1
# Copyright (C) 2005-2010 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 bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
22
branch as _mod_branch,
23
conflicts as _mod_conflicts,
30
revision as _mod_revision,
40
from bzrlib.cleanup import OperationWithCleanups
47
from bzrlib.symbol_versioning import (
51
# TODO: Report back as changes are merged in
54
def transform_tree(from_tree, to_tree, interesting_ids=None):
55
from_tree.lock_tree_write()
56
operation = OperationWithCleanups(merge_inner)
57
operation.add_cleanup(from_tree.unlock)
58
operation.run_simple(from_tree.branch, to_tree, from_tree,
59
ignore_zero=True, interesting_ids=interesting_ids, this_tree=from_tree)
62
class MergeHooks(hooks.Hooks):
65
hooks.Hooks.__init__(self)
66
self.create_hook(hooks.HookPoint('merge_file_content',
67
"Called with a bzrlib.merge.Merger object to create a per file "
68
"merge object when starting a merge. "
69
"Should return either None or a subclass of "
70
"``bzrlib.merge.AbstractPerFileMerger``. "
71
"Such objects will then be called per file "
72
"that needs to be merged (including when one "
73
"side has deleted the file and the other has changed it). "
74
"See the AbstractPerFileMerger API docs for details on how it is "
79
class AbstractPerFileMerger(object):
80
"""PerFileMerger objects are used by plugins extending merge for bzrlib.
82
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
84
:ivar merger: The Merge3Merger performing the merge.
87
def __init__(self, merger):
88
"""Create a PerFileMerger for use with merger."""
91
def merge_contents(self, merge_params):
92
"""Attempt to merge the contents of a single file.
94
:param merge_params: A bzrlib.merge.MergeHookParams
95
:return : A tuple of (status, chunks), where status is one of
96
'not_applicable', 'success', 'conflicted', or 'delete'. If status
97
is 'success' or 'conflicted', then chunks should be an iterable of
98
strings for the new file contents.
100
return ('not applicable', None)
103
class PerFileMerger(AbstractPerFileMerger):
104
"""Merge individual files when self.file_matches returns True.
106
This class is intended to be subclassed. The file_matches and
107
merge_matching methods should be overridden with concrete implementations.
110
def file_matches(self, params):
111
"""Return True if merge_matching should be called on this file.
113
Only called with merges of plain files with no clear winner.
115
Subclasses must override this.
117
raise NotImplementedError(self.file_matches)
119
def get_filename(self, params, tree):
120
"""Lookup the filename (i.e. basename, not path), given a Tree (e.g.
121
self.merger.this_tree) and a MergeHookParams.
123
return osutils.basename(tree.id2path(params.file_id))
125
def get_filepath(self, params, tree):
126
"""Calculate the path to the file in a tree.
128
:param params: A MergeHookParams describing the file to merge
129
:param tree: a Tree, e.g. self.merger.this_tree.
131
return tree.id2path(params.file_id)
133
def merge_contents(self, params):
134
"""Merge the contents of a single file."""
135
# Check whether this custom merge logic should be used.
137
# OTHER is a straight winner, rely on default merge.
138
params.winner == 'other' or
139
# THIS and OTHER aren't both files.
140
not params.is_file_merge() or
141
# The filename doesn't match *.xml
142
not self.file_matches(params)):
143
return 'not_applicable', None
144
return self.merge_matching(params)
146
def merge_matching(self, params):
147
"""Merge the contents of a single file that has matched the criteria
148
in PerFileMerger.merge_contents (is a conflict, is a file,
149
self.file_matches is True).
151
Subclasses must override this.
153
raise NotImplementedError(self.merge_matching)
156
class ConfigurableFileMerger(PerFileMerger):
157
"""Merge individual files when configured via a .conf file.
159
This is a base class for concrete custom file merging logic. Concrete
160
classes should implement ``merge_text``.
162
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
164
:ivar affected_files: The configured file paths to merge.
166
:cvar name_prefix: The prefix to use when looking up configuration
167
details. <name_prefix>_merge_files describes the files targeted by the
170
:cvar default_files: The default file paths to merge when no configuration
177
def __init__(self, merger):
178
super(ConfigurableFileMerger, self).__init__(merger)
179
self.affected_files = None
180
self.default_files = self.__class__.default_files or []
181
self.name_prefix = self.__class__.name_prefix
182
if self.name_prefix is None:
183
raise ValueError("name_prefix must be set.")
185
def file_matches(self, params):
186
"""Check whether the file should call the merge hook.
188
<name_prefix>_merge_files configuration variable is a list of files
189
that should use the hook.
191
affected_files = self.affected_files
192
if affected_files is None:
193
config = self.merger.this_branch.get_config()
194
# Until bzr provides a better policy for caching the config, we
195
# just add the part we're interested in to the params to avoid
196
# reading the config files repeatedly (bazaar.conf, location.conf,
198
config_key = self.name_prefix + '_merge_files'
199
affected_files = config.get_user_option_as_list(config_key)
200
if affected_files is None:
201
# If nothing was specified in the config, use the default.
202
affected_files = self.default_files
203
self.affected_files = affected_files
205
filepath = self.get_filepath(params, self.merger.this_tree)
206
if filepath in affected_files:
210
def merge_matching(self, params):
211
return self.merge_text(params)
213
def merge_text(self, params):
214
"""Merge the byte contents of a single file.
216
This is called after checking that the merge should be performed in
217
merge_contents, and it should behave as per
218
``bzrlib.merge.AbstractPerFileMerger.merge_contents``.
220
raise NotImplementedError(self.merge_text)
223
class MergeHookParams(object):
224
"""Object holding parameters passed to merge_file_content hooks.
226
There are some fields hooks can access:
228
:ivar file_id: the file ID of the file being merged
229
:ivar trans_id: the transform ID for the merge of this file
230
:ivar this_kind: kind of file_id in 'this' tree
231
:ivar other_kind: kind of file_id in 'other' tree
232
:ivar winner: one of 'this', 'other', 'conflict'
235
def __init__(self, merger, file_id, trans_id, this_kind, other_kind,
237
self._merger = merger
238
self.file_id = file_id
239
self.trans_id = trans_id
240
self.this_kind = this_kind
241
self.other_kind = other_kind
244
def is_file_merge(self):
245
"""True if this_kind and other_kind are both 'file'."""
246
return self.this_kind == 'file' and self.other_kind == 'file'
248
@decorators.cachedproperty
249
def base_lines(self):
250
"""The lines of the 'base' version of the file."""
251
return self._merger.get_lines(self._merger.base_tree, self.file_id)
253
@decorators.cachedproperty
254
def this_lines(self):
255
"""The lines of the 'this' version of the file."""
256
return self._merger.get_lines(self._merger.this_tree, self.file_id)
258
@decorators.cachedproperty
259
def other_lines(self):
260
"""The lines of the 'other' version of the file."""
261
return self._merger.get_lines(self._merger.other_tree, self.file_id)
264
class Merger(object):
268
def __init__(self, this_branch, other_tree=None, base_tree=None,
269
this_tree=None, pb=None, change_reporter=None,
270
recurse='down', revision_graph=None):
271
object.__init__(self)
272
self.this_branch = this_branch
273
self.this_basis = _mod_revision.ensure_null(
274
this_branch.last_revision())
275
self.this_rev_id = None
276
self.this_tree = this_tree
277
self.this_revision_tree = None
278
self.this_basis_tree = None
279
self.other_tree = other_tree
280
self.other_branch = None
281
self.base_tree = base_tree
282
self.ignore_zero = False
283
self.backup_files = False
284
self.interesting_ids = None
285
self.interesting_files = None
286
self.show_base = False
287
self.reprocess = False
289
warnings.warn("pb parameter to Merger() is deprecated and ignored")
291
self.recurse = recurse
292
self.change_reporter = change_reporter
293
self._cached_trees = {}
294
self._revision_graph = revision_graph
295
self._base_is_ancestor = None
296
self._base_is_other_ancestor = None
297
self._is_criss_cross = None
298
self._lca_trees = None
300
def cache_trees_with_revision_ids(self, trees):
301
"""Cache any tree in trees if it has a revision_id."""
302
for maybe_tree in trees:
303
if maybe_tree is None:
306
rev_id = maybe_tree.get_revision_id()
307
except AttributeError:
309
self._cached_trees[rev_id] = maybe_tree
312
def revision_graph(self):
313
if self._revision_graph is None:
314
self._revision_graph = self.this_branch.repository.get_graph()
315
return self._revision_graph
317
def _set_base_is_ancestor(self, value):
318
self._base_is_ancestor = value
320
def _get_base_is_ancestor(self):
321
if self._base_is_ancestor is None:
322
self._base_is_ancestor = self.revision_graph.is_ancestor(
323
self.base_rev_id, self.this_basis)
324
return self._base_is_ancestor
326
base_is_ancestor = property(_get_base_is_ancestor, _set_base_is_ancestor)
328
def _set_base_is_other_ancestor(self, value):
329
self._base_is_other_ancestor = value
331
def _get_base_is_other_ancestor(self):
332
if self._base_is_other_ancestor is None:
333
if self.other_basis is None:
335
self._base_is_other_ancestor = self.revision_graph.is_ancestor(
336
self.base_rev_id, self.other_basis)
337
return self._base_is_other_ancestor
339
base_is_other_ancestor = property(_get_base_is_other_ancestor,
340
_set_base_is_other_ancestor)
343
def from_uncommitted(tree, other_tree, pb=None, base_tree=None):
344
"""Return a Merger for uncommitted changes in other_tree.
346
:param tree: The tree to merge into
347
:param other_tree: The tree to get uncommitted changes from
348
:param pb: A progress indicator
349
:param base_tree: The basis to use for the merge. If unspecified,
350
other_tree.basis_tree() will be used.
352
if base_tree is None:
353
base_tree = other_tree.basis_tree()
354
merger = Merger(tree.branch, other_tree, base_tree, tree, pb)
355
merger.base_rev_id = merger.base_tree.get_revision_id()
356
merger.other_rev_id = None
357
merger.other_basis = merger.base_rev_id
361
def from_mergeable(klass, tree, mergeable, pb):
362
"""Return a Merger for a bundle or merge directive.
