1
# Copyright (C) 2005-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
22
branch as _mod_branch,
24
conflicts as _mod_conflicts,
31
revision as _mod_revision,
47
from bzrlib.symbol_versioning import (
51
# TODO: Report back as changes are merged in
54
def transform_tree(from_tree, to_tree, interesting_ids=None):
55
from_tree.lock_tree_write()
56
operation = cleanup.OperationWithCleanups(merge_inner)
57
operation.add_cleanup(from_tree.unlock)
58
operation.run_simple(from_tree.branch, to_tree, from_tree,
59
ignore_zero=True, interesting_ids=interesting_ids, this_tree=from_tree)
62
class MergeHooks(hooks.Hooks):
65
hooks.Hooks.__init__(self, "bzrlib.merge", "Merger.hooks")
66
self.add_hook('merge_file_content',
67
"Called with a bzrlib.merge.Merger object to create a per file "
68
"merge object when starting a merge. "
69
"Should return either None or a subclass of "
70
"``bzrlib.merge.AbstractPerFileMerger``. "
71
"Such objects will then be called per file "
72
"that needs to be merged (including when one "
73
"side has deleted the file and the other has changed it). "
74
"See the AbstractPerFileMerger API docs for details on how it is "
79
class AbstractPerFileMerger(object):
80
"""PerFileMerger objects are used by plugins extending merge for bzrlib.
82
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
84
:ivar merger: The Merge3Merger performing the merge.
87
def __init__(self, merger):
88
"""Create a PerFileMerger for use with merger."""
91
def merge_contents(self, merge_params):
92
"""Attempt to merge the contents of a single file.
94
:param merge_params: A bzrlib.merge.MergeHookParams
95
:return : A tuple of (status, chunks), where status is one of
96
'not_applicable', 'success', 'conflicted', or 'delete'. If status
97
is 'success' or 'conflicted', then chunks should be an iterable of
98
strings for the new file contents.
100
return ('not applicable', None)
103
class PerFileMerger(AbstractPerFileMerger):
104
"""Merge individual files when self.file_matches returns True.
106
This class is intended to be subclassed. The file_matches and
107
merge_matching methods should be overridden with concrete implementations.
110
def file_matches(self, params):
111
"""Return True if merge_matching should be called on this file.
113
Only called with merges of plain files with no clear winner.
115
Subclasses must override this.
117
raise NotImplementedError(self.file_matches)
119
def get_filename(self, params, tree):
120
"""Lookup the filename (i.e. basename, not path), given a Tree (e.g.
121
self.merger.this_tree) and a MergeHookParams.
123
return osutils.basename(tree.id2path(params.file_id))
125
def get_filepath(self, params, tree):
126
"""Calculate the path to the file in a tree.
128
:param params: A MergeHookParams describing the file to merge
129
:param tree: a Tree, e.g. self.merger.this_tree.
131
return tree.id2path(params.file_id)
133
def merge_contents(self, params):
134
"""Merge the contents of a single file."""
135
# Check whether this custom merge logic should be used.
137
# OTHER is a straight winner, rely on default merge.
138
params.winner == 'other' or
139
# THIS and OTHER aren't both files.
140
not params.is_file_merge() or
141
# The filename doesn't match *.xml
142
not self.file_matches(params)):
143
return 'not_applicable', None
144
return self.merge_matching(params)
146
def merge_matching(self, params):
147
"""Merge the contents of a single file that has matched the criteria
148
in PerFileMerger.merge_contents (is a conflict, is a file,
149
self.file_matches is True).
151
Subclasses must override this.
153
raise NotImplementedError(self.merge_matching)
156
class ConfigurableFileMerger(PerFileMerger):
157
"""Merge individual files when configured via a .conf file.
159
This is a base class for concrete custom file merging logic. Concrete
160
classes should implement ``merge_text``.
162
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
164
:ivar affected_files: The configured file paths to merge.
166
:cvar name_prefix: The prefix to use when looking up configuration
167
details. <name_prefix>_merge_files describes the files targeted by the
170
:cvar default_files: The default file paths to merge when no configuration
177
def __init__(self, merger):
178
super(ConfigurableFileMerger, self).__init__(merger)
179
self.affected_files = None
180
self.default_files = self.__class__.default_files or []
181
self.name_prefix = self.__class__.name_prefix
182
if self.name_prefix is None:
183
raise ValueError("name_prefix must be set.")
185
def file_matches(self, params):
186
"""Check whether the file should call the merge hook.
188
<name_prefix>_merge_files configuration variable is a list of files
189
that should use the hook.
191
affected_files = self.affected_files
192
if affected_files is None:
193
config = self.merger.this_branch.get_config()
194
# Until bzr provides a better policy for caching the config, we
195
# just add the part we're interested in to the params to avoid
196
# reading the config files repeatedly (bazaar.conf, location.conf,
198
config_key = self.name_prefix + '_merge_files'
199
affected_files = config.get_user_option_as_list(config_key)
200
if affected_files is None:
201
# If nothing was specified in the config, use the default.
202
affected_files = self.default_files
203
self.affected_files = affected_files
205
filepath = self.get_filepath(params, self.merger.this_tree)
206
if filepath in affected_files:
210
def merge_matching(self, params):
211
return self.merge_text(params)
213
def merge_text(self, params):
214
"""Merge the byte contents of a single file.
216
This is called after checking that the merge should be performed in
217
merge_contents, and it should behave as per
218
``bzrlib.merge.AbstractPerFileMerger.merge_contents``.
220
raise NotImplementedError(self.merge_text)
223
class MergeHookParams(object):
224
"""Object holding parameters passed to merge_file_content hooks.
226
There are some fields hooks can access:
228
:ivar file_id: the file ID of the file being merged
229
:ivar trans_id: the transform ID for the merge of this file
230
:ivar this_kind: kind of file_id in 'this' tree
231
:ivar other_kind: kind of file_id in 'other' tree
232
:ivar winner: one of 'this', 'other', 'conflict'
235
def __init__(self, merger, file_id, trans_id, this_kind, other_kind,
237
self._merger = merger
238
self.file_id = file_id
239
self.trans_id = trans_id
240
self.this_kind = this_kind
241
self.other_kind = other_kind
244
def is_file_merge(self):
245
"""True if this_kind and other_kind are both 'file'."""
246
return self.this_kind == 'file' and self.other_kind == 'file'
248
@decorators.cachedproperty
249
def base_lines(self):
250
"""The lines of the 'base' version of the file."""
251
return self._merger.get_lines(self._merger.base_tree, self.file_id)
253
@decorators.cachedproperty
254
def this_lines(self):
255
"""The lines of the 'this' version of the file."""
256
return self._merger.get_lines(self._merger.this_tree, self.file_id)
258
@decorators.cachedproperty
259
def other_lines(self):
260
"""The lines of the 'other' version of the file."""
261
return self._merger.get_lines(self._merger.other_tree, self.file_id)
264
class Merger(object):
268
def __init__(self, this_branch, other_tree=None, base_tree=None,
269
this_tree=None, pb=None, change_reporter=None,
270
recurse='down', revision_graph=None):
271
object.__init__(self)
272
self.this_branch = this_branch
273
self.this_basis = _mod_revision.ensure_null(
274
this_branch.last_revision())
275
self.this_rev_id = None
276
self.this_tree = this_tree
277
self.this_revision_tree = None
278
self.this_basis_tree = None
279
self.other_tree = other_tree
280
self.other_branch = None
281
self.base_tree = base_tree
282
self.ignore_zero = False
283
self.backup_files = False
284
self.interesting_ids = None
285
self.interesting_files = None
286
self.show_base = False
287
self.reprocess = False
289
warnings.warn("pb parameter to Merger() is deprecated and ignored")
291
self.recurse = recurse
292
self.change_reporter = change_reporter
293
self._cached_trees = {}
294
self._revision_graph = revision_graph
295
self._base_is_ancestor = None
296
self._base_is_other_ancestor = None
297
self._is_criss_cross = None
298
self._lca_trees = None
300
def cache_trees_with_revision_ids(self, trees):
301
"""Cache any tree in trees if it has a revision_id."""
302
for maybe_tree in trees:
303
if maybe_tree is None:
306
rev_id = maybe_tree.get_revision_id()
307
except AttributeError:
309
self._cached_trees[rev_id] = maybe_tree
312
def revision_graph(self):
313
if self._revision_graph is None:
314
self._revision_graph = self.this_branch.repository.get_graph()
315
return self._revision_graph
317
def _set_base_is_ancestor(self, value):
318
self._base_is_ancestor = value
320
def _get_base_is_ancestor(self):
321
if self._base_is_ancestor is None:
322
self._base_is_ancestor = self.revision_graph.is_ancestor(
323
self.base_rev_id, self.this_basis)
324
return self._base_is_ancestor
326
base_is_ancestor = property(_get_base_is_ancestor, _set_base_is_ancestor)
328
def _set_base_is_other_ancestor(self, value):
329
self._base_is_other_ancestor = value
331
def _get_base_is_other_ancestor(self):
332
if self._base_is_other_ancestor is None:
333
if self.other_basis is None:
335
self._base_is_other_ancestor = self.revision_graph.is_ancestor(
336
self.base_rev_id, self.other_basis)
337
return self._base_is_other_ancestor
339
base_is_other_ancestor = property(_get_base_is_other_ancestor,
340
_set_base_is_other_ancestor)
343
def from_uncommitted(tree, other_tree, pb=None, base_tree=None):
344
"""Return a Merger for uncommitted changes in other_tree.
346
:param tree: The tree to merge into
347
:param other_tree: The tree to get uncommitted changes from
348
:param pb: A progress indicator
349
:param base_tree: The basis to use for the merge. If unspecified,
350
other_tree.basis_tree() will be used.
352
if base_tree is None:
353
base_tree = other_tree.basis_tree()
354
merger = Merger(tree.branch, other_tree, base_tree, tree, pb)
355
merger.base_rev_id = merger.base_tree.get_revision_id()
356
merger.other_rev_id = None
357
merger.other_basis = merger.base_rev_id
361
def from_mergeable(klass, tree, mergeable, pb):
362
"""Return a Merger for a bundle or merge directive.
