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
17
from __future__ import absolute_import
19
from .lazy_import import lazy_import
20
lazy_import(globals(), """
24
branch as _mod_branch,
26
conflicts as _mod_conflicts,
31
revision as _mod_revision,
40
from breezy.bzr import (
44
from breezy.i18n import gettext
55
# TODO: Report back as changes are merged in
58
def transform_tree(from_tree, to_tree, interesting_files=None):
59
with from_tree.lock_tree_write():
60
merge_inner(from_tree.branch, to_tree, from_tree,
61
ignore_zero=True, this_tree=from_tree,
62
interesting_files=interesting_files)
65
class MergeHooks(hooks.Hooks):
68
hooks.Hooks.__init__(self, "breezy.merge", "Merger.hooks")
69
self.add_hook('merge_file_content',
70
"Called with a breezy.merge.Merger object to create a per file "
71
"merge object when starting a merge. "
72
"Should return either None or a subclass of "
73
"``breezy.merge.AbstractPerFileMerger``. "
74
"Such objects will then be called per file "
75
"that needs to be merged (including when one "
76
"side has deleted the file and the other has changed it). "
77
"See the AbstractPerFileMerger API docs for details on how it is "
80
self.add_hook('pre_merge',
81
'Called before a merge. '
82
'Receives a Merger object as the single argument.',
84
self.add_hook('post_merge',
85
'Called after a merge. '
86
'Receives a Merger object as the single argument. '
87
'The return value is ignored.',
91
class AbstractPerFileMerger(object):
92
"""PerFileMerger objects are used by plugins extending merge for breezy.
94
See ``breezy.plugins.news_merge.news_merge`` for an example concrete class.
96
:ivar merger: The Merge3Merger performing the merge.
99
def __init__(self, merger):
100
"""Create a PerFileMerger for use with merger."""
103
def merge_contents(self, merge_params):
104
"""Attempt to merge the contents of a single file.
106
:param merge_params: A breezy.merge.MergeFileHookParams
107
:return: A tuple of (status, chunks), where status is one of
108
'not_applicable', 'success', 'conflicted', or 'delete'. If status
109
is 'success' or 'conflicted', then chunks should be an iterable of
110
strings for the new file contents.
112
return ('not applicable', None)
115
class PerFileMerger(AbstractPerFileMerger):
116
"""Merge individual files when self.file_matches returns True.
118
This class is intended to be subclassed. The file_matches and
119
merge_matching methods should be overridden with concrete implementations.
122
def file_matches(self, params):
123
"""Return True if merge_matching should be called on this file.
125
Only called with merges of plain files with no clear winner.
127
Subclasses must override this.
129
raise NotImplementedError(self.file_matches)
131
def merge_contents(self, params):
132
"""Merge the contents of a single file."""
133
# Check whether this custom merge logic should be used.
135
# OTHER is a straight winner, rely on default merge.
136
params.winner == 'other' or
137
# THIS and OTHER aren't both files.
138
not params.is_file_merge() or
139
# The filename doesn't match
140
not self.file_matches(params)):
141
return 'not_applicable', None
142
return self.merge_matching(params)
144
def merge_matching(self, params):
145
"""Merge the contents of a single file that has matched the criteria
146
in PerFileMerger.merge_contents (is a conflict, is a file,
147
self.file_matches is True).
149
Subclasses must override this.
151
raise NotImplementedError(self.merge_matching)
154
class ConfigurableFileMerger(PerFileMerger):
155
"""Merge individual files when configured via a .conf file.
157
This is a base class for concrete custom file merging logic. Concrete
158
classes should implement ``merge_text``.
160
See ``breezy.plugins.news_merge.news_merge`` for an example concrete class.
162
:ivar affected_files: The configured file paths to merge.
164
:cvar name_prefix: The prefix to use when looking up configuration
165
details. <name_prefix>_merge_files describes the files targeted by the
168
:cvar default_files: The default file paths to merge when no configuration
175
def __init__(self, merger):
176
super(ConfigurableFileMerger, self).__init__(merger)
177
self.affected_files = None
178
self.default_files = self.__class__.default_files or []
179
self.name_prefix = self.__class__.name_prefix
180
if self.name_prefix is None:
181
raise ValueError("name_prefix must be set.")
183
def file_matches(self, params):
184
"""Check whether the file should call the merge hook.
186
<name_prefix>_merge_files configuration variable is a list of files
187
that should use the hook.
189
affected_files = self.affected_files
190
if affected_files is None:
191
config = self.merger.this_branch.get_config()
192
# Until bzr provides a better policy for caching the config, we
193
# just add the part we're interested in to the params to avoid
194
# reading the config files repeatedly (breezy.conf, location.conf,
196
config_key = self.name_prefix + '_merge_files'
197
affected_files = config.get_user_option_as_list(config_key)
198
if affected_files is None:
199
# If nothing was specified in the config, use the default.
200
affected_files = self.default_files
201
self.affected_files = affected_files
203
filepath = params.this_path
204
if filepath in affected_files:
208
def merge_matching(self, params):
209
return self.merge_text(params)
211
def merge_text(self, params):
212
"""Merge the byte contents of a single file.
214
This is called after checking that the merge should be performed in
215
merge_contents, and it should behave as per
216
``breezy.merge.AbstractPerFileMerger.merge_contents``.
218
raise NotImplementedError(self.merge_text)
221
class MergeFileHookParams(object):
222
"""Object holding parameters passed to merge_file_content hooks.
224
There are some fields hooks can access:
226
:ivar file_id: the file ID of the file being merged
227
:ivar base_path: Path in base tree
228
:ivar other_path: Path in other tree
229
:ivar this_path: Path in this tree
230
:ivar trans_id: the transform ID for the merge of this file
231
:ivar this_kind: kind of file_id in 'this' tree
232
:ivar other_kind: kind of file_id in 'other' tree
233
:ivar winner: one of 'this', 'other', 'conflict'
236
def __init__(self, merger, file_id, paths, trans_id, this_kind, other_kind,
238
self._merger = merger
239
self.file_id = file_id
241
self.base_path, self.other_path, self.this_path = paths
242
self.trans_id = trans_id
243
self.this_kind = this_kind
244
self.other_kind = other_kind
247
def is_file_merge(self):
248
"""True if this_kind and other_kind are both 'file'."""
249
return self.this_kind == 'file' and self.other_kind == 'file'
251
@decorators.cachedproperty
252
def base_lines(self):
253
"""The lines of the 'base' version of the file."""
254
return self._merger.get_lines(self._merger.base_tree, self.base_path)
256
@decorators.cachedproperty
257
def this_lines(self):
258
"""The lines of the 'this' version of the file."""
259
return self._merger.get_lines(self._merger.this_tree, self.this_path)
261
@decorators.cachedproperty
262
def other_lines(self):
263
"""The lines of the 'other' version of the file."""
264
return self._merger.get_lines(self._merger.other_tree, self.other_path)
267
class Merger(object):
271
def __init__(self, this_branch, other_tree=None, base_tree=None,
272
this_tree=None, change_reporter=None,
273
recurse='down', revision_graph=None):
274
object.__init__(self)
275
self.this_branch = this_branch
276
self.this_basis = _mod_revision.ensure_null(
277
this_branch.last_revision())
278
self.this_rev_id = None
279
self.this_tree = this_tree
280
self.this_revision_tree = None
281
self.this_basis_tree = None
282
self.other_tree = other_tree
283
self.other_branch = None
284
self.base_tree = base_tree
285
self.ignore_zero = False
286
self.backup_files = False
287
self.interesting_files = None
288
self.show_base = False
289
self.reprocess = False
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, 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 base_tree: The basis to use for the merge. If unspecified,
349
other_tree.basis_tree() will be used.
351
if base_tree is None:
352
base_tree = other_tree.basis_tree()
353
merger = Merger(tree.branch, other_tree, base_tree, tree)
354
merger.base_rev_id = merger.base_tree.get_revision_id()
355
merger.other_rev_id = None
356
merger.other_basis = merger.base_rev_id
360
def from_mergeable(klass, tree, mergeable):
361
"""Return a Merger for a bundle or merge directive.
363
:param tree: The tree to merge changes into
364
:param mergeable: A merge directive or bundle
366
mergeable.install_revisions(tree.branch.repository)
367
base_revision_id, other_revision_id, verified =\
368
mergeable.get_merge_request(tree.branch.repository)
369
revision_graph = tree.branch.repository.get_graph()
370
if base_revision_id is not None:
371
if (base_revision_id != _mod_revision.NULL_REVISION and
372
revision_graph.is_ancestor(
373
base_revision_id, tree.branch.last_revision())):
374
base_revision_id = None
376
trace.warning('Performing cherrypick')
377
merger = klass.from_revision_ids(tree, other_revision_id,
378
base_revision_id, revision_graph=revision_graph)
379
return merger, verified
382
def from_revision_ids(tree, other, base=None, other_branch=None,
383
base_branch=None, revision_graph=None,
385
"""Return a Merger for revision-ids.
387
:param tree: The tree to merge changes into
388
:param other: The revision-id to use as OTHER
389
:param base: The revision-id to use as BASE. If not specified, will
391
:param other_branch: A branch containing the other revision-id. If
392
not supplied, tree.branch is used.
393
:param base_branch: A branch containing the base revision-id. If
394
not supplied, other_branch or tree.branch will be used.
395
:param revision_graph: If you have a revision_graph precomputed, pass
396
it in, otherwise it will be created for you.
397
:param tree_branch: The branch associated with tree. If not supplied,
398
tree.branch will be used.
