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,
41
from bzrlib.i18n import gettext
48
from bzrlib.symbol_versioning import (
52
# TODO: Report back as changes are merged in
55
def transform_tree(from_tree, to_tree, interesting_ids=None):
56
from_tree.lock_tree_write()
57
operation = cleanup.OperationWithCleanups(merge_inner)
58
operation.add_cleanup(from_tree.unlock)
59
operation.run_simple(from_tree.branch, to_tree, from_tree,
60
ignore_zero=True, interesting_ids=interesting_ids, this_tree=from_tree)
63
class MergeHooks(hooks.Hooks):
66
hooks.Hooks.__init__(self, "bzrlib.merge", "Merger.hooks")
67
self.add_hook('merge_file_content',
68
"Called with a bzrlib.merge.Merger object to create a per file "
69
"merge object when starting a merge. "
70
"Should return either None or a subclass of "
71
"``bzrlib.merge.AbstractPerFileMerger``. "
72
"Such objects will then be called per file "
73
"that needs to be merged (including when one "
74
"side has deleted the file and the other has changed it). "
75
"See the AbstractPerFileMerger API docs for details on how it is "
80
class AbstractPerFileMerger(object):
81
"""PerFileMerger objects are used by plugins extending merge for bzrlib.
83
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
85
:ivar merger: The Merge3Merger performing the merge.
88
def __init__(self, merger):
89
"""Create a PerFileMerger for use with merger."""
92
def merge_contents(self, merge_params):
93
"""Attempt to merge the contents of a single file.
95
:param merge_params: A bzrlib.merge.MergeHookParams
96
:return: A tuple of (status, chunks), where status is one of
97
'not_applicable', 'success', 'conflicted', or 'delete'. If status
98
is 'success' or 'conflicted', then chunks should be an iterable of
99
strings for the new file contents.
101
return ('not applicable', None)
104
class PerFileMerger(AbstractPerFileMerger):
105
"""Merge individual files when self.file_matches returns True.
107
This class is intended to be subclassed. The file_matches and
108
merge_matching methods should be overridden with concrete implementations.
111
def file_matches(self, params):
112
"""Return True if merge_matching should be called on this file.
114
Only called with merges of plain files with no clear winner.
116
Subclasses must override this.
118
raise NotImplementedError(self.file_matches)
120
def get_filename(self, params, tree):
121
"""Lookup the filename (i.e. basename, not path), given a Tree (e.g.
122
self.merger.this_tree) and a MergeHookParams.
124
return osutils.basename(tree.id2path(params.file_id))
126
def get_filepath(self, params, tree):
127
"""Calculate the path to the file in a tree.
129
:param params: A MergeHookParams describing the file to merge
130
:param tree: a Tree, e.g. self.merger.this_tree.
132
return tree.id2path(params.file_id)
134
def merge_contents(self, params):
135
"""Merge the contents of a single file."""
136
# Check whether this custom merge logic should be used.
138
# OTHER is a straight winner, rely on default merge.
139
params.winner == 'other' or
140
# THIS and OTHER aren't both files.
141
not params.is_file_merge() or
142
# The filename doesn't match *.xml
143
not self.file_matches(params)):
144
return 'not_applicable', None
145
return self.merge_matching(params)
147
def merge_matching(self, params):
148
"""Merge the contents of a single file that has matched the criteria
149
in PerFileMerger.merge_contents (is a conflict, is a file,
150
self.file_matches is True).
152
Subclasses must override this.
154
raise NotImplementedError(self.merge_matching)
157
class ConfigurableFileMerger(PerFileMerger):
158
"""Merge individual files when configured via a .conf file.
160
This is a base class for concrete custom file merging logic. Concrete
161
classes should implement ``merge_text``.
163
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
165
:ivar affected_files: The configured file paths to merge.
167
:cvar name_prefix: The prefix to use when looking up configuration
168
details. <name_prefix>_merge_files describes the files targeted by the
171
:cvar default_files: The default file paths to merge when no configuration
178
def __init__(self, merger):
179
super(ConfigurableFileMerger, self).__init__(merger)
180
self.affected_files = None
181
self.default_files = self.__class__.default_files or []
182
self.name_prefix = self.__class__.name_prefix
183
if self.name_prefix is None:
184
raise ValueError("name_prefix must be set.")
186
def file_matches(self, params):
187
"""Check whether the file should call the merge hook.
189
<name_prefix>_merge_files configuration variable is a list of files
190
that should use the hook.
192
affected_files = self.affected_files
193
if affected_files is None:
194
config = self.merger.this_branch.get_config()
195
# Until bzr provides a better policy for caching the config, we
196
# just add the part we're interested in to the params to avoid
197
# reading the config files repeatedly (bazaar.conf, location.conf,
199
config_key = self.name_prefix + '_merge_files'
200
affected_files = config.get_user_option_as_list(config_key)
201
if affected_files is None:
202
# If nothing was specified in the config, use the default.
203
affected_files = self.default_files
204
self.affected_files = affected_files
206
filepath = self.get_filepath(params, self.merger.this_tree)
207
if filepath in affected_files:
211
def merge_matching(self, params):
212
return self.merge_text(params)
214
def merge_text(self, params):
215
"""Merge the byte contents of a single file.
217
This is called after checking that the merge should be performed in
218
merge_contents, and it should behave as per
219
``bzrlib.merge.AbstractPerFileMerger.merge_contents``.
221
raise NotImplementedError(self.merge_text)
224
class MergeHookParams(object):
225
"""Object holding parameters passed to merge_file_content hooks.
227
There are some fields hooks can access:
229
:ivar file_id: the file ID of the file being merged
230
:ivar trans_id: the transform ID for the merge of this file
231
:ivar this_kind: kind of file_id in 'this' tree
232
:ivar other_kind: kind of file_id in 'other' tree
233
:ivar winner: one of 'this', 'other', 'conflict'
236
def __init__(self, merger, file_id, trans_id, this_kind, other_kind,
238
self._merger = merger
239
self.file_id = file_id
240
self.trans_id = trans_id
241
self.this_kind = this_kind
242
self.other_kind = other_kind
245
def is_file_merge(self):
246
"""True if this_kind and other_kind are both 'file'."""
247
return self.this_kind == 'file' and self.other_kind == 'file'
249
@decorators.cachedproperty
250
def base_lines(self):
251
"""The lines of the 'base' version of the file."""
252
return self._merger.get_lines(self._merger.base_tree, self.file_id)
254
@decorators.cachedproperty
255
def this_lines(self):
256
"""The lines of the 'this' version of the file."""
257
return self._merger.get_lines(self._merger.this_tree, self.file_id)
259
@decorators.cachedproperty
260
def other_lines(self):
261
"""The lines of the 'other' version of the file."""
262
return self._merger.get_lines(self._merger.other_tree, self.file_id)
265
class Merger(object):
269
def __init__(self, this_branch, other_tree=None, base_tree=None,
270
this_tree=None, pb=None, change_reporter=None,
271
recurse='down', revision_graph=None):
272
object.__init__(self)
273
self.this_branch = this_branch
274
self.this_basis = _mod_revision.ensure_null(
275
this_branch.last_revision())
276
self.this_rev_id = None
277
self.this_tree = this_tree
278
self.this_revision_tree = None
279
self.this_basis_tree = None
280
self.other_tree = other_tree
281
self.other_branch = None
282
self.base_tree = base_tree
283
self.ignore_zero = False
284
self.backup_files = False
285
self.interesting_ids = None
286
self.interesting_files = None
287
self.show_base = False
288
self.reprocess = False
290
warnings.warn("pb parameter to Merger() is deprecated and ignored")
292
self.recurse = recurse
293
self.change_reporter = change_reporter
294
self._cached_trees = {}
295
self._revision_graph = revision_graph
296
self._base_is_ancestor = None
297
self._base_is_other_ancestor = None
298
self._is_criss_cross = None
299
self._lca_trees = None
301
def cache_trees_with_revision_ids(self, trees):
302
"""Cache any tree in trees if it has a revision_id."""
303
for maybe_tree in trees:
304
if maybe_tree is None:
307
rev_id = maybe_tree.get_revision_id()
308
except AttributeError:
310
self._cached_trees[rev_id] = maybe_tree
313
def revision_graph(self):
314
if self._revision_graph is None:
315
self._revision_graph = self.this_branch.repository.get_graph()
316
return self._revision_graph
318
def _set_base_is_ancestor(self, value):
319
self._base_is_ancestor = value
321
def _get_base_is_ancestor(self):
322
if self._base_is_ancestor is None:
323
self._base_is_ancestor = self.revision_graph.is_ancestor(
324
self.base_rev_id, self.this_basis)
325
return self._base_is_ancestor
327
base_is_ancestor = property(_get_base_is_ancestor, _set_base_is_ancestor)
329
def _set_base_is_other_ancestor(self, value):
330
self._base_is_other_ancestor = value
332
def _get_base_is_other_ancestor(self):
333
if self._base_is_other_ancestor is None:
334
if self.other_basis is None:
336
self._base_is_other_ancestor = self.revision_graph.is_ancestor(
337
self.base_rev_id, self.other_basis)
338
return self._base_is_other_ancestor
340
base_is_other_ancestor = property(_get_base_is_other_ancestor,
341
_set_base_is_other_ancestor)
344
def from_uncommitted(tree, other_tree, pb=None, base_tree=None):
345
"""Return a Merger for uncommitted changes in other_tree.
347
:param tree: The tree to merge into
348
:param other_tree: The tree to get uncommitted changes from
349
:param pb: A progress indicator
350
:param base_tree: The basis to use for the merge. If unspecified,
351
other_tree.basis_tree() will be used.
353
if base_tree is None:
354
base_tree = other_tree.basis_tree()
355
merger = Merger(tree.branch, other_tree, base_tree, tree, pb)
356
merger.base_rev_id = merger.base_tree.get_revision_id()
357
merger.other_rev_id = None
358
merger.other_basis = merger.base_rev_id
362
def from_mergeable(klass, tree, mergeable, pb):
363
"""Return a Merger for a bundle or merge directive.
