~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/commit.py

  • Committer: Robert Collins
  • Date: 2007-08-02 03:17:46 UTC
  • mto: (2592.3.65 repository)
  • mto: This revision was merged to the branch mainline in revision 2667.
  • Revision ID: robertc@robertcollins.net-20070802031746-mpnoaxym829719w6
* ``bzrlib.pack.make_readv_reader`` allows readv based access to pack
  files that are stored on a transport. (Robert Collins)

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2005, 2006, 2007, 2008 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
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
16
16
 
17
17
 
18
18
# The newly committed revision is going to have a shape corresponding
19
 
# to that of the working tree.  Files that are not in the
 
19
# to that of the working inventory.  Files that are not in the
20
20
# working tree and that were in the predecessor are reported as
21
21
# removed --- this can include files that were either removed from the
22
22
# inventory or deleted in the working tree.  If they were only
25
25
# We then consider the remaining entries, which will be in the new
26
26
# version.  Directory entries are simply copied across.  File entries
27
27
# must be checked to see if a new version of the file should be
28
 
# recorded.  For each parent revision tree, we check to see what
 
28
# recorded.  For each parent revision inventory, we check to see what
29
29
# version of the file was present.  If the file was present in at
30
30
# least one tree, and if it was the same version in all the trees,
31
31
# then we can just refer to that version.  Otherwise, a new version
59
59
from bzrlib import (
60
60
    debug,
61
61
    errors,
62
 
    revision,
63
 
    trace,
 
62
    inventory,
64
63
    tree,
65
 
    xml_serializer,
66
64
    )
67
65
from bzrlib.branch import Branch
68
66
import bzrlib.config
70
68
                           ConflictsInTree,
71
69
                           StrictCommitFailed
72
70
                           )
73
 
from bzrlib.osutils import (get_user_encoding,
74
 
                            kind_marker, isdir,isfile, is_inside_any,
 
71
from bzrlib.osutils import (kind_marker, isdir,isfile, is_inside_any, 
75
72
                            is_inside_or_parent_of_any,
76
 
                            minimum_path_selection,
77
 
                            quotefn, sha_file, split_lines,
78
 
                            splitpath,
79
 
                            )
 
73
                            quotefn, sha_file, split_lines)
80
74
from bzrlib.testament import Testament
81
 
from bzrlib.trace import mutter, note, warning, is_quiet
82
 
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
 
75
from bzrlib.trace import mutter, note, warning
 
76
from bzrlib.xml5 import serializer_v5
 
77
from bzrlib.inventory import Inventory, InventoryEntry
83
78
from bzrlib import symbol_versioning
84
79
from bzrlib.symbol_versioning import (deprecated_passed,
85
80
        deprecated_function,
86
81
        DEPRECATED_PARAMETER)
87
82
from bzrlib.workingtree import WorkingTree
88
 
from bzrlib.urlutils import unescape_for_display
89
83
import bzrlib.ui
90
84
 
91
85
 
92
86
class NullCommitReporter(object):
93
87
    """I report on progress of a commit."""
94
88
 
95
 
    def started(self, revno, revid, location=None):
96
 
        if location is None:
97
 
            symbol_versioning.warn("As of bzr 1.0 you must pass a location "
98
 
                                   "to started.", DeprecationWarning,
99
 
                                   stacklevel=2)
100
 
        pass
101
 
 
102
89
    def snapshot_change(self, change, path):
103
90
        pass
104
91
 
105
92
    def completed(self, revno, rev_id):
106
93
        pass
107
94
 
108
 
    def deleted(self, path):
 
95
    def deleted(self, file_id):
109
96
        pass
110
97
 
111
98
    def escaped(self, escape_count, message):
117
104
    def renamed(self, change, old_path, new_path):
118
105
        pass
119
106
 
120
 
    def is_verbose(self):
121
 
        return False
122
 
 
123
107
 
124
108
class ReportCommitToLog(NullCommitReporter):
125
109
 
131
115
        note(format, *args)
132
116
 
133
117
    def snapshot_change(self, change, path):
134
 
        if path == '' and change in ('added', 'modified'):
 
118
        if change == 'unchanged':
 
119
            return
 
120
        if change == 'added' and path == '':
135
121
            return
136
122
        self._note("%s %s", change, path)
137
123
 
138
 
    def started(self, revno, rev_id, location=None):
139
 
        if location is not None:
140
 
            location = ' to: ' + unescape_for_display(location, 'utf-8')
141
 
        else:
142
 
            # When started was added, location was only made optional by
143
 
            # accident.  Matt Nordhoff 20071129
144
 
            symbol_versioning.warn("As of bzr 1.0 you must pass a location "
145
 
                                   "to started.", DeprecationWarning,
146
 
                                   stacklevel=2)
147
 
            location = ''
148
 
        self._note('Committing%s', location)
149
 
 
150
124
    def completed(self, revno, rev_id):
151
125
        self._note('Committed revision %d.', revno)
152
 
 
153
 
    def deleted(self, path):
154
 
        self._note('deleted %s', path)
 
126
    
 
127
    def deleted(self, file_id):
 
128
        self._note('deleted %s', file_id)
155
129
 
156
130
    def escaped(self, escape_count, message):
157
131
        self._note("replaced %d control characters in message", escape_count)
162
136
    def renamed(self, change, old_path, new_path):
163
137
        self._note('%s %s => %s', change, old_path, new_path)
164
138
 
165
 
    def is_verbose(self):
166
 
        return True
167
 
 
168
139
 
169
140
class Commit(object):
170
141
    """Task of committing a new revision.
181
152
    def __init__(self,
182
153
                 reporter=None,
183
154
                 config=None):
184
 
