« back to all changes in this revision
Viewing changes to bzrlib/commit.py
- browse files at revision 1248
- compare with another revision
- compare with revision 5814
- download diff
- download tarball
- view history from revision 1248
- Committer: Martin Pool
- Date: 2005-09-15 04:32:41 UTC
- Revision ID: mbp@sourcefrog.net-20050915043239-f8f8978f485c1b85
- new weave based cleanup [broken]
- don't do a simple comparison to the basis version, but rather take
file merges into account.
- do more generation of XML in memory rather than to temporary files
- can now commit basic merges (not totally correctly though)
- track file parent versions
- new revision id derived from the supplied timestamp (if given),
not the current timestamp
- only write working inventory if there were changes caused by
deleted files
- put back WorkingTree.get_file_size
- methods to serialize to/from strings, as well as files.
Doesn't handle selective commit or proper merges yet.
- don't do a simple comparison to the basis version, but rather take
file merges into account.
- do more generation of XML in memory rather than to temporary files
- can now commit basic merges (not totally correctly though)
- track file parent versions
- new revision id derived from the supplied timestamp (if given),
not the current timestamp
- only write working inventory if there were changes caused by
deleted files
- put back WorkingTree.get_file_size
- methods to serialize to/from strings, as well as files.
Doesn't handle selective commit or proper merges yet.
- files modified:
- bzrlib/commit.py
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?
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
29
## This is not quite safe if the working copy changes during the
30
## commit; for the moment that is simply not allowed. A better
31
## approach is to make a temporary copy of the files before
32
## computing their hashes, and then add those hashes in turn to
33
## the inventory. This should mean at least that there are no
34
## broken hash pointers. There is no way we can get a snapshot
35
## of the whole directory at an instant. This would also have to
36
## be robust against files disappearing, moving, etc. So the
37
## whole thing is a bit hard.
38
39
## The newly committed revision is going to have a shape corresponding
40
## to that of the working inventory. Files that are not in the
41
## working tree and that were in the predecessor are reported as
42
## removed -- this can include files that were either removed from the
43
## inventory or deleted in the working tree. If they were only
44
## deleted from disk, they are removed from the working inventory.
45
46
## We then consider the remaining entries, which will be in the new
47
## version. Directory entries are simply copied across. File entries
48
## must be checked to see if a new version of the file should be
49
## recorded. For each parent revision inventory, we check to see what
50
## version of the file was present. If the file was present in at
51
## least one tree, and if it was the same version in all the trees,
52
## then we can just refer to that version. Otherwise, a new version
53
## representing the merger of the file versions must be added.
54
55
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?
56
46
57
47
58
48
59
49
import os
60
50
import sys
61
51
import time
62
import tempfile
63
import sha
52
import pdb
64
53
65
54
from binascii import hexlify
66
55
from cStringIO import StringIO
68
57
from bzrlib.osutils import (local_time_offset, username,
69
58
rand_bytes, compact_date, user_email,
70
59
kind_marker, is_inside_any, quotefn,
71
sha_string, sha_strings, sha_file, isdir, isfile)
60
sha_string, sha_strings, sha_file, isdir, isfile,
61
split_lines)
72
62
from bzrlib.branch import gen_file_id, INVENTORY_FILEID, ANCESTRY_FILEID
73
63
from bzrlib.errors import BzrError, PointlessCommit
74
64
from bzrlib.revision import Revision, RevisionReference
75
65
from bzrlib.trace import mutter, note
76
66
from bzrlib.xml5 import serializer_v5
77
67
from bzrlib.inventory import Inventory
78
from bzrlib.delta import compare_trees
79
68
from bzrlib.weave import Weave
80
69
from bzrlib.weavefile import read_weave, write_weave_v5
81
70
from bzrlib.atomicfile import AtomicFile
166
155
"""
167
156
168
157
self.branch = branch
158
self.weave_store = branch.weave_store
169
159
self.rev_id = rev_id
170
160
self.specific_files = specific_files
171
161
self.allow_pointless = allow_pointless
175
165
else:
176
166
self.timestamp = long(timestamp)
177
167
168
if rev_id is None:
169
self.rev_id = _gen_revision_id(self.branch, self.timestamp)
170
else:
171
self.rev_id = rev_id
172
178
173
if committer is None:
179
174
self.committer = username(self.branch)
180
175
else:
191
186
192
187
self.branch.lock_write()
193
188
try:
194
# First walk over the working inventory; and both update that
195
# and also build a new revision inventory. The revision
196
# inventory needs to hold the text-id, sha1 and size of the
197
# actual file versions committed in the revision. (These are
198
# not present in the working inventory.) We also need to
199
# detect missing/deleted files, and remove them from the
200
# working inventory.
201
202
189
self.work_tree = self.branch.working_tree()
203
190
self.work_inv = self.work_tree.inventory
204
191
self.basis_tree = self.branch.basis_tree()
206
193
207
194
self._gather_parents()
208
195
209
if self.rev_id is None:
210
self.rev_id = _gen_revision_id(self.branch, time.time())
211
212
self._remove_deletions()
196
any_deletions = self._remove_deleted()
197
self.new_inv = self.work_inv.copy()
198
any_changes = self._store_files()
199
self._report_deletes()
213
200
214
201
# TODO: update hashcache
215
self.delta = compare_trees(self.basis_tree, self.work_tree,
216
specific_files=self.specific_files)
217
218
if not (self.delta.has_changed()
219
or self.allow_pointless
202
if not (self.allow_pointless
203
or any_changes
204
or any_deletions
220
205
or len(self.parents) != 1):
221
206
raise PointlessCommit()
222
207
223
self.new_inv = self.basis_inv.copy()
224
225
## FIXME: Don't write to stdout!
226
self.delta.show(sys.stdout)
227
228
self._remove_deleted()
229
self._store_files()
230
231
self.branch._write_inventory(self.work_inv)
208
if any_deletions:
209
self.branch._write_inventory(self.work_inv)
232
210
self._record_inventory()
233
211
self._record_ancestry()
234
235
212
self._make_revision()
236
213
note('committted r%d {%s}', (self.branch.revno() + 1),
237
214
self.rev_id)
242
219
243
220
244
221
245
def _remove_deletions(self):
246
"""Remove deleted files from the working inventory."""
247
pass
248
249
250
251
222
def _record_inventory(self):
252
223
"""Store the inventory for the new revision."""
253
inv_tmp = StringIO()
254
serializer_v5.write_inventory(self.new_inv, inv_tmp)
255
inv_tmp.seek(0)
256
self.inv_sha1 = sha_string(inv_tmp.getvalue())
257
inv_lines = inv_tmp.readlines()
258
self.branch.weave_store.add_text(INVENTORY_FILEID, self.rev_id,
259
inv_lines, self.parents)
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)
260
228
261
229
262
230
def _record_ancestry(self):
263
231
"""Append merged revision ancestry to the ancestry file."""
264
if len(self.parents) > 1:
265
raise NotImplementedError("sorry, can't commit merges yet")
266
w = self.branch.weave_store.get_weave_or_empty(ANCESTRY_FILEID)
232
w = self.weave_store.get_weave_or_empty(ANCESTRY_FILEID)
267
233
if self.parents:
268
234
lines = w.get(w.lookup(self.parents[0]))
269
235
else:
271
237
lines.append(self.rev_id + '\n')
272
238
parent_idxs = map(w.lookup, self.parents)
273
239
w.add(self.rev_id, parent_idxs, lines)
274
self.branch.weave_store.put_weave(ANCESTRY_FILEID, w)
240
self.weave_store.put_weave(ANCESTRY_FILEID, w)
275
241
276
242
277
243
def _gather_parents(self):
278
244
pending_merges = self.branch.pending_merges()
279
if pending_merges:
280
raise NotImplementedError("sorry, can't commit merges to the weave format yet")
281
245
self.parents = []
282
246
precursor_id = self.branch.last_revision()
283
247
if precursor_id:
284
248
self.parents.append(precursor_id)
285
249
self.parents += pending_merges
250
self.parent_trees = map(self.branch.revision_tree, self.parents)
286
251
287
252
288
253
def _make_revision(self):
294
259
inventory_sha1=self.inv_sha1,
295
260
revision_id=self.rev_id)
296
261
self.rev.parents = map(RevisionReference, self.parents)
297
rev_tmp = tempfile.TemporaryFile()
262
rev_tmp = StringIO()
298
263
serializer_v5.write_revision(self.rev, rev_tmp)
299
264
rev_tmp.seek(0)
300
265
self.branch.revision_store.add(rev_tmp, self.rev_id)
302
267
303
268
304
269
def _remove_deleted(self):
305
"""Remove deleted files from the working and stored inventories."""
306
for path, id, kind in self.delta.removed:
307
if self.work_inv.has_id(id):
308
del self.work_inv[id]
309
if self.new_inv.has_id(id):
310
del self.new_inv[id]
311
270
"""Remove deleted files from the working inventories.
271
272
This is done prior to taking the working inventory as the
273
basis for the new committed inventory.
274
275
This returns true if any files
276
*that existed in the basis inventory* were deleted.
277
Files that were added and deleted
278
in the working copy don't matter.
279
"""
280
any_deletes = False
281
for file_id in list(iter(self.work_inv)): # snapshot for deletion
282
if not self.work_tree.has_id(file_id):
283
note('missing %s', self.work_inv.id2path(file_id))
284
del self.work_inv[file_id]
285
if self.basis_inv.has_id(file_id):
286
any_deletes = True
287
return any_deletes
288
289
290
def _find_file_parents(self, file_id):
291
"""Return the text versions and hashes for all file parents.
292
293
Returned as a map from text version to text sha1.
