will be included. References to nodes outside of
1789
the ancestor set will also be removed.
1791
# TODO: this would be simpler if find_unique_ancestors took a list
1792
# instead of a single tip, internally it supports it, but it
1793
# isn't a "backwards compatible" api change.
1794
if base_key is None:
1795
parent_map = dict(self.graph.iter_ancestry(tip_keys))
1796
# We remove NULL_REVISION because it isn't a proper tuple key, and
1797
# thus confuses things like _get_interesting_texts, and our logic
1798
# to add the texts into the memory weave.
1799
if NULL_REVISION in parent_map:
1800
parent_map.pop(NULL_REVISION)
1803
for tip in tip_keys:
1805
self.graph.find_unique_ancestors(tip, [base_key]))
1806
parent_map = self.graph.get_parent_map(interesting)
1807
parent_map[base_key] = ()
1808
culled_parent_map, child_map, tails = self._remove_external_references(
1810
# Remove all the tails but base_key
1811
if base_key is not None:
1812
tails.remove(base_key)
1813
self._prune_tails(culled_parent_map, child_map, tails)
1814
# Now remove all the uninteresting 'linear' regions
1815
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
1819
def _remove_external_references(parent_map):
1820
"""Remove references that go outside of the parent map.
1822
:param parent_map: Something returned from Graph.get_parent_map(keys)
1823
:return: (filtered_parent_map, child_map, tails)
1824
filtered_parent_map is parent_map without external references
1825
child_map is the {parent_key: [child_keys]} mapping
1826
tails is a list of nodes that do not have any parents in the map
1828
# TODO: The basic effect of this function seems more generic than
1829
# _PlanMerge. But the specific details of building a child_map,
1830
# and computing tails seems very specific to _PlanMerge.
1831
# Still, should this be in Graph land?
1832
filtered_parent_map = {}
1835
for key, parent_keys in parent_map.iteritems():
1836
culled_parent_keys = [p for p in parent_keys if p in parent_map]
1837
if not culled_parent_keys:
1839
for parent_key in culled_parent_keys:
1840
child_map.setdefault(parent_key, []).append(key)
1841
# TODO: Do we want to do this, it adds overhead for every node,
1842
# just to say that the node has no children
1843
child_map.setdefault(key, [])
1844
filtered_parent_map[key] = culled_parent_keys
1845
return filtered_parent_map, child_map, tails
1848
def _prune_tails(parent_map, child_map, tails_to_remove):
1849
"""Remove tails from the parent map.
1851
This will remove the supplied revisions until no more children have 0
1854
:param parent_map: A dict of {child: [parents]}, this dictionary will
1855
be modified in place.
1856
:param tails_to_remove: A list of tips that should be removed,
1857
this list will be consumed
1858
:param child_map: The reverse dict of parent_map ({parent: [children]})
1859
this dict will be modified
1860
:return: None, parent_map will be modified in place.
1862
while tails_to_remove:
1863
next = tails_to_remove.pop()
1864
parent_map.pop(next)
1865
children = child_map.pop(next)
1866
for child in children:
1867
child_parents = parent_map[child]
1868
child_parents.remove(next)
1869
if len(child_parents) == 0:
1870
tails_to_remove.append(child)
1872
def _get_interesting_texts(self, parent_map):
1873
"""Return a dict of texts we are interested in.
1875
Note that the input is in key tuples, but the output is in plain
1878
:param parent_map: The output from _find_recursive_lcas
1879
:return: A dict of {'revision_id':lines} as returned by
1880
_PlanMergeBase.get_lines()
1882
all_revision_keys = set(parent_map)
1883
all_revision_keys.add(self.a_key)
1884
all_revision_keys.add(self.b_key)
1886
# Everything else is in 'keys' but get_lines is in 'revision_ids'
1887
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
1890
def _build_weave(self):
1891
from bzrlib import weave
1892
self._weave = weave.Weave(weave_name='in_memory_weave',
1893
allow_reserved=True)
1894
parent_map = self._find_recursive_lcas()
1896
all_texts = self._get_interesting_texts(parent_map)
1898
# Note: Unfortunately, the order given by topo_sort will effect the
1899
# ordering resolution in the output. Specifically, if you add A then B,
1900
# then in the output text A lines will show up before B lines. And, of
1901
# course, topo_sort doesn't guarantee any real ordering.
1902
# So we use merge_sort, and add a fake node on the tip.
1903
# This ensures that left-hand parents will always be inserted into the
1904
# weave before right-hand parents.
1905
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
1906
parent_map[tip_key] = (self.a_key, self.b_key)
1908
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
1912
# for key in tsort.topo_sort(parent_map):
1913
parent_keys = parent_map[key]
1914
revision_id = key[-1]
1915
parent_ids = [k[-1] for k in parent_keys]
1916
self._weave.add_lines(revision_id, parent_ids,
1917
all_texts[revision_id])
1919
def plan_merge(self):
1920
"""Generate a 'plan' for merging the two revisions.
1922
This involves comparing their texts and determining the cause of
1923
differences. If text A has a line and text B does not, then either the
1924
line was added to text A, or it was deleted from B. Once the causes
1925
are combined, they are written out in the format described in
1926
VersionedFile.plan_merge
1928
if self._head_key is not None: # There was a single head
1929
if self._head_key == self.a_key:
1932
if self._head_key != self.b_key:
1933
raise AssertionError('There was an invalid head: %s != %s'
1934
% (self.b_key, self._head_key))
1936
head_rev = self._head_key[-1]
1937
lines = self.get_lines([head_rev])[head_rev]
1938
return ((plan, line) for line in lines)
1939
return self._weave.plan_merge(self.a_rev, self.b_rev)
1942
class _PlanLCAMerge(_PlanMergeBase):
1944
This merge algorithm differs from _PlanMerge in that:
1945
1. comparisons are done against LCAs only
1946
2. cases where a contested line is new versus one LCA but old versus
1947
another are marked as conflicts, by emitting the line as conflicted-a
1950
This is faster, and hopefully produces more useful output.
1953
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
1954
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
1955
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
1958
if lca == NULL_REVISION:
1961
self.lcas.add(lca[-1])
1962
for lca in self.lcas:
1963
if _mod_revision.is_null(lca):
1966
lca_lines = self.get_lines([lca])[lca]
1967
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
1969
blocks = list(matcher.get_matching_blocks())
1970
self._cached_matching_blocks[(a_rev, lca)] = blocks
1971
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
1973
blocks = list(matcher.get_matching_blocks())
1974
self._cached_matching_blocks[(b_rev, lca)] = blocks
1976
def _determine_status(self, revision_id, unique_line_numbers):
1977
"""Determines the status unique lines versus all lcas.
1979
Basically, determines why the line is unique to this revision.
1981
A line may be determined new, killed, or both.
1983
If a line is determined new, that means it was not present in at least
1984
one LCA, and is not present in the other merge revision.
1986
If a line is determined killed, that means the line was present in
1989
If a line is killed and new, this indicates that the two merge
1990
revisions contain differing conflict resolutions.
1991
:param revision_id: The id of the revision in which the lines are
1993
:param unique_line_numbers: The line numbers of unique lines.
1994
:return a tuple of (new_this, killed_other):
1998
unique_line_numbers = set(unique_line_numbers)
1999
for lca in self.lcas:
2000
blocks = self._get_matching_blocks(revision_id, lca)
2001
unique_vs_lca, _ignored = self._unique_lines(blocks)
2002
new.update(unique_line_numbers.intersection(unique_vs_lca))
2003
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
