1
# Copyright (C) 2005, 2006, 2008 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19
from itertools import chain
30
revision as _mod_revision,
34
from bzrlib.branch import Branch
35
from bzrlib.conflicts import ConflictList, Conflict
36
from bzrlib.errors import (BzrCommandError,
46
WorkingTreeNotRevision,
49
from bzrlib.graph import Graph
50
from bzrlib.merge3 import Merge3
51
from bzrlib.osutils import rename, pathjoin
52
from progress import DummyProgress, ProgressPhase
53
from bzrlib.revision import (NULL_REVISION, ensure_null)
54
from bzrlib.textfile import check_text_lines
55
from bzrlib.trace import mutter, warning, note, is_quiet
56
from bzrlib.transform import (TransformPreview, TreeTransform,
57
resolve_conflicts, cook_conflicts,
58
conflict_pass, FinalPaths, create_from_tree,
59
unique_add, ROOT_PARENT)
60
from bzrlib.versionedfile import PlanWeaveMerge
63
# TODO: Report back as changes are merged in
66
def transform_tree(from_tree, to_tree, interesting_ids=None):
67
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=None, 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
96
self.recurse = recurse
97
self.change_reporter = change_reporter
98
self._cached_trees = {}
99
self._revision_graph = revision_graph
100
self._base_is_ancestor = None
101
self._base_is_other_ancestor = None
102
self._is_criss_cross = None
103
self._lca_trees = None
106
def revision_graph(self):
107
if self._revision_graph is None:
108
self._revision_graph = self.this_branch.repository.get_graph()
109
return self._revision_graph
111
def _set_base_is_ancestor(self, value):
112
self._base_is_ancestor = value
114
def _get_base_is_ancestor(self):
115
if self._base_is_ancestor is None:
116
self._base_is_ancestor = self.revision_graph.is_ancestor(
117
self.base_rev_id, self.this_basis)
118
return self._base_is_ancestor
120
base_is_ancestor = property(_get_base_is_ancestor, _set_base_is_ancestor)
122
def _set_base_is_other_ancestor(self, value):
123
self._base_is_other_ancestor = value
125
def _get_base_is_other_ancestor(self):
126
if self._base_is_other_ancestor is None:
127
if self.other_basis is None:
129
self._base_is_other_ancestor = self.revision_graph.is_ancestor(
130
self.base_rev_id, self.other_basis)
131
return self._base_is_other_ancestor
133
base_is_other_ancestor = property(_get_base_is_other_ancestor,
134
_set_base_is_other_ancestor)
137
def from_uncommitted(tree, other_tree, pb=None, base_tree=None):
138
"""Return a Merger for uncommitted changes in other_tree.
140
:param tree: The tree to merge into
141
:param other_tree: The tree to get uncommitted changes from
142
:param pb: A progress indicator
143
:param base_tree: The basis to use for the merge. If unspecified,
144
other_tree.basis_tree() will be used.
146
if base_tree is None:
147
base_tree = other_tree.basis_tree()
148
merger = Merger(tree.branch, other_tree, base_tree, tree, pb)
149
merger.base_rev_id = merger.base_tree.get_revision_id()
150
merger.other_rev_id = None
151
merger.other_basis = merger.base_rev_id
155
def from_mergeable(klass, tree, mergeable, pb):
156
"""Return a Merger for a bundle or merge directive.
158
:param tree: The tree to merge changes into
159
:param mergeable: A merge directive or bundle
160
:param pb: A progress indicator
162
mergeable.install_revisions(tree.branch.repository)
163
base_revision_id, other_revision_id, verified =\
164
mergeable.get_merge_request(tree.branch.repository)
165
revision_graph = tree.branch.repository.get_graph()
166
if base_revision_id is not None:
167
if (base_revision_id != _mod_revision.NULL_REVISION and
168
revision_graph.is_ancestor(
169
base_revision_id, tree.branch.last_revision())):
170
base_revision_id = None
172
warning('Performing cherrypick')
173
merger = klass.from_revision_ids(pb, tree, other_revision_id,
174
base_revision_id, revision_graph=
176
return merger, verified
179
def from_revision_ids(pb, tree, other, base=None, other_branch=None,
180
base_branch=None, revision_graph=None,
182
"""Return a Merger for revision-ids.
184
:param pb: A progress indicator
185
:param tree: The tree to merge changes into
186
:param other: The revision-id to use as OTHER
187
:param base: The revision-id to use as BASE. If not specified, will
189
:param other_branch: A branch containing the other revision-id. If
190
not supplied, tree.branch is used.
191
:param base_branch: A branch containing the base revision-id. If
192
not supplied, other_branch or tree.branch will be used.
193
:param revision_graph: If you have a revision_graph precomputed, pass
194
it in, otherwise it will be created for you.
195
:param tree_branch: The branch associated with tree. If not supplied,
196
tree.branch will be used.
198
if tree_branch is None:
199
tree_branch = tree.branch
200
merger = Merger(tree_branch, this_tree=tree, pb=pb,
201
revision_graph=revision_graph)
202
if other_branch is None:
203
other_branch = tree.branch
204
merger.set_other_revision(other, other_branch)
208
if base_branch is None:
209
base_branch = other_branch
210
merger.set_base_revision(base, base_branch)
213
def revision_tree(self, revision_id, branch=None):
214
if revision_id not in self._cached_trees:
216
branch = self.this_branch
218
tree = self.this_tree.revision_tree(revision_id)
219
except errors.NoSuchRevisionInTree:
220
tree = branch.repository.revision_tree(revision_id)
221
self._cached_trees[revision_id] = tree
222
return self._cached_trees[revision_id]
224
def _get_tree(self, treespec, possible_transports=None):
225
from bzrlib import workingtree
226
location, revno = treespec
228
tree = workingtree.WorkingTree.open_containing(location)[0]
229
return tree.branch, tree
230
branch = Branch.open_containing(location, possible_transports)[0]
232
revision_id = branch.last_revision()
234
revision_id = branch.get_rev_id(revno)
235
revision_id = ensure_null(revision_id)
236
return branch, self.revision_tree(revision_id, branch)
238
def ensure_revision_trees(self):
239
if self.this_revision_tree is None:
240
self.this_basis_tree = self.revision_tree(self.this_basis)
241
if self.this_basis == self.this_rev_id:
242
self.this_revision_tree = self.this_basis_tree
244
if self.other_rev_id is None:
245
other_basis_tree = self.revision_tree(self.other_basis)
246
changes = other_basis_tree.changes_from(self.other_tree)
247
if changes.has_changed():
248
raise WorkingTreeNotRevision(self.this_tree)
249
other_rev_id = self.other_basis
250
self.other_tree = other_basis_tree
252
def file_revisions(self, file_id):
253
self.ensure_revision_trees()
254
def get_id(tree, file_id):
255
revision_id = tree.inventory[file_id].revision
257
if self.this_rev_id is None:
258
if self.this_basis_tree.get_file_sha1(file_id) != \
259
self.this_tree.get_file_sha1(file_id):
260
raise WorkingTreeNotRevision(self.this_tree)
262
trees = (self.this_basis_tree, self.other_tree)
263
return [get_id(tree, file_id) for tree in trees]
265
def check_basis(self, check_clean, require_commits=True):
266
if self.this_basis is None and require_commits is True:
267
raise BzrCommandError("This branch has no commits."
268
" (perhaps you would prefer 'bzr pull')")
271
if self.this_basis != self.this_rev_id:
272
raise errors.UncommittedChanges(self.this_tree)
274
def compare_basis(self):
276
basis_tree = self.revision_tree(self.this_tree.last_revision())
277
except errors.NoSuchRevision:
278
basis_tree = self.this_tree.basis_tree()
279
changes = self.this_tree.changes_from(basis_tree)
280
if not changes.has_changed():
281
self.this_rev_id = self.this_basis
283
def set_interesting_files(self, file_list):
284
self.interesting_files = file_list
286
def set_pending(self):
287
if not self.base_is_ancestor or not self.base_is_other_ancestor or self.other_rev_id is None:
291
def _add_parent(self):
292
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
293
new_parent_trees = []
294
for revision_id in new_parents:
296
tree = self.revision_tree(revision_id)
297
except errors.NoSuchRevision:
301
new_parent_trees.append((revision_id, tree))
303
self.this_tree.set_parent_trees(new_parent_trees,
304
allow_leftmost_as_ghost=True)
306
for _revision_id, tree in new_parent_trees:
310
def set_other(self, other_revision, possible_transports=None):
311
"""Set the revision and tree to merge from.
