1
# Copyright (C) 2005-2011 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,
24
conflicts as _mod_conflicts,
31
revision as _mod_revision,
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 = cleanup.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, "bzrlib.merge", "Merger.hooks")
66
self.add_hook('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
if self.this_rev_id is None:
462
if self.this_basis_tree.get_file_sha1(file_id) != \
463
self.this_tree.get_file_sha1(file_id):
464
raise errors.WorkingTreeNotRevision(self.this_tree)
466
trees = (self.this_basis_tree, self.other_tree)
467
return [tree.get_file_revision(file_id) for tree in trees]
469
@deprecated_method(deprecated_in((2, 1, 0)))
470
def check_basis(self, check_clean, require_commits=True):
471
if self.this_basis is None and require_commits is True:
472
raise errors.BzrCommandError(
473
"This branch has no commits."
474
" (perhaps you would prefer 'bzr pull')")
477
if self.this_basis != self.this_rev_id:
478
raise errors.UncommittedChanges(self.this_tree)
480
@deprecated_method(deprecated_in((2, 1, 0)))
481
def compare_basis(self):
483
basis_tree = self.revision_tree(self.this_tree.last_revision())
484
except errors.NoSuchRevision:
485
basis_tree = self.this_tree.basis_tree()
486
if not self.this_tree.has_changes(basis_tree):
487
self.this_rev_id = self.this_basis
489
def set_interesting_files(self, file_list):
490
self.interesting_files = file_list
492
def set_pending(self):
493
if (not self.base_is_ancestor or not self.base_is_other_ancestor
494
or self.other_rev_id is None):
498
def _add_parent(self):
499
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
500
new_parent_trees = []
501
operation = cleanup.OperationWithCleanups(
502
self.this_tree.set_parent_trees)
503
for revision_id in new_parents:
505
tree = self.revision_tree(revision_id)
506
except errors.NoSuchRevision:
510
operation.add_cleanup(tree.unlock)
511
new_parent_trees.append((revision_id, tree))
512
operation.run_simple(new_parent_trees, allow_leftmost_as_ghost=True)
514
def set_other(self, other_revision, possible_transports=None):
515
"""Set the revision and tree to merge from.
517
This sets the other_tree, other_rev_id, other_basis attributes.
519
:param other_revision: The [path, revision] list to merge from.
521
self.other_branch, self.other_tree = self._get_tree(other_revision,
523
if other_revision[1] == -1:
524
self.other_rev_id = _mod_revision.ensure_null(
525
self.other_branch.last_revision())
526
if _mod_revision.is_null(self.other_rev_id):
527
raise errors.NoCommits(self.other_branch)
528
self.other_basis = self.other_rev_id
529
elif other_revision[1] is not None:
530
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
531
self.other_basis = self.other_rev_id
533
self.other_rev_id = None
534
self.other_basis = self.other_branch.last_revision()
535
if self.other_basis is None:
536
raise errors.NoCommits(self.other_branch)
537
if self.other_rev_id is not None:
538
self._cached_trees[self.other_rev_id] = self.other_tree
539
self._maybe_fetch(self.other_branch,self.this_branch, self.other_basis)
541
def set_other_revision(self, revision_id, other_branch):
542
"""Set 'other' based on a branch and revision id
544
:param revision_id: The revision to use for a tree
545
:param other_branch: The branch containing this tree
547
self.other_rev_id = revision_id
548
self.other_branch = other_branch
549
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
550
self.other_tree = self.revision_tree(revision_id)
551
self.other_basis = revision_id
553
def set_base_revision(self, revision_id, branch):
554
"""Set 'base' based on a branch and revision id
556
:param revision_id: The revision to use for a tree
557
:param branch: The branch containing this tree
559
self.base_rev_id = revision_id
560
self.base_branch = branch
561
self._maybe_fetch(branch, self.this_branch, revision_id)
562
self.base_tree = self.revision_tree(revision_id)
564
def _maybe_fetch(self, source, target, revision_id):
565
if not source.repository.has_same_location(target.repository):
566
target.fetch(source, revision_id)
569
revisions = [_mod_revision.ensure_null(self.this_basis),
570
_mod_revision.ensure_null(self.other_basis)]
571
if _mod_revision.NULL_REVISION in revisions:
572
self.base_rev_id = _mod_revision.NULL_REVISION
573
self.base_tree = self.revision_tree(self.base_rev_id)
574
self._is_criss_cross = False
576
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
577
self._is_criss_cross = False
579
self.base_rev_id = _mod_revision.NULL_REVISION
581
self.base_rev_id = list(lcas)[0]
582
else: # len(lcas) > 1
583
self._is_criss_cross = True
585
# find_unique_lca can only handle 2 nodes, so we have to
586
# start back at the beginning. It is a shame to traverse
587
# the graph again, but better than re-implementing
589
self.base_rev_id = self.revision_graph.find_unique_lca(
590
revisions[0], revisions[1])
592
self.base_rev_id = self.revision_graph.find_unique_lca(
594
sorted_lca_keys = self.revision_graph.find_merge_order(
596
if self.base_rev_id == _mod_revision.NULL_REVISION:
597
self.base_rev_id = sorted_lca_keys[0]
599
if self.base_rev_id == _mod_revision.NULL_REVISION:
600
raise errors.UnrelatedBranches()
601
if self._is_criss_cross:
602
trace.warning('Warning: criss-cross merge encountered. See bzr'
603
' help criss-cross.')
604
trace.mutter('Criss-cross lcas: %r' % lcas)
605
if self.base_rev_id in lcas:
606
trace.mutter('Unable to find unique lca. '
607
'Fallback %r as best option.' % self.base_rev_id)
608
interesting_revision_ids = set(lcas)
609
interesting_revision_ids.add(self.base_rev_id)
610
interesting_trees = dict((t.get_revision_id(), t)
611
for t in self.this_branch.repository.revision_trees(
612
interesting_revision_ids))
613
self._cached_trees.update(interesting_trees)
614
if self.base_rev_id in lcas:
615
self.base_tree = interesting_trees[self.base_rev_id]
617
self.base_tree = interesting_trees.pop(self.base_rev_id)
618
self._lca_trees = [interesting_trees[key]
619
for key in sorted_lca_keys]
621
self.base_tree = self.revision_tree(self.base_rev_id)
622
self.base_is_ancestor = True
623
self.base_is_other_ancestor = True
624
trace.mutter('Base revid: %r' % self.base_rev_id)
626
def set_base(self, base_revision):
627
"""Set the base revision to use for the merge.
629
:param base_revision: A 2-list containing a path and revision number.
631
trace.mutter("doing merge() with no base_revision specified")
632
if base_revision == [None, None]:
635
base_branch, self.base_tree = self._get_tree(base_revision)
636
if base_revision[1] == -1:
637
self.base_rev_id = base_branch.last_revision()
638
elif base_revision[1] is None:
639
self.base_rev_id = _mod_revision.NULL_REVISION
641
self.base_rev_id = _mod_revision.ensure_null(
642
base_branch.get_rev_id(base_revision[1]))
643
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
645
def make_merger(self):
646
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
647
'other_tree': self.other_tree,
648
'interesting_ids': self.interesting_ids,
649
'interesting_files': self.interesting_files,
650
'this_branch': self.this_branch,
652
if self.merge_type.requires_base:
653
kwargs['base_tree'] = self.base_tree
654
if self.merge_type.supports_reprocess:
655
kwargs['reprocess'] = self.reprocess
657
raise errors.BzrError(
658
"Conflict reduction is not supported for merge"
659
" type %s." % self.merge_type)
660
if self.merge_type.supports_show_base:
661
kwargs['show_base'] = self.show_base
663
raise errors.BzrError("Showing base is not supported for this"
664
" merge type. %s" % self.merge_type)
665
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
666
and not self.base_is_other_ancestor):
667
raise errors.CannotReverseCherrypick()
668
if self.merge_type.supports_cherrypick:
669
kwargs['cherrypick'] = (not self.base_is_ancestor or
670
not self.base_is_other_ancestor)
671
if self._is_criss_cross and getattr(self.merge_type,
672
'supports_lca_trees', False):
673
kwargs['lca_trees'] = self._lca_trees
674
return self.merge_type(pb=None,
675
change_reporter=self.change_reporter,
678
def _do_merge_to(self):
679
merge = self.make_merger()
680
if self.other_branch is not None:
681
self.other_branch.update_references(self.this_branch)
683
if self.recurse == 'down':
684
for relpath, file_id in self.this_tree.iter_references():
685
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
686
other_revision = self.other_tree.get_reference_revision(
688
if other_revision == sub_tree.last_revision():
690
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
691
sub_merge.merge_type = self.merge_type
692
other_branch = self.other_branch.reference_parent(file_id, relpath)
693
sub_merge.set_other_revision(other_revision, other_branch)
694
base_revision = self.base_tree.get_reference_revision(file_id)
695
sub_merge.base_tree = \
696
sub_tree.branch.repository.revision_tree(base_revision)
697
sub_merge.base_rev_id = base_revision
702
operation = cleanup.OperationWithCleanups(self._do_merge_to)
703
self.this_tree.lock_tree_write()
704
operation.add_cleanup(self.this_tree.unlock)
705
if self.base_tree is not None:
706
self.base_tree.lock_read()
707
operation.add_cleanup(self.base_tree.unlock)
708
if self.other_tree is not None:
709
self.other_tree.lock_read()
710
operation.add_cleanup(self.other_tree.unlock)
711
merge = operation.run_simple()
712
if len(merge.cooked_conflicts) == 0:
713
if not self.ignore_zero and not trace.is_quiet():
714
trace.note("All changes applied successfully.")
