1
# Copyright (C) 2005 Canonical Ltd
1
# Copyright (C) 2005, 2006, 2007, 2008 Canonical Ltd
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
5
5
# the Free Software Foundation; either version 2 of the License, or
6
6
# (at your option) any later version.
8
8
# This program is distributed in the hope that it will be useful,
9
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
11
# GNU General Public License for more details.
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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18
## XXX: Can we do any better about making interrupted commits change
21
## XXX: If we merged two versions of a file then we still need to
22
## create a new version representing that merge, even if it didn't
23
## change from the parent.
25
## TODO: Read back the just-generated changeset, and make sure it
26
## applies and recreates the right state.
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
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
20
# working tree and that were in the predecessor are reported as
21
# removed --- this can include files that were either removed from the
22
# inventory or deleted in the working tree. If they were only
23
# deleted from disk, they are removed from the working inventory.
25
# We then consider the remaining entries, which will be in the new
26
# version. Directory entries are simply copied across. File entries
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
29
# version of the file was present. If the file was present in at
30
# least one tree, and if it was the same version in all the trees,
31
# then we can just refer to that version. Otherwise, a new version
32
# representing the merger of the file versions must be added.
34
# TODO: Update hashcache before and after - or does the WorkingTree
37
# TODO: Rather than mashing together the ancestry and storing it back,
38
# perhaps the weave should have single method which does it all in one
39
# go, avoiding a lot of redundant work.
41
# TODO: Perhaps give a warning if one of the revisions marked as
42
# merged is already in the ancestry, and then don't record it as a
45
# TODO: If the file is newly merged but unchanged from the version it
46
# merges from, then it should still be reported as newly added
47
# relative to the basis revision.
49
# TODO: Change the parameter 'rev_id' to 'revision_id' to be consistent with
50
# the rest of the code; add a deprecation of the old name.
37
from binascii import hexlify
38
57
from cStringIO import StringIO
40
from bzrlib.osutils import (local_time_offset, username,
41
rand_bytes, compact_date, user_email,
42
kind_marker, is_inside_any, quotefn,
43
sha_string, sha_file, isdir, isfile)
44
from bzrlib.branch import gen_file_id, INVENTORY_FILEID, ANCESTRY_FILEID
45
from bzrlib.errors import BzrError, PointlessCommit
46
from bzrlib.revision import Revision, RevisionReference
47
from bzrlib.trace import mutter, note
48
from bzrlib.xml5 import serializer_v5
49
from bzrlib.inventory import Inventory
50
from bzrlib.delta import compare_trees
51
from bzrlib.weave import Weave
52
from bzrlib.weavefile import read_weave, write_weave_v5
53
from bzrlib.atomicfile import AtomicFile
56
def commit(*args, **kwargs):
57
"""Commit a new revision to a branch.
59
Function-style interface for convenience of old callers.
61
New code should use the Commit class instead.
63
Commit().commit(*args, **kwargs)
67
from bzrlib.branch import Branch
69
from bzrlib.errors import (BzrError, PointlessCommit,
73
from bzrlib.osutils import (get_user_encoding,
74
kind_marker, isdir,isfile, is_inside_any,
75
is_inside_or_parent_of_any,
76
minimum_path_selection,
77
quotefn, sha_file, split_lines,
80
from bzrlib.testament import Testament
81
from bzrlib.trace import mutter, note, warning, is_quiet
82
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
83
from bzrlib import symbol_versioning
84
from bzrlib.symbol_versioning import (deprecated_passed,
87
from bzrlib.workingtree import WorkingTree
88
from bzrlib.urlutils import unescape_for_display
66
92
class NullCommitReporter(object):
67
93
"""I report on progress of a commit."""
68
def added(self, path):
71
def removed(self, path):
74
def renamed(self, old_path, new_path):
95
def started(self, revno, revid, location=None):
97
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
98
"to started.", DeprecationWarning,
102
def snapshot_change(self, change, path):
105
def completed(self, revno, rev_id):
108
def deleted(self, path):
111
def missing(self, path):
114
def renamed(self, change, old_path, new_path):
117
def is_verbose(self):
78
121
class ReportCommitToLog(NullCommitReporter):
79
def added(self, path):
80
note('added %s', path)
82
def removed(self, path):
83
note('removed %s', path)
85
def renamed(self, old_path, new_path):
86
note('renamed %s => %s', old_path, new_path)
123
def _note(self, format, *args):
126
Subclasses may choose to override this method.