        """Create a Commit object.
185
 
 
186
 
        :param reporter: the default reporter to use or None to decide later
187
 
        """
188
 
        self.reporter = reporter
 
155
        if reporter is not None:
 
156
            self.reporter = reporter
 
157
        else:
 
158
            self.reporter = NullCommitReporter()
189
159
        self.config = config
190
 
 
 
160
        
191
161
    def commit(self,
192
162
               message=None,
193
163
               timestamp=None,
204
174
               reporter=None,
205
175
               config=None,
206
176
               message_callback=None,
207
 
               recursive='down',
208
 
               exclude=None,
209
 
               possible_master_transports=None):
 
177
               recursive='down'):
210
178
        """Commit working copy as a new revision.
211
179
 
212
180
        :param message: the commit message (it or message_callback is required)
230
198
 
231
199
        :param revprops: Properties for new revision
232
200
        :param local: Perform a local only commit.
233
 
        :param reporter: the reporter to use or None for the default
234
 
        :param verbose: if True and the reporter is not None, report everything
235
201
        :param recursive: If set to 'down', commit in any subtrees that have
236
202
            pending changes of any sort during this commit.
237
 
        :param exclude: None or a list of relative paths to exclude from the
238
 
            commit. Pending changes to excluded files will be ignored by the
239
 
            commit.
240
203
        """
241
204
        mutter('preparing to commit')
242
205
 
251
214
        if message_callback is None:
252
215
            if message is not None:
253
216
                if isinstance(message, str):
254
 
                    message = message.decode(get_user_encoding())
 
217
                    message = message.decode(bzrlib.user_encoding)
255
218
                message_callback = lambda x: message
256
219
            else:
257
220
                raise BzrError("The message or message_callback keyword"
258
221
                               " parameter is required for commit().")
259
222
 
260
223
        self.bound_branch = None
261
 
        self.any_entries_deleted = False
262
 
        if exclude is not None:
263
 
            self.exclude = sorted(
264
 
                minimum_path_selection(exclude))
265
 
        else:
266
 
            self.exclude = []
267
224
        self.local = local
268
225
        self.master_branch = None
269
226
        self.master_locked = False
270
 
        self.recursive = recursive
271
227
        self.rev_id = None
272
 
        if specific_files is not None:
273
 
            self.specific_files = sorted(
274
 
                minimum_path_selection(specific_files))
275
 
        else:
276
 
            self.specific_files = None
277
 
            
 
228
        self.specific_files = specific_files
278
229
        self.allow_pointless = allow_pointless
 
230
        self.recursive = recursive
279
231
        self.revprops = revprops
280
232
        self.message_callback = message_callback
281
233
        self.timestamp = timestamp
284
236
        self.strict = strict
285
237
        self.verbose = verbose
286
238
 
 
239
        if reporter is None and self.reporter is None:
 
240
            self.reporter = NullCommitReporter()
 
241
        elif reporter is not None:
 
242
            self.reporter = reporter
 
243
 
287
244
        self.work_tree.lock_write()
288
 
        self.parents = self.work_tree.get_parent_ids()
289
 
        # We can use record_iter_changes IFF iter_changes is compatible with
290
 
        # the command line parameters, and the repository has fast delta
291
 
        # generation. See bug 347649.
292
 
        self.use_record_iter_changes = (
293
 
            not self.specific_files and
294
 
            not self.exclude and 
295
 
            not self.branch.repository._format.supports_tree_reference and
296
 
            (self.branch.repository._format.fast_deltas or
297
 
             len(self.parents) < 2))
298
245
        self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
299
 
        self.basis_revid = self.work_tree.last_revision()
300
246
        self.basis_tree = self.work_tree.basis_tree()
301
247
        self.basis_tree.lock_read()
302
248
        try:
305
251
                raise ConflictsInTree
306
252
 
307
253
            # Setup the bound branch variables as needed.
308
 
            self._check_bound_branch(possible_master_transports)
 
254
            self._check_bound_branch()
309
255
 
310
256
            # Check that the working tree is up to date
311
 
            old_revno, new_revno = self._check_out_of_date_tree()
 
257
            old_revno,new_revno = self._check_out_of_date_tree()
312
258
 
313
 
            # Complete configuration setup
314
 
            if reporter is not None:
315
 
                self.reporter = reporter
316
 
            elif self.reporter is None:
317
 
                self.reporter = self._select_reporter()
318
259
            if self.config is None:
319
260
                self.config = self.branch.get_config()
320
261
 
321
 
            self._set_specific_file_ids()
 
262
            # If provided, ensure the specified files are versioned
 
263
            if specific_files is not None:
 
264
                # Note: We don't actually need the IDs here. This routine
 
265
                # is being called because it raises PathNotVerisonedError
 
266
                # as a side effect of finding the IDs.
 
267
                # XXX: Dont we have filter_unversioned to do this more
 
268
                # cheaply?
 
269
                tree.find_ids_across_trees(specific_files,
 
270
                                           [self.basis_tree, self.work_tree])
322
271
 
323
272
            # Setup the progress bar. As the number of files that need to be
324
273
            # committed in unknown, progress is reported as stages.
326
275
            # information in the progress bar during the relevant stages.
327
276
            self.pb_stage_name = ""
328
277
            self.pb_stage_count = 0
329
 
            self.pb_stage_total = 5
 
278
            self.pb_stage_total = 4
330
279
            if self.bound_branch:
331
280
                self.pb_stage_total += 1
332
281
            self.pb.show_pct = False
335
284
            self.pb.show_count = True
336
285
            self.pb.show_bar = True
337
286
 
338
 
            self._gather_parents()
339
287
            # After a merge, a selected file commit is not supported.
340
288
            # See 'bzr help merge' for an explanation as to why.
 