716
trace.note("%d conflicts encountered."
717
% len(merge.cooked_conflicts))
719
return len(merge.cooked_conflicts)
722
class _InventoryNoneEntry(object):
723
"""This represents an inventory entry which *isn't there*.
725
It simplifies the merging logic if we always have an InventoryEntry, even
726
if it isn't actually present
733
symlink_target = None
736
_none_entry = _InventoryNoneEntry()
739
class Merge3Merger(object):
740
"""Three-way merger that uses the merge3 text merger"""
742
supports_reprocess = True
743
supports_show_base = True
744
history_based = False
745
supports_cherrypick = True
746
supports_reverse_cherrypick = True
747
winner_idx = {"this": 2, "other": 1, "conflict": 1}
748
supports_lca_trees = True
750
def __init__(self, working_tree, this_tree, base_tree, other_tree,
751
interesting_ids=None, reprocess=False, show_base=False,
752
pb=None, pp=None, change_reporter=None,
753
interesting_files=None, do_merge=True,
754
cherrypick=False, lca_trees=None, this_branch=None):
755
"""Initialize the merger object and perform the merge.
757
:param working_tree: The working tree to apply the merge to
758
:param this_tree: The local tree in the merge operation
759
:param base_tree: The common tree in the merge operation
760
:param other_tree: The other tree to merge changes from
761
:param this_branch: The branch associated with this_tree. Defaults to
762
this_tree.branch if not supplied.
763
:param interesting_ids: The file_ids of files that should be
764
participate in the merge. May not be combined with
766
:param: reprocess If True, perform conflict-reduction processing.
767
:param show_base: If True, show the base revision in text conflicts.
768
(incompatible with reprocess)
770
:param pp: A ProgressPhase object
771
:param change_reporter: An object that should report changes made
772
:param interesting_files: The tree-relative paths of files that should
773
participate in the merge. If these paths refer to directories,
774
the contents of those directories will also be included. May not
775
be combined with interesting_ids. If neither interesting_files nor
776
interesting_ids is specified, all files may participate in the
778
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
779
if the ancestry was found to include a criss-cross merge.
780
Otherwise should be None.
782
object.__init__(self)
783
if interesting_files is not None and interesting_ids is not None:
785
'specify either interesting_ids or interesting_files')
786
if this_branch is None:
787
this_branch = this_tree.branch
788
self.interesting_ids = interesting_ids
789
self.interesting_files = interesting_files
790
self.this_tree = working_tree
791
self.base_tree = base_tree
792
self.other_tree = other_tree
793
self.this_branch = this_branch
794
self._raw_conflicts = []
795
self.cooked_conflicts = []
796
self.reprocess = reprocess
797
self.show_base = show_base
798
self._lca_trees = lca_trees
799
# Uncommenting this will change the default algorithm to always use
800
# _entries_lca. This can be useful for running the test suite and
801
# making sure we haven't missed any corner cases.
802
# if lca_trees is None:
803
# self._lca_trees = [self.base_tree]
804
self.change_reporter = change_reporter
805
self.cherrypick = cherrypick
809
warnings.warn("pp argument to Merge3Merger is deprecated")
811
warnings.warn("pb argument to Merge3Merger is deprecated")
814
operation = cleanup.OperationWithCleanups(self._do_merge)
815
self.this_tree.lock_tree_write()
816
operation.add_cleanup(self.this_tree.unlock)
817
self.base_tree.lock_read()
818
operation.add_cleanup(self.base_tree.unlock)
819
self.other_tree.lock_read()
820
operation.add_cleanup(self.other_tree.unlock)
823
def _do_merge(self, operation):
824
self.tt = transform.TreeTransform(self.this_tree, None)
825
operation.add_cleanup(self.tt.finalize)
826
self._compute_transform()
827
results = self.tt.apply(no_conflicts=True)
828
self.write_modified(results)
830
self.this_tree.add_conflicts(self.cooked_conflicts)
831
except errors.UnsupportedOperation:
834
def make_preview_transform(self):
835
operation = cleanup.OperationWithCleanups(self._make_preview_transform)
836
self.base_tree.lock_read()
837
operation.add_cleanup(self.base_tree.unlock)
838
self.other_tree.lock_read()
839
operation.add_cleanup(self.other_tree.unlock)
840
return operation.run_simple()
842
def _make_preview_transform(self):
843
self.tt = transform.TransformPreview(self.this_tree)
844
self._compute_transform()
847
def _compute_transform(self):
848
if self._lca_trees is None:
849
entries = self._entries3()
850
resolver = self._three_way
852
entries = self._entries_lca()
853
resolver = self._lca_multi_way
854
child_pb = ui.ui_factory.nested_progress_bar()
856
factories = Merger.hooks['merge_file_content']
857
hooks = [factory(self) for factory in factories] + [self]
858
self.active_hooks = [hook for hook in hooks if hook is not None]
859
for num, (file_id, changed, parents3, names3,
860
executable3) in enumerate(entries):
861
child_pb.update('Preparing file merge', num, len(entries))
862
self._merge_names(file_id, parents3, names3, resolver=resolver)
864
file_status = self._do_merge_contents(file_id)
866
file_status = 'unmodified'
867
self._merge_executable(file_id,
868
executable3, file_status, resolver=resolver)
872
self._finish_computing_transform()
874
def _finish_computing_transform(self):
875
"""Finalize the transform and report the changes.
877
This is the second half of _compute_transform.
879
child_pb = ui.ui_factory.nested_progress_bar()
881
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
882
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
885
if self.change_reporter is not None:
886
from bzrlib import delta
887
delta.report_changes(
888
self.tt.iter_changes(), self.change_reporter)
889
self.cook_conflicts(fs_conflicts)
890
for conflict in self.cooked_conflicts:
891
trace.warning(conflict)
894
"""Gather data about files modified between three trees.
896
Return a list of tuples of file_id, changed, parents3, names3,
897
executable3. changed is a boolean indicating whether the file contents
898
or kind were changed. parents3 is a tuple of parent ids for base,
899
other and this. names3 is a tuple of names for base, other and this.
900
executable3 is a tuple of execute-bit values for base, other and this.
903
iterator = self.other_tree.iter_changes(self.base_tree,
904
specific_files=self.interesting_files,
905
extra_trees=[self.this_tree])
906
this_entries = dict((e.file_id, e) for p, e in
907
self.this_tree.iter_entries_by_dir(
908
self.interesting_ids))
909
for (file_id, paths, changed, versioned, parents, names, kind,
910
executable) in iterator:
911
if (self.interesting_ids is not None and
912
file_id not in self.interesting_ids):
914
entry = this_entries.get(file_id)
915
if entry is not None:
916
this_name = entry.name
917
this_parent = entry.parent_id
918
this_executable = entry.executable
922
this_executable = None
923
parents3 = parents + (this_parent,)
924
names3 = names + (this_name,)
925
executable3 = executable + (this_executable,)
926
result.append((file_id, changed, parents3, names3, executable3))
929
def _entries_lca(self):
930
"""Gather data about files modified between multiple trees.
932
This compares OTHER versus all LCA trees, and for interesting entries,
933
it then compares with THIS and BASE.