364
:param tree: The tree to merge changes into
365
:param mergeable: A merge directive or bundle
366
:param pb: A progress indicator
368
mergeable.install_revisions(tree.branch.repository)
369
base_revision_id, other_revision_id, verified =\
370
mergeable.get_merge_request(tree.branch.repository)
371
revision_graph = tree.branch.repository.get_graph()
372
if base_revision_id is not None:
373
if (base_revision_id != _mod_revision.NULL_REVISION and
374
revision_graph.is_ancestor(
375
base_revision_id, tree.branch.last_revision())):
376
base_revision_id = None
378
trace.warning('Performing cherrypick')
379
merger = klass.from_revision_ids(pb, tree, other_revision_id,
380
base_revision_id, revision_graph=
382
return merger, verified
385
def from_revision_ids(pb, tree, other, base=None, other_branch=None,
386
base_branch=None, revision_graph=None,
388
"""Return a Merger for revision-ids.
390
:param pb: A progress indicator
391
:param tree: The tree to merge changes into
392
:param other: The revision-id to use as OTHER
393
:param base: The revision-id to use as BASE. If not specified, will
395
:param other_branch: A branch containing the other revision-id. If
396
not supplied, tree.branch is used.
397
:param base_branch: A branch containing the base revision-id. If
398
not supplied, other_branch or tree.branch will be used.
399
:param revision_graph: If you have a revision_graph precomputed, pass
400
it in, otherwise it will be created for you.
401
:param tree_branch: The branch associated with tree. If not supplied,
402
tree.branch will be used.
404
if tree_branch is None:
405
tree_branch = tree.branch
406
merger = Merger(tree_branch, this_tree=tree, pb=pb,
407
revision_graph=revision_graph)
408
if other_branch is None:
409
other_branch = tree.branch
410
merger.set_other_revision(other, other_branch)
414
if base_branch is None:
415
base_branch = other_branch
416
merger.set_base_revision(base, base_branch)
419
def revision_tree(self, revision_id, branch=None):
420
if revision_id not in self._cached_trees:
422
branch = self.this_branch
424
tree = self.this_tree.revision_tree(revision_id)
425
except errors.NoSuchRevisionInTree:
426
tree = branch.repository.revision_tree(revision_id)
427
self._cached_trees[revision_id] = tree
428
return self._cached_trees[revision_id]
430
def _get_tree(self, treespec, possible_transports=None):
431
location, revno = treespec
433
tree = workingtree.WorkingTree.open_containing(location)[0]
434
return tree.branch, tree
435
branch = _mod_branch.Branch.open_containing(
436
location, possible_transports)[0]
438
revision_id = branch.last_revision()
440
revision_id = branch.get_rev_id(revno)
441
revision_id = _mod_revision.ensure_null(revision_id)
442
return branch, self.revision_tree(revision_id, branch)
444
@deprecated_method(deprecated_in((2, 1, 0)))
445
def ensure_revision_trees(self):
446
if self.this_revision_tree is None:
447
self.this_basis_tree = self.revision_tree(self.this_basis)
448
if self.this_basis == self.this_rev_id:
449
self.this_revision_tree = self.this_basis_tree
451
if self.other_rev_id is None:
452
other_basis_tree = self.revision_tree(self.other_basis)
453
if other_basis_tree.has_changes(self.other_tree):
454
raise errors.WorkingTreeNotRevision(self.this_tree)
455
other_rev_id = self.other_basis
456
self.other_tree = other_basis_tree
458
@deprecated_method(deprecated_in((2, 1, 0)))
459
def file_revisions(self, file_id):
460
self.ensure_revision_trees()
461
def get_id(tree, file_id):
462
revision_id = tree.inventory[file_id].revision
464
if self.this_rev_id is None:
465
if self.this_basis_tree.get_file_sha1(file_id) != \
466
self.this_tree.get_file_sha1(file_id):
467
raise errors.WorkingTreeNotRevision(self.this_tree)
469
trees = (self.this_basis_tree, self.other_tree)
470
return [get_id(tree, file_id) for tree in trees]
472
@deprecated_method(deprecated_in((2, 1, 0)))
473
def check_basis(self, check_clean, require_commits=True):
474
if self.this_basis is None and require_commits is True:
475
raise errors.BzrCommandError(
476
"This branch has no commits."
477
" (perhaps you would prefer 'bzr pull')")
480
if self.this_basis != self.this_rev_id:
481
raise errors.UncommittedChanges(self.this_tree)
483
@deprecated_method(deprecated_in((2, 1, 0)))
484
def compare_basis(self):
486
basis_tree = self.revision_tree(self.this_tree.last_revision())
487
except errors.NoSuchRevision:
488
basis_tree = self.this_tree.basis_tree()
489
if not self.this_tree.has_changes(basis_tree):
490
self.this_rev_id = self.this_basis
492
def set_interesting_files(self, file_list):
493
self.interesting_files = file_list
495
def set_pending(self):
496
if (not self.base_is_ancestor or not self.base_is_other_ancestor
497
or self.other_rev_id is None):
501
def _add_parent(self):
502
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
503
new_parent_trees = []
504
operation = OperationWithCleanups(self.this_tree.set_parent_trees)
505
for revision_id in new_parents:
507
tree = self.revision_tree(revision_id)
508
except errors.NoSuchRevision:
512
operation.add_cleanup(tree.unlock)
513
new_parent_trees.append((revision_id, tree))
514
operation.run_simple(new_parent_trees, allow_leftmost_as_ghost=True)
516
def set_other(self, other_revision, possible_transports=None):
517
"""Set the revision and tree to merge from.
519
This sets the other_tree, other_rev_id, other_basis attributes.
521
:param other_revision: The [path, revision] list to merge from.
523
self.other_branch, self.other_tree = self._get_tree(other_revision,
525
if other_revision[1] == -1:
526
self.other_rev_id = _mod_revision.ensure_null(
527
self.other_branch.last_revision())
528
if _mod_revision.is_null(self.other_rev_id):
529
raise errors.NoCommits(self.other_branch)
530
self.other_basis = self.other_rev_id
531
elif other_revision[1] is not None:
532
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
533
self.other_basis = self.other_rev_id
535
self.other_rev_id = None
536
self.other_basis = self.other_branch.last_revision()
537
if self.other_basis is None:
538
raise errors.NoCommits(self.other_branch)
539
if self.other_rev_id is not None:
540
self._cached_trees[self.other_rev_id] = self.other_tree
541
self._maybe_fetch(self.other_branch,self.this_branch, self.other_basis)
543
def set_other_revision(self, revision_id, other_branch):
544
"""Set 'other' based on a branch and revision id
546
:param revision_id: The revision to use for a tree
547
:param other_branch: The branch containing this tree
549
self.other_rev_id = revision_id
550
self.other_branch = other_branch
551
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
552
self.other_tree = self.revision_tree(revision_id)
553
self.other_basis = revision_id
555
def set_base_revision(self, revision_id, branch):
556
"""Set 'base' based on a branch and revision id
558
:param revision_id: The revision to use for a tree
559
:param branch: The branch containing this tree
561
self.base_rev_id = revision_id
562
self.base_branch = branch
563
self._maybe_fetch(branch, self.this_branch, revision_id)
564
self.base_tree = self.revision_tree(revision_id)
566
def _maybe_fetch(self, source, target, revision_id):
567
if not source.repository.has_same_location(target.repository):
569
tags_to_fetch = set(source.tags.get_reverse_tag_dict())
570
except errors.TagsNotSupported:
572
fetch_spec = _mod_graph.NotInOtherForRevs(target.repository,
573
source.repository, [revision_id],
574
if_present_ids=tags_to_fetch).execute()
575
target.fetch(source, fetch_spec=fetch_spec)
578
revisions = [_mod_revision.ensure_null(self.this_basis),
579
_mod_revision.ensure_null(self.other_basis)]
580
if _mod_revision.NULL_REVISION in revisions:
581
self.base_rev_id = _mod_revision.NULL_REVISION
582
self.base_tree = self.revision_tree(self.base_rev_id)
583
self._is_criss_cross = False
585
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
586
self._is_criss_cross = False
588
self.base_rev_id = _mod_revision.NULL_REVISION
590
self.base_rev_id = list(lcas)[0]
591
else: # len(lcas) > 1
592
self._is_criss_cross = True
594
# find_unique_lca can only handle 2 nodes, so we have to
595
# start back at the beginning. It is a shame to traverse
596
# the graph again, but better than re-implementing
598
self.base_rev_id = self.revision_graph.find_unique_lca(
599
revisions[0], revisions[1])
601
self.base_rev_id = self.revision_graph.find_unique_lca(
603
sorted_lca_keys = self.revision_graph.find_merge_order(
605
if self.base_rev_id == _mod_revision.NULL_REVISION:
606
self.base_rev_id = sorted_lca_keys[0]
608
if self.base_rev_id == _mod_revision.NULL_REVISION:
609
raise errors.UnrelatedBranches()
610
if self._is_criss_cross:
611
trace.warning('Warning: criss-cross merge encountered. See bzr'
612
' help criss-cross.')
613
trace.mutter('Criss-cross lcas: %r' % lcas)
614
if self.base_rev_id in lcas:
615
trace.mutter('Unable to find unique lca. '
616
'Fallback %r as best option.' % self.base_rev_id)
617
interesting_revision_ids = set(lcas)
618
interesting_revision_ids.add(self.base_rev_id)
619
interesting_trees = dict((t.get_revision_id(), t)
620
for t in self.this_branch.repository.revision_trees(
621
interesting_revision_ids))
622
self._cached_trees.update(interesting_trees)
623
if self.base_rev_id in lcas:
624
self.base_tree = interesting_trees[self.base_rev_id]
626
self.base_tree = interesting_trees.pop(self.base_rev_id)
627
self._lca_trees = [interesting_trees[key]
628
for key in sorted_lca_keys]
630
self.base_tree = self.revision_tree(self.base_rev_id)
631
self.base_is_ancestor = True
632
self.base_is_other_ancestor = True
633
trace.mutter('Base revid: %r' % self.base_rev_id)
635
def set_base(self, base_revision):
636
"""Set the base revision to use for the merge.
638
:param base_revision: A 2-list containing a path and revision number.
