1
# Copyright (C) 2005-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
22
branch as _mod_branch,
24
conflicts as _mod_conflicts,
31
revision as _mod_revision,
47
from bzrlib.symbol_versioning import (
51
# TODO: Report back as changes are merged in
54
def transform_tree(from_tree, to_tree, interesting_ids=None):
55
from_tree.lock_tree_write()
56
operation = cleanup.OperationWithCleanups(merge_inner)
57
operation.add_cleanup(from_tree.unlock)
58
operation.run_simple(from_tree.branch, to_tree, from_tree,
59
ignore_zero=True, interesting_ids=interesting_ids, this_tree=from_tree)
62
class MergeHooks(hooks.Hooks):
65
hooks.Hooks.__init__(self, "bzrlib.merge", "Merger.hooks")
66
self.add_hook('merge_file_content',
67
"Called with a bzrlib.merge.Merger object to create a per file "
68
"merge object when starting a merge. "
69
"Should return either None or a subclass of "
70
"``bzrlib.merge.AbstractPerFileMerger``. "
71
"Such objects will then be called per file "
72
"that needs to be merged (including when one "
73
"side has deleted the file and the other has changed it). "
74
"See the AbstractPerFileMerger API docs for details on how it is "
79
class AbstractPerFileMerger(object):
80
"""PerFileMerger objects are used by plugins extending merge for bzrlib.
82
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
84
:ivar merger: The Merge3Merger performing the merge.
87
def __init__(self, merger):
88
"""Create a PerFileMerger for use with merger."""
91
def merge_contents(self, merge_params):
92
"""Attempt to merge the contents of a single file.
94
:param merge_params: A bzrlib.merge.MergeHookParams
95
:return: A tuple of (status, chunks), where status is one of
96
'not_applicable', 'success', 'conflicted', or 'delete'. If status
97
is 'success' or 'conflicted', then chunks should be an iterable of
98
strings for the new file contents.
100
return ('not applicable', None)
103
class PerFileMerger(AbstractPerFileMerger):
104
"""Merge individual files when self.file_matches returns True.
106
This class is intended to be subclassed. The file_matches and
107
merge_matching methods should be overridden with concrete implementations.
110
def file_matches(self, params):
111
"""Return True if merge_matching should be called on this file.
113
Only called with merges of plain files with no clear winner.
115
Subclasses must override this.
117
raise NotImplementedError(self.file_matches)
119
def get_filename(self, params, tree):
120
"""Lookup the filename (i.e. basename, not path), given a Tree (e.g.
121
self.merger.this_tree) and a MergeHookParams.
123
return osutils.basename(tree.id2path(params.file_id))
125
def get_filepath(self, params, tree):
126
"""Calculate the path to the file in a tree.
128
:param params: A MergeHookParams describing the file to merge
129
:param tree: a Tree, e.g. self.merger.this_tree.
131
return tree.id2path(params.file_id)
133
def merge_contents(self, params):
134
"""Merge the contents of a single file."""
135
# Check whether this custom merge logic should be used.
137
# OTHER is a straight winner, rely on default merge.
138
params.winner == 'other' or
139
# THIS and OTHER aren't both files.
140
not params.is_file_merge() or
141
# The filename doesn't match *.xml
142
not self.file_matches(params)):
143
return 'not_applicable', None
144
return self.merge_matching(params)
146
def merge_matching(self, params):
147
"""Merge the contents of a single file that has matched the criteria
148
in PerFileMerger.merge_contents (is a conflict, is a file,
149
self.file_matches is True).
151
Subclasses must override this.
153
raise NotImplementedError(self.merge_matching)
156
class ConfigurableFileMerger(PerFileMerger):
157
"""Merge individual files when configured via a .conf file.
159
This is a base class for concrete custom file merging logic. Concrete
160
classes should implement ``merge_text``.
162
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
164
:ivar affected_files: The configured file paths to merge.
166
:cvar name_prefix: The prefix to use when looking up configuration
167
details. <name_prefix>_merge_files describes the files targeted by the
170
:cvar default_files: The default file paths to merge when no configuration
177
def __init__(self, merger):
178
super(ConfigurableFileMerger, self).__init__(merger)
179
self.affected_files = None
180
self.default_files = self.__class__.default_files or []
181
self.name_prefix = self.__class__.name_prefix
182
if self.name_prefix is None:
183
raise ValueError("name_prefix must be set.")
185
def file_matches(self, params):
186
"""Check whether the file should call the merge hook.
188
<name_prefix>_merge_files configuration variable is a list of files
189
that should use the hook.
191
affected_files = self.affected_files
192
if affected_files is None:
193
config = self.merger.this_branch.get_config()
194
# Until bzr provides a better policy for caching the config, we
195
# just add the part we're interested in to the params to avoid
196
# reading the config files repeatedly (bazaar.conf, location.conf,
198
config_key = self.name_prefix + '_merge_files'
199
affected_files = config.get_user_option_as_list(config_key)
200
if affected_files is None:
201
# If nothing was specified in the config, use the default.
202
affected_files = self.default_files
203
self.affected_files = affected_files
205
filepath = self.get_filepath(params, self.merger.this_tree)
206
if filepath in affected_files:
210
def merge_matching(self, params):
211
return self.merge_text(params)
213
def merge_text(self, params):
214
"""Merge the byte contents of a single file.
216
This is called after checking that the merge should be performed in
217
merge_contents, and it should behave as per
218
``bzrlib.merge.AbstractPerFileMerger.merge_contents``.
220
raise NotImplementedError(self.merge_text)
223
class MergeHookParams(object):
224
"""Object holding parameters passed to merge_file_content hooks.
226
There are some fields hooks can access:
228
:ivar file_id: the file ID of the file being merged
229
:ivar trans_id: the transform ID for the merge of this file
230
:ivar this_kind: kind of file_id in 'this' tree
231
:ivar other_kind: kind of file_id in 'other' tree
232
:ivar winner: one of 'this', 'other', 'conflict'
235
def __init__(self, merger, file_id, trans_id, this_kind, other_kind,
237
self._merger = merger
238
self.file_id = file_id
239
self.trans_id = trans_id
240
self.this_kind = this_kind
241
self.other_kind = other_kind
244
def is_file_merge(self):
245
"""True if this_kind and other_kind are both 'file'."""
246
return self.this_kind == 'file' and self.other_kind == 'file'
248
@decorators.cachedproperty
249
def base_lines(self):
250
"""The lines of the 'base' version of the file."""
251
return self._merger.get_lines(self._merger.base_tree, self.file_id)
253
@decorators.cachedproperty
254
def this_lines(self):
255
"""The lines of the 'this' version of the file."""
256
return self._merger.get_lines(self._merger.this_tree, self.file_id)
258
@decorators.cachedproperty
259
def other_lines(self):
260
"""The lines of the 'other' version of the file."""
261
return self._merger.get_lines(self._merger.other_tree, self.file_id)
264
class Merger(object):
268
def __init__(self, this_branch, other_tree=None, base_tree=None,
269
this_tree=None, pb=None, change_reporter=None,
270
recurse='down', revision_graph=None):
271
object.__init__(self)
272
self.this_branch = this_branch
273
self.this_basis = _mod_revision.ensure_null(
274
this_branch.last_revision())
275
self.this_rev_id = None
276
self.this_tree = this_tree
277
self.this_revision_tree = None
278
self.this_basis_tree = None
279
self.other_tree = other_tree
280
self.other_branch = None
281
self.base_tree = base_tree
282
self.ignore_zero = False
283
self.backup_files = False
284
self.interesting_ids = None
285
self.interesting_files = None
286
self.show_base = False
287
self.reprocess = False
289
warnings.warn("pb parameter to Merger() is deprecated and ignored")
291
self.recurse = recurse
292
self.change_reporter = change_reporter
293
self._cached_trees = {}
294
self._revision_graph = revision_graph
295
self._base_is_ancestor = None
296
self._base_is_other_ancestor = None
297
self._is_criss_cross = None
298
self._lca_trees = None
300
def cache_trees_with_revision_ids(self, trees):
301
"""Cache any tree in trees if it has a revision_id."""
302
for maybe_tree in trees:
303
if maybe_tree is None:
306
rev_id = maybe_tree.get_revision_id()
307
except AttributeError:
309
self._cached_trees[rev_id] = maybe_tree
312
def revision_graph(self):
313
if self._revision_graph is None:
314
self._revision_graph = self.this_branch.repository.get_graph()
315
return self._revision_graph
317
def _set_base_is_ancestor(self, value):
318
self._base_is_ancestor = value
320
def _get_base_is_ancestor(self):
321
if self._base_is_ancestor is None:
322
self._base_is_ancestor = self.revision_graph.is_ancestor(
323
self.base_rev_id, self.this_basis)
324
return self._base_is_ancestor
326
base_is_ancestor = property(_get_base_is_ancestor, _set_base_is_ancestor)
328
def _set_base_is_other_ancestor(self, value):
329
self._base_is_other_ancestor = value
331
def _get_base_is_other_ancestor(self):
332
if self._base_is_other_ancestor is None:
333
if self.other_basis is None:
335
self._base_is_other_ancestor = self.revision_graph.is_ancestor(
336
self.base_rev_id, self.other_basis)
337
return self._base_is_other_ancestor
339
base_is_other_ancestor = property(_get_base_is_other_ancestor,
340
_set_base_is_other_ancestor)
343
def from_uncommitted(tree, other_tree, pb=None, base_tree=None):
344
"""Return a Merger for uncommitted changes in other_tree.
346
:param tree: The tree to merge into
347
:param other_tree: The tree to get uncommitted changes from
348
:param pb: A progress indicator
349
:param base_tree: The basis to use for the merge. If unspecified,
350
other_tree.basis_tree() will be used.
352
if base_tree is None:
353
base_tree = other_tree.basis_tree()
354
merger = Merger(tree.branch, other_tree, base_tree, tree, pb)
355
merger.base_rev_id = merger.base_tree.get_revision_id()
356
merger.other_rev_id = None
357
merger.other_basis = merger.base_rev_id
361
def from_mergeable(klass, tree, mergeable, pb):
362
"""Return a Merger for a bundle or merge directive.
364
:param tree: The tree to merge changes into
365
:param mergeable: A merge directive or bundle
366
:param pb: A progress indicator
368
mergeable.install_revisions(tree.branch.repository)
369
base_revision_id, other_revision_id, verified =\
370
mergeable.get_merge_request(tree.branch.repository)
371
revision_graph = tree.branch.repository.get_graph()
372
if base_revision_id is not None:
373
if (base_revision_id != _mod_revision.NULL_REVISION and
374
revision_graph.is_ancestor(
375
base_revision_id, tree.branch.last_revision())):
376
base_revision_id = None
378
trace.warning('Performing cherrypick')
379
merger = klass.from_revision_ids(pb, tree, other_revision_id,
380
base_revision_id, revision_graph=
382
return merger, verified
385
def from_revision_ids(pb, tree, other, base=None, other_branch=None,
386
base_branch=None, revision_graph=None,
388
"""Return a Merger for revision-ids.
390
:param pb: A progress indicator
391
:param tree: The tree to merge changes into
392
:param other: The revision-id to use as OTHER
393
:param base: The revision-id to use as BASE. If not specified, will
395
:param other_branch: A branch containing the other revision-id. If
396
not supplied, tree.branch is used.
