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,
40
from bzrlib.cleanup import OperationWithCleanups
47
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
51
56
# TODO: Report back as changes are merged in
54
59
def transform_tree(from_tree, to_tree, interesting_ids=None):
55
from_tree.lock_tree_write()
56
operation = 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)
66
self.create_hook(hooks.HookPoint('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)
60
merge_inner(from_tree.branch, to_tree, from_tree, ignore_zero=True,
61
interesting_ids=interesting_ids, this_tree=from_tree)
264
64
class Merger(object):
268
65
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):
66
this_tree=None, pb=DummyProgress(), change_reporter=None,
271
68
object.__init__(self)
69
assert this_tree is not None, "this_tree is required"
272
70
self.this_branch = this_branch
273
71
self.this_basis = _mod_revision.ensure_null(
274
72
this_branch.last_revision())
451
183
if self.other_rev_id is None:
452
184
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)
185
changes = other_basis_tree.changes_from(self.other_tree)
186
if changes.has_changed():
187
raise WorkingTreeNotRevision(self.this_tree)
455
188
other_rev_id = self.other_basis
456
189
self.other_tree = other_basis_tree
458
@deprecated_method(deprecated_in((2, 1, 0)))
459
191
def file_revisions(self, file_id):
460
192
self.ensure_revision_trees()
461
193
def get_id(tree, file_id):
462
194
revision_id = tree.inventory[file_id].revision
195
assert revision_id is not None
463
196
return revision_id
464
197
if self.this_rev_id is None:
465
198
if self.this_basis_tree.get_file_sha1(file_id) != \
466
199
self.this_tree.get_file_sha1(file_id):
467
raise errors.WorkingTreeNotRevision(self.this_tree)
200
raise WorkingTreeNotRevision(self.this_tree)
469
202
trees = (self.this_basis_tree, self.other_tree)
470
203
return [get_id(tree, file_id) for tree in trees]
472
@deprecated_method(deprecated_in((2, 1, 0)))
473
205
def check_basis(self, check_clean, require_commits=True):
474
206
if self.this_basis is None and require_commits is True:
475
raise errors.BzrCommandError(
476
"This branch has no commits."
477
" (perhaps you would prefer 'bzr pull')")
207
raise BzrCommandError("This branch has no commits."
208
" (perhaps you would prefer 'bzr pull')")
479
210
self.compare_basis()
480
211
if self.this_basis != self.this_rev_id:
481
212
raise errors.UncommittedChanges(self.this_tree)
483
@deprecated_method(deprecated_in((2, 1, 0)))
484
214
def compare_basis(self):
486
216
basis_tree = self.revision_tree(self.this_tree.last_revision())
487
except errors.NoSuchRevision:
217
except errors.RevisionNotPresent:
488
218
basis_tree = self.this_tree.basis_tree()
489
if not self.this_tree.has_changes(basis_tree):
219
changes = self.this_tree.changes_from(basis_tree)
220
if not changes.has_changed():
490
221
self.this_rev_id = self.this_basis
492
223
def set_interesting_files(self, file_list):
493
224
self.interesting_files = file_list
495
226
def set_pending(self):
496
if (not self.base_is_ancestor or not self.base_is_other_ancestor
497
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:
499
229
self._add_parent()
501
231
def _add_parent(self):
502
232
new_parents = self.this_tree.get_parent_ids() + [self.other_rev_id]
503
233
new_parent_trees = []
504
operation = OperationWithCleanups(self.this_tree.set_parent_trees)
505
234
for revision_id in new_parents:
507
236
tree = self.revision_tree(revision_id)
508
except errors.NoSuchRevision:
237
except errors.RevisionNotPresent:
512
operation.add_cleanup(tree.unlock)
513
241
new_parent_trees.append((revision_id, tree))
514
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:
516
250
def set_other(self, other_revision, possible_transports=None):
517
251
"""Set the revision and tree to merge from.
634
340
self.base_rev_id = _mod_revision.ensure_null(
635
341
base_branch.get_rev_id(base_revision[1]))
636
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,
638
def make_merger(self):
639
351
kwargs = {'working_tree':self.this_tree, 'this_tree': self.this_tree,
640
352
'other_tree': self.other_tree,
641
353
'interesting_ids': self.interesting_ids,
642
354
'interesting_files': self.interesting_files,
643
'this_branch': self.this_branch,
645
356
if self.merge_type.requires_base:
646
357
kwargs['base_tree'] = self.base_tree
647
358
if self.merge_type.supports_reprocess:
648
359
kwargs['reprocess'] = self.reprocess
649
360
elif self.reprocess:
650
raise errors.BzrError(
651
"Conflict reduction is not supported for merge"
652
" type %s." % self.merge_type)
361
raise BzrError("Conflict reduction is not supported for merge"
362
" type %s." % self.merge_type)
653
363
if self.merge_type.supports_show_base:
654
364
kwargs['show_base'] = self.show_base
655
365
elif self.show_base:
656
raise errors.BzrError("Showing base is not supported for this"
366
raise BzrError("Showing base is not supported for this"
657
367
" merge type. %s" % self.merge_type)
658
if (not getattr(self.merge_type, 'supports_reverse_cherrypick', True)
659
and not self.base_is_other_ancestor):
660
raise errors.CannotReverseCherrypick()
661
if self.merge_type.supports_cherrypick:
662
kwargs['cherrypick'] = (not self.base_is_ancestor or
663
not self.base_is_other_ancestor)
664
if self._is_criss_cross and getattr(self.merge_type,
665
'supports_lca_trees', False):
666
kwargs['lca_trees'] = self._lca_trees
667
return self.merge_type(pb=None,
668
change_reporter=self.change_reporter,
671
def _do_merge_to(self):
672
merge = self.make_merger()
673
if self.other_branch is not None:
674
self.other_branch.update_references(self.this_branch)
676
if self.recurse == 'down':
677
for relpath, file_id in self.this_tree.iter_references():
678
sub_tree = self.this_tree.get_nested_tree(file_id, relpath)
679
other_revision = self.other_tree.get_reference_revision(
681
if other_revision == sub_tree.last_revision():
683
sub_merge = Merger(sub_tree.branch, this_tree=sub_tree)
684
sub_merge.merge_type = self.merge_type
685
other_branch = self.other_branch.reference_parent(file_id, relpath)
686
sub_merge.set_other_revision(other_revision, other_branch)
687
base_revision = self.base_tree.get_reference_revision(file_id)
688
sub_merge.base_tree = \
689
sub_tree.branch.repository.revision_tree(base_revision)
690
sub_merge.base_rev_id = base_revision
695
operation = OperationWithCleanups(self._do_merge_to)
696
368
self.this_tree.lock_tree_write()
697
operation.add_cleanup(self.this_tree.unlock)
698
369
if self.base_tree is not None:
699
370
self.base_tree.lock_read()
700
operation.add_cleanup(self.base_tree.unlock)
701
371
if self.other_tree is not None:
702
372
self.other_tree.lock_read()
703
operation.add_cleanup(self.other_tree.unlock)
704
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()
705
400
if len(merge.cooked_conflicts) == 0:
706
if not self.ignore_zero and not trace.is_quiet():
707
trace.note("All changes applied successfully.")
