1
# Copyright (C) 2005 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
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
19
# nothing? Perhaps the best approach is to integrate commit of
20
# AtomicFiles with releasing the lock on the Branch.
22
# TODO: Separate 'prepare' phase where we find a list of potentially
23
# committed files. We then can then pause the commit to prompt for a
24
# commit message, knowing the summary will be the same as what's
25
# actually used for the commit. (But perhaps simpler to simply get
26
# the tree status, then use that for a selective commit?)
28
# The newly committed revision is going to have a shape corresponding
29
# to that of the working inventory. Files that are not in the
30
# working tree and that were in the predecessor are reported as
31
# removed --- this can include files that were either removed from the
32
# inventory or deleted in the working tree. If they were only
33
# deleted from disk, they are removed from the working inventory.
35
# We then consider the remaining entries, which will be in the new
36
# version. Directory entries are simply copied across. File entries
37
# must be checked to see if a new version of the file should be
38
# recorded. For each parent revision inventory, we check to see what
39
# version of the file was present. If the file was present in at
40
# least one tree, and if it was the same version in all the trees,
41
# then we can just refer to that version. Otherwise, a new version
42
# representing the merger of the file versions must be added.
44
# TODO: Update hashcache before and after - or does the WorkingTree
47
# This code requires all merge parents to be present in the branch.
48
# We could relax this but for the sake of simplicity the constraint is
49
# here for now. It's not totally clear to me how we'd know which file
50
# need new text versions if some parents are absent. -- mbp 20050915
52
# TODO: Rather than mashing together the ancestry and storing it back,
53
# perhaps the weave should have single method which does it all in one
54
# go, avoiding a lot of redundant work.
56
# TODO: Perhaps give a warning if one of the revisions marked as
57
# merged is already in the ancestry, and then don't record it as a
60
# TODO: If the file is newly merged but unchanged from the version it
61
# merges from, then it should still be reported as newly added
62
# relative to the basis revision.
71
from binascii import hexlify
72
from cStringIO import StringIO
74
from bzrlib.osutils import (local_time_offset, username,
75
rand_bytes, compact_date, user_email,
76
kind_marker, is_inside_any, quotefn,
77
sha_string, sha_strings, sha_file, isdir, isfile,
79
from bzrlib.branch import gen_file_id
80
from bzrlib.errors import (BzrError, PointlessCommit,
83
from bzrlib.revision import Revision
84
from bzrlib.trace import mutter, note, warning
85
from bzrlib.xml5 import serializer_v5
86
from bzrlib.inventory import Inventory
87
from bzrlib.weave import Weave
88
from bzrlib.weavefile import read_weave, write_weave_v5
89
from bzrlib.atomicfile import AtomicFile
92
def commit(*args, **kwargs):
93
"""Commit a new revision to a branch.
95
Function-style interface for convenience of old callers.
97
New code should use the Commit class instead.
99
## XXX: Remove this in favor of Branch.commit?
100
Commit().commit(*args, **kwargs)
103
class NullCommitReporter(object):
104
"""I report on progress of a commit."""
105
def added(self, path):
108
def removed(self, path):
111
def renamed(self, old_path, new_path):
115
class ReportCommitToLog(NullCommitReporter):
116
def added(self, path):
117
note('added %s', path)
119
def removed(self, path):
120
note('removed %s', path)
122
def renamed(self, old_path, new_path):
123
note('renamed %s => %s', old_path, new_path)
126
class Commit(object):
127
"""Task of committing a new revision.
129
This is a MethodObject: it accumulates state as the commit is
130
prepared, and then it is discarded. It doesn't represent
131
historical revisions, just the act of recording a new one.
134
Modified to hold a list of files that have been deleted from
135
the working directory; these should be removed from the
140
if reporter is not None:
141
self.reporter = reporter
143
self.reporter = NullCommitReporter()
153
allow_pointless=True,
155
"""Commit working copy as a new revision.