364
:param tree: The tree to merge changes into
365
:param mergeable: A merge directive or bundle
366
:param pb: A progress indicator
368
mergeable.install_revisions(tree.branch.repository)
369
base_revision_id, other_revision_id, verified =\
370
mergeable.get_merge_request(tree.branch.repository)
371
revision_graph = tree.branch.repository.get_graph()
372
if base_revision_id is not None:
373
if (base_revision_id != _mod_revision.NULL_REVISION and
374
revision_graph.is_ancestor(
375
base_revision_id, tree.branch.last_revision())):
376
base_revision_id = None
378
trace.warning('Performing cherrypick')
379
merger = klass.from_revision_ids(pb, tree, other_revision_id,
380
base_revision_id, revision_graph=
382
return merger, verified
385
def from_revision_ids(pb, tree, other, base=None, other_branch=None,
386
base_branch=None, revision_graph=None,
388
"""Return a Merger for revision-ids.
390
:param pb: A progress indicator
391
:param tree: The tree to merge changes into
392
:param other: The revision-id to use as OTHER
393
:param base: The revision-id to use as BASE. If not specified, will
395
:param other_branch: A branch containing the other revision-id. If
396
not supplied, tree.branch is used.
397
:param base_branch: A branch containing the base revision-id. If
398
not supplied, other_branch or tree.branch will be used.
399
:param revision_graph: If you have a revision_graph precomputed, pass
400
it in, otherwise it will be created for you.
401
:param tree_branch: The branch associated with tree. If not supplied,
402
tree.branch will be used.
404
if tree_branch is None:
405
tree_branch = tree.branch
406
merger = Merger(tree_branch, this_tree=tree, pb=pb,
407
revision_graph=revision_graph)
408
if other_branch is None:
409
other_branch = tree.branch
410
merger.set_other_revision(other, other_branch)
414
if base_branch is None:
415
base_branch = other_branch
416
merger.set_base_revision(base, base_branch)
419
def revision_tree(self, revision_id, branch=None):
420
if revision_id not in self._cached_trees:
422
branch = self.this_branch
424
tree = self.this_tree.revision_tree(revision_id)
425
except errors.NoSuchRevisionInTree:
426
tree = branch.repository.revision_tree(revision_id)
427
self._cached_trees[revision_id] = tree
428
return self._cached_trees[revision_id]
430
def _get_tree(self, treespec, possible_transports=None):
431
location, revno = treespec
433
tree = workingtree.WorkingTree.open_containing(location)[0]
434
return tree.branch, tree
435
branch = _mod_branch.Branch.open_containing(
436
location, possible_transports)[0]
438
revision_id = branch.last_revision()
440
revision_id = branch.get_rev_id(revno)
441
revision_id = _mod_revision.ensure_null(revision_id)
442
return branch, self.revision_tree(revision_id, branch)
444
@deprecated_method(deprecated_in((2, 1, 0)))
445
def ensure_revision_trees(self):
446
if self.this_revision_tree is None:
447
self.this_basis_tree = self.revision_tree(self.this_basis)
448
if self.this_basis == self.this_rev_id:
449
self.this_revision_tree = self.this_basis_tree
451
if self.other_rev_id is None:
452
other_basis_tree = self.revision_tree(self.other_basis)
453
if other_basis_tree.has_changes(self.other_tree):
454
raise errors.WorkingTreeNotRevision(self.this_tree)
455
other_rev_id = self.other_basis
456
self.other_tree = other_basis_tree
458
@deprecated_method(deprecated_in((2, 1, 0)))
459
def file_revisions(self, file_id):
460
self.ensure_revision_trees()
461
if self.this_rev_id is None:
462
if self.this_basis_tree.get_file_sha1(file_id) != \
463
self.this_tree.get_file_sha1(file_id):
464
raise errors.WorkingTreeNotRevision(self.this_tree)
466
trees = (self.this_basis_tree, self.other_tree)
467
return [tree.get_file_revision(file_id) for tree in trees]
469
@deprecated_method(deprecated_in((2, 1, 0)))
470
def check_basis(self, check_clean, require_commits=True):
471
if self.this_basis is None and require_commits is True:
472
raise errors.BzrCommandError(
473
"This branch has no commits."
474
" (perhaps you would prefer 'bzr pull')")
477
if self.this_basis != self.this_rev_id:
478
raise errors.UncommittedChanges(self.this_tree)
480
@deprecated_method(deprecated_in((2, 1, 0)))
481
def compare_basis(self):
483
basis_tree = self.revision_tree(self.this_tree.last_revision())
484
except errors.NoSuchRevision:
485
basis_tree = self.this_tree.basis_tree()
486
if not self.this_tree.has_changes(basis_tree):
487
self.this_rev_id = self.this_basis
489
def set_interesting_files(self, file_list):
490
self.interesting_files = file_list
492
def set_pending(self):
493
if (not self.base_is_ancestor or not self.base_is_other_ancestor
494
or self.other_rev_id is None):
498
def _add_parent(self):
499
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
500
new_parent_trees = []
501
operation = cleanup.OperationWithCleanups(
502
self.this_tree.set_parent_trees)
503
for revision_id in new_parents:
505
tree = self.revision_tree(revision_id)
506
except errors.NoSuchRevision:
510
operation.add_cleanup(tree.unlock)
511
new_parent_trees.append((revision_id, tree))
512
operation.run_simple(new_parent_trees, allow_leftmost_as_ghost=True)
514
def set_other(self, other_revision, possible_transports=None):
515
"""Set the revision and tree to merge from.
517
This sets the other_tree, other_rev_id, other_basis attributes.
519
:param other_revision: The [path, revision] list to merge from.
521
self.other_branch, self.other_tree = self._get_tree(other_revision,
523
if other_revision[1] == -1:
524
self.other_rev_id = _mod_revision.ensure_null(
525
self.other_branch.last_revision())
526
if _mod_revision.is_null(self.other_rev_id):
527
raise errors.NoCommits(self.other_branch)
528
self.other_basis = self.other_rev_id
529
elif other_revision[1] is not None:
530
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
531
self.other_basis = self.other_rev_id
533
self.other_rev_id = None
534
self.other_basis = self.other_branch.last_revision()
535
if self.other_basis is None:
536
raise errors.NoCommits(self.other_branch)
537
if self.other_rev_id is not None:
538
self._cached_trees[self.other_rev_id] = self.other_tree
539
self._maybe_fetch(self.other_branch,self.this_branch, self.other_basis)
541
def set_other_revision(self, revision_id, other_branch):
542
"""Set 'other' based on a branch and revision id
544
:param revision_id: The revision to use for a tree
545
:param other_branch: The branch containing this tree
547
self.other_rev_id = revision_id
548
self.other_branch = other_branch
549
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
550
self.other_tree = self.revision_tree(revision_id)
551
self.other_basis = revision_id
553
def set_base_revision(self, revision_id, branch):
554
"""Set 'base' based on a branch and revision id
556
:param revision_id: The revision to use for a tree
557
:param branch: The branch containing this tree
559
self.base_rev_id = revision_id
560
self.base_branch = branch
561
self._maybe_fetch(branch, self.this_branch, revision_id)
562
self.base_tree = self.revision_tree(revision_id)
564
def _maybe_fetch(self, source, target, revision_id):
565
if not source.repository.has_same_location(target.repository):
567
tags_to_fetch = set(source.tags.get_reverse_tag_dict())
568
except errors.TagsNotSupported:
570
fetch_spec = _mod_graph.NotInOtherForRevs(target.repository,
571
source.repository, [revision_id],
572
if_present_ids=tags_to_fetch).execute()
573
target.fetch(source, fetch_spec=fetch_spec)
576
revisions = [_mod_revision.ensure_null(self.this_basis),
577
_mod_revision.ensure_null(self.other_basis)]
578
if _mod_revision.NULL_REVISION in revisions:
579
self.base_rev_id = _mod_revision.NULL_REVISION
580
self.base_tree = self.revision_tree(self.base_rev_id)
581
self._is_criss_cross = False
583
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
584
self._is_criss_cross = False
586
self.base_rev_id = _mod_revision.NULL_REVISION
588
self.base_rev_id = list(lcas)[0]
589
else: # len(lcas) > 1
590
self._is_criss_cross = True
592
# find_unique_lca can only handle 2 nodes, so we have to
593
# start back at the beginning. It is a shame to traverse
594
# the graph again, but better than re-implementing
596
self.base_rev_id = self.revision_graph.find_unique_lca(
597
revisions[0], revisions[1])
599
self.base_rev_id = self.revision_graph.find_unique_lca(
601
sorted_lca_keys = self.revision_graph.find_merge_order(
603
if self.base_rev_id == _mod_revision.NULL_REVISION:
604
self.base_rev_id = sorted_lca_keys[0]
606
if self.base_rev_id == _mod_revision.NULL_REVISION:
607
raise errors.UnrelatedBranches()
608
if self._is_criss_cross:
609
trace.warning('Warning: criss-cross merge encountered. See bzr'
610
' help criss-cross.')
611
trace.mutter('Criss-cross lcas: %r' % lcas)
612
if self.base_rev_id in lcas:
613
trace.mutter('Unable to find unique lca. '
614
'Fallback %r as best option.' % self.base_rev_id)
615
interesting_revision_ids = set(lcas)
616
interesting_revision_ids.add(self.base_rev_id)
617
interesting_trees = dict((t.get_revision_id(), t)
618
for t in self.this_branch.repository.revision_trees(
619
interesting_revision_ids))
620
self._cached_trees.update(interesting_trees)
621
if self.base_rev_id in lcas:
622
self.base_tree = interesting_trees[self.base_rev_id]
624
self.base_tree = interesting_trees.pop(self.base_rev_id)
625
self._lca_trees = [interesting_trees[key]
626
for key in sorted_lca_keys]
628
self.base_tree = self.revision_tree(self.base_rev_id)
629
self.base_is_ancestor = True
630
self.base_is_other_ancestor = True
631
trace.mutter('Base revid: %r' % self.base_rev_id)
633
def set_base(self, base_revision):
634
"""Set the base revision to use for the merge.
636
:param base_revision: A 2-list containing a path and revision number.