935
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
936
:return: [(file_id, changed, parents, names, executable)]
937
file_id Simple file_id of the entry
938
changed Boolean, True if the kind or contents changed
940
parents ((base, [parent_id, in, lcas]), parent_id_other,
942
names ((base, [name, in, lcas]), name_in_other, name_in_this)
943
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
945
if self.interesting_files is not None:
946
lookup_trees = [self.this_tree, self.base_tree]
947
lookup_trees.extend(self._lca_trees)
948
# I think we should include the lca trees as well
949
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
952
interesting_ids = self.interesting_ids
954
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
956
base_inventory = self.base_tree.inventory
957
this_inventory = self.this_tree.inventory
958
for path, file_id, other_ie, lca_values in walker.iter_all():
959
# Is this modified at all from any of the other trees?
961
other_ie = _none_entry
962
if interesting_ids is not None and file_id not in interesting_ids:
965
# If other_revision is found in any of the lcas, that means this
966
# node is uninteresting. This is because when merging, if there are
967
# multiple heads(), we have to create a new node. So if we didn't,
968
# we know that the ancestry is linear, and that OTHER did not
970
# See doc/developers/lca_merge_resolution.txt for details
971
other_revision = other_ie.revision
972
if other_revision is not None:
973
# We can't use this shortcut when other_revision is None,
974
# because it may be None because things are WorkingTrees, and
975
# not because it is *actually* None.
976
is_unmodified = False
977
for lca_path, ie in lca_values:
978
if ie is not None and ie.revision == other_revision:
985
for lca_path, lca_ie in lca_values:
987
lca_entries.append(_none_entry)
989
lca_entries.append(lca_ie)
991
if file_id in base_inventory:
992
base_ie = base_inventory[file_id]
994
base_ie = _none_entry
996
if file_id in this_inventory:
997
this_ie = this_inventory[file_id]
999
this_ie = _none_entry
1005
for lca_ie in lca_entries:
1006
lca_kinds.append(lca_ie.kind)
1007
lca_parent_ids.append(lca_ie.parent_id)
1008
lca_names.append(lca_ie.name)
1009
lca_executable.append(lca_ie.executable)
1011
kind_winner = self._lca_multi_way(
1012
(base_ie.kind, lca_kinds),
1013
other_ie.kind, this_ie.kind)
1014
parent_id_winner = self._lca_multi_way(
1015
(base_ie.parent_id, lca_parent_ids),
1016
other_ie.parent_id, this_ie.parent_id)
1017
name_winner = self._lca_multi_way(
1018
(base_ie.name, lca_names),
1019
other_ie.name, this_ie.name)
1021
content_changed = True
1022
if kind_winner == 'this':
1023
# No kind change in OTHER, see if there are *any* changes
1024
if other_ie.kind == 'directory':
1025
if parent_id_winner == 'this' and name_winner == 'this':
1026
# No change for this directory in OTHER, skip
1028
content_changed = False
1029
elif other_ie.kind is None or other_ie.kind == 'file':
1030
def get_sha1(ie, tree):
1031
if ie.kind != 'file':
1033
return tree.get_file_sha1(file_id)
1034
base_sha1 = get_sha1(base_ie, self.base_tree)
1035
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1036
in zip(lca_entries, self._lca_trees)]
1037
this_sha1 = get_sha1(this_ie, self.this_tree)
1038
other_sha1 = get_sha1(other_ie, self.other_tree)
1039
sha1_winner = self._lca_multi_way(
1040
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1041
allow_overriding_lca=False)
1042
exec_winner = self._lca_multi_way(
1043
(base_ie.executable, lca_executable),
1044
other_ie.executable, this_ie.executable)
1045
if (parent_id_winner == 'this' and name_winner == 'this'
1046
and sha1_winner == 'this' and exec_winner == 'this'):
1047
# No kind, parent, name, exec, or content change for
1048
# OTHER, so this node is not considered interesting
1050
if sha1_winner == 'this':
1051
content_changed = False
1052
elif other_ie.kind == 'symlink':
1053
def get_target(ie, tree):
1054
if ie.kind != 'symlink':
1056
return tree.get_symlink_target(file_id)
1057
base_target = get_target(base_ie, self.base_tree)
1058
lca_targets = [get_target(ie, tree) for ie, tree
1059
in zip(lca_entries, self._lca_trees)]
1060
this_target = get_target(this_ie, self.this_tree)
1061
other_target = get_target(other_ie, self.other_tree)
1062
target_winner = self._lca_multi_way(
1063
(base_target, lca_targets),
1064
other_target, this_target)
1065
if (parent_id_winner == 'this' and name_winner == 'this'
1066
and target_winner == 'this'):
1067
# No kind, parent, name, or symlink target change
1070
if target_winner == 'this':
1071
content_changed = False
1072
elif other_ie.kind == 'tree-reference':
1073
# The 'changed' information seems to be handled at a higher
1074
# level. At least, _entries3 returns False for content
1075
# changed, even when at a new revision_id.
1076
content_changed = False
1077
if (parent_id_winner == 'this' and name_winner == 'this'):
1078
# Nothing interesting
1081
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1083
# If we have gotten this far, that means something has changed
1084
result.append((file_id, content_changed,
1085
((base_ie.parent_id, lca_parent_ids),
1086
other_ie.parent_id, this_ie.parent_id),
1087
((base_ie.name, lca_names),
1088
other_ie.name, this_ie.name),
1089
((base_ie.executable, lca_executable),
1090
other_ie.executable, this_ie.executable)
1095
if self.tt.final_kind(self.tt.root) is None:
1096
self.tt.cancel_deletion(self.tt.root)
1097
if self.tt.final_file_id(self.tt.root) is None:
1098
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
1100
other_root_file_id = self.other_tree.get_root_id()
1101
if other_root_file_id is None:
1103
other_root = self.tt.trans_id_file_id(other_root_file_id)
1104
if other_root == self.tt.root:
1106
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
1107
# the other tree's root is a non-root in the current tree (as when
1108
# a previously unrelated branch is merged into another)
1110
if self.tt.final_kind(other_root) is not None:
1111
other_root_is_present = True
1113
# other_root doesn't have a physical representation. We still need
1114
# to move any references to the actual root of the tree.
1115
other_root_is_present = False
1116
# 'other_tree.inventory.root' is not present in this tree. We are
1117
# calling adjust_path for children which *want* to be present with a
1118
# correct place to go.
1119
for _, child in self.other_tree.inventory.root.children.iteritems():
1120
trans_id = self.tt.trans_id_file_id(child.file_id)
1121
if not other_root_is_present:
1122
if self.tt.final_kind(trans_id) is not None:
1123
# The item exist in the final tree and has a defined place
1126
# Move the item into the root
1128
final_name = self.tt.final_name(trans_id)
1129
except errors.NoFinalPath:
1130
# This file is not present anymore, ignore it.
1132
self.tt.adjust_path(final_name, self.tt.root, trans_id)
1133
if other_root_is_present:
1134
self.tt.cancel_creation(other_root)
1135
self.tt.cancel_versioning(other_root)
1137
def write_modified(self, results):
1138
modified_hashes = {}
1139
for path in results.modified_paths:
1140
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
1143
hash = self.this_tree.get_file_sha1(file_id)
1146
modified_hashes[file_id] = hash
1147
self.this_tree.set_merge_modified(modified_hashes)
1150
def parent(entry, file_id):
1151
"""Determine the parent for a file_id (used as a key method)"""
1154
return entry.parent_id
1157
def name(entry, file_id):
1158
"""Determine the name for a file_id (used as a key method)"""
1164
def contents_sha1(tree, file_id):
1165
"""Determine the sha1 of the file contents (used as a key method)."""
1166
if file_id not in tree:
1168
return tree.get_file_sha1(file_id)
1171
def executable(tree, file_id):
1172
"""Determine the executability of a file-id (used as a key method)."""
1173
if not tree.has_id(file_id):
1175
if tree.kind(file_id) != "file":
1177
return tree.is_executable(file_id)
1180
def kind(tree, file_id):
1181
"""Determine the kind of a file-id (used as a key method)."""
1182
if not tree.has_id(file_id):
1184
return tree.kind(file_id)
1187
def _three_way(base, other, this):
1189
# if 'base == other', either they all agree, or only 'this' has
1192
elif this not in (base, other):
1193
# 'this' is neither 'base' nor 'other', so both sides changed
1196
# "Ambiguous clean merge" -- both sides have made the same change.
1199
# this == base: only other has changed.
1203
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
1204
"""Consider LCAs when determining whether a change has occurred.
1206
If LCAS are all identical, this is the same as a _three_way comparison.