157
timestamp -- if not None, seconds-since-epoch for a
158
postdated/predated commit.
160
specific_files -- If true, commit only those files.
162
rev_id -- If set, use this as the new revision id.
163
Useful for test or import commands that need to tightly
164
control what revisions are assigned. If you duplicate
165
a revision id that exists elsewhere it is your own fault.
166
If null (default), a time/random revision id is generated.
168
allow_pointless -- If true (default), commit even if nothing
169
has changed and no merges are recorded.
171
mutter('preparing to commit')
174
self.weave_store = branch.weave_store
176
self.specific_files = specific_files
177
self.allow_pointless = allow_pointless
179
if timestamp is None:
180
self.timestamp = time.time()
182
self.timestamp = long(timestamp)
185
self.rev_id = _gen_revision_id(self.branch, self.timestamp)
189
if committer is None:
190
self.committer = username(self.branch)
192
assert isinstance(committer, basestring), type(committer)
193
self.committer = committer
196
self.timezone = local_time_offset()
198
self.timezone = int(timezone)
200
assert isinstance(message, basestring), type(message)
201
self.message = message
202
self._escape_commit_message()
204
self.branch.lock_write()
206
self.work_tree = self.branch.working_tree()
207
self.work_inv = self.work_tree.inventory
208
self.basis_tree = self.branch.basis_tree()
209
self.basis_inv = self.basis_tree.inventory
211
self._gather_parents()
212
if len(self.parents) > 1 and self.specific_files:
213
raise NotImplementedError('selected-file commit of merges is not supported yet')
214
self._check_parents_present()
216
self._remove_deleted()
217
self._populate_new_inv()
218
self._store_snapshot()
219
self._report_deletes()
221
if not (self.allow_pointless
222
or len(self.parents) > 1
223
or self.new_inv != self.basis_inv):
224
raise PointlessCommit()
226
self._record_inventory()
227
self._record_ancestry()
228
self._make_revision()
229
note('committed r%d {%s}', (self.branch.revno() + 1),
231
self.branch.append_revision(self.rev_id)
232
self.branch.set_pending_merges([])
238
def _record_inventory(self):
239
"""Store the inventory for the new revision."""
240
inv_text = serializer_v5.write_inventory_to_string(self.new_inv)
241
self.inv_sha1 = sha_string(inv_text)
242
s = self.branch.control_weaves
243
s.add_text('inventory', self.rev_id,
244
split_lines(inv_text), self.parents)
246
def _escape_commit_message(self):
247
"""Replace xml-incompatible control characters."""
248
# Python strings can include characters that can't be
249
# represented in well-formed XML; escape characters that
250
# aren't listed in the XML specification
251
# (http://www.w3.org/TR/REC-xml/#NT-Char).
252
if isinstance(self.message, unicode):
253
char_pattern = u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]'
255
# Use a regular 'str' as pattern to avoid having re.subn
256
# return 'unicode' results.
257
char_pattern = '[^x09\x0A\x0D\x20-\xFF]'
258
self.message, escape_count = re.subn(
260
lambda match: match.group(0).encode('unicode_escape'),
263
note("replaced %d control characters in message", escape_count)
265
def _record_ancestry(self):
266
"""Append merged revision ancestry to the ancestry file.
268
This should be the merged ancestry of all parents, plus the
270
s = self.branch.control_weaves
271
w = s.get_weave_or_empty('ancestry')
272
lines = self._make_ancestry(w)
273
w.add(self.rev_id, self.parents, lines)
274
s.put_weave('ancestry', w)
277
def _make_ancestry(self, ancestry_weave):
278
"""Return merged ancestry lines.
280
The lines are revision-ids followed by newlines."""