313
This sets the other_tree, other_rev_id, other_basis attributes.
315
:param other_revision: The [path, revision] list to merge from.
317
self.other_branch, self.other_tree = self._get_tree(other_revision,
319
if other_revision[1] == -1:
320
self.other_rev_id = _mod_revision.ensure_null(
321
self.other_branch.last_revision())
322
if _mod_revision.is_null(self.other_rev_id):
323
raise NoCommits(self.other_branch)
324
self.other_basis = self.other_rev_id
325
elif other_revision[1] is not None:
326
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
327
self.other_basis = self.other_rev_id
329
self.other_rev_id = None
330
self.other_basis = self.other_branch.last_revision()
331
if self.other_basis is None:
332
raise NoCommits(self.other_branch)
333
if self.other_rev_id is not None:
334
self._cached_trees[self.other_rev_id] = self.other_tree
335
self._maybe_fetch(self.other_branch,self.this_branch, self.other_basis)
337
def set_other_revision(self, revision_id, other_branch):
338
"""Set 'other' based on a branch and revision id
340
:param revision_id: The revision to use for a tree
341
:param other_branch: The branch containing this tree
343
self.other_rev_id = revision_id
344
self.other_branch = other_branch
345
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
346
self.other_tree = self.revision_tree(revision_id)
347
self.other_basis = revision_id
349
def set_base_revision(self, revision_id, branch):
350
"""Set 'base' based on a branch and revision id
352
:param revision_id: The revision to use for a tree
353
:param branch: The branch containing this tree
355
self.base_rev_id = revision_id
356
self.base_branch = branch
357
self._maybe_fetch(branch, self.this_branch, revision_id)
358
self.base_tree = self.revision_tree(revision_id)
360
def _maybe_fetch(self, source, target, revision_id):
361
if not source.repository.has_same_location(target.repository):
362
target.fetch(source, revision_id)
365
revisions = [ensure_null(self.this_basis),
366
ensure_null(self.other_basis)]
367
if NULL_REVISION in revisions:
368
self.base_rev_id = NULL_REVISION
369
self.base_tree = self.revision_tree(self.base_rev_id)
370
self._is_criss_cross = False
372
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
373
self._is_criss_cross = False
375
self.base_rev_id = NULL_REVISION
377
self.base_rev_id = list(lcas)[0]
378
else: # len(lcas) > 1
380
# find_unique_lca can only handle 2 nodes, so we have to
381
# start back at the beginning. It is a shame to traverse
382
# the graph again, but better than re-implementing
384
self.base_rev_id = self.revision_graph.find_unique_lca(
385
revisions[0], revisions[1])
387
self.base_rev_id = self.revision_graph.find_unique_lca(
389
self._is_criss_cross = True
390
if self.base_rev_id == NULL_REVISION:
391
raise UnrelatedBranches()
392
if self._is_criss_cross:
393
warning('Warning: criss-cross merge encountered. See bzr'
394
' help criss-cross.')
395
mutter('Criss-cross lcas: %r' % lcas)
396
interesting_revision_ids = [self.base_rev_id]
397
interesting_revision_ids.extend(lcas)
398
interesting_trees = dict((t.get_revision_id(), t)
399
for t in self.this_branch.repository.revision_trees(
400
interesting_revision_ids))
401
self._cached_trees.update(interesting_trees)
402
self.base_tree = interesting_trees.pop(self.base_rev_id)
403
sorted_lca_keys = self.revision_graph.find_merge_order(
405
self._lca_trees = [interesting_trees[key]
406
for key in sorted_lca_keys]
408
self.base_tree = self.revision_tree(self.base_rev_id)
409
self.base_is_ancestor = True
410
self.base_is_other_ancestor = True
411
mutter('Base revid: %r' % self.base_rev_id)
413
def set_base(self, base_revision):
414
"""Set the base revision to use for the merge.
416
:param base_revision: A 2-list containing a path and revision number.
418
mutter("doing merge() with no base_revision specified")
419
if base_revision == [None, None]:
422
base_branch, self.base_tree = self._get_tree(base_revision)
423
if base_revision[1] == -1:
424
self.base_rev_id = base_branch.last_revision()
425
elif base_revision[1] is None:
426
self.base_rev_id = _mod_revision.NULL_REVISION
428
self.base_rev_id = _mod_revision.ensure_null(
429
base_branch.get_rev_id(base_revision[1]))
430
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
432
def make_merger(self):
433
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
434
'other_tree': self.other_tree,
435
'interesting_ids': self.interesting_ids,
436
'interesting_files': self.interesting_files,
439
if self.merge_type.requires_base:
440
kwargs['base_tree'] = self.base_tree
441
if self.merge_type.supports_reprocess:
442
kwargs['reprocess'] = self.reprocess
444
raise BzrError("Conflict reduction is not supported for merge"
445
" type %s." % self.merge_type)
446
if self.merge_type.supports_show_base:
447
kwargs['show_base'] = self.show_base
449
raise BzrError("Showing base is not supported for this"
450
" merge type. %s" % self.merge_type)
451
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
452
and not self.base_is_other_ancestor):
453
raise errors.CannotReverseCherrypick()
454
if self.merge_type.supports_cherrypick:
455
kwargs['cherrypick'] = (not self.base_is_ancestor or
456
not self.base_is_other_ancestor)
457
if self._is_criss_cross and getattr(self.merge_type,
458
'supports_lca_trees', False):
459
kwargs['lca_trees'] = self._lca_trees
460
return self.merge_type(pb=self._pb,
461
change_reporter=self.change_reporter,
464
def _do_merge_to(self, merge):
465
if self.other_branch is not None:
466
self.other_branch.update_references(self.this_branch)
468
if self.recurse == 'down':
469
for relpath, file_id in self.this_tree.iter_references():
470
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
471
other_revision = self.other_tree.get_reference_revision(
473
if other_revision == sub_tree.last_revision():
475
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
476
sub_merge.merge_type = self.merge_type
477
other_branch = self.other_branch.reference_parent(file_id, relpath)
478
sub_merge.set_other_revision(other_revision, other_branch)
479
base_revision = self.base_tree.get_reference_revision(file_id)
480
sub_merge.base_tree = \
481
sub_tree.branch.repository.revision_tree(base_revision)
482
sub_merge.base_rev_id = base_revision
486
self.this_tree.lock_tree_write()
488
if self.base_tree is not None:
489
self.base_tree.lock_read()
491
if self.other_tree is not None:
492
self.other_tree.lock_read()
494
merge = self.make_merger()
495
self._do_merge_to(merge)
497
if self.other_tree is not None:
498
self.other_tree.unlock()
500
if self.base_tree is not None:
501
self.base_tree.unlock()
503
self.this_tree.unlock()
504
if len(merge.cooked_conflicts) == 0:
505
if not self.ignore_zero and not is_quiet():
506
note("All changes applied successfully.")
508
note("%d conflicts encountered." % len(merge.cooked_conflicts))
510
return len(merge.cooked_conflicts)
513
class _InventoryNoneEntry(object):
514
"""This represents an inventory entry which *isn't there*.
