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)
66
from bzrlib.branch import Branch
68
from bzrlib.errors import (BzrError, PointlessCommit,
72
from bzrlib.osutils import (get_user_encoding,
73
kind_marker, isdir,isfile, is_inside_any,
74
is_inside_or_parent_of_any,
75
minimum_path_selection,
76
quotefn, sha_file, split_lines,
79
from bzrlib.testament import Testament
80
from bzrlib.trace import mutter, note, warning, is_quiet
81
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
82
from bzrlib import symbol_versioning
83
from bzrlib.symbol_versioning import (deprecated_passed,
86
from bzrlib.workingtree import WorkingTree
87
from bzrlib.urlutils import unescape_for_display
66
91
class NullCommitReporter(object):
67
92
"""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):
94
def started(self, revno, revid, location=None):
96
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
97
"to started.", DeprecationWarning,
101
def snapshot_change(self, change, path):
104
def completed(self, revno, rev_id):
107
def deleted(self, file_id):
110
def escaped(self, escape_count, message):
113
def missing(self, path):
116
def renamed(self, change, old_path, new_path):
119
def is_verbose(self):
78
123
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)
125
def _note(self, format, *args):
128
Subclasses may choose to override this method.
132
def snapshot_change(self, change, path):
133
if change == 'unchanged':
135
if change == 'added' and path == '':
137
self._note("%s %s", change, path)
139
def started(self, revno, rev_id, location=None):
140
if location is not None:
141
location = ' to: ' + unescape_for_display(location, 'utf-8')
143
# When started was added, location was only made optional by
144
# accident. Matt Nordhoff 20071129
145
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
146
"to started.", DeprecationWarning,
149
self._note('Committing%s', location)
151
def completed(self, revno, rev_id):
152
self._note('Committed revision %d.', revno)
154
def deleted(self, file_id):
155
self._note('deleted %s', file_id)
157
def escaped(self, escape_count, message):
158
self._note("replaced %d control characters in message", escape_count)
160
def missing(self, path):
161
self._note('missing %s', path)
163
def renamed(self, change, old_path, new_path):
164
self._note('%s %s => %s', change, old_path, new_path)
166
def is_verbose(self):
89
170
class Commit(object):
99
180
working inventory.
101
182
def __init__(self,
103
if reporter is not None:
104
self.reporter = reporter
106
self.reporter = NullCommitReporter()
185
"""Create a Commit object.
187
:param reporter: the default reporter to use or None to decide later
189
self.reporter = reporter
114
197
specific_files=None,
116
allow_pointless=True):
199
allow_pointless=True,
207
message_callback=None,
210
possible_master_transports=None):
117
211
"""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.
213
:param message: the commit message (it or message_callback is required)
215
:param timestamp: if not None, seconds-since-epoch for a
216
postdated/predated commit.
218
:param specific_files: If true, commit only those files.
220
:param rev_id: If set, use this as the new revision id.
144
221
Useful for test or import commands that need to tightly
145
222
control what revisions are assigned. If you duplicate
146
223
a revision id that exists elsewhere it is your own fault.
147
224
If null (default), a time/random revision id is generated.
226
:param allow_pointless: If true (default), commit even if nothing
227
has changed and no merges are recorded.
229
:param strict: If true, don't allow a commit if the working tree
230
contains unknown files.
232
:param revprops: Properties for new revision
233
:param local: Perform a local only commit.
234
:param reporter: the reporter to use or None for the default
235
:param verbose: if True and the reporter is not None, report everything
236
:param recursive: If set to 'down', commit in any subtrees that have
237
pending changes of any sort during this commit.
238
:param exclude: None or a list of relative paths to exclude from the
239
commit. Pending changes to excluded files will be ignored by the
151
self.branch.lock_write()
153
self.specific_files = specific_files
242
mutter('preparing to commit')
244
if working_tree is None:
245
raise BzrError("working_tree must be passed into commit().")
247
self.work_tree = working_tree
248
self.branch = self.work_tree.branch
249
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
250
if not self.branch.repository.supports_rich_root():
251
raise errors.RootNotRich()
252
if message_callback is None:
253
if message is not None:
254
if isinstance(message, str):
255
message = message.decode(get_user_encoding())
256
message_callback = lambda x: message
258
raise BzrError("The message or message_callback keyword"
259
" parameter is required for commit().")
