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
21
from bzrlib.lazy_import import lazy_import
22
lazy_import(globals(), """
24
branch as _mod_branch,
26
conflicts as _mod_conflicts,
33
revision as _mod_revision,
43
from bzrlib.i18n import gettext
51
from bzrlib.symbol_versioning import (
55
# TODO: Report back as changes are merged in
58
def transform_tree(from_tree, to_tree, interesting_ids=None):
59
from_tree.lock_tree_write()
60
operation = cleanup.OperationWithCleanups(merge_inner)
61
operation.add_cleanup(from_tree.unlock)
62
operation.run_simple(from_tree.branch, to_tree, from_tree,
63
ignore_zero=True, interesting_ids=interesting_ids, this_tree=from_tree)
66
class MergeHooks(hooks.Hooks):
69
hooks.Hooks.__init__(self, "bzrlib.merge", "Merger.hooks")
70
self.add_hook('merge_file_content',
71
"Called with a bzrlib.merge.Merger object to create a per file "
72
"merge object when starting a merge. "
73
"Should return either None or a subclass of "
74
"``bzrlib.merge.AbstractPerFileMerger``. "
75
"Such objects will then be called per file "
76
"that needs to be merged (including when one "
77
"side has deleted the file and the other has changed it). "
78
"See the AbstractPerFileMerger API docs for details on how it is "
81
self.add_hook('pre_merge',
82
'Called before a merge. '
83
'Receives a Merger object as the single argument.',
85
self.add_hook('post_merge',
86
'Called after a merge. '
87
'Receives a Merger object as the single argument. '
88
'The return value is ignored.',
92
class AbstractPerFileMerger(object):
93
"""PerFileMerger objects are used by plugins extending merge for bzrlib.
95
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
97
:ivar merger: The Merge3Merger performing the merge.
100
def __init__(self, merger):
101
"""Create a PerFileMerger for use with merger."""
104
def merge_contents(self, merge_params):
105
"""Attempt to merge the contents of a single file.
107
:param merge_params: A bzrlib.merge.MergeFileHookParams
108
:return: A tuple of (status, chunks), where status is one of
109
'not_applicable', 'success', 'conflicted', or 'delete'. If status
110
is 'success' or 'conflicted', then chunks should be an iterable of
111
strings for the new file contents.
113
return ('not applicable', None)
116
class PerFileMerger(AbstractPerFileMerger):
117
"""Merge individual files when self.file_matches returns True.
119
This class is intended to be subclassed. The file_matches and
120
merge_matching methods should be overridden with concrete implementations.
123
def file_matches(self, params):
124
"""Return True if merge_matching should be called on this file.
126
Only called with merges of plain files with no clear winner.
128
Subclasses must override this.
130
raise NotImplementedError(self.file_matches)
132
def get_filename(self, params, tree):
133
"""Lookup the filename (i.e. basename, not path), given a Tree (e.g.
134
self.merger.this_tree) and a MergeFileHookParams.
136
return osutils.basename(tree.id2path(params.file_id))
138
def get_filepath(self, params, tree):
139
"""Calculate the path to the file in a tree.
141
:param params: A MergeFileHookParams describing the file to merge
142
:param tree: a Tree, e.g. self.merger.this_tree.
144
return tree.id2path(params.file_id)
146
def merge_contents(self, params):
147
"""Merge the contents of a single file."""
148
# Check whether this custom merge logic should be used.
150
# OTHER is a straight winner, rely on default merge.
151
params.winner == 'other' or
152
# THIS and OTHER aren't both files.
153
not params.is_file_merge() or
154
# The filename doesn't match
155
not self.file_matches(params)):
156
return 'not_applicable', None
157
return self.merge_matching(params)
159
def merge_matching(self, params):
160
"""Merge the contents of a single file that has matched the criteria
161
in PerFileMerger.merge_contents (is a conflict, is a file,
162
self.file_matches is True).
164
Subclasses must override this.
166
raise NotImplementedError(self.merge_matching)
169
class ConfigurableFileMerger(PerFileMerger):
170
"""Merge individual files when configured via a .conf file.
172
This is a base class for concrete custom file merging logic. Concrete
173
classes should implement ``merge_text``.
175
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
177
:ivar affected_files: The configured file paths to merge.
179
:cvar name_prefix: The prefix to use when looking up configuration
180
details. <name_prefix>_merge_files describes the files targeted by the
183
:cvar default_files: The default file paths to merge when no configuration
190
def __init__(self, merger):
191
super(ConfigurableFileMerger, self).__init__(merger)
192
self.affected_files = None
193
self.default_files = self.__class__.default_files or []
194
self.name_prefix = self.__class__.name_prefix
195
if self.name_prefix is None:
196
raise ValueError("name_prefix must be set.")
198
def file_matches(self, params):
199
"""Check whether the file should call the merge hook.
201
<name_prefix>_merge_files configuration variable is a list of files
202
that should use the hook.
204
affected_files = self.affected_files
205
if affected_files is None:
206
config = self.merger.this_branch.get_config()
207
# Until bzr provides a better policy for caching the config, we
208
# just add the part we're interested in to the params to avoid
209
# reading the config files repeatedly (bazaar.conf, location.conf,
211
config_key = self.name_prefix + '_merge_files'
212
affected_files = config.get_user_option_as_list(config_key)
213
if affected_files is None:
214
# If nothing was specified in the config, use the default.
215
affected_files = self.default_files
216
self.affected_files = affected_files
218
filepath = self.get_filepath(params, self.merger.this_tree)
219
if filepath in affected_files:
223
def merge_matching(self, params):
224
return self.merge_text(params)
226
def merge_text(self, params):
227
"""Merge the byte contents of a single file.
229
This is called after checking that the merge should be performed in
230
merge_contents, and it should behave as per
231
``bzrlib.merge.AbstractPerFileMerger.merge_contents``.
233
raise NotImplementedError(self.merge_text)
236
class MergeFileHookParams(object):
237
"""Object holding parameters passed to merge_file_content hooks.
239
There are some fields hooks can access:
241
:ivar file_id: the file ID of the file being merged
242
:ivar trans_id: the transform ID for the merge of this file
243
:ivar this_kind: kind of file_id in 'this' tree
244
:ivar other_kind: kind of file_id in 'other' tree
245
:ivar winner: one of 'this', 'other', 'conflict'
248
def __init__(self, merger, file_id, trans_id, this_kind, other_kind,
250
self._merger = merger
251
self.file_id = file_id
252
self.trans_id = trans_id
253
self.this_kind = this_kind
254
self.other_kind = other_kind
257
def is_file_merge(self):
258
"""True if this_kind and other_kind are both 'file'."""
259
return self.this_kind == 'file' and self.other_kind == 'file'
261
@decorators.cachedproperty
262
def base_lines(self):
263
"""The lines of the 'base' version of the file."""
264
return self._merger.get_lines(self._merger.base_tree, self.file_id)
266
@decorators.cachedproperty
267
def this_lines(self):
268
"""The lines of the 'this' version of the file."""
269
return self._merger.get_lines(self._merger.this_tree, self.file_id)
271
@decorators.cachedproperty
272
def other_lines(self):
273
"""The lines of the 'other' version of the file."""
274
return self._merger.get_lines(self._merger.other_tree, self.file_id)
277
class Merger(object):
281
def __init__(self, this_branch, other_tree=None, base_tree=None,
282
this_tree=None, pb=None, change_reporter=None,
283
recurse='down', revision_graph=None):
284
object.__init__(self)
285
self.this_branch = this_branch
286
self.this_basis = _mod_revision.ensure_null(
287
this_branch.last_revision())
288
self.this_rev_id = None
289
self.this_tree = this_tree
290
self.this_revision_tree = None
291
self.this_basis_tree = None
292
self.other_tree = other_tree
293
self.other_branch = None
294
self.base_tree = base_tree
295
self.ignore_zero = False
296
self.backup_files = False
297
self.interesting_ids = None
298
self.interesting_files = None
299
self.show_base = False
300
self.reprocess = False
302
warnings.warn("pb parameter to Merger() is deprecated and ignored")
304
self.recurse = recurse
305
self.change_reporter = change_reporter
306
self._cached_trees = {}
307
self._revision_graph = revision_graph
308
self._base_is_ancestor = None
309
self._base_is_other_ancestor = None
310
self._is_criss_cross = None
311
self._lca_trees = None
313
def cache_trees_with_revision_ids(self, trees):
314
"""Cache any tree in trees if it has a revision_id."""
315
for maybe_tree in trees:
316
if maybe_tree is None:
319
rev_id = maybe_tree.get_revision_id()
320
except AttributeError:
322
self._cached_trees[rev_id] = maybe_tree
325
def revision_graph(self):
326
if self._revision_graph is None:
327
self._revision_graph = self.this_branch.repository.get_graph()
328
return self._revision_graph
330
def _set_base_is_ancestor(self, value):
331
self._base_is_ancestor = value
333
def _get_base_is_ancestor(self):
334
if self._base_is_ancestor is None:
335
self._base_is_ancestor = self.revision_graph.is_ancestor(
336
self.base_rev_id, self.this_basis)
337
return self._base_is_ancestor
339
base_is_ancestor = property(_get_base_is_ancestor, _set_base_is_ancestor)
341
def _set_base_is_other_ancestor(self, value):
342
self._base_is_other_ancestor = value
344
def _get_base_is_other_ancestor(self):
345
if self._base_is_other_ancestor is None:
346
if self.other_basis is None:
348
self._base_is_other_ancestor = self.revision_graph.is_ancestor(
349
self.base_rev_id, self.other_basis)
350
return self._base_is_other_ancestor
352
base_is_other_ancestor = property(_get_base_is_other_ancestor,
353
_set_base_is_other_ancestor)
356
def from_uncommitted(tree, other_tree, pb=None, base_tree=None):
357
"""Return a Merger for uncommitted changes in other_tree.
359
:param tree: The tree to merge into
360
:param other_tree: The tree to get uncommitted changes from
361
:param pb: A progress indicator
362
:param base_tree: The basis to use for the merge. If unspecified,
363
other_tree.basis_tree() will be used.
365
if base_tree is None:
366
base_tree = other_tree.basis_tree()
367
merger = Merger(tree.branch, other_tree, base_tree, tree, pb)
368
merger.base_rev_id = merger.base_tree.get_revision_id()
369
merger.other_rev_id = None
370
merger.other_basis = merger.base_rev_id
374
def from_mergeable(klass, tree, mergeable, pb):
375
"""Return a Merger for a bundle or merge directive.