397
:param base_branch: A branch containing the base revision-id. If
398
not supplied, other_branch or tree.branch will be used.
399
:param revision_graph: If you have a revision_graph precomputed, pass
400
it in, otherwise it will be created for you.
401
:param tree_branch: The branch associated with tree. If not supplied,
402
tree.branch will be used.
404
if tree_branch is None:
405
tree_branch = tree.branch
406
merger = Merger(tree_branch, this_tree=tree, pb=pb,
407
revision_graph=revision_graph)
408
if other_branch is None:
409
other_branch = tree.branch
410
merger.set_other_revision(other, other_branch)
414
if base_branch is None:
415
base_branch = other_branch
416
merger.set_base_revision(base, base_branch)
419
def revision_tree(self, revision_id, branch=None):
420
if revision_id not in self._cached_trees:
422
branch = self.this_branch
424
tree = self.this_tree.revision_tree(revision_id)
425
except errors.NoSuchRevisionInTree:
426
tree = branch.repository.revision_tree(revision_id)
427
self._cached_trees[revision_id] = tree
428
return self._cached_trees[revision_id]
430
def _get_tree(self, treespec, possible_transports=None):
431
location, revno = treespec
433
tree = workingtree.WorkingTree.open_containing(location)[0]
434
return tree.branch, tree
435
branch = _mod_branch.Branch.open_containing(
436
location, possible_transports)[0]
438
revision_id = branch.last_revision()
440
revision_id = branch.get_rev_id(revno)
441
revision_id = _mod_revision.ensure_null(revision_id)
442
return branch, self.revision_tree(revision_id, branch)
444
@deprecated_method(deprecated_in((2, 1, 0)))
445
def ensure_revision_trees(self):
446
if self.this_revision_tree is None:
447
self.this_basis_tree = self.revision_tree(self.this_basis)
448
if self.this_basis == self.this_rev_id:
449
self.this_revision_tree = self.this_basis_tree
451
if self.other_rev_id is None:
452
other_basis_tree = self.revision_tree(self.other_basis)
453
if other_basis_tree.has_changes(self.other_tree):
454
raise errors.WorkingTreeNotRevision(self.this_tree)
455
other_rev_id = self.other_basis
456
self.other_tree = other_basis_tree
458
@deprecated_method(deprecated_in((2, 1, 0)))
459
def file_revisions(self, file_id):
460
self.ensure_revision_trees()
461
if self.this_rev_id is None:
462
if self.this_basis_tree.get_file_sha1(file_id) != \
463
self.this_tree.get_file_sha1(file_id):
464
raise errors.WorkingTreeNotRevision(self.this_tree)
466
trees = (self.this_basis_tree, self.other_tree)
467
return [tree.get_file_revision(file_id) for tree in trees]
469
@deprecated_method(deprecated_in((2, 1, 0)))
470
def check_basis(self, check_clean, require_commits=True):
471
if self.this_basis is None and require_commits is True:
472
raise errors.BzrCommandError(
473
"This branch has no commits."
474
" (perhaps you would prefer 'bzr pull')")
477
if self.this_basis != self.this_rev_id:
478
raise errors.UncommittedChanges(self.this_tree)
480
@deprecated_method(deprecated_in((2, 1, 0)))
481
def compare_basis(self):
483
basis_tree = self.revision_tree(self.this_tree.last_revision())
484
except errors.NoSuchRevision:
485
basis_tree = self.this_tree.basis_tree()
486
if not self.this_tree.has_changes(basis_tree):
487
self.this_rev_id = self.this_basis
489
def set_interesting_files(self, file_list):
490
self.interesting_files = file_list
492
def set_pending(self):
493
if (not self.base_is_ancestor or not self.base_is_other_ancestor
494
or self.other_rev_id is None):
498
def _add_parent(self):
499
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
500
new_parent_trees = []
501
operation = cleanup.OperationWithCleanups(
502
self.this_tree.set_parent_trees)
503
for revision_id in new_parents:
505
tree = self.revision_tree(revision_id)
506
except errors.NoSuchRevision:
510
operation.add_cleanup(tree.unlock)
511
new_parent_trees.append((revision_id, tree))
512
operation.run_simple(new_parent_trees, allow_leftmost_as_ghost=True)
514
def set_other(self, other_revision, possible_transports=None):
515
"""Set the revision and tree to merge from.
517
This sets the other_tree, other_rev_id, other_basis attributes.
519
:param other_revision: The [path, revision] list to merge from.
521
self.other_branch, self.other_tree = self._get_tree(other_revision,
523
if other_revision[1] == -1:
524
self.other_rev_id = _mod_revision.ensure_null(
525
self.other_branch.last_revision())
526
if _mod_revision.is_null(self.other_rev_id):
527
raise errors.NoCommits(self.other_branch)
528
self.other_basis = self.other_rev_id
529
elif other_revision[1] is not None:
530
self.other_rev_id = self.other_branch.get_rev_id(other_revision[1])
531
self.other_basis = self.other_rev_id
533
self.other_rev_id = None
534
self.other_basis = self.other_branch.last_revision()
535
if self.other_basis is None:
536
raise errors.NoCommits(self.other_branch)
537
if self.other_rev_id is not None:
538
self._cached_trees[self.other_rev_id] = self.other_tree
539
self._maybe_fetch(self.other_branch,self.this_branch, self.other_basis)
541
def set_other_revision(self, revision_id, other_branch):
542
"""Set 'other' based on a branch and revision id
544
:param revision_id: The revision to use for a tree
545
:param other_branch: The branch containing this tree
547
self.other_rev_id = revision_id
548
self.other_branch = other_branch
549
self._maybe_fetch(other_branch, self.this_branch, self.other_rev_id)
550
self.other_tree = self.revision_tree(revision_id)
551
self.other_basis = revision_id
553
def set_base_revision(self, revision_id, branch):
554
"""Set 'base' based on a branch and revision id
556
:param revision_id: The revision to use for a tree
557
:param branch: The branch containing this tree
559
self.base_rev_id = revision_id
560
self.base_branch = branch
561
self._maybe_fetch(branch, self.this_branch, revision_id)
562
self.base_tree = self.revision_tree(revision_id)
564
def _maybe_fetch(self, source, target, revision_id):
565
if not source.repository.has_same_location(target.repository):
566
target.fetch(source, revision_id)
569
revisions = [_mod_revision.ensure_null(self.this_basis),
570
_mod_revision.ensure_null(self.other_basis)]
571
if _mod_revision.NULL_REVISION in revisions:
572
self.base_rev_id = _mod_revision.NULL_REVISION
573
self.base_tree = self.revision_tree(self.base_rev_id)
574
self._is_criss_cross = False
576
lcas = self.revision_graph.find_lca(revisions[0], revisions[1])
577
self._is_criss_cross = False
579
self.base_rev_id = _mod_revision.NULL_REVISION
581
self.base_rev_id = list(lcas)[0]
582
else: # len(lcas) > 1
583
self._is_criss_cross = True
585
# find_unique_lca can only handle 2 nodes, so we have to
586
# start back at the beginning. It is a shame to traverse
587
# the graph again, but better than re-implementing
589
self.base_rev_id = self.revision_graph.find_unique_lca(
590
revisions[0], revisions[1])
592
self.base_rev_id = self.revision_graph.find_unique_lca(
594
sorted_lca_keys = self.revision_graph.find_merge_order(
596
if self.base_rev_id == _mod_revision.NULL_REVISION:
597
self.base_rev_id = sorted_lca_keys[0]
599
if self.base_rev_id == _mod_revision.NULL_REVISION:
600
raise errors.UnrelatedBranches()
601
if self._is_criss_cross:
602
trace.warning('Warning: criss-cross merge encountered. See bzr'
603
' help criss-cross.')
604
trace.mutter('Criss-cross lcas: %r' % lcas)
605
if self.base_rev_id in lcas:
606
trace.mutter('Unable to find unique lca. '
607
'Fallback %r as best option.' % self.base_rev_id)
608
interesting_revision_ids = set(lcas)
609
interesting_revision_ids.add(self.base_rev_id)
610
interesting_trees = dict((t.get_revision_id(), t)
611
for t in self.this_branch.repository.revision_trees(
612
interesting_revision_ids))
613
self._cached_trees.update(interesting_trees)
614
if self.base_rev_id in lcas:
615
self.base_tree = interesting_trees[self.base_rev_id]
617
self.base_tree = interesting_trees.pop(self.base_rev_id)
618
self._lca_trees = [interesting_trees[key]
619
for key in sorted_lca_keys]
621
self.base_tree = self.revision_tree(self.base_rev_id)
622
self.base_is_ancestor = True
623
self.base_is_other_ancestor = True
624
trace.mutter('Base revid: %r' % self.base_rev_id)
626
def set_base(self, base_revision):
627
"""Set the base revision to use for the merge.
629
:param base_revision: A 2-list containing a path and revision number.
631
trace.mutter("doing merge() with no base_revision specified")
632
if base_revision == [None, None]:
635
base_branch, self.base_tree = self._get_tree(base_revision)
636
if base_revision[1] == -1:
637
self.base_rev_id = base_branch.last_revision()
638
elif base_revision[1] is None:
639
self.base_rev_id = _mod_revision.NULL_REVISION
641
self.base_rev_id = _mod_revision.ensure_null(
642
base_branch.get_rev_id(base_revision[1]))
643
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
645
def make_merger(self):
646
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
647
'other_tree': self.other_tree,
648
'interesting_ids': self.interesting_ids,
649
'interesting_files': self.interesting_files,
650
'this_branch': self.this_branch,
652
if self.merge_type.requires_base:
653
kwargs['base_tree'] = self.base_tree
654
if self.merge_type.supports_reprocess:
655
kwargs['reprocess'] = self.reprocess
657
raise errors.BzrError(
658
"Conflict reduction is not supported for merge"
659
" type %s." % self.merge_type)
660
if self.merge_type.supports_show_base:
661
kwargs['show_base'] = self.show_base
663
raise errors.BzrError("Showing base is not supported for this"
664
" merge type. %s" % self.merge_type)
665
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
666
and not self.base_is_other_ancestor):
667
raise errors.CannotReverseCherrypick()
668
if self.merge_type.supports_cherrypick:
669
kwargs['cherrypick'] = (not self.base_is_ancestor or
670
not self.base_is_other_ancestor)
671
if self._is_criss_cross and getattr(self.merge_type,
672
'supports_lca_trees', False):
673
kwargs['lca_trees'] = self._lca_trees
674
return self.merge_type(pb=None,
675
change_reporter=self.change_reporter,
678
def _do_merge_to(self):
679
merge = self.make_merger()
680
if self.other_branch is not None:
681
self.other_branch.update_references(self.this_branch)
683
if self.recurse == 'down':
684
for relpath, file_id in self.this_tree.iter_references():
685
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
686
other_revision = self.other_tree.get_reference_revision(
688
if other_revision == sub_tree.last_revision():
690
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
691
sub_merge.merge_type = self.merge_type
692
other_branch = self.other_branch.reference_parent(file_id, relpath)
693
sub_merge.set_other_revision(other_revision, other_branch)
694
base_revision = self.base_tree.get_reference_revision(file_id)
695
sub_merge.base_tree = \
696
sub_tree.branch.repository.revision_tree(base_revision)
697
sub_merge.base_rev_id = base_revision
702
operation = cleanup.OperationWithCleanups(self._do_merge_to)
703
self.this_tree.lock_tree_write()
704
operation.add_cleanup(self.this_tree.unlock)
705
if self.base_tree is not None:
706
self.base_tree.lock_read()
707
operation.add_cleanup(self.base_tree.unlock)
708
if self.other_tree is not None:
709
self.other_tree.lock_read()
710
operation.add_cleanup(self.other_tree.unlock)
711
merge = operation.run_simple()
712
if len(merge.cooked_conflicts) == 0:
713
if not self.ignore_zero and not trace.is_quiet():
714
trace.note("All changes applied successfully.")
