1
# Copyright (C) 2005-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
from __future__ import absolute_import
21
from bzrlib.lazy_import import lazy_import
22
lazy_import(globals(), """
24
branch as _mod_branch,
26
conflicts as _mod_conflicts,
33
revision as _mod_revision,
43
from bzrlib.i18n import gettext
51
from bzrlib.symbol_versioning import (
55
# TODO: Report back as changes are merged in
58
def transform_tree(from_tree, to_tree, interesting_ids=None):
59
from_tree.lock_tree_write()
60
operation = cleanup.OperationWithCleanups(merge_inner)
61
operation.add_cleanup(from_tree.unlock)
62
operation.run_simple(from_tree.branch, to_tree, from_tree,
63
ignore_zero=True, interesting_ids=interesting_ids, this_tree=from_tree)
66
class MergeHooks(hooks.Hooks):
69
hooks.Hooks.__init__(self, "bzrlib.merge", "Merger.hooks")
70
self.add_hook('merge_file_content',
71
"Called with a bzrlib.merge.Merger object to create a per file "
72
"merge object when starting a merge. "
73
"Should return either None or a subclass of "
74
"``bzrlib.merge.AbstractPerFileMerger``. "
75
"Such objects will then be called per file "
76
"that needs to be merged (including when one "
77
"side has deleted the file and the other has changed it). "
78
"See the AbstractPerFileMerger API docs for details on how it is "
83
class AbstractPerFileMerger(object):
84
"""PerFileMerger objects are used by plugins extending merge for bzrlib.
86
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
88
:ivar merger: The Merge3Merger performing the merge.
91
def __init__(self, merger):
92
"""Create a PerFileMerger for use with merger."""
95
def merge_contents(self, merge_params):
96
"""Attempt to merge the contents of a single file.
98
:param merge_params: A bzrlib.merge.MergeHookParams
99
:return: A tuple of (status, chunks), where status is one of
100
'not_applicable', 'success', 'conflicted', or 'delete'. If status
101
is 'success' or 'conflicted', then chunks should be an iterable of
102
strings for the new file contents.
104
return ('not applicable', None)
107
class PerFileMerger(AbstractPerFileMerger):
108
"""Merge individual files when self.file_matches returns True.
110
This class is intended to be subclassed. The file_matches and
111
merge_matching methods should be overridden with concrete implementations.
114
def file_matches(self, params):
115
"""Return True if merge_matching should be called on this file.
117
Only called with merges of plain files with no clear winner.
119
Subclasses must override this.
121
raise NotImplementedError(self.file_matches)
123
def get_filename(self, params, tree):
124
"""Lookup the filename (i.e. basename, not path), given a Tree (e.g.
125
self.merger.this_tree) and a MergeHookParams.
127
return osutils.basename(tree.id2path(params.file_id))
129
def get_filepath(self, params, tree):
130
"""Calculate the path to the file in a tree.
132
:param params: A MergeHookParams describing the file to merge
133
:param tree: a Tree, e.g. self.merger.this_tree.
135
return tree.id2path(params.file_id)
137
def merge_contents(self, params):
138
"""Merge the contents of a single file."""
139
# Check whether this custom merge logic should be used.
141
# OTHER is a straight winner, rely on default merge.
142
params.winner == 'other' or
143
# THIS and OTHER aren't both files.
144
not params.is_file_merge() or
145
# The filename doesn't match
146
not self.file_matches(params)):
147
return 'not_applicable', None
148
return self.merge_matching(params)
150
def merge_matching(self, params):
151
"""Merge the contents of a single file that has matched the criteria
152
in PerFileMerger.merge_contents (is a conflict, is a file,
153
self.file_matches is True).
155
Subclasses must override this.
157
raise NotImplementedError(self.merge_matching)
160
class ConfigurableFileMerger(PerFileMerger):
161
"""Merge individual files when configured via a .conf file.
163
This is a base class for concrete custom file merging logic. Concrete
164
classes should implement ``merge_text``.
166
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
168
:ivar affected_files: The configured file paths to merge.
170
:cvar name_prefix: The prefix to use when looking up configuration
171
details. <name_prefix>_merge_files describes the files targeted by the
174
:cvar default_files: The default file paths to merge when no configuration
181
def __init__(self, merger):
182
super(ConfigurableFileMerger, self).__init__(merger)
183
self.affected_files = None
184
self.default_files = self.__class__.default_files or []
185
self.name_prefix = self.__class__.name_prefix
186
if self.name_prefix is None:
187
raise ValueError("name_prefix must be set.")
189
def file_matches(self, params):
190
"""Check whether the file should call the merge hook.
192
<name_prefix>_merge_files configuration variable is a list of files
193
that should use the hook.
195
affected_files = self.affected_files
196
if affected_files is None:
197
config = self.merger.this_branch.get_config()
198
# Until bzr provides a better policy for caching the config, we
199
# just add the part we're interested in to the params to avoid
200
# reading the config files repeatedly (bazaar.conf, location.conf,
202
config_key = self.name_prefix + '_merge_files'
203
affected_files = config.get_user_option_as_list(config_key)
204
if affected_files is None:
205
# If nothing was specified in the config, use the default.
206
affected_files = self.default_files
207
self.affected_files = affected_files
209
filepath = self.get_filepath(params, self.merger.this_tree)
210
if filepath in affected_files:
214
def merge_matching(self, params):
215
return self.merge_text(params)
217
def merge_text(self, params):
218
"""Merge the byte contents of a single file.
220
This is called after checking that the merge should be performed in
221
merge_contents, and it should behave as per
222
``bzrlib.merge.AbstractPerFileMerger.merge_contents``.
224
raise NotImplementedError(self.merge_text)
227
class MergeHookParams(object):
228
"""Object holding parameters passed to merge_file_content hooks.
230
There are some fields hooks can access:
232
:ivar file_id: the file ID of the file being merged
233
:ivar trans_id: the transform ID for the merge of this file
234
:ivar this_kind: kind of file_id in 'this' tree
235
:ivar other_kind: kind of file_id in 'other' tree
236
:ivar winner: one of 'this', 'other', 'conflict'
239
def __init__(self, merger, file_id, trans_id, this_kind, other_kind,
241
self._merger = merger
242
self.file_id = file_id
243
self.trans_id = trans_id
244
self.this_kind = this_kind
245
self.other_kind = other_kind
248
def is_file_merge(self):
249
"""True if this_kind and other_kind are both 'file'."""
250
return self.this_kind == 'file' and self.other_kind == 'file'
252
@decorators.cachedproperty
253
def base_lines(self):
254
"""The lines of the 'base' version of the file."""
255
return self._merger.get_lines(self._merger.base_tree, self.file_id)
257
@decorators.cachedproperty
258
def this_lines(self):
259
"""The lines of the 'this' version of the file."""
260
return self._merger.get_lines(self._merger.this_tree, self.file_id)
262
@decorators.cachedproperty
263
def other_lines(self):
264
"""The lines of the 'other' version of the file."""
265
return self._merger.get_lines(self._merger.other_tree, self.file_id)
268
class Merger(object):
272
def __init__(self, this_branch, other_tree=None, base_tree=None,
273
this_tree=None, pb=None, change_reporter=None,
274
recurse='down', revision_graph=None):
275
object.__init__(self)
276
self.this_branch = this_branch
277
self.this_basis = _mod_revision.ensure_null(
278
this_branch.last_revision())
279
self.this_rev_id = None
280
self.this_tree = this_tree
281
self.this_revision_tree = None
282
self.this_basis_tree = None
283
self.other_tree = other_tree
284
self.other_branch = None
285
self.base_tree = base_tree
286
self.ignore_zero = False
287
self.backup_files = False
288
self.interesting_ids = None
289
self.interesting_files = None
290
self.show_base = False
291
self.reprocess = False
293
warnings.warn("pb parameter to Merger() is deprecated and ignored")
295
self.recurse = recurse
296
self.change_reporter = change_reporter
297
self._cached_trees = {}
298
self._revision_graph = revision_graph
299
self._base_is_ancestor = None
300
self._base_is_other_ancestor = None
301
self._is_criss_cross = None
302
self._lca_trees = None
304
def cache_trees_with_revision_ids(self, trees):
305
"""Cache any tree in trees if it has a revision_id."""
306
for maybe_tree in trees:
307
if maybe_tree is None:
310
rev_id = maybe_tree.get_revision_id()
311
except AttributeError:
313
self._cached_trees[rev_id] = maybe_tree
316
def revision_graph(self):
317
if self._revision_graph is None:
318
self._revision_graph = self.this_branch.repository.get_graph()
319
return self._revision_graph
321
def _set_base_is_ancestor(self, value):
322
self._base_is_ancestor = value
324
def _get_base_is_ancestor(self):
325
if self._base_is_ancestor is None:
326
self._base_is_ancestor = self.revision_graph.is_ancestor(
327
self.base_rev_id, self.this_basis)
328
return self._base_is_ancestor
330
base_is_ancestor = property(_get_base_is_ancestor, _set_base_is_ancestor)
332
def _set_base_is_other_ancestor(self, value):
333
self._base_is_other_ancestor = value
335
def _get_base_is_other_ancestor(self):
336
if self._base_is_other_ancestor is None:
337
if self.other_basis is None:
339
self._base_is_other_ancestor = self.revision_graph.is_ancestor(
340
self.base_rev_id, self.other_basis)
341
return self._base_is_other_ancestor
343
base_is_other_ancestor = property(_get_base_is_other_ancestor,
344
_set_base_is_other_ancestor)
347
def from_uncommitted(tree, other_tree, pb=None, base_tree=None):
348
"""Return a Merger for uncommitted changes in other_tree.
350
:param tree: The tree to merge into
351
:param other_tree: The tree to get uncommitted changes from
352
:param pb: A progress indicator
353
:param base_tree: The basis to use for the merge. If unspecified,
354
other_tree.basis_tree() will be used.
356
if base_tree is None:
357
base_tree = other_tree.basis_tree()
358
merger = Merger(tree.branch, other_tree, base_tree, tree, pb)
359
merger.base_rev_id = merger.base_tree.get_revision_id()
360
merger.other_rev_id = None
361
merger.other_basis = merger.base_rev_id
365
def from_mergeable(klass, tree, mergeable, pb):
366
"""Return a Merger for a bundle or merge directive.