377
:param tree: The tree to merge changes into
378
:param mergeable: A merge directive or bundle
379
:param pb: A progress indicator
381
mergeable.install_revisions(tree.branch.repository)
382
base_revision_id, other_revision_id, verified =\
383
mergeable.get_merge_request(tree.branch.repository)
384
revision_graph = tree.branch.repository.get_graph()
385
if base_revision_id is not None:
386
if (base_revision_id != _mod_revision.NULL_REVISION and
387
revision_graph.is_ancestor(
388
base_revision_id, tree.branch.last_revision())):
389
base_revision_id = None
391
trace.warning('Performing cherrypick')
392
merger = klass.from_revision_ids(pb, tree, other_revision_id,
393
base_revision_id, revision_graph=
395
return merger, verified
398
def from_revision_ids(pb, tree, other, base=None, other_branch=None,
399
base_branch=None, revision_graph=None,
401
"""Return a Merger for revision-ids.
403
:param pb: A progress indicator
404
:param tree: The tree to merge changes into
405
:param other: The revision-id to use as OTHER
406
:param base: The revision-id to use as BASE. If not specified, will
408
:param other_branch: A branch containing the other revision-id. If
409
not supplied, tree.branch is used.
410
:param base_branch: A branch containing the base revision-id. If
411
not supplied, other_branch or tree.branch will be used.
412
:param revision_graph: If you have a revision_graph precomputed, pass
413
it in, otherwise it will be created for you.
414
:param tree_branch: The branch associated with tree. If not supplied,
415
tree.branch will be used.
417
if tree_branch is None:
418
tree_branch = tree.branch
419
merger = Merger(tree_branch, this_tree=tree, pb=pb,
420
revision_graph=revision_graph)
421
if other_branch is None:
422
other_branch = tree.branch
423
merger.set_other_revision(other, other_branch)
427
if base_branch is None:
428
base_branch = other_branch
429
merger.set_base_revision(base, base_branch)
432
def revision_tree(self, revision_id, branch=None):
433
if revision_id not in self._cached_trees:
435
branch = self.this_branch
437
tree = self.this_tree.revision_tree(revision_id)
438
except errors.NoSuchRevisionInTree:
439
tree = branch.repository.revision_tree(revision_id)
440
self._cached_trees[revision_id] = tree
441
return self._cached_trees[revision_id]
443
def _get_tree(self, treespec, possible_transports=None):
444
location, revno = treespec
446
tree = workingtree.WorkingTree.open_containing(location)[0]
447
return tree.branch, tree
448
branch = _mod_branch.Branch.open_containing(
449
location, possible_transports)[0]
451
revision_id = branch.last_revision()
453
revision_id = branch.get_rev_id(revno)
454
revision_id = _mod_revision.ensure_null(revision_id)
455
return branch, self.revision_tree(revision_id, branch)
457
def set_interesting_files(self, file_list):
458
self.interesting_files = file_list
460
def set_pending(self):
461
if (not self.base_is_ancestor or not self.base_is_other_ancestor
462
or self.other_rev_id is None):
466
def _add_parent(self):
467
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
468
new_parent_trees = []
469
operation = cleanup.OperationWithCleanups(
470
self.this_tree.set_parent_trees)
471
for revision_id in new_parents:
473
tree = self.revision_tree(revision_id)
474
except errors.NoSuchRevision:
478
operation.add_cleanup(tree.unlock)
479
new_parent_trees.append((revision_id, tree))
480
operation.run_simple(new_parent_trees, allow_leftmost_as_ghost=True)
482
def set_other(self, other_revision, possible_transports=None):
483
"""Set the revision and tree to merge from.
485
This sets the other_tree, other_rev_id, other_basis attributes.
487
:param other_revision: The [path, revision] list to merge from.
489
self.other_branch, self.other_tree = self._get_tree(other_revision,
491
if other_revision[1] == -1:
492
self.other_rev_id = _mod_revision.ensure_null(
493
self.other_branch.last_revision())
494
if _mod_revision.is_null(self.other_rev_id):
495
raise errors.NoCommits(self.other_branch)
496
self.other_basis = self.other_rev_id
497
elif other_revision[1] is not None:
498
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
499
self.other_basis = self.other_rev_id
501
self.other_rev_id = None
502
self.other_basis = self.other_branch.last_revision()
503
if self.other_basis is None:
504
raise errors.NoCommits(self.other_branch)
505
if self.other_rev_id is not None:
506
self._cached_trees[self.other_rev_id] = self.other_tree
507
self._maybe_fetch(self.other_branch, self.this_branch, self.other_basis)
509
def set_other_revision(self, revision_id, other_branch):
510
"""Set 'other' based on a branch and revision id
512
:param revision_id: The revision to use for a tree
513
:param other_branch: The branch containing this tree
515
self.other_rev_id = revision_id
516
self.other_branch = other_branch
517
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
518
self.other_tree = self.revision_tree(revision_id)
519
self.other_basis = revision_id
521
def set_base_revision(self, revision_id, branch):
522
"""Set 'base' based on a branch and revision id
524
:param revision_id: The revision to use for a tree
525
:param branch: The branch containing this tree
527
self.base_rev_id = revision_id
528
self.base_branch = branch
529
self._maybe_fetch(branch, self.this_branch, revision_id)
530
self.base_tree = self.revision_tree(revision_id)
532
def _maybe_fetch(self, source, target, revision_id):
533
if not source.repository.has_same_location(target.repository):
534
target.fetch(source, revision_id)
537
revisions = [_mod_revision.ensure_null(self.this_basis),
538
_mod_revision.ensure_null(self.other_basis)]
539
if _mod_revision.NULL_REVISION in revisions:
540
self.base_rev_id = _mod_revision.NULL_REVISION
541
self.base_tree = self.revision_tree(self.base_rev_id)
542
self._is_criss_cross = False
544
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
545
self._is_criss_cross = False
547
self.base_rev_id = _mod_revision.NULL_REVISION
549
self.base_rev_id = list(lcas)[0]
550
else: # len(lcas) > 1
551
self._is_criss_cross = True
553
# find_unique_lca can only handle 2 nodes, so we have to
554
# start back at the beginning. It is a shame to traverse
555
# the graph again, but better than re-implementing
557
self.base_rev_id = self.revision_graph.find_unique_lca(
558
revisions[0], revisions[1])
560
self.base_rev_id = self.revision_graph.find_unique_lca(
562
sorted_lca_keys = self.revision_graph.find_merge_order(
564
if self.base_rev_id == _mod_revision.NULL_REVISION:
565
self.base_rev_id = sorted_lca_keys[0]
567
if self.base_rev_id == _mod_revision.NULL_REVISION:
568
raise errors.UnrelatedBranches()
569
if self._is_criss_cross:
570
trace.warning('Warning: criss-cross merge encountered. See bzr'
571
' help criss-cross.')
572
trace.mutter('Criss-cross lcas: %r' % lcas)
573
if self.base_rev_id in lcas:
574
trace.mutter('Unable to find unique lca. '
575
'Fallback %r as best option.'
577
interesting_revision_ids = set(lcas)
578
interesting_revision_ids.add(self.base_rev_id)
579
interesting_trees = dict((t.get_revision_id(), t)
580
for t in self.this_branch.repository.revision_trees(
581
interesting_revision_ids))
582
self._cached_trees.update(interesting_trees)
583
if self.base_rev_id in lcas:
584
self.base_tree = interesting_trees[self.base_rev_id]
586
self.base_tree = interesting_trees.pop(self.base_rev_id)
587
self._lca_trees = [interesting_trees[key]
588
for key in sorted_lca_keys]
590
self.base_tree = self.revision_tree(self.base_rev_id)
591
self.base_is_ancestor = True
592
self.base_is_other_ancestor = True
593
trace.mutter('Base revid: %r' % self.base_rev_id)
595
def set_base(self, base_revision):
596
"""Set the base revision to use for the merge.
598
:param base_revision: A 2-list containing a path and revision number.
600
trace.mutter("doing merge() with no base_revision specified")
601
if base_revision == [None, None]:
604
base_branch, self.base_tree = self._get_tree(base_revision)
605
if base_revision[1] == -1:
606
self.base_rev_id = base_branch.last_revision()
607
elif base_revision[1] is None:
608
self.base_rev_id = _mod_revision.NULL_REVISION
610
self.base_rev_id = _mod_revision.ensure_null(
611
base_branch.get_rev_id(base_revision[1]))
612
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
614
def make_merger(self):
615
kwargs = {'working_tree': self.this_tree, 'this_tree': self.this_tree,
616
'other_tree': self.other_tree,
617
'interesting_ids': self.interesting_ids,
618
'interesting_files': self.interesting_files,
619
'this_branch': self.this_branch,
621
if self.merge_type.requires_base:
622
kwargs['base_tree'] = self.base_tree
623
if self.merge_type.supports_reprocess:
624
kwargs['reprocess'] = self.reprocess
626
raise errors.BzrError(
627
"Conflict reduction is not supported for merge"
628
" type %s." % self.merge_type)
629
if self.merge_type.supports_show_base:
630
kwargs['show_base'] = self.show_base
632
raise errors.BzrError("Showing base is not supported for this"
633
" merge type. %s" % self.merge_type)
634
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
635
and not self.base_is_other_ancestor):
636
raise errors.CannotReverseCherrypick()
637
if self.merge_type.supports_cherrypick:
638
kwargs['cherrypick'] = (not self.base_is_ancestor or
639
not self.base_is_other_ancestor)
640
if self._is_criss_cross and getattr(self.merge_type,
641
'supports_lca_trees', False):
642
kwargs['lca_trees'] = self._lca_trees
643
return self.merge_type(pb=None,
644
change_reporter=self.change_reporter,
647
def _do_merge_to(self):
648
merge = self.make_merger()
649
if self.other_branch is not None:
650
self.other_branch.update_references(self.this_branch)
651
for hook in Merger.hooks['pre_merge']:
654
for hook in Merger.hooks['post_merge']:
656
if self.recurse == 'down':
657
for relpath, file_id in self.this_tree.iter_references():
658
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
659
other_revision = self.other_tree.get_reference_revision(
661
if other_revision == sub_tree.last_revision():
663
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
664
sub_merge.merge_type = self.merge_type
665
other_branch = self.other_branch.reference_parent(file_id,
667
sub_merge.set_other_revision(other_revision, other_branch)
668
base_revision = self.base_tree.get_reference_revision(file_id)
669
sub_merge.base_tree = \
670
sub_tree.branch.repository.revision_tree(base_revision)
671
sub_merge.base_rev_id = base_revision
676
operation = cleanup.OperationWithCleanups(self._do_merge_to)
677
self.this_tree.lock_tree_write()
678
operation.add_cleanup(self.this_tree.unlock)
679
if self.base_tree is not None:
680
self.base_tree.lock_read()
681
operation.add_cleanup(self.base_tree.unlock)
682
if self.other_tree is not None:
683
self.other_tree.lock_read()
684
operation.add_cleanup(self.other_tree.unlock)
685
merge = operation.run_simple()
686
if len(merge.cooked_conflicts) == 0:
687
if not self.ignore_zero and not trace.is_quiet():
688
trace.note(gettext("All changes applied successfully."))