365
:param tree: The tree to merge changes into
366
:param mergeable: A merge directive or bundle
367
:param pb: A progress indicator
369
mergeable.install_revisions(tree.branch.repository)
370
base_revision_id, other_revision_id, verified =\
371
mergeable.get_merge_request(tree.branch.repository)
372
revision_graph = tree.branch.repository.get_graph()
373
if base_revision_id is not None:
374
if (base_revision_id != _mod_revision.NULL_REVISION and
375
revision_graph.is_ancestor(
376
base_revision_id, tree.branch.last_revision())):
377
base_revision_id = None
379
trace.warning('Performing cherrypick')
380
merger = klass.from_revision_ids(pb, tree, other_revision_id,
381
base_revision_id, revision_graph=
383
return merger, verified
386
def from_revision_ids(pb, tree, other, base=None, other_branch=None,
387
base_branch=None, revision_graph=None,
389
"""Return a Merger for revision-ids.
391
:param pb: A progress indicator
392
:param tree: The tree to merge changes into
393
:param other: The revision-id to use as OTHER
394
:param base: The revision-id to use as BASE. If not specified, will
396
:param other_branch: A branch containing the other revision-id. If
397
not supplied, tree.branch is used.
398
:param base_branch: A branch containing the base revision-id. If
399
not supplied, other_branch or tree.branch will be used.
400
:param revision_graph: If you have a revision_graph precomputed, pass
401
it in, otherwise it will be created for you.
402
:param tree_branch: The branch associated with tree. If not supplied,
403
tree.branch will be used.
405
if tree_branch is None:
406
tree_branch = tree.branch
407
merger = Merger(tree_branch, this_tree=tree, pb=pb,
408
revision_graph=revision_graph)
409
if other_branch is None:
410
other_branch = tree.branch
411
merger.set_other_revision(other, other_branch)
415
if base_branch is None:
416
base_branch = other_branch
417
merger.set_base_revision(base, base_branch)
420
def revision_tree(self, revision_id, branch=None):
421
if revision_id not in self._cached_trees:
423
branch = self.this_branch
425
tree = self.this_tree.revision_tree(revision_id)
426
except errors.NoSuchRevisionInTree:
427
tree = branch.repository.revision_tree(revision_id)
428
self._cached_trees[revision_id] = tree
429
return self._cached_trees[revision_id]
431
def _get_tree(self, treespec, possible_transports=None):
432
location, revno = treespec
434
tree = workingtree.WorkingTree.open_containing(location)[0]
435
return tree.branch, tree
436
branch = _mod_branch.Branch.open_containing(
437
location, possible_transports)[0]
439
revision_id = branch.last_revision()
441
revision_id = branch.get_rev_id(revno)
442
revision_id = _mod_revision.ensure_null(revision_id)
443
return branch, self.revision_tree(revision_id, branch)
445
@deprecated_method(deprecated_in((2, 1, 0)))
446
def ensure_revision_trees(self):
447
if self.this_revision_tree is None:
448
self.this_basis_tree = self.revision_tree(self.this_basis)
449
if self.this_basis == self.this_rev_id:
450
self.this_revision_tree = self.this_basis_tree
452
if self.other_rev_id is None:
453
other_basis_tree = self.revision_tree(self.other_basis)
454
if other_basis_tree.has_changes(self.other_tree):
455
raise errors.WorkingTreeNotRevision(self.this_tree)
456
other_rev_id = self.other_basis
457
self.other_tree = other_basis_tree
459
@deprecated_method(deprecated_in((2, 1, 0)))
460
def file_revisions(self, file_id):
461
self.ensure_revision_trees()
462
if self.this_rev_id is None:
463
if self.this_basis_tree.get_file_sha1(file_id) != \
464
self.this_tree.get_file_sha1(file_id):
465
raise errors.WorkingTreeNotRevision(self.this_tree)
467
trees = (self.this_basis_tree, self.other_tree)
468
return [tree.get_file_revision(file_id) for tree in trees]
470
@deprecated_method(deprecated_in((2, 1, 0)))
471
def check_basis(self, check_clean, require_commits=True):
472
if self.this_basis is None and require_commits is True:
473
raise errors.BzrCommandError(
474
"This branch has no commits."
475
" (perhaps you would prefer 'bzr pull')")
478
if self.this_basis != self.this_rev_id:
479
raise errors.UncommittedChanges(self.this_tree)
481
@deprecated_method(deprecated_in((2, 1, 0)))
482
def compare_basis(self):
484
basis_tree = self.revision_tree(self.this_tree.last_revision())
485
except errors.NoSuchRevision:
486
basis_tree = self.this_tree.basis_tree()
487
if not self.this_tree.has_changes(basis_tree):
488
self.this_rev_id = self.this_basis
490
def set_interesting_files(self, file_list):
491
self.interesting_files = file_list
493
def set_pending(self):
494
if (not self.base_is_ancestor or not self.base_is_other_ancestor
495
or self.other_rev_id is None):
499
def _add_parent(self):
500
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
501
new_parent_trees = []
502
operation = cleanup.OperationWithCleanups(
503
self.this_tree.set_parent_trees)
504
for revision_id in new_parents:
506
tree = self.revision_tree(revision_id)
507
except errors.NoSuchRevision:
511
operation.add_cleanup(tree.unlock)
512
new_parent_trees.append((revision_id, tree))
513
operation.run_simple(new_parent_trees, allow_leftmost_as_ghost=True)
515
def set_other(self, other_revision, possible_transports=None):
516
"""Set the revision and tree to merge from.
518
This sets the other_tree, other_rev_id, other_basis attributes.
520
:param other_revision: The [path, revision] list to merge from.
522
self.other_branch, self.other_tree = self._get_tree(other_revision,
524
if other_revision[1] == -1:
525
self.other_rev_id = _mod_revision.ensure_null(
526
self.other_branch.last_revision())
527
if _mod_revision.is_null(self.other_rev_id):
528
raise errors.NoCommits(self.other_branch)
529
self.other_basis = self.other_rev_id
530
elif other_revision[1] is not None:
531
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
532
self.other_basis = self.other_rev_id
534
self.other_rev_id = None
535
self.other_basis = self.other_branch.last_revision()
536
if self.other_basis is None:
537
raise errors.NoCommits(self.other_branch)
538
if self.other_rev_id is not None:
539
self._cached_trees[self.other_rev_id] = self.other_tree
540
self._maybe_fetch(self.other_branch,self.this_branch, self.other_basis)
542
def set_other_revision(self, revision_id, other_branch):
543
"""Set 'other' based on a branch and revision id
545
:param revision_id: The revision to use for a tree
546
:param other_branch: The branch containing this tree
548
self.other_rev_id = revision_id
549
self.other_branch = other_branch
550
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
551
self.other_tree = self.revision_tree(revision_id)
552
self.other_basis = revision_id
554
def set_base_revision(self, revision_id, branch):
555
"""Set 'base' based on a branch and revision id
557
:param revision_id: The revision to use for a tree
558
:param branch: The branch containing this tree
560
self.base_rev_id = revision_id
561
self.base_branch = branch
562
self._maybe_fetch(branch, self.this_branch, revision_id)
563
self.base_tree = self.revision_tree(revision_id)
565
def _maybe_fetch(self, source, target, revision_id):
566
if not source.repository.has_same_location(target.repository):
567
target.fetch(source, revision_id)
570
revisions = [_mod_revision.ensure_null(self.this_basis),
571
_mod_revision.ensure_null(self.other_basis)]
572
if _mod_revision.NULL_REVISION in revisions:
573
self.base_rev_id = _mod_revision.NULL_REVISION
574
self.base_tree = self.revision_tree(self.base_rev_id)
575
self._is_criss_cross = False
577
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
578
self._is_criss_cross = False
580
self.base_rev_id = _mod_revision.NULL_REVISION
582
self.base_rev_id = list(lcas)[0]
583
else: # len(lcas) > 1
584
self._is_criss_cross = True
586
# find_unique_lca can only handle 2 nodes, so we have to
587
# start back at the beginning. It is a shame to traverse
588
# the graph again, but better than re-implementing
590
self.base_rev_id = self.revision_graph.find_unique_lca(
591
revisions[0], revisions[1])
593
self.base_rev_id = self.revision_graph.find_unique_lca(
595
sorted_lca_keys = self.revision_graph.find_merge_order(
597
if self.base_rev_id == _mod_revision.NULL_REVISION:
598
self.base_rev_id = sorted_lca_keys[0]
600
if self.base_rev_id == _mod_revision.NULL_REVISION:
601
raise errors.UnrelatedBranches()
602
if self._is_criss_cross:
603
trace.warning('Warning: criss-cross merge encountered. See bzr'
604
' help criss-cross.')
605
trace.mutter('Criss-cross lcas: %r' % lcas)
606
if self.base_rev_id in lcas:
607
trace.mutter('Unable to find unique lca. '
608
'Fallback %r as best option.' % self.base_rev_id)
609
interesting_revision_ids = set(lcas)
610
interesting_revision_ids.add(self.base_rev_id)
611
interesting_trees = dict((t.get_revision_id(), t)
612
for t in self.this_branch.repository.revision_trees(
613
interesting_revision_ids))
614
self._cached_trees.update(interesting_trees)
615
if self.base_rev_id in lcas:
616
self.base_tree = interesting_trees[self.base_rev_id]
618
self.base_tree = interesting_trees.pop(self.base_rev_id)
619
self._lca_trees = [interesting_trees[key]
620
for key in sorted_lca_keys]
622
self.base_tree = self.revision_tree(self.base_rev_id)
623
self.base_is_ancestor = True
624
self.base_is_other_ancestor = True
625
trace.mutter('Base revid: %r' % self.base_rev_id)
627
def set_base(self, base_revision):
628
"""Set the base revision to use for the merge.
630
:param base_revision: A 2-list containing a path and revision number.