130
def snapshot_change(self, change, path):
131
if path == '' and change in ('added', 'modified'):
133
self._note("%s %s", change, path)
135
def started(self, revno, rev_id, location=None):
136
if location is not None:
137
location = ' to: ' + unescape_for_display(location, 'utf-8')
139
# When started was added, location was only made optional by
140
# accident. Matt Nordhoff 20071129
141
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
142
"to started.", DeprecationWarning,
145
self._note('Committing%s', location)
147
def completed(self, revno, rev_id):
148
self._note('Committed revision %d.', revno)
150
def deleted(self, path):
151
self._note('deleted %s', path)
153
def missing(self, path):
154
self._note('missing %s', path)
156
def renamed(self, change, old_path, new_path):
157
self._note('%s %s => %s', change, old_path, new_path)
159
def is_verbose(self):
89
163
class Commit(object):
99
173
working inventory.
101
175
def __init__(self,
103
if reporter is not None:
104
self.reporter = reporter
106
self.reporter = NullCommitReporter()
178
"""Create a Commit object.
180
:param reporter: the default reporter to use or None to decide later
182
self.reporter = reporter
114
190
specific_files=None,
116
allow_pointless=True):
192
allow_pointless=True,
200
message_callback=None,
203
possible_master_transports=None):
117
204
"""Commit working copy as a new revision.
119
The basic approach is to add all the file texts into the
120
store, then the inventory, then make a new revision pointing
121
to that inventory and store that.
123
This is not quite safe if the working copy changes during the
124
commit; for the moment that is simply not allowed. A better
125
approach is to make a temporary copy of the files before
126
computing their hashes, and then add those hashes in turn to
127
the inventory. This should mean at least that there are no
128
broken hash pointers. There is no way we can get a snapshot
129
of the whole directory at an instant. This would also have to
130
be robust against files disappearing, moving, etc. So the
131
whole thing is a bit hard.
133
This raises PointlessCommit if there are no changes, no new merges,
134
and allow_pointless is false.
136
timestamp -- if not None, seconds-since-epoch for a
137
postdated/predated commit.
140
If true, commit only those files.
143
If set, use this as the new revision id.
206
:param message: the commit message (it or message_callback is required)
207
:param message_callback: A callback: message = message_callback(cmt_obj)
209
:param timestamp: if not None, seconds-since-epoch for a
210
postdated/predated commit.
212
:param specific_files: If true, commit only those files.
214
:param rev_id: If set, use this as the new revision id.
144
215
Useful for test or import commands that need to tightly
145
216
control what revisions are assigned. If you duplicate
146
217
a revision id that exists elsewhere it is your own fault.
147
218
If null (default), a time/random revision id is generated.
220
:param allow_pointless: If true (default), commit even if nothing
221
has changed and no merges are recorded.
223
:param strict: If true, don't allow a commit if the working tree
224
contains unknown files.
226
:param revprops: Properties for new revision
227
:param local: Perform a local only commit.
228
:param reporter: the reporter to use or None for the default
229
:param verbose: if True and the reporter is not None, report everything
230
:param recursive: If set to 'down', commit in any subtrees that have
231
pending changes of any sort during this commit.
232
:param exclude: None or a list of relative paths to exclude from the
233
commit. Pending changes to excluded files will be ignored by the
151
self.branch.lock_write()
153
self.specific_files = specific_files
236
mutter('preparing to commit')
238
if working_tree is None:
239
raise BzrError("working_tree must be passed into commit().")
241
self.work_tree = working_tree
242
self.branch = self.work_tree.branch
243
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
244
if not self.branch.repository.supports_rich_root():
245
raise errors.RootNotRich()
246
if message_callback is None:
247
if message is not None:
248
if isinstance(message, str):
249
message = message.decode(get_user_encoding())
250
message_callback = lambda x: message
252
raise BzrError("The message or message_callback keyword"
253
" parameter is required for commit().")
255
self.bound_branch = None
256
self.any_entries_deleted = False
257
if exclude is not None:
258
self.exclude = sorted(
259
minimum_path_selection(exclude))
263
self.master_branch = None
264
self.master_locked = False
265
self.recursive = recursive
267
if specific_files is not None:
268
self.specific_files = sorted(
269
minimum_path_selection(specific_files))
271
self.specific_files = None
154
273
self.allow_pointless = allow_pointless
156
if timestamp is None:
157
self.timestamp = time.time()
159
self.timestamp = long(timestamp)
161
if committer is None:
162
self.committer = username(self.branch)
164
assert isinstance(committer, basestring), type(committer)
165
self.committer = committer
168
self.timezone = local_time_offset()
170
self.timezone = int(timezone)
172
assert isinstance(message, basestring), type(message)
173
self.message = message
274
self.revprops = revprops
275
self.message_callback = message_callback
276
self.timestamp = timestamp
277
self.timezone = timezone
278
self.committer = committer
280
self.verbose = verbose
282
self.work_tree.lock_write()
283
self.parents = self.work_tree.get_parent_ids()