716
trace.note("%d conflicts encountered."
717
% len(merge.cooked_conflicts))
719
return len(merge.cooked_conflicts)
722
class _InventoryNoneEntry(object):
723
"""This represents an inventory entry which *isn't there*.
725
It simplifies the merging logic if we always have an InventoryEntry, even
726
if it isn't actually present
733
symlink_target = None
736
_none_entry = _InventoryNoneEntry()
739
class Merge3Merger(object):
740
"""Three-way merger that uses the merge3 text merger"""
742
supports_reprocess = True
743
supports_show_base = True
744
history_based = False
745
supports_cherrypick = True
746
supports_reverse_cherrypick = True
747
winner_idx = {"this": 2, "other": 1, "conflict": 1}
748
supports_lca_trees = True
750
def __init__(self, working_tree, this_tree, base_tree, other_tree,
751
interesting_ids=None, reprocess=False, show_base=False,
752
pb=None, pp=None, change_reporter=None,
753
interesting_files=None, do_merge=True,
754
cherrypick=False, lca_trees=None, this_branch=None):
755
"""Initialize the merger object and perform the merge.
757
:param working_tree: The working tree to apply the merge to
758
:param this_tree: The local tree in the merge operation
759
:param base_tree: The common tree in the merge operation
760
:param other_tree: The other tree to merge changes from
761
:param this_branch: The branch associated with this_tree. Defaults to
762
this_tree.branch if not supplied.
763
:param interesting_ids: The file_ids of files that should be
764
participate in the merge. May not be combined with
766
:param: reprocess If True, perform conflict-reduction processing.
767
:param show_base: If True, show the base revision in text conflicts.
768
(incompatible with reprocess)
770
:param pp: A ProgressPhase object
771
:param change_reporter: An object that should report changes made
772
:param interesting_files: The tree-relative paths of files that should
773
participate in the merge. If these paths refer to directories,
774
the contents of those directories will also be included. May not
775
be combined with interesting_ids. If neither interesting_files nor
776
interesting_ids is specified, all files may participate in the
778
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
779
if the ancestry was found to include a criss-cross merge.
780
Otherwise should be None.
782
object.__init__(self)
783
if interesting_files is not None and interesting_ids is not None:
785
'specify either interesting_ids or interesting_files')
786
if this_branch is None:
787
this_branch = this_tree.branch
788
self.interesting_ids = interesting_ids
789
self.interesting_files = interesting_files
790
self.this_tree = working_tree
791
self.base_tree = base_tree
792
self.other_tree = other_tree
793
self.this_branch = this_branch
794
self._raw_conflicts = []
795
self.cooked_conflicts = []
796
self.reprocess = reprocess
797
self.show_base = show_base
798
self._lca_trees = lca_trees
799
# Uncommenting this will change the default algorithm to always use
800
# _entries_lca. This can be useful for running the test suite and
801
# making sure we haven't missed any corner cases.
802
# if lca_trees is None:
803
# self._lca_trees = [self.base_tree]
804
self.change_reporter = change_reporter
805
self.cherrypick = cherrypick
809
warnings.warn("pp argument to Merge3Merger is deprecated")
811
warnings.warn("pb argument to Merge3Merger is deprecated")
814
operation = cleanup.OperationWithCleanups(self._do_merge)
815
self.this_tree.lock_tree_write()
816
operation.add_cleanup(self.this_tree.unlock)
817
self.base_tree.lock_read()
818
operation.add_cleanup(self.base_tree.unlock)
819
self.other_tree.lock_read()
820
operation.add_cleanup(self.other_tree.unlock)
823
def _do_merge(self, operation):
824
self.tt = transform.TreeTransform(self.this_tree, None)
825
operation.add_cleanup(self.tt.finalize)
826
self._compute_transform()
827
results = self.tt.apply(no_conflicts=True)
828
self.write_modified(results)
830
self.this_tree.add_conflicts(self.cooked_conflicts)
831
except errors.UnsupportedOperation:
834
def make_preview_transform(self):
835
operation = cleanup.OperationWithCleanups(self._make_preview_transform)
836
self.base_tree.lock_read()
837
operation.add_cleanup(self.base_tree.unlock)
838
self.other_tree.lock_read()
839
operation.add_cleanup(self.other_tree.unlock)
840
return operation.run_simple()
842
def _make_preview_transform(self):
843
self.tt = transform.TransformPreview(self.this_tree)
844
self._compute_transform()
847
def _compute_transform(self):
848
if self._lca_trees is None:
849
entries = self._entries3()
850
resolver = self._three_way
852
entries = self._entries_lca()
853
resolver = self._lca_multi_way
854
child_pb = ui.ui_factory.nested_progress_bar()
856
factories = Merger.hooks['merge_file_content']
857
hooks = [factory(self) for factory in factories] + [self]
858
self.active_hooks = [hook for hook in hooks if hook is not None]
859
for num, (file_id, changed, parents3, names3,
860
executable3) in enumerate(entries):
861
child_pb.update('Preparing file merge', num, len(entries))
862
self._merge_names(file_id, parents3, names3, resolver=resolver)
864
file_status = self._do_merge_contents(file_id)
866
file_status = 'unmodified'
867
self._merge_executable(file_id,
868
executable3, file_status, resolver=resolver)
872
self._finish_computing_transform()
874
def _finish_computing_transform(self):
875
"""Finalize the transform and report the changes.
877
This is the second half of _compute_transform.
879
child_pb = ui.ui_factory.nested_progress_bar()
881
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
882
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
885
if self.change_reporter is not None:
886
from bzrlib import delta
887
delta.report_changes(
888
self.tt.iter_changes(), self.change_reporter)
889
self.cook_conflicts(fs_conflicts)
890
for conflict in self.cooked_conflicts:
891
trace.warning(unicode(conflict))
894
"""Gather data about files modified between three trees.
896
Return a list of tuples of file_id, changed, parents3, names3,
897
executable3. changed is a boolean indicating whether the file contents
898
or kind were changed. parents3 is a tuple of parent ids for base,
899
other and this. names3 is a tuple of names for base, other and this.
900
executable3 is a tuple of execute-bit values for base, other and this.
903
iterator = self.other_tree.iter_changes(self.base_tree,
904
specific_files=self.interesting_files,
905
extra_trees=[self.this_tree])
906
this_entries = dict((e.file_id, e) for p, e in
907
self.this_tree.iter_entries_by_dir(
908
self.interesting_ids))
909
for (file_id, paths, changed, versioned, parents, names, kind,
910
executable) in iterator:
911
if (self.interesting_ids is not None and
912
file_id not in self.interesting_ids):
914
entry = this_entries.get(file_id)
915
if entry is not None:
916
this_name = entry.name
917
this_parent = entry.parent_id
918
this_executable = entry.executable
922
this_executable = None
923
parents3 = parents + (this_parent,)
924
names3 = names + (this_name,)
925
executable3 = executable + (this_executable,)
926
result.append((file_id, changed, parents3, names3, executable3))
929
def _entries_lca(self):
930
"""Gather data about files modified between multiple trees.
932
This compares OTHER versus all LCA trees, and for interesting entries,
933
it then compares with THIS and BASE.
935
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
937
:return: [(file_id, changed, parents, names, executable)], where:
939
* file_id: Simple file_id of the entry
940
* changed: Boolean, True if the kind or contents changed else False
941
* parents: ((base, [parent_id, in, lcas]), parent_id_other,
943
* names: ((base, [name, in, lcas]), name_in_other, name_in_this)
944
* executable: ((base, [exec, in, lcas]), exec_in_other,
947
if self.interesting_files is not None:
948
lookup_trees = [self.this_tree, self.base_tree]
949
lookup_trees.extend(self._lca_trees)
950
# I think we should include the lca trees as well
951
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
954
interesting_ids = self.interesting_ids
956
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
958
base_inventory = self.base_tree.inventory
959
this_inventory = self.this_tree.inventory
960
for path, file_id, other_ie, lca_values in walker.iter_all():
961
# Is this modified at all from any of the other trees?
963
other_ie = _none_entry
964
if interesting_ids is not None and file_id not in interesting_ids:
967
# If other_revision is found in any of the lcas, that means this
968
# node is uninteresting. This is because when merging, if there are
969
# multiple heads(), we have to create a new node. So if we didn't,
970
# we know that the ancestry is linear, and that OTHER did not
972
# See doc/developers/lca_merge_resolution.txt for details
973
other_revision = other_ie.revision
974
if other_revision is not None:
975
# We can't use this shortcut when other_revision is None,
976
# because it may be None because things are WorkingTrees, and
977
# not because it is *actually* None.
978
is_unmodified = False
979
for lca_path, ie in lca_values:
980
if ie is not None and ie.revision == other_revision:
987
for lca_path, lca_ie in lca_values:
989
lca_entries.append(_none_entry)
991
lca_entries.append(lca_ie)
993
if file_id in base_inventory:
994
base_ie = base_inventory[file_id]
996
base_ie = _none_entry
998
if file_id in this_inventory:
999
this_ie = this_inventory[file_id]
1001
this_ie = _none_entry
1007
for lca_ie in lca_entries:
1008
lca_kinds.append(lca_ie.kind)
1009
lca_parent_ids.append(lca_ie.parent_id)
1010
lca_names.append(lca_ie.name)
1011
lca_executable.append(lca_ie.executable)
1013
kind_winner = self._lca_multi_way(
1014
(base_ie.kind, lca_kinds),
1015
other_ie.kind, this_ie.kind)
1016
parent_id_winner = self._lca_multi_way(
1017
(base_ie.parent_id, lca_parent_ids),
1018
other_ie.parent_id, this_ie.parent_id)
1019
name_winner = self._lca_multi_way(
1020
(base_ie.name, lca_names),
1021
other_ie.name, this_ie.name)
1023
content_changed = True
1024
if kind_winner == 'this':
1025
# No kind change in OTHER, see if there are *any* changes
1026
if other_ie.kind == 'directory':
1027
if parent_id_winner == 'this' and name_winner == 'this':
1028
# No change for this directory in OTHER, skip
1030
content_changed = False
1031
elif other_ie.kind is None or other_ie.kind == 'file':
1032
def get_sha1(ie, tree):
1033
if ie.kind != 'file':
1035
return tree.get_file_sha1(file_id)
1036
base_sha1 = get_sha1(base_ie, self.base_tree)
1037
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1038
in zip(lca_entries, self._lca_trees)]
1039
this_sha1 = get_sha1(this_ie, self.this_tree)
1040
other_sha1 = get_sha1(other_ie, self.other_tree)
1041
sha1_winner = self._lca_multi_way(
1042
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1043
allow_overriding_lca=False)
1044
exec_winner = self._lca_multi_way(
1045
(base_ie.executable, lca_executable),
1046
other_ie.executable, this_ie.executable)
1047
if (parent_id_winner == 'this' and name_winner == 'this'
1048
and sha1_winner == 'this' and exec_winner == 'this'):
1049
# No kind, parent, name, exec, or content change for
1050
# OTHER, so this node is not considered interesting
1052
if sha1_winner == 'this':
1053
content_changed = False
1054
elif other_ie.kind == 'symlink':
1055
def get_target(ie, tree):
1056
if ie.kind != 'symlink':
1058
return tree.get_symlink_target(file_id)
1059
base_target = get_target(base_ie, self.base_tree)
1060
lca_targets = [get_target(ie, tree) for ie, tree
1061
in zip(lca_entries, self._lca_trees)]
1062
this_target = get_target(this_ie, self.this_tree)
1063
other_target = get_target(other_ie, self.other_tree)
1064
target_winner = self._lca_multi_way(
1065
(base_target, lca_targets),
1066
other_target, this_target)
1067
if (parent_id_winner == 'this' and name_winner == 'this'
1068
and target_winner == 'this'):
1069
# No kind, parent, name, or symlink target change
1072
if target_winner == 'this':
1073
content_changed = False
1074
elif other_ie.kind == 'tree-reference':
1075
# The 'changed' information seems to be handled at a higher
1076
# level. At least, _entries3 returns False for content
1077
# changed, even when at a new revision_id.
1078
content_changed = False
1079
if (parent_id_winner == 'this' and name_winner == 'this'):
1080
# Nothing interesting
1083
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1085
# If we have gotten this far, that means something has changed
1086
result.append((file_id, content_changed,
1087
((base_ie.parent_id, lca_parent_ids),
1088
other_ie.parent_id, this_ie.parent_id),
1089
((base_ie.name, lca_names),
1090
other_ie.name, this_ie.name),
1091
((base_ie.executable, lca_executable),
1092
other_ie.executable, this_ie.executable)
1097
if self.tt.final_kind(self.tt.root) is None:
1098
self.tt.cancel_deletion(self.tt.root)
1099
if self.tt.final_file_id(self.tt.root) is None:
1100
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
1102
other_root_file_id = self.other_tree.get_root_id()
1103
if other_root_file_id is None:
1105
other_root = self.tt.trans_id_file_id(other_root_file_id)
1106
if other_root == self.tt.root:
1108
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
1109
# the other tree's root is a non-root in the current tree (as when
1110
# a previously unrelated branch is merged into another)
1112
if self.tt.final_kind(other_root) is not None:
1113
other_root_is_present = True
1115
# other_root doesn't have a physical representation. We still need
1116
# to move any references to the actual root of the tree.
1117
other_root_is_present = False
1118
# 'other_tree.inventory.root' is not present in this tree. We are
1119
# calling adjust_path for children which *want* to be present with a
1120
# correct place to go.
1121
for _, child in self.other_tree.inventory.root.children.iteritems():
1122
trans_id = self.tt.trans_id_file_id(child.file_id)
1123
if not other_root_is_present:
1124
if self.tt.final_kind(trans_id) is not None:
1125
# The item exist in the final tree and has a defined place
1128
# Move the item into the root
1130
final_name = self.tt.final_name(trans_id)
1131
except errors.NoFinalPath:
1132
# This file is not present anymore, ignore it.
1134
self.tt.adjust_path(final_name, self.tt.root, trans_id)
1135
if other_root_is_present:
1136
self.tt.cancel_creation(other_root)
1137
self.tt.cancel_versioning(other_root)
1139
def write_modified(self, results):
1140
modified_hashes = {}
1141
for path in results.modified_paths:
1142
file_id = self.this_tree.path2id(self.this_tree.relpath(path))
1145
hash = self.this_tree.get_file_sha1(file_id)
1148
modified_hashes[file_id] = hash
1149
self.this_tree.set_merge_modified(modified_hashes)
1152
def parent(entry, file_id):
1153
"""Determine the parent for a file_id (used as a key method)"""
1156
return entry.parent_id
1159
def name(entry, file_id):
1160
"""Determine the name for a file_id (used as a key method)"""
1166
def contents_sha1(tree, file_id):
1167
"""Determine the sha1 of the file contents (used as a key method)."""
1168
if file_id not in tree:
1170
return tree.get_file_sha1(file_id)
1173
def executable(tree, file_id):
1174
"""Determine the executability of a file-id (used as a key method)."""
1175
if not tree.has_id(file_id):
1177
if tree.kind(file_id) != "file":
1179
return tree.is_executable(file_id)
1182
def kind(tree, file_id):
1183
"""Determine the kind of a file-id (used as a key method)."""
1184
if not tree.has_id(file_id):
1186
return tree.kind(file_id)
1189
def _three_way(base, other, this):
1191
# if 'base == other', either they all agree, or only 'this' has
1194
elif this not in (base, other):
1195
# 'this' is neither 'base' nor 'other', so both sides changed
1198
# "Ambiguous clean merge" -- both sides have made the same change.
1201
# this == base: only other has changed.
1205
def _lca_multi_way(bases, other, this, allow_overriding_lca=True):
1206
"""Consider LCAs when determining whether a change has occurred.
1208
If LCAS are all identical, this is the same as a _three_way comparison.
1210
:param bases: value in (BASE, [LCAS])
1211
:param other: value in OTHER
1212
:param this: value in THIS
1213
:param allow_overriding_lca: If there is more than one unique lca
1214
value, allow OTHER to override THIS if it has a new value, and
1215
THIS only has an lca value, or vice versa. This is appropriate for
1216
truly scalar values, not as much for non-scalars.
1217
:return: 'this', 'other', or 'conflict' depending on whether an entry
1220
# See doc/developers/lca_tree_merging.txt for details about this
1223
# Either Ambiguously clean, or nothing was actually changed. We
1226
base_val, lca_vals = bases
1227
# Remove 'base_val' from the lca_vals, because it is not interesting
1228
filtered_lca_vals = [lca_val for lca_val in lca_vals
1229
if lca_val != base_val]
1230
if len(filtered_lca_vals) == 0:
1231
return Merge3Merger._three_way(base_val, other, this)
1233
unique_lca_vals = set(filtered_lca_vals)
1234
if len(unique_lca_vals) == 1:
1235
return Merge3Merger._three_way(unique_lca_vals.pop(), other, this)
1237
if allow_overriding_lca:
1238
if other in unique_lca_vals:
1239
if this in unique_lca_vals:
1240
# Each side picked a different lca, conflict
1243
# This has a value which supersedes both lca values, and
1244
# other only has an lca value
1246
elif this in unique_lca_vals:
1247
# OTHER has a value which supersedes both lca values, and this
1248
# only has an lca value
1251
# At this point, the lcas disagree, and the tip disagree
1255
@deprecated_method(deprecated_in((2, 2, 0)))
1256
def scalar_three_way(this_tree, base_tree, other_tree, file_id, key):
1257
"""Do a three-way test on a scalar.
1258
Return "this", "other" or "conflict", depending whether a value wins.
1260
key_base = key(base_tree, file_id)
1261
key_other = key(other_tree, file_id)
1262
#if base == other, either they all agree, or only THIS has changed.
1263
if key_base == key_other:
1265
key_this = key(this_tree, file_id)
1266
# "Ambiguous clean merge"
1267
if key_this == key_other:
1269
elif key_this == key_base:
1274
def merge_names(self, file_id):
1275
def get_entry(tree):
1276
if tree.has_id(file_id):
1277
return tree.inventory[file_id]
1280
this_entry = get_entry(self.this_tree)
1281
other_entry = get_entry(self.other_tree)
1282
base_entry = get_entry(self.base_tree)
1283
entries = (base_entry, other_entry, this_entry)
1286
for entry in entries:
1289
parents.append(None)
1291
names.append(entry.name)
1292
parents.append(entry.parent_id)
1293
return self._merge_names(file_id, parents, names,
1294
resolver=self._three_way)
1296
def _merge_names(self, file_id, parents, names, resolver):
1297
"""Perform a merge on file_id names and parents"""
1298
base_name, other_name, this_name = names
1299
base_parent, other_parent, this_parent = parents
1301
name_winner = resolver(*names)
1303
parent_id_winner = resolver(*parents)
1304
if this_name is None:
1305
if name_winner == "this":
1306
name_winner = "other"
1307
if parent_id_winner == "this":
1308
parent_id_winner = "other"
1309
if name_winner == "this" and parent_id_winner == "this":
1311
if name_winner == 'conflict' or parent_id_winner == 'conflict':
1312
# Creating helpers (.OTHER or .THIS) here cause problems down the
1313
# road if a ContentConflict needs to be created so we should not do
1315
trans_id = self.tt.trans_id_file_id(file_id)
1316
self._raw_conflicts.append(('path conflict', trans_id, file_id,
1317
this_parent, this_name,
1318
other_parent, other_name))
1319
if other_name is None:
1320
# it doesn't matter whether the result was 'other' or
1321
# 'conflict'-- if there's no 'other', we leave it alone.
1323
parent_id = parents[self.winner_idx[parent_id_winner]]
1324
if parent_id is not None:
1325
# if we get here, name_winner and parent_winner are set to safe
1327
self.tt.adjust_path(names[self.winner_idx[name_winner]],
1328
self.tt.trans_id_file_id(parent_id),
1329
self.tt.trans_id_file_id(file_id))
1331
def _do_merge_contents(self, file_id):
1332
"""Performs a merge on file_id contents."""