640
trace.mutter("doing merge() with no base_revision specified")
641
if base_revision == [None, None]:
644
base_branch, self.base_tree = self._get_tree(base_revision)
645
if base_revision[1] == -1:
646
self.base_rev_id = base_branch.last_revision()
647
elif base_revision[1] is None:
648
self.base_rev_id = _mod_revision.NULL_REVISION
650
self.base_rev_id = _mod_revision.ensure_null(
651
base_branch.get_rev_id(base_revision[1]))
652
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
654
def make_merger(self):
655
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
656
'other_tree': self.other_tree,
657
'interesting_ids': self.interesting_ids,
658
'interesting_files': self.interesting_files,
659
'this_branch': self.this_branch,
661
if self.merge_type.requires_base:
662
kwargs['base_tree'] = self.base_tree
663
if self.merge_type.supports_reprocess:
664
kwargs['reprocess'] = self.reprocess
666
raise errors.BzrError(
667
"Conflict reduction is not supported for merge"
668
" type %s." % self.merge_type)
669
if self.merge_type.supports_show_base:
670
kwargs['show_base'] = self.show_base
672
raise errors.BzrError("Showing base is not supported for this"
673
" merge type. %s" % self.merge_type)
674
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
675
and not self.base_is_other_ancestor):
676
raise errors.CannotReverseCherrypick()
677
if self.merge_type.supports_cherrypick:
678
kwargs['cherrypick'] = (not self.base_is_ancestor or
679
not self.base_is_other_ancestor)
680
if self._is_criss_cross and getattr(self.merge_type,
681
'supports_lca_trees', False):
682
kwargs['lca_trees'] = self._lca_trees
683
return self.merge_type(pb=None,
684
change_reporter=self.change_reporter,
687
def _do_merge_to(self):
688
merge = self.make_merger()
689
if self.other_branch is not None:
690
self.other_branch.update_references(self.this_branch)
692
if self.recurse == 'down':
693
for relpath, file_id in self.this_tree.iter_references():
694
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
695
other_revision = self.other_tree.get_reference_revision(
697
if other_revision == sub_tree.last_revision():
699
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
700
sub_merge.merge_type = self.merge_type
701
other_branch = self.other_branch.reference_parent(file_id, relpath)
702
sub_merge.set_other_revision(other_revision, other_branch)
703
base_revision = self.base_tree.get_reference_revision(file_id)
704
sub_merge.base_tree = \
705
sub_tree.branch.repository.revision_tree(base_revision)
706
sub_merge.base_rev_id = base_revision
711
operation = OperationWithCleanups(self._do_merge_to)
712
self.this_tree.lock_tree_write()
713
operation.add_cleanup(self.this_tree.unlock)
714
if self.base_tree is not None:
715
self.base_tree.lock_read()
716
operation.add_cleanup(self.base_tree.unlock)
717
if self.other_tree is not None:
718
self.other_tree.lock_read()
719
operation.add_cleanup(self.other_tree.unlock)
720
merge = operation.run_simple()
721
if len(merge.cooked_conflicts) == 0:
722
if not self.ignore_zero and not trace.is_quiet():
723
trace.note("All changes applied successfully.")
725
trace.note("%d conflicts encountered."
726
% len(merge.cooked_conflicts))
728
return len(merge.cooked_conflicts)
731
class _InventoryNoneEntry(object):
732
"""This represents an inventory entry which *isn't there*.
734
It simplifies the merging logic if we always have an InventoryEntry, even
735
if it isn't actually present
742
symlink_target = None
745
_none_entry = _InventoryNoneEntry()
748
class Merge3Merger(object):
749
"""Three-way merger that uses the merge3 text merger"""
751
supports_reprocess = True
752
supports_show_base = True
753
history_based = False
754
supports_cherrypick = True
755
supports_reverse_cherrypick = True
756
winner_idx = {"this": 2, "other": 1, "conflict": 1}
757
supports_lca_trees = True
759
def __init__(self, working_tree, this_tree, base_tree, other_tree,
760
interesting_ids=None, reprocess=False, show_base=False,
761
pb=None, pp=None, change_reporter=None,
762
interesting_files=None, do_merge=True,
763
cherrypick=False, lca_trees=None, this_branch=None):
764
"""Initialize the merger object and perform the merge.
766
:param working_tree: The working tree to apply the merge to
767
:param this_tree: The local tree in the merge operation
768
:param base_tree: The common tree in the merge operation
769
:param other_tree: The other tree to merge changes from
770
:param this_branch: The branch associated with this_tree. Defaults to
771
this_tree.branch if not supplied.
772
:param interesting_ids: The file_ids of files that should be
773
participate in the merge. May not be combined with
775
:param: reprocess If True, perform conflict-reduction processing.
776
:param show_base: If True, show the base revision in text conflicts.
777
(incompatible with reprocess)
779
:param pp: A ProgressPhase object
780
:param change_reporter: An object that should report changes made
781
:param interesting_files: The tree-relative paths of files that should
782
participate in the merge. If these paths refer to directories,
783
the contents of those directories will also be included. May not
784
be combined with interesting_ids. If neither interesting_files nor
785
interesting_ids is specified, all files may participate in the
787
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
788
if the ancestry was found to include a criss-cross merge.
789
Otherwise should be None.
791
object.__init__(self)
792
if interesting_files is not None and interesting_ids is not None:
794
'specify either interesting_ids or interesting_files')
795
if this_branch is None:
796
this_branch = this_tree.branch
797
self.interesting_ids = interesting_ids
798
self.interesting_files = interesting_files
799
self.this_tree = working_tree
800
self.base_tree = base_tree
801
self.other_tree = other_tree
802
self.this_branch = this_branch
803
self._raw_conflicts = []
804
self.cooked_conflicts = []
805
self.reprocess = reprocess
806
self.show_base = show_base
807
self._lca_trees = lca_trees
808
# Uncommenting this will change the default algorithm to always use
809
# _entries_lca. This can be useful for running the test suite and
810
# making sure we haven't missed any corner cases.
811
# if lca_trees is None:
812
# self._lca_trees = [self.base_tree]
813
self.change_reporter = change_reporter
814
self.cherrypick = cherrypick
818
warnings.warn("pp argument to Merge3Merger is deprecated")
820
warnings.warn("pb argument to Merge3Merger is deprecated")
823
operation = OperationWithCleanups(self._do_merge)
824
self.this_tree.lock_tree_write()
825
operation.add_cleanup(self.this_tree.unlock)
826
self.base_tree.lock_read()
827
operation.add_cleanup(self.base_tree.unlock)
828
self.other_tree.lock_read()
829
operation.add_cleanup(self.other_tree.unlock)
832
def _do_merge(self, operation):
833
self.tt = transform.TreeTransform(self.this_tree, None)
834
operation.add_cleanup(self.tt.finalize)
835
self._compute_transform()
836
results = self.tt.apply(no_conflicts=True)
837
self.write_modified(results)
839
self.this_tree.add_conflicts(self.cooked_conflicts)
840
except errors.UnsupportedOperation:
843
def make_preview_transform(self):
844
operation = OperationWithCleanups(self._make_preview_transform)
845
self.base_tree.lock_read()
846
operation.add_cleanup(self.base_tree.unlock)
847
self.other_tree.lock_read()
848
operation.add_cleanup(self.other_tree.unlock)
849
return operation.run_simple()
851
def _make_preview_transform(self):
852
self.tt = transform.TransformPreview(self.this_tree)
853
self._compute_transform()
856
def _compute_transform(self):
857
if self._lca_trees is None:
858
entries = self._entries3()
859
resolver = self._three_way
861
entries = self._entries_lca()
862
resolver = self._lca_multi_way
863
child_pb = ui.ui_factory.nested_progress_bar()
865
factories = Merger.hooks['merge_file_content']
866
hooks = [factory(self) for factory in factories] + [self]
867
self.active_hooks = [hook for hook in hooks if hook is not None]
868
for num, (file_id, changed, parents3, names3,
869
executable3) in enumerate(entries):
870
child_pb.update('Preparing file merge', num, len(entries))
871
self._merge_names(file_id, parents3, names3, resolver=resolver)
873
file_status = self._do_merge_contents(file_id)
875
file_status = 'unmodified'
876
self._merge_executable(file_id,
877
executable3, file_status, resolver=resolver)
881
self._finish_computing_transform()
883
def _finish_computing_transform(self):
884
"""Finalize the transform and report the changes.
886
This is the second half of _compute_transform.
888
child_pb = ui.ui_factory.nested_progress_bar()
890
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
891
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
894
if self.change_reporter is not None:
895
from bzrlib import delta
896
delta.report_changes(
897
self.tt.iter_changes(), self.change_reporter)
898
self.cook_conflicts(fs_conflicts)
899
for conflict in self.cooked_conflicts:
900
trace.warning(conflict)
903
"""Gather data about files modified between three trees.
905
Return a list of tuples of file_id, changed, parents3, names3,
906
executable3. changed is a boolean indicating whether the file contents
907
or kind were changed. parents3 is a tuple of parent ids for base,
908
other and this. names3 is a tuple of names for base, other and this.
909
executable3 is a tuple of execute-bit values for base, other and this.
912
iterator = self.other_tree.iter_changes(self.base_tree,
913
include_unchanged=True, specific_files=self.interesting_files,
914
extra_trees=[self.this_tree])
915
this_entries = dict((e.file_id, e) for p, e in
916
self.this_tree.iter_entries_by_dir(
917
self.interesting_ids))
918
for (file_id, paths, changed, versioned, parents, names, kind,
919
executable) in iterator:
920
if (self.interesting_ids is not None and
921
file_id not in self.interesting_ids):
923
entry = this_entries.get(file_id)
924
if entry is not None:
925
this_name = entry.name
926
this_parent = entry.parent_id
927
this_executable = entry.executable
931
this_executable = None
932
parents3 = parents + (this_parent,)
933
names3 = names + (this_name,)
934
executable3 = executable + (this_executable,)
935
result.append((file_id, changed, parents3, names3, executable3))
938
def _entries_lca(self):
939
"""Gather data about files modified between multiple trees.
941
This compares OTHER versus all LCA trees, and for interesting entries,
942
it then compares with THIS and BASE.