289
            self.basis_inv = self.basis_tree.inventory
 
290
            self._gather_parents()
341
291
            if len(self.parents) > 1 and self.specific_files:
342
292
                raise errors.CannotCommitSelectedFileMerge(self.specific_files)
343
 
            # Excludes are a form of selected file commit.
344
 
            if len(self.parents) > 1 and self.exclude:
345
 
                raise errors.CannotCommitSelectedFileMerge(self.exclude)
346
 
 
 
293
            
347
294
            # Collect the changes
348
 
            self._set_progress_stage("Collecting changes", counter=True)
 
295
            self._set_progress_stage("Collecting changes",
 
296
                    entries_title="Directory")
349
297
            self.builder = self.branch.get_commit_builder(self.parents,
350
298
                self.config, timestamp, timezone, committer, revprops, rev_id)
351
 
 
352
 
            try:
353
 
                self.builder.will_record_deletes()
354
 
                # find the location being committed to
355
 
                if self.bound_branch:
356
 
                    master_location = self.master_branch.base
357
 
                else:
358
 
                    master_location = self.branch.base
359
 
 
360
 
                # report the start of the commit
361
 
                self.reporter.started(new_revno, self.rev_id, master_location)
362
 
 
363
 
                self._update_builder_with_changes()
364
 
                self._check_pointless()
365
 
 
366
 
                # TODO: Now the new inventory is known, check for conflicts.
367
 
                # ADHB 2006-08-08: If this is done, populate_new_inv should not add
368
 
                # weave lines, because nothing should be recorded until it is known
369
 
                # that commit will succeed.
370
 
                self._set_progress_stage("Saving data locally")
371
 
                self.builder.finish_inventory()
372
 
 
373
 
                # Prompt the user for a commit message if none provided
374
 
                message = message_callback(self)
375
 
                self.message = message
376
 
                self._escape_commit_message()
377
 
 
378
 
                # Add revision data to the local branch
379
 
                self.rev_id = self.builder.commit(self.message)
380
 
 
381
 
            except Exception, e:
382
 
                mutter("aborting commit write group because of exception:")
383
 
                trace.log_exception_quietly()
384
 
                note("aborting commit write group: %r" % (e,))
385
 
                self.builder.abort()
386
 
                raise
387
 
 
388
 
            self._process_pre_hooks(old_revno, new_revno)
389
 
 
 
299
            self._update_builder_with_changes()
 
300
            self._check_pointless()
 
301
 
 
302
            # TODO: Now the new inventory is known, check for conflicts.
 
303
            # ADHB 2006-08-08: If this is done, populate_new_inv should not add
 
304
            # weave lines, because nothing should be recorded until it is known
 
305
            # that commit will succeed.
 
306
            self._set_progress_stage("Saving data locally")
 
307
            self.builder.finish_inventory()
 
308
 
 
309
            # Prompt the user for a commit message if none provided
 
310
            message = message_callback(self)
 
311
            assert isinstance(message, unicode), type(message)
 
312
            self.message = message
 
313
            self._escape_commit_message()
 
314
 
 
315
            # Add revision data to the local branch
 
316
            self.rev_id = self.builder.commit(self.message)
 
317
            
390
318
            # Upload revision data to the master.
391
319
            # this will propagate merged revisions too if needed.
392
320
            if self.bound_branch:
393
321
                self._set_progress_stage("Uploading data to master branch")
 
322
                self.master_branch.repository.fetch(self.branch.repository,
 
323
                                                    revision_id=self.rev_id)
 
324
                # now the master has the revision data
394
325
                # 'commit' to the master first so a timeout here causes the
395
326
                # local branch to be out of date
396
 
                self.master_branch.import_last_revision_info(
397
 
                    self.branch.repository, new_revno, self.rev_id)
 
327
                self.master_branch.set_last_revision_info(new_revno,
 
328
                                                          self.rev_id)
398
329
 
399
330
            # and now do the commit locally.
400
331
            self.branch.set_last_revision_info(new_revno, self.rev_id)
401
332
 
402
333
            # Make the working tree up to date with the branch
403
334
            self._set_progress_stage("Updating the working tree")
404
 
            self.work_tree.update_basis_by_delta(self.rev_id,
405
 
                 self.builder.get_basis_delta())
 
335
            rev_tree = self.builder.revision_tree()
 
336
            self.work_tree.set_parent_trees([(self.rev_id, rev_tree)])
406
337
            self.reporter.completed(new_revno, self.rev_id)
407
 
            self._process_post_hooks(old_revno, new_revno)
 
338
            self._process_hooks(old_revno, new_revno)
408
339
        finally:
409
340
            self._cleanup()
410
341
        return self.rev_id
411
342
 
412
 
    def _select_reporter(self):
413
 
        """Select the CommitReporter to use."""
414
 
        if is_quiet():
415
 
            return NullCommitReporter()
416
 
        return ReportCommitToLog()
 
343
    def _any_real_changes(self):
 
344
        """Are there real changes between new_inventory and basis?
 
345
 
 
346
        For trees without rich roots, inv.root.revision changes every commit.
 
347
        But if that is the only change, we want to treat it as though there
 
348
        are *no* changes.
 