1333
def contents_pair(tree):
1334
if file_id not in tree:
1336
kind = tree.kind(file_id)
1338
contents = tree.get_file_sha1(file_id)
1339
elif kind == "symlink":
1340
contents = tree.get_symlink_target(file_id)
1343
return kind, contents
1345
# See SPOT run. run, SPOT, run.
1346
# So we're not QUITE repeating ourselves; we do tricky things with
1348
base_pair = contents_pair(self.base_tree)
1349
other_pair = contents_pair(self.other_tree)
1351
this_pair = contents_pair(self.this_tree)
1352
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1353
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1354
this_pair, allow_overriding_lca=False)
1356
if base_pair == other_pair:
1359
# We delayed evaluating this_pair as long as we can to avoid
1360
# unnecessary sha1 calculation
1361
this_pair = contents_pair(self.this_tree)
1362
winner = self._three_way(base_pair, other_pair, this_pair)
1363
if winner == 'this':
1364
# No interesting changes introduced by OTHER
1366
# We have a hypothetical conflict, but if we have files, then we
1367
# can try to merge the content
1368
trans_id = self.tt.trans_id_file_id(file_id)
1369
params = MergeHookParams(self, file_id, trans_id, this_pair[0],
1370
other_pair[0], winner)
1371
hooks = self.active_hooks
1372
hook_status = 'not_applicable'
1374
hook_status, lines = hook.merge_contents(params)
1375
if hook_status != 'not_applicable':
1376
# Don't try any more hooks, this one applies.
1379
if hook_status == 'not_applicable':
1380
# This is a contents conflict, because none of the available
1381
# functions could merge it.
1383
name = self.tt.final_name(trans_id)
1384
parent_id = self.tt.final_parent(trans_id)
1385
if self.this_tree.has_id(file_id):
1386
self.tt.unversion_file(trans_id)
1387
file_group = self._dump_conflicts(name, parent_id, file_id,
1389
self._raw_conflicts.append(('contents conflict', file_group))
1390
elif hook_status == 'success':
1391
self.tt.create_file(lines, trans_id)
1392
elif hook_status == 'conflicted':
1393
# XXX: perhaps the hook should be able to provide
1394
# the BASE/THIS/OTHER files?
1395
self.tt.create_file(lines, trans_id)
1396
self._raw_conflicts.append(('text conflict', trans_id))
1397
name = self.tt.final_name(trans_id)
1398
parent_id = self.tt.final_parent(trans_id)
1399
self._dump_conflicts(name, parent_id, file_id)
1400
elif hook_status == 'delete':
1401
self.tt.unversion_file(trans_id)
1403
elif hook_status == 'done':
1404
# The hook function did whatever it needs to do directly, no
1405
# further action needed here.
1408
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1409
if not self.this_tree.has_id(file_id) and result == "modified":
1410
self.tt.version_file(file_id, trans_id)
1411
# The merge has been performed, so the old contents should not be
1413
self.tt.delete_contents(trans_id)
1416
def _default_other_winner_merge(self, merge_hook_params):
1417
"""Replace this contents with other."""
1418
file_id = merge_hook_params.file_id
1419
trans_id = merge_hook_params.trans_id
1420
file_in_this = self.this_tree.has_id(file_id)
1421
if self.other_tree.has_id(file_id):
1422
# OTHER changed the file
1424
if wt.supports_content_filtering():
1425
# We get the path from the working tree if it exists.
1426
# That fails though when OTHER is adding a file, so
1427
# we fall back to the other tree to find the path if
1428
# it doesn't exist locally.
1430
filter_tree_path = wt.id2path(file_id)
1431
except errors.NoSuchId:
1432
filter_tree_path = self.other_tree.id2path(file_id)
1434
# Skip the id2path lookup for older formats
1435
filter_tree_path = None
1436
transform.create_from_tree(self.tt, trans_id,
1437
self.other_tree, file_id,
1438
filter_tree_path=filter_tree_path)
1441
# OTHER deleted the file
1442
return 'delete', None
1444
raise AssertionError(
1445
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1448
def merge_contents(self, merge_hook_params):
1449
"""Fallback merge logic after user installed hooks."""
1450
# This function is used in merge hooks as the fallback instance.
1451
# Perhaps making this function and the functions it calls be a
1452
# a separate class would be better.
1453
if merge_hook_params.winner == 'other':
1454
# OTHER is a straight winner, so replace this contents with other
1455
return self._default_other_winner_merge(merge_hook_params)
1456
elif merge_hook_params.is_file_merge():
1457
# THIS and OTHER are both files, so text merge. Either
1458
# BASE is a file, or both converted to files, so at least we
1459
# have agreement that output should be a file.
1461
self.text_merge(merge_hook_params.file_id,
1462
merge_hook_params.trans_id)
1463
except errors.BinaryFile:
1464
return 'not_applicable', None
1467
return 'not_applicable', None
1469
def get_lines(self, tree, file_id):
1470
"""Return the lines in a file, or an empty list."""
1471
if tree.has_id(file_id):
1472
return tree.get_file_lines(file_id)
1476
def text_merge(self, file_id, trans_id):
1477
"""Perform a three-way text merge on a file_id"""
1478
# it's possible that we got here with base as a different type.
1479
# if so, we just want two-way text conflicts.
1480
if self.base_tree.has_id(file_id) and \
1481
self.base_tree.kind(file_id) == "file":
1482
base_lines = self.get_lines(self.base_tree, file_id)
1485
other_lines = self.get_lines(self.other_tree, file_id)
1486
this_lines = self.get_lines(self.this_tree, file_id)
1487
m3 = merge3.Merge3(base_lines, this_lines, other_lines,
1488
is_cherrypick=self.cherrypick)
1489
start_marker = "!START OF MERGE CONFLICT!" + "I HOPE THIS IS UNIQUE"
1490
if self.show_base is True:
1491
base_marker = '|' * 7
1495
def iter_merge3(retval):
1496
retval["text_conflicts"] = False
1497
for line in m3.merge_lines(name_a = "TREE",
1498
name_b = "MERGE-SOURCE",
1499
name_base = "BASE-REVISION",
1500
start_marker=start_marker,
1501
base_marker=base_marker,
1502
reprocess=self.reprocess):
1503
if line.startswith(start_marker):
1504
retval["text_conflicts"] = True
1505
yield line.replace(start_marker, '<' * 7)
1509
merge3_iterator = iter_merge3(retval)
1510
self.tt.create_file(merge3_iterator, trans_id)
1511
if retval["text_conflicts"] is True:
1512
self._raw_conflicts.append(('text conflict', trans_id))
1513
name = self.tt.final_name(trans_id)
1514
parent_id = self.tt.final_parent(trans_id)
1515
file_group = self._dump_conflicts(name, parent_id, file_id,
1516
this_lines, base_lines,
1518
file_group.append(trans_id)
1520
def _dump_conflicts(self, name, parent_id, file_id, this_lines=None,
1521
base_lines=None, other_lines=None, set_version=False,
1523
"""Emit conflict files.
1524
If this_lines, base_lines, or other_lines are omitted, they will be
1525
determined automatically. If set_version is true, the .OTHER, .THIS
1526
or .BASE (in that order) will be created as versioned files.
1528
data = [('OTHER', self.other_tree, other_lines),
1529
('THIS', self.this_tree, this_lines)]
1531
data.append(('BASE', self.base_tree, base_lines))
1533
# We need to use the actual path in the working tree of the file here,
1534
# ignoring the conflict suffixes
1536
if wt.supports_content_filtering():
1538
filter_tree_path = wt.id2path(file_id)
1539
except errors.NoSuchId:
1540
# file has been deleted
1541
filter_tree_path = None
1543
# Skip the id2path lookup for older formats
1544
filter_tree_path = None
1548
for suffix, tree, lines in data:
1549
if tree.has_id(file_id):
1550
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1551
suffix, lines, filter_tree_path)
1552
file_group.append(trans_id)
1553
if set_version and not versioned:
1554
self.tt.version_file(file_id, trans_id)
1558
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1559
lines=None, filter_tree_path=None):
1560
"""Emit a single conflict file."""
1561
name = name + '.' + suffix
1562
trans_id = self.tt.create_path(name, parent_id)
1563
transform.create_from_tree(self.tt, trans_id, tree, file_id, lines,
1567
def merge_executable(self, file_id, file_status):
1568
"""Perform a merge on the execute bit."""
1569
executable = [self.executable(t, file_id) for t in (self.base_tree,
1570
self.other_tree, self.this_tree)]
1571
self._merge_executable(file_id, executable, file_status,
1572
resolver=self._three_way)
1574
def _merge_executable(self, file_id, executable, file_status,
1576
"""Perform a merge on the execute bit."""
1577
base_executable, other_executable, this_executable = executable
1578
if file_status == "deleted":
1580
winner = resolver(*executable)
1581
if winner == "conflict":
1582
# There must be a None in here, if we have a conflict, but we
1583
# need executability since file status was not deleted.
1584
if self.executable(self.other_tree, file_id) is None:
1588
if winner == 'this' and file_status != "modified":
1590
trans_id = self.tt.trans_id_file_id(file_id)
1591
if self.tt.final_kind(trans_id) != "file":
1593
if winner == "this":
1594
executability = this_executable
1596
if self.other_tree.has_id(file_id):
1597
executability = other_executable
1598
elif self.this_tree.has_id(file_id):
1599
executability = this_executable
1600
elif self.base_tree_has_id(file_id):
1601
executability = base_executable
1602
if executability is not None:
1603
trans_id = self.tt.trans_id_file_id(file_id)
1604
self.tt.set_executability(executability, trans_id)
1606
def cook_conflicts(self, fs_conflicts):
1607
"""Convert all conflicts into a form that doesn't depend on trans_id"""
1608
self.cooked_conflicts.extend(transform.cook_conflicts(
1609
fs_conflicts, self.tt))
1610
fp = transform.FinalPaths(self.tt)
1611
for conflict in self._raw_conflicts:
1612
conflict_type = conflict[0]
1613
if conflict_type == 'path conflict':
1615
this_parent, this_name,
1616
other_parent, other_name) = conflict[1:]
1617
if this_parent is None or this_name is None:
1618
this_path = '<deleted>'
1620
parent_path = fp.get_path(
1621
self.tt.trans_id_file_id(this_parent))
1622
this_path = osutils.pathjoin(parent_path, this_name)
1623
if other_parent is None or other_name is None:
1624
other_path = '<deleted>'
1626
parent_path = fp.get_path(
1627
self.tt.trans_id_file_id(other_parent))
1628
other_path = osutils.pathjoin(parent_path, other_name)
1629
c = _mod_conflicts.Conflict.factory(
1630
'path conflict', path=this_path,
1631
conflict_path=other_path,
1633
elif conflict_type == 'contents conflict':
1634
for trans_id in conflict[1]:
1635
file_id = self.tt.final_file_id(trans_id)
1636
if file_id is not None:
1638
path = fp.get_path(trans_id)
1639
for suffix in ('.BASE', '.THIS', '.OTHER'):
1640
if path.endswith(suffix):
1641
path = path[:-len(suffix)]
1643
c = _mod_conflicts.Conflict.factory(conflict_type,
1644
path=path, file_id=file_id)
1645
elif conflict_type == 'text conflict':
1646
trans_id = conflict[1]
1647
path = fp.get_path(trans_id)
1648
file_id = self.tt.final_file_id(trans_id)
1649
c = _mod_conflicts.Conflict.factory(conflict_type,
1650
path=path, file_id=file_id)
1652
raise AssertionError('bad conflict type: %r' % (conflict,))
1653
self.cooked_conflicts.append(c)
1654
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
1657
class WeaveMerger(Merge3Merger):
1658
"""Three-way tree merger, text weave merger."""
1659
supports_reprocess = True
1660
supports_show_base = False
1661
supports_reverse_cherrypick = False
1662
history_based = True
1664
def _generate_merge_plan(self, file_id, base):
1665
return self.this_tree.plan_file_merge(file_id, self.other_tree,
1668
def _merged_lines(self, file_id):
1669
"""Generate the merged lines.
1670
There is no distinction between lines that are meant to contain <<<<<<<
1674
base = self.base_tree
1677
plan = self._generate_merge_plan(file_id, base)
1678
if 'merge' in debug.debug_flags:
1680
trans_id = self.tt.trans_id_file_id(file_id)
1681
name = self.tt.final_name(trans_id) + '.plan'
1682
contents = ('%11s|%s' % l for l in plan)
1683
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1684
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1685
'>>>>>>> MERGE-SOURCE\n')
1686
lines, conflicts = textmerge.merge_lines(self.reprocess)
1688
base_lines = textmerge.base_from_plan()
1691
return lines, base_lines
1693
def text_merge(self, file_id, trans_id):
1694
"""Perform a (weave) text merge for a given file and file-id.
