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):
568
target.fetch(source, revision_id)
571
revisions = [_mod_revision.ensure_null(self.this_basis),
572
_mod_revision.ensure_null(self.other_basis)]
573
if _mod_revision.NULL_REVISION in revisions:
574
self.base_rev_id = _mod_revision.NULL_REVISION
575
self.base_tree = self.revision_tree(self.base_rev_id)
576
self._is_criss_cross = False
578
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
579
self._is_criss_cross = False
581
self.base_rev_id = _mod_revision.NULL_REVISION
583
self.base_rev_id = list(lcas)[0]
584
else: # len(lcas) > 1
586
# find_unique_lca can only handle 2 nodes, so we have to
587
# start back at the beginning. It is a shame to traverse
588
# the graph again, but better than re-implementing
590
self.base_rev_id = self.revision_graph.find_unique_lca(
591
revisions[0], revisions[1])
593
self.base_rev_id = self.revision_graph.find_unique_lca(
595
self._is_criss_cross = True
596
if self.base_rev_id == _mod_revision.NULL_REVISION:
597
raise errors.UnrelatedBranches()
598
if self._is_criss_cross:
599
trace.warning('Warning: criss-cross merge encountered. See bzr'
600
' help criss-cross.')
601
trace.mutter('Criss-cross lcas: %r' % lcas)
602
interesting_revision_ids = [self.base_rev_id]
603
interesting_revision_ids.extend(lcas)
604
interesting_trees = dict((t.get_revision_id(), t)
605
for t in self.this_branch.repository.revision_trees(
606
interesting_revision_ids))
607
self._cached_trees.update(interesting_trees)
608
self.base_tree = interesting_trees.pop(self.base_rev_id)
609
sorted_lca_keys = self.revision_graph.find_merge_order(
611
self._lca_trees = [interesting_trees[key]
612
for key in sorted_lca_keys]
614
self.base_tree = self.revision_tree(self.base_rev_id)
615
self.base_is_ancestor = True
616
self.base_is_other_ancestor = True
617
trace.mutter('Base revid: %r' % self.base_rev_id)
619
def set_base(self, base_revision):
620
"""Set the base revision to use for the merge.
622
:param base_revision: A 2-list containing a path and revision number.
624
trace.mutter("doing merge() with no base_revision specified")
625
if base_revision == [None, None]:
628
base_branch, self.base_tree = self._get_tree(base_revision)
629
if base_revision[1] == -1:
630
self.base_rev_id = base_branch.last_revision()
631
elif base_revision[1] is None:
632
self.base_rev_id = _mod_revision.NULL_REVISION
634
self.base_rev_id = _mod_revision.ensure_null(
635
base_branch.get_rev_id(base_revision[1]))
636
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
638
def make_merger(self):
639
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
640
'other_tree': self.other_tree,
641
'interesting_ids': self.interesting_ids,
642
'interesting_files': self.interesting_files,
643
'this_branch': self.this_branch,
645
if self.merge_type.requires_base:
646
kwargs['base_tree'] = self.base_tree
647
if self.merge_type.supports_reprocess:
648
kwargs['reprocess'] = self.reprocess
650
raise errors.BzrError(
651
"Conflict reduction is not supported for merge"
652
" type %s." % self.merge_type)
653
if self.merge_type.supports_show_base:
654
kwargs['show_base'] = self.show_base
656
raise errors.BzrError("Showing base is not supported for this"
657
" merge type. %s" % self.merge_type)
658
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
659
and not self.base_is_other_ancestor):
660
raise errors.CannotReverseCherrypick()
661
if self.merge_type.supports_cherrypick:
662
kwargs['cherrypick'] = (not self.base_is_ancestor or
663
not self.base_is_other_ancestor)
664
if self._is_criss_cross and getattr(self.merge_type,
665
'supports_lca_trees', False):
666
kwargs['lca_trees'] = self._lca_trees
667
return self.merge_type(pb=None,
668
change_reporter=self.change_reporter,
671
def _do_merge_to(self):
672
merge = self.make_merger()
673
if self.other_branch is not None:
674
self.other_branch.update_references(self.this_branch)
676
if self.recurse == 'down':
677
for relpath, file_id in self.this_tree.iter_references():
678
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
679
other_revision = self.other_tree.get_reference_revision(
681
if other_revision == sub_tree.last_revision():
683
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
684
sub_merge.merge_type = self.merge_type
685
other_branch = self.other_branch.reference_parent(file_id, relpath)
686
sub_merge.set_other_revision(other_revision, other_branch)
687
base_revision = self.base_tree.get_reference_revision(file_id)
688
sub_merge.base_tree = \
689
sub_tree.branch.repository.revision_tree(base_revision)
690
sub_merge.base_rev_id = base_revision
695
operation = OperationWithCleanups(self._do_merge_to)
696
self.this_tree.lock_tree_write()
697
operation.add_cleanup(self.this_tree.unlock)
698
if self.base_tree is not None:
699
self.base_tree.lock_read()
700
operation.add_cleanup(self.base_tree.unlock)
701
if self.other_tree is not None:
702
self.other_tree.lock_read()
703
operation.add_cleanup(self.other_tree.unlock)
704
merge = operation.run_simple()
705
if len(merge.cooked_conflicts) == 0:
706
if not self.ignore_zero and not trace.is_quiet():
707
trace.note("All changes applied successfully.")
709
trace.note("%d conflicts encountered."
710
% len(merge.cooked_conflicts))
712
return len(merge.cooked_conflicts)
715
class _InventoryNoneEntry(object):
716
"""This represents an inventory entry which *isn't there*.
718
It simplifies the merging logic if we always have an InventoryEntry, even
719
if it isn't actually present
726
symlink_target = None
729
_none_entry = _InventoryNoneEntry()
732
class Merge3Merger(object):
733
"""Three-way merger that uses the merge3 text merger"""
735
supports_reprocess = True
736
supports_show_base = True
737
history_based = False
738
supports_cherrypick = True
739
supports_reverse_cherrypick = True
740
winner_idx = {"this": 2, "other": 1, "conflict": 1}
741
supports_lca_trees = True
743
def __init__(self, working_tree, this_tree, base_tree, other_tree,
744
interesting_ids=None, reprocess=False, show_base=False,
745
pb=None, pp=None, change_reporter=None,
746
interesting_files=None, do_merge=True,
747
cherrypick=False, lca_trees=None, this_branch=None):
748
"""Initialize the merger object and perform the merge.
750
:param working_tree: The working tree to apply the merge to
751
:param this_tree: The local tree in the merge operation
752
:param base_tree: The common tree in the merge operation
753
:param other_tree: The other tree to merge changes from
754
:param this_branch: The branch associated with this_tree. Defaults to
755
this_tree.branch if not supplied.
756
:param interesting_ids: The file_ids of files that should be
757
participate in the merge. May not be combined with
759
:param: reprocess If True, perform conflict-reduction processing.
760
:param show_base: If True, show the base revision in text conflicts.
761
(incompatible with reprocess)
763
:param pp: A ProgressPhase object
764
:param change_reporter: An object that should report changes made
765
:param interesting_files: The tree-relative paths of files that should
766
participate in the merge. If these paths refer to directories,
767
the contents of those directories will also be included. May not
768
be combined with interesting_ids. If neither interesting_files nor
769
interesting_ids is specified, all files may participate in the
771
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
772
if the ancestry was found to include a criss-cross merge.
773
Otherwise should be None.