400
if tree_branch is None:
401
tree_branch = tree.branch
402
merger = Merger(tree_branch, this_tree=tree,
403
revision_graph=revision_graph)
404
if other_branch is None:
405
other_branch = tree.branch
406
merger.set_other_revision(other, other_branch)
410
if base_branch is None:
411
base_branch = other_branch
412
merger.set_base_revision(base, base_branch)
415
def revision_tree(self, revision_id, branch=None):
416
if revision_id not in self._cached_trees:
418
branch = self.this_branch
420
tree = self.this_tree.revision_tree(revision_id)
421
except errors.NoSuchRevisionInTree:
422
tree = branch.repository.revision_tree(revision_id)
423
self._cached_trees[revision_id] = tree
424
return self._cached_trees[revision_id]
426
def _get_tree(self, treespec, possible_transports=None):
427
location, revno = treespec
429
tree = workingtree.WorkingTree.open_containing(location)[0]
430
return tree.branch, tree
431
branch = _mod_branch.Branch.open_containing(
432
location, possible_transports)[0]
434
revision_id = branch.last_revision()
436
revision_id = branch.get_rev_id(revno)
437
revision_id = _mod_revision.ensure_null(revision_id)
438
return branch, self.revision_tree(revision_id, branch)
440
def set_interesting_files(self, file_list):
441
self.interesting_files = file_list
443
def set_pending(self):
444
if (not self.base_is_ancestor or not self.base_is_other_ancestor
445
or self.other_rev_id is None):
449
def _add_parent(self):
450
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
451
new_parent_trees = []
452
operation = cleanup.OperationWithCleanups(
453
self.this_tree.set_parent_trees)
454
for revision_id in new_parents:
456
tree = self.revision_tree(revision_id)
457
except errors.NoSuchRevision:
461
operation.add_cleanup(tree.unlock)
462
new_parent_trees.append((revision_id, tree))
463
operation.run_simple(new_parent_trees, allow_leftmost_as_ghost=True)
465
def set_other(self, other_revision, possible_transports=None):
466
"""Set the revision and tree to merge from.
468
This sets the other_tree, other_rev_id, other_basis attributes.
470
:param other_revision: The [path, revision] list to merge from.
472
self.other_branch, self.other_tree = self._get_tree(other_revision,
474
if other_revision[1] == -1:
475
self.other_rev_id = _mod_revision.ensure_null(
476
self.other_branch.last_revision())
477
if _mod_revision.is_null(self.other_rev_id):
478
raise errors.NoCommits(self.other_branch)
479
self.other_basis = self.other_rev_id
480
elif other_revision[1] is not None:
481
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
482
self.other_basis = self.other_rev_id
484
self.other_rev_id = None
485
self.other_basis = self.other_branch.last_revision()
486
if self.other_basis is None:
487
raise errors.NoCommits(self.other_branch)
488
if self.other_rev_id is not None:
489
self._cached_trees[self.other_rev_id] = self.other_tree
490
self._maybe_fetch(self.other_branch,
491
self.this_branch, self.other_basis)
493
def set_other_revision(self, revision_id, other_branch):
494
"""Set 'other' based on a branch and revision id
496
:param revision_id: The revision to use for a tree
497
:param other_branch: The branch containing this tree
499
self.other_rev_id = revision_id
500
self.other_branch = other_branch
501
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
502
self.other_tree = self.revision_tree(revision_id)
503
self.other_basis = revision_id
505
def set_base_revision(self, revision_id, branch):
506
"""Set 'base' based on a branch and revision id
508
:param revision_id: The revision to use for a tree
509
:param branch: The branch containing this tree
511
self.base_rev_id = revision_id
512
self.base_branch = branch
513
self._maybe_fetch(branch, self.this_branch, revision_id)
514
self.base_tree = self.revision_tree(revision_id)
516
def _maybe_fetch(self, source, target, revision_id):
517
if not source.repository.has_same_location(target.repository):
518
target.fetch(source, revision_id)
521
revisions = [_mod_revision.ensure_null(self.this_basis),
522
_mod_revision.ensure_null(self.other_basis)]
523
if _mod_revision.NULL_REVISION in revisions:
524
self.base_rev_id = _mod_revision.NULL_REVISION
525
self.base_tree = self.revision_tree(self.base_rev_id)
526
self._is_criss_cross = False
528
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
529
self._is_criss_cross = False
531
self.base_rev_id = _mod_revision.NULL_REVISION
533
self.base_rev_id = list(lcas)[0]
534
else: # len(lcas) > 1
535
self._is_criss_cross = True
537
# find_unique_lca can only handle 2 nodes, so we have to
538
# start back at the beginning. It is a shame to traverse
539
# the graph again, but better than re-implementing
541
self.base_rev_id = self.revision_graph.find_unique_lca(
542
revisions[0], revisions[1])
544
self.base_rev_id = self.revision_graph.find_unique_lca(
546
sorted_lca_keys = self.revision_graph.find_merge_order(
548
if self.base_rev_id == _mod_revision.NULL_REVISION:
549
self.base_rev_id = sorted_lca_keys[0]
551
if self.base_rev_id == _mod_revision.NULL_REVISION:
552
raise errors.UnrelatedBranches()
553
if self._is_criss_cross:
554
trace.warning('Warning: criss-cross merge encountered. See bzr'
555
' help criss-cross.')
556
trace.mutter('Criss-cross lcas: %r' % lcas)
557
if self.base_rev_id in lcas:
558
trace.mutter('Unable to find unique lca. '
559
'Fallback %r as best option.'
561
interesting_revision_ids = set(lcas)
562
interesting_revision_ids.add(self.base_rev_id)
563
interesting_trees = dict((t.get_revision_id(), t)
564
for t in self.this_branch.repository.revision_trees(
565
interesting_revision_ids))
566
self._cached_trees.update(interesting_trees)
567
if self.base_rev_id in lcas:
568
self.base_tree = interesting_trees[self.base_rev_id]
570
self.base_tree = interesting_trees.pop(self.base_rev_id)
571
self._lca_trees = [interesting_trees[key]
572
for key in sorted_lca_keys]
574
self.base_tree = self.revision_tree(self.base_rev_id)
575
self.base_is_ancestor = True
576
self.base_is_other_ancestor = True
577
trace.mutter('Base revid: %r' % self.base_rev_id)
579
def set_base(self, base_revision):
580
"""Set the base revision to use for the merge.
582
:param base_revision: A 2-list containing a path and revision number.
584
trace.mutter("doing merge() with no base_revision specified")
585
if base_revision == [None, None]:
588
base_branch, self.base_tree = self._get_tree(base_revision)
589
if base_revision[1] == -1:
590
self.base_rev_id = base_branch.last_revision()
591
elif base_revision[1] is None:
592
self.base_rev_id = _mod_revision.NULL_REVISION
594
self.base_rev_id = _mod_revision.ensure_null(
595
base_branch.get_rev_id(base_revision[1]))
596
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
598
def make_merger(self):
599
kwargs = {'working_tree': self.this_tree, 'this_tree': self.this_tree,
600
'other_tree': self.other_tree,
601
'interesting_files': self.interesting_files,
602
'this_branch': self.this_branch,
603
'other_branch': self.other_branch,
605
if self.merge_type.requires_base:
606
kwargs['base_tree'] = self.base_tree
607
if self.merge_type.supports_reprocess:
608
kwargs['reprocess'] = self.reprocess
610
raise errors.BzrError(
611
"Conflict reduction is not supported for merge"
612
" type %s." % self.merge_type)
613
if self.merge_type.supports_show_base:
614
kwargs['show_base'] = self.show_base
616
raise errors.BzrError("Showing base is not supported for this"
617
" merge type. %s" % self.merge_type)
618
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
619
and not self.base_is_other_ancestor):
620
raise errors.CannotReverseCherrypick()
621
if self.merge_type.supports_cherrypick:
622
kwargs['cherrypick'] = (not self.base_is_ancestor or
623
not self.base_is_other_ancestor)
624
if self._is_criss_cross and getattr(self.merge_type,
625
'supports_lca_trees', False):
626
kwargs['lca_trees'] = self._lca_trees
627
return self.merge_type(change_reporter=self.change_reporter,
630
def _do_merge_to(self):
631
merge = self.make_merger()
632
if self.other_branch is not None:
633
self.other_branch.update_references(self.this_branch)
634
for hook in Merger.hooks['pre_merge']:
637
for hook in Merger.hooks['post_merge']:
639
if self.recurse == 'down':
640
for relpath, file_id in self.this_tree.iter_references():
641
sub_tree = self.this_tree.get_nested_tree(relpath)
642
other_revision = self.other_tree.get_reference_revision(
644
if other_revision == sub_tree.last_revision():
646
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
647
sub_merge.merge_type = self.merge_type
648
other_branch = self.other_branch.reference_parent(
650
sub_merge.set_other_revision(other_revision, other_branch)
651
base_tree_path = _mod_tree.find_previous_path(
652
self.this_tree, self.base_tree, relpath)
653
base_revision = self.base_tree.get_reference_revision(
655
sub_merge.base_tree = \
656
sub_tree.branch.repository.revision_tree(base_revision)
657
sub_merge.base_rev_id = base_revision
662
operation = cleanup.OperationWithCleanups(self._do_merge_to)
663
self.this_tree.lock_tree_write()
664
operation.add_cleanup(self.this_tree.unlock)
665
if self.base_tree is not None:
666
self.base_tree.lock_read()
667
operation.add_cleanup(self.base_tree.unlock)
668
if self.other_tree is not None:
669
self.other_tree.lock_read()
670
operation.add_cleanup(self.other_tree.unlock)
671
merge = operation.run_simple()
672
if len(merge.cooked_conflicts) == 0:
673
if not self.ignore_zero and not trace.is_quiet():
674
trace.note(gettext("All changes applied successfully."))
676
trace.note(gettext("%d conflicts encountered.")
677
% len(merge.cooked_conflicts))
679
return len(merge.cooked_conflicts)
682
class _InventoryNoneEntry(object):
683
"""This represents an inventory entry which *isn't there*.
