13
13
# You should have received a copy of the GNU General Public License
14
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
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
21
22
from bzrlib import (
22
branch as _mod_branch,
23
conflicts as _mod_conflicts,
30
27
revision as _mod_revision,
39
from bzrlib.cleanup import OperationWithCleanups
45
from bzrlib.symbol_versioning import (
29
from bzrlib.branch import Branch
30
from bzrlib.conflicts import ConflictList, Conflict
31
from bzrlib.errors import (BzrCommandError,
41
WorkingTreeNotRevision,
44
from bzrlib.merge3 import Merge3
45
from bzrlib.osutils import rename, pathjoin
46
from progress import DummyProgress, ProgressPhase
47
from bzrlib.revision import (is_ancestor, NULL_REVISION, ensure_null)
48
from bzrlib.textfile import check_text_lines
49
from bzrlib.trace import mutter, warning, note
50
from bzrlib.transform import (TreeTransform, resolve_conflicts, cook_conflicts,
51
conflict_pass, FinalPaths, create_by_entry,
52
unique_add, ROOT_PARENT)
53
from bzrlib.versionedfile import PlanWeaveMerge
49
56
# TODO: Report back as changes are merged in
52
59
def transform_tree(from_tree, to_tree, interesting_ids=None):
53
from_tree.lock_tree_write()
54
operation = OperationWithCleanups(merge_inner)
55
operation.add_cleanup(from_tree.unlock)
56
operation.run_simple(from_tree.branch, to_tree, from_tree,
57
ignore_zero=True, interesting_ids=interesting_ids, this_tree=from_tree)
60
class MergeHooks(hooks.Hooks):
63
hooks.Hooks.__init__(self)
64
self.create_hook(hooks.HookPoint('merge_file_content',
65
"Called with a bzrlib.merge.Merger object to create a per file "
66
"merge object when starting a merge. "
67
"Should return either None or a subclass of "
68
"``bzrlib.merge.AbstractPerFileMerger``. "
69
"Such objects will then be called per file "
70
"that needs to be merged (including when one "
71
"side has deleted the file and the other has changed it). "
72
"See the AbstractPerFileMerger API docs for details on how it is "
77
class AbstractPerFileMerger(object):
78
"""PerFileMerger objects are used by plugins extending merge for bzrlib.
80
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
82
:ivar merger: The Merge3Merger performing the merge.
85
def __init__(self, merger):
86
"""Create a PerFileMerger for use with merger."""
89
def merge_contents(self, merge_params):
90
"""Attempt to merge the contents of a single file.
92
:param merge_params: A bzrlib.merge.MergeHookParams
93
:return : A tuple of (status, chunks), where status is one of
94
'not_applicable', 'success', 'conflicted', or 'delete'. If status
95
is 'success' or 'conflicted', then chunks should be an iterable of
96
strings for the new file contents.
98
return ('not applicable', None)
101
class PerFileMerger(AbstractPerFileMerger):
102
"""Merge individual files when self.file_matches returns True.
104
This class is intended to be subclassed. The file_matches and
105
merge_matching methods should be overridden with concrete implementations.
108
def file_matches(self, params):
109
"""Return True if merge_matching should be called on this file.
111
Only called with merges of plain files with no clear winner.
113
Subclasses must override this.
115
raise NotImplementedError(self.file_matches)
117
def get_filename(self, params, tree):
118
"""Lookup the filename (i.e. basename, not path), given a Tree (e.g.
119
self.merger.this_tree) and a MergeHookParams.
121
return osutils.basename(tree.id2path(params.file_id))
123
def get_filepath(self, params, tree):
124
"""Calculate the path to the file in a tree.
126
:param params: A MergeHookParams describing the file to merge
127
:param tree: a Tree, e.g. self.merger.this_tree.
129
return tree.id2path(params.file_id)
131
def merge_contents(self, params):
132
"""Merge the contents of a single file."""
133
# Check whether this custom merge logic should be used.
135
# OTHER is a straight winner, rely on default merge.
136
params.winner == 'other' or
137
# THIS and OTHER aren't both files.
138
not params.is_file_merge() or
139
# The filename doesn't match *.xml
140
not self.file_matches(params)):
141
return 'not_applicable', None
142
return self.merge_matching(params)
144
def merge_matching(self, params):
145
"""Merge the contents of a single file that has matched the criteria
146
in PerFileMerger.merge_contents (is a conflict, is a file,
147
self.file_matches is True).
149
Subclasses must override this.
151
raise NotImplementedError(self.merge_matching)
154
class ConfigurableFileMerger(PerFileMerger):
155
"""Merge individual files when configured via a .conf file.
157
This is a base class for concrete custom file merging logic. Concrete
158
classes should implement ``merge_text``.
160
See ``bzrlib.plugins.news_merge.news_merge`` for an example concrete class.
162
:ivar affected_files: The configured file paths to merge.
164
:cvar name_prefix: The prefix to use when looking up configuration
165
details. <name_prefix>_merge_files describes the files targeted by the
168
:cvar default_files: The default file paths to merge when no configuration
175
def __init__(self, merger):
176
super(ConfigurableFileMerger, self).__init__(merger)
177
self.affected_files = None
178
self.default_files = self.__class__.default_files or []
179
self.name_prefix = self.__class__.name_prefix
180
if self.name_prefix is None:
181
raise ValueError("name_prefix must be set.")
183
def file_matches(self, params):
184
"""Check whether the file should call the merge hook.
186
<name_prefix>_merge_files configuration variable is a list of files
187
that should use the hook.
189
affected_files = self.affected_files
190
if affected_files is None:
191
config = self.merger.this_branch.get_config()
192
# Until bzr provides a better policy for caching the config, we
193
# just add the part we're interested in to the params to avoid
194
# reading the config files repeatedly (bazaar.conf, location.conf,
196
config_key = self.name_prefix + '_merge_files'
197
affected_files = config.get_user_option_as_list(config_key)
198
if affected_files is None:
199
# If nothing was specified in the config, use the default.
200
affected_files = self.default_files
201
self.affected_files = affected_files
203
filepath = self.get_filepath(params, self.merger.this_tree)
204
if filepath in affected_files:
208
def merge_matching(self, params):
209
return self.merge_text(params)
211
def merge_text(self, params):
212
"""Merge the byte contents of a single file.
214
This is called after checking that the merge should be performed in
215
merge_contents, and it should behave as per
216
``bzrlib.merge.AbstractPerFileMerger.merge_contents``.
218
raise NotImplementedError(self.merge_text)
221
class MergeHookParams(object):
222
"""Object holding parameters passed to merge_file_content hooks.
224
There are some fields hooks can access:
226
:ivar file_id: the file ID of the file being merged
227
:ivar trans_id: the transform ID for the merge of this file
228
:ivar this_kind: kind of file_id in 'this' tree
229
:ivar other_kind: kind of file_id in 'other' tree
230
:ivar winner: one of 'this', 'other', 'conflict'
233
def __init__(self, merger, file_id, trans_id, this_kind, other_kind,
235
self._merger = merger
236
self.file_id = file_id
237
self.trans_id = trans_id
238
self.this_kind = this_kind
239
self.other_kind = other_kind
242
def is_file_merge(self):
243
"""True if this_kind and other_kind are both 'file'."""
244
return self.this_kind == 'file' and self.other_kind == 'file'
246
@decorators.cachedproperty
247
def base_lines(self):
248
"""The lines of the 'base' version of the file."""
249
return self._merger.get_lines(self._merger.base_tree, self.file_id)
251
@decorators.cachedproperty
252
def this_lines(self):
253
"""The lines of the 'this' version of the file."""
254
return self._merger.get_lines(self._merger.this_tree, self.file_id)
256
@decorators.cachedproperty
257
def other_lines(self):
258
"""The lines of the 'other' version of the file."""