775
object.__init__(self)
776
if interesting_files is not None and interesting_ids is not None:
778
'specify either interesting_ids or interesting_files')
779
if this_branch is None:
780
this_branch = this_tree.branch
781
self.interesting_ids = interesting_ids
782
self.interesting_files = interesting_files
783
self.this_tree = working_tree
784
self.base_tree = base_tree
785
self.other_tree = other_tree
786
self.this_branch = this_branch
787
self._raw_conflicts = []
788
self.cooked_conflicts = []
789
self.reprocess = reprocess
790
self.show_base = show_base
791
self._lca_trees = lca_trees
792
# Uncommenting this will change the default algorithm to always use
793
# _entries_lca. This can be useful for running the test suite and
794
# making sure we haven't missed any corner cases.
795
# if lca_trees is None:
796
# self._lca_trees = [self.base_tree]
797
self.change_reporter = change_reporter
798
self.cherrypick = cherrypick
802
warnings.warn("pp argument to Merge3Merger is deprecated")
804
warnings.warn("pb argument to Merge3Merger is deprecated")
807
operation = OperationWithCleanups(self._do_merge)
808
self.this_tree.lock_tree_write()
809
operation.add_cleanup(self.this_tree.unlock)
810
self.base_tree.lock_read()
811
operation.add_cleanup(self.base_tree.unlock)
812
self.other_tree.lock_read()
813
operation.add_cleanup(self.other_tree.unlock)
816
def _do_merge(self, operation):
817
self.tt = transform.TreeTransform(self.this_tree, None)
818
operation.add_cleanup(self.tt.finalize)
819
self._compute_transform()
820
results = self.tt.apply(no_conflicts=True)
821
self.write_modified(results)
823
self.this_tree.add_conflicts(self.cooked_conflicts)
824
except errors.UnsupportedOperation:
827
def make_preview_transform(self):
828
operation = OperationWithCleanups(self._make_preview_transform)
829
self.base_tree.lock_read()
830
operation.add_cleanup(self.base_tree.unlock)
831
self.other_tree.lock_read()
832
operation.add_cleanup(self.other_tree.unlock)
833
return operation.run_simple()
835
def _make_preview_transform(self):
836
self.tt = transform.TransformPreview(self.this_tree)
837
self._compute_transform()
840
def _compute_transform(self):
841
if self._lca_trees is None:
842
entries = self._entries3()
843
resolver = self._three_way
845
entries = self._entries_lca()
846
resolver = self._lca_multi_way
847
child_pb = ui.ui_factory.nested_progress_bar()
849
factories = Merger.hooks['merge_file_content']
850
hooks = [factory(self) for factory in factories] + [self]
851
self.active_hooks = [hook for hook in hooks if hook is not None]
852
for num, (file_id, changed, parents3, names3,
853
executable3) in enumerate(entries):
854
child_pb.update('Preparing file merge', num, len(entries))
855
self._merge_names(file_id, parents3, names3, resolver=resolver)
857
file_status = self._do_merge_contents(file_id)
859
file_status = 'unmodified'
860
self._merge_executable(file_id,
861
executable3, file_status, resolver=resolver)
865
self._finish_computing_transform()
867
def _finish_computing_transform(self):
868
"""Finalize the transform and report the changes.
870
This is the second half of _compute_transform.
872
child_pb = ui.ui_factory.nested_progress_bar()
874
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
875
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
878
if self.change_reporter is not None:
879
from bzrlib import delta
880
delta.report_changes(
881
self.tt.iter_changes(), self.change_reporter)
882
self.cook_conflicts(fs_conflicts)
883
for conflict in self.cooked_conflicts:
884
trace.warning(conflict)
887
"""Gather data about files modified between three trees.
889
Return a list of tuples of file_id, changed, parents3, names3,
890
executable3. changed is a boolean indicating whether the file contents
891
or kind were changed. parents3 is a tuple of parent ids for base,
892
other and this. names3 is a tuple of names for base, other and this.
893
executable3 is a tuple of execute-bit values for base, other and this.
896
iterator = self.other_tree.iter_changes(self.base_tree,
897
include_unchanged=True, specific_files=self.interesting_files,
898
extra_trees=[self.this_tree])
899
this_entries = dict((e.file_id, e) for p, e in
900
self.this_tree.iter_entries_by_dir(
901
self.interesting_ids))
902
for (file_id, paths, changed, versioned, parents, names, kind,
903
executable) in iterator:
904
if (self.interesting_ids is not None and
905
file_id not in self.interesting_ids):
907
entry = this_entries.get(file_id)
908
if entry is not None:
909
this_name = entry.name
910
this_parent = entry.parent_id
911
this_executable = entry.executable
915
this_executable = None
916
parents3 = parents + (this_parent,)
917
names3 = names + (this_name,)
918
executable3 = executable + (this_executable,)
919
result.append((file_id, changed, parents3, names3, executable3))
922
def _entries_lca(self):
923
"""Gather data about files modified between multiple trees.
925
This compares OTHER versus all LCA trees, and for interesting entries,
926
it then compares with THIS and BASE.
928
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
929
:return: [(file_id, changed, parents, names, executable)]
930
file_id Simple file_id of the entry
931
changed Boolean, True if the kind or contents changed
933
parents ((base, [parent_id, in, lcas]), parent_id_other,
935
names ((base, [name, in, lcas]), name_in_other, name_in_this)
936
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
938
if self.interesting_files is not None:
939
lookup_trees = [self.this_tree, self.base_tree]
940
lookup_trees.extend(self._lca_trees)
941
# I think we should include the lca trees as well
942
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
945
interesting_ids = self.interesting_ids
947
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
949
base_inventory = self.base_tree.inventory
950
this_inventory = self.this_tree.inventory
951
for path, file_id, other_ie, lca_values in walker.iter_all():
952
# Is this modified at all from any of the other trees?
954
other_ie = _none_entry
955
if interesting_ids is not None and file_id not in interesting_ids:
958
# If other_revision is found in any of the lcas, that means this
959
# node is uninteresting. This is because when merging, if there are
960
# multiple heads(), we have to create a new node. So if we didn't,
961
# we know that the ancestry is linear, and that OTHER did not
963
# See doc/developers/lca_merge_resolution.txt for details
964
other_revision = other_ie.revision
965
if other_revision is not None:
966
# We can't use this shortcut when other_revision is None,
967
# because it may be None because things are WorkingTrees, and
968
# not because it is *actually* None.
969
is_unmodified = False
970
for lca_path, ie in lca_values:
971
if ie is not None and ie.revision == other_revision:
978
for lca_path, lca_ie in lca_values:
980
lca_entries.append(_none_entry)
982
lca_entries.append(lca_ie)
984
if file_id in base_inventory:
985
base_ie = base_inventory[file_id]
987
base_ie = _none_entry
989
if file_id in this_inventory:
990
this_ie = this_inventory[file_id]
992
this_ie = _none_entry
998
for lca_ie in lca_entries:
999
lca_kinds.append(lca_ie.kind)
1000
lca_parent_ids.append(lca_ie.parent_id)
1001
lca_names.append(lca_ie.name)
1002
lca_executable.append(lca_ie.executable)
1004
kind_winner = self._lca_multi_way(
1005
(base_ie.kind, lca_kinds),
1006
other_ie.kind, this_ie.kind)
1007
parent_id_winner = self._lca_multi_way(
1008
(base_ie.parent_id, lca_parent_ids),
1009
other_ie.parent_id, this_ie.parent_id)
1010
name_winner = self._lca_multi_way(
1011
(base_ie.name, lca_names),
1012
other_ie.name, this_ie.name)
1014
content_changed = True
1015
if kind_winner == 'this':
1016
# No kind change in OTHER, see if there are *any* changes
1017
if other_ie.kind == 'directory':
1018
if parent_id_winner == 'this' and name_winner == 'this':
1019
# No change for this directory in OTHER, skip
1021
content_changed = False
1022
elif other_ie.kind is None or other_ie.kind == 'file':
1023
def get_sha1(ie, tree):
1024
if ie.kind != 'file':
1026
return tree.get_file_sha1(file_id)
1027
base_sha1 = get_sha1(base_ie, self.base_tree)
1028
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1029
in zip(lca_entries, self._lca_trees)]
1030
this_sha1 = get_sha1(this_ie, self.this_tree)
1031
other_sha1 = get_sha1(other_ie, self.other_tree)
1032
sha1_winner = self._lca_multi_way(
1033
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1034
allow_overriding_lca=False)
1035
exec_winner = self._lca_multi_way(
1036
(base_ie.executable, lca_executable),
1037
other_ie.executable, this_ie.executable)
1038
if (parent_id_winner == 'this' and name_winner == 'this'
1039
and sha1_winner == 'this' and exec_winner == 'this'):
1040
# No kind, parent, name, exec, or content change for
1041
# OTHER, so this node is not considered interesting
1043
if sha1_winner == 'this':
1044
content_changed = False
1045
elif other_ie.kind == 'symlink':
1046
def get_target(ie, tree):
1047
if ie.kind != 'symlink':
1049
return tree.get_symlink_target(file_id)
1050
base_target = get_target(base_ie, self.base_tree)
1051
lca_targets = [get_target(ie, tree) for ie, tree
1052
in zip(lca_entries, self._lca_trees)]
1053
this_target = get_target(this_ie, self.this_tree)
1054
other_target = get_target(other_ie, self.other_tree)
1055
target_winner = self._lca_multi_way(
1056
(base_target, lca_targets),
1057
other_target, this_target)
1058
if (parent_id_winner == 'this' and name_winner == 'this'
1059
and target_winner == 'this'):
1060
# No kind, parent, name, or symlink target change
1063
if target_winner == 'this':
1064
content_changed = False
1065
elif other_ie.kind == 'tree-reference':