284
# We can use record_iter_changes IFF iter_changes is compatible with
285
# the command line parameters, and the repository has fast delta
286
# generation. See bug 347649.
287
self.use_record_iter_changes = (
288
not self.specific_files and
290
not self.branch.repository._format.supports_tree_reference and
291
(self.branch.repository._format.fast_deltas or
292
len(self.parents) < 2))
293
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
294
self.basis_revid = self.work_tree.last_revision()
295
self.basis_tree = self.work_tree.basis_tree()
296
self.basis_tree.lock_read()
176
# First walk over the working inventory; and both update that
177
# and also build a new revision inventory. The revision
178
# inventory needs to hold the text-id, sha1 and size of the
179
# actual file versions committed in the revision. (These are
180
# not present in the working inventory.) We also need to
181
# detect missing/deleted files, and remove them from the
184
self.work_tree = self.branch.working_tree()
185
self.work_inv = self.work_tree.inventory
186
self.basis_tree = self.branch.basis_tree()
187
self.basis_inv = self.basis_tree.inventory
298
# Cannot commit with conflicts present.
299
if len(self.work_tree.conflicts()) > 0:
300
raise ConflictsInTree
302
# Setup the bound branch variables as needed.
303
self._check_bound_branch(possible_master_transports)
305
# Check that the working tree is up to date
306
old_revno, new_revno = self._check_out_of_date_tree()
308
# Complete configuration setup
309
if reporter is not None:
310
self.reporter = reporter
311
elif self.reporter is None:
312
self.reporter = self._select_reporter()
313
if self.config is None:
314
self.config = self.branch.get_config()
316
self._set_specific_file_ids()
318
# Setup the progress bar. As the number of files that need to be
319
# committed in unknown, progress is reported as stages.
320
# We keep track of entries separately though and include that
321
# information in the progress bar during the relevant stages.
322
self.pb_stage_name = ""
323
self.pb_stage_count = 0
324
self.pb_stage_total = 5
325
if self.bound_branch:
326
self.pb_stage_total += 1
327
self.pb.show_pct = False
328
self.pb.show_spinner = False
329
self.pb.show_eta = False
330
self.pb.show_count = True
331
self.pb.show_bar = True
189
333
self._gather_parents()
191
if self.rev_id is None:
192
self.rev_id = _gen_revision_id(self.branch, time.time())
194
# TODO: update hashcache
195
self.delta = compare_trees(self.basis_tree, self.work_tree,
196
specific_files=self.specific_files)
198
if not (self.delta.has_changed()
199
or self.allow_pointless
200
or len(self.parents) != 1):
201
raise PointlessCommit()
203
self.new_inv = self.basis_inv.copy()
205
self.delta.show(sys.stdout)
207
self._remove_deleted()
210
self.branch._write_inventory(self.work_inv)
211
self._record_inventory()
212
self._record_ancestry()
214
self._make_revision()
215
note('committted r%d {%s}', (self.branch.revno() + 1),
217
self.branch.append_revision(self.rev_id)
218
self.branch.set_pending_merges([])
334
# After a merge, a selected file commit is not supported.
335
# See 'bzr help merge' for an explanation as to why.
336
if len(self.parents) > 1 and self.specific_files:
337
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
338
# Excludes are a form of selected file commit.
339
if len(self.parents) > 1 and self.exclude:
340
raise errors.CannotCommitSelectedFileMerge(self.exclude)
342
# Collect the changes
343
self._set_progress_stage("Collecting changes", counter=True)
344
self.builder = self.branch.get_commit_builder(self.parents,
345
self.config, timestamp, timezone, committer, revprops, rev_id)
348
self.builder.will_record_deletes()
349
# find the location being committed to
350
if self.bound_branch:
351
master_location = self.master_branch.base
353
master_location = self.branch.base
355
# report the start of the commit
356
self.reporter.started(new_revno, self.rev_id, master_location)
358
self._update_builder_with_changes()
359
self._check_pointless()
361
# TODO: Now the new inventory is known, check for conflicts.
362
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
363
# weave lines, because nothing should be recorded until it is known
364
# that commit will succeed.
365
self._set_progress_stage("Saving data locally")
366
self.builder.finish_inventory()
368
# Prompt the user for a commit message if none provided
369
message = message_callback(self)
370
self.message = message
372
# Add revision data to the local branch
373
self.rev_id = self.builder.commit(self.message)
376
mutter("aborting commit write group because of exception:")
377
trace.log_exception_quietly()
378
note("aborting commit write group: %r" % (e,))
382
self._process_pre_hooks(old_revno, new_revno)
384
# Upload revision data to the master.
385
# this will propagate merged revisions too if needed.
386
if self.bound_branch:
387
self._set_progress_stage("Uploading data to master branch")
388
# 'commit' to the master first so a timeout here causes the
389
# local branch to be out of date
390
self.master_branch.import_last_revision_info(
391
self.branch.repository, new_revno, self.rev_id)
393
# and now do the commit locally.
394
self.branch.set_last_revision_info(new_revno, self.rev_id)
396
# Make the working tree be up to date with the branch. This
397
# includes automatic changes scheduled to be made to the tree, such
398
# as updating its basis and unversioning paths that were missing.
399
self.work_tree.unversion(self.deleted_ids)
400
self._set_progress_stage("Updating the working tree")
401
self.work_tree.update_basis_by_delta(self.rev_id,
402
self.builder.get_basis_delta())
403
self.reporter.completed(new_revno, self.rev_id)
404
self._process_post_hooks(old_revno, new_revno)
223
def _record_inventory(self):
224
"""Store the inventory for the new revision."""
