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
20
branch as _mod_branch,
21
conflicts as _mod_conflicts,
30
revision as _mod_revision,
39
from bzrlib.cleanup import OperationWithCleanups
40
from bzrlib.symbol_versioning import (
44
# TODO: Report back as changes are merged in
47
def transform_tree(from_tree, to_tree, interesting_ids=None):
48
from_tree.lock_tree_write()
49
operation = OperationWithCleanups(merge_inner)
50
operation.add_cleanup(from_tree.unlock)
51
operation.run_simple(from_tree.branch, to_tree, from_tree,
52
ignore_zero=True, interesting_ids=interesting_ids, this_tree=from_tree)
55
class MergeHooks(hooks.Hooks):
58
hooks.Hooks.__init__(self)
59
self.create_hook(hooks.HookPoint('merge_file_content',
60
"Called with a bzrlib.merge.Merger object to create a per file "
61
"merge object when starting a merge. "
62
"Should return either None or a subclass of "
63
"``bzrlib.merge.AbstractPerFileMerger``. "
64
"Such objects will then be called per file "
65
"that needs to be merged (including when one "
66
"side has deleted the file and the other has changed it). "
67
"See the AbstractPerFileMerger API docs for details on how it is "
72
class AbstractPerFileMerger(object):
73
"""PerFileMerger objects are used by plugins extending merge for bzrlib.
75
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
77
:ivar merger: The Merge3Merger performing the merge.
80
def __init__(self, merger):
81
"""Create a PerFileMerger for use with merger."""
84
def merge_contents(self, merge_params):
85
"""Attempt to merge the contents of a single file.
87
:param merge_params: A bzrlib.merge.MergeHookParams
88
:return : A tuple of (status, chunks), where status is one of
89
'not_applicable', 'success', 'conflicted', or 'delete'. If status
90
is 'success' or 'conflicted', then chunks should be an iterable of
91
strings for the new file contents.
93
return ('not applicable', None)
96
class PerFileMerger(AbstractPerFileMerger):
97
"""Merge individual files when self.file_matches returns True.
99
This class is intended to be subclassed. The file_matches and
100
merge_matching methods should be overridden with concrete implementations.
103
def file_matches(self, params):
104
"""Return True if merge_matching should be called on this file.
106
Only called with merges of plain files with no clear winner.
108
Subclasses must override this.
110
raise NotImplementedError(self.file_matches)
112
def get_filename(self, params, tree):
113
"""Lookup the filename (i.e. basename, not path), given a Tree (e.g.
114
self.merger.this_tree) and a MergeHookParams.
116
return osutils.basename(tree.id2path(params.file_id))
118
def get_filepath(self, params, tree):
119
"""Calculate the path to the file in a tree.
121
:param params: A MergeHookParams describing the file to merge
122
:param tree: a Tree, e.g. self.merger.this_tree.
124
return tree.id2path(params.file_id)
126
def merge_contents(self, params):
127
"""Merge the contents of a single file."""
128
# Check whether this custom merge logic should be used.
130
# OTHER is a straight winner, rely on default merge.
131
params.winner == 'other' or
132
# THIS and OTHER aren't both files.
133
not params.is_file_merge() or
134
# The filename doesn't match *.xml
135
not self.file_matches(params)):
136
return 'not_applicable', None
137
return self.merge_matching(params)
139
def merge_matching(self, params):
140
"""Merge the contents of a single file that has matched the criteria
141
in PerFileMerger.merge_contents (is a conflict, is a file,
142
self.file_matches is True).
144
Subclasses must override this.
146
raise NotImplementedError(self.merge_matching)
149
class ConfigurableFileMerger(PerFileMerger):
150
"""Merge individual files when configured via a .conf file.
152
This is a base class for concrete custom file merging logic. Concrete
153
classes should implement ``merge_text``.
155
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
157
:ivar affected_files: The configured file paths to merge.
159
:cvar name_prefix: The prefix to use when looking up configuration
160
details. <name_prefix>_merge_files describes the files targeted by the
163
:cvar default_files: The default file paths to merge when no configuration
170
def __init__(self, merger):
171
super(ConfigurableFileMerger, self).__init__(merger)
172
self.affected_files = None
173
self.default_files = self.__class__.default_files or []
174
self.name_prefix = self.__class__.name_prefix
175
if self.name_prefix is None:
176
raise ValueError("name_prefix must be set.")
178
def file_matches(self, params):
179
"""Check whether the file should call the merge hook.
181
<name_prefix>_merge_files configuration variable is a list of files
182
that should use the hook.
184
affected_files = self.affected_files
185
if affected_files is None:
186
config = self.merger.this_branch.get_config()
187
# Until bzr provides a better policy for caching the config, we
188
# just add the part we're interested in to the params to avoid
189
# reading the config files repeatedly (bazaar.conf, location.conf,
191
config_key = self.name_prefix + '_merge_files'
192
affected_files = config.get_user_option_as_list(config_key)
193
if affected_files is None:
194
# If nothing was specified in the config, use the default.
195
affected_files = self.default_files
196
self.affected_files = affected_files
198
filepath = self.get_filepath(params, self.merger.this_tree)
199
if filepath in affected_files:
203
def merge_matching(self, params):
204
return self.merge_text(params)
206
def merge_text(self, params):
207
"""Merge the byte contents of a single file.
209
This is called after checking that the merge should be performed in
210
merge_contents, and it should behave as per
211
``bzrlib.merge.AbstractPerFileMerger.merge_contents``.
213
raise NotImplementedError(self.merge_text)
216
class MergeHookParams(object):
217
"""Object holding parameters passed to merge_file_content hooks.
219
There are some fields hooks can access:
221
:ivar file_id: the file ID of the file being merged
222
:ivar trans_id: the transform ID for the merge of this file
223
:ivar this_kind: kind of file_id in 'this' tree
224
:ivar other_kind: kind of file_id in 'other' tree
225
:ivar winner: one of 'this', 'other', 'conflict'
228
def __init__(self, merger, file_id, trans_id, this_kind, other_kind,
230
self._merger = merger
231
self.file_id = file_id
232
self.trans_id = trans_id
233
self.this_kind = this_kind
234
self.other_kind = other_kind
237
def is_file_merge(self):
238
"""True if this_kind and other_kind are both 'file'."""
239
return self.this_kind == 'file' and self.other_kind == 'file'
241
@decorators.cachedproperty
242
def base_lines(self):
243
"""The lines of the 'base' version of the file."""
244
return self._merger.get_lines(self._merger.base_tree, self.file_id)
246
@decorators.cachedproperty
247
def this_lines(self):
248
"""The lines of the 'this' version of the file."""
249
return self._merger.get_lines(self._merger.this_tree, self.file_id)
251
@decorators.cachedproperty
252
def other_lines(self):
253
"""The lines of the 'other' version of the file."""
254
return self._merger.get_lines(self._merger.other_tree, self.file_id)
257
class Merger(object):
261
def __init__(self, this_branch, other_tree=None, base_tree=None,
262
this_tree=None, pb=None, change_reporter=None,
263
recurse='down', revision_graph=None):
264
object.__init__(self)
265
self.this_branch = this_branch
266
self.this_basis = _mod_revision.ensure_null(
267
this_branch.last_revision())
268
self.this_rev_id = None
269
self.this_tree = this_tree
270
self.this_revision_tree = None
271
self.this_basis_tree = None
272
self.other_tree = other_tree
273
self.other_branch = None
274
self.base_tree = base_tree
275
self.ignore_zero = False
276
self.backup_files = False
277
self.interesting_ids = None
278
self.interesting_files = None
279
self.show_base = False
280
self.reprocess = False
282
warnings.warn("pb parameter to Merger() is deprecated and ignored")
284
self.recurse = recurse
285
self.change_reporter = change_reporter
286
self._cached_trees = {}
287
self._revision_graph = revision_graph
288
self._base_is_ancestor = None
289
self._base_is_other_ancestor = None
290
self._is_criss_cross = None
291
self._lca_trees = None
293
def cache_trees_with_revision_ids(self, trees):
294
"""Cache any tree in trees if it has a revision_id."""
295
for maybe_tree in trees:
296
if maybe_tree is None:
299
rev_id = maybe_tree.get_revision_id()
300
except AttributeError:
302
self._cached_trees[rev_id] = maybe_tree
305
def revision_graph(self):
306
if self._revision_graph is None:
307
self._revision_graph = self.this_branch.repository.get_graph()
308
return self._revision_graph
310
def _set_base_is_ancestor(self, value):
311
self._base_is_ancestor = value
313
def _get_base_is_ancestor(self):
314
if self._base_is_ancestor is None:
315
self._base_is_ancestor = self.revision_graph.is_ancestor(
316
self.base_rev_id, self.this_basis)
317
return self._base_is_ancestor
319
base_is_ancestor = property(_get_base_is_ancestor, _set_base_is_ancestor)
321
def _set_base_is_other_ancestor(self, value):
322
self._base_is_other_ancestor = value
324
def _get_base_is_other_ancestor(self):
325
if self._base_is_other_ancestor is None:
326
if self.other_basis is None:
328
self._base_is_other_ancestor = self.revision_graph.is_ancestor(
329
self.base_rev_id, self.other_basis)
330
return self._base_is_other_ancestor
332
base_is_other_ancestor = property(_get_base_is_other_ancestor,
333
_set_base_is_other_ancestor)
336
def from_uncommitted(tree, other_tree, pb=None, base_tree=None):
337
"""Return a Merger for uncommitted changes in other_tree.
339
:param tree: The tree to merge into
340
:param other_tree: The tree to get uncommitted changes from
341
:param pb: A progress indicator
342
:param base_tree: The basis to use for the merge. If unspecified,
343
other_tree.basis_tree() will be used.
345
if base_tree is None:
346
base_tree = other_tree.basis_tree()
347
merger = Merger(tree.branch, other_tree, base_tree, tree, pb)
348
merger.base_rev_id = merger.base_tree.get_revision_id()
349
merger.other_rev_id = None
350
merger.other_basis = merger.base_rev_id
354
def from_mergeable(klass, tree, mergeable, pb):
355
"""Return a Merger for a bundle or merge directive.
