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(), """
22
branch as _mod_branch,
24
conflicts as _mod_conflicts,
31
revision as _mod_revision,
40
from breezy.bzr import (
43
from breezy.i18n import gettext
54
# TODO: Report back as changes are merged in
57
def transform_tree(from_tree, to_tree, interesting_files=None):
58
with from_tree.lock_tree_write():
59
merge_inner(from_tree.branch, to_tree, from_tree,
60
ignore_zero=True, this_tree=from_tree,
61
interesting_files=interesting_files)
64
class MergeHooks(hooks.Hooks):
67
hooks.Hooks.__init__(self, "breezy.merge", "Merger.hooks")
68
self.add_hook('merge_file_content',
69
"Called with a breezy.merge.Merger object to create a per file "
70
"merge object when starting a merge. "
71
"Should return either None or a subclass of "
72
"``breezy.merge.AbstractPerFileMerger``. "
73
"Such objects will then be called per file "
74
"that needs to be merged (including when one "
75
"side has deleted the file and the other has changed it). "
76
"See the AbstractPerFileMerger API docs for details on how it is "
79
self.add_hook('pre_merge',
80
'Called before a merge. '
81
'Receives a Merger object as the single argument.',
83
self.add_hook('post_merge',
84
'Called after a merge. '
85
'Receives a Merger object as the single argument. '
86
'The return value is ignored.',
90
class AbstractPerFileMerger(object):
91
"""PerFileMerger objects are used by plugins extending merge for breezy.
93
See ``breezy.plugins.news_merge.news_merge`` for an example concrete class.
95
:ivar merger: The Merge3Merger performing the merge.
98
def __init__(self, merger):
99
"""Create a PerFileMerger for use with merger."""
102
def merge_contents(self, merge_params):
103
"""Attempt to merge the contents of a single file.
105
:param merge_params: A breezy.merge.MergeFileHookParams
106
:return: A tuple of (status, chunks), where status is one of
107
'not_applicable', 'success', 'conflicted', or 'delete'. If status
108
is 'success' or 'conflicted', then chunks should be an iterable of
109
strings for the new file contents.
111
return ('not applicable', None)
114
class PerFileMerger(AbstractPerFileMerger):
115
"""Merge individual files when self.file_matches returns True.
117
This class is intended to be subclassed. The file_matches and
118
merge_matching methods should be overridden with concrete implementations.
121
def file_matches(self, params):
122
"""Return True if merge_matching should be called on this file.
124
Only called with merges of plain files with no clear winner.
126
Subclasses must override this.
128
raise NotImplementedError(self.file_matches)
130
def merge_contents(self, params):
131
"""Merge the contents of a single file."""
132
# Check whether this custom merge logic should be used.
134
# OTHER is a straight winner, rely on default merge.
135
params.winner == 'other' or
136
# THIS and OTHER aren't both files.
137
not params.is_file_merge() or
138
# The filename doesn't match
139
not self.file_matches(params)):
140
return 'not_applicable', None
141
return self.merge_matching(params)
143
def merge_matching(self, params):
144
"""Merge the contents of a single file that has matched the criteria
145
in PerFileMerger.merge_contents (is a conflict, is a file,
146
self.file_matches is True).
148
Subclasses must override this.
150
raise NotImplementedError(self.merge_matching)
153
class ConfigurableFileMerger(PerFileMerger):
154
"""Merge individual files when configured via a .conf file.
156
This is a base class for concrete custom file merging logic. Concrete
157
classes should implement ``merge_text``.
159
See ``breezy.plugins.news_merge.news_merge`` for an example concrete class.
161
:ivar affected_files: The configured file paths to merge.
163
:cvar name_prefix: The prefix to use when looking up configuration
164
details. <name_prefix>_merge_files describes the files targeted by the
167
:cvar default_files: The default file paths to merge when no configuration
174
def __init__(self, merger):
175
super(ConfigurableFileMerger, self).__init__(merger)
176
self.affected_files = None
177
self.default_files = self.__class__.default_files or []
178
self.name_prefix = self.__class__.name_prefix
179
if self.name_prefix is None:
180
raise ValueError("name_prefix must be set.")
182
def file_matches(self, params):
183
"""Check whether the file should call the merge hook.
185
<name_prefix>_merge_files configuration variable is a list of files
186
that should use the hook.
188
affected_files = self.affected_files
189
if affected_files is None:
190
config = self.merger.this_branch.get_config()
191
# Until bzr provides a better policy for caching the config, we
192
# just add the part we're interested in to the params to avoid
193
# reading the config files repeatedly (breezy.conf, location.conf,
195
config_key = self.name_prefix + '_merge_files'
196
affected_files = config.get_user_option_as_list(config_key)
197
if affected_files is None:
198
# If nothing was specified in the config, use the default.
199
affected_files = self.default_files
200
self.affected_files = affected_files
202
filepath = params.this_path
203
if filepath in affected_files:
207
def merge_matching(self, params):
208
return self.merge_text(params)
210
def merge_text(self, params):
211
"""Merge the byte contents of a single file.
213
This is called after checking that the merge should be performed in
214
merge_contents, and it should behave as per
215
``breezy.merge.AbstractPerFileMerger.merge_contents``.
217
raise NotImplementedError(self.merge_text)
220
class MergeFileHookParams(object):
221
"""Object holding parameters passed to merge_file_content hooks.
223
There are some fields hooks can access:
225
:ivar file_id: the file ID of the file being merged
226
:ivar base_path: Path in base tree
227
:ivar other_path: Path in other tree
228
:ivar this_path: Path in this tree
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, paths, trans_id, this_kind, other_kind,
237
self._merger = merger
238
self.file_id = file_id
240
self.base_path, self.other_path, self.this_path = paths
241
self.trans_id = trans_id
242
self.this_kind = this_kind
243
self.other_kind = other_kind
246
def is_file_merge(self):
247
"""True if this_kind and other_kind are both 'file'."""
248
return self.this_kind == 'file' and self.other_kind == 'file'
250
@decorators.cachedproperty
251
def base_lines(self):
252
"""The lines of the 'base' version of the file."""
253
return self._merger.get_lines(self._merger.base_tree, self.base_path, self.file_id)
255
@decorators.cachedproperty
256
def this_lines(self):
257
"""The lines of the 'this' version of the file."""
258
return self._merger.get_lines(self._merger.this_tree, self.this_path, self.file_id)
260
@decorators.cachedproperty
261
def other_lines(self):
262
"""The lines of the 'other' version of the file."""
263
return self._merger.get_lines(self._merger.other_tree, self.other_path, self.file_id)
266
class Merger(object):
270
def __init__(self, this_branch, other_tree=None, base_tree=None,
271
this_tree=None, change_reporter=None,
272
recurse='down', revision_graph=None):
273
object.__init__(self)
274
self.this_branch = this_branch
275
self.this_basis = _mod_revision.ensure_null(
276
this_branch.last_revision())
277
self.this_rev_id = None
278
self.this_tree = this_tree
279
self.this_revision_tree = None
280
self.this_basis_tree = None
281
self.other_tree = other_tree
282
self.other_branch = None
283
self.base_tree = base_tree
284
self.ignore_zero = False
285
self.backup_files = False
286
self.interesting_files = None
287
self.show_base = False
288
self.reprocess = False
290
self.recurse = recurse
291
self.change_reporter = change_reporter
292
self._cached_trees = {}
293
self._revision_graph = revision_graph
294
self._base_is_ancestor = None
295
self._base_is_other_ancestor = None
296
self._is_criss_cross = None
297
self._lca_trees = None
299
def cache_trees_with_revision_ids(self, trees):
300
"""Cache any tree in trees if it has a revision_id."""
301
for maybe_tree in trees:
302
if maybe_tree is None:
305
rev_id = maybe_tree.get_revision_id()
306
except AttributeError:
308
self._cached_trees[rev_id] = maybe_tree
311
def revision_graph(self):
312
if self._revision_graph is None:
313
self._revision_graph = self.this_branch.repository.get_graph()
314
return self._revision_graph
316
def _set_base_is_ancestor(self, value):
317
self._base_is_ancestor = value
319
def _get_base_is_ancestor(self):
320
if self._base_is_ancestor is None:
321
self._base_is_ancestor = self.revision_graph.is_ancestor(
322
self.base_rev_id, self.this_basis)
323
return self._base_is_ancestor
325
base_is_ancestor = property(_get_base_is_ancestor, _set_base_is_ancestor)
327
def _set_base_is_other_ancestor(self, value):
328
self._base_is_other_ancestor = value
330
def _get_base_is_other_ancestor(self):
331
if self._base_is_other_ancestor is None:
332
if self.other_basis is None:
334
self._base_is_other_ancestor = self.revision_graph.is_ancestor(
335
self.base_rev_id, self.other_basis)
336
return self._base_is_other_ancestor
338
base_is_other_ancestor = property(_get_base_is_other_ancestor,
339
_set_base_is_other_ancestor)
342
def from_uncommitted(tree, other_tree, base_tree=None):
343
"""Return a Merger for uncommitted changes in other_tree.
345
:param tree: The tree to merge into
346
:param other_tree: The tree to get uncommitted changes from
347
:param base_tree: The basis to use for the merge. If unspecified,
348
other_tree.basis_tree() will be used.
350
if base_tree is None:
351
base_tree = other_tree.basis_tree()
352
merger = Merger(tree.branch, other_tree, base_tree, tree)
353
merger.base_rev_id = merger.base_tree.get_revision_id()
354
merger.other_rev_id = None
355
merger.other_basis = merger.base_rev_id
359
def from_mergeable(klass, tree, mergeable):
360
"""Return a Merger for a bundle or merge directive.
362
:param tree: The tree to merge changes into
363
:param mergeable: A merge directive or bundle
365
mergeable.install_revisions(tree.branch.repository)
366
base_revision_id, other_revision_id, verified =\
367
mergeable.get_merge_request(tree.branch.repository)
368
revision_graph = tree.branch.repository.get_graph()
369
if base_revision_id is not None:
370
if (base_revision_id != _mod_revision.NULL_REVISION and
371
revision_graph.is_ancestor(
372
base_revision_id, tree.branch.last_revision())):
373
base_revision_id = None
375
trace.warning('Performing cherrypick')
376
merger = klass.from_revision_ids(tree, other_revision_id,
377
base_revision_id, revision_graph=revision_graph)
378
return merger, verified
381
def from_revision_ids(tree, other, base=None, other_branch=None,
382
base_branch=None, revision_graph=None,
384
"""Return a Merger for revision-ids.