226
serializer_v5.write_inventory(self.new_inv, inv_tmp)
228
self.inv_sha1 = sha_string(inv_tmp.getvalue())
229
inv_lines = inv_tmp.readlines()
230
self.branch.weave_store.add_text(INVENTORY_FILEID, self.rev_id,
231
inv_lines, self.parents)
234
def _record_ancestry(self):
235
"""Append merged revision ancestry to the ancestry file."""
409
def _select_reporter(self):
410
"""Select the CommitReporter to use."""
412
return NullCommitReporter()
413
return ReportCommitToLog()
415
def _check_pointless(self):
416
if self.allow_pointless:
418
# A merge with no effect on files
236
419
if len(self.parents) > 1:
237
raise NotImplementedError("sorry, can't commit merges yet")
238
w = self.branch.weave_store.get_weave_or_empty(ANCESTRY_FILEID)
421
# TODO: we could simplify this by using self.builder.basis_delta.
423
# The initial commit adds a root directory, but this in itself is not
424
# a worthwhile commit.
425
if (self.basis_revid == revision.NULL_REVISION and
426
((self.builder.new_inventory is not None and
427
len(self.builder.new_inventory) == 1) or
428
len(self.builder._basis_delta) == 1)):
429
raise PointlessCommit()
430
if self.builder.any_changes():
432
raise PointlessCommit()
434
def _check_bound_branch(self, possible_master_transports=None):
435
"""Check to see if the local branch is bound.
437
If it is bound, then most of the commit will actually be
438
done using the remote branch as the target branch.
439
Only at the end will the local branch be updated.
441
if self.local and not self.branch.get_bound_location():
442
raise errors.LocalRequiresBoundBranch()
445
self.master_branch = self.branch.get_master_branch(
446
possible_master_transports)
448
if not self.master_branch:
449
# make this branch the reference branch for out of date checks.
450
self.master_branch = self.branch
453
# If the master branch is bound, we must fail
454
master_bound_location = self.master_branch.get_bound_location()
455
if master_bound_location:
456
raise errors.CommitToDoubleBoundBranch(self.branch,
457
self.master_branch, master_bound_location)
459
# TODO: jam 20051230 We could automatically push local
460
# commits to the remote branch if they would fit.
461
# But for now, just require remote to be identical
464
# Make sure the local branch is identical to the master
465
master_info = self.master_branch.last_revision_info()
466
local_info = self.branch.last_revision_info()
467
if local_info != master_info:
468
raise errors.BoundBranchOutOfDate(self.branch,
471
# Now things are ready to change the master branch
473
self.bound_branch = self.branch
474
self.master_branch.lock_write()
475
self.master_locked = True
477
def _check_out_of_date_tree(self):
478
"""Check that the working tree is up to date.
480
:return: old_revision_number,new_revision_number tuple
483
first_tree_parent = self.work_tree.get_parent_ids()[0]
485
# if there are no parents, treat our parent as 'None'
486
# this is so that we still consider the master branch
487
# - in a checkout scenario the tree may have no
488
# parents but the branch may do.
489
first_tree_parent = bzrlib.revision.NULL_REVISION
490
old_revno, master_last = self.master_branch.last_revision_info()
491
if master_last != first_tree_parent:
492
if master_last != bzrlib.revision.NULL_REVISION:
493
raise errors.OutOfDateTree(self.work_tree)
494
if self.branch.repository.has_revision(first_tree_parent):
495
new_revno = old_revno + 1
497
# ghost parents never appear in revision history.
499
return old_revno,new_revno
501
def _process_pre_hooks(self, old_revno, new_revno):
502
"""Process any registered pre commit hooks."""
503
self._set_progress_stage("Running pre_commit hooks")
504
self._process_hooks("pre_commit", old_revno, new_revno)
506
def _process_post_hooks(self, old_revno, new_revno):
507
"""Process any registered post commit hooks."""