638
trace.mutter("doing merge() with no base_revision specified")
639
if base_revision == [None, None]:
642
base_branch, self.base_tree = self._get_tree(base_revision)
643
if base_revision[1] == -1:
644
self.base_rev_id = base_branch.last_revision()
645
elif base_revision[1] is None:
646
self.base_rev_id = _mod_revision.NULL_REVISION
648
self.base_rev_id = _mod_revision.ensure_null(
649
base_branch.get_rev_id(base_revision[1]))
650
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
652
def make_merger(self):
653
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
654
'other_tree': self.other_tree,
655
'interesting_ids': self.interesting_ids,
656
'interesting_files': self.interesting_files,
657
'this_branch': self.this_branch,
659
if self.merge_type.requires_base:
660
kwargs['base_tree'] = self.base_tree
661
if self.merge_type.supports_reprocess:
662
kwargs['reprocess'] = self.reprocess
664
raise errors.BzrError(
665
"Conflict reduction is not supported for merge"
666
" type %s." % self.merge_type)
667
if self.merge_type.supports_show_base:
668
kwargs['show_base'] = self.show_base
670
raise errors.BzrError("Showing base is not supported for this"
671
" merge type. %s" % self.merge_type)
672
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
673
and not self.base_is_other_ancestor):
674
raise errors.CannotReverseCherrypick()
675
if self.merge_type.supports_cherrypick:
676
kwargs['cherrypick'] = (not self.base_is_ancestor or
677
not self.base_is_other_ancestor)
678
if self._is_criss_cross and getattr(self.merge_type,
679
'supports_lca_trees', False):
680
kwargs['lca_trees'] = self._lca_trees
681
return self.merge_type(pb=None,
682
change_reporter=self.change_reporter,
685
def _do_merge_to(self):
686
merge = self.make_merger()
687
if self.other_branch is not None:
688
self.other_branch.update_references(self.this_branch)
690
if self.recurse == 'down':
691
for relpath, file_id in self.this_tree.iter_references():
692
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
693
other_revision = self.other_tree.get_reference_revision(
695
if other_revision == sub_tree.last_revision():
697
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
698
sub_merge.merge_type = self.merge_type
699
other_branch = self.other_branch.reference_parent(file_id, relpath)
700
sub_merge.set_other_revision(other_revision, other_branch)
701
base_revision = self.base_tree.get_reference_revision(file_id)
702
sub_merge.base_tree = \
703
sub_tree.branch.repository.revision_tree(base_revision)
704
sub_merge.base_rev_id = base_revision
709
operation = cleanup.OperationWithCleanups(self._do_merge_to)
710
self.this_tree.lock_tree_write()
711
operation.add_cleanup(self.this_tree.unlock)
712
if self.base_tree is not None:
713
self.base_tree.lock_read()
714
operation.add_cleanup(self.base_tree.unlock)
715
if self.other_tree is not None:
716
self.other_tree.lock_read()
717
operation.add_cleanup(self.other_tree.unlock)
718
merge = operation.run_simple()
719
if len(merge.cooked_conflicts) == 0:
720
if not self.ignore_zero and not trace.is_quiet():
721
trace.note("All changes applied successfully.")
723
trace.note("%d conflicts encountered."
724
% len(merge.cooked_conflicts))
726
return len(merge.cooked_conflicts)
729
class _InventoryNoneEntry(object):
730
"""This represents an inventory entry which *isn't there*.
732
It simplifies the merging logic if we always have an InventoryEntry, even
733
if it isn't actually present
740
symlink_target = None
743
_none_entry = _InventoryNoneEntry()
746
class Merge3Merger(object):
747
"""Three-way merger that uses the merge3 text merger"""
749
supports_reprocess = True
750
supports_show_base = True
751
history_based = False
752
supports_cherrypick = True
753
supports_reverse_cherrypick = True
754
winner_idx = {"this": 2, "other": 1, "conflict": 1}
755
supports_lca_trees = True
757
def __init__(self, working_tree, this_tree, base_tree, other_tree,
758
interesting_ids=None, reprocess=False, show_base=False,
759
pb=None, pp=None, change_reporter=None,
760
interesting_files=None, do_merge=True,
761
cherrypick=False, lca_trees=None, this_branch=None):
762
"""Initialize the merger object and perform the merge.
764
:param working_tree: The working tree to apply the merge to
765
:param this_tree: The local tree in the merge operation
766
:param base_tree: The common tree in the merge operation
767
:param other_tree: The other tree to merge changes from
768
:param this_branch: The branch associated with this_tree. Defaults to
769
this_tree.branch if not supplied.
770
:param interesting_ids: The file_ids of files that should be
771
participate in the merge. May not be combined with
773
:param: reprocess If True, perform conflict-reduction processing.
774
:param show_base: If True, show the base revision in text conflicts.
775
(incompatible with reprocess)
777
:param pp: A ProgressPhase object
778
:param change_reporter: An object that should report changes made
779
:param interesting_files: The tree-relative paths of files that should
780
participate in the merge. If these paths refer to directories,
781
the contents of those directories will also be included. May not
782
be combined with interesting_ids. If neither interesting_files nor
783
interesting_ids is specified, all files may participate in the
785
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
786
if the ancestry was found to include a criss-cross merge.
787
Otherwise should be None.
789
object.__init__(self)
790
if interesting_files is not None and interesting_ids is not None:
792
'specify either interesting_ids or interesting_files')
793
if this_branch is None:
794
this_branch = this_tree.branch
795
self.interesting_ids = interesting_ids
796
self.interesting_files = interesting_files
797
self.this_tree = working_tree
798
self.base_tree = base_tree
799
self.other_tree = other_tree
800
self.this_branch = this_branch
801
self._raw_conflicts = []
802
self.cooked_conflicts = []
803
self.reprocess = reprocess
804
self.show_base = show_base
805
self._lca_trees = lca_trees
806
# Uncommenting this will change the default algorithm to always use
807
# _entries_lca. This can be useful for running the test suite and
808
# making sure we haven't missed any corner cases.
809
# if lca_trees is None:
810
# self._lca_trees = [self.base_tree]
811
self.change_reporter = change_reporter
812
self.cherrypick = cherrypick
816
warnings.warn("pp argument to Merge3Merger is deprecated")
818
warnings.warn("pb argument to Merge3Merger is deprecated")
821
operation = cleanup.OperationWithCleanups(self._do_merge)
822
self.this_tree.lock_tree_write()
823
operation.add_cleanup(self.this_tree.unlock)
824
self.base_tree.lock_read()
825
operation.add_cleanup(self.base_tree.unlock)
826
self.other_tree.lock_read()
827
operation.add_cleanup(self.other_tree.unlock)
830
def _do_merge(self, operation):
831
self.tt = transform.TreeTransform(self.this_tree, None)
832
operation.add_cleanup(self.tt.finalize)
833
self._compute_transform()
834
results = self.tt.apply(no_conflicts=True)
835
self.write_modified(results)
837
self.this_tree.add_conflicts(self.cooked_conflicts)
838
except errors.UnsupportedOperation:
841
def make_preview_transform(self):
842
operation = cleanup.OperationWithCleanups(self._make_preview_transform)
843
self.base_tree.lock_read()
844
operation.add_cleanup(self.base_tree.unlock)
845
self.other_tree.lock_read()
846
operation.add_cleanup(self.other_tree.unlock)
847
return operation.run_simple()
849
def _make_preview_transform(self):
850
self.tt = transform.TransformPreview(self.this_tree)
851
self._compute_transform()
854
def _compute_transform(self):
855
if self._lca_trees is None:
856
entries = self._entries3()
857
resolver = self._three_way
859
entries = self._entries_lca()
860
resolver = self._lca_multi_way
861
child_pb = ui.ui_factory.nested_progress_bar()
863
factories = Merger.hooks['merge_file_content']
864
hooks = [factory(self) for factory in factories] + [self]
865
self.active_hooks = [hook for hook in hooks if hook is not None]
866
for num, (file_id, changed, parents3, names3,
867
executable3) in enumerate(entries):
868
child_pb.update('Preparing file merge', num, len(entries))
869
self._merge_names(file_id, parents3, names3, resolver=resolver)
871
file_status = self._do_merge_contents(file_id)
873
file_status = 'unmodified'
874
self._merge_executable(file_id,
875
executable3, file_status, resolver=resolver)
879
self._finish_computing_transform()
881
def _finish_computing_transform(self):
882
"""Finalize the transform and report the changes.
884
This is the second half of _compute_transform.
886
child_pb = ui.ui_factory.nested_progress_bar()
888
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
889
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
892
if self.change_reporter is not None:
893
from bzrlib import delta
894
delta.report_changes(
895
self.tt.iter_changes(), self.change_reporter)
896
self.cook_conflicts(fs_conflicts)
897
for conflict in self.cooked_conflicts:
898
trace.warning(conflict)
901
"""Gather data about files modified between three trees.
903
Return a list of tuples of file_id, changed, parents3, names3,
904
executable3. changed is a boolean indicating whether the file contents
905
or kind were changed. parents3 is a tuple of parent ids for base,
906
other and this. names3 is a tuple of names for base, other and this.
907
executable3 is a tuple of execute-bit values for base, other and this.
910
iterator = self.other_tree.iter_changes(self.base_tree,
911
specific_files=self.interesting_files,
912
extra_trees=[self.this_tree])
913
this_entries = dict((e.file_id, e) for p, e in
914
self.this_tree.iter_entries_by_dir(
915
self.interesting_ids))
916
for (file_id, paths, changed, versioned, parents, names, kind,
917
executable) in iterator:
918
if (self.interesting_ids is not None and
919
file_id not in self.interesting_ids):
921
entry = this_entries.get(file_id)
922
if entry is not None:
923
this_name = entry.name
924
this_parent = entry.parent_id
925
this_executable = entry.executable
929
this_executable = None
930
parents3 = parents + (this_parent,)
931
names3 = names + (this_name,)
932
executable3 = executable + (this_executable,)
933
result.append((file_id, changed, parents3, names3, executable3))
936
def _entries_lca(self):
937
"""Gather data about files modified between multiple trees.
939
This compares OTHER versus all LCA trees, and for interesting entries,
940
it then compares with THIS and BASE.