368
:param tree: The tree to merge changes into
369
:param mergeable: A merge directive or bundle
370
:param pb: A progress indicator
372
mergeable.install_revisions(tree.branch.repository)
373
base_revision_id, other_revision_id, verified =\
374
mergeable.get_merge_request(tree.branch.repository)
375
revision_graph = tree.branch.repository.get_graph()
376
if base_revision_id is not None:
377
if (base_revision_id != _mod_revision.NULL_REVISION and
378
revision_graph.is_ancestor(
379
base_revision_id, tree.branch.last_revision())):
380
base_revision_id = None
382
trace.warning('Performing cherrypick')
383
merger = klass.from_revision_ids(pb, tree, other_revision_id,
384
base_revision_id, revision_graph=
386
return merger, verified
389
def from_revision_ids(pb, tree, other, base=None, other_branch=None,
390
base_branch=None, revision_graph=None,
392
"""Return a Merger for revision-ids.
394
:param pb: A progress indicator
395
:param tree: The tree to merge changes into
396
:param other: The revision-id to use as OTHER
397
:param base: The revision-id to use as BASE. If not specified, will
399
:param other_branch: A branch containing the other revision-id. If
400
not supplied, tree.branch is used.
401
:param base_branch: A branch containing the base revision-id. If
402
not supplied, other_branch or tree.branch will be used.
403
:param revision_graph: If you have a revision_graph precomputed, pass
404
it in, otherwise it will be created for you.
405
:param tree_branch: The branch associated with tree. If not supplied,
406
tree.branch will be used.
408
if tree_branch is None:
409
tree_branch = tree.branch
410
merger = Merger(tree_branch, this_tree=tree, pb=pb,
411
revision_graph=revision_graph)
412
if other_branch is None:
413
other_branch = tree.branch
414
merger.set_other_revision(other, other_branch)
418
if base_branch is None:
419
base_branch = other_branch
420
merger.set_base_revision(base, base_branch)
423
def revision_tree(self, revision_id, branch=None):
424
if revision_id not in self._cached_trees:
426
branch = self.this_branch
428
tree = self.this_tree.revision_tree(revision_id)
429
except errors.NoSuchRevisionInTree:
430
tree = branch.repository.revision_tree(revision_id)
431
self._cached_trees[revision_id] = tree
432
return self._cached_trees[revision_id]
434
def _get_tree(self, treespec, possible_transports=None):
435
location, revno = treespec
437
tree = workingtree.WorkingTree.open_containing(location)[0]
438
return tree.branch, tree
439
branch = _mod_branch.Branch.open_containing(
440
location, possible_transports)[0]
442
revision_id = branch.last_revision()
444
revision_id = branch.get_rev_id(revno)
445
revision_id = _mod_revision.ensure_null(revision_id)
446
return branch, self.revision_tree(revision_id, branch)
448
@deprecated_method(deprecated_in((2, 1, 0)))
449
def ensure_revision_trees(self):
450
if self.this_revision_tree is None:
451
self.this_basis_tree = self.revision_tree(self.this_basis)
452
if self.this_basis == self.this_rev_id:
453
self.this_revision_tree = self.this_basis_tree
455
if self.other_rev_id is None:
456
other_basis_tree = self.revision_tree(self.other_basis)
457
if other_basis_tree.has_changes(self.other_tree):
458
raise errors.WorkingTreeNotRevision(self.this_tree)
459
other_rev_id = self.other_basis
460
self.other_tree = other_basis_tree
462
@deprecated_method(deprecated_in((2, 1, 0)))
463
def file_revisions(self, file_id):
464
self.ensure_revision_trees()
465
if self.this_rev_id is None:
466
if self.this_basis_tree.get_file_sha1(file_id) != \
467
self.this_tree.get_file_sha1(file_id):
468
raise errors.WorkingTreeNotRevision(self.this_tree)
470
trees = (self.this_basis_tree, self.other_tree)
471
return [tree.get_file_revision(file_id) for tree in trees]
473
@deprecated_method(deprecated_in((2, 1, 0)))
474
def check_basis(self, check_clean, require_commits=True):
475
if self.this_basis is None and require_commits is True:
476
raise errors.BzrCommandError(
477
"This branch has no commits."
478
" (perhaps you would prefer 'bzr pull')")
481
if self.this_basis != self.this_rev_id:
482
raise errors.UncommittedChanges(self.this_tree)
484
@deprecated_method(deprecated_in((2, 1, 0)))
485
def compare_basis(self):
487
basis_tree = self.revision_tree(self.this_tree.last_revision())
488
except errors.NoSuchRevision:
489
basis_tree = self.this_tree.basis_tree()
490
if not self.this_tree.has_changes(basis_tree):
491
self.this_rev_id = self.this_basis
493
def set_interesting_files(self, file_list):
494
self.interesting_files = file_list
496
def set_pending(self):
497
if (not self.base_is_ancestor or not self.base_is_other_ancestor
498
or self.other_rev_id is None):
502
def _add_parent(self):
503
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
504
new_parent_trees = []
505
operation = cleanup.OperationWithCleanups(
506
self.this_tree.set_parent_trees)
507
for revision_id in new_parents:
509
tree = self.revision_tree(revision_id)
510
except errors.NoSuchRevision:
514
operation.add_cleanup(tree.unlock)
515
new_parent_trees.append((revision_id, tree))
516
operation.run_simple(new_parent_trees, allow_leftmost_as_ghost=True)
518
def set_other(self, other_revision, possible_transports=None):
519
"""Set the revision and tree to merge from.
521
This sets the other_tree, other_rev_id, other_basis attributes.
523
:param other_revision: The [path, revision] list to merge from.
525
self.other_branch, self.other_tree = self._get_tree(other_revision,
527
if other_revision[1] == -1:
528
self.other_rev_id = _mod_revision.ensure_null(
529
self.other_branch.last_revision())
530
if _mod_revision.is_null(self.other_rev_id):
531
raise errors.NoCommits(self.other_branch)
532
self.other_basis = self.other_rev_id
533
elif other_revision[1] is not None:
534
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
535
self.other_basis = self.other_rev_id
537
self.other_rev_id = None
538
self.other_basis = self.other_branch.last_revision()
539
if self.other_basis is None:
540
raise errors.NoCommits(self.other_branch)
541
if self.other_rev_id is not None:
542
self._cached_trees[self.other_rev_id] = self.other_tree
543
self._maybe_fetch(self.other_branch,self.this_branch, self.other_basis)
545
def set_other_revision(self, revision_id, other_branch):
546
"""Set 'other' based on a branch and revision id
548
:param revision_id: The revision to use for a tree
549
:param other_branch: The branch containing this tree
551
self.other_rev_id = revision_id
552
self.other_branch = other_branch
553
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
554
self.other_tree = self.revision_tree(revision_id)
555
self.other_basis = revision_id
557
def set_base_revision(self, revision_id, branch):
558
"""Set 'base' based on a branch and revision id
560
:param revision_id: The revision to use for a tree
561
:param branch: The branch containing this tree
563
self.base_rev_id = revision_id
564
self.base_branch = branch
565
self._maybe_fetch(branch, self.this_branch, revision_id)
566
self.base_tree = self.revision_tree(revision_id)
568
def _maybe_fetch(self, source, target, revision_id):
569
if not source.repository.has_same_location(target.repository):
570
target.fetch(source, revision_id)
573
revisions = [_mod_revision.ensure_null(self.this_basis),
574
_mod_revision.ensure_null(self.other_basis)]
575
if _mod_revision.NULL_REVISION in revisions:
576
self.base_rev_id = _mod_revision.NULL_REVISION
577
self.base_tree = self.revision_tree(self.base_rev_id)
578
self._is_criss_cross = False
580
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
581
self._is_criss_cross = False
583
self.base_rev_id = _mod_revision.NULL_REVISION
585
self.base_rev_id = list(lcas)[0]
586
else: # len(lcas) > 1
587
self._is_criss_cross = True
589
# find_unique_lca can only handle 2 nodes, so we have to
590
# start back at the beginning. It is a shame to traverse
591
# the graph again, but better than re-implementing
593
self.base_rev_id = self.revision_graph.find_unique_lca(
594
revisions[0], revisions[1])
596
self.base_rev_id = self.revision_graph.find_unique_lca(
598
sorted_lca_keys = self.revision_graph.find_merge_order(
600
if self.base_rev_id == _mod_revision.NULL_REVISION:
601
self.base_rev_id = sorted_lca_keys[0]
603
if self.base_rev_id == _mod_revision.NULL_REVISION:
604
raise errors.UnrelatedBranches()
605
if self._is_criss_cross:
606
trace.warning('Warning: criss-cross merge encountered. See bzr'
607
' help criss-cross.')
608
trace.mutter('Criss-cross lcas: %r' % lcas)
609
if self.base_rev_id in lcas:
610
trace.mutter('Unable to find unique lca. '
611
'Fallback %r as best option.'
613
interesting_revision_ids = set(lcas)
614
interesting_revision_ids.add(self.base_rev_id)
615
interesting_trees = dict((t.get_revision_id(), t)
616
for t in self.this_branch.repository.revision_trees(
617
interesting_revision_ids))
618
self._cached_trees.update(interesting_trees)
619
if self.base_rev_id in lcas:
620
self.base_tree = interesting_trees[self.base_rev_id]
622
self.base_tree = interesting_trees.pop(self.base_rev_id)
623
self._lca_trees = [interesting_trees[key]
624
for key in sorted_lca_keys]
626
self.base_tree = self.revision_tree(self.base_rev_id)
627
self.base_is_ancestor = True
628
self.base_is_other_ancestor = True
629
trace.mutter('Base revid: %r' % self.base_rev_id)
631
def set_base(self, base_revision):
632
"""Set the base revision to use for the merge.
634
:param base_revision: A 2-list containing a path and revision number.
