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,
620
'other_branch': self.other_branch,
622
if self.merge_type.requires_base:
623
kwargs['base_tree'] = self.base_tree
624
if self.merge_type.supports_reprocess:
625
kwargs['reprocess'] = self.reprocess
627
raise errors.BzrError(
628
"Conflict reduction is not supported for merge"
629
" type %s." % self.merge_type)
630
if self.merge_type.supports_show_base:
631
kwargs['show_base'] = self.show_base
633
raise errors.BzrError("Showing base is not supported for this"
634
" merge type. %s" % self.merge_type)
635
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
636
and not self.base_is_other_ancestor):
637
raise errors.CannotReverseCherrypick()
638
if self.merge_type.supports_cherrypick:
639
kwargs['cherrypick'] = (not self.base_is_ancestor or
640
not self.base_is_other_ancestor)
641
if self._is_criss_cross and getattr(self.merge_type,
642
'supports_lca_trees', False):
643
kwargs['lca_trees'] = self._lca_trees
644
return self.merge_type(pb=None,
645
change_reporter=self.change_reporter,
648
def _do_merge_to(self):
649
merge = self.make_merger()
650
if self.other_branch is not None:
651
self.other_branch.update_references(self.this_branch)
652
for hook in Merger.hooks['pre_merge']:
655
for hook in Merger.hooks['post_merge']:
657
if self.recurse == 'down':
658
for relpath, file_id in self.this_tree.iter_references():
659
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
660
other_revision = self.other_tree.get_reference_revision(
662
if other_revision == sub_tree.last_revision():
664
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
665
sub_merge.merge_type = self.merge_type
666
other_branch = self.other_branch.reference_parent(file_id,
668
sub_merge.set_other_revision(other_revision, other_branch)
669
base_revision = self.base_tree.get_reference_revision(file_id)
670
sub_merge.base_tree = \
671
sub_tree.branch.repository.revision_tree(base_revision)
672
sub_merge.base_rev_id = base_revision
677
operation = cleanup.OperationWithCleanups(self._do_merge_to)
678
self.this_tree.lock_tree_write()
679
operation.add_cleanup(self.this_tree.unlock)
680
if self.base_tree is not None:
681
self.base_tree.lock_read()
682
operation.add_cleanup(self.base_tree.unlock)
683
if self.other_tree is not None:
684
self.other_tree.lock_read()
685
operation.add_cleanup(self.other_tree.unlock)
686
merge = operation.run_simple()
687
if len(merge.cooked_conflicts) == 0:
688
if not self.ignore_zero and not trace.is_quiet():
689
trace.note(gettext("All changes applied successfully."))
691
trace.note(gettext("%d conflicts encountered.")
692
% len(merge.cooked_conflicts))
694
return len(merge.cooked_conflicts)
697
class _InventoryNoneEntry(object):
698
"""This represents an inventory entry which *isn't there*.
700
It simplifies the merging logic if we always have an InventoryEntry, even
701
if it isn't actually present
708
symlink_target = None
711
_none_entry = _InventoryNoneEntry()
714
class Merge3Merger(object):
715
"""Three-way merger that uses the merge3 text merger"""
717
supports_reprocess = True
718
supports_show_base = True
719
history_based = False
720
supports_cherrypick = True
721
supports_reverse_cherrypick = True
722
winner_idx = {"this": 2, "other": 1, "conflict": 1}
723
supports_lca_trees = True
725
def __init__(self, working_tree, this_tree, base_tree, other_tree,
726
interesting_ids=None, reprocess=False, show_base=False,
727
pb=None, pp=None, change_reporter=None,
728
interesting_files=None, do_merge=True,
729
cherrypick=False, lca_trees=None, this_branch=None,
731
"""Initialize the merger object and perform the merge.
733
:param working_tree: The working tree to apply the merge to
734
:param this_tree: The local tree in the merge operation
735
:param base_tree: The common tree in the merge operation
736
:param other_tree: The other tree to merge changes from
737
:param this_branch: The branch associated with this_tree. Defaults to
738
this_tree.branch if not supplied.
739
:param other_branch: The branch associated with other_tree, if any.
740
:param interesting_ids: The file_ids of files that should be
741
participate in the merge. May not be combined with
743
:param: reprocess If True, perform conflict-reduction processing.
744
:param show_base: If True, show the base revision in text conflicts.
745
(incompatible with reprocess)
747
:param pp: A ProgressPhase object
748
:param change_reporter: An object that should report changes made
749
:param interesting_files: The tree-relative paths of files that should
750
participate in the merge. If these paths refer to directories,
751
the contents of those directories will also be included. May not
752
be combined with interesting_ids. If neither interesting_files nor
753
interesting_ids is specified, all files may participate in the
755
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
756
if the ancestry was found to include a criss-cross merge.
757
Otherwise should be None.
759
object.__init__(self)
760
if interesting_files is not None and interesting_ids is not None:
762
'specify either interesting_ids or interesting_files')
763
if this_branch is None:
764
this_branch = this_tree.branch
765
self.interesting_ids = interesting_ids
766
self.interesting_files = interesting_files
767
self.working_tree = working_tree
768
self.this_tree = this_tree
769
self.base_tree = base_tree
770
self.other_tree = other_tree
771
self.this_branch = this_branch
772
self.other_branch = other_branch
773
self._raw_conflicts = []
774
self.cooked_conflicts = []
775
self.reprocess = reprocess
776
self.show_base = show_base
777
self._lca_trees = lca_trees
778
# Uncommenting this will change the default algorithm to always use
779
# _entries_lca. This can be useful for running the test suite and
780
# making sure we haven't missed any corner cases.
781
# if lca_trees is None:
782
# self._lca_trees = [self.base_tree]
783
self.change_reporter = change_reporter
784
self.cherrypick = cherrypick
788
warnings.warn("pp argument to Merge3Merger is deprecated")
790
warnings.warn("pb argument to Merge3Merger is deprecated")
793
operation = cleanup.OperationWithCleanups(self._do_merge)
794
self.working_tree.lock_tree_write()
795
operation.add_cleanup(self.working_tree.unlock)
796
self.this_tree.lock_read()
797
operation.add_cleanup(self.this_tree.unlock)
798
self.base_tree.lock_read()
799
operation.add_cleanup(self.base_tree.unlock)
800
self.other_tree.lock_read()
801
operation.add_cleanup(self.other_tree.unlock)
804
def _do_merge(self, operation):
805
self.tt = transform.TreeTransform(self.working_tree, None)
806
operation.add_cleanup(self.tt.finalize)
807
self._compute_transform()
808
results = self.tt.apply(no_conflicts=True)
809
self.write_modified(results)
811
self.working_tree.add_conflicts(self.cooked_conflicts)
812
except errors.UnsupportedOperation:
815
def make_preview_transform(self):
816
operation = cleanup.OperationWithCleanups(self._make_preview_transform)
817
self.base_tree.lock_read()
818
operation.add_cleanup(self.base_tree.unlock)
819
self.other_tree.lock_read()
820
operation.add_cleanup(self.other_tree.unlock)
821
return operation.run_simple()
823
def _make_preview_transform(self):
824
self.tt = transform.TransformPreview(self.working_tree)
825
self._compute_transform()
828
def _compute_transform(self):
829
if self._lca_trees is None:
830
entries = self._entries3()
831
resolver = self._three_way
833
entries = self._entries_lca()
834
resolver = self._lca_multi_way
835
# Prepare merge hooks
836
factories = Merger.hooks['merge_file_content']
837
# One hook for each registered one plus our default merger
838
hooks = [factory(self) for factory in factories] + [self]
839
self.active_hooks = [hook for hook in hooks if hook is not None]
840
child_pb = ui.ui_factory.nested_progress_bar()
842
for num, (file_id, changed, parents3, names3,
843
executable3) in enumerate(entries):
844
# Try merging each entry
845
child_pb.update(gettext('Preparing file merge'),
847
self._merge_names(file_id, parents3, names3, resolver=resolver)
849
file_status = self._do_merge_contents(file_id)
851
file_status = 'unmodified'
852
self._merge_executable(file_id,
853
executable3, file_status, resolver=resolver)
856
self.tt.fixup_new_roots()
857
self._finish_computing_transform()
859
def _finish_computing_transform(self):
860
"""Finalize the transform and report the changes.