942
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
943
:return: [(file_id, changed, parents, names, executable)]
944
file_id Simple file_id of the entry
945
changed Boolean, True if the kind or contents changed
947
parents ((base, [parent_id, in, lcas]), parent_id_other,
949
names ((base, [name, in, lcas]), name_in_other, name_in_this)
950
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
952
if self.interesting_files is not None:
953
lookup_trees = [self.this_tree, self.base_tree]
954
lookup_trees.extend(self._lca_trees)
955
# I think we should include the lca trees as well
956
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
959
interesting_ids = self.interesting_ids
961
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
963
base_inventory = self.base_tree.inventory
964
this_inventory = self.this_tree.inventory
965
for path, file_id, other_ie, lca_values in walker.iter_all():
966
# Is this modified at all from any of the other trees?
968
other_ie = _none_entry
969
if interesting_ids is not None and file_id not in interesting_ids:
972
# If other_revision is found in any of the lcas, that means this
973
# node is uninteresting. This is because when merging, if there are
974
# multiple heads(), we have to create a new node. So if we didn't,
975
# we know that the ancestry is linear, and that OTHER did not
977
# See doc/developers/lca_merge_resolution.txt for details
978
other_revision = other_ie.revision
979
if other_revision is not None:
980
# We can't use this shortcut when other_revision is None,
981
# because it may be None because things are WorkingTrees, and
982
# not because it is *actually* None.
983
is_unmodified = False
984
for lca_path, ie in lca_values:
985
if ie is not None and ie.revision == other_revision:
992
for lca_path, lca_ie in lca_values:
994
lca_entries.append(_none_entry)
996
lca_entries.append(lca_ie)
998
if file_id in base_inventory:
999
base_ie = base_inventory[file_id]
1001
base_ie = _none_entry
1003
if file_id in this_inventory:
1004
this_ie = this_inventory[file_id]
1006
this_ie = _none_entry
1012
for lca_ie in lca_entries:
1013
lca_kinds.append(lca_ie.kind)
1014
lca_parent_ids.append(lca_ie.parent_id)
1015
lca_names.append(lca_ie.name)
1016
lca_executable.append(lca_ie.executable)
1018
kind_winner = self._lca_multi_way(
1019
(base_ie.kind, lca_kinds),
1020
other_ie.kind, this_ie.kind)
1021
parent_id_winner = self._lca_multi_way(
1022
(base_ie.parent_id, lca_parent_ids),
1023
other_ie.parent_id, this_ie.parent_id)
1024
name_winner = self._lca_multi_way(
1025
(base_ie.name, lca_names),
1026
other_ie.name, this_ie.name)
1028
content_changed = True
1029
if kind_winner == 'this':
1030
# No kind change in OTHER, see if there are *any* changes
1031
if other_ie.kind == 'directory':
1032
if parent_id_winner == 'this' and name_winner == 'this':
1033
# No change for this directory in OTHER, skip
1035
content_changed = False
1036
elif other_ie.kind is None or other_ie.kind == 'file':
1037
def get_sha1(ie, tree):
1038
if ie.kind != 'file':
1040
return tree.get_file_sha1(file_id)
1041
base_sha1 = get_sha1(base_ie, self.base_tree)
1042
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1043
in zip(lca_entries, self._lca_trees)]
1044
this_sha1 = get_sha1(this_ie, self.this_tree)
1045
other_sha1 = get_sha1(other_ie, self.other_tree)
1046
sha1_winner = self._lca_multi_way(
1047
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1048
allow_overriding_lca=False)
1049
exec_winner = self._lca_multi_way(
1050
(base_ie.executable, lca_executable),
1051
other_ie.executable, this_ie.executable)
1052
if (parent_id_winner == 'this' and name_winner == 'this'
1053
and sha1_winner == 'this' and exec_winner == 'this'):
1054
# No kind, parent, name, exec, or content change for
1055
# OTHER, so this node is not considered interesting
1057
if sha1_winner == 'this':
1058
content_changed = False
1059
elif other_ie.kind == 'symlink':
1060
def get_target(ie, tree):
1061
if ie.kind != 'symlink':
1063
return tree.get_symlink_target(file_id)
1064
base_target = get_target(base_ie, self.base_tree)
1065
lca_targets = [get_target(ie, tree) for ie, tree
1066
in zip(lca_entries, self._lca_trees)]
1067
this_target = get_target(this_ie, self.this_tree)
1068
other_target = get_target(other_ie, self.other_tree)
1069
target_winner = self._lca_multi_way(
1070
(base_target, lca_targets),
1071
other_target, this_target)
1072
if (parent_id_winner == 'this' and name_winner == 'this'
1073
and target_winner == 'this'):
1074
# No kind, parent, name, or symlink target change
1077
if target_winner == 'this':
1078
content_changed = False
1079
elif other_ie.kind == 'tree-reference':
1080
# The 'changed' information seems to be handled at a higher
1081
# level. At least, _entries3 returns False for content
1082
# changed, even when at a new revision_id.
1083
content_changed = False
1084
if (parent_id_winner == 'this' and name_winner == 'this'):
1085
# Nothing interesting
1088
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1090
# If we have gotten this far, that means something has changed
1091
result.append((file_id, content_changed,
1092
((base_ie.parent_id, lca_parent_ids),
1093
other_ie.parent_id, this_ie.parent_id),
1094
((base_ie.name, lca_names),
1095
other_ie.name, this_ie.name),
1096
((base_ie.executable, lca_executable),
1097
other_ie.executable, this_ie.executable)
1102
if self.tt.final_kind(self.tt.root) is None:
1103
self.tt.cancel_deletion(self.tt.root)
1104
if self.tt.final_file_id(self.tt.root) is None:
1105
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
1107
other_root_file_id = self.other_tree.get_root_id()
1108
if other_root_file_id is None:
1110
other_root = self.tt.trans_id_file_id(other_root_file_id)
1111
if other_root == self.tt.root:
1113
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
1114
# the other tree's root is a non-root in the current tree (as when
1115
# a previously unrelated branch is merged into another)
1117
if self.tt.final_kind(other_root) is not None:
1118
other_root_is_present = True
1120
# other_root doesn't have a physical representation. We still need
1121
# to move any references to the actual root of the tree.
1122
other_root_is_present = False
1123
# 'other_tree.inventory.root' is not present in this tree. We are
1124
# calling adjust_path for children which *want* to be present with a
1125
# correct place to go.
1126
for _, child in self.other_tree.inventory.root.children.iteritems():
1127
trans_id = self.tt.trans_id_file_id(child.file_id)
1128
if not other_root_is_present:
1129
if self.tt.final_kind(trans_id) is not None:
1130
# The item exist in the final tree and has a defined place
1133
# Move the item into the root
1135
final_name = self.tt.final_name(trans_id)
1136
except errors.NoFinalPath:
1137
# This file is not present anymore, ignore it.
1139
self.tt.adjust_path(final_name, self.tt.root, trans_id)
1140
if other_root_is_present:
1141
self.tt.cancel_creation(other_root)
1142
self.tt.cancel_versioning(other_root)
1144
def write_modified(self, results):
1145
modified_hashes = {}
1146
for path in results.modified_paths:
1147
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
1150
hash = self.this_tree.get_file_sha1(file_id)
1153
modified_hashes[file_id] = hash
1154
self.this_tree.set_merge_modified(modified_hashes)
1157
def parent(entry, file_id):
1158
"""Determine the parent for a file_id (used as a key method)"""
1161
return entry.parent_id
1164
def name(entry, file_id):
1165
"""Determine the name for a file_id (used as a key method)"""
1171
def contents_sha1(tree, file_id):
1172
"""Determine the sha1 of the file contents (used as a key method)."""
1173
if file_id not in tree:
1175
return tree.get_file_sha1(file_id)
1178
def executable(tree, file_id):
1179
"""Determine the executability of a file-id (used as a key method)."""
1180
if not tree.has_id(file_id):
1182
if tree.kind(file_id) != "file":
1184
return tree.is_executable(file_id)
1187
def kind(tree, file_id):
1188
"""Determine the kind of a file-id (used as a key method)."""
1189
if not tree.has_id(file_id):
1191
return tree.kind(file_id)
1194
def _three_way(base, other, this):
1196
# if 'base == other', either they all agree, or only 'this' has
1199
elif this not in (base, other):
1200
# 'this' is neither 'base' nor 'other', so both sides changed
1203
# "Ambiguous clean merge" -- both sides have made the same change.
1206
# this == base: only other has changed.
1210
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
1211
"""Consider LCAs when determining whether a change has occurred.
1213
If LCAS are all identical, this is the same as a _three_way comparison.
1215
:param bases: value in (BASE, [LCAS])
1216
:param other: value in OTHER
1217
:param this: value in THIS
1218
:param allow_overriding_lca: If there is more than one unique lca
1219
value, allow OTHER to override THIS if it has a new value, and
1220
THIS only has an lca value, or vice versa. This is appropriate for
1221
truly scalar values, not as much for non-scalars.
1222
:return: 'this', 'other', or 'conflict' depending on whether an entry
1225
# See doc/developers/lca_tree_merging.txt for details about this
1228
# Either Ambiguously clean, or nothing was actually changed. We
1231
base_val, lca_vals = bases
1232
# Remove 'base_val' from the lca_vals, because it is not interesting
1233
filtered_lca_vals = [lca_val for lca_val in lca_vals
1234
if lca_val != base_val]
1235
if len(filtered_lca_vals) == 0:
1236
return Merge3Merger._three_way(base_val, other, this)
1238
unique_lca_vals = set(filtered_lca_vals)
1239
if len(unique_lca_vals) == 1:
1240
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1242
if allow_overriding_lca:
1243
if other in unique_lca_vals:
1244
if this in unique_lca_vals:
1245
# Each side picked a different lca, conflict
1248
# This has a value which supersedes both lca values, and
1249
# other only has an lca value
1251
elif this in unique_lca_vals:
1252
# OTHER has a value which supersedes both lca values, and this
1253
# only has an lca value
1256
# At this point, the lcas disagree, and the tip disagree
1260
@deprecated_method(deprecated_in((2, 2, 0)))
1261
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1262
"""Do a three-way test on a scalar.
1263
Return "this", "other" or "conflict", depending whether a value wins.