261
self.bound_branch = None
262
self.any_entries_deleted = False
263
if exclude is not None:
264
self.exclude = sorted(
265
minimum_path_selection(exclude))
269
self.master_branch = None
270
self.master_locked = False
271
self.recursive = recursive
273
if specific_files is not None:
274
self.specific_files = sorted(
275
minimum_path_selection(specific_files))
277
self.specific_files = None
278
self.specific_file_ids = None
154
279
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
280
self.revprops = revprops
281
self.message_callback = message_callback
282
self.timestamp = timestamp
283
self.timezone = timezone
284
self.committer = committer
286
self.verbose = verbose
288
self.work_tree.lock_write()
289
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
290
self.basis_revid = self.work_tree.last_revision()
291
self.basis_tree = self.work_tree.basis_tree()
292
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()
294
# Cannot commit with conflicts present.
295
if len(self.work_tree.conflicts()) > 0:
296
raise ConflictsInTree
298
# Setup the bound branch variables as needed.
299
self._check_bound_branch(possible_master_transports)
301
# Check that the working tree is up to date
302
old_revno, new_revno = self._check_out_of_date_tree()
304
# Complete configuration setup
305
if reporter is not None:
306
self.reporter = reporter
307
elif self.reporter is None:
308
self.reporter = self._select_reporter()
309
if self.config is None:
310
self.config = self.branch.get_config()
312
# If provided, ensure the specified files are versioned
313
if self.specific_files is not None:
314
# Note: This routine is being called because it raises
315
# PathNotVersionedError as a side effect of finding the IDs. We
316
# later use the ids we found as input to the working tree
317
# inventory iterator, so we only consider those ids rather than
318
# examining the whole tree again.
319
# XXX: Dont we have filter_unversioned to do this more
321
self.specific_file_ids = tree.find_ids_across_trees(
322
specific_files, [self.basis_tree, self.work_tree])
324
# Setup the progress bar. As the number of files that need to be
325
# committed in unknown, progress is reported as stages.
326
# We keep track of entries separately though and include that
327
# information in the progress bar during the relevant stages.
328
self.pb_stage_name = ""
329
self.pb_stage_count = 0
330
self.pb_stage_total = 5
331
if self.bound_branch:
332
self.pb_stage_total += 1
333
self.pb.show_pct = False
334
self.pb.show_spinner = False
335
self.pb.show_eta = False
336
self.pb.show_count = True
337
self.pb.show_bar = True
187
339
self.basis_inv = self.basis_tree.inventory
189
340
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([])
341
# After a merge, a selected file commit is not supported.
342
# See 'bzr help merge' for an explanation as to why.
343
if len(self.parents) > 1 and self.specific_files:
344
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
345
# Excludes are a form of selected file commit.
346
if len(self.parents) > 1 and self.exclude:
347
raise errors.CannotCommitSelectedFileMerge(self.exclude)
349
# Collect the changes
350
self._set_progress_stage("Collecting changes",
351
entries_title="Directory")
352
self.builder = self.branch.get_commit_builder(self.parents,
353
self.config, timestamp, timezone, committer, revprops, rev_id)
356
self.builder.will_record_deletes()
357
# find the location being committed to
358
if self.bound_branch:
359
master_location = self.master_branch.base
361
master_location = self.branch.base
363
# report the start of the commit
364
self.reporter.started(new_revno, self.rev_id, master_location)
366
self._update_builder_with_changes()
367
self._report_and_accumulate_deletes()
368
self._check_pointless()
370
# TODO: Now the new inventory is known, check for conflicts.
371
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
372
# weave lines, because nothing should be recorded until it is known
373
# that commit will succeed.
374
self._set_progress_stage("Saving data locally")
375
self.builder.finish_inventory()
377
# Prompt the user for a commit message if none provided
378
message = message_callback(self)
379
self.message = message
380
self._escape_commit_message()
382
# Add revision data to the local branch
383
self.rev_id = self.builder.commit(self.message)
386
mutter("aborting commit write group because of exception:")
387
trace.log_exception_quietly()
388
note("aborting commit write group: %r" % (e,))
392
self._process_pre_hooks(old_revno, new_revno)
394
# Upload revision data to the master.
395
# this will propagate merged revisions too if needed.
396
if self.bound_branch:
397
self._set_progress_stage("Uploading data to master branch")
398
# 'commit' to the master first so a timeout here causes the
399
# local branch to be out of date
400
self.master_branch.import_last_revision_info(
401
self.branch.repository, new_revno, self.rev_id)
403
# and now do the commit locally.
404
self.branch.set_last_revision_info(new_revno, self.rev_id)
406
# Make the working tree up to date with the branch
407
self._set_progress_stage("Updating the working tree")
408
self.work_tree.update_basis_by_delta(self.rev_id,
409
self.builder.get_basis_delta())
410
self.reporter.completed(new_revno, self.rev_id)
411
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."""