357
:param tree: The tree to merge changes into
358
:param mergeable: A merge directive or bundle
359
:param pb: A progress indicator
361
mergeable.install_revisions(tree.branch.repository)
362
base_revision_id, other_revision_id, verified =\
363
mergeable.get_merge_request(tree.branch.repository)
364
revision_graph = tree.branch.repository.get_graph()
365
if base_revision_id is not None:
366
if (base_revision_id != _mod_revision.NULL_REVISION and
367
revision_graph.is_ancestor(
368
base_revision_id, tree.branch.last_revision())):
369
base_revision_id = None
371
trace.warning('Performing cherrypick')
372
merger = klass.from_revision_ids(pb, tree, other_revision_id,
373
base_revision_id, revision_graph=
375
return merger, verified
378
def from_revision_ids(pb, tree, other, base=None, other_branch=None,
379
base_branch=None, revision_graph=None,
381
"""Return a Merger for revision-ids.
383
:param pb: A progress indicator
384
:param tree: The tree to merge changes into
385
:param other: The revision-id to use as OTHER
386
:param base: The revision-id to use as BASE. If not specified, will
388
:param other_branch: A branch containing the other revision-id. If
389
not supplied, tree.branch is used.
390
:param base_branch: A branch containing the base revision-id. If
391
not supplied, other_branch or tree.branch will be used.
392
:param revision_graph: If you have a revision_graph precomputed, pass
393
it in, otherwise it will be created for you.
394
:param tree_branch: The branch associated with tree. If not supplied,
395
tree.branch will be used.
397
if tree_branch is None:
398
tree_branch = tree.branch
399
merger = Merger(tree_branch, this_tree=tree, pb=pb,
400
revision_graph=revision_graph)
401
if other_branch is None:
402
other_branch = tree.branch
403
merger.set_other_revision(other, other_branch)
407
if base_branch is None:
408
base_branch = other_branch
409
merger.set_base_revision(base, base_branch)
412
def revision_tree(self, revision_id, branch=None):
413
if revision_id not in self._cached_trees:
415
branch = self.this_branch
417
tree = self.this_tree.revision_tree(revision_id)
418
except errors.NoSuchRevisionInTree:
419
tree = branch.repository.revision_tree(revision_id)
420
self._cached_trees[revision_id] = tree
421
return self._cached_trees[revision_id]
423
def _get_tree(self, treespec, possible_transports=None):
424
from bzrlib import workingtree
425
location, revno = treespec
427
tree = workingtree.WorkingTree.open_containing(location)[0]
428
return tree.branch, tree
429
branch = _mod_branch.Branch.open_containing(
430
location, possible_transports)[0]
432
revision_id = branch.last_revision()
434
revision_id = branch.get_rev_id(revno)
435
revision_id = _mod_revision.ensure_null(revision_id)
436
return branch, self.revision_tree(revision_id, branch)
438
@deprecated_method(deprecated_in((2, 1, 0)))
439
def ensure_revision_trees(self):
440
if self.this_revision_tree is None:
441
self.this_basis_tree = self.revision_tree(self.this_basis)
442
if self.this_basis == self.this_rev_id:
443
self.this_revision_tree = self.this_basis_tree
445
if self.other_rev_id is None:
446
other_basis_tree = self.revision_tree(self.other_basis)
447
if other_basis_tree.has_changes(self.other_tree):
448
raise errors.WorkingTreeNotRevision(self.this_tree)
449
other_rev_id = self.other_basis
450
self.other_tree = other_basis_tree
452
@deprecated_method(deprecated_in((2, 1, 0)))
453
def file_revisions(self, file_id):
454
self.ensure_revision_trees()
455
def get_id(tree, file_id):
456
revision_id = tree.inventory[file_id].revision
458
if self.this_rev_id is None:
459
if self.this_basis_tree.get_file_sha1(file_id) != \
460
self.this_tree.get_file_sha1(file_id):
461
raise errors.WorkingTreeNotRevision(self.this_tree)
463
trees = (self.this_basis_tree, self.other_tree)
464
return [get_id(tree, file_id) for tree in trees]
466
@deprecated_method(deprecated_in((2, 1, 0)))
467
def check_basis(self, check_clean, require_commits=True):
468
if self.this_basis is None and require_commits is True:
469
raise errors.BzrCommandError(
470
"This branch has no commits."
471
" (perhaps you would prefer 'bzr pull')")
474
if self.this_basis != self.this_rev_id:
475
raise errors.UncommittedChanges(self.this_tree)
477
@deprecated_method(deprecated_in((2, 1, 0)))
478
def compare_basis(self):
480
basis_tree = self.revision_tree(self.this_tree.last_revision())
481
except errors.NoSuchRevision:
482
basis_tree = self.this_tree.basis_tree()
483
if not self.this_tree.has_changes(basis_tree):
484
self.this_rev_id = self.this_basis
486
def set_interesting_files(self, file_list):
487
self.interesting_files = file_list
489
def set_pending(self):
490
if (not self.base_is_ancestor or not self.base_is_other_ancestor
491
or self.other_rev_id is None):
495
def _add_parent(self):
496
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
497
new_parent_trees = []
498
operation = OperationWithCleanups(self.this_tree.set_parent_trees)
499
for revision_id in new_parents:
501
tree = self.revision_tree(revision_id)
502
except errors.NoSuchRevision:
506
operation.add_cleanup(tree.unlock)
507
new_parent_trees.append((revision_id, tree))
508
operation.run_simple(new_parent_trees, allow_leftmost_as_ghost=True)
510
def set_other(self, other_revision, possible_transports=None):
511
"""Set the revision and tree to merge from.
513
This sets the other_tree, other_rev_id, other_basis attributes.
515
:param other_revision: The [path, revision] list to merge from.
517
self.other_branch, self.other_tree = self._get_tree(other_revision,
519
if other_revision[1] == -1:
520
self.other_rev_id = _mod_revision.ensure_null(
521
self.other_branch.last_revision())
522
if _mod_revision.is_null(self.other_rev_id):
523
raise errors.NoCommits(self.other_branch)
524
self.other_basis = self.other_rev_id
525
elif other_revision[1] is not None:
526
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
527
self.other_basis = self.other_rev_id
529
self.other_rev_id = None
530
self.other_basis = self.other_branch.last_revision()
531
if self.other_basis is None:
532
raise errors.NoCommits(self.other_branch)
533
if self.other_rev_id is not None:
534
self._cached_trees[self.other_rev_id] = self.other_tree
535
self._maybe_fetch(self.other_branch,self.this_branch, self.other_basis)
537
def set_other_revision(self, revision_id, other_branch):
538
"""Set 'other' based on a branch and revision id
540
:param revision_id: The revision to use for a tree
541
:param other_branch: The branch containing this tree
543
self.other_rev_id = revision_id
544
self.other_branch = other_branch
545
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
546
self.other_tree = self.revision_tree(revision_id)
547
self.other_basis = revision_id
549
def set_base_revision(self, revision_id, branch):
550
"""Set 'base' based on a branch and revision id
552
:param revision_id: The revision to use for a tree
553
:param branch: The branch containing this tree
555
self.base_rev_id = revision_id
556
self.base_branch = branch
557
self._maybe_fetch(branch, self.this_branch, revision_id)
558
self.base_tree = self.revision_tree(revision_id)
560
def _maybe_fetch(self, source, target, revision_id):
561
if not source.repository.has_same_location(target.repository):
562
target.fetch(source, revision_id)
565
revisions = [_mod_revision.ensure_null(self.this_basis),
566
_mod_revision.ensure_null(self.other_basis)]
567
if _mod_revision.NULL_REVISION in revisions:
568
self.base_rev_id = _mod_revision.NULL_REVISION
569
self.base_tree = self.revision_tree(self.base_rev_id)
570
self._is_criss_cross = False
572
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
573
self._is_criss_cross = False
575
self.base_rev_id = _mod_revision.NULL_REVISION
577
self.base_rev_id = list(lcas)[0]
578
else: # len(lcas) > 1
580
# find_unique_lca can only handle 2 nodes, so we have to
581
# start back at the beginning. It is a shame to traverse
582
# the graph again, but better than re-implementing
584
self.base_rev_id = self.revision_graph.find_unique_lca(
585
revisions[0], revisions[1])
587
self.base_rev_id = self.revision_graph.find_unique_lca(
589
self._is_criss_cross = True
590
if self.base_rev_id == _mod_revision.NULL_REVISION:
591
raise errors.UnrelatedBranches()
592
if self._is_criss_cross:
593
trace.warning('Warning: criss-cross merge encountered. See bzr'
594
' help criss-cross.')
595
trace.mutter('Criss-cross lcas: %r' % lcas)
596
interesting_revision_ids = [self.base_rev_id]
597
interesting_revision_ids.extend(lcas)
598
interesting_trees = dict((t.get_revision_id(), t)
599
for t in self.this_branch.repository.revision_trees(
600
interesting_revision_ids))
601
self._cached_trees.update(interesting_trees)
602
self.base_tree = interesting_trees.pop(self.base_rev_id)
603
sorted_lca_keys = self.revision_graph.find_merge_order(
605
self._lca_trees = [interesting_trees[key]
606
for key in sorted_lca_keys]
608
self.base_tree = self.revision_tree(self.base_rev_id)
609
self.base_is_ancestor = True
610
self.base_is_other_ancestor = True
611
trace.mutter('Base revid: %r' % self.base_rev_id)
613
def set_base(self, base_revision):
614
"""Set the base revision to use for the merge.
616
:param base_revision: A 2-list containing a path and revision number.