401
if not self.ignore_zero:
402
note("All changes applied successfully.")
709
trace.note("%d conflicts encountered."
710
% len(merge.cooked_conflicts))
404
note("%d conflicts encountered." % len(merge.cooked_conflicts))
712
406
return len(merge.cooked_conflicts)
715
class _InventoryNoneEntry(object):
716
"""This represents an inventory entry which *isn't there*.
718
It simplifies the merging logic if we always have an InventoryEntry, even
719
if it isn't actually present
726
symlink_target = None
729
_none_entry = _InventoryNoneEntry()
732
409
class Merge3Merger(object):
733
410
"""Three-way merger that uses the merge3 text merger"""
734
411
requires_base = True
735
412
supports_reprocess = True
736
413
supports_show_base = True
737
414
history_based = False
738
supports_cherrypick = True
739
supports_reverse_cherrypick = True
740
415
winner_idx = {"this": 2, "other": 1, "conflict": 1}
741
supports_lca_trees = True
743
def __init__(self, working_tree, this_tree, base_tree, other_tree,
417
def __init__(self, working_tree, this_tree, base_tree, other_tree,
744
418
interesting_ids=None, reprocess=False, show_base=False,
745
pb=None, pp=None, change_reporter=None,
746
interesting_files=None, do_merge=True,
747
cherrypick=False, lca_trees=None, this_branch=None):
419
pb=DummyProgress(), pp=None, change_reporter=None,
420
interesting_files=None):
748
421
"""Initialize the merger object and perform the merge.
750
423
:param working_tree: The working tree to apply the merge to
751
424
:param this_tree: The local tree in the merge operation
752
425
:param base_tree: The common tree in the merge operation
753
:param other_tree: The other tree to merge changes from
754
:param this_branch: The branch associated with this_tree. Defaults to
755
this_tree.branch if not supplied.
426
:param other_tree: The other other tree to merge changes from
756
427
:param interesting_ids: The file_ids of files that should be
757
428
participate in the merge. May not be combined with
758
429
interesting_files.
759
430
:param: reprocess If True, perform conflict-reduction processing.
760
431
:param show_base: If True, show the base revision in text conflicts.
761
432
(incompatible with reprocess)
433
:param pb: A Progress bar
763
434
:param pp: A ProgressPhase object
764
435
:param change_reporter: An object that should report changes made
765
436
:param interesting_files: The tree-relative paths of files that should
768
439
be combined with interesting_ids. If neither interesting_files nor
769
440
interesting_ids is specified, all files may participate in the
771
:param lca_trees: Can be set to a dictionary of {revision_id:rev_tree}
772
if the ancestry was found to include a criss-cross merge.
773
Otherwise should be None.
775
443
object.__init__(self)
776
if interesting_files is not None and interesting_ids is not None:
778
'specify either interesting_ids or interesting_files')
779
if this_branch is None:
780
this_branch = this_tree.branch
444
if interesting_files is not None:
445
assert interesting_ids is None
781
446
self.interesting_ids = interesting_ids
782
447
self.interesting_files = interesting_files
783
448
self.this_tree = working_tree
449
self.this_tree.lock_tree_write()
784
450
self.base_tree = base_tree
451
self.base_tree.lock_read()
785
452
self.other_tree = other_tree
786
self.this_branch = this_branch
453
self.other_tree.lock_read()
787
454
self._raw_conflicts = []
788
455
self.cooked_conflicts = []
789
456
self.reprocess = reprocess
790
457
self.show_base = show_base
791
self._lca_trees = lca_trees
792
# Uncommenting this will change the default algorithm to always use
793
# _entries_lca. This can be useful for running the test suite and
794
# making sure we haven't missed any corner cases.
795
# if lca_trees is None:
796
# self._lca_trees = [self.base_tree]
797
460
self.change_reporter = change_reporter
798
self.cherrypick = cherrypick
802
warnings.warn("pp argument to Merge3Merger is deprecated")
804
warnings.warn("pb argument to Merge3Merger is deprecated")
807
operation = OperationWithCleanups(self._do_merge)
808
self.this_tree.lock_tree_write()
809
operation.add_cleanup(self.this_tree.unlock)
810
self.base_tree.lock_read()
811
operation.add_cleanup(self.base_tree.unlock)
812
self.other_tree.lock_read()
813
operation.add_cleanup(self.other_tree.unlock)
816
def _do_merge(self, operation):
817
self.tt = transform.TreeTransform(self.this_tree, None)
818
operation.add_cleanup(self.tt.finalize)
819
self._compute_transform()
820
results = self.tt.apply(no_conflicts=True)
821
self.write_modified(results)
462
self.pp = ProgressPhase("Merge phase", 3, self.pb)
464
self.tt = TreeTransform(working_tree, self.pb)
823
self.this_tree.add_conflicts(self.cooked_conflicts)
824
except errors.UnsupportedOperation:
827
def make_preview_transform(self):
828
operation = OperationWithCleanups(self._make_preview_transform)
829
self.base_tree.lock_read()
830
operation.add_cleanup(self.base_tree.unlock)
831
self.other_tree.lock_read()
832
operation.add_cleanup(self.other_tree.unlock)
833
return operation.run_simple()
835
def _make_preview_transform(self):
836
self.tt = transform.TransformPreview(self.this_tree)
837
self._compute_transform()
840
def _compute_transform(self):
841
if self._lca_trees is None:
842
467
entries = self._entries3()
843
resolver = self._three_way
845
entries = self._entries_lca()
846
resolver = self._lca_multi_way
847
child_pb = ui.ui_factory.nested_progress_bar()
849
factories = Merger.hooks['merge_file_content']
850
hooks = [factory(self) for factory in factories] + [self]
851
self.active_hooks = [hook for hook in hooks if hook is not None]
852
for num, (file_id, changed, parents3, names3,
853
executable3) in enumerate(entries):
854
child_pb.update('Preparing file merge', num, len(entries))
855
self._merge_names(file_id, parents3, names3, resolver=resolver)
857
file_status = self._do_merge_contents(file_id)
859
file_status = 'unmodified'
860
self._merge_executable(file_id,
861
executable3, file_status, resolver=resolver)
865
self._finish_computing_transform()
867
def _finish_computing_transform(self):
868
"""Finalize the transform and report the changes.
870
This is the second half of _compute_transform.
872
child_pb = ui.ui_factory.nested_progress_bar()
874
fs_conflicts = transform.resolve_conflicts(self.tt, child_pb,
875
lambda t, c: transform.conflict_pass(t, c, self.other_tree))
878
if self.change_reporter is not None:
879
from bzrlib import delta
880
delta.report_changes(
881
self.tt.iter_changes(), self.change_reporter)
882
self.cook_conflicts(fs_conflicts)
883
for conflict in self.cooked_conflicts:
884
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()
886
510
def _entries3(self):
887
511
"""Gather data about files modified between three trees.