632
trace.mutter("doing merge() with no base_revision specified")
633
if base_revision == [None, None]:
636
base_branch, self.base_tree = self._get_tree(base_revision)
637
if base_revision[1] == -1:
638
self.base_rev_id = base_branch.last_revision()
639
elif base_revision[1] is None:
640
self.base_rev_id = _mod_revision.NULL_REVISION
642
self.base_rev_id = _mod_revision.ensure_null(
643
base_branch.get_rev_id(base_revision[1]))
644
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
646
def make_merger(self):
647
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
648
'other_tree': self.other_tree,
649
'interesting_ids': self.interesting_ids,
650
'interesting_files': self.interesting_files,
651
'this_branch': self.this_branch,
653
if self.merge_type.requires_base:
654
kwargs['base_tree'] = self.base_tree
655
if self.merge_type.supports_reprocess:
656
kwargs['reprocess'] = self.reprocess
658
raise errors.BzrError(
659
"Conflict reduction is not supported for merge"
660
" type %s." % self.merge_type)
661
if self.merge_type.supports_show_base:
662
kwargs['show_base'] = self.show_base
664
raise errors.BzrError("Showing base is not supported for this"
665
" merge type. %s" % self.merge_type)
666
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
667
and not self.base_is_other_ancestor):
668
raise errors.CannotReverseCherrypick()
669
if self.merge_type.supports_cherrypick:
670
kwargs['cherrypick'] = (not self.base_is_ancestor or
671
not self.base_is_other_ancestor)
672
if self._is_criss_cross and getattr(self.merge_type,
673
'supports_lca_trees', False):
674
kwargs['lca_trees'] = self._lca_trees
675
return self.merge_type(pb=None,
676
change_reporter=self.change_reporter,
679
def _do_merge_to(self):
680
merge = self.make_merger()
681
if self.other_branch is not None:
682
self.other_branch.update_references(self.this_branch)
684
if self.recurse == 'down':
685
for relpath, file_id in self.this_tree.iter_references():
686
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
687
other_revision = self.other_tree.get_reference_revision(
689
if other_revision == sub_tree.last_revision():
691
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
692
sub_merge.merge_type = self.merge_type
693
other_branch = self.other_branch.reference_parent(file_id, relpath)
694
sub_merge.set_other_revision(other_revision, other_branch)
695
base_revision = self.base_tree.get_reference_revision(file_id)
696
sub_merge.base_tree = \
697
sub_tree.branch.repository.revision_tree(base_revision)
698
sub_merge.base_rev_id = base_revision
703
operation = cleanup.OperationWithCleanups(self._do_merge_to)
704
self.this_tree.lock_tree_write()
705
operation.add_cleanup(self.this_tree.unlock)
706
if self.base_tree is not None:
707
self.base_tree.lock_read()
708
operation.add_cleanup(self.base_tree.unlock)
709
if self.other_tree is not None:
710
self.other_tree.lock_read()
711
operation.add_cleanup(self.other_tree.unlock)
712
merge = operation.run_simple()
713
if len(merge.cooked_conflicts) == 0:
714
if not self.ignore_zero and not trace.is_quiet():
715
trace.note(gettext("All changes applied successfully."))
717
trace.note(gettext("%d conflicts encountered.")
718
% len(merge.cooked_conflicts))
720
return len(merge.cooked_conflicts)
723
class _InventoryNoneEntry(object):
724
"""This represents an inventory entry which *isn't there*.
726
It simplifies the merging logic if we always have an InventoryEntry, even
727
if it isn't actually present
734
symlink_target = None
737
_none_entry = _InventoryNoneEntry()
740
class Merge3Merger(object):
741
"""Three-way merger that uses the merge3 text merger"""
743
supports_reprocess = True
744
supports_show_base = True
745
history_based = False
746
supports_cherrypick = True
747
supports_reverse_cherrypick = True
748
winner_idx = {"this": 2, "other": 1, "conflict": 1}
749
supports_lca_trees = True
751
def __init__(self, working_tree, this_tree, base_tree, other_tree,
752
interesting_ids=None, reprocess=False, show_base=False,
753
pb=None, pp=None, change_reporter=None,
754
interesting_files=None, do_merge=True,
755
cherrypick=False, lca_trees=None, this_branch=None):
756
"""Initialize the merger object and perform the merge.
758
:param working_tree: The working tree to apply the merge to
759
:param this_tree: The local tree in the merge operation
760
:param base_tree: The common tree in the merge operation
761
:param other_tree: The other tree to merge changes from
762
:param this_branch: The branch associated with this_tree. Defaults to
763
this_tree.branch if not supplied.
764
:param interesting_ids: The file_ids of files that should be
765
participate in the merge. May not be combined with
767
:param: reprocess If True, perform conflict-reduction processing.
768
:param show_base: If True, show the base revision in text conflicts.
769
(incompatible with reprocess)
771
:param pp: A ProgressPhase object
772
:param change_reporter: An object that should report changes made
773
:param interesting_files: The tree-relative paths of files that should
774
participate in the merge. If these paths refer to directories,
775
the contents of those directories will also be included. May not
776
be combined with interesting_ids. If neither interesting_files nor
777
interesting_ids is specified, all files may participate in the
779
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
780
if the ancestry was found to include a criss-cross merge.
781
Otherwise should be None.
783
object.__init__(self)
784
if interesting_files is not None and interesting_ids is not None:
786
'specify either interesting_ids or interesting_files')
787
if this_branch is None:
788
this_branch = this_tree.branch
789
self.interesting_ids = interesting_ids
790
self.interesting_files = interesting_files
791
self.this_tree = working_tree
792
self.base_tree = base_tree
793
self.other_tree = other_tree
794
self.this_branch = this_branch
795
self._raw_conflicts = []
796
self.cooked_conflicts = []
797
self.reprocess = reprocess
798
self.show_base = show_base
799
self._lca_trees = lca_trees
800
# Uncommenting this will change the default algorithm to always use
801
# _entries_lca. This can be useful for running the test suite and
802
# making sure we haven't missed any corner cases.
803
# if lca_trees is None:
804
# self._lca_trees = [self.base_tree]
805
self.change_reporter = change_reporter
806
self.cherrypick = cherrypick
810
warnings.warn("pp argument to Merge3Merger is deprecated")
812
warnings.warn("pb argument to Merge3Merger is deprecated")
815
operation = cleanup.OperationWithCleanups(self._do_merge)
816
self.this_tree.lock_tree_write()
817
operation.add_cleanup(self.this_tree.unlock)
818
self.base_tree.lock_read()
819
operation.add_cleanup(self.base_tree.unlock)
820
self.other_tree.lock_read()
821
operation.add_cleanup(self.other_tree.unlock)
824
def _do_merge(self, operation):
825
self.tt = transform.TreeTransform(self.this_tree, None)
826
operation.add_cleanup(self.tt.finalize)
827
self._compute_transform()
828
results = self.tt.apply(no_conflicts=True)
829
self.write_modified(results)
831
self.this_tree.add_conflicts(self.cooked_conflicts)
832
except errors.UnsupportedOperation:
835
def make_preview_transform(self):
836
operation = cleanup.OperationWithCleanups(self._make_preview_transform)
837
self.base_tree.lock_read()
838
operation.add_cleanup(self.base_tree.unlock)
839
self.other_tree.lock_read()
840
operation.add_cleanup(self.other_tree.unlock)
841
return operation.run_simple()
843
def _make_preview_transform(self):
844
self.tt = transform.TransformPreview(self.this_tree)
845
self._compute_transform()
848
def _compute_transform(self):
849
if self._lca_trees is None:
850
entries = self._entries3()
851
resolver = self._three_way
853
entries = self._entries_lca()
854
resolver = self._lca_multi_way
855
child_pb = ui.ui_factory.nested_progress_bar()
857
factories = Merger.hooks['merge_file_content']
858
hooks = [factory(self) for factory in factories] + [self]
859
self.active_hooks = [hook for hook in hooks if hook is not None]
860
for num, (file_id, changed, parents3, names3,
861
executable3) in enumerate(entries):
862
child_pb.update(gettext('Preparing file merge'), num, len(entries))
863
self._merge_names(file_id, parents3, names3, resolver=resolver)
865
file_status = self._do_merge_contents(file_id)
867
file_status = 'unmodified'
868
self._merge_executable(file_id,
869
executable3, file_status, resolver=resolver)
872
self.tt.fixup_new_roots()
873
self._finish_computing_transform()
875
def _finish_computing_transform(self):
876
"""Finalize the transform and report the changes.
878
This is the second half of _compute_transform.
880
child_pb = ui.ui_factory.nested_progress_bar()
882
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
883
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
886
if self.change_reporter is not None:
887
from bzrlib import delta
888
delta.report_changes(
889
self.tt.iter_changes(), self.change_reporter)
890
self.cook_conflicts(fs_conflicts)
891
for conflict in self.cooked_conflicts:
892
trace.warning(unicode(conflict))
895
"""Gather data about files modified between three trees.
897
Return a list of tuples of file_id, changed, parents3, names3,
898
executable3. changed is a boolean indicating whether the file contents
899
or kind were changed. parents3 is a tuple of parent ids for base,
900
other and this. names3 is a tuple of names for base, other and this.
901
executable3 is a tuple of execute-bit values for base, other and this.
904
iterator = self.other_tree.iter_changes(self.base_tree,
905
specific_files=self.interesting_files,
906
extra_trees=[self.this_tree])
907
this_entries = dict((e.file_id, e) for p, e in
908
self.this_tree.iter_entries_by_dir(
909
self.interesting_ids))
910
for (file_id, paths, changed, versioned, parents, names, kind,
911
executable) in iterator:
912
if (self.interesting_ids is not None and
913
file_id not in self.interesting_ids):
915
entry = this_entries.get(file_id)
916
if entry is not None:
917
this_name = entry.name
918
this_parent = entry.parent_id
919
this_executable = entry.executable
923
this_executable = None
924
parents3 = parents + (this_parent,)
925
names3 = names + (this_name,)
926
executable3 = executable + (this_executable,)
927
result.append((file_id, changed, parents3, names3, executable3))
930
def _entries_lca(self):
931
"""Gather data about files modified between multiple trees.
933
This compares OTHER versus all LCA trees, and for interesting entries,
934
it then compares with THIS and BASE.