1208
:param bases: value in (BASE, [LCAS])
1209
:param other: value in OTHER
1210
:param this: value in THIS
1211
:param allow_overriding_lca: If there is more than one unique lca
1212
value, allow OTHER to override THIS if it has a new value, and
1213
THIS only has an lca value, or vice versa. This is appropriate for
1214
truly scalar values, not as much for non-scalars.
1215
:return: 'this', 'other', or 'conflict' depending on whether an entry
1218
# See doc/developers/lca_tree_merging.txt for details about this
1221
# Either Ambiguously clean, or nothing was actually changed. We
1224
base_val, lca_vals = bases
1225
# Remove 'base_val' from the lca_vals, because it is not interesting
1226
filtered_lca_vals = [lca_val for lca_val in lca_vals
1227
if lca_val != base_val]
1228
if len(filtered_lca_vals) == 0:
1229
return Merge3Merger._three_way(base_val, other, this)
1231
unique_lca_vals = set(filtered_lca_vals)
1232
if len(unique_lca_vals) == 1:
1233
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1235
if allow_overriding_lca:
1236
if other in unique_lca_vals:
1237
if this in unique_lca_vals:
1238
# Each side picked a different lca, conflict
1241
# This has a value which supersedes both lca values, and
1242
# other only has an lca value
1244
elif this in unique_lca_vals:
1245
# OTHER has a value which supersedes both lca values, and this
1246
# only has an lca value
1249
# At this point, the lcas disagree, and the tip disagree
1253
@deprecated_method(deprecated_in((2, 2, 0)))
1254
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1255
"""Do a three-way test on a scalar.
1256
Return "this", "other" or "conflict", depending whether a value wins.
1258
key_base = key(base_tree, file_id)
1259
key_other = key(other_tree, file_id)
1260
#if base == other, either they all agree, or only THIS has changed.
1261
if key_base == key_other:
1263
key_this = key(this_tree, file_id)
1264
# "Ambiguous clean merge"
1265
if key_this == key_other:
1267
elif key_this == key_base:
1272
def merge_names(self, file_id):
1273
def get_entry(tree):
1274
if tree.has_id(file_id):
1275
return tree.inventory[file_id]
1278
this_entry = get_entry(self.this_tree)
1279
other_entry = get_entry(self.other_tree)
1280
base_entry = get_entry(self.base_tree)
1281
entries = (base_entry, other_entry, this_entry)
1284
for entry in entries:
1287
parents.append(None)
1289
names.append(entry.name)
1290
parents.append(entry.parent_id)
1291
return self._merge_names(file_id, parents, names,
1292
resolver=self._three_way)
1294
def _merge_names(self, file_id, parents, names, resolver):
1295
"""Perform a merge on file_id names and parents"""
1296
base_name, other_name, this_name = names
1297
base_parent, other_parent, this_parent = parents
1299
name_winner = resolver(*names)
1301
parent_id_winner = resolver(*parents)
1302
if this_name is None:
1303
if name_winner == "this":
1304
name_winner = "other"
1305
if parent_id_winner == "this":
1306
parent_id_winner = "other"
1307
if name_winner == "this" and parent_id_winner == "this":
1309
if name_winner == 'conflict' or parent_id_winner == 'conflict':
1310
# Creating helpers (.OTHER or .THIS) here cause problems down the
1311
# road if a ContentConflict needs to be created so we should not do
1313
trans_id = self.tt.trans_id_file_id(file_id)
1314
self._raw_conflicts.append(('path conflict', trans_id, file_id,
1315
this_parent, this_name,
1316
other_parent, other_name))
1317
if other_name is None:
1318
# it doesn't matter whether the result was 'other' or
1319
# 'conflict'-- if there's no 'other', we leave it alone.
1321
parent_id = parents[self.winner_idx[parent_id_winner]]
1322
if parent_id is not None:
1323
# if we get here, name_winner and parent_winner are set to safe
1325
self.tt.adjust_path(names[self.winner_idx[name_winner]],
1326
self.tt.trans_id_file_id(parent_id),
1327
self.tt.trans_id_file_id(file_id))
1329
def _do_merge_contents(self, file_id):
1330
"""Performs a merge on file_id contents."""
1331
def contents_pair(tree):
1332
if file_id not in tree:
1334
kind = tree.kind(file_id)
1336
contents = tree.get_file_sha1(file_id)
1337
elif kind == "symlink":
1338
contents = tree.get_symlink_target(file_id)
1341
return kind, contents
1343
# See SPOT run. run, SPOT, run.
1344
# So we're not QUITE repeating ourselves; we do tricky things with
1346
base_pair = contents_pair(self.base_tree)
1347
other_pair = contents_pair(self.other_tree)
1349
this_pair = contents_pair(self.this_tree)
1350
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1351
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1352
this_pair, allow_overriding_lca=False)
1354
if base_pair == other_pair:
1357
# We delayed evaluating this_pair as long as we can to avoid
1358
# unnecessary sha1 calculation
1359
this_pair = contents_pair(self.this_tree)
1360
winner = self._three_way(base_pair, other_pair, this_pair)
1361
if winner == 'this':
1362
# No interesting changes introduced by OTHER
1364
# We have a hypothetical conflict, but if we have files, then we
1365
# can try to merge the content
1366
trans_id = self.tt.trans_id_file_id(file_id)
1367
params = MergeHookParams(self, file_id, trans_id, this_pair[0],
1368
other_pair[0], winner)
1369
hooks = self.active_hooks
1370
hook_status = 'not_applicable'
1372
hook_status, lines = hook.merge_contents(params)
1373
if hook_status != 'not_applicable':
1374
# Don't try any more hooks, this one applies.
1377
if hook_status == 'not_applicable':
1378
# This is a contents conflict, because none of the available
1379
# functions could merge it.
1381
name = self.tt.final_name(trans_id)
1382
parent_id = self.tt.final_parent(trans_id)
1383
if self.this_tree.has_id(file_id):
1384
self.tt.unversion_file(trans_id)
1385
file_group = self._dump_conflicts(name, parent_id, file_id,
1387
self._raw_conflicts.append(('contents conflict', file_group))
1388
elif hook_status == 'success':
1389
self.tt.create_file(lines, trans_id)
1390
elif hook_status == 'conflicted':
1391
# XXX: perhaps the hook should be able to provide
1392
# the BASE/THIS/OTHER files?
1393
self.tt.create_file(lines, trans_id)
1394
self._raw_conflicts.append(('text conflict', trans_id))
1395
name = self.tt.final_name(trans_id)
1396
parent_id = self.tt.final_parent(trans_id)
1397
self._dump_conflicts(name, parent_id, file_id)
1398
elif hook_status == 'delete':
1399
self.tt.unversion_file(trans_id)
1401
elif hook_status == 'done':
1402
# The hook function did whatever it needs to do directly, no
1403
# further action needed here.
1406
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1407
if not self.this_tree.has_id(file_id) and result == "modified":
1408
self.tt.version_file(file_id, trans_id)
1409
# The merge has been performed, so the old contents should not be
1411
self.tt.delete_contents(trans_id)
1414
def _default_other_winner_merge(self, merge_hook_params):
1415
"""Replace this contents with other."""
1416
file_id = merge_hook_params.file_id
1417
trans_id = merge_hook_params.trans_id
1418
file_in_this = self.this_tree.has_id(file_id)
1419
if self.other_tree.has_id(file_id):
1420
# OTHER changed the file
1422
if wt.supports_content_filtering():
1423
# We get the path from the working tree if it exists.
1424
# That fails though when OTHER is adding a file, so
1425
# we fall back to the other tree to find the path if
1426
# it doesn't exist locally.
1428
filter_tree_path = wt.id2path(file_id)
1429
except errors.NoSuchId:
1430
filter_tree_path = self.other_tree.id2path(file_id)
1432
# Skip the id2path lookup for older formats
1433
filter_tree_path = None
1434
transform.create_from_tree(self.tt, trans_id,
1435
self.other_tree, file_id,
1436
filter_tree_path=filter_tree_path)
1439
# OTHER deleted the file
1440
return 'delete', None
1442
raise AssertionError(
1443
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1446
def merge_contents(self, merge_hook_params):
1447
"""Fallback merge logic after user installed hooks."""