690
trace.note(gettext("%d conflicts encountered.")
691
% len(merge.cooked_conflicts))
693
return len(merge.cooked_conflicts)
696
class _InventoryNoneEntry(object):
697
"""This represents an inventory entry which *isn't there*.
699
It simplifies the merging logic if we always have an InventoryEntry, even
700
if it isn't actually present
707
symlink_target = None
710
_none_entry = _InventoryNoneEntry()
713
class Merge3Merger(object):
714
"""Three-way merger that uses the merge3 text merger"""
716
supports_reprocess = True
717
supports_show_base = True
718
history_based = False
719
supports_cherrypick = True
720
supports_reverse_cherrypick = True
721
winner_idx = {"this": 2, "other": 1, "conflict": 1}
722
supports_lca_trees = True
724
def __init__(self, working_tree, this_tree, base_tree, other_tree,
725
interesting_ids=None, reprocess=False, show_base=False,
726
pb=None, pp=None, change_reporter=None,
727
interesting_files=None, do_merge=True,
728
cherrypick=False, lca_trees=None, this_branch=None):
729
"""Initialize the merger object and perform the merge.
731
:param working_tree: The working tree to apply the merge to
732
:param this_tree: The local tree in the merge operation
733
:param base_tree: The common tree in the merge operation
734
:param other_tree: The other tree to merge changes from
735
:param this_branch: The branch associated with this_tree. Defaults to
736
this_tree.branch if not supplied.
737
:param interesting_ids: The file_ids of files that should be
738
participate in the merge. May not be combined with
740
:param: reprocess If True, perform conflict-reduction processing.
741
:param show_base: If True, show the base revision in text conflicts.
742
(incompatible with reprocess)
744
:param pp: A ProgressPhase object
745
:param change_reporter: An object that should report changes made
746
:param interesting_files: The tree-relative paths of files that should
747
participate in the merge. If these paths refer to directories,
748
the contents of those directories will also be included. May not
749
be combined with interesting_ids. If neither interesting_files nor
750
interesting_ids is specified, all files may participate in the
752
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
753
if the ancestry was found to include a criss-cross merge.
754
Otherwise should be None.
756
object.__init__(self)
757
if interesting_files is not None and interesting_ids is not None:
759
'specify either interesting_ids or interesting_files')
760
if this_branch is None:
761
this_branch = this_tree.branch
762
self.interesting_ids = interesting_ids
763
self.interesting_files = interesting_files
764
self.this_tree = working_tree
765
self.base_tree = base_tree
766
self.other_tree = other_tree
767
self.this_branch = this_branch
768
self._raw_conflicts = []
769
self.cooked_conflicts = []
770
self.reprocess = reprocess
771
self.show_base = show_base
772
self._lca_trees = lca_trees
773
# Uncommenting this will change the default algorithm to always use
774
# _entries_lca. This can be useful for running the test suite and
775
# making sure we haven't missed any corner cases.
776
# if lca_trees is None:
777
# self._lca_trees = [self.base_tree]
778
self.change_reporter = change_reporter
779
self.cherrypick = cherrypick
783
warnings.warn("pp argument to Merge3Merger is deprecated")
785
warnings.warn("pb argument to Merge3Merger is deprecated")
788
operation = cleanup.OperationWithCleanups(self._do_merge)
789
self.this_tree.lock_tree_write()
790
operation.add_cleanup(self.this_tree.unlock)
791
self.base_tree.lock_read()
792
operation.add_cleanup(self.base_tree.unlock)
793
self.other_tree.lock_read()
794
operation.add_cleanup(self.other_tree.unlock)
797
def _do_merge(self, operation):
798
self.tt = transform.TreeTransform(self.this_tree, None)
799
operation.add_cleanup(self.tt.finalize)
800
self._compute_transform()
801
results = self.tt.apply(no_conflicts=True)
802
self.write_modified(results)
804
self.this_tree.add_conflicts(self.cooked_conflicts)
805
except errors.UnsupportedOperation:
808
def make_preview_transform(self):
809
operation = cleanup.OperationWithCleanups(self._make_preview_transform)
810
self.base_tree.lock_read()
811
operation.add_cleanup(self.base_tree.unlock)
812
self.other_tree.lock_read()
813
operation.add_cleanup(self.other_tree.unlock)
814
return operation.run_simple()
816
def _make_preview_transform(self):
817
self.tt = transform.TransformPreview(self.this_tree)
818
self._compute_transform()
821
def _compute_transform(self):
822
if self._lca_trees is None:
823
entries = self._entries3()
824
resolver = self._three_way
826
entries = self._entries_lca()
827
resolver = self._lca_multi_way
828
# Prepare merge hooks
829
factories = Merger.hooks['merge_file_content']
830
# One hook for each registered one plus our default merger
831
hooks = [factory(self) for factory in factories] + [self]
832
self.active_hooks = [hook for hook in hooks if hook is not None]
833
child_pb = ui.ui_factory.nested_progress_bar()
835
for num, (file_id, changed, parents3, names3,
836
executable3) in enumerate(entries):
837
# Try merging each entry
838
child_pb.update(gettext('Preparing file merge'),
840
self._merge_names(file_id, parents3, names3, resolver=resolver)
842
file_status = self._do_merge_contents(file_id)
844
file_status = 'unmodified'
845
self._merge_executable(file_id,
846
executable3, file_status, resolver=resolver)
849
self.tt.fixup_new_roots()
850
self._finish_computing_transform()
852
def _finish_computing_transform(self):
853
"""Finalize the transform and report the changes.
855
This is the second half of _compute_transform.
857
child_pb = ui.ui_factory.nested_progress_bar()
859
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
860
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
863
if self.change_reporter is not None:
864
from bzrlib import delta
865
delta.report_changes(
866
self.tt.iter_changes(), self.change_reporter)
867
self.cook_conflicts(fs_conflicts)
868
for conflict in self.cooked_conflicts:
869
trace.warning(unicode(conflict))
872
"""Gather data about files modified between three trees.
874
Return a list of tuples of file_id, changed, parents3, names3,
875
executable3. changed is a boolean indicating whether the file contents
876
or kind were changed. parents3 is a tuple of parent ids for base,
877
other and this. names3 is a tuple of names for base, other and this.
878
executable3 is a tuple of execute-bit values for base, other and this.
881
iterator = self.other_tree.iter_changes(self.base_tree,
882
specific_files=self.interesting_files,
883
extra_trees=[self.this_tree])
884
this_entries = dict((e.file_id, e) for p, e in
885
self.this_tree.iter_entries_by_dir(
886
self.interesting_ids))
887
for (file_id, paths, changed, versioned, parents, names, kind,
888
executable) in iterator:
889
if (self.interesting_ids is not None and
890
file_id not in self.interesting_ids):
892
entry = this_entries.get(file_id)
893
if entry is not None:
894
this_name = entry.name
895
this_parent = entry.parent_id
896
this_executable = entry.executable
900
this_executable = None
901
parents3 = parents + (this_parent,)
902
names3 = names + (this_name,)
903
executable3 = executable + (this_executable,)
904
result.append((file_id, changed, parents3, names3, executable3))
907
def _entries_lca(self):
908
"""Gather data about files modified between multiple trees.
910
This compares OTHER versus all LCA trees, and for interesting entries,
911
it then compares with THIS and BASE.
913
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
915
:return: [(file_id, changed, parents, names, executable)], where:
917
* file_id: Simple file_id of the entry
918
* changed: Boolean, True if the kind or contents changed else False
919
* parents: ((base, [parent_id, in, lcas]), parent_id_other,
921
* names: ((base, [name, in, lcas]), name_in_other, name_in_this)
922
* executable: ((base, [exec, in, lcas]), exec_in_other,
925
if self.interesting_files is not None:
926
lookup_trees = [self.this_tree, self.base_tree]
927
lookup_trees.extend(self._lca_trees)
928
# I think we should include the lca trees as well
929
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
932
interesting_ids = self.interesting_ids
934
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
936
base_inventory = self.base_tree.inventory
937
this_inventory = self.this_tree.inventory
938
for path, file_id, other_ie, lca_values in walker.iter_all():
939
# Is this modified at all from any of the other trees?
941
other_ie = _none_entry
942
if interesting_ids is not None and file_id not in interesting_ids:
945
# If other_revision is found in any of the lcas, that means this
946
# node is uninteresting. This is because when merging, if there are
947
# multiple heads(), we have to create a new node. So if we didn't,
948
# we know that the ancestry is linear, and that OTHER did not
950
# See doc/developers/lca_merge_resolution.txt for details
951
other_revision = other_ie.revision
952
if other_revision is not None:
953
# We can't use this shortcut when other_revision is None,
954
# because it may be None because things are WorkingTrees, and
955
# not because it is *actually* None.
956
is_unmodified = False
957
for lca_path, ie in lca_values:
958
if ie is not None and ie.revision == other_revision:
965
for lca_path, lca_ie in lca_values:
967
lca_entries.append(_none_entry)
969
lca_entries.append(lca_ie)
971
if base_inventory.has_id(file_id):
972
base_ie = base_inventory[file_id]
974
base_ie = _none_entry
976
if this_inventory.has_id(file_id):
977
this_ie = this_inventory[file_id]
979
this_ie = _none_entry
985
for lca_ie in lca_entries:
986
lca_kinds.append(lca_ie.kind)
987
lca_parent_ids.append(lca_ie.parent_id)
988
lca_names.append(lca_ie.name)
989
lca_executable.append(lca_ie.executable)
991
kind_winner = self._lca_multi_way(
992
(base_ie.kind, lca_kinds),
993
other_ie.kind, this_ie.kind)
994
parent_id_winner = self._lca_multi_way(
995
(base_ie.parent_id, lca_parent_ids),
996
other_ie.parent_id, this_ie.parent_id)
997
name_winner = self._lca_multi_way(
998
(base_ie.name, lca_names),
999
other_ie.name, this_ie.name)
1001
content_changed = True
1002
if kind_winner == 'this':
1003
# No kind change in OTHER, see if there are *any* changes
1004
if other_ie.kind == 'directory':
1005
if parent_id_winner == 'this' and name_winner == 'this':
1006
# No change for this directory in OTHER, skip
1008
content_changed = False
1009
elif other_ie.kind is None or other_ie.kind == 'file':
1010
def get_sha1(ie, tree):
1011
if ie.kind != 'file':
1013
return tree.get_file_sha1(file_id)
1014
base_sha1 = get_sha1(base_ie, self.base_tree)
1015
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1016
in zip(lca_entries, self._lca_trees)]
1017
this_sha1 = get_sha1(this_ie, self.this_tree)
1018
other_sha1 = get_sha1(other_ie, self.other_tree)
1019
sha1_winner = self._lca_multi_way(
1020
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1021
allow_overriding_lca=False)
1022
exec_winner = self._lca_multi_way(
1023
(base_ie.executable, lca_executable),
1024
other_ie.executable, this_ie.executable)
1025
if (parent_id_winner == 'this' and name_winner == 'this'
1026
and sha1_winner == 'this' and exec_winner == 'this'):
1027
# No kind, parent, name, exec, or content change for
1028
# OTHER, so this node is not considered interesting
1030
if sha1_winner == 'this':
1031
content_changed = False
1032
elif other_ie.kind == 'symlink':
1033
def get_target(ie, tree):
1034
if ie.kind != 'symlink':
1036
return tree.get_symlink_target(file_id)
1037
base_target = get_target(base_ie, self.base_tree)
1038
lca_targets = [get_target(ie, tree) for ie, tree
1039
in zip(lca_entries, self._lca_trees)]
1040
this_target = get_target(this_ie, self.this_tree)
1041
other_target = get_target(other_ie, self.other_tree)
1042
target_winner = self._lca_multi_way(
1043
(base_target, lca_targets),
1044
other_target, this_target)
1045
if (parent_id_winner == 'this' and name_winner == 'this'
1046
and target_winner == 'this'):
1047
# No kind, parent, name, or symlink target change
1050
if target_winner == 'this':
1051
content_changed = False
1052
elif other_ie.kind == 'tree-reference':
1053
# The 'changed' information seems to be handled at a higher
1054
# level. At least, _entries3 returns False for content
1055
# changed, even when at a new revision_id.
1056
content_changed = False
1057
if (parent_id_winner == 'this' and name_winner == 'this'):
1058
# Nothing interesting
1061
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1063
# If we have gotten this far, that means something has changed
1064
result.append((file_id, content_changed,
1065
((base_ie.parent_id, lca_parent_ids),
1066
other_ie.parent_id, this_ie.parent_id),
1067
((base_ie.name, lca_names),
1068
other_ie.name, this_ie.name),
1069
((base_ie.executable, lca_executable),
1070
other_ie.executable, this_ie.executable)
1074
@deprecated_method(deprecated_in((2, 4, 0)))
1076
if self.tt.final_kind(self.tt.root) is None:
1077
self.tt.cancel_deletion(self.tt.root)
1078
if self.tt.final_file_id(self.tt.root) is None:
1079
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
1081
other_root_file_id = self.other_tree.get_root_id()
1082
if other_root_file_id is None:
1084
other_root = self.tt.trans_id_file_id(other_root_file_id)
1085
if other_root == self.tt.root:
1087
if self.this_tree.inventory.has_id(
1088
self.other_tree.inventory.root.file_id):
1089
# the other tree's root is a non-root in the current tree (as
1090
# when a previously unrelated branch is merged into another)
1092
if self.tt.final_kind(other_root) is not None:
1093
other_root_is_present = True
1095
# other_root doesn't have a physical representation. We still need
1096
# to move any references to the actual root of the tree.
1097
other_root_is_present = False
1098
# 'other_tree.inventory.root' is not present in this tree. We are
1099
# calling adjust_path for children which *want* to be present with a
1100
# correct place to go.
1101
for _, child in self.other_tree.inventory.root.children.iteritems():
1102
trans_id = self.tt.trans_id_file_id(child.file_id)
1103
if not other_root_is_present:
1104
if self.tt.final_kind(trans_id) is not None:
1105
# The item exist in the final tree and has a defined place
1108
# Move the item into the root
1110
final_name = self.tt.final_name(trans_id)
1111
except errors.NoFinalPath:
1112
# This file is not present anymore, ignore it.
1114
self.tt.adjust_path(final_name, self.tt.root, trans_id)
1115
if other_root_is_present:
1116
self.tt.cancel_creation(other_root)
1117
self.tt.cancel_versioning(other_root)
1119
def write_modified(self, results):
1120
modified_hashes = {}
1121
for path in results.modified_paths:
1122
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
1125
hash = self.this_tree.get_file_sha1(file_id)
1128
modified_hashes[file_id] = hash
1129
self.this_tree.set_merge_modified(modified_hashes)
1132
def parent(entry, file_id):
1133
"""Determine the parent for a file_id (used as a key method)"""
1136
return entry.parent_id
1139
def name(entry, file_id):
1140
"""Determine the name for a file_id (used as a key method)"""
1146
def contents_sha1(tree, file_id):
1147
"""Determine the sha1 of the file contents (used as a key method)."""
1148
if not tree.has_id(file_id):
1150
return tree.get_file_sha1(file_id)
1153
def executable(tree, file_id):
1154
"""Determine the executability of a file-id (used as a key method)."""
1155
if not tree.has_id(file_id):
1157
if tree.kind(file_id) != "file":
1159
return tree.is_executable(file_id)
1162
def kind(tree, file_id):
1163
"""Determine the kind of a file-id (used as a key method)."""
1164
if not tree.has_id(file_id):
1166
return tree.kind(file_id)
1169
def _three_way(base, other, this):
1171
# if 'base == other', either they all agree, or only 'this' has
1174
elif this not in (base, other):
1175
# 'this' is neither 'base' nor 'other', so both sides changed
1178
# "Ambiguous clean merge" -- both sides have made the same change.
1181
# this == base: only other has changed.
1185
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
1186
"""Consider LCAs when determining whether a change has occurred.
1188
If LCAS are all identical, this is the same as a _three_way comparison.
1190
:param bases: value in (BASE, [LCAS])
1191
:param other: value in OTHER
1192
:param this: value in THIS
1193
:param allow_overriding_lca: If there is more than one unique lca
1194
value, allow OTHER to override THIS if it has a new value, and
1195
THIS only has an lca value, or vice versa. This is appropriate for
1196
truly scalar values, not as much for non-scalars.
1197
:return: 'this', 'other', or 'conflict' depending on whether an entry
1200
# See doc/developers/lca_tree_merging.txt for details about this
1203
# Either Ambiguously clean, or nothing was actually changed. We
1206
base_val, lca_vals = bases
1207
# Remove 'base_val' from the lca_vals, because it is not interesting
1208
filtered_lca_vals = [lca_val for lca_val in lca_vals
1209
if lca_val != base_val]
1210
if len(filtered_lca_vals) == 0:
1211
return Merge3Merger._three_way(base_val, other, this)
1213
unique_lca_vals = set(filtered_lca_vals)
1214
if len(unique_lca_vals) == 1:
1215
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1217
if allow_overriding_lca:
1218
if other in unique_lca_vals:
1219
if this in unique_lca_vals:
1220
# Each side picked a different lca, conflict
1223
# This has a value which supersedes both lca values, and
1224
# other only has an lca value
1226
elif this in unique_lca_vals:
1227
# OTHER has a value which supersedes both lca values, and this
1228
# only has an lca value
1231
# At this point, the lcas disagree, and the tip disagree
1235
@deprecated_method(deprecated_in((2, 2, 0)))
1236
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1237
"""Do a three-way test on a scalar.
1238
Return "this", "other" or "conflict", depending whether a value wins.
1240
key_base = key(base_tree, file_id)
1241
key_other = key(other_tree, file_id)
1242
#if base == other, either they all agree, or only THIS has changed.
1243
if key_base == key_other:
1245
key_this = key(this_tree, file_id)
1246
# "Ambiguous clean merge"
1247
if key_this == key_other:
1249
elif key_this == key_base:
1254
def merge_names(self, file_id):
1255
def get_entry(tree):
1256
if tree.has_id(file_id):
1257
return tree.inventory[file_id]
1260
this_entry = get_entry(self.this_tree)
1261
other_entry = get_entry(self.other_tree)
1262
base_entry = get_entry(self.base_tree)
1263
entries = (base_entry, other_entry, this_entry)
1266
for entry in entries:
1269
parents.append(None)
1271
names.append(entry.name)
1272
parents.append(entry.parent_id)
1273
return self._merge_names(file_id, parents, names,
1274
resolver=self._three_way)
1276
def _merge_names(self, file_id, parents, names, resolver):
1277
"""Perform a merge on file_id names and parents"""
1278
base_name, other_name, this_name = names
1279
base_parent, other_parent, this_parent = parents
1281
name_winner = resolver(*names)
1283
parent_id_winner = resolver(*parents)
1284
if this_name is None:
1285
if name_winner == "this":
1286
name_winner = "other"
1287
if parent_id_winner == "this":
1288
parent_id_winner = "other"
1289
if name_winner == "this" and parent_id_winner == "this":
1291
if name_winner == 'conflict' or parent_id_winner == 'conflict':
1292
# Creating helpers (.OTHER or .THIS) here cause problems down the
1293
# road if a ContentConflict needs to be created so we should not do
1295
trans_id = self.tt.trans_id_file_id(file_id)
1296
self._raw_conflicts.append(('path conflict', trans_id, file_id,
1297
this_parent, this_name,
1298
other_parent, other_name))
1299
if not self.other_tree.has_id(file_id):
1300
# it doesn't matter whether the result was 'other' or
1301
# 'conflict'-- if it has no file id, we leave it alone.
1303
parent_id = parents[self.winner_idx[parent_id_winner]]
1304
name = names[self.winner_idx[name_winner]]
1305
if parent_id is not None or name is not None:
1306
# if we get here, name_winner and parent_winner are set to safe
1308
if parent_id is None and name is not None:
1309
# if parent_id is None and name is non-None, current file is
1311
if names[self.winner_idx[parent_id_winner]] != '':
1312
raise AssertionError(
1313
'File looks like a root, but named %s' %
1314
names[self.winner_idx[parent_id_winner]])
1315
parent_trans_id = transform.ROOT_PARENT
1317
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1318
self.tt.adjust_path(name, parent_trans_id,
1319
self.tt.trans_id_file_id(file_id))
1321
def _do_merge_contents(self, file_id):
1322
"""Performs a merge on file_id contents."""