936
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
938
:return: [(file_id, changed, parents, names, executable)], where:
940
* file_id: Simple file_id of the entry
941
* changed: Boolean, True if the kind or contents changed else False
942
* parents: ((base, [parent_id, in, lcas]), parent_id_other,
944
* names: ((base, [name, in, lcas]), name_in_other, name_in_this)
945
* executable: ((base, [exec, in, lcas]), exec_in_other,
948
if self.interesting_files is not None:
949
lookup_trees = [self.this_tree, self.base_tree]
950
lookup_trees.extend(self._lca_trees)
951
# I think we should include the lca trees as well
952
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
955
interesting_ids = self.interesting_ids
957
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
959
base_inventory = self.base_tree.inventory
960
this_inventory = self.this_tree.inventory
961
for path, file_id, other_ie, lca_values in walker.iter_all():
962
# Is this modified at all from any of the other trees?
964
other_ie = _none_entry
965
if interesting_ids is not None and file_id not in interesting_ids:
968
# If other_revision is found in any of the lcas, that means this
969
# node is uninteresting. This is because when merging, if there are
970
# multiple heads(), we have to create a new node. So if we didn't,
971
# we know that the ancestry is linear, and that OTHER did not
973
# See doc/developers/lca_merge_resolution.txt for details
974
other_revision = other_ie.revision
975
if other_revision is not None:
976
# We can't use this shortcut when other_revision is None,
977
# because it may be None because things are WorkingTrees, and
978
# not because it is *actually* None.
979
is_unmodified = False
980
for lca_path, ie in lca_values:
981
if ie is not None and ie.revision == other_revision:
988
for lca_path, lca_ie in lca_values:
990
lca_entries.append(_none_entry)
992
lca_entries.append(lca_ie)
994
if base_inventory.has_id(file_id):
995
base_ie = base_inventory[file_id]
997
base_ie = _none_entry
999
if this_inventory.has_id(file_id):
1000
this_ie = this_inventory[file_id]
1002
this_ie = _none_entry
1008
for lca_ie in lca_entries:
1009
lca_kinds.append(lca_ie.kind)
1010
lca_parent_ids.append(lca_ie.parent_id)
1011
lca_names.append(lca_ie.name)
1012
lca_executable.append(lca_ie.executable)
1014
kind_winner = self._lca_multi_way(
1015
(base_ie.kind, lca_kinds),
1016
other_ie.kind, this_ie.kind)
1017
parent_id_winner = self._lca_multi_way(
1018
(base_ie.parent_id, lca_parent_ids),
1019
other_ie.parent_id, this_ie.parent_id)
1020
name_winner = self._lca_multi_way(
1021
(base_ie.name, lca_names),
1022
other_ie.name, this_ie.name)
1024
content_changed = True
1025
if kind_winner == 'this':
1026
# No kind change in OTHER, see if there are *any* changes
1027
if other_ie.kind == 'directory':
1028
if parent_id_winner == 'this' and name_winner == 'this':
1029
# No change for this directory in OTHER, skip
1031
content_changed = False
1032
elif other_ie.kind is None or other_ie.kind == 'file':
1033
def get_sha1(ie, tree):
1034
if ie.kind != 'file':
1036
return tree.get_file_sha1(file_id)
1037
base_sha1 = get_sha1(base_ie, self.base_tree)
1038
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1039
in zip(lca_entries, self._lca_trees)]
1040
this_sha1 = get_sha1(this_ie, self.this_tree)
1041
other_sha1 = get_sha1(other_ie, self.other_tree)
1042
sha1_winner = self._lca_multi_way(
1043
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1044
allow_overriding_lca=False)
1045
exec_winner = self._lca_multi_way(
1046
(base_ie.executable, lca_executable),
1047
other_ie.executable, this_ie.executable)
1048
if (parent_id_winner == 'this' and name_winner == 'this'
1049
and sha1_winner == 'this' and exec_winner == 'this'):
1050
# No kind, parent, name, exec, or content change for
1051
# OTHER, so this node is not considered interesting
1053
if sha1_winner == 'this':
1054
content_changed = False
1055
elif other_ie.kind == 'symlink':
1056
def get_target(ie, tree):
1057
if ie.kind != 'symlink':
1059
return tree.get_symlink_target(file_id)
1060
base_target = get_target(base_ie, self.base_tree)
1061
lca_targets = [get_target(ie, tree) for ie, tree
1062
in zip(lca_entries, self._lca_trees)]
1063
this_target = get_target(this_ie, self.this_tree)
1064
other_target = get_target(other_ie, self.other_tree)
1065
target_winner = self._lca_multi_way(
1066
(base_target, lca_targets),
1067
other_target, this_target)
1068
if (parent_id_winner == 'this' and name_winner == 'this'
1069
and target_winner == 'this'):
1070
# No kind, parent, name, or symlink target change
1073
if target_winner == 'this':
1074
content_changed = False
1075
elif other_ie.kind == 'tree-reference':
1076
# The 'changed' information seems to be handled at a higher
1077
# level. At least, _entries3 returns False for content
1078
# changed, even when at a new revision_id.
1079
content_changed = False
1080
if (parent_id_winner == 'this' and name_winner == 'this'):
1081
# Nothing interesting
1084
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1086
# If we have gotten this far, that means something has changed
1087
result.append((file_id, content_changed,
1088
((base_ie.parent_id, lca_parent_ids),
1089
other_ie.parent_id, this_ie.parent_id),
1090
((base_ie.name, lca_names),
1091
other_ie.name, this_ie.name),
1092
((base_ie.executable, lca_executable),
1093
other_ie.executable, this_ie.executable)
1097
@deprecated_method(deprecated_in((2, 4, 0)))
1099
if self.tt.final_kind(self.tt.root) is None:
1100
self.tt.cancel_deletion(self.tt.root)
1101
if self.tt.final_file_id(self.tt.root) is None:
1102
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
1104
other_root_file_id = self.other_tree.get_root_id()
1105
if other_root_file_id is None:
1107
other_root = self.tt.trans_id_file_id(other_root_file_id)
1108
if other_root == self.tt.root:
1110
if self.this_tree.inventory.has_id(
1111
self.other_tree.inventory.root.file_id):
1112
# the other tree's root is a non-root in the current tree (as
1113
# when a previously unrelated branch is merged into another)
1115
if self.tt.final_kind(other_root) is not None:
1116
other_root_is_present = True
1118
# other_root doesn't have a physical representation. We still need
1119
# to move any references to the actual root of the tree.
1120
other_root_is_present = False
1121
# 'other_tree.inventory.root' is not present in this tree. We are
1122
# calling adjust_path for children which *want* to be present with a
1123
# correct place to go.
1124
for _, child in self.other_tree.inventory.root.children.iteritems():
1125
trans_id = self.tt.trans_id_file_id(child.file_id)
1126
if not other_root_is_present:
1127
if self.tt.final_kind(trans_id) is not None:
1128
# The item exist in the final tree and has a defined place
1131
# Move the item into the root
1133
final_name = self.tt.final_name(trans_id)
1134
except errors.NoFinalPath:
1135
# This file is not present anymore, ignore it.
1137
self.tt.adjust_path(final_name, self.tt.root, trans_id)
1138
if other_root_is_present:
1139
self.tt.cancel_creation(other_root)
1140
self.tt.cancel_versioning(other_root)
1142
def write_modified(self, results):
1143
modified_hashes = {}
1144
for path in results.modified_paths:
1145
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
1148
hash = self.this_tree.get_file_sha1(file_id)
1151
modified_hashes[file_id] = hash
1152
self.this_tree.set_merge_modified(modified_hashes)
1155
def parent(entry, file_id):
1156
"""Determine the parent for a file_id (used as a key method)"""
1159
return entry.parent_id
1162
def name(entry, file_id):
1163
"""Determine the name for a file_id (used as a key method)"""
1169
def contents_sha1(tree, file_id):
1170
"""Determine the sha1 of the file contents (used as a key method)."""
1171
if not tree.has_id(file_id):
1173
return tree.get_file_sha1(file_id)
1176
def executable(tree, file_id):
1177
"""Determine the executability of a file-id (used as a key method)."""
1178
if not tree.has_id(file_id):
1180
if tree.kind(file_id) != "file":
1182
return tree.is_executable(file_id)
1185
def kind(tree, file_id):
1186
"""Determine the kind of a file-id (used as a key method)."""
1187
if not tree.has_id(file_id):
1189
return tree.kind(file_id)
1192
def _three_way(base, other, this):
1194
# if 'base == other', either they all agree, or only 'this' has
1197
elif this not in (base, other):
1198
# 'this' is neither 'base' nor 'other', so both sides changed
1201
# "Ambiguous clean merge" -- both sides have made the same change.
1204
# this == base: only other has changed.
1208
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
1209
"""Consider LCAs when determining whether a change has occurred.
1211
If LCAS are all identical, this is the same as a _three_way comparison.
1213
:param bases: value in (BASE, [LCAS])
1214
:param other: value in OTHER
1215
:param this: value in THIS
1216
:param allow_overriding_lca: If there is more than one unique lca
1217
value, allow OTHER to override THIS if it has a new value, and
1218
THIS only has an lca value, or vice versa. This is appropriate for
1219
truly scalar values, not as much for non-scalars.
1220
:return: 'this', 'other', or 'conflict' depending on whether an entry
1223
# See doc/developers/lca_tree_merging.txt for details about this
1226
# Either Ambiguously clean, or nothing was actually changed. We
1229
base_val, lca_vals = bases
1230
# Remove 'base_val' from the lca_vals, because it is not interesting
1231
filtered_lca_vals = [lca_val for lca_val in lca_vals
1232
if lca_val != base_val]
1233
if len(filtered_lca_vals) == 0:
1234
return Merge3Merger._three_way(base_val, other, this)
1236
unique_lca_vals = set(filtered_lca_vals)
1237
if len(unique_lca_vals) == 1:
1238
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1240
if allow_overriding_lca:
1241
if other in unique_lca_vals:
1242
if this in unique_lca_vals:
1243
# Each side picked a different lca, conflict
1246
# This has a value which supersedes both lca values, and
1247
# other only has an lca value
1249
elif this in unique_lca_vals:
1250
# OTHER has a value which supersedes both lca values, and this
1251
# only has an lca value
1254
# At this point, the lcas disagree, and the tip disagree
1258
@deprecated_method(deprecated_in((2, 2, 0)))
1259
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1260
"""Do a three-way test on a scalar.
1261
Return "this", "other" or "conflict", depending whether a value wins.
1263
key_base = key(base_tree, file_id)
1264
key_other = key(other_tree, file_id)
1265
#if base == other, either they all agree, or only THIS has changed.