1448
# This function is used in merge hooks as the fallback instance.
1449
# Perhaps making this function and the functions it calls be a
1450
# a separate class would be better.
1451
if merge_hook_params.winner == 'other':
1452
# OTHER is a straight winner, so replace this contents with other
1453
return self._default_other_winner_merge(merge_hook_params)
1454
elif merge_hook_params.is_file_merge():
1455
# THIS and OTHER are both files, so text merge. Either
1456
# BASE is a file, or both converted to files, so at least we
1457
# have agreement that output should be a file.
1459
self.text_merge(merge_hook_params.file_id,
1460
merge_hook_params.trans_id)
1461
except errors.BinaryFile:
1462
return 'not_applicable', None
1465
return 'not_applicable', None
1467
def get_lines(self, tree, file_id):
1468
"""Return the lines in a file, or an empty list."""
1469
if tree.has_id(file_id):
1470
return tree.get_file_lines(file_id)
1474
def text_merge(self, file_id, trans_id):
1475
"""Perform a three-way text merge on a file_id"""
1476
# it's possible that we got here with base as a different type.
1477
# if so, we just want two-way text conflicts.
1478
if self.base_tree.has_id(file_id) and \
1479
self.base_tree.kind(file_id) == "file":
1480
base_lines = self.get_lines(self.base_tree, file_id)
1483
other_lines = self.get_lines(self.other_tree, file_id)
1484
this_lines = self.get_lines(self.this_tree, file_id)
1485
m3 = merge3.Merge3(base_lines, this_lines, other_lines,
1486
is_cherrypick=self.cherrypick)
1487
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1488
if self.show_base is True:
1489
base_marker = '|' * 7
1493
def iter_merge3(retval):
1494
retval["text_conflicts"] = False
1495
for line in m3.merge_lines(name_a = "TREE",
1496
name_b = "MERGE-SOURCE",
1497
name_base = "BASE-REVISION",
1498
start_marker=start_marker,
1499
base_marker=base_marker,
1500
reprocess=self.reprocess):
1501
if line.startswith(start_marker):
1502
retval["text_conflicts"] = True
1503
yield line.replace(start_marker, '<' * 7)
1507
merge3_iterator = iter_merge3(retval)
1508
self.tt.create_file(merge3_iterator, trans_id)
1509
if retval["text_conflicts"] is True:
1510
self._raw_conflicts.append(('text conflict', trans_id))
1511
name = self.tt.final_name(trans_id)
1512
parent_id = self.tt.final_parent(trans_id)
1513
file_group = self._dump_conflicts(name, parent_id, file_id,
1514
this_lines, base_lines,
1516
file_group.append(trans_id)
1518
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1519
base_lines=None, other_lines=None, set_version=False,
1521
"""Emit conflict files.
1522
If this_lines, base_lines, or other_lines are omitted, they will be
1523
determined automatically. If set_version is true, the .OTHER, .THIS
1524
or .BASE (in that order) will be created as versioned files.
1526
data = [('OTHER', self.other_tree, other_lines),
1527
('THIS', self.this_tree, this_lines)]
1529
data.append(('BASE', self.base_tree, base_lines))
1531
# We need to use the actual path in the working tree of the file here,
1532
# ignoring the conflict suffixes
1534
if wt.supports_content_filtering():
1536
filter_tree_path = wt.id2path(file_id)
1537
except errors.NoSuchId:
1538
# file has been deleted
1539
filter_tree_path = None
1541
# Skip the id2path lookup for older formats
1542
filter_tree_path = None
1546
for suffix, tree, lines in data:
1547
if tree.has_id(file_id):
1548
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1549
suffix, lines, filter_tree_path)
1550
file_group.append(trans_id)
1551
if set_version and not versioned:
1552
self.tt.version_file(file_id, trans_id)
1556
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1557
lines=None, filter_tree_path=None):
1558
"""Emit a single conflict file."""
1559
name = name + '.' + suffix
1560
trans_id = self.tt.create_path(name, parent_id)
1561
transform.create_from_tree(self.tt, trans_id, tree, file_id, lines,
1565
def merge_executable(self, file_id, file_status):
1566
"""Perform a merge on the execute bit."""
1567
executable = [self.executable(t, file_id) for t in (self.base_tree,
1568
self.other_tree, self.this_tree)]
1569
self._merge_executable(file_id, executable, file_status,
1570
resolver=self._three_way)
1572
def _merge_executable(self, file_id, executable, file_status,
1574
"""Perform a merge on the execute bit."""
1575
base_executable, other_executable, this_executable = executable
1576
if file_status == "deleted":
1578
winner = resolver(*executable)
1579
if winner == "conflict":
1580
# There must be a None in here, if we have a conflict, but we
1581
# need executability since file status was not deleted.
1582
if self.executable(self.other_tree, file_id) is None:
1586
if winner == 'this' and file_status != "modified":
1588
trans_id = self.tt.trans_id_file_id(file_id)
1589
if self.tt.final_kind(trans_id) != "file":
1591
if winner == "this":
1592
executability = this_executable
1594
if self.other_tree.has_id(file_id):
1595
executability = other_executable
1596
elif self.this_tree.has_id(file_id):
1597
executability = this_executable
1598
elif self.base_tree_has_id(file_id):
1599
executability = base_executable
1600
if executability is not None:
1601
trans_id = self.tt.trans_id_file_id(file_id)
1602
self.tt.set_executability(executability, trans_id)
1604
def cook_conflicts(self, fs_conflicts):
1605
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1606
self.cooked_conflicts.extend(transform.cook_conflicts(
1607
fs_conflicts, self.tt))
1608
fp = transform.FinalPaths(self.tt)
1609
for conflict in self._raw_conflicts:
1610
conflict_type = conflict[0]
1611
if conflict_type == 'path conflict':
1613
this_parent, this_name,
1614
other_parent, other_name) = conflict[1:]
1615
if this_parent is None or this_name is None:
1616
this_path = '<deleted>'
1618
parent_path = fp.get_path(
1619
self.tt.trans_id_file_id(this_parent))
1620
this_path = osutils.pathjoin(parent_path, this_name)
1621
if other_parent is None or other_name is None:
1622
other_path = '<deleted>'
1624
parent_path = fp.get_path(
1625
self.tt.trans_id_file_id(other_parent))
1626
other_path = osutils.pathjoin(parent_path, other_name)
1627
c = _mod_conflicts.Conflict.factory(
1628
'path conflict', path=this_path,
1629
conflict_path=other_path,
1631
elif conflict_type == 'contents conflict':
1632
for trans_id in conflict[1]:
1633
file_id = self.tt.final_file_id(trans_id)
1634
if file_id is not None:
1636
path = fp.get_path(trans_id)
1637
for suffix in ('.BASE', '.THIS', '.OTHER'):
1638
if path.endswith(suffix):
1639
path = path[:-len(suffix)]
1641
c = _mod_conflicts.Conflict.factory(conflict_type,
1642
path=path, file_id=file_id)
1643
elif conflict_type == 'text conflict':
1644
trans_id = conflict[1]
1645
path = fp.get_path(trans_id)
1646
file_id = self.tt.final_file_id(trans_id)
1647
c = _mod_conflicts.Conflict.factory(conflict_type,
1648
path=path, file_id=file_id)
1650
raise AssertionError('bad conflict type: %r' % (conflict,))
1651
self.cooked_conflicts.append(c)
1652
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
1655
class WeaveMerger(Merge3Merger):
1656
"""Three-way tree merger, text weave merger."""
1657
supports_reprocess = True
1658
supports_show_base = False
1659
supports_reverse_cherrypick = False
1660
history_based = True
1662
def _generate_merge_plan(self, file_id, base):
1663
return self.this_tree.plan_file_merge(file_id, self.other_tree,
1666
def _merged_lines(self, file_id):
1667
"""Generate the merged lines.
1668
There is no distinction between lines that are meant to contain <<<<<<<
1672
base = self.base_tree
1675
plan = self._generate_merge_plan(file_id, base)
1676
if 'merge' in debug.debug_flags:
1678
trans_id = self.tt.trans_id_file_id(file_id)
1679
name = self.tt.final_name(trans_id) + '.plan'
1680
contents = ('%11s|%s' % l for l in plan)
1681
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1682
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1683
'>>>>>>> MERGE-SOURCE\n')
1684
lines, conflicts = textmerge.merge_lines(self.reprocess)
1686
base_lines = textmerge.base_from_plan()
1689
return lines, base_lines
1691
def text_merge(self, file_id, trans_id):
1692
"""Perform a (weave) text merge for a given file and file-id.
1693
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1694
and a conflict will be noted.