416
def _select_reporter(self):
417
"""Select the CommitReporter to use."""
419
return NullCommitReporter()
420
return ReportCommitToLog()
422
def _check_pointless(self):
423
if self.allow_pointless:
425
# A merge with no effect on files
236
426
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)
428
# TODO: we could simplify this by using self.builder.basis_delta.
430
# The initial commit adds a root directory, but this in itself is not
431
# a worthwhile commit.
432
if (self.basis_revid == revision.NULL_REVISION and
433
len(self.builder.new_inventory) == 1):
434
raise PointlessCommit()
435
# If length == 1, then we only have the root entry. Which means
436
# that there is no real difference (only the root could be different)
437
# unless deletes occured, in which case the length is irrelevant.
438
if (self.any_entries_deleted or
439
(len(self.builder.new_inventory) != 1 and
440
self.builder.any_changes())):
442
raise PointlessCommit()
444
def _check_bound_branch(self, possible_master_transports=None):
445
"""Check to see if the local branch is bound.
447
If it is bound, then most of the commit will actually be
448
done using the remote branch as the target branch.
449
Only at the end will the local branch be updated.
451
if self.local and not self.branch.get_bound_location():
452
raise errors.LocalRequiresBoundBranch()
455
self.master_branch = self.branch.get_master_branch(
456
possible_master_transports)
458
if not self.master_branch:
459
# make this branch the reference branch for out of date checks.
460
self.master_branch = self.branch
463
# If the master branch is bound, we must fail
464
master_bound_location = self.master_branch.get_bound_location()
465
if master_bound_location:
466
raise errors.CommitToDoubleBoundBranch(self.branch,
467
self.master_branch, master_bound_location)
469
# TODO: jam 20051230 We could automatically push local
470
# commits to the remote branch if they would fit.
471
# But for now, just require remote to be identical
474
# Make sure the local branch is identical to the master
475
master_info = self.master_branch.last_revision_info()
476
local_info = self.branch.last_revision_info()
477
if local_info != master_info:
478
raise errors.BoundBranchOutOfDate(self.branch,
481
# Now things are ready to change the master branch
483
self.bound_branch = self.branch
484
self.master_branch.lock_write()
485
self.master_locked = True
487
def _check_out_of_date_tree(self):
488
"""Check that the working tree is up to date.
490
:return: old_revision_number,new_revision_number tuple
493
first_tree_parent = self.work_tree.get_parent_ids()[0]
495
# if there are no parents, treat our parent as 'None'
496
# this is so that we still consider the master branch
497
# - in a checkout scenario the tree may have no
498
# parents but the branch may do.
499
first_tree_parent = bzrlib.revision.NULL_REVISION
500
old_revno, master_last = self.master_branch.last_revision_info()
501
if master_last != first_tree_parent:
502
if master_last != bzrlib.revision.NULL_REVISION:
503
raise errors.OutOfDateTree(self.work_tree)
504
if self.branch.repository.has_revision(first_tree_parent):
505
new_revno = old_revno + 1
507
# ghost parents never appear in revision history.
509
return old_revno,new_revno
511
def _process_pre_hooks(self, old_revno, new_revno):
512
"""Process any registered pre commit hooks."""
513
self._set_progress_stage("Running pre_commit hooks")
514
self._process_hooks("pre_commit", old_revno, new_revno)
516
def _process_post_hooks(self, old_revno, new_revno):
517
"""Process any registered post commit hooks."""
518
# Process the post commit hooks, if any
519
self._set_progress_stage("Running post_commit hooks")
520
# old style commit hooks - should be deprecated ? (obsoleted in
522
if self.config.post_commit() is not None:
523
hooks = self.config.post_commit().split(' ')
524
# this would be nicer with twisted.python.reflect.namedAny
526
result = eval(hook + '(branch, rev_id)',
527
{'branch':self.branch,
529
'rev_id':self.rev_id})
530
# process new style post commit hooks
531
self._process_hooks("post_commit", old_revno, new_revno)
533
def _process_hooks(self, hook_name, old_revno, new_revno):
534
if not Branch.hooks[hook_name]:
537
# new style commit hooks:
538
if not self.bound_branch:
539
hook_master = self.branch
542
hook_master = self.master_branch
543
hook_local = self.branch
544
# With bound branches, when the master is behind the local branch,
545
# the 'old_revno' and old_revid values here are incorrect.
546
# XXX: FIXME ^. RBC 20060206
240
lines = w.get(w.lookup(self.parents[0]))
548
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)
550
old_revid = bzrlib.revision.NULL_REVISION
552
if hook_name == "pre_commit":
553
future_tree = self.builder.revision_tree()
554
tree_delta = future_tree.changes_from(self.basis_tree,
557
for hook in Branch.hooks[hook_name]:
558
# show the running hook in the progress bar. As hooks may
559
# end up doing nothing (e.g. because they are not configured by
560
# the user) this is still showing progress, not showing overall
561
# actions - its up to each plugin to show a UI if it want's to
562
# (such as 'Emailing diff to foo@example.com').
563
self.pb_stage_name = "Running %s hooks [%s]" % \
564
(hook_name, Branch.hooks.get_hook_name(hook))
565
self._emit_progress()
566
if 'hooks' in debug.debug_flags:
567
mutter("Invoking commit hook: %r", hook)
568
if hook_name == "post_commit":
569
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
571
elif hook_name == "pre_commit":
572
hook(hook_local, hook_master,
573
old_revno, old_revid, new_revno, self.rev_id,
574
tree_delta, future_tree)
577
"""Cleanup any open locks, progress bars etc."""
578
cleanups = [self._cleanup_bound_branch,
579
self.basis_tree.unlock,
580
self.work_tree.unlock,
582
found_exception = None
583
for cleanup in cleanups:
586
# we want every cleanup to run no matter what.
587
# so we have a catchall here, but we will raise the
588
# last encountered exception up the stack: and
589
# typically this will be useful enough.
592
if found_exception is not None:
593
# don't do a plan raise, because the last exception may have been
594
# trashed, e is our sure-to-work exception even though it loses the
595
# full traceback. XXX: RBC 20060421 perhaps we could check the
596
# exc_info and if its the same one do a plain raise otherwise
597
# 'raise e' as we do now.
600
def _cleanup_bound_branch(self):
601
"""Executed at the end of a try/finally to cleanup a bound branch.
603
If the branch wasn't bound, this is a no-op.
604
If it was, it resents self.branch to the local branch, instead
607
if not self.bound_branch:
609
if self.master_locked:
610
self.master_branch.unlock()
612
def _escape_commit_message(self):
613
"""Replace xml-incompatible control characters."""
614
# FIXME: RBC 20060419 this should be done by the revision
615
# serialiser not by commit. Then we can also add an unescaper
616
# in the deserializer and start roundtripping revision messages
617
# precisely. See repository_implementations/test_repository.py
619
# Python strings can include characters that can't be
620
# represented in well-formed XML; escape characters that
621
# aren't listed in the XML specification
622
# (http://www.w3.org/TR/REC-xml/#NT-Char).
623
self.message, escape_count = re.subn(
624
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
625
lambda match: match.group(0).encode('unicode_escape'),
628
self.reporter.escaped(escape_count, self.message)
249
630
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)
631
"""Record the parents of a merge for merge detection."""
632
# TODO: Make sure that this list doesn't contain duplicate
633
# entries and the order is preserved when doing this.
634
self.parents = self.work_tree.get_parent_ids()
635
self.parent_invs = [self.basis_inv]
636
for revision in self.parents[1:]:
637
if self.branch.repository.has_revision(revision):
638
mutter('commit parent revision {%s}', revision)
639
inventory = self.branch.repository.get_inventory(revision)
640
self.parent_invs.append(inventory)
642
mutter('commit parent ghost revision {%s}', revision)
644
def _update_builder_with_changes(self):
645
"""Update the commit builder with the data about what has changed.