1266
if key_base == key_other:
1268
key_this = key(this_tree, file_id)
1269
# "Ambiguous clean merge"
1270
if key_this == key_other:
1272
elif key_this == key_base:
1277
def merge_names(self, file_id):
1278
def get_entry(tree):
1279
if tree.has_id(file_id):
1280
return tree.inventory[file_id]
1283
this_entry = get_entry(self.this_tree)
1284
other_entry = get_entry(self.other_tree)
1285
base_entry = get_entry(self.base_tree)
1286
entries = (base_entry, other_entry, this_entry)
1289
for entry in entries:
1292
parents.append(None)
1294
names.append(entry.name)
1295
parents.append(entry.parent_id)
1296
return self._merge_names(file_id, parents, names,
1297
resolver=self._three_way)
1299
def _merge_names(self, file_id, parents, names, resolver):
1300
"""Perform a merge on file_id names and parents"""
1301
base_name, other_name, this_name = names
1302
base_parent, other_parent, this_parent = parents
1304
name_winner = resolver(*names)
1306
parent_id_winner = resolver(*parents)
1307
if this_name is None:
1308
if name_winner == "this":
1309
name_winner = "other"
1310
if parent_id_winner == "this":
1311
parent_id_winner = "other"
1312
if name_winner == "this" and parent_id_winner == "this":
1314
if name_winner == 'conflict' or parent_id_winner == 'conflict':
1315
# Creating helpers (.OTHER or .THIS) here cause problems down the
1316
# road if a ContentConflict needs to be created so we should not do
1318
trans_id = self.tt.trans_id_file_id(file_id)
1319
self._raw_conflicts.append(('path conflict', trans_id, file_id,
1320
this_parent, this_name,
1321
other_parent, other_name))
1322
if not self.other_tree.has_id(file_id):
1323
# it doesn't matter whether the result was 'other' or
1324
# 'conflict'-- if it has no file id, we leave it alone.
1326
parent_id = parents[self.winner_idx[parent_id_winner]]
1327
name = names[self.winner_idx[name_winner]]
1328
if parent_id is not None or name is not None:
1329
# if we get here, name_winner and parent_winner are set to safe
1331
if parent_id is None and name is not None:
1332
# if parent_id is None and name is non-None, current file is
1334
if names[self.winner_idx[parent_id_winner]] != '':
1335
raise AssertionError(
1336
'File looks like a root, but named %s' %
1337
names[self.winner_idx[parent_id_winner]])
1338
parent_trans_id = transform.ROOT_PARENT
1340
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1341
self.tt.adjust_path(name, parent_trans_id,
1342
self.tt.trans_id_file_id(file_id))
1344
def _do_merge_contents(self, file_id):
1345
"""Performs a merge on file_id contents."""
1346
def contents_pair(tree):
1347
if not tree.has_id(file_id):
1349
kind = tree.kind(file_id)
1351
contents = tree.get_file_sha1(file_id)
1352
elif kind == "symlink":
1353
contents = tree.get_symlink_target(file_id)
1356
return kind, contents
1358
# See SPOT run. run, SPOT, run.
1359
# So we're not QUITE repeating ourselves; we do tricky things with
1361
base_pair = contents_pair(self.base_tree)
1362
other_pair = contents_pair(self.other_tree)
1364
this_pair = contents_pair(self.this_tree)
1365
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1366
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1367
this_pair, allow_overriding_lca=False)
1369
if base_pair == other_pair:
1372
# We delayed evaluating this_pair as long as we can to avoid
1373
# unnecessary sha1 calculation
1374
this_pair = contents_pair(self.this_tree)
1375
winner = self._three_way(base_pair, other_pair, this_pair)
1376
if winner == 'this':
1377
# No interesting changes introduced by OTHER
1379
# We have a hypothetical conflict, but if we have files, then we
1380
# can try to merge the content
1381
trans_id = self.tt.trans_id_file_id(file_id)
1382
params = MergeHookParams(self, file_id, trans_id, this_pair[0],
1383
other_pair[0], winner)
1384
hooks = self.active_hooks
1385
hook_status = 'not_applicable'
1387
hook_status, lines = hook.merge_contents(params)
1388
if hook_status != 'not_applicable':
1389
# Don't try any more hooks, this one applies.
1392
if hook_status == 'not_applicable':
1393
# This is a contents conflict, because none of the available
1394
# functions could merge it.
1396
name = self.tt.final_name(trans_id)
1397
parent_id = self.tt.final_parent(trans_id)
1398
if self.this_tree.has_id(file_id):
1399
self.tt.unversion_file(trans_id)
1400
file_group = self._dump_conflicts(name, parent_id, file_id,
1402
self._raw_conflicts.append(('contents conflict', file_group))
1403
elif hook_status == 'success':
1404
self.tt.create_file(lines, trans_id)
1405
elif hook_status == 'conflicted':
1406
# XXX: perhaps the hook should be able to provide
1407
# the BASE/THIS/OTHER files?
1408
self.tt.create_file(lines, trans_id)
1409
self._raw_conflicts.append(('text conflict', trans_id))
1410
name = self.tt.final_name(trans_id)
1411
parent_id = self.tt.final_parent(trans_id)
1412
self._dump_conflicts(name, parent_id, file_id)
1413
elif hook_status == 'delete':
1414
self.tt.unversion_file(trans_id)
1416
elif hook_status == 'done':
1417
# The hook function did whatever it needs to do directly, no
1418
# further action needed here.
1421
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1422
if not self.this_tree.has_id(file_id) and result == "modified":
1423
self.tt.version_file(file_id, trans_id)
1424
# The merge has been performed, so the old contents should not be
1426
self.tt.delete_contents(trans_id)
1429
def _default_other_winner_merge(self, merge_hook_params):
1430
"""Replace this contents with other."""
1431
file_id = merge_hook_params.file_id
1432
trans_id = merge_hook_params.trans_id
1433
file_in_this = self.this_tree.has_id(file_id)
1434
if self.other_tree.has_id(file_id):
1435
# OTHER changed the file
1437
if wt.supports_content_filtering():
1438
# We get the path from the working tree if it exists.
1439
# That fails though when OTHER is adding a file, so
1440
# we fall back to the other tree to find the path if
1441
# it doesn't exist locally.
1443
filter_tree_path = wt.id2path(file_id)
1444
except errors.NoSuchId:
1445
filter_tree_path = self.other_tree.id2path(file_id)
1447
# Skip the id2path lookup for older formats
1448
filter_tree_path = None
1449
transform.create_from_tree(self.tt, trans_id,
1450
self.other_tree, file_id,
1451
filter_tree_path=filter_tree_path)
1454
# OTHER deleted the file
1455
return 'delete', None
1457
raise AssertionError(
1458
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1461
def merge_contents(self, merge_hook_params):
1462
"""Fallback merge logic after user installed hooks."""
1463
# This function is used in merge hooks as the fallback instance.
1464
# Perhaps making this function and the functions it calls be a
1465
# a separate class would be better.
1466
if merge_hook_params.winner == 'other':
1467
# OTHER is a straight winner, so replace this contents with other
1468
return self._default_other_winner_merge(merge_hook_params)
1469
elif merge_hook_params.is_file_merge():
1470
# THIS and OTHER are both files, so text merge. Either
1471
# BASE is a file, or both converted to files, so at least we
1472
# have agreement that output should be a file.
1474
self.text_merge(merge_hook_params.file_id,
1475
merge_hook_params.trans_id)
1476
except errors.BinaryFile:
1477
return 'not_applicable', None
1480
return 'not_applicable', None
1482
def get_lines(self, tree, file_id):
1483
"""Return the lines in a file, or an empty list."""
1484
if tree.has_id(file_id):
1485
return tree.get_file_lines(file_id)
1489
def text_merge(self, file_id, trans_id):
1490
"""Perform a three-way text merge on a file_id"""
1491
# it's possible that we got here with base as a different type.
1492
# if so, we just want two-way text conflicts.
1493
if self.base_tree.has_id(file_id) and \
1494
self.base_tree.kind(file_id) == "file":
1495
base_lines = self.get_lines(self.base_tree, file_id)
1498
other_lines = self.get_lines(self.other_tree, file_id)
1499
this_lines = self.get_lines(self.this_tree, file_id)
1500
m3 = merge3.Merge3(base_lines, this_lines, other_lines,
1501
is_cherrypick=self.cherrypick)
1502
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1503
if self.show_base is True:
1504
base_marker = '|' * 7
1508
def iter_merge3(retval):
1509
retval["text_conflicts"] = False
1510
for line in m3.merge_lines(name_a = "TREE",
1511
name_b = "MERGE-SOURCE",
1512
name_base = "BASE-REVISION",
1513
start_marker=start_marker,
1514
base_marker=base_marker,
1515
reprocess=self.reprocess):
1516
if line.startswith(start_marker):
1517
retval["text_conflicts"] = True
1518
yield line.replace(start_marker, '<' * 7)
1522
merge3_iterator = iter_merge3(retval)
1523
self.tt.create_file(merge3_iterator, trans_id)
1524
if retval["text_conflicts"] is True:
1525
self._raw_conflicts.append(('text conflict', trans_id))
1526
name = self.tt.final_name(trans_id)
1527
parent_id = self.tt.final_parent(trans_id)
1528
file_group = self._dump_conflicts(name, parent_id, file_id,
1529
this_lines, base_lines,
1531
file_group.append(trans_id)
1533
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1534
base_lines=None, other_lines=None, set_version=False,
1536
"""Emit conflict files.
1537
If this_lines, base_lines, or other_lines are omitted, they will be
1538
determined automatically. If set_version is true, the .OTHER, .THIS
1539
or .BASE (in that order) will be created as versioned files.
1541
data = [('OTHER', self.other_tree, other_lines),
1542
('THIS', self.this_tree, this_lines)]
1544
data.append(('BASE', self.base_tree, base_lines))
1546
# We need to use the actual path in the working tree of the file here,
1547
# ignoring the conflict suffixes
1549
if wt.supports_content_filtering():
1551
filter_tree_path = wt.id2path(file_id)
1552
except errors.NoSuchId:
1553
# file has been deleted
1554
filter_tree_path = None
1556
# Skip the id2path lookup for older formats
1557
filter_tree_path = None
1561
for suffix, tree, lines in data:
1562
if tree.has_id(file_id):
1563
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1564
suffix, lines, filter_tree_path)
1565
file_group.append(trans_id)
1566
if set_version and not versioned:
1567
self.tt.version_file(file_id, trans_id)
1571
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1572
lines=None, filter_tree_path=None):
1573
"""Emit a single conflict file."""
1574
name = name + '.' + suffix
1575
trans_id = self.tt.create_path(name, parent_id)
1576
transform.create_from_tree(self.tt, trans_id, tree, file_id, lines,
1580
def merge_executable(self, file_id, file_status):
1581
"""Perform a merge on the execute bit."""