1066
# The 'changed' information seems to be handled at a higher
1067
# level. At least, _entries3 returns False for content
1068
# changed, even when at a new revision_id.
1069
content_changed = False
1070
if (parent_id_winner == 'this' and name_winner == 'this'):
1071
# Nothing interesting
1074
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1076
# If we have gotten this far, that means something has changed
1077
result.append((file_id, content_changed,
1078
((base_ie.parent_id, lca_parent_ids),
1079
other_ie.parent_id, this_ie.parent_id),
1080
((base_ie.name, lca_names),
1081
other_ie.name, this_ie.name),
1082
((base_ie.executable, lca_executable),
1083
other_ie.executable, this_ie.executable)
1088
if self.tt.final_kind(self.tt.root) is None:
1089
self.tt.cancel_deletion(self.tt.root)
1090
if self.tt.final_file_id(self.tt.root) is None:
1091
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
1093
other_root_file_id = self.other_tree.get_root_id()
1094
if other_root_file_id is None:
1096
other_root = self.tt.trans_id_file_id(other_root_file_id)
1097
if other_root == self.tt.root:
1099
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
1100
# the other tree's root is a non-root in the current tree (as when
1101
# a previously unrelated branch is merged into another)
1103
if self.tt.final_kind(other_root) is not None:
1104
other_root_is_present = True
1106
# other_root doesn't have a physical representation. We still need
1107
# to move any references to the actual root of the tree.
1108
other_root_is_present = False
1109
# 'other_tree.inventory.root' is not present in this tree. We are
1110
# calling adjust_path for children which *want* to be present with a
1111
# correct place to go.
1112
for thing, child in self.other_tree.inventory.root.children.iteritems():
1113
trans_id = self.tt.trans_id_file_id(child.file_id)
1114
if not other_root_is_present:
1115
if self.tt.final_kind(trans_id) is not None:
1116
# The item exist in the final tree and has a defined place
1119
# Move the item into the root
1120
self.tt.adjust_path(self.tt.final_name(trans_id),
1121
self.tt.root, trans_id)
1122
if other_root_is_present:
1123
self.tt.cancel_creation(other_root)
1124
self.tt.cancel_versioning(other_root)
1126
def write_modified(self, results):
1127
modified_hashes = {}
1128
for path in results.modified_paths:
1129
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
1132
hash = self.this_tree.get_file_sha1(file_id)
1135
modified_hashes[file_id] = hash
1136
self.this_tree.set_merge_modified(modified_hashes)
1139
def parent(entry, file_id):
1140
"""Determine the parent for a file_id (used as a key method)"""
1143
return entry.parent_id
1146
def name(entry, file_id):
1147
"""Determine the name for a file_id (used as a key method)"""
1153
def contents_sha1(tree, file_id):
1154
"""Determine the sha1 of the file contents (used as a key method)."""
1155
if file_id not in tree:
1157
return tree.get_file_sha1(file_id)
1160
def executable(tree, file_id):
1161
"""Determine the executability of a file-id (used as a key method)."""
1162
if not tree.has_id(file_id):
1164
if tree.kind(file_id) != "file":
1166
return tree.is_executable(file_id)
1169
def kind(tree, file_id):
1170
"""Determine the kind of a file-id (used as a key method)."""
1171
if not tree.has_id(file_id):
1173
return tree.kind(file_id)
1176
def _three_way(base, other, this):
1178
# if 'base == other', either they all agree, or only 'this' has
1181
elif this not in (base, other):
1182
# 'this' is neither 'base' nor 'other', so both sides changed
1185
# "Ambiguous clean merge" -- both sides have made the same change.
1188
# this == base: only other has changed.
1192
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
1193
"""Consider LCAs when determining whether a change has occurred.
1195
If LCAS are all identical, this is the same as a _three_way comparison.
1197
:param bases: value in (BASE, [LCAS])
1198
:param other: value in OTHER
1199
:param this: value in THIS
1200
:param allow_overriding_lca: If there is more than one unique lca
1201
value, allow OTHER to override THIS if it has a new value, and
1202
THIS only has an lca value, or vice versa. This is appropriate for
1203
truly scalar values, not as much for non-scalars.
1204
:return: 'this', 'other', or 'conflict' depending on whether an entry
1207
# See doc/developers/lca_tree_merging.txt for details about this
1210
# Either Ambiguously clean, or nothing was actually changed. We
1213
base_val, lca_vals = bases
1214
# Remove 'base_val' from the lca_vals, because it is not interesting
1215
filtered_lca_vals = [lca_val for lca_val in lca_vals
1216
if lca_val != base_val]
1217
if len(filtered_lca_vals) == 0:
1218
return Merge3Merger._three_way(base_val, other, this)
1220
unique_lca_vals = set(filtered_lca_vals)
1221
if len(unique_lca_vals) == 1:
1222
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1224
if allow_overriding_lca:
1225
if other in unique_lca_vals:
1226
if this in unique_lca_vals:
1227
# Each side picked a different lca, conflict
1230
# This has a value which supersedes both lca values, and
1231
# other only has an lca value
1233
elif this in unique_lca_vals:
1234
# OTHER has a value which supersedes both lca values, and this
1235
# only has an lca value
1238
# At this point, the lcas disagree, and the tip disagree
1242
@deprecated_method(deprecated_in((2, 2, 0)))
1243
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1244
"""Do a three-way test on a scalar.
1245
Return "this", "other" or "conflict", depending whether a value wins.
1247
key_base = key(base_tree, file_id)
1248
key_other = key(other_tree, file_id)
1249
#if base == other, either they all agree, or only THIS has changed.