259
return self._merger.get_lines(self._merger.other_tree, self.file_id)
60
merge_inner(from_tree.branch, to_tree, from_tree, ignore_zero=True,
61
interesting_ids=interesting_ids, this_tree=from_tree)
262
64
class Merger(object):
266
65
def __init__(self, this_branch, other_tree=None, base_tree=None,
267
this_tree=None, pb=None, change_reporter=None,
268
recurse='down', revision_graph=None):
66
this_tree=None, pb=DummyProgress(), change_reporter=None,
269
68
object.__init__(self)
69
assert this_tree is not None, "this_tree is required"
270
70
self.this_branch = this_branch
271
71
self.this_basis = _mod_revision.ensure_null(
272
72
this_branch.last_revision())
450
183
if self.other_rev_id is None:
451
184
other_basis_tree = self.revision_tree(self.other_basis)
452
if other_basis_tree.has_changes(self.other_tree):
453
raise errors.WorkingTreeNotRevision(self.this_tree)
185
changes = other_basis_tree.changes_from(self.other_tree)
186
if changes.has_changed():
187
raise WorkingTreeNotRevision(self.this_tree)
454
188
other_rev_id = self.other_basis
455
189
self.other_tree = other_basis_tree
457
@deprecated_method(deprecated_in((2, 1, 0)))
458
191
def file_revisions(self, file_id):
459
192
self.ensure_revision_trees()
460
193
def get_id(tree, file_id):
461
194
revision_id = tree.inventory[file_id].revision
195
assert revision_id is not None
462
196
return revision_id
463
197
if self.this_rev_id is None:
464
198
if self.this_basis_tree.get_file_sha1(file_id) != \
465
199
self.this_tree.get_file_sha1(file_id):
466
raise errors.WorkingTreeNotRevision(self.this_tree)
200
raise WorkingTreeNotRevision(self.this_tree)
468
202
trees = (self.this_basis_tree, self.other_tree)
469
203
return [get_id(tree, file_id) for tree in trees]
471
@deprecated_method(deprecated_in((2, 1, 0)))
472
205
def check_basis(self, check_clean, require_commits=True):
473
206
if self.this_basis is None and require_commits is True:
474
raise errors.BzrCommandError(
475
"This branch has no commits."
476
" (perhaps you would prefer 'bzr pull')")
207
raise BzrCommandError("This branch has no commits."
208
" (perhaps you would prefer 'bzr pull')")
478
210
self.compare_basis()
479
211
if self.this_basis != self.this_rev_id:
480
raise errors.UncommittedChanges(self.this_tree)
212
raise BzrCommandError("Working tree has uncommitted changes.")
482
@deprecated_method(deprecated_in((2, 1, 0)))
483
214
def compare_basis(self):
485
216
basis_tree = self.revision_tree(self.this_tree.last_revision())
486
except errors.NoSuchRevision:
217
except errors.RevisionNotPresent:
487
218
basis_tree = self.this_tree.basis_tree()
488
if not self.this_tree.has_changes(basis_tree):
219
changes = self.this_tree.changes_from(basis_tree)
220
if not changes.has_changed():
489
221
self.this_rev_id = self.this_basis
491
223
def set_interesting_files(self, file_list):
492
224
self.interesting_files = file_list
494
226
def set_pending(self):
495
if (not self.base_is_ancestor or not self.base_is_other_ancestor
496
or self.other_rev_id is None):
227
if not self.base_is_ancestor or not self.base_is_other_ancestor or self.other_rev_id is None:
498
229
self._add_parent()
500
231
def _add_parent(self):
501
232
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
502
233
new_parent_trees = []
503
operation = OperationWithCleanups(self.this_tree.set_parent_trees)
504
234
for revision_id in new_parents:
506
236
tree = self.revision_tree(revision_id)
507
except errors.NoSuchRevision:
237
except errors.RevisionNotPresent:
511
operation.add_cleanup(tree.unlock)
512
241
new_parent_trees.append((revision_id, tree))
513
operation.run_simple(new_parent_trees, allow_leftmost_as_ghost=True)
243
self.this_tree.set_parent_trees(new_parent_trees,
244
allow_leftmost_as_ghost=True)
246
for _revision_id, tree in new_parent_trees:
515
250
def set_other(self, other_revision, possible_transports=None):
516
251
"""Set the revision and tree to merge from.
633
340
self.base_rev_id = _mod_revision.ensure_null(
634
341
base_branch.get_rev_id(base_revision[1]))
635
342
self._maybe_fetch(base_branch, self.this_branch, self.base_rev_id)
343
self.base_is_ancestor = is_ancestor(self.this_basis,
346
self.base_is_other_ancestor = is_ancestor(self.other_basis,
637
def make_merger(self):
638
351
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
639
352
'other_tree': self.other_tree,
640
353
'interesting_ids': self.interesting_ids,
641
354
'interesting_files': self.interesting_files,
642
'this_branch': self.this_branch,
644
356
if self.merge_type.requires_base:
645
357
kwargs['base_tree'] = self.base_tree
646
358
if self.merge_type.supports_reprocess:
647
359
kwargs['reprocess'] = self.reprocess
648
360
elif self.reprocess:
649
raise errors.BzrError(
650
"Conflict reduction is not supported for merge"
651
" type %s." % self.merge_type)
361
raise BzrError("Conflict reduction is not supported for merge"
362
" type %s." % self.merge_type)
652
363
if self.merge_type.supports_show_base:
653
364
kwargs['show_base'] = self.show_base
654
365
elif self.show_base:
655
raise errors.BzrError("Showing base is not supported for this"
366
raise BzrError("Showing base is not supported for this"
656
367
" merge type. %s" % self.merge_type)
657
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
658
and not self.base_is_other_ancestor):
659
raise errors.CannotReverseCherrypick()
660
if self.merge_type.supports_cherrypick:
661
kwargs['cherrypick'] = (not self.base_is_ancestor or
662
not self.base_is_other_ancestor)
663
if self._is_criss_cross and getattr(self.merge_type,
664
'supports_lca_trees', False):
665
kwargs['lca_trees'] = self._lca_trees
666
return self.merge_type(pb=None,
667
change_reporter=self.change_reporter,
670
def _do_merge_to(self):
671
merge = self.make_merger()
672
if self.other_branch is not None:
673
self.other_branch.update_references(self.this_branch)
675
if self.recurse == 'down':
676
for relpath, file_id in self.this_tree.iter_references():
677
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
678
other_revision = self.other_tree.get_reference_revision(
680
if other_revision == sub_tree.last_revision():
682
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
683
sub_merge.merge_type = self.merge_type
684
other_branch = self.other_branch.reference_parent(file_id, relpath)
685
sub_merge.set_other_revision(other_revision, other_branch)
686
base_revision = self.base_tree.get_reference_revision(file_id)
687
sub_merge.base_tree = \
688
sub_tree.branch.repository.revision_tree(base_revision)
689
sub_merge.base_rev_id = base_revision
694
operation = OperationWithCleanups(self._do_merge_to)
695
368
self.this_tree.lock_tree_write()
696
operation.add_cleanup(self.this_tree.unlock)
697
369
if self.base_tree is not None:
698
370
self.base_tree.lock_read()
699
operation.add_cleanup(self.base_tree.unlock)
700
371
if self.other_tree is not None:
701
372
self.other_tree.lock_read()
702
operation.add_cleanup(self.other_tree.unlock)
703
merge = operation.run_simple()
374
merge = self.merge_type(pb=self._pb,
375
change_reporter=self.change_reporter,
377
if self.recurse == 'down':
378
for path, file_id in self.this_tree.iter_references():
379
sub_tree = self.this_tree.get_nested_tree(file_id, path)
380
other_revision = self.other_tree.get_reference_revision(
382
if other_revision == sub_tree.last_revision():
384
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
385
sub_merge.merge_type = self.merge_type
386
relpath = self.this_tree.relpath(path)
387
other_branch = self.other_branch.reference_parent(file_id, relpath)
388
sub_merge.set_other_revision(other_revision, other_branch)
389
base_revision = self.base_tree.get_reference_revision(file_id)
390
sub_merge.base_tree = \
391
sub_tree.branch.repository.revision_tree(base_revision)
395
if self.other_tree is not None:
396
self.other_tree.unlock()
397
if self.base_tree is not None:
398
self.base_tree.unlock()
399
self.this_tree.unlock()
704
400
if len(merge.cooked_conflicts) == 0:
705
if not self.ignore_zero and not trace.is_quiet():
706
trace.note("All changes applied successfully.")