636
trace.mutter("doing merge() with no base_revision specified")
637
if base_revision == [None, None]:
640
base_branch, self.base_tree = self._get_tree(base_revision)
641
if base_revision[1] == -1:
642
self.base_rev_id = base_branch.last_revision()
643
elif base_revision[1] is None:
644
self.base_rev_id = _mod_revision.NULL_REVISION
646
self.base_rev_id = _mod_revision.ensure_null(
647
base_branch.get_rev_id(base_revision[1]))
648
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
650
def make_merger(self):
651
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
652
'other_tree': self.other_tree,
653
'interesting_ids': self.interesting_ids,
654
'interesting_files': self.interesting_files,
655
'this_branch': self.this_branch,
657
if self.merge_type.requires_base:
658
kwargs['base_tree'] = self.base_tree
659
if self.merge_type.supports_reprocess:
660
kwargs['reprocess'] = self.reprocess
662
raise errors.BzrError(
663
"Conflict reduction is not supported for merge"
664
" type %s." % self.merge_type)
665
if self.merge_type.supports_show_base:
666
kwargs['show_base'] = self.show_base
668
raise errors.BzrError("Showing base is not supported for this"
669
" merge type. %s" % self.merge_type)
670
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
671
and not self.base_is_other_ancestor):
672
raise errors.CannotReverseCherrypick()
673
if self.merge_type.supports_cherrypick:
674
kwargs['cherrypick'] = (not self.base_is_ancestor or
675
not self.base_is_other_ancestor)
676
if self._is_criss_cross and getattr(self.merge_type,
677
'supports_lca_trees', False):
678
kwargs['lca_trees'] = self._lca_trees
679
return self.merge_type(pb=None,
680
change_reporter=self.change_reporter,
683
def _do_merge_to(self):
684
merge = self.make_merger()
685
if self.other_branch is not None:
686
self.other_branch.update_references(self.this_branch)
688
if self.recurse == 'down':
689
for relpath, file_id in self.this_tree.iter_references():
690
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
691
other_revision = self.other_tree.get_reference_revision(
693
if other_revision == sub_tree.last_revision():
695
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
696
sub_merge.merge_type = self.merge_type
697
other_branch = self.other_branch.reference_parent(file_id,
699
sub_merge.set_other_revision(other_revision, other_branch)
700
base_revision = self.base_tree.get_reference_revision(file_id)
701
sub_merge.base_tree = \
702
sub_tree.branch.repository.revision_tree(base_revision)
703
sub_merge.base_rev_id = base_revision
708
operation = cleanup.OperationWithCleanups(self._do_merge_to)
709
self.this_tree.lock_tree_write()
710
operation.add_cleanup(self.this_tree.unlock)
711
if self.base_tree is not None:
712
self.base_tree.lock_read()
713
operation.add_cleanup(self.base_tree.unlock)
714
if self.other_tree is not None:
715
self.other_tree.lock_read()
716
operation.add_cleanup(self.other_tree.unlock)
717
merge = operation.run_simple()
718
if len(merge.cooked_conflicts) == 0:
719
if not self.ignore_zero and not trace.is_quiet():
720
trace.note(gettext("All changes applied successfully."))
722
trace.note(gettext("%d conflicts encountered.")
723
% len(merge.cooked_conflicts))
725
return len(merge.cooked_conflicts)
728
class _InventoryNoneEntry(object):
729
"""This represents an inventory entry which *isn't there*.
731
It simplifies the merging logic if we always have an InventoryEntry, even
732
if it isn't actually present
739
symlink_target = None
742
_none_entry = _InventoryNoneEntry()
745
class Merge3Merger(object):
746
"""Three-way merger that uses the merge3 text merger"""
748
supports_reprocess = True
749
supports_show_base = True
750
history_based = False
751
supports_cherrypick = True
752
supports_reverse_cherrypick = True
753
winner_idx = {"this": 2, "other": 1, "conflict": 1}
754
supports_lca_trees = True
756
def __init__(self, working_tree, this_tree, base_tree, other_tree,
757
interesting_ids=None, reprocess=False, show_base=False,
758
pb=None, pp=None, change_reporter=None,
759
interesting_files=None, do_merge=True,
760
cherrypick=False, lca_trees=None, this_branch=None):
761
"""Initialize the merger object and perform the merge.
763
:param working_tree: The working tree to apply the merge to
764
:param this_tree: The local tree in the merge operation
765
:param base_tree: The common tree in the merge operation
766
:param other_tree: The other tree to merge changes from
767
:param this_branch: The branch associated with this_tree. Defaults to
768
this_tree.branch if not supplied.
769
:param interesting_ids: The file_ids of files that should be
770
participate in the merge. May not be combined with
772
:param: reprocess If True, perform conflict-reduction processing.
773
:param show_base: If True, show the base revision in text conflicts.
774
(incompatible with reprocess)
776
:param pp: A ProgressPhase object
777
:param change_reporter: An object that should report changes made
778
:param interesting_files: The tree-relative paths of files that should
779
participate in the merge. If these paths refer to directories,
780
the contents of those directories will also be included. May not
781
be combined with interesting_ids. If neither interesting_files nor
782
interesting_ids is specified, all files may participate in the
784
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
785
if the ancestry was found to include a criss-cross merge.
786
Otherwise should be None.
788
object.__init__(self)
789
if interesting_files is not None and interesting_ids is not None:
791
'specify either interesting_ids or interesting_files')
792
if this_branch is None:
793
this_branch = this_tree.branch
794
self.interesting_ids = interesting_ids
795
self.interesting_files = interesting_files
796
self.this_tree = working_tree
797
self.base_tree = base_tree
798
self.other_tree = other_tree
799
self.this_branch = this_branch
800
self._raw_conflicts = []
801
self.cooked_conflicts = []
802
self.reprocess = reprocess
803
self.show_base = show_base
804
self._lca_trees = lca_trees
805
# Uncommenting this will change the default algorithm to always use
806
# _entries_lca. This can be useful for running the test suite and
807
# making sure we haven't missed any corner cases.
808
# if lca_trees is None:
809
# self._lca_trees = [self.base_tree]
810
self.change_reporter = change_reporter
811
self.cherrypick = cherrypick
815
warnings.warn("pp argument to Merge3Merger is deprecated")
817
warnings.warn("pb argument to Merge3Merger is deprecated")
820
operation = cleanup.OperationWithCleanups(self._do_merge)
821
self.this_tree.lock_tree_write()
822
operation.add_cleanup(self.this_tree.unlock)
823
self.base_tree.lock_read()
824
operation.add_cleanup(self.base_tree.unlock)
825
self.other_tree.lock_read()
826
operation.add_cleanup(self.other_tree.unlock)
829
def _do_merge(self, operation):
830
self.tt = transform.TreeTransform(self.this_tree, None)
831
operation.add_cleanup(self.tt.finalize)
832
self._compute_transform()
833
results = self.tt.apply(no_conflicts=True)
834
self.write_modified(results)
836
self.this_tree.add_conflicts(self.cooked_conflicts)
837
except errors.UnsupportedOperation:
840
def make_preview_transform(self):
841
operation = cleanup.OperationWithCleanups(self._make_preview_transform)
842
self.base_tree.lock_read()
843
operation.add_cleanup(self.base_tree.unlock)
844
self.other_tree.lock_read()
845
operation.add_cleanup(self.other_tree.unlock)
846
return operation.run_simple()
848
def _make_preview_transform(self):
849
self.tt = transform.TransformPreview(self.this_tree)
850
self._compute_transform()
853
def _compute_transform(self):
854
if self._lca_trees is None:
855
entries = self._entries3()
856
resolver = self._three_way
858
entries = self._entries_lca()
859
resolver = self._lca_multi_way
860
# Prepare merge hooks
861
factories = Merger.hooks['merge_file_content']
862
# One hook for each registered one plus our default merger
863
hooks = [factory(self) for factory in factories] + [self]
864
self.active_hooks = [hook for hook in hooks if hook is not None]
865
child_pb = ui.ui_factory.nested_progress_bar()
867
for num, (file_id, changed, parents3, names3,
868
executable3) in enumerate(entries):
869
# Try merging each entry
870
child_pb.update(gettext('Preparing file merge'),
872
self._merge_names(file_id, parents3, names3, resolver=resolver)
874
file_status = self._do_merge_contents(file_id)
876
file_status = 'unmodified'
877
self._merge_executable(file_id,
878
executable3, file_status, resolver=resolver)
881
self.tt.fixup_new_roots()
882
self._finish_computing_transform()
884
def _finish_computing_transform(self):
885
"""Finalize the transform and report the changes.
887
This is the second half of _compute_transform.
889
child_pb = ui.ui_factory.nested_progress_bar()
891
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
892
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
895
if self.change_reporter is not None:
896
from bzrlib import delta
897
delta.report_changes(
898
self.tt.iter_changes(), self.change_reporter)
899
self.cook_conflicts(fs_conflicts)
900
for conflict in self.cooked_conflicts:
901
trace.warning(unicode(conflict))
904
"""Gather data about files modified between three trees.
906
Return a list of tuples of file_id, changed, parents3, names3,
907
executable3. changed is a boolean indicating whether the file contents
908
or kind were changed. parents3 is a tuple of parent ids for base,
909
other and this. names3 is a tuple of names for base, other and this.
910
executable3 is a tuple of execute-bit values for base, other and this.
913
iterator = self.other_tree.iter_changes(self.base_tree,
914
specific_files=self.interesting_files,
915
extra_trees=[self.this_tree])
916
this_entries = dict((e.file_id, e) for p, e in
917
self.this_tree.iter_entries_by_dir(
918
self.interesting_ids))
919
for (file_id, paths, changed, versioned, parents, names, kind,
920
executable) in iterator:
921
if (self.interesting_ids is not None and
922
file_id not in self.interesting_ids):
924
entry = this_entries.get(file_id)
925
if entry is not None:
926
this_name = entry.name
927
this_parent = entry.parent_id
928
this_executable = entry.executable
932
this_executable = None
933
parents3 = parents + (this_parent,)
934
names3 = names + (this_name,)
935
executable3 = executable + (this_executable,)
936
result.append((file_id, changed, parents3, names3, executable3))
939
def _entries_lca(self):
940
"""Gather data about files modified between multiple trees.
942
This compares OTHER versus all LCA trees, and for interesting entries,
943
it then compares with THIS and BASE.