919
540
result.append((file_id, changed, parents3, names3, executable3))
922
def _entries_lca(self):
923
"""Gather data about files modified between multiple trees.
925
This compares OTHER versus all LCA trees, and for interesting entries,
926
it then compares with THIS and BASE.
928
For the multi-valued entries, the format will be (BASE, [lca1, lca2])
929
:return: [(file_id, changed, parents, names, executable)]
930
file_id Simple file_id of the entry
931
changed Boolean, True if the kind or contents changed
933
parents ((base, [parent_id, in, lcas]), parent_id_other,
935
names ((base, [name, in, lcas]), name_in_other, name_in_this)
936
executable ((base, [exec, in, lcas]), exec_in_other, exec_in_this)
938
if self.interesting_files is not None:
939
lookup_trees = [self.this_tree, self.base_tree]
940
lookup_trees.extend(self._lca_trees)
941
# I think we should include the lca trees as well
942
interesting_ids = self.other_tree.paths2ids(self.interesting_files,
945
interesting_ids = self.interesting_ids
947
walker = _mod_tree.MultiWalker(self.other_tree, self._lca_trees)
949
base_inventory = self.base_tree.inventory
950
this_inventory = self.this_tree.inventory
951
for path, file_id, other_ie, lca_values in walker.iter_all():
952
# Is this modified at all from any of the other trees?
954
other_ie = _none_entry
955
if interesting_ids is not None and file_id not in interesting_ids:
958
# If other_revision is found in any of the lcas, that means this
959
# node is uninteresting. This is because when merging, if there are
960
# multiple heads(), we have to create a new node. So if we didn't,
961
# we know that the ancestry is linear, and that OTHER did not
963
# See doc/developers/lca_merge_resolution.txt for details
964
other_revision = other_ie.revision
965
if other_revision is not None:
966
# We can't use this shortcut when other_revision is None,
967
# because it may be None because things are WorkingTrees, and
968
# not because it is *actually* None.
969
is_unmodified = False
970
for lca_path, ie in lca_values:
971
if ie is not None and ie.revision == other_revision:
978
for lca_path, lca_ie in lca_values:
980
lca_entries.append(_none_entry)
982
lca_entries.append(lca_ie)
984
if file_id in base_inventory:
985
base_ie = base_inventory[file_id]
987
base_ie = _none_entry
989
if file_id in this_inventory:
990
this_ie = this_inventory[file_id]
992
this_ie = _none_entry
998
for lca_ie in lca_entries:
999
lca_kinds.append(lca_ie.kind)
1000
lca_parent_ids.append(lca_ie.parent_id)
1001
lca_names.append(lca_ie.name)
1002
lca_executable.append(lca_ie.executable)
1004
kind_winner = self._lca_multi_way(
1005
(base_ie.kind, lca_kinds),
1006
other_ie.kind, this_ie.kind)
1007
parent_id_winner = self._lca_multi_way(
1008
(base_ie.parent_id, lca_parent_ids),
1009
other_ie.parent_id, this_ie.parent_id)
1010
name_winner = self._lca_multi_way(
1011
(base_ie.name, lca_names),
1012
other_ie.name, this_ie.name)
1014
content_changed = True
1015
if kind_winner == 'this':
1016
# No kind change in OTHER, see if there are *any* changes
1017
if other_ie.kind == 'directory':
1018
if parent_id_winner == 'this' and name_winner == 'this':
1019
# No change for this directory in OTHER, skip
1021
content_changed = False
1022
elif other_ie.kind is None or other_ie.kind == 'file':
1023
def get_sha1(ie, tree):
1024
if ie.kind != 'file':
1026
return tree.get_file_sha1(file_id)
1027
base_sha1 = get_sha1(base_ie, self.base_tree)
1028
lca_sha1s = [get_sha1(ie, tree) for ie, tree
1029
in zip(lca_entries, self._lca_trees)]
1030
this_sha1 = get_sha1(this_ie, self.this_tree)
1031
other_sha1 = get_sha1(other_ie, self.other_tree)
1032
sha1_winner = self._lca_multi_way(
1033
(base_sha1, lca_sha1s), other_sha1, this_sha1,
1034
allow_overriding_lca=False)
1035
exec_winner = self._lca_multi_way(
1036
(base_ie.executable, lca_executable),
1037
other_ie.executable, this_ie.executable)
1038
if (parent_id_winner == 'this' and name_winner == 'this'
1039
and sha1_winner == 'this' and exec_winner == 'this'):
1040
# No kind, parent, name, exec, or content change for
1041
# OTHER, so this node is not considered interesting
1043
if sha1_winner == 'this':
1044
content_changed = False
1045
elif other_ie.kind == 'symlink':
1046
def get_target(ie, tree):
1047
if ie.kind != 'symlink':
1049
return tree.get_symlink_target(file_id)
1050
base_target = get_target(base_ie, self.base_tree)
1051
lca_targets = [get_target(ie, tree) for ie, tree
1052
in zip(lca_entries, self._lca_trees)]
1053
this_target = get_target(this_ie, self.this_tree)
1054
other_target = get_target(other_ie, self.other_tree)
1055
target_winner = self._lca_multi_way(
1056
(base_target, lca_targets),
1057
other_target, this_target)
1058
if (parent_id_winner == 'this' and name_winner == 'this'
1059
and target_winner == 'this'):
1060
# No kind, parent, name, or symlink target change
1063
if target_winner == 'this':
1064
content_changed = False
1065
elif other_ie.kind == 'tree-reference':
1066
# The 'changed' information seems to be handled at a higher
1067
# level. At least, _entries3 returns False for content
1068
# changed, even when at a new revision_id.
1069
content_changed = False
1070
if (parent_id_winner == 'this' and name_winner == 'this'):
1071
# Nothing interesting
1074
raise AssertionError('unhandled kind: %s' % other_ie.kind)
1076
# If we have gotten this far, that means something has changed
1077
result.append((file_id, content_changed,
1078
((base_ie.parent_id, lca_parent_ids),
1079
other_ie.parent_id, this_ie.parent_id),
1080
((base_ie.name, lca_names),
1081
other_ie.name, this_ie.name),
1082
((base_ie.executable, lca_executable),
1083
other_ie.executable, this_ie.executable)
1087
543
def fix_root(self):
1089
545
self.tt.final_kind(self.tt.root)
1090
except errors.NoSuchFile:
1091
547
self.tt.cancel_deletion(self.tt.root)
1092
548
if self.tt.final_file_id(self.tt.root) is None:
1093
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),
1095
other_root_file_id = self.other_tree.get_root_id()
1096
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
1098
554
other_root = self.tt.trans_id_file_id(other_root_file_id)
1099
555
if other_root == self.tt.root:
1101
if self.other_tree.inventory.root.file_id in self.this_tree.inventory:
1102
# the other tree's root is a non-root in the current tree (as when
1103
# a previously unrelated branch is merged into another)
1106
558
self.tt.final_kind(other_root)
1107
other_root_is_present = True
1108
except errors.NoSuchFile:
1109
# other_root doesn't have a physical representation. We still need
1110
# to move any references to the actual root of the tree.
1111
other_root_is_present = False
1112
# 'other_tree.inventory.root' is not present in this tree. We are
1113
# calling adjust_path for children which *want* to be present with a
1114
# correct place to go.
1115
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():
1116
567
trans_id = self.tt.trans_id_file_id(child.file_id)
1117
if not other_root_is_present:
1118
# FIXME: Make final_kind returns None instead of raising
1119
# NoSuchFile to avoid the ugly construct below -- vila 20100402
1121
self.tt.final_kind(trans_id)
1122
# The item exist in the final tree and has a defined place
1125
except errors.NoSuchFile, e:
1127
# Move the item into the root
1128
self.tt.adjust_path(self.tt.final_name(trans_id),
1129
self.tt.root, trans_id)
1130
if other_root_is_present:
1131
self.tt.cancel_creation(other_root)
1132
self.tt.cancel_versioning(other_root)
568
self.tt.adjust_path(self.tt.final_name(trans_id), target, trans_id)
1134
570
def write_modified(self, results):
1135
571
modified_hashes = {}
1338
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))
1340
735
# See SPOT run. run, SPOT, run.
1341
736
# So we're not QUITE repeating ourselves; we do tricky things with
1343
738
base_pair = contents_pair(self.base_tree)
1344
739
other_pair = contents_pair(self.other_tree)
1346
this_pair = contents_pair(self.this_tree)
1347
lca_pairs = [contents_pair(tree) for tree in self._lca_trees]
1348
winner = self._lca_multi_way((base_pair, lca_pairs), other_pair,
1349
this_pair, allow_overriding_lca=False)
1351
if base_pair == other_pair:
1354
# We delayed evaluating this_pair as long as we can to avoid
1355
# unnecessary sha1 calculation
1356
this_pair = contents_pair(self.this_tree)
1357
winner = self._three_way(base_pair, other_pair, this_pair)
1358
if winner == 'this':
1359
# No interesting changes introduced by OTHER
1361
# We have a hypothetical conflict, but if we have files, then we
1362
# can try to merge the content
1363
trans_id = self.tt.trans_id_file_id(file_id)
1364
params = MergeHookParams(self, file_id, trans_id, this_pair[0],
1365
other_pair[0], winner)
1366
hooks = self.active_hooks
1367
hook_status = 'not_applicable'
1369
hook_status, lines = hook.merge_contents(params)
1370
if hook_status != 'not_applicable':
1371
# Don't try any more hooks, this one applies.
1374
if hook_status == 'not_applicable':
1375
# This is a contents conflict, because none of the available
1376
# functions could merge it.
1378
name = self.tt.final_name(trans_id)
1379
parent_id = self.tt.final_parent(trans_id)
1380
if self.this_tree.has_id(file_id):
1381
self.tt.unversion_file(trans_id)
1382
file_group = self._dump_conflicts(name, parent_id, file_id,
1384
self._raw_conflicts.append(('contents conflict', file_group))
1385
elif hook_status == 'success':
1386
self.tt.create_file(lines, trans_id)
1387
elif hook_status == 'conflicted':
1388
# XXX: perhaps the hook should be able to provide
1389
# the BASE/THIS/OTHER files?
1390
self.tt.create_file(lines, trans_id)
1391
self._raw_conflicts.append(('text conflict', trans_id))
1392
name = self.tt.final_name(trans_id)
1393
parent_id = self.tt.final_parent(trans_id)
1394
self._dump_conflicts(name, parent_id, file_id)
1395
elif hook_status == 'delete':
1396
self.tt.unversion_file(trans_id)
1398
elif hook_status == 'done':
1399
# The hook function did whatever it needs to do directly, no
1400
# further action needed here.
1403
raise AssertionError('unknown hook_status: %r' % (hook_status,))
1404
if not self.this_tree.has_id(file_id) and result == "modified":
1405
self.tt.version_file(file_id, trans_id)
1406
# The merge has been performed, so the old contents should not be
1409
self.tt.delete_contents(trans_id)
1410
except errors.NoSuchFile:
1414
def _default_other_winner_merge(self, merge_hook_params):
1415
"""Replace this contents with other."""
1416
file_id = merge_hook_params.file_id
1417
trans_id = merge_hook_params.trans_id
1418
file_in_this = self.this_tree.has_id(file_id)
1419
if self.other_tree.has_id(file_id):
1420
# OTHER changed the file
1422
if wt.supports_content_filtering():
1423
# We get the path from the working tree if it exists.
1424
# That fails though when OTHER is adding a file, so
1425
# we fall back to the other tree to find the path if
1426
# it doesn't exist locally.
1428
filter_tree_path = wt.id2path(file_id)
1429
except errors.NoSuchId:
1430
filter_tree_path = self.other_tree.id2path(file_id)
1432
# Skip the id2path lookup for older formats
1433
filter_tree_path = None
1434
transform.create_from_tree(self.tt, trans_id,
1435
self.other_tree, file_id,
1436
filter_tree_path=filter_tree_path)
1439
# OTHER deleted the file
1440
return 'delete', None
1442
raise AssertionError(
1443
'winner is OTHER, but file_id %r not in THIS or OTHER tree'
1446
def merge_contents(self, merge_hook_params):
1447
"""Fallback merge logic after user installed hooks."""
1448
# This function is used in merge hooks as the fallback instance.
1449
# Perhaps making this function and the functions it calls be a
1450
# a separate class would be better.
1451
if merge_hook_params.winner == 'other':
1452
# OTHER is a straight winner, so replace this contents with other
1453
return self._default_other_winner_merge(merge_hook_params)
1454
elif merge_hook_params.is_file_merge():
1455
# THIS and OTHER are both files, so text merge. Either
1456
# BASE is a file, or both converted to files, so at least we
1457
# have agreement that output should be a file.
1459
self.text_merge(merge_hook_params.file_id,
1460
merge_hook_params.trans_id)
1461
except errors.BinaryFile:
1462
return 'not_applicable', None
1465
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()
1467
787
def get_lines(self, tree, file_id):
1468
788
"""Return the lines in a file, or an empty list."""
1469
if tree.has_id(file_id):
1470
return tree.get_file_lines(file_id)
790
return tree.get_file(file_id).readlines()
1523
842
determined automatically. If set_version is true, the .OTHER, .THIS
1524
843
or .BASE (in that order) will be created as versioned files.