1250
if key_base == key_other:
1252
key_this = key(this_tree, file_id)
1253
# "Ambiguous clean merge"
1254
if key_this == key_other:
1256
elif key_this == key_base:
1261
def merge_names(self, file_id):
1262
def get_entry(tree):
1263
if tree.has_id(file_id):
1264
return tree.inventory[file_id]
1267
this_entry = get_entry(self.this_tree)
1268
other_entry = get_entry(self.other_tree)
1269
base_entry = get_entry(self.base_tree)
1270
entries = (base_entry, other_entry, this_entry)
1273
for entry in entries:
1276
parents.append(None)
1278
names.append(entry.name)
1279
parents.append(entry.parent_id)
1280
return self._merge_names(file_id, parents, names,
1281
resolver=self._three_way)
1283
def _merge_names(self, file_id, parents, names, resolver):
1284
"""Perform a merge on file_id names and parents"""
1285
base_name, other_name, this_name = names
1286
base_parent, other_parent, this_parent = parents
1288
name_winner = resolver(*names)
1290
parent_id_winner = resolver(*parents)
1291
if this_name is None:
1292
if name_winner == "this":
1293
name_winner = "other"
1294
if parent_id_winner == "this":
1295
parent_id_winner = "other"
1296
if name_winner == "this" and parent_id_winner == "this":
1298
if name_winner == 'conflict' or parent_id_winner == 'conflict':
1299
# Creating helpers (.OTHER or .THIS) here cause problems down the
1300
# road if a ContentConflict needs to be created so we should not do
1302
trans_id = self.tt.trans_id_file_id(file_id)
1303
self._raw_conflicts.append(('path conflict', trans_id, file_id,
1304
this_parent, this_name,
1305
other_parent, other_name))
1306
if other_name is None:
1307
# it doesn't matter whether the result was 'other' or
1308
# 'conflict'-- if there's no 'other', we leave it alone.
1310
parent_id = parents[self.winner_idx[parent_id_winner]]
1311
if parent_id is not None:
1312
# if we get here, name_winner and parent_winner are set to safe
1314
self.tt.adjust_path(names[self.winner_idx[name_winner]],
1315
self.tt.trans_id_file_id(parent_id),
1316
self.tt.trans_id_file_id(file_id))
1318
def _do_merge_contents(self, file_id):
1319
"""Performs a merge on file_id contents."""
1320
def contents_pair(tree):
1321
if file_id not in tree:
1323
kind = tree.kind(file_id)
1325
contents = tree.get_file_sha1(file_id)
1326
elif kind == "symlink":
1327
contents = tree.get_symlink_target(file_id)
1330
return kind, contents
1332
# See SPOT run. run, SPOT, run.
1333
# So we're not QUITE repeating ourselves; we do tricky things with
1335
base_pair = contents_pair(self.base_tree)
1336
other_pair = contents_pair(self.other_tree)
1338
this_pair = contents_pair(self.this_tree)
1339
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1340
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1341
this_pair, allow_overriding_lca=False)
1343
if base_pair == other_pair:
1346
# We delayed evaluating this_pair as long as we can to avoid
1347
# unnecessary sha1 calculation
1348
this_pair = contents_pair(self.this_tree)
1349
winner = self._three_way(base_pair, other_pair, this_pair)
1350
if winner == 'this':
1351
# No interesting changes introduced by OTHER
1353
# We have a hypothetical conflict, but if we have files, then we
1354
# can try to merge the content
1355
trans_id = self.tt.trans_id_file_id(file_id)
1356
params = MergeHookParams(self, file_id, trans_id, this_pair[0],
1357
other_pair[0], winner)
1358
hooks = self.active_hooks
1359
hook_status = 'not_applicable'
1361
hook_status, lines = hook.merge_contents(params)
1362
if hook_status != 'not_applicable':
1363
# Don't try any more hooks, this one applies.
1366
if hook_status == 'not_applicable':
1367
# This is a contents conflict, because none of the available
1368
# functions could merge it.
1370
name = self.tt.final_name(trans_id)
1371
parent_id = self.tt.final_parent(trans_id)
1372
if self.this_tree.has_id(file_id):
1373
self.tt.unversion_file(trans_id)
1374
file_group = self._dump_conflicts(name, parent_id, file_id,
1376
self._raw_conflicts.append(('contents conflict', file_group))
1377
elif hook_status == 'success':
1378
self.tt.create_file(lines, trans_id)
1379
elif hook_status == 'conflicted':
1380
# XXX: perhaps the hook should be able to provide
1381
# the BASE/THIS/OTHER files?
1382
self.tt.create_file(lines, trans_id)
1383
self._raw_conflicts.append(('text conflict', trans_id))
1384
name = self.tt.final_name(trans_id)
1385
parent_id = self.tt.final_parent(trans_id)
1386
self._dump_conflicts(name, parent_id, file_id)
1387
elif hook_status == 'delete':
1388
self.tt.unversion_file(trans_id)
1390
elif hook_status == 'done':
1391
# The hook function did whatever it needs to do directly, no
1392
# further action needed here.
1395
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1396
if not self.this_tree.has_id(file_id) and result == "modified":
1397
self.tt.version_file(file_id, trans_id)
1398
# The merge has been performed, so the old contents should not be
1400
self.tt.delete_contents(trans_id)
1403
def _default_other_winner_merge(self, merge_hook_params):
1404
"""Replace this contents with other."""
1405
file_id = merge_hook_params.file_id
1406
trans_id = merge_hook_params.trans_id
1407
file_in_this = self.this_tree.has_id(file_id)
1408
if self.other_tree.has_id(file_id):
1409
# OTHER changed the file
1411
if wt.supports_content_filtering():
1412
# We get the path from the working tree if it exists.
1413
# That fails though when OTHER is adding a file, so
1414
# we fall back to the other tree to find the path if
1415
# it doesn't exist locally.
1417
filter_tree_path = wt.id2path(file_id)
1418
except errors.NoSuchId:
1419
filter_tree_path = self.other_tree.id2path(file_id)
1421
# Skip the id2path lookup for older formats
1422
filter_tree_path = None
1423
transform.create_from_tree(self.tt, trans_id,
1424
self.other_tree, file_id,
1425
filter_tree_path=filter_tree_path)
1428
# OTHER deleted the file
1429
return 'delete', None
1431
raise AssertionError(
1432
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1435
def merge_contents(self, merge_hook_params):
1436
"""Fallback merge logic after user installed hooks."""
1437
# This function is used in merge hooks as the fallback instance.
1438
# Perhaps making this function and the functions it calls be a
1439
# a separate class would be better.
1440
if merge_hook_params.winner == 'other':
1441
# OTHER is a straight winner, so replace this contents with other
1442
return self._default_other_winner_merge(merge_hook_params)
1443
elif merge_hook_params.is_file_merge():
1444
# THIS and OTHER are both files, so text merge. Either
1445
# BASE is a file, or both converted to files, so at least we
1446
# have agreement that output should be a file.
1448
self.text_merge(merge_hook_params.file_id,
1449
merge_hook_params.trans_id)
1450
except errors.BinaryFile:
1451
return 'not_applicable', None
1454
return 'not_applicable', None
1456
def get_lines(self, tree, file_id):
1457
"""Return the lines in a file, or an empty list."""
1458
if tree.has_id(file_id):
1459
return tree.get_file_lines(file_id)
1463
def text_merge(self, file_id, trans_id):
1464
"""Perform a three-way text merge on a file_id"""
1465
# it's possible that we got here with base as a different type.
1466
# if so, we just want two-way text conflicts.
1467
if self.base_tree.has_id(file_id) and \
1468
self.base_tree.kind(file_id) == "file":
1469
base_lines = self.get_lines(self.base_tree, file_id)
1472
other_lines = self.get_lines(self.other_tree, file_id)
1473
this_lines = self.get_lines(self.this_tree, file_id)
1474
m3 = merge3.Merge3(base_lines, this_lines, other_lines,
1475
is_cherrypick=self.cherrypick)
1476
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1477
if self.show_base is True:
1478
base_marker = '|' * 7
1482
def iter_merge3(retval):
1483
retval["text_conflicts"] = False
1484
for line in m3.merge_lines(name_a = "TREE",
1485
name_b = "MERGE-SOURCE",
1486
name_base = "BASE-REVISION",
1487
start_marker=start_marker,
1488
base_marker=base_marker,
1489
reprocess=self.reprocess):
1490
if line.startswith(start_marker):
1491
retval["text_conflicts"] = True
1492
yield line.replace(start_marker, '<' * 7)
1496
merge3_iterator = iter_merge3(retval)
1497
self.tt.create_file(merge3_iterator, trans_id)
1498
if retval["text_conflicts"] is True:
1499
self._raw_conflicts.append(('text conflict', trans_id))
1500
name = self.tt.final_name(trans_id)
1501
parent_id = self.tt.final_parent(trans_id)
1502
file_group = self._dump_conflicts(name, parent_id, file_id,
1503
this_lines, base_lines,
1505
file_group.append(trans_id)
1507
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1508
base_lines=None, other_lines=None, set_version=False,
1510
"""Emit conflict files.