516
It simplifies the merging logic if we always have an InventoryEntry, even
517
if it isn't actually present
524
symlink_target = None
527
_none_entry = _InventoryNoneEntry()
530
class Merge3Merger(object):
531
"""Three-way merger that uses the merge3 text merger"""
533
supports_reprocess = True
534
supports_show_base = True
535
history_based = False
536
supports_cherrypick = True
537
supports_reverse_cherrypick = True
538
winner_idx = {"this": 2, "other": 1, "conflict": 1}
539
supports_lca_trees = True
541
def __init__(self, working_tree, this_tree, base_tree, other_tree,
542
interesting_ids=None, reprocess=False, show_base=False,
543
pb=DummyProgress(), pp=None, change_reporter=None,
544
interesting_files=None, do_merge=True,
545
cherrypick=False, lca_trees=None):
546
"""Initialize the merger object and perform the merge.
548
:param working_tree: The working tree to apply the merge to
549
:param this_tree: The local tree in the merge operation
550
:param base_tree: The common tree in the merge operation
551
:param other_tree: The other tree to merge changes from
552
:param interesting_ids: The file_ids of files that should be
553
participate in the merge. May not be combined with
555
:param: reprocess If True, perform conflict-reduction processing.
556
:param show_base: If True, show the base revision in text conflicts.
557
(incompatible with reprocess)
558
:param pb: A Progress bar
559
:param pp: A ProgressPhase object
560
:param change_reporter: An object that should report changes made
561
:param interesting_files: The tree-relative paths of files that should
562
participate in the merge. If these paths refer to directories,
563
the contents of those directories will also be included. May not
564
be combined with interesting_ids. If neither interesting_files nor
565
interesting_ids is specified, all files may participate in the
567
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
568
if the ancestry was found to include a criss-cross merge.
569
Otherwise should be None.
571
object.__init__(self)
572
if interesting_files is not None and interesting_ids is not None:
574
'specify either interesting_ids or interesting_files')
575
self.interesting_ids = interesting_ids
576
self.interesting_files = interesting_files
577
self.this_tree = working_tree
578
self.base_tree = base_tree
579
self.other_tree = other_tree
580
self._raw_conflicts = []
581
self.cooked_conflicts = []
582
self.reprocess = reprocess
583
self.show_base = show_base
584
self._lca_trees = lca_trees
585
# Uncommenting this will change the default algorithm to always use
586
# _entries_lca. This can be useful for running the test suite and
587
# making sure we haven't missed any corner cases.
588
# if lca_trees is None:
589
# self._lca_trees = [self.base_tree]
592
self.change_reporter = change_reporter
593
self.cherrypick = cherrypick
595
self.pp = ProgressPhase("Merge phase", 3, self.pb)
600
self.this_tree.lock_tree_write()
601
self.base_tree.lock_read()
602
self.other_tree.lock_read()
603
self.tt = TreeTransform(self.this_tree, self.pb)
606
self._compute_transform()
608
results = self.tt.apply(no_conflicts=True)
609
self.write_modified(results)
611
self.this_tree.add_conflicts(self.cooked_conflicts)
612
except UnsupportedOperation:
616
self.other_tree.unlock()
617
self.base_tree.unlock()
618
self.this_tree.unlock()
621
def make_preview_transform(self):
622
self.base_tree.lock_read()
623
self.other_tree.lock_read()
624
self.tt = TransformPreview(self.this_tree)
627
self._compute_transform()
630
self.other_tree.unlock()
631
self.base_tree.unlock()
635
def _compute_transform(self):
636
if self._lca_trees is None:
637
entries = self._entries3()
638
resolver = self._three_way
640
entries = self._entries_lca()
641
resolver = self._lca_multi_way
642
child_pb = ui.ui_factory.nested_progress_bar()
644
for num, (file_id, changed, parents3, names3,
645
executable3) in enumerate(entries):
646
child_pb.update('Preparing file merge', num, len(entries))
647
self._merge_names(file_id, parents3, names3, resolver=resolver)
649
file_status = self.merge_contents(file_id)
651
file_status = 'unmodified'
652
self._merge_executable(file_id,
653
executable3, file_status, resolver=resolver)
658
child_pb = ui.ui_factory.nested_progress_bar()
660
fs_conflicts = resolve_conflicts(self.tt, child_pb,
661
lambda t, c: conflict_pass(t, c, self.other_tree))
664
if self.change_reporter is not None:
665
from bzrlib import delta
666
delta.report_changes(
667
self.tt.iter_changes(), self.change_reporter)
668
self.cook_conflicts(fs_conflicts)
669
for conflict in self.cooked_conflicts:
673
"""Gather data about files modified between three trees.
675
Return a list of tuples of file_id, changed, parents3, names3,
676
executable3. changed is a boolean indicating whether the file contents
677
or kind were changed. parents3 is a tuple of parent ids for base,
678
other and this. names3 is a tuple of names for base, other and this.
679
executable3 is a tuple of execute-bit values for base, other and this.
682
iterator = self.other_tree.iter_changes(self.base_tree,
683
include_unchanged=True, specific_files=self.interesting_files,
684
extra_trees=[self.this_tree])
685
this_entries = dict((e.file_id, e) for p, e in
686
self.this_tree.iter_entries_by_dir(
687
self.interesting_ids))
688
for (file_id, paths, changed, versioned, parents, names, kind,
689
executable) in iterator:
690
if (self.interesting_ids is not None and
691
file_id not in self.interesting_ids):
693
entry = this_entries.get(file_id)
694
if entry is not None:
695
this_name = entry.name
696
this_parent = entry.parent_id
697
this_executable = entry.executable
701
this_executable = None
702
parents3 = parents + (this_parent,)
703
names3 = names + (this_name,)
704
executable3 = executable + (this_executable,)
705
result.append((file_id, changed, parents3, names3, executable3))
708
def _entries_lca(self):
709
"""Gather data about files modified between multiple trees.
711
This compares OTHER versus all LCA trees, and for interesting entries,
712
it then compares with THIS and BASE.
714
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
715
:return: [(file_id, changed, parents, names, executable)]
716
file_id Simple file_id of the entry
717
changed Boolean, True if the kind or contents changed
719
parents ((base, [parent_id, in, lcas]), parent_id_other,
721
names ((base, [name, in, lcas]), name_in_other, name_in_this)
722
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
724
if self.interesting_files is not None:
725
lookup_trees = [self.this_tree, self.base_tree]
726
lookup_trees.extend(self._lca_trees)
727
# I think we should include the lca trees as well
728
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
731
interesting_ids = self.interesting_ids
733
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
735
base_inventory = self.base_tree.inventory
736
this_inventory = self.this_tree.inventory
737
for path, file_id, other_ie, lca_values in walker.iter_all():
738
# Is this modified at all from any of the other trees?
740
other_ie = _none_entry
741
if interesting_ids is not None and file_id not in interesting_ids:
744
# If other_revision is found in any of the lcas, that means this
745
# node is uninteresting. This is because when merging, if there are
746
# multiple heads(), we have to create a new node. So if we didn't,
747
# we know that the ancestry is linear, and that OTHER did not
749
# See doc/developers/lca_merge_resolution.txt for details
750
other_revision = other_ie.revision
751
if other_revision is not None:
752
# We can't use this shortcut when other_revision is None,
753
# because it may be None because things are WorkingTrees, and
754
# not because it is *actually* None.
755
is_unmodified = False
756
for lca_path, ie in lca_values:
757
if ie is not None and ie.revision == other_revision:
764
for lca_path, lca_ie in lca_values:
766
lca_entries.append(_none_entry)
768
lca_entries.append(lca_ie)
770
if file_id in base_inventory:
771
base_ie = base_inventory[file_id]
773
base_ie = _none_entry
775
if file_id in this_inventory:
776
this_ie = this_inventory[file_id]
778
this_ie = _none_entry
784
for lca_ie in lca_entries:
785
lca_kinds.append(lca_ie.kind)
786
lca_parent_ids.append(lca_ie.parent_id)
787
lca_names.append(lca_ie.name)
788
lca_executable.append(lca_ie.executable)
790
kind_winner = self._lca_multi_way(
791
(base_ie.kind, lca_kinds),
792
other_ie.kind, this_ie.kind)
793
parent_id_winner = self._lca_multi_way(
794
(base_ie.parent_id, lca_parent_ids),
795
other_ie.parent_id, this_ie.parent_id)
796
name_winner = self._lca_multi_way(
797
(base_ie.name, lca_names),
798
other_ie.name, this_ie.name)
800
content_changed = True
801
if kind_winner == 'this':
802
# No kind change in OTHER, see if there are *any* changes
803
if other_ie.kind == 'directory':
804
if parent_id_winner == 'this' and name_winner == 'this':
805
# No change for this directory in OTHER, skip
807
content_changed = False
808
elif other_ie.kind is None or other_ie.kind == 'file':
809
def get_sha1(ie, tree):
810
if ie.kind != 'file':
812
return tree.get_file_sha1(file_id)
813
base_sha1 = get_sha1(base_ie, self.base_tree)
814
lca_sha1s = [get_sha1(ie, tree) for ie, tree
815
in zip(lca_entries, self._lca_trees)]
816
this_sha1 = get_sha1(this_ie, self.this_tree)
817
other_sha1 = get_sha1(other_ie, self.other_tree)
818
sha1_winner = self._lca_multi_way(
819
(base_sha1, lca_sha1s), other_sha1, this_sha1,
820
allow_overriding_lca=False)
821
exec_winner = self._lca_multi_way(
822
(base_ie.executable, lca_executable),
823
other_ie.executable, this_ie.executable)
824
if (parent_id_winner == 'this' and name_winner == 'this'
825
and sha1_winner == 'this' and exec_winner == 'this'):
826
# No kind, parent, name, exec, or content change for
827
# OTHER, so this node is not considered interesting
829
if sha1_winner == 'this':
830
content_changed = False
831
elif other_ie.kind == 'symlink':
832
def get_target(ie, tree):
833
if ie.kind != 'symlink':
835
return tree.get_symlink_target(file_id)
836
base_target = get_target(base_ie, self.base_tree)
837
lca_targets = [get_target(ie, tree) for ie, tree
838
in zip(lca_entries, self._lca_trees)]
839
this_target = get_target(this_ie, self.this_tree)
840
other_target = get_target(other_ie, self.other_tree)
841
target_winner = self._lca_multi_way(
842
(base_target, lca_targets),
843
other_target, this_target)
844
if (parent_id_winner == 'this' and name_winner == 'this'
845
and target_winner == 'this'):
846
# No kind, parent, name, or symlink target change
849
if target_winner == 'this':
850
content_changed = False
851
elif other_ie.kind == 'tree-reference':
852
# The 'changed' information seems to be handled at a higher
853
# level. At least, _entries3 returns False for content
854
# changed, even when at a new revision_id.
855
content_changed = False
856
if (parent_id_winner == 'this' and name_winner == 'this'):
857
# Nothing interesting
860
raise AssertionError('unhandled kind: %s' % other_ie.kind)
861
# XXX: We need to handle kind == 'symlink'
863
# If we have gotten this far, that means something has changed
864
result.append((file_id, content_changed,
865
((base_ie.parent_id, lca_parent_ids),
866
other_ie.parent_id, this_ie.parent_id),
867
((base_ie.name, lca_names),
868
other_ie.name, this_ie.name),
869
((base_ie.executable, lca_executable),
870
other_ie.executable, this_ie.executable)
877
self.tt.final_kind(self.tt.root)
879
self.tt.cancel_deletion(self.tt.root)
880
if self.tt.final_file_id(self.tt.root) is None:
881
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
883
other_root_file_id = self.other_tree.get_root_id()
884
if other_root_file_id is None:
886
other_root = self.tt.trans_id_file_id(other_root_file_id)
887
if other_root == self.tt.root:
890
self.tt.final_kind(other_root)
893
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
894
# the other tree's root is a non-root in the current tree
896
self.reparent_children(self.other_tree.inventory.root, self.tt.root)
897
self.tt.cancel_creation(other_root)
898
self.tt.cancel_versioning(other_root)
900
def reparent_children(self, ie, target):
901
for thing, child in ie.children.iteritems():
902
trans_id = self.tt.trans_id_file_id(child.file_id)
903
self.tt.adjust_path(self.tt.final_name(trans_id), target, trans_id)
905
def write_modified(self, results):
907
for path in results.modified_paths:
908
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
911
hash = self.this_tree.get_file_sha1(file_id)
914
modified_hashes[file_id] = hash
915
self.this_tree.set_merge_modified(modified_hashes)
918
def parent(entry, file_id):
919
"""Determine the parent for a file_id (used as a key method)"""
922
return entry.parent_id
925
def name(entry, file_id):
926
"""Determine the name for a file_id (used as a key method)"""
932
def contents_sha1(tree, file_id):
933
"""Determine the sha1 of the file contents (used as a key method)."""
934
if file_id not in tree:
936
return tree.get_file_sha1(file_id)
939
def executable(tree, file_id):
940
"""Determine the executability of a file-id (used as a key method)."""
941
if file_id not in tree:
943
if tree.kind(file_id) != "file":
945
return tree.is_executable(file_id)
948
def kind(tree, file_id):
949
"""Determine the kind of a file-id (used as a key method)."""
950
if file_id not in tree:
952
return tree.kind(file_id)
955
def _three_way(base, other, this):
956
#if base == other, either they all agree, or only THIS has changed.
959
elif this not in (base, other):
961
# "Ambiguous clean merge" -- both sides have made the same change.
964
# this == base: only other has changed.
969
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
970
"""Consider LCAs when determining whether a change has occurred.
972
If LCAS are all identical, this is the same as a _three_way comparison.
974
:param bases: value in (BASE, [LCAS])
975
:param other: value in OTHER
976
:param this: value in THIS
977
:param allow_overriding_lca: If there is more than one unique lca
978
value, allow OTHER to override THIS if it has a new value, and
979
THIS only has an lca value, or vice versa. This is appropriate for
980
truly scalar values, not as much for non-scalars.
981
:return: 'this', 'other', or 'conflict' depending on whether an entry
984
# See doc/developers/lca_tree_merging.txt for details about this
987
# Either Ambiguously clean, or nothing was actually changed. We
990
base_val, lca_vals = bases
991
# Remove 'base_val' from the lca_vals, because it is not interesting
992
filtered_lca_vals = [lca_val for lca_val in lca_vals
993
if lca_val != base_val]
994
if len(filtered_lca_vals) == 0:
995
return Merge3Merger._three_way(base_val, other, this)
997
unique_lca_vals = set(filtered_lca_vals)
998
if len(unique_lca_vals) == 1:
999
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1001
if allow_overriding_lca:
1002
if other in unique_lca_vals:
1003
if this in unique_lca_vals:
1004
# Each side picked a different lca, conflict
1007
# This has a value which supersedes both lca values, and
1008
# other only has an lca value
1010
elif this in unique_lca_vals:
1011
# OTHER has a value which supersedes both lca values, and this
1012
# only has an lca value
1015
# At this point, the lcas disagree, and the tips disagree
1019
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1020
"""Do a three-way test on a scalar.
1021
Return "this", "other" or "conflict", depending whether a value wins.
1023
key_base = key(base_tree, file_id)
1024
key_other = key(other_tree, file_id)
1025
#if base == other, either they all agree, or only THIS has changed.