1696
lines, base_lines = self._merged_lines(file_id)
1698
# Note we're checking whether the OUTPUT is binary in this case,
1699
# because we don't want to get into weave merge guts.
1700
textfile.check_text_lines(lines)
1701
self.tt.create_file(lines, trans_id)
1702
if base_lines is not None:
1704
self._raw_conflicts.append(('text conflict', trans_id))
1705
name = self.tt.final_name(trans_id)
1706
parent_id = self.tt.final_parent(trans_id)
1707
file_group = self._dump_conflicts(name, parent_id, file_id,
1709
base_lines=base_lines)
1710
file_group.append(trans_id)
1713
class LCAMerger(WeaveMerger):
1715
def _generate_merge_plan(self, file_id, base):
1716
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1719
class Diff3Merger(Merge3Merger):
1720
"""Three-way merger using external diff3 for text merging"""
1722
def dump_file(self, temp_dir, name, tree, file_id):
1723
out_path = osutils.pathjoin(temp_dir, name)
1724
out_file = open(out_path, "wb")
1726
in_file = tree.get_file(file_id)
1727
for line in in_file:
1728
out_file.write(line)
1733
def text_merge(self, file_id, trans_id):
1734
"""Perform a diff3 merge using a specified file-id and trans-id.
1735
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1736
will be dumped, and a will be conflict noted.
1739
temp_dir = osutils.mkdtemp(prefix="bzr-")
1741
new_file = osutils.pathjoin(temp_dir, "new")
1742
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1743
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1744
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1745
status = bzrlib.patch.diff3(new_file, this, base, other)
1746
if status not in (0, 1):
1747
raise errors.BzrError("Unhandled diff3 exit code")
1748
f = open(new_file, 'rb')
1750
self.tt.create_file(f, trans_id)
1754
name = self.tt.final_name(trans_id)
1755
parent_id = self.tt.final_parent(trans_id)
1756
self._dump_conflicts(name, parent_id, file_id)
1757
self._raw_conflicts.append(('text conflict', trans_id))
1759
osutils.rmtree(temp_dir)
1762
class PathNotInTree(errors.BzrError):
1764
_fmt = """Merge-into failed because %(tree)s does not contain %(path)s."""
1766
def __init__(self, path, tree):
1767
errors.BzrError.__init__(self, path=path, tree=tree)
1770
class MergeIntoMerger(Merger):
1771
"""Merger that understands other_tree will be merged into a subdir.
1773
This also changes the Merger api so that it uses real Branch, revision_id,
1774
and RevisonTree objects, rather than using revision specs.
1777
def __init__(self, this_tree, other_branch, other_tree, target_subdir,
1778
source_subpath, other_rev_id=None):
1779
"""Create a new MergeIntoMerger object.
1781
source_subpath in other_tree will be effectively copied to
1782
target_subdir in this_tree.
1784
:param this_tree: The tree that we will be merging into.
1785
:param other_branch: The Branch we will be merging from.
1786
:param other_tree: The RevisionTree object we want to merge.
1787
:param target_subdir: The relative path where we want to merge
1788
other_tree into this_tree
1789
:param source_subpath: The relative path specifying the subtree of
1790
other_tree to merge into this_tree.
1792
# It is assumed that we are merging a tree that is not in our current
1793
# ancestry, which means we are using the "EmptyTree" as our basis.
1794
null_ancestor_tree = this_tree.branch.repository.revision_tree(
1795
_mod_revision.NULL_REVISION)
1796
super(MergeIntoMerger, self).__init__(
1797
this_branch=this_tree.branch,
1798
this_tree=this_tree,
1799
other_tree=other_tree,
1800
base_tree=null_ancestor_tree,
1802
self._target_subdir = target_subdir
1803
self._source_subpath = source_subpath
1804
self.other_branch = other_branch
1805
if other_rev_id is None:
1806
other_rev_id = other_tree.get_revision_id()
1807
self.other_rev_id = self.other_basis = other_rev_id
1808
self.base_is_ancestor = True
1809
self.backup_files = True
1810
self.merge_type = Merge3Merger
1811
self.show_base = False
1812
self.reprocess = False
1813
self.interesting_ids = None
1814
self.merge_type = _MergeTypeParameterizer(MergeIntoMergeType,
1815
target_subdir=self._target_subdir,
1816
source_subpath=self._source_subpath)
1817
if self._source_subpath != '':
1818
# If this isn't a partial merge make sure the revisions will be
1820
self._maybe_fetch(self.other_branch, self.this_branch,
1823
def set_pending(self):
1824
if self._source_subpath != '':
1826
Merger.set_pending(self)
1829
class _MergeTypeParameterizer(object):
1830
"""Wrap a merge-type class to provide extra parameters.
1832
This is hack used by MergeIntoMerger to pass some extra parameters to its
1833
merge_type. Merger.do_merge() sets up its own set of parameters to pass to
1834
the 'merge_type' member. It is difficult override do_merge without
1835
re-writing the whole thing, so instead we create a wrapper which will pass
1836
the extra parameters.
1839
def __init__(self, merge_type, **kwargs):
1840
self._extra_kwargs = kwargs
1841
self._merge_type = merge_type
1843
def __call__(self, *args, **kwargs):
1844
kwargs.update(self._extra_kwargs)
1845
return self._merge_type(*args, **kwargs)
1847
def __getattr__(self, name):
1848
return getattr(self._merge_type, name)
1851
class MergeIntoMergeType(Merge3Merger):
1852
"""Merger that incorporates a tree (or part of a tree) into another."""
1854
def __init__(self, *args, **kwargs):
1855
"""Initialize the merger object.
1857
:param args: See Merge3Merger.__init__'s args.
1858
:param kwargs: See Merge3Merger.__init__'s keyword args, except for
1859
source_subpath and target_subdir.
1860
:keyword source_subpath: The relative path specifying the subtree of
1861
other_tree to merge into this_tree.
1862
:keyword target_subdir: The relative path where we want to merge
1863
other_tree into this_tree
1865
# All of the interesting work happens during Merge3Merger.__init__(),
1866
# so we have have to hack in to get our extra parameters set.
1867
self._source_subpath = kwargs.pop('source_subpath')
1868
self._target_subdir = kwargs.pop('target_subdir')
1869
super(MergeIntoMergeType, self).__init__(*args, **kwargs)
1871
def _compute_transform(self):
1872
child_pb = ui.ui_factory.nested_progress_bar()
1874
entries = self._entries_to_incorporate()
1875
entries = list(entries)
1876
for num, (entry, parent_id) in enumerate(entries):
1877
child_pb.update('Preparing file merge', num, len(entries))
1878
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1879
trans_id = transform.new_by_entry(self.tt, entry,
1880
parent_trans_id, self.other_tree)
1883
self._finish_computing_transform()
1885
def _entries_to_incorporate(self):
1886
"""Yields pairs of (inventory_entry, new_parent)."""
1887
other_inv = self.other_tree.inventory
1888
subdir_id = other_inv.path2id(self._source_subpath)
1889
if subdir_id is None:
1890
# XXX: The error would be clearer if it gave the URL of the source
1891
# branch, but we don't have a reference to that here.
1892
raise PathNotInTree(self._source_subpath, "Source tree")
1893
subdir = other_inv[subdir_id]
1894
parent_in_target = osutils.dirname(self._target_subdir)
1895
target_id = self.this_tree.inventory.path2id(parent_in_target)
1896
if target_id is None:
1897
raise PathNotInTree(self._target_subdir, "Target tree")
1898
name_in_target = osutils.basename(self._target_subdir)
1899
merge_into_root = subdir.copy()
1900
merge_into_root.name = name_in_target
1901
if merge_into_root.file_id in self.this_tree.inventory:
1902
# Give the root a new file-id.
1903
# This can happen fairly easily if the directory we are
1904
# incorporating is the root, and both trees have 'TREE_ROOT' as
1905
# their root_id. Users will expect this to Just Work, so we
1906
# change the file-id here.
1907
# Non-root file-ids could potentially conflict too. That's really
1908
# an edge case, so we don't do anything special for those. We let
1909
# them cause conflicts.
1910
merge_into_root.file_id = generate_ids.gen_file_id(name_in_target)
1911
yield (merge_into_root, target_id)
1912
if subdir.kind != 'directory':
1913
# No children, so we are done.
1915
for ignored_path, entry in other_inv.iter_entries_by_dir(subdir_id):
1916
parent_id = entry.parent_id
1917
if parent_id == subdir.file_id:
1918
# The root's parent ID has changed, so make sure children of
1919
# the root refer to the new ID.
1920
parent_id = merge_into_root.file_id
1921
yield (entry, parent_id)
1924
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1926
merge_type=Merge3Merger,
1927
interesting_ids=None,
1931
interesting_files=None,
1934
change_reporter=None):
1935
"""Primary interface for merging.