618
trace.mutter("doing merge() with no base_revision specified")
619
if base_revision == [None, None]:
622
base_branch, self.base_tree = self._get_tree(base_revision)
623
if base_revision[1] == -1:
624
self.base_rev_id = base_branch.last_revision()
625
elif base_revision[1] is None:
626
self.base_rev_id = _mod_revision.NULL_REVISION
628
self.base_rev_id = _mod_revision.ensure_null(
629
base_branch.get_rev_id(base_revision[1]))
630
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
632
def make_merger(self):
633
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
634
'other_tree': self.other_tree,
635
'interesting_ids': self.interesting_ids,
636
'interesting_files': self.interesting_files,
637
'this_branch': self.this_branch,
639
if self.merge_type.requires_base:
640
kwargs['base_tree'] = self.base_tree
641
if self.merge_type.supports_reprocess:
642
kwargs['reprocess'] = self.reprocess
644
raise errors.BzrError(
645
"Conflict reduction is not supported for merge"
646
" type %s." % self.merge_type)
647
if self.merge_type.supports_show_base:
648
kwargs['show_base'] = self.show_base
650
raise errors.BzrError("Showing base is not supported for this"
651
" merge type. %s" % self.merge_type)
652
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
653
and not self.base_is_other_ancestor):
654
raise errors.CannotReverseCherrypick()
655
if self.merge_type.supports_cherrypick:
656
kwargs['cherrypick'] = (not self.base_is_ancestor or
657
not self.base_is_other_ancestor)
658
if self._is_criss_cross and getattr(self.merge_type,
659
'supports_lca_trees', False):
660
kwargs['lca_trees'] = self._lca_trees
661
return self.merge_type(pb=None,
662
change_reporter=self.change_reporter,
665
def _do_merge_to(self):
666
merge = self.make_merger()
667
if self.other_branch is not None:
668
self.other_branch.update_references(self.this_branch)
670
if self.recurse == 'down':
671
for relpath, file_id in self.this_tree.iter_references():
672
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
673
other_revision = self.other_tree.get_reference_revision(
675
if other_revision == sub_tree.last_revision():
677
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
678
sub_merge.merge_type = self.merge_type
679
other_branch = self.other_branch.reference_parent(file_id, relpath)
680
sub_merge.set_other_revision(other_revision, other_branch)
681
base_revision = self.base_tree.get_reference_revision(file_id)
682
sub_merge.base_tree = \
683
sub_tree.branch.repository.revision_tree(base_revision)
684
sub_merge.base_rev_id = base_revision
689
operation = OperationWithCleanups(self._do_merge_to)
690
self.this_tree.lock_tree_write()
691
operation.add_cleanup(self.this_tree.unlock)
692
if self.base_tree is not None:
693
self.base_tree.lock_read()
694
operation.add_cleanup(self.base_tree.unlock)
695
if self.other_tree is not None:
696
self.other_tree.lock_read()
697
operation.add_cleanup(self.other_tree.unlock)
698
merge = operation.run_simple()
699
if len(merge.cooked_conflicts) == 0:
700
if not self.ignore_zero and not trace.is_quiet():
701
trace.note("All changes applied successfully.")
703
trace.note("%d conflicts encountered."
704
% len(merge.cooked_conflicts))
706
return len(merge.cooked_conflicts)
709
class _InventoryNoneEntry(object):
710
"""This represents an inventory entry which *isn't there*.
712
It simplifies the merging logic if we always have an InventoryEntry, even
713
if it isn't actually present
720
symlink_target = None
723
_none_entry = _InventoryNoneEntry()
726
class Merge3Merger(object):
727
"""Three-way merger that uses the merge3 text merger"""
729
supports_reprocess = True
730
supports_show_base = True
731
history_based = False
732
supports_cherrypick = True
733
supports_reverse_cherrypick = True
734
winner_idx = {"this": 2, "other": 1, "conflict": 1}
735
supports_lca_trees = True
737
def __init__(self, working_tree, this_tree, base_tree, other_tree,
738
interesting_ids=None, reprocess=False, show_base=False,
739
pb=None, pp=None, change_reporter=None,
740
interesting_files=None, do_merge=True,
741
cherrypick=False, lca_trees=None, this_branch=None):
742
"""Initialize the merger object and perform the merge.
744
:param working_tree: The working tree to apply the merge to
745
:param this_tree: The local tree in the merge operation
746
:param base_tree: The common tree in the merge operation
747
:param other_tree: The other tree to merge changes from
748
:param this_branch: The branch associated with this_tree. Defaults to
749
this_tree.branch if not supplied.
750
:param interesting_ids: The file_ids of files that should be
751
participate in the merge. May not be combined with
753
:param: reprocess If True, perform conflict-reduction processing.
754
:param show_base: If True, show the base revision in text conflicts.
755
(incompatible with reprocess)
757
:param pp: A ProgressPhase object
758
:param change_reporter: An object that should report changes made
759
:param interesting_files: The tree-relative paths of files that should
760
participate in the merge. If these paths refer to directories,
761
the contents of those directories will also be included. May not
762
be combined with interesting_ids. If neither interesting_files nor
763
interesting_ids is specified, all files may participate in the
765
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
766
if the ancestry was found to include a criss-cross merge.
767
Otherwise should be None.
769
object.__init__(self)
770
if interesting_files is not None and interesting_ids is not None:
772
'specify either interesting_ids or interesting_files')
773
if this_branch is None:
774
this_branch = this_tree.branch
775
self.interesting_ids = interesting_ids
776
self.interesting_files = interesting_files
777
self.this_tree = working_tree
778
self.base_tree = base_tree
779
self.other_tree = other_tree
780
self.this_branch = this_branch
781
self._raw_conflicts = []
782
self.cooked_conflicts = []
783
self.reprocess = reprocess
784
self.show_base = show_base
785
self._lca_trees = lca_trees
786
# Uncommenting this will change the default algorithm to always use
787
# _entries_lca. This can be useful for running the test suite and
788
# making sure we haven't missed any corner cases.
789
# if lca_trees is None:
790
# self._lca_trees = [self.base_tree]
791
self.change_reporter = change_reporter
792
self.cherrypick = cherrypick
796
warnings.warn("pp argument to Merge3Merger is deprecated")
798
warnings.warn("pb argument to Merge3Merger is deprecated")
801
operation = OperationWithCleanups(self._do_merge)
802
self.this_tree.lock_tree_write()
803
operation.add_cleanup(self.this_tree.unlock)
804
self.base_tree.lock_read()
805
operation.add_cleanup(self.base_tree.unlock)
806
self.other_tree.lock_read()
807
operation.add_cleanup(self.other_tree.unlock)
810
def _do_merge(self, operation):
811
self.tt = transform.TreeTransform(self.this_tree, None)
812
operation.add_cleanup(self.tt.finalize)
813
self._compute_transform()
814
results = self.tt.apply(no_conflicts=True)
815
self.write_modified(results)
817
self.this_tree.add_conflicts(self.cooked_conflicts)
818
except errors.UnsupportedOperation:
821
def make_preview_transform(self):
822
operation = OperationWithCleanups(self._make_preview_transform)
823
self.base_tree.lock_read()
824
operation.add_cleanup(self.base_tree.unlock)
825
self.other_tree.lock_read()
826
operation.add_cleanup(self.other_tree.unlock)
827
return operation.run_simple()
829
def _make_preview_transform(self):
830
self.tt = transform.TransformPreview(self.this_tree)
831
self._compute_transform()
834
def _compute_transform(self):
835
if self._lca_trees is None:
836
entries = self._entries3()
837
resolver = self._three_way
839
entries = self._entries_lca()
840
resolver = self._lca_multi_way
841
child_pb = ui.ui_factory.nested_progress_bar()
843
factories = Merger.hooks['merge_file_content']
844
hooks = [factory(self) for factory in factories] + [self]
845
self.active_hooks = [hook for hook in hooks if hook is not None]
846
for num, (file_id, changed, parents3, names3,
847
executable3) in enumerate(entries):
848
child_pb.update('Preparing file merge', num, len(entries))
849
self._merge_names(file_id, parents3, names3, resolver=resolver)
851
file_status = self._do_merge_contents(file_id)
853
file_status = 'unmodified'
854
self._merge_executable(file_id,
855
executable3, file_status, resolver=resolver)
859
child_pb = ui.ui_factory.nested_progress_bar()
861
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
862
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
865
if self.change_reporter is not None:
866
from bzrlib import delta
867
delta.report_changes(
868
self.tt.iter_changes(), self.change_reporter)
869
self.cook_conflicts(fs_conflicts)
870
for conflict in self.cooked_conflicts:
871
trace.warning(conflict)
874
"""Gather data about files modified between three trees.
876
Return a list of tuples of file_id, changed, parents3, names3,
877
executable3. changed is a boolean indicating whether the file contents
878
or kind were changed. parents3 is a tuple of parent ids for base,
879
other and this. names3 is a tuple of names for base, other and this.
880
executable3 is a tuple of execute-bit values for base, other and this.
883
iterator = self.other_tree.iter_changes(self.base_tree,
884
include_unchanged=True, specific_files=self.interesting_files,
885
extra_trees=[self.this_tree])
886
this_entries = dict((e.file_id, e) for p, e in
887
self.this_tree.iter_entries_by_dir(
888
self.interesting_ids))
889
for (file_id, paths, changed, versioned, parents, names, kind,
890
executable) in iterator:
891
if (self.interesting_ids is not None and
892
file_id not in self.interesting_ids):
894
entry = this_entries.get(file_id)
895
if entry is not None:
896
this_name = entry.name
897
this_parent = entry.parent_id
898
this_executable = entry.executable
902
this_executable = None
903
parents3 = parents + (this_parent,)
904
names3 = names + (this_name,)
905
executable3 = executable + (this_executable,)
906
result.append((file_id, changed, parents3, names3, executable3))
909
def _entries_lca(self):
910
"""Gather data about files modified between multiple trees.
912
This compares OTHER versus all LCA trees, and for interesting entries,
913
it then compares with THIS and BASE.