1265
key_base = key(base_tree, file_id)
1266
key_other = key(other_tree, file_id)
1267
#if base == other, either they all agree, or only THIS has changed.
1268
if key_base == key_other:
1270
key_this = key(this_tree, file_id)
1271
# "Ambiguous clean merge"
1272
if key_this == key_other:
1274
elif key_this == key_base:
1279
def merge_names(self, file_id):
1280
def get_entry(tree):
1281
if tree.has_id(file_id):
1282
return tree.inventory[file_id]
1285
this_entry = get_entry(self.this_tree)
1286
other_entry = get_entry(self.other_tree)
1287
base_entry = get_entry(self.base_tree)
1288
entries = (base_entry, other_entry, this_entry)
1291
for entry in entries:
1294
parents.append(None)
1296
names.append(entry.name)
1297
parents.append(entry.parent_id)
1298
return self._merge_names(file_id, parents, names,
1299
resolver=self._three_way)
1301
def _merge_names(self, file_id, parents, names, resolver):
1302
"""Perform a merge on file_id names and parents"""
1303
base_name, other_name, this_name = names
1304
base_parent, other_parent, this_parent = parents
1306
name_winner = resolver(*names)
1308
parent_id_winner = resolver(*parents)
1309
if this_name is None:
1310
if name_winner == "this":
1311
name_winner = "other"
1312
if parent_id_winner == "this":
1313
parent_id_winner = "other"
1314
if name_winner == "this" and parent_id_winner == "this":
1316
if name_winner == 'conflict' or parent_id_winner == 'conflict':
1317
# Creating helpers (.OTHER or .THIS) here cause problems down the
1318
# road if a ContentConflict needs to be created so we should not do
1320
trans_id = self.tt.trans_id_file_id(file_id)
1321
self._raw_conflicts.append(('path conflict', trans_id, file_id,
1322
this_parent, this_name,
1323
other_parent, other_name))
1324
if other_name is None:
1325
# it doesn't matter whether the result was 'other' or
1326
# 'conflict'-- if there's no 'other', we leave it alone.
1328
parent_id = parents[self.winner_idx[parent_id_winner]]
1329
if parent_id is not None:
1330
# if we get here, name_winner and parent_winner are set to safe
1332
self.tt.adjust_path(names[self.winner_idx[name_winner]],
1333
self.tt.trans_id_file_id(parent_id),
1334
self.tt.trans_id_file_id(file_id))
1336
def _do_merge_contents(self, file_id):
1337
"""Performs a merge on file_id contents."""
1338
def contents_pair(tree):
1339
if file_id not in tree:
1341
kind = tree.kind(file_id)
1343
contents = tree.get_file_sha1(file_id)
1344
elif kind == "symlink":
1345
contents = tree.get_symlink_target(file_id)
1348
return kind, contents
1350
# See SPOT run. run, SPOT, run.
1351
# So we're not QUITE repeating ourselves; we do tricky things with
1353
base_pair = contents_pair(self.base_tree)
1354
other_pair = contents_pair(self.other_tree)
1356
this_pair = contents_pair(self.this_tree)
1357
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1358
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1359
this_pair, allow_overriding_lca=False)
1361
if base_pair == other_pair:
1364
# We delayed evaluating this_pair as long as we can to avoid
1365
# unnecessary sha1 calculation
1366
this_pair = contents_pair(self.this_tree)
1367
winner = self._three_way(base_pair, other_pair, this_pair)
1368
if winner == 'this':
1369
# No interesting changes introduced by OTHER
1371
# We have a hypothetical conflict, but if we have files, then we
1372
# can try to merge the content
1373
trans_id = self.tt.trans_id_file_id(file_id)
1374
params = MergeHookParams(self, file_id, trans_id, this_pair[0],
1375
other_pair[0], winner)
1376
hooks = self.active_hooks
1377
hook_status = 'not_applicable'
1379
hook_status, lines = hook.merge_contents(params)
1380
if hook_status != 'not_applicable':
1381
# Don't try any more hooks, this one applies.
1384
if hook_status == 'not_applicable':
1385
# This is a contents conflict, because none of the available
1386
# functions could merge it.
1388
name = self.tt.final_name(trans_id)
1389
parent_id = self.tt.final_parent(trans_id)
1390
if self.this_tree.has_id(file_id):
1391
self.tt.unversion_file(trans_id)
1392
file_group = self._dump_conflicts(name, parent_id, file_id,
1394
self._raw_conflicts.append(('contents conflict', file_group))
1395
elif hook_status == 'success':
1396
self.tt.create_file(lines, trans_id)
1397
elif hook_status == 'conflicted':
1398
# XXX: perhaps the hook should be able to provide
1399
# the BASE/THIS/OTHER files?
1400
self.tt.create_file(lines, trans_id)
1401
self._raw_conflicts.append(('text conflict', trans_id))
1402
name = self.tt.final_name(trans_id)
1403
parent_id = self.tt.final_parent(trans_id)
1404
self._dump_conflicts(name, parent_id, file_id)
1405
elif hook_status == 'delete':
1406
self.tt.unversion_file(trans_id)
1408
elif hook_status == 'done':
1409
# The hook function did whatever it needs to do directly, no
1410
# further action needed here.
1413
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1414
if not self.this_tree.has_id(file_id) and result == "modified":
1415
self.tt.version_file(file_id, trans_id)
1416
# The merge has been performed, so the old contents should not be
1418
self.tt.delete_contents(trans_id)
1421
def _default_other_winner_merge(self, merge_hook_params):
1422
"""Replace this contents with other."""
1423
file_id = merge_hook_params.file_id
1424
trans_id = merge_hook_params.trans_id
1425
file_in_this = self.this_tree.has_id(file_id)
1426
if self.other_tree.has_id(file_id):
1427
# OTHER changed the file
1429
if wt.supports_content_filtering():
1430
# We get the path from the working tree if it exists.
1431
# That fails though when OTHER is adding a file, so
1432
# we fall back to the other tree to find the path if
1433
# it doesn't exist locally.
1435
filter_tree_path = wt.id2path(file_id)
1436
except errors.NoSuchId:
1437
filter_tree_path = self.other_tree.id2path(file_id)
1439
# Skip the id2path lookup for older formats
1440
filter_tree_path = None
1441
transform.create_from_tree(self.tt, trans_id,
1442
self.other_tree, file_id,
1443
filter_tree_path=filter_tree_path)
1446
# OTHER deleted the file
1447
return 'delete', None
1449
raise AssertionError(
1450
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1453
def merge_contents(self, merge_hook_params):
1454
"""Fallback merge logic after user installed hooks."""
1455
# This function is used in merge hooks as the fallback instance.
1456
# Perhaps making this function and the functions it calls be a
1457
# a separate class would be better.
1458
if merge_hook_params.winner == 'other':
1459
# OTHER is a straight winner, so replace this contents with other
1460
return self._default_other_winner_merge(merge_hook_params)
1461
elif merge_hook_params.is_file_merge():
1462
# THIS and OTHER are both files, so text merge. Either
1463
# BASE is a file, or both converted to files, so at least we
1464
# have agreement that output should be a file.
1466
self.text_merge(merge_hook_params.file_id,
1467
merge_hook_params.trans_id)
1468
except errors.BinaryFile:
1469
return 'not_applicable', None
1472
return 'not_applicable', None
1474
def get_lines(self, tree, file_id):
1475
"""Return the lines in a file, or an empty list."""
1476
if tree.has_id(file_id):
1477
return tree.get_file_lines(file_id)
1481
def text_merge(self, file_id, trans_id):
1482
"""Perform a three-way text merge on a file_id"""
1483
# it's possible that we got here with base as a different type.
1484
# if so, we just want two-way text conflicts.
1485
if self.base_tree.has_id(file_id) and \
1486
self.base_tree.kind(file_id) == "file":
1487
base_lines = self.get_lines(self.base_tree, file_id)
1490
other_lines = self.get_lines(self.other_tree, file_id)
1491
this_lines = self.get_lines(self.this_tree, file_id)
1492
m3 = merge3.Merge3(base_lines, this_lines, other_lines,
1493
is_cherrypick=self.cherrypick)
1494
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1495
if self.show_base is True:
1496
base_marker = '|' * 7
1500
def iter_merge3(retval):
1501
retval["text_conflicts"] = False
1502
for line in m3.merge_lines(name_a = "TREE",
1503
name_b = "MERGE-SOURCE",
1504
name_base = "BASE-REVISION",
1505
start_marker=start_marker,
1506
base_marker=base_marker,
1507
reprocess=self.reprocess):
1508
if line.startswith(start_marker):
1509
retval["text_conflicts"] = True
1510
yield line.replace(start_marker, '<' * 7)
1514
merge3_iterator = iter_merge3(retval)
1515
self.tt.create_file(merge3_iterator, trans_id)
1516
if retval["text_conflicts"] is True:
1517
self._raw_conflicts.append(('text conflict', trans_id))
1518
name = self.tt.final_name(trans_id)
1519
parent_id = self.tt.final_parent(trans_id)
1520
file_group = self._dump_conflicts(name, parent_id, file_id,
1521
this_lines, base_lines,
1523
file_group.append(trans_id)
1525
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1526
base_lines=None, other_lines=None, set_version=False,
1528
"""Emit conflict files.
1529
If this_lines, base_lines, or other_lines are omitted, they will be
1530
determined automatically. If set_version is true, the .OTHER, .THIS
1531
or .BASE (in that order) will be created as versioned files.
1533
data = [('OTHER', self.other_tree, other_lines),
1534
('THIS', self.this_tree, this_lines)]
1536
data.append(('BASE', self.base_tree, base_lines))
1538
# We need to use the actual path in the working tree of the file here,
1539
# ignoring the conflict suffixes
1541
if wt.supports_content_filtering():
1543
filter_tree_path = wt.id2path(file_id)
1544
except errors.NoSuchId:
1545
# file has been deleted
1546
filter_tree_path = None
1548
# Skip the id2path lookup for older formats
1549
filter_tree_path = None
1553
for suffix, tree, lines in data:
1554
if tree.has_id(file_id):
1555
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1556
suffix, lines, filter_tree_path)
1557
file_group.append(trans_id)
1558
if set_version and not versioned:
1559
self.tt.version_file(file_id, trans_id)
1563
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1564
lines=None, filter_tree_path=None):
1565
"""Emit a single conflict file."""