1937
Typical use is probably::
1939
merge_inner(branch, branch.get_revision_tree(other_revision),
1940
branch.get_revision_tree(base_revision))
1942
if this_tree is None:
1943
raise errors.BzrError("bzrlib.merge.merge_inner requires a this_tree "
1944
"parameter as of bzrlib version 0.8.")
1945
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1946
pb=pb, change_reporter=change_reporter)
1947
merger.backup_files = backup_files
1948
merger.merge_type = merge_type
1949
merger.interesting_ids = interesting_ids
1950
merger.ignore_zero = ignore_zero
1951
if interesting_files:
1953
raise ValueError('Only supply interesting_ids'
1954
' or interesting_files')
1955
merger.interesting_files = interesting_files
1956
merger.show_base = show_base
1957
merger.reprocess = reprocess
1958
merger.other_rev_id = other_rev_id
1959
merger.other_basis = other_rev_id
1960
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1961
if get_revision_id is None:
1962
get_revision_id = base_tree.last_revision
1963
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
1964
merger.set_base_revision(get_revision_id(), this_branch)
1965
return merger.do_merge()
1967
def get_merge_type_registry():
1968
"""Merge type registry is in bzrlib.option to avoid circular imports.
1970
This method provides a sanctioned way to retrieve it.
1972
from bzrlib import option
1973
return option._merge_type_registry
1976
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1977
def status_a(revision, text):
1978
if revision in ancestors_b:
1979
return 'killed-b', text
1981
return 'new-a', text
1983
def status_b(revision, text):
1984
if revision in ancestors_a:
1985
return 'killed-a', text
1987
return 'new-b', text
1989
plain_a = [t for (a, t) in annotated_a]
1990
plain_b = [t for (a, t) in annotated_b]
1991
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1992
blocks = matcher.get_matching_blocks()
1995
for ai, bi, l in blocks:
1996
# process all mismatched sections
1997
# (last mismatched section is handled because blocks always
1998
# includes a 0-length last block)
1999
for revision, text in annotated_a[a_cur:ai]:
2000
yield status_a(revision, text)
2001
for revision, text in annotated_b[b_cur:bi]:
2002
yield status_b(revision, text)
2003
# and now the matched section
2006
for text_a in plain_a[ai:a_cur]:
2007
yield "unchanged", text_a
2010
class _PlanMergeBase(object):
2012
def __init__(self, a_rev, b_rev, vf, key_prefix):
2015
:param a_rev: Revision-id of one revision to merge
2016
:param b_rev: Revision-id of the other revision to merge
2017
:param vf: A VersionedFiles containing both revisions
2018
:param key_prefix: A prefix for accessing keys in vf, typically
2024
self._last_lines = None
2025
self._last_lines_revision_id = None
2026
self._cached_matching_blocks = {}
2027
self._key_prefix = key_prefix
2028
self._precache_tip_lines()
2030
def _precache_tip_lines(self):
2031
lines = self.get_lines([self.a_rev, self.b_rev])
2032
self.lines_a = lines[self.a_rev]
2033
self.lines_b = lines[self.b_rev]
2035
def get_lines(self, revisions):
2036
"""Get lines for revisions from the backing VersionedFiles.
2038
:raises RevisionNotPresent: on absent texts.
2040
keys = [(self._key_prefix + (rev,)) for rev in revisions]
2042
for record in self.vf.get_record_stream(keys, 'unordered', True):
2043
if record.storage_kind == 'absent':
2044
raise errors.RevisionNotPresent(record.key, self.vf)
2045
result[record.key[-1]] = osutils.chunks_to_lines(
2046
record.get_bytes_as('chunked'))
2049
def plan_merge(self):
2050
"""Generate a 'plan' for merging the two revisions.
2052
This involves comparing their texts and determining the cause of
2053
differences. If text A has a line and text B does not, then either the
2054
line was added to text A, or it was deleted from B. Once the causes
2055
are combined, they are written out in the format described in
2056
VersionedFile.plan_merge
2058
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
2059
unique_a, unique_b = self._unique_lines(blocks)
2060
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
2061
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
2062
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
2064
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
2067
for i, j, n in blocks:
2068
for a_index in range(last_i, i):
2069
if a_index in new_a:
2070
if a_index in killed_b:
2071
yield 'conflicted-a', self.lines_a[a_index]
2073
yield 'new-a', self.lines_a[a_index]
2075
yield 'killed-b', self.lines_a[a_index]
2076
for b_index in range(last_j, j):
2077
if b_index in new_b:
2078
if b_index in killed_a:
2079
yield 'conflicted-b', self.lines_b[b_index]
2081
yield 'new-b', self.lines_b[b_index]
2083
yield 'killed-a', self.lines_b[b_index]
2084
# handle common lines
2085
for a_index in range(i, i+n):
2086
yield 'unchanged', self.lines_a[a_index]
2090
def _get_matching_blocks(self, left_revision, right_revision):
2091
"""Return a description of which sections of two revisions match.
2093
See SequenceMatcher.get_matching_blocks
2095
cached = self._cached_matching_blocks.get((left_revision,
2097
if cached is not None:
2099
if self._last_lines_revision_id == left_revision:
2100
left_lines = self._last_lines
2101
right_lines = self.get_lines([right_revision])[right_revision]
2103
lines = self.get_lines([left_revision, right_revision])
2104
left_lines = lines[left_revision]
2105
right_lines = lines[right_revision]
2106
self._last_lines = right_lines
2107
self._last_lines_revision_id = right_revision
2108
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
2110
return matcher.get_matching_blocks()
2112
def _unique_lines(self, matching_blocks):
2113
"""Analyse matching_blocks to determine which lines are unique
2115
:return: a tuple of (unique_left, unique_right), where the values are
2116
sets of line numbers of unique lines.
2122
for i, j, n in matching_blocks:
2123
unique_left.extend(range(last_i, i))
2124
unique_right.extend(range(last_j, j))
2127
return unique_left, unique_right
2130
def _subtract_plans(old_plan, new_plan):
2131
"""Remove changes from new_plan that came from old_plan.
2133
It is assumed that the difference between the old_plan and new_plan
2134
is their choice of 'b' text.
2136
All lines from new_plan that differ from old_plan are emitted
2137
verbatim. All lines from new_plan that match old_plan but are
2138
not about the 'b' revision are emitted verbatim.
2140
Lines that match and are about the 'b' revision are the lines we
2141
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
2142
is skipped entirely.
2144
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
2147
for i, j, n in matcher.get_matching_blocks():
2148
for jj in range(last_j, j):
2150
for jj in range(j, j+n):
2151
plan_line = new_plan[jj]
2152
if plan_line[0] == 'new-b':
2154
elif plan_line[0] == 'killed-b':
2155
yield 'unchanged', plan_line[1]
2161
class _PlanMerge(_PlanMergeBase):
2162
"""Plan an annotate merge using on-the-fly annotation"""
2164
def __init__(self, a_rev, b_rev, vf, key_prefix):
2165
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
2166
self.a_key = self._key_prefix + (self.a_rev,)
2167
self.b_key = self._key_prefix + (self.b_rev,)
2168
self.graph = _mod_graph.Graph(self.vf)
2169
heads = self.graph.heads((self.a_key, self.b_key))
2171
# one side dominates, so we can just return its values, yay for
2173
# Ideally we would know that before we get this far
2174
self._head_key = heads.pop()
2175
if self._head_key == self.a_key:
2179
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2180
self._head_key[-1], other)
2183
self._head_key = None
2186
def _precache_tip_lines(self):
2187
# Turn this into a no-op, because we will do this later
2190
def _find_recursive_lcas(self):
2191
"""Find all the ancestors back to a unique lca"""
2192
cur_ancestors = (self.a_key, self.b_key)
2193
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2194
# rather than a key tuple. We will just map that directly to no common
2198
next_lcas = self.graph.find_lca(*cur_ancestors)
2199
# Map a plain NULL_REVISION to a simple no-ancestors
2200
if next_lcas == set([_mod_revision.NULL_REVISION]):
2202
# Order the lca's based on when they were merged into the tip
2203
# While the actual merge portion of weave merge uses a set() of
2204
# active revisions, the order of insertion *does* effect the
2205
# implicit ordering of the texts.
2206
for rev_key in cur_ancestors:
2207
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2209
parent_map[rev_key] = ordered_parents
2210
if len(next_lcas) == 0:
2212
elif len(next_lcas) == 1:
2213
parent_map[list(next_lcas)[0]] = ()
2215
elif len(next_lcas) > 2:
2216
# More than 2 lca's, fall back to grabbing all nodes between
2217
# this and the unique lca.
2218
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2220
self.a_key, self.b_key, cur_ancestors)
2221
cur_lcas = next_lcas
2222
while len(cur_lcas) > 1:
2223
cur_lcas = self.graph.find_lca(*cur_lcas)
2224
if len(cur_lcas) == 0:
2225
# No common base to find, use the full ancestry
2228
unique_lca = list(cur_lcas)[0]
2229
if unique_lca == _mod_revision.NULL_REVISION:
2230
# find_lca will return a plain 'NULL_REVISION' rather
2231
# than a key tuple when there is no common ancestor, we
2232
# prefer to just use None, because it doesn't confuse
2233
# _get_interesting_texts()
2235
parent_map.update(self._find_unique_parents(next_lcas,
2238
cur_ancestors = next_lcas
2241
def _find_unique_parents(self, tip_keys, base_key):
2242
"""Find ancestors of tip that aren't ancestors of base.