945
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
947
:return: [(file_id, changed, parents, names, executable)], where:
949
* file_id: Simple file_id of the entry
950
* changed: Boolean, True if the kind or contents changed else False
951
* parents: ((base, [parent_id, in, lcas]), parent_id_other,
953
* names: ((base, [name, in, lcas]), name_in_other, name_in_this)
954
* executable: ((base, [exec, in, lcas]), exec_in_other,
957
if self.interesting_files is not None:
958
lookup_trees = [self.this_tree, self.base_tree]
959
lookup_trees.extend(self._lca_trees)
960
# I think we should include the lca trees as well
961
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
964
interesting_ids = self.interesting_ids
966
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
968
base_inventory = self.base_tree.inventory
969
this_inventory = self.this_tree.inventory
970
for path, file_id, other_ie, lca_values in walker.iter_all():
971
# Is this modified at all from any of the other trees?
973
other_ie = _none_entry
974
if interesting_ids is not None and file_id not in interesting_ids:
977
# If other_revision is found in any of the lcas, that means this
978
# node is uninteresting. This is because when merging, if there are
979
# multiple heads(), we have to create a new node. So if we didn't,
980
# we know that the ancestry is linear, and that OTHER did not
982
# See doc/developers/lca_merge_resolution.txt for details
983
other_revision = other_ie.revision
984
if other_revision is not None:
985
# We can't use this shortcut when other_revision is None,
986
# because it may be None because things are WorkingTrees, and
987
# not because it is *actually* None.
988
is_unmodified = False
989
for lca_path, ie in lca_values:
990
if ie is not None and ie.revision == other_revision:
997
for lca_path, lca_ie in lca_values:
999
lca_entries.append(_none_entry)
1001
lca_entries.append(lca_ie)
1003
if base_inventory.has_id(file_id):
1004
base_ie = base_inventory[file_id]
1006
base_ie = _none_entry
1008
if this_inventory.has_id(file_id):
1009
this_ie = this_inventory[file_id]
1011
this_ie = _none_entry
1017
for lca_ie in lca_entries:
1018
lca_kinds.append(lca_ie.kind)
1019
lca_parent_ids.append(lca_ie.parent_id)
1020
lca_names.append(lca_ie.name)
1021
lca_executable.append(lca_ie.executable)
1023
kind_winner = self._lca_multi_way(
1024
(base_ie.kind, lca_kinds),
1025
other_ie.kind, this_ie.kind)
1026
parent_id_winner = self._lca_multi_way(
1027
(base_ie.parent_id, lca_parent_ids),
1028
other_ie.parent_id, this_ie.parent_id)
1029
name_winner = self._lca_multi_way(
1030
(base_ie.name, lca_names),
1031
other_ie.name, this_ie.name)
1033
content_changed = True
1034
if kind_winner == 'this':
1035
# No kind change in OTHER, see if there are *any* changes
1036
if other_ie.kind == 'directory':
1037
if parent_id_winner == 'this' and name_winner == 'this':
1038
# No change for this directory in OTHER, skip
1040
content_changed = False
1041
elif other_ie.kind is None or other_ie.kind == 'file':
1042
def get_sha1(ie, tree):
1043
if ie.kind != 'file':
1045
return tree.get_file_sha1(file_id)
1046
base_sha1 = get_sha1(base_ie, self.base_tree)
1047
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1048
in zip(lca_entries, self._lca_trees)]
1049
this_sha1 = get_sha1(this_ie, self.this_tree)
1050
other_sha1 = get_sha1(other_ie, self.other_tree)
1051
sha1_winner = self._lca_multi_way(
1052
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1053
allow_overriding_lca=False)
1054
exec_winner = self._lca_multi_way(
1055
(base_ie.executable, lca_executable),
1056
other_ie.executable, this_ie.executable)
1057
if (parent_id_winner == 'this' and name_winner == 'this'
1058
and sha1_winner == 'this' and exec_winner == 'this'):
1059
# No kind, parent, name, exec, or content change for
1060
# OTHER, so this node is not considered interesting
1062
if sha1_winner == 'this':
1063
content_changed = False
1064
elif other_ie.kind == 'symlink':
1065
def get_target(ie, tree):
1066
if ie.kind != 'symlink':
1068
return tree.get_symlink_target(file_id)
1069
base_target = get_target(base_ie, self.base_tree)
1070
lca_targets = [get_target(ie, tree) for ie, tree
1071
in zip(lca_entries, self._lca_trees)]
1072
this_target = get_target(this_ie, self.this_tree)
1073
other_target = get_target(other_ie, self.other_tree)
1074
target_winner = self._lca_multi_way(
1075
(base_target, lca_targets),
1076
other_target, this_target)
1077
if (parent_id_winner == 'this' and name_winner == 'this'
1078
and target_winner == 'this'):
1079
# No kind, parent, name, or symlink target change
1082
if target_winner == 'this':
1083
content_changed = False
1084
elif other_ie.kind == 'tree-reference':
1085
# The 'changed' information seems to be handled at a higher
1086
# level. At least, _entries3 returns False for content
1087
# changed, even when at a new revision_id.
1088
content_changed = False
1089
if (parent_id_winner == 'this' and name_winner == 'this'):
1090
# Nothing interesting
1093
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1095
# If we have gotten this far, that means something has changed
1096
result.append((file_id, content_changed,
1097
((base_ie.parent_id, lca_parent_ids),
1098
other_ie.parent_id, this_ie.parent_id),
1099
((base_ie.name, lca_names),
1100
other_ie.name, this_ie.name),
1101
((base_ie.executable, lca_executable),
1102
other_ie.executable, this_ie.executable)
1106
@deprecated_method(deprecated_in((2, 4, 0)))
1108
if self.tt.final_kind(self.tt.root) is None:
1109
self.tt.cancel_deletion(self.tt.root)
1110
if self.tt.final_file_id(self.tt.root) is None:
1111
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
1113
other_root_file_id = self.other_tree.get_root_id()
1114
if other_root_file_id is None:
1116
other_root = self.tt.trans_id_file_id(other_root_file_id)
1117
if other_root == self.tt.root:
1119
if self.this_tree.inventory.has_id(
1120
self.other_tree.inventory.root.file_id):
1121
# the other tree's root is a non-root in the current tree (as
1122
# when a previously unrelated branch is merged into another)
1124
if self.tt.final_kind(other_root) is not None:
1125
other_root_is_present = True
1127
# other_root doesn't have a physical representation. We still need
1128
# to move any references to the actual root of the tree.
1129
other_root_is_present = False
1130
# 'other_tree.inventory.root' is not present in this tree. We are
1131
# calling adjust_path for children which *want* to be present with a
1132
# correct place to go.
1133
for _, child in self.other_tree.inventory.root.children.iteritems():
1134
trans_id = self.tt.trans_id_file_id(child.file_id)
1135
if not other_root_is_present:
1136
if self.tt.final_kind(trans_id) is not None:
1137
# The item exist in the final tree and has a defined place
1140
# Move the item into the root
1142
final_name = self.tt.final_name(trans_id)
1143
except errors.NoFinalPath:
1144
# This file is not present anymore, ignore it.
1146
self.tt.adjust_path(final_name, self.tt.root, trans_id)
1147
if other_root_is_present:
1148
self.tt.cancel_creation(other_root)
1149
self.tt.cancel_versioning(other_root)
1151
def write_modified(self, results):
1152
modified_hashes = {}
1153
for path in results.modified_paths:
1154
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
1157
hash = self.this_tree.get_file_sha1(file_id)
1160
modified_hashes[file_id] = hash
1161
self.this_tree.set_merge_modified(modified_hashes)
1164
def parent(entry, file_id):
1165
"""Determine the parent for a file_id (used as a key method)"""
1168
return entry.parent_id
1171
def name(entry, file_id):
1172
"""Determine the name for a file_id (used as a key method)"""
1178
def contents_sha1(tree, file_id):
1179
"""Determine the sha1 of the file contents (used as a key method)."""
1180
if not tree.has_id(file_id):
1182
return tree.get_file_sha1(file_id)
1185
def executable(tree, file_id):
1186
"""Determine the executability of a file-id (used as a key method)."""
1187
if not tree.has_id(file_id):
1189
if tree.kind(file_id) != "file":
1191
return tree.is_executable(file_id)
1194
def kind(tree, file_id):
1195
"""Determine the kind of a file-id (used as a key method)."""
1196
if not tree.has_id(file_id):
1198
return tree.kind(file_id)
1201
def _three_way(base, other, this):
1203
# if 'base == other', either they all agree, or only 'this' has
1206
elif this not in (base, other):
1207
# 'this' is neither 'base' nor 'other', so both sides changed
1210
# "Ambiguous clean merge" -- both sides have made the same change.
1213
# this == base: only other has changed.
1217
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
1218
"""Consider LCAs when determining whether a change has occurred.
1220
If LCAS are all identical, this is the same as a _three_way comparison.
1222
:param bases: value in (BASE, [LCAS])
1223
:param other: value in OTHER
1224
:param this: value in THIS
1225
:param allow_overriding_lca: If there is more than one unique lca
1226
value, allow OTHER to override THIS if it has a new value, and
1227
THIS only has an lca value, or vice versa. This is appropriate for
1228
truly scalar values, not as much for non-scalars.
1229
:return: 'this', 'other', or 'conflict' depending on whether an entry
1232
# See doc/developers/lca_tree_merging.txt for details about this
1235
# Either Ambiguously clean, or nothing was actually changed. We
1238
base_val, lca_vals = bases
1239
# Remove 'base_val' from the lca_vals, because it is not interesting
1240
filtered_lca_vals = [lca_val for lca_val in lca_vals
1241
if lca_val != base_val]
1242
if len(filtered_lca_vals) == 0:
1243
return Merge3Merger._three_way(base_val, other, this)
1245
unique_lca_vals = set(filtered_lca_vals)
1246
if len(unique_lca_vals) == 1:
1247
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1249
if allow_overriding_lca:
1250
if other in unique_lca_vals:
1251
if this in unique_lca_vals:
1252
# Each side picked a different lca, conflict
1255
# This has a value which supersedes both lca values, and
1256
# other only has an lca value
1258
elif this in unique_lca_vals:
1259
# OTHER has a value which supersedes both lca values, and this
1260
# only has an lca value
1263
# At this point, the lcas disagree, and the tip disagree
1267
@deprecated_method(deprecated_in((2, 2, 0)))
1268
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1269
"""Do a three-way test on a scalar.
1270
Return "this", "other" or "conflict", depending whether a value wins.
1272
key_base = key(base_tree, file_id)
1273
key_other = key(other_tree, file_id)
1274
#if base == other, either they all agree, or only THIS has changed.