401
if not self.ignore_zero:
402
note("All changes applied successfully.")
708
trace.note("%d conflicts encountered."
709
% len(merge.cooked_conflicts))
404
note("%d conflicts encountered." % len(merge.cooked_conflicts))
711
406
return len(merge.cooked_conflicts)
714
class _InventoryNoneEntry(object):
715
"""This represents an inventory entry which *isn't there*.
717
It simplifies the merging logic if we always have an InventoryEntry, even
718
if it isn't actually present
725
symlink_target = None
728
_none_entry = _InventoryNoneEntry()
731
409
class Merge3Merger(object):
732
410
"""Three-way merger that uses the merge3 text merger"""
733
411
requires_base = True
734
412
supports_reprocess = True
735
413
supports_show_base = True
736
414
history_based = False
737
supports_cherrypick = True
738
supports_reverse_cherrypick = True
739
415
winner_idx = {"this": 2, "other": 1, "conflict": 1}
740
supports_lca_trees = True
742
def __init__(self, working_tree, this_tree, base_tree, other_tree,
417
def __init__(self, working_tree, this_tree, base_tree, other_tree,
743
418
interesting_ids=None, reprocess=False, show_base=False,
744
pb=None, pp=None, change_reporter=None,
745
interesting_files=None, do_merge=True,
746
cherrypick=False, lca_trees=None, this_branch=None):
419
pb=DummyProgress(), pp=None, change_reporter=None,
420
interesting_files=None):
747
421
"""Initialize the merger object and perform the merge.
749
423
:param working_tree: The working tree to apply the merge to
750
424
:param this_tree: The local tree in the merge operation
751
425
:param base_tree: The common tree in the merge operation
752
:param other_tree: The other tree to merge changes from
753
:param this_branch: The branch associated with this_tree. Defaults to
754
this_tree.branch if not supplied.
426
:param other_tree: The other other tree to merge changes from
755
427
:param interesting_ids: The file_ids of files that should be
756
428
participate in the merge. May not be combined with
757
429
interesting_files.
758
430
:param: reprocess If True, perform conflict-reduction processing.
759
431
:param show_base: If True, show the base revision in text conflicts.
760
432
(incompatible with reprocess)
433
:param pb: A Progress bar
762
434
:param pp: A ProgressPhase object
763
435
:param change_reporter: An object that should report changes made
764
436
:param interesting_files: The tree-relative paths of files that should
767
439
be combined with interesting_ids. If neither interesting_files nor
768
440
interesting_ids is specified, all files may participate in the
770
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
771
if the ancestry was found to include a criss-cross merge.
772
Otherwise should be None.
774
443
object.__init__(self)
775
if interesting_files is not None and interesting_ids is not None:
777
'specify either interesting_ids or interesting_files')
778
if this_branch is None:
779
this_branch = this_tree.branch
444
if interesting_files is not None:
445
assert interesting_ids is None
780
446
self.interesting_ids = interesting_ids
781
447
self.interesting_files = interesting_files
782
448
self.this_tree = working_tree
449
self.this_tree.lock_tree_write()
783
450
self.base_tree = base_tree
451
self.base_tree.lock_read()
784
452
self.other_tree = other_tree
785
self.this_branch = this_branch
453
self.other_tree.lock_read()
786
454
self._raw_conflicts = []
787
455
self.cooked_conflicts = []
788
456
self.reprocess = reprocess
789
457
self.show_base = show_base
790
self._lca_trees = lca_trees
791
# Uncommenting this will change the default algorithm to always use
792
# _entries_lca. This can be useful for running the test suite and
793
# making sure we haven't missed any corner cases.
794
# if lca_trees is None:
795
# self._lca_trees = [self.base_tree]
796
460
self.change_reporter = change_reporter
797
self.cherrypick = cherrypick
801
warnings.warn("pp argument to Merge3Merger is deprecated")
803
warnings.warn("pb argument to Merge3Merger is deprecated")
806
operation = OperationWithCleanups(self._do_merge)
807
self.this_tree.lock_tree_write()
808
operation.add_cleanup(self.this_tree.unlock)
809
self.base_tree.lock_read()
810
operation.add_cleanup(self.base_tree.unlock)
811
self.other_tree.lock_read()
812
operation.add_cleanup(self.other_tree.unlock)
815
def _do_merge(self, operation):
816
self.tt = transform.TreeTransform(self.this_tree, None)
817
operation.add_cleanup(self.tt.finalize)
818
self._compute_transform()
819
results = self.tt.apply(no_conflicts=True)
820
self.write_modified(results)
462
self.pp = ProgressPhase("Merge phase", 3, self.pb)
464
self.tt = TreeTransform(working_tree, self.pb)
822
self.this_tree.add_conflicts(self.cooked_conflicts)
823
except errors.UnsupportedOperation:
826
def make_preview_transform(self):
827
operation = OperationWithCleanups(self._make_preview_transform)
828
self.base_tree.lock_read()
829
operation.add_cleanup(self.base_tree.unlock)
830
self.other_tree.lock_read()
831
operation.add_cleanup(self.other_tree.unlock)
832
return operation.run_simple()
834
def _make_preview_transform(self):
835
self.tt = transform.TransformPreview(self.this_tree)
836
self._compute_transform()
839
def _compute_transform(self):
840
if self._lca_trees is None:
841
467
entries = self._entries3()
842
resolver = self._three_way
844
entries = self._entries_lca()
845
resolver = self._lca_multi_way
846
child_pb = ui.ui_factory.nested_progress_bar()
848
factories = Merger.hooks['merge_file_content']
849
hooks = [factory(self) for factory in factories] + [self]
850
self.active_hooks = [hook for hook in hooks if hook is not None]
851
for num, (file_id, changed, parents3, names3,
852
executable3) in enumerate(entries):
853
child_pb.update('Preparing file merge', num, len(entries))
854
self._merge_names(file_id, parents3, names3, resolver=resolver)
856
file_status = self._do_merge_contents(file_id)
858
file_status = 'unmodified'
859
self._merge_executable(file_id,
860
executable3, file_status, resolver=resolver)
864
child_pb = ui.ui_factory.nested_progress_bar()
866
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
867
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
870
if self.change_reporter is not None:
871
from bzrlib import delta
872
delta.report_changes(
873
self.tt.iter_changes(), self.change_reporter)
874
self.cook_conflicts(fs_conflicts)
875
for conflict in self.cooked_conflicts:
876
trace.warning(conflict)
468
child_pb = ui.ui_factory.nested_progress_bar()
470
for num, (file_id, changed, parents3, names3,
471
executable3) in enumerate(entries):
472
child_pb.update('Preparing file merge', num, len(entries))
473
self._merge_names(file_id, parents3, names3)
475
file_status = self.merge_contents(file_id)
477
file_status = 'unmodified'
478
self._merge_executable(file_id,
479
executable3, file_status)
484
child_pb = ui.ui_factory.nested_progress_bar()
486
fs_conflicts = resolve_conflicts(self.tt, child_pb,
487
lambda t, c: conflict_pass(t, c, self.other_tree))
490
if change_reporter is not None:
491
from bzrlib import delta
492
delta.report_changes(self.tt._iter_changes(), change_reporter)
493
self.cook_conflicts(fs_conflicts)
494
for conflict in self.cooked_conflicts:
497
results = self.tt.apply(no_conflicts=True)
498
self.write_modified(results)
500
working_tree.add_conflicts(self.cooked_conflicts)
501
except UnsupportedOperation:
505
self.other_tree.unlock()
506
self.base_tree.unlock()
507
self.this_tree.unlock()
878
510
def _entries3(self):
879
511
"""Gather data about files modified between three trees.