862
This is the second half of _compute_transform.
864
child_pb = ui.ui_factory.nested_progress_bar()
866
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
867
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
870
if self.change_reporter is not None:
871
from bzrlib import delta
872
delta.report_changes(
873
self.tt.iter_changes(), self.change_reporter)
874
self.cook_conflicts(fs_conflicts)
875
for conflict in self.cooked_conflicts:
876
trace.warning(unicode(conflict))
879
"""Gather data about files modified between three trees.
881
Return a list of tuples of file_id, changed, parents3, names3,
882
executable3. changed is a boolean indicating whether the file contents
883
or kind were changed. parents3 is a tuple of parent ids for base,
884
other and this. names3 is a tuple of names for base, other and this.
885
executable3 is a tuple of execute-bit values for base, other and this.
888
iterator = self.other_tree.iter_changes(self.base_tree,
889
specific_files=self.interesting_files,
890
extra_trees=[self.this_tree])
891
this_entries = dict((e.file_id, e) for p, e in
892
self.this_tree.iter_entries_by_dir(
893
self.interesting_ids))
894
for (file_id, paths, changed, versioned, parents, names, kind,
895
executable) in iterator:
896
if (self.interesting_ids is not None and
897
file_id not in self.interesting_ids):
899
entry = this_entries.get(file_id)
900
if entry is not None:
901
this_name = entry.name
902
this_parent = entry.parent_id
903
this_executable = entry.executable
907
this_executable = None
908
parents3 = parents + (this_parent,)
909
names3 = names + (this_name,)
910
executable3 = executable + (this_executable,)
911
result.append((file_id, changed, parents3, names3, executable3))
914
def _entries_lca(self):
915
"""Gather data about files modified between multiple trees.
917
This compares OTHER versus all LCA trees, and for interesting entries,
918
it then compares with THIS and BASE.
920
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
922
:return: [(file_id, changed, parents, names, executable)], where:
924
* file_id: Simple file_id of the entry
925
* changed: Boolean, True if the kind or contents changed else False
926
* parents: ((base, [parent_id, in, lcas]), parent_id_other,
928
* names: ((base, [name, in, lcas]), name_in_other, name_in_this)
929
* executable: ((base, [exec, in, lcas]), exec_in_other,
932
if self.interesting_files is not None:
933
lookup_trees = [self.this_tree, self.base_tree]
934
lookup_trees.extend(self._lca_trees)
935
# I think we should include the lca trees as well
936
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
939
interesting_ids = self.interesting_ids
941
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
943
base_inventory = self.base_tree.root_inventory
944
this_inventory = self.this_tree.root_inventory
945
for path, file_id, other_ie, lca_values in walker.iter_all():
946
# Is this modified at all from any of the other trees?
948
other_ie = _none_entry
949
if interesting_ids is not None and file_id not in interesting_ids:
952
# If other_revision is found in any of the lcas, that means this
953
# node is uninteresting. This is because when merging, if there are
954
# multiple heads(), we have to create a new node. So if we didn't,
955
# we know that the ancestry is linear, and that OTHER did not
957
# See doc/developers/lca_merge_resolution.txt for details
958
other_revision = other_ie.revision
959
if other_revision is not None:
960
# We can't use this shortcut when other_revision is None,
961
# because it may be None because things are WorkingTrees, and
962
# not because it is *actually* None.
963
is_unmodified = False
964
for lca_path, ie in lca_values:
965
if ie is not None and ie.revision == other_revision:
972
for lca_path, lca_ie in lca_values:
974
lca_entries.append(_none_entry)
976
lca_entries.append(lca_ie)
978
if base_inventory.has_id(file_id):
979
base_ie = base_inventory[file_id]
981
base_ie = _none_entry
983
if this_inventory.has_id(file_id):
984
this_ie = this_inventory[file_id]
986
this_ie = _none_entry
992
for lca_ie in lca_entries:
993
lca_kinds.append(lca_ie.kind)
994
lca_parent_ids.append(lca_ie.parent_id)
995
lca_names.append(lca_ie.name)
996
lca_executable.append(lca_ie.executable)
998
kind_winner = self._lca_multi_way(
999
(base_ie.kind, lca_kinds),
1000
other_ie.kind, this_ie.kind)
1001
parent_id_winner = self._lca_multi_way(
1002
(base_ie.parent_id, lca_parent_ids),
1003
other_ie.parent_id, this_ie.parent_id)
1004
name_winner = self._lca_multi_way(
1005
(base_ie.name, lca_names),
1006
other_ie.name, this_ie.name)
1008
content_changed = True
1009
if kind_winner == 'this':
1010
# No kind change in OTHER, see if there are *any* changes
1011
if other_ie.kind == 'directory':
1012
if parent_id_winner == 'this' and name_winner == 'this':
1013
# No change for this directory in OTHER, skip
1015
content_changed = False
1016
elif other_ie.kind is None or other_ie.kind == 'file':
1017
def get_sha1(ie, tree):
1018
if ie.kind != 'file':
1020
return tree.get_file_sha1(file_id)
1021
base_sha1 = get_sha1(base_ie, self.base_tree)
1022
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1023
in zip(lca_entries, self._lca_trees)]
1024
this_sha1 = get_sha1(this_ie, self.this_tree)
1025
other_sha1 = get_sha1(other_ie, self.other_tree)
1026
sha1_winner = self._lca_multi_way(
1027
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1028
allow_overriding_lca=False)
1029
exec_winner = self._lca_multi_way(
1030
(base_ie.executable, lca_executable),
1031
other_ie.executable, this_ie.executable)
1032
if (parent_id_winner == 'this' and name_winner == 'this'
1033
and sha1_winner == 'this' and exec_winner == 'this'):
1034
# No kind, parent, name, exec, or content change for
1035
# OTHER, so this node is not considered interesting
1037
if sha1_winner == 'this':
1038
content_changed = False
1039
elif other_ie.kind == 'symlink':
1040
def get_target(ie, tree):
1041
if ie.kind != 'symlink':
1043
return tree.get_symlink_target(file_id)
1044
base_target = get_target(base_ie, self.base_tree)
1045
lca_targets = [get_target(ie, tree) for ie, tree
1046
in zip(lca_entries, self._lca_trees)]
1047
this_target = get_target(this_ie, self.this_tree)
1048
other_target = get_target(other_ie, self.other_tree)
1049
target_winner = self._lca_multi_way(
1050
(base_target, lca_targets),
1051
other_target, this_target)
1052
if (parent_id_winner == 'this' and name_winner == 'this'
1053
and target_winner == 'this'):
1054
# No kind, parent, name, or symlink target change
1057
if target_winner == 'this':
1058
content_changed = False
1059
elif other_ie.kind == 'tree-reference':
1060
# The 'changed' information seems to be handled at a higher
1061
# level. At least, _entries3 returns False for content
1062
# changed, even when at a new revision_id.
1063
content_changed = False
1064
if (parent_id_winner == 'this' and name_winner == 'this'):
1065
# Nothing interesting
1068
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1070
# If we have gotten this far, that means something has changed
1071
result.append((file_id, content_changed,
1072
((base_ie.parent_id, lca_parent_ids),
1073
other_ie.parent_id, this_ie.parent_id),
1074
((base_ie.name, lca_names),
1075
other_ie.name, this_ie.name),
1076
((base_ie.executable, lca_executable),
1077
other_ie.executable, this_ie.executable)
1081
@deprecated_method(deprecated_in((2, 4, 0)))
1083
if self.tt.final_kind(self.tt.root) is None:
1084
self.tt.cancel_deletion(self.tt.root)
1085
if self.tt.final_file_id(self.tt.root) is None:
1086
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
1088
other_root_file_id = self.other_tree.get_root_id()
1089
if other_root_file_id is None:
1091
other_root = self.tt.trans_id_file_id(other_root_file_id)
1092
if other_root == self.tt.root:
1094
if self.this_tree.has_id(
1095
self.other_tree.get_root_id()):
1096
# the other tree's root is a non-root in the current tree (as
1097
# when a previously unrelated branch is merged into another)
1099
if self.tt.final_kind(other_root) is not None:
1100
other_root_is_present = True
1102
# other_root doesn't have a physical representation. We still need
1103
# to move any references to the actual root of the tree.
1104
other_root_is_present = False
1105
# the other tree root is not present in this tree. We are
1106
# calling adjust_path for children which *want* to be present with a
1107
# correct place to go.
1108
for child_id in self.other_tree.iter_children(
1109
self.other_tree.get_root_id()):
1110
trans_id = self.tt.trans_id_file_id(child_id)
1111
if not other_root_is_present:
1112
if self.tt.final_kind(trans_id) is not None:
1113
# The item exist in the final tree and has a defined place
1116
# Move the item into the root
1118
final_name = self.tt.final_name(trans_id)
1119
except errors.NoFinalPath:
1120
# This file is not present anymore, ignore it.
1122
self.tt.adjust_path(final_name, self.tt.root, trans_id)
1123
if other_root_is_present:
1124
self.tt.cancel_creation(other_root)
1125
self.tt.cancel_versioning(other_root)
1127
def write_modified(self, results):
1128
modified_hashes = {}
1129
for path in results.modified_paths:
1130
file_id = self.working_tree.path2id(self.working_tree.relpath(path))
1133
hash = self.working_tree.get_file_sha1(file_id)
1136
modified_hashes[file_id] = hash
1137
self.working_tree.set_merge_modified(modified_hashes)
1140
def parent(entry, file_id):
1141
"""Determine the parent for a file_id (used as a key method)"""
1144
return entry.parent_id
1147
def name(entry, file_id):
1148
"""Determine the name for a file_id (used as a key method)"""
1154
def contents_sha1(tree, file_id):
1155
"""Determine the sha1 of the file contents (used as a key method)."""
1156
if not tree.has_id(file_id):
1158
return tree.get_file_sha1(file_id)
1161
def executable(tree, file_id):
1162
"""Determine the executability of a file-id (used as a key method)."""
1163
if not tree.has_id(file_id):
1165
if tree.kind(file_id) != "file":
1167
return tree.is_executable(file_id)
1170
def kind(tree, file_id):
1171
"""Determine the kind of a file-id (used as a key method)."""
1172
if not tree.has_id(file_id):
1174
return tree.kind(file_id)
1177
def _three_way(base, other, this):
1179
# if 'base == other', either they all agree, or only 'this' has
1182
elif this not in (base, other):
1183
# 'this' is neither 'base' nor 'other', so both sides changed
1186
# "Ambiguous clean merge" -- both sides have made the same change.
1189
# this == base: only other has changed.
1193
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
1194
"""Consider LCAs when determining whether a change has occurred.
1196
If LCAS are all identical, this is the same as a _three_way comparison.
1198
:param bases: value in (BASE, [LCAS])
1199
:param other: value in OTHER
1200
:param this: value in THIS
1201
:param allow_overriding_lca: If there is more than one unique lca
1202
value, allow OTHER to override THIS if it has a new value, and
1203
THIS only has an lca value, or vice versa. This is appropriate for
1204
truly scalar values, not as much for non-scalars.
1205
:return: 'this', 'other', or 'conflict' depending on whether an entry
1208
# See doc/developers/lca_tree_merging.txt for details about this
1211
# Either Ambiguously clean, or nothing was actually changed. We
1214
base_val, lca_vals = bases
1215
# Remove 'base_val' from the lca_vals, because it is not interesting
1216
filtered_lca_vals = [lca_val for lca_val in lca_vals
1217
if lca_val != base_val]
1218
if len(filtered_lca_vals) == 0:
1219
return Merge3Merger._three_way(base_val, other, this)
1221
unique_lca_vals = set(filtered_lca_vals)
1222
if len(unique_lca_vals) == 1:
1223
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1225
if allow_overriding_lca:
1226
if other in unique_lca_vals:
1227
if this in unique_lca_vals:
1228
# Each side picked a different lca, conflict
1231
# This has a value which supersedes both lca values, and
1232
# other only has an lca value
1234
elif this in unique_lca_vals:
1235
# OTHER has a value which supersedes both lca values, and this
1236
# only has an lca value
1239
# At this point, the lcas disagree, and the tip disagree
1242
def merge_names(self, file_id):
1243
def get_entry(tree):
1245
return tree.root_inventory[file_id]
1246
except errors.NoSuchId:
1248
this_entry = get_entry(self.this_tree)
1249
other_entry = get_entry(self.other_tree)
1250
base_entry = get_entry(self.base_tree)
1251
entries = (base_entry, other_entry, this_entry)
1254
for entry in entries:
1257
parents.append(None)
1259
names.append(entry.name)
1260
parents.append(entry.parent_id)
1261
return self._merge_names(file_id, parents, names,
1262
resolver=self._three_way)
1264
def _merge_names(self, file_id, parents, names, resolver):
1265
"""Perform a merge on file_id names and parents"""
1266
base_name, other_name, this_name = names
1267
base_parent, other_parent, this_parent = parents
1269
name_winner = resolver(*names)
1271
parent_id_winner = resolver(*parents)
1272
if this_name is None:
1273
if name_winner == "this":
1274
name_winner = "other"
1275
if parent_id_winner == "this":
1276
parent_id_winner = "other"
1277
if name_winner == "this" and parent_id_winner == "this":
1279
if name_winner == 'conflict' or parent_id_winner == 'conflict':
1280
# Creating helpers (.OTHER or .THIS) here cause problems down the
1281
# road if a ContentConflict needs to be created so we should not do
1283
trans_id = self.tt.trans_id_file_id(file_id)
1284
self._raw_conflicts.append(('path conflict', trans_id, file_id,
1285
this_parent, this_name,
1286
other_parent, other_name))
1287
if not self.other_tree.has_id(file_id):
1288
# it doesn't matter whether the result was 'other' or
1289
# 'conflict'-- if it has no file id, we leave it alone.
1291
parent_id = parents[self.winner_idx[parent_id_winner]]
1292
name = names[self.winner_idx[name_winner]]
1293
if parent_id is not None or name is not None:
1294
# if we get here, name_winner and parent_winner are set to safe
1296
if parent_id is None and name is not None:
1297
# if parent_id is None and name is non-None, current file is
1299
if names[self.winner_idx[parent_id_winner]] != '':
1300
raise AssertionError(
1301
'File looks like a root, but named %s' %
1302
names[self.winner_idx[parent_id_winner]])
1303
parent_trans_id = transform.ROOT_PARENT
1305
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1306
self.tt.adjust_path(name, parent_trans_id,
1307
self.tt.trans_id_file_id(file_id))
1309
def _do_merge_contents(self, file_id):
1310
"""Performs a merge on file_id contents."""
