« back to all changes in this revision
Viewing changes to bzrlib/commit.py
- browse files at revision 1219
- compare with another revision
- download diff
- download tarball
- view history from revision 1219
- Committer: Martin Pool
- Date: 2005-09-13 01:37:23 UTC
- Revision ID: mbp@sourcefrog.net-20050913013723-7e0026b48cbf08ff
- BROKEN: start refactoring fetch code to work well with weaves
- files removed:
- bzrlib/selftest/test_ancestry.py
- files renamed:
- bzrlib/selftest/test_weave.py => tools/testweave.py
- files modified:
- .bzrignore
added
removed
bzrlib/commit.py
1
1
# Copyright (C) 2005 Canonical Ltd
2
#
2
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
5
5
# the Free Software Foundation; either version 2 of the License, or
6
6
# (at your option) any later version.
7
#
7
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.
12
#
12
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
16
16
17
17
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.
21
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?)
27
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.
34
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.
43
44
# TODO: Update hashcache before and after - or does the WorkingTree
45
# look after that?
46
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
18
## XXX: Can we do any better about making interrupted commits change
19
## nothing?
20
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.
24
25
## TODO: Read back the just-generated changeset, and make sure it
26
## applies and recreates the right state.
27
28
51
29
52
30
53
31
import os
54
32
import sys
55
33
import time
56
import pdb
34
import tempfile
35
import sha
57
36
58
37
from binascii import hexlify
59
38
from cStringIO import StringIO
61
40
from bzrlib.osutils import (local_time_offset, username,
62
41
rand_bytes, compact_date, user_email,
63
42
kind_marker, is_inside_any, quotefn,
64
sha_string, sha_strings, sha_file, isdir, isfile,
65
split_lines)
66
from bzrlib.branch import gen_file_id, INVENTORY_FILEID, ANCESTRY_FILEID
67
from bzrlib.errors import (BzrError, PointlessCommit,
68
HistoryMissing,
69
)
43
sha_string, sha_file, isdir, isfile)
44
from bzrlib.branch import gen_file_id
45
from bzrlib.errors import BzrError, PointlessCommit
70
46
from bzrlib.revision import Revision, RevisionReference
71
from bzrlib.trace import mutter, note, warning
47
from bzrlib.trace import mutter, note
72
48
from bzrlib.xml5 import serializer_v5
73
49
from bzrlib.inventory import Inventory
50
from bzrlib.delta import compare_trees
74
51
from bzrlib.weave import Weave
75
52
from bzrlib.weavefile import read_weave, write_weave_v5
76
53
from bzrlib.atomicfile import AtomicFile
83
60
84
61
New code should use the Commit class instead.
85
62
"""
86
## XXX: Remove this in favor of Branch.commit?
87
63
Commit().commit(*args, **kwargs)
88
64
89
65
137
113
committer=None,
138
114
specific_files=None,
139
115
rev_id=None,
140
allow_pointless=True,
141
verbose=False):
116
allow_pointless=True):
142
117
"""Commit working copy as a new revision.
143
118
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.
122
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.
132
133
This raises PointlessCommit if there are no changes, no new merges,
134
and allow_pointless is false.
135
144
136
timestamp -- if not None, seconds-since-epoch for a
145
137
postdated/predated commit.
146
138
147
specific_files -- If true, commit only those files.
139
specific_files
140
If true, commit only those files.
148
141
149
rev_id -- If set, use this as the new revision id.
142
rev_id
143
If set, use this as the new revision id.
150
144
Useful for test or import commands that need to tightly
151
145
control what revisions are assigned. If you duplicate
152
146
a revision id that exists elsewhere it is your own fault.
153
147
If null (default), a time/random revision id is generated.
154
155
allow_pointless -- If true (default), commit even if nothing
156
has changed and no merges are recorded.