1323
def contents_pair(tree):
1324
if not tree.has_id(file_id):
1326
kind = tree.kind(file_id)
1328
contents = tree.get_file_sha1(file_id)
1329
elif kind == "symlink":
1330
contents = tree.get_symlink_target(file_id)
1333
return kind, contents
1335
# See SPOT run. run, SPOT, run.
1336
# So we're not QUITE repeating ourselves; we do tricky things with
1338
base_pair = contents_pair(self.base_tree)
1339
other_pair = contents_pair(self.other_tree)
1341
this_pair = contents_pair(self.this_tree)
1342
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1343
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1344
this_pair, allow_overriding_lca=False)
1346
if base_pair == other_pair:
1349
# We delayed evaluating this_pair as long as we can to avoid
1350
# unnecessary sha1 calculation
1351
this_pair = contents_pair(self.this_tree)
1352
winner = self._three_way(base_pair, other_pair, this_pair)
1353
if winner == 'this':
1354
# No interesting changes introduced by OTHER
1356
# We have a hypothetical conflict, but if we have files, then we
1357
# can try to merge the content
1358
trans_id = self.tt.trans_id_file_id(file_id)
1359
params = MergeFileHookParams(self, file_id, trans_id, this_pair[0],
1360
other_pair[0], winner)
1361
hooks = self.active_hooks
1362
hook_status = 'not_applicable'
1364
hook_status, lines = hook.merge_contents(params)
1365
if hook_status != 'not_applicable':
1366
# Don't try any more hooks, this one applies.
1368
# If the merge ends up replacing the content of the file, we get rid of
1369
# it at the end of this method (this variable is used to track the
1370
# exceptions to this rule).
1373
if hook_status == 'not_applicable':
1374
# No merge hook was able to resolve the situation. Two cases exist:
1375
# a content conflict or a duplicate one.
1377
name = self.tt.final_name(trans_id)
1378
parent_id = self.tt.final_parent(trans_id)
1380
inhibit_content_conflict = False
1381
if params.this_kind is None: # file_id is not in THIS
1382
# Is the name used for a different file_id ?
1383
dupe_path = self.other_tree.id2path(file_id)
1384
this_id = self.this_tree.path2id(dupe_path)
1385
if this_id is not None:
1386
# Two entries for the same path
1388
# versioning the merged file will trigger a duplicate
1390
self.tt.version_file(file_id, trans_id)
1391
transform.create_from_tree(
1392
self.tt, trans_id, self.other_tree, file_id,
1393
filter_tree_path=self._get_filter_tree_path(file_id))
1394
inhibit_content_conflict = True
1395
elif params.other_kind is None: # file_id is not in OTHER
1396
# Is the name used for a different file_id ?
1397
dupe_path = self.this_tree.id2path(file_id)
1398
other_id = self.other_tree.path2id(dupe_path)
1399
if other_id is not None:
1400
# Two entries for the same path again, but here, the other
1401
# entry will also be merged. We simply inhibit the
1402
# 'content' conflict creation because we know OTHER will
1403
# create (or has already created depending on ordering) an
1404
# entry at the same path. This will trigger a 'duplicate'
1407
inhibit_content_conflict = True
1408
if not inhibit_content_conflict:
1409
if params.this_kind is not None:
1410
self.tt.unversion_file(trans_id)
1411
# This is a contents conflict, because none of the available
1412
# functions could merge it.
1413
file_group = self._dump_conflicts(name, parent_id, file_id,
1415
self._raw_conflicts.append(('contents conflict', file_group))
1416
elif hook_status == 'success':
1417
self.tt.create_file(lines, trans_id)
1418
elif hook_status == 'conflicted':
1419
# XXX: perhaps the hook should be able to provide
1420
# the BASE/THIS/OTHER files?
1421
self.tt.create_file(lines, trans_id)
1422
self._raw_conflicts.append(('text conflict', trans_id))
1423
name = self.tt.final_name(trans_id)
1424
parent_id = self.tt.final_parent(trans_id)
1425
self._dump_conflicts(name, parent_id, file_id)
1426
elif hook_status == 'delete':
1427
self.tt.unversion_file(trans_id)
1429
elif hook_status == 'done':
1430
# The hook function did whatever it needs to do directly, no
1431
# further action needed here.
1434
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1435
if not self.this_tree.has_id(file_id) and result == "modified":
1436
self.tt.version_file(file_id, trans_id)
1438
# The merge has been performed and produced a new content, so the
1439
# old contents should not be retained.
1440
self.tt.delete_contents(trans_id)
1443
def _default_other_winner_merge(self, merge_hook_params):
1444
"""Replace this contents with other."""
1445
file_id = merge_hook_params.file_id
1446
trans_id = merge_hook_params.trans_id
1447
if self.other_tree.has_id(file_id):
1448
# OTHER changed the file
1449
transform.create_from_tree(
1450
self.tt, trans_id, self.other_tree, file_id,
1451
filter_tree_path=self._get_filter_tree_path(file_id))
1453
elif self.this_tree.has_id(file_id):
1454
# OTHER deleted the file
1455
return 'delete', None
1457
raise AssertionError(
1458
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1461
def merge_contents(self, merge_hook_params):
1462
"""Fallback merge logic after user installed hooks."""
1463
# This function is used in merge hooks as the fallback instance.
1464
# Perhaps making this function and the functions it calls be a
1465
# a separate class would be better.
1466
if merge_hook_params.winner == 'other':
1467
# OTHER is a straight winner, so replace this contents with other
1468
return self._default_other_winner_merge(merge_hook_params)
1469
elif merge_hook_params.is_file_merge():
1470
# THIS and OTHER are both files, so text merge. Either
1471
# BASE is a file, or both converted to files, so at least we
1472
# have agreement that output should be a file.
1474
self.text_merge(merge_hook_params.file_id,
1475
merge_hook_params.trans_id)
1476
except errors.BinaryFile:
1477
return 'not_applicable', None
1480
return 'not_applicable', None
1482
def get_lines(self, tree, file_id):
1483
"""Return the lines in a file, or an empty list."""
1484
if tree.has_id(file_id):
1485
return tree.get_file_lines(file_id)
1489
def text_merge(self, file_id, trans_id):
1490
"""Perform a three-way text merge on a file_id"""
1491
# it's possible that we got here with base as a different type.
1492
# if so, we just want two-way text conflicts.
1493
if self.base_tree.has_id(file_id) and \
1494
self.base_tree.kind(file_id) == "file":
1495
base_lines = self.get_lines(self.base_tree, file_id)
1498
other_lines = self.get_lines(self.other_tree, file_id)
1499
this_lines = self.get_lines(self.this_tree, file_id)
1500
m3 = merge3.Merge3(base_lines, this_lines, other_lines,
1501
is_cherrypick=self.cherrypick)
1502
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1503
if self.show_base is True:
1504
base_marker = '|' * 7
1508
def iter_merge3(retval):
1509
retval["text_conflicts"] = False
1510
for line in m3.merge_lines(name_a = "TREE",
1511
name_b = "MERGE-SOURCE",
1512
name_base = "BASE-REVISION",
1513
start_marker=start_marker,
1514
base_marker=base_marker,
1515
reprocess=self.reprocess):
1516
if line.startswith(start_marker):
1517
retval["text_conflicts"] = True
1518
yield line.replace(start_marker, '<' * 7)
1522
merge3_iterator = iter_merge3(retval)
1523
self.tt.create_file(merge3_iterator, trans_id)
1524
if retval["text_conflicts"] is True:
1525
self._raw_conflicts.append(('text conflict', trans_id))
1526
name = self.tt.final_name(trans_id)
1527
parent_id = self.tt.final_parent(trans_id)
1528
file_group = self._dump_conflicts(name, parent_id, file_id,
1529
this_lines, base_lines,
1531
file_group.append(trans_id)
1534
def _get_filter_tree_path(self, file_id):
1535
if self.this_tree.supports_content_filtering():
1536
# We get the path from the working tree if it exists.
1537
# That fails though when OTHER is adding a file, so
1538
# we fall back to the other tree to find the path if
1539
# it doesn't exist locally.
1541
return self.this_tree.id2path(file_id)
1542
except errors.NoSuchId:
1543
return self.other_tree.id2path(file_id)
1544
# Skip the id2path lookup for older formats
1547
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1548
base_lines=None, other_lines=None, set_version=False,
1550
"""Emit conflict files.
1551
If this_lines, base_lines, or other_lines are omitted, they will be
1552
determined automatically. If set_version is true, the .OTHER, .THIS
1553
or .BASE (in that order) will be created as versioned files.
1555
data = [('OTHER', self.other_tree, other_lines),
1556
('THIS', self.this_tree, this_lines)]
1558
data.append(('BASE', self.base_tree, base_lines))
1560
# We need to use the actual path in the working tree of the file here,
1561
# ignoring the conflict suffixes
1563
if wt.supports_content_filtering():
1565
filter_tree_path = wt.id2path(file_id)
1566
except errors.NoSuchId:
1567
# file has been deleted
1568
filter_tree_path = None
1570
# Skip the id2path lookup for older formats
1571
filter_tree_path = None
1575
for suffix, tree, lines in data:
1576
if tree.has_id(file_id):
1577
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1578
suffix, lines, filter_tree_path)
1579
file_group.append(trans_id)
1580
if set_version and not versioned:
1581
self.tt.version_file(file_id, trans_id)
1585
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1586
lines=None, filter_tree_path=None):
1587
"""Emit a single conflict file."""
1588
name = name + '.' + suffix
1589
trans_id = self.tt.create_path(name, parent_id)
1590
transform.create_from_tree(self.tt, trans_id, tree, file_id, lines,
1594
def merge_executable(self, file_id, file_status):
1595
"""Perform a merge on the execute bit."""