1526
data = [('OTHER', self.other_tree, other_lines),
845
data = [('OTHER', self.other_tree, other_lines),
1527
846
('THIS', self.this_tree, this_lines)]
1529
848
data.append(('BASE', self.base_tree, base_lines))
1531
# We need to use the actual path in the working tree of the file here,
1532
# ignoring the conflict suffixes
1534
if wt.supports_content_filtering():
1536
filter_tree_path = wt.id2path(file_id)
1537
except errors.NoSuchId:
1538
# file has been deleted
1539
filter_tree_path = None
1541
# Skip the id2path lookup for older formats
1542
filter_tree_path = None
1544
849
versioned = False
1546
851
for suffix, tree, lines in data:
1547
if tree.has_id(file_id):
1548
853
trans_id = self._conflict_file(name, parent_id, tree, file_id,
1549
suffix, lines, filter_tree_path)
1550
855
file_group.append(trans_id)
1551
856
if set_version and not versioned:
1552
857
self.tt.version_file(file_id, trans_id)
1553
858
versioned = True
1554
859
return file_group
1556
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1557
lines=None, filter_tree_path=None):
861
def _conflict_file(self, name, parent_id, tree, file_id, suffix,
1558
863
"""Emit a single conflict file."""
1559
864
name = name + '.' + suffix
1560
865
trans_id = self.tt.create_path(name, parent_id)
1561
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)
1565
870
def merge_executable(self, file_id, file_status):
1566
871
"""Perform a merge on the execute bit."""
1567
872
executable = [self.executable(t, file_id) for t in (self.base_tree,
1568
873
self.other_tree, self.this_tree)]
1569
self._merge_executable(file_id, executable, file_status,
1570
resolver=self._three_way)
874
self._merge_executable(file_id, executable, file_status)
1572
def _merge_executable(self, file_id, executable, file_status,
876
def _merge_executable(self, file_id, executable, file_status):
1574
877
"""Perform a merge on the execute bit."""
1575
878
base_executable, other_executable, this_executable = executable
1576
879
if file_status == "deleted":
1578
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)
1579
888
if winner == "conflict":
1580
889
# There must be a None in here, if we have a conflict, but we
1581
890
# need executability since file status was not deleted.
1641
935
if path.endswith(suffix):
1642
936
path = path[:-len(suffix)]
1644
c = _mod_conflicts.Conflict.factory(conflict_type,
1645
path=path, file_id=file_id)
1646
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':
1647
941
trans_id = conflict[1]
1648
942
path = fp.get_path(trans_id)
1649
943
file_id = self.tt.final_file_id(trans_id)
1650
c = _mod_conflicts.Conflict.factory(conflict_type,
1651
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)
1653
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)
1654
969
self.cooked_conflicts.append(c)
1655
self.cooked_conflicts.sort(key=_mod_conflicts.Conflict.sort_key)
970
self.cooked_conflicts.sort(key=Conflict.sort_key)
1658
973
class WeaveMerger(Merge3Merger):
1659
974
"""Three-way tree merger, text weave merger."""
1660
975
supports_reprocess = True
1661
976
supports_show_base = False
1662
supports_reverse_cherrypick = False
1663
history_based = True
1665
def _generate_merge_plan(self, file_id, base):
1666
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)
1669
988
def _merged_lines(self, file_id):
1670
989
"""Generate the merged lines.
1671
990
There is no distinction between lines that are meant to contain <<<<<<<
1675
base = self.base_tree
1678
plan = self._generate_merge_plan(file_id, base)
1679
if 'merge' in debug.debug_flags:
1681
trans_id = self.tt.trans_id_file_id(file_id)
1682
name = self.tt.final_name(trans_id) + '.plan'
1683
contents = ('%11s|%s' % l for l in plan)
1684
self.tt.new_file(name, self.tt.final_parent(trans_id), contents)
1685
textmerge = versionedfile.PlanWeaveMerge(plan, '<<<<<<< TREE\n',
1686
'>>>>>>> MERGE-SOURCE\n')
1687
lines, conflicts = textmerge.merge_lines(self.reprocess)
1689
base_lines = textmerge.base_from_plan()
1692
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)
1694
998
def text_merge(self, file_id, trans_id):
1695
999
"""Perform a (weave) text merge for a given file and file-id.
1696
1000
If conflicts are encountered, .THIS and .OTHER files will be emitted,
1697
1001
and a conflict will be noted.
1699
lines, base_lines = self._merged_lines(file_id)
1003
lines, conflicts = self._merged_lines(file_id)
1700
1004
lines = list(lines)
1701
# 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,
1702
1006
# because we don't want to get into weave merge guts.
1703
textfile.check_text_lines(lines)
1007
check_text_lines(lines)
1704
1008
self.tt.create_file(lines, trans_id)
1705
if base_lines is not None:
1707
1010
self._raw_conflicts.append(('text conflict', trans_id))
1708
1011
name = self.tt.final_name(trans_id)
1709
1012
parent_id = self.tt.final_parent(trans_id)
1710
file_group = self._dump_conflicts(name, parent_id, file_id,
1712
base_lines=base_lines)
1013
file_group = self._dump_conflicts(name, parent_id, file_id,
1713
1015
file_group.append(trans_id)
1716
class LCAMerger(WeaveMerger):
1718
def _generate_merge_plan(self, file_id, base):
1719
return self.this_tree.plan_file_lca_merge(file_id, self.other_tree,
1722
1018
class Diff3Merger(Merge3Merger):
1723
1019
"""Three-way merger using external diff3 for text merging"""
1725
1021
def dump_file(self, temp_dir, name, tree, file_id):
1726
out_path = osutils.pathjoin(temp_dir, name)
1022
out_path = pathjoin(temp_dir, name)
1727
1023
out_file = open(out_path, "wb")
1729
1025
in_file = tree.get_file(file_id)
1762
1058
osutils.rmtree(temp_dir)
1765
class PathNotInTree(errors.BzrError):
1767
_fmt = """Merge-into failed because %(tree)s does not contain %(path)s."""
1769
def __init__(self, path, tree):
1770
errors.BzrError.__init__(self, path=path, tree=tree)
1773
class MergeIntoMerger(Merger):
1774
"""Merger that understands other_tree will be merged into a subdir.
1776
This also changes the Merger api so that it uses real Branch, revision_id,
1777
and RevisonTree objects, rather than using revision specs.
1780
def __init__(self, this_tree, other_branch, other_tree, target_subdir,
1781
source_subpath, other_rev_id=None):
1782
"""Create a new MergeIntoMerger object.
1784
source_subpath in other_tree will be effectively copied to
1785
target_subdir in this_tree.
1787
:param this_tree: The tree that we will be merging into.
1788
:param other_branch: The Branch we will be merging from.
1789
:param other_tree: The RevisionTree object we want to merge.
1790
:param target_subdir: The relative path where we want to merge
1791
other_tree into this_tree
1792
:param source_subpath: The relative path specifying the subtree of
1793
other_tree to merge into this_tree.