685
It simplifies the merging logic if we always have an InventoryEntry, even
686
if it isn't actually present
693
symlink_target = None
697
_none_entry = _InventoryNoneEntry()
700
class Merge3Merger(object):
701
"""Three-way merger that uses the merge3 text merger"""
703
supports_reprocess = True
704
supports_show_base = True
705
history_based = False
706
supports_cherrypick = True
707
supports_reverse_cherrypick = True
708
winner_idx = {"this": 2, "other": 1, "conflict": 1}
709
supports_lca_trees = True
710
requires_file_merge_plan = False
712
def __init__(self, working_tree, this_tree, base_tree, other_tree,
713
reprocess=False, show_base=False,
714
change_reporter=None, interesting_files=None, do_merge=True,
715
cherrypick=False, lca_trees=None, this_branch=None,
717
"""Initialize the merger object and perform the merge.
719
:param working_tree: The working tree to apply the merge to
720
:param this_tree: The local tree in the merge operation
721
:param base_tree: The common tree in the merge operation
722
:param other_tree: The other tree to merge changes from
723
:param this_branch: The branch associated with this_tree. Defaults to
724
this_tree.branch if not supplied.
725
:param other_branch: The branch associated with other_tree, if any.
726
:param: reprocess If True, perform conflict-reduction processing.
727
:param show_base: If True, show the base revision in text conflicts.
728
(incompatible with reprocess)
729
:param change_reporter: An object that should report changes made
730
:param interesting_files: The tree-relative paths of files that should
731
participate in the merge. If these paths refer to directories,
732
the contents of those directories will also be included. If not
733
specified, all files may participate in the
735
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
736
if the ancestry was found to include a criss-cross merge.
737
Otherwise should be None.
739
object.__init__(self)
740
if this_branch is None:
741
this_branch = this_tree.branch
742
self.interesting_files = interesting_files
743
self.working_tree = working_tree
744
self.this_tree = this_tree
745
self.base_tree = base_tree
746
self.other_tree = other_tree
747
self.this_branch = this_branch
748
self.other_branch = other_branch
749
self._raw_conflicts = []
750
self.cooked_conflicts = []
751
self.reprocess = reprocess
752
self.show_base = show_base
753
self._lca_trees = lca_trees
754
# Uncommenting this will change the default algorithm to always use
755
# _entries_lca. This can be useful for running the test suite and
756
# making sure we haven't missed any corner cases.
757
# if lca_trees is None:
758
# self._lca_trees = [self.base_tree]
759
self.change_reporter = change_reporter
760
self.cherrypick = cherrypick
765
operation = cleanup.OperationWithCleanups(self._do_merge)
766
self.working_tree.lock_tree_write()
767
operation.add_cleanup(self.working_tree.unlock)
768
self.this_tree.lock_read()
769
operation.add_cleanup(self.this_tree.unlock)
770
self.base_tree.lock_read()
771
operation.add_cleanup(self.base_tree.unlock)
772
self.other_tree.lock_read()
773
operation.add_cleanup(self.other_tree.unlock)
776
def _do_merge(self, operation):
777
self.tt = transform.TreeTransform(self.working_tree, None)
778
operation.add_cleanup(self.tt.finalize)
779
self._compute_transform()
780
results = self.tt.apply(no_conflicts=True)
781
self.write_modified(results)
783
self.working_tree.add_conflicts(self.cooked_conflicts)
784
except errors.UnsupportedOperation:
787
def make_preview_transform(self):
788
with self.base_tree.lock_read(), self.other_tree.lock_read():
789
self.tt = transform.TransformPreview(self.working_tree)
790
self._compute_transform()
793
def _compute_transform(self):
794
if self._lca_trees is None:
795
entries = self._entries3()
796
resolver = self._three_way
798
entries = self._entries_lca()
799
resolver = self._lca_multi_way
800
# Prepare merge hooks
801
factories = Merger.hooks['merge_file_content']
802
# One hook for each registered one plus our default merger
803
hooks = [factory(self) for factory in factories] + [self]
804
self.active_hooks = [hook for hook in hooks if hook is not None]
805
with ui.ui_factory.nested_progress_bar() as child_pb:
806
for num, (file_id, changed, paths3, parents3, names3,
807
executable3) in enumerate(entries):
808
# Try merging each entry
809
child_pb.update(gettext('Preparing file merge'),
811
self._merge_names(file_id, paths3, parents3,
812
names3, resolver=resolver)
814
file_status = self._do_merge_contents(paths3, file_id)
816
file_status = 'unmodified'
817
self._merge_executable(paths3, file_id, executable3,
818
file_status, resolver=resolver)
819
self.tt.fixup_new_roots()
820
self._finish_computing_transform()
822
def _finish_computing_transform(self):
823
"""Finalize the transform and report the changes.
825
This is the second half of _compute_transform.
827
with ui.ui_factory.nested_progress_bar() as child_pb:
828
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
829
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
830
if self.change_reporter is not None:
831
from breezy import delta
832
delta.report_changes(
833
self.tt.iter_changes(), self.change_reporter)
834
self.cook_conflicts(fs_conflicts)
835
for conflict in self.cooked_conflicts:
836
trace.warning('%s', conflict.describe())
839
"""Gather data about files modified between three trees.
841
Return a list of tuples of file_id, changed, parents3, names3,
842
executable3. changed is a boolean indicating whether the file contents
843
or kind were changed. parents3 is a tuple of parent ids for base,
844
other and this. names3 is a tuple of names for base, other and this.
845
executable3 is a tuple of execute-bit values for base, other and this.
848
iterator = self.other_tree.iter_changes(self.base_tree,
849
specific_files=self.interesting_files,
850
extra_trees=[self.this_tree])
851
this_interesting_files = self.this_tree.find_related_paths_across_trees(
852
self.interesting_files, trees=[self.other_tree])
853
this_entries = dict(self.this_tree.iter_entries_by_dir(
854
specific_files=this_interesting_files))
855
for (file_id, paths, changed, versioned, parents, names, kind,
856
executable) in iterator:
857
if paths[0] is not None:
858
this_path = _mod_tree.find_previous_path(
859
self.base_tree, self.this_tree, paths[0])
861
this_path = _mod_tree.find_previous_path(
862
self.other_tree, self.this_tree, paths[1])
863
this_entry = this_entries.get(this_path)
864
if this_entry is not None:
865
this_name = this_entry.name
866
this_parent = this_entry.parent_id
867
this_executable = this_entry.executable
871
this_executable = None
872
parents3 = parents + (this_parent,)
873
names3 = names + (this_name,)
874
paths3 = paths + (this_path, )
875
executable3 = executable + (this_executable,)
876
result.append((file_id, changed, paths3,
877
parents3, names3, executable3))
880
def _entries_lca(self):
881
"""Gather data about files modified between multiple trees.
883
This compares OTHER versus all LCA trees, and for interesting entries,
884
it then compares with THIS and BASE.
886
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
888
:return: [(file_id, changed, paths, parents, names, executable)], where:
890
* file_id: Simple file_id of the entry
891
* changed: Boolean, True if the kind or contents changed else False
892
* paths: ((base, [path, in, lcas]), path_other, path_this)
893
* parents: ((base, [parent_id, in, lcas]), parent_id_other,
895
* names: ((base, [name, in, lcas]), name_in_other, name_in_this)
896
* executable: ((base, [exec, in, lcas]), exec_in_other,
899
if self.interesting_files is not None:
900
lookup_trees = [self.this_tree, self.base_tree]
901
lookup_trees.extend(self._lca_trees)
902
# I think we should include the lca trees as well
903
interesting_files = self.other_tree.find_related_paths_across_trees(
904
self.interesting_files, lookup_trees)
906
interesting_files = None
908
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
910
base_inventory = self.base_tree.root_inventory
911
this_inventory = self.this_tree.root_inventory
912
for path, file_id, other_ie, lca_values in walker.iter_all():
913
# Is this modified at all from any of the other trees?
915
other_ie = _none_entry
918
other_path = self.other_tree.id2path(file_id)
919
if interesting_files is not None and other_path not in interesting_files:
922
# If other_revision is found in any of the lcas, that means this
923
# node is uninteresting. This is because when merging, if there are
924
# multiple heads(), we have to create a new node. So if we didn't,
925
# we know that the ancestry is linear, and that OTHER did not
927
# See doc/developers/lca_merge_resolution.txt for details
928
other_revision = other_ie.revision
929
if other_revision is not None:
930
# We can't use this shortcut when other_revision is None,
931
# because it may be None because things are WorkingTrees, and
932
# not because it is *actually* None.
933
is_unmodified = False
934
for lca_path, ie in lca_values:
935
if ie is not None and ie.revision == other_revision:
943
for lca_path, lca_ie in lca_values:
945
lca_entries.append(_none_entry)
946
lca_paths.append(None)
948
lca_entries.append(lca_ie)
949
lca_paths.append(lca_path)
952
base_ie = base_inventory.get_entry(file_id)
953
except errors.NoSuchId:
954
base_ie = _none_entry
957
base_path = self.base_tree.id2path(file_id)
960
this_ie = this_inventory.get_entry(file_id)
961
except errors.NoSuchId:
962
this_ie = _none_entry
965
this_path = self.this_tree.id2path(file_id)
971
for lca_ie in lca_entries:
972
lca_kinds.append(lca_ie.kind)
973
lca_parent_ids.append(lca_ie.parent_id)
974
lca_names.append(lca_ie.name)
975
lca_executable.append(lca_ie.executable)
977
kind_winner = self._lca_multi_way(
978
(base_ie.kind, lca_kinds),
979
other_ie.kind, this_ie.kind)
980
parent_id_winner = self._lca_multi_way(
981
(base_ie.parent_id, lca_parent_ids),
982
other_ie.parent_id, this_ie.parent_id)
983
name_winner = self._lca_multi_way(
984
(base_ie.name, lca_names),
985
other_ie.name, this_ie.name)
987
content_changed = True
988
if kind_winner == 'this':
989
# No kind change in OTHER, see if there are *any* changes
990
if other_ie.kind == 'directory':
991
if parent_id_winner == 'this' and name_winner == 'this':
992
# No change for this directory in OTHER, skip
994
content_changed = False
995
elif other_ie.kind is None or other_ie.kind == 'file':
996
def get_sha1(tree, path):
1000
return tree.get_file_sha1(path)
1001
except errors.NoSuchFile:
1003
base_sha1 = get_sha1(self.base_tree, base_path)
1004
lca_sha1s = [get_sha1(tree, lca_path)
1006
in zip(self._lca_trees, lca_paths)]
1007
this_sha1 = get_sha1(self.this_tree, this_path)
1008
other_sha1 = get_sha1(self.other_tree, other_path)
1009
sha1_winner = self._lca_multi_way(
1010
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1011
allow_overriding_lca=False)
1012
exec_winner = self._lca_multi_way(
1013
(base_ie.executable, lca_executable),
1014
other_ie.executable, this_ie.executable)
1015
if (parent_id_winner == 'this' and name_winner == 'this'
1016
and sha1_winner == 'this' and exec_winner == 'this'):
1017
# No kind, parent, name, exec, or content change for
1018
# OTHER, so this node is not considered interesting
1020
if sha1_winner == 'this':
1021
content_changed = False
1022
elif other_ie.kind == 'symlink':
1023
def get_target(ie, tree, path):
1024
if ie.kind != 'symlink':
1026
return tree.get_symlink_target(path)
1027
base_target = get_target(base_ie, self.base_tree, base_path)
1028
lca_targets = [get_target(ie, tree, lca_path) for ie, tree, lca_path
1029
in zip(lca_entries, self._lca_trees, lca_paths)]
1030
this_target = get_target(
1031
this_ie, self.this_tree, this_path)
1032
other_target = get_target(
1033
other_ie, self.other_tree, other_path)
1034
target_winner = self._lca_multi_way(
1035
(base_target, lca_targets),
1036
other_target, this_target)
1037
if (parent_id_winner == 'this' and name_winner == 'this'
1038
and target_winner == 'this'):
1039
# No kind, parent, name, or symlink target change
1042
if target_winner == 'this':
1043
content_changed = False
1044
elif other_ie.kind == 'tree-reference':
1045
# The 'changed' information seems to be handled at a higher
1046
# level. At least, _entries3 returns False for content
1047
# changed, even when at a new revision_id.
1048
content_changed = False
1049
if (parent_id_winner == 'this' and name_winner == 'this'):
1050
# Nothing interesting
1053
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1055
# If we have gotten this far, that means something has changed
1056
result.append((file_id, content_changed,
1057
((base_path, lca_paths),
1058
other_path, this_path),
1059
((base_ie.parent_id, lca_parent_ids),
1060
other_ie.parent_id, this_ie.parent_id),
1061
((base_ie.name, lca_names),
1062
other_ie.name, this_ie.name),
1063
((base_ie.executable, lca_executable),
1064
other_ie.executable, this_ie.executable)
1068
def write_modified(self, results):
1069
if not self.working_tree.supports_merge_modified():
1071
modified_hashes = {}
1072
for path in results.modified_paths:
1073
wt_relpath = self.working_tree.relpath(path)
1074
file_id = self.working_tree.path2id(wt_relpath)
1077
hash = self.working_tree.get_file_sha1(wt_relpath)
1080
modified_hashes[file_id] = hash
1081
self.working_tree.set_merge_modified(modified_hashes)
1085
"""Determine the parent for a file_id (used as a key method)"""
1088
return entry.parent_id
1092
"""Determine the name for a file_id (used as a key method)"""
1098
def contents_sha1(tree, path):
1099
"""Determine the sha1 of the file contents (used as a key method)."""
1101
return tree.get_file_sha1(path)
1102
except errors.NoSuchFile:
1106
def executable(tree, path):
1107
"""Determine the executability of a file-id (used as a key method)."""
1109
if tree.kind(path) != "file":
1111
except errors.NoSuchFile:
1113
return tree.is_executable(path)
1116
def kind(tree, path):
1117
"""Determine the kind of a file-id (used as a key method)."""
1119
return tree.kind(path)
1120
except errors.NoSuchFile:
1124
def _three_way(base, other, this):
1126
# if 'base == other', either they all agree, or only 'this' has
1129
elif this not in (base, other):
1130
# 'this' is neither 'base' nor 'other', so both sides changed
1133
# "Ambiguous clean merge" -- both sides have made the same change.
1136
# this == base: only other has changed.
1140
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
1141
"""Consider LCAs when determining whether a change has occurred.
1143
If LCAS are all identical, this is the same as a _three_way comparison.
1145
:param bases: value in (BASE, [LCAS])
1146
:param other: value in OTHER
1147
:param this: value in THIS
1148
:param allow_overriding_lca: If there is more than one unique lca
1149
value, allow OTHER to override THIS if it has a new value, and
1150
THIS only has an lca value, or vice versa. This is appropriate for
1151
truly scalar values, not as much for non-scalars.
1152
:return: 'this', 'other', or 'conflict' depending on whether an entry
1155
# See doc/developers/lca_tree_merging.txt for details about this
1158
# Either Ambiguously clean, or nothing was actually changed. We
1161
base_val, lca_vals = bases
1162
# Remove 'base_val' from the lca_vals, because it is not interesting
1163
filtered_lca_vals = [lca_val for lca_val in lca_vals
1164
if lca_val != base_val]
1165
if len(filtered_lca_vals) == 0:
1166
return Merge3Merger._three_way(base_val, other, this)
1168
unique_lca_vals = set(filtered_lca_vals)
1169
if len(unique_lca_vals) == 1:
1170
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1172
if allow_overriding_lca:
1173
if other in unique_lca_vals:
1174
if this in unique_lca_vals:
1175
# Each side picked a different lca, conflict
1178
# This has a value which supersedes both lca values, and
1179
# other only has an lca value
1181
elif this in unique_lca_vals:
1182
# OTHER has a value which supersedes both lca values, and this
1183
# only has an lca value
1186
# At this point, the lcas disagree, and the tip disagree
1189
def _merge_names(self, file_id, paths, parents, names, resolver):
1190
"""Perform a merge on file_id names and parents"""
1191
base_name, other_name, this_name = names
1192
base_parent, other_parent, this_parent = parents
1193
unused_base_path, other_path, this_path = paths
1195
name_winner = resolver(*names)
1197
parent_id_winner = resolver(*parents)
1198
if this_name is None:
1199
if name_winner == "this":
1200
name_winner = "other"
1201
if parent_id_winner == "this":
1202
parent_id_winner = "other"
1203
if name_winner == "this" and parent_id_winner == "this":
1205
if name_winner == 'conflict' or parent_id_winner == 'conflict':
1206
# Creating helpers (.OTHER or .THIS) here cause problems down the
1207
# road if a ContentConflict needs to be created so we should not do
1209
trans_id = self.tt.trans_id_file_id(file_id)
1210
self._raw_conflicts.append(('path conflict', trans_id, file_id,
1211
this_parent, this_name,
1212
other_parent, other_name))
1213
if other_path is None:
1214
# it doesn't matter whether the result was 'other' or
1215
# 'conflict'-- if it has no file id, we leave it alone.
1217
parent_id = parents[self.winner_idx[parent_id_winner]]
1218
name = names[self.winner_idx[name_winner]]
1219
if parent_id is not None or name is not None:
1220
# if we get here, name_winner and parent_winner are set to safe
1222
if parent_id is None and name is not None:
1223
# if parent_id is None and name is non-None, current file is
1225
if names[self.winner_idx[parent_id_winner]] != '':
1226
raise AssertionError(
1227
'File looks like a root, but named %s' %
1228
names[self.winner_idx[parent_id_winner]])
1229
parent_trans_id = transform.ROOT_PARENT
1231
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1232
self.tt.adjust_path(name, parent_trans_id,
1233
self.tt.trans_id_file_id(file_id))
1235
def _do_merge_contents(self, paths, file_id):
1236
"""Performs a merge on file_id contents."""
1237
def contents_pair(tree, path):
1241
kind = tree.kind(path)
1242
except errors.NoSuchFile:
1245
contents = tree.get_file_sha1(path)
1246
elif kind == "symlink":
1247
contents = tree.get_symlink_target(path)
1250
return kind, contents
1252
base_path, other_path, this_path = paths
1253
# See SPOT run. run, SPOT, run.
1254
# So we're not QUITE repeating ourselves; we do tricky things with
1256
other_pair = contents_pair(self.other_tree, other_path)
1257
this_pair = contents_pair(self.this_tree, this_path)
1259
(base_path, lca_paths) = base_path
1260
base_pair = contents_pair(self.base_tree, base_path)
1261
lca_pairs = [contents_pair(tree, path)
1262
for tree, path in zip(self._lca_trees, lca_paths)]
1263
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1264
this_pair, allow_overriding_lca=False)
1266
base_pair = contents_pair(self.base_tree, base_path)
1267
if base_pair == other_pair:
1270
# We delayed evaluating this_pair as long as we can to avoid
1271
# unnecessary sha1 calculation
1272
this_pair = contents_pair(self.this_tree, this_path)
1273
winner = self._three_way(base_pair, other_pair, this_pair)
1274
if winner == 'this':
1275
# No interesting changes introduced by OTHER
1277
# We have a hypothetical conflict, but if we have files, then we
1278
# can try to merge the content
1279
trans_id = self.tt.trans_id_file_id(file_id)
1280
params = MergeFileHookParams(
1281
self, file_id, (base_path, other_path,
1282
this_path), trans_id, this_pair[0],
1283
other_pair[0], winner)
1284
hooks = self.active_hooks
1285
hook_status = 'not_applicable'
1287
hook_status, lines = hook.merge_contents(params)
1288
if hook_status != 'not_applicable':
1289
# Don't try any more hooks, this one applies.
1291
# If the merge ends up replacing the content of the file, we get rid of
1292
# it at the end of this method (this variable is used to track the
1293
# exceptions to this rule).
1296
if hook_status == 'not_applicable':
1297
# No merge hook was able to resolve the situation. Two cases exist:
1298
# a content conflict or a duplicate one.
1300
name = self.tt.final_name(trans_id)
1301
parent_id = self.tt.final_parent(trans_id)