1596
executable = [self.executable(t, file_id) for t in (self.base_tree,
1597
self.other_tree, self.this_tree)]
1598
self._merge_executable(file_id, executable, file_status,
1599
resolver=self._three_way)
1601
def _merge_executable(self, file_id, executable, file_status,
1603
"""Perform a merge on the execute bit."""
1604
base_executable, other_executable, this_executable = executable
1605
if file_status == "deleted":
1607
winner = resolver(*executable)
1608
if winner == "conflict":
1609
# There must be a None in here, if we have a conflict, but we
1610
# need executability since file status was not deleted.
1611
if self.executable(self.other_tree, file_id) is None:
1615
if winner == 'this' and file_status != "modified":
1617
trans_id = self.tt.trans_id_file_id(file_id)
1618
if self.tt.final_kind(trans_id) != "file":
1620
if winner == "this":
1621
executability = this_executable
1623
if self.other_tree.has_id(file_id):
1624
executability = other_executable
1625
elif self.this_tree.has_id(file_id):
1626
executability = this_executable
1627
elif self.base_tree_has_id(file_id):
1628
executability = base_executable
1629
if executability is not None:
1630
trans_id = self.tt.trans_id_file_id(file_id)
1631
self.tt.set_executability(executability, trans_id)
1633
def cook_conflicts(self, fs_conflicts):
1634
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1635
content_conflict_file_ids = set()
1636
cooked_conflicts = transform.cook_conflicts(fs_conflicts, self.tt)
1637
fp = transform.FinalPaths(self.tt)
1638
for conflict in self._raw_conflicts:
1639
conflict_type = conflict[0]
1640
if conflict_type == 'path conflict':
1642
this_parent, this_name,
1643
other_parent, other_name) = conflict[1:]
1644
if this_parent is None or this_name is None:
1645
this_path = '<deleted>'
1647
parent_path = fp.get_path(
1648
self.tt.trans_id_file_id(this_parent))
1649
this_path = osutils.pathjoin(parent_path, this_name)
1650
if other_parent is None or other_name is None:
1651
other_path = '<deleted>'
1653
if other_parent == self.other_tree.get_root_id():
1654
# The tree transform doesn't know about the other root,
1655
# so we special case here to avoid a NoFinalPath
1659
parent_path = fp.get_path(
1660
self.tt.trans_id_file_id(other_parent))
1661
other_path = osutils.pathjoin(parent_path, other_name)
1662
c = _mod_conflicts.Conflict.factory(
1663
'path conflict', path=this_path,
1664
conflict_path=other_path,
1666
elif conflict_type == 'contents conflict':
1667
for trans_id in conflict[1]:
1668
file_id = self.tt.final_file_id(trans_id)
1669
if file_id is not None:
1670
# Ok we found the relevant file-id
1672
path = fp.get_path(trans_id)
1673
for suffix in ('.BASE', '.THIS', '.OTHER'):
1674
if path.endswith(suffix):
1675
# Here is the raw path
1676
path = path[:-len(suffix)]
1678
c = _mod_conflicts.Conflict.factory(conflict_type,
1679
path=path, file_id=file_id)
1680
content_conflict_file_ids.add(file_id)
1681
elif conflict_type == 'text conflict':
1682
trans_id = conflict[1]
1683
path = fp.get_path(trans_id)
1684
file_id = self.tt.final_file_id(trans_id)
1685
c = _mod_conflicts.Conflict.factory(conflict_type,
1686
path=path, file_id=file_id)
1688
raise AssertionError('bad conflict type: %r' % (conflict,))
1689
cooked_conflicts.append(c)
1691
self.cooked_conflicts = []
1692
# We want to get rid of path conflicts when a corresponding contents
1693
# conflict exists. This can occur when one branch deletes a file while
1694
# the other renames *and* modifies it. In this case, the content
1695
# conflict is enough.
1696
for c in cooked_conflicts:
1697
if (c.typestring == 'path conflict'
1698
and c.file_id in content_conflict_file_ids):
1700
self.cooked_conflicts.append(c)
1701
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
1704
class WeaveMerger(Merge3Merger):
1705
"""Three-way tree merger, text weave merger."""
1706
supports_reprocess = True
1707
supports_show_base = False
1708
supports_reverse_cherrypick = False
1709
history_based = True
1711
def _generate_merge_plan(self, file_id, base):
1712
return self.this_tree.plan_file_merge(file_id, self.other_tree,
1715
def _merged_lines(self, file_id):
1716
"""Generate the merged lines.
1717
There is no distinction between lines that are meant to contain <<<<<<<
1721
base = self.base_tree
1724
plan = self._generate_merge_plan(file_id, base)
1725
if 'merge' in debug.debug_flags:
1727
trans_id = self.tt.trans_id_file_id(file_id)
1728
name = self.tt.final_name(trans_id) + '.plan'
1729
contents = ('%11s|%s' % l for l in plan)
1730
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1731
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1732
'>>>>>>> MERGE-SOURCE\n')
1733
lines, conflicts = textmerge.merge_lines(self.reprocess)
1735
base_lines = textmerge.base_from_plan()
1738
return lines, base_lines
1740
def text_merge(self, file_id, trans_id):
1741
"""Perform a (weave) text merge for a given file and file-id.
1742
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1743
and a conflict will be noted.
1745
lines, base_lines = self._merged_lines(file_id)
1747
# Note we're checking whether the OUTPUT is binary in this case,
1748
# because we don't want to get into weave merge guts.
1749
textfile.check_text_lines(lines)
1750
self.tt.create_file(lines, trans_id)
1751
if base_lines is not None:
1753
self._raw_conflicts.append(('text conflict', trans_id))
1754
name = self.tt.final_name(trans_id)
1755
parent_id = self.tt.final_parent(trans_id)
1756
file_group = self._dump_conflicts(name, parent_id, file_id,
1758
base_lines=base_lines)
1759
file_group.append(trans_id)
1762
class LCAMerger(WeaveMerger):
1764
def _generate_merge_plan(self, file_id, base):
1765
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1768
class Diff3Merger(Merge3Merger):
1769
"""Three-way merger using external diff3 for text merging"""
1771
def dump_file(self, temp_dir, name, tree, file_id):
1772
out_path = osutils.pathjoin(temp_dir, name)
1773
out_file = open(out_path, "wb")
1775
in_file = tree.get_file(file_id)
1776
for line in in_file:
1777
out_file.write(line)
1782
def text_merge(self, file_id, trans_id):
1783
"""Perform a diff3 merge using a specified file-id and trans-id.
1784
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1785
will be dumped, and a will be conflict noted.
1788
temp_dir = osutils.mkdtemp(prefix="bzr-")
1790
new_file = osutils.pathjoin(temp_dir, "new")
1791
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1792
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1793
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1794
status = bzrlib.patch.diff3(new_file, this, base, other)
1795
if status not in (0, 1):
1796
raise errors.BzrError("Unhandled diff3 exit code")
1797
f = open(new_file, 'rb')
1799
self.tt.create_file(f, trans_id)
1803
name = self.tt.final_name(trans_id)
1804
parent_id = self.tt.final_parent(trans_id)
1805
self._dump_conflicts(name, parent_id, file_id)
1806
self._raw_conflicts.append(('text conflict', trans_id))
1808
osutils.rmtree(temp_dir)
1811
class PathNotInTree(errors.BzrError):
1813
_fmt = """Merge-into failed because %(tree)s does not contain %(path)s."""
1815
def __init__(self, path, tree):
1816
errors.BzrError.__init__(self, path=path, tree=tree)
1819
class MergeIntoMerger(Merger):
1820
"""Merger that understands other_tree will be merged into a subdir.
1822
This also changes the Merger api so that it uses real Branch, revision_id,
1823
and RevisonTree objects, rather than using revision specs.
1826
def __init__(self, this_tree, other_branch, other_tree, target_subdir,
1827
source_subpath, other_rev_id=None):
1828
"""Create a new MergeIntoMerger object.
1830
source_subpath in other_tree will be effectively copied to
1831
target_subdir in this_tree.
1833
:param this_tree: The tree that we will be merging into.
1834
:param other_branch: The Branch we will be merging from.
1835
:param other_tree: The RevisionTree object we want to merge.
1836
:param target_subdir: The relative path where we want to merge
1837
other_tree into this_tree
1838
:param source_subpath: The relative path specifying the subtree of
1839
other_tree to merge into this_tree.
1841
# It is assumed that we are merging a tree that is not in our current
1842
# ancestry, which means we are using the "EmptyTree" as our basis.
1843
null_ancestor_tree = this_tree.branch.repository.revision_tree(
1844
_mod_revision.NULL_REVISION)
1845
super(MergeIntoMerger, self).__init__(
1846
this_branch=this_tree.branch,
1847
this_tree=this_tree,
1848
other_tree=other_tree,
1849
base_tree=null_ancestor_tree,
1851
self._target_subdir = target_subdir
1852
self._source_subpath = source_subpath
1853
self.other_branch = other_branch
1854
if other_rev_id is None:
1855
other_rev_id = other_tree.get_revision_id()
1856
self.other_rev_id = self.other_basis = other_rev_id
1857
self.base_is_ancestor = True
1858
self.backup_files = True
1859
self.merge_type = Merge3Merger
1860
self.show_base = False
1861
self.reprocess = False
1862
self.interesting_ids = None
1863
self.merge_type = _MergeTypeParameterizer(MergeIntoMergeType,
1864
target_subdir=self._target_subdir,
1865
source_subpath=self._source_subpath)
1866
if self._source_subpath != '':
1867
# If this isn't a partial merge make sure the revisions will be
1869
self._maybe_fetch(self.other_branch, self.this_branch,
1872
def set_pending(self):
1873
if self._source_subpath != '':
1875
Merger.set_pending(self)
1878
class _MergeTypeParameterizer(object):
1879
"""Wrap a merge-type class to provide extra parameters.
1881
This is hack used by MergeIntoMerger to pass some extra parameters to its
1882
merge_type. Merger.do_merge() sets up its own set of parameters to pass to
1883
the 'merge_type' member. It is difficult override do_merge without
1884
re-writing the whole thing, so instead we create a wrapper which will pass
1885
the extra parameters.
1888
def __init__(self, merge_type, **kwargs):
1889
self._extra_kwargs = kwargs
1890
self._merge_type = merge_type
1892
def __call__(self, *args, **kwargs):
1893
kwargs.update(self._extra_kwargs)
1894
return self._merge_type(*args, **kwargs)
1896
def __getattr__(self, name):
1897
return getattr(self._merge_type, name)
1900
class MergeIntoMergeType(Merge3Merger):
1901
"""Merger that incorporates a tree (or part of a tree) into another."""