386
:param tree: The tree to merge changes into
387
:param other: The revision-id to use as OTHER
388
:param base: The revision-id to use as BASE. If not specified, will
390
:param other_branch: A branch containing the other revision-id. If
391
not supplied, tree.branch is used.
392
:param base_branch: A branch containing the base revision-id. If
393
not supplied, other_branch or tree.branch will be used.
394
:param revision_graph: If you have a revision_graph precomputed, pass
395
it in, otherwise it will be created for you.
396
:param tree_branch: The branch associated with tree. If not supplied,
397
tree.branch will be used.
399
if tree_branch is None:
400
tree_branch = tree.branch
401
merger = Merger(tree_branch, this_tree=tree,
402
revision_graph=revision_graph)
403
if other_branch is None:
404
other_branch = tree.branch
405
merger.set_other_revision(other, other_branch)
409
if base_branch is None:
410
base_branch = other_branch
411
merger.set_base_revision(base, base_branch)
414
def revision_tree(self, revision_id, branch=None):
415
if revision_id not in self._cached_trees:
417
branch = self.this_branch
419
tree = self.this_tree.revision_tree(revision_id)
420
except errors.NoSuchRevisionInTree:
421
tree = branch.repository.revision_tree(revision_id)
422
self._cached_trees[revision_id] = tree
423
return self._cached_trees[revision_id]
425
def _get_tree(self, treespec, possible_transports=None):
426
location, revno = treespec
428
tree = workingtree.WorkingTree.open_containing(location)[0]
429
return tree.branch, tree
430
branch = _mod_branch.Branch.open_containing(
431
location, possible_transports)[0]
433
revision_id = branch.last_revision()
435
revision_id = branch.get_rev_id(revno)
436
revision_id = _mod_revision.ensure_null(revision_id)
437
return branch, self.revision_tree(revision_id, branch)
439
def set_interesting_files(self, file_list):
440
self.interesting_files = file_list
442
def set_pending(self):
443
if (not self.base_is_ancestor or not self.base_is_other_ancestor
444
or self.other_rev_id is None):
448
def _add_parent(self):
449
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
450
new_parent_trees = []
451
operation = cleanup.OperationWithCleanups(
452
self.this_tree.set_parent_trees)
453
for revision_id in new_parents:
455
tree = self.revision_tree(revision_id)
456
except errors.NoSuchRevision:
460
operation.add_cleanup(tree.unlock)
461
new_parent_trees.append((revision_id, tree))
462
operation.run_simple(new_parent_trees, allow_leftmost_as_ghost=True)
464
def set_other(self, other_revision, possible_transports=None):
465
"""Set the revision and tree to merge from.
467
This sets the other_tree, other_rev_id, other_basis attributes.
469
:param other_revision: The [path, revision] list to merge from.
471
self.other_branch, self.other_tree = self._get_tree(other_revision,
473
if other_revision[1] == -1:
474
self.other_rev_id = _mod_revision.ensure_null(
475
self.other_branch.last_revision())
476
if _mod_revision.is_null(self.other_rev_id):
477
raise errors.NoCommits(self.other_branch)
478
self.other_basis = self.other_rev_id
479
elif other_revision[1] is not None:
480
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
481
self.other_basis = self.other_rev_id
483
self.other_rev_id = None
484
self.other_basis = self.other_branch.last_revision()
485
if self.other_basis is None:
486
raise errors.NoCommits(self.other_branch)
487
if self.other_rev_id is not None:
488
self._cached_trees[self.other_rev_id] = self.other_tree
489
self._maybe_fetch(self.other_branch,
490
self.this_branch, self.other_basis)
492
def set_other_revision(self, revision_id, other_branch):
493
"""Set 'other' based on a branch and revision id
495
:param revision_id: The revision to use for a tree
496
:param other_branch: The branch containing this tree
498
self.other_rev_id = revision_id
499
self.other_branch = other_branch
500
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
501
self.other_tree = self.revision_tree(revision_id)
502
self.other_basis = revision_id
504
def set_base_revision(self, revision_id, branch):
505
"""Set 'base' based on a branch and revision id
507
:param revision_id: The revision to use for a tree
508
:param branch: The branch containing this tree
510
self.base_rev_id = revision_id
511
self.base_branch = branch
512
self._maybe_fetch(branch, self.this_branch, revision_id)
513
self.base_tree = self.revision_tree(revision_id)
515
def _maybe_fetch(self, source, target, revision_id):
516
if not source.repository.has_same_location(target.repository):
517
target.fetch(source, revision_id)
520
revisions = [_mod_revision.ensure_null(self.this_basis),
521
_mod_revision.ensure_null(self.other_basis)]
522
if _mod_revision.NULL_REVISION in revisions:
523
self.base_rev_id = _mod_revision.NULL_REVISION
524
self.base_tree = self.revision_tree(self.base_rev_id)
525
self._is_criss_cross = False
527
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
528
self._is_criss_cross = False
530
self.base_rev_id = _mod_revision.NULL_REVISION
532
self.base_rev_id = list(lcas)[0]
533
else: # len(lcas) > 1
534
self._is_criss_cross = True
536
# find_unique_lca can only handle 2 nodes, so we have to
537
# start back at the beginning. It is a shame to traverse
538
# the graph again, but better than re-implementing
540
self.base_rev_id = self.revision_graph.find_unique_lca(
541
revisions[0], revisions[1])
543
self.base_rev_id = self.revision_graph.find_unique_lca(
545
sorted_lca_keys = self.revision_graph.find_merge_order(
547
if self.base_rev_id == _mod_revision.NULL_REVISION:
548
self.base_rev_id = sorted_lca_keys[0]
550
if self.base_rev_id == _mod_revision.NULL_REVISION:
551
raise errors.UnrelatedBranches()
552
if self._is_criss_cross:
553
trace.warning('Warning: criss-cross merge encountered. See bzr'
554
' help criss-cross.')
555
trace.mutter('Criss-cross lcas: %r' % lcas)
556
if self.base_rev_id in lcas:
557
trace.mutter('Unable to find unique lca. '
558
'Fallback %r as best option.'
560
interesting_revision_ids = set(lcas)
561
interesting_revision_ids.add(self.base_rev_id)
562
interesting_trees = dict((t.get_revision_id(), t)
563
for t in self.this_branch.repository.revision_trees(
564
interesting_revision_ids))
565
self._cached_trees.update(interesting_trees)
566
if self.base_rev_id in lcas:
567
self.base_tree = interesting_trees[self.base_rev_id]
569
self.base_tree = interesting_trees.pop(self.base_rev_id)
570
self._lca_trees = [interesting_trees[key]
571
for key in sorted_lca_keys]
573
self.base_tree = self.revision_tree(self.base_rev_id)
574
self.base_is_ancestor = True
575
self.base_is_other_ancestor = True
576
trace.mutter('Base revid: %r' % self.base_rev_id)
578
def set_base(self, base_revision):
579
"""Set the base revision to use for the merge.
581
:param base_revision: A 2-list containing a path and revision number.
583
trace.mutter("doing merge() with no base_revision specified")
584
if base_revision == [None, None]:
587
base_branch, self.base_tree = self._get_tree(base_revision)
588
if base_revision[1] == -1:
589
self.base_rev_id = base_branch.last_revision()
590
elif base_revision[1] is None:
591
self.base_rev_id = _mod_revision.NULL_REVISION
593
self.base_rev_id = _mod_revision.ensure_null(
594
base_branch.get_rev_id(base_revision[1]))
595
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
597
def make_merger(self):
598
kwargs = {'working_tree': self.this_tree, 'this_tree': self.this_tree,
599
'other_tree': self.other_tree,
600
'interesting_files': self.interesting_files,
601
'this_branch': self.this_branch,
602
'other_branch': self.other_branch,
604
if self.merge_type.requires_base:
605
kwargs['base_tree'] = self.base_tree
606
if self.merge_type.supports_reprocess:
607
kwargs['reprocess'] = self.reprocess
609
raise errors.BzrError(
610
"Conflict reduction is not supported for merge"
611
" type %s." % self.merge_type)
612
if self.merge_type.supports_show_base:
613
kwargs['show_base'] = self.show_base
615
raise errors.BzrError("Showing base is not supported for this"
616
" merge type. %s" % self.merge_type)
617
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
618
and not self.base_is_other_ancestor):
619
raise errors.CannotReverseCherrypick()
620
if self.merge_type.supports_cherrypick:
621
kwargs['cherrypick'] = (not self.base_is_ancestor or
622
not self.base_is_other_ancestor)
623
if self._is_criss_cross and getattr(self.merge_type,
624
'supports_lca_trees', False):
625
kwargs['lca_trees'] = self._lca_trees
626
return self.merge_type(change_reporter=self.change_reporter,
629
def _do_merge_to(self):
630
merge = self.make_merger()
631
if self.other_branch is not None:
632
self.other_branch.update_references(self.this_branch)
633
for hook in Merger.hooks['pre_merge']:
636
for hook in Merger.hooks['post_merge']:
638
if self.recurse == 'down':
639
for relpath, file_id in self.this_tree.iter_references():
640
sub_tree = self.this_tree.get_nested_tree(relpath)
641
other_revision = self.other_tree.get_reference_revision(
643
if other_revision == sub_tree.last_revision():
645
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
646
sub_merge.merge_type = self.merge_type
647
other_branch = self.other_branch.reference_parent(
649
sub_merge.set_other_revision(other_revision, other_branch)
650
base_tree_path = _mod_tree.find_previous_path(
651
self.this_tree, self.base_tree, relpath)
652
base_revision = self.base_tree.get_reference_revision(
654
sub_merge.base_tree = \
655
sub_tree.branch.repository.revision_tree(base_revision)
656
sub_merge.base_rev_id = base_revision
661
operation = cleanup.OperationWithCleanups(self._do_merge_to)
662
self.this_tree.lock_tree_write()
663
operation.add_cleanup(self.this_tree.unlock)
664
if self.base_tree is not None:
665
self.base_tree.lock_read()
666
operation.add_cleanup(self.base_tree.unlock)
667
if self.other_tree is not None:
668
self.other_tree.lock_read()
669
operation.add_cleanup(self.other_tree.unlock)
670
merge = operation.run_simple()
671
if len(merge.cooked_conflicts) == 0:
672
if not self.ignore_zero and not trace.is_quiet():
673
trace.note(gettext("All changes applied successfully."))