1302
inhibit_content_conflict = False
1303
if params.this_kind is None: # file_id is not in THIS
1304
# Is the name used for a different file_id ?
1305
if self.this_tree.is_versioned(other_path):
1306
# Two entries for the same path
1308
# versioning the merged file will trigger a duplicate
1310
self.tt.version_file(file_id, trans_id)
1311
transform.create_from_tree(
1312
self.tt, trans_id, self.other_tree,
1313
other_path, file_id=file_id,
1314
filter_tree_path=self._get_filter_tree_path(file_id))
1315
inhibit_content_conflict = True
1316
elif params.other_kind is None: # file_id is not in OTHER
1317
# Is the name used for a different file_id ?
1318
if self.other_tree.is_versioned(this_path):
1319
# Two entries for the same path again, but here, the other
1320
# entry will also be merged. We simply inhibit the
1321
# 'content' conflict creation because we know OTHER will
1322
# create (or has already created depending on ordering) an
1323
# entry at the same path. This will trigger a 'duplicate'
1326
inhibit_content_conflict = True
1327
if not inhibit_content_conflict:
1328
if params.this_kind is not None:
1329
self.tt.unversion_file(trans_id)
1330
# This is a contents conflict, because none of the available
1331
# functions could merge it.
1332
file_group = self._dump_conflicts(
1333
name, (base_path, other_path, this_path), parent_id,
1334
file_id, set_version=True)
1335
self._raw_conflicts.append(('contents conflict', file_group))
1336
elif hook_status == 'success':
1337
self.tt.create_file(lines, trans_id)
1338
elif hook_status == 'conflicted':
1339
# XXX: perhaps the hook should be able to provide
1340
# the BASE/THIS/OTHER files?
1341
self.tt.create_file(lines, trans_id)
1342
self._raw_conflicts.append(('text conflict', trans_id))
1343
name = self.tt.final_name(trans_id)
1344
parent_id = self.tt.final_parent(trans_id)
1345
self._dump_conflicts(
1346
name, (base_path, other_path, this_path), parent_id, file_id)
1347
elif hook_status == 'delete':
1348
self.tt.unversion_file(trans_id)
1350
elif hook_status == 'done':
1351
# The hook function did whatever it needs to do directly, no
1352
# further action needed here.
1355
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1356
if not this_path and result == "modified":
1357
self.tt.version_file(file_id, trans_id)
1359
# The merge has been performed and produced a new content, so the
1360
# old contents should not be retained.
1361
self.tt.delete_contents(trans_id)
1364
def _default_other_winner_merge(self, merge_hook_params):
1365
"""Replace this contents with other."""
1366
file_id = merge_hook_params.file_id
1367
trans_id = merge_hook_params.trans_id
1368
if merge_hook_params.other_path is not None:
1369
# OTHER changed the file
1370
transform.create_from_tree(
1371
self.tt, trans_id, self.other_tree,
1372
merge_hook_params.other_path, file_id=file_id,
1373
filter_tree_path=self._get_filter_tree_path(file_id))
1375
elif merge_hook_params.this_path is not None:
1376
# OTHER deleted the file
1377
return 'delete', None
1379
raise AssertionError(
1380
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1383
def merge_contents(self, merge_hook_params):
1384
"""Fallback merge logic after user installed hooks."""
1385
# This function is used in merge hooks as the fallback instance.
1386
# Perhaps making this function and the functions it calls be a
1387
# a separate class would be better.
1388
if merge_hook_params.winner == 'other':
1389
# OTHER is a straight winner, so replace this contents with other
1390
return self._default_other_winner_merge(merge_hook_params)
1391
elif merge_hook_params.is_file_merge():
1392
# THIS and OTHER are both files, so text merge. Either
1393
# BASE is a file, or both converted to files, so at least we
1394
# have agreement that output should be a file.
1396
self.text_merge(merge_hook_params.trans_id,
1397
merge_hook_params.paths, merge_hook_params.file_id)
1398
except errors.BinaryFile:
1399
return 'not_applicable', None
1402
return 'not_applicable', None
1404
def get_lines(self, tree, path):
1405
"""Return the lines in a file, or an empty list."""
1409
kind = tree.kind(path)
1410
except errors.NoSuchFile:
1415
return tree.get_file_lines(path)
1417
def text_merge(self, trans_id, paths, file_id):
1418
"""Perform a three-way text merge on a file_id"""
1419
# it's possible that we got here with base as a different type.
1420
# if so, we just want two-way text conflicts.
1421
base_path, other_path, this_path = paths
1422
base_lines = self.get_lines(self.base_tree, base_path)
1423
other_lines = self.get_lines(self.other_tree, other_path)
1424
this_lines = self.get_lines(self.this_tree, this_path)
1425
m3 = merge3.Merge3(base_lines, this_lines, other_lines,
1426
is_cherrypick=self.cherrypick)
1427
start_marker = b"!START OF MERGE CONFLICT!" + b"I HOPE THIS IS UNIQUE"
1428
if self.show_base is True:
1429
base_marker = b'|' * 7
1433
def iter_merge3(retval):
1434
retval["text_conflicts"] = False
1435
for line in m3.merge_lines(name_a=b"TREE",
1436
name_b=b"MERGE-SOURCE",
1437
name_base=b"BASE-REVISION",
1438
start_marker=start_marker,
1439
base_marker=base_marker,
1440
reprocess=self.reprocess):
1441
if line.startswith(start_marker):
1442
retval["text_conflicts"] = True
1443
yield line.replace(start_marker, b'<' * 7)
1447
merge3_iterator = iter_merge3(retval)
1448
self.tt.create_file(merge3_iterator, trans_id)
1449
if retval["text_conflicts"] is True:
1450
self._raw_conflicts.append(('text conflict', trans_id))
1451
name = self.tt.final_name(trans_id)
1452
parent_id = self.tt.final_parent(trans_id)
1453
file_group = self._dump_conflicts(name, paths, parent_id, file_id,
1454
this_lines, base_lines,
1456
file_group.append(trans_id)
1458
def _get_filter_tree_path(self, file_id):
1459
if self.this_tree.supports_content_filtering():
1460
# We get the path from the working tree if it exists.
1461
# That fails though when OTHER is adding a file, so
1462
# we fall back to the other tree to find the path if
1463
# it doesn't exist locally.
1465
return self.this_tree.id2path(file_id)
1466
except errors.NoSuchId:
1467
return self.other_tree.id2path(file_id)
1468
# Skip the id2path lookup for older formats
1471
def _dump_conflicts(self, name, paths, parent_id, file_id, this_lines=None,
1472
base_lines=None, other_lines=None, set_version=False,
1474
"""Emit conflict files.
1475
If this_lines, base_lines, or other_lines are omitted, they will be
1476
determined automatically. If set_version is true, the .OTHER, .THIS
1477
or .BASE (in that order) will be created as versioned files.
1479
base_path, other_path, this_path = paths
1480
data = [('OTHER', self.other_tree, other_path, other_lines),
1481
('THIS', self.this_tree, this_path, this_lines)]
1483
data.append(('BASE', self.base_tree, base_path, base_lines))
1485
# We need to use the actual path in the working tree of the file here,
1486
# ignoring the conflict suffixes
1488
if wt.supports_content_filtering():
1490
filter_tree_path = wt.id2path(file_id)
1491
except errors.NoSuchId:
1492
# file has been deleted
1493
filter_tree_path = None
1495
# Skip the id2path lookup for older formats
1496
filter_tree_path = None
1500
for suffix, tree, path, lines in data:
1501
if path is not None:
1502
trans_id = self._conflict_file(
1503
name, parent_id, path, tree, file_id, suffix, lines,
1505
file_group.append(trans_id)
1506
if set_version and not versioned:
1507
self.tt.version_file(file_id, trans_id)
1511
def _conflict_file(self, name, parent_id, path, tree, file_id, suffix,
1512
lines=None, filter_tree_path=None):
1513
"""Emit a single conflict file."""
1514
name = name + '.' + suffix
1515
trans_id = self.tt.create_path(name, parent_id)
1516
transform.create_from_tree(
1517
self.tt, trans_id, tree, path,
1518
file_id=file_id, chunks=lines,
1519
filter_tree_path=filter_tree_path)
1522
def merge_executable(self, paths, file_id, file_status):
1523
"""Perform a merge on the execute bit."""
1524
executable = [self.executable(t, p, file_id)
1525
for t, p in zip([self.base_tree, self.other_tree, self.this_tree], paths)]
1526
self._merge_executable(paths, file_id, executable, file_status,
1527
resolver=self._three_way)
1529
def _merge_executable(self, paths, file_id, executable, file_status,
1531
"""Perform a merge on the execute bit."""