508
# Process the post commit hooks, if any
509
self._set_progress_stage("Running post_commit hooks")
510
# old style commit hooks - should be deprecated ? (obsoleted in
512
if self.config.post_commit() is not None:
513
hooks = self.config.post_commit().split(' ')
514
# this would be nicer with twisted.python.reflect.namedAny
516
result = eval(hook + '(branch, rev_id)',
517
{'branch':self.branch,
519
'rev_id':self.rev_id})
520
# process new style post commit hooks
521
self._process_hooks("post_commit", old_revno, new_revno)
523
def _process_hooks(self, hook_name, old_revno, new_revno):
524
if not Branch.hooks[hook_name]:
527
# new style commit hooks:
528
if not self.bound_branch:
529
hook_master = self.branch
532
hook_master = self.master_branch
533
hook_local = self.branch
534
# With bound branches, when the master is behind the local branch,
535
# the 'old_revno' and old_revid values here are incorrect.
536
# XXX: FIXME ^. RBC 20060206
240
lines = w.get(w.lookup(self.parents[0]))
538
old_revid = self.parents[0]
243
lines.append(self.rev_id + '\n')
244
parent_idxs = map(w.lookup, self.parents)
245
w.add(self.rev_id, parent_idxs, lines)
246
self.branch.weave_store.put_weave(ANCESTRY_FILEID, w)
540
old_revid = bzrlib.revision.NULL_REVISION
542
if hook_name == "pre_commit":
543
future_tree = self.builder.revision_tree()
544
tree_delta = future_tree.changes_from(self.basis_tree,
547
for hook in Branch.hooks[hook_name]:
548
# show the running hook in the progress bar. As hooks may
549
# end up doing nothing (e.g. because they are not configured by
550
# the user) this is still showing progress, not showing overall
551
# actions - its up to each plugin to show a UI if it want's to
552
# (such as 'Emailing diff to foo@example.com').
553
self.pb_stage_name = "Running %s hooks [%s]" % \
554
(hook_name, Branch.hooks.get_hook_name(hook))
555
self._emit_progress()
556
if 'hooks' in debug.debug_flags:
557
mutter("Invoking commit hook: %r", hook)
558
if hook_name == "post_commit":
559
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
561
elif hook_name == "pre_commit":
562
hook(hook_local, hook_master,
563
old_revno, old_revid, new_revno, self.rev_id,
564
tree_delta, future_tree)
567
"""Cleanup any open locks, progress bars etc."""
568
cleanups = [self._cleanup_bound_branch,
569
self.basis_tree.unlock,
570
self.work_tree.unlock,
572
found_exception = None
573
for cleanup in cleanups:
576
# we want every cleanup to run no matter what.
577
# so we have a catchall here, but we will raise the
578
# last encountered exception up the stack: and
579
# typically this will be useful enough.
582
if found_exception is not None:
583
# don't do a plan raise, because the last exception may have been
584
# trashed, e is our sure-to-work exception even though it loses the
585
# full traceback. XXX: RBC 20060421 perhaps we could check the
586
# exc_info and if its the same one do a plain raise otherwise
587
# 'raise e' as we do now.
590
def _cleanup_bound_branch(self):
591
"""Executed at the end of a try/finally to cleanup a bound branch.
593
If the branch wasn't bound, this is a no-op.
594
If it was, it resents self.branch to the local branch, instead
597
if not self.bound_branch:
599
if self.master_locked:
600
self.master_branch.unlock()
249
602
def _gather_parents(self):
250
pending_merges = self.branch.pending_merges()
252
raise NotImplementedError("sorry, can't commit merges to the weave format yet")
254
precursor_id = self.branch.last_patch()
256
self.parents.append(precursor_id)
257
self.parents += pending_merges
260
def _make_revision(self):
261
"""Record a new revision object for this commit."""
262
self.rev = Revision(timestamp=self.timestamp,
263
timezone=self.timezone,
264
committer=self.committer,
265
message=self.message,
266
inventory_sha1=self.inv_sha1,
267
revision_id=self.rev_id)
268
self.rev.parents = map(RevisionReference, self.parents)
269
rev_tmp = tempfile.TemporaryFile()
270
serializer_v5.write_revision(self.rev, rev_tmp)
272
self.branch.revision_store.add(rev_tmp, self.rev_id)
273
mutter('new revision_id is {%s}', self.rev_id)
276
def _remove_deleted(self):
277
"""Remove deleted files from the working and stored inventories."""
278
for path, id, kind in self.delta.removed:
279
if self.work_inv.has_id(id):
280
del self.work_inv[id]
281
if self.new_inv.has_id(id):
286
def _store_files(self):
287
"""Store new texts of modified/added files."""
288
# We must make sure that directories are added before anything
289
# inside them is added. the files within the delta report are
290
# sorted by path so we know the directory will come before its
292
for path, file_id, kind in self.delta.added:
294
ie = self.work_inv[file_id].copy()
297
self._store_file_text(file_id)
299
for path, file_id, kind in self.delta.modified:
302
self._store_file_text(file_id)
304
for old_path, new_path, file_id, kind, text_modified in self.delta.renamed:
307
if not text_modified:
309
self._store_file_text(file_id)
312
def _store_file_text(self, file_id):
313
"""Store updated text for one modified or added file."""