1795
# It is assumed that we are merging a tree that is not in our current
1796
# ancestry, which means we are using the "EmptyTree" as our basis.
1797
null_ancestor_tree = this_tree.branch.repository.revision_tree(
1798
_mod_revision.NULL_REVISION)
1799
super(MergeIntoMerger, self).__init__(
1800
this_branch=this_tree.branch,
1801
this_tree=this_tree,
1802
other_tree=other_tree,
1803
base_tree=null_ancestor_tree,
1805
self._target_subdir = target_subdir
1806
self._source_subpath = source_subpath
1807
self.other_branch = other_branch
1808
if other_rev_id is None:
1809
other_rev_id = other_tree.get_revision_id()
1810
self.other_rev_id = self.other_basis = other_rev_id
1811
self.base_is_ancestor = True
1812
self.backup_files = True
1813
self.merge_type = Merge3Merger
1814
self.show_base = False
1815
self.reprocess = False
1816
self.interesting_ids = None
1817
self.merge_type = _MergeTypeParameterizer(MergeIntoMergeType,
1818
target_subdir=self._target_subdir,
1819
source_subpath=self._source_subpath)
1820
if self._source_subpath != '':
1821
# If this isn't a partial merge make sure the revisions will be
1823
self._maybe_fetch(self.other_branch, self.this_branch,
1826
def set_pending(self):
1827
if self._source_subpath != '':
1829
Merger.set_pending(self)
1832
class _MergeTypeParameterizer(object):
1833
"""Wrap a merge-type class to provide extra parameters.
1835
This is hack used by MergeIntoMerger to pass some extra parameters to its
1836
merge_type. Merger.do_merge() sets up its own set of parameters to pass to
1837
the 'merge_type' member. It is difficult override do_merge without
1838
re-writing the whole thing, so instead we create a wrapper which will pass
1839
the extra parameters.
1842
def __init__(self, merge_type, **kwargs):
1843
self._extra_kwargs = kwargs
1844
self._merge_type = merge_type
1846
def __call__(self, *args, **kwargs):
1847
kwargs.update(self._extra_kwargs)
1848
return self._merge_type(*args, **kwargs)
1850
def __getattr__(self, name):
1851
return getattr(self._merge_type, name)
1854
class MergeIntoMergeType(Merge3Merger):
1855
"""Merger that incorporates a tree (or part of a tree) into another."""
1857
def __init__(self, *args, **kwargs):
1858
"""Initialize the merger object.
1860
:param args: See Merge3Merger.__init__'s args.
1861
:param kwargs: See Merge3Merger.__init__'s keyword args, except for
1862
source_subpath and target_subdir.
1863
:keyword source_subpath: The relative path specifying the subtree of
1864
other_tree to merge into this_tree.
1865
:keyword target_subdir: The relative path where we want to merge
1866
other_tree into this_tree
1868
# All of the interesting work happens during Merge3Merger.__init__(),
1869
# so we have have to hack in to get our extra parameters set.
1870
self._source_subpath = kwargs.pop('source_subpath')
1871
self._target_subdir = kwargs.pop('target_subdir')
1872
super(MergeIntoMergeType, self).__init__(*args, **kwargs)
1874
def _compute_transform(self):
1875
child_pb = ui.ui_factory.nested_progress_bar()
1877
entries = self._entries_to_incorporate()
1878
entries = list(entries)
1879
for num, (entry, parent_id) in enumerate(entries):
1880
child_pb.update('Preparing file merge', num, len(entries))
1881
parent_trans_id = self.tt.trans_id_file_id(parent_id)
1882
trans_id = transform.new_by_entry(self.tt, entry,
1883
parent_trans_id, self.other_tree)
1886
self._finish_computing_transform()
1888
def _entries_to_incorporate(self):
1889
"""Yields pairs of (inventory_entry, new_parent)."""
1890
other_inv = self.other_tree.inventory
1891
subdir_id = other_inv.path2id(self._source_subpath)
1892
if subdir_id is None:
1893
# XXX: The error would be clearer if it gave the URL of the source
1894
# branch, but we don't have a reference to that here.
1895
raise PathNotInTree(self._source_subpath, "Source tree")
1896
subdir = other_inv[subdir_id]
1897
parent_in_target = osutils.dirname(self._target_subdir)
1898
target_id = self.this_tree.inventory.path2id(parent_in_target)
1899
if target_id is None:
1900
raise PathNotInTree(self._target_subdir, "Target tree")
1901
name_in_target = osutils.basename(self._target_subdir)
1902
merge_into_root = subdir.copy()
1903
merge_into_root.name = name_in_target
1904
if merge_into_root.file_id in self.this_tree.inventory:
1905
# Give the root a new file-id.
1906
# This can happen fairly easily if the directory we are
1907
# incorporating is the root, and both trees have 'TREE_ROOT' as
1908
# their root_id. Users will expect this to Just Work, so we
1909
# change the file-id here.
1910
# Non-root file-ids could potentially conflict too. That's really
1911
# an edge case, so we don't do anything special for those. We let
1912
# them cause conflicts.
1913
merge_into_root.file_id = generate_ids.gen_file_id(name_in_target)
1914
yield (merge_into_root, target_id)
1915
if subdir.kind != 'directory':
1916
# No children, so we are done.
1918
for ignored_path, entry in other_inv.iter_entries_by_dir(subdir_id):
1919
parent_id = entry.parent_id
1920
if parent_id == subdir.file_id:
1921
# The root's parent ID has changed, so make sure children of
1922
# the root refer to the new ID.
1923
parent_id = merge_into_root.file_id
1924
yield (entry, parent_id)
1927
1061
def merge_inner(this_branch, other_tree, base_tree, ignore_zero=False,
1928
1062
backup_files=False,
1929
1063
merge_type=Merge3Merger,
2003
1130
yield status_a(revision, text)
2004
1131
for revision, text in annotated_b[b_cur:bi]:
2005
1132
yield status_b(revision, text)
2006
1134
# and now the matched section
2009
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
2010
1139
yield "unchanged", text_a
2013
class _PlanMergeBase(object):
2015
def __init__(self, a_rev, b_rev, vf, key_prefix):
2018
:param a_rev: Revision-id of one revision to merge
2019
:param b_rev: Revision-id of the other revision to merge
2020
:param vf: A VersionedFiles containing both revisions
2021
:param key_prefix: A prefix for accessing keys in vf, typically
2027
self._last_lines = None
2028
self._last_lines_revision_id = None
2029
self._cached_matching_blocks = {}
2030
self._key_prefix = key_prefix
2031
self._precache_tip_lines()
2033
def _precache_tip_lines(self):
2034
lines = self.get_lines([self.a_rev, self.b_rev])
2035
self.lines_a = lines[self.a_rev]
2036
self.lines_b = lines[self.b_rev]
2038
def get_lines(self, revisions):
2039
"""Get lines for revisions from the backing VersionedFiles.
2041
:raises RevisionNotPresent: on absent texts.