1026
if key_base == key_other:
1028
key_this = key(this_tree, file_id)
1029
# "Ambiguous clean merge"
1030
if key_this == key_other:
1032
elif key_this == key_base:
1037
def merge_names(self, file_id):
1038
def get_entry(tree):
1039
if file_id in tree.inventory:
1040
return tree.inventory[file_id]
1043
this_entry = get_entry(self.this_tree)
1044
other_entry = get_entry(self.other_tree)
1045
base_entry = get_entry(self.base_tree)
1046
entries = (base_entry, other_entry, this_entry)
1049
for entry in entries:
1052
parents.append(None)
1054
names.append(entry.name)
1055
parents.append(entry.parent_id)
1056
return self._merge_names(file_id, parents, names,
1057
resolver=self._three_way)
1059
def _merge_names(self, file_id, parents, names, resolver):
1060
"""Perform a merge on file_id names and parents"""
1061
base_name, other_name, this_name = names
1062
base_parent, other_parent, this_parent = parents
1064
name_winner = resolver(*names)
1066
parent_id_winner = resolver(*parents)
1067
if this_name is None:
1068
if name_winner == "this":
1069
name_winner = "other"
1070
if parent_id_winner == "this":
1071
parent_id_winner = "other"
1072
if name_winner == "this" and parent_id_winner == "this":
1074
if name_winner == "conflict":
1075
trans_id = self.tt.trans_id_file_id(file_id)
1076
self._raw_conflicts.append(('name conflict', trans_id,
1077
this_name, other_name))
1078
if parent_id_winner == "conflict":
1079
trans_id = self.tt.trans_id_file_id(file_id)
1080
self._raw_conflicts.append(('parent conflict', trans_id,
1081
this_parent, other_parent))
1082
if other_name is None:
1083
# it doesn't matter whether the result was 'other' or
1084
# 'conflict'-- if there's no 'other', we leave it alone.
1086
# if we get here, name_winner and parent_winner are set to safe values.
1087
trans_id = self.tt.trans_id_file_id(file_id)
1088
parent_id = parents[self.winner_idx[parent_id_winner]]
1089
if parent_id is not None:
1090
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1091
self.tt.adjust_path(names[self.winner_idx[name_winner]],
1092
parent_trans_id, trans_id)
1094
def merge_contents(self, file_id):
1095
"""Performs a merge on file_id contents."""
1096
def contents_pair(tree):
1097
if file_id not in tree:
1099
kind = tree.kind(file_id)
1101
contents = tree.get_file_sha1(file_id)
1102
elif kind == "symlink":
1103
contents = tree.get_symlink_target(file_id)
1106
return kind, contents
1108
def contents_conflict():
1109
trans_id = self.tt.trans_id_file_id(file_id)
1110
name = self.tt.final_name(trans_id)
1111
parent_id = self.tt.final_parent(trans_id)
1112
if file_id in self.this_tree.inventory:
1113
self.tt.unversion_file(trans_id)
1114
if file_id in self.this_tree:
1115
self.tt.delete_contents(trans_id)
1116
file_group = self._dump_conflicts(name, parent_id, file_id,
1118
self._raw_conflicts.append(('contents conflict', file_group))
1120
# See SPOT run. run, SPOT, run.
1121
# So we're not QUITE repeating ourselves; we do tricky things with
1123
base_pair = contents_pair(self.base_tree)
1124
other_pair = contents_pair(self.other_tree)
1126
this_pair = contents_pair(self.this_tree)
1127
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1128
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1129
this_pair, allow_overriding_lca=False)
1131
if base_pair == other_pair:
1134
# We delayed evaluating this_pair as long as we can to avoid
1135
# unnecessary sha1 calculation
1136
this_pair = contents_pair(self.this_tree)
1137
winner = self._three_way(base_pair, other_pair, this_pair)
1138
if winner == 'this':
1139
# No interesting changes introduced by OTHER
1141
trans_id = self.tt.trans_id_file_id(file_id)
1142
if winner == 'other':
1143
# OTHER is a straight winner, so replace this contents with other
1144
file_in_this = file_id in self.this_tree
1146
# Remove any existing contents
1147
self.tt.delete_contents(trans_id)
1148
if file_id in self.other_tree:
1149
# OTHER changed the file
1150
create_from_tree(self.tt, trans_id,
1151
self.other_tree, file_id)
1152
if not file_in_this:
1153
self.tt.version_file(file_id, trans_id)
1156
# OTHER deleted the file
1157
self.tt.unversion_file(trans_id)
1160
# We have a hypothetical conflict, but if we have files, then we
1161
# can try to merge the content
1162
if this_pair[0] == 'file' and other_pair[0] == 'file':
1163
# THIS and OTHER are both files, so text merge. Either
1164
# BASE is a file, or both converted to files, so at least we
1165
# have agreement that output should be a file.
1167
self.text_merge(file_id, trans_id)
1169
return contents_conflict()
1170
if file_id not in self.this_tree:
1171
self.tt.version_file(file_id, trans_id)
1173
self.tt.tree_kind(trans_id)
1174
self.tt.delete_contents(trans_id)
1179
return contents_conflict()
1181
def get_lines(self, tree, file_id):
1182
"""Return the lines in a file, or an empty list."""
1184
return tree.get_file(file_id).readlines()
1188
def text_merge(self, file_id, trans_id):
1189
"""Perform a three-way text merge on a file_id"""
1190
# it's possible that we got here with base as a different type.
1191
# if so, we just want two-way text conflicts.
1192
if file_id in self.base_tree and \
1193
self.base_tree.kind(file_id) == "file":
1194
base_lines = self.get_lines(self.base_tree, file_id)
1197
other_lines = self.get_lines(self.other_tree, file_id)
1198
this_lines = self.get_lines(self.this_tree, file_id)
1199
m3 = Merge3(base_lines, this_lines, other_lines,
1200
is_cherrypick=self.cherrypick)
1201
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1202
if self.show_base is True:
1203
base_marker = '|' * 7
1207
def iter_merge3(retval):
1208
retval["text_conflicts"] = False
1209
for line in m3.merge_lines(name_a = "TREE",
1210
name_b = "MERGE-SOURCE",
1211
name_base = "BASE-REVISION",
1212
start_marker=start_marker,
1213
base_marker=base_marker,
1214
reprocess=self.reprocess):
1215
if line.startswith(start_marker):
1216
retval["text_conflicts"] = True
1217
yield line.replace(start_marker, '<' * 7)
1221
merge3_iterator = iter_merge3(retval)
1222
self.tt.create_file(merge3_iterator, trans_id)
1223
if retval["text_conflicts"] is True:
1224
self._raw_conflicts.append(('text conflict', trans_id))
1225
name = self.tt.final_name(trans_id)
1226
parent_id = self.tt.final_parent(trans_id)
1227
file_group = self._dump_conflicts(name, parent_id, file_id,
1228
this_lines, base_lines,
1230
file_group.append(trans_id)
1232
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1233
base_lines=None, other_lines=None, set_version=False,
1235
"""Emit conflict files.
1236
If this_lines, base_lines, or other_lines are omitted, they will be
1237
determined automatically. If set_version is true, the .OTHER, .THIS
1238
or .BASE (in that order) will be created as versioned files.
1240
data = [('OTHER', self.other_tree, other_lines),
1241
('THIS', self.this_tree, this_lines)]
1243
data.append(('BASE', self.base_tree, base_lines))
1246
for suffix, tree, lines in data:
1248
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1250
file_group.append(trans_id)
1251
if set_version and not versioned:
1252
self.tt.version_file(file_id, trans_id)
1256
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1258
"""Emit a single conflict file."""
1259
name = name + '.' + suffix
1260
trans_id = self.tt.create_path(name, parent_id)
1261
create_from_tree(self.tt, trans_id, tree, file_id, lines)
1264
def merge_executable(self, file_id, file_status):
1265
"""Perform a merge on the execute bit."""
1266
executable = [self.executable(t, file_id) for t in (self.base_tree,
1267
self.other_tree, self.this_tree)]
1268
self._merge_executable(file_id, executable, file_status,
1269
resolver=self._three_way)
1271
def _merge_executable(self, file_id, executable, file_status,
1273
"""Perform a merge on the execute bit."""