1532
base_executable, other_executable, this_executable = executable
1533
base_path, other_path, this_path = paths
1534
if file_status == "deleted":
1536
winner = resolver(*executable)
1537
if winner == "conflict":
1538
# There must be a None in here, if we have a conflict, but we
1539
# need executability since file status was not deleted.
1540
if other_path is None:
1544
if winner == 'this' and file_status != "modified":
1546
trans_id = self.tt.trans_id_file_id(file_id)
1547
if self.tt.final_kind(trans_id) != "file":
1549
if winner == "this":
1550
executability = this_executable
1552
if other_path is not None:
1553
executability = other_executable
1554
elif this_path is not None:
1555
executability = this_executable
1556
elif base_path is not None:
1557
executability = base_executable
1558
if executability is not None:
1559
trans_id = self.tt.trans_id_file_id(file_id)
1560
self.tt.set_executability(executability, trans_id)
1562
def cook_conflicts(self, fs_conflicts):
1563
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1564
content_conflict_file_ids = set()
1565
cooked_conflicts = transform.cook_conflicts(fs_conflicts, self.tt)
1566
fp = transform.FinalPaths(self.tt)
1567
for conflict in self._raw_conflicts:
1568
conflict_type = conflict[0]
1569
if conflict_type == 'path conflict':
1571
this_parent, this_name,
1572
other_parent, other_name) = conflict[1:]
1573
if this_parent is None or this_name is None:
1574
this_path = '<deleted>'
1576
parent_path = fp.get_path(
1577
self.tt.trans_id_file_id(this_parent))
1578
this_path = osutils.pathjoin(parent_path, this_name)
1579
if other_parent is None or other_name is None:
1580
other_path = '<deleted>'
1582
if other_parent == self.other_tree.get_root_id():
1583
# The tree transform doesn't know about the other root,
1584
# so we special case here to avoid a NoFinalPath
1588
parent_path = fp.get_path(
1589
self.tt.trans_id_file_id(other_parent))
1590
other_path = osutils.pathjoin(parent_path, other_name)
1591
c = _mod_conflicts.Conflict.factory(
1592
'path conflict', path=this_path,
1593
conflict_path=other_path,
1595
elif conflict_type == 'contents conflict':
1596
for trans_id in conflict[1]:
1597
file_id = self.tt.final_file_id(trans_id)
1598
if file_id is not None:
1599
# Ok we found the relevant file-id
1601
path = fp.get_path(trans_id)
1602
for suffix in ('.BASE', '.THIS', '.OTHER'):
1603
if path.endswith(suffix):
1604
# Here is the raw path
1605
path = path[:-len(suffix)]
1607
c = _mod_conflicts.Conflict.factory(conflict_type,
1608
path=path, file_id=file_id)
1609
content_conflict_file_ids.add(file_id)
1610
elif conflict_type == 'text conflict':
1611
trans_id = conflict[1]
1612
path = fp.get_path(trans_id)
1613
file_id = self.tt.final_file_id(trans_id)
1614
c = _mod_conflicts.Conflict.factory(conflict_type,
1615
path=path, file_id=file_id)
1617
raise AssertionError('bad conflict type: %r' % (conflict,))
1618
cooked_conflicts.append(c)
1620
self.cooked_conflicts = []
1621
# We want to get rid of path conflicts when a corresponding contents
1622
# conflict exists. This can occur when one branch deletes a file while
1623
# the other renames *and* modifies it. In this case, the content
1624
# conflict is enough.
1625
for c in cooked_conflicts:
1626
if (c.typestring == 'path conflict'
1627
and c.file_id in content_conflict_file_ids):
1629
self.cooked_conflicts.append(c)
1630
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
1633
class WeaveMerger(Merge3Merger):
1634
"""Three-way tree merger, text weave merger."""
1635
supports_reprocess = True
1636
supports_show_base = False
1637
supports_reverse_cherrypick = False
1638
history_based = True
1639
requires_file_merge_plan = True
1641
def _generate_merge_plan(self, file_id, base):
1642
return self.this_tree.plan_file_merge(file_id, self.other_tree,
1645
def _merged_lines(self, file_id):
1646
"""Generate the merged lines.
1647
There is no distinction between lines that are meant to contain <<<<<<<
1651
base = self.base_tree
1654
plan = self._generate_merge_plan(file_id, base)
1655
if 'merge' in debug.debug_flags:
1657
trans_id = self.tt.trans_id_file_id(file_id)
1658
name = self.tt.final_name(trans_id) + '.plan'
1659
contents = (b'%11s|%s' % l for l in plan)
1660
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1661
textmerge = versionedfile.PlanWeaveMerge(plan, b'<<<<<<< TREE\n',
1662
b'>>>>>>> MERGE-SOURCE\n')
1663
lines, conflicts = textmerge.merge_lines(self.reprocess)
1665
base_lines = textmerge.base_from_plan()
1668
return lines, base_lines
1670
def text_merge(self, trans_id, paths, file_id):
1671
"""Perform a (weave) text merge for a given file and file-id.
1672
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1673
and a conflict will be noted.
1675
base_path, other_path, this_path = paths
1676
lines, base_lines = self._merged_lines(file_id)
1678
# Note we're checking whether the OUTPUT is binary in this case,
1679
# because we don't want to get into weave merge guts.
1680
textfile.check_text_lines(lines)
1681
self.tt.create_file(lines, trans_id)
1682
if base_lines is not None:
1684
self._raw_conflicts.append(('text conflict', trans_id))
1685
name = self.tt.final_name(trans_id)
1686
parent_id = self.tt.final_parent(trans_id)
1687
file_group = self._dump_conflicts(name, paths, parent_id, file_id,
1689
base_lines=base_lines)
1690
file_group.append(trans_id)
1693
class LCAMerger(WeaveMerger):
1695
requires_file_merge_plan = True
1697
def _generate_merge_plan(self, file_id, base):
1698
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1702
class Diff3Merger(Merge3Merger):
1703
"""Three-way merger using external diff3 for text merging"""
1705
requires_file_merge_plan = False
1707
def dump_file(self, temp_dir, name, tree, path):
1708
out_path = osutils.pathjoin(temp_dir, name)
1709
with open(out_path, "wb") as out_file:
1710
in_file = tree.get_file(path)
1711
for line in in_file:
1712
out_file.write(line)
1715
def text_merge(self, trans_id, paths, file_id):
1716
"""Perform a diff3 merge using a specified file-id and trans-id.
1717
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1718
will be dumped, and a will be conflict noted.
1721
base_path, other_path, this_path = paths
1722
temp_dir = osutils.mkdtemp(prefix="bzr-")
1724
new_file = osutils.pathjoin(temp_dir, "new")
1725
this = self.dump_file(
1726
temp_dir, "this", self.this_tree, this_path)
1727
base = self.dump_file(
1728
temp_dir, "base", self.base_tree, base_path)
1729
other = self.dump_file(
1730
temp_dir, "other", self.other_tree, other_path)
1731
status = breezy.patch.diff3(new_file, this, base, other)
1732
if status not in (0, 1):
1733
raise errors.BzrError("Unhandled diff3 exit code")
1734
with open(new_file, 'rb') as f:
1735
self.tt.create_file(f, trans_id)
1737
name = self.tt.final_name(trans_id)
1738
parent_id = self.tt.final_parent(trans_id)
1739
self._dump_conflicts(name, paths, parent_id, file_id)
1740
self._raw_conflicts.append(('text conflict', trans_id))
1742
osutils.rmtree(temp_dir)
1745
class PathNotInTree(errors.BzrError):
1747
_fmt = """Merge-into failed because %(tree)s does not contain %(path)s."""
1749
def __init__(self, path, tree):
1750
errors.BzrError.__init__(self, path=path, tree=tree)
1753
class MergeIntoMerger(Merger):
1754
"""Merger that understands other_tree will be merged into a subdir.
1756
This also changes the Merger api so that it uses real Branch, revision_id,
1757
and RevisonTree objects, rather than using revision specs.
1760
def __init__(self, this_tree, other_branch, other_tree, target_subdir,
1761
source_subpath, other_rev_id=None):
1762
"""Create a new MergeIntoMerger object.
1764
source_subpath in other_tree will be effectively copied to
1765
target_subdir in this_tree.
1767
:param this_tree: The tree that we will be merging into.
1768
:param other_branch: The Branch we will be merging from.
1769
:param other_tree: The RevisionTree object we want to merge.
1770
:param target_subdir: The relative path where we want to merge
1771
other_tree into this_tree
1772
:param source_subpath: The relative path specifying the subtree of
1773
other_tree to merge into this_tree.
1775
# It is assumed that we are merging a tree that is not in our current
1776
# ancestry, which means we are using the "EmptyTree" as our basis.
1777
null_ancestor_tree = this_tree.branch.repository.revision_tree(
1778
_mod_revision.NULL_REVISION)
1779
super(MergeIntoMerger, self).__init__(
1780
this_branch=this_tree.branch,
1781
this_tree=this_tree,
1782
other_tree=other_tree,
1783
base_tree=null_ancestor_tree,
1785
self._target_subdir = target_subdir
1786
self._source_subpath = source_subpath
1787
self.other_branch = other_branch
1788
if other_rev_id is None:
1789
other_rev_id = other_tree.get_revision_id()
1790
self.other_rev_id = self.other_basis = other_rev_id
1791
self.base_is_ancestor = True
1792
self.backup_files = True
1793
self.merge_type = Merge3Merger
1794
self.show_base = False
1795
self.reprocess = False
1796
self.interesting_files = None
1797
self.merge_type = _MergeTypeParameterizer(MergeIntoMergeType,
1798
target_subdir=self._target_subdir,
1799
source_subpath=self._source_subpath)
1800
if self._source_subpath != '':
1801
# If this isn't a partial merge make sure the revisions will be
1803
self._maybe_fetch(self.other_branch, self.this_branch,
1806
def set_pending(self):
1807
if self._source_subpath != '':
1809
Merger.set_pending(self)
1812
class _MergeTypeParameterizer(object):
1813
"""Wrap a merge-type class to provide extra parameters.
1815
This is hack used by MergeIntoMerger to pass some extra parameters to its
1816
merge_type. Merger.do_merge() sets up its own set of parameters to pass to
1817
the 'merge_type' member. It is difficult override do_merge without
1818
re-writing the whole thing, so instead we create a wrapper which will pass
1819
the extra parameters.
1822
def __init__(self, merge_type, **kwargs):
1823
self._extra_kwargs = kwargs
1824
self._merge_type = merge_type
1826
def __call__(self, *args, **kwargs):
1827
kwargs.update(self._extra_kwargs)
1828
return self._merge_type(*args, **kwargs)
1830
def __getattr__(self, name):
1831
return getattr(self._merge_type, name)
1834
class MergeIntoMergeType(Merge3Merger):
1835
"""Merger that incorporates a tree (or part of a tree) into another."""
1837
def __init__(self, *args, **kwargs):
1838
"""Initialize the merger object.
1840
:param args: See Merge3Merger.__init__'s args.
1841
:param kwargs: See Merge3Merger.__init__'s keyword args, except for
1842
source_subpath and target_subdir.
1843
:keyword source_subpath: The relative path specifying the subtree of
1844
other_tree to merge into this_tree.