2043
keys = [(self._key_prefix + (rev,)) for rev in revisions]
2045
for record in self.vf.get_record_stream(keys, 'unordered', True):
2046
if record.storage_kind == 'absent':
2047
raise errors.RevisionNotPresent(record.key, self.vf)
2048
result[record.key[-1]] = osutils.chunks_to_lines(
2049
record.get_bytes_as('chunked'))
2052
def plan_merge(self):
2053
"""Generate a 'plan' for merging the two revisions.
2055
This involves comparing their texts and determining the cause of
2056
differences. If text A has a line and text B does not, then either the
2057
line was added to text A, or it was deleted from B. Once the causes
2058
are combined, they are written out in the format described in
2059
VersionedFile.plan_merge
2061
blocks = self._get_matching_blocks(self.a_rev, self.b_rev)
2062
unique_a, unique_b = self._unique_lines(blocks)
2063
new_a, killed_b = self._determine_status(self.a_rev, unique_a)
2064
new_b, killed_a = self._determine_status(self.b_rev, unique_b)
2065
return self._iter_plan(blocks, new_a, killed_b, new_b, killed_a)
2067
def _iter_plan(self, blocks, new_a, killed_b, new_b, killed_a):
2070
for i, j, n in blocks:
2071
for a_index in range(last_i, i):
2072
if a_index in new_a:
2073
if a_index in killed_b:
2074
yield 'conflicted-a', self.lines_a[a_index]
2076
yield 'new-a', self.lines_a[a_index]
2078
yield 'killed-b', self.lines_a[a_index]
2079
for b_index in range(last_j, j):
2080
if b_index in new_b:
2081
if b_index in killed_a:
2082
yield 'conflicted-b', self.lines_b[b_index]
2084
yield 'new-b', self.lines_b[b_index]
2086
yield 'killed-a', self.lines_b[b_index]
2087
# handle common lines
2088
for a_index in range(i, i+n):
2089
yield 'unchanged', self.lines_a[a_index]
2093
def _get_matching_blocks(self, left_revision, right_revision):
2094
"""Return a description of which sections of two revisions match.
2096
See SequenceMatcher.get_matching_blocks
2098
cached = self._cached_matching_blocks.get((left_revision,
2100
if cached is not None:
2102
if self._last_lines_revision_id == left_revision:
2103
left_lines = self._last_lines
2104
right_lines = self.get_lines([right_revision])[right_revision]
2106
lines = self.get_lines([left_revision, right_revision])
2107
left_lines = lines[left_revision]
2108
right_lines = lines[right_revision]
2109
self._last_lines = right_lines
2110
self._last_lines_revision_id = right_revision
2111
matcher = patiencediff.PatienceSequenceMatcher(None, left_lines,
2113
return matcher.get_matching_blocks()
2115
def _unique_lines(self, matching_blocks):
2116
"""Analyse matching_blocks to determine which lines are unique
2118
:return: a tuple of (unique_left, unique_right), where the values are
2119
sets of line numbers of unique lines.
2125
for i, j, n in matching_blocks:
2126
unique_left.extend(range(last_i, i))
2127
unique_right.extend(range(last_j, j))
2130
return unique_left, unique_right
2133
def _subtract_plans(old_plan, new_plan):
2134
"""Remove changes from new_plan that came from old_plan.
2136
It is assumed that the difference between the old_plan and new_plan
2137
is their choice of 'b' text.
2139
All lines from new_plan that differ from old_plan are emitted
2140
verbatim. All lines from new_plan that match old_plan but are
2141
not about the 'b' revision are emitted verbatim.
2143
Lines that match and are about the 'b' revision are the lines we
2144
don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
2145
is skipped entirely.
2147
matcher = patiencediff.PatienceSequenceMatcher(None, old_plan,
2150
for i, j, n in matcher.get_matching_blocks():
2151
for jj in range(last_j, j):
2153
for jj in range(j, j+n):
2154
plan_line = new_plan[jj]
2155
if plan_line[0] == 'new-b':
2157
elif plan_line[0] == 'killed-b':
2158
yield 'unchanged', plan_line[1]
2164
class _PlanMerge(_PlanMergeBase):
2165
"""Plan an annotate merge using on-the-fly annotation"""
2167
def __init__(self, a_rev, b_rev, vf, key_prefix):
2168
super(_PlanMerge, self).__init__(a_rev, b_rev, vf, key_prefix)
2169
self.a_key = self._key_prefix + (self.a_rev,)
2170
self.b_key = self._key_prefix + (self.b_rev,)
2171
self.graph = _mod_graph.Graph(self.vf)
2172
heads = self.graph.heads((self.a_key, self.b_key))
2174
# one side dominates, so we can just return its values, yay for
2176
# Ideally we would know that before we get this far
2177
self._head_key = heads.pop()
2178
if self._head_key == self.a_key:
2182
trace.mutter('found dominating revision for %s\n%s > %s', self.vf,
2183
self._head_key[-1], other)
2186
self._head_key = None
2189
def _precache_tip_lines(self):
2190
# Turn this into a no-op, because we will do this later
2193
def _find_recursive_lcas(self):
2194
"""Find all the ancestors back to a unique lca"""
2195
cur_ancestors = (self.a_key, self.b_key)
2196
# graph.find_lca(uncommon, keys) now returns plain NULL_REVISION,
2197
# rather than a key tuple. We will just map that directly to no common
2201
next_lcas = self.graph.find_lca(*cur_ancestors)
2202
# Map a plain NULL_REVISION to a simple no-ancestors
2203
if next_lcas == set([_mod_revision.NULL_REVISION]):
2205
# Order the lca's based on when they were merged into the tip
2206
# While the actual merge portion of weave merge uses a set() of
2207
# active revisions, the order of insertion *does* effect the
2208
# implicit ordering of the texts.
2209
for rev_key in cur_ancestors:
2210
ordered_parents = tuple(self.graph.find_merge_order(rev_key,
2212
parent_map[rev_key] = ordered_parents
2213
if len(next_lcas) == 0:
2215
elif len(next_lcas) == 1:
2216
parent_map[list(next_lcas)[0]] = ()
2218
elif len(next_lcas) > 2:
2219
# More than 2 lca's, fall back to grabbing all nodes between
2220
# this and the unique lca.
2221
trace.mutter('More than 2 LCAs, falling back to all nodes for:'
2223
self.a_key, self.b_key, cur_ancestors)
2224
cur_lcas = next_lcas
2225
while len(cur_lcas) > 1:
2226
cur_lcas = self.graph.find_lca(*cur_lcas)
2227
if len(cur_lcas) == 0:
2228
# No common base to find, use the full ancestry
2231
unique_lca = list(cur_lcas)[0]
2232
if unique_lca == _mod_revision.NULL_REVISION:
2233
# find_lca will return a plain 'NULL_REVISION' rather
2234
# than a key tuple when there is no common ancestor, we
2235
# prefer to just use None, because it doesn't confuse
2236
# _get_interesting_texts()
2238
parent_map.update(self._find_unique_parents(next_lcas,
2241
cur_ancestors = next_lcas
2244
def _find_unique_parents(self, tip_keys, base_key):
2245
"""Find ancestors of tip that aren't ancestors of base.
