To get this branch, use:
bzr branch
http://gegoxaren.bato24.eu/bzr/brz/remove-bazaar
« back to all changes in this revision
Viewing changes to bzrlib/commit.py
- browse files at revision 1249
- compare with another revision
- download diff
- download tarball
- view history from revision 1249
- Committer: Martin Pool
- Date: 2005-09-15 05:15:58 UTC
- Revision ID: mbp@sourcefrog.net-20050915051558-0d2738d29f376682
- improvements to weave commit [broken]
- work out pointless commits by just comparing the inventories (much
simpler)
- fix up selective commit: start with a blank inventory and fill it
from either the basis or working inventory depending on which
files are selected
- work out pointless commits by just comparing the inventories (much
simpler)
- fix up selective commit: start with a blank inventory and fill it
from either the basis or working inventory depending on which
files are selected
- files added:
- HACKING
- bzrlib/util
- files removed:
- plugins/changeset
- files renamed:
- plugins/ => bzrlib/plugins/
- tools/testweave.py => bzrlib/selftest/test_weave.py
- effbot/ => bzrlib/util/effbot/
- elementtree/ => bzrlib/util/elementtree/
- urlgrabber/ => bzrlib/util/urlgrabber/
- bzrlib/newinventory.py => contrib/newinventory.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
19
# FIXME: "bzr commit doc/format" commits doc/format.txt!
20
21
def commit(branch, message,
22
timestamp=None,
23
timezone=None,
24
committer=None,
25
verbose=True,
26
specific_files=None,
27
rev_id=None,
28
allow_pointless=True):
29
"""Commit working copy as a new revision.
30
31
The basic approach is to add all the file texts into the
32
store, then the inventory, then make a new revision pointing
33
to that inventory and store that.
34
35
This is not quite safe if the working copy changes during the
36
commit; for the moment that is simply not allowed. A better
37
approach is to make a temporary copy of the files before
38
computing their hashes, and then add those hashes in turn to
39
the inventory. This should mean at least that there are no
40
broken hash pointers. There is no way we can get a snapshot
41
of the whole directory at an instant. This would also have to
42
be robust against files disappearing, moving, etc. So the
43
whole thing is a bit hard.
44
45
This raises PointlessCommit if there are no changes, no new merges,
46
and allow_pointless is false.
47
48
timestamp -- if not None, seconds-since-epoch for a
49
postdated/predated commit.
50
51
specific_files
52
If true, commit only those files.
53
54
rev_id
55
If set, use this as the new revision id.
56
Useful for test or import commands that need to tightly
57
control what revisions are assigned. If you duplicate
58
a revision id that exists elsewhere it is your own fault.
59
If null (default), a time/random revision id is generated.
60
"""
61
62
import time, tempfile
63
64
from bzrlib.osutils import local_time_offset, username
65
from bzrlib.branch import gen_file_id
66
from bzrlib.errors import BzrError, PointlessCommit
67
from bzrlib.revision import Revision, RevisionReference
68
from bzrlib.trace import mutter, note
69
from bzrlib.xml import pack_xml
70
71
branch.lock_write()
72
73
try:
74
# First walk over the working inventory; and both update that
75
# and also build a new revision inventory. The revision
76
# inventory needs to hold the text-id, sha1 and size of the
77
# actual file versions committed in the revision. (These are
78
# not present in the working inventory.) We also need to
79
# detect missing/deleted files, and remove them from the
80
# working inventory.
81
82
work_tree = branch.working_tree()
83
work_inv = work_tree.inventory
84
basis = branch.basis_tree()
85
basis_inv = basis.inventory
86
87
if verbose:
88
note('looking for changes...')