675
trace.note(gettext("%d conflicts encountered.")
676
% len(merge.cooked_conflicts))
678
return len(merge.cooked_conflicts)
681
class _InventoryNoneEntry(object):
682
"""This represents an inventory entry which *isn't there*.
684
It simplifies the merging logic if we always have an InventoryEntry, even
685
if it isn't actually present
692
symlink_target = None
696
_none_entry = _InventoryNoneEntry()
699
class Merge3Merger(object):
700
"""Three-way merger that uses the merge3 text merger"""
702
supports_reprocess = True
703
supports_show_base = True
704
history_based = False
705
supports_cherrypick = True
706
supports_reverse_cherrypick = True
707
winner_idx = {"this": 2, "other": 1, "conflict": 1}
708
supports_lca_trees = True
709
requires_file_merge_plan = False
711
def __init__(self, working_tree, this_tree, base_tree, other_tree,
712
reprocess=False, show_base=False,
713
change_reporter=None, interesting_files=None, do_merge=True,
714
cherrypick=False, lca_trees=None, this_branch=None,
716
"""Initialize the merger object and perform the merge.
718
:param working_tree: The working tree to apply the merge to
719
:param this_tree: The local tree in the merge operation
720
:param base_tree: The common tree in the merge operation
721
:param other_tree: The other tree to merge changes from
722
:param this_branch: The branch associated with this_tree. Defaults to
723
this_tree.branch if not supplied.
724
:param other_branch: The branch associated with other_tree, if any.
725
:param: reprocess If True, perform conflict-reduction processing.
726
:param show_base: If True, show the base revision in text conflicts.
727
(incompatible with reprocess)
728
:param change_reporter: An object that should report changes made
729
:param interesting_files: The tree-relative paths of files that should
730
participate in the merge. If these paths refer to directories,
731
the contents of those directories will also be included. If not
732
specified, all files may participate in the
734
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
735
if the ancestry was found to include a criss-cross merge.
736
Otherwise should be None.
738
object.__init__(self)
739
if this_branch is None:
740
this_branch = this_tree.branch
741
self.interesting_files = interesting_files
742
self.working_tree = working_tree
743
self.this_tree = this_tree
744
self.base_tree = base_tree
745
self.other_tree = other_tree
746
self.this_branch = this_branch
747
self.other_branch = other_branch
748
self._raw_conflicts = []
749
self.cooked_conflicts = []
750
self.reprocess = reprocess
751
self.show_base = show_base
752
self._lca_trees = lca_trees
753
# Uncommenting this will change the default algorithm to always use
754
# _entries_lca. This can be useful for running the test suite and
755
# making sure we haven't missed any corner cases.
756
# if lca_trees is None:
757
# self._lca_trees = [self.base_tree]
758
self.change_reporter = change_reporter
759
self.cherrypick = cherrypick
764
operation = cleanup.OperationWithCleanups(self._do_merge)
765
self.working_tree.lock_tree_write()
766
operation.add_cleanup(self.working_tree.unlock)
767
self.this_tree.lock_read()
768
operation.add_cleanup(self.this_tree.unlock)
769
self.base_tree.lock_read()
770
operation.add_cleanup(self.base_tree.unlock)
771
self.other_tree.lock_read()
772
operation.add_cleanup(self.other_tree.unlock)
775
def _do_merge(self, operation):
776
self.tt = transform.TreeTransform(self.working_tree, None)
777
operation.add_cleanup(self.tt.finalize)
778
self._compute_transform()
779
results = self.tt.apply(no_conflicts=True)
780
self.write_modified(results)
782
self.working_tree.add_conflicts(self.cooked_conflicts)
783
except errors.UnsupportedOperation:
786
def make_preview_transform(self):
787
with self.base_tree.lock_read(), self.other_tree.lock_read():
788
self.tt = transform.TransformPreview(self.working_tree)
789
self._compute_transform()
792
def _compute_transform(self):
793
if self._lca_trees is None:
794
entries = self._entries3()
795
resolver = self._three_way
797
entries = self._entries_lca()
798
resolver = self._lca_multi_way
799
# Prepare merge hooks
800
factories = Merger.hooks['merge_file_content']
801
# One hook for each registered one plus our default merger
802
hooks = [factory(self) for factory in factories] + [self]
803
self.active_hooks = [hook for hook in hooks if hook is not None]
804
with ui.ui_factory.nested_progress_bar() as child_pb:
805
for num, (file_id, changed, paths3, parents3, names3,
806
executable3) in enumerate(entries):
807
# Try merging each entry
808
child_pb.update(gettext('Preparing file merge'),
810
self._merge_names(file_id, paths3, parents3,
811
names3, resolver=resolver)
813
file_status = self._do_merge_contents(paths3, file_id)
815
file_status = 'unmodified'
816
self._merge_executable(paths3, file_id, executable3,
817
file_status, resolver=resolver)
818
self.tt.fixup_new_roots()
819
self._finish_computing_transform()
821
def _finish_computing_transform(self):
822
"""Finalize the transform and report the changes.
824
This is the second half of _compute_transform.
826
with ui.ui_factory.nested_progress_bar() as child_pb:
827
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
828
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
829
if self.change_reporter is not None:
830
from breezy import delta
831
delta.report_changes(
832
self.tt.iter_changes(), self.change_reporter)
833
self.cook_conflicts(fs_conflicts)
834
for conflict in self.cooked_conflicts:
835
trace.warning('%s', conflict.describe())
838
"""Gather data about files modified between three trees.
840
Return a list of tuples of file_id, changed, parents3, names3,
841
executable3. changed is a boolean indicating whether the file contents
842
or kind were changed. parents3 is a tuple of parent ids for base,
843
other and this. names3 is a tuple of names for base, other and this.
844
executable3 is a tuple of execute-bit values for base, other and this.
847
iterator = self.other_tree.iter_changes(self.base_tree,
848
specific_files=self.interesting_files,
849
extra_trees=[self.this_tree])
850
this_interesting_files = self.this_tree.find_related_paths_across_trees(
851
self.interesting_files, trees=[self.other_tree])
852
this_entries = dict(self.this_tree.iter_entries_by_dir(
853
specific_files=this_interesting_files))
854
for (file_id, paths, changed, versioned, parents, names, kind,
855
executable) in iterator:
856
if paths[0] is not None:
857
this_path = _mod_tree.find_previous_path(
858
self.base_tree, self.this_tree, paths[0])
860
this_path = _mod_tree.find_previous_path(
861
self.other_tree, self.this_tree, paths[1])
862
this_entry = this_entries.get(this_path)
863
if this_entry is not None:
864
this_name = this_entry.name
865
this_parent = this_entry.parent_id
866
this_executable = this_entry.executable
870
this_executable = None
871
parents3 = parents + (this_parent,)
872
names3 = names + (this_name,)
873
paths3 = paths + (this_path, )
874
executable3 = executable + (this_executable,)
875
result.append((file_id, changed, paths3,
876
parents3, names3, executable3))
879
def _entries_lca(self):
880
"""Gather data about files modified between multiple trees.
882
This compares OTHER versus all LCA trees, and for interesting entries,
883
it then compares with THIS and BASE.
885
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
887
:return: [(file_id, changed, paths, parents, names, executable)], where:
889
* file_id: Simple file_id of the entry
890
* changed: Boolean, True if the kind or contents changed else False
891
* paths: ((base, [path, in, lcas]), path_other, path_this)
892
* parents: ((base, [parent_id, in, lcas]), parent_id_other,
894
* names: ((base, [name, in, lcas]), name_in_other, name_in_this)
895
* executable: ((base, [exec, in, lcas]), exec_in_other,
898
if self.interesting_files is not None:
899
lookup_trees = [self.this_tree, self.base_tree]
900
lookup_trees.extend(self._lca_trees)
901
# I think we should include the lca trees as well
902
interesting_files = self.other_tree.find_related_paths_across_trees(
903
self.interesting_files, lookup_trees)
905
interesting_files = None
907
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
909
base_inventory = self.base_tree.root_inventory
910
this_inventory = self.this_tree.root_inventory
911
for path, file_id, other_ie, lca_values in walker.iter_all():
912
# Is this modified at all from any of the other trees?
914
other_ie = _none_entry
917
other_path = self.other_tree.id2path(file_id)
918
if interesting_files is not None and other_path not in interesting_files:
921
# If other_revision is found in any of the lcas, that means this
922
# node is uninteresting. This is because when merging, if there are
923
# multiple heads(), we have to create a new node. So if we didn't,
924
# we know that the ancestry is linear, and that OTHER did not
926
# See doc/developers/lca_merge_resolution.txt for details
927
other_revision = other_ie.revision
928
if other_revision is not None:
929
# We can't use this shortcut when other_revision is None,
930
# because it may be None because things are WorkingTrees, and
931
# not because it is *actually* None.
932
is_unmodified = False
933
for lca_path, ie in lca_values:
934
if ie is not None and ie.revision == other_revision:
942
for lca_path, lca_ie in lca_values:
944
lca_entries.append(_none_entry)
945
lca_paths.append(None)
947
lca_entries.append(lca_ie)
948
lca_paths.append(lca_path)
951
base_ie = base_inventory.get_entry(file_id)
952
except errors.NoSuchId:
953
base_ie = _none_entry
956
base_path = self.base_tree.id2path(file_id)
959
this_ie = this_inventory.get_entry(file_id)
960
except errors.NoSuchId:
961
this_ie = _none_entry
964
this_path = self.this_tree.id2path(file_id)
970
for lca_ie in lca_entries:
971
lca_kinds.append(lca_ie.kind)
972
lca_parent_ids.append(lca_ie.parent_id)
973
lca_names.append(lca_ie.name)
974
lca_executable.append(lca_ie.executable)
976
kind_winner = self._lca_multi_way(
977
(base_ie.kind, lca_kinds),
978
other_ie.kind, this_ie.kind)
979
parent_id_winner = self._lca_multi_way(
980
(base_ie.parent_id, lca_parent_ids),
981
other_ie.parent_id, this_ie.parent_id)
982
name_winner = self._lca_multi_way(
983
(base_ie.name, lca_names),
984
other_ie.name, this_ie.name)
986
content_changed = True
987
if kind_winner == 'this':
988
# No kind change in OTHER, see if there are *any* changes
989
if other_ie.kind == 'directory':
990
if parent_id_winner == 'this' and name_winner == 'this':
991
# No change for this directory in OTHER, skip
993
content_changed = False
994
elif other_ie.kind is None or other_ie.kind == 'file':
995
def get_sha1(tree, path):
999
return tree.get_file_sha1(path)
1000
except errors.NoSuchFile:
1002
base_sha1 = get_sha1(self.base_tree, base_path)
1003
lca_sha1s = [get_sha1(tree, lca_path)
1005
in zip(self._lca_trees, lca_paths)]
1006
this_sha1 = get_sha1(self.this_tree, this_path)
1007
other_sha1 = get_sha1(self.other_tree, other_path)
1008
sha1_winner = self._lca_multi_way(
1009
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1010
allow_overriding_lca=False)
1011
exec_winner = self._lca_multi_way(
1012
(base_ie.executable, lca_executable),
1013
other_ie.executable, this_ie.executable)
1014
if (parent_id_winner == 'this' and name_winner == 'this'
1015
and sha1_winner == 'this' and exec_winner == 'this'):
1016
# No kind, parent, name, exec, or content change for
1017
# OTHER, so this node is not considered interesting
1019
if sha1_winner == 'this':
1020
content_changed = False
1021
elif other_ie.kind == 'symlink':
1022
def get_target(ie, tree, path):
1023
if ie.kind != 'symlink':
1025
return tree.get_symlink_target(path)
1026
base_target = get_target(base_ie, self.base_tree, base_path)
1027
lca_targets = [get_target(ie, tree, lca_path) for ie, tree, lca_path
1028
in zip(lca_entries, self._lca_trees, lca_paths)]
1029
this_target = get_target(
1030
this_ie, self.this_tree, this_path)
1031
other_target = get_target(
1032
other_ie, self.other_tree, other_path)
1033
target_winner = self._lca_multi_way(
1034
(base_target, lca_targets),
1035
other_target, this_target)
1036
if (parent_id_winner == 'this' and name_winner == 'this'
1037
and target_winner == 'this'):
1038
# No kind, parent, name, or symlink target change
1041
if target_winner == 'this':
1042
content_changed = False
1043
elif other_ie.kind == 'tree-reference':
1044
# The 'changed' information seems to be handled at a higher
1045
# level. At least, _entries3 returns False for content
1046
# changed, even when at a new revision_id.
1047
content_changed = False
1048
if (parent_id_winner == 'this' and name_winner == 'this'):
1049
# Nothing interesting
1052
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1054
# If we have gotten this far, that means something has changed
1055
result.append((file_id, content_changed,
1056
((base_path, lca_paths),
1057
other_path, this_path),
1058
((base_ie.parent_id, lca_parent_ids),
1059
other_ie.parent_id, this_ie.parent_id),
1060
((base_ie.name, lca_names),
1061
other_ie.name, this_ie.name),
1062
((base_ie.executable, lca_executable),
1063
other_ie.executable, this_ie.executable)
1067
def write_modified(self, results):
1068
if not self.working_tree.supports_merge_modified():
1070
modified_hashes = {}
1071
for path in results.modified_paths:
1072
wt_relpath = self.working_tree.relpath(path)
1073
file_id = self.working_tree.path2id(wt_relpath)
1076
hash = self.working_tree.get_file_sha1(wt_relpath)
1079
modified_hashes[file_id] = hash
1080
self.working_tree.set_merge_modified(modified_hashes)
1083
def parent(entry, file_id):
1084
"""Determine the parent for a file_id (used as a key method)"""
1087
return entry.parent_id
1090
def name(entry, file_id):
1091
"""Determine the name for a file_id (used as a key method)"""
1097
def contents_sha1(tree, path, file_id=None):
1098
"""Determine the sha1 of the file contents (used as a key method)."""
1100
return tree.get_file_sha1(path, file_id)
1101
except errors.NoSuchFile:
1105
def executable(tree, path, file_id=None):
1106
"""Determine the executability of a file-id (used as a key method)."""
1108
if tree.kind(path, file_id) != "file":
1110
except errors.NoSuchFile:
1112
return tree.is_executable(path)
1115
def kind(tree, path, file_id=None):
1116
"""Determine the kind of a file-id (used as a key method)."""
1118
return tree.kind(path, file_id)
1119
except errors.NoSuchFile:
1123
def _three_way(base, other, this):
1125
# if 'base == other', either they all agree, or only 'this' has
1128
elif this not in (base, other):
1129
# 'this' is neither 'base' nor 'other', so both sides changed
1132
# "Ambiguous clean merge" -- both sides have made the same change.
1135
# this == base: only other has changed.
1139
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
1140
"""Consider LCAs when determining whether a change has occurred.
1142
If LCAS are all identical, this is the same as a _three_way comparison.
1144
:param bases: value in (BASE, [LCAS])
1145
:param other: value in OTHER
1146
:param this: value in THIS
1147
:param allow_overriding_lca: If there is more than one unique lca
1148
value, allow OTHER to override THIS if it has a new value, and
1149
THIS only has an lca value, or vice versa. This is appropriate for
1150
truly scalar values, not as much for non-scalars.
1151
:return: 'this', 'other', or 'conflict' depending on whether an entry
1154
# See doc/developers/lca_tree_merging.txt for details about this
1157
# Either Ambiguously clean, or nothing was actually changed. We
1160
base_val, lca_vals = bases
1161
# Remove 'base_val' from the lca_vals, because it is not interesting
1162
filtered_lca_vals = [lca_val for lca_val in lca_vals
1163
if lca_val != base_val]
1164
if len(filtered_lca_vals) == 0:
1165
return Merge3Merger._three_way(base_val, other, this)
1167
unique_lca_vals = set(filtered_lca_vals)
1168
if len(unique_lca_vals) == 1:
1169
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1171
if allow_overriding_lca:
1172
if other in unique_lca_vals:
1173
if this in unique_lca_vals:
1174
# Each side picked a different lca, conflict
1177
# This has a value which supersedes both lca values, and
1178
# other only has an lca value
1180
elif this in unique_lca_vals:
1181
# OTHER has a value which supersedes both lca values, and this
1182
# only has an lca value
1185
# At this point, the lcas disagree, and the tip disagree
1188
def _merge_names(self, file_id, paths, parents, names, resolver):
1189
"""Perform a merge on file_id names and parents"""
1190
base_name, other_name, this_name = names
1191
base_parent, other_parent, this_parent = parents
1192
unused_base_path, other_path, this_path = paths
1194
name_winner = resolver(*names)
1196
parent_id_winner = resolver(*parents)
1197
if this_name is None:
1198
if name_winner == "this":
1199
name_winner = "other"
1200
if parent_id_winner == "this":
1201
parent_id_winner = "other"
1202
if name_winner == "this" and parent_id_winner == "this":
1204
if name_winner == 'conflict' or parent_id_winner == 'conflict':
1205
# Creating helpers (.OTHER or .THIS) here cause problems down the
1206
# road if a ContentConflict needs to be created so we should not do
1208
trans_id = self.tt.trans_id_file_id(file_id)
1209
self._raw_conflicts.append(('path conflict', trans_id, file_id,
1210
this_parent, this_name,
1211
other_parent, other_name))
1212
if other_path is None:
1213
# it doesn't matter whether the result was 'other' or
1214
# 'conflict'-- if it has no file id, we leave it alone.
1216
parent_id = parents[self.winner_idx[parent_id_winner]]
1217
name = names[self.winner_idx[name_winner]]
1218
if parent_id is not None or name is not None:
1219
# if we get here, name_winner and parent_winner are set to safe
1221
if parent_id is None and name is not None:
1222
# if parent_id is None and name is non-None, current file is
1224
if names[self.winner_idx[parent_id_winner]] != '':
1225
raise AssertionError(
1226
'File looks like a root, but named %s' %
1227
names[self.winner_idx[parent_id_winner]])
1228
parent_trans_id = transform.ROOT_PARENT
1230
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1231
self.tt.adjust_path(name, parent_trans_id,
1232
self.tt.trans_id_file_id(file_id))
1234
def _do_merge_contents(self, paths, file_id):
1235
"""Performs a merge on file_id contents."""
1236
def contents_pair(tree, path):
1240
kind = tree.kind(path)
1241
except errors.NoSuchFile:
1244
contents = tree.get_file_sha1(path)
1245
elif kind == "symlink":
1246
contents = tree.get_symlink_target(path)
1249
return kind, contents
1251
base_path, other_path, this_path = paths
1252
# See SPOT run. run, SPOT, run.
1253
# So we're not QUITE repeating ourselves; we do tricky things with
1255
other_pair = contents_pair(self.other_tree, other_path)
1256
this_pair = contents_pair(self.this_tree, this_path)
1258
(base_path, lca_paths) = base_path
1259
base_pair = contents_pair(self.base_tree, base_path)
1260
lca_pairs = [contents_pair(tree, path)
1261
for tree, path in zip(self._lca_trees, lca_paths)]
1262
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1263
this_pair, allow_overriding_lca=False)
1265
base_pair = contents_pair(self.base_tree, base_path)
1266
if base_pair == other_pair:
1269
# We delayed evaluating this_pair as long as we can to avoid
1270
# unnecessary sha1 calculation
1271
this_pair = contents_pair(self.this_tree, this_path)
1272
winner = self._three_way(base_pair, other_pair, this_pair)
1273
if winner == 'this':
1274
# No interesting changes introduced by OTHER
1276
# We have a hypothetical conflict, but if we have files, then we
1277
# can try to merge the content
1278
trans_id = self.tt.trans_id_file_id(file_id)
1279
params = MergeFileHookParams(
1280
self, file_id, (base_path, other_path,
1281
this_path), trans_id, this_pair[0],
1282
other_pair[0], winner)
1283
hooks = self.active_hooks
1284
hook_status = 'not_applicable'
1286
hook_status, lines = hook.merge_contents(params)
1287
if hook_status != 'not_applicable':