1511
If this_lines, base_lines, or other_lines are omitted, they will be
1512
determined automatically. If set_version is true, the .OTHER, .THIS
1513
or .BASE (in that order) will be created as versioned files.
1515
data = [('OTHER', self.other_tree, other_lines),
1516
('THIS', self.this_tree, this_lines)]
1518
data.append(('BASE', self.base_tree, base_lines))
1520
# We need to use the actual path in the working tree of the file here,
1521
# ignoring the conflict suffixes
1523
if wt.supports_content_filtering():
1525
filter_tree_path = wt.id2path(file_id)
1526
except errors.NoSuchId:
1527
# file has been deleted
1528
filter_tree_path = None
1530
# Skip the id2path lookup for older formats
1531
filter_tree_path = None
1535
for suffix, tree, lines in data:
1536
if tree.has_id(file_id):
1537
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1538
suffix, lines, filter_tree_path)
1539
file_group.append(trans_id)
1540
if set_version and not versioned:
1541
self.tt.version_file(file_id, trans_id)
1545
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1546
lines=None, filter_tree_path=None):
1547
"""Emit a single conflict file."""
1548
name = name + '.' + suffix
1549
trans_id = self.tt.create_path(name, parent_id)
1550
transform.create_from_tree(self.tt, trans_id, tree, file_id, lines,
1554
def merge_executable(self, file_id, file_status):
1555
"""Perform a merge on the execute bit."""
1556
executable = [self.executable(t, file_id) for t in (self.base_tree,
1557
self.other_tree, self.this_tree)]
1558
self._merge_executable(file_id, executable, file_status,
1559
resolver=self._three_way)
1561
def _merge_executable(self, file_id, executable, file_status,
1563
"""Perform a merge on the execute bit."""
1564
base_executable, other_executable, this_executable = executable
1565
if file_status == "deleted":
1567
winner = resolver(*executable)
1568
if winner == "conflict":
1569
# There must be a None in here, if we have a conflict, but we
1570
# need executability since file status was not deleted.
1571
if self.executable(self.other_tree, file_id) is None:
1575
if winner == 'this' and file_status != "modified":
1577
trans_id = self.tt.trans_id_file_id(file_id)
1578
if self.tt.final_kind(trans_id) != "file":
1580
if winner == "this":
1581
executability = this_executable
1583
if self.other_tree.has_id(file_id):
1584
executability = other_executable
1585
elif self.this_tree.has_id(file_id):
1586
executability = this_executable
1587
elif self.base_tree_has_id(file_id):
1588
executability = base_executable
1589
if executability is not None:
1590
trans_id = self.tt.trans_id_file_id(file_id)
1591
self.tt.set_executability(executability, trans_id)
1593
def cook_conflicts(self, fs_conflicts):
1594
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1595
self.cooked_conflicts.extend(transform.cook_conflicts(
1596
fs_conflicts, self.tt))
1597
fp = transform.FinalPaths(self.tt)
1598
for conflict in self._raw_conflicts:
1599
conflict_type = conflict[0]
1600
if conflict_type == 'path conflict':
1602
this_parent, this_name,
1603
other_parent, other_name) = conflict[1:]
1604
if this_parent is None or this_name is None:
1605
this_path = '<deleted>'
1607
parent_path = fp.get_path(
1608
self.tt.trans_id_file_id(this_parent))
1609
this_path = osutils.pathjoin(parent_path, this_name)
1610
if other_parent is None or other_name is None:
1611
other_path = '<deleted>'
1613
parent_path = fp.get_path(
1614
self.tt.trans_id_file_id(other_parent))
1615
other_path = osutils.pathjoin(parent_path, other_name)
1616
c = _mod_conflicts.Conflict.factory(
1617
'path conflict', path=this_path,
1618
conflict_path=other_path,
1620
elif conflict_type == 'contents conflict':
1621
for trans_id in conflict[1]:
1622
file_id = self.tt.final_file_id(trans_id)
1623
if file_id is not None:
1625
path = fp.get_path(trans_id)
1626
for suffix in ('.BASE', '.THIS', '.OTHER'):
1627
if path.endswith(suffix):
1628
path = path[:-len(suffix)]
1630
c = _mod_conflicts.Conflict.factory(conflict_type,
1631
path=path, file_id=file_id)
1632
elif conflict_type == 'text conflict':
1633
trans_id = conflict[1]
1634
path = fp.get_path(trans_id)
1635
file_id = self.tt.final_file_id(trans_id)
1636
c = _mod_conflicts.Conflict.factory(conflict_type,
1637
path=path, file_id=file_id)
1639
raise AssertionError('bad conflict type: %r' % (conflict,))
1640
self.cooked_conflicts.append(c)
1641
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
1644
class WeaveMerger(Merge3Merger):
1645
"""Three-way tree merger, text weave merger."""
1646
supports_reprocess = True
1647
supports_show_base = False
1648
supports_reverse_cherrypick = False
1649
history_based = True
1651
def _generate_merge_plan(self, file_id, base):
1652
return self.this_tree.plan_file_merge(file_id, self.other_tree,
1655
def _merged_lines(self, file_id):
1656
"""Generate the merged lines.
1657
There is no distinction between lines that are meant to contain <<<<<<<
1661
base = self.base_tree
1664
plan = self._generate_merge_plan(file_id, base)
1665
if 'merge' in debug.debug_flags:
1667
trans_id = self.tt.trans_id_file_id(file_id)
1668
name = self.tt.final_name(trans_id) + '.plan'
1669
contents = ('%11s|%s' % l for l in plan)
1670
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1671
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1672
'>>>>>>> MERGE-SOURCE\n')
1673
lines, conflicts = textmerge.merge_lines(self.reprocess)
1675
base_lines = textmerge.base_from_plan()
1678
return lines, base_lines
1680
def text_merge(self, file_id, trans_id):
1681
"""Perform a (weave) text merge for a given file and file-id.
1682
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1683
and a conflict will be noted.
1685
lines, base_lines = self._merged_lines(file_id)
1687
# Note we're checking whether the OUTPUT is binary in this case,
1688
# because we don't want to get into weave merge guts.
1689
textfile.check_text_lines(lines)
1690
self.tt.create_file(lines, trans_id)
1691
if base_lines is not None:
1693
self._raw_conflicts.append(('text conflict', trans_id))
1694
name = self.tt.final_name(trans_id)
1695
parent_id = self.tt.final_parent(trans_id)
1696
file_group = self._dump_conflicts(name, parent_id, file_id,
1698
base_lines=base_lines)
1699
file_group.append(trans_id)
1702
class LCAMerger(WeaveMerger):
1704
def _generate_merge_plan(self, file_id, base):
1705
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1708
class Diff3Merger(Merge3Merger):
1709
"""Three-way merger using external diff3 for text merging"""
1711
def dump_file(self, temp_dir, name, tree, file_id):
1712
out_path = osutils.pathjoin(temp_dir, name)
1713
out_file = open(out_path, "wb")
1715
in_file = tree.get_file(file_id)
1716
for line in in_file:
1717
out_file.write(line)
1722
def text_merge(self, file_id, trans_id):
1723
"""Perform a diff3 merge using a specified file-id and trans-id.
1724
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1725
will be dumped, and a will be conflict noted.
1728
temp_dir = osutils.mkdtemp(prefix="bzr-")
1730
new_file = osutils.pathjoin(temp_dir, "new")
1731
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1732
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1733
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1734
status = bzrlib.patch.diff3(new_file, this, base, other)
1735
if status not in (0, 1):
1736
raise errors.BzrError("Unhandled diff3 exit code")
1737
f = open(new_file, 'rb')
1739
self.tt.create_file(f, trans_id)
1743
name = self.tt.final_name(trans_id)
1744
parent_id = self.tt.final_parent(trans_id)
1745
self._dump_conflicts(name, parent_id, file_id)
1746
self._raw_conflicts.append(('text conflict', trans_id))
1748
osutils.rmtree(temp_dir)
1751
class PathNotInTree(errors.BzrError):
1753
_fmt = """Merge-into failed because %(tree)s does not contain %(path)s."""