915
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
916
:return: [(file_id, changed, parents, names, executable)]
917
file_id Simple file_id of the entry
918
changed Boolean, True if the kind or contents changed
920
parents ((base, [parent_id, in, lcas]), parent_id_other,
922
names ((base, [name, in, lcas]), name_in_other, name_in_this)
923
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
925
if self.interesting_files is not None:
926
lookup_trees = [self.this_tree, self.base_tree]
927
lookup_trees.extend(self._lca_trees)
928
# I think we should include the lca trees as well
929
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
932
interesting_ids = self.interesting_ids
934
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
936
base_inventory = self.base_tree.inventory
937
this_inventory = self.this_tree.inventory
938
for path, file_id, other_ie, lca_values in walker.iter_all():
939
# Is this modified at all from any of the other trees?
941
other_ie = _none_entry
942
if interesting_ids is not None and file_id not in interesting_ids:
945
# If other_revision is found in any of the lcas, that means this
946
# node is uninteresting. This is because when merging, if there are
947
# multiple heads(), we have to create a new node. So if we didn't,
948
# we know that the ancestry is linear, and that OTHER did not
950
# See doc/developers/lca_merge_resolution.txt for details
951
other_revision = other_ie.revision
952
if other_revision is not None:
953
# We can't use this shortcut when other_revision is None,
954
# because it may be None because things are WorkingTrees, and
955
# not because it is *actually* None.
956
is_unmodified = False
957
for lca_path, ie in lca_values:
958
if ie is not None and ie.revision == other_revision:
965
for lca_path, lca_ie in lca_values:
967
lca_entries.append(_none_entry)
969
lca_entries.append(lca_ie)
971
if file_id in base_inventory:
972
base_ie = base_inventory[file_id]
974
base_ie = _none_entry
976
if file_id in this_inventory:
977
this_ie = this_inventory[file_id]
979
this_ie = _none_entry
985
for lca_ie in lca_entries:
986
lca_kinds.append(lca_ie.kind)
987
lca_parent_ids.append(lca_ie.parent_id)
988
lca_names.append(lca_ie.name)
989
lca_executable.append(lca_ie.executable)
991
kind_winner = self._lca_multi_way(
992
(base_ie.kind, lca_kinds),
993
other_ie.kind, this_ie.kind)
994
parent_id_winner = self._lca_multi_way(
995
(base_ie.parent_id, lca_parent_ids),
996
other_ie.parent_id, this_ie.parent_id)
997
name_winner = self._lca_multi_way(
998
(base_ie.name, lca_names),
999
other_ie.name, this_ie.name)
1001
content_changed = True
1002
if kind_winner == 'this':
1003
# No kind change in OTHER, see if there are *any* changes
1004
if other_ie.kind == 'directory':
1005
if parent_id_winner == 'this' and name_winner == 'this':
1006
# No change for this directory in OTHER, skip
1008
content_changed = False
1009
elif other_ie.kind is None or other_ie.kind == 'file':
1010
def get_sha1(ie, tree):
1011
if ie.kind != 'file':
1013
return tree.get_file_sha1(file_id)
1014
base_sha1 = get_sha1(base_ie, self.base_tree)
1015
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1016
in zip(lca_entries, self._lca_trees)]
1017
this_sha1 = get_sha1(this_ie, self.this_tree)
1018
other_sha1 = get_sha1(other_ie, self.other_tree)
1019
sha1_winner = self._lca_multi_way(
1020
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1021
allow_overriding_lca=False)
1022
exec_winner = self._lca_multi_way(
1023
(base_ie.executable, lca_executable),
1024
other_ie.executable, this_ie.executable)
1025
if (parent_id_winner == 'this' and name_winner == 'this'
1026
and sha1_winner == 'this' and exec_winner == 'this'):
1027
# No kind, parent, name, exec, or content change for
1028
# OTHER, so this node is not considered interesting
1030
if sha1_winner == 'this':
1031
content_changed = False
1032
elif other_ie.kind == 'symlink':
1033
def get_target(ie, tree):
1034
if ie.kind != 'symlink':
1036
return tree.get_symlink_target(file_id)
1037
base_target = get_target(base_ie, self.base_tree)
1038
lca_targets = [get_target(ie, tree) for ie, tree
1039
in zip(lca_entries, self._lca_trees)]
1040
this_target = get_target(this_ie, self.this_tree)
1041
other_target = get_target(other_ie, self.other_tree)
1042
target_winner = self._lca_multi_way(
1043
(base_target, lca_targets),
1044
other_target, this_target)
1045
if (parent_id_winner == 'this' and name_winner == 'this'
1046
and target_winner == 'this'):
1047
# No kind, parent, name, or symlink target change
1050
if target_winner == 'this':
1051
content_changed = False
1052
elif other_ie.kind == 'tree-reference':
1053
# The 'changed' information seems to be handled at a higher
1054
# level. At least, _entries3 returns False for content
1055
# changed, even when at a new revision_id.
1056
content_changed = False
1057
if (parent_id_winner == 'this' and name_winner == 'this'):
1058
# Nothing interesting
1061
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1063
# If we have gotten this far, that means something has changed
1064
result.append((file_id, content_changed,
1065
((base_ie.parent_id, lca_parent_ids),
1066
other_ie.parent_id, this_ie.parent_id),
1067
((base_ie.name, lca_names),
1068
other_ie.name, this_ie.name),
1069
((base_ie.executable, lca_executable),
1070
other_ie.executable, this_ie.executable)
1077
self.tt.final_kind(self.tt.root)
1078
except errors.NoSuchFile:
1079
self.tt.cancel_deletion(self.tt.root)
1080
if self.tt.final_file_id(self.tt.root) is None:
1081
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
1083
other_root_file_id = self.other_tree.get_root_id()
1084
if other_root_file_id is None:
1086
other_root = self.tt.trans_id_file_id(other_root_file_id)
1087
if other_root == self.tt.root:
1089
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
1090
# the other tree's root is a non-root in the current tree (as when
1091
# a previously unrelated branch is merged into another)
1094
self.tt.final_kind(other_root)
1095
other_root_is_present = True
1096
except errors.NoSuchFile:
1097
# other_root doesn't have a physical representation. We still need
1098
# to move any references to the actual root of the tree.
1099
other_root_is_present = False
1100
# 'other_tree.inventory.root' is not present in this tree. We are
1101
# calling adjust_path for children which *want* to be present with a
1102
# correct place to go.
1103
for thing, child in self.other_tree.inventory.root.children.iteritems():
1104
trans_id = self.tt.trans_id_file_id(child.file_id)
1105
if not other_root_is_present:
1106
# FIXME: Make final_kind returns None instead of raising
1107
# NoSuchFile to avoid the ugly construct below -- vila 20100402
1109
self.tt.final_kind(trans_id)
1110
# The item exist in the final tree and has a defined place
1113
except errors.NoSuchFile, e:
1115
# Move the item into the root
1116
self.tt.adjust_path(self.tt.final_name(trans_id),
1117
self.tt.root, trans_id)
1118
if other_root_is_present:
1119
self.tt.cancel_creation(other_root)
1120
self.tt.cancel_versioning(other_root)
1122
def write_modified(self, results):
1123
modified_hashes = {}
1124
for path in results.modified_paths:
1125
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
1128
hash = self.this_tree.get_file_sha1(file_id)
1131
modified_hashes[file_id] = hash
1132
self.this_tree.set_merge_modified(modified_hashes)
1135
def parent(entry, file_id):
1136
"""Determine the parent for a file_id (used as a key method)"""
1139
return entry.parent_id
1142
def name(entry, file_id):
1143
"""Determine the name for a file_id (used as a key method)"""
1149
def contents_sha1(tree, file_id):
1150
"""Determine the sha1 of the file contents (used as a key method)."""
1151
if file_id not in tree:
1153
return tree.get_file_sha1(file_id)
1156
def executable(tree, file_id):
1157
"""Determine the executability of a file-id (used as a key method)."""
1158
if not tree.has_id(file_id):
1160
if tree.kind(file_id) != "file":
1162
return tree.is_executable(file_id)
1165
def kind(tree, file_id):
1166
"""Determine the kind of a file-id (used as a key method)."""
1167
if not tree.has_id(file_id):
1169
return tree.kind(file_id)
1172
def _three_way(base, other, this):
1174
# if 'base == other', either they all agree, or only 'this' has
1177
elif this not in (base, other):
1178
# 'this' is neither 'base' nor 'other', so both sides changed
1181
# "Ambiguous clean merge" -- both sides have made the same change.
1184
# this == base: only other has changed.
1188
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
1189
"""Consider LCAs when determining whether a change has occurred.
1191
If LCAS are all identical, this is the same as a _three_way comparison.
1193
:param bases: value in (BASE, [LCAS])
1194
:param other: value in OTHER
1195
:param this: value in THIS
1196
:param allow_overriding_lca: If there is more than one unique lca
1197
value, allow OTHER to override THIS if it has a new value, and
1198
THIS only has an lca value, or vice versa. This is appropriate for
1199
truly scalar values, not as much for non-scalars.
1200
:return: 'this', 'other', or 'conflict' depending on whether an entry
1203
# See doc/developers/lca_tree_merging.txt for details about this
1206
# Either Ambiguously clean, or nothing was actually changed. We
1209
base_val, lca_vals = bases
1210
# Remove 'base_val' from the lca_vals, because it is not interesting
1211
filtered_lca_vals = [lca_val for lca_val in lca_vals
1212
if lca_val != base_val]
1213
if len(filtered_lca_vals) == 0:
1214
return Merge3Merger._three_way(base_val, other, this)
1216
unique_lca_vals = set(filtered_lca_vals)
1217
if len(unique_lca_vals) == 1:
1218
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1220
if allow_overriding_lca:
1221
if other in unique_lca_vals:
1222
if this in unique_lca_vals:
1223
# Each side picked a different lca, conflict
1226
# This has a value which supersedes both lca values, and
1227
# other only has an lca value
1229
elif this in unique_lca_vals:
1230
# OTHER has a value which supersedes both lca values, and this
1231
# only has an lca value
1234
# At this point, the lcas disagree, and the tip disagree
1238
@deprecated_method(deprecated_in((2, 2, 0)))
1239
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1240
"""Do a three-way test on a scalar.
1241
Return "this", "other" or "conflict", depending whether a value wins.
1243
key_base = key(base_tree, file_id)
1244
key_other = key(other_tree, file_id)
1245
#if base == other, either they all agree, or only THIS has changed.