1582
executable = [self.executable(t, file_id) for t in (self.base_tree,
1583
self.other_tree, self.this_tree)]
1584
self._merge_executable(file_id, executable, file_status,
1585
resolver=self._three_way)
1587
def _merge_executable(self, file_id, executable, file_status,
1589
"""Perform a merge on the execute bit."""
1590
base_executable, other_executable, this_executable = executable
1591
if file_status == "deleted":
1593
winner = resolver(*executable)
1594
if winner == "conflict":
1595
# There must be a None in here, if we have a conflict, but we
1596
# need executability since file status was not deleted.
1597
if self.executable(self.other_tree, file_id) is None:
1601
if winner == 'this' and file_status != "modified":
1603
trans_id = self.tt.trans_id_file_id(file_id)
1604
if self.tt.final_kind(trans_id) != "file":
1606
if winner == "this":
1607
executability = this_executable
1609
if self.other_tree.has_id(file_id):
1610
executability = other_executable
1611
elif self.this_tree.has_id(file_id):
1612
executability = this_executable
1613
elif self.base_tree_has_id(file_id):
1614
executability = base_executable
1615
if executability is not None:
1616
trans_id = self.tt.trans_id_file_id(file_id)
1617
self.tt.set_executability(executability, trans_id)
1619
def cook_conflicts(self, fs_conflicts):
1620
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1621
content_conflict_file_ids = set()
1622
cooked_conflicts = transform.cook_conflicts(fs_conflicts, self.tt)
1623
fp = transform.FinalPaths(self.tt)
1624
for conflict in self._raw_conflicts:
1625
conflict_type = conflict[0]
1626
if conflict_type == 'path conflict':
1628
this_parent, this_name,
1629
other_parent, other_name) = conflict[1:]
1630
if this_parent is None or this_name is None:
1631
this_path = '<deleted>'
1633
parent_path = fp.get_path(
1634
self.tt.trans_id_file_id(this_parent))
1635
this_path = osutils.pathjoin(parent_path, this_name)
1636
if other_parent is None or other_name is None:
1637
other_path = '<deleted>'
1639
if other_parent == self.other_tree.get_root_id():
1640
# The tree transform doesn't know about the other root,
1641
# so we special case here to avoid a NoFinalPath
1645
parent_path = fp.get_path(
1646
self.tt.trans_id_file_id(other_parent))
1647
other_path = osutils.pathjoin(parent_path, other_name)
1648
c = _mod_conflicts.Conflict.factory(
1649
'path conflict', path=this_path,
1650
conflict_path=other_path,
1652
elif conflict_type == 'contents conflict':
1653
for trans_id in conflict[1]:
1654
file_id = self.tt.final_file_id(trans_id)
1655
if file_id is not None:
1657
path = fp.get_path(trans_id)
1658
for suffix in ('.BASE', '.THIS', '.OTHER'):
1659
if path.endswith(suffix):
1660
path = path[:-len(suffix)]
1662
c = _mod_conflicts.Conflict.factory(conflict_type,
1663
path=path, file_id=file_id)
1664
content_conflict_file_ids.add(file_id)
1665
elif conflict_type == 'text conflict':
1666
trans_id = conflict[1]
1667
path = fp.get_path(trans_id)
1668
file_id = self.tt.final_file_id(trans_id)
1669
c = _mod_conflicts.Conflict.factory(conflict_type,
1670
path=path, file_id=file_id)
1672
raise AssertionError('bad conflict type: %r' % (conflict,))
1673
cooked_conflicts.append(c)
1675
self.cooked_conflicts = []
1676
# We want to get rid of path conflicts when a corresponding contents
1677
# conflict exists. This can occur when one branch deletes a file while
1678
# the other renames *and* modifies it. In this case, the content
1679
# conflict is enough.
1680
for c in cooked_conflicts:
1681
if (c.typestring == 'path conflict'
1682
and c.file_id in content_conflict_file_ids):
1684
self.cooked_conflicts.append(c)
1685
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
1688
class WeaveMerger(Merge3Merger):
1689
"""Three-way tree merger, text weave merger."""
1690
supports_reprocess = True
1691
supports_show_base = False
1692
supports_reverse_cherrypick = False
1693
history_based = True
1695
def _generate_merge_plan(self, file_id, base):
1696
return self.this_tree.plan_file_merge(file_id, self.other_tree,
1699
def _merged_lines(self, file_id):
1700
"""Generate the merged lines.
1701
There is no distinction between lines that are meant to contain <<<<<<<
1705
base = self.base_tree
1708
plan = self._generate_merge_plan(file_id, base)
1709
if 'merge' in debug.debug_flags:
1711
trans_id = self.tt.trans_id_file_id(file_id)
1712
name = self.tt.final_name(trans_id) + '.plan'
1713
contents = ('%11s|%s' % l for l in plan)
1714
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1715
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1716
'>>>>>>> MERGE-SOURCE\n')
1717
lines, conflicts = textmerge.merge_lines(self.reprocess)
1719
base_lines = textmerge.base_from_plan()
1722
return lines, base_lines
1724
def text_merge(self, file_id, trans_id):
1725
"""Perform a (weave) text merge for a given file and file-id.
1726
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1727
and a conflict will be noted.
1729
lines, base_lines = self._merged_lines(file_id)
1731
# Note we're checking whether the OUTPUT is binary in this case,
1732
# because we don't want to get into weave merge guts.
1733
textfile.check_text_lines(lines)
1734
self.tt.create_file(lines, trans_id)
1735
if base_lines is not None:
1737
self._raw_conflicts.append(('text conflict', trans_id))
1738
name = self.tt.final_name(trans_id)
1739
parent_id = self.tt.final_parent(trans_id)
1740
file_group = self._dump_conflicts(name, parent_id, file_id,
1742
base_lines=base_lines)
1743
file_group.append(trans_id)
1746
class LCAMerger(WeaveMerger):
1748
def _generate_merge_plan(self, file_id, base):
1749
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1752
class Diff3Merger(Merge3Merger):
1753
"""Three-way merger using external diff3 for text merging"""
1755
def dump_file(self, temp_dir, name, tree, file_id):
1756
out_path = osutils.pathjoin(temp_dir, name)
1757
out_file = open(out_path, "wb")
1759
in_file = tree.get_file(file_id)
1760
for line in in_file:
1761
out_file.write(line)
1766
def text_merge(self, file_id, trans_id):
1767
"""Perform a diff3 merge using a specified file-id and trans-id.
1768
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1769
will be dumped, and a will be conflict noted.
1772
temp_dir = osutils.mkdtemp(prefix="bzr-")
1774
new_file = osutils.pathjoin(temp_dir, "new")
1775
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1776
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1777
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1778
status = bzrlib.patch.diff3(new_file, this, base, other)
1779
if status not in (0, 1):
1780
raise errors.BzrError("Unhandled diff3 exit code")
1781
f = open(new_file, 'rb')
1783
self.tt.create_file(f, trans_id)
1787
name = self.tt.final_name(trans_id)
1788
parent_id = self.tt.final_parent(trans_id)
1789
self._dump_conflicts(name, parent_id, file_id)
1790
self._raw_conflicts.append(('text conflict', trans_id))
1792
osutils.rmtree(temp_dir)
1795
class PathNotInTree(errors.BzrError):
1797
_fmt = """Merge-into failed because %(tree)s does not contain %(path)s."""
1799
def __init__(self, path, tree):
1800
errors.BzrError.__init__(self, path=path, tree=tree)
1803
class MergeIntoMerger(Merger):
1804
"""Merger that understands other_tree will be merged into a subdir.
1806
This also changes the Merger api so that it uses real Branch, revision_id,
1807
and RevisonTree objects, rather than using revision specs.
1810
def __init__(self, this_tree, other_branch, other_tree, target_subdir,
1811
source_subpath, other_rev_id=None):
1812
"""Create a new MergeIntoMerger object.
1814
source_subpath in other_tree will be effectively copied to
1815
target_subdir in this_tree.
1817
:param this_tree: The tree that we will be merging into.
1818
:param other_branch: The Branch we will be merging from.
1819
:param other_tree: The RevisionTree object we want to merge.
1820
:param target_subdir: The relative path where we want to merge
1821
other_tree into this_tree
1822
:param source_subpath: The relative path specifying the subtree of
1823
other_tree to merge into this_tree.
1825
# It is assumed that we are merging a tree that is not in our current
1826
# ancestry, which means we are using the "EmptyTree" as our basis.
1827
null_ancestor_tree = this_tree.branch.repository.revision_tree(
1828
_mod_revision.NULL_REVISION)
1829
super(MergeIntoMerger, self).__init__(
1830
this_branch=this_tree.branch,
1831
this_tree=this_tree,
1832
other_tree=other_tree,
1833
base_tree=null_ancestor_tree,
1835
self._target_subdir = target_subdir
1836
self._source_subpath = source_subpath
1837
self.other_branch = other_branch
1838
if other_rev_id is None:
1839
other_rev_id = other_tree.get_revision_id()
1840
self.other_rev_id = self.other_basis = other_rev_id
1841
self.base_is_ancestor = True
1842
self.backup_files = True
1843
self.merge_type = Merge3Merger
1844
self.show_base = False
1845
self.reprocess = False
1846
self.interesting_ids = None
1847
self.merge_type = _MergeTypeParameterizer(MergeIntoMergeType,
1848
target_subdir=self._target_subdir,
1849
source_subpath=self._source_subpath)
1850
if self._source_subpath != '':
1851
# If this isn't a partial merge make sure the revisions will be
1853
self._maybe_fetch(self.other_branch, self.this_branch,
1856
def set_pending(self):
1857
if self._source_subpath != '':
1859
Merger.set_pending(self)
1862
class _MergeTypeParameterizer(object):
1863
"""Wrap a merge-type class to provide extra parameters.
1865
This is hack used by MergeIntoMerger to pass some extra parameters to its
1866
merge_type. Merger.do_merge() sets up its own set of parameters to pass to
1867
the 'merge_type' member. It is difficult override do_merge without
1868
re-writing the whole thing, so instead we create a wrapper which will pass
1869
the extra parameters.