911
540
result.append((file_id, changed, parents3, names3, executable3))
914
def _entries_lca(self):
915
"""Gather data about files modified between multiple trees.
917
This compares OTHER versus all LCA trees, and for interesting entries,
918
it then compares with THIS and BASE.
920
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
921
:return: [(file_id, changed, parents, names, executable)]
922
file_id Simple file_id of the entry
923
changed Boolean, True if the kind or contents changed
925
parents ((base, [parent_id, in, lcas]), parent_id_other,
927
names ((base, [name, in, lcas]), name_in_other, name_in_this)
928
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
930
if self.interesting_files is not None:
931
lookup_trees = [self.this_tree, self.base_tree]
932
lookup_trees.extend(self._lca_trees)
933
# I think we should include the lca trees as well
934
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
937
interesting_ids = self.interesting_ids
939
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
941
base_inventory = self.base_tree.inventory
942
this_inventory = self.this_tree.inventory
943
for path, file_id, other_ie, lca_values in walker.iter_all():
944
# Is this modified at all from any of the other trees?
946
other_ie = _none_entry
947
if interesting_ids is not None and file_id not in interesting_ids:
950
# If other_revision is found in any of the lcas, that means this
951
# node is uninteresting. This is because when merging, if there are
952
# multiple heads(), we have to create a new node. So if we didn't,
953
# we know that the ancestry is linear, and that OTHER did not
955
# See doc/developers/lca_merge_resolution.txt for details
956
other_revision = other_ie.revision
957
if other_revision is not None:
958
# We can't use this shortcut when other_revision is None,
959
# because it may be None because things are WorkingTrees, and
960
# not because it is *actually* None.
961
is_unmodified = False
962
for lca_path, ie in lca_values:
963
if ie is not None and ie.revision == other_revision:
970
for lca_path, lca_ie in lca_values:
972
lca_entries.append(_none_entry)
974
lca_entries.append(lca_ie)
976
if file_id in base_inventory:
977
base_ie = base_inventory[file_id]
979
base_ie = _none_entry
981
if file_id in this_inventory:
982
this_ie = this_inventory[file_id]
984
this_ie = _none_entry
990
for lca_ie in lca_entries:
991
lca_kinds.append(lca_ie.kind)
992
lca_parent_ids.append(lca_ie.parent_id)
993
lca_names.append(lca_ie.name)
994
lca_executable.append(lca_ie.executable)
996
kind_winner = self._lca_multi_way(
997
(base_ie.kind, lca_kinds),
998
other_ie.kind, this_ie.kind)
999
parent_id_winner = self._lca_multi_way(
1000
(base_ie.parent_id, lca_parent_ids),
1001
other_ie.parent_id, this_ie.parent_id)
1002
name_winner = self._lca_multi_way(
1003
(base_ie.name, lca_names),
1004
other_ie.name, this_ie.name)
1006
content_changed = True
1007
if kind_winner == 'this':
1008
# No kind change in OTHER, see if there are *any* changes
1009
if other_ie.kind == 'directory':
1010
if parent_id_winner == 'this' and name_winner == 'this':
1011
# No change for this directory in OTHER, skip
1013
content_changed = False
1014
elif other_ie.kind is None or other_ie.kind == 'file':
1015
def get_sha1(ie, tree):
1016
if ie.kind != 'file':
1018
return tree.get_file_sha1(file_id)
1019
base_sha1 = get_sha1(base_ie, self.base_tree)
1020
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1021
in zip(lca_entries, self._lca_trees)]
1022
this_sha1 = get_sha1(this_ie, self.this_tree)
1023
other_sha1 = get_sha1(other_ie, self.other_tree)
1024
sha1_winner = self._lca_multi_way(
1025
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1026
allow_overriding_lca=False)
1027
exec_winner = self._lca_multi_way(
1028
(base_ie.executable, lca_executable),
1029
other_ie.executable, this_ie.executable)
1030
if (parent_id_winner == 'this' and name_winner == 'this'
1031
and sha1_winner == 'this' and exec_winner == 'this'):
1032
# No kind, parent, name, exec, or content change for
1033
# OTHER, so this node is not considered interesting
1035
if sha1_winner == 'this':
1036
content_changed = False
1037
elif other_ie.kind == 'symlink':
1038
def get_target(ie, tree):
1039
if ie.kind != 'symlink':
1041
return tree.get_symlink_target(file_id)
1042
base_target = get_target(base_ie, self.base_tree)
1043
lca_targets = [get_target(ie, tree) for ie, tree
1044
in zip(lca_entries, self._lca_trees)]
1045
this_target = get_target(this_ie, self.this_tree)
1046
other_target = get_target(other_ie, self.other_tree)
1047
target_winner = self._lca_multi_way(
1048
(base_target, lca_targets),
1049
other_target, this_target)
1050
if (parent_id_winner == 'this' and name_winner == 'this'
1051
and target_winner == 'this'):
1052
# No kind, parent, name, or symlink target change
1055
if target_winner == 'this':
1056
content_changed = False
1057
elif other_ie.kind == 'tree-reference':
1058
# The 'changed' information seems to be handled at a higher
1059
# level. At least, _entries3 returns False for content
1060
# changed, even when at a new revision_id.
1061
content_changed = False
1062
if (parent_id_winner == 'this' and name_winner == 'this'):
1063
# Nothing interesting
1066
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1068
# If we have gotten this far, that means something has changed
1069
result.append((file_id, content_changed,
1070
((base_ie.parent_id, lca_parent_ids),
1071
other_ie.parent_id, this_ie.parent_id),
1072
((base_ie.name, lca_names),
1073
other_ie.name, this_ie.name),
1074
((base_ie.executable, lca_executable),
1075
other_ie.executable, this_ie.executable)
1080
543
def fix_root(self):
1082
545
self.tt.final_kind(self.tt.root)
1083
except errors.NoSuchFile:
1084
547
self.tt.cancel_deletion(self.tt.root)
1085
548
if self.tt.final_file_id(self.tt.root) is None:
1086
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
549
self.tt.version_file(self.tt.tree_file_id(self.tt.root),
1088
other_root_file_id = self.other_tree.get_root_id()
1089
if other_root_file_id is None:
551
if self.other_tree.inventory.root is None:
553
other_root_file_id = self.other_tree.inventory.root.file_id
1091
554
other_root = self.tt.trans_id_file_id(other_root_file_id)
1092
555
if other_root == self.tt.root:
1094
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
1095
# the other tree's root is a non-root in the current tree (as when
1096
# a previously unrelated branch is merged into another)
1099
558
self.tt.final_kind(other_root)
1100
other_root_is_present = True
1101
except errors.NoSuchFile:
1102
# other_root doesn't have a physical representation. We still need
1103
# to move any references to the actual root of the tree.
1104
other_root_is_present = False
1105
# 'other_tree.inventory.root' is not present in this tree. We are
1106
# calling adjust_path for children which *want* to be present with a
1107
# correct place to go.
1108
for thing, child in self.other_tree.inventory.root.children.iteritems():
561
self.reparent_children(self.other_tree.inventory.root, self.tt.root)
562
self.tt.cancel_creation(other_root)
563
self.tt.cancel_versioning(other_root)
565
def reparent_children(self, ie, target):
566
for thing, child in ie.children.iteritems():
1109
567
trans_id = self.tt.trans_id_file_id(child.file_id)
1110
if not other_root_is_present:
1111
# FIXME: Make final_kind returns None instead of raising
1112
# NoSuchFile to avoid the ugly construct below -- vila 20100402
1114
self.tt.final_kind(trans_id)
1115
# The item exist in the final tree and has a defined place
1118
except errors.NoSuchFile, e:
1120
# Move the item into the root
1121
self.tt.adjust_path(self.tt.final_name(trans_id),
1122
self.tt.root, trans_id)
1123
if other_root_is_present:
1124
self.tt.cancel_creation(other_root)
1125
self.tt.cancel_versioning(other_root)
568
self.tt.adjust_path(self.tt.final_name(trans_id), target, trans_id)
1127
570
def write_modified(self, results):
1128
571
modified_hashes = {}
1331
721
return kind, contents
723
def contents_conflict():
724
trans_id = self.tt.trans_id_file_id(file_id)
725
name = self.tt.final_name(trans_id)
726
parent_id = self.tt.final_parent(trans_id)
727
if file_id in self.this_tree.inventory:
728
self.tt.unversion_file(trans_id)
729
if file_id in self.this_tree:
730
self.tt.delete_contents(trans_id)
731
file_group = self._dump_conflicts(name, parent_id, file_id,
733
self._raw_conflicts.append(('contents conflict', file_group))
1333
735
# See SPOT run. run, SPOT, run.
1334
736
# So we're not QUITE repeating ourselves; we do tricky things with
1336
738
base_pair = contents_pair(self.base_tree)
1337
739
other_pair = contents_pair(self.other_tree)
1339
this_pair = contents_pair(self.this_tree)
1340
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1341
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1342
this_pair, allow_overriding_lca=False)
1344
if base_pair == other_pair:
1347
# We delayed evaluating this_pair as long as we can to avoid
1348
# unnecessary sha1 calculation
1349
this_pair = contents_pair(self.this_tree)
1350
winner = self._three_way(base_pair, other_pair, this_pair)
1351
if winner == 'this':
1352
# No interesting changes introduced by OTHER
1354
# We have a hypothetical conflict, but if we have files, then we
1355
# can try to merge the content
1356
trans_id = self.tt.trans_id_file_id(file_id)
1357
params = MergeHookParams(self, file_id, trans_id, this_pair[0],
1358
other_pair[0], winner)
1359
hooks = self.active_hooks
1360
hook_status = 'not_applicable'
1362
hook_status, lines = hook.merge_contents(params)
1363
if hook_status != 'not_applicable':
1364
# Don't try any more hooks, this one applies.
1367
if hook_status == 'not_applicable':
1368
# This is a contents conflict, because none of the available
1369
# functions could merge it.
1371
name = self.tt.final_name(trans_id)
1372
parent_id = self.tt.final_parent(trans_id)
1373
if self.this_tree.has_id(file_id):
1374
self.tt.unversion_file(trans_id)
1375
file_group = self._dump_conflicts(name, parent_id, file_id,
1377
self._raw_conflicts.append(('contents conflict', file_group))
1378
elif hook_status == 'success':
1379
self.tt.create_file(lines, trans_id)
1380
elif hook_status == 'conflicted':
1381
# XXX: perhaps the hook should be able to provide
1382
# the BASE/THIS/OTHER files?
1383
self.tt.create_file(lines, trans_id)
1384
self._raw_conflicts.append(('text conflict', trans_id))
1385
name = self.tt.final_name(trans_id)
1386
parent_id = self.tt.final_parent(trans_id)
1387
self._dump_conflicts(name, parent_id, file_id)
1388
elif hook_status == 'delete':
1389
self.tt.unversion_file(trans_id)
1391
elif hook_status == 'done':
1392
# The hook function did whatever it needs to do directly, no
1393
# further action needed here.
1396
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1397
if not self.this_tree.has_id(file_id) and result == "modified":
1398
self.tt.version_file(file_id, trans_id)
1399
# The merge has been performed, so the old contents should not be
1402
self.tt.delete_contents(trans_id)
1403
except errors.NoSuchFile:
1407
def _default_other_winner_merge(self, merge_hook_params):
1408
"""Replace this contents with other."""
1409
file_id = merge_hook_params.file_id
1410
trans_id = merge_hook_params.trans_id
1411
file_in_this = self.this_tree.has_id(file_id)
1412
if self.other_tree.has_id(file_id):
1413
# OTHER changed the file
1415
if wt.supports_content_filtering():
1416
# We get the path from the working tree if it exists.
1417
# That fails though when OTHER is adding a file, so
1418
# we fall back to the other tree to find the path if
1419
# it doesn't exist locally.
1421
filter_tree_path = wt.id2path(file_id)
1422
except errors.NoSuchId:
1423
filter_tree_path = self.other_tree.id2path(file_id)
1425
# Skip the id2path lookup for older formats
1426
filter_tree_path = None
1427
transform.create_from_tree(self.tt, trans_id,
1428
self.other_tree, file_id,
1429
filter_tree_path=filter_tree_path)
1432
# OTHER deleted the file
1433
return 'delete', None
1435
raise AssertionError(
1436
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1439
def merge_contents(self, merge_hook_params):
1440
"""Fallback merge logic after user installed hooks."""
1441
# This function is used in merge hooks as the fallback instance.
1442
# Perhaps making this function and the functions it calls be a
1443
# a separate class would be better.
1444
if merge_hook_params.winner == 'other':
1445
# OTHER is a straight winner, so replace this contents with other
1446
return self._default_other_winner_merge(merge_hook_params)
1447
elif merge_hook_params.is_file_merge():
1448
# THIS and OTHER are both files, so text merge. Either
1449
# BASE is a file, or both converted to files, so at least we
1450
# have agreement that output should be a file.
1452
self.text_merge(merge_hook_params.file_id,
1453
merge_hook_params.trans_id)
1454
except errors.BinaryFile:
1455
return 'not_applicable', None
1458
return 'not_applicable', None
740
if base_pair == other_pair:
741
# OTHER introduced no changes
743
this_pair = contents_pair(self.this_tree)
744
if this_pair == other_pair:
745
# THIS and OTHER introduced the same changes
748
trans_id = self.tt.trans_id_file_id(file_id)
749
if this_pair == base_pair:
750
# only OTHER introduced changes
751
if file_id in self.this_tree:
752
# Remove any existing contents
753
self.tt.delete_contents(trans_id)
754
if file_id in self.other_tree:
755
# OTHER changed the file
756
create_by_entry(self.tt,
757
self.other_tree.inventory[file_id],
758
self.other_tree, trans_id)
759
if file_id not in self.this_tree.inventory:
760
self.tt.version_file(file_id, trans_id)
762
elif file_id in self.this_tree.inventory:
763
# OTHER deleted the file
764
self.tt.unversion_file(trans_id)
766
#BOTH THIS and OTHER introduced changes; scalar conflict
767
elif this_pair[0] == "file" and other_pair[0] == "file":
768
# THIS and OTHER are both files, so text merge. Either
769
# BASE is a file, or both converted to files, so at least we
770
# have agreement that output should be a file.
772
self.text_merge(file_id, trans_id)
774
return contents_conflict()
775
if file_id not in self.this_tree.inventory:
776
self.tt.version_file(file_id, trans_id)
778
self.tt.tree_kind(trans_id)
779
self.tt.delete_contents(trans_id)
784
# Scalar conflict, can't text merge. Dump conflicts
785
return contents_conflict()
1460
787
def get_lines(self, tree, file_id):
1461
788
"""Return the lines in a file, or an empty list."""
1462
if tree.has_id(file_id):
1463
return tree.get_file_lines(file_id)
790
return tree.get_file(file_id).readlines()
1516
842
determined automatically. If set_version is true, the .OTHER, .THIS
1517
843
or .BASE (in that order) will be created as versioned files.