1275
if key_base == key_other:
1277
key_this = key(this_tree, file_id)
1278
# "Ambiguous clean merge"
1279
if key_this == key_other:
1281
elif key_this == key_base:
1286
def merge_names(self, file_id):
1287
def get_entry(tree):
1288
if tree.has_id(file_id):
1289
return tree.inventory[file_id]
1292
this_entry = get_entry(self.this_tree)
1293
other_entry = get_entry(self.other_tree)
1294
base_entry = get_entry(self.base_tree)
1295
entries = (base_entry, other_entry, this_entry)
1298
for entry in entries:
1301
parents.append(None)
1303
names.append(entry.name)
1304
parents.append(entry.parent_id)
1305
return self._merge_names(file_id, parents, names,
1306
resolver=self._three_way)
1308
def _merge_names(self, file_id, parents, names, resolver):
1309
"""Perform a merge on file_id names and parents"""
1310
base_name, other_name, this_name = names
1311
base_parent, other_parent, this_parent = parents
1313
name_winner = resolver(*names)
1315
parent_id_winner = resolver(*parents)
1316
if this_name is None:
1317
if name_winner == "this":
1318
name_winner = "other"
1319
if parent_id_winner == "this":
1320
parent_id_winner = "other"
1321
if name_winner == "this" and parent_id_winner == "this":
1323
if name_winner == 'conflict' or parent_id_winner == 'conflict':
1324
# Creating helpers (.OTHER or .THIS) here cause problems down the
1325
# road if a ContentConflict needs to be created so we should not do
1327
trans_id = self.tt.trans_id_file_id(file_id)
1328
self._raw_conflicts.append(('path conflict', trans_id, file_id,
1329
this_parent, this_name,
1330
other_parent, other_name))
1331
if not self.other_tree.has_id(file_id):
1332
# it doesn't matter whether the result was 'other' or
1333
# 'conflict'-- if it has no file id, we leave it alone.
1335
parent_id = parents[self.winner_idx[parent_id_winner]]
1336
name = names[self.winner_idx[name_winner]]
1337
if parent_id is not None or name is not None:
1338
# if we get here, name_winner and parent_winner are set to safe
1340
if parent_id is None and name is not None:
1341
# if parent_id is None and name is non-None, current file is
1343
if names[self.winner_idx[parent_id_winner]] != '':
1344
raise AssertionError(
1345
'File looks like a root, but named %s' %
1346
names[self.winner_idx[parent_id_winner]])
1347
parent_trans_id = transform.ROOT_PARENT
1349
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1350
self.tt.adjust_path(name, parent_trans_id,
1351
self.tt.trans_id_file_id(file_id))
1353
def _do_merge_contents(self, file_id):
1354
"""Performs a merge on file_id contents."""
1355
def contents_pair(tree):
1356
if not tree.has_id(file_id):
1358
kind = tree.kind(file_id)
1360
contents = tree.get_file_sha1(file_id)
1361
elif kind == "symlink":
1362
contents = tree.get_symlink_target(file_id)
1365
return kind, contents
1367
# See SPOT run. run, SPOT, run.
1368
# So we're not QUITE repeating ourselves; we do tricky things with
1370
base_pair = contents_pair(self.base_tree)
1371
other_pair = contents_pair(self.other_tree)
1373
this_pair = contents_pair(self.this_tree)
1374
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1375
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1376
this_pair, allow_overriding_lca=False)
1378
if base_pair == other_pair:
1381
# We delayed evaluating this_pair as long as we can to avoid
1382
# unnecessary sha1 calculation
1383
this_pair = contents_pair(self.this_tree)
1384
winner = self._three_way(base_pair, other_pair, this_pair)
1385
if winner == 'this':
1386
# No interesting changes introduced by OTHER
1388
# We have a hypothetical conflict, but if we have files, then we
1389
# can try to merge the content
1390
trans_id = self.tt.trans_id_file_id(file_id)
1391
params = MergeHookParams(self, file_id, trans_id, this_pair[0],
1392
other_pair[0], winner)
1393
hooks = self.active_hooks
1394
hook_status = 'not_applicable'
1396
hook_status, lines = hook.merge_contents(params)
1397
if hook_status != 'not_applicable':
1398
# Don't try any more hooks, this one applies.
1400
# If the merge ends up replacing the content of the file, we get rid of
1401
# it at the end of this method (this variable is used to track the
1402
# exceptions to this rule).
1405
if hook_status == 'not_applicable':
1406
# No merge hook was able to resolve the situation. Two cases exist:
1407
# a content conflict or a duplicate one.
1409
name = self.tt.final_name(trans_id)
1410
parent_id = self.tt.final_parent(trans_id)
1412
inhibit_content_conflict = False
1413
if params.this_kind is None: # file_id is not in THIS
1414
# Is the name used for a different file_id ?
1415
dupe_path = self.other_tree.id2path(file_id)
1416
this_id = self.this_tree.path2id(dupe_path)
1417
if this_id is not None:
1418
# Two entries for the same path
1420
# versioning the merged file will trigger a duplicate
1422
self.tt.version_file(file_id, trans_id)
1423
transform.create_from_tree(
1424
self.tt, trans_id, self.other_tree, file_id,
1425
filter_tree_path=self._get_filter_tree_path(file_id))
1426
inhibit_content_conflict = True
1427
elif params.other_kind is None: # file_id is not in OTHER
1428
# Is the name used for a different file_id ?
1429
dupe_path = self.this_tree.id2path(file_id)
1430
other_id = self.other_tree.path2id(dupe_path)
1431
if other_id is not None:
1432
# Two entries for the same path again, but here, the other
1433
# entry will also be merged. We simply inhibit the
1434
# 'content' conflict creation because we know OTHER will
1435
# create (or has already created depending on ordering) an
1436
# entry at the same path. This will trigger a 'duplicate'
1439
inhibit_content_conflict = True
1440
if not inhibit_content_conflict:
1441
if params.this_kind is not None:
1442
self.tt.unversion_file(trans_id)
1443
# This is a contents conflict, because none of the available
1444
# functions could merge it.
1445
file_group = self._dump_conflicts(name, parent_id, file_id,
1447
self._raw_conflicts.append(('contents conflict', file_group))
1448
elif hook_status == 'success':
1449
self.tt.create_file(lines, trans_id)
1450
elif hook_status == 'conflicted':
1451
# XXX: perhaps the hook should be able to provide
1452
# the BASE/THIS/OTHER files?
1453
self.tt.create_file(lines, trans_id)
1454
self._raw_conflicts.append(('text conflict', trans_id))
1455
name = self.tt.final_name(trans_id)
1456
parent_id = self.tt.final_parent(trans_id)
1457
self._dump_conflicts(name, parent_id, file_id)
1458
elif hook_status == 'delete':
1459
self.tt.unversion_file(trans_id)
1461
elif hook_status == 'done':
1462
# The hook function did whatever it needs to do directly, no
1463
# further action needed here.
1466
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1467
if not self.this_tree.has_id(file_id) and result == "modified":
1468
self.tt.version_file(file_id, trans_id)
1470
# The merge has been performed and produced a new content, so the
1471
# old contents should not be retained.
1472
self.tt.delete_contents(trans_id)
1475
def _default_other_winner_merge(self, merge_hook_params):
1476
"""Replace this contents with other."""
1477
file_id = merge_hook_params.file_id
1478
trans_id = merge_hook_params.trans_id
1479
if self.other_tree.has_id(file_id):
1480
# OTHER changed the file
1481
transform.create_from_tree(
1482
self.tt, trans_id, self.other_tree, file_id,
1483
filter_tree_path=self._get_filter_tree_path(file_id))
1485
elif self.this_tree.has_id(file_id):
1486
# OTHER deleted the file
1487
return 'delete', None
1489
raise AssertionError(
1490
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1493
def merge_contents(self, merge_hook_params):
1494
"""Fallback merge logic after user installed hooks."""
1495
# This function is used in merge hooks as the fallback instance.
1496
# Perhaps making this function and the functions it calls be a
1497
# a separate class would be better.
1498
if merge_hook_params.winner == 'other':
1499
# OTHER is a straight winner, so replace this contents with other
1500
return self._default_other_winner_merge(merge_hook_params)
1501
elif merge_hook_params.is_file_merge():
1502
# THIS and OTHER are both files, so text merge. Either
1503
# BASE is a file, or both converted to files, so at least we
1504
# have agreement that output should be a file.
1506
self.text_merge(merge_hook_params.file_id,
1507
merge_hook_params.trans_id)
1508
except errors.BinaryFile:
1509
return 'not_applicable', None
1512
return 'not_applicable', None
1514
def get_lines(self, tree, file_id):
1515
"""Return the lines in a file, or an empty list."""
1516
if tree.has_id(file_id):
1517
return tree.get_file_lines(file_id)
1521
def text_merge(self, file_id, trans_id):
1522
"""Perform a three-way text merge on a file_id"""
1523
# it's possible that we got here with base as a different type.
1524
# if so, we just want two-way text conflicts.
1525
if self.base_tree.has_id(file_id) and \
1526
self.base_tree.kind(file_id) == "file":
1527
base_lines = self.get_lines(self.base_tree, file_id)
1530
other_lines = self.get_lines(self.other_tree, file_id)
1531
this_lines = self.get_lines(self.this_tree, file_id)
1532
m3 = merge3.Merge3(base_lines, this_lines, other_lines,
1533
is_cherrypick=self.cherrypick)
1534
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1535
if self.show_base is True:
1536
base_marker = '|' * 7
1540
def iter_merge3(retval):
1541
retval["text_conflicts"] = False
1542
for line in m3.merge_lines(name_a = "TREE",
1543
name_b = "MERGE-SOURCE",
1544
name_base = "BASE-REVISION",
1545
start_marker=start_marker,
1546
base_marker=base_marker,
1547
reprocess=self.reprocess):
1548
if line.startswith(start_marker):
1549
retval["text_conflicts"] = True
1550
yield line.replace(start_marker, '<' * 7)
1554
merge3_iterator = iter_merge3(retval)
1555
self.tt.create_file(merge3_iterator, trans_id)
1556
if retval["text_conflicts"] is True:
1557
self._raw_conflicts.append(('text conflict', trans_id))
1558
name = self.tt.final_name(trans_id)
1559
parent_id = self.tt.final_parent(trans_id)
1560
file_group = self._dump_conflicts(name, parent_id, file_id,
1561
this_lines, base_lines,
1563
file_group.append(trans_id)
1566
def _get_filter_tree_path(self, file_id):
1567
if self.this_tree.supports_content_filtering():
1568
# We get the path from the working tree if it exists.
1569
# That fails though when OTHER is adding a file, so
1570
# we fall back to the other tree to find the path if
1571
# it doesn't exist locally.
1573
return self.this_tree.id2path(file_id)
1574
except errors.NoSuchId:
1575
return self.other_tree.id2path(file_id)
1576
# Skip the id2path lookup for older formats
1579
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1580
base_lines=None, other_lines=None, set_version=False,
1582
"""Emit conflict files.
1583
If this_lines, base_lines, or other_lines are omitted, they will be
1584
determined automatically. If set_version is true, the .OTHER, .THIS
1585
or .BASE (in that order) will be created as versioned files.
1587
data = [('OTHER', self.other_tree, other_lines),
1588
('THIS', self.this_tree, this_lines)]
1590
data.append(('BASE', self.base_tree, base_lines))
1592
# We need to use the actual path in the working tree of the file here,
1593
# ignoring the conflict suffixes
1595
if wt.supports_content_filtering():
1597
filter_tree_path = wt.id2path(file_id)
1598
except errors.NoSuchId:
1599
# file has been deleted
1600
filter_tree_path = None
1602
# Skip the id2path lookup for older formats
1603
filter_tree_path = None
1607
for suffix, tree, lines in data:
1608
if tree.has_id(file_id):
1609
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1610
suffix, lines, filter_tree_path)
1611
file_group.append(trans_id)
1612
if set_version and not versioned:
1613
self.tt.version_file(file_id, trans_id)
1617
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1618
lines=None, filter_tree_path=None):
1619
"""Emit a single conflict file."""