89
90
pending_merges = branch.pending_merges()
91
92
missing_ids, new_inv, any_changes = \
93
_gather_commit(branch,
94
work_tree,
95
work_inv,
96
basis_inv,
97
specific_files,
98
verbose)
99
100
if not (any_changes or allow_pointless or pending_merges):
101
raise PointlessCommit()
102
103
for file_id in missing_ids:
104
# Any files that have been deleted are now removed from the
105
# working inventory. Files that were not selected for commit
106
# are left as they were in the working inventory and ommitted
107
# from the revision inventory.
108
109
# have to do this later so we don't mess up the iterator.
110
# since parents may be removed before their children we
111
# have to test.
112
113
# FIXME: There's probably a better way to do this; perhaps
114
# the workingtree should know how to filter itbranch.
115
if work_inv.has_id(file_id):
116
del work_inv[file_id]
117
118
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
48
49
import os
50
import sys
51
import time
52
import pdb
53
54
from binascii import hexlify
55
from cStringIO import StringIO
56
57
from bzrlib.osutils import (local_time_offset, username,
58
rand_bytes, compact_date, user_email,
59
kind_marker, is_inside_any, quotefn,
60
sha_string, sha_strings, sha_file, isdir, isfile,
61
split_lines)
62
from bzrlib.branch import gen_file_id, INVENTORY_FILEID, ANCESTRY_FILEID
63
from bzrlib.errors import BzrError, PointlessCommit
64
from bzrlib.revision import Revision, RevisionReference
65
from bzrlib.trace import mutter, note
66
from bzrlib.xml5 import serializer_v5
67
from bzrlib.inventory import Inventory
68
from bzrlib.weave import Weave
69
from bzrlib.weavefile import read_weave, write_weave_v5
70
from bzrlib.atomicfile import AtomicFile
71
72
73
def commit(*args, **kwargs):
74
"""Commit a new revision to a branch.
75
76
Function-style interface for convenience of old callers.
77
78
New code should use the Commit class instead.
79
"""
80
Commit().commit(*args, **kwargs)
81
82
83
class NullCommitReporter(object):
84
"""I report on progress of a commit."""
85
def added(self, path):
86
pass
87
88
def removed(self, path):
89
pass
90
91
def renamed(self, old_path, new_path):
92
pass
93
94
95
class ReportCommitToLog(NullCommitReporter):
96
def added(self, path):
97
note('added %s', path)
98
99
def removed(self, path):
100
note('removed %s', path)
101
102
def renamed(self, old_path, new_path):
103
note('renamed %s => %s', old_path, new_path)
104
105
106
class Commit(object):
107
"""Task of committing a new revision.
108
109
This is a MethodObject: it accumulates state as the commit is
110
prepared, and then it is discarded. It doesn't represent
111
historical revisions, just the act of recording a new one.
112
113
missing_ids
114
Modified to hold a list of files that have been deleted from
115
the working directory; these should be removed from the
116
working inventory.
117
"""
118
def __init__(self,
119
reporter=None):
120
if reporter is not None:
121
self.reporter = reporter
122
else:
123
self.reporter = NullCommitReporter()
124
125
126
def commit(self,
127
branch, message,
128
timestamp=None,
129
timezone=None,
130
committer=None,
131
specific_files=None,
132
rev_id=None,
133
allow_pointless=True):
134
"""Commit working copy as a new revision.
135
136
The basic approach is to add all the file texts into the
137
store, then the inventory, then make a new revision pointing
138
to that inventory and store that.
139
140
This raises PointlessCommit if there are no changes, no new merges,
141
and allow_pointless is false.
142
143
timestamp -- if not None, seconds-since-epoch for a
144
postdated/predated commit.
145
146
specific_files
147
If true, commit only those files.
148
149
rev_id
150
If set, use this as the new revision id.
151
Useful for test or import commands that need to tightly
152
control what revisions are assigned. If you duplicate
153
a revision id that exists elsewhere it is your own fault.
154
If null (default), a time/random revision id is generated.
155
"""
156
self.any_changes = False
157
158
self.branch = branch
159
self.weave_store = branch.weave_store
160
self.rev_id = rev_id
161
self.specific_files = specific_files
162
self.allow_pointless = allow_pointless
163