281
parent_ancestries = [ancestry_weave.get(p) for p in self.parents]
282
new_lines = merge_ancestry_lines(self.rev_id, parent_ancestries)
283
mutter('merged ancestry of {%s}:\n%s', self.rev_id, ''.join(new_lines))
287
def _gather_parents(self):
288
pending_merges = self.branch.pending_merges()
290
self.parent_trees = []
291
precursor_id = self.branch.last_revision()
293
self.parents.append(precursor_id)
294
self.parent_trees.append(self.basis_tree)
295
self.parents += pending_merges
296
self.parent_trees.extend(map(self.branch.revision_tree, pending_merges))
299
def _check_parents_present(self):
300
for parent_id in self.parents:
301
mutter('commit parent revision {%s}', parent_id)
302
if not self.branch.has_revision(parent_id):
303
warning("can't commit a merge from an absent parent")
304
raise HistoryMissing(self.branch, 'revision', parent_id)
307
def _make_revision(self):
308
"""Record a new revision object for this commit."""
309
self.rev = Revision(timestamp=self.timestamp,
310
timezone=self.timezone,
311
committer=self.committer,
312
message=self.message,
313
inventory_sha1=self.inv_sha1,
314
revision_id=self.rev_id)
315
self.rev.parent_ids = self.parents
317
serializer_v5.write_revision(self.rev, rev_tmp)
319
self.branch.revision_store.add(rev_tmp, self.rev_id, compressed=True)
320
mutter('new revision_id is {%s}', self.rev_id)
323
def _remove_deleted(self):
324
"""Remove deleted files from the working inventories.
326
This is done prior to taking the working inventory as the
327
basis for the new committed inventory.
329
This returns true if any files
330
*that existed in the basis inventory* were deleted.
331
Files that were added and deleted
332
in the working copy don't matter.
334
specific = self.specific_files
336
for path, ie in self.work_inv.iter_entries():
337
if specific and not is_inside_any(specific, path):
339
if not self.work_tree.has_filename(path):
340
note('missing %s', path)
341
deleted_ids.append(ie.file_id)
343
for file_id in deleted_ids:
344
del self.work_inv[file_id]
345
self.branch._write_inventory(self.work_inv)
348
def _find_entry_parents(self, file_id):
349
"""Return the text versions and hashes for all file parents.
351
Returned as a map from text version to inventory entry.
353
This is a set containing the file versions in all parents
354
revisions containing the file. If the file is new, the set
357
for tree in self.parent_trees:
358
if file_id in tree.inventory:
359
ie = tree.inventory[file_id]
360
assert ie.file_id == file_id
362
assert r[ie.revision] == ie
367
def _snapshot_entry(self, path, ie, previous_entries):
368
"""Store a single possibly changed inventory entry in the branch."""
369
mutter('parents of %s are %r', path, previous_entries)
370
if ie.revision is not None:
371
# not selected for commit
373
if ie.kind == 'symlink':
374
ie.read_symlink_target(self.branch.abspath(path))
375
if len(previous_entries) == 1:
376
# cannot be unchanged unless there is only one parent file rev.
377
parent_ie = previous_entries.values()[0]
378
unchanged = ie.unchanged(parent_ie, self.work_tree)
380
mutter("found unchanged entry")
381
ie.revision = parent_ie.revision
382
self.report_entry_status(previous_entries, path, ie)
384
mutter('new revision for {%s}', ie.file_id)
385
ie.revision = self.rev_id
386
if ie.kind != 'file':
387
self.report_entry_status(previous_entries, path, ie)
389
# file is either new, or a file merge; need to record
391
self.report_entry_status(previous_entries, path, ie)
393
self._commit_file(ie, previous_entries)
395
def report_entry_status(self, previous_entries, path, ie):
396
if len(previous_entries) > 1:
397
note('merged %s', path)
398
elif len(previous_entries) == 0:
399
note('added %s', path)
400
elif ie.revision == self.rev_id:
401
note('modified/renamed/reparented%s', path)
403
note('unchanged %s', path)
405
def _store_snapshot(self):
406
"""Pass over inventory and record a snapshot.