1872
def __init__(self, merge_type, **kwargs):
1873
self._extra_kwargs = kwargs
1874
self._merge_type = merge_type
1876
def __call__(self, *args, **kwargs):
1877
kwargs.update(self._extra_kwargs)
1878
return self._merge_type(*args, **kwargs)
1880
def __getattr__(self, name):
1881
return getattr(self._merge_type, name)
1884
class MergeIntoMergeType(Merge3Merger):
1885
"""Merger that incorporates a tree (or part of a tree) into another."""
1887
def __init__(self, *args, **kwargs):
1888
"""Initialize the merger object.
1890
:param args: See Merge3Merger.__init__'s args.
1891
:param kwargs: See Merge3Merger.__init__'s keyword args, except for
1892
source_subpath and target_subdir.
1893
:keyword source_subpath: The relative path specifying the subtree of
1894
other_tree to merge into this_tree.
1895
:keyword target_subdir: The relative path where we want to merge
1896
other_tree into this_tree
1898
# All of the interesting work happens during Merge3Merger.__init__(),
1899
# so we have have to hack in to get our extra parameters set.
1900
self._source_subpath = kwargs.pop('source_subpath')
1901
self._target_subdir = kwargs.pop('target_subdir')
1902
super(MergeIntoMergeType, self).__init__(*args, **kwargs)
1904
def _compute_transform(self):
1905
child_pb = ui.ui_factory.nested_progress_bar()
1907
entries = self._entries_to_incorporate()
1908
entries = list(entries)
1909
for num, (entry, parent_id) in enumerate(entries):
1910
child_pb.update(gettext('Preparing file merge'), num, len(entries))
1911
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1912
trans_id = transform.new_by_entry(self.tt, entry,
1913
parent_trans_id, self.other_tree)
1916
self._finish_computing_transform()
1918
def _entries_to_incorporate(self):
1919
"""Yields pairs of (inventory_entry, new_parent)."""
1920
other_inv = self.other_tree.inventory
1921
subdir_id = other_inv.path2id(self._source_subpath)
1922
if subdir_id is None:
1923
# XXX: The error would be clearer if it gave the URL of the source
1924
# branch, but we don't have a reference to that here.
1925
raise PathNotInTree(self._source_subpath, "Source tree")
1926
subdir = other_inv[subdir_id]
1927
parent_in_target = osutils.dirname(self._target_subdir)
1928
target_id = self.this_tree.inventory.path2id(parent_in_target)
1929
if target_id is None:
1930
raise PathNotInTree(self._target_subdir, "Target tree")
1931
name_in_target = osutils.basename(self._target_subdir)
1932
merge_into_root = subdir.copy()
1933
merge_into_root.name = name_in_target
1934
if self.this_tree.inventory.has_id(merge_into_root.file_id):
1935
# Give the root a new file-id.
1936
# This can happen fairly easily if the directory we are
1937
# incorporating is the root, and both trees have 'TREE_ROOT' as
1938
# their root_id. Users will expect this to Just Work, so we
1939
# change the file-id here.
1940
# Non-root file-ids could potentially conflict too. That's really
1941
# an edge case, so we don't do anything special for those. We let
1942
# them cause conflicts.
1943
merge_into_root.file_id = generate_ids.gen_file_id(name_in_target)
1944
yield (merge_into_root, target_id)
1945
if subdir.kind != 'directory':
1946
# No children, so we are done.
1948
for ignored_path, entry in other_inv.iter_entries_by_dir(subdir_id):
1949
parent_id = entry.parent_id
1950
if parent_id == subdir.file_id:
1951
# The root's parent ID has changed, so make sure children of
1952
# the root refer to the new ID.
1953
parent_id = merge_into_root.file_id
1954
yield (entry, parent_id)
1957
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1959
merge_type=Merge3Merger,
1960
interesting_ids=None,
1964
interesting_files=None,
1967
change_reporter=None):
1968
"""Primary interface for merging.
1970
Typical use is probably::
1972
merge_inner(branch, branch.get_revision_tree(other_revision),
1973
branch.get_revision_tree(base_revision))
1975
if this_tree is None:
1976
raise errors.BzrError("bzrlib.merge.merge_inner requires a this_tree "
1978
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1979
pb=pb, change_reporter=change_reporter)
1980
merger.backup_files = backup_files
1981
merger.merge_type = merge_type
1982
merger.interesting_ids = interesting_ids
1983
merger.ignore_zero = ignore_zero
1984
if interesting_files:
1986
raise ValueError('Only supply interesting_ids'
1987
' or interesting_files')
1988
merger.interesting_files = interesting_files
1989
merger.show_base = show_base
1990
merger.reprocess = reprocess
1991
merger.other_rev_id = other_rev_id
1992
merger.other_basis = other_rev_id
1993
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1994
if get_revision_id is None:
1995
get_revision_id = base_tree.last_revision
1996
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
1997
merger.set_base_revision(get_revision_id(), this_branch)
1998
return merger.do_merge()
2000
def get_merge_type_registry():
2001
"""Merge type registry is in bzrlib.option to avoid circular imports.
2003
This method provides a sanctioned way to retrieve it.
2005
from bzrlib import option
2006
return option._merge_type_registry
2009
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
2010
def status_a(revision, text):
2011
if revision in ancestors_b:
2012
return 'killed-b', text
2014
return 'new-a', text
2016
def status_b(revision, text):
2017
if revision in ancestors_a:
2018
return 'killed-a', text
2020
return 'new-b', text
2022
plain_a = [t for (a, t) in annotated_a]
2023
plain_b = [t for (a, t) in annotated_b]
2024
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
2025
blocks = matcher.get_matching_blocks()
2028
for ai, bi, l in blocks:
2029
# process all mismatched sections
2030
# (last mismatched section is handled because blocks always
2031
# includes a 0-length last block)
2032
for revision, text in annotated_a[a_cur:ai]:
2033
yield status_a(revision, text)
2034
for revision, text in annotated_b[b_cur:bi]:
2035
yield status_b(revision, text)
2036
# and now the matched section
2039
for text_a in plain_a[ai:a_cur]:
2040
yield "unchanged", text_a
2043
class _PlanMergeBase(object):
2045
def __init__(self, a_rev, b_rev, vf, key_prefix):
2048
:param a_rev: Revision-id of one revision to merge
2049
:param b_rev: Revision-id of the other revision to merge
2050
:param vf: A VersionedFiles containing both revisions
2051
:param key_prefix: A prefix for accessing keys in vf, typically
2057
self._last_lines = None
2058
self._last_lines_revision_id = None
2059
self._cached_matching_blocks = {}
2060
self._key_prefix = key_prefix
2061
self._precache_tip_lines()
2063
def _precache_tip_lines(self):
2064
lines = self.get_lines([self.a_rev, self.b_rev])
2065
self.lines_a = lines[self.a_rev]
2066
self.lines_b = lines[self.b_rev]
2068
def get_lines(self, revisions):
2069
"""Get lines for revisions from the backing VersionedFiles.
2071
:raises RevisionNotPresent: on absent texts.
2073
keys = [(self._key_prefix + (rev,)) for rev in revisions]
2075
for record in self.vf.get_record_stream(keys, 'unordered', True):
2076
if record.storage_kind == 'absent':
2077
raise errors.RevisionNotPresent(record.key, self.vf)
2078
result[record.key[-1]] = osutils.chunks_to_lines(
2079
record.get_bytes_as('chunked'))
2082
def plan_merge(self):
2083
"""Generate a 'plan' for merging the two revisions.
2085
This involves comparing their texts and determining the cause of
2086
differences. If text A has a line and text B does not, then either the
2087
line was added to text A, or it was deleted from B. Once the causes
2088
are combined, they are written out in the format described in
2089
VersionedFile.plan_merge
2091
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
2092
unique_a, unique_b = self._unique_lines(blocks)
2093
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
2094
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
2095
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
2097
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
2100
for i, j, n in blocks:
2101
for a_index in range(last_i, i):
2102
if a_index in new_a:
2103
if a_index in killed_b:
2104
yield 'conflicted-a', self.lines_a[a_index]
2106
yield 'new-a', self.lines_a[a_index]
2108
yield 'killed-b', self.lines_a[a_index]
2109
for b_index in range(last_j, j):
2110
if b_index in new_b:
2111
if b_index in killed_a:
2112
yield 'conflicted-b', self.lines_b[b_index]
2114
yield 'new-b', self.lines_b[b_index]
2116
yield 'killed-a', self.lines_b[b_index]
2117
# handle common lines
2118
for a_index in range(i, i+n):
2119
yield 'unchanged', self.lines_a[a_index]
2123
def _get_matching_blocks(self, left_revision, right_revision):
2124
"""Return a description of which sections of two revisions match.
2126
See SequenceMatcher.get_matching_blocks
2128
cached = self._cached_matching_blocks.get((left_revision,
2130
if cached is not None:
2132
if self._last_lines_revision_id == left_revision:
2133
left_lines = self._last_lines
2134
right_lines = self.get_lines([right_revision])[right_revision]
2136
lines = self.get_lines([left_revision, right_revision])
2137
left_lines = lines[left_revision]
2138
right_lines = lines[right_revision]
2139
self._last_lines = right_lines
2140
self._last_lines_revision_id = right_revision
2141
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
2143
return matcher.get_matching_blocks()
2145
def _unique_lines(self, matching_blocks):
2146
"""Analyse matching_blocks to determine which lines are unique
2148
:return: a tuple of (unique_left, unique_right), where the values are
2149
sets of line numbers of unique lines.
2155
for i, j, n in matching_blocks:
2156
unique_left.extend(range(last_i, i))
2157
unique_right.extend(range(last_j, j))
2160
return unique_left, unique_right
2163
def _subtract_plans(old_plan, new_plan):
2164
"""Remove changes from new_plan that came from old_plan.
2166
It is assumed that the difference between the old_plan and new_plan
2167
is their choice of 'b' text.
2169
All lines from new_plan that differ from old_plan are emitted
2170
verbatim. All lines from new_plan that match old_plan but are
2171
not about the 'b' revision are emitted verbatim.
2173
Lines that match and are about the 'b' revision are the lines we
2174
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
2175
is skipped entirely.