1566
name = name + '.' + suffix
1567
trans_id = self.tt.create_path(name, parent_id)
1568
transform.create_from_tree(self.tt, trans_id, tree, file_id, lines,
1572
def merge_executable(self, file_id, file_status):
1573
"""Perform a merge on the execute bit."""
1574
executable = [self.executable(t, file_id) for t in (self.base_tree,
1575
self.other_tree, self.this_tree)]
1576
self._merge_executable(file_id, executable, file_status,
1577
resolver=self._three_way)
1579
def _merge_executable(self, file_id, executable, file_status,
1581
"""Perform a merge on the execute bit."""
1582
base_executable, other_executable, this_executable = executable
1583
if file_status == "deleted":
1585
winner = resolver(*executable)
1586
if winner == "conflict":
1587
# There must be a None in here, if we have a conflict, but we
1588
# need executability since file status was not deleted.
1589
if self.executable(self.other_tree, file_id) is None:
1593
if winner == 'this' and file_status != "modified":
1595
trans_id = self.tt.trans_id_file_id(file_id)
1596
if self.tt.final_kind(trans_id) != "file":
1598
if winner == "this":
1599
executability = this_executable
1601
if self.other_tree.has_id(file_id):
1602
executability = other_executable
1603
elif self.this_tree.has_id(file_id):
1604
executability = this_executable
1605
elif self.base_tree_has_id(file_id):
1606
executability = base_executable
1607
if executability is not None:
1608
trans_id = self.tt.trans_id_file_id(file_id)
1609
self.tt.set_executability(executability, trans_id)
1611
def cook_conflicts(self, fs_conflicts):
1612
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1613
self.cooked_conflicts.extend(transform.cook_conflicts(
1614
fs_conflicts, self.tt))
1615
fp = transform.FinalPaths(self.tt)
1616
for conflict in self._raw_conflicts:
1617
conflict_type = conflict[0]
1618
if conflict_type == 'path conflict':
1620
this_parent, this_name,
1621
other_parent, other_name) = conflict[1:]
1622
if this_parent is None or this_name is None:
1623
this_path = '<deleted>'
1625
parent_path = fp.get_path(
1626
self.tt.trans_id_file_id(this_parent))
1627
this_path = osutils.pathjoin(parent_path, this_name)
1628
if other_parent is None or other_name is None:
1629
other_path = '<deleted>'
1631
parent_path = fp.get_path(
1632
self.tt.trans_id_file_id(other_parent))
1633
other_path = osutils.pathjoin(parent_path, other_name)
1634
c = _mod_conflicts.Conflict.factory(
1635
'path conflict', path=this_path,
1636
conflict_path=other_path,
1638
elif conflict_type == 'contents conflict':
1639
for trans_id in conflict[1]:
1640
file_id = self.tt.final_file_id(trans_id)
1641
if file_id is not None:
1643
path = fp.get_path(trans_id)
1644
for suffix in ('.BASE', '.THIS', '.OTHER'):
1645
if path.endswith(suffix):
1646
path = path[:-len(suffix)]
1648
c = _mod_conflicts.Conflict.factory(conflict_type,
1649
path=path, file_id=file_id)
1650
elif conflict_type == 'text conflict':
1651
trans_id = conflict[1]
1652
path = fp.get_path(trans_id)
1653
file_id = self.tt.final_file_id(trans_id)
1654
c = _mod_conflicts.Conflict.factory(conflict_type,
1655
path=path, file_id=file_id)
1657
raise AssertionError('bad conflict type: %r' % (conflict,))
1658
self.cooked_conflicts.append(c)
1659
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
1662
class WeaveMerger(Merge3Merger):
1663
"""Three-way tree merger, text weave merger."""
1664
supports_reprocess = True
1665
supports_show_base = False
1666
supports_reverse_cherrypick = False
1667
history_based = True
1669
def _generate_merge_plan(self, file_id, base):
1670
return self.this_tree.plan_file_merge(file_id, self.other_tree,
1673
def _merged_lines(self, file_id):
1674
"""Generate the merged lines.
1675
There is no distinction between lines that are meant to contain <<<<<<<
1679
base = self.base_tree
1682
plan = self._generate_merge_plan(file_id, base)
1683
if 'merge' in debug.debug_flags:
1685
trans_id = self.tt.trans_id_file_id(file_id)
1686
name = self.tt.final_name(trans_id) + '.plan'
1687
contents = ('%11s|%s' % l for l in plan)
1688
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1689
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1690
'>>>>>>> MERGE-SOURCE\n')
1691
lines, conflicts = textmerge.merge_lines(self.reprocess)
1693
base_lines = textmerge.base_from_plan()
1696
return lines, base_lines
1698
def text_merge(self, file_id, trans_id):
1699
"""Perform a (weave) text merge for a given file and file-id.
1700
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1701
and a conflict will be noted.
1703
lines, base_lines = self._merged_lines(file_id)
1705
# Note we're checking whether the OUTPUT is binary in this case,
1706
# because we don't want to get into weave merge guts.
1707
textfile.check_text_lines(lines)
1708
self.tt.create_file(lines, trans_id)
1709
if base_lines is not None:
1711
self._raw_conflicts.append(('text conflict', trans_id))
1712
name = self.tt.final_name(trans_id)
1713
parent_id = self.tt.final_parent(trans_id)
1714
file_group = self._dump_conflicts(name, parent_id, file_id,
1716
base_lines=base_lines)
1717
file_group.append(trans_id)
1720
class LCAMerger(WeaveMerger):
1722
def _generate_merge_plan(self, file_id, base):
1723
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1726
class Diff3Merger(Merge3Merger):
1727
"""Three-way merger using external diff3 for text merging"""
1729
def dump_file(self, temp_dir, name, tree, file_id):
1730
out_path = osutils.pathjoin(temp_dir, name)
1731
out_file = open(out_path, "wb")
1733
in_file = tree.get_file(file_id)
1734
for line in in_file:
1735
out_file.write(line)
1740
def text_merge(self, file_id, trans_id):
1741
"""Perform a diff3 merge using a specified file-id and trans-id.
1742
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1743
will be dumped, and a will be conflict noted.
1746
temp_dir = osutils.mkdtemp(prefix="bzr-")
1748
new_file = osutils.pathjoin(temp_dir, "new")
1749
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1750
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1751
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1752
status = bzrlib.patch.diff3(new_file, this, base, other)
1753
if status not in (0, 1):
1754
raise errors.BzrError("Unhandled diff3 exit code")
1755
f = open(new_file, 'rb')
1757
self.tt.create_file(f, trans_id)
1761
name = self.tt.final_name(trans_id)
1762
parent_id = self.tt.final_parent(trans_id)
1763
self._dump_conflicts(name, parent_id, file_id)
1764
self._raw_conflicts.append(('text conflict', trans_id))
1766
osutils.rmtree(temp_dir)
1769
class PathNotInTree(errors.BzrError):
1771
_fmt = """Merge-into failed because %(tree)s does not contain %(path)s."""
1773
def __init__(self, path, tree):
1774
errors.BzrError.__init__(self, path=path, tree=tree)
1777
class MergeIntoMerger(Merger):
1778
"""Merger that understands other_tree will be merged into a subdir.
1780
This also changes the Merger api so that it uses real Branch, revision_id,
1781
and RevisonTree objects, rather than using revision specs.
1784
def __init__(self, this_tree, other_branch, other_tree, target_subdir,
1785
source_subpath, other_rev_id=None):
1786
"""Create a new MergeIntoMerger object.
1788
source_subpath in other_tree will be effectively copied to
1789
target_subdir in this_tree.
1791
:param this_tree: The tree that we will be merging into.
1792
:param other_branch: The Branch we will be merging from.
1793
:param other_tree: The RevisionTree object we want to merge.
1794
:param target_subdir: The relative path where we want to merge
1795
other_tree into this_tree
1796
:param source_subpath: The relative path specifying the subtree of
1797
other_tree to merge into this_tree.
1799
# It is assumed that we are merging a tree that is not in our current
1800
# ancestry, which means we are using the "EmptyTree" as our basis.
1801
null_ancestor_tree = this_tree.branch.repository.revision_tree(
1802
_mod_revision.NULL_REVISION)
1803
super(MergeIntoMerger, self).__init__(
1804
this_branch=this_tree.branch,
1805
this_tree=this_tree,
1806
other_tree=other_tree,
1807
base_tree=null_ancestor_tree,
1809
self._target_subdir = target_subdir
1810
self._source_subpath = source_subpath
1811
self.other_branch = other_branch
1812
if other_rev_id is None:
1813
other_rev_id = other_tree.get_revision_id()
1814
self.other_rev_id = self.other_basis = other_rev_id
1815
self.base_is_ancestor = True
1816
self.backup_files = True
1817
self.merge_type = Merge3Merger
1818
self.show_base = False
1819
self.reprocess = False
1820
self.interesting_ids = None
1821
self.merge_type = _MergeTypeParameterizer(MergeIntoMergeType,
1822
target_subdir=self._target_subdir,
1823
source_subpath=self._source_subpath)
1824
if self._source_subpath != '':
1825
# If this isn't a partial merge make sure the revisions will be
1827
self._maybe_fetch(self.other_branch, self.this_branch,
1830
def set_pending(self):
1831
if self._source_subpath != '':
1833
Merger.set_pending(self)
1836
class _MergeTypeParameterizer(object):
1837
"""Wrap a merge-type class to provide extra parameters.
1839
This is hack used by MergeIntoMerger to pass some extra parameters to its
1840
merge_type. Merger.do_merge() sets up its own set of parameters to pass to
1841
the 'merge_type' member. It is difficult override do_merge without
1842
re-writing the whole thing, so instead we create a wrapper which will pass
1843
the extra parameters.
1846
def __init__(self, merge_type, **kwargs):
1847
self._extra_kwargs = kwargs
1848
self._merge_type = merge_type
1850
def __call__(self, *args, **kwargs):
1851
kwargs.update(self._extra_kwargs)
1852
return self._merge_type(*args, **kwargs)
1854
def __getattr__(self, name):
1855
return getattr(self._merge_type, name)
1858
class MergeIntoMergeType(Merge3Merger):
1859
"""Merger that incorporates a tree (or part of a tree) into another."""