944
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
945
:return: [(file_id, changed, parents, names, executable)]
946
file_id Simple file_id of the entry
947
changed Boolean, True if the kind or contents changed
949
parents ((base, [parent_id, in, lcas]), parent_id_other,
951
names ((base, [name, in, lcas]), name_in_other, name_in_this)
952
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
954
if self.interesting_files is not None:
955
lookup_trees = [self.this_tree, self.base_tree]
956
lookup_trees.extend(self._lca_trees)
957
# I think we should include the lca trees as well
958
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
961
interesting_ids = self.interesting_ids
963
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
965
base_inventory = self.base_tree.inventory
966
this_inventory = self.this_tree.inventory
967
for path, file_id, other_ie, lca_values in walker.iter_all():
968
# Is this modified at all from any of the other trees?
970
other_ie = _none_entry
971
if interesting_ids is not None and file_id not in interesting_ids:
974
# If other_revision is found in any of the lcas, that means this
975
# node is uninteresting. This is because when merging, if there are
976
# multiple heads(), we have to create a new node. So if we didn't,
977
# we know that the ancestry is linear, and that OTHER did not
979
# See doc/developers/lca_merge_resolution.txt for details
980
other_revision = other_ie.revision
981
if other_revision is not None:
982
# We can't use this shortcut when other_revision is None,
983
# because it may be None because things are WorkingTrees, and
984
# not because it is *actually* None.
985
is_unmodified = False
986
for lca_path, ie in lca_values:
987
if ie is not None and ie.revision == other_revision:
994
for lca_path, lca_ie in lca_values:
996
lca_entries.append(_none_entry)
998
lca_entries.append(lca_ie)
1000
if file_id in base_inventory:
1001
base_ie = base_inventory[file_id]
1003
base_ie = _none_entry
1005
if file_id in this_inventory:
1006
this_ie = this_inventory[file_id]
1008
this_ie = _none_entry
1014
for lca_ie in lca_entries:
1015
lca_kinds.append(lca_ie.kind)
1016
lca_parent_ids.append(lca_ie.parent_id)
1017
lca_names.append(lca_ie.name)
1018
lca_executable.append(lca_ie.executable)
1020
kind_winner = self._lca_multi_way(
1021
(base_ie.kind, lca_kinds),
1022
other_ie.kind, this_ie.kind)
1023
parent_id_winner = self._lca_multi_way(
1024
(base_ie.parent_id, lca_parent_ids),
1025
other_ie.parent_id, this_ie.parent_id)
1026
name_winner = self._lca_multi_way(
1027
(base_ie.name, lca_names),
1028
other_ie.name, this_ie.name)
1030
content_changed = True
1031
if kind_winner == 'this':
1032
# No kind change in OTHER, see if there are *any* changes
1033
if other_ie.kind == 'directory':
1034
if parent_id_winner == 'this' and name_winner == 'this':
1035
# No change for this directory in OTHER, skip
1037
content_changed = False
1038
elif other_ie.kind is None or other_ie.kind == 'file':
1039
def get_sha1(ie, tree):
1040
if ie.kind != 'file':
1042
return tree.get_file_sha1(file_id)
1043
base_sha1 = get_sha1(base_ie, self.base_tree)
1044
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1045
in zip(lca_entries, self._lca_trees)]
1046
this_sha1 = get_sha1(this_ie, self.this_tree)
1047
other_sha1 = get_sha1(other_ie, self.other_tree)
1048
sha1_winner = self._lca_multi_way(
1049
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1050
allow_overriding_lca=False)
1051
exec_winner = self._lca_multi_way(
1052
(base_ie.executable, lca_executable),
1053
other_ie.executable, this_ie.executable)
1054
if (parent_id_winner == 'this' and name_winner == 'this'
1055
and sha1_winner == 'this' and exec_winner == 'this'):
1056
# No kind, parent, name, exec, or content change for
1057
# OTHER, so this node is not considered interesting
1059
if sha1_winner == 'this':
1060
content_changed = False
1061
elif other_ie.kind == 'symlink':
1062
def get_target(ie, tree):
1063
if ie.kind != 'symlink':
1065
return tree.get_symlink_target(file_id)
1066
base_target = get_target(base_ie, self.base_tree)
1067
lca_targets = [get_target(ie, tree) for ie, tree
1068
in zip(lca_entries, self._lca_trees)]
1069
this_target = get_target(this_ie, self.this_tree)
1070
other_target = get_target(other_ie, self.other_tree)
1071
target_winner = self._lca_multi_way(
1072
(base_target, lca_targets),
1073
other_target, this_target)
1074
if (parent_id_winner == 'this' and name_winner == 'this'
1075
and target_winner == 'this'):
1076
# No kind, parent, name, or symlink target change
1079
if target_winner == 'this':
1080
content_changed = False
1081
elif other_ie.kind == 'tree-reference':
1082
# The 'changed' information seems to be handled at a higher
1083
# level. At least, _entries3 returns False for content
1084
# changed, even when at a new revision_id.
1085
content_changed = False
1086
if (parent_id_winner == 'this' and name_winner == 'this'):
1087
# Nothing interesting
1090
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1092
# If we have gotten this far, that means something has changed
1093
result.append((file_id, content_changed,
1094
((base_ie.parent_id, lca_parent_ids),
1095
other_ie.parent_id, this_ie.parent_id),
1096
((base_ie.name, lca_names),
1097
other_ie.name, this_ie.name),
1098
((base_ie.executable, lca_executable),
1099
other_ie.executable, this_ie.executable)
1104
if self.tt.final_kind(self.tt.root) is None:
1105
self.tt.cancel_deletion(self.tt.root)
1106
if self.tt.final_file_id(self.tt.root) is None:
1107
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
1109
other_root_file_id = self.other_tree.get_root_id()
1110
if other_root_file_id is None:
1112
other_root = self.tt.trans_id_file_id(other_root_file_id)
1113
if other_root == self.tt.root:
1115
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
1116
# the other tree's root is a non-root in the current tree (as when
1117
# a previously unrelated branch is merged into another)
1119
if self.tt.final_kind(other_root) is not None:
1120
other_root_is_present = True
1122
# other_root doesn't have a physical representation. We still need
1123
# to move any references to the actual root of the tree.
1124
other_root_is_present = False
1125
# 'other_tree.inventory.root' is not present in this tree. We are
1126
# calling adjust_path for children which *want* to be present with a
1127
# correct place to go.
1128
for thing, child in self.other_tree.inventory.root.children.iteritems():
1129
trans_id = self.tt.trans_id_file_id(child.file_id)
1130
if not other_root_is_present:
1131
if self.tt.final_kind(trans_id) is not None:
1132
# The item exist in the final tree and has a defined place
1135
# Move the item into the root
1136
self.tt.adjust_path(self.tt.final_name(trans_id),
1137
self.tt.root, trans_id)
1138
if other_root_is_present:
1139
self.tt.cancel_creation(other_root)
1140
self.tt.cancel_versioning(other_root)
1142
def write_modified(self, results):
1143
modified_hashes = {}
1144
for path in results.modified_paths:
1145
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
1148
hash = self.this_tree.get_file_sha1(file_id)
1151
modified_hashes[file_id] = hash
1152
self.this_tree.set_merge_modified(modified_hashes)
1155
def parent(entry, file_id):
1156
"""Determine the parent for a file_id (used as a key method)"""
1159
return entry.parent_id
1162
def name(entry, file_id):
1163
"""Determine the name for a file_id (used as a key method)"""
1169
def contents_sha1(tree, file_id):
1170
"""Determine the sha1 of the file contents (used as a key method)."""
1171
if file_id not in tree:
1173
return tree.get_file_sha1(file_id)
1176
def executable(tree, file_id):
1177
"""Determine the executability of a file-id (used as a key method)."""
1178
if not tree.has_id(file_id):
1180
if tree.kind(file_id) != "file":
1182
return tree.is_executable(file_id)
1185
def kind(tree, file_id):
1186
"""Determine the kind of a file-id (used as a key method)."""
1187
if not tree.has_id(file_id):
1189
return tree.kind(file_id)
1192
def _three_way(base, other, this):
1194
# if 'base == other', either they all agree, or only 'this' has
1197
elif this not in (base, other):
1198
# 'this' is neither 'base' nor 'other', so both sides changed
1201
# "Ambiguous clean merge" -- both sides have made the same change.
1204
# this == base: only other has changed.
1208
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
1209
"""Consider LCAs when determining whether a change has occurred.
1211
If LCAS are all identical, this is the same as a _three_way comparison.
1213
:param bases: value in (BASE, [LCAS])
1214
:param other: value in OTHER
1215
:param this: value in THIS
1216
:param allow_overriding_lca: If there is more than one unique lca
1217
value, allow OTHER to override THIS if it has a new value, and
1218
THIS only has an lca value, or vice versa. This is appropriate for
1219
truly scalar values, not as much for non-scalars.
1220
:return: 'this', 'other', or 'conflict' depending on whether an entry
1223
# See doc/developers/lca_tree_merging.txt for details about this
1226
# Either Ambiguously clean, or nothing was actually changed. We
1229
base_val, lca_vals = bases
1230
# Remove 'base_val' from the lca_vals, because it is not interesting
1231
filtered_lca_vals = [lca_val for lca_val in lca_vals
1232
if lca_val != base_val]
1233
if len(filtered_lca_vals) == 0:
1234
return Merge3Merger._three_way(base_val, other, this)
1236
unique_lca_vals = set(filtered_lca_vals)
1237
if len(unique_lca_vals) == 1:
1238
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1240
if allow_overriding_lca:
1241
if other in unique_lca_vals:
1242
if this in unique_lca_vals:
1243
# Each side picked a different lca, conflict
1246
# This has a value which supersedes both lca values, and
1247
# other only has an lca value
1249
elif this in unique_lca_vals:
1250
# OTHER has a value which supersedes both lca values, and this
1251
# only has an lca value
1254
# At this point, the lcas disagree, and the tip disagree
1258
@deprecated_method(deprecated_in((2, 2, 0)))
1259
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1260
"""Do a three-way test on a scalar.
1261
Return "this", "other" or "conflict", depending whether a value wins.