1311
def contents_pair(tree):
1312
if not tree.has_id(file_id):
1314
kind = tree.kind(file_id)
1316
contents = tree.get_file_sha1(file_id)
1317
elif kind == "symlink":
1318
contents = tree.get_symlink_target(file_id)
1321
return kind, contents
1323
# See SPOT run. run, SPOT, run.
1324
# So we're not QUITE repeating ourselves; we do tricky things with
1326
base_pair = contents_pair(self.base_tree)
1327
other_pair = contents_pair(self.other_tree)
1329
this_pair = contents_pair(self.this_tree)
1330
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1331
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1332
this_pair, allow_overriding_lca=False)
1334
if base_pair == other_pair:
1337
# We delayed evaluating this_pair as long as we can to avoid
1338
# unnecessary sha1 calculation
1339
this_pair = contents_pair(self.this_tree)
1340
winner = self._three_way(base_pair, other_pair, this_pair)
1341
if winner == 'this':
1342
# No interesting changes introduced by OTHER
1344
# We have a hypothetical conflict, but if we have files, then we
1345
# can try to merge the content
1346
trans_id = self.tt.trans_id_file_id(file_id)
1347
params = MergeFileHookParams(self, file_id, trans_id, this_pair[0],
1348
other_pair[0], winner)
1349
hooks = self.active_hooks
1350
hook_status = 'not_applicable'
1352
hook_status, lines = hook.merge_contents(params)
1353
if hook_status != 'not_applicable':
1354
# Don't try any more hooks, this one applies.
1356
# If the merge ends up replacing the content of the file, we get rid of
1357
# it at the end of this method (this variable is used to track the
1358
# exceptions to this rule).
1361
if hook_status == 'not_applicable':
1362
# No merge hook was able to resolve the situation. Two cases exist:
1363
# a content conflict or a duplicate one.
1365
name = self.tt.final_name(trans_id)
1366
parent_id = self.tt.final_parent(trans_id)
1368
inhibit_content_conflict = False
1369
if params.this_kind is None: # file_id is not in THIS
1370
# Is the name used for a different file_id ?
1371
dupe_path = self.other_tree.id2path(file_id)
1372
this_id = self.this_tree.path2id(dupe_path)
1373
if this_id is not None:
1374
# Two entries for the same path
1376
# versioning the merged file will trigger a duplicate
1378
self.tt.version_file(file_id, trans_id)
1379
transform.create_from_tree(
1380
self.tt, trans_id, self.other_tree, file_id,
1381
filter_tree_path=self._get_filter_tree_path(file_id))
1382
inhibit_content_conflict = True
1383
elif params.other_kind is None: # file_id is not in OTHER
1384
# Is the name used for a different file_id ?
1385
dupe_path = self.this_tree.id2path(file_id)
1386
other_id = self.other_tree.path2id(dupe_path)
1387
if other_id is not None:
1388
# Two entries for the same path again, but here, the other
1389
# entry will also be merged. We simply inhibit the
1390
# 'content' conflict creation because we know OTHER will
1391
# create (or has already created depending on ordering) an
1392
# entry at the same path. This will trigger a 'duplicate'
1395
inhibit_content_conflict = True
1396
if not inhibit_content_conflict:
1397
if params.this_kind is not None:
1398
self.tt.unversion_file(trans_id)
1399
# This is a contents conflict, because none of the available
1400
# functions could merge it.
1401
file_group = self._dump_conflicts(name, parent_id, file_id,
1403
self._raw_conflicts.append(('contents conflict', file_group))
1404
elif hook_status == 'success':
1405
self.tt.create_file(lines, trans_id)
1406
elif hook_status == 'conflicted':
1407
# XXX: perhaps the hook should be able to provide
1408
# the BASE/THIS/OTHER files?
1409
self.tt.create_file(lines, trans_id)
1410
self._raw_conflicts.append(('text conflict', trans_id))
1411
name = self.tt.final_name(trans_id)
1412
parent_id = self.tt.final_parent(trans_id)
1413
self._dump_conflicts(name, parent_id, file_id)
1414
elif hook_status == 'delete':
1415
self.tt.unversion_file(trans_id)
1417
elif hook_status == 'done':
1418
# The hook function did whatever it needs to do directly, no
1419
# further action needed here.
1422
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1423
if not self.this_tree.has_id(file_id) and result == "modified":
1424
self.tt.version_file(file_id, trans_id)
1426
# The merge has been performed and produced a new content, so the
1427
# old contents should not be retained.
1428
self.tt.delete_contents(trans_id)
1431
def _default_other_winner_merge(self, merge_hook_params):
1432
"""Replace this contents with other."""
1433
file_id = merge_hook_params.file_id
1434
trans_id = merge_hook_params.trans_id
1435
if self.other_tree.has_id(file_id):
1436
# OTHER changed the file
1437
transform.create_from_tree(
1438
self.tt, trans_id, self.other_tree, file_id,
1439
filter_tree_path=self._get_filter_tree_path(file_id))
1441
elif self.this_tree.has_id(file_id):
1442
# OTHER deleted the file
1443
return 'delete', None
1445
raise AssertionError(
1446
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1449
def merge_contents(self, merge_hook_params):
1450
"""Fallback merge logic after user installed hooks."""
1451
# This function is used in merge hooks as the fallback instance.
1452
# Perhaps making this function and the functions it calls be a
1453
# a separate class would be better.
1454
if merge_hook_params.winner == 'other':
1455
# OTHER is a straight winner, so replace this contents with other
1456
return self._default_other_winner_merge(merge_hook_params)
1457
elif merge_hook_params.is_file_merge():
1458
# THIS and OTHER are both files, so text merge. Either
1459
# BASE is a file, or both converted to files, so at least we
1460
# have agreement that output should be a file.
1462
self.text_merge(merge_hook_params.file_id,
1463
merge_hook_params.trans_id)
1464
except errors.BinaryFile:
1465
return 'not_applicable', None
1468
return 'not_applicable', None
1470
def get_lines(self, tree, file_id):
1471
"""Return the lines in a file, or an empty list."""
1472
if tree.has_id(file_id):
1473
return tree.get_file_lines(file_id)
1477
def text_merge(self, file_id, trans_id):
1478
"""Perform a three-way text merge on a file_id"""
1479
# it's possible that we got here with base as a different type.
1480
# if so, we just want two-way text conflicts.
1481
if self.base_tree.has_id(file_id) and \
1482
self.base_tree.kind(file_id) == "file":
1483
base_lines = self.get_lines(self.base_tree, file_id)
1486
other_lines = self.get_lines(self.other_tree, file_id)
1487
this_lines = self.get_lines(self.this_tree, file_id)
1488
m3 = merge3.Merge3(base_lines, this_lines, other_lines,
1489
is_cherrypick=self.cherrypick)
1490
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1491
if self.show_base is True:
1492
base_marker = '|' * 7
1496
def iter_merge3(retval):
1497
retval["text_conflicts"] = False
1498
for line in m3.merge_lines(name_a = "TREE",
1499
name_b = "MERGE-SOURCE",
1500
name_base = "BASE-REVISION",
1501
start_marker=start_marker,
1502
base_marker=base_marker,
1503
reprocess=self.reprocess):
1504
if line.startswith(start_marker):
1505
retval["text_conflicts"] = True
1506
yield line.replace(start_marker, '<' * 7)
1510
merge3_iterator = iter_merge3(retval)
1511
self.tt.create_file(merge3_iterator, trans_id)
1512
if retval["text_conflicts"] is True:
1513
self._raw_conflicts.append(('text conflict', trans_id))
1514
name = self.tt.final_name(trans_id)
1515
parent_id = self.tt.final_parent(trans_id)
1516
file_group = self._dump_conflicts(name, parent_id, file_id,
1517
this_lines, base_lines,
1519
file_group.append(trans_id)
1522
def _get_filter_tree_path(self, file_id):
1523
if self.this_tree.supports_content_filtering():
1524
# We get the path from the working tree if it exists.
1525
# That fails though when OTHER is adding a file, so
1526
# we fall back to the other tree to find the path if
1527
# it doesn't exist locally.
1529
return self.this_tree.id2path(file_id)
1530
except errors.NoSuchId:
1531
return self.other_tree.id2path(file_id)
1532
# Skip the id2path lookup for older formats
1535
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1536
base_lines=None, other_lines=None, set_version=False,
1538
"""Emit conflict files.
1539
If this_lines, base_lines, or other_lines are omitted, they will be
1540
determined automatically. If set_version is true, the .OTHER, .THIS
1541
or .BASE (in that order) will be created as versioned files.
1543
data = [('OTHER', self.other_tree, other_lines),
1544
('THIS', self.this_tree, this_lines)]
1546
data.append(('BASE', self.base_tree, base_lines))
1548
# We need to use the actual path in the working tree of the file here,
1549
# ignoring the conflict suffixes
1551
if wt.supports_content_filtering():
1553
filter_tree_path = wt.id2path(file_id)
1554
except errors.NoSuchId:
1555
# file has been deleted
1556
filter_tree_path = None
1558
# Skip the id2path lookup for older formats
1559
filter_tree_path = None
1563
for suffix, tree, lines in data:
1564
if tree.has_id(file_id):
1565
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1566
suffix, lines, filter_tree_path)
1567
file_group.append(trans_id)
1568
if set_version and not versioned:
1569
self.tt.version_file(file_id, trans_id)
1573
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1574
lines=None, filter_tree_path=None):
1575
"""Emit a single conflict file."""