1755
def __init__(self, path, tree):
1756
errors.BzrError.__init__(self, path=path, tree=tree)
1759
class MergeIntoMerger(Merger):
1760
"""Merger that understands other_tree will be merged into a subdir.
1762
This also changes the Merger api so that it uses real Branch, revision_id,
1763
and RevisonTree objects, rather than using revision specs.
1766
def __init__(self, this_tree, other_branch, other_tree, target_subdir,
1767
source_subpath, other_rev_id=None):
1768
"""Create a new MergeIntoMerger object.
1770
source_subpath in other_tree will be effectively copied to
1771
target_subdir in this_tree.
1773
:param this_tree: The tree that we will be merging into.
1774
:param other_branch: The Branch we will be merging from.
1775
:param other_tree: The RevisionTree object we want to merge.
1776
:param target_subdir: The relative path where we want to merge
1777
other_tree into this_tree
1778
:param source_subpath: The relative path specifying the subtree of
1779
other_tree to merge into this_tree.
1781
# It is assumed that we are merging a tree that is not in our current
1782
# ancestry, which means we are using the "EmptyTree" as our basis.
1783
null_ancestor_tree = this_tree.branch.repository.revision_tree(
1784
_mod_revision.NULL_REVISION)
1785
super(MergeIntoMerger, self).__init__(
1786
this_branch=this_tree.branch,
1787
this_tree=this_tree,
1788
other_tree=other_tree,
1789
base_tree=null_ancestor_tree,
1791
self._target_subdir = target_subdir
1792
self._source_subpath = source_subpath
1793
self.other_branch = other_branch
1794
if other_rev_id is None:
1795
other_rev_id = other_tree.get_revision_id()
1796
self.other_rev_id = self.other_basis = other_rev_id
1797
self.base_is_ancestor = True
1798
self.backup_files = True
1799
self.merge_type = Merge3Merger
1800
self.show_base = False
1801
self.reprocess = False
1802
self.interesting_ids = None
1803
self.merge_type = _MergeTypeParameterizer(MergeIntoMergeType,
1804
target_subdir=self._target_subdir,
1805
source_subpath=self._source_subpath)
1806
if self._source_subpath != '':
1807
# If this isn't a partial merge make sure the revisions will be
1809
self._maybe_fetch(self.other_branch, self.this_branch,
1812
def set_pending(self):
1813
if self._source_subpath != '':
1815
Merger.set_pending(self)
1818
class _MergeTypeParameterizer(object):
1819
"""Wrap a merge-type class to provide extra parameters.
1821
This is hack used by MergeIntoMerger to pass some extra parameters to its
1822
merge_type. Merger.do_merge() sets up its own set of parameters to pass to
1823
the 'merge_type' member. It is difficult override do_merge without
1824
re-writing the whole thing, so instead we create a wrapper which will pass
1825
the extra parameters.
1828
def __init__(self, merge_type, **kwargs):
1829
self._extra_kwargs = kwargs
1830
self._merge_type = merge_type
1832
def __call__(self, *args, **kwargs):
1833
kwargs.update(self._extra_kwargs)
1834
return self._merge_type(*args, **kwargs)
1836
def __getattr__(self, name):
1837
return getattr(self._merge_type, name)
1840
class MergeIntoMergeType(Merge3Merger):
1841
"""Merger that incorporates a tree (or part of a tree) into another."""
1843
def __init__(self, *args, **kwargs):
1844
"""Initialize the merger object.
1846
:param args: See Merge3Merger.__init__'s args.
1847
:param kwargs: See Merge3Merger.__init__'s keyword args, except for
1848
source_subpath and target_subdir.
1849
:keyword source_subpath: The relative path specifying the subtree of
1850
other_tree to merge into this_tree.
1851
:keyword target_subdir: The relative path where we want to merge
1852
other_tree into this_tree
1854
# All of the interesting work happens during Merge3Merger.__init__(),
1855
# so we have have to hack in to get our extra parameters set.
1856
self._source_subpath = kwargs.pop('source_subpath')
1857
self._target_subdir = kwargs.pop('target_subdir')
1858
super(MergeIntoMergeType, self).__init__(*args, **kwargs)
1860
def _compute_transform(self):
1861
child_pb = ui.ui_factory.nested_progress_bar()
1863
entries = self._entries_to_incorporate()
1864
entries = list(entries)
1865
for num, (entry, parent_id) in enumerate(entries):
1866
child_pb.update('Preparing file merge', num, len(entries))
1867
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1868
trans_id = transform.new_by_entry(self.tt, entry,
1869
parent_trans_id, self.other_tree)
1872
self._finish_computing_transform()
1874
def _entries_to_incorporate(self):
1875
"""Yields pairs of (inventory_entry, new_parent)."""
1876
other_inv = self.other_tree.inventory
1877
subdir_id = other_inv.path2id(self._source_subpath)
1878
if subdir_id is None:
1879
# XXX: The error would be clearer if it gave the URL of the source
1880
# branch, but we don't have a reference to that here.
1881
raise PathNotInTree(self._source_subpath, "Source tree")
1882
subdir = other_inv[subdir_id]
1883
parent_in_target = osutils.dirname(self._target_subdir)
1884
target_id = self.this_tree.inventory.path2id(parent_in_target)
1885
if target_id is None:
1886
raise PathNotInTree(self._target_subdir, "Target tree")
1887
name_in_target = osutils.basename(self._target_subdir)
1888
merge_into_root = subdir.copy()
1889
merge_into_root.name = name_in_target
1890
if merge_into_root.file_id in self.this_tree.inventory:
1891
# Give the root a new file-id.
1892
# This can happen fairly easily if the directory we are
1893
# incorporating is the root, and both trees have 'TREE_ROOT' as
1894
# their root_id. Users will expect this to Just Work, so we
1895
# change the file-id here.
1896
# Non-root file-ids could potentially conflict too. That's really
1897
# an edge case, so we don't do anything special for those. We let
1898
# them cause conflicts.
1899
merge_into_root.file_id = generate_ids.gen_file_id(name_in_target)
1900
yield (merge_into_root, target_id)
1901
if subdir.kind != 'directory':
1902
# No children, so we are done.
1904
for ignored_path, entry in other_inv.iter_entries_by_dir(subdir_id):
1905
parent_id = entry.parent_id
1906
if parent_id == subdir.file_id:
1907
# The root's parent ID has changed, so make sure children of
1908
# the root refer to the new ID.
1909
parent_id = merge_into_root.file_id
1910
yield (entry, parent_id)
1913
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1915
merge_type=Merge3Merger,
1916
interesting_ids=None,
1920
interesting_files=None,
1923
change_reporter=None):
1924
"""Primary interface for merging.
1926
Typical use is probably::
1928
merge_inner(branch, branch.get_revision_tree(other_revision),
1929
branch.get_revision_tree(base_revision))
1931
if this_tree is None:
1932
raise errors.BzrError("bzrlib.merge.merge_inner requires a this_tree "
1933
"parameter as of bzrlib version 0.8.")
1934
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1935
pb=pb, change_reporter=change_reporter)
1936
merger.backup_files = backup_files
1937
merger.merge_type = merge_type
1938
merger.interesting_ids = interesting_ids
1939
merger.ignore_zero = ignore_zero
1940
if interesting_files:
1942
raise ValueError('Only supply interesting_ids'
1943
' or interesting_files')
1944
merger.interesting_files = interesting_files
1945
merger.show_base = show_base
1946
merger.reprocess = reprocess
1947
merger.other_rev_id = other_rev_id
1948
merger.other_basis = other_rev_id
1949
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1950
if get_revision_id is None:
1951
get_revision_id = base_tree.last_revision
1952
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
1953
merger.set_base_revision(get_revision_id(), this_branch)
1954
return merger.do_merge()
1956
def get_merge_type_registry():
1957
"""Merge type registry is in bzrlib.option to avoid circular imports.