1274
base_executable, other_executable, this_executable = executable
1275
if file_status == "deleted":
1277
winner = resolver(*executable)
1278
if winner == "conflict":
1279
# There must be a None in here, if we have a conflict, but we
1280
# need executability since file status was not deleted.
1281
if self.executable(self.other_tree, file_id) is None:
1285
if winner == 'this' and file_status != "modified":
1287
trans_id = self.tt.trans_id_file_id(file_id)
1289
if self.tt.final_kind(trans_id) != "file":
1293
if winner == "this":
1294
executability = this_executable
1296
if file_id in self.other_tree:
1297
executability = other_executable
1298
elif file_id in self.this_tree:
1299
executability = this_executable
1300
elif file_id in self.base_tree:
1301
executability = base_executable
1302
if executability is not None:
1303
trans_id = self.tt.trans_id_file_id(file_id)
1304
self.tt.set_executability(executability, trans_id)
1306
def cook_conflicts(self, fs_conflicts):
1307
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1308
from conflicts import Conflict
1310
self.cooked_conflicts.extend(cook_conflicts(fs_conflicts, self.tt))
1311
fp = FinalPaths(self.tt)
1312
for conflict in self._raw_conflicts:
1313
conflict_type = conflict[0]
1314
if conflict_type in ('name conflict', 'parent conflict'):
1315
trans_id = conflict[1]
1316
conflict_args = conflict[2:]
1317
if trans_id not in name_conflicts:
1318
name_conflicts[trans_id] = {}
1319
unique_add(name_conflicts[trans_id], conflict_type,
1321
if conflict_type == 'contents conflict':
1322
for trans_id in conflict[1]:
1323
file_id = self.tt.final_file_id(trans_id)
1324
if file_id is not None:
1326
path = fp.get_path(trans_id)
1327
for suffix in ('.BASE', '.THIS', '.OTHER'):
1328
if path.endswith(suffix):
1329
path = path[:-len(suffix)]
1331
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
1332
self.cooked_conflicts.append(c)
1333
if conflict_type == 'text conflict':
1334
trans_id = conflict[1]
1335
path = fp.get_path(trans_id)
1336
file_id = self.tt.final_file_id(trans_id)
1337
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
1338
self.cooked_conflicts.append(c)
1340
for trans_id, conflicts in name_conflicts.iteritems():
1342
this_parent, other_parent = conflicts['parent conflict']
1343
if this_parent == other_parent:
1344
raise AssertionError()
1346
this_parent = other_parent = \
1347
self.tt.final_file_id(self.tt.final_parent(trans_id))
1349
this_name, other_name = conflicts['name conflict']
1350
if this_name == other_name:
1351
raise AssertionError()
1353
this_name = other_name = self.tt.final_name(trans_id)
1354
other_path = fp.get_path(trans_id)
1355
if this_parent is not None and this_name is not None:
1356
this_parent_path = \
1357
fp.get_path(self.tt.trans_id_file_id(this_parent))
1358
this_path = pathjoin(this_parent_path, this_name)
1360
this_path = "<deleted>"
1361
file_id = self.tt.final_file_id(trans_id)
1362
c = Conflict.factory('path conflict', path=this_path,
1363
conflict_path=other_path, file_id=file_id)
1364
self.cooked_conflicts.append(c)
1365
self.cooked_conflicts.sort(key=Conflict.sort_key)
1368
class WeaveMerger(Merge3Merger):
1369
"""Three-way tree merger, text weave merger."""
1370
supports_reprocess = True
1371
supports_show_base = False
1372
supports_reverse_cherrypick = False
1373
history_based = True
1375
def _merged_lines(self, file_id):
1376
"""Generate the merged lines.
1377
There is no distinction between lines that are meant to contain <<<<<<<
1381
base = self.base_tree
1384
plan = self.this_tree.plan_file_merge(file_id, self.other_tree,
1386
if 'merge' in debug.debug_flags:
1388
trans_id = self.tt.trans_id_file_id(file_id)
1389
name = self.tt.final_name(trans_id) + '.plan'
1390
contents = ('%10s|%s' % l for l in plan)
1391
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1392
textmerge = PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1393
'>>>>>>> MERGE-SOURCE\n')
1394
return textmerge.merge_lines(self.reprocess)
1396
def text_merge(self, file_id, trans_id):
1397
"""Perform a (weave) text merge for a given file and file-id.
1398
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1399
and a conflict will be noted.
1401
lines, conflicts = self._merged_lines(file_id)
1403
# Note we're checking whether the OUTPUT is binary in this case,
1404
# because we don't want to get into weave merge guts.
1405
check_text_lines(lines)
1406
self.tt.create_file(lines, trans_id)
1408
self._raw_conflicts.append(('text conflict', trans_id))
1409
name = self.tt.final_name(trans_id)
1410
parent_id = self.tt.final_parent(trans_id)
1411
file_group = self._dump_conflicts(name, parent_id, file_id,
1413
file_group.append(trans_id)
1416
class LCAMerger(WeaveMerger):
1418
def _merged_lines(self, file_id):
1419
"""Generate the merged lines.
1420
There is no distinction between lines that are meant to contain <<<<<<<
1424
base = self.base_tree
1427
plan = self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1429
if 'merge' in debug.debug_flags:
1431
trans_id = self.tt.trans_id_file_id(file_id)
1432
name = self.tt.final_name(trans_id) + '.plan'
1433
contents = ('%10s|%s' % l for l in plan)
1434
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1435
textmerge = PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1436
'>>>>>>> MERGE-SOURCE\n')
1437
return textmerge.merge_lines(self.reprocess)
1440
class Diff3Merger(Merge3Merger):
1441
"""Three-way merger using external diff3 for text merging"""
1443
def dump_file(self, temp_dir, name, tree, file_id):
1444
out_path = pathjoin(temp_dir, name)
1445
out_file = open(out_path, "wb")
1447
in_file = tree.get_file(file_id)
1448
for line in in_file:
1449
out_file.write(line)
1454
def text_merge(self, file_id, trans_id):
1455
"""Perform a diff3 merge using a specified file-id and trans-id.
1456
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1457
will be dumped, and a will be conflict noted.
1460
temp_dir = osutils.mkdtemp(prefix="bzr-")
1462
new_file = pathjoin(temp_dir, "new")
1463
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1464
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1465
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1466
status = bzrlib.patch.diff3(new_file, this, base, other)
1467
if status not in (0, 1):
1468
raise BzrError("Unhandled diff3 exit code")
1469
f = open(new_file, 'rb')
1471
self.tt.create_file(f, trans_id)
1475
name = self.tt.final_name(trans_id)
1476
parent_id = self.tt.final_parent(trans_id)
1477
self._dump_conflicts(name, parent_id, file_id)
1478
self._raw_conflicts.append(('text conflict', trans_id))
1480
osutils.rmtree(temp_dir)
1483
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1485
merge_type=Merge3Merger,
1486
interesting_ids=None,
1490
interesting_files=None,
1493
change_reporter=None):
1494
"""Primary interface for merging.
1496
typical use is probably
1497
'merge_inner(branch, branch.get_revision_tree(other_revision),
1498
branch.get_revision_tree(base_revision))'
1500
if this_tree is None:
1501
raise BzrError("bzrlib.merge.merge_inner requires a this_tree "
1502
"parameter as of bzrlib version 0.8.")
1503
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1504
pb=pb, change_reporter=change_reporter)
1505
merger.backup_files = backup_files
1506
merger.merge_type = merge_type
1507
merger.interesting_ids = interesting_ids
1508
merger.ignore_zero = ignore_zero
1509
if interesting_files:
1511
raise ValueError('Only supply interesting_ids'
1512
' or interesting_files')
1513
merger.interesting_files = interesting_files
1514
merger.show_base = show_base
1515
merger.reprocess = reprocess
1516
merger.other_rev_id = other_rev_id
1517
merger.other_basis = other_rev_id
1518
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1519
if get_revision_id is None:
1520
get_revision_id = base_tree.last_revision
1521
merger.set_base_revision(get_revision_id(), this_branch)
1522
return merger.do_merge()
1524
def get_merge_type_registry():
1525
"""Merge type registry is in bzrlib.option to avoid circular imports.
1527
This method provides a sanctioned way to retrieve it.