1576
name = name + '.' + suffix
1577
trans_id = self.tt.create_path(name, parent_id)
1578
transform.create_from_tree(self.tt, trans_id, tree, file_id, lines,
1582
def merge_executable(self, file_id, file_status):
1583
"""Perform a merge on the execute bit."""
1584
executable = [self.executable(t, file_id) for t in (self.base_tree,
1585
self.other_tree, self.this_tree)]
1586
self._merge_executable(file_id, executable, file_status,
1587
resolver=self._three_way)
1589
def _merge_executable(self, file_id, executable, file_status,
1591
"""Perform a merge on the execute bit."""
1592
base_executable, other_executable, this_executable = executable
1593
if file_status == "deleted":
1595
winner = resolver(*executable)
1596
if winner == "conflict":
1597
# There must be a None in here, if we have a conflict, but we
1598
# need executability since file status was not deleted.
1599
if self.executable(self.other_tree, file_id) is None:
1603
if winner == 'this' and file_status != "modified":
1605
trans_id = self.tt.trans_id_file_id(file_id)
1606
if self.tt.final_kind(trans_id) != "file":
1608
if winner == "this":
1609
executability = this_executable
1611
if self.other_tree.has_id(file_id):
1612
executability = other_executable
1613
elif self.this_tree.has_id(file_id):
1614
executability = this_executable
1615
elif self.base_tree_has_id(file_id):
1616
executability = base_executable
1617
if executability is not None:
1618
trans_id = self.tt.trans_id_file_id(file_id)
1619
self.tt.set_executability(executability, trans_id)
1621
def cook_conflicts(self, fs_conflicts):
1622
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1623
content_conflict_file_ids = set()
1624
cooked_conflicts = transform.cook_conflicts(fs_conflicts, self.tt)
1625
fp = transform.FinalPaths(self.tt)
1626
for conflict in self._raw_conflicts:
1627
conflict_type = conflict[0]
1628
if conflict_type == 'path conflict':
1630
this_parent, this_name,
1631
other_parent, other_name) = conflict[1:]
1632
if this_parent is None or this_name is None:
1633
this_path = '<deleted>'
1635
parent_path = fp.get_path(
1636
self.tt.trans_id_file_id(this_parent))
1637
this_path = osutils.pathjoin(parent_path, this_name)
1638
if other_parent is None or other_name is None:
1639
other_path = '<deleted>'
1641
if other_parent == self.other_tree.get_root_id():
1642
# The tree transform doesn't know about the other root,
1643
# so we special case here to avoid a NoFinalPath
1647
parent_path = fp.get_path(
1648
self.tt.trans_id_file_id(other_parent))
1649
other_path = osutils.pathjoin(parent_path, other_name)
1650
c = _mod_conflicts.Conflict.factory(
1651
'path conflict', path=this_path,
1652
conflict_path=other_path,
1654
elif conflict_type == 'contents conflict':
1655
for trans_id in conflict[1]:
1656
file_id = self.tt.final_file_id(trans_id)
1657
if file_id is not None:
1658
# Ok we found the relevant file-id
1660
path = fp.get_path(trans_id)
1661
for suffix in ('.BASE', '.THIS', '.OTHER'):
1662
if path.endswith(suffix):
1663
# Here is the raw path
1664
path = path[:-len(suffix)]
1666
c = _mod_conflicts.Conflict.factory(conflict_type,
1667
path=path, file_id=file_id)
1668
content_conflict_file_ids.add(file_id)
1669
elif conflict_type == 'text conflict':
1670
trans_id = conflict[1]
1671
path = fp.get_path(trans_id)
1672
file_id = self.tt.final_file_id(trans_id)
1673
c = _mod_conflicts.Conflict.factory(conflict_type,
1674
path=path, file_id=file_id)
1676
raise AssertionError('bad conflict type: %r' % (conflict,))
1677
cooked_conflicts.append(c)
1679
self.cooked_conflicts = []
1680
# We want to get rid of path conflicts when a corresponding contents
1681
# conflict exists. This can occur when one branch deletes a file while
1682
# the other renames *and* modifies it. In this case, the content
1683
# conflict is enough.
1684
for c in cooked_conflicts:
1685
if (c.typestring == 'path conflict'
1686
and c.file_id in content_conflict_file_ids):
1688
self.cooked_conflicts.append(c)
1689
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
1692
class WeaveMerger(Merge3Merger):
1693
"""Three-way tree merger, text weave merger."""
1694
supports_reprocess = True
1695
supports_show_base = False
1696
supports_reverse_cherrypick = False
1697
history_based = True
1699
def _generate_merge_plan(self, file_id, base):
1700
return self.this_tree.plan_file_merge(file_id, self.other_tree,
1703
def _merged_lines(self, file_id):
1704
"""Generate the merged lines.
1705
There is no distinction between lines that are meant to contain <<<<<<<
1709
base = self.base_tree
1712
plan = self._generate_merge_plan(file_id, base)
1713
if 'merge' in debug.debug_flags:
1715
trans_id = self.tt.trans_id_file_id(file_id)
1716
name = self.tt.final_name(trans_id) + '.plan'
1717
contents = ('%11s|%s' % l for l in plan)
1718
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1719
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1720
'>>>>>>> MERGE-SOURCE\n')
1721
lines, conflicts = textmerge.merge_lines(self.reprocess)
1723
base_lines = textmerge.base_from_plan()
1726
return lines, base_lines
1728
def text_merge(self, file_id, trans_id):
1729
"""Perform a (weave) text merge for a given file and file-id.
1730
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1731
and a conflict will be noted.
1733
lines, base_lines = self._merged_lines(file_id)
1735
# Note we're checking whether the OUTPUT is binary in this case,
1736
# because we don't want to get into weave merge guts.
1737
textfile.check_text_lines(lines)
1738
self.tt.create_file(lines, trans_id)
1739
if base_lines is not None:
1741
self._raw_conflicts.append(('text conflict', trans_id))
1742
name = self.tt.final_name(trans_id)
1743
parent_id = self.tt.final_parent(trans_id)
1744
file_group = self._dump_conflicts(name, parent_id, file_id,
1746
base_lines=base_lines)
1747
file_group.append(trans_id)
1750
class LCAMerger(WeaveMerger):
1752
def _generate_merge_plan(self, file_id, base):
1753
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1756
class Diff3Merger(Merge3Merger):
1757
"""Three-way merger using external diff3 for text merging"""
1759
def dump_file(self, temp_dir, name, tree, file_id):
1760
out_path = osutils.pathjoin(temp_dir, name)
1761
out_file = open(out_path, "wb")
1763
in_file = tree.get_file(file_id)
1764
for line in in_file:
1765
out_file.write(line)
1770
def text_merge(self, file_id, trans_id):
1771
"""Perform a diff3 merge using a specified file-id and trans-id.
1772
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1773
will be dumped, and a will be conflict noted.
1776
temp_dir = osutils.mkdtemp(prefix="bzr-")
1778
new_file = osutils.pathjoin(temp_dir, "new")
1779
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1780
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1781
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1782
status = bzrlib.patch.diff3(new_file, this, base, other)
1783
if status not in (0, 1):
1784
raise errors.BzrError("Unhandled diff3 exit code")
1785
f = open(new_file, 'rb')
1787
self.tt.create_file(f, trans_id)
1791
name = self.tt.final_name(trans_id)
1792
parent_id = self.tt.final_parent(trans_id)
1793
self._dump_conflicts(name, parent_id, file_id)
1794
self._raw_conflicts.append(('text conflict', trans_id))
1796
osutils.rmtree(temp_dir)
1799
class PathNotInTree(errors.BzrError):
1801
_fmt = """Merge-into failed because %(tree)s does not contain %(path)s."""