1903
def __init__(self, *args, **kwargs):
1904
"""Initialize the merger object.
1906
:param args: See Merge3Merger.__init__'s args.
1907
:param kwargs: See Merge3Merger.__init__'s keyword args, except for
1908
source_subpath and target_subdir.
1909
:keyword source_subpath: The relative path specifying the subtree of
1910
other_tree to merge into this_tree.
1911
:keyword target_subdir: The relative path where we want to merge
1912
other_tree into this_tree
1914
# All of the interesting work happens during Merge3Merger.__init__(),
1915
# so we have have to hack in to get our extra parameters set.
1916
self._source_subpath = kwargs.pop('source_subpath')
1917
self._target_subdir = kwargs.pop('target_subdir')
1918
super(MergeIntoMergeType, self).__init__(*args, **kwargs)
1920
def _compute_transform(self):
1921
child_pb = ui.ui_factory.nested_progress_bar()
1923
entries = self._entries_to_incorporate()
1924
entries = list(entries)
1925
for num, (entry, parent_id) in enumerate(entries):
1926
child_pb.update(gettext('Preparing file merge'), num, len(entries))
1927
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1928
trans_id = transform.new_by_entry(self.tt, entry,
1929
parent_trans_id, self.other_tree)
1932
self._finish_computing_transform()
1934
def _entries_to_incorporate(self):
1935
"""Yields pairs of (inventory_entry, new_parent)."""
1936
other_inv = self.other_tree.inventory
1937
subdir_id = other_inv.path2id(self._source_subpath)
1938
if subdir_id is None:
1939
# XXX: The error would be clearer if it gave the URL of the source
1940
# branch, but we don't have a reference to that here.
1941
raise PathNotInTree(self._source_subpath, "Source tree")
1942
subdir = other_inv[subdir_id]
1943
parent_in_target = osutils.dirname(self._target_subdir)
1944
target_id = self.this_tree.inventory.path2id(parent_in_target)
1945
if target_id is None:
1946
raise PathNotInTree(self._target_subdir, "Target tree")
1947
name_in_target = osutils.basename(self._target_subdir)
1948
merge_into_root = subdir.copy()
1949
merge_into_root.name = name_in_target
1950
if self.this_tree.inventory.has_id(merge_into_root.file_id):
1951
# Give the root a new file-id.
1952
# This can happen fairly easily if the directory we are
1953
# incorporating is the root, and both trees have 'TREE_ROOT' as
1954
# their root_id. Users will expect this to Just Work, so we
1955
# change the file-id here.
1956
# Non-root file-ids could potentially conflict too. That's really
1957
# an edge case, so we don't do anything special for those. We let
1958
# them cause conflicts.
1959
merge_into_root.file_id = generate_ids.gen_file_id(name_in_target)
1960
yield (merge_into_root, target_id)
1961
if subdir.kind != 'directory':
1962
# No children, so we are done.
1964
for ignored_path, entry in other_inv.iter_entries_by_dir(subdir_id):
1965
parent_id = entry.parent_id
1966
if parent_id == subdir.file_id:
1967
# The root's parent ID has changed, so make sure children of
1968
# the root refer to the new ID.
1969
parent_id = merge_into_root.file_id
1970
yield (entry, parent_id)
1973
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1975
merge_type=Merge3Merger,
1976
interesting_ids=None,
1980
interesting_files=None,
1983
change_reporter=None):
1984
"""Primary interface for merging.
1986
Typical use is probably::
1988
merge_inner(branch, branch.get_revision_tree(other_revision),
1989
branch.get_revision_tree(base_revision))
1991
if this_tree is None:
1992
raise errors.BzrError("bzrlib.merge.merge_inner requires a this_tree "
1994
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1995
pb=pb, change_reporter=change_reporter)
1996
merger.backup_files = backup_files
1997
merger.merge_type = merge_type
1998
merger.interesting_ids = interesting_ids
1999
merger.ignore_zero = ignore_zero
2000
if interesting_files:
2002
raise ValueError('Only supply interesting_ids'
2003
' or interesting_files')
2004
merger.interesting_files = interesting_files
2005
merger.show_base = show_base
2006
merger.reprocess = reprocess
2007
merger.other_rev_id = other_rev_id
2008
merger.other_basis = other_rev_id
2009
get_revision_id = getattr(base_tree, 'get_revision_id', None)
2010
if get_revision_id is None:
2011
get_revision_id = base_tree.last_revision
2012
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
2013
merger.set_base_revision(get_revision_id(), this_branch)
2014
return merger.do_merge()
2017
merge_type_registry = registry.Registry()
2018
merge_type_registry.register('diff3', Diff3Merger,
2019
"Merge using external diff3.")
2020
merge_type_registry.register('lca', LCAMerger,
2021
"LCA-newness merge.")
2022
merge_type_registry.register('merge3', Merge3Merger,
2023
"Native diff3-style merge.")
2024
merge_type_registry.register('weave', WeaveMerger,
2025
"Weave-based merge.")
2028
def get_merge_type_registry():
2029
"""Merge type registry was previously in bzrlib.option
2031
This method provides a backwards compatible way to retrieve it.
2033
return merge_type_registry
2036
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
2037
def status_a(revision, text):
2038
if revision in ancestors_b:
2039
return 'killed-b', text
2041
return 'new-a', text
2043
def status_b(revision, text):
2044
if revision in ancestors_a:
2045
return 'killed-a', text
2047
return 'new-b', text
2049
plain_a = [t for (a, t) in annotated_a]
2050
plain_b = [t for (a, t) in annotated_b]
2051
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
2052
blocks = matcher.get_matching_blocks()
2055
for ai, bi, l in blocks:
2056
# process all mismatched sections
2057
# (last mismatched section is handled because blocks always
2058
# includes a 0-length last block)
2059
for revision, text in annotated_a[a_cur:ai]:
2060
yield status_a(revision, text)
2061
for revision, text in annotated_b[b_cur:bi]:
2062
yield status_b(revision, text)
2063
# and now the matched section
2066
for text_a in plain_a[ai:a_cur]:
2067
yield "unchanged", text_a
2070
class _PlanMergeBase(object):
2072
def __init__(self, a_rev, b_rev, vf, key_prefix):
2075
:param a_rev: Revision-id of one revision to merge
2076
:param b_rev: Revision-id of the other revision to merge
2077
:param vf: A VersionedFiles containing both revisions
2078
:param key_prefix: A prefix for accessing keys in vf, typically
2084
self._last_lines = None
2085
self._last_lines_revision_id = None
2086
self._cached_matching_blocks = {}
2087
self._key_prefix = key_prefix
2088
self._precache_tip_lines()
2090
def _precache_tip_lines(self):
2091
lines = self.get_lines([self.a_rev, self.b_rev])
2092
self.lines_a = lines[self.a_rev]
2093
self.lines_b = lines[self.b_rev]
2095
def get_lines(self, revisions):
2096
"""Get lines for revisions from the backing VersionedFiles.
2098
:raises RevisionNotPresent: on absent texts.
2100
keys = [(self._key_prefix + (rev,)) for rev in revisions]
2102
for record in self.vf.get_record_stream(keys, 'unordered', True):
2103
if record.storage_kind == 'absent':
2104
raise errors.RevisionNotPresent(record.key, self.vf)
2105
result[record.key[-1]] = osutils.chunks_to_lines(
2106
record.get_bytes_as('chunked'))
2109
def plan_merge(self):
2110
"""Generate a 'plan' for merging the two revisions.
2112
This involves comparing their texts and determining the cause of
2113
differences. If text A has a line and text B does not, then either the
2114
line was added to text A, or it was deleted from B. Once the causes
2115
are combined, they are written out in the format described in
2116
VersionedFile.plan_merge
2118
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
2119
unique_a, unique_b = self._unique_lines(blocks)
2120
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
2121
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
2122
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
2124
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
2127
for i, j, n in blocks:
2128
for a_index in range(last_i, i):
2129
if a_index in new_a:
2130
if a_index in killed_b:
2131
yield 'conflicted-a', self.lines_a[a_index]
2133
yield 'new-a', self.lines_a[a_index]
2135
yield 'killed-b', self.lines_a[a_index]
2136
for b_index in range(last_j, j):
2137
if b_index in new_b:
2138
if b_index in killed_a:
2139
yield 'conflicted-b', self.lines_b[b_index]
2141
yield 'new-b', self.lines_b[b_index]
2143
yield 'killed-a', self.lines_b[b_index]
2144
# handle common lines
2145
for a_index in range(i, i+n):
2146
yield 'unchanged', self.lines_a[a_index]
2150
def _get_matching_blocks(self, left_revision, right_revision):
2151
"""Return a description of which sections of two revisions match.
2153
See SequenceMatcher.get_matching_blocks
2155
cached = self._cached_matching_blocks.get((left_revision,
2157
if cached is not None:
2159
if self._last_lines_revision_id == left_revision:
2160
left_lines = self._last_lines
2161
right_lines = self.get_lines([right_revision])[right_revision]
2163
lines = self.get_lines([left_revision, right_revision])
2164
left_lines = lines[left_revision]
2165
right_lines = lines[right_revision]
2166
self._last_lines = right_lines
2167
self._last_lines_revision_id = right_revision
2168
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
2170
return matcher.get_matching_blocks()
2172
def _unique_lines(self, matching_blocks):
2173
"""Analyse matching_blocks to determine which lines are unique
2175
:return: a tuple of (unique_left, unique_right), where the values are
2176
sets of line numbers of unique lines.
2182
for i, j, n in matching_blocks:
2183
unique_left.extend(range(last_i, i))
2184
unique_right.extend(range(last_j, j))
2187
return unique_left, unique_right
2190
def _subtract_plans(old_plan, new_plan):
2191
"""Remove changes from new_plan that came from old_plan.
2193
It is assumed that the difference between the old_plan and new_plan
2194
is their choice of 'b' text.
2196
All lines from new_plan that differ from old_plan are emitted
2197
verbatim. All lines from new_plan that match old_plan but are
2198
not about the 'b' revision are emitted verbatim.
2200
Lines that match and are about the 'b' revision are the lines we
2201
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
2202
is skipped entirely.