1246
if key_base == key_other:
1248
key_this = key(this_tree, file_id)
1249
# "Ambiguous clean merge"
1250
if key_this == key_other:
1252
elif key_this == key_base:
1257
def merge_names(self, file_id):
1258
def get_entry(tree):
1259
if tree.has_id(file_id):
1260
return tree.inventory[file_id]
1263
this_entry = get_entry(self.this_tree)
1264
other_entry = get_entry(self.other_tree)
1265
base_entry = get_entry(self.base_tree)
1266
entries = (base_entry, other_entry, this_entry)
1269
for entry in entries:
1272
parents.append(None)
1274
names.append(entry.name)
1275
parents.append(entry.parent_id)
1276
return self._merge_names(file_id, parents, names,
1277
resolver=self._three_way)
1279
def _merge_names(self, file_id, parents, names, resolver):
1280
"""Perform a merge on file_id names and parents"""
1281
base_name, other_name, this_name = names
1282
base_parent, other_parent, this_parent = parents
1284
name_winner = resolver(*names)
1286
parent_id_winner = resolver(*parents)
1287
if this_name is None:
1288
if name_winner == "this":
1289
name_winner = "other"
1290
if parent_id_winner == "this":
1291
parent_id_winner = "other"
1292
if name_winner == "this" and parent_id_winner == "this":
1294
if name_winner == 'conflict' or parent_id_winner == 'conflict':
1295
# Creating helpers (.OTHER or .THIS) here cause problems down the
1296
# road if a ContentConflict needs to be created so we should not do
1298
trans_id = self.tt.trans_id_file_id(file_id)
1299
self._raw_conflicts.append(('path conflict', trans_id, file_id,
1300
this_parent, this_name,
1301
other_parent, other_name))
1302
if other_name is None:
1303
# it doesn't matter whether the result was 'other' or
1304
# 'conflict'-- if there's no 'other', we leave it alone.
1306
parent_id = parents[self.winner_idx[parent_id_winner]]
1307
if parent_id is not None:
1308
# if we get here, name_winner and parent_winner are set to safe
1310
self.tt.adjust_path(names[self.winner_idx[name_winner]],
1311
self.tt.trans_id_file_id(parent_id),
1312
self.tt.trans_id_file_id(file_id))
1314
def _do_merge_contents(self, file_id):
1315
"""Performs a merge on file_id contents."""
1316
def contents_pair(tree):
1317
if file_id not in tree:
1319
kind = tree.kind(file_id)
1321
contents = tree.get_file_sha1(file_id)
1322
elif kind == "symlink":
1323
contents = tree.get_symlink_target(file_id)
1326
return kind, contents
1328
# See SPOT run. run, SPOT, run.
1329
# So we're not QUITE repeating ourselves; we do tricky things with
1331
base_pair = contents_pair(self.base_tree)
1332
other_pair = contents_pair(self.other_tree)
1334
this_pair = contents_pair(self.this_tree)
1335
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1336
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1337
this_pair, allow_overriding_lca=False)
1339
if base_pair == other_pair:
1342
# We delayed evaluating this_pair as long as we can to avoid
1343
# unnecessary sha1 calculation
1344
this_pair = contents_pair(self.this_tree)
1345
winner = self._three_way(base_pair, other_pair, this_pair)
1346
if winner == 'this':
1347
# No interesting changes introduced by OTHER
1349
# We have a hypothetical conflict, but if we have files, then we
1350
# can try to merge the content
1351
trans_id = self.tt.trans_id_file_id(file_id)
1352
params = MergeHookParams(self, file_id, trans_id, this_pair[0],
1353
other_pair[0], winner)
1354
hooks = self.active_hooks
1355
hook_status = 'not_applicable'
1357
hook_status, lines = hook.merge_contents(params)
1358
if hook_status != 'not_applicable':
1359
# Don't try any more hooks, this one applies.
1362
if hook_status == 'not_applicable':
1363
# This is a contents conflict, because none of the available
1364
# functions could merge it.
1366
name = self.tt.final_name(trans_id)
1367
parent_id = self.tt.final_parent(trans_id)
1368
if self.this_tree.has_id(file_id):
1369
self.tt.unversion_file(trans_id)
1370
file_group = self._dump_conflicts(name, parent_id, file_id,
1372
self._raw_conflicts.append(('contents conflict', file_group))
1373
elif hook_status == 'success':
1374
self.tt.create_file(lines, trans_id)
1375
elif hook_status == 'conflicted':
1376
# XXX: perhaps the hook should be able to provide
1377
# the BASE/THIS/OTHER files?
1378
self.tt.create_file(lines, trans_id)
1379
self._raw_conflicts.append(('text conflict', trans_id))
1380
name = self.tt.final_name(trans_id)
1381
parent_id = self.tt.final_parent(trans_id)
1382
self._dump_conflicts(name, parent_id, file_id)
1383
elif hook_status == 'delete':
1384
self.tt.unversion_file(trans_id)
1386
elif hook_status == 'done':
1387
# The hook function did whatever it needs to do directly, no
1388
# further action needed here.
1391
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1392
if not self.this_tree.has_id(file_id) and result == "modified":
1393
self.tt.version_file(file_id, trans_id)
1394
# The merge has been performed, so the old contents should not be
1397
self.tt.delete_contents(trans_id)
1398
except errors.NoSuchFile:
1402
def _default_other_winner_merge(self, merge_hook_params):
1403
"""Replace this contents with other."""
1404
file_id = merge_hook_params.file_id
1405
trans_id = merge_hook_params.trans_id
1406
file_in_this = self.this_tree.has_id(file_id)
1407
if self.other_tree.has_id(file_id):
1408
# OTHER changed the file
1410
if wt.supports_content_filtering():
1411
# We get the path from the working tree if it exists.
1412
# That fails though when OTHER is adding a file, so
1413
# we fall back to the other tree to find the path if
1414
# it doesn't exist locally.
1416
filter_tree_path = wt.id2path(file_id)
1417
except errors.NoSuchId:
1418
filter_tree_path = self.other_tree.id2path(file_id)
1420
# Skip the id2path lookup for older formats
1421
filter_tree_path = None
1422
transform.create_from_tree(self.tt, trans_id,
1423
self.other_tree, file_id,
1424
filter_tree_path=filter_tree_path)
1427
# OTHER deleted the file
1428
return 'delete', None
1430
raise AssertionError(
1431
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1434
def merge_contents(self, merge_hook_params):
1435
"""Fallback merge logic after user installed hooks."""
1436
# This function is used in merge hooks as the fallback instance.
1437
# Perhaps making this function and the functions it calls be a
1438
# a separate class would be better.
1439
if merge_hook_params.winner == 'other':
1440
# OTHER is a straight winner, so replace this contents with other
1441
return self._default_other_winner_merge(merge_hook_params)
1442
elif merge_hook_params.is_file_merge():
1443
# THIS and OTHER are both files, so text merge. Either
1444
# BASE is a file, or both converted to files, so at least we
1445
# have agreement that output should be a file.
1447
self.text_merge(merge_hook_params.file_id,
1448
merge_hook_params.trans_id)
1449
except errors.BinaryFile:
1450
return 'not_applicable', None
1453
return 'not_applicable', None
1455
def get_lines(self, tree, file_id):
1456
"""Return the lines in a file, or an empty list."""
1457
if tree.has_id(file_id):
1458
return tree.get_file(file_id).readlines()
1462
def text_merge(self, file_id, trans_id):
1463
"""Perform a three-way text merge on a file_id"""
1464
# it's possible that we got here with base as a different type.
1465
# if so, we just want two-way text conflicts.
1466
if self.base_tree.has_id(file_id) and \
1467
self.base_tree.kind(file_id) == "file":
1468
base_lines = self.get_lines(self.base_tree, file_id)
1471
other_lines = self.get_lines(self.other_tree, file_id)
1472
this_lines = self.get_lines(self.this_tree, file_id)
1473
m3 = merge3.Merge3(base_lines, this_lines, other_lines,
1474
is_cherrypick=self.cherrypick)
1475
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1476
if self.show_base is True:
1477
base_marker = '|' * 7
1481
def iter_merge3(retval):
1482
retval["text_conflicts"] = False
1483
for line in m3.merge_lines(name_a = "TREE",
1484
name_b = "MERGE-SOURCE",
1485
name_base = "BASE-REVISION",
1486
start_marker=start_marker,
1487
base_marker=base_marker,
1488
reprocess=self.reprocess):
1489
if line.startswith(start_marker):
1490
retval["text_conflicts"] = True
1491
yield line.replace(start_marker, '<' * 7)
1495
merge3_iterator = iter_merge3(retval)
1496
self.tt.create_file(merge3_iterator, trans_id)
1497
if retval["text_conflicts"] is True:
1498
self._raw_conflicts.append(('text conflict', trans_id))
1499
name = self.tt.final_name(trans_id)
1500
parent_id = self.tt.final_parent(trans_id)
1501
file_group = self._dump_conflicts(name, parent_id, file_id,
1502
this_lines, base_lines,
1504
file_group.append(trans_id)
1506
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1507
base_lines=None, other_lines=None, set_version=False,
1509
"""Emit conflict files.
1510
If this_lines, base_lines, or other_lines are omitted, they will be
1511
determined automatically. If set_version is true, the .OTHER, .THIS
1512
or .BASE (in that order) will be created as versioned files.
1514
data = [('OTHER', self.other_tree, other_lines),
1515
('THIS', self.this_tree, this_lines)]
1517
data.append(('BASE', self.base_tree, base_lines))
1519
# We need to use the actual path in the working tree of the file here,
1520
# ignoring the conflict suffixes
1522
if wt.supports_content_filtering():
1524
filter_tree_path = wt.id2path(file_id)
1525
except errors.NoSuchId:
1526
# file has been deleted
1527
filter_tree_path = None
1529
# Skip the id2path lookup for older formats
1530
filter_tree_path = None
1534
for suffix, tree, lines in data:
1535
if tree.has_id(file_id):
1536
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1537
suffix, lines, filter_tree_path)
1538
file_group.append(trans_id)
1539
if set_version and not versioned:
1540
self.tt.version_file(file_id, trans_id)
1544
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1545
lines=None, filter_tree_path=None):
1546
"""Emit a single conflict file."""
1547
name = name + '.' + suffix
1548
trans_id = self.tt.create_path(name, parent_id)
1549
transform.create_from_tree(self.tt, trans_id, tree, file_id, lines,
1553
def merge_executable(self, file_id, file_status):
1554
"""Perform a merge on the execute bit."""