349
        """
 
350
        new_entries = self.builder.new_inventory.iter_entries()
 
351
        basis_entries = self.basis_inv.iter_entries()
 
352
        new_path, new_root_ie = new_entries.next()
 
353
        basis_path, basis_root_ie = basis_entries.next()
 
354
 
 
355
        # This is a copy of InventoryEntry.__eq__ only leaving out .revision
 
356
        def ie_equal_no_revision(this, other):
 
357
            return ((this.file_id == other.file_id)
 
358
                    and (this.name == other.name)
 
359
                    and (this.symlink_target == other.symlink_target)
 
360
                    and (this.text_sha1 == other.text_sha1)
 
361
                    and (this.text_size == other.text_size)
 
362
                    and (this.text_id == other.text_id)
 
363
                    and (this.parent_id == other.parent_id)
 
364
                    and (this.kind == other.kind)
 
365
                    and (this.executable == other.executable)
 
366
                    and (this.reference_revision == other.reference_revision)
 
367
                    )
 
368
        if not ie_equal_no_revision(new_root_ie, basis_root_ie):
 
369
            return True
 
370
 
 
371
        for new_ie, basis_ie in zip(new_entries, basis_entries):
 
372
            if new_ie != basis_ie:
 
373
                return True
 
374
 
 
375
        # No actual changes present
 
376
        return False
417
377
 
418
378
    def _check_pointless(self):
419
379
        if self.allow_pointless:
421
381
        # A merge with no effect on files
422
382
        if len(self.parents) > 1:
423
383
            return
424
 
        # TODO: we could simplify this by using self.builder.basis_delta.
425
 
 
426
 
        # The initial commit adds a root directory, but this in itself is not
427
 
        # a worthwhile commit.
428
 
        if (self.basis_revid == revision.NULL_REVISION and
429
 
            ((self.builder.new_inventory is not None and
430
 
             len(self.builder.new_inventory) == 1) or
431
 
            len(self.builder._basis_delta) == 1)):
 
384
        # work around the fact that a newly-initted tree does differ from its
 
385
        # basis
 
386
        if len(self.basis_inv) == 0 and len(self.builder.new_inventory) == 1:
432
387
            raise PointlessCommit()
433
 
        if self.builder.any_changes():
 
388
        # Shortcut, if the number of entries changes, then we obviously have
 
389
        # a change
 
390
        if len(self.builder.new_inventory) != len(self.basis_inv):
 
391
            return
 
392
        # If length == 1, then we only have the root entry. Which means
 
393
        # that there is no real difference (only the root could be different)
 
394
        if (len(self.builder.new_inventory) != 1 and self._any_real_changes()):
434
395
            return
435
396
        raise PointlessCommit()
436
397
 
437
 
    def _check_bound_branch(self, possible_master_transports=None):
 
398
    def _check_bound_branch(self):
438
399
        """Check to see if the local branch is bound.
439
400
 
440
401
        If it is bound, then most of the commit will actually be
445
406
            raise errors.LocalRequiresBoundBranch()
446
407
 
447
408
        if not self.local:
448
 
            self.master_branch = self.branch.get_master_branch(
449
 
                possible_master_transports)
 
409
            self.master_branch = self.branch.get_master_branch()
450
410
 
451
411
        if not self.master_branch:
452
412
            # make this branch the reference branch for out of date checks.
463
423
        #       commits to the remote branch if they would fit.
464
424
        #       But for now, just require remote to be identical
465
425
        #       to local.
466
 
 
 
426
        
467
427
        # Make sure the local branch is identical to the master
468
428
        master_info = self.master_branch.last_revision_info()
469
429
        local_info = self.branch.last_revision_info()
501
461
            new_revno = 1
502
462
        return old_revno,new_revno
503
463
 
504
 
    def _process_pre_hooks(self, old_revno, new_revno):
505
 
        """Process any registered pre commit hooks."""
506
 
        self._set_progress_stage("Running pre_commit hooks")
507
 
        self._process_hooks("pre_commit", old_revno, new_revno)
508
 
 
509
 
    def _process_post_hooks(self, old_revno, new_revno):
510
 
        """Process any registered post commit hooks."""
 
464
    def _process_hooks(self, old_revno, new_revno):
 
465
        """Process any registered commit hooks."""
511
466
        # Process the post commit hooks, if any
512
 
        self._set_progress_stage("Running post_commit hooks")
 
467
        self._set_progress_stage("Running post commit hooks")
513
468
        # old style commit hooks - should be deprecated ? (obsoleted in
514
469
        # 0.15)
515
470
        if self.config.post_commit() is not None:
520
475
                              {'branch':self.branch,
521
476
                               'bzrlib':bzrlib,
522
477
                               'rev_id':self.rev_id})
523
 
        # process new style post commit hooks
524
 
        self._process_hooks("post_commit", old_revno, new_revno)
525
 
 
526
 
    def _process_hooks(self, hook_name, old_revno, new_revno):
527
 
        if not Branch.hooks[hook_name]:
528
 
            return
529
 
 
530
478
        # new style commit hooks:
531
479
        if not self.bound_branch:
532
480
            hook_master = self.branch
541
489
            old_revid = self.parents[0]
542
490
        else:
543
491
            old_revid = bzrlib.revision.NULL_REVISION
544
 
 
545
 
        if hook_name == "pre_commit":
546
 
            future_tree = self.builder.revision_tree()
547
 
            tree_delta = future_tree.changes_from(self.basis_tree,
548
 
                                             include_root=True)
549
 
 
550
 
        for hook in Branch.hooks[hook_name]:
 
492
        for hook in Branch.hooks['post_commit']:
551
493
            # show the running hook in the progress bar. As hooks may
552
494
            # end up doing nothing (e.g. because they are not configured by
553
495
            # the user) this is still showing progress, not showing overall
554
496
            # actions - its up to each plugin to show a UI if it want's to
555
497
            # (such as 'Emailing diff to foo@example.com').
556
 
            self.pb_stage_name = "Running %s hooks [%s]" % \
557
 
                (hook_name, Branch.hooks.get_hook_name(hook))
 
498
            self.pb_stage_name = "Running post commit hooks [%s]" % \
 
499
                Branch.hooks.get_hook_name(hook)
558
500
            self._emit_progress()
559
501
            if 'hooks' in debug.debug_flags:
560
502
                mutter("Invoking commit hook: %r", hook)
561
 
            if hook_name == "post_commit":
562
 
                hook(hook_local, hook_master, old_revno, old_revid, new_revno,
563
 
                     self.rev_id)
564
 
            elif hook_name == "pre_commit":
565
 
                hook(hook_local, hook_master,
566
 
                     old_revno, old_revid, new_revno, self.rev_id,
567
 
                     tree_delta, future_tree)
 