1845
:keyword target_subdir: The relative path where we want to merge
1846
other_tree into this_tree
1848
# All of the interesting work happens during Merge3Merger.__init__(),
1849
# so we have have to hack in to get our extra parameters set.
1850
self._source_subpath = kwargs.pop('source_subpath')
1851
self._target_subdir = kwargs.pop('target_subdir')
1852
super(MergeIntoMergeType, self).__init__(*args, **kwargs)
1854
def _compute_transform(self):
1855
with ui.ui_factory.nested_progress_bar() as child_pb:
1856
entries = self._entries_to_incorporate()
1857
entries = list(entries)
1858
for num, (entry, parent_id, relpath) in enumerate(entries):
1859
child_pb.update(gettext('Preparing file merge'),
1861
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1862
path = osutils.pathjoin(self._source_subpath, relpath)
1863
trans_id = transform.new_by_entry(path, self.tt, entry,
1864
parent_trans_id, self.other_tree)
1865
self._finish_computing_transform()
1867
def _entries_to_incorporate(self):
1868
"""Yields pairs of (inventory_entry, new_parent)."""
1869
other_inv = self.other_tree.root_inventory
1870
subdir_id = other_inv.path2id(self._source_subpath)
1871
if subdir_id is None:
1872
# XXX: The error would be clearer if it gave the URL of the source
1873
# branch, but we don't have a reference to that here.
1874
raise PathNotInTree(self._source_subpath, "Source tree")
1875
subdir = other_inv.get_entry(subdir_id)
1876
parent_in_target = osutils.dirname(self._target_subdir)
1877
target_id = self.this_tree.path2id(parent_in_target)
1878
if target_id is None:
1879
raise PathNotInTree(self._target_subdir, "Target tree")
1880
name_in_target = osutils.basename(self._target_subdir)
1881
merge_into_root = subdir.copy()
1882
merge_into_root.name = name_in_target
1883
if self.this_tree.has_id(merge_into_root.file_id):
1884
# Give the root a new file-id.
1885
# This can happen fairly easily if the directory we are
1886
# incorporating is the root, and both trees have 'TREE_ROOT' as
1887
# their root_id. Users will expect this to Just Work, so we
1888
# change the file-id here.
1889
# Non-root file-ids could potentially conflict too. That's really
1890
# an edge case, so we don't do anything special for those. We let
1891
# them cause conflicts.
1892
merge_into_root.file_id = generate_ids.gen_file_id(name_in_target)
1893
yield (merge_into_root, target_id, '')
1894
if subdir.kind != 'directory':
1895
# No children, so we are done.
1897
for path, entry in other_inv.iter_entries_by_dir(subdir_id):
1898
parent_id = entry.parent_id
1899
if parent_id == subdir.file_id:
1900
# The root's parent ID has changed, so make sure children of
1901
# the root refer to the new ID.
1902
parent_id = merge_into_root.file_id
1903
yield (entry, parent_id, path)
1906
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1908
merge_type=Merge3Merger,
1912
interesting_files=None,
1914
change_reporter=None):
1915
"""Primary interface for merging.
1917
Typical use is probably::
1919
merge_inner(branch, branch.get_revision_tree(other_revision),
1920
branch.get_revision_tree(base_revision))
1922
if this_tree is None:
1923
raise errors.BzrError("breezy.merge.merge_inner requires a this_tree "
1925
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1926
change_reporter=change_reporter)
1927
merger.backup_files = backup_files
1928
merger.merge_type = merge_type
1929
merger.ignore_zero = ignore_zero
1930
merger.interesting_files = interesting_files
1931
merger.show_base = show_base
1932
merger.reprocess = reprocess
1933
merger.other_rev_id = other_rev_id
1934
merger.other_basis = other_rev_id
1935
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1936
if get_revision_id is None:
1937
get_revision_id = base_tree.last_revision
1938
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
1939
merger.set_base_revision(get_revision_id(), this_branch)
1940
return merger.do_merge()
1943
merge_type_registry = registry.Registry()
1944
merge_type_registry.register('diff3', Diff3Merger,
1945
"Merge using external diff3.")
1946
merge_type_registry.register('lca', LCAMerger,
1947
"LCA-newness merge.")
1948
merge_type_registry.register('merge3', Merge3Merger,
1949
"Native diff3-style merge.")
1950
merge_type_registry.register('weave', WeaveMerger,
1951
"Weave-based merge.")
1954
def get_merge_type_registry():
1955
"""Merge type registry was previously in breezy.option
1957
This method provides a backwards compatible way to retrieve it.
1959
return merge_type_registry
1962
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1963
def status_a(revision, text):
1964
if revision in ancestors_b:
1965
return 'killed-b', text
1967
return 'new-a', text
1969
def status_b(revision, text):
1970
if revision in ancestors_a:
1971
return 'killed-a', text
1973
return 'new-b', text
1975
plain_a = [t for (a, t) in annotated_a]
1976
plain_b = [t for (a, t) in annotated_b]
1977
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1978
blocks = matcher.get_matching_blocks()
1981
for ai, bi, l in blocks:
1982
# process all mismatched sections
1983
# (last mismatched section is handled because blocks always
1984
# includes a 0-length last block)
1985
for revision, text in annotated_a[a_cur:ai]:
1986
yield status_a(revision, text)
1987
for revision, text in annotated_b[b_cur:bi]:
1988
yield status_b(revision, text)
1989
# and now the matched section
1992
for text_a in plain_a[ai:a_cur]:
1993
yield "unchanged", text_a
1996
class _PlanMergeBase(object):
1998
def __init__(self, a_rev, b_rev, vf, key_prefix):
2001
:param a_rev: Revision-id of one revision to merge
2002
:param b_rev: Revision-id of the other revision to merge
2003
:param vf: A VersionedFiles containing both revisions
2004
:param key_prefix: A prefix for accessing keys in vf, typically
2010
self._last_lines = None
2011
self._last_lines_revision_id = None
2012
self._cached_matching_blocks = {}
2013
self._key_prefix = key_prefix
2014
self._precache_tip_lines()
2016
def _precache_tip_lines(self):
2017
lines = self.get_lines([self.a_rev, self.b_rev])
2018
self.lines_a = lines[self.a_rev]
2019
self.lines_b = lines[self.b_rev]
2021
def get_lines(self, revisions):
2022
"""Get lines for revisions from the backing VersionedFiles.
2024
:raises RevisionNotPresent: on absent texts.
2026
keys = [(self._key_prefix + (rev,)) for rev in revisions]
2028
for record in self.vf.get_record_stream(keys, 'unordered', True):
2029
if record.storage_kind == 'absent':
2030
raise errors.RevisionNotPresent(record.key, self.vf)
2031
result[record.key[-1]] = osutils.chunks_to_lines(
2032
record.get_bytes_as('chunked'))
2035
def plan_merge(self):
2036
"""Generate a 'plan' for merging the two revisions.
2038
This involves comparing their texts and determining the cause of
2039
differences. If text A has a line and text B does not, then either the
2040
line was added to text A, or it was deleted from B. Once the causes
2041
are combined, they are written out in the format described in
2042
VersionedFile.plan_merge
2044
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
2045
unique_a, unique_b = self._unique_lines(blocks)
2046
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
2047
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
2048
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
2050
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
2053
for i, j, n in blocks:
2054
for a_index in range(last_i, i):
2055
if a_index in new_a:
2056
if a_index in killed_b:
2057
yield 'conflicted-a', self.lines_a[a_index]
2059
yield 'new-a', self.lines_a[a_index]
2061
yield 'killed-b', self.lines_a[a_index]
2062
for b_index in range(last_j, j):
2063
if b_index in new_b:
2064
if b_index in killed_a:
2065
yield 'conflicted-b', self.lines_b[b_index]
2067
yield 'new-b', self.lines_b[b_index]
2069
yield 'killed-a', self.lines_b[b_index]
2070
# handle common lines
2071
for a_index in range(i, i + n):
2072
yield 'unchanged', self.lines_a[a_index]
2076
def _get_matching_blocks(self, left_revision, right_revision):
2077
"""Return a description of which sections of two revisions match.
2079
See SequenceMatcher.get_matching_blocks
2081
cached = self._cached_matching_blocks.get((left_revision,
2083
if cached is not None:
2085
if self._last_lines_revision_id == left_revision:
2086
left_lines = self._last_lines
2087
right_lines = self.get_lines([right_revision])[right_revision]
2089
lines = self.get_lines([left_revision, right_revision])
2090
left_lines = lines[left_revision]
2091
right_lines = lines[right_revision]
2092
self._last_lines = right_lines
2093
self._last_lines_revision_id = right_revision
2094
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
2096
return matcher.get_matching_blocks()
2098
def _unique_lines(self, matching_blocks):
2099
"""Analyse matching_blocks to determine which lines are unique
2101
:return: a tuple of (unique_left, unique_right), where the values are
2102
sets of line numbers of unique lines.
2108
for i, j, n in matching_blocks:
2109
unique_left.extend(range(last_i, i))
2110
unique_right.extend(range(last_j, j))
2113
return unique_left, unique_right
2116
def _subtract_plans(old_plan, new_plan):
2117
"""Remove changes from new_plan that came from old_plan.
2119
It is assumed that the difference between the old_plan and new_plan
2120
is their choice of 'b' text.
2122
All lines from new_plan that differ from old_plan are emitted
2123
verbatim. All lines from new_plan that match old_plan but are
2124
not about the 'b' revision are emitted verbatim.
2126
Lines that match and are about the 'b' revision are the lines we
2127
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
2128
is skipped entirely.