2247
:param tip_keys: Nodes that are interesting
2248
:param base_key: Cull all ancestors of this node
2249
:return: The parent map for all revisions between tip_keys and
2250
base_key. base_key will be included. References to nodes outside of
2251
the ancestor set will also be removed.
2253
# TODO: this would be simpler if find_unique_ancestors took a list
2254
# instead of a single tip, internally it supports it, but it
2255
# isn't a "backwards compatible" api change.
2256
if base_key is None:
2257
parent_map = dict(self.graph.iter_ancestry(tip_keys))
2258
# We remove NULL_REVISION because it isn't a proper tuple key, and
2259
# thus confuses things like _get_interesting_texts, and our logic
2260
# to add the texts into the memory weave.
2261
if _mod_revision.NULL_REVISION in parent_map:
2262
parent_map.pop(_mod_revision.NULL_REVISION)
2265
for tip in tip_keys:
2267
self.graph.find_unique_ancestors(tip, [base_key]))
2268
parent_map = self.graph.get_parent_map(interesting)
2269
parent_map[base_key] = ()
2270
culled_parent_map, child_map, tails = self._remove_external_references(
2272
# Remove all the tails but base_key
2273
if base_key is not None:
2274
tails.remove(base_key)
2275
self._prune_tails(culled_parent_map, child_map, tails)
2276
# Now remove all the uninteresting 'linear' regions
2277
simple_map = _mod_graph.collapse_linear_regions(culled_parent_map)
2281
def _remove_external_references(parent_map):
2282
"""Remove references that go outside of the parent map.
2284
:param parent_map: Something returned from Graph.get_parent_map(keys)
2285
:return: (filtered_parent_map, child_map, tails)
2286
filtered_parent_map is parent_map without external references
2287
child_map is the {parent_key: [child_keys]} mapping
2288
tails is a list of nodes that do not have any parents in the map
2290
# TODO: The basic effect of this function seems more generic than
2291
# _PlanMerge. But the specific details of building a child_map,
2292
# and computing tails seems very specific to _PlanMerge.
2293
# Still, should this be in Graph land?
2294
filtered_parent_map = {}
2297
for key, parent_keys in parent_map.iteritems():
2298
culled_parent_keys = [p for p in parent_keys if p in parent_map]
2299
if not culled_parent_keys:
2301
for parent_key in culled_parent_keys:
2302
child_map.setdefault(parent_key, []).append(key)
2303
# TODO: Do we want to do this, it adds overhead for every node,
2304
# just to say that the node has no children
2305
child_map.setdefault(key, [])
2306
filtered_parent_map[key] = culled_parent_keys
2307
return filtered_parent_map, child_map, tails
2310
def _prune_tails(parent_map, child_map, tails_to_remove):
2311
"""Remove tails from the parent map.
2313
This will remove the supplied revisions until no more children have 0
2316
:param parent_map: A dict of {child: [parents]}, this dictionary will
2317
be modified in place.
2318
:param tails_to_remove: A list of tips that should be removed,
2319
this list will be consumed
2320
:param child_map: The reverse dict of parent_map ({parent: [children]})
2321
this dict will be modified
2322
:return: None, parent_map will be modified in place.
2324
while tails_to_remove:
2325
next = tails_to_remove.pop()
2326
parent_map.pop(next)
2327
children = child_map.pop(next)
2328
for child in children:
2329
child_parents = parent_map[child]
2330
child_parents.remove(next)
2331
if len(child_parents) == 0:
2332
tails_to_remove.append(child)
2334
def _get_interesting_texts(self, parent_map):
2335
"""Return a dict of texts we are interested in.
2337
Note that the input is in key tuples, but the output is in plain
2340
:param parent_map: The output from _find_recursive_lcas
2341
:return: A dict of {'revision_id':lines} as returned by
2342
_PlanMergeBase.get_lines()
2344
all_revision_keys = set(parent_map)
2345
all_revision_keys.add(self.a_key)
2346
all_revision_keys.add(self.b_key)
2348
# Everything else is in 'keys' but get_lines is in 'revision_ids'
2349
all_texts = self.get_lines([k[-1] for k in all_revision_keys])
2352
def _build_weave(self):
2353
from bzrlib import weave
2354
self._weave = weave.Weave(weave_name='in_memory_weave',
2355
allow_reserved=True)
2356
parent_map = self._find_recursive_lcas()
2358
all_texts = self._get_interesting_texts(parent_map)
2360
# Note: Unfortunately, the order given by topo_sort will effect the
2361
# ordering resolution in the output. Specifically, if you add A then B,
2362
# then in the output text A lines will show up before B lines. And, of
2363
# course, topo_sort doesn't guarantee any real ordering.
2364
# So we use merge_sort, and add a fake node on the tip.
2365
# This ensures that left-hand parents will always be inserted into the
2366
# weave before right-hand parents.
2367
tip_key = self._key_prefix + (_mod_revision.CURRENT_REVISION,)
2368
parent_map[tip_key] = (self.a_key, self.b_key)
2370
for seq_num, key, depth, eom in reversed(tsort.merge_sort(parent_map,
2374
# for key in tsort.topo_sort(parent_map):
2375
parent_keys = parent_map[key]
2376
revision_id = key[-1]
2377
parent_ids = [k[-1] for k in parent_keys]
2378
self._weave.add_lines(revision_id, parent_ids,
2379
all_texts[revision_id])
2381
def plan_merge(self):
2382
"""Generate a 'plan' for merging the two revisions.
2384
This involves comparing their texts and determining the cause of
2385
differences. If text A has a line and text B does not, then either the
2386
line was added to text A, or it was deleted from B. Once the causes
2387
are combined, they are written out in the format described in
2388
VersionedFile.plan_merge
2390
if self._head_key is not None: # There was a single head
2391
if self._head_key == self.a_key:
2394
if self._head_key != self.b_key:
2395
raise AssertionError('There was an invalid head: %s != %s'
2396
% (self.b_key, self._head_key))
2398
head_rev = self._head_key[-1]
2399
lines = self.get_lines([head_rev])[head_rev]
2400
return ((plan, line) for line in lines)
2401
return self._weave.plan_merge(self.a_rev, self.b_rev)
2404
class _PlanLCAMerge(_PlanMergeBase):
2406
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.
2453
:param revision_id: The id of the revision in which the lines are
2455
:param unique_line_numbers: The line numbers of unique lines.
2456
:return a tuple of (new_this, killed_other):
2460
unique_line_numbers = set(unique_line_numbers)
2461
for lca in self.lcas:
2462
blocks = self._get_matching_blocks(revision_id, lca)
2463
unique_vs_lca, _ignored = self._unique_lines(blocks)
2464
new.update(unique_line_numbers.intersection(unique_vs_lca))
2465
killed.update(unique_line_numbers.difference(unique_vs_lca))