314
note('store new text for {%s} in revision {%s}',
315
file_id, self.rev_id)
316
new_lines = self.work_tree.get_file(file_id).readlines()
317
if file_id in self.new_inv: # was in basis inventory
318
ie = self.new_inv[file_id]
319
assert ie.file_id == file_id
320
assert file_id in self.basis_inv
321
assert self.basis_inv[file_id].kind == 'file'
322
old_version = self.basis_inv[file_id].text_version
323
file_parents = [old_version]
324
else: # new in this revision
325
ie = self.work_inv[file_id].copy()
327
assert file_id not in self.basis_inv
329
assert ie.kind == 'file'
330
self._add_text_to_weave(file_id, new_lines, file_parents)
331
# make a new inventory entry for this file, using whatever
332
# it had in the working copy, plus details on the new text
333
ie.text_sha1 = _sha_strings(new_lines)
334
ie.text_size = sum(map(len, new_lines))
335
ie.text_version = self.rev_id
336
ie.entry_version = self.rev_id
339
def _add_text_to_weave(self, file_id, new_lines, parents):
340
if file_id.startswith('__'):
341
raise ValueError('illegal file-id %r for text file' % file_id)
342
self.branch.weave_store.add_text(file_id, self.rev_id, new_lines, parents)
345
def _gen_revision_id(branch, when):
346
"""Return new revision-id."""
347
s = '%s-%s-' % (user_email(branch), compact_date(when))
348
s += hexlify(rand_bytes(8))
352
def _sha_strings(strings):
353
"""Return the sha-1 of concatenation of strings"""
355
map(s.update, strings)
603
"""Record the parents of a merge for merge detection."""
604
# TODO: Make sure that this list doesn't contain duplicate
605
# entries and the order is preserved when doing this.
606
if self.use_record_iter_changes:
608
self.basis_inv = self.basis_tree.inventory
609
self.parent_invs = [self.basis_inv]
610
for revision in self.parents[1:]:
611
if self.branch.repository.has_revision(revision):
612
mutter('commit parent revision {%s}', revision)
613
inventory = self.branch.repository.get_inventory(revision)
614
self.parent_invs.append(inventory)
616
mutter('commit parent ghost revision {%s}', revision)
618
def _update_builder_with_changes(self):
619
"""Update the commit builder with the data about what has changed.
621
exclude = self.exclude
622
specific_files = self.specific_files or []
623
mutter("Selecting files for commit with filter %s", specific_files)
626
if self.use_record_iter_changes:
627
iter_changes = self.work_tree.iter_changes(self.basis_tree)
628
iter_changes = self._filter_iter_changes(iter_changes)
629
for file_id, path, fs_hash in self.builder.record_iter_changes(
630
self.work_tree, self.basis_revid, iter_changes):
631
self.work_tree._observed_sha1(file_id, path, fs_hash)
633
# Build the new inventory
634
self._populate_from_inventory()
635
self._record_unselected()
636
self._report_and_accumulate_deletes()
638
def _filter_iter_changes(self, iter_changes):
639
"""Process iter_changes.
641
This method reports on the changes in iter_changes to the user, and
642
converts 'missing' entries in the iter_changes iterator to 'deleted'
643
entries. 'missing' entries have their
645
:param iter_changes: An iter_changes to process.
646
:return: A generator of changes.
648
reporter = self.reporter
649
report_changes = reporter.is_verbose()
651
for change in iter_changes:
653
old_path = change[1][0]
654
new_path = change[1][1]
655
versioned = change[3][1]
657
versioned = change[3][1]
658
if kind is None and versioned:
661
reporter.missing(new_path)
662
deleted_ids.append(change[0])
663
# Reset the new path (None) and new versioned flag (False)
664
change = (change[0], (change[1][0], None), change[2],
665
(change[3][0], False)) + change[4:]
666
elif kind == 'tree-reference':
667
if self.recursive == 'down':
668
self._commit_nested_tree(change[0], change[1][1])
669
if change[3][0] or change[3][1]:
673
reporter.deleted(old_path)
674
elif old_path is None:
675
reporter.snapshot_change('added', new_path)
676
elif old_path != new_path:
677
reporter.renamed('renamed', old_path, new_path)
680
self.work_tree.branch.repository._format.rich_root_data):
681
# Don't report on changes to '' in non rich root
683
reporter.snapshot_change('modified', new_path)
684
self._next_progress_entry()
685
# Unversion IDs that were found to be deleted
686
self.deleted_ids = deleted_ids
688
def _record_unselected(self):
689
# If specific files are selected, then all un-selected files must be
690
# recorded in their previous state. For more details, see
691
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
692
if self.specific_files or self.exclude:
693
specific_files = self.specific_files or []
694
for path, old_ie in self.basis_inv.iter_entries():
695
if old_ie.file_id in self.builder.new_inventory:
696
# already added - skip.
698
if (is_inside_any(specific_files, path)
699
and not is_inside_any(self.exclude, path)):
700
# was inside the selected path, and not excluded - if not
701
# present it has been deleted so skip.
703
# From here down it was either not selected, or was excluded:
704
# We preserve the entry unaltered.
706
# Note: specific file commits after a merge are currently
707
# prohibited. This test is for sanity/safety in case it's
708
# required after that changes.
709
if len(self.parents) > 1:
711
self.builder.record_entry_contents(ie, self.parent_invs, path,
712
self.basis_tree, None)
714
def _report_and_accumulate_deletes(self):
715
if (isinstance(self.basis_inv, Inventory)
716
and isinstance(self.builder.new_inventory, Inventory)):
717
# the older Inventory classes provide a _byid dict, and building a
718
# set from the keys of this dict is substantially faster than even
719
# getting a set of ids from the inventory
721
# <lifeless> set(dict) is roughly the same speed as
722
# set(iter(dict)) and both are significantly slower than
724
deleted_ids = set(self.basis_inv._byid.keys()) - \
725
set(self.builder.new_inventory._byid.keys())
727
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
729
self.any_entries_deleted = True
730
deleted = [(self.basis_tree.id2path(file_id), file_id)
731
for file_id in deleted_ids]
733
# XXX: this is not quite directory-order sorting
734
for path, file_id in deleted:
735
self.builder.record_delete(path, file_id)
736
self.reporter.deleted(path)
738
def _check_strict(self):
739
# XXX: when we use iter_changes this would likely be faster if
740
# iter_changes would check for us (even in the presence of
743
# raise an exception as soon as we find a single unknown.
744
for unknown in self.work_tree.unknowns():
745
raise StrictCommitFailed()
747
def _populate_from_inventory(self):
748
"""Populate the CommitBuilder by walking the working tree inventory."""
749
# Build the revision inventory.
751
# This starts by creating a new empty inventory. Depending on
752
# which files are selected for commit, and what is present in the
753
# current tree, the new inventory is populated. inventory entries
754
# which are candidates for modification have their revision set to
755
# None; inventory entries that are carried over untouched have their
756
# revision set to their prior value.
758
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
759
# results to create a new inventory at the same time, which results
760
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
763
specific_files = self.specific_files
764
exclude = self.exclude
765
report_changes = self.reporter.is_verbose()
767
# A tree of paths that have been deleted. E.g. if foo/bar has been
768
# deleted, then we have {'foo':{'bar':{}}}
770
# XXX: Note that entries may have the wrong kind because the entry does
771
# not reflect the status on disk.
772
work_inv = self.work_tree.inventory
773
# NB: entries will include entries within the excluded ids/paths
774
# because iter_entries_by_dir has no 'exclude' facility today.
775
entries = work_inv.iter_entries_by_dir(
776
specific_file_ids=self.specific_file_ids, yield_parents=True)
777
for path, existing_ie in entries:
778
file_id = existing_ie.file_id
779
name = existing_ie.name
780
parent_id = existing_ie.parent_id
781
kind = existing_ie.kind
782
# Skip files that have been deleted from the working tree.
783
# The deleted path ids are also recorded so they can be explicitly
786
path_segments = splitpath(path)
787
deleted_dict = deleted_paths
788
for segment in path_segments:
789
deleted_dict = deleted_dict.get(segment, None)
791
# We either took a path not present in the dict
792
# (deleted_dict was None), or we've reached an empty
793
# child dir in the dict, so are now a sub-path.
797
if deleted_dict is not None:
798
# the path has a deleted parent, do not add it.
800
if exclude and is_inside_any(exclude, path):
801
# Skip excluded paths. Excluded paths are processed by
802
# _update_builder_with_changes.
804
content_summary = self.work_tree.path_content_summary(path)
805
# Note that when a filter of specific files is given, we must only
806
# skip/record deleted files matching that filter.
807
if not specific_files or is_inside_any(specific_files, path):
808
if content_summary[0] == 'missing':
809
if not deleted_paths:
810
# path won't have been split yet.
811
path_segments = splitpath(path)
812
deleted_dict = deleted_paths
813
for segment in path_segments:
814
deleted_dict = deleted_dict.setdefault(segment, {})
815
self.reporter.missing(path)
816
self._next_progress_entry()
817
deleted_ids.append(file_id)
819
# TODO: have the builder do the nested commit just-in-time IF and
821
if content_summary[0] == 'tree-reference':