1695
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1696
and a conflict will be noted.
1698
lines, base_lines = self._merged_lines(file_id)
1700
# Note we're checking whether the OUTPUT is binary in this case,
1701
# because we don't want to get into weave merge guts.
1702
textfile.check_text_lines(lines)
1703
self.tt.create_file(lines, trans_id)
1704
if base_lines is not None:
1706
self._raw_conflicts.append(('text conflict', trans_id))
1707
name = self.tt.final_name(trans_id)
1708
parent_id = self.tt.final_parent(trans_id)
1709
file_group = self._dump_conflicts(name, parent_id, file_id,
1711
base_lines=base_lines)
1712
file_group.append(trans_id)
1715
class LCAMerger(WeaveMerger):
1717
def _generate_merge_plan(self, file_id, base):
1718
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1721
class Diff3Merger(Merge3Merger):
1722
"""Three-way merger using external diff3 for text merging"""
1724
def dump_file(self, temp_dir, name, tree, file_id):
1725
out_path = osutils.pathjoin(temp_dir, name)
1726
out_file = open(out_path, "wb")
1728
in_file = tree.get_file(file_id)
1729
for line in in_file:
1730
out_file.write(line)
1735
def text_merge(self, file_id, trans_id):
1736
"""Perform a diff3 merge using a specified file-id and trans-id.
1737
If conflicts are encountered, .BASE, .THIS. and .OTHER conflict files
1738
will be dumped, and a will be conflict noted.
1741
temp_dir = osutils.mkdtemp(prefix="bzr-")
1743
new_file = osutils.pathjoin(temp_dir, "new")
1744
this = self.dump_file(temp_dir, "this", self.this_tree, file_id)
1745
base = self.dump_file(temp_dir, "base", self.base_tree, file_id)
1746
other = self.dump_file(temp_dir, "other", self.other_tree, file_id)
1747
status = bzrlib.patch.diff3(new_file, this, base, other)
1748
if status not in (0, 1):
1749
raise errors.BzrError("Unhandled diff3 exit code")
1750
f = open(new_file, 'rb')
1752
self.tt.create_file(f, trans_id)
1756
name = self.tt.final_name(trans_id)
1757
parent_id = self.tt.final_parent(trans_id)
1758
self._dump_conflicts(name, parent_id, file_id)
1759
self._raw_conflicts.append(('text conflict', trans_id))
1761
osutils.rmtree(temp_dir)
1764
class PathNotInTree(errors.BzrError):
1766
_fmt = """Merge-into failed because %(tree)s does not contain %(path)s."""
1768
def __init__(self, path, tree):
1769
errors.BzrError.__init__(self, path=path, tree=tree)
1772
class MergeIntoMerger(Merger):
1773
"""Merger that understands other_tree will be merged into a subdir.
1775
This also changes the Merger api so that it uses real Branch, revision_id,
1776
and RevisonTree objects, rather than using revision specs.
1779
def __init__(self, this_tree, other_branch, other_tree, target_subdir,
1780
source_subpath, other_rev_id=None):
1781
"""Create a new MergeIntoMerger object.
1783
source_subpath in other_tree will be effectively copied to
1784
target_subdir in this_tree.
1786
:param this_tree: The tree that we will be merging into.
1787
:param other_branch: The Branch we will be merging from.
1788
:param other_tree: The RevisionTree object we want to merge.
1789
:param target_subdir: The relative path where we want to merge
1790
other_tree into this_tree
1791
:param source_subpath: The relative path specifying the subtree of
1792
other_tree to merge into this_tree.
1794
# It is assumed that we are merging a tree that is not in our current
1795
# ancestry, which means we are using the "EmptyTree" as our basis.
1796
null_ancestor_tree = this_tree.branch.repository.revision_tree(
1797
_mod_revision.NULL_REVISION)
1798
super(MergeIntoMerger, self).__init__(
1799
this_branch=this_tree.branch,
1800
this_tree=this_tree,
1801
other_tree=other_tree,
1802
base_tree=null_ancestor_tree,
1804
self._target_subdir = target_subdir
1805
self._source_subpath = source_subpath
1806
self.other_branch = other_branch
1807
if other_rev_id is None:
1808
other_rev_id = other_tree.get_revision_id()
1809
self.other_rev_id = self.other_basis = other_rev_id
1810
self.base_is_ancestor = True
1811
self.backup_files = True
1812
self.merge_type = Merge3Merger
1813
self.show_base = False
1814
self.reprocess = False
1815
self.interesting_ids = None
1816
self.merge_type = _MergeTypeParameterizer(MergeIntoMergeType,
1817
target_subdir=self._target_subdir,
1818
source_subpath=self._source_subpath)
1819
if self._source_subpath != '':
1820
# If this isn't a partial merge make sure the revisions will be
1822
self._maybe_fetch(self.other_branch, self.this_branch,
1825
def set_pending(self):
1826
if self._source_subpath != '':
1828
Merger.set_pending(self)
1831
class _MergeTypeParameterizer(object):
1832
"""Wrap a merge-type class to provide extra parameters.
1834
This is hack used by MergeIntoMerger to pass some extra parameters to its
1835
merge_type. Merger.do_merge() sets up its own set of parameters to pass to
1836
the 'merge_type' member. It is difficult override do_merge without
1837
re-writing the whole thing, so instead we create a wrapper which will pass
1838
the extra parameters.
1841
def __init__(self, merge_type, **kwargs):
1842
self._extra_kwargs = kwargs
1843
self._merge_type = merge_type
1845
def __call__(self, *args, **kwargs):
1846
kwargs.update(self._extra_kwargs)
1847
return self._merge_type(*args, **kwargs)
1849
def __getattr__(self, name):
1850
return getattr(self._merge_type, name)
1853
class MergeIntoMergeType(Merge3Merger):
1854
"""Merger that incorporates a tree (or part of a tree) into another."""
1856
def __init__(self, *args, **kwargs):
1857
"""Initialize the merger object.
1859
:param args: See Merge3Merger.__init__'s args.
1860
:param kwargs: See Merge3Merger.__init__'s keyword args, except for
1861
source_subpath and target_subdir.
1862
:keyword source_subpath: The relative path specifying the subtree of
1863
other_tree to merge into this_tree.
1864
:keyword target_subdir: The relative path where we want to merge
1865
other_tree into this_tree
1867
# All of the interesting work happens during Merge3Merger.__init__(),
1868
# so we have have to hack in to get our extra parameters set.
1869
self._source_subpath = kwargs.pop('source_subpath')
1870
self._target_subdir = kwargs.pop('target_subdir')
1871
super(MergeIntoMergeType, self).__init__(*args, **kwargs)
1873
def _compute_transform(self):
1874
child_pb = ui.ui_factory.nested_progress_bar()
1876
entries = self._entries_to_incorporate()
1877
entries = list(entries)
1878
for num, (entry, parent_id) in enumerate(entries):
1879
child_pb.update('Preparing file merge', num, len(entries))
1880
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1881
trans_id = transform.new_by_entry(self.tt, entry,
1882
parent_trans_id, self.other_tree)
1885
self._finish_computing_transform()
1887
def _entries_to_incorporate(self):
1888
"""Yields pairs of (inventory_entry, new_parent)."""
1889
other_inv = self.other_tree.inventory
1890
subdir_id = other_inv.path2id(self._source_subpath)
1891
if subdir_id is None:
1892
# XXX: The error would be clearer if it gave the URL of the source
1893
# branch, but we don't have a reference to that here.
1894
raise PathNotInTree(self._source_subpath, "Source tree")
1895
subdir = other_inv[subdir_id]
1896
parent_in_target = osutils.dirname(self._target_subdir)
1897
target_id = self.this_tree.inventory.path2id(parent_in_target)
1898
if target_id is None:
1899
raise PathNotInTree(self._target_subdir, "Target tree")
1900
name_in_target = osutils.basename(self._target_subdir)
1901
merge_into_root = subdir.copy()
1902
merge_into_root.name = name_in_target
1903
if merge_into_root.file_id in self.this_tree.inventory:
1904
# Give the root a new file-id.
1905
# This can happen fairly easily if the directory we are
1906
# incorporating is the root, and both trees have 'TREE_ROOT' as
1907
# their root_id. Users will expect this to Just Work, so we
1908
# change the file-id here.
1909
# Non-root file-ids could potentially conflict too. That's really
1910
# an edge case, so we don't do anything special for those. We let
1911
# them cause conflicts.
1912
merge_into_root.file_id = generate_ids.gen_file_id(name_in_target)
1913
yield (merge_into_root, target_id)
1914
if subdir.kind != 'directory':
1915
# No children, so we are done.
1917
for ignored_path, entry in other_inv.iter_entries_by_dir(subdir_id):
1918
parent_id = entry.parent_id
1919
if parent_id == subdir.file_id:
1920
# The root's parent ID has changed, so make sure children of
1921
# the root refer to the new ID.
1922
parent_id = merge_into_root.file_id
1923
yield (entry, parent_id)
1926
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1928
merge_type=Merge3Merger,
1929
interesting_ids=None,
1933
interesting_files=None,
1936
change_reporter=None):
1937
"""Primary interface for merging.
1939
Typical use is probably::
1941
merge_inner(branch, branch.get_revision_tree(other_revision),
1942
branch.get_revision_tree(base_revision))
1944
if this_tree is None:
1945
raise errors.BzrError("bzrlib.merge.merge_inner requires a this_tree "
1946
"parameter as of bzrlib version 0.8.")
1947
merger = Merger(this_branch, other_tree, base_tree, this_tree=this_tree,
1948
pb=pb, change_reporter=change_reporter)
1949
merger.backup_files = backup_files
1950
merger.merge_type = merge_type
1951
merger.interesting_ids = interesting_ids
1952
merger.ignore_zero = ignore_zero
1953
if interesting_files:
1955
raise ValueError('Only supply interesting_ids'
1956
' or interesting_files')
1957
merger.interesting_files = interesting_files
1958
merger.show_base = show_base
1959
merger.reprocess = reprocess
1960
merger.other_rev_id = other_rev_id
1961
merger.other_basis = other_rev_id
1962
get_revision_id = getattr(base_tree, 'get_revision_id', None)
1963
if get_revision_id is None:
1964
get_revision_id = base_tree.last_revision
1965
merger.cache_trees_with_revision_ids([other_tree, base_tree, this_tree])
1966
merger.set_base_revision(get_revision_id(), this_branch)
1967
return merger.do_merge()
1969
def get_merge_type_registry():
1970
"""Merge type registry is in bzrlib.option to avoid circular imports.
1972
This method provides a sanctioned way to retrieve it.