408
Entries get a new revision when they are modified in
409
any way, which includes a merge with a new set of
410
parents that have the same entry. Currently we do not
411
check for that set being ancestors of each other - and
412
we should - only parallel children should count for this
413
test see find_entry_parents to correct this. FIXME <---
414
I.e. if we are merging in revision FOO, and our
415
copy of file id BAR is identical to FOO.BAR, we should
416
generate a new revision of BAR IF and only IF FOO is
417
neither a child of our current tip, nor an ancestor of
418
our tip. The presence of FOO in our store should not
419
affect this logic UNLESS we are doing a merge of FOO,
422
# XXX: Need to think more here about when the user has
423
# made a specific decision on a particular value -- c.f.
425
for path, ie in self.new_inv.iter_entries():
426
previous_entries = self._find_entry_parents(ie. file_id)
427
self._snapshot_entry(path, ie, previous_entries)
429
def _populate_new_inv(self):
430
"""Build revision inventory.
432
This creates a new empty inventory. Depending on
433
which files are selected for commit, and what is present in the
434
current tree, the new inventory is populated. inventory entries
435
which are candidates for modification have their revision set to
436
None; inventory entries that are carried over untouched have their
437
revision set to their prior value.
439
mutter("Selecting files for commit with filter %s", self.specific_files)
440
self.new_inv = Inventory()
441
for path, new_ie in self.work_inv.iter_entries():
442
file_id = new_ie.file_id
443
mutter('check %s {%s}', path, new_ie.file_id)
444
if self.specific_files:
445
if not is_inside_any(self.specific_files, path):
446
mutter('%s not selected for commit', path)
447
self._carry_file(file_id)
449
mutter('%s selected for commit', path)
454
def _carry_file(self, file_id):
455
"""Carry the file unchanged from the basis revision."""
456
if self.basis_inv.has_id(file_id):
457
self.new_inv.add(self.basis_inv[file_id].copy())
459
def _report_deletes(self):
460
for file_id in self.basis_inv:
461
if file_id not in self.new_inv:
462
note('deleted %s', self.basis_inv.id2path(file_id))
464
def _commit_file(self, new_ie, file_parents):
465
mutter('storing file {%s} in revision {%s}',
466
new_ie.file_id, new_ie.revision)
467
# special case to avoid diffing on renames or
469
if (len(file_parents) == 1
470
and new_ie.text_sha1 == file_parents.values()[0].text_sha1
471
and new_ie.text_size == file_parents.values()[0].text_size):
472
previous_ie = file_parents.values()[0]
473
self.weave_store.add_identical_text(
474
new_ie.file_id, previous_ie.revision,
475
new_ie.revision, file_parents)
477
new_lines = self.work_tree.get_file(new_ie.file_id).readlines()
478
self._add_text_to_weave(new_ie.file_id, new_lines, file_parents)
479
new_ie.text_sha1 = sha_strings(new_lines)
480
new_ie.text_size = sum(map(len, new_lines))
482
def _add_text_to_weave(self, file_id, new_lines, parents):
483
self.weave_store.add_text(file_id, self.rev_id, new_lines, parents)
485
def _gen_revision_id(branch, when):
486
"""Return new revision-id."""
487
s = '%s-%s-' % (user_email(branch), compact_date(when))
488
s += hexlify(rand_bytes(8))
494
def merge_ancestry_lines(rev_id, ancestries):
495
"""Return merged ancestry lines.
497
rev_id -- id of the new revision
499
ancestries -- a sequence of ancestries for parent revisions,
500
as newline-terminated line lists.
502
if len(ancestries) == 0:
503
return [rev_id + '\n']
504
seen = set(ancestries[0])
505
ancs = ancestries[0][:]
506
for parent_ancestry in ancestries[1:]:
507
for line in parent_ancestry:
508
assert line[-1] == '\n'
added
removed
bzrlib/commit.py