157
148
"""
158
mutter('preparing to commit')
159
149
160
150
self.branch = branch
161
self.weave_store = branch.weave_store
151
self.branch.lock_write()
162
152
self.rev_id = rev_id
163
153
self.specific_files = specific_files
164
154
self.allow_pointless = allow_pointless
168
158
else:
169
159
self.timestamp = long(timestamp)
170
160
171
if rev_id is None:
172
self.rev_id = _gen_revision_id(self.branch, self.timestamp)
173
else:
174
self.rev_id = rev_id
175
176
161
if committer is None:
177
162
self.committer = username(self.branch)
178
163
else:
187
172
assert isinstance(message, basestring), type(message)
188
173
self.message = message
189
174
190
self.branch.lock_write()
191
175
try:
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
182
# working inventory.
183
192
184
self.work_tree = self.branch.working_tree()
193
185
self.work_inv = self.work_tree.inventory
194
186
self.basis_tree = self.branch.basis_tree()
195
187
self.basis_inv = self.basis_tree.inventory
196
188
197
self._gather_parents()
198
self._check_parents_present()
189
self.pending_merges = self.branch.pending_merges()
190
if self.pending_merges:
191
raise NotImplementedError("sorry, can't commit merges to the weave format yet")
199
192
193
if self.rev_id is None:
194
self.rev_id = _gen_revision_id(self.branch, time.time())
195
196
# todo: update hashcache
197
self.delta = compare_trees(self.basis_tree, self.work_tree,
198
specific_files=self.specific_files)
199
200
if not (self.delta.has_changed()
201
or self.allow_pointless
202
or self.pending_merges):
203
raise PointlessCommit()
204
205
self.new_inv = self.basis_inv.copy()
206
207
self.delta.show(sys.stdout)
208
200
209
self._remove_deleted()
201
self.new_inv = Inventory()
202
self._store_entries()
203
self._report_deletes()
204
205
if not (self.allow_pointless
206
or len(self.parents) > 1
207
or self.new_inv != self.basis_inv):
208
raise PointlessCommit()
209
210
self._store_files()
211
212
self.branch._write_inventory(self.work_inv)
210
213
self._record_inventory()
211
self._record_ancestry()
214
212
215
self._make_revision()
213
note('committed r%d {%s}', (self.branch.revno() + 1),
216
note('committted r%d {%s}', (self.branch.revno() + 1),
214
217
self.rev_id)
215
218
self.branch.append_revision(self.rev_id)
216
219
self.branch.set_pending_merges([])
218
221
self.branch.unlock()
219
222
220
223
221
222
224
def _record_inventory(self):
223
"""Store the inventory for the new revision."""
224
inv_text = serializer_v5.write_inventory_to_string(self.new_inv)
225
self.inv_sha1 = sha_string(inv_text)
226
self.weave_store.add_text(INVENTORY_FILEID, self.rev_id,
227
split_lines(inv_text), self.parents)
228
229
230
def _record_ancestry(self):
231
"""Append merged revision ancestry to the ancestry file.
232
233
This should be the merged ancestry of all parents, plus the
234
new revision id."""
235
w = self.weave_store.get_weave_or_empty(ANCESTRY_FILEID)
236
lines = self._merge_ancestry_lines(w)
237
w.add(self.rev_id, self.parents, lines)
238
self.weave_store.put_weave(ANCESTRY_FILEID, w)
239
240
241
def _merge_ancestry_lines(self, ancestry_weave):
242
"""Return merged ancestry lines.
243
244
The lines are revision-ids followed by newlines."""
245
seen = set()
246
ancs = []
247
for parent_id in self.parents:
248
for line in ancestry_weave.get(parent_id):
249
assert line[-1] == '\n'
250
if line not in seen:
251
ancs.append(line)
252
seen.add(line)
253
r = self.rev_id + '\n'
254
assert r not in ancs
255
ancs.append(r)
256
mutter('merged ancestry of {%s}:\n%s', self.rev_id, ''.join(ancs))
257
return ancs
258
259
260
def _gather_parents(self):
261
pending_merges = self.branch.pending_merges()
262
self.parents = []
263
self.parent_trees = []
264
precursor_id = self.branch.last_revision()
265
if precursor_id:
266
self.parents.append(precursor_id)
267
self.parent_trees.append(self.basis_tree)
268
self.parents += pending_merges
269
self.parent_trees.extend(map(self.branch.revision_tree, pending_merges))
270
271
272
def _check_parents_present(self):
273
for parent_id in self.parents:
274
mutter('commit parent revision {%s}', parent_id)
275
if not self.branch.has_revision(parent_id):
276
warning("can't commit a merge from an absent parent")
277
raise HistoryMissing(self.branch, 'revision', parent_id)
278
279
225
inv_tmp = StringIO()
226
serializer_v5.write_inventory(self.new_inv, inv_tmp)
227
self.inv_sha1 = sha_string(inv_tmp.getvalue())
228
inv_tmp.seek(0)
229
self.branch.inventory_store.add(inv_tmp, self.rev_id)
230
231
280
232
def _make_revision(self):
281
233
"""Record a new revision object for this commit."""