647
# Build the revision inventory.
649
# This starts by creating a new empty inventory. Depending on
650
# which files are selected for commit, and what is present in the
651
# current tree, the new inventory is populated. inventory entries
652
# which are candidates for modification have their revision set to
653
# None; inventory entries that are carried over untouched have their
654
# revision set to their prior value.
656
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
657
# results to create a new inventory at the same time, which results
658
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
661
exclude = self.exclude
662
specific_files = self.specific_files or []
663
mutter("Selecting files for commit with filter %s", specific_files)
665
# Build the new inventory
666
self._populate_from_inventory()
668
# If specific files are selected, then all un-selected files must be
669
# recorded in their previous state. For more details, see
670
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
671
if specific_files or exclude:
672
for path, old_ie in self.basis_inv.iter_entries():
673
if old_ie.file_id in self.builder.new_inventory:
674
# already added - skip.
676
if (is_inside_any(specific_files, path)
677
and not is_inside_any(exclude, path)):
678
# was inside the selected path, and not excluded - if not
679
# present it has been deleted so skip.
681
# From here down it was either not selected, or was excluded:
682
if old_ie.kind == 'directory':
683
self._next_progress_entry()
684
# We preserve the entry unaltered.
686
# Note: specific file commits after a merge are currently
687
# prohibited. This test is for sanity/safety in case it's
688
# required after that changes.
689
if len(self.parents) > 1:
691
self.builder.record_entry_contents(ie, self.parent_invs, path,
692
self.basis_tree, None)
694
def _report_and_accumulate_deletes(self):
695
# XXX: Could the list of deleted paths and ids be instead taken from
696
# _populate_from_inventory?
697
if (isinstance(self.basis_inv, Inventory)
698
and isinstance(self.builder.new_inventory, Inventory)):
699
# the older Inventory classes provide a _byid dict, and building a
700
# set from the keys of this dict is substantially faster than even
701
# getting a set of ids from the inventory
703
# <lifeless> set(dict) is roughly the same speed as
704
# set(iter(dict)) and both are significantly slower than
706
deleted_ids = set(self.basis_inv._byid.keys()) - \
707
set(self.builder.new_inventory._byid.keys())
709
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
711
self.any_entries_deleted = True
712
deleted = [(self.basis_tree.id2path(file_id), file_id)
713
for file_id in deleted_ids]
715
# XXX: this is not quite directory-order sorting
716
for path, file_id in deleted:
717
self.builder.record_delete(path, file_id)
718
self.reporter.deleted(path)
720
def _populate_from_inventory(self):
721
"""Populate the CommitBuilder by walking the working tree inventory."""
723
# raise an exception as soon as we find a single unknown.
724
for unknown in self.work_tree.unknowns():
725
raise StrictCommitFailed()
727
specific_files = self.specific_files
728
exclude = self.exclude
729
report_changes = self.reporter.is_verbose()
731
# A tree of paths that have been deleted. E.g. if foo/bar has been
732
# deleted, then we have {'foo':{'bar':{}}}
734
# XXX: Note that entries may have the wrong kind because the entry does
735
# not reflect the status on disk.
736
work_inv = self.work_tree.inventory
737
# NB: entries will include entries within the excluded ids/paths
738
# because iter_entries_by_dir has no 'exclude' facility today.
739
entries = work_inv.iter_entries_by_dir(
740
specific_file_ids=self.specific_file_ids, yield_parents=True)
741
for path, existing_ie in entries:
742
file_id = existing_ie.file_id
743
name = existing_ie.name
744
parent_id = existing_ie.parent_id
745
kind = existing_ie.kind
746
if kind == 'directory':
747
self._next_progress_entry()
748
# Skip files that have been deleted from the working tree.
749
# The deleted path ids are also recorded so they can be explicitly
752
path_segments = splitpath(path)
753
deleted_dict = deleted_paths
754
for segment in path_segments:
755
deleted_dict = deleted_dict.get(segment, None)
757
# We either took a path not present in the dict
758
# (deleted_dict was None), or we've reached an empty
759
# child dir in the dict, so are now a sub-path.
763
if deleted_dict is not None:
764
# the path has a deleted parent, do not add it.
766
if exclude and is_inside_any(exclude, path):
767
# Skip excluded paths. Excluded paths are processed by
768
# _update_builder_with_changes.
770
content_summary = self.work_tree.path_content_summary(path)
771
# Note that when a filter of specific files is given, we must only
772
# skip/record deleted files matching that filter.
773
if not specific_files or is_inside_any(specific_files, path):
774
if content_summary[0] == 'missing':
775
if not deleted_paths:
776
# path won't have been split yet.
777
path_segments = splitpath(path)
778
deleted_dict = deleted_paths
779
for segment in path_segments:
780
deleted_dict = deleted_dict.setdefault(segment, {})
781
self.reporter.missing(path)
782
deleted_ids.append(file_id)
784
# TODO: have the builder do the nested commit just-in-time IF and
786
if content_summary[0] == 'tree-reference':