2204
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
2207
for i, j, n in matcher.get_matching_blocks():
2208
for jj in range(last_j, j):
2210
for jj in range(j, j+n):
2211
plan_line = new_plan[jj]
2212
if plan_line[0] == 'new-b':
2214
elif plan_line[0] == 'killed-b':
2215
yield 'unchanged', plan_line[1]
2221
class _PlanMerge(_PlanMergeBase):
2222
"""Plan an annotate merge using on-the-fly annotation"""
2224
def __init__(self, a_rev, b_rev, vf, key_prefix):
2225
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
2226
self.a_key = self._key_prefix + (self.a_rev,)
2227
self.b_key = self._key_prefix + (self.b_rev,)
2228
self.graph = _mod_graph.Graph(self.vf)
2229
heads = self.graph.heads((self.a_key, self.b_key))
2231
# one side dominates, so we can just return its values, yay for
2233
# Ideally we would know that before we get this far
2234
self._head_key = heads.pop()
2235
if self._head_key == self.a_key:
2239
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2240
self._head_key[-1], other)
2243
self._head_key = None
2246
def _precache_tip_lines(self):
2247
# Turn this into a no-op, because we will do this later
2250
def _find_recursive_lcas(self):
2251
"""Find all the ancestors back to a unique lca"""
2252
cur_ancestors = (self.a_key, self.b_key)
2253
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2254
# rather than a key tuple. We will just map that directly to no common
2258
next_lcas = self.graph.find_lca(*cur_ancestors)
2259
# Map a plain NULL_REVISION to a simple no-ancestors
2260
if next_lcas == set([_mod_revision.NULL_REVISION]):
2262
# Order the lca's based on when they were merged into the tip
2263
# While the actual merge portion of weave merge uses a set() of
2264
# active revisions, the order of insertion *does* effect the
2265
# implicit ordering of the texts.
2266
for rev_key in cur_ancestors:
2267
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2269
parent_map[rev_key] = ordered_parents
2270
if len(next_lcas) == 0:
2272
elif len(next_lcas) == 1:
2273
parent_map[list(next_lcas)[0]] = ()
2275
elif len(next_lcas) > 2:
2276
# More than 2 lca's, fall back to grabbing all nodes between
2277
# this and the unique lca.
2278
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2280
self.a_key, self.b_key, cur_ancestors)
2281
cur_lcas = next_lcas
2282
while len(cur_lcas) > 1:
2283
cur_lcas = self.graph.find_lca(*cur_lcas)
2284
if len(cur_lcas) == 0:
2285
# No common base to find, use the full ancestry
2288
unique_lca = list(cur_lcas)[0]
2289
if unique_lca == _mod_revision.NULL_REVISION:
2290
# find_lca will return a plain 'NULL_REVISION' rather
2291
# than a key tuple when there is no common ancestor, we
2292
# prefer to just use None, because it doesn't confuse
2293
# _get_interesting_texts()
2295
parent_map.update(self._find_unique_parents(next_lcas,
2298
cur_ancestors = next_lcas
2301
def _find_unique_parents(self, tip_keys, base_key):
2302
"""Find ancestors of tip that aren't ancestors of base.
2304
:param tip_keys: Nodes that are interesting
2305
:param base_key: Cull all ancestors of this node
2306
:return: The parent map for all revisions between tip_keys and
2307
base_key. base_key will be included. References to nodes outside of
2308
the ancestor set will also be removed.
2310
# TODO: this would be simpler if find_unique_ancestors took a list
2311
# instead of a single tip, internally it supports it, but it
2312
# isn't a "backwards compatible" api change.
2313
if base_key is None:
2314
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2315
# We remove NULL_REVISION because it isn't a proper tuple key, and
2316
# thus confuses things like _get_interesting_texts, and our logic
2317
# to add the texts into the memory weave.
2318
if _mod_revision.NULL_REVISION in parent_map:
2319
parent_map.pop(_mod_revision.NULL_REVISION)
2322
for tip in tip_keys:
2324
self.graph.find_unique_ancestors(tip, [base_key]))
2325
parent_map = self.graph.get_parent_map(interesting)
2326
parent_map[base_key] = ()
2327
culled_parent_map, child_map, tails = self._remove_external_references(
2329
# Remove all the tails but base_key
2330
if base_key is not None:
2331
tails.remove(base_key)
2332
self._prune_tails(culled_parent_map, child_map, tails)
2333
# Now remove all the uninteresting 'linear' regions
2334
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2338
def _remove_external_references(parent_map):
2339
"""Remove references that go outside of the parent map.
2341
:param parent_map: Something returned from Graph.get_parent_map(keys)
2342
:return: (filtered_parent_map, child_map, tails)
2343
filtered_parent_map is parent_map without external references
2344
child_map is the {parent_key: [child_keys]} mapping
2345
tails is a list of nodes that do not have any parents in the map
2347
# TODO: The basic effect of this function seems more generic than
2348
# _PlanMerge. But the specific details of building a child_map,
2349
# and computing tails seems very specific to _PlanMerge.
2350
# Still, should this be in Graph land?
2351
filtered_parent_map = {}
2354
for key, parent_keys in parent_map.iteritems():
2355
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2356
if not culled_parent_keys:
2358
for parent_key in culled_parent_keys:
2359
child_map.setdefault(parent_key, []).append(key)
2360
# TODO: Do we want to do this, it adds overhead for every node,
2361
# just to say that the node has no children
2362
child_map.setdefault(key, [])
2363
filtered_parent_map[key] = culled_parent_keys
2364
return filtered_parent_map, child_map, tails
2367
def _prune_tails(parent_map, child_map, tails_to_remove):
2368
"""Remove tails from the parent map.
2370
This will remove the supplied revisions until no more children have 0
2373
:param parent_map: A dict of {child: [parents]}, this dictionary will
2374
be modified in place.
2375
:param tails_to_remove: A list of tips that should be removed,
2376
this list will be consumed
2377
:param child_map: The reverse dict of parent_map ({parent: [children]})
2378
this dict will be modified
2379
:return: None, parent_map will be modified in place.
2381
while tails_to_remove:
2382
next = tails_to_remove.pop()
2383
parent_map.pop(next)
2384
children = child_map.pop(next)
2385
for child in children:
2386
child_parents = parent_map[child]
2387
child_parents.remove(next)
2388
if len(child_parents) == 0:
2389
tails_to_remove.append(child)
2391
def _get_interesting_texts(self, parent_map):
2392
"""Return a dict of texts we are interested in.
2394
Note that the input is in key tuples, but the output is in plain
2397
:param parent_map: The output from _find_recursive_lcas
2398
:return: A dict of {'revision_id':lines} as returned by
2399
_PlanMergeBase.get_lines()
2401
all_revision_keys = set(parent_map)
2402
all_revision_keys.add(self.a_key)
2403
all_revision_keys.add(self.b_key)
2405
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2406
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2409
def _build_weave(self):
2410
from bzrlib import weave
2411
self._weave = weave.Weave(weave_name='in_memory_weave',
2412
allow_reserved=True)
2413
parent_map = self._find_recursive_lcas()
2415
all_texts = self._get_interesting_texts(parent_map)
2417
# Note: Unfortunately, the order given by topo_sort will effect the
2418
# ordering resolution in the output. Specifically, if you add A then B,
2419
# then in the output text A lines will show up before B lines. And, of
2420
# course, topo_sort doesn't guarantee any real ordering.
2421
# So we use merge_sort, and add a fake node on the tip.
2422
# This ensures that left-hand parents will always be inserted into the
2423
# weave before right-hand parents.
2424
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2425
parent_map[tip_key] = (self.a_key, self.b_key)
2427
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2431
# for key in tsort.topo_sort(parent_map):
2432
parent_keys = parent_map[key]
2433
revision_id = key[-1]
2434
parent_ids = [k[-1] for k in parent_keys]
2435
self._weave.add_lines(revision_id, parent_ids,
2436
all_texts[revision_id])
2438
def plan_merge(self):
2439
"""Generate a 'plan' for merging the two revisions.
2441
This involves comparing their texts and determining the cause of
2442
differences. If text A has a line and text B does not, then either the
2443
line was added to text A, or it was deleted from B. Once the causes
2444
are combined, they are written out in the format described in
2445
VersionedFile.plan_merge
2447
if self._head_key is not None: # There was a single head
2448
if self._head_key == self.a_key:
2451
if self._head_key != self.b_key:
2452
raise AssertionError('There was an invalid head: %s != %s'
2453
% (self.b_key, self._head_key))
2455
head_rev = self._head_key[-1]
2456
lines = self.get_lines([head_rev])[head_rev]
2457
return ((plan, line) for line in lines)
2458
return self._weave.plan_merge(self.a_rev, self.b_rev)
2461
class _PlanLCAMerge(_PlanMergeBase):
2463
This merge algorithm differs from _PlanMerge in that:
2465
1. comparisons are done against LCAs only
2466
2. cases where a contested line is new versus one LCA but old versus
2467
another are marked as conflicts, by emitting the line as conflicted-a
2470
This is faster, and hopefully produces more useful output.
2473
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2474
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2475
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2478
if lca == _mod_revision.NULL_REVISION:
2481
self.lcas.add(lca[-1])
2482
for lca in self.lcas:
2483
if _mod_revision.is_null(lca):
2486
lca_lines = self.get_lines([lca])[lca]
2487
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2489
blocks = list(matcher.get_matching_blocks())
2490
self._cached_matching_blocks[(a_rev, lca)] = blocks
2491
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2493
blocks = list(matcher.get_matching_blocks())
2494
self._cached_matching_blocks[(b_rev, lca)] = blocks
2496
def _determine_status(self, revision_id, unique_line_numbers):
2497
"""Determines the status unique lines versus all lcas.
2499
Basically, determines why the line is unique to this revision.
2501
A line may be determined new, killed, or both.
2503
If a line is determined new, that means it was not present in at least
2504
one LCA, and is not present in the other merge revision.
2506
If a line is determined killed, that means the line was present in
2509
If a line is killed and new, this indicates that the two merge
2510
revisions contain differing conflict resolutions.
2512
:param revision_id: The id of the revision in which the lines are
2514
:param unique_line_numbers: The line numbers of unique lines.
2515
:return: a tuple of (new_this, killed_other)
2519
unique_line_numbers = set(unique_line_numbers)
2520
for lca in self.lcas:
2521
blocks = self._get_matching_blocks(revision_id, lca)
2522
unique_vs_lca, _ignored = self._unique_lines(blocks)
2523
new.update(unique_line_numbers.intersection(unique_vs_lca))
2524
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