1288
# Don't try any more hooks, this one applies.
1290
# If the merge ends up replacing the content of the file, we get rid of
1291
# it at the end of this method (this variable is used to track the
1292
# exceptions to this rule).
1295
if hook_status == 'not_applicable':
1296
# No merge hook was able to resolve the situation. Two cases exist:
1297
# a content conflict or a duplicate one.
1299
name = self.tt.final_name(trans_id)
1300
parent_id = self.tt.final_parent(trans_id)
1301
inhibit_content_conflict = False
1302
if params.this_kind is None: # file_id is not in THIS
1303
# Is the name used for a different file_id ?
1304
if self.this_tree.is_versioned(other_path):
1305
# Two entries for the same path
1307
# versioning the merged file will trigger a duplicate
1309
self.tt.version_file(file_id, trans_id)
1310
transform.create_from_tree(
1311
self.tt, trans_id, self.other_tree,
1312
other_path, file_id=file_id,
1313
filter_tree_path=self._get_filter_tree_path(file_id))
1314
inhibit_content_conflict = True
1315
elif params.other_kind is None: # file_id is not in OTHER
1316
# Is the name used for a different file_id ?
1317
if self.other_tree.is_versioned(this_path):
1318
# Two entries for the same path again, but here, the other
1319
# entry will also be merged. We simply inhibit the
1320
# 'content' conflict creation because we know OTHER will
1321
# create (or has already created depending on ordering) an
1322
# entry at the same path. This will trigger a 'duplicate'
1325
inhibit_content_conflict = True
1326
if not inhibit_content_conflict:
1327
if params.this_kind is not None:
1328
self.tt.unversion_file(trans_id)
1329
# This is a contents conflict, because none of the available
1330
# functions could merge it.
1331
file_group = self._dump_conflicts(
1332
name, (base_path, other_path, this_path), parent_id,
1333
file_id, set_version=True)
1334
self._raw_conflicts.append(('contents conflict', file_group))
1335
elif hook_status == 'success':
1336
self.tt.create_file(lines, trans_id)
1337
elif hook_status == 'conflicted':
1338
# XXX: perhaps the hook should be able to provide
1339
# the BASE/THIS/OTHER files?
1340
self.tt.create_file(lines, trans_id)
1341
self._raw_conflicts.append(('text conflict', trans_id))
1342
name = self.tt.final_name(trans_id)
1343
parent_id = self.tt.final_parent(trans_id)
1344
self._dump_conflicts(
1345
name, (base_path, other_path, this_path), parent_id, file_id)
1346
elif hook_status == 'delete':
1347
self.tt.unversion_file(trans_id)
1349
elif hook_status == 'done':
1350
# The hook function did whatever it needs to do directly, no
1351
# further action needed here.
1354
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1355
if not this_path and result == "modified":
1356
self.tt.version_file(file_id, trans_id)
1358
# The merge has been performed and produced a new content, so the
1359
# old contents should not be retained.
1360
self.tt.delete_contents(trans_id)
1363
def _default_other_winner_merge(self, merge_hook_params):
1364
"""Replace this contents with other."""
1365
file_id = merge_hook_params.file_id
1366
trans_id = merge_hook_params.trans_id
1367
if merge_hook_params.other_path is not None:
1368
# OTHER changed the file
1369
transform.create_from_tree(
1370
self.tt, trans_id, self.other_tree,
1371
merge_hook_params.other_path, file_id=file_id,
1372
filter_tree_path=self._get_filter_tree_path(file_id))
1374
elif merge_hook_params.this_path is not None:
1375
# OTHER deleted the file
1376
return 'delete', None
1378
raise AssertionError(
1379
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1382
def merge_contents(self, merge_hook_params):
1383
"""Fallback merge logic after user installed hooks."""
1384
# This function is used in merge hooks as the fallback instance.
1385
# Perhaps making this function and the functions it calls be a
1386
# a separate class would be better.
1387
if merge_hook_params.winner == 'other':
1388
# OTHER is a straight winner, so replace this contents with other
1389
return self._default_other_winner_merge(merge_hook_params)
1390
elif merge_hook_params.is_file_merge():
1391
# THIS and OTHER are both files, so text merge. Either
1392
# BASE is a file, or both converted to files, so at least we
1393
# have agreement that output should be a file.
1395
self.text_merge(merge_hook_params.trans_id,
1396
merge_hook_params.paths, merge_hook_params.file_id)
1397
except errors.BinaryFile:
1398
return 'not_applicable', None
1401
return 'not_applicable', None
1403
def get_lines(self, tree, path, file_id=None):
1404
"""Return the lines in a file, or an empty list."""
1408
kind = tree.kind(path)
1409
except errors.NoSuchFile:
1414
return tree.get_file_lines(path)
1416
def text_merge(self, trans_id, paths, file_id):
1417
"""Perform a three-way text merge on a file_id"""
1418
# it's possible that we got here with base as a different type.
1419
# if so, we just want two-way text conflicts.
1420
base_path, other_path, this_path = paths
1421
base_lines = self.get_lines(self.base_tree, base_path, file_id)
1422
other_lines = self.get_lines(self.other_tree, other_path, file_id)
1423
this_lines = self.get_lines(self.this_tree, this_path, file_id)
1424
m3 = merge3.Merge3(base_lines, this_lines, other_lines,
1425
is_cherrypick=self.cherrypick)
1426
start_marker = b"!START OF MERGE CONFLICT!" + b"I HOPE THIS IS UNIQUE"
1427
if self.show_base is True:
1428
base_marker = b'|' * 7
1432
def iter_merge3(retval):
1433
retval["text_conflicts"] = False
1434
for line in m3.merge_lines(name_a=b"TREE",
1435
name_b=b"MERGE-SOURCE",
1436
name_base=b"BASE-REVISION",
1437
start_marker=start_marker,
1438
base_marker=base_marker,
1439
reprocess=self.reprocess):
1440
if line.startswith(start_marker):
1441
retval["text_conflicts"] = True
1442
yield line.replace(start_marker, b'<' * 7)
1446
merge3_iterator = iter_merge3(retval)
1447
self.tt.create_file(merge3_iterator, trans_id)
1448
if retval["text_conflicts"] is True:
1449
self._raw_conflicts.append(('text conflict', trans_id))
1450
name = self.tt.final_name(trans_id)
1451
parent_id = self.tt.final_parent(trans_id)
1452
file_group = self._dump_conflicts(name, paths, parent_id, file_id,
1453
this_lines, base_lines,
1455
file_group.append(trans_id)
1457
def _get_filter_tree_path(self, file_id):
1458
if self.this_tree.supports_content_filtering():
1459
# We get the path from the working tree if it exists.
1460
# That fails though when OTHER is adding a file, so
1461
# we fall back to the other tree to find the path if
1462
# it doesn't exist locally.
1464
return self.this_tree.id2path(file_id)
1465
except errors.NoSuchId:
1466
return self.other_tree.id2path(file_id)
1467
# Skip the id2path lookup for older formats
1470
def _dump_conflicts(self, name, paths, parent_id, file_id, this_lines=None,
1471
base_lines=None, other_lines=None, set_version=False,
1473
"""Emit conflict files.
1474
If this_lines, base_lines, or other_lines are omitted, they will be
1475
determined automatically. If set_version is true, the .OTHER, .THIS
1476
or .BASE (in that order) will be created as versioned files.
1478
base_path, other_path, this_path = paths
1479
data = [('OTHER', self.other_tree, other_path, other_lines),
1480
('THIS', self.this_tree, this_path, this_lines)]
1482
data.append(('BASE', self.base_tree, base_path, base_lines))
1484
# We need to use the actual path in the working tree of the file here,
1485
# ignoring the conflict suffixes
1487
if wt.supports_content_filtering():
1489
filter_tree_path = wt.id2path(file_id)
1490
except errors.NoSuchId:
1491
# file has been deleted
1492
filter_tree_path = None
1494
# Skip the id2path lookup for older formats
1495
filter_tree_path = None
1499
for suffix, tree, path, lines in data:
1500
if path is not None:
1501
trans_id = self._conflict_file(
1502
name, parent_id, path, tree, file_id, suffix, lines,
1504
file_group.append(trans_id)
1505
if set_version and not versioned:
1506
self.tt.version_file(file_id, trans_id)
1510
def _conflict_file(self, name, parent_id, path, tree, file_id, suffix,
1511
lines=None, filter_tree_path=None):
1512
"""Emit a single conflict file."""
1513
name = name + '.' + suffix
1514
trans_id = self.tt.create_path(name, parent_id)
1515
transform.create_from_tree(
1516
self.tt, trans_id, tree, path,
1517
file_id=file_id, chunks=lines,
1518
filter_tree_path=filter_tree_path)
1521
def merge_executable(self, paths, file_id, file_status):
1522
"""Perform a merge on the execute bit."""
1523
executable = [self.executable(t, p, file_id)
1524
for t, p in zip([self.base_tree, self.other_tree, self.this_tree], paths)]
1525
self._merge_executable(paths, file_id, executable, file_status,
1526
resolver=self._three_way)
1528
def _merge_executable(self, paths, file_id, executable, file_status,
1530
"""Perform a merge on the execute bit."""