1555
executable = [self.executable(t, file_id) for t in (self.base_tree,
1556
self.other_tree, self.this_tree)]
1557
self._merge_executable(file_id, executable, file_status,
1558
resolver=self._three_way)
1560
def _merge_executable(self, file_id, executable, file_status,
1562
"""Perform a merge on the execute bit."""
1563
base_executable, other_executable, this_executable = executable
1564
if file_status == "deleted":
1566
winner = resolver(*executable)
1567
if winner == "conflict":
1568
# There must be a None in here, if we have a conflict, but we
1569
# need executability since file status was not deleted.
1570
if self.executable(self.other_tree, file_id) is None:
1574
if winner == 'this' and file_status != "modified":
1576
trans_id = self.tt.trans_id_file_id(file_id)
1578
if self.tt.final_kind(trans_id) != "file":
1580
except errors.NoSuchFile:
1582
if winner == "this":
1583
executability = this_executable
1585
if self.other_tree.has_id(file_id):
1586
executability = other_executable
1587
elif self.this_tree.has_id(file_id):
1588
executability = this_executable
1589
elif self.base_tree_has_id(file_id):
1590
executability = base_executable
1591
if executability is not None:
1592
trans_id = self.tt.trans_id_file_id(file_id)
1593
self.tt.set_executability(executability, trans_id)
1595
def cook_conflicts(self, fs_conflicts):
1596
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1597
self.cooked_conflicts.extend(transform.cook_conflicts(
1598
fs_conflicts, self.tt))
1599
fp = transform.FinalPaths(self.tt)
1600
for conflict in self._raw_conflicts:
1601
conflict_type = conflict[0]
1602
if conflict_type == 'path conflict':
1604
this_parent, this_name,
1605
other_parent, other_name) = conflict[1:]
1606
if this_parent is None or this_name is None:
1607
this_path = '<deleted>'
1609
parent_path = fp.get_path(
1610
self.tt.trans_id_file_id(this_parent))
1611
this_path = osutils.pathjoin(parent_path, this_name)
1612
if other_parent is None or other_name is None:
1613
other_path = '<deleted>'
1615
parent_path = fp.get_path(
1616
self.tt.trans_id_file_id(other_parent))
1617
other_path = osutils.pathjoin(parent_path, other_name)
1618
c = _mod_conflicts.Conflict.factory(
1619
'path conflict', path=this_path,
1620
conflict_path=other_path,
1622
elif conflict_type == 'contents conflict':
1623
for trans_id in conflict[1]:
1624
file_id = self.tt.final_file_id(trans_id)
1625
if file_id is not None:
1627
path = fp.get_path(trans_id)
1628
for suffix in ('.BASE', '.THIS', '.OTHER'):
1629
if path.endswith(suffix):
1630
path = path[:-len(suffix)]
1632
c = _mod_conflicts.Conflict.factory(conflict_type,
1633
path=path, file_id=file_id)
1634
elif conflict_type == 'text conflict':
1635
trans_id = conflict[1]
1636
path = fp.get_path(trans_id)
1637
file_id = self.tt.final_file_id(trans_id)
1638
c = _mod_conflicts.Conflict.factory(conflict_type,
1639
path=path, file_id=file_id)
1641
raise AssertionError('bad conflict type: %r' % (conflict,))
1642
self.cooked_conflicts.append(c)
1643
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
1646
class WeaveMerger(Merge3Merger):
1647
"""Three-way tree merger, text weave merger."""
1648
supports_reprocess = True
1649
supports_show_base = False
1650
supports_reverse_cherrypick = False
1651
history_based = True
1653
def _generate_merge_plan(self, file_id, base):
1654
return self.this_tree.plan_file_merge(file_id, self.other_tree,
1657
def _merged_lines(self, file_id):
1658
"""Generate the merged lines.
1659
There is no distinction between lines that are meant to contain <<<<<<<
1663
base = self.base_tree
1666
plan = self._generate_merge_plan(file_id, base)
1667
if 'merge' in debug.debug_flags:
1669
trans_id = self.tt.trans_id_file_id(file_id)
1670
name = self.tt.final_name(trans_id) + '.plan'
1671
contents = ('%11s|%s' % l for l in plan)
1672
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1673
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1674
'>>>>>>> MERGE-SOURCE\n')
1675
lines, conflicts = textmerge.merge_lines(self.reprocess)
1677
base_lines = textmerge.base_from_plan()
1680
return lines, base_lines
1682
def text_merge(self, file_id, trans_id):
1683
"""Perform a (weave) text merge for a given file and file-id.
1684
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1685
and a conflict will be noted.
1687
lines, base_lines = self._merged_lines(file_id)
1689
# Note we're checking whether the OUTPUT is binary in this case,
1690
# because we don't want to get into weave merge guts.
1691
textfile.check_text_lines(lines)
1692
self.tt.create_file(lines, trans_id)
1693
if base_lines is not None:
1695
self._raw_conflicts.append(('text conflict', trans_id))
1696
name = self.tt.final_name(trans_id)
1697
parent_id = self.tt.final_parent(trans_id)
1698
file_group = self._dump_conflicts(name, parent_id, file_id,
1700
base_lines=base_lines)
1701
file_group.append(trans_id)
1704
class LCAMerger(WeaveMerger):
1706
def _generate_merge_plan(self, file_id, base):
1707
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1710
class Diff3Merger(Merge3Merger):
1711
"""Three-way merger using external diff3 for text merging"""
1713
def dump_file(self, temp_dir, name, tree, file_id):
1714
out_path = osutils.pathjoin(temp_dir, name)
1715
out_file = open(out_path, "wb")
1717
in_file = tree.get_file(file_id)
1718
for line in in_file:
1719
out_file.write(line)
1724
def text_merge(self, file_id, trans_id):
1725
"""Perform a diff3 merge using a specified file-id and trans-id.
1726
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1727
will be dumped, and a will be conflict noted.
1730
temp_dir = osutils.mkdtemp(prefix="bzr-")
1732
new_file = osutils.pathjoin(temp_dir, "new")
1733
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1734
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1735
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1736
status = bzrlib.patch.diff3(new_file, this, base, other)
1737
if status not in (0, 1):
1738
raise errors.BzrError("Unhandled diff3 exit code")
1739
f = open(new_file, 'rb')
1741
self.tt.create_file(f, trans_id)
1745
name = self.tt.final_name(trans_id)
1746
parent_id = self.tt.final_parent(trans_id)
1747
self._dump_conflicts(name, parent_id, file_id)
1748
self._raw_conflicts.append(('text conflict', trans_id))
1750
osutils.rmtree(temp_dir)
1753
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1755
merge_type=Merge3Merger,
1756
interesting_ids=None,
1760
interesting_files=None,
1763
change_reporter=None):
1764
"""Primary interface for merging.
1766
typical use is probably
1767
'merge_inner(branch, branch.get_revision_tree(other_revision),
1768
branch.get_revision_tree(base_revision))'
1770
if this_tree is None:
1771
raise errors.BzrError("bzrlib.merge.merge_inner requires a this_tree "
1772
"parameter as of bzrlib version 0.8.")
1773
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1774
pb=pb, change_reporter=change_reporter)
1775
merger.backup_files = backup_files
1776
merger.merge_type = merge_type
1777
merger.interesting_ids = interesting_ids
1778
merger.ignore_zero = ignore_zero
1779
if interesting_files:
1781
raise ValueError('Only supply interesting_ids'
1782
' or interesting_files')
1783
merger.interesting_files = interesting_files
1784
merger.show_base = show_base
1785
merger.reprocess = reprocess
1786
merger.other_rev_id = other_rev_id
1787
merger.other_basis = other_rev_id
1788
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1789
if get_revision_id is None:
1790
get_revision_id = base_tree.last_revision
1791
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
1792
merger.set_base_revision(get_revision_id(), this_branch)
1793
return merger.do_merge()
1795
def get_merge_type_registry():
1796
"""Merge type registry is in bzrlib.option to avoid circular imports.
1798
This method provides a sanctioned way to retrieve it.
1800
from bzrlib import option
1801
return option._merge_type_registry
1804
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1805
def status_a(revision, text):
1806
if revision in ancestors_b:
1807
return 'killed-b', text
1809
return 'new-a', text
1811
def status_b(revision, text):
1812
if revision in ancestors_a:
1813
return 'killed-a', text
1815
return 'new-b', text
1817
plain_a = [t for (a, t) in annotated_a]
1818
plain_b = [t for (a, t) in annotated_b]
1819
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1820
blocks = matcher.get_matching_blocks()
1823
for ai, bi, l in blocks:
1824
# process all mismatched sections
1825
# (last mismatched section is handled because blocks always
1826
# includes a 0-length last block)
1827
for revision, text in annotated_a[a_cur:ai]:
1828
yield status_a(revision, text)
1829
for revision, text in annotated_b[b_cur:bi]:
1830
yield status_b(revision, text)
1831
# and now the matched section
1834
for text_a in plain_a[ai:a_cur]:
1835
yield "unchanged", text_a
1838
class _PlanMergeBase(object):
1840
def __init__(self, a_rev, b_rev, vf, key_prefix):
1843
:param a_rev: Revision-id of one revision to merge
1844
:param b_rev: Revision-id of the other revision to merge
1845
:param vf: A VersionedFiles containing both revisions
1846
:param key_prefix: A prefix for accessing keys in vf, typically
1852
self._last_lines = None
1853
self._last_lines_revision_id = None
1854
self._cached_matching_blocks = {}
1855
self._key_prefix = key_prefix
1856
self._precache_tip_lines()
1858
def _precache_tip_lines(self):
1859
lines = self.get_lines([self.a_rev, self.b_rev])
1860
self.lines_a = lines[self.a_rev]
1861
self.lines_b = lines[self.b_rev]
1863
def get_lines(self, revisions):
1864
"""Get lines for revisions from the backing VersionedFiles.
1866
:raises RevisionNotPresent: on absent texts.
1868
keys = [(self._key_prefix + (rev,)) for rev in revisions]
1870
for record in self.vf.get_record_stream(keys, 'unordered', True):
1871
if record.storage_kind == 'absent':
1872
raise errors.RevisionNotPresent(record.key, self.vf)
1873
result[record.key[-1]] = osutils.chunks_to_lines(
1874
record.get_bytes_as('chunked'))
1877
def plan_merge(self):
1878
"""Generate a 'plan' for merging the two revisions.