1620
name = name + '.' + suffix
1621
trans_id = self.tt.create_path(name, parent_id)
1622
transform.create_from_tree(self.tt, trans_id, tree, file_id, lines,
1626
def merge_executable(self, file_id, file_status):
1627
"""Perform a merge on the execute bit."""
1628
executable = [self.executable(t, file_id) for t in (self.base_tree,
1629
self.other_tree, self.this_tree)]
1630
self._merge_executable(file_id, executable, file_status,
1631
resolver=self._three_way)
1633
def _merge_executable(self, file_id, executable, file_status,
1635
"""Perform a merge on the execute bit."""
1636
base_executable, other_executable, this_executable = executable
1637
if file_status == "deleted":
1639
winner = resolver(*executable)
1640
if winner == "conflict":
1641
# There must be a None in here, if we have a conflict, but we
1642
# need executability since file status was not deleted.
1643
if self.executable(self.other_tree, file_id) is None:
1647
if winner == 'this' and file_status != "modified":
1649
trans_id = self.tt.trans_id_file_id(file_id)
1650
if self.tt.final_kind(trans_id) != "file":
1652
if winner == "this":
1653
executability = this_executable
1655
if self.other_tree.has_id(file_id):
1656
executability = other_executable
1657
elif self.this_tree.has_id(file_id):
1658
executability = this_executable
1659
elif self.base_tree_has_id(file_id):
1660
executability = base_executable
1661
if executability is not None:
1662
trans_id = self.tt.trans_id_file_id(file_id)
1663
self.tt.set_executability(executability, trans_id)
1665
def cook_conflicts(self, fs_conflicts):
1666
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1667
content_conflict_file_ids = set()
1668
cooked_conflicts = transform.cook_conflicts(fs_conflicts, self.tt)
1669
fp = transform.FinalPaths(self.tt)
1670
for conflict in self._raw_conflicts:
1671
conflict_type = conflict[0]
1672
if conflict_type == 'path conflict':
1674
this_parent, this_name,
1675
other_parent, other_name) = conflict[1:]
1676
if this_parent is None or this_name is None:
1677
this_path = '<deleted>'
1679
parent_path = fp.get_path(
1680
self.tt.trans_id_file_id(this_parent))
1681
this_path = osutils.pathjoin(parent_path, this_name)
1682
if other_parent is None or other_name is None:
1683
other_path = '<deleted>'
1685
if other_parent == self.other_tree.get_root_id():
1686
# The tree transform doesn't know about the other root,
1687
# so we special case here to avoid a NoFinalPath
1691
parent_path = fp.get_path(
1692
self.tt.trans_id_file_id(other_parent))
1693
other_path = osutils.pathjoin(parent_path, other_name)
1694
c = _mod_conflicts.Conflict.factory(
1695
'path conflict', path=this_path,
1696
conflict_path=other_path,
1698
elif conflict_type == 'contents conflict':
1699
for trans_id in conflict[1]:
1700
file_id = self.tt.final_file_id(trans_id)
1701
if file_id is not None:
1702
# Ok we found the relevant file-id
1704
path = fp.get_path(trans_id)
1705
for suffix in ('.BASE', '.THIS', '.OTHER'):
1706
if path.endswith(suffix):
1707
# Here is the raw path
1708
path = path[:-len(suffix)]
1710
c = _mod_conflicts.Conflict.factory(conflict_type,
1711
path=path, file_id=file_id)
1712
content_conflict_file_ids.add(file_id)
1713
elif conflict_type == 'text conflict':
1714
trans_id = conflict[1]
1715
path = fp.get_path(trans_id)
1716
file_id = self.tt.final_file_id(trans_id)
1717
c = _mod_conflicts.Conflict.factory(conflict_type,
1718
path=path, file_id=file_id)
1720
raise AssertionError('bad conflict type: %r' % (conflict,))
1721
cooked_conflicts.append(c)
1723
self.cooked_conflicts = []
1724
# We want to get rid of path conflicts when a corresponding contents
1725
# conflict exists. This can occur when one branch deletes a file while
1726
# the other renames *and* modifies it. In this case, the content
1727
# conflict is enough.
1728
for c in cooked_conflicts:
1729
if (c.typestring == 'path conflict'
1730
and c.file_id in content_conflict_file_ids):
1732
self.cooked_conflicts.append(c)
1733
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
1736
class WeaveMerger(Merge3Merger):
1737
"""Three-way tree merger, text weave merger."""
1738
supports_reprocess = True
1739
supports_show_base = False
1740
supports_reverse_cherrypick = False
1741
history_based = True
1743
def _generate_merge_plan(self, file_id, base):
1744
return self.this_tree.plan_file_merge(file_id, self.other_tree,
1747
def _merged_lines(self, file_id):
1748
"""Generate the merged lines.
1749
There is no distinction between lines that are meant to contain <<<<<<<
1753
base = self.base_tree
1756
plan = self._generate_merge_plan(file_id, base)
1757
if 'merge' in debug.debug_flags:
1759
trans_id = self.tt.trans_id_file_id(file_id)
1760
name = self.tt.final_name(trans_id) + '.plan'
1761
contents = ('%11s|%s' % l for l in plan)
1762
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1763
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1764
'>>>>>>> MERGE-SOURCE\n')
1765
lines, conflicts = textmerge.merge_lines(self.reprocess)
1767
base_lines = textmerge.base_from_plan()
1770
return lines, base_lines
1772
def text_merge(self, file_id, trans_id):
1773
"""Perform a (weave) text merge for a given file and file-id.
1774
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1775
and a conflict will be noted.
1777
lines, base_lines = self._merged_lines(file_id)
1779
# Note we're checking whether the OUTPUT is binary in this case,
1780
# because we don't want to get into weave merge guts.
1781
textfile.check_text_lines(lines)
1782
self.tt.create_file(lines, trans_id)
1783
if base_lines is not None:
1785
self._raw_conflicts.append(('text conflict', trans_id))
1786
name = self.tt.final_name(trans_id)
1787
parent_id = self.tt.final_parent(trans_id)
1788
file_group = self._dump_conflicts(name, parent_id, file_id,
1790
base_lines=base_lines)
1791
file_group.append(trans_id)
1794
class LCAMerger(WeaveMerger):
1796
def _generate_merge_plan(self, file_id, base):
1797
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1800
class Diff3Merger(Merge3Merger):
1801
"""Three-way merger using external diff3 for text merging"""
1803
def dump_file(self, temp_dir, name, tree, file_id):
1804
out_path = osutils.pathjoin(temp_dir, name)
1805
out_file = open(out_path, "wb")
1807
in_file = tree.get_file(file_id)
1808
for line in in_file:
1809
out_file.write(line)
1814
def text_merge(self, file_id, trans_id):
1815
"""Perform a diff3 merge using a specified file-id and trans-id.
1816
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1817
will be dumped, and a will be conflict noted.
1820
temp_dir = osutils.mkdtemp(prefix="bzr-")
1822
new_file = osutils.pathjoin(temp_dir, "new")
1823
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1824
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1825
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1826
status = bzrlib.patch.diff3(new_file, this, base, other)
1827
if status not in (0, 1):
1828
raise errors.BzrError("Unhandled diff3 exit code")
1829
f = open(new_file, 'rb')
1831
self.tt.create_file(f, trans_id)
1835
name = self.tt.final_name(trans_id)
1836
parent_id = self.tt.final_parent(trans_id)
1837
self._dump_conflicts(name, parent_id, file_id)
1838
self._raw_conflicts.append(('text conflict', trans_id))
1840
osutils.rmtree(temp_dir)
1843
class PathNotInTree(errors.BzrError):
1845
_fmt = """Merge-into failed because %(tree)s does not contain %(path)s."""
1847
def __init__(self, path, tree):
1848
errors.BzrError.__init__(self, path=path, tree=tree)
1851
class MergeIntoMerger(Merger):
1852
"""Merger that understands other_tree will be merged into a subdir.
1854
This also changes the Merger api so that it uses real Branch, revision_id,
1855
and RevisonTree objects, rather than using revision specs.
1858
def __init__(self, this_tree, other_branch, other_tree, target_subdir,
1859
source_subpath, other_rev_id=None):
1860
"""Create a new MergeIntoMerger object.
1862
source_subpath in other_tree will be effectively copied to
1863
target_subdir in this_tree.
1865
:param this_tree: The tree that we will be merging into.
1866
:param other_branch: The Branch we will be merging from.
1867
:param other_tree: The RevisionTree object we want to merge.
1868
:param target_subdir: The relative path where we want to merge
1869
other_tree into this_tree
1870
:param source_subpath: The relative path specifying the subtree of
1871
other_tree to merge into this_tree.
1873
# It is assumed that we are merging a tree that is not in our current
1874
# ancestry, which means we are using the "EmptyTree" as our basis.
1875
null_ancestor_tree = this_tree.branch.repository.revision_tree(
1876
_mod_revision.NULL_REVISION)
1877
super(MergeIntoMerger, self).__init__(
1878
this_branch=this_tree.branch,
1879
this_tree=this_tree,
1880
other_tree=other_tree,
1881
base_tree=null_ancestor_tree,
1883
self._target_subdir = target_subdir
1884
self._source_subpath = source_subpath
1885
self.other_branch = other_branch
1886
if other_rev_id is None:
1887
other_rev_id = other_tree.get_revision_id()
1888
self.other_rev_id = self.other_basis = other_rev_id
1889
self.base_is_ancestor = True
1890
self.backup_files = True
1891
self.merge_type = Merge3Merger
1892
self.show_base = False
1893
self.reprocess = False
1894
self.interesting_ids = None
1895
self.merge_type = _MergeTypeParameterizer(MergeIntoMergeType,
1896
target_subdir=self._target_subdir,
1897
source_subpath=self._source_subpath)
1898
if self._source_subpath != '':
1899
# If this isn't a partial merge make sure the revisions will be
1901
self._maybe_fetch(self.other_branch, self.this_branch,
1904
def set_pending(self):
1905
if self._source_subpath != '':
1907
Merger.set_pending(self)
1910
class _MergeTypeParameterizer(object):
1911
"""Wrap a merge-type class to provide extra parameters.
1913
This is hack used by MergeIntoMerger to pass some extra parameters to its
1914
merge_type. Merger.do_merge() sets up its own set of parameters to pass to
1915
the 'merge_type' member. It is difficult override do_merge without
1916
re-writing the whole thing, so instead we create a wrapper which will pass
1917
the extra parameters.