282
234
self.rev = Revision(timestamp=self.timestamp,
285
237
message=self.message,
286
238
inventory_sha1=self.inv_sha1,
287
239
revision_id=self.rev_id)
288
self.rev.parents = map(RevisionReference, self.parents)
289
rev_tmp = StringIO()
240
241
self.rev.parents = []
242
precursor_id = self.branch.last_patch()
243
if precursor_id:
244
self.rev.parents.append(RevisionReference(precursor_id))
245
for merge_rev in self.pending_merges:
246
rev.parents.append(RevisionReference(merge_rev))
247
248
rev_tmp = tempfile.TemporaryFile()
290
249
serializer_v5.write_revision(self.rev, rev_tmp)
291
250
rev_tmp.seek(0)
292
251
self.branch.revision_store.add(rev_tmp, self.rev_id)
294
253
295
254
296
255
def _remove_deleted(self):
297
"""Remove deleted files from the working inventories.
298
299
This is done prior to taking the working inventory as the
300
basis for the new committed inventory.
301
302
This returns true if any files
303
*that existed in the basis inventory* were deleted.
304
Files that were added and deleted
305
in the working copy don't matter.
306
"""
307
specific = self.specific_files
308
deleted_ids = []
309
for path, ie in self.work_inv.iter_entries():
310
if specific and not is_inside_any(specific, path):
311
continue
312
if not self.work_tree.has_filename(path):
313
note('missing %s', path)
314
deleted_ids.append(ie.file_id)
315
if deleted_ids:
316
for file_id in deleted_ids:
317
del self.work_inv[file_id]
318
self.branch._write_inventory(self.work_inv)
319
320
321
def _find_file_parents(self, file_id):
322
"""Return the text versions and hashes for all file parents.
323
324
Returned as a map from text version to inventory entry.
325
326
This is a set containing the file versions in all parents
327
revisions containing the file. If the file is new, the set
328
will be empty."""
329
r = {}
330
for tree in self.parent_trees:
331
if file_id in tree.inventory:
332
ie = tree.inventory[file_id]
333
assert ie.kind == 'file'
334
assert ie.file_id == file_id
335
if ie.text_version in r:
336
assert r[ie.text_version] == ie
337
else:
338
r[ie.text_version] = ie
339
return r
340
341
342
def _store_entries(self):
343
"""Build revision inventory and store modified files.
344
345
This is called with new_inv a new empty inventory. Depending on
346
which files are selected for commit, and which ones have
347
been modified or merged, new inventory entries are built
348
based on the working and parent inventories.
349
350
As a side-effect this stores new text versions for committed
351
files with text changes or merges.
352
353
Each entry can have one of several things happen:
354
355
carry_file -- carried from the previous version (if not
356
selected for commit)
357
358
commit_nonfile -- no text to worry about
359
360
commit_old_text -- same text, may have moved
361
362
commit_file -- new text version
363
"""
364
for path, new_ie in self.work_inv.iter_entries():
365
file_id = new_ie.file_id
366
mutter('check %s {%s}', path, new_ie.file_id)
367
if self.specific_files:
368
if not is_inside_any(self.specific_files, path):
369
mutter('%s not selected for commit', path)
370
self._carry_file(file_id)
371
continue
372
if new_ie.kind != 'file':
373
self._commit_nonfile(file_id)
374
continue
375
376
file_parents = self._find_file_parents(file_id)
377
if len(file_parents) == 1:
378
parent_ie = file_parents.values()[0]
379
wc_sha1 = self.work_tree.get_file_sha1(file_id)
380
if parent_ie.text_sha1 == wc_sha1:
381
# text not changed or merged
382
self._commit_old_text(file_id, parent_ie)
383
continue
384
385
mutter('parents of %s are %r', path, file_parents)
386
387
# file is either new, or a file merge; need to record
388
# a new version
389
if len(file_parents) > 1:
390
note('merged %s', path)
391
elif len(file_parents) == 0:
392
note('added %s', path)
256
"""Remove deleted files from the working and stored inventories."""
257
for path, id, kind in self.delta.removed:
258
if self.work_inv.has_id(id):
259
del self.work_inv[id]
260
if self.new_inv.has_id(id):
261
del self.new_inv[id]
262
263
264
265
def _store_files(self):
266
"""Store new texts of modified/added files."""