1861
def __init__(self, *args, **kwargs):
1862
"""Initialize the merger object.
1864
:param args: See Merge3Merger.__init__'s args.
1865
:param kwargs: See Merge3Merger.__init__'s keyword args, except for
1866
source_subpath and target_subdir.
1867
:keyword source_subpath: The relative path specifying the subtree of
1868
other_tree to merge into this_tree.
1869
:keyword target_subdir: The relative path where we want to merge
1870
other_tree into this_tree
1872
# All of the interesting work happens during Merge3Merger.__init__(),
1873
# so we have have to hack in to get our extra parameters set.
1874
self._source_subpath = kwargs.pop('source_subpath')
1875
self._target_subdir = kwargs.pop('target_subdir')
1876
super(MergeIntoMergeType, self).__init__(*args, **kwargs)
1878
def _compute_transform(self):
1879
child_pb = ui.ui_factory.nested_progress_bar()
1881
entries = self._entries_to_incorporate()
1882
entries = list(entries)
1883
for num, (entry, parent_id) in enumerate(entries):
1884
child_pb.update('Preparing file merge', num, len(entries))
1885
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1886
trans_id = transform.new_by_entry(self.tt, entry,
1887
parent_trans_id, self.other_tree)
1890
self._finish_computing_transform()
1892
def _entries_to_incorporate(self):
1893
"""Yields pairs of (inventory_entry, new_parent)."""
1894
other_inv = self.other_tree.inventory
1895
subdir_id = other_inv.path2id(self._source_subpath)
1896
if subdir_id is None:
1897
# XXX: The error would be clearer if it gave the URL of the source
1898
# branch, but we don't have a reference to that here.
1899
raise PathNotInTree(self._source_subpath, "Source tree")
1900
subdir = other_inv[subdir_id]
1901
parent_in_target = osutils.dirname(self._target_subdir)
1902
target_id = self.this_tree.inventory.path2id(parent_in_target)
1903
if target_id is None:
1904
raise PathNotInTree(self._target_subdir, "Target tree")
1905
name_in_target = osutils.basename(self._target_subdir)
1906
merge_into_root = subdir.copy()
1907
merge_into_root.name = name_in_target
1908
if merge_into_root.file_id in self.this_tree.inventory:
1909
# Give the root a new file-id.
1910
# This can happen fairly easily if the directory we are
1911
# incorporating is the root, and both trees have 'TREE_ROOT' as
1912
# their root_id. Users will expect this to Just Work, so we
1913
# change the file-id here.
1914
# Non-root file-ids could potentially conflict too. That's really
1915
# an edge case, so we don't do anything special for those. We let
1916
# them cause conflicts.
1917
merge_into_root.file_id = generate_ids.gen_file_id(name_in_target)
1918
yield (merge_into_root, target_id)
1919
if subdir.kind != 'directory':
1920
# No children, so we are done.
1922
for ignored_path, entry in other_inv.iter_entries_by_dir(subdir_id):
1923
parent_id = entry.parent_id
1924
if parent_id == subdir.file_id:
1925
# The root's parent ID has changed, so make sure children of
1926
# the root refer to the new ID.
1927
parent_id = merge_into_root.file_id
1928
yield (entry, parent_id)
1931
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1933
merge_type=Merge3Merger,
1934
interesting_ids=None,
1938
interesting_files=None,
1941
change_reporter=None):
1942
"""Primary interface for merging.
1944
Typical use is probably::
1946
merge_inner(branch, branch.get_revision_tree(other_revision),
1947
branch.get_revision_tree(base_revision))
1949
if this_tree is None:
1950
raise errors.BzrError("bzrlib.merge.merge_inner requires a this_tree "
1951
"parameter as of bzrlib version 0.8.")
1952
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1953
pb=pb, change_reporter=change_reporter)
1954
merger.backup_files = backup_files
1955
merger.merge_type = merge_type
1956
merger.interesting_ids = interesting_ids
1957
merger.ignore_zero = ignore_zero
1958
if interesting_files:
1960
raise ValueError('Only supply interesting_ids'
1961
' or interesting_files')
1962
merger.interesting_files = interesting_files
1963
merger.show_base = show_base
1964
merger.reprocess = reprocess
1965
merger.other_rev_id = other_rev_id
1966
merger.other_basis = other_rev_id
1967
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1968
if get_revision_id is None:
1969
get_revision_id = base_tree.last_revision
1970
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
1971
merger.set_base_revision(get_revision_id(), this_branch)
1972
return merger.do_merge()
1974
def get_merge_type_registry():
1975
"""Merge type registry is in bzrlib.option to avoid circular imports.
1977
This method provides a sanctioned way to retrieve it.
1979
from bzrlib import option
1980
return option._merge_type_registry
1983
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1984
def status_a(revision, text):
1985
if revision in ancestors_b:
1986
return 'killed-b', text
1988
return 'new-a', text
1990
def status_b(revision, text):
1991
if revision in ancestors_a:
1992
return 'killed-a', text
1994
return 'new-b', text
1996
plain_a = [t for (a, t) in annotated_a]
1997
plain_b = [t for (a, t) in annotated_b]
1998
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1999
blocks = matcher.get_matching_blocks()
2002
for ai, bi, l in blocks:
2003
# process all mismatched sections
2004
# (last mismatched section is handled because blocks always
2005
# includes a 0-length last block)
2006
for revision, text in annotated_a[a_cur:ai]:
2007
yield status_a(revision, text)
2008
for revision, text in annotated_b[b_cur:bi]:
2009
yield status_b(revision, text)
2010
# and now the matched section
2013
for text_a in plain_a[ai:a_cur]:
2014
yield "unchanged", text_a
2017
class _PlanMergeBase(object):
2019
def __init__(self, a_rev, b_rev, vf, key_prefix):
2022
:param a_rev: Revision-id of one revision to merge
2023
:param b_rev: Revision-id of the other revision to merge
2024
:param vf: A VersionedFiles containing both revisions
2025
:param key_prefix: A prefix for accessing keys in vf, typically
2031
self._last_lines = None
2032
self._last_lines_revision_id = None
2033
self._cached_matching_blocks = {}
2034
self._key_prefix = key_prefix
2035
self._precache_tip_lines()
2037
def _precache_tip_lines(self):
2038
lines = self.get_lines([self.a_rev, self.b_rev])
2039
self.lines_a = lines[self.a_rev]
2040
self.lines_b = lines[self.b_rev]
2042
def get_lines(self, revisions):
2043
"""Get lines for revisions from the backing VersionedFiles.
2045
:raises RevisionNotPresent: on absent texts.
2047
keys = [(self._key_prefix + (rev,)) for rev in revisions]
2049
for record in self.vf.get_record_stream(keys, 'unordered', True):
2050
if record.storage_kind == 'absent':
2051
raise errors.RevisionNotPresent(record.key, self.vf)
2052
result[record.key[-1]] = osutils.chunks_to_lines(
2053
record.get_bytes_as('chunked'))
2056
def plan_merge(self):
2057
"""Generate a 'plan' for merging the two revisions.
2059
This involves comparing their texts and determining the cause of
2060
differences. If text A has a line and text B does not, then either the
2061
line was added to text A, or it was deleted from B. Once the causes
2062
are combined, they are written out in the format described in
2063
VersionedFile.plan_merge
2065
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
2066
unique_a, unique_b = self._unique_lines(blocks)
2067
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
2068
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
2069
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
2071
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
2074
for i, j, n in blocks:
2075
for a_index in range(last_i, i):
2076
if a_index in new_a:
2077
if a_index in killed_b:
2078
yield 'conflicted-a', self.lines_a[a_index]
2080
yield 'new-a', self.lines_a[a_index]
2082
yield 'killed-b', self.lines_a[a_index]
2083
for b_index in range(last_j, j):
2084
if b_index in new_b:
2085
if b_index in killed_a:
2086
yield 'conflicted-b', self.lines_b[b_index]
2088
yield 'new-b', self.lines_b[b_index]
2090
yield 'killed-a', self.lines_b[b_index]
2091
# handle common lines
2092
for a_index in range(i, i+n):
2093
yield 'unchanged', self.lines_a[a_index]
2097
def _get_matching_blocks(self, left_revision, right_revision):
2098
"""Return a description of which sections of two revisions match.
2100
See SequenceMatcher.get_matching_blocks
2102
cached = self._cached_matching_blocks.get((left_revision,
2104
if cached is not None:
2106
if self._last_lines_revision_id == left_revision:
2107
left_lines = self._last_lines
2108
right_lines = self.get_lines([right_revision])[right_revision]
2110
lines = self.get_lines([left_revision, right_revision])
2111
left_lines = lines[left_revision]
2112
right_lines = lines[right_revision]
2113
self._last_lines = right_lines
2114
self._last_lines_revision_id = right_revision
2115
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
2117
return matcher.get_matching_blocks()
2119
def _unique_lines(self, matching_blocks):
2120
"""Analyse matching_blocks to determine which lines are unique
2122
:return: a tuple of (unique_left, unique_right), where the values are
2123
sets of line numbers of unique lines.
2129
for i, j, n in matching_blocks:
2130
unique_left.extend(range(last_i, i))
2131
unique_right.extend(range(last_j, j))
2134
return unique_left, unique_right
2137
def _subtract_plans(old_plan, new_plan):
2138
"""Remove changes from new_plan that came from old_plan.
2140
It is assumed that the difference between the old_plan and new_plan
2141
is their choice of 'b' text.
2143
All lines from new_plan that differ from old_plan are emitted
2144
verbatim. All lines from new_plan that match old_plan but are
2145
not about the 'b' revision are emitted verbatim.
2147
Lines that match and are about the 'b' revision are the lines we
2148
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
2149
is skipped entirely.