294
295
This is a set containing the file versions in all parents
296
revisions containing the file. If the file is new, the set
297
will be empty."""
298
r = {}
299
for tree in self.parent_trees:
300
if file_id in tree.inventory:
301
ie = tree.inventory[file_id]
302
assert ie.kind == 'file'
303
assert ie.file_id == file_id
304
if ie.text_version in r:
305
assert r[ie.text_version] == ie.text_sha1
306
else:
307
r[ie.text_version] = ie.text_sha1
308
return r
312
309
313
310
314
311
def _store_files(self):
315
"""Store new texts of modified/added files."""
316
# We must make sure that directories are added before anything
317
# inside them is added. the files within the delta report are
318
# sorted by path so we know the directory will come before its
319
# contents.
320
for path, file_id, kind in self.delta.added:
321
if kind != 'file':
322
ie = self.work_inv[file_id].copy()
323
self.new_inv.add(ie)
312
"""Store new texts of modified/added files.
313
314
This is called with new_inv set to a copy of the working
315
inventory, with deleted/removed files already cut out. So
316
this code only needs to deal with setting text versions, and
317
possibly recording new file texts."""
318
any_changes = False
319
for path, new_ie in self.new_inv.iter_entries():
320
if self.specific_files:
321
if not is_inside_any(self.specific_files, path):
322
# Not done yet
323
pass
324
mutter('check %s {%s}', path, new_ie.file_id)
325
if new_ie.kind != 'file':
326
# only regular files have texts to update
327
continue
328
file_id = new_ie.file_id
329
file_parents = self._find_file_parents(file_id)
330
wc_sha1 = self.work_tree.get_file_sha1(file_id)
331
wc_len = self.work_tree.get_file_size(file_id)
332
if (len(file_parents) == 1
333
and file_parents.values()[0] == wc_sha1):
334
# same as the single previous version, can reuse that
335
text_version = file_parents.keys()[0]
324
336
else:
325
self._store_file_text(file_id)
326
327
for path, file_id, kind in self.delta.modified:
328
if kind != 'file':
329
continue
330
self._store_file_text(file_id)
331
332
for old_path, new_path, file_id, kind, text_modified in self.delta.renamed:
333
if kind != 'file':
334
continue
335
if not text_modified:
336
continue
337
self._store_file_text(file_id)
338
339
340
def _store_file_text(self, file_id):
341
"""Store updated text for one modified or added file."""
342
note('store new text for {%s} in revision {%s}',
343
file_id, self.rev_id)
337
# file is either new, or a file merge; need to record
338
# a new version
339
if len(file_parents) > 1:
340
note('merged %s', path)
341
elif len(file_parents) == 0:
342
note('added %s', path)
343
else:
344
note('modified %s', path)
345
self._store_text(file_id, file_parents)
346
text_version = self.rev_id
347
any_changes = True
348
new_ie.text_version = text_version
349
new_ie.text_sha1 = wc_sha1
350
new_ie.text_size = wc_len
351
return any_changes
352
353
354
def _report_deletes(self):
355
for file_id in self.basis_inv:
356
if file_id not in self.new_inv:
357
note('deleted %s', self.basis_inv.id2path(file_id))
358
359
360
def _store_text(self, file_id, file_parents):
361
mutter('store new text for {%s} in revision {%s}',
362
file_id, self.rev_id)
344
363
new_lines = self.work_tree.get_file(file_id).readlines()
345
if file_id in self.new_inv: # was in basis inventory
346
ie = self.new_inv[file_id]
347
assert ie.file_id == file_id
348
assert file_id in self.basis_inv
349
assert self.basis_inv[file_id].kind == 'file'
350
old_version = self.basis_inv[file_id].text_version
351
file_parents = [old_version]
352
else: # new in this revision
353
ie = self.work_inv[file_id].copy()
354
self.new_inv.add(ie)
355
assert file_id not in self.basis_inv
356
file_parents = []
357
assert ie.kind == 'file'
358
364
self._add_text_to_weave(file_id, new_lines, file_parents)
359
# make a new inventory entry for this file, using whatever
360
# it had in the working copy, plus details on the new text
361
ie.text_sha1 = sha_strings(new_lines)
362
ie.text_size = sum(map(len, new_lines))
363
ie.text_version = self.rev_id
364
ie.entry_version = self.rev_id
365
365
366
366
367
367
def _add_text_to_weave(self, file_id, new_lines, parents):
368
368
if file_id.startswith('__'):
369
369
raise ValueError('illegal file-id %r for text file' % file_id)
370
self.branch.weave_store.add_text(file_id, self.rev_id, new_lines, parents)
370
self.weave_store.add_text(file_id, self.rev_id, new_lines, parents)
371
371
372
372
373
373
def _gen_revision_id(branch, when):
376
376
s += hexlify(rand_bytes(8))
377
377
return s
378
378
379
380
381
Loggerhead is a web-based interface for Breezy
Version: 2.0.1