1920
def __init__(self, merge_type, **kwargs):
1921
self._extra_kwargs = kwargs
1922
self._merge_type = merge_type
1924
def __call__(self, *args, **kwargs):
1925
kwargs.update(self._extra_kwargs)
1926
return self._merge_type(*args, **kwargs)
1928
def __getattr__(self, name):
1929
return getattr(self._merge_type, name)
1932
class MergeIntoMergeType(Merge3Merger):
1933
"""Merger that incorporates a tree (or part of a tree) into another."""
1935
def __init__(self, *args, **kwargs):
1936
"""Initialize the merger object.
1938
:param args: See Merge3Merger.__init__'s args.
1939
:param kwargs: See Merge3Merger.__init__'s keyword args, except for
1940
source_subpath and target_subdir.
1941
:keyword source_subpath: The relative path specifying the subtree of
1942
other_tree to merge into this_tree.
1943
:keyword target_subdir: The relative path where we want to merge
1944
other_tree into this_tree
1946
# All of the interesting work happens during Merge3Merger.__init__(),
1947
# so we have have to hack in to get our extra parameters set.
1948
self._source_subpath = kwargs.pop('source_subpath')
1949
self._target_subdir = kwargs.pop('target_subdir')
1950
super(MergeIntoMergeType, self).__init__(*args, **kwargs)
1952
def _compute_transform(self):
1953
child_pb = ui.ui_factory.nested_progress_bar()
1955
entries = self._entries_to_incorporate()
1956
entries = list(entries)
1957
for num, (entry, parent_id) in enumerate(entries):
1958
child_pb.update(gettext('Preparing file merge'), num, len(entries))
1959
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1960
trans_id = transform.new_by_entry(self.tt, entry,
1961
parent_trans_id, self.other_tree)
1964
self._finish_computing_transform()
1966
def _entries_to_incorporate(self):
1967
"""Yields pairs of (inventory_entry, new_parent)."""
1968
other_inv = self.other_tree.inventory
1969
subdir_id = other_inv.path2id(self._source_subpath)
1970
if subdir_id is None:
1971
# XXX: The error would be clearer if it gave the URL of the source
1972
# branch, but we don't have a reference to that here.
1973
raise PathNotInTree(self._source_subpath, "Source tree")
1974
subdir = other_inv[subdir_id]
1975
parent_in_target = osutils.dirname(self._target_subdir)
1976
target_id = self.this_tree.inventory.path2id(parent_in_target)
1977
if target_id is None:
1978
raise PathNotInTree(self._target_subdir, "Target tree")
1979
name_in_target = osutils.basename(self._target_subdir)
1980
merge_into_root = subdir.copy()
1981
merge_into_root.name = name_in_target
1982
if self.this_tree.inventory.has_id(merge_into_root.file_id):
1983
# Give the root a new file-id.
1984
# This can happen fairly easily if the directory we are
1985
# incorporating is the root, and both trees have 'TREE_ROOT' as
1986
# their root_id. Users will expect this to Just Work, so we
1987
# change the file-id here.
1988
# Non-root file-ids could potentially conflict too. That's really
1989
# an edge case, so we don't do anything special for those. We let
1990
# them cause conflicts.
1991
merge_into_root.file_id = generate_ids.gen_file_id(name_in_target)
1992
yield (merge_into_root, target_id)
1993
if subdir.kind != 'directory':
1994
# No children, so we are done.
1996
for ignored_path, entry in other_inv.iter_entries_by_dir(subdir_id):
1997
parent_id = entry.parent_id
1998
if parent_id == subdir.file_id:
1999
# The root's parent ID has changed, so make sure children of
2000
# the root refer to the new ID.
2001
parent_id = merge_into_root.file_id
2002
yield (entry, parent_id)
2005
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
2007
merge_type=Merge3Merger,
2008
interesting_ids=None,
2012
interesting_files=None,
2015
change_reporter=None):
2016
"""Primary interface for merging.
2018
Typical use is probably::
2020
merge_inner(branch, branch.get_revision_tree(other_revision),
2021
branch.get_revision_tree(base_revision))
2023
if this_tree is None:
2024
raise errors.BzrError("bzrlib.merge.merge_inner requires a this_tree "
2026
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
2027
pb=pb, change_reporter=change_reporter)
2028
merger.backup_files = backup_files
2029
merger.merge_type = merge_type
2030
merger.interesting_ids = interesting_ids
2031
merger.ignore_zero = ignore_zero
2032
if interesting_files:
2034
raise ValueError('Only supply interesting_ids'
2035
' or interesting_files')
2036
merger.interesting_files = interesting_files
2037
merger.show_base = show_base
2038
merger.reprocess = reprocess
2039
merger.other_rev_id = other_rev_id
2040
merger.other_basis = other_rev_id
2041
get_revision_id = getattr(base_tree, 'get_revision_id', None)
2042
if get_revision_id is None:
2043
get_revision_id = base_tree.last_revision
2044
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
2045
merger.set_base_revision(get_revision_id(), this_branch)
2046
return merger.do_merge()
2049
merge_type_registry = registry.Registry()
2050
merge_type_registry.register('diff3', Diff3Merger,
2051
"Merge using external diff3.")
2052
merge_type_registry.register('lca', LCAMerger,
2053
"LCA-newness merge.")
2054
merge_type_registry.register('merge3', Merge3Merger,
2055
"Native diff3-style merge.")
2056
merge_type_registry.register('weave', WeaveMerger,
2057
"Weave-based merge.")
2060
def get_merge_type_registry():
2061
"""Merge type registry was previously in bzrlib.option
2063
This method provides a backwards compatible way to retrieve it.
2065
return merge_type_registry
2068
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
2069
def status_a(revision, text):
2070
if revision in ancestors_b:
2071
return 'killed-b', text
2073
return 'new-a', text
2075
def status_b(revision, text):
2076
if revision in ancestors_a:
2077
return 'killed-a', text
2079
return 'new-b', text
2081
plain_a = [t for (a, t) in annotated_a]
2082
plain_b = [t for (a, t) in annotated_b]
2083
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
2084
blocks = matcher.get_matching_blocks()
2087
for ai, bi, l in blocks:
2088
# process all mismatched sections
2089
# (last mismatched section is handled because blocks always
2090
# includes a 0-length last block)
2091
for revision, text in annotated_a[a_cur:ai]:
2092
yield status_a(revision, text)
2093
for revision, text in annotated_b[b_cur:bi]:
2094
yield status_b(revision, text)
2095
# and now the matched section
2098
for text_a in plain_a[ai:a_cur]:
2099
yield "unchanged", text_a
2102
class _PlanMergeBase(object):
2104
def __init__(self, a_rev, b_rev, vf, key_prefix):
2107
:param a_rev: Revision-id of one revision to merge
2108
:param b_rev: Revision-id of the other revision to merge
2109
:param vf: A VersionedFiles containing both revisions
2110
:param key_prefix: A prefix for accessing keys in vf, typically
2116
self._last_lines = None
2117
self._last_lines_revision_id = None
2118
self._cached_matching_blocks = {}
2119
self._key_prefix = key_prefix
2120
self._precache_tip_lines()
2122
def _precache_tip_lines(self):
2123
lines = self.get_lines([self.a_rev, self.b_rev])
2124
self.lines_a = lines[self.a_rev]
2125
self.lines_b = lines[self.b_rev]
2127
def get_lines(self, revisions):
2128
"""Get lines for revisions from the backing VersionedFiles.
2130
:raises RevisionNotPresent: on absent texts.
2132
keys = [(self._key_prefix + (rev,)) for rev in revisions]
2134
for record in self.vf.get_record_stream(keys, 'unordered', True):
2135
if record.storage_kind == 'absent':
2136
raise errors.RevisionNotPresent(record.key, self.vf)
2137
result[record.key[-1]] = osutils.chunks_to_lines(
2138
record.get_bytes_as('chunked'))
2141
def plan_merge(self):
2142
"""Generate a 'plan' for merging the two revisions.
2144
This involves comparing their texts and determining the cause of
2145
differences. If text A has a line and text B does not, then either the
2146
line was added to text A, or it was deleted from B. Once the causes
2147
are combined, they are written out in the format described in
2148
VersionedFile.plan_merge
2150
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
2151
unique_a, unique_b = self._unique_lines(blocks)
2152
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
2153
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
2154
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
2156
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
2159
for i, j, n in blocks:
2160
for a_index in range(last_i, i):
2161
if a_index in new_a:
2162
if a_index in killed_b:
2163
yield 'conflicted-a', self.lines_a[a_index]
2165
yield 'new-a', self.lines_a[a_index]
2167
yield 'killed-b', self.lines_a[a_index]
2168
for b_index in range(last_j, j):
2169
if b_index in new_b:
2170
if b_index in killed_a:
2171
yield 'conflicted-b', self.lines_b[b_index]
2173
yield 'new-b', self.lines_b[b_index]
2175
yield 'killed-a', self.lines_b[b_index]
2176
# handle common lines
2177
for a_index in range(i, i+n):
2178
yield 'unchanged', self.lines_a[a_index]
2182
def _get_matching_blocks(self, left_revision, right_revision):
2183
"""Return a description of which sections of two revisions match.
2185
See SequenceMatcher.get_matching_blocks
2187
cached = self._cached_matching_blocks.get((left_revision,
2189
if cached is not None:
2191
if self._last_lines_revision_id == left_revision:
2192
left_lines = self._last_lines
2193
right_lines = self.get_lines([right_revision])[right_revision]
2195
lines = self.get_lines([left_revision, right_revision])
2196
left_lines = lines[left_revision]
2197
right_lines = lines[right_revision]
2198
self._last_lines = right_lines
2199
self._last_lines_revision_id = right_revision
2200
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
2202
return matcher.get_matching_blocks()
2204
def _unique_lines(self, matching_blocks):
2205
"""Analyse matching_blocks to determine which lines are unique
2207
:return: a tuple of (unique_left, unique_right), where the values are
2208
sets of line numbers of unique lines.
2214
for i, j, n in matching_blocks:
2215
unique_left.extend(range(last_i, i))
2216
unique_right.extend(range(last_j, j))
2219
return unique_left, unique_right
2222
def _subtract_plans(old_plan, new_plan):
2223
"""Remove changes from new_plan that came from old_plan.
2225
It is assumed that the difference between the old_plan and new_plan
2226
is their choice of 'b' text.
2228
All lines from new_plan that differ from old_plan are emitted
2229
verbatim. All lines from new_plan that match old_plan but are
2230
not about the 'b' revision are emitted verbatim.
2232
Lines that match and are about the 'b' revision are the lines we
2233
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
2234
is skipped entirely.