1263
key_base = key(base_tree, file_id)
1264
key_other = key(other_tree, file_id)
1265
#if base == other, either they all agree, or only THIS has changed.
1266
if key_base == key_other:
1268
key_this = key(this_tree, file_id)
1269
# "Ambiguous clean merge"
1270
if key_this == key_other:
1272
elif key_this == key_base:
1277
def merge_names(self, file_id):
1278
def get_entry(tree):
1279
if tree.has_id(file_id):
1280
return tree.inventory[file_id]
1283
this_entry = get_entry(self.this_tree)
1284
other_entry = get_entry(self.other_tree)
1285
base_entry = get_entry(self.base_tree)
1286
entries = (base_entry, other_entry, this_entry)
1289
for entry in entries:
1292
parents.append(None)
1294
names.append(entry.name)
1295
parents.append(entry.parent_id)
1296
return self._merge_names(file_id, parents, names,
1297
resolver=self._three_way)
1299
def _merge_names(self, file_id, parents, names, resolver):
1300
"""Perform a merge on file_id names and parents"""
1301
base_name, other_name, this_name = names
1302
base_parent, other_parent, this_parent = parents
1304
name_winner = resolver(*names)
1306
parent_id_winner = resolver(*parents)
1307
if this_name is None:
1308
if name_winner == "this":
1309
name_winner = "other"
1310
if parent_id_winner == "this":
1311
parent_id_winner = "other"
1312
if name_winner == "this" and parent_id_winner == "this":
1314
if name_winner == 'conflict' or parent_id_winner == 'conflict':
1315
# Creating helpers (.OTHER or .THIS) here cause problems down the
1316
# road if a ContentConflict needs to be created so we should not do
1318
trans_id = self.tt.trans_id_file_id(file_id)
1319
self._raw_conflicts.append(('path conflict', trans_id, file_id,
1320
this_parent, this_name,
1321
other_parent, other_name))
1322
if other_name is None:
1323
# it doesn't matter whether the result was 'other' or
1324
# 'conflict'-- if there's no 'other', we leave it alone.
1326
parent_id = parents[self.winner_idx[parent_id_winner]]
1327
if parent_id is not None:
1328
# if we get here, name_winner and parent_winner are set to safe
1330
self.tt.adjust_path(names[self.winner_idx[name_winner]],
1331
self.tt.trans_id_file_id(parent_id),
1332
self.tt.trans_id_file_id(file_id))
1334
def _do_merge_contents(self, file_id):
1335
"""Performs a merge on file_id contents."""
1336
def contents_pair(tree):
1337
if file_id not in tree:
1339
kind = tree.kind(file_id)
1341
contents = tree.get_file_sha1(file_id)
1342
elif kind == "symlink":
1343
contents = tree.get_symlink_target(file_id)
1346
return kind, contents
1348
# See SPOT run. run, SPOT, run.
1349
# So we're not QUITE repeating ourselves; we do tricky things with
1351
base_pair = contents_pair(self.base_tree)
1352
other_pair = contents_pair(self.other_tree)
1354
this_pair = contents_pair(self.this_tree)
1355
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1356
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1357
this_pair, allow_overriding_lca=False)
1359
if base_pair == other_pair:
1362
# We delayed evaluating this_pair as long as we can to avoid
1363
# unnecessary sha1 calculation
1364
this_pair = contents_pair(self.this_tree)
1365
winner = self._three_way(base_pair, other_pair, this_pair)
1366
if winner == 'this':
1367
# No interesting changes introduced by OTHER
1369
# We have a hypothetical conflict, but if we have files, then we
1370
# can try to merge the content
1371
trans_id = self.tt.trans_id_file_id(file_id)
1372
params = MergeHookParams(self, file_id, trans_id, this_pair[0],
1373
other_pair[0], winner)
1374
hooks = self.active_hooks
1375
hook_status = 'not_applicable'
1377
hook_status, lines = hook.merge_contents(params)
1378
if hook_status != 'not_applicable':
1379
# Don't try any more hooks, this one applies.
1382
if hook_status == 'not_applicable':
1383
# This is a contents conflict, because none of the available
1384
# functions could merge it.
1386
name = self.tt.final_name(trans_id)
1387
parent_id = self.tt.final_parent(trans_id)
1388
if self.this_tree.has_id(file_id):
1389
self.tt.unversion_file(trans_id)
1390
file_group = self._dump_conflicts(name, parent_id, file_id,
1392
self._raw_conflicts.append(('contents conflict', file_group))
1393
elif hook_status == 'success':
1394
self.tt.create_file(lines, trans_id)
1395
elif hook_status == 'conflicted':
1396
# XXX: perhaps the hook should be able to provide
1397
# the BASE/THIS/OTHER files?
1398
self.tt.create_file(lines, trans_id)
1399
self._raw_conflicts.append(('text conflict', trans_id))
1400
name = self.tt.final_name(trans_id)
1401
parent_id = self.tt.final_parent(trans_id)
1402
self._dump_conflicts(name, parent_id, file_id)
1403
elif hook_status == 'delete':
1404
self.tt.unversion_file(trans_id)
1406
elif hook_status == 'done':
1407
# The hook function did whatever it needs to do directly, no
1408
# further action needed here.
1411
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1412
if not self.this_tree.has_id(file_id) and result == "modified":
1413
self.tt.version_file(file_id, trans_id)
1414
# The merge has been performed, so the old contents should not be
1416
self.tt.delete_contents(trans_id)
1419
def _default_other_winner_merge(self, merge_hook_params):
1420
"""Replace this contents with other."""
1421
file_id = merge_hook_params.file_id
1422
trans_id = merge_hook_params.trans_id
1423
file_in_this = self.this_tree.has_id(file_id)
1424
if self.other_tree.has_id(file_id):
1425
# OTHER changed the file
1427
if wt.supports_content_filtering():
1428
# We get the path from the working tree if it exists.
1429
# That fails though when OTHER is adding a file, so
1430
# we fall back to the other tree to find the path if
1431
# it doesn't exist locally.
1433
filter_tree_path = wt.id2path(file_id)
1434
except errors.NoSuchId:
1435
filter_tree_path = self.other_tree.id2path(file_id)
1437
# Skip the id2path lookup for older formats
1438
filter_tree_path = None
1439
transform.create_from_tree(self.tt, trans_id,
1440
self.other_tree, file_id,
1441
filter_tree_path=filter_tree_path)
1444
# OTHER deleted the file
1445
return 'delete', None
1447
raise AssertionError(
1448
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1451
def merge_contents(self, merge_hook_params):
1452
"""Fallback merge logic after user installed hooks."""
1453
# This function is used in merge hooks as the fallback instance.
1454
# Perhaps making this function and the functions it calls be a
1455
# a separate class would be better.
1456
if merge_hook_params.winner == 'other':
1457
# OTHER is a straight winner, so replace this contents with other
1458
return self._default_other_winner_merge(merge_hook_params)
1459
elif merge_hook_params.is_file_merge():
1460
# THIS and OTHER are both files, so text merge. Either
1461
# BASE is a file, or both converted to files, so at least we
1462
# have agreement that output should be a file.
1464
self.text_merge(merge_hook_params.file_id,
1465
merge_hook_params.trans_id)
1466
except errors.BinaryFile:
1467
return 'not_applicable', None
1470
return 'not_applicable', None
1472
def get_lines(self, tree, file_id):
1473
"""Return the lines in a file, or an empty list."""
1474
if tree.has_id(file_id):
1475
return tree.get_file_lines(file_id)
1479
def text_merge(self, file_id, trans_id):
1480
"""Perform a three-way text merge on a file_id"""
1481
# it's possible that we got here with base as a different type.
1482
# if so, we just want two-way text conflicts.
1483
if self.base_tree.has_id(file_id) and \
1484
self.base_tree.kind(file_id) == "file":
1485
base_lines = self.get_lines(self.base_tree, file_id)
1488
other_lines = self.get_lines(self.other_tree, file_id)
1489
this_lines = self.get_lines(self.this_tree, file_id)
1490
m3 = merge3.Merge3(base_lines, this_lines, other_lines,
1491
is_cherrypick=self.cherrypick)
1492
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1493
if self.show_base is True:
1494
base_marker = '|' * 7
1498
def iter_merge3(retval):
1499
retval["text_conflicts"] = False
1500
for line in m3.merge_lines(name_a = "TREE",
1501
name_b = "MERGE-SOURCE",
1502
name_base = "BASE-REVISION",
1503
start_marker=start_marker,
1504
base_marker=base_marker,
1505
reprocess=self.reprocess):
1506
if line.startswith(start_marker):
1507
retval["text_conflicts"] = True
1508
yield line.replace(start_marker, '<' * 7)
1512
merge3_iterator = iter_merge3(retval)
1513
self.tt.create_file(merge3_iterator, trans_id)
1514
if retval["text_conflicts"] is True:
1515
self._raw_conflicts.append(('text conflict', trans_id))
1516
name = self.tt.final_name(trans_id)
1517
parent_id = self.tt.final_parent(trans_id)
1518
file_group = self._dump_conflicts(name, parent_id, file_id,
1519
this_lines, base_lines,
1521
file_group.append(trans_id)
1523
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1524
base_lines=None, other_lines=None, set_version=False,
1526
"""Emit conflict files.
1527
If this_lines, base_lines, or other_lines are omitted, they will be
1528
determined automatically. If set_version is true, the .OTHER, .THIS
1529
or .BASE (in that order) will be created as versioned files.
1531
data = [('OTHER', self.other_tree, other_lines),
1532
('THIS', self.this_tree, this_lines)]
1534
data.append(('BASE', self.base_tree, base_lines))
1536
# We need to use the actual path in the working tree of the file here,
1537
# ignoring the conflict suffixes
1539
if wt.supports_content_filtering():
1541
filter_tree_path = wt.id2path(file_id)
1542
except errors.NoSuchId:
1543
# file has been deleted
1544
filter_tree_path = None
1546
# Skip the id2path lookup for older formats
1547
filter_tree_path = None
1551
for suffix, tree, lines in data:
1552
if tree.has_id(file_id):
1553
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1554
suffix, lines, filter_tree_path)
1555
file_group.append(trans_id)
1556
if set_version and not versioned:
1557
self.tt.version_file(file_id, trans_id)
1561
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1562
lines=None, filter_tree_path=None):
1563
"""Emit a single conflict file."""