1531
base_executable, other_executable, this_executable = executable
1532
base_path, other_path, this_path = paths
1533
if file_status == "deleted":
1535
winner = resolver(*executable)
1536
if winner == "conflict":
1537
# There must be a None in here, if we have a conflict, but we
1538
# need executability since file status was not deleted.
1539
if other_path is None:
1543
if winner == 'this' and file_status != "modified":
1545
trans_id = self.tt.trans_id_file_id(file_id)
1546
if self.tt.final_kind(trans_id) != "file":
1548
if winner == "this":
1549
executability = this_executable
1551
if other_path is not None:
1552
executability = other_executable
1553
elif this_path is not None:
1554
executability = this_executable
1555
elif base_path is not None:
1556
executability = base_executable
1557
if executability is not None:
1558
trans_id = self.tt.trans_id_file_id(file_id)
1559
self.tt.set_executability(executability, trans_id)
1561
def cook_conflicts(self, fs_conflicts):
1562
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1563
content_conflict_file_ids = set()
1564
cooked_conflicts = transform.cook_conflicts(fs_conflicts, self.tt)
1565
fp = transform.FinalPaths(self.tt)
1566
for conflict in self._raw_conflicts:
1567
conflict_type = conflict[0]
1568
if conflict_type == 'path conflict':
1570
this_parent, this_name,
1571
other_parent, other_name) = conflict[1:]
1572
if this_parent is None or this_name is None:
1573
this_path = '<deleted>'
1575
parent_path = fp.get_path(
1576
self.tt.trans_id_file_id(this_parent))
1577
this_path = osutils.pathjoin(parent_path, this_name)
1578
if other_parent is None or other_name is None:
1579
other_path = '<deleted>'
1581
if other_parent == self.other_tree.get_root_id():
1582
# The tree transform doesn't know about the other root,
1583
# so we special case here to avoid a NoFinalPath
1587
parent_path = fp.get_path(
1588
self.tt.trans_id_file_id(other_parent))
1589
other_path = osutils.pathjoin(parent_path, other_name)
1590
c = _mod_conflicts.Conflict.factory(
1591
'path conflict', path=this_path,
1592
conflict_path=other_path,
1594
elif conflict_type == 'contents conflict':
1595
for trans_id in conflict[1]:
1596
file_id = self.tt.final_file_id(trans_id)
1597
if file_id is not None:
1598
# Ok we found the relevant file-id
1600
path = fp.get_path(trans_id)
1601
for suffix in ('.BASE', '.THIS', '.OTHER'):
1602
if path.endswith(suffix):
1603
# Here is the raw path
1604
path = path[:-len(suffix)]
1606
c = _mod_conflicts.Conflict.factory(conflict_type,
1607
path=path, file_id=file_id)
1608
content_conflict_file_ids.add(file_id)
1609
elif conflict_type == 'text conflict':
1610
trans_id = conflict[1]
1611
path = fp.get_path(trans_id)
1612
file_id = self.tt.final_file_id(trans_id)
1613
c = _mod_conflicts.Conflict.factory(conflict_type,
1614
path=path, file_id=file_id)
1616
raise AssertionError('bad conflict type: %r' % (conflict,))
1617
cooked_conflicts.append(c)
1619
self.cooked_conflicts = []
1620
# We want to get rid of path conflicts when a corresponding contents
1621
# conflict exists. This can occur when one branch deletes a file while
1622
# the other renames *and* modifies it. In this case, the content
1623
# conflict is enough.
1624
for c in cooked_conflicts:
1625
if (c.typestring == 'path conflict'
1626
and c.file_id in content_conflict_file_ids):
1628
self.cooked_conflicts.append(c)
1629
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
1632
class WeaveMerger(Merge3Merger):
1633
"""Three-way tree merger, text weave merger."""
1634
supports_reprocess = True
1635
supports_show_base = False
1636
supports_reverse_cherrypick = False
1637
history_based = True
1638
requires_file_merge_plan = True
1640
def _generate_merge_plan(self, file_id, base):
1641
return self.this_tree.plan_file_merge(file_id, self.other_tree,
1644
def _merged_lines(self, file_id):
1645
"""Generate the merged lines.
1646
There is no distinction between lines that are meant to contain <<<<<<<
1650
base = self.base_tree
1653
plan = self._generate_merge_plan(file_id, base)
1654
if 'merge' in debug.debug_flags:
1656
trans_id = self.tt.trans_id_file_id(file_id)
1657
name = self.tt.final_name(trans_id) + '.plan'
1658
contents = (b'%11s|%s' % l for l in plan)
1659
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1660
textmerge = versionedfile.PlanWeaveMerge(plan, b'<<<<<<< TREE\n',
1661
b'>>>>>>> MERGE-SOURCE\n')
1662
lines, conflicts = textmerge.merge_lines(self.reprocess)
1664
base_lines = textmerge.base_from_plan()
1667
return lines, base_lines
1669
def text_merge(self, trans_id, paths, file_id):
1670
"""Perform a (weave) text merge for a given file and file-id.
1671
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1672
and a conflict will be noted.
1674
base_path, other_path, this_path = paths
1675
lines, base_lines = self._merged_lines(file_id)
1677
# Note we're checking whether the OUTPUT is binary in this case,
1678
# because we don't want to get into weave merge guts.
1679
textfile.check_text_lines(lines)
1680
self.tt.create_file(lines, trans_id)
1681
if base_lines is not None:
1683
self._raw_conflicts.append(('text conflict', trans_id))
1684
name = self.tt.final_name(trans_id)
1685
parent_id = self.tt.final_parent(trans_id)
1686
file_group = self._dump_conflicts(name, paths, parent_id, file_id,
1688
base_lines=base_lines)
1689
file_group.append(trans_id)
1692
class LCAMerger(WeaveMerger):
1694
requires_file_merge_plan = True
1696
def _generate_merge_plan(self, file_id, base):
1697
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1701
class Diff3Merger(Merge3Merger):
1702
"""Three-way merger using external diff3 for text merging"""
1704
requires_file_merge_plan = False
1706
def dump_file(self, temp_dir, name, tree, path, file_id=None):
1707
out_path = osutils.pathjoin(temp_dir, name)
1708
with open(out_path, "wb") as out_file:
1709
in_file = tree.get_file(path)
1710
for line in in_file:
1711
out_file.write(line)
1714
def text_merge(self, trans_id, paths, file_id):
1715
"""Perform a diff3 merge using a specified file-id and trans-id.
1716
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1717
will be dumped, and a will be conflict noted.
1720
base_path, other_path, this_path = paths
1721
temp_dir = osutils.mkdtemp(prefix="bzr-")
1723
new_file = osutils.pathjoin(temp_dir, "new")
1724
this = self.dump_file(
1725
temp_dir, "this", self.this_tree, this_path, file_id)
1726
base = self.dump_file(
1727
temp_dir, "base", self.base_tree, base_path, file_id)
1728
other = self.dump_file(
1729
temp_dir, "other", self.other_tree, other_path, file_id)
1730
status = breezy.patch.diff3(new_file, this, base, other)
1731
if status not in (0, 1):
1732
raise errors.BzrError("Unhandled diff3 exit code")
1733
with open(new_file, 'rb') as f:
1734
self.tt.create_file(f, trans_id)
1736
name = self.tt.final_name(trans_id)
1737
parent_id = self.tt.final_parent(trans_id)
1738
self._dump_conflicts(name, paths, parent_id, file_id)
1739
self._raw_conflicts.append(('text conflict', trans_id))
1741
osutils.rmtree(temp_dir)
1744
class PathNotInTree(errors.BzrError):
1746
_fmt = """Merge-into failed because %(tree)s does not contain %(path)s."""
1748
def __init__(self, path, tree):
1749
errors.BzrError.__init__(self, path=path, tree=tree)
1752
class MergeIntoMerger(Merger):
1753
"""Merger that understands other_tree will be merged into a subdir.
1755
This also changes the Merger api so that it uses real Branch, revision_id,
1756
and RevisonTree objects, rather than using revision specs.
1759
def __init__(self, this_tree, other_branch, other_tree, target_subdir,
1760
source_subpath, other_rev_id=None):
1761
"""Create a new MergeIntoMerger object.
1763
source_subpath in other_tree will be effectively copied to
1764
target_subdir in this_tree.
1766
:param this_tree: The tree that we will be merging into.
1767
:param other_branch: The Branch we will be merging from.
1768
:param other_tree: The RevisionTree object we want to merge.
1769
:param target_subdir: The relative path where we want to merge
1770
other_tree into this_tree
1771
:param source_subpath: The relative path specifying the subtree of
1772
other_tree to merge into this_tree.
1774
# It is assumed that we are merging a tree that is not in our current
1775
# ancestry, which means we are using the "EmptyTree" as our basis.
1776
null_ancestor_tree = this_tree.branch.repository.revision_tree(
1777
_mod_revision.NULL_REVISION)
1778
super(MergeIntoMerger, self).__init__(
1779
this_branch=this_tree.branch,
1780
this_tree=this_tree,
1781
other_tree=other_tree,
1782
base_tree=null_ancestor_tree,
1784
self._target_subdir = target_subdir
1785
self._source_subpath = source_subpath
1786
self.other_branch = other_branch
1787
if other_rev_id is None:
1788
other_rev_id = other_tree.get_revision_id()
1789
self.other_rev_id = self.other_basis = other_rev_id
1790
self.base_is_ancestor = True
1791
self.backup_files = True
1792
self.merge_type = Merge3Merger
1793
self.show_base = False
1794
self.reprocess = False
1795
self.interesting_files = None
1796
self.merge_type = _MergeTypeParameterizer(MergeIntoMergeType,
1797
target_subdir=self._target_subdir,
1798
source_subpath=self._source_subpath)
1799
if self._source_subpath != '':
1800
# If this isn't a partial merge make sure the revisions will be
1802
self._maybe_fetch(self.other_branch, self.this_branch,
1805
def set_pending(self):
1806
if self._source_subpath != '':
1808
Merger.set_pending(self)
1811
class _MergeTypeParameterizer(object):
1812
"""Wrap a merge-type class to provide extra parameters.
1814
This is hack used by MergeIntoMerger to pass some extra parameters to its
1815
merge_type. Merger.do_merge() sets up its own set of parameters to pass to
1816
the 'merge_type' member. It is difficult override do_merge without
1817
re-writing the whole thing, so instead we create a wrapper which will pass
1818
the extra parameters.
1821
def __init__(self, merge_type, **kwargs):
1822
self._extra_kwargs = kwargs
1823
self._merge_type = merge_type
1825
def __call__(self, *args, **kwargs):
1826
kwargs.update(self._extra_kwargs)
1827
return self._merge_type(*args, **kwargs)
1829
def __getattr__(self, name):
1830
return getattr(self._merge_type, name)
1833
class MergeIntoMergeType(Merge3Merger):
1834
"""Merger that incorporates a tree (or part of a tree) into another."""
1836
def __init__(self, *args, **kwargs):
1837
"""Initialize the merger object.