787
# enforce repository nested tree policy.
788
if (not self.work_tree.supports_tree_reference() or
789
# repository does not support it either.
790
not self.branch.repository._format.supports_tree_reference):
791
content_summary = ('directory',) + content_summary[1:]
792
kind = content_summary[0]
793
# TODO: specific_files filtering before nested tree processing
794
if kind == 'tree-reference':
795
if self.recursive == 'down':
796
nested_revision_id = self._commit_nested_tree(
798
content_summary = content_summary[:3] + (
801
content_summary = content_summary[:3] + (
802
self.work_tree.get_reference_revision(file_id),)
804
# Record an entry for this item
805
# Note: I don't particularly want to have the existing_ie
806
# parameter but the test suite currently (28-Jun-07) breaks
807
# without it thanks to a unicode normalisation issue. :-(
808
definitely_changed = kind != existing_ie.kind
809
self._record_entry(path, file_id, specific_files, kind, name,
810
parent_id, definitely_changed, existing_ie, report_changes,
813
# Unversion IDs that were found to be deleted
814
self.work_tree.unversion(deleted_ids)
816
def _commit_nested_tree(self, file_id, path):
817
"Commit a nested tree."
818
sub_tree = self.work_tree.get_nested_tree(file_id, path)
819
# FIXME: be more comprehensive here:
820
# this works when both trees are in --trees repository,
821
# but when both are bound to a different repository,
822
# it fails; a better way of approaching this is to
823
# finally implement the explicit-caches approach design
824
# a while back - RBC 20070306.
825
if sub_tree.branch.repository.has_same_location(
826
self.work_tree.branch.repository):
827
sub_tree.branch.repository = \
828
self.work_tree.branch.repository
830
return sub_tree.commit(message=None, revprops=self.revprops,
831
recursive=self.recursive,
832
message_callback=self.message_callback,
833
timestamp=self.timestamp, timezone=self.timezone,
834
committer=self.committer,
835
allow_pointless=self.allow_pointless,
836
strict=self.strict, verbose=self.verbose,
837
local=self.local, reporter=self.reporter)
838
except errors.PointlessCommit:
839
return self.work_tree.get_reference_revision(file_id)
841
def _record_entry(self, path, file_id, specific_files, kind, name,
842
parent_id, definitely_changed, existing_ie, report_changes,
844
"Record the new inventory entry for a path if any."
845
# mutter('check %s {%s}', path, file_id)
846
# mutter('%s selected for commit', path)
847
if definitely_changed or existing_ie is None:
848
ie = make_entry(kind, name, parent_id, file_id)
850
ie = existing_ie.copy()
852
# For carried over entries we don't care about the fs hash - the repo
853
# isn't generating a sha, so we're not saving computation time.
854
_, _, fs_hash = self.builder.record_entry_contents(
855
ie, self.parent_invs, path, self.work_tree, content_summary)
857
self._report_change(ie, path)
859
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
862
def _report_change(self, ie, path):
863
"""Report a change to the user.
865
The change that has occurred is described relative to the basis
868
if (self.basis_inv.has_id(ie.file_id)):
869
basis_ie = self.basis_inv[ie.file_id]
872
change = ie.describe_change(basis_ie, ie)
873
if change in (InventoryEntry.RENAMED,
874
InventoryEntry.MODIFIED_AND_RENAMED):
875
old_path = self.basis_inv.id2path(ie.file_id)
876
self.reporter.renamed(change, old_path, path)
878
self.reporter.snapshot_change(change, path)
880
def _set_progress_stage(self, name, entries_title=None):
881
"""Set the progress stage and emit an update to the progress bar."""
882
self.pb_stage_name = name
883
self.pb_stage_count += 1
884
self.pb_entries_title = entries_title
885
if entries_title is not None:
886
self.pb_entries_count = 0
887
self.pb_entries_total = '?'
888
self._emit_progress()
890
def _next_progress_entry(self):
891
"""Emit an update to the progress bar and increment the entry count."""
892
self.pb_entries_count += 1
893
self._emit_progress()
895
def _emit_progress(self):
896
if self.pb_entries_title:
897
if self.pb_entries_total == '?':
898
text = "%s [%s %d] - Stage" % (self.pb_stage_name,
899
self.pb_entries_title, self.pb_entries_count)
901
text = "%s [%s %d/%s] - Stage" % (self.pb_stage_name,
902
self.pb_entries_title, self.pb_entries_count,
903
str(self.pb_entries_total))
905
text = "%s - Stage" % (self.pb_stage_name)
906
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
added
removed
bzrlib/commit.py