1564
name = name + '.' + suffix
1565
trans_id = self.tt.create_path(name, parent_id)
1566
transform.create_from_tree(self.tt, trans_id, tree, file_id, lines,
1570
def merge_executable(self, file_id, file_status):
1571
"""Perform a merge on the execute bit."""
1572
executable = [self.executable(t, file_id) for t in (self.base_tree,
1573
self.other_tree, self.this_tree)]
1574
self._merge_executable(file_id, executable, file_status,
1575
resolver=self._three_way)
1577
def _merge_executable(self, file_id, executable, file_status,
1579
"""Perform a merge on the execute bit."""
1580
base_executable, other_executable, this_executable = executable
1581
if file_status == "deleted":
1583
winner = resolver(*executable)
1584
if winner == "conflict":
1585
# There must be a None in here, if we have a conflict, but we
1586
# need executability since file status was not deleted.
1587
if self.executable(self.other_tree, file_id) is None:
1591
if winner == 'this' and file_status != "modified":
1593
trans_id = self.tt.trans_id_file_id(file_id)
1594
if self.tt.final_kind(trans_id) != "file":
1596
if winner == "this":
1597
executability = this_executable
1599
if self.other_tree.has_id(file_id):
1600
executability = other_executable
1601
elif self.this_tree.has_id(file_id):
1602
executability = this_executable
1603
elif self.base_tree_has_id(file_id):
1604
executability = base_executable
1605
if executability is not None:
1606
trans_id = self.tt.trans_id_file_id(file_id)
1607
self.tt.set_executability(executability, trans_id)
1609
def cook_conflicts(self, fs_conflicts):
1610
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1611
self.cooked_conflicts.extend(transform.cook_conflicts(
1612
fs_conflicts, self.tt))
1613
fp = transform.FinalPaths(self.tt)
1614
for conflict in self._raw_conflicts:
1615
conflict_type = conflict[0]
1616
if conflict_type == 'path conflict':
1618
this_parent, this_name,
1619
other_parent, other_name) = conflict[1:]
1620
if this_parent is None or this_name is None:
1621
this_path = '<deleted>'
1623
parent_path = fp.get_path(
1624
self.tt.trans_id_file_id(this_parent))
1625
this_path = osutils.pathjoin(parent_path, this_name)
1626
if other_parent is None or other_name is None:
1627
other_path = '<deleted>'
1629
parent_path = fp.get_path(
1630
self.tt.trans_id_file_id(other_parent))
1631
other_path = osutils.pathjoin(parent_path, other_name)
1632
c = _mod_conflicts.Conflict.factory(
1633
'path conflict', path=this_path,
1634
conflict_path=other_path,
1636
elif conflict_type == 'contents conflict':
1637
for trans_id in conflict[1]:
1638
file_id = self.tt.final_file_id(trans_id)
1639
if file_id is not None:
1641
path = fp.get_path(trans_id)
1642
for suffix in ('.BASE', '.THIS', '.OTHER'):
1643
if path.endswith(suffix):
1644
path = path[:-len(suffix)]
1646
c = _mod_conflicts.Conflict.factory(conflict_type,
1647
path=path, file_id=file_id)
1648
elif conflict_type == 'text conflict':
1649
trans_id = conflict[1]
1650
path = fp.get_path(trans_id)
1651
file_id = self.tt.final_file_id(trans_id)
1652
c = _mod_conflicts.Conflict.factory(conflict_type,
1653
path=path, file_id=file_id)
1655
raise AssertionError('bad conflict type: %r' % (conflict,))
1656
self.cooked_conflicts.append(c)
1657
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
1660
class WeaveMerger(Merge3Merger):
1661
"""Three-way tree merger, text weave merger."""
1662
supports_reprocess = True
1663
supports_show_base = False
1664
supports_reverse_cherrypick = False
1665
history_based = True
1667
def _generate_merge_plan(self, file_id, base):
1668
return self.this_tree.plan_file_merge(file_id, self.other_tree,
1671
def _merged_lines(self, file_id):
1672
"""Generate the merged lines.
1673
There is no distinction between lines that are meant to contain <<<<<<<
1677
base = self.base_tree
1680
plan = self._generate_merge_plan(file_id, base)
1681
if 'merge' in debug.debug_flags:
1683
trans_id = self.tt.trans_id_file_id(file_id)
1684
name = self.tt.final_name(trans_id) + '.plan'
1685
contents = ('%11s|%s' % l for l in plan)
1686
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1687
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1688
'>>>>>>> MERGE-SOURCE\n')
1689
lines, conflicts = textmerge.merge_lines(self.reprocess)
1691
base_lines = textmerge.base_from_plan()
1694
return lines, base_lines
1696
def text_merge(self, file_id, trans_id):
1697
"""Perform a (weave) text merge for a given file and file-id.
1698
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1699
and a conflict will be noted.
1701
lines, base_lines = self._merged_lines(file_id)
1703
# Note we're checking whether the OUTPUT is binary in this case,
1704
# because we don't want to get into weave merge guts.
1705
textfile.check_text_lines(lines)
1706
self.tt.create_file(lines, trans_id)
1707
if base_lines is not None:
1709
self._raw_conflicts.append(('text conflict', trans_id))
1710
name = self.tt.final_name(trans_id)
1711
parent_id = self.tt.final_parent(trans_id)
1712
file_group = self._dump_conflicts(name, parent_id, file_id,
1714
base_lines=base_lines)
1715
file_group.append(trans_id)
1718
class LCAMerger(WeaveMerger):
1720
def _generate_merge_plan(self, file_id, base):
1721
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1724
class Diff3Merger(Merge3Merger):
1725
"""Three-way merger using external diff3 for text merging"""
1727
def dump_file(self, temp_dir, name, tree, file_id):
1728
out_path = osutils.pathjoin(temp_dir, name)
1729
out_file = open(out_path, "wb")
1731
in_file = tree.get_file(file_id)
1732
for line in in_file:
1733
out_file.write(line)
1738
def text_merge(self, file_id, trans_id):
1739
"""Perform a diff3 merge using a specified file-id and trans-id.
1740
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1741
will be dumped, and a will be conflict noted.
1744
temp_dir = osutils.mkdtemp(prefix="bzr-")
1746
new_file = osutils.pathjoin(temp_dir, "new")
1747
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1748
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1749
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1750
status = bzrlib.patch.diff3(new_file, this, base, other)
1751
if status not in (0, 1):
1752
raise errors.BzrError("Unhandled diff3 exit code")
1753
f = open(new_file, 'rb')
1755
self.tt.create_file(f, trans_id)
1759
name = self.tt.final_name(trans_id)
1760
parent_id = self.tt.final_parent(trans_id)
1761
self._dump_conflicts(name, parent_id, file_id)
1762
self._raw_conflicts.append(('text conflict', trans_id))
1764
osutils.rmtree(temp_dir)
1767
class PathNotInTree(errors.BzrError):
1769
_fmt = """Merge-into failed because %(tree)s does not contain %(path)s."""
1771
def __init__(self, path, tree):
1772
errors.BzrError.__init__(self, path=path, tree=tree)
1775
class MergeIntoMerger(Merger):
1776
"""Merger that understands other_tree will be merged into a subdir.
1778
This also changes the Merger api so that it uses real Branch, revision_id,
1779
and RevisonTree objects, rather than using revision specs.
1782
def __init__(self, this_tree, other_branch, other_tree, target_subdir,
1783
source_subpath, other_rev_id=None):
1784
"""Create a new MergeIntoMerger object.
1786
source_subpath in other_tree will be effectively copied to
1787
target_subdir in this_tree.
1789
:param this_tree: The tree that we will be merging into.
1790
:param other_branch: The Branch we will be merging from.
1791
:param other_tree: The RevisionTree object we want to merge.
1792
:param target_subdir: The relative path where we want to merge
1793
other_tree into this_tree
1794
:param source_subpath: The relative path specifying the subtree of
1795
other_tree to merge into this_tree.
1797
# It is assumed that we are merging a tree that is not in our current
1798
# ancestry, which means we are using the "EmptyTree" as our basis.
1799
null_ancestor_tree = this_tree.branch.repository.revision_tree(
1800
_mod_revision.NULL_REVISION)
1801
super(MergeIntoMerger, self).__init__(
1802
this_branch=this_tree.branch,
1803
this_tree=this_tree,
1804
other_tree=other_tree,
1805
base_tree=null_ancestor_tree,
1807
self._target_subdir = target_subdir
1808
self._source_subpath = source_subpath
1809
self.other_branch = other_branch
1810
if other_rev_id is None:
1811
other_rev_id = other_tree.get_revision_id()
1812
self.other_rev_id = self.other_basis = other_rev_id
1813
self.base_is_ancestor = True
1814
self.backup_files = True
1815
self.merge_type = Merge3Merger
1816
self.show_base = False
1817
self.reprocess = False
1818
self.interesting_ids = None
1819
self.merge_type = _MergeTypeParameterizer(MergeIntoMergeType,
1820
target_subdir=self._target_subdir,
1821
source_subpath=self._source_subpath)
1822
if self._source_subpath != '':
1823
# If this isn't a partial merge make sure the revisions will be
1825
self._maybe_fetch(self.other_branch, self.this_branch,
1828
def set_pending(self):
1829
if self._source_subpath != '':
1831
Merger.set_pending(self)
1834
class _MergeTypeParameterizer(object):
1835
"""Wrap a merge-type class to provide extra parameters.
1837
This is hack used by MergeIntoMerger to pass some extra parameters to its
1838
merge_type. Merger.do_merge() sets up its own set of parameters to pass to
1839
the 'merge_type' member. It is difficult override do_merge without
1840
re-writing the whole thing, so instead we create a wrapper which will pass
1841
the extra parameters.
1844
def __init__(self, merge_type, **kwargs):
1845
self._extra_kwargs = kwargs
1846
self._merge_type = merge_type
1848
def __call__(self, *args, **kwargs):
1849
kwargs.update(self._extra_kwargs)
1850
return self._merge_type(*args, **kwargs)
1852
def __getattr__(self, name):
1853
return getattr(self._merge_type, name)
1856
class MergeIntoMergeType(Merge3Merger):
1857
"""Merger that incorporates a tree (or part of a tree) into another."""