2151
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
2154
for i, j, n in matcher.get_matching_blocks():
2155
for jj in range(last_j, j):
2157
for jj in range(j, j+n):
2158
plan_line = new_plan[jj]
2159
if plan_line[0] == 'new-b':
2161
elif plan_line[0] == 'killed-b':
2162
yield 'unchanged', plan_line[1]
2168
class _PlanMerge(_PlanMergeBase):
2169
"""Plan an annotate merge using on-the-fly annotation"""
2171
def __init__(self, a_rev, b_rev, vf, key_prefix):
2172
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
2173
self.a_key = self._key_prefix + (self.a_rev,)
2174
self.b_key = self._key_prefix + (self.b_rev,)
2175
self.graph = _mod_graph.Graph(self.vf)
2176
heads = self.graph.heads((self.a_key, self.b_key))
2178
# one side dominates, so we can just return its values, yay for
2180
# Ideally we would know that before we get this far
2181
self._head_key = heads.pop()
2182
if self._head_key == self.a_key:
2186
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2187
self._head_key[-1], other)
2190
self._head_key = None
2193
def _precache_tip_lines(self):
2194
# Turn this into a no-op, because we will do this later
2197
def _find_recursive_lcas(self):
2198
"""Find all the ancestors back to a unique lca"""
2199
cur_ancestors = (self.a_key, self.b_key)
2200
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2201
# rather than a key tuple. We will just map that directly to no common
2205
next_lcas = self.graph.find_lca(*cur_ancestors)
2206
# Map a plain NULL_REVISION to a simple no-ancestors
2207
if next_lcas == set([_mod_revision.NULL_REVISION]):
2209
# Order the lca's based on when they were merged into the tip
2210
# While the actual merge portion of weave merge uses a set() of
2211
# active revisions, the order of insertion *does* effect the
2212
# implicit ordering of the texts.
2213
for rev_key in cur_ancestors:
2214
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2216
parent_map[rev_key] = ordered_parents
2217
if len(next_lcas) == 0:
2219
elif len(next_lcas) == 1:
2220
parent_map[list(next_lcas)[0]] = ()
2222
elif len(next_lcas) > 2:
2223
# More than 2 lca's, fall back to grabbing all nodes between
2224
# this and the unique lca.
2225
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2227
self.a_key, self.b_key, cur_ancestors)
2228
cur_lcas = next_lcas
2229
while len(cur_lcas) > 1:
2230
cur_lcas = self.graph.find_lca(*cur_lcas)
2231
if len(cur_lcas) == 0:
2232
# No common base to find, use the full ancestry
2235
unique_lca = list(cur_lcas)[0]
2236
if unique_lca == _mod_revision.NULL_REVISION:
2237
# find_lca will return a plain 'NULL_REVISION' rather
2238
# than a key tuple when there is no common ancestor, we
2239
# prefer to just use None, because it doesn't confuse
2240
# _get_interesting_texts()
2242
parent_map.update(self._find_unique_parents(next_lcas,
2245
cur_ancestors = next_lcas
2248
def _find_unique_parents(self, tip_keys, base_key):
2249
"""Find ancestors of tip that aren't ancestors of base.
2251
:param tip_keys: Nodes that are interesting
2252
:param base_key: Cull all ancestors of this node
2253
:return: The parent map for all revisions between tip_keys and
2254
base_key. base_key will be included. References to nodes outside of
2255
the ancestor set will also be removed.
2257
# TODO: this would be simpler if find_unique_ancestors took a list
2258
# instead of a single tip, internally it supports it, but it
2259
# isn't a "backwards compatible" api change.
2260
if base_key is None:
2261
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2262
# We remove NULL_REVISION because it isn't a proper tuple key, and
2263
# thus confuses things like _get_interesting_texts, and our logic
2264
# to add the texts into the memory weave.
2265
if _mod_revision.NULL_REVISION in parent_map:
2266
parent_map.pop(_mod_revision.NULL_REVISION)
2269
for tip in tip_keys:
2271
self.graph.find_unique_ancestors(tip, [base_key]))
2272
parent_map = self.graph.get_parent_map(interesting)
2273
parent_map[base_key] = ()
2274
culled_parent_map, child_map, tails = self._remove_external_references(
2276
# Remove all the tails but base_key
2277
if base_key is not None:
2278
tails.remove(base_key)
2279
self._prune_tails(culled_parent_map, child_map, tails)
2280
# Now remove all the uninteresting 'linear' regions
2281
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2285
def _remove_external_references(parent_map):
2286
"""Remove references that go outside of the parent map.
2288
:param parent_map: Something returned from Graph.get_parent_map(keys)
2289
:return: (filtered_parent_map, child_map, tails)
2290
filtered_parent_map is parent_map without external references
2291
child_map is the {parent_key: [child_keys]} mapping
2292
tails is a list of nodes that do not have any parents in the map
2294
# TODO: The basic effect of this function seems more generic than
2295
# _PlanMerge. But the specific details of building a child_map,
2296
# and computing tails seems very specific to _PlanMerge.
2297
# Still, should this be in Graph land?
2298
filtered_parent_map = {}
2301
for key, parent_keys in parent_map.iteritems():
2302
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2303
if not culled_parent_keys:
2305
for parent_key in culled_parent_keys:
2306
child_map.setdefault(parent_key, []).append(key)
2307
# TODO: Do we want to do this, it adds overhead for every node,
2308
# just to say that the node has no children
2309
child_map.setdefault(key, [])
2310
filtered_parent_map[key] = culled_parent_keys
2311
return filtered_parent_map, child_map, tails
2314
def _prune_tails(parent_map, child_map, tails_to_remove):
2315
"""Remove tails from the parent map.
2317
This will remove the supplied revisions until no more children have 0
2320
:param parent_map: A dict of {child: [parents]}, this dictionary will
2321
be modified in place.
2322
:param tails_to_remove: A list of tips that should be removed,
2323
this list will be consumed
2324
:param child_map: The reverse dict of parent_map ({parent: [children]})
2325
this dict will be modified
2326
:return: None, parent_map will be modified in place.
2328
while tails_to_remove:
2329
next = tails_to_remove.pop()
2330
parent_map.pop(next)
2331
children = child_map.pop(next)
2332
for child in children:
2333
child_parents = parent_map[child]
2334
child_parents.remove(next)
2335
if len(child_parents) == 0:
2336
tails_to_remove.append(child)
2338
def _get_interesting_texts(self, parent_map):
2339
"""Return a dict of texts we are interested in.
2341
Note that the input is in key tuples, but the output is in plain
2344
:param parent_map: The output from _find_recursive_lcas
2345
:return: A dict of {'revision_id':lines} as returned by
2346
_PlanMergeBase.get_lines()
2348
all_revision_keys = set(parent_map)
2349
all_revision_keys.add(self.a_key)
2350
all_revision_keys.add(self.b_key)
2352
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2353
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2356
def _build_weave(self):
2357
from bzrlib import weave
2358
self._weave = weave.Weave(weave_name='in_memory_weave',
2359
allow_reserved=True)
2360
parent_map = self._find_recursive_lcas()
2362
all_texts = self._get_interesting_texts(parent_map)
2364
# Note: Unfortunately, the order given by topo_sort will effect the
2365
# ordering resolution in the output. Specifically, if you add A then B,
2366
# then in the output text A lines will show up before B lines. And, of
2367
# course, topo_sort doesn't guarantee any real ordering.
2368
# So we use merge_sort, and add a fake node on the tip.
2369
# This ensures that left-hand parents will always be inserted into the
2370
# weave before right-hand parents.
2371
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2372
parent_map[tip_key] = (self.a_key, self.b_key)
2374
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2378
# for key in tsort.topo_sort(parent_map):
2379
parent_keys = parent_map[key]
2380
revision_id = key[-1]
2381
parent_ids = [k[-1] for k in parent_keys]
2382
self._weave.add_lines(revision_id, parent_ids,
2383
all_texts[revision_id])
2385
def plan_merge(self):
2386
"""Generate a 'plan' for merging the two revisions.
2388
This involves comparing their texts and determining the cause of
2389
differences. If text A has a line and text B does not, then either the
2390
line was added to text A, or it was deleted from B. Once the causes
2391
are combined, they are written out in the format described in
2392
VersionedFile.plan_merge
2394
if self._head_key is not None: # There was a single head
2395
if self._head_key == self.a_key:
2398
if self._head_key != self.b_key:
2399
raise AssertionError('There was an invalid head: %s != %s'
2400
% (self.b_key, self._head_key))
2402
head_rev = self._head_key[-1]
2403
lines = self.get_lines([head_rev])[head_rev]
2404
return ((plan, line) for line in lines)
2405
return self._weave.plan_merge(self.a_rev, self.b_rev)
2408
class _PlanLCAMerge(_PlanMergeBase):
2410
This merge algorithm differs from _PlanMerge in that:
2411
1. comparisons are done against LCAs only
2412
2. cases where a contested line is new versus one LCA but old versus
2413
another are marked as conflicts, by emitting the line as conflicted-a
2416
This is faster, and hopefully produces more useful output.
2419
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2420
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2421
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2424
if lca == _mod_revision.NULL_REVISION:
2427
self.lcas.add(lca[-1])
2428
for lca in self.lcas:
2429
if _mod_revision.is_null(lca):
2432
lca_lines = self.get_lines([lca])[lca]
2433
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2435
blocks = list(matcher.get_matching_blocks())
2436
self._cached_matching_blocks[(a_rev, lca)] = blocks
2437
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2439
blocks = list(matcher.get_matching_blocks())
2440
self._cached_matching_blocks[(b_rev, lca)] = blocks
2442
def _determine_status(self, revision_id, unique_line_numbers):
2443
"""Determines the status unique lines versus all lcas.
2445
Basically, determines why the line is unique to this revision.
2447
A line may be determined new, killed, or both.
2449
If a line is determined new, that means it was not present in at least
2450
one LCA, and is not present in the other merge revision.
2452
If a line is determined killed, that means the line was present in
2455
If a line is killed and new, this indicates that the two merge
2456
revisions contain differing conflict resolutions.
2457
:param revision_id: The id of the revision in which the lines are
2459
:param unique_line_numbers: The line numbers of unique lines.
2460
:return a tuple of (new_this, killed_other):
2464
unique_line_numbers = set(unique_line_numbers)
2465
for lca in self.lcas:
2466
blocks = self._get_matching_blocks(revision_id, lca)
2467
unique_vs_lca, _ignored = self._unique_lines(blocks)
2468
new.update(unique_line_numbers.intersection(unique_vs_lca))
2469
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