1974
from bzrlib import option
1975
return option._merge_type_registry
1978
def _plan_annotate_merge(annotated_a, annotated_b, ancestors_a, ancestors_b):
1979
def status_a(revision, text):
1980
if revision in ancestors_b:
1981
return 'killed-b', text
1983
return 'new-a', text
1985
def status_b(revision, text):
1986
if revision in ancestors_a:
1987
return 'killed-a', text
1989
return 'new-b', text
1991
plain_a = [t for (a, t) in annotated_a]
1992
plain_b = [t for (a, t) in annotated_b]
1993
matcher = patiencediff.PatienceSequenceMatcher(None, plain_a, plain_b)
1994
blocks = matcher.get_matching_blocks()
1997
for ai, bi, l in blocks:
1998
# process all mismatched sections
1999
# (last mismatched section is handled because blocks always
2000
# includes a 0-length last block)
2001
for revision, text in annotated_a[a_cur:ai]:
2002
yield status_a(revision, text)
2003
for revision, text in annotated_b[b_cur:bi]:
2004
yield status_b(revision, text)
2005
# and now the matched section
2008
for text_a in plain_a[ai:a_cur]:
2009
yield "unchanged", text_a
2012
class _PlanMergeBase(object):
2014
def __init__(self, a_rev, b_rev, vf, key_prefix):
2017
:param a_rev: Revision-id of one revision to merge
2018
:param b_rev: Revision-id of the other revision to merge
2019
:param vf: A VersionedFiles containing both revisions
2020
:param key_prefix: A prefix for accessing keys in vf, typically
2026
self._last_lines = None
2027
self._last_lines_revision_id = None
2028
self._cached_matching_blocks = {}
2029
self._key_prefix = key_prefix
2030
self._precache_tip_lines()
2032
def _precache_tip_lines(self):
2033
lines = self.get_lines([self.a_rev, self.b_rev])
2034
self.lines_a = lines[self.a_rev]
2035
self.lines_b = lines[self.b_rev]
2037
def get_lines(self, revisions):
2038
"""Get lines for revisions from the backing VersionedFiles.
2040
:raises RevisionNotPresent: on absent texts.
2042
keys = [(self._key_prefix + (rev,)) for rev in revisions]
2044
for record in self.vf.get_record_stream(keys, 'unordered', True):
2045
if record.storage_kind == 'absent':
2046
raise errors.RevisionNotPresent(record.key, self.vf)
2047
result[record.key[-1]] = osutils.chunks_to_lines(
2048
record.get_bytes_as('chunked'))
2051
def plan_merge(self):
2052
"""Generate a 'plan' for merging the two revisions.
2054
This involves comparing their texts and determining the cause of
2055
differences. If text A has a line and text B does not, then either the
2056
line was added to text A, or it was deleted from B. Once the causes
2057
are combined, they are written out in the format described in
2058
VersionedFile.plan_merge
2060
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
2061
unique_a, unique_b = self._unique_lines(blocks)
2062
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
2063
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
2064
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
2066
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
2069
for i, j, n in blocks:
2070
for a_index in range(last_i, i):
2071
if a_index in new_a:
2072
if a_index in killed_b:
2073
yield 'conflicted-a', self.lines_a[a_index]
2075
yield 'new-a', self.lines_a[a_index]
2077
yield 'killed-b', self.lines_a[a_index]
2078
for b_index in range(last_j, j):
2079
if b_index in new_b:
2080
if b_index in killed_a:
2081
yield 'conflicted-b', self.lines_b[b_index]
2083
yield 'new-b', self.lines_b[b_index]
2085
yield 'killed-a', self.lines_b[b_index]
2086
# handle common lines
2087
for a_index in range(i, i+n):
2088
yield 'unchanged', self.lines_a[a_index]
2092
def _get_matching_blocks(self, left_revision, right_revision):
2093
"""Return a description of which sections of two revisions match.
2095
See SequenceMatcher.get_matching_blocks
2097
cached = self._cached_matching_blocks.get((left_revision,
2099
if cached is not None:
2101
if self._last_lines_revision_id == left_revision:
2102
left_lines = self._last_lines
2103
right_lines = self.get_lines([right_revision])[right_revision]
2105
lines = self.get_lines([left_revision, right_revision])
2106
left_lines = lines[left_revision]
2107
right_lines = lines[right_revision]
2108
self._last_lines = right_lines
2109
self._last_lines_revision_id = right_revision
2110
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
2112
return matcher.get_matching_blocks()
2114
def _unique_lines(self, matching_blocks):
2115
"""Analyse matching_blocks to determine which lines are unique
2117
:return: a tuple of (unique_left, unique_right), where the values are
2118
sets of line numbers of unique lines.
2124
for i, j, n in matching_blocks:
2125
unique_left.extend(range(last_i, i))
2126
unique_right.extend(range(last_j, j))
2129
return unique_left, unique_right
2132
def _subtract_plans(old_plan, new_plan):
2133
"""Remove changes from new_plan that came from old_plan.
2135
It is assumed that the difference between the old_plan and new_plan
2136
is their choice of 'b' text.
2138
All lines from new_plan that differ from old_plan are emitted
2139
verbatim. All lines from new_plan that match old_plan but are
2140
not about the 'b' revision are emitted verbatim.
2142
Lines that match and are about the 'b' revision are the lines we
2143
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
2144
is skipped entirely.
2146
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
2149
for i, j, n in matcher.get_matching_blocks():
2150
for jj in range(last_j, j):
2152
for jj in range(j, j+n):
2153
plan_line = new_plan[jj]
2154
if plan_line[0] == 'new-b':
2156
elif plan_line[0] == 'killed-b':
2157
yield 'unchanged', plan_line[1]
2163
class _PlanMerge(_PlanMergeBase):
2164
"""Plan an annotate merge using on-the-fly annotation"""
2166
def __init__(self, a_rev, b_rev, vf, key_prefix):
2167
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
2168
self.a_key = self._key_prefix + (self.a_rev,)
2169
self.b_key = self._key_prefix + (self.b_rev,)
2170
self.graph = _mod_graph.Graph(self.vf)
2171
heads = self.graph.heads((self.a_key, self.b_key))
2173
# one side dominates, so we can just return its values, yay for
2175
# Ideally we would know that before we get this far
2176
self._head_key = heads.pop()
2177
if self._head_key == self.a_key:
2181
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2182
self._head_key[-1], other)
2185
self._head_key = None
2188
def _precache_tip_lines(self):
2189
# Turn this into a no-op, because we will do this later
2192
def _find_recursive_lcas(self):
2193
"""Find all the ancestors back to a unique lca"""
2194
cur_ancestors = (self.a_key, self.b_key)
2195
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2196
# rather than a key tuple. We will just map that directly to no common
2200
next_lcas = self.graph.find_lca(*cur_ancestors)
2201
# Map a plain NULL_REVISION to a simple no-ancestors
2202
if next_lcas == set([_mod_revision.NULL_REVISION]):
2204
# Order the lca's based on when they were merged into the tip
2205
# While the actual merge portion of weave merge uses a set() of
2206
# active revisions, the order of insertion *does* effect the
2207
# implicit ordering of the texts.
2208
for rev_key in cur_ancestors:
2209
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2211
parent_map[rev_key] = ordered_parents
2212
if len(next_lcas) == 0:
2214
elif len(next_lcas) == 1:
2215
parent_map[list(next_lcas)[0]] = ()
2217
elif len(next_lcas) > 2:
2218
# More than 2 lca's, fall back to grabbing all nodes between
2219
# this and the unique lca.
2220
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2222
self.a_key, self.b_key, cur_ancestors)
2223
cur_lcas = next_lcas
2224
while len(cur_lcas) > 1:
2225
cur_lcas = self.graph.find_lca(*cur_lcas)
2226
if len(cur_lcas) == 0:
2227
# No common base to find, use the full ancestry
2230
unique_lca = list(cur_lcas)[0]
2231
if unique_lca == _mod_revision.NULL_REVISION:
2232
# find_lca will return a plain 'NULL_REVISION' rather
2233
# than a key tuple when there is no common ancestor, we
2234
# prefer to just use None, because it doesn't confuse
2235
# _get_interesting_texts()
2237
parent_map.update(self._find_unique_parents(next_lcas,
2240
cur_ancestors = next_lcas
2243
def _find_unique_parents(self, tip_keys, base_key):
2244
"""Find ancestors of tip that aren't ancestors of base.
2246
:param tip_keys: Nodes that are interesting
2247
:param base_key: Cull all ancestors of this node
2248
:return: The parent map for all revisions between tip_keys and
2249
base_key. base_key will be included. References to nodes outside of
2250
the ancestor set will also be removed.
2252
# TODO: this would be simpler if find_unique_ancestors took a list
2253
# instead of a single tip, internally it supports it, but it
2254
# isn't a "backwards compatible" api change.
2255
if base_key is None:
2256
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2257
# We remove NULL_REVISION because it isn't a proper tuple key, and
2258
# thus confuses things like _get_interesting_texts, and our logic
2259
# to add the texts into the memory weave.
2260
if _mod_revision.NULL_REVISION in parent_map:
2261
parent_map.pop(_mod_revision.NULL_REVISION)
2264
for tip in tip_keys:
2266
self.graph.find_unique_ancestors(tip, [base_key]))
2267
parent_map = self.graph.get_parent_map(interesting)
2268
parent_map[base_key] = ()
2269
culled_parent_map, child_map, tails = self._remove_external_references(
2271
# Remove all the tails but base_key
2272
if base_key is not None:
2273
tails.remove(base_key)
2274
self._prune_tails(culled_parent_map, child_map, tails)
2275
# Now remove all the uninteresting 'linear' regions
2276
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2280
def _remove_external_references(parent_map):
2281
"""Remove references that go outside of the parent map.
2283
:param parent_map: Something returned from Graph.get_parent_map(keys)
2284
:return: (filtered_parent_map, child_map, tails)
2285
filtered_parent_map is parent_map without external references
2286
child_map is the {parent_key: [child_keys]} mapping
2287
tails is a list of nodes that do not have any parents in the map
2289
# TODO: The basic effect of this function seems more generic than
2290
# _PlanMerge. But the specific details of building a child_map,
2291
# and computing tails seems very specific to _PlanMerge.
2292
# Still, should this be in Graph land?
2293
filtered_parent_map = {}
2296
for key, parent_keys in parent_map.iteritems():
2297
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2298
if not culled_parent_keys:
2300
for parent_key in culled_parent_keys:
2301
child_map.setdefault(parent_key, []).append(key)
2302
# TODO: Do we want to do this, it adds overhead for every node,
2303
# just to say that the node has no children
2304
child_map.setdefault(key, [])
2305
filtered_parent_map[key] = culled_parent_keys
2306
return filtered_parent_map, child_map, tails
2309
def _prune_tails(parent_map, child_map, tails_to_remove):
2310
"""Remove tails from the parent map.
2312
This will remove the supplied revisions until no more children have 0
2315
:param parent_map: A dict of {child: [parents]}, this dictionary will
2316
be modified in place.
2317
:param tails_to_remove: A list of tips that should be removed,
2318
this list will be consumed
2319
:param child_map: The reverse dict of parent_map ({parent: [children]})
2320
this dict will be modified
2321
:return: None, parent_map will be modified in place.
2323
while tails_to_remove:
2324
next = tails_to_remove.pop()
2325
parent_map.pop(next)
2326
children = child_map.pop(next)
2327
for child in children:
2328
child_parents = parent_map[child]
2329
child_parents.remove(next)
2330
if len(child_parents) == 0:
2331
tails_to_remove.append(child)
2333
def _get_interesting_texts(self, parent_map):
2334
"""Return a dict of texts we are interested in.