503
            hook(hook_local, hook_master, old_revno, old_revid, new_revno,
 
504
                self.rev_id)
568
505
 
569
506
    def _cleanup(self):
570
507
        """Cleanup any open locks, progress bars etc."""
582
519
            # typically this will be useful enough.
583
520
            except Exception, e:
584
521
                found_exception = e
585
 
        if found_exception is not None:
 
522
        if found_exception is not None: 
586
523
            # don't do a plan raise, because the last exception may have been
587
524
            # trashed, e is our sure-to-work exception even though it loses the
588
525
            # full traceback. XXX: RBC 20060421 perhaps we could check the
589
 
            # exc_info and if its the same one do a plain raise otherwise
 
526
            # exc_info and if its the same one do a plain raise otherwise 
590
527
            # 'raise e' as we do now.
591
528
            raise e
592
529
 
608
545
        # serialiser not by commit. Then we can also add an unescaper
609
546
        # in the deserializer and start roundtripping revision messages
610
547
        # precisely. See repository_implementations/test_repository.py
611
 
        self.message, escape_count = xml_serializer.escape_invalid_chars(
 
548
        
 
549
        # Python strings can include characters that can't be
 
550
        # represented in well-formed XML; escape characters that
 
551
        # aren't listed in the XML specification
 
552
        # (http://www.w3.org/TR/REC-xml/#NT-Char).
 
553
        self.message, escape_count = re.subn(
 
554
            u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
 
555
            lambda match: match.group(0).encode('unicode_escape'),
612
556
            self.message)
613
557
        if escape_count:
614
558
            self.reporter.escaped(escape_count, self.message)
615
559
 
616
560
    def _gather_parents(self):
617
561
        """Record the parents of a merge for merge detection."""
618
 
        # TODO: Make sure that this list doesn't contain duplicate
 
562
        # TODO: Make sure that this list doesn't contain duplicate 
619
563
        # entries and the order is preserved when doing this.
620
 
        if self.use_record_iter_changes:
621
 
            return
622
 
        self.basis_inv = self.basis_tree.inventory
 
564
        self.parents = self.work_tree.get_parent_ids()
623
565
        self.parent_invs = [self.basis_inv]
624
566
        for revision in self.parents[1:]:
625
567
            if self.branch.repository.has_revision(revision):
632
574
    def _update_builder_with_changes(self):
633
575
        """Update the commit builder with the data about what has changed.
634
576
        """
635
 