1529
from bzrlib import option
1530
return option._merge_type_registry
1533
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1534
def status_a(revision, text):
1535
if revision in ancestors_b:
1536
return 'killed-b', text
1538
return 'new-a', text
1540
def status_b(revision, text):
1541
if revision in ancestors_a:
1542
return 'killed-a', text
1544
return 'new-b', text
1546
plain_a = [t for (a, t) in annotated_a]
1547
plain_b = [t for (a, t) in annotated_b]
1548
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1549
blocks = matcher.get_matching_blocks()
1552
for ai, bi, l in blocks:
1553
# process all mismatched sections
1554
# (last mismatched section is handled because blocks always
1555
# includes a 0-length last block)
1556
for revision, text in annotated_a[a_cur:ai]:
1557
yield status_a(revision, text)
1558
for revision, text in annotated_b[b_cur:bi]:
1559
yield status_b(revision, text)
1560
# and now the matched section
1563
for text_a in plain_a[ai:a_cur]:
1564
yield "unchanged", text_a
1567
class _PlanMergeBase(object):
1569
def __init__(self, a_rev, b_rev, vf, key_prefix):
1572
:param a_rev: Revision-id of one revision to merge
1573
:param b_rev: Revision-id of the other revision to merge
1574
:param vf: A VersionedFiles containing both revisions
1575
:param key_prefix: A prefix for accessing keys in vf, typically
1581
self._last_lines = None
1582
self._last_lines_revision_id = None
1583
self._cached_matching_blocks = {}
1584
self._key_prefix = key_prefix
1585
self._precache_tip_lines()
1587
def _precache_tip_lines(self):
1588
lines = self.get_lines([self.a_rev, self.b_rev])
1589
self.lines_a = lines[self.a_rev]
1590
self.lines_b = lines[self.b_rev]
1592
def get_lines(self, revisions):
1593
"""Get lines for revisions from the backing VersionedFiles.
1595
:raises RevisionNotPresent: on absent texts.
1597
keys = [(self._key_prefix + (rev,)) for rev in revisions]
1599
for record in self.vf.get_record_stream(keys, 'unordered', True):
1600
if record.storage_kind == 'absent':
1601
raise errors.RevisionNotPresent(record.key, self.vf)
1602
result[record.key[-1]] = osutils.chunks_to_lines(
1603
record.get_bytes_as('chunked'))
1606
def plan_merge(self):
1607
"""Generate a 'plan' for merging the two revisions.
1609
This involves comparing their texts and determining the cause of
1610
differences. If text A has a line and text B does not, then either the
1611
line was added to text A, or it was deleted from B. Once the causes
1612
are combined, they are written out in the format described in
1613
VersionedFile.plan_merge
1615
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
1616
unique_a, unique_b = self._unique_lines(blocks)
1617
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
1618
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
1619
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
1621
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
1624
for i, j, n in blocks:
1625
for a_index in range(last_i, i):
1626
if a_index in new_a:
1627
if a_index in killed_b:
1628
yield 'conflicted-a', self.lines_a[a_index]
1630
yield 'new-a', self.lines_a[a_index]
1632
yield 'killed-b', self.lines_a[a_index]
1633
for b_index in range(last_j, j):
1634
if b_index in new_b:
1635
if b_index in killed_a:
1636
yield 'conflicted-b', self.lines_b[b_index]
1638
yield 'new-b', self.lines_b[b_index]
1640
yield 'killed-a', self.lines_b[b_index]
1641
# handle common lines
1642
for a_index in range(i, i+n):
1643
yield 'unchanged', self.lines_a[a_index]
1647
def _get_matching_blocks(self, left_revision, right_revision):
1648
"""Return a description of which sections of two revisions match.
1650
See SequenceMatcher.get_matching_blocks
1652
cached = self._cached_matching_blocks.get((left_revision,
1654
if cached is not None:
1656
if self._last_lines_revision_id == left_revision:
1657
left_lines = self._last_lines
1658
right_lines = self.get_lines([right_revision])[right_revision]
1660
lines = self.get_lines([left_revision, right_revision])
1661
left_lines = lines[left_revision]
1662
right_lines = lines[right_revision]
1663
self._last_lines = right_lines
1664
self._last_lines_revision_id = right_revision
1665
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
1667
return matcher.get_matching_blocks()
1669
def _unique_lines(self, matching_blocks):
1670
"""Analyse matching_blocks to determine which lines are unique
1672
:return: a tuple of (unique_left, unique_right), where the values are
1673
sets of line numbers of unique lines.
1679
for i, j, n in matching_blocks:
1680
unique_left.extend(range(last_i, i))
1681
unique_right.extend(range(last_j, j))
1684
return unique_left, unique_right
1687
def _subtract_plans(old_plan, new_plan):
1688
"""Remove changes from new_plan that came from old_plan.
1690
It is assumed that the difference between the old_plan and new_plan
1691
is their choice of 'b' text.
1693
All lines from new_plan that differ from old_plan are emitted
1694
verbatim. All lines from new_plan that match old_plan but are
1695
not about the 'b' revision are emitted verbatim.
1697
Lines that match and are about the 'b' revision are the lines we
1698
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
1699
is skipped entirely.
1701
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
1704
for i, j, n in matcher.get_matching_blocks():
1705
for jj in range(last_j, j):
1707
for jj in range(j, j+n):
1708
plan_line = new_plan[jj]
1709
if plan_line[0] == 'new-b':
1711
elif plan_line[0] == 'killed-b':
1712
yield 'unchanged', plan_line[1]
1718
class _PlanMerge(_PlanMergeBase):
1719
"""Plan an annotate merge using on-the-fly annotation"""
1721
def __init__(self, a_rev, b_rev, vf, key_prefix):
1722
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
1723
self.a_key = self._key_prefix + (self.a_rev,)
1724
self.b_key = self._key_prefix + (self.b_rev,)
1725
self.graph = Graph(self.vf)
1726
heads = self.graph.heads((self.a_key, self.b_key))
1728
# one side dominates, so we can just return its values, yay for
1730
# Ideally we would know that before we get this far
1731
self._head_key = heads.pop()
1732
if self._head_key == self.a_key:
1736
mutter('found dominating revision for %s\n%s > %s', self.vf,
1737
self._head_key[-1], other)
1740
self._head_key = None
1743
def _precache_tip_lines(self):
1744
# Turn this into a no-op, because we will do this later
1747
def _find_recursive_lcas(self):
1748
"""Find all the ancestors back to a unique lca"""
1749
cur_ancestors = (self.a_key, self.b_key)
1750
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
1751
# rather than a key tuple. We will just map that directly to no common
1755
next_lcas = self.graph.find_lca(*cur_ancestors)
1756
# Map a plain NULL_REVISION to a simple no-ancestors
1757
if next_lcas == set([NULL_REVISION]):
1759
# Order the lca's based on when they were merged into the tip
1760
# While the actual merge portion of weave merge uses a set() of
1761
# active revisions, the order of insertion *does* effect the
1762
# implicit ordering of the texts.
1763
for rev_key in cur_ancestors:
1764
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
1766
parent_map[rev_key] = ordered_parents
1767
if len(next_lcas) == 0:
1769
elif len(next_lcas) == 1:
1770
parent_map[list(next_lcas)[0]] = ()
1772
elif len(next_lcas) > 2:
1773
# More than 2 lca's, fall back to grabbing all nodes between
1774
# this and the unique lca.
1775
mutter('More than 2 LCAs, falling back to all nodes for:'
1776
' %s, %s\n=> %s', self.a_key, self.b_key, cur_ancestors)
1777
cur_lcas = next_lcas
1778
while len(cur_lcas) > 1:
1779
cur_lcas = self.graph.find_lca(*cur_lcas)
1780
if len(cur_lcas) == 0:
1781
# No common base to find, use the full ancestry
1784
unique_lca = list(cur_lcas)[0]
1785
if unique_lca == NULL_REVISION:
1786
# find_lca will return a plain 'NULL_REVISION' rather
1787
# than a key tuple when there is no common ancestor, we
1788
# prefer to just use None, because it doesn't confuse
1789
# _get_interesting_texts()
1791
parent_map.update(self._find_unique_parents(next_lcas,
1794
cur_ancestors = next_lcas
1797
def _find_unique_parents(self, tip_keys, base_key):
1798
"""Find ancestors of tip that aren't ancestors of base.