2236
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
2239
for i, j, n in matcher.get_matching_blocks():
2240
for jj in range(last_j, j):
2242
for jj in range(j, j+n):
2243
plan_line = new_plan[jj]
2244
if plan_line[0] == 'new-b':
2246
elif plan_line[0] == 'killed-b':
2247
yield 'unchanged', plan_line[1]
2253
class _PlanMerge(_PlanMergeBase):
2254
"""Plan an annotate merge using on-the-fly annotation"""
2256
def __init__(self, a_rev, b_rev, vf, key_prefix):
2257
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
2258
self.a_key = self._key_prefix + (self.a_rev,)
2259
self.b_key = self._key_prefix + (self.b_rev,)
2260
self.graph = _mod_graph.Graph(self.vf)
2261
heads = self.graph.heads((self.a_key, self.b_key))
2263
# one side dominates, so we can just return its values, yay for
2265
# Ideally we would know that before we get this far
2266
self._head_key = heads.pop()
2267
if self._head_key == self.a_key:
2271
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2272
self._head_key[-1], other)
2275
self._head_key = None
2278
def _precache_tip_lines(self):
2279
# Turn this into a no-op, because we will do this later
2282
def _find_recursive_lcas(self):
2283
"""Find all the ancestors back to a unique lca"""
2284
cur_ancestors = (self.a_key, self.b_key)
2285
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2286
# rather than a key tuple. We will just map that directly to no common
2290
next_lcas = self.graph.find_lca(*cur_ancestors)
2291
# Map a plain NULL_REVISION to a simple no-ancestors
2292
if next_lcas == set([_mod_revision.NULL_REVISION]):
2294
# Order the lca's based on when they were merged into the tip
2295
# While the actual merge portion of weave merge uses a set() of
2296
# active revisions, the order of insertion *does* effect the
2297
# implicit ordering of the texts.
2298
for rev_key in cur_ancestors:
2299
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2301
parent_map[rev_key] = ordered_parents
2302
if len(next_lcas) == 0:
2304
elif len(next_lcas) == 1:
2305
parent_map[list(next_lcas)[0]] = ()
2307
elif len(next_lcas) > 2:
2308
# More than 2 lca's, fall back to grabbing all nodes between
2309
# this and the unique lca.
2310
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2312
self.a_key, self.b_key, cur_ancestors)
2313
cur_lcas = next_lcas
2314
while len(cur_lcas) > 1:
2315
cur_lcas = self.graph.find_lca(*cur_lcas)
2316
if len(cur_lcas) == 0:
2317
# No common base to find, use the full ancestry
2320
unique_lca = list(cur_lcas)[0]
2321
if unique_lca == _mod_revision.NULL_REVISION:
2322
# find_lca will return a plain 'NULL_REVISION' rather
2323
# than a key tuple when there is no common ancestor, we
2324
# prefer to just use None, because it doesn't confuse
2325
# _get_interesting_texts()
2327
parent_map.update(self._find_unique_parents(next_lcas,
2330
cur_ancestors = next_lcas
2333
def _find_unique_parents(self, tip_keys, base_key):
2334
"""Find ancestors of tip that aren't ancestors of base.
2336
:param tip_keys: Nodes that are interesting
2337
:param base_key: Cull all ancestors of this node
2338
:return: The parent map for all revisions between tip_keys and
2339
base_key. base_key will be included. References to nodes outside of
2340
the ancestor set will also be removed.
2342
# TODO: this would be simpler if find_unique_ancestors took a list
2343
# instead of a single tip, internally it supports it, but it
2344
# isn't a "backwards compatible" api change.
2345
if base_key is None:
2346
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2347
# We remove NULL_REVISION because it isn't a proper tuple key, and
2348
# thus confuses things like _get_interesting_texts, and our logic
2349
# to add the texts into the memory weave.
2350
if _mod_revision.NULL_REVISION in parent_map:
2351
parent_map.pop(_mod_revision.NULL_REVISION)
2354
for tip in tip_keys:
2356
self.graph.find_unique_ancestors(tip, [base_key]))
2357
parent_map = self.graph.get_parent_map(interesting)
2358
parent_map[base_key] = ()
2359
culled_parent_map, child_map, tails = self._remove_external_references(
2361
# Remove all the tails but base_key
2362
if base_key is not None:
2363
tails.remove(base_key)
2364
self._prune_tails(culled_parent_map, child_map, tails)
2365
# Now remove all the uninteresting 'linear' regions
2366
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2370
def _remove_external_references(parent_map):
2371
"""Remove references that go outside of the parent map.
2373
:param parent_map: Something returned from Graph.get_parent_map(keys)
2374
:return: (filtered_parent_map, child_map, tails)
2375
filtered_parent_map is parent_map without external references
2376
child_map is the {parent_key: [child_keys]} mapping
2377
tails is a list of nodes that do not have any parents in the map
2379
# TODO: The basic effect of this function seems more generic than
2380
# _PlanMerge. But the specific details of building a child_map,
2381
# and computing tails seems very specific to _PlanMerge.
2382
# Still, should this be in Graph land?
2383
filtered_parent_map = {}
2386
for key, parent_keys in parent_map.iteritems():
2387
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2388
if not culled_parent_keys:
2390
for parent_key in culled_parent_keys:
2391
child_map.setdefault(parent_key, []).append(key)
2392
# TODO: Do we want to do this, it adds overhead for every node,
2393
# just to say that the node has no children
2394
child_map.setdefault(key, [])
2395
filtered_parent_map[key] = culled_parent_keys
2396
return filtered_parent_map, child_map, tails
2399
def _prune_tails(parent_map, child_map, tails_to_remove):
2400
"""Remove tails from the parent map.
2402
This will remove the supplied revisions until no more children have 0
2405
:param parent_map: A dict of {child: [parents]}, this dictionary will
2406
be modified in place.
2407
:param tails_to_remove: A list of tips that should be removed,
2408
this list will be consumed
2409
:param child_map: The reverse dict of parent_map ({parent: [children]})
2410
this dict will be modified
2411
:return: None, parent_map will be modified in place.
2413
while tails_to_remove:
2414
next = tails_to_remove.pop()
2415
parent_map.pop(next)
2416
children = child_map.pop(next)
2417
for child in children:
2418
child_parents = parent_map[child]
2419
child_parents.remove(next)
2420
if len(child_parents) == 0:
2421
tails_to_remove.append(child)
2423
def _get_interesting_texts(self, parent_map):
2424
"""Return a dict of texts we are interested in.
2426
Note that the input is in key tuples, but the output is in plain
2429
:param parent_map: The output from _find_recursive_lcas
2430
:return: A dict of {'revision_id':lines} as returned by
2431
_PlanMergeBase.get_lines()
2433
all_revision_keys = set(parent_map)
2434
all_revision_keys.add(self.a_key)
2435
all_revision_keys.add(self.b_key)
2437
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2438
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2441
def _build_weave(self):
2442
from bzrlib import weave
2443
self._weave = weave.Weave(weave_name='in_memory_weave',
2444
allow_reserved=True)
2445
parent_map = self._find_recursive_lcas()
2447
all_texts = self._get_interesting_texts(parent_map)
2449
# Note: Unfortunately, the order given by topo_sort will effect the
2450
# ordering resolution in the output. Specifically, if you add A then B,
2451
# then in the output text A lines will show up before B lines. And, of
2452
# course, topo_sort doesn't guarantee any real ordering.
2453
# So we use merge_sort, and add a fake node on the tip.
2454
# This ensures that left-hand parents will always be inserted into the
2455
# weave before right-hand parents.
2456
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2457
parent_map[tip_key] = (self.a_key, self.b_key)
2459
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2463
# for key in tsort.topo_sort(parent_map):
2464
parent_keys = parent_map[key]
2465
revision_id = key[-1]
2466
parent_ids = [k[-1] for k in parent_keys]
2467
self._weave.add_lines(revision_id, parent_ids,
2468
all_texts[revision_id])
2470
def plan_merge(self):
2471
"""Generate a 'plan' for merging the two revisions.
2473
This involves comparing their texts and determining the cause of
2474
differences. If text A has a line and text B does not, then either the
2475
line was added to text A, or it was deleted from B. Once the causes
2476
are combined, they are written out in the format described in
2477
VersionedFile.plan_merge
2479
if self._head_key is not None: # There was a single head
2480
if self._head_key == self.a_key:
2483
if self._head_key != self.b_key:
2484
raise AssertionError('There was an invalid head: %s != %s'
2485
% (self.b_key, self._head_key))
2487
head_rev = self._head_key[-1]
2488
lines = self.get_lines([head_rev])[head_rev]
2489
return ((plan, line) for line in lines)
2490
return self._weave.plan_merge(self.a_rev, self.b_rev)
2493
class _PlanLCAMerge(_PlanMergeBase):
2495
This merge algorithm differs from _PlanMerge in that:
2497
1. comparisons are done against LCAs only
2498
2. cases where a contested line is new versus one LCA but old versus
2499
another are marked as conflicts, by emitting the line as conflicted-a
2502
This is faster, and hopefully produces more useful output.
2505
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2506
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2507
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2510
if lca == _mod_revision.NULL_REVISION:
2513
self.lcas.add(lca[-1])
2514
for lca in self.lcas:
2515
if _mod_revision.is_null(lca):
2518
lca_lines = self.get_lines([lca])[lca]
2519
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2521
blocks = list(matcher.get_matching_blocks())
2522
self._cached_matching_blocks[(a_rev, lca)] = blocks
2523
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2525
blocks = list(matcher.get_matching_blocks())
2526
self._cached_matching_blocks[(b_rev, lca)] = blocks
2528
def _determine_status(self, revision_id, unique_line_numbers):
2529
"""Determines the status unique lines versus all lcas.
2531
Basically, determines why the line is unique to this revision.
2533
A line may be determined new, killed, or both.
2535
If a line is determined new, that means it was not present in at least
2536
one LCA, and is not present in the other merge revision.
2538
If a line is determined killed, that means the line was present in
2541
If a line is killed and new, this indicates that the two merge
2542
revisions contain differing conflict resolutions.
2544
:param revision_id: The id of the revision in which the lines are
2546
:param unique_line_numbers: The line numbers of unique lines.
2547
:return: a tuple of (new_this, killed_other)
2551
unique_line_numbers = set(unique_line_numbers)
2552
for lca in self.lcas:
2553
blocks = self._get_matching_blocks(revision_id, lca)
2554
unique_vs_lca, _ignored = self._unique_lines(blocks)
2555
new.update(unique_line_numbers.intersection(unique_vs_lca))
2556
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