1959
This method provides a sanctioned way to retrieve it.
1961
from bzrlib import option
1962
return option._merge_type_registry
1965
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1966
def status_a(revision, text):
1967
if revision in ancestors_b:
1968
return 'killed-b', text
1970
return 'new-a', text
1972
def status_b(revision, text):
1973
if revision in ancestors_a:
1974
return 'killed-a', text
1976
return 'new-b', text
1978
plain_a = [t for (a, t) in annotated_a]
1979
plain_b = [t for (a, t) in annotated_b]
1980
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1981
blocks = matcher.get_matching_blocks()
1984
for ai, bi, l in blocks:
1985
# process all mismatched sections
1986
# (last mismatched section is handled because blocks always
1987
# includes a 0-length last block)
1988
for revision, text in annotated_a[a_cur:ai]:
1989
yield status_a(revision, text)
1990
for revision, text in annotated_b[b_cur:bi]:
1991
yield status_b(revision, text)
1992
# and now the matched section
1995
for text_a in plain_a[ai:a_cur]:
1996
yield "unchanged", text_a
1999
class _PlanMergeBase(object):
2001
def __init__(self, a_rev, b_rev, vf, key_prefix):
2004
:param a_rev: Revision-id of one revision to merge
2005
:param b_rev: Revision-id of the other revision to merge
2006
:param vf: A VersionedFiles containing both revisions
2007
:param key_prefix: A prefix for accessing keys in vf, typically
2013
self._last_lines = None
2014
self._last_lines_revision_id = None
2015
self._cached_matching_blocks = {}
2016
self._key_prefix = key_prefix
2017
self._precache_tip_lines()
2019
def _precache_tip_lines(self):
2020
lines = self.get_lines([self.a_rev, self.b_rev])
2021
self.lines_a = lines[self.a_rev]
2022
self.lines_b = lines[self.b_rev]
2024
def get_lines(self, revisions):
2025
"""Get lines for revisions from the backing VersionedFiles.
2027
:raises RevisionNotPresent: on absent texts.
2029
keys = [(self._key_prefix + (rev,)) for rev in revisions]
2031
for record in self.vf.get_record_stream(keys, 'unordered', True):
2032
if record.storage_kind == 'absent':
2033
raise errors.RevisionNotPresent(record.key, self.vf)
2034
result[record.key[-1]] = osutils.chunks_to_lines(
2035
record.get_bytes_as('chunked'))
2038
def plan_merge(self):
2039
"""Generate a 'plan' for merging the two revisions.
2041
This involves comparing their texts and determining the cause of
2042
differences. If text A has a line and text B does not, then either the
2043
line was added to text A, or it was deleted from B. Once the causes
2044
are combined, they are written out in the format described in
2045
VersionedFile.plan_merge
2047
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
2048
unique_a, unique_b = self._unique_lines(blocks)
2049
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
2050
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
2051
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
2053
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
2056
for i, j, n in blocks:
2057
for a_index in range(last_i, i):
2058
if a_index in new_a:
2059
if a_index in killed_b:
2060
yield 'conflicted-a', self.lines_a[a_index]
2062
yield 'new-a', self.lines_a[a_index]
2064
yield 'killed-b', self.lines_a[a_index]
2065
for b_index in range(last_j, j):
2066
if b_index in new_b:
2067
if b_index in killed_a:
2068
yield 'conflicted-b', self.lines_b[b_index]
2070
yield 'new-b', self.lines_b[b_index]
2072
yield 'killed-a', self.lines_b[b_index]
2073
# handle common lines
2074
for a_index in range(i, i+n):
2075
yield 'unchanged', self.lines_a[a_index]
2079
def _get_matching_blocks(self, left_revision, right_revision):
2080
"""Return a description of which sections of two revisions match.
2082
See SequenceMatcher.get_matching_blocks
2084
cached = self._cached_matching_blocks.get((left_revision,
2086
if cached is not None:
2088
if self._last_lines_revision_id == left_revision:
2089
left_lines = self._last_lines
2090
right_lines = self.get_lines([right_revision])[right_revision]
2092
lines = self.get_lines([left_revision, right_revision])
2093
left_lines = lines[left_revision]
2094
right_lines = lines[right_revision]
2095
self._last_lines = right_lines
2096
self._last_lines_revision_id = right_revision
2097
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
2099
return matcher.get_matching_blocks()
2101
def _unique_lines(self, matching_blocks):
2102
"""Analyse matching_blocks to determine which lines are unique
2104
:return: a tuple of (unique_left, unique_right), where the values are
2105
sets of line numbers of unique lines.
2111
for i, j, n in matching_blocks:
2112
unique_left.extend(range(last_i, i))
2113
unique_right.extend(range(last_j, j))
2116
return unique_left, unique_right
2119
def _subtract_plans(old_plan, new_plan):
2120
"""Remove changes from new_plan that came from old_plan.
2122
It is assumed that the difference between the old_plan and new_plan
2123
is their choice of 'b' text.
2125
All lines from new_plan that differ from old_plan are emitted
2126
verbatim. All lines from new_plan that match old_plan but are
2127
not about the 'b' revision are emitted verbatim.
2129
Lines that match and are about the 'b' revision are the lines we
2130
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
2131
is skipped entirely.
2133
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
2136
for i, j, n in matcher.get_matching_blocks():
2137
for jj in range(last_j, j):
2139
for jj in range(j, j+n):
2140
plan_line = new_plan[jj]
2141
if plan_line[0] == 'new-b':
2143
elif plan_line[0] == 'killed-b':
2144
yield 'unchanged', plan_line[1]
2150
class _PlanMerge(_PlanMergeBase):
2151
"""Plan an annotate merge using on-the-fly annotation"""
2153
def __init__(self, a_rev, b_rev, vf, key_prefix):
2154
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
2155
self.a_key = self._key_prefix + (self.a_rev,)
2156
self.b_key = self._key_prefix + (self.b_rev,)
2157
self.graph = _mod_graph.Graph(self.vf)
2158
heads = self.graph.heads((self.a_key, self.b_key))
2160
# one side dominates, so we can just return its values, yay for
2162
# Ideally we would know that before we get this far
2163
self._head_key = heads.pop()
2164
if self._head_key == self.a_key:
2168
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2169
self._head_key[-1], other)
2172
self._head_key = None
2175
def _precache_tip_lines(self):
2176
# Turn this into a no-op, because we will do this later
2179
def _find_recursive_lcas(self):
2180
"""Find all the ancestors back to a unique lca"""
2181
cur_ancestors = (self.a_key, self.b_key)
2182
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2183
# rather than a key tuple. We will just map that directly to no common
2187
next_lcas = self.graph.find_lca(*cur_ancestors)
2188
# Map a plain NULL_REVISION to a simple no-ancestors
2189
if next_lcas == set([_mod_revision.NULL_REVISION]):
2191
# Order the lca's based on when they were merged into the tip
2192
# While the actual merge portion of weave merge uses a set() of
2193
# active revisions, the order of insertion *does* effect the
2194
# implicit ordering of the texts.
2195
for rev_key in cur_ancestors:
2196
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2198
parent_map[rev_key] = ordered_parents
2199
if len(next_lcas) == 0:
2201
elif len(next_lcas) == 1:
2202
parent_map[list(next_lcas)[0]] = ()
2204
elif len(next_lcas) > 2:
2205
# More than 2 lca's, fall back to grabbing all nodes between
2206
# this and the unique lca.
2207
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2209
self.a_key, self.b_key, cur_ancestors)
2210
cur_lcas = next_lcas
2211
while len(cur_lcas) > 1:
2212
cur_lcas = self.graph.find_lca(*cur_lcas)
2213
if len(cur_lcas) == 0:
2214
# No common base to find, use the full ancestry
2217
unique_lca = list(cur_lcas)[0]
2218
if unique_lca == _mod_revision.NULL_REVISION:
2219
# find_lca will return a plain 'NULL_REVISION' rather
2220
# than a key tuple when there is no common ancestor, we
2221
# prefer to just use None, because it doesn't confuse
2222
# _get_interesting_texts()
2224
parent_map.update(self._find_unique_parents(next_lcas,
2227
cur_ancestors = next_lcas
2230
def _find_unique_parents(self, tip_keys, base_key):
2231
"""Find ancestors of tip that aren't ancestors of base.