822
# enforce repository nested tree policy.
823
if (not self.work_tree.supports_tree_reference() or
824
# repository does not support it either.
825
not self.branch.repository._format.supports_tree_reference):
826
content_summary = ('directory',) + content_summary[1:]
827
kind = content_summary[0]
828
# TODO: specific_files filtering before nested tree processing
829
if kind == 'tree-reference':
830
if self.recursive == 'down':
831
nested_revision_id = self._commit_nested_tree(
833
content_summary = content_summary[:3] + (
836
content_summary = content_summary[:3] + (
837
self.work_tree.get_reference_revision(file_id),)
839
# Record an entry for this item
840
# Note: I don't particularly want to have the existing_ie
841
# parameter but the test suite currently (28-Jun-07) breaks
842
# without it thanks to a unicode normalisation issue. :-(
843
definitely_changed = kind != existing_ie.kind
844
self._record_entry(path, file_id, specific_files, kind, name,
845
parent_id, definitely_changed, existing_ie, report_changes,
848
# Unversion IDs that were found to be deleted
849
self.deleted_ids = deleted_ids
851
def _commit_nested_tree(self, file_id, path):
852
"Commit a nested tree."
853
sub_tree = self.work_tree.get_nested_tree(file_id, path)
854
# FIXME: be more comprehensive here:
855
# this works when both trees are in --trees repository,
856
# but when both are bound to a different repository,
857
# it fails; a better way of approaching this is to
858
# finally implement the explicit-caches approach design
859
# a while back - RBC 20070306.
860
if sub_tree.branch.repository.has_same_location(
861
self.work_tree.branch.repository):
862
sub_tree.branch.repository = \
863
self.work_tree.branch.repository
865
return sub_tree.commit(message=None, revprops=self.revprops,
866
recursive=self.recursive,
867
message_callback=self.message_callback,
868
timestamp=self.timestamp, timezone=self.timezone,
869
committer=self.committer,
870
allow_pointless=self.allow_pointless,
871
strict=self.strict, verbose=self.verbose,
872
local=self.local, reporter=self.reporter)
873
except errors.PointlessCommit:
874
return self.work_tree.get_reference_revision(file_id)
876
def _record_entry(self, path, file_id, specific_files, kind, name,
877
parent_id, definitely_changed, existing_ie, report_changes,
879
"Record the new inventory entry for a path if any."
880
# mutter('check %s {%s}', path, file_id)
881
# mutter('%s selected for commit', path)
882
if definitely_changed or existing_ie is None:
883
ie = make_entry(kind, name, parent_id, file_id)
885
ie = existing_ie.copy()
887
# For carried over entries we don't care about the fs hash - the repo
888
# isn't generating a sha, so we're not saving computation time.
889
_, _, fs_hash = self.builder.record_entry_contents(
890
ie, self.parent_invs, path, self.work_tree, content_summary)
892
self._report_change(ie, path)
894
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
897
def _report_change(self, ie, path):
898
"""Report a change to the user.
900
The change that has occurred is described relative to the basis
903
if (self.basis_inv.has_id(ie.file_id)):
904
basis_ie = self.basis_inv[ie.file_id]
907
change = ie.describe_change(basis_ie, ie)
908
if change in (InventoryEntry.RENAMED,
909
InventoryEntry.MODIFIED_AND_RENAMED):
910
old_path = self.basis_inv.id2path(ie.file_id)
911
self.reporter.renamed(change, old_path, path)
912
self._next_progress_entry()
914
if change == 'unchanged':
916
self.reporter.snapshot_change(change, path)
917
self._next_progress_entry()
919
def _set_progress_stage(self, name, counter=False):
920
"""Set the progress stage and emit an update to the progress bar."""
921
self.pb_stage_name = name
922
self.pb_stage_count += 1
924
self.pb_entries_count = 0
926
self.pb_entries_count = None
927
self._emit_progress()
929
def _next_progress_entry(self):
930
"""Emit an update to the progress bar and increment the entry count."""
931
self.pb_entries_count += 1
932
self._emit_progress()
934
def _emit_progress(self):
935
if self.pb_entries_count is not None:
936
text = "%s [%d] - Stage" % (self.pb_stage_name,
937
self.pb_entries_count)
939
text = "%s - Stage" % (self.pb_stage_name, )
940
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
942
def _set_specific_file_ids(self):
943
"""populate self.specific_file_ids if we will use it."""
944
if not self.use_record_iter_changes:
945
# If provided, ensure the specified files are versioned
946
if self.specific_files is not None:
947
# Note: This routine is being called because it raises
948
# PathNotVersionedError as a side effect of finding the IDs. We
949
# later use the ids we found as input to the working tree
950
# inventory iterator, so we only consider those ids rather than
951
# examining the whole tree again.
952
# XXX: Dont we have filter_unversioned to do this more
954
self.specific_file_ids = tree.find_ids_across_trees(
955
self.specific_files, [self.basis_tree, self.work_tree])
957
self.specific_file_ids = None
added
removed
bzrlib/commit.py