1519
data = [('OTHER', self.other_tree, other_lines),
845
data = [('OTHER', self.other_tree, other_lines),
1520
846
('THIS', self.this_tree, this_lines)]
1522
848
data.append(('BASE', self.base_tree, base_lines))
1524
# We need to use the actual path in the working tree of the file here,
1525
# ignoring the conflict suffixes
1527
if wt.supports_content_filtering():
1529
filter_tree_path = wt.id2path(file_id)
1530
except errors.NoSuchId:
1531
# file has been deleted
1532
filter_tree_path = None
1534
# Skip the id2path lookup for older formats
1535
filter_tree_path = None
1537
849
versioned = False
1539
851
for suffix, tree, lines in data:
1540
if tree.has_id(file_id):
1541
853
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1542
suffix, lines, filter_tree_path)
1543
855
file_group.append(trans_id)
1544
856
if set_version and not versioned:
1545
857
self.tt.version_file(file_id, trans_id)
1546
858
versioned = True
1547
859
return file_group
1549
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1550
lines=None, filter_tree_path=None):
861
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1551
863
"""Emit a single conflict file."""
1552
864
name = name + '.' + suffix
1553
865
trans_id = self.tt.create_path(name, parent_id)
1554
transform.create_from_tree(self.tt, trans_id, tree, file_id, lines,
866
entry = tree.inventory[file_id]
867
create_by_entry(self.tt, entry, tree, trans_id, lines)
1558
870
def merge_executable(self, file_id, file_status):
1559
871
"""Perform a merge on the execute bit."""
1560
872
executable = [self.executable(t, file_id) for t in (self.base_tree,
1561
873
self.other_tree, self.this_tree)]
1562
self._merge_executable(file_id, executable, file_status,
1563
resolver=self._three_way)
874
self._merge_executable(file_id, executable, file_status)
1565
def _merge_executable(self, file_id, executable, file_status,
876
def _merge_executable(self, file_id, executable, file_status):
1567
877
"""Perform a merge on the execute bit."""
1568
878
base_executable, other_executable, this_executable = executable
1569
879
if file_status == "deleted":
1571
winner = resolver(*executable)
881
trans_id = self.tt.trans_id_file_id(file_id)
883
if self.tt.final_kind(trans_id) != "file":
887
winner = self._three_way(*executable)
1572
888
if winner == "conflict":
1573
889
# There must be a None in here, if we have a conflict, but we
1574
890
# need executability since file status was not deleted.
1634
935
if path.endswith(suffix):
1635
936
path = path[:-len(suffix)]
1637
c = _mod_conflicts.Conflict.factory(conflict_type,
1638
path=path, file_id=file_id)
1639
elif conflict_type == 'text conflict':
938
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
939
self.cooked_conflicts.append(c)
940
if conflict_type == 'text conflict':
1640
941
trans_id = conflict[1]
1641
942
path = fp.get_path(trans_id)
1642
943
file_id = self.tt.final_file_id(trans_id)
1643
c = _mod_conflicts.Conflict.factory(conflict_type,
1644
path=path, file_id=file_id)
944
c = Conflict.factory(conflict_type, path=path, file_id=file_id)
945
self.cooked_conflicts.append(c)
947
for trans_id, conflicts in name_conflicts.iteritems():
949
this_parent, other_parent = conflicts['parent conflict']
950
assert this_parent != other_parent
952
this_parent = other_parent = \
953
self.tt.final_file_id(self.tt.final_parent(trans_id))
955
this_name, other_name = conflicts['name conflict']
956
assert this_name != other_name
958
this_name = other_name = self.tt.final_name(trans_id)
959
other_path = fp.get_path(trans_id)
960
if this_parent is not None and this_name is not None:
962
fp.get_path(self.tt.trans_id_file_id(this_parent))
963
this_path = pathjoin(this_parent_path, this_name)
1646
raise AssertionError('bad conflict type: %r' % (conflict,))
965
this_path = "<deleted>"
966
file_id = self.tt.final_file_id(trans_id)
967
c = Conflict.factory('path conflict', path=this_path,
968
conflict_path=other_path, file_id=file_id)
1647
969
self.cooked_conflicts.append(c)
1648
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
970
self.cooked_conflicts.sort(key=Conflict.sort_key)
1651
973
class WeaveMerger(Merge3Merger):
1652
974
"""Three-way tree merger, text weave merger."""
1653
975
supports_reprocess = True
1654
976
supports_show_base = False
1655
supports_reverse_cherrypick = False
1656
history_based = True
1658
def _generate_merge_plan(self, file_id, base):
1659
return self.this_tree.plan_file_merge(file_id, self.other_tree,
978
def __init__(self, working_tree, this_tree, base_tree, other_tree,
979
interesting_ids=None, pb=DummyProgress(), pp=None,
980
reprocess=False, change_reporter=None,
981
interesting_files=None):
982
super(WeaveMerger, self).__init__(working_tree, this_tree,
983
base_tree, other_tree,
984
interesting_ids=interesting_ids,
985
pb=pb, pp=pp, reprocess=reprocess,
986
change_reporter=change_reporter)
1662
988
def _merged_lines(self, file_id):
1663
989
"""Generate the merged lines.
1664
990
There is no distinction between lines that are meant to contain <<<<<<<
1668
base = self.base_tree
1671
plan = self._generate_merge_plan(file_id, base)
1672
if 'merge' in debug.debug_flags:
1674
trans_id = self.tt.trans_id_file_id(file_id)
1675
name = self.tt.final_name(trans_id) + '.plan'
1676
contents = ('%11s|%s' % l for l in plan)
1677
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1678
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1679
'>>>>>>> MERGE-SOURCE\n')
1680
lines, conflicts = textmerge.merge_lines(self.reprocess)
1682
base_lines = textmerge.base_from_plan()
1685
return lines, base_lines
993
plan = self.this_tree.plan_file_merge(file_id, self.other_tree)
994
textmerge = PlanWeaveMerge(plan, '<<<<<<< TREE\n',
995
'>>>>>>> MERGE-SOURCE\n')
996
return textmerge.merge_lines(self.reprocess)
1687
998
def text_merge(self, file_id, trans_id):
1688
999
"""Perform a (weave) text merge for a given file and file-id.
1689
1000
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1690
1001
and a conflict will be noted.
1692
lines, base_lines = self._merged_lines(file_id)
1003
lines, conflicts = self._merged_lines(file_id)
1693
1004
lines = list(lines)