1803
def __init__(self, path, tree):
1804
errors.BzrError.__init__(self, path=path, tree=tree)
1807
class MergeIntoMerger(Merger):
1808
"""Merger that understands other_tree will be merged into a subdir.
1810
This also changes the Merger api so that it uses real Branch, revision_id,
1811
and RevisonTree objects, rather than using revision specs.
1814
def __init__(self, this_tree, other_branch, other_tree, target_subdir,
1815
source_subpath, other_rev_id=None):
1816
"""Create a new MergeIntoMerger object.
1818
source_subpath in other_tree will be effectively copied to
1819
target_subdir in this_tree.
1821
:param this_tree: The tree that we will be merging into.
1822
:param other_branch: The Branch we will be merging from.
1823
:param other_tree: The RevisionTree object we want to merge.
1824
:param target_subdir: The relative path where we want to merge
1825
other_tree into this_tree
1826
:param source_subpath: The relative path specifying the subtree of
1827
other_tree to merge into this_tree.
1829
# It is assumed that we are merging a tree that is not in our current
1830
# ancestry, which means we are using the "EmptyTree" as our basis.
1831
null_ancestor_tree = this_tree.branch.repository.revision_tree(
1832
_mod_revision.NULL_REVISION)
1833
super(MergeIntoMerger, self).__init__(
1834
this_branch=this_tree.branch,
1835
this_tree=this_tree,
1836
other_tree=other_tree,
1837
base_tree=null_ancestor_tree,
1839
self._target_subdir = target_subdir
1840
self._source_subpath = source_subpath
1841
self.other_branch = other_branch
1842
if other_rev_id is None:
1843
other_rev_id = other_tree.get_revision_id()
1844
self.other_rev_id = self.other_basis = other_rev_id
1845
self.base_is_ancestor = True
1846
self.backup_files = True
1847
self.merge_type = Merge3Merger
1848
self.show_base = False
1849
self.reprocess = False
1850
self.interesting_ids = None
1851
self.merge_type = _MergeTypeParameterizer(MergeIntoMergeType,
1852
target_subdir=self._target_subdir,
1853
source_subpath=self._source_subpath)
1854
if self._source_subpath != '':
1855
# If this isn't a partial merge make sure the revisions will be
1857
self._maybe_fetch(self.other_branch, self.this_branch,
1860
def set_pending(self):
1861
if self._source_subpath != '':
1863
Merger.set_pending(self)
1866
class _MergeTypeParameterizer(object):
1867
"""Wrap a merge-type class to provide extra parameters.
1869
This is hack used by MergeIntoMerger to pass some extra parameters to its
1870
merge_type. Merger.do_merge() sets up its own set of parameters to pass to
1871
the 'merge_type' member. It is difficult override do_merge without
1872
re-writing the whole thing, so instead we create a wrapper which will pass
1873
the extra parameters.
1876
def __init__(self, merge_type, **kwargs):
1877
self._extra_kwargs = kwargs
1878
self._merge_type = merge_type
1880
def __call__(self, *args, **kwargs):
1881
kwargs.update(self._extra_kwargs)
1882
return self._merge_type(*args, **kwargs)
1884
def __getattr__(self, name):
1885
return getattr(self._merge_type, name)
1888
class MergeIntoMergeType(Merge3Merger):
1889
"""Merger that incorporates a tree (or part of a tree) into another."""
1891
def __init__(self, *args, **kwargs):
1892
"""Initialize the merger object.
1894
:param args: See Merge3Merger.__init__'s args.
1895
:param kwargs: See Merge3Merger.__init__'s keyword args, except for
1896
source_subpath and target_subdir.
1897
:keyword source_subpath: The relative path specifying the subtree of
1898
other_tree to merge into this_tree.
1899
:keyword target_subdir: The relative path where we want to merge
1900
other_tree into this_tree
1902
# All of the interesting work happens during Merge3Merger.__init__(),
1903
# so we have have to hack in to get our extra parameters set.
1904
self._source_subpath = kwargs.pop('source_subpath')
1905
self._target_subdir = kwargs.pop('target_subdir')
1906
super(MergeIntoMergeType, self).__init__(*args, **kwargs)
1908
def _compute_transform(self):
1909
child_pb = ui.ui_factory.nested_progress_bar()
1911
entries = self._entries_to_incorporate()
1912
entries = list(entries)
1913
for num, (entry, parent_id) in enumerate(entries):
1914
child_pb.update(gettext('Preparing file merge'), num, len(entries))
1915
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1916
trans_id = transform.new_by_entry(self.tt, entry,
1917
parent_trans_id, self.other_tree)
1920
self._finish_computing_transform()
1922
def _entries_to_incorporate(self):
1923
"""Yields pairs of (inventory_entry, new_parent)."""
1924
other_inv = self.other_tree.root_inventory
1925
subdir_id = other_inv.path2id(self._source_subpath)
1926
if subdir_id is None:
1927
# XXX: The error would be clearer if it gave the URL of the source
1928
# branch, but we don't have a reference to that here.
1929
raise PathNotInTree(self._source_subpath, "Source tree")
1930
subdir = other_inv[subdir_id]
1931
parent_in_target = osutils.dirname(self._target_subdir)
1932
target_id = self.this_tree.path2id(parent_in_target)
1933
if target_id is None:
1934
raise PathNotInTree(self._target_subdir, "Target tree")
1935
name_in_target = osutils.basename(self._target_subdir)
1936
merge_into_root = subdir.copy()
1937
merge_into_root.name = name_in_target
1938
if self.this_tree.has_id(merge_into_root.file_id):
1939
# Give the root a new file-id.
1940
# This can happen fairly easily if the directory we are
1941
# incorporating is the root, and both trees have 'TREE_ROOT' as
1942
# their root_id. Users will expect this to Just Work, so we
1943
# change the file-id here.
1944
# Non-root file-ids could potentially conflict too. That's really
1945
# an edge case, so we don't do anything special for those. We let
1946
# them cause conflicts.
1947
merge_into_root.file_id = generate_ids.gen_file_id(name_in_target)
1948
yield (merge_into_root, target_id)
1949
if subdir.kind != 'directory':
1950
# No children, so we are done.
1952
for ignored_path, entry in other_inv.iter_entries_by_dir(subdir_id):
1953
parent_id = entry.parent_id
1954
if parent_id == subdir.file_id:
1955
# The root's parent ID has changed, so make sure children of
1956
# the root refer to the new ID.
1957
parent_id = merge_into_root.file_id
1958
yield (entry, parent_id)
1961
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1963
merge_type=Merge3Merger,
1964
interesting_ids=None,
1968
interesting_files=None,
1971
change_reporter=None):
1972
"""Primary interface for merging.
1974
Typical use is probably::
1976
merge_inner(branch, branch.get_revision_tree(other_revision),
1977
branch.get_revision_tree(base_revision))
1979
if this_tree is None:
1980
raise errors.BzrError("bzrlib.merge.merge_inner requires a this_tree "
1982
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1983
pb=pb, change_reporter=change_reporter)
1984
merger.backup_files = backup_files
1985
merger.merge_type = merge_type
1986
merger.interesting_ids = interesting_ids
1987
merger.ignore_zero = ignore_zero
1988
if interesting_files:
1990
raise ValueError('Only supply interesting_ids'
1991
' or interesting_files')
1992
merger.interesting_files = interesting_files
1993
merger.show_base = show_base
1994
merger.reprocess = reprocess
1995
merger.other_rev_id = other_rev_id
1996
merger.other_basis = other_rev_id
1997
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1998
if get_revision_id is None:
1999
get_revision_id = base_tree.last_revision
2000
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
2001
merger.set_base_revision(get_revision_id(), this_branch)
2002
return merger.do_merge()
2005
merge_type_registry = registry.Registry()
2006
merge_type_registry.register('diff3', Diff3Merger,
2007
"Merge using external diff3.")
2008
merge_type_registry.register('lca', LCAMerger,
2009
"LCA-newness merge.")