1839
:param args: See Merge3Merger.__init__'s args.
1840
:param kwargs: See Merge3Merger.__init__'s keyword args, except for
1841
source_subpath and target_subdir.
1842
:keyword source_subpath: The relative path specifying the subtree of
1843
other_tree to merge into this_tree.
1844
:keyword target_subdir: The relative path where we want to merge
1845
other_tree into this_tree
1847
# All of the interesting work happens during Merge3Merger.__init__(),
1848
# so we have have to hack in to get our extra parameters set.
1849
self._source_subpath = kwargs.pop('source_subpath')
1850
self._target_subdir = kwargs.pop('target_subdir')
1851
super(MergeIntoMergeType, self).__init__(*args, **kwargs)
1853
def _compute_transform(self):
1854
with ui.ui_factory.nested_progress_bar() as child_pb:
1855
entries = self._entries_to_incorporate()
1856
entries = list(entries)
1857
for num, (entry, parent_id, relpath) in enumerate(entries):
1858
child_pb.update(gettext('Preparing file merge'),
1860
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1861
path = osutils.pathjoin(self._source_subpath, relpath)
1862
trans_id = transform.new_by_entry(path, self.tt, entry,
1863
parent_trans_id, self.other_tree)
1864
self._finish_computing_transform()
1866
def _entries_to_incorporate(self):
1867
"""Yields pairs of (inventory_entry, new_parent)."""
1868
other_inv = self.other_tree.root_inventory
1869
subdir_id = other_inv.path2id(self._source_subpath)
1870
if subdir_id is None:
1871
# XXX: The error would be clearer if it gave the URL of the source
1872
# branch, but we don't have a reference to that here.
1873
raise PathNotInTree(self._source_subpath, "Source tree")
1874
subdir = other_inv.get_entry(subdir_id)
1875
parent_in_target = osutils.dirname(self._target_subdir)
1876
target_id = self.this_tree.path2id(parent_in_target)
1877
if target_id is None:
1878
raise PathNotInTree(self._target_subdir, "Target tree")
1879
name_in_target = osutils.basename(self._target_subdir)
1880
merge_into_root = subdir.copy()
1881
merge_into_root.name = name_in_target
1882
if self.this_tree.has_id(merge_into_root.file_id):
1883
# Give the root a new file-id.
1884
# This can happen fairly easily if the directory we are
1885
# incorporating is the root, and both trees have 'TREE_ROOT' as
1886
# their root_id. Users will expect this to Just Work, so we
1887
# change the file-id here.
1888
# Non-root file-ids could potentially conflict too. That's really
1889
# an edge case, so we don't do anything special for those. We let
1890
# them cause conflicts.
1891
merge_into_root.file_id = generate_ids.gen_file_id(name_in_target)
1892
yield (merge_into_root, target_id, '')
1893
if subdir.kind != 'directory':
1894
# No children, so we are done.
1896
for path, entry in other_inv.iter_entries_by_dir(subdir_id):
1897
parent_id = entry.parent_id
1898
if parent_id == subdir.file_id:
1899
# The root's parent ID has changed, so make sure children of
1900
# the root refer to the new ID.
1901
parent_id = merge_into_root.file_id
1902
yield (entry, parent_id, path)
1905
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1907
merge_type=Merge3Merger,
1911
interesting_files=None,
1913
change_reporter=None):
1914
"""Primary interface for merging.
1916
Typical use is probably::
1918
merge_inner(branch, branch.get_revision_tree(other_revision),
1919
branch.get_revision_tree(base_revision))
1921
if this_tree is None:
1922
raise errors.BzrError("breezy.merge.merge_inner requires a this_tree "
1924
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1925
change_reporter=change_reporter)
1926
merger.backup_files = backup_files
1927
merger.merge_type = merge_type
1928
merger.ignore_zero = ignore_zero
1929
merger.interesting_files = interesting_files
1930
merger.show_base = show_base
1931
merger.reprocess = reprocess
1932
merger.other_rev_id = other_rev_id
1933
merger.other_basis = other_rev_id
1934
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1935
if get_revision_id is None:
1936
get_revision_id = base_tree.last_revision
1937
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
1938
merger.set_base_revision(get_revision_id(), this_branch)
1939
return merger.do_merge()
1942
merge_type_registry = registry.Registry()
1943
merge_type_registry.register('diff3', Diff3Merger,
1944
"Merge using external diff3.")
1945
merge_type_registry.register('lca', LCAMerger,
1946
"LCA-newness merge.")
1947
merge_type_registry.register('merge3', Merge3Merger,
1948
"Native diff3-style merge.")
1949
merge_type_registry.register('weave', WeaveMerger,
1950
"Weave-based merge.")
1953
def get_merge_type_registry():
1954
"""Merge type registry was previously in breezy.option
1956
This method provides a backwards compatible way to retrieve it.
1958
return merge_type_registry
1961
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1962
def status_a(revision, text):
1963
if revision in ancestors_b:
1964
return 'killed-b', text
1966
return 'new-a', text
1968
def status_b(revision, text):
1969
if revision in ancestors_a:
1970
return 'killed-a', text
1972
return 'new-b', text
1974
plain_a = [t for (a, t) in annotated_a]
1975
plain_b = [t for (a, t) in annotated_b]
1976
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1977
blocks = matcher.get_matching_blocks()
1980
for ai, bi, l in blocks:
1981
# process all mismatched sections
1982
# (last mismatched section is handled because blocks always
1983
# includes a 0-length last block)
1984
for revision, text in annotated_a[a_cur:ai]:
1985
yield status_a(revision, text)
1986
for revision, text in annotated_b[b_cur:bi]:
1987
yield status_b(revision, text)
1988
# and now the matched section
1991
for text_a in plain_a[ai:a_cur]:
1992
yield "unchanged", text_a
1995
class _PlanMergeBase(object):
1997
def __init__(self, a_rev, b_rev, vf, key_prefix):
2000
:param a_rev: Revision-id of one revision to merge
2001
:param b_rev: Revision-id of the other revision to merge
2002
:param vf: A VersionedFiles containing both revisions
2003
:param key_prefix: A prefix for accessing keys in vf, typically
2009
self._last_lines = None
2010
self._last_lines_revision_id = None
2011
self._cached_matching_blocks = {}
2012
self._key_prefix = key_prefix
2013
self._precache_tip_lines()
2015
def _precache_tip_lines(self):
2016
lines = self.get_lines([self.a_rev, self.b_rev])
2017
self.lines_a = lines[self.a_rev]
2018
self.lines_b = lines[self.b_rev]
2020
def get_lines(self, revisions):
2021
"""Get lines for revisions from the backing VersionedFiles.
2023
:raises RevisionNotPresent: on absent texts.
2025
keys = [(self._key_prefix + (rev,)) for rev in revisions]
2027
for record in self.vf.get_record_stream(keys, 'unordered', True):
2028
if record.storage_kind == 'absent':
2029
raise errors.RevisionNotPresent(record.key, self.vf)
2030
result[record.key[-1]] = osutils.chunks_to_lines(
2031
record.get_bytes_as('chunked'))
2034
def plan_merge(self):
2035
"""Generate a 'plan' for merging the two revisions.
2037
This involves comparing their texts and determining the cause of
2038
differences. If text A has a line and text B does not, then either the
2039
line was added to text A, or it was deleted from B. Once the causes
2040
are combined, they are written out in the format described in
2041
VersionedFile.plan_merge
2043
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
2044
unique_a, unique_b = self._unique_lines(blocks)
2045
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
2046
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
2047
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
2049
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
2052
for i, j, n in blocks:
2053
for a_index in range(last_i, i):
2054
if a_index in new_a:
2055
if a_index in killed_b:
2056
yield 'conflicted-a', self.lines_a[a_index]
2058
yield 'new-a', self.lines_a[a_index]
2060
yield 'killed-b', self.lines_a[a_index]
2061
for b_index in range(last_j, j):
2062
if b_index in new_b:
2063
if b_index in killed_a:
2064
yield 'conflicted-b', self.lines_b[b_index]
2066
yield 'new-b', self.lines_b[b_index]
2068
yield 'killed-a', self.lines_b[b_index]
2069
# handle common lines
2070
for a_index in range(i, i + n):
2071
yield 'unchanged', self.lines_a[a_index]
2075
def _get_matching_blocks(self, left_revision, right_revision):
2076
"""Return a description of which sections of two revisions match.
2078
See SequenceMatcher.get_matching_blocks
2080
cached = self._cached_matching_blocks.get((left_revision,
2082
if cached is not None:
2084
if self._last_lines_revision_id == left_revision:
2085
left_lines = self._last_lines
2086
right_lines = self.get_lines([right_revision])[right_revision]
2088
lines = self.get_lines([left_revision, right_revision])
2089
left_lines = lines[left_revision]
2090
right_lines = lines[right_revision]
2091
self._last_lines = right_lines
2092
self._last_lines_revision_id = right_revision
2093
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
2095
return matcher.get_matching_blocks()
2097
def _unique_lines(self, matching_blocks):
2098
"""Analyse matching_blocks to determine which lines are unique
2100
:return: a tuple of (unique_left, unique_right), where the values are
2101
sets of line numbers of unique lines.
2107
for i, j, n in matching_blocks:
2108
unique_left.extend(range(last_i, i))
2109
unique_right.extend(range(last_j, j))
2112
return unique_left, unique_right
2115
def _subtract_plans(old_plan, new_plan):
2116
"""Remove changes from new_plan that came from old_plan.
2118
It is assumed that the difference between the old_plan and new_plan
2119
is their choice of 'b' text.
2121
All lines from new_plan that differ from old_plan are emitted
2122
verbatim. All lines from new_plan that match old_plan but are
2123
not about the 'b' revision are emitted verbatim.
2125
Lines that match and are about the 'b' revision are the lines we
2126
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
2127
is skipped entirely.