1694
# Note we're checking whether the OUTPUT is binary in this case,
1005
# Note we're checking whether the OUTPUT is binary in this case,
1695
1006
# because we don't want to get into weave merge guts.
1696
textfile.check_text_lines(lines)
1007
check_text_lines(lines)
1697
1008
self.tt.create_file(lines, trans_id)
1698
if base_lines is not None:
1700
1010
self._raw_conflicts.append(('text conflict', trans_id))
1701
1011
name = self.tt.final_name(trans_id)
1702
1012
parent_id = self.tt.final_parent(trans_id)
1703
file_group = self._dump_conflicts(name, parent_id, file_id,
1705
base_lines=base_lines)
1013
file_group = self._dump_conflicts(name, parent_id, file_id,
1706
1015
file_group.append(trans_id)
1709
class LCAMerger(WeaveMerger):
1711
def _generate_merge_plan(self, file_id, base):
1712
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1715
1018
class Diff3Merger(Merge3Merger):
1716
1019
"""Three-way merger using external diff3 for text merging"""
1718
1021
def dump_file(self, temp_dir, name, tree, file_id):
1719
out_path = osutils.pathjoin(temp_dir, name)
1022
out_path = pathjoin(temp_dir, name)
1720
1023
out_file = open(out_path, "wb")
1722
1025
in_file = tree.get_file(file_id)
1833
1130
yield status_a(revision, text)
1834
1131
for revision, text in annotated_b[b_cur:bi]:
1835
1132
yield status_b(revision, text)
1836
1134
# and now the matched section
1839
for text_a in plain_a[ai:a_cur]:
1137
for text_a, text_b in zip(plain_a[ai:a_cur], plain_b[bi:b_cur]):
1138
assert text_a == text_b
1840
1139
yield "unchanged", text_a
1843
class _PlanMergeBase(object):
1845
def __init__(self, a_rev, b_rev, vf, key_prefix):
1848
:param a_rev: Revision-id of one revision to merge
1849
:param b_rev: Revision-id of the other revision to merge
1850
:param vf: A VersionedFiles containing both revisions
1851
:param key_prefix: A prefix for accessing keys in vf, typically
1857
self._last_lines = None
1858
self._last_lines_revision_id = None
1859
self._cached_matching_blocks = {}
1860
self._key_prefix = key_prefix
1861
self._precache_tip_lines()
1863
def _precache_tip_lines(self):
1864
lines = self.get_lines([self.a_rev, self.b_rev])
1865
self.lines_a = lines[self.a_rev]
1866
self.lines_b = lines[self.b_rev]
1868
def get_lines(self, revisions):
1869
"""Get lines for revisions from the backing VersionedFiles.
1871
:raises RevisionNotPresent: on absent texts.
1873
keys = [(self._key_prefix + (rev,)) for rev in revisions]
1875
for record in self.vf.get_record_stream(keys, 'unordered', True):
1876
if record.storage_kind == 'absent':
1877
raise errors.RevisionNotPresent(record.key, self.vf)
1878
result[record.key[-1]] = osutils.chunks_to_lines(
1879
record.get_bytes_as('chunked'))
1882
def plan_merge(self):
1883
"""Generate a 'plan' for merging the two revisions.
1885
This involves comparing their texts and determining the cause of
1886
differences. If text A has a line and text B does not, then either the
1887
line was added to text A, or it was deleted from B. Once the causes
1888
are combined, they are written out in the format described in
1889
VersionedFile.plan_merge
1891
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
1892
unique_a, unique_b = self._unique_lines(blocks)
1893
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
1894
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
1895
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
1897
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
1900
for i, j, n in blocks:
1901
for a_index in range(last_i, i):
1902
if a_index in new_a:
1903
if a_index in killed_b:
1904
yield 'conflicted-a', self.lines_a[a_index]
1906
yield 'new-a', self.lines_a[a_index]
1908
yield 'killed-b', self.lines_a[a_index]
1909
for b_index in range(last_j, j):
1910
if b_index in new_b:
1911
if b_index in killed_a:
1912
yield 'conflicted-b', self.lines_b[b_index]
1914
yield 'new-b', self.lines_b[b_index]
1916
yield 'killed-a', self.lines_b[b_index]
1917
# handle common lines
1918
for a_index in range(i, i+n):
1919
yield 'unchanged', self.lines_a[a_index]
1923
def _get_matching_blocks(self, left_revision, right_revision):
1924
"""Return a description of which sections of two revisions match.
1926
See SequenceMatcher.get_matching_blocks
1928
cached = self._cached_matching_blocks.get((left_revision,
1930
if cached is not None:
1932
if self._last_lines_revision_id == left_revision:
1933
left_lines = self._last_lines
1934
right_lines = self.get_lines([right_revision])[right_revision]
1936
lines = self.get_lines([left_revision, right_revision])
1937
left_lines = lines[left_revision]
1938
right_lines = lines[right_revision]
1939
self._last_lines = right_lines
1940
self._last_lines_revision_id = right_revision
1941
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
1943
return matcher.get_matching_blocks()
1945
def _unique_lines(self, matching_blocks):
1946
"""Analyse matching_blocks to determine which lines are unique
1948
:return: a tuple of (unique_left, unique_right), where the values are
1949
sets of line numbers of unique lines.
1955
for i, j, n in matching_blocks:
1956
unique_left.extend(range(last_i, i))
1957
unique_right.extend(range(last_j, j))
1960
return unique_left, unique_right
1963
def _subtract_plans(old_plan, new_plan):
1964
"""Remove changes from new_plan that came from old_plan.
1966
It is assumed that the difference between the old_plan and new_plan
1967
is their choice of 'b' text.
1969
All lines from new_plan that differ from old_plan are emitted
1970
verbatim. All lines from new_plan that match old_plan but are
1971
not about the 'b' revision are emitted verbatim.
1973
Lines that match and are about the 'b' revision are the lines we
1974
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
1975
is skipped entirely.
1977
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
1980
for i, j, n in matcher.get_matching_blocks():
1981
for jj in range(last_j, j):
1983
for jj in range(j, j+n):
1984
plan_line = new_plan[jj]
1985
if plan_line[0] == 'new-b':
1987
elif plan_line[0] == 'killed-b':
1988
yield 'unchanged', plan_line[1]
1994
class _PlanMerge(_PlanMergeBase):
1995
"""Plan an annotate merge using on-the-fly annotation"""
1997
def __init__(self, a_rev, b_rev, vf, key_prefix):
1998
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
1999
self.a_key = self._key_prefix + (self.a_rev,)
2000
self.b_key = self._key_prefix + (self.b_rev,)
2001
self.graph = _mod_graph.Graph(self.vf)
2002
heads = self.graph.heads((self.a_key, self.b_key))
2004
# one side dominates, so we can just return its values, yay for
2006
# Ideally we would know that before we get this far
2007
self._head_key = heads.pop()
2008
if self._head_key == self.a_key:
2012
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2013
self._head_key[-1], other)
2016
self._head_key = None
2019
def _precache_tip_lines(self):
2020
# Turn this into a no-op, because we will do this later
2023
def _find_recursive_lcas(self):
2024
"""Find all the ancestors back to a unique lca"""
2025
cur_ancestors = (self.a_key, self.b_key)
2026
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2027
# rather than a key tuple. We will just map that directly to no common
2031
next_lcas = self.graph.find_lca(*cur_ancestors)
2032
# Map a plain NULL_REVISION to a simple no-ancestors
2033
if next_lcas == set([_mod_revision.NULL_REVISION]):
2035
# Order the lca's based on when they were merged into the tip
2036
# While the actual merge portion of weave merge uses a set() of
2037
# active revisions, the order of insertion *does* effect the
2038
# implicit ordering of the texts.
2039
for rev_key in cur_ancestors:
2040
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2042
parent_map[rev_key] = ordered_parents
2043
if len(next_lcas) == 0:
2045
elif len(next_lcas) == 1:
2046
parent_map[list(next_lcas)[0]] = ()
2048
elif len(next_lcas) > 2:
2049
# More than 2 lca's, fall back to grabbing all nodes between
2050
# this and the unique lca.
2051
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2053
self.a_key, self.b_key, cur_ancestors)
2054
cur_lcas = next_lcas
2055
while len(cur_lcas) > 1:
2056
cur_lcas = self.graph.find_lca(*cur_lcas)
2057
if len(cur_lcas) == 0:
2058
# No common base to find, use the full ancestry
2061
unique_lca = list(cur_lcas)[0]
2062
if unique_lca == _mod_revision.NULL_REVISION:
2063
# find_lca will return a plain 'NULL_REVISION' rather
2064
# than a key tuple when there is no common ancestor, we
2065
# prefer to just use None, because it doesn't confuse
2066
# _get_interesting_texts()
2068
parent_map.update(self._find_unique_parents(next_lcas,
2071
cur_ancestors = next_lcas
2074
def _find_unique_parents(self, tip_keys, base_key):
2075
"""Find ancestors of tip that aren't ancestors of base.
2077
:param tip_keys: Nodes that are interesting
2078
:param base_key: Cull all ancestors of this node
2079
:return: The parent map for all revisions between tip_keys and
2080
base_key. base_key will be included. References to nodes outside of
2081
the ancestor set will also be removed.