2177
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
2180
for i, j, n in matcher.get_matching_blocks():
2181
for jj in range(last_j, j):
2183
for jj in range(j, j+n):
2184
plan_line = new_plan[jj]
2185
if plan_line[0] == 'new-b':
2187
elif plan_line[0] == 'killed-b':
2188
yield 'unchanged', plan_line[1]
2194
class _PlanMerge(_PlanMergeBase):
2195
"""Plan an annotate merge using on-the-fly annotation"""
2197
def __init__(self, a_rev, b_rev, vf, key_prefix):
2198
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
2199
self.a_key = self._key_prefix + (self.a_rev,)
2200
self.b_key = self._key_prefix + (self.b_rev,)
2201
self.graph = _mod_graph.Graph(self.vf)
2202
heads = self.graph.heads((self.a_key, self.b_key))
2204
# one side dominates, so we can just return its values, yay for
2206
# Ideally we would know that before we get this far
2207
self._head_key = heads.pop()
2208
if self._head_key == self.a_key:
2212
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2213
self._head_key[-1], other)
2216
self._head_key = None
2219
def _precache_tip_lines(self):
2220
# Turn this into a no-op, because we will do this later
2223
def _find_recursive_lcas(self):
2224
"""Find all the ancestors back to a unique lca"""
2225
cur_ancestors = (self.a_key, self.b_key)
2226
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2227
# rather than a key tuple. We will just map that directly to no common
2231
next_lcas = self.graph.find_lca(*cur_ancestors)
2232
# Map a plain NULL_REVISION to a simple no-ancestors
2233
if next_lcas == set([_mod_revision.NULL_REVISION]):
2235
# Order the lca's based on when they were merged into the tip
2236
# While the actual merge portion of weave merge uses a set() of
2237
# active revisions, the order of insertion *does* effect the
2238
# implicit ordering of the texts.
2239
for rev_key in cur_ancestors:
2240
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2242
parent_map[rev_key] = ordered_parents
2243
if len(next_lcas) == 0:
2245
elif len(next_lcas) == 1:
2246
parent_map[list(next_lcas)[0]] = ()
2248
elif len(next_lcas) > 2:
2249
# More than 2 lca's, fall back to grabbing all nodes between
2250
# this and the unique lca.
2251
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2253
self.a_key, self.b_key, cur_ancestors)
2254
cur_lcas = next_lcas
2255
while len(cur_lcas) > 1:
2256
cur_lcas = self.graph.find_lca(*cur_lcas)
2257
if len(cur_lcas) == 0:
2258
# No common base to find, use the full ancestry
2261
unique_lca = list(cur_lcas)[0]
2262
if unique_lca == _mod_revision.NULL_REVISION:
2263
# find_lca will return a plain 'NULL_REVISION' rather
2264
# than a key tuple when there is no common ancestor, we
2265
# prefer to just use None, because it doesn't confuse
2266
# _get_interesting_texts()
2268
parent_map.update(self._find_unique_parents(next_lcas,
2271
cur_ancestors = next_lcas
2274
def _find_unique_parents(self, tip_keys, base_key):
2275
"""Find ancestors of tip that aren't ancestors of base.
2277
:param tip_keys: Nodes that are interesting
2278
:param base_key: Cull all ancestors of this node
2279
:return: The parent map for all revisions between tip_keys and
2280
base_key. base_key will be included. References to nodes outside of
2281
the ancestor set will also be removed.
2283
# TODO: this would be simpler if find_unique_ancestors took a list
2284
# instead of a single tip, internally it supports it, but it
2285
# isn't a "backwards compatible" api change.
2286
if base_key is None:
2287
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2288
# We remove NULL_REVISION because it isn't a proper tuple key, and
2289
# thus confuses things like _get_interesting_texts, and our logic
2290
# to add the texts into the memory weave.
2291
if _mod_revision.NULL_REVISION in parent_map:
2292
parent_map.pop(_mod_revision.NULL_REVISION)
2295
for tip in tip_keys:
2297
self.graph.find_unique_ancestors(tip, [base_key]))
2298
parent_map = self.graph.get_parent_map(interesting)
2299
parent_map[base_key] = ()
2300
culled_parent_map, child_map, tails = self._remove_external_references(
2302
# Remove all the tails but base_key
2303
if base_key is not None:
2304
tails.remove(base_key)
2305
self._prune_tails(culled_parent_map, child_map, tails)
2306
# Now remove all the uninteresting 'linear' regions
2307
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2311
def _remove_external_references(parent_map):
2312
"""Remove references that go outside of the parent map.
2314
:param parent_map: Something returned from Graph.get_parent_map(keys)
2315
:return: (filtered_parent_map, child_map, tails)
2316
filtered_parent_map is parent_map without external references
2317
child_map is the {parent_key: [child_keys]} mapping
2318
tails is a list of nodes that do not have any parents in the map
2320
# TODO: The basic effect of this function seems more generic than
2321
# _PlanMerge. But the specific details of building a child_map,
2322
# and computing tails seems very specific to _PlanMerge.
2323
# Still, should this be in Graph land?
2324
filtered_parent_map = {}
2327
for key, parent_keys in parent_map.iteritems():
2328
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2329
if not culled_parent_keys:
2331
for parent_key in culled_parent_keys:
2332
child_map.setdefault(parent_key, []).append(key)
2333
# TODO: Do we want to do this, it adds overhead for every node,
2334
# just to say that the node has no children
2335
child_map.setdefault(key, [])
2336
filtered_parent_map[key] = culled_parent_keys
2337
return filtered_parent_map, child_map, tails
2340
def _prune_tails(parent_map, child_map, tails_to_remove):
2341
"""Remove tails from the parent map.
2343
This will remove the supplied revisions until no more children have 0
2346
:param parent_map: A dict of {child: [parents]}, this dictionary will
2347
be modified in place.
2348
:param tails_to_remove: A list of tips that should be removed,
2349
this list will be consumed
2350
:param child_map: The reverse dict of parent_map ({parent: [children]})
2351
this dict will be modified
2352
:return: None, parent_map will be modified in place.
2354
while tails_to_remove:
2355
next = tails_to_remove.pop()
2356
parent_map.pop(next)
2357
children = child_map.pop(next)
2358
for child in children:
2359
child_parents = parent_map[child]
2360
child_parents.remove(next)
2361
if len(child_parents) == 0:
2362
tails_to_remove.append(child)
2364
def _get_interesting_texts(self, parent_map):
2365
"""Return a dict of texts we are interested in.
2367
Note that the input is in key tuples, but the output is in plain
2370
:param parent_map: The output from _find_recursive_lcas
2371
:return: A dict of {'revision_id':lines} as returned by
2372
_PlanMergeBase.get_lines()
2374
all_revision_keys = set(parent_map)
2375
all_revision_keys.add(self.a_key)
2376
all_revision_keys.add(self.b_key)
2378
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2379
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2382
def _build_weave(self):
2383
from bzrlib import weave
2384
self._weave = weave.Weave(weave_name='in_memory_weave',
2385
allow_reserved=True)
2386
parent_map = self._find_recursive_lcas()
2388
all_texts = self._get_interesting_texts(parent_map)
2390
# Note: Unfortunately, the order given by topo_sort will effect the
2391
# ordering resolution in the output. Specifically, if you add A then B,
2392
# then in the output text A lines will show up before B lines. And, of
2393
# course, topo_sort doesn't guarantee any real ordering.
2394
# So we use merge_sort, and add a fake node on the tip.
2395
# This ensures that left-hand parents will always be inserted into the
2396
# weave before right-hand parents.
2397
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2398
parent_map[tip_key] = (self.a_key, self.b_key)
2400
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2404
# for key in tsort.topo_sort(parent_map):
2405
parent_keys = parent_map[key]
2406
revision_id = key[-1]
2407
parent_ids = [k[-1] for k in parent_keys]
2408
self._weave.add_lines(revision_id, parent_ids,
2409
all_texts[revision_id])
2411
def plan_merge(self):
2412
"""Generate a 'plan' for merging the two revisions.
2414
This involves comparing their texts and determining the cause of
2415
differences. If text A has a line and text B does not, then either the
2416
line was added to text A, or it was deleted from B. Once the causes
2417
are combined, they are written out in the format described in
2418
VersionedFile.plan_merge
2420
if self._head_key is not None: # There was a single head
2421
if self._head_key == self.a_key:
2424
if self._head_key != self.b_key:
2425
raise AssertionError('There was an invalid head: %s != %s'
2426
% (self.b_key, self._head_key))
2428
head_rev = self._head_key[-1]
2429
lines = self.get_lines([head_rev])[head_rev]
2430
return ((plan, line) for line in lines)
2431
return self._weave.plan_merge(self.a_rev, self.b_rev)
2434
class _PlanLCAMerge(_PlanMergeBase):
2436
This merge algorithm differs from _PlanMerge in that:
2438
1. comparisons are done against LCAs only
2439
2. cases where a contested line is new versus one LCA but old versus
2440
another are marked as conflicts, by emitting the line as conflicted-a
2443
This is faster, and hopefully produces more useful output.
2446
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2447
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2448
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2451
if lca == _mod_revision.NULL_REVISION:
2454
self.lcas.add(lca[-1])
2455
for lca in self.lcas:
2456
if _mod_revision.is_null(lca):
2459
lca_lines = self.get_lines([lca])[lca]
2460
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2462
blocks = list(matcher.get_matching_blocks())
2463
self._cached_matching_blocks[(a_rev, lca)] = blocks
2464
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2466
blocks = list(matcher.get_matching_blocks())
2467
self._cached_matching_blocks[(b_rev, lca)] = blocks
2469
def _determine_status(self, revision_id, unique_line_numbers):
2470
"""Determines the status unique lines versus all lcas.
2472
Basically, determines why the line is unique to this revision.
2474
A line may be determined new, killed, or both.
2476
If a line is determined new, that means it was not present in at least
2477
one LCA, and is not present in the other merge revision.
2479
If a line is determined killed, that means the line was present in
2482
If a line is killed and new, this indicates that the two merge
2483
revisions contain differing conflict resolutions.
2485
:param revision_id: The id of the revision in which the lines are
2487
:param unique_line_numbers: The line numbers of unique lines.
2488
:return: a tuple of (new_this, killed_other)
2492
unique_line_numbers = set(unique_line_numbers)
2493
for lca in self.lcas:
2494
blocks = self._get_matching_blocks(revision_id, lca)
2495
unique_vs_lca, _ignored = self._unique_lines(blocks)
2496
new.update(unique_line_numbers.intersection(unique_vs_lca))
2497
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