2244
:param tip_keys: Nodes that are interesting
2245
:param base_key: Cull all ancestors of this node
2246
:return: The parent map for all revisions between tip_keys and
2247
base_key. base_key will be included. References to nodes outside of
2248
the ancestor set will also be removed.
2250
# TODO: this would be simpler if find_unique_ancestors took a list
2251
# instead of a single tip, internally it supports it, but it
2252
# isn't a "backwards compatible" api change.
2253
if base_key is None:
2254
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2255
# We remove NULL_REVISION because it isn't a proper tuple key, and
2256
# thus confuses things like _get_interesting_texts, and our logic
2257
# to add the texts into the memory weave.
2258
if _mod_revision.NULL_REVISION in parent_map:
2259
parent_map.pop(_mod_revision.NULL_REVISION)
2262
for tip in tip_keys:
2264
self.graph.find_unique_ancestors(tip, [base_key]))
2265
parent_map = self.graph.get_parent_map(interesting)
2266
parent_map[base_key] = ()
2267
culled_parent_map, child_map, tails = self._remove_external_references(
2269
# Remove all the tails but base_key
2270
if base_key is not None:
2271
tails.remove(base_key)
2272
self._prune_tails(culled_parent_map, child_map, tails)
2273
# Now remove all the uninteresting 'linear' regions
2274
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2278
def _remove_external_references(parent_map):
2279
"""Remove references that go outside of the parent map.
2281
:param parent_map: Something returned from Graph.get_parent_map(keys)
2282
:return: (filtered_parent_map, child_map, tails)
2283
filtered_parent_map is parent_map without external references
2284
child_map is the {parent_key: [child_keys]} mapping
2285
tails is a list of nodes that do not have any parents in the map
2287
# TODO: The basic effect of this function seems more generic than
2288
# _PlanMerge. But the specific details of building a child_map,
2289
# and computing tails seems very specific to _PlanMerge.
2290
# Still, should this be in Graph land?
2291
filtered_parent_map = {}
2294
for key, parent_keys in parent_map.iteritems():
2295
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2296
if not culled_parent_keys:
2298
for parent_key in culled_parent_keys:
2299
child_map.setdefault(parent_key, []).append(key)
2300
# TODO: Do we want to do this, it adds overhead for every node,
2301
# just to say that the node has no children
2302
child_map.setdefault(key, [])
2303
filtered_parent_map[key] = culled_parent_keys
2304
return filtered_parent_map, child_map, tails
2307
def _prune_tails(parent_map, child_map, tails_to_remove):
2308
"""Remove tails from the parent map.
2310
This will remove the supplied revisions until no more children have 0
2313
:param parent_map: A dict of {child: [parents]}, this dictionary will
2314
be modified in place.
2315
:param tails_to_remove: A list of tips that should be removed,
2316
this list will be consumed
2317
:param child_map: The reverse dict of parent_map ({parent: [children]})
2318
this dict will be modified
2319
:return: None, parent_map will be modified in place.
2321
while tails_to_remove:
2322
next = tails_to_remove.pop()
2323
parent_map.pop(next)
2324
children = child_map.pop(next)
2325
for child in children:
2326
child_parents = parent_map[child]
2327
child_parents.remove(next)
2328
if len(child_parents) == 0:
2329
tails_to_remove.append(child)
2331
def _get_interesting_texts(self, parent_map):
2332
"""Return a dict of texts we are interested in.
2334
Note that the input is in key tuples, but the output is in plain
2337
:param parent_map: The output from _find_recursive_lcas
2338
:return: A dict of {'revision_id':lines} as returned by
2339
_PlanMergeBase.get_lines()
2341
all_revision_keys = set(parent_map)
2342
all_revision_keys.add(self.a_key)
2343
all_revision_keys.add(self.b_key)
2345
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2346
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2349
def _build_weave(self):
2350
from bzrlib import weave
2351
self._weave = weave.Weave(weave_name='in_memory_weave',
2352
allow_reserved=True)
2353
parent_map = self._find_recursive_lcas()
2355
all_texts = self._get_interesting_texts(parent_map)
2357
# Note: Unfortunately, the order given by topo_sort will effect the
2358
# ordering resolution in the output. Specifically, if you add A then B,
2359
# then in the output text A lines will show up before B lines. And, of
2360
# course, topo_sort doesn't guarantee any real ordering.
2361
# So we use merge_sort, and add a fake node on the tip.
2362
# This ensures that left-hand parents will always be inserted into the
2363
# weave before right-hand parents.
2364
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2365
parent_map[tip_key] = (self.a_key, self.b_key)
2367
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2371
# for key in tsort.topo_sort(parent_map):
2372
parent_keys = parent_map[key]
2373
revision_id = key[-1]
2374
parent_ids = [k[-1] for k in parent_keys]
2375
self._weave.add_lines(revision_id, parent_ids,
2376
all_texts[revision_id])
2378
def plan_merge(self):
2379
"""Generate a 'plan' for merging the two revisions.
2381
This involves comparing their texts and determining the cause of
2382
differences. If text A has a line and text B does not, then either the
2383
line was added to text A, or it was deleted from B. Once the causes
2384
are combined, they are written out in the format described in
2385
VersionedFile.plan_merge
2387
if self._head_key is not None: # There was a single head
2388
if self._head_key == self.a_key:
2391
if self._head_key != self.b_key:
2392
raise AssertionError('There was an invalid head: %s != %s'
2393
% (self.b_key, self._head_key))
2395
head_rev = self._head_key[-1]
2396
lines = self.get_lines([head_rev])[head_rev]
2397
return ((plan, line) for line in lines)
2398
return self._weave.plan_merge(self.a_rev, self.b_rev)
2401
class _PlanLCAMerge(_PlanMergeBase):
2403
This merge algorithm differs from _PlanMerge in that:
2404
1. comparisons are done against LCAs only
2405
2. cases where a contested line is new versus one LCA but old versus
2406
another are marked as conflicts, by emitting the line as conflicted-a
2409
This is faster, and hopefully produces more useful output.
2412
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2413
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2414
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2417
if lca == _mod_revision.NULL_REVISION:
2420
self.lcas.add(lca[-1])
2421
for lca in self.lcas:
2422
if _mod_revision.is_null(lca):
2425
lca_lines = self.get_lines([lca])[lca]
2426
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2428
blocks = list(matcher.get_matching_blocks())
2429
self._cached_matching_blocks[(a_rev, lca)] = blocks
2430
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2432
blocks = list(matcher.get_matching_blocks())
2433
self._cached_matching_blocks[(b_rev, lca)] = blocks
2435
def _determine_status(self, revision_id, unique_line_numbers):
2436
"""Determines the status unique lines versus all lcas.
2438
Basically, determines why the line is unique to this revision.
2440
A line may be determined new, killed, or both.
2442
If a line is determined new, that means it was not present in at least
2443
one LCA, and is not present in the other merge revision.
2445
If a line is determined killed, that means the line was present in
2448
If a line is killed and new, this indicates that the two merge
2449
revisions contain differing conflict resolutions.
2450
:param revision_id: The id of the revision in which the lines are
2452
:param unique_line_numbers: The line numbers of unique lines.
2453
:return a tuple of (new_this, killed_other):
2457
unique_line_numbers = set(unique_line_numbers)
2458
for lca in self.lcas:
2459
blocks = self._get_matching_blocks(revision_id, lca)
2460
unique_vs_lca, _ignored = self._unique_lines(blocks)
2461
new.update(unique_line_numbers.intersection(unique_vs_lca))
2462
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