2130
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
2133
for i, j, n in matcher.get_matching_blocks():
2134
for jj in range(last_j, j):
2136
for jj in range(j, j + n):
2137
plan_line = new_plan[jj]
2138
if plan_line[0] == 'new-b':
2140
elif plan_line[0] == 'killed-b':
2141
yield 'unchanged', plan_line[1]
2147
class _PlanMerge(_PlanMergeBase):
2148
"""Plan an annotate merge using on-the-fly annotation"""
2150
def __init__(self, a_rev, b_rev, vf, key_prefix):
2151
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
2152
self.a_key = self._key_prefix + (self.a_rev,)
2153
self.b_key = self._key_prefix + (self.b_rev,)
2154
self.graph = _mod_graph.Graph(self.vf)
2155
heads = self.graph.heads((self.a_key, self.b_key))
2157
# one side dominates, so we can just return its values, yay for
2159
# Ideally we would know that before we get this far
2160
self._head_key = heads.pop()
2161
if self._head_key == self.a_key:
2165
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2166
self._head_key[-1], other)
2169
self._head_key = None
2172
def _precache_tip_lines(self):
2173
# Turn this into a no-op, because we will do this later
2176
def _find_recursive_lcas(self):
2177
"""Find all the ancestors back to a unique lca"""
2178
cur_ancestors = (self.a_key, self.b_key)
2179
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2180
# rather than a key tuple. We will just map that directly to no common
2184
next_lcas = self.graph.find_lca(*cur_ancestors)
2185
# Map a plain NULL_REVISION to a simple no-ancestors
2186
if next_lcas == {_mod_revision.NULL_REVISION}:
2188
# Order the lca's based on when they were merged into the tip
2189
# While the actual merge portion of weave merge uses a set() of
2190
# active revisions, the order of insertion *does* effect the
2191
# implicit ordering of the texts.
2192
for rev_key in cur_ancestors:
2193
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2195
parent_map[rev_key] = ordered_parents
2196
if len(next_lcas) == 0:
2198
elif len(next_lcas) == 1:
2199
parent_map[list(next_lcas)[0]] = ()
2201
elif len(next_lcas) > 2:
2202
# More than 2 lca's, fall back to grabbing all nodes between
2203
# this and the unique lca.
2204
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2206
self.a_key, self.b_key, cur_ancestors)
2207
cur_lcas = next_lcas
2208
while len(cur_lcas) > 1:
2209
cur_lcas = self.graph.find_lca(*cur_lcas)
2210
if len(cur_lcas) == 0:
2211
# No common base to find, use the full ancestry
2214
unique_lca = list(cur_lcas)[0]
2215
if unique_lca == _mod_revision.NULL_REVISION:
2216
# find_lca will return a plain 'NULL_REVISION' rather
2217
# than a key tuple when there is no common ancestor, we
2218
# prefer to just use None, because it doesn't confuse
2219
# _get_interesting_texts()
2221
parent_map.update(self._find_unique_parents(next_lcas,
2224
cur_ancestors = next_lcas
2227
def _find_unique_parents(self, tip_keys, base_key):
2228
"""Find ancestors of tip that aren't ancestors of base.
2230
:param tip_keys: Nodes that are interesting
2231
:param base_key: Cull all ancestors of this node
2232
:return: The parent map for all revisions between tip_keys and
2233
base_key. base_key will be included. References to nodes outside of
2234
the ancestor set will also be removed.
2236
# TODO: this would be simpler if find_unique_ancestors took a list
2237
# instead of a single tip, internally it supports it, but it
2238
# isn't a "backwards compatible" api change.
2239
if base_key is None:
2240
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2241
# We remove NULL_REVISION because it isn't a proper tuple key, and
2242
# thus confuses things like _get_interesting_texts, and our logic
2243
# to add the texts into the memory weave.
2244
if _mod_revision.NULL_REVISION in parent_map:
2245
parent_map.pop(_mod_revision.NULL_REVISION)
2248
for tip in tip_keys:
2250
self.graph.find_unique_ancestors(tip, [base_key]))
2251
parent_map = self.graph.get_parent_map(interesting)
2252
parent_map[base_key] = ()
2253
culled_parent_map, child_map, tails = self._remove_external_references(
2255
# Remove all the tails but base_key
2256
if base_key is not None:
2257
tails.remove(base_key)
2258
self._prune_tails(culled_parent_map, child_map, tails)
2259
# Now remove all the uninteresting 'linear' regions
2260
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2264
def _remove_external_references(parent_map):
2265
"""Remove references that go outside of the parent map.
2267
:param parent_map: Something returned from Graph.get_parent_map(keys)
2268
:return: (filtered_parent_map, child_map, tails)
2269
filtered_parent_map is parent_map without external references
2270
child_map is the {parent_key: [child_keys]} mapping
2271
tails is a list of nodes that do not have any parents in the map
2273
# TODO: The basic effect of this function seems more generic than
2274
# _PlanMerge. But the specific details of building a child_map,
2275
# and computing tails seems very specific to _PlanMerge.
2276
# Still, should this be in Graph land?
2277
filtered_parent_map = {}
2280
for key, parent_keys in viewitems(parent_map):
2281
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2282
if not culled_parent_keys:
2284
for parent_key in culled_parent_keys:
2285
child_map.setdefault(parent_key, []).append(key)
2286
# TODO: Do we want to do this, it adds overhead for every node,
2287
# just to say that the node has no children
2288
child_map.setdefault(key, [])
2289
filtered_parent_map[key] = culled_parent_keys
2290
return filtered_parent_map, child_map, tails
2293
def _prune_tails(parent_map, child_map, tails_to_remove):
2294
"""Remove tails from the parent map.
2296
This will remove the supplied revisions until no more children have 0
2299
:param parent_map: A dict of {child: [parents]}, this dictionary will
2300
be modified in place.
2301
:param tails_to_remove: A list of tips that should be removed,
2302
this list will be consumed
2303
:param child_map: The reverse dict of parent_map ({parent: [children]})
2304
this dict will be modified
2305
:return: None, parent_map will be modified in place.
2307
while tails_to_remove:
2308
next = tails_to_remove.pop()
2309
parent_map.pop(next)
2310
children = child_map.pop(next)
2311
for child in children:
2312
child_parents = parent_map[child]
2313
child_parents.remove(next)
2314
if len(child_parents) == 0:
2315
tails_to_remove.append(child)
2317
def _get_interesting_texts(self, parent_map):
2318
"""Return a dict of texts we are interested in.
2320
Note that the input is in key tuples, but the output is in plain
2323
:param parent_map: The output from _find_recursive_lcas
2324
:return: A dict of {'revision_id':lines} as returned by
2325
_PlanMergeBase.get_lines()
2327
all_revision_keys = set(parent_map)
2328
all_revision_keys.add(self.a_key)
2329
all_revision_keys.add(self.b_key)
2331
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2332
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2335
def _build_weave(self):
2336
from .bzr import weave
2337
self._weave = weave.Weave(weave_name='in_memory_weave',
2338
allow_reserved=True)
2339
parent_map = self._find_recursive_lcas()
2341
all_texts = self._get_interesting_texts(parent_map)
2343
# Note: Unfortunately, the order given by topo_sort will effect the
2344
# ordering resolution in the output. Specifically, if you add A then B,
2345
# then in the output text A lines will show up before B lines. And, of
2346
# course, topo_sort doesn't guarantee any real ordering.
2347
# So we use merge_sort, and add a fake node on the tip.
2348
# This ensures that left-hand parents will always be inserted into the
2349
# weave before right-hand parents.
2350
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2351
parent_map[tip_key] = (self.a_key, self.b_key)
2353
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2357
# for key in tsort.topo_sort(parent_map):
2358
parent_keys = parent_map[key]
2359
revision_id = key[-1]
2360
parent_ids = [k[-1] for k in parent_keys]
2361
self._weave.add_lines(revision_id, parent_ids,
2362
all_texts[revision_id])
2364
def plan_merge(self):
2365
"""Generate a 'plan' for merging the two revisions.
2367
This involves comparing their texts and determining the cause of
2368
differences. If text A has a line and text B does not, then either the
2369
line was added to text A, or it was deleted from B. Once the causes
2370
are combined, they are written out in the format described in
2371
VersionedFile.plan_merge
2373
if self._head_key is not None: # There was a single head
2374
if self._head_key == self.a_key:
2377
if self._head_key != self.b_key:
2378
raise AssertionError('There was an invalid head: %s != %s'
2379
% (self.b_key, self._head_key))
2381
head_rev = self._head_key[-1]
2382
lines = self.get_lines([head_rev])[head_rev]
2383
return ((plan, line) for line in lines)
2384
return self._weave.plan_merge(self.a_rev, self.b_rev)
2387
class _PlanLCAMerge(_PlanMergeBase):
2389
This merge algorithm differs from _PlanMerge in that:
2391
1. comparisons are done against LCAs only
2392
2. cases where a contested line is new versus one LCA but old versus
2393
another are marked as conflicts, by emitting the line as conflicted-a
2396
This is faster, and hopefully produces more useful output.
2399
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2400
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2401
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2404
if lca == _mod_revision.NULL_REVISION:
2407
self.lcas.add(lca[-1])
2408
for lca in self.lcas:
2409
if _mod_revision.is_null(lca):
2412
lca_lines = self.get_lines([lca])[lca]
2413
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2415
blocks = list(matcher.get_matching_blocks())
2416
self._cached_matching_blocks[(a_rev, lca)] = blocks
2417
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2419
blocks = list(matcher.get_matching_blocks())
2420
self._cached_matching_blocks[(b_rev, lca)] = blocks
2422
def _determine_status(self, revision_id, unique_line_numbers):
2423
"""Determines the status unique lines versus all lcas.
2425
Basically, determines why the line is unique to this revision.
2427
A line may be determined new, killed, or both.
2429
If a line is determined new, that means it was not present in at least
2430
one LCA, and is not present in the other merge revision.
2432
If a line is determined killed, that means the line was present in
2435
If a line is killed and new, this indicates that the two merge
2436
revisions contain differing conflict resolutions.
2438
:param revision_id: The id of the revision in which the lines are
2440
:param unique_line_numbers: The line numbers of unique lines.
2441
:return: a tuple of (new_this, killed_other)
2445
unique_line_numbers = set(unique_line_numbers)
2446
for lca in self.lcas:
2447
blocks = self._get_matching_blocks(revision_id, lca)
2448
unique_vs_lca, _ignored = self._unique_lines(blocks)
2449
new.update(unique_line_numbers.intersection(unique_vs_lca))
2450
killed.update(unique_line_numbers.difference(unique_vs_lca))