        exclude = self.exclude
636
 
        specific_files = self.specific_files or []
637
 
        mutter("Selecting files for commit with filter %s", specific_files)
638
 
 
639
 
        self._check_strict()
640
 
        if self.use_record_iter_changes:
641
 
            iter_changes = self.work_tree.iter_changes(self.basis_tree)
642
 
            iter_changes = self._filter_iter_changes(iter_changes)
643
 
            for file_id, path, fs_hash in self.builder.record_iter_changes(
644
 
                self.work_tree, self.basis_revid, iter_changes):
645
 
                self.work_tree._observed_sha1(file_id, path, fs_hash)
646
 
        else:
647
 
            # Build the new inventory
648
 
            self._populate_from_inventory()
649
 
            self._record_unselected()
650
 
            self._report_and_accumulate_deletes()
651
 
 
652
 
    def _filter_iter_changes(self, iter_changes):
653
 
        """Process iter_changes.
654
 
 
655
 
        This method reports on the changes in iter_changes to the user, and 
656
 
        converts 'missing' entries in the iter_changes iterator to 'deleted'
657
 
        entries. 'missing' entries have their
658
 
 
659
 
        :param iter_changes: An iter_changes to process.
660
 
        :return: A generator of changes.
661
 
        """
662
 
        reporter = self.reporter
663
 
        report_changes = reporter.is_verbose()
664
 
        deleted_ids = []
665
 
        for change in iter_changes:
666
 
            if report_changes:
667
 
                old_path = change[1][0]
668
 
                new_path = change[1][1]
669
 
                versioned = change[3][1]
670
 
            kind = change[6][1]
671
 
            versioned = change[3][1]
672
 
            if kind is None and versioned:
673
 
                # 'missing' path
674
 
                if report_changes:
675
 
                    reporter.missing(new_path)
676
 
                deleted_ids.append(change[0])
677
 
                # Reset the new path (None) and new versioned flag (False)
678
 
                change = (change[0], (change[1][0], None), change[2],
679
 
                    (change[3][0], False)) + change[4:]
680
 
            elif kind == 'tree-reference':
681
 
                if self.recursive == 'down':
682
 
                    self._commit_nested_tree(change[0], change[1][1])
683
 
            if change[3][0] or change[3][1]:
684
 
                yield change
685
 
                if report_changes:
686
 
                    if new_path is None:
687
 
                        reporter.deleted(old_path)
688
 
                    elif old_path is None:
689
 
                        reporter.snapshot_change('added', new_path)
690
 
                    elif old_path != new_path:
691
 
                        reporter.renamed('renamed', old_path, new_path)
692
 
                    else:
693
 
                        if (new_path or 
694
 
                            self.work_tree.branch.repository._format.rich_root_data):
695
 
                            # Don't report on changes to '' in non rich root
696
 
                            # repositories.
697
 
                            reporter.snapshot_change('modified', new_path)
698
 
            self._next_progress_entry()
699
 
        # Unversion IDs that were found to be deleted
700
 
        self.work_tree.unversion(deleted_ids)
701
 
 
702
 
    def _record_unselected(self):
703
 
        # If specific files are selected, then all un-selected files must be
704
 
        # recorded in their previous state. For more details, see
705
 
        # https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
706
 
        if self.specific_files or self.exclude:
707
 
            specific_files = self.specific_files or []
708
 
            for path, old_ie in self.basis_inv.iter_entries():
709
 
                if old_ie.file_id in self.builder.new_inventory:
710
 
                    # already added - skip.
711
 
                    continue
712
 
                if (is_inside_any(specific_files, path)
713
 
                    and not is_inside_any(self.exclude, path)):
714
 
                    # was inside the selected path, and not excluded - if not
715
 
                    # present it has been deleted so skip.
716
 
                    continue
717
 
                # From here down it was either not selected, or was excluded:
718
 
                # We preserve the entry unaltered.
719
 
                ie = old_ie.copy()
720
 
                # Note: specific file commits after a merge are currently
721
 
                # prohibited. This test is for sanity/safety in case it's
722
 
                # required after that changes.
723
 
                if len(self.parents) > 1:
724
 
                    ie.revision = None
725
 
                self.builder.record_entry_contents(ie, self.parent_invs, path,
726
 
                    self.basis_tree, None)
727
 
 
728
 
    def _report_and_accumulate_deletes(self):
729
 
        if (isinstance(self.basis_inv, Inventory)
730
 
            and isinstance(self.builder.new_inventory, Inventory)):
731
 
            # the older Inventory classes provide a _byid dict, and building a
732
 
            # set from the keys of this dict is substantially faster than even
733
 
            # getting a set of ids from the inventory
734
 
            #
735
 
            # <lifeless> set(dict) is roughly the same speed as
736
 
            # set(iter(dict)) and both are significantly slower than
737
 
            # set(dict.keys())
738
 
            deleted_ids = set(self.basis_inv._byid.keys()) - \
739
 
               set(self.builder.new_inventory._byid.keys())
740
 
        else:
741
 
            deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
742
 
        if deleted_ids:
743
 
            self.any_entries_deleted = True
744
 
            deleted = [(self.basis_tree.id2path(file_id), file_id)
745
 
                for file_id in deleted_ids]
746
 
            deleted.sort()
747
 
            # XXX: this is not quite directory-order sorting
748
 
            for path, file_id in deleted:
749
 
                self.builder.record_delete(path, file_id)
750
 
                self.reporter.deleted(path)
751
 
 
752
 
    def _check_strict(self):
753
 
        # XXX: when we use iter_changes this would likely be faster if
754
 
        # iter_changes would check for us (even in the presence of
755
 
        # selected_files).
756
 
        if self.strict:
757
 
            # raise an exception as soon as we find a single unknown.
758
 
            for unknown in self.work_tree.unknowns():
759
 
                raise StrictCommitFailed()
760
 
 
761
 
    def _populate_from_inventory(self):
762
 
        """Populate the CommitBuilder by walking the working tree inventory."""
763
577
        # Build the revision inventory.
764
578
        #
765
579
        # This starts by creating a new empty inventory. Depending on
766
580
        # which files are selected for commit, and what is present in the
767
 
        # current tree, the new inventory is populated. inventory entries
 
581
        # current tree, the new inventory is populated. inventory entries 
768
582
        # which are candidates for modification have their revision set to
769
583
        # None; inventory entries that are carried over untouched have their
770
584
        # revision set to their prior value.
775
589
        # ADHB 11-07-2006
776
590
 
777
591
        specific_files = self.specific_files
778
 
        exclude = self.exclude
779
 
        report_changes = self.reporter.is_verbose()
 
592
        mutter("Selecting files for commit with filter %s", specific_files)
 
593
 
 
594
        # Check and warn about old CommitBuilders
 
595
        if not self.builder.record_root_entry:
 
596
            symbol_versioning.warn('CommitBuilders should support recording'
 
597
                ' the root entry as of bzr 0.10.', DeprecationWarning, 
 
598
                stacklevel=1)
 
599
            self.builder.new_inventory.add(self.basis_inv.root.copy())
 
600
 
 
601
        # Build the new inventory
 
602
        self._populate_from_inventory(specific_files)
 
603
 
 
604
        # If specific files are selected, then all un-selected files must be
 
605
        # recorded in their previous state. For more details, see
 
606
        # https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
 
607
        if specific_files:
 
608
            for path, new_ie in self.basis_inv.iter_entries():
 
609
                if new_ie.file_id in self.builder.new_inventory:
 
610
                    continue
 
611
                if is_inside_any(specific_files, path):
 
612
                    continue
 
613
                ie = new_ie.copy()
 
614
                ie.revision = None
 
615
                self.builder.record_entry_contents(ie, self.parent_invs, path,
 
616
                                                   self.basis_tree)
 
617
 
 
618
        # Report what was deleted. We could skip this when no deletes are
 
619
        # detected to gain a performance win, but it arguably serves as a
 
620
        # 'safety check' by informing the user whenever anything disappears.
 
621
        for path, ie in self.basis_inv.iter_entries():
 
622
            if ie.file_id not in self.builder.new_inventory:
 
623
                self.reporter.deleted(path)
 
624
 
 
625
    def _populate_from_inventory(self, specific_files):
 
626
        """Populate the CommitBuilder by walking the working tree inventory."""
 
627
        if self.strict:
 
628
            # raise an exception as soon as we find a single unknown.
 