1859
def __init__(self, *args, **kwargs):
1860
"""Initialize the merger object.
1862
:param args: See Merge3Merger.__init__'s args.
1863
:param kwargs: See Merge3Merger.__init__'s keyword args, except for
1864
source_subpath and target_subdir.
1865
:keyword source_subpath: The relative path specifying the subtree of
1866
other_tree to merge into this_tree.
1867
:keyword target_subdir: The relative path where we want to merge
1868
other_tree into this_tree
1870
# All of the interesting work happens during Merge3Merger.__init__(),
1871
# so we have have to hack in to get our extra parameters set.
1872
self._source_subpath = kwargs.pop('source_subpath')
1873
self._target_subdir = kwargs.pop('target_subdir')
1874
super(MergeIntoMergeType, self).__init__(*args, **kwargs)
1876
def _compute_transform(self):
1877
child_pb = ui.ui_factory.nested_progress_bar()
1879
entries = self._entries_to_incorporate()
1880
entries = list(entries)
1881
for num, (entry, parent_id) in enumerate(entries):
1882
child_pb.update('Preparing file merge', num, len(entries))
1883
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1884
trans_id = transform.new_by_entry(self.tt, entry,
1885
parent_trans_id, self.other_tree)
1888
self._finish_computing_transform()
1890
def _entries_to_incorporate(self):
1891
"""Yields pairs of (inventory_entry, new_parent)."""
1892
other_inv = self.other_tree.inventory
1893
subdir_id = other_inv.path2id(self._source_subpath)
1894
if subdir_id is None:
1895
# XXX: The error would be clearer if it gave the URL of the source
1896
# branch, but we don't have a reference to that here.
1897
raise PathNotInTree(self._source_subpath, "Source tree")
1898
subdir = other_inv[subdir_id]
1899
parent_in_target = osutils.dirname(self._target_subdir)
1900
target_id = self.this_tree.inventory.path2id(parent_in_target)
1901
if target_id is None:
1902
raise PathNotInTree(self._target_subdir, "Target tree")
1903
name_in_target = osutils.basename(self._target_subdir)
1904
merge_into_root = subdir.copy()
1905
merge_into_root.name = name_in_target
1906
if merge_into_root.file_id in self.this_tree.inventory:
1907
# Give the root a new file-id.
1908
# This can happen fairly easily if the directory we are
1909
# incorporating is the root, and both trees have 'TREE_ROOT' as
1910
# their root_id. Users will expect this to Just Work, so we
1911
# change the file-id here.
1912
# Non-root file-ids could potentially conflict too. That's really
1913
# an edge case, so we don't do anything special for those. We let
1914
# them cause conflicts.
1915
merge_into_root.file_id = generate_ids.gen_file_id(name_in_target)
1916
yield (merge_into_root, target_id)
1917
if subdir.kind != 'directory':
1918
# No children, so we are done.
1920
for ignored_path, entry in other_inv.iter_entries_by_dir(subdir_id):
1921
parent_id = entry.parent_id
1922
if parent_id == subdir.file_id:
1923
# The root's parent ID has changed, so make sure children of
1924
# the root refer to the new ID.
1925
parent_id = merge_into_root.file_id
1926
yield (entry, parent_id)
1929
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1931
merge_type=Merge3Merger,
1932
interesting_ids=None,
1936
interesting_files=None,
1939
change_reporter=None):
1940
"""Primary interface for merging.
1942
Typical use is probably::
1944
merge_inner(branch, branch.get_revision_tree(other_revision),
1945
branch.get_revision_tree(base_revision))
1947
if this_tree is None:
1948
raise errors.BzrError("bzrlib.merge.merge_inner requires a this_tree "
1949
"parameter as of bzrlib version 0.8.")
1950
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1951
pb=pb, change_reporter=change_reporter)
1952
merger.backup_files = backup_files
1953
merger.merge_type = merge_type
1954
merger.interesting_ids = interesting_ids
1955
merger.ignore_zero = ignore_zero
1956
if interesting_files:
1958
raise ValueError('Only supply interesting_ids'
1959
' or interesting_files')
1960
merger.interesting_files = interesting_files
1961
merger.show_base = show_base
1962
merger.reprocess = reprocess
1963
merger.other_rev_id = other_rev_id
1964
merger.other_basis = other_rev_id
1965
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1966
if get_revision_id is None:
1967
get_revision_id = base_tree.last_revision
1968
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
1969
merger.set_base_revision(get_revision_id(), this_branch)
1970
return merger.do_merge()
1972
def get_merge_type_registry():
1973
"""Merge type registry is in bzrlib.option to avoid circular imports.
1975
This method provides a sanctioned way to retrieve it.
1977
from bzrlib import option
1978
return option._merge_type_registry
1981
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1982
def status_a(revision, text):
1983
if revision in ancestors_b:
1984
return 'killed-b', text
1986
return 'new-a', text
1988
def status_b(revision, text):
1989
if revision in ancestors_a:
1990
return 'killed-a', text
1992
return 'new-b', text
1994
plain_a = [t for (a, t) in annotated_a]
1995
plain_b = [t for (a, t) in annotated_b]
1996
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1997
blocks = matcher.get_matching_blocks()
2000
for ai, bi, l in blocks:
2001
# process all mismatched sections
2002
# (last mismatched section is handled because blocks always
2003
# includes a 0-length last block)
2004
for revision, text in annotated_a[a_cur:ai]:
2005
yield status_a(revision, text)
2006
for revision, text in annotated_b[b_cur:bi]:
2007
yield status_b(revision, text)
2008
# and now the matched section
2011
for text_a in plain_a[ai:a_cur]:
2012
yield "unchanged", text_a
2015
class _PlanMergeBase(object):
2017
def __init__(self, a_rev, b_rev, vf, key_prefix):
2020
:param a_rev: Revision-id of one revision to merge
2021
:param b_rev: Revision-id of the other revision to merge
2022
:param vf: A VersionedFiles containing both revisions
2023
:param key_prefix: A prefix for accessing keys in vf, typically
2029
self._last_lines = None
2030
self._last_lines_revision_id = None
2031
self._cached_matching_blocks = {}
2032
self._key_prefix = key_prefix
2033
self._precache_tip_lines()
2035
def _precache_tip_lines(self):
2036
lines = self.get_lines([self.a_rev, self.b_rev])
2037
self.lines_a = lines[self.a_rev]
2038
self.lines_b = lines[self.b_rev]
2040
def get_lines(self, revisions):
2041
"""Get lines for revisions from the backing VersionedFiles.
2043
:raises RevisionNotPresent: on absent texts.
2045
keys = [(self._key_prefix + (rev,)) for rev in revisions]
2047
for record in self.vf.get_record_stream(keys, 'unordered', True):
2048
if record.storage_kind == 'absent':
2049
raise errors.RevisionNotPresent(record.key, self.vf)
2050
result[record.key[-1]] = osutils.chunks_to_lines(
2051
record.get_bytes_as('chunked'))
2054
def plan_merge(self):
2055
"""Generate a 'plan' for merging the two revisions.
2057
This involves comparing their texts and determining the cause of
2058
differences. If text A has a line and text B does not, then either the
2059
line was added to text A, or it was deleted from B. Once the causes
2060
are combined, they are written out in the format described in
2061
VersionedFile.plan_merge
2063
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
2064
unique_a, unique_b = self._unique_lines(blocks)
2065
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
2066
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
2067
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
2069
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
2072
for i, j, n in blocks:
2073
for a_index in range(last_i, i):
2074
if a_index in new_a:
2075
if a_index in killed_b:
2076
yield 'conflicted-a', self.lines_a[a_index]
2078
yield 'new-a', self.lines_a[a_index]
2080
yield 'killed-b', self.lines_a[a_index]
2081
for b_index in range(last_j, j):
2082
if b_index in new_b:
2083
if b_index in killed_a:
2084
yield 'conflicted-b', self.lines_b[b_index]
2086
yield 'new-b', self.lines_b[b_index]
2088
yield 'killed-a', self.lines_b[b_index]
2089
# handle common lines
2090
for a_index in range(i, i+n):
2091
yield 'unchanged', self.lines_a[a_index]
2095
def _get_matching_blocks(self, left_revision, right_revision):
2096
"""Return a description of which sections of two revisions match.
2098
See SequenceMatcher.get_matching_blocks
2100
cached = self._cached_matching_blocks.get((left_revision,
2102
if cached is not None:
2104
if self._last_lines_revision_id == left_revision:
2105
left_lines = self._last_lines
2106
right_lines = self.get_lines([right_revision])[right_revision]
2108
lines = self.get_lines([left_revision, right_revision])
2109
left_lines = lines[left_revision]
2110
right_lines = lines[right_revision]
2111
self._last_lines = right_lines
2112
self._last_lines_revision_id = right_revision
2113
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
2115
return matcher.get_matching_blocks()
2117
def _unique_lines(self, matching_blocks):
2118
"""Analyse matching_blocks to determine which lines are unique
2120
:return: a tuple of (unique_left, unique_right), where the values are
2121
sets of line numbers of unique lines.
2127
for i, j, n in matching_blocks:
2128
unique_left.extend(range(last_i, i))
2129
unique_right.extend(range(last_j, j))
2132
return unique_left, unique_right
2135
def _subtract_plans(old_plan, new_plan):
2136
"""Remove changes from new_plan that came from old_plan.
2138
It is assumed that the difference between the old_plan and new_plan
2139
is their choice of 'b' text.
2141
All lines from new_plan that differ from old_plan are emitted
2142
verbatim. All lines from new_plan that match old_plan but are
2143
not about the 'b' revision are emitted verbatim.
2145
Lines that match and are about the 'b' revision are the lines we
2146
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
2147
is skipped entirely.