1880
This involves comparing their texts and determining the cause of
1881
differences. If text A has a line and text B does not, then either the
1882
line was added to text A, or it was deleted from B. Once the causes
1883
are combined, they are written out in the format described in
1884
VersionedFile.plan_merge
1886
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
1887
unique_a, unique_b = self._unique_lines(blocks)
1888
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
1889
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
1890
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
1892
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
1895
for i, j, n in blocks:
1896
for a_index in range(last_i, i):
1897
if a_index in new_a:
1898
if a_index in killed_b:
1899
yield 'conflicted-a', self.lines_a[a_index]
1901
yield 'new-a', self.lines_a[a_index]
1903
yield 'killed-b', self.lines_a[a_index]
1904
for b_index in range(last_j, j):
1905
if b_index in new_b:
1906
if b_index in killed_a:
1907
yield 'conflicted-b', self.lines_b[b_index]
1909
yield 'new-b', self.lines_b[b_index]
1911
yield 'killed-a', self.lines_b[b_index]
1912
# handle common lines
1913
for a_index in range(i, i+n):
1914
yield 'unchanged', self.lines_a[a_index]
1918
def _get_matching_blocks(self, left_revision, right_revision):
1919
"""Return a description of which sections of two revisions match.
1921
See SequenceMatcher.get_matching_blocks
1923
cached = self._cached_matching_blocks.get((left_revision,
1925
if cached is not None:
1927
if self._last_lines_revision_id == left_revision:
1928
left_lines = self._last_lines
1929
right_lines = self.get_lines([right_revision])[right_revision]
1931
lines = self.get_lines([left_revision, right_revision])
1932
left_lines = lines[left_revision]
1933
right_lines = lines[right_revision]
1934
self._last_lines = right_lines
1935
self._last_lines_revision_id = right_revision
1936
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
1938
return matcher.get_matching_blocks()
1940
def _unique_lines(self, matching_blocks):
1941
"""Analyse matching_blocks to determine which lines are unique
1943
:return: a tuple of (unique_left, unique_right), where the values are
1944
sets of line numbers of unique lines.
1950
for i, j, n in matching_blocks:
1951
unique_left.extend(range(last_i, i))
1952
unique_right.extend(range(last_j, j))
1955
return unique_left, unique_right
1958
def _subtract_plans(old_plan, new_plan):
1959
"""Remove changes from new_plan that came from old_plan.
1961
It is assumed that the difference between the old_plan and new_plan
1962
is their choice of 'b' text.
1964
All lines from new_plan that differ from old_plan are emitted
1965
verbatim. All lines from new_plan that match old_plan but are
1966
not about the 'b' revision are emitted verbatim.
1968
Lines that match and are about the 'b' revision are the lines we
1969
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
1970
is skipped entirely.
1972
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
1975
for i, j, n in matcher.get_matching_blocks():
1976
for jj in range(last_j, j):
1978
for jj in range(j, j+n):
1979
plan_line = new_plan[jj]
1980
if plan_line[0] == 'new-b':
1982
elif plan_line[0] == 'killed-b':
1983
yield 'unchanged', plan_line[1]
1989
class _PlanMerge(_PlanMergeBase):
1990
"""Plan an annotate merge using on-the-fly annotation"""
1992
def __init__(self, a_rev, b_rev, vf, key_prefix):
1993
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
1994
self.a_key = self._key_prefix + (self.a_rev,)
1995
self.b_key = self._key_prefix + (self.b_rev,)
1996
self.graph = _mod_graph.Graph(self.vf)
1997
heads = self.graph.heads((self.a_key, self.b_key))
1999
# one side dominates, so we can just return its values, yay for
2001
# Ideally we would know that before we get this far
2002
self._head_key = heads.pop()
2003
if self._head_key == self.a_key:
2007
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2008
self._head_key[-1], other)
2011
self._head_key = None
2014
def _precache_tip_lines(self):
2015
# Turn this into a no-op, because we will do this later
2018
def _find_recursive_lcas(self):
2019
"""Find all the ancestors back to a unique lca"""
2020
cur_ancestors = (self.a_key, self.b_key)
2021
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2022
# rather than a key tuple. We will just map that directly to no common
2026
next_lcas = self.graph.find_lca(*cur_ancestors)
2027
# Map a plain NULL_REVISION to a simple no-ancestors
2028
if next_lcas == set([_mod_revision.NULL_REVISION]):
2030
# Order the lca's based on when they were merged into the tip
2031
# While the actual merge portion of weave merge uses a set() of
2032
# active revisions, the order of insertion *does* effect the
2033
# implicit ordering of the texts.
2034
for rev_key in cur_ancestors:
2035
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2037
parent_map[rev_key] = ordered_parents
2038
if len(next_lcas) == 0:
2040
elif len(next_lcas) == 1:
2041
parent_map[list(next_lcas)[0]] = ()
2043
elif len(next_lcas) > 2:
2044
# More than 2 lca's, fall back to grabbing all nodes between
2045
# this and the unique lca.
2046
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2048
self.a_key, self.b_key, cur_ancestors)
2049
cur_lcas = next_lcas
2050
while len(cur_lcas) > 1:
2051
cur_lcas = self.graph.find_lca(*cur_lcas)
2052
if len(cur_lcas) == 0:
2053
# No common base to find, use the full ancestry
2056
unique_lca = list(cur_lcas)[0]
2057
if unique_lca == _mod_revision.NULL_REVISION:
2058
# find_lca will return a plain 'NULL_REVISION' rather
2059
# than a key tuple when there is no common ancestor, we
2060
# prefer to just use None, because it doesn't confuse
2061
# _get_interesting_texts()
2063
parent_map.update(self._find_unique_parents(next_lcas,
2066
cur_ancestors = next_lcas
2069
def _find_unique_parents(self, tip_keys, base_key):
2070
"""Find ancestors of tip that aren't ancestors of base.
2072
:param tip_keys: Nodes that are interesting
2073
:param base_key: Cull all ancestors of this node
2074
:return: The parent map for all revisions between tip_keys and
2075
base_key. base_key will be included. References to nodes outside of
2076
the ancestor set will also be removed.
2078
# TODO: this would be simpler if find_unique_ancestors took a list
2079
# instead of a single tip, internally it supports it, but it
2080
# isn't a "backwards compatible" api change.
2081
if base_key is None:
2082
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2083
# We remove NULL_REVISION because it isn't a proper tuple key, and
2084
# thus confuses things like _get_interesting_texts, and our logic
2085
# to add the texts into the memory weave.
2086
if _mod_revision.NULL_REVISION in parent_map:
2087
parent_map.pop(_mod_revision.NULL_REVISION)
2090
for tip in tip_keys:
2092
self.graph.find_unique_ancestors(tip, [base_key]))
2093
parent_map = self.graph.get_parent_map(interesting)
2094
parent_map[base_key] = ()
2095
culled_parent_map, child_map, tails = self._remove_external_references(
2097
# Remove all the tails but base_key
2098
if base_key is not None:
2099
tails.remove(base_key)
2100
self._prune_tails(culled_parent_map, child_map, tails)
2101
# Now remove all the uninteresting 'linear' regions
2102
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2106
def _remove_external_references(parent_map):
2107
"""Remove references that go outside of the parent map.
2109
:param parent_map: Something returned from Graph.get_parent_map(keys)
2110
:return: (filtered_parent_map, child_map, tails)
2111
filtered_parent_map is parent_map without external references
2112
child_map is the {parent_key: [child_keys]} mapping
2113
tails is a list of nodes that do not have any parents in the map
2115
# TODO: The basic effect of this function seems more generic than
2116
# _PlanMerge. But the specific details of building a child_map,
2117
# and computing tails seems very specific to _PlanMerge.
2118
# Still, should this be in Graph land?
2119
filtered_parent_map = {}
2122
for key, parent_keys in parent_map.iteritems():
2123
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2124
if not culled_parent_keys:
2126
for parent_key in culled_parent_keys:
2127
child_map.setdefault(parent_key, []).append(key)
2128
# TODO: Do we want to do this, it adds overhead for every node,
2129
# just to say that the node has no children
2130
child_map.setdefault(key, [])
2131
filtered_parent_map[key] = culled_parent_keys
2132
return filtered_parent_map, child_map, tails
2135
def _prune_tails(parent_map, child_map, tails_to_remove):
2136
"""Remove tails from the parent map.
2138
This will remove the supplied revisions until no more children have 0
2141
:param parent_map: A dict of {child: [parents]}, this dictionary will
2142
be modified in place.
2143
:param tails_to_remove: A list of tips that should be removed,
2144
this list will be consumed
2145
:param child_map: The reverse dict of parent_map ({parent: [children]})
2146
this dict will be modified
2147
:return: None, parent_map will be modified in place.
2149
while tails_to_remove:
2150
next = tails_to_remove.pop()
2151
parent_map.pop(next)
2152
children = child_map.pop(next)
2153
for child in children:
2154
child_parents = parent_map[child]
2155
child_parents.remove(next)
2156
if len(child_parents) == 0:
2157
tails_to_remove.append(child)
2159
def _get_interesting_texts(self, parent_map):
2160
"""Return a dict of texts we are interested in.
2162
Note that the input is in key tuples, but the output is in plain
2165
:param parent_map: The output from _find_recursive_lcas
2166
:return: A dict of {'revision_id':lines} as returned by
2167
_PlanMergeBase.get_lines()
2169
all_revision_keys = set(parent_map)
2170
all_revision_keys.add(self.a_key)
2171
all_revision_keys.add(self.b_key)
2173
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2174
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2177
def _build_weave(self):
2178
from bzrlib import weave
2179
self._weave = weave.Weave(weave_name='in_memory_weave',
2180
allow_reserved=True)
2181
parent_map = self._find_recursive_lcas()
2183
all_texts = self._get_interesting_texts(parent_map)
2185
# Note: Unfortunately, the order given by topo_sort will effect the
2186
# ordering resolution in the output. Specifically, if you add A then B,
2187
# then in the output text A lines will show up before B lines. And, of
2188
# course, topo_sort doesn't guarantee any real ordering.
2189
# So we use merge_sort, and add a fake node on the tip.
2190
# This ensures that left-hand parents will always be inserted into the
2191
# weave before right-hand parents.
2192
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2193
parent_map[tip_key] = (self.a_key, self.b_key)
2195
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2199
# for key in tsort.topo_sort(parent_map):
2200
parent_keys = parent_map[key]
2201
revision_id = key[-1]
2202
parent_ids = [k[-1] for k in parent_keys]
2203
self._weave.add_lines(revision_id, parent_ids,
2204
all_texts[revision_id])
2206
def plan_merge(self):
2207
"""Generate a 'plan' for merging the two revisions.