267
# We must make sure that directories are added before anything
268
# inside them is added. the files within the delta report are
269
# sorted by path so we know the directory will come before its
270
# contents.
271
for path, file_id, kind in self.delta.added:
272
if kind != 'file':
273
ie = self.work_inv[file_id].copy()
274
self.new_inv.add(ie)
393
275
else:
394
note('modified %s', path)
395
self._commit_file(new_ie, file_id, file_parents)
396
397
398
def _commit_nonfile(self, file_id):
399
self.new_inv.add(self.work_inv[file_id].copy())
400
401
402
def _carry_file(self, file_id):
403
"""Carry the file unchanged from the basis revision."""
404
if self.basis_inv.has_id(file_id):
405
self.new_inv.add(self.basis_inv[file_id].copy())
406
407
408
def _commit_old_text(self, file_id, parent_ie):
409
"""Keep the same text as last time, but possibly a different name."""
410
ie = self.work_inv[file_id].copy()
411
ie.text_version = parent_ie.text_version
412
ie.text_size = parent_ie.text_size
413
ie.text_sha1 = parent_ie.text_sha1
414
self.new_inv.add(ie)
415
416
417
def _report_deletes(self):
418
for file_id in self.basis_inv:
419
if file_id not in self.new_inv:
420
note('deleted %s', self.basis_inv.id2path(file_id))
421
422
423
def _commit_file(self, new_ie, file_id, file_parents):
424
mutter('store new text for {%s} in revision {%s}',
425
file_id, self.rev_id)
276
self._store_file_text(file_id)
277
278
for path, file_id, kind in self.delta.modified:
279
if kind != 'file':
280
continue
281
self._store_file_text(file_id)
282
283
for old_path, new_path, file_id, kind, text_modified in self.delta.renamed:
284
if kind != 'file':
285
continue
286
if not text_modified:
287
continue
288
self._store_file_text(file_id)
289
290
291
def _store_file_text(self, file_id):
292
"""Store updated text for one modified or added file."""
293
note('store new text for {%s} in revision {%s}',
294
file_id, self.rev_id)
426
295
new_lines = self.work_tree.get_file(file_id).readlines()
296
if file_id in self.new_inv: # was in basis inventory
297
ie = self.new_inv[file_id]
298
assert ie.file_id == file_id
299
assert file_id in self.basis_inv
300
assert self.basis_inv[file_id].kind == 'file'
301
old_version = self.basis_inv[file_id].text_version
302
file_parents = [old_version]
303
else: # new in this revision
304
ie = self.work_inv[file_id].copy()
305
self.new_inv.add(ie)
306
assert file_id not in self.basis_inv
307
file_parents = []
308
assert ie.kind == 'file'
427
309
self._add_text_to_weave(file_id, new_lines, file_parents)
428
new_ie.text_version = self.rev_id
429
new_ie.text_sha1 = sha_strings(new_lines)
430
new_ie.text_size = sum(map(len, new_lines))
431
self.new_inv.add(new_ie)
310
# make a new inventory entry for this file, using whatever
311
# it had in the working copy, plus details on the new text
312
ie.text_sha1 = _sha_strings(new_lines)
313
ie.text_size = sum(map(len, new_lines))
314
ie.text_version = self.rev_id
315
ie.entry_version = self.rev_id
432
316
433
317
434
318
def _add_text_to_weave(self, file_id, new_lines, parents):
435
if file_id.startswith('__'):
436
raise ValueError('illegal file-id %r for text file' % file_id)
437
self.weave_store.add_text(file_id, self.rev_id, new_lines, parents)
319
weave_fn = self.branch.controlfilename(['weaves', file_id+'.weave'])
320
if os.path.exists(weave_fn):
321
w = read_weave(file(weave_fn, 'rb'))
322
else:
323
w = Weave()
324
# XXX: Should set the appropriate parents by looking for this file_id
325
# in all revision parents
326
parent_idxs = map(w.lookup, parents)
327
w.add(self.rev_id, parent_idxs, new_lines)
328
af = AtomicFile(weave_fn)
329
try:
330
write_weave_v5(w, af)
331
af.commit()
332
finally:
333
af.close()
438
334
439
335
440
336
def _gen_revision_id(branch, when):
444
340
return s
445
341
446
342
447
448
343
def _sha_strings(strings):
344
"""Return the sha-1 of concatenation of strings"""
345
s = sha.new()
346
map(s.update, strings)
347
return s.hexdigest()
Loggerhead is a web-based interface for Breezy
Version: 2.0.1