1800
:param tip_keys: Nodes that are interesting
1801
:param base_key: Cull all ancestors of this node
1802
:return: The parent map for all revisions between tip_keys and
1803
base_key. base_key will be included. References to nodes outside of
1804
the ancestor set will also be removed.
1806
# TODO: this would be simpler if find_unique_ancestors took a list
1807
# instead of a single tip, internally it supports it, but it
1808
# isn't a "backwards compatible" api change.
1809
if base_key is None:
1810
parent_map = dict(self.graph.iter_ancestry(tip_keys))
1811
# We remove NULL_REVISION because it isn't a proper tuple key, and
1812
# thus confuses things like _get_interesting_texts, and our logic
1813
# to add the texts into the memory weave.
1814
if NULL_REVISION in parent_map:
1815
parent_map.pop(NULL_REVISION)
1818
for tip in tip_keys:
1820
self.graph.find_unique_ancestors(tip, [base_key]))
1821
parent_map = self.graph.get_parent_map(interesting)
1822
parent_map[base_key] = ()
1823
culled_parent_map, child_map, tails = self._remove_external_references(
1825
# Remove all the tails but base_key
1826
if base_key is not None:
1827
tails.remove(base_key)
1828
self._prune_tails(culled_parent_map, child_map, tails)
1829
# Now remove all the uninteresting 'linear' regions
1830
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
1834
def _remove_external_references(parent_map):
1835
"""Remove references that go outside of the parent map.
1837
:param parent_map: Something returned from Graph.get_parent_map(keys)
1838
:return: (filtered_parent_map, child_map, tails)
1839
filtered_parent_map is parent_map without external references
1840
child_map is the {parent_key: [child_keys]} mapping
1841
tails is a list of nodes that do not have any parents in the map
1843
# TODO: The basic effect of this function seems more generic than
1844
# _PlanMerge. But the specific details of building a child_map,
1845
# and computing tails seems very specific to _PlanMerge.
1846
# Still, should this be in Graph land?
1847
filtered_parent_map = {}
1850
for key, parent_keys in parent_map.iteritems():
1851
culled_parent_keys = [p for p in parent_keys if p in parent_map]
1852
if not culled_parent_keys:
1854
for parent_key in culled_parent_keys:
1855
child_map.setdefault(parent_key, []).append(key)
1856
# TODO: Do we want to do this, it adds overhead for every node,
1857
# just to say that the node has no children
1858
child_map.setdefault(key, [])
1859
filtered_parent_map[key] = culled_parent_keys
1860
return filtered_parent_map, child_map, tails
1863
def _prune_tails(parent_map, child_map, tails_to_remove):
1864
"""Remove tails from the parent map.
1866
This will remove the supplied revisions until no more children have 0
1869
:param parent_map: A dict of {child: [parents]}, this dictionary will
1870
be modified in place.
1871
:param tails_to_remove: A list of tips that should be removed,
1872
this list will be consumed
1873
:param child_map: The reverse dict of parent_map ({parent: [children]})
1874
this dict will be modified
1875
:return: None, parent_map will be modified in place.
1877
while tails_to_remove:
1878
next = tails_to_remove.pop()
1879
parent_map.pop(next)
1880
children = child_map.pop(next)
1881
for child in children:
1882
child_parents = parent_map[child]
1883
child_parents.remove(next)
1884
if len(child_parents) == 0:
1885
tails_to_remove.append(child)
1887
def _get_interesting_texts(self, parent_map):
1888
"""Return a dict of texts we are interested in.
1890
Note that the input is in key tuples, but the output is in plain
1893
:param parent_map: The output from _find_recursive_lcas
1894
:return: A dict of {'revision_id':lines} as returned by
1895
_PlanMergeBase.get_lines()
1897
all_revision_keys = set(parent_map)
1898
all_revision_keys.add(self.a_key)
1899
all_revision_keys.add(self.b_key)
1901
# Everything else is in 'keys' but get_lines is in 'revision_ids'
1902
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
1905
def _build_weave(self):
1906
from bzrlib import weave
1907
self._weave = weave.Weave(weave_name='in_memory_weave',
1908
allow_reserved=True)
1909
parent_map = self._find_recursive_lcas()
1911
all_texts = self._get_interesting_texts(parent_map)
1913
# Note: Unfortunately, the order given by topo_sort will effect the
1914
# ordering resolution in the output. Specifically, if you add A then B,
1915
# then in the output text A lines will show up before B lines. And, of
1916
# course, topo_sort doesn't guarantee any real ordering.
1917
# So we use merge_sort, and add a fake node on the tip.
1918
# This ensures that left-hand parents will always be inserted into the
1919
# weave before right-hand parents.
1920
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
1921
parent_map[tip_key] = (self.a_key, self.b_key)
1923
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
1927
# for key in tsort.topo_sort(parent_map):
1928
parent_keys = parent_map[key]
1929
revision_id = key[-1]
1930
parent_ids = [k[-1] for k in parent_keys]
1931
self._weave.add_lines(revision_id, parent_ids,
1932
all_texts[revision_id])
1934
def plan_merge(self):
1935
"""Generate a 'plan' for merging the two revisions.
1937
This involves comparing their texts and determining the cause of
1938
differences. If text A has a line and text B does not, then either the
1939
line was added to text A, or it was deleted from B. Once the causes
1940
are combined, they are written out in the format described in
1941
VersionedFile.plan_merge
1943
if self._head_key is not None: # There was a single head
1944
if self._head_key == self.a_key:
1947
if self._head_key != self.b_key:
1948
raise AssertionError('There was an invalid head: %s != %s'
1949
% (self.b_key, self._head_key))
1951
head_rev = self._head_key[-1]
1952
lines = self.get_lines([head_rev])[head_rev]
1953
return ((plan, line) for line in lines)
1954
return self._weave.plan_merge(self.a_rev, self.b_rev)
1957
class _PlanLCAMerge(_PlanMergeBase):
1959
This merge algorithm differs from _PlanMerge in that:
1960
1. comparisons are done against LCAs only
1961
2. cases where a contested line is new versus one LCA but old versus
1962
another are marked as conflicts, by emitting the line as conflicted-a
1965
This is faster, and hopefully produces more useful output.
1968
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
1969
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
1970
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
1973
if lca == NULL_REVISION:
1976
self.lcas.add(lca[-1])
1977
for lca in self.lcas:
1978
if _mod_revision.is_null(lca):
1981
lca_lines = self.get_lines([lca])[lca]
1982
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
1984
blocks = list(matcher.get_matching_blocks())
1985
self._cached_matching_blocks[(a_rev, lca)] = blocks
1986
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
1988
blocks = list(matcher.get_matching_blocks())
1989
self._cached_matching_blocks[(b_rev, lca)] = blocks
1991
def _determine_status(self, revision_id, unique_line_numbers):
1992
"""Determines the status unique lines versus all lcas.
1994
Basically, determines why the line is unique to this revision.
1996
A line may be determined new, killed, or both.
1998
If a line is determined new, that means it was not present in at least
1999
one LCA, and is not present in the other merge revision.
2001
If a line is determined killed, that means the line was present in
2004
If a line is killed and new, this indicates that the two merge
2005
revisions contain differing conflict resolutions.
2006
:param revision_id: The id of the revision in which the lines are
2008
:param unique_line_numbers: The line numbers of unique lines.
2009
:return a tuple of (new_this, killed_other):
2013
unique_line_numbers = set(unique_line_numbers)
2014
for lca in self.lcas:
2015
blocks = self._get_matching_blocks(revision_id, lca)
2016
unique_vs_lca, _ignored = self._unique_lines(blocks)
2017
new.update(unique_line_numbers.intersection(unique_vs_lca))
2018
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