2233
:param tip_keys: Nodes that are interesting
2234
:param base_key: Cull all ancestors of this node
2235
:return: The parent map for all revisions between tip_keys and
2236
base_key. base_key will be included. References to nodes outside of
2237
the ancestor set will also be removed.
2239
# TODO: this would be simpler if find_unique_ancestors took a list
2240
# instead of a single tip, internally it supports it, but it
2241
# isn't a "backwards compatible" api change.
2242
if base_key is None:
2243
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2244
# We remove NULL_REVISION because it isn't a proper tuple key, and
2245
# thus confuses things like _get_interesting_texts, and our logic
2246
# to add the texts into the memory weave.
2247
if _mod_revision.NULL_REVISION in parent_map:
2248
parent_map.pop(_mod_revision.NULL_REVISION)
2251
for tip in tip_keys:
2253
self.graph.find_unique_ancestors(tip, [base_key]))
2254
parent_map = self.graph.get_parent_map(interesting)
2255
parent_map[base_key] = ()
2256
culled_parent_map, child_map, tails = self._remove_external_references(
2258
# Remove all the tails but base_key
2259
if base_key is not None:
2260
tails.remove(base_key)
2261
self._prune_tails(culled_parent_map, child_map, tails)
2262
# Now remove all the uninteresting 'linear' regions
2263
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2267
def _remove_external_references(parent_map):
2268
"""Remove references that go outside of the parent map.
2270
:param parent_map: Something returned from Graph.get_parent_map(keys)
2271
:return: (filtered_parent_map, child_map, tails)
2272
filtered_parent_map is parent_map without external references
2273
child_map is the {parent_key: [child_keys]} mapping
2274
tails is a list of nodes that do not have any parents in the map
2276
# TODO: The basic effect of this function seems more generic than
2277
# _PlanMerge. But the specific details of building a child_map,
2278
# and computing tails seems very specific to _PlanMerge.
2279
# Still, should this be in Graph land?
2280
filtered_parent_map = {}
2283
for key, parent_keys in parent_map.iteritems():
2284
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2285
if not culled_parent_keys:
2287
for parent_key in culled_parent_keys:
2288
child_map.setdefault(parent_key, []).append(key)
2289
# TODO: Do we want to do this, it adds overhead for every node,
2290
# just to say that the node has no children
2291
child_map.setdefault(key, [])
2292
filtered_parent_map[key] = culled_parent_keys
2293
return filtered_parent_map, child_map, tails
2296
def _prune_tails(parent_map, child_map, tails_to_remove):
2297
"""Remove tails from the parent map.
2299
This will remove the supplied revisions until no more children have 0
2302
:param parent_map: A dict of {child: [parents]}, this dictionary will
2303
be modified in place.
2304
:param tails_to_remove: A list of tips that should be removed,
2305
this list will be consumed
2306
:param child_map: The reverse dict of parent_map ({parent: [children]})
2307
this dict will be modified
2308
:return: None, parent_map will be modified in place.
2310
while tails_to_remove:
2311
next = tails_to_remove.pop()
2312
parent_map.pop(next)
2313
children = child_map.pop(next)
2314
for child in children:
2315
child_parents = parent_map[child]
2316
child_parents.remove(next)
2317
if len(child_parents) == 0:
2318
tails_to_remove.append(child)
2320
def _get_interesting_texts(self, parent_map):
2321
"""Return a dict of texts we are interested in.
2323
Note that the input is in key tuples, but the output is in plain
2326
:param parent_map: The output from _find_recursive_lcas
2327
:return: A dict of {'revision_id':lines} as returned by
2328
_PlanMergeBase.get_lines()
2330
all_revision_keys = set(parent_map)
2331
all_revision_keys.add(self.a_key)
2332
all_revision_keys.add(self.b_key)
2334
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2335
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2338
def _build_weave(self):
2339
from bzrlib import weave
2340
self._weave = weave.Weave(weave_name='in_memory_weave',
2341
allow_reserved=True)
2342
parent_map = self._find_recursive_lcas()
2344
all_texts = self._get_interesting_texts(parent_map)
2346
# Note: Unfortunately, the order given by topo_sort will effect the
2347
# ordering resolution in the output. Specifically, if you add A then B,
2348
# then in the output text A lines will show up before B lines. And, of
2349
# course, topo_sort doesn't guarantee any real ordering.
2350
# So we use merge_sort, and add a fake node on the tip.
2351
# This ensures that left-hand parents will always be inserted into the
2352
# weave before right-hand parents.
2353
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2354
parent_map[tip_key] = (self.a_key, self.b_key)
2356
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2360
# for key in tsort.topo_sort(parent_map):
2361
parent_keys = parent_map[key]
2362
revision_id = key[-1]
2363
parent_ids = [k[-1] for k in parent_keys]
2364
self._weave.add_lines(revision_id, parent_ids,
2365
all_texts[revision_id])
2367
def plan_merge(self):
2368
"""Generate a 'plan' for merging the two revisions.
2370
This involves comparing their texts and determining the cause of
2371
differences. If text A has a line and text B does not, then either the
2372
line was added to text A, or it was deleted from B. Once the causes
2373
are combined, they are written out in the format described in
2374
VersionedFile.plan_merge
2376
if self._head_key is not None: # There was a single head
2377
if self._head_key == self.a_key:
2380
if self._head_key != self.b_key:
2381
raise AssertionError('There was an invalid head: %s != %s'
2382
% (self.b_key, self._head_key))
2384
head_rev = self._head_key[-1]
2385
lines = self.get_lines([head_rev])[head_rev]
2386
return ((plan, line) for line in lines)
2387
return self._weave.plan_merge(self.a_rev, self.b_rev)
2390
class _PlanLCAMerge(_PlanMergeBase):
2392
This merge algorithm differs from _PlanMerge in that:
2393
1. comparisons are done against LCAs only
2394
2. cases where a contested line is new versus one LCA but old versus
2395
another are marked as conflicts, by emitting the line as conflicted-a
2398
This is faster, and hopefully produces more useful output.
2401
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2402
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2403
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2406
if lca == _mod_revision.NULL_REVISION:
2409
self.lcas.add(lca[-1])
2410
for lca in self.lcas:
2411
if _mod_revision.is_null(lca):
2414
lca_lines = self.get_lines([lca])[lca]
2415
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2417
blocks = list(matcher.get_matching_blocks())
2418
self._cached_matching_blocks[(a_rev, lca)] = blocks
2419
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2421
blocks = list(matcher.get_matching_blocks())
2422
self._cached_matching_blocks[(b_rev, lca)] = blocks
2424
def _determine_status(self, revision_id, unique_line_numbers):
2425
"""Determines the status unique lines versus all lcas.
2427
Basically, determines why the line is unique to this revision.
2429
A line may be determined new, killed, or both.
2431
If a line is determined new, that means it was not present in at least
2432
one LCA, and is not present in the other merge revision.
2434
If a line is determined killed, that means the line was present in
2437
If a line is killed and new, this indicates that the two merge
2438
revisions contain differing conflict resolutions.
2439
:param revision_id: The id of the revision in which the lines are
2441
:param unique_line_numbers: The line numbers of unique lines.
2442
:return a tuple of (new_this, killed_other):
2446
unique_line_numbers = set(unique_line_numbers)
2447
for lca in self.lcas:
2448
blocks = self._get_matching_blocks(revision_id, lca)
2449
unique_vs_lca, _ignored = self._unique_lines(blocks)
2450
new.update(unique_line_numbers.intersection(unique_vs_lca))
2451
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