2336
Note that the input is in key tuples, but the output is in plain
2339
:param parent_map: The output from _find_recursive_lcas
2340
:return: A dict of {'revision_id':lines} as returned by
2341
_PlanMergeBase.get_lines()
2343
all_revision_keys = set(parent_map)
2344
all_revision_keys.add(self.a_key)
2345
all_revision_keys.add(self.b_key)
2347
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2348
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2351
def _build_weave(self):
2352
from bzrlib import weave
2353
self._weave = weave.Weave(weave_name='in_memory_weave',
2354
allow_reserved=True)
2355
parent_map = self._find_recursive_lcas()
2357
all_texts = self._get_interesting_texts(parent_map)
2359
# Note: Unfortunately, the order given by topo_sort will effect the
2360
# ordering resolution in the output. Specifically, if you add A then B,
2361
# then in the output text A lines will show up before B lines. And, of
2362
# course, topo_sort doesn't guarantee any real ordering.
2363
# So we use merge_sort, and add a fake node on the tip.
2364
# This ensures that left-hand parents will always be inserted into the
2365
# weave before right-hand parents.
2366
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2367
parent_map[tip_key] = (self.a_key, self.b_key)
2369
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2373
# for key in tsort.topo_sort(parent_map):
2374
parent_keys = parent_map[key]
2375
revision_id = key[-1]
2376
parent_ids = [k[-1] for k in parent_keys]
2377
self._weave.add_lines(revision_id, parent_ids,
2378
all_texts[revision_id])
2380
def plan_merge(self):
2381
"""Generate a 'plan' for merging the two revisions.
2383
This involves comparing their texts and determining the cause of
2384
differences. If text A has a line and text B does not, then either the
2385
line was added to text A, or it was deleted from B. Once the causes
2386
are combined, they are written out in the format described in
2387
VersionedFile.plan_merge
2389
if self._head_key is not None: # There was a single head
2390
if self._head_key == self.a_key:
2393
if self._head_key != self.b_key:
2394
raise AssertionError('There was an invalid head: %s != %s'
2395
% (self.b_key, self._head_key))
2397
head_rev = self._head_key[-1]
2398
lines = self.get_lines([head_rev])[head_rev]
2399
return ((plan, line) for line in lines)
2400
return self._weave.plan_merge(self.a_rev, self.b_rev)
2403
class _PlanLCAMerge(_PlanMergeBase):
2405
This merge algorithm differs from _PlanMerge in that:
2407
1. comparisons are done against LCAs only
2408
2. cases where a contested line is new versus one LCA but old versus
2409
another are marked as conflicts, by emitting the line as conflicted-a
2412
This is faster, and hopefully produces more useful output.
2415
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2416
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2417
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2420
if lca == _mod_revision.NULL_REVISION:
2423
self.lcas.add(lca[-1])
2424
for lca in self.lcas:
2425
if _mod_revision.is_null(lca):
2428
lca_lines = self.get_lines([lca])[lca]
2429
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2431
blocks = list(matcher.get_matching_blocks())
2432
self._cached_matching_blocks[(a_rev, lca)] = blocks
2433
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2435
blocks = list(matcher.get_matching_blocks())
2436
self._cached_matching_blocks[(b_rev, lca)] = blocks
2438
def _determine_status(self, revision_id, unique_line_numbers):
2439
"""Determines the status unique lines versus all lcas.
2441
Basically, determines why the line is unique to this revision.
2443
A line may be determined new, killed, or both.
2445
If a line is determined new, that means it was not present in at least
2446
one LCA, and is not present in the other merge revision.
2448
If a line is determined killed, that means the line was present in
2451
If a line is killed and new, this indicates that the two merge
2452
revisions contain differing conflict resolutions.
2454
:param revision_id: The id of the revision in which the lines are
2456
:param unique_line_numbers: The line numbers of unique lines.
2457
:return: a tuple of (new_this, killed_other)
2461
unique_line_numbers = set(unique_line_numbers)
2462
for lca in self.lcas:
2463
blocks = self._get_matching_blocks(revision_id, lca)
2464
unique_vs_lca, _ignored = self._unique_lines(blocks)
2465
new.update(unique_line_numbers.intersection(unique_vs_lca))
2466
killed.update(unique_line_numbers.difference(unique_vs_lca))
1
from merge_core import merge_flex
2
from changeset import generate_changeset, ExceptionConflictHandler
3
from changeset import Inventory
4
from bzrlib import Branch
6
from trace import mutter
12
class MergeConflictHandler(ExceptionConflictHandler):
13
"""Handle conflicts encountered while merging"""
14
def copy(self, source, dest):
15
"""Copy the text and mode of a file
16
:param source: The path of the file to copy
17
:param dest: The distination file to create
19
s_file = file(source, "rb")
20
d_file = file(dest, "wb")
23
os.chmod(dest, 0777 & os.stat(source).st_mode)
25
def add_suffix(self, name, suffix, last_new_name=None):
26
"""Rename a file to append a suffix. If the new name exists, the
27
suffix is added repeatedly until a non-existant name is found
29
:param name: The path of the file
30
:param suffix: The suffix to append
31
:param last_new_name: (used for recursive calls) the last name tried
33
if last_new_name is None:
35
new_name = last_new_name+suffix
37
os.rename(name, new_name)
39
if e.errno != errno.EEXIST and e.errno != errno.ENOTEMPTY:
41
self.add_suffix(name, suffix, last_new_name=new_name)
43
def merge_conflict(self, new_file, this_path, base_path, other_path):
45
Handle diff3 conflicts by producing a .THIS, .BASE and .OTHER. The
46
main file will be a version with diff3 conflicts.
47
:param new_file: Path to the output file with diff3 markers
48
:param this_path: Path to the file text for the THIS tree
49
:param base_path: Path to the file text for the BASE tree
50
:param other_path: Path to the file text for the OTHER tree
52
self.add_suffix(this_path, ".THIS")
53
self.copy(base_path, this_path+".BASE")
54
self.copy(other_path, this_path+".OTHER")
55
os.rename(new_file, this_path)
57
def target_exists(self, entry, target, old_path):
58
"""Handle the case when the target file or dir exists"""
59
self.add_suffix(target, ".moved")
62
def __init__(self, path, id, present=None, isdir=None):
65
self.present = present
67
self.interesting = True
70
return "SourceFile(%s, %s)" % (self.path, self.id)
72
def get_tree(treespec, temp_root, label):
76
base_tree = branch.working_tree()
78
base_tree = branch.basis_tree()
80
base_tree = branch.revision_tree(branch.lookup_revision(revno))
81
temp_path = os.path.join(temp_root, label)
83
return MergeTree(base_tree, temp_path)
86
def abspath(tree, file_id):
87
path = tree.inventory.id2path(file_id)
92
def file_exists(tree, file_id):
93
return tree.has_filename(tree.id2path(file_id))
95
def inventory_map(tree):
97
for file_id in tree.inventory:
98
if not file_exists(tree, file_id):
100
path = abspath(tree, file_id)
101
inventory[path] = SourceFile(path, file_id)
105
class MergeTree(object):
106
def __init__(self, tree, tempdir):
107
object.__init__(self)
108
if hasattr(tree, "basedir"):
109
self.root = tree.basedir
112
self.inventory = inventory_map(tree)
114
self.tempdir = tempdir
115
os.mkdir(os.path.join(self.tempdir, "texts"))
118
def readonly_path(self, id):
119
if self.root is not None:
120
return self.tree.abspath(self.tree.id2path(id))
122
if self.tree.inventory[id].kind in ("directory", "root_directory"):
124
if not self.cached.has_key(id):
125
path = os.path.join(self.tempdir, "texts", id)
126
outfile = file(path, "wb")
127
outfile.write(self.tree.get_file(id).read())
128
assert(os.path.exists(path))
129
self.cached[id] = path
130
return self.cached[id]
132
def merge(other_revision, base_revision):
133
tempdir = tempfile.mkdtemp(prefix="bzr-")
135
this_branch = Branch('.')
136
other_tree = get_tree(other_revision, tempdir, "other")
137
base_tree = get_tree(base_revision, tempdir, "base")
138
merge_inner(this_branch, other_tree, base_tree, tempdir)
140
shutil.rmtree(tempdir)
143
def generate_cset_optimized(tree_a, tree_b, inventory_a, inventory_b):
144
"""Generate a changeset, using the text_id to mark really-changed files.
145
This permits blazing comparisons when text_ids are present. It also
146
disables metadata comparison for files with identical texts.
148
for file_id in tree_a.tree.inventory:
149
if file_id not in tree_b.tree.inventory:
151
entry_a = tree_a.tree.inventory[file_id]
152
entry_b = tree_b.tree.inventory[file_id]
153
if (entry_a.kind, entry_b.kind) != ("file", "file"):
155
if None in (entry_a.text_id, entry_b.text_id):
157
if entry_a.text_id != entry_b.text_id:
159
inventory_a[abspath(tree_a.tree, file_id)].interesting = False
160
inventory_b[abspath(tree_b.tree, file_id)].interesting = False
161
cset = generate_changeset(tree_a, tree_b, inventory_a, inventory_b)
162
for entry in cset.entries.itervalues():
163
entry.metadata_change = None
167
def merge_inner(this_branch, other_tree, base_tree, tempdir):
168
this_tree = get_tree(('.', None), tempdir, "this")
170
def get_inventory(tree):
171
return tree.inventory
173
inv_changes = merge_flex(this_tree, base_tree, other_tree,
174
generate_cset_optimized, get_inventory,
175
MergeConflictHandler(base_tree.root))
178
for id, path in inv_changes.iteritems():
183
assert path.startswith('./')
185
adjust_ids.append((path, id))
186
this_branch.set_inventory(regen_inventory(this_branch, this_tree.root, adjust_ids))
189
def regen_inventory(this_branch, root, new_entries):
190
old_entries = this_branch.read_working_inventory()
193
for file_id in old_entries:
194
entry = old_entries[file_id]
195
path = old_entries.id2path(file_id)
196
new_inventory[file_id] = (path, file_id, entry.parent_id, entry.kind)
197
by_path[path] = file_id
202
for path, file_id in new_entries:
204
del new_inventory[file_id]
207
new_path_list.append((path, file_id))
208
if file_id not in old_entries:
210
# Ensure no file is added before its parent
212
for path, file_id in new_path_list:
216
parent = by_path[os.path.dirname(path)]
217
kind = bzrlib.osutils.file_kind(os.path.join(root, path))
218
new_inventory[file_id] = (path, file_id, parent, kind)
219
by_path[path] = file_id
221
# Get a list in insertion order
222
new_inventory_list = new_inventory.values()
223
mutter ("""Inventory regeneration:
224
old length: %i insertions: %i deletions: %i new_length: %i"""\
225
% (len(old_entries), insertions, deletions, len(new_inventory_list)))
226
assert len(new_inventory_list) == len(old_entries) + insertions - deletions
227
new_inventory_list.sort()
228
return new_inventory_list
added
removed
bzrlib/merge.py