2129
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
2132
for i, j, n in matcher.get_matching_blocks():
2133
for jj in range(last_j, j):
2135
for jj in range(j, j + n):
2136
plan_line = new_plan[jj]
2137
if plan_line[0] == 'new-b':
2139
elif plan_line[0] == 'killed-b':
2140
yield 'unchanged', plan_line[1]
2146
class _PlanMerge(_PlanMergeBase):
2147
"""Plan an annotate merge using on-the-fly annotation"""
2149
def __init__(self, a_rev, b_rev, vf, key_prefix):
2150
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
2151
self.a_key = self._key_prefix + (self.a_rev,)
2152
self.b_key = self._key_prefix + (self.b_rev,)
2153
self.graph = _mod_graph.Graph(self.vf)
2154
heads = self.graph.heads((self.a_key, self.b_key))
2156
# one side dominates, so we can just return its values, yay for
2158
# Ideally we would know that before we get this far
2159
self._head_key = heads.pop()
2160
if self._head_key == self.a_key:
2164
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2165
self._head_key[-1], other)
2168
self._head_key = None
2171
def _precache_tip_lines(self):
2172
# Turn this into a no-op, because we will do this later
2175
def _find_recursive_lcas(self):
2176
"""Find all the ancestors back to a unique lca"""
2177
cur_ancestors = (self.a_key, self.b_key)
2178
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2179
# rather than a key tuple. We will just map that directly to no common
2183
next_lcas = self.graph.find_lca(*cur_ancestors)
2184
# Map a plain NULL_REVISION to a simple no-ancestors
2185
if next_lcas == {_mod_revision.NULL_REVISION}:
2187
# Order the lca's based on when they were merged into the tip
2188
# While the actual merge portion of weave merge uses a set() of
2189
# active revisions, the order of insertion *does* effect the
2190
# implicit ordering of the texts.
2191
for rev_key in cur_ancestors:
2192
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2194
parent_map[rev_key] = ordered_parents
2195
if len(next_lcas) == 0:
2197
elif len(next_lcas) == 1:
2198
parent_map[list(next_lcas)[0]] = ()
2200
elif len(next_lcas) > 2:
2201
# More than 2 lca's, fall back to grabbing all nodes between
2202
# this and the unique lca.
2203
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2205
self.a_key, self.b_key, cur_ancestors)
2206
cur_lcas = next_lcas
2207
while len(cur_lcas) > 1:
2208
cur_lcas = self.graph.find_lca(*cur_lcas)
2209
if len(cur_lcas) == 0:
2210
# No common base to find, use the full ancestry
2213
unique_lca = list(cur_lcas)[0]
2214
if unique_lca == _mod_revision.NULL_REVISION:
2215
# find_lca will return a plain 'NULL_REVISION' rather
2216
# than a key tuple when there is no common ancestor, we
2217
# prefer to just use None, because it doesn't confuse
2218
# _get_interesting_texts()
2220
parent_map.update(self._find_unique_parents(next_lcas,
2223
cur_ancestors = next_lcas
2226
def _find_unique_parents(self, tip_keys, base_key):
2227
"""Find ancestors of tip that aren't ancestors of base.
2229
:param tip_keys: Nodes that are interesting
2230
:param base_key: Cull all ancestors of this node
2231
:return: The parent map for all revisions between tip_keys and
2232
base_key. base_key will be included. References to nodes outside of
2233
the ancestor set will also be removed.
2235
# TODO: this would be simpler if find_unique_ancestors took a list
2236
# instead of a single tip, internally it supports it, but it
2237
# isn't a "backwards compatible" api change.
2238
if base_key is None:
2239
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2240
# We remove NULL_REVISION because it isn't a proper tuple key, and
2241
# thus confuses things like _get_interesting_texts, and our logic
2242
# to add the texts into the memory weave.
2243
if _mod_revision.NULL_REVISION in parent_map:
2244
parent_map.pop(_mod_revision.NULL_REVISION)
2247
for tip in tip_keys:
2249
self.graph.find_unique_ancestors(tip, [base_key]))
2250
parent_map = self.graph.get_parent_map(interesting)
2251
parent_map[base_key] = ()
2252
culled_parent_map, child_map, tails = self._remove_external_references(
2254
# Remove all the tails but base_key
2255
if base_key is not None:
2256
tails.remove(base_key)
2257
self._prune_tails(culled_parent_map, child_map, tails)
2258
# Now remove all the uninteresting 'linear' regions
2259
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2263
def _remove_external_references(parent_map):
2264
"""Remove references that go outside of the parent map.
2266
:param parent_map: Something returned from Graph.get_parent_map(keys)
2267
:return: (filtered_parent_map, child_map, tails)
2268
filtered_parent_map is parent_map without external references
2269
child_map is the {parent_key: [child_keys]} mapping
2270
tails is a list of nodes that do not have any parents in the map
2272
# TODO: The basic effect of this function seems more generic than
2273
# _PlanMerge. But the specific details of building a child_map,
2274
# and computing tails seems very specific to _PlanMerge.
2275
# Still, should this be in Graph land?
2276
filtered_parent_map = {}
2279
for key, parent_keys in viewitems(parent_map):
2280
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2281
if not culled_parent_keys:
2283
for parent_key in culled_parent_keys:
2284
child_map.setdefault(parent_key, []).append(key)
2285
# TODO: Do we want to do this, it adds overhead for every node,
2286
# just to say that the node has no children
2287
child_map.setdefault(key, [])
2288
filtered_parent_map[key] = culled_parent_keys
2289
return filtered_parent_map, child_map, tails
2292
def _prune_tails(parent_map, child_map, tails_to_remove):
2293
"""Remove tails from the parent map.
2295
This will remove the supplied revisions until no more children have 0
2298
:param parent_map: A dict of {child: [parents]}, this dictionary will
2299
be modified in place.
2300
:param tails_to_remove: A list of tips that should be removed,
2301
this list will be consumed
2302
:param child_map: The reverse dict of parent_map ({parent: [children]})
2303
this dict will be modified
2304
:return: None, parent_map will be modified in place.
2306
while tails_to_remove:
2307
next = tails_to_remove.pop()
2308
parent_map.pop(next)
2309
children = child_map.pop(next)
2310
for child in children:
2311
child_parents = parent_map[child]
2312
child_parents.remove(next)
2313
if len(child_parents) == 0:
2314
tails_to_remove.append(child)
2316
def _get_interesting_texts(self, parent_map):
2317
"""Return a dict of texts we are interested in.
2319
Note that the input is in key tuples, but the output is in plain
2322
:param parent_map: The output from _find_recursive_lcas
2323
:return: A dict of {'revision_id':lines} as returned by
2324
_PlanMergeBase.get_lines()
2326
all_revision_keys = set(parent_map)
2327
all_revision_keys.add(self.a_key)
2328
all_revision_keys.add(self.b_key)
2330
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2331
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2334
def _build_weave(self):
2335
from .bzr import weave
2336
self._weave = weave.Weave(weave_name='in_memory_weave',
2337
allow_reserved=True)
2338
parent_map = self._find_recursive_lcas()
2340
all_texts = self._get_interesting_texts(parent_map)
2342
# Note: Unfortunately, the order given by topo_sort will effect the
2343
# ordering resolution in the output. Specifically, if you add A then B,
2344
# then in the output text A lines will show up before B lines. And, of
2345
# course, topo_sort doesn't guarantee any real ordering.
2346
# So we use merge_sort, and add a fake node on the tip.
2347
# This ensures that left-hand parents will always be inserted into the
2348
# weave before right-hand parents.
2349
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2350
parent_map[tip_key] = (self.a_key, self.b_key)
2352
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2356
# for key in tsort.topo_sort(parent_map):
2357
parent_keys = parent_map[key]
2358
revision_id = key[-1]
2359
parent_ids = [k[-1] for k in parent_keys]
2360
self._weave.add_lines(revision_id, parent_ids,
2361
all_texts[revision_id])
2363
def plan_merge(self):
2364
"""Generate a 'plan' for merging the two revisions.
2366
This involves comparing their texts and determining the cause of
2367
differences. If text A has a line and text B does not, then either the
2368
line was added to text A, or it was deleted from B. Once the causes
2369
are combined, they are written out in the format described in
2370
VersionedFile.plan_merge
2372
if self._head_key is not None: # There was a single head
2373
if self._head_key == self.a_key:
2376
if self._head_key != self.b_key:
2377
raise AssertionError('There was an invalid head: %s != %s'
2378
% (self.b_key, self._head_key))
2380
head_rev = self._head_key[-1]
2381
lines = self.get_lines([head_rev])[head_rev]
2382
return ((plan, line) for line in lines)
2383
return self._weave.plan_merge(self.a_rev, self.b_rev)
2386
class _PlanLCAMerge(_PlanMergeBase):
2388
This merge algorithm differs from _PlanMerge in that:
2390
1. comparisons are done against LCAs only
2391
2. cases where a contested line is new versus one LCA but old versus
2392
another are marked as conflicts, by emitting the line as conflicted-a
2395
This is faster, and hopefully produces more useful output.
2398
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2399
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2400
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2403
if lca == _mod_revision.NULL_REVISION:
2406
self.lcas.add(lca[-1])
2407
for lca in self.lcas:
2408
if _mod_revision.is_null(lca):
2411
lca_lines = self.get_lines([lca])[lca]
2412
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2414
blocks = list(matcher.get_matching_blocks())
2415
self._cached_matching_blocks[(a_rev, lca)] = blocks
2416
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2418
blocks = list(matcher.get_matching_blocks())
2419
self._cached_matching_blocks[(b_rev, lca)] = blocks
2421
def _determine_status(self, revision_id, unique_line_numbers):
2422
"""Determines the status unique lines versus all lcas.
2424
Basically, determines why the line is unique to this revision.
2426
A line may be determined new, killed, or both.
2428
If a line is determined new, that means it was not present in at least
2429
one LCA, and is not present in the other merge revision.
2431
If a line is determined killed, that means the line was present in
2434
If a line is killed and new, this indicates that the two merge
2435
revisions contain differing conflict resolutions.
2437
:param revision_id: The id of the revision in which the lines are
2439
:param unique_line_numbers: The line numbers of unique lines.
2440
:return: a tuple of (new_this, killed_other)
2444
unique_line_numbers = set(unique_line_numbers)
2445
for lca in self.lcas:
2446
blocks = self._get_matching_blocks(revision_id, lca)
2447
unique_vs_lca, _ignored = self._unique_lines(blocks)
2448
new.update(unique_line_numbers.intersection(unique_vs_lca))
2449
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
breezy/merge.py