2149
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
2152
for i, j, n in matcher.get_matching_blocks():
2153
for jj in range(last_j, j):
2155
for jj in range(j, j+n):
2156
plan_line = new_plan[jj]
2157
if plan_line[0] == 'new-b':
2159
elif plan_line[0] == 'killed-b':
2160
yield 'unchanged', plan_line[1]
2166
class _PlanMerge(_PlanMergeBase):
2167
"""Plan an annotate merge using on-the-fly annotation"""
2169
def __init__(self, a_rev, b_rev, vf, key_prefix):
2170
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
2171
self.a_key = self._key_prefix + (self.a_rev,)
2172
self.b_key = self._key_prefix + (self.b_rev,)
2173
self.graph = _mod_graph.Graph(self.vf)
2174
heads = self.graph.heads((self.a_key, self.b_key))
2176
# one side dominates, so we can just return its values, yay for
2178
# Ideally we would know that before we get this far
2179
self._head_key = heads.pop()
2180
if self._head_key == self.a_key:
2184
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2185
self._head_key[-1], other)
2188
self._head_key = None
2191
def _precache_tip_lines(self):
2192
# Turn this into a no-op, because we will do this later
2195
def _find_recursive_lcas(self):
2196
"""Find all the ancestors back to a unique lca"""
2197
cur_ancestors = (self.a_key, self.b_key)
2198
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2199
# rather than a key tuple. We will just map that directly to no common
2203
next_lcas = self.graph.find_lca(*cur_ancestors)
2204
# Map a plain NULL_REVISION to a simple no-ancestors
2205
if next_lcas == set([_mod_revision.NULL_REVISION]):
2207
# Order the lca's based on when they were merged into the tip
2208
# While the actual merge portion of weave merge uses a set() of
2209
# active revisions, the order of insertion *does* effect the
2210
# implicit ordering of the texts.
2211
for rev_key in cur_ancestors:
2212
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2214
parent_map[rev_key] = ordered_parents
2215
if len(next_lcas) == 0:
2217
elif len(next_lcas) == 1:
2218
parent_map[list(next_lcas)[0]] = ()
2220
elif len(next_lcas) > 2:
2221
# More than 2 lca's, fall back to grabbing all nodes between
2222
# this and the unique lca.
2223
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2225
self.a_key, self.b_key, cur_ancestors)
2226
cur_lcas = next_lcas
2227
while len(cur_lcas) > 1:
2228
cur_lcas = self.graph.find_lca(*cur_lcas)
2229
if len(cur_lcas) == 0:
2230
# No common base to find, use the full ancestry
2233
unique_lca = list(cur_lcas)[0]
2234
if unique_lca == _mod_revision.NULL_REVISION:
2235
# find_lca will return a plain 'NULL_REVISION' rather
2236
# than a key tuple when there is no common ancestor, we
2237
# prefer to just use None, because it doesn't confuse
2238
# _get_interesting_texts()
2240
parent_map.update(self._find_unique_parents(next_lcas,
2243
cur_ancestors = next_lcas
2246
def _find_unique_parents(self, tip_keys, base_key):
2247
"""Find ancestors of tip that aren't ancestors of base.
2249
:param tip_keys: Nodes that are interesting
2250
:param base_key: Cull all ancestors of this node
2251
:return: The parent map for all revisions between tip_keys and
2252
base_key. base_key will be included. References to nodes outside of
2253
the ancestor set will also be removed.
2255
# TODO: this would be simpler if find_unique_ancestors took a list
2256
# instead of a single tip, internally it supports it, but it
2257
# isn't a "backwards compatible" api change.
2258
if base_key is None:
2259
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2260
# We remove NULL_REVISION because it isn't a proper tuple key, and
2261
# thus confuses things like _get_interesting_texts, and our logic
2262
# to add the texts into the memory weave.
2263
if _mod_revision.NULL_REVISION in parent_map:
2264
parent_map.pop(_mod_revision.NULL_REVISION)
2267
for tip in tip_keys:
2269
self.graph.find_unique_ancestors(tip, [base_key]))
2270
parent_map = self.graph.get_parent_map(interesting)
2271
parent_map[base_key] = ()
2272
culled_parent_map, child_map, tails = self._remove_external_references(
2274
# Remove all the tails but base_key
2275
if base_key is not None:
2276
tails.remove(base_key)
2277
self._prune_tails(culled_parent_map, child_map, tails)
2278
# Now remove all the uninteresting 'linear' regions
2279
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2283
def _remove_external_references(parent_map):
2284
"""Remove references that go outside of the parent map.
2286
:param parent_map: Something returned from Graph.get_parent_map(keys)
2287
:return: (filtered_parent_map, child_map, tails)
2288
filtered_parent_map is parent_map without external references
2289
child_map is the {parent_key: [child_keys]} mapping
2290
tails is a list of nodes that do not have any parents in the map
2292
# TODO: The basic effect of this function seems more generic than
2293
# _PlanMerge. But the specific details of building a child_map,
2294
# and computing tails seems very specific to _PlanMerge.
2295
# Still, should this be in Graph land?
2296
filtered_parent_map = {}
2299
for key, parent_keys in parent_map.iteritems():
2300
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2301
if not culled_parent_keys:
2303
for parent_key in culled_parent_keys:
2304
child_map.setdefault(parent_key, []).append(key)
2305
# TODO: Do we want to do this, it adds overhead for every node,
2306
# just to say that the node has no children
2307
child_map.setdefault(key, [])
2308
filtered_parent_map[key] = culled_parent_keys
2309
return filtered_parent_map, child_map, tails
2312
def _prune_tails(parent_map, child_map, tails_to_remove):
2313
"""Remove tails from the parent map.
2315
This will remove the supplied revisions until no more children have 0
2318
:param parent_map: A dict of {child: [parents]}, this dictionary will
2319
be modified in place.
2320
:param tails_to_remove: A list of tips that should be removed,
2321
this list will be consumed
2322
:param child_map: The reverse dict of parent_map ({parent: [children]})
2323
this dict will be modified
2324
:return: None, parent_map will be modified in place.
2326
while tails_to_remove:
2327
next = tails_to_remove.pop()
2328
parent_map.pop(next)
2329
children = child_map.pop(next)
2330
for child in children:
2331
child_parents = parent_map[child]
2332
child_parents.remove(next)
2333
if len(child_parents) == 0:
2334
tails_to_remove.append(child)
2336
def _get_interesting_texts(self, parent_map):
2337
"""Return a dict of texts we are interested in.
2339
Note that the input is in key tuples, but the output is in plain
2342
:param parent_map: The output from _find_recursive_lcas
2343
:return: A dict of {'revision_id':lines} as returned by
2344
_PlanMergeBase.get_lines()
2346
all_revision_keys = set(parent_map)
2347
all_revision_keys.add(self.a_key)
2348
all_revision_keys.add(self.b_key)
2350
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2351
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2354
def _build_weave(self):
2355
from bzrlib import weave
2356
self._weave = weave.Weave(weave_name='in_memory_weave',
2357
allow_reserved=True)
2358
parent_map = self._find_recursive_lcas()
2360
all_texts = self._get_interesting_texts(parent_map)
2362
# Note: Unfortunately, the order given by topo_sort will effect the
2363
# ordering resolution in the output. Specifically, if you add A then B,
2364
# then in the output text A lines will show up before B lines. And, of
2365
# course, topo_sort doesn't guarantee any real ordering.
2366
# So we use merge_sort, and add a fake node on the tip.
2367
# This ensures that left-hand parents will always be inserted into the
2368
# weave before right-hand parents.
2369
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2370
parent_map[tip_key] = (self.a_key, self.b_key)
2372
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2376
# for key in tsort.topo_sort(parent_map):
2377
parent_keys = parent_map[key]
2378
revision_id = key[-1]
2379
parent_ids = [k[-1] for k in parent_keys]
2380
self._weave.add_lines(revision_id, parent_ids,
2381
all_texts[revision_id])
2383
def plan_merge(self):
2384
"""Generate a 'plan' for merging the two revisions.
2386
This involves comparing their texts and determining the cause of
2387
differences. If text A has a line and text B does not, then either the
2388
line was added to text A, or it was deleted from B. Once the causes
2389
are combined, they are written out in the format described in
2390
VersionedFile.plan_merge
2392
if self._head_key is not None: # There was a single head
2393
if self._head_key == self.a_key:
2396
if self._head_key != self.b_key:
2397
raise AssertionError('There was an invalid head: %s != %s'
2398
% (self.b_key, self._head_key))
2400
head_rev = self._head_key[-1]
2401
lines = self.get_lines([head_rev])[head_rev]
2402
return ((plan, line) for line in lines)
2403
return self._weave.plan_merge(self.a_rev, self.b_rev)
2406
class _PlanLCAMerge(_PlanMergeBase):
2408
This merge algorithm differs from _PlanMerge in that:
2409
1. comparisons are done against LCAs only
2410
2. cases where a contested line is new versus one LCA but old versus
2411
another are marked as conflicts, by emitting the line as conflicted-a
2414
This is faster, and hopefully produces more useful output.
2417
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2418
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2419
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2422
if lca == _mod_revision.NULL_REVISION:
2425
self.lcas.add(lca[-1])
2426
for lca in self.lcas:
2427
if _mod_revision.is_null(lca):
2430
lca_lines = self.get_lines([lca])[lca]
2431
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2433
blocks = list(matcher.get_matching_blocks())
2434
self._cached_matching_blocks[(a_rev, lca)] = blocks
2435
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2437
blocks = list(matcher.get_matching_blocks())
2438
self._cached_matching_blocks[(b_rev, lca)] = blocks
2440
def _determine_status(self, revision_id, unique_line_numbers):
2441
"""Determines the status unique lines versus all lcas.
2443
Basically, determines why the line is unique to this revision.
2445
A line may be determined new, killed, or both.
2447
If a line is determined new, that means it was not present in at least
2448
one LCA, and is not present in the other merge revision.
2450
If a line is determined killed, that means the line was present in
2453
If a line is killed and new, this indicates that the two merge
2454
revisions contain differing conflict resolutions.
2455
:param revision_id: The id of the revision in which the lines are
2457
:param unique_line_numbers: The line numbers of unique lines.
2458
:return a tuple of (new_this, killed_other):
2462
unique_line_numbers = set(unique_line_numbers)
2463
for lca in self.lcas:
2464
blocks = self._get_matching_blocks(revision_id, lca)
2465
unique_vs_lca, _ignored = self._unique_lines(blocks)
2466
new.update(unique_line_numbers.intersection(unique_vs_lca))
2467
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