629
            for unknown in self.work_tree.unknowns():
 
630
                raise StrictCommitFailed()
 
631
               
780
632
        deleted_ids = []
781
 
        # A tree of paths that have been deleted. E.g. if foo/bar has been
782
 
        # deleted, then we have {'foo':{'bar':{}}}
783
 
        deleted_paths = {}
784
 
        # XXX: Note that entries may have the wrong kind because the entry does
785
 
        # not reflect the status on disk.
 
633
        deleted_paths = set()
786
634
        work_inv = self.work_tree.inventory
787
 
        # NB: entries will include entries within the excluded ids/paths
788
 
        # because iter_entries_by_dir has no 'exclude' facility today.
789
 
        entries = work_inv.iter_entries_by_dir(
790
 
            specific_file_ids=self.specific_file_ids, yield_parents=True)
 
635
        assert work_inv.root is not None
 
636
        entries = work_inv.iter_entries()
 
637
        if not self.builder.record_root_entry:
 
638
            entries.next()
791
639
        for path, existing_ie in entries:
792
640
            file_id = existing_ie.file_id
793
641
            name = existing_ie.name
794
642
            parent_id = existing_ie.parent_id
795
643
            kind = existing_ie.kind
 
644
            if kind == 'directory':
 
645
                self._next_progress_entry()
 
646
 
796
647
            # Skip files that have been deleted from the working tree.
797
 
            # The deleted path ids are also recorded so they can be explicitly
798
 
            # unversioned later.
799
 
            if deleted_paths:
800
 
                path_segments = splitpath(path)
801
 
                deleted_dict = deleted_paths
802
 
                for segment in path_segments:
803
 
                    deleted_dict = deleted_dict.get(segment, None)
804
 
                    if not deleted_dict:
805
 
                        # We either took a path not present in the dict
806
 
                        # (deleted_dict was None), or we've reached an empty
807
 
                        # child dir in the dict, so are now a sub-path.
808
 
                        break
809
 
                else:
810
 
                    deleted_dict = None
811
 
                if deleted_dict is not None:
812
 
                    # the path has a deleted parent, do not add it.
813
 
                    continue
814
 
            if exclude and is_inside_any(exclude, path):
815
 
                # Skip excluded paths. Excluded paths are processed by
816
 
                # _update_builder_with_changes.
 
648
            # The deleted files/directories are also recorded so they
 
649
            # can be explicitly unversioned later. Note that when a
 
650
            # filter of specific files is given, we must only skip/record
 
651
            # deleted files matching that filter.
 
652
            if is_inside_any(deleted_paths, path):
817
653
                continue
818
 
            content_summary = self.work_tree.path_content_summary(path)
819
 
            # Note that when a filter of specific files is given, we must only
820
 
            # skip/record deleted files matching that filter.
821
654
            if not specific_files or is_inside_any(specific_files, path):
822
 
                if content_summary[0] == 'missing':
823
 
                    if not deleted_paths:
824
 
                        # path won't have been split yet.
825
 
                        path_segments = splitpath(path)
826
 
                    deleted_dict = deleted_paths
827
 
                    for segment in path_segments:
828
 
                        deleted_dict = deleted_dict.setdefault(segment, {})
 
655
                if not self.work_tree.has_filename(path):
 
656
                    deleted_paths.add(path)
829
657
                    self.reporter.missing(path)
830
 
                    self._next_progress_entry()
831
658
                    deleted_ids.append(file_id)
832
659
                    continue
833
 
            # TODO: have the builder do the nested commit just-in-time IF and
834
 
            # only if needed.
835
 
            if content_summary[0] == 'tree-reference':
836
 
                # enforce repository nested tree policy.
837
 
                if (not self.work_tree.supports_tree_reference() or
838
 
                    # repository does not support it either.
839
 
                    not self.branch.repository._format.supports_tree_reference):
840
 
                    content_summary = ('directory',) + content_summary[1:]
841
 
            kind = content_summary[0]
842
 
            # TODO: specific_files filtering before nested tree processing
843
 
            if kind == 'tree-reference':
844
 
                if self.recursive == 'down':
845
 
                    nested_revision_id = self._commit_nested_tree(
846
 
                        file_id, path)
847
 
                    content_summary = content_summary[:3] + (
848
 
                        nested_revision_id,)
849
 
                else:
850
 
                    content_summary = content_summary[:3] + (
851
 
                        self.work_tree.get_reference_revision(file_id),)
 
660
            try:
 
661
                kind = self.work_tree.kind(file_id)
 
662
                # TODO: specific_files filtering before nested tree processing
 
663
                if kind == 'tree-reference' and self.recursive == 'down':
 
664
                    self._commit_nested_tree(file_id, path)
 
665
            except errors.NoSuchFile:
 
666
                pass
852
667
 
853
668
            # Record an entry for this item
854
669
            # Note: I don't particularly want to have the existing_ie
855
670
            # parameter but the test suite currently (28-Jun-07) breaks
856
671
            # without it thanks to a unicode normalisation issue. :-(
857
 
            definitely_changed = kind != existing_ie.kind
 
672
            definitely_changed = kind != existing_ie.kind 
858
673
            self._record_entry(path, file_id, specific_files, kind, name,
859
 
                parent_id, definitely_changed, existing_ie, report_changes,
860
 
                content_summary)
 
674
                parent_id, definitely_changed, existing_ie)
861
675
 
862
676
        # Unversion IDs that were found to be deleted
863
677
        self.work_tree.unversion(deleted_ids)
868
682
        # FIXME: be more comprehensive here:
869
683
        # this works when both trees are in --trees repository,
870
684
        # but when both are bound to a different repository,
871
 
        # it fails; a better way of approaching this is to
 
685
        # it fails; a better way of approaching this is to 
872
686
        # finally implement the explicit-caches approach design
873
687
        # a while back - RBC 20070306.
874
 
        if sub_tree.branch.repository.has_same_location(
875
 
            self.work_tree.branch.repository):
 