2010
merge_type_registry.register('merge3', Merge3Merger,
2011
"Native diff3-style merge.")
2012
merge_type_registry.register('weave', WeaveMerger,
2013
"Weave-based merge.")
2016
def get_merge_type_registry():
2017
"""Merge type registry was previously in bzrlib.option
2019
This method provides a backwards compatible way to retrieve it.
2021
return merge_type_registry
2024
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
2025
def status_a(revision, text):
2026
if revision in ancestors_b:
2027
return 'killed-b', text
2029
return 'new-a', text
2031
def status_b(revision, text):
2032
if revision in ancestors_a:
2033
return 'killed-a', text
2035
return 'new-b', text
2037
plain_a = [t for (a, t) in annotated_a]
2038
plain_b = [t for (a, t) in annotated_b]
2039
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
2040
blocks = matcher.get_matching_blocks()
2043
for ai, bi, l in blocks:
2044
# process all mismatched sections
2045
# (last mismatched section is handled because blocks always
2046
# includes a 0-length last block)
2047
for revision, text in annotated_a[a_cur:ai]:
2048
yield status_a(revision, text)
2049
for revision, text in annotated_b[b_cur:bi]:
2050
yield status_b(revision, text)
2051
# and now the matched section
2054
for text_a in plain_a[ai:a_cur]:
2055
yield "unchanged", text_a
2058
class _PlanMergeBase(object):
2060
def __init__(self, a_rev, b_rev, vf, key_prefix):
2063
:param a_rev: Revision-id of one revision to merge
2064
:param b_rev: Revision-id of the other revision to merge
2065
:param vf: A VersionedFiles containing both revisions
2066
:param key_prefix: A prefix for accessing keys in vf, typically
2072
self._last_lines = None
2073
self._last_lines_revision_id = None
2074
self._cached_matching_blocks = {}
2075
self._key_prefix = key_prefix
2076
self._precache_tip_lines()
2078
def _precache_tip_lines(self):
2079
lines = self.get_lines([self.a_rev, self.b_rev])
2080
self.lines_a = lines[self.a_rev]
2081
self.lines_b = lines[self.b_rev]
2083
def get_lines(self, revisions):
2084
"""Get lines for revisions from the backing VersionedFiles.
2086
:raises RevisionNotPresent: on absent texts.
2088
keys = [(self._key_prefix + (rev,)) for rev in revisions]
2090
for record in self.vf.get_record_stream(keys, 'unordered', True):
2091
if record.storage_kind == 'absent':
2092
raise errors.RevisionNotPresent(record.key, self.vf)
2093
result[record.key[-1]] = osutils.chunks_to_lines(
2094
record.get_bytes_as('chunked'))
2097
def plan_merge(self):
2098
"""Generate a 'plan' for merging the two revisions.
2100
This involves comparing their texts and determining the cause of
2101
differences. If text A has a line and text B does not, then either the
2102
line was added to text A, or it was deleted from B. Once the causes
2103
are combined, they are written out in the format described in
2104
VersionedFile.plan_merge
2106
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
2107
unique_a, unique_b = self._unique_lines(blocks)
2108
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
2109
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
2110
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
2112
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
2115
for i, j, n in blocks:
2116
for a_index in range(last_i, i):
2117
if a_index in new_a:
2118
if a_index in killed_b:
2119
yield 'conflicted-a', self.lines_a[a_index]
2121
yield 'new-a', self.lines_a[a_index]
2123
yield 'killed-b', self.lines_a[a_index]
2124
for b_index in range(last_j, j):
2125
if b_index in new_b:
2126
if b_index in killed_a:
2127
yield 'conflicted-b', self.lines_b[b_index]
2129
yield 'new-b', self.lines_b[b_index]
2131
yield 'killed-a', self.lines_b[b_index]
2132
# handle common lines
2133
for a_index in range(i, i+n):
2134
yield 'unchanged', self.lines_a[a_index]
2138
def _get_matching_blocks(self, left_revision, right_revision):
2139
"""Return a description of which sections of two revisions match.
2141
See SequenceMatcher.get_matching_blocks
2143
cached = self._cached_matching_blocks.get((left_revision,
2145
if cached is not None:
2147
if self._last_lines_revision_id == left_revision:
2148
left_lines = self._last_lines
2149
right_lines = self.get_lines([right_revision])[right_revision]
2151
lines = self.get_lines([left_revision, right_revision])
2152
left_lines = lines[left_revision]
2153
right_lines = lines[right_revision]
2154
self._last_lines = right_lines
2155
self._last_lines_revision_id = right_revision
2156
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
2158
return matcher.get_matching_blocks()
2160
def _unique_lines(self, matching_blocks):
2161
"""Analyse matching_blocks to determine which lines are unique
2163
:return: a tuple of (unique_left, unique_right), where the values are
2164
sets of line numbers of unique lines.
2170
for i, j, n in matching_blocks:
2171
unique_left.extend(range(last_i, i))
2172
unique_right.extend(range(last_j, j))
2175
return unique_left, unique_right
2178
def _subtract_plans(old_plan, new_plan):
2179
"""Remove changes from new_plan that came from old_plan.
2181
It is assumed that the difference between the old_plan and new_plan
2182
is their choice of 'b' text.
2184
All lines from new_plan that differ from old_plan are emitted
2185
verbatim. All lines from new_plan that match old_plan but are
2186
not about the 'b' revision are emitted verbatim.
2188
Lines that match and are about the 'b' revision are the lines we
2189
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
2190
is skipped entirely.
2192
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
2195
for i, j, n in matcher.get_matching_blocks():
2196
for jj in range(last_j, j):
2198
for jj in range(j, j+n):
2199
plan_line = new_plan[jj]
2200
if plan_line[0] == 'new-b':
2202
elif plan_line[0] == 'killed-b':
2203
yield 'unchanged', plan_line[1]
2209
class _PlanMerge(_PlanMergeBase):
2210
"""Plan an annotate merge using on-the-fly annotation"""
2212
def __init__(self, a_rev, b_rev, vf, key_prefix):
2213
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
2214
self.a_key = self._key_prefix + (self.a_rev,)
2215
self.b_key = self._key_prefix + (self.b_rev,)
2216
self.graph = _mod_graph.Graph(self.vf)
2217
heads = self.graph.heads((self.a_key, self.b_key))
2219
# one side dominates, so we can just return its values, yay for
2221
# Ideally we would know that before we get this far
2222
self._head_key = heads.pop()
2223
if self._head_key == self.a_key:
2227
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2228
self._head_key[-1], other)
2231
self._head_key = None
2234
def _precache_tip_lines(self):
2235
# Turn this into a no-op, because we will do this later
2238
def _find_recursive_lcas(self):
2239
"""Find all the ancestors back to a unique lca"""
2240
cur_ancestors = (self.a_key, self.b_key)
2241
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2242
# rather than a key tuple. We will just map that directly to no common
2246
next_lcas = self.graph.find_lca(*cur_ancestors)
2247
# Map a plain NULL_REVISION to a simple no-ancestors
2248
if next_lcas == set([_mod_revision.NULL_REVISION]):
2250
# Order the lca's based on when they were merged into the tip
2251
# While the actual merge portion of weave merge uses a set() of
2252
# active revisions, the order of insertion *does* effect the
2253
# implicit ordering of the texts.
2254
for rev_key in cur_ancestors:
2255
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2257
parent_map[rev_key] = ordered_parents
2258
if len(next_lcas) == 0:
2260
elif len(next_lcas) == 1:
2261
parent_map[list(next_lcas)[0]] = ()
2263
elif len(next_lcas) > 2:
2264
# More than 2 lca's, fall back to grabbing all nodes between
2265
# this and the unique lca.
2266
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2268
self.a_key, self.b_key, cur_ancestors)
2269
cur_lcas = next_lcas
2270
while len(cur_lcas) > 1:
2271
cur_lcas = self.graph.find_lca(*cur_lcas)
2272
if len(cur_lcas) == 0:
2273
# No common base to find, use the full ancestry
2276
unique_lca = list(cur_lcas)[0]
2277
if unique_lca == _mod_revision.NULL_REVISION:
2278
# find_lca will return a plain 'NULL_REVISION' rather
2279
# than a key tuple when there is no common ancestor, we
2280
# prefer to just use None, because it doesn't confuse
2281
# _get_interesting_texts()
2283
parent_map.update(self._find_unique_parents(next_lcas,
2286
cur_ancestors = next_lcas
2289
def _find_unique_parents(self, tip_keys, base_key):
2290
"""Find ancestors of tip that aren't ancestors of base.