2209
This involves comparing their texts and determining the cause of
2210
differences. If text A has a line and text B does not, then either the
2211
line was added to text A, or it was deleted from B. Once the causes
2212
are combined, they are written out in the format described in
2213
VersionedFile.plan_merge
2215
if self._head_key is not None: # There was a single head
2216
if self._head_key == self.a_key:
2219
if self._head_key != self.b_key:
2220
raise AssertionError('There was an invalid head: %s != %s'
2221
% (self.b_key, self._head_key))
2223
head_rev = self._head_key[-1]
2224
lines = self.get_lines([head_rev])[head_rev]
2225
return ((plan, line) for line in lines)
2226
return self._weave.plan_merge(self.a_rev, self.b_rev)
2229
class _PlanLCAMerge(_PlanMergeBase):
2231
This merge algorithm differs from _PlanMerge in that:
2232
1. comparisons are done against LCAs only
2233
2. cases where a contested line is new versus one LCA but old versus
2234
another are marked as conflicts, by emitting the line as conflicted-a
2237
This is faster, and hopefully produces more useful output.
2240
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2241
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2242
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2245
if lca == _mod_revision.NULL_REVISION:
2248
self.lcas.add(lca[-1])
2249
for lca in self.lcas:
2250
if _mod_revision.is_null(lca):
2253
lca_lines = self.get_lines([lca])[lca]
2254
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2256
blocks = list(matcher.get_matching_blocks())
2257
self._cached_matching_blocks[(a_rev, lca)] = blocks
2258
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2260
blocks = list(matcher.get_matching_blocks())
2261
self._cached_matching_blocks[(b_rev, lca)] = blocks
2263
def _determine_status(self, revision_id, unique_line_numbers):
2264
"""Determines the status unique lines versus all lcas.
2266
Basically, determines why the line is unique to this revision.
2268
A line may be determined new, killed, or both.
2270
If a line is determined new, that means it was not present in at least
2271
one LCA, and is not present in the other merge revision.
2273
If a line is determined killed, that means the line was present in
2276
If a line is killed and new, this indicates that the two merge
2277
revisions contain differing conflict resolutions.
2278
:param revision_id: The id of the revision in which the lines are
2280
:param unique_line_numbers: The line numbers of unique lines.
2281
:return a tuple of (new_this, killed_other):
2285
unique_line_numbers = set(unique_line_numbers)
2286
for lca in self.lcas:
2287
blocks = self._get_matching_blocks(revision_id, lca)
2288
unique_vs_lca, _ignored = self._unique_lines(blocks)
2289
new.update(unique_line_numbers.intersection(unique_vs_lca))
2290
killed.update(unique_line_numbers.difference(unique_vs_lca))
1
from merge_core import merge_flex
2
from changeset import generate_changeset, ExceptionConflictHandler
3
from changeset import Inventory
4
from bzrlib import Branch
6
from trace import mutter
12
class MergeConflictHandler(ExceptionConflictHandler):
13
"""Handle conflicts encountered while merging"""
14
def copy(self, source, dest):
15
"""Copy the text and mode of a file
16
:param source: The path of the file to copy
17
:param dest: The distination file to create
19
s_file = file(source, "rb")
20
d_file = file(dest, "wb")
23
os.chmod(dest, 0777 & os.stat(source).st_mode)
25
def add_suffix(self, name, suffix, last_new_name=None):
26
"""Rename a file to append a suffix. If the new name exists, the
27
suffix is added repeatedly until a non-existant name is found
29
:param name: The path of the file
30
:param suffix: The suffix to append
31
:param last_new_name: (used for recursive calls) the last name tried
33
if last_new_name is None:
35
new_name = last_new_name+suffix
37
os.rename(name, new_name)
39
if e.errno != errno.EEXIST and e.errno != errno.ENOTEMPTY:
41
self.add_suffix(name, suffix, last_new_name=new_name)
43
def merge_conflict(self, new_file, this_path, base_path, other_path):
45
Handle diff3 conflicts by producing a .THIS, .BASE and .OTHER. The
46
main file will be a version with diff3 conflicts.
47
:param new_file: Path to the output file with diff3 markers
48
:param this_path: Path to the file text for the THIS tree
49
:param base_path: Path to the file text for the BASE tree
50
:param other_path: Path to the file text for the OTHER tree
52
self.add_suffix(this_path, ".THIS")
53
self.copy(base_path, this_path+".BASE")
54
self.copy(other_path, this_path+".OTHER")
55
os.rename(new_file, this_path)
57
def target_exists(self, entry, target, old_path):
58
"""Handle the case when the target file or dir exists"""
59
self.add_suffix(target, ".moved")
62
def __init__(self, path, id, present=None, isdir=None):
65
self.present = present
67
self.interesting = True
70
return "SourceFile(%s, %s)" % (self.path, self.id)
72
def get_tree(treespec, temp_root, label):
76
base_tree = branch.working_tree()
78
base_tree = branch.basis_tree()
80
base_tree = branch.revision_tree(branch.lookup_revision(revno))
81
temp_path = os.path.join(temp_root, label)
83
return MergeTree(base_tree, temp_path)
86
def abspath(tree, file_id):
87
path = tree.inventory.id2path(file_id)
92
def file_exists(tree, file_id):
93
return tree.has_filename(tree.id2path(file_id))
95
def inventory_map(tree):
97
for file_id in tree.inventory:
98
if not file_exists(tree, file_id):
100
path = abspath(tree, file_id)
101
inventory[path] = SourceFile(path, file_id)
105
class MergeTree(object):
106
def __init__(self, tree, tempdir):
107
object.__init__(self)
108
if hasattr(tree, "basedir"):
109
self.root = tree.basedir
112
self.inventory = inventory_map(tree)
114
self.tempdir = tempdir
115
os.mkdir(os.path.join(self.tempdir, "texts"))
118
def readonly_path(self, id):
119
if self.root is not None:
120
return self.tree.abspath(self.tree.id2path(id))
122
if self.tree.inventory[id].kind in ("directory", "root_directory"):
124
if not self.cached.has_key(id):
125
path = os.path.join(self.tempdir, "texts", id)
126
outfile = file(path, "wb")
127
outfile.write(self.tree.get_file(id).read())
128
assert(os.path.exists(path))
129
self.cached[id] = path
130
return self.cached[id]
132
def merge(other_revision, base_revision):
133
tempdir = tempfile.mkdtemp(prefix="bzr-")
135
this_branch = Branch('.')
136
other_tree = get_tree(other_revision, tempdir, "other")
137
base_tree = get_tree(base_revision, tempdir, "base")
138
merge_inner(this_branch, other_tree, base_tree, tempdir)
140
shutil.rmtree(tempdir)
143
def generate_cset_optimized(tree_a, tree_b, inventory_a, inventory_b):
144
"""Generate a changeset, using the text_id to mark really-changed files.
145
This permits blazing comparisons when text_ids are present. It also
146
disables metadata comparison for files with identical texts.
148
for file_id in tree_a.tree.inventory:
149
if file_id not in tree_b.tree.inventory:
151
entry_a = tree_a.tree.inventory[file_id]
152
entry_b = tree_b.tree.inventory[file_id]
153
if (entry_a.kind, entry_b.kind) != ("file", "file"):
155
if None in (entry_a.text_id, entry_b.text_id):
157
if entry_a.text_id != entry_b.text_id:
159
inventory_a[abspath(tree_a.tree, file_id)].interesting = False
160
inventory_b[abspath(tree_b.tree, file_id)].interesting = False
161
cset = generate_changeset(tree_a, tree_b, inventory_a, inventory_b)
162
for entry in cset.entries.itervalues():
163
entry.metadata_change = None
167
def merge_inner(this_branch, other_tree, base_tree, tempdir):
168
this_tree = get_tree(('.', None), tempdir, "this")
170
def get_inventory(tree):
171
return tree.inventory
173
inv_changes = merge_flex(this_tree, base_tree, other_tree,
174
generate_cset_optimized, get_inventory,
175
MergeConflictHandler(base_tree.root))
178
for id, path in inv_changes.iteritems():
183
assert path.startswith('./')
185
adjust_ids.append((path, id))
186
this_branch.set_inventory(regen_inventory(this_branch, this_tree.root, adjust_ids))
189
def regen_inventory(this_branch, root, new_entries):
190
old_entries = this_branch.read_working_inventory()
193
for file_id in old_entries:
194
entry = old_entries[file_id]
195
path = old_entries.id2path(file_id)
196
new_inventory[file_id] = (path, file_id, entry.parent_id, entry.kind)
197
by_path[path] = file_id
202
for path, file_id in new_entries:
204
del new_inventory[file_id]
207
new_path_list.append((path, file_id))
208
if file_id not in old_entries:
210
# Ensure no file is added before its parent
212
for path, file_id in new_path_list:
216
parent = by_path[os.path.dirname(path)]
217
kind = bzrlib.osutils.file_kind(os.path.join(root, path))
218
new_inventory[file_id] = (path, file_id, parent, kind)
219
by_path[path] = file_id
221
# Get a list in insertion order
222
new_inventory_list = new_inventory.values()
223
mutter ("""Inventory regeneration:
224
old length: %i insertions: %i deletions: %i new_length: %i"""\
225
% (len(old_entries), insertions, deletions, len(new_inventory_list)))
226
assert len(new_inventory_list) == len(old_entries) + insertions - deletions
227
new_inventory_list.sort()
228
return new_inventory_list
added
removed
bzrlib/merge.py