688
        if (sub_tree.branch.repository.bzrdir.root_transport.base
 
689
            ==
 
690
            self.work_tree.branch.repository.bzrdir.root_transport.base):
876
691
            sub_tree.branch.repository = \
877
692
                self.work_tree.branch.repository
878
693
        try:
879
 
            return sub_tree.commit(message=None, revprops=self.revprops,
 
694
            sub_tree.commit(message=None, revprops=self.revprops,
880
695
                recursive=self.recursive,
881
696
                message_callback=self.message_callback,
882
697
                timestamp=self.timestamp, timezone=self.timezone,
885
700
                strict=self.strict, verbose=self.verbose,
886
701
                local=self.local, reporter=self.reporter)
887
702
        except errors.PointlessCommit:
888
 
            return self.work_tree.get_reference_revision(file_id)
 
703
            pass
889
704
 
890
705
    def _record_entry(self, path, file_id, specific_files, kind, name,
891
 
        parent_id, definitely_changed, existing_ie, report_changes,
892
 
        content_summary):
 
706
                      parent_id, definitely_changed, existing_ie=None):
893
707
        "Record the new inventory entry for a path if any."
894
708
        # mutter('check %s {%s}', path, file_id)
895
 
        # mutter('%s selected for commit', path)
896
 
        if definitely_changed or existing_ie is None:
897
 
            ie = make_entry(kind, name, parent_id, file_id)
 
709
        if (not specific_files or 
 
710
            is_inside_or_parent_of_any(specific_files, path)):
 
711
                # mutter('%s selected for commit', path)
 
712
                if definitely_changed or existing_ie is None:
 
713
                    ie = inventory.make_entry(kind, name, parent_id, file_id)
 
714
                else:
 
715
                    ie = existing_ie.copy()
 
716
                    ie.revision = None
898
717
        else:
899
 
            ie = existing_ie.copy()
900
 
            ie.revision = None
901
 
        # For carried over entries we don't care about the fs hash - the repo
902
 
        # isn't generating a sha, so we're not saving computation time.
903
 
        _, _, fs_hash = self.builder.record_entry_contents(
904
 
            ie, self.parent_invs, path, self.work_tree, content_summary)
905
 
        if report_changes:
 
718
            # mutter('%s not selected for commit', path)
 
719
            if self.basis_inv.has_id(file_id):
 
720
                ie = self.basis_inv[file_id].copy()
 
721
            else:
 
722
                # this entry is new and not being committed
 
723
                ie = None
 
724
        if ie is not None:
 
725
            self.builder.record_entry_contents(ie, self.parent_invs, 
 
726
                path, self.work_tree)
906
727
            self._report_change(ie, path)
907
 
        if fs_hash:
908
 
            self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
909
728
        return ie
910
729
 
911
730
    def _report_change(self, ie, path):
919
738
        else:
920
739
            basis_ie = None
921
740
        change = ie.describe_change(basis_ie, ie)
922
 
        if change in (InventoryEntry.RENAMED,
 
741
        if change in (InventoryEntry.RENAMED, 
923
742
            InventoryEntry.MODIFIED_AND_RENAMED):
924
743
            old_path = self.basis_inv.id2path(ie.file_id)
925
744
            self.reporter.renamed(change, old_path, path)
926
 
            self._next_progress_entry()
927
745
        else:
928
 
            if change == 'unchanged':
929
 
                return
930
746
            self.reporter.snapshot_change(change, path)
931
 
            self._next_progress_entry()
932
747
 
933
 
    def _set_progress_stage(self, name, counter=False):
 
748
    def _set_progress_stage(self, name, entries_title=None):
934
749
        """Set the progress stage and emit an update to the progress bar."""
935
750
        self.pb_stage_name = name
936
751
        self.pb_stage_count += 1
937
 
        if counter:
 
752
        self.pb_entries_title = entries_title
 
753
        if entries_title is not None:
938
754
            self.pb_entries_count = 0
939
 
        else:
940
 
            self.pb_entries_count = None
 
755
            self.pb_entries_total = '?'
941
756
        self._emit_progress()
942
757
 
943
758
    def _next_progress_entry(self):
946
761
        self._emit_progress()
947
762
 
948
763
    def _emit_progress(self):
949
 
        if self.pb_entries_count is not None:
950
 
            text = "%s [%d] - Stage" % (self.pb_stage_name,
951
 
                self.pb_entries_count)
 
764
        if self.pb_entries_title:
 
765
            if self.pb_entries_total == '?':
 
766
                text = "%s [%s %d] - Stage" % (self.pb_stage_name,
 
767
                    self.pb_entries_title, self.pb_entries_count)
 
768
            else:
 
769
                text = "%s [%s %d/%s] - Stage" % (self.pb_stage_name,
 
770
                    self.pb_entries_title, self.pb_entries_count,
 
771
                    str(self.pb_entries_total))
952
772
        else:
953
 
            text = "%s - Stage" % (self.pb_stage_name, )
 
773
            text = "%s - Stage" % (self.pb_stage_name)
954
774
        self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
955
775
 
956
 
    def _set_specific_file_ids(self):
957
 
        """populate self.specific_file_ids if we will use it."""
958
 
        if not self.use_record_iter_changes:
959
 
            # If provided, ensure the specified files are versioned
960
 
            if self.specific_files is not None:
961
 
                # Note: This routine is being called because it raises
962
 
                # PathNotVersionedError as a side effect of finding the IDs. We
963
 
                # later use the ids we found as input to the working tree
964
 
                # inventory iterator, so we only consider those ids rather than
965
 
                # examining the whole tree again.
966
 
                # XXX: Dont we have filter_unversioned to do this more
967
 
                # cheaply?
968
 
                self.specific_file_ids = tree.find_ids_across_trees(
969
 
                    self.specific_files, [self.basis_tree, self.work_tree])
970
 
            else:
971
 
                self.specific_file_ids = None