2083
# TODO: this would be simpler if find_unique_ancestors took a list
2084
# instead of a single tip, internally it supports it, but it
2085
# isn't a "backwards compatible" api change.
2086
if base_key is None:
2087
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2088
# We remove NULL_REVISION because it isn't a proper tuple key, and
2089
# thus confuses things like _get_interesting_texts, and our logic
2090
# to add the texts into the memory weave.
2091
if _mod_revision.NULL_REVISION in parent_map:
2092
parent_map.pop(_mod_revision.NULL_REVISION)
2095
for tip in tip_keys:
2097
self.graph.find_unique_ancestors(tip, [base_key]))
2098
parent_map = self.graph.get_parent_map(interesting)
2099
parent_map[base_key] = ()
2100
culled_parent_map, child_map, tails = self._remove_external_references(
2102
# Remove all the tails but base_key
2103
if base_key is not None:
2104
tails.remove(base_key)
2105
self._prune_tails(culled_parent_map, child_map, tails)
2106
# Now remove all the uninteresting 'linear' regions
2107
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2111
def _remove_external_references(parent_map):
2112
"""Remove references that go outside of the parent map.
2114
:param parent_map: Something returned from Graph.get_parent_map(keys)
2115
:return: (filtered_parent_map, child_map, tails)
2116
filtered_parent_map is parent_map without external references
2117
child_map is the {parent_key: [child_keys]} mapping
2118
tails is a list of nodes that do not have any parents in the map
2120
# TODO: The basic effect of this function seems more generic than
2121
# _PlanMerge. But the specific details of building a child_map,
2122
# and computing tails seems very specific to _PlanMerge.
2123
# Still, should this be in Graph land?
2124
filtered_parent_map = {}
2127
for key, parent_keys in parent_map.iteritems():
2128
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2129
if not culled_parent_keys:
2131
for parent_key in culled_parent_keys:
2132
child_map.setdefault(parent_key, []).append(key)
2133
# TODO: Do we want to do this, it adds overhead for every node,
2134
# just to say that the node has no children
2135
child_map.setdefault(key, [])
2136
filtered_parent_map[key] = culled_parent_keys
2137
return filtered_parent_map, child_map, tails
2140
def _prune_tails(parent_map, child_map, tails_to_remove):
2141
"""Remove tails from the parent map.
2143
This will remove the supplied revisions until no more children have 0
2146
:param parent_map: A dict of {child: [parents]}, this dictionary will
2147
be modified in place.
2148
:param tails_to_remove: A list of tips that should be removed,
2149
this list will be consumed
2150
:param child_map: The reverse dict of parent_map ({parent: [children]})
2151
this dict will be modified
2152
:return: None, parent_map will be modified in place.
2154
while tails_to_remove:
2155
next = tails_to_remove.pop()
2156
parent_map.pop(next)
2157
children = child_map.pop(next)
2158
for child in children:
2159
child_parents = parent_map[child]
2160
child_parents.remove(next)
2161
if len(child_parents) == 0:
2162
tails_to_remove.append(child)
2164
def _get_interesting_texts(self, parent_map):
2165
"""Return a dict of texts we are interested in.
2167
Note that the input is in key tuples, but the output is in plain
2170
:param parent_map: The output from _find_recursive_lcas
2171
:return: A dict of {'revision_id':lines} as returned by
2172
_PlanMergeBase.get_lines()
2174
all_revision_keys = set(parent_map)
2175
all_revision_keys.add(self.a_key)
2176
all_revision_keys.add(self.b_key)
2178
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2179
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2182
def _build_weave(self):
2183
from bzrlib import weave
2184
self._weave = weave.Weave(weave_name='in_memory_weave',
2185
allow_reserved=True)
2186
parent_map = self._find_recursive_lcas()
2188
all_texts = self._get_interesting_texts(parent_map)
2190
# Note: Unfortunately, the order given by topo_sort will effect the
2191
# ordering resolution in the output. Specifically, if you add A then B,
2192
# then in the output text A lines will show up before B lines. And, of
2193
# course, topo_sort doesn't guarantee any real ordering.
2194
# So we use merge_sort, and add a fake node on the tip.
2195
# This ensures that left-hand parents will always be inserted into the
2196
# weave before right-hand parents.
2197
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2198
parent_map[tip_key] = (self.a_key, self.b_key)
2200
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2204
# for key in tsort.topo_sort(parent_map):
2205
parent_keys = parent_map[key]
2206
revision_id = key[-1]
2207
parent_ids = [k[-1] for k in parent_keys]
2208
self._weave.add_lines(revision_id, parent_ids,
2209
all_texts[revision_id])
2211
def plan_merge(self):
2212
"""Generate a 'plan' for merging the two revisions.
2214
This involves comparing their texts and determining the cause of
2215
differences. If text A has a line and text B does not, then either the
2216
line was added to text A, or it was deleted from B. Once the causes
2217
are combined, they are written out in the format described in
2218
VersionedFile.plan_merge
2220
if self._head_key is not None: # There was a single head
2221
if self._head_key == self.a_key:
2224
if self._head_key != self.b_key:
2225
raise AssertionError('There was an invalid head: %s != %s'
2226
% (self.b_key, self._head_key))
2228
head_rev = self._head_key[-1]
2229
lines = self.get_lines([head_rev])[head_rev]
2230
return ((plan, line) for line in lines)
2231
return self._weave.plan_merge(self.a_rev, self.b_rev)
2234
class _PlanLCAMerge(_PlanMergeBase):
2236
This merge algorithm differs from _PlanMerge in that:
2237
1. comparisons are done against LCAs only
2238
2. cases where a contested line is new versus one LCA but old versus
2239
another are marked as conflicts, by emitting the line as conflicted-a
2242
This is faster, and hopefully produces more useful output.
2245
def __init__(self, a_rev, b_rev, vf, key_prefix, graph):
2246
_PlanMergeBase.__init__(self, a_rev, b_rev, vf, key_prefix)
2247
lcas = graph.find_lca(key_prefix + (a_rev,), key_prefix + (b_rev,))
2250
if lca == _mod_revision.NULL_REVISION:
2253
self.lcas.add(lca[-1])
2254
for lca in self.lcas:
2255
if _mod_revision.is_null(lca):
2258
lca_lines = self.get_lines([lca])[lca]
2259
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_a,
2261
blocks = list(matcher.get_matching_blocks())
2262
self._cached_matching_blocks[(a_rev, lca)] = blocks
2263
matcher = patiencediff.PatienceSequenceMatcher(None, self.lines_b,
2265
blocks = list(matcher.get_matching_blocks())
2266
self._cached_matching_blocks[(b_rev, lca)] = blocks
2268
def _determine_status(self, revision_id, unique_line_numbers):
2269
"""Determines the status unique lines versus all lcas.
2271
Basically, determines why the line is unique to this revision.
2273
A line may be determined new, killed, or both.
2275
If a line is determined new, that means it was not present in at least
2276
one LCA, and is not present in the other merge revision.
2278
If a line is determined killed, that means the line was present in
2281
If a line is killed and new, this indicates that the two merge
2282
revisions contain differing conflict resolutions.
2283
:param revision_id: The id of the revision in which the lines are
2285
:param unique_line_numbers: The line numbers of unique lines.
2286
:return a tuple of (new_this, killed_other):
2290
unique_line_numbers = set(unique_line_numbers)
2291
for lca in self.lcas:
2292
blocks = self._get_matching_blocks(revision_id, lca)
2293
unique_vs_lca, _ignored = self._unique_lines(blocks)
2294
new.update(unique_line_numbers.intersection(unique_vs_lca))
2295
killed.update(unique_line_numbers.difference(unique_vs_lca))
added
removed
bzrlib/merge.py