2292
:param tip_keys: Nodes that are interesting
2293
:param base_key: Cull all ancestors of this node
2294
:return: The parent map for all revisions between tip_keys and
2295
base_key. base_key will be included. References to nodes outside of
2296
the ancestor set will also be removed.
2298
# TODO: this would be simpler if find_unique_ancestors took a list
2299
# instead of a single tip, internally it supports it, but it
2300
# isn't a "backwards compatible" api change.
2301
if base_key is None:
2302
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2303
# We remove NULL_REVISION because it isn't a proper tuple key, and
2304
# thus confuses things like _get_interesting_texts, and our logic
2305
# to add the texts into the memory weave.
2306
if _mod_revision.NULL_REVISION in parent_map:
2307
parent_map.pop(_mod_revision.NULL_REVISION)
2310
for tip in tip_keys:
2312
self.graph.find_unique_ancestors(tip, [base_key]))
2313
parent_map = self.graph.get_parent_map(interesting)
2314
parent_map[base_key] = ()
2315
culled_parent_map, child_map, tails = self._remove_external_references(
2317
# Remove all the tails but base_key
2318
if base_key is not None:
2319
tails.remove(base_key)
2320
self._prune_tails(culled_parent_map, child_map, tails)
2321
# Now remove all the uninteresting 'linear' regions
2322
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2326
def _remove_external_references(parent_map):
2327
"""Remove references that go outside of the parent map.
2329
:param parent_map: Something returned from Graph.get_parent_map(keys)
2330
:return: (filtered_parent_map, child_map, tails)
2331
filtered_parent_map is parent_map without external references
2332
child_map is the {parent_key: [child_keys]} mapping
2333
tails is a list of nodes that do not have any parents in the map
2335
# TODO: The basic effect of this function seems more generic than
2336
# _PlanMerge. But the specific details of building a child_map,
2337
# and computing tails seems very specific to _PlanMerge.
2338
# Still, should this be in Graph land?
2339
filtered_parent_map = {}
2342
for key, parent_keys in parent_map.iteritems():
2343
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2344
if not culled_parent_keys:
2346
for parent_key in culled_parent_keys:
2347
child_map.setdefault(parent_key, []).append(key)
2348
# TODO: Do we want to do this, it adds overhead for every node,
2349
# just to say that the node has no children
2350
child_map.setdefault(key, [])
2351
filtered_parent_map[key] = culled_parent_keys
2352
return filtered_parent_map, child_map, tails
2355
def _prune_tails(parent_map, child_map, tails_to_remove):
2356
"""Remove tails from the parent map.
2358
This will remove the supplied revisions until no more children have 0
2361
:param parent_map: A dict of {child: [parents]}, this dictionary will
2362
be modified in place.
2363
:param tails_to_remove: A list of tips that should be removed,
2364
this list will be consumed
2365
:param child_map: The reverse dict of parent_map ({parent: [children]})
2366
this dict will be modified
2367
:return: None, parent_map will be modified in place.
2369
while tails_to_remove:
2370
next = tails_to_remove.pop()
2371
parent_map.pop(next)
2372
children = child_map.pop(next)
2373
for child in children:
2374
child_parents = parent_map[child]
2375
child_parents.remove(next)
2376
if len(child_parents) == 0:
2377
tails_to_remove.append(child)
2379
def _get_interesting_texts(self, parent_map):
2380
"""Return a dict of texts we are interested in.
2382
Note that the input is in key tuples, but the output is in plain
2385
:param parent_map: The output from _find_recursive_lcas
2386
:return: A dict of {'revision_id':lines} as returned by
2387
_PlanMergeBase.get_lines()
2389
all_revision_keys = set(parent_map)
2390
all_revision_keys.add(self.a_key)
2391
all_revision_keys.add(self.b_key)
2393
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2394
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2397
def _build_weave(self):
2398
from bzrlib import weave
2399
self._weave = weave.Weave(weave_name='in_memory_weave',
2400
allow_reserved=True)
2401
parent_map = self._find_recursive_lcas()
2403
all_texts = self._get_interesting_texts(parent_map)
2405
# Note: Unfortunately, the order given by topo_sort will effect the
2406
# ordering resolution in the output. Specifically, if you add A then B,
2407
# then in the output text A lines will show up before B lines. And, of
2408
# course, topo_sort doesn't guarantee any real ordering.
2409
# So we use merge_sort, and add a fake node on the tip.
2410
# This ensures that left-hand parents will always be inserted into the
2411
# weave before right-hand parents.
2412
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2413
parent_map[tip_key] = (self.a_key, self.b_key)
2415
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2419
# for key in tsort.topo_sort(parent_map):
2420
parent_keys = parent_map[key]
2421
revision_id = key[-1]
2422
parent_ids = [k[-1] for k in parent_keys]
2423
self._weave.add_lines(revision_id, parent_ids,
2424
all_texts[revision_id])
2426
def plan_merge(self):
2427
"""Generate a 'plan' for merging the two revisions.
2429
This involves comparing their texts and determining the cause of
2430
differences. If text A has a line and text B does not, then either the
2431
line was added to text A, or it was deleted from B. Once the causes
2432
are combined, they are written out in the format described in
2433
VersionedFile.plan_merge
2435
if self._head_key is not None: # There was a single head
2436
if self._head_key == self.a_key:
2439
if self._head_key != self.b_key:
2440
raise AssertionError('There was an invalid head: %s != %s'
2441
% (self.b_key, self._head_key))
2443
head_rev = self._head_key[-1]
2444
lines = self.get_lines([head_rev])[head_rev]
2445
return ((plan, line) for line in lines)
2446
return self._weave.plan_merge(self.a_rev, self.b_rev)
2449
class _PlanLCAMerge(_PlanMergeBase):
2451
This merge algorithm differs from _PlanMerge in that:
2453
1. comparisons are done against LCAs only
2454
2. cases where a contested line is new versus one LCA but old versus
2455
another are marked as conflicts, by emitting the line as conflicted-a
2458
This is faster, and hopefully produces more useful output.
2461
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2462
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2463
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2466
if lca == _mod_revision.NULL_REVISION:
2469
self.lcas.add(lca[-1])
2470
for lca in self.lcas:
2471
if _mod_revision.is_null(lca):
2474
lca_lines = self.get_lines([lca])[lca]
2475
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2477
blocks = list(matcher.get_matching_blocks())
2478
self._cached_matching_blocks[(a_rev, lca)] = blocks
2479
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2481
blocks = list(matcher.get_matching_blocks())
2482
self._cached_matching_blocks[(b_rev, lca)] = blocks
2484
def _determine_status(self, revision_id, unique_line_numbers):
2485
"""Determines the status unique lines versus all lcas.
2487
Basically, determines why the line is unique to this revision.
2489
A line may be determined new, killed, or both.
2491
If a line is determined new, that means it was not present in at least
2492
one LCA, and is not present in the other merge revision.
2494
If a line is determined killed, that means the line was present in
2497
If a line is killed and new, this indicates that the two merge
2498
revisions contain differing conflict resolutions.
2500
:param revision_id: The id of the revision in which the lines are
2502
:param unique_line_numbers: The line numbers of unique lines.
2503
:return: a tuple of (new_this, killed_other)
2507
unique_line_numbers = set(unique_line_numbers)
2508
for lca in self.lcas:
2509
blocks = self._get_matching_blocks(revision_id, lca)
2510
unique_vs_lca, _ignored = self._unique_lines(blocks)
2511
new.update(unique_line_numbers.intersection(unique_vs_lca))
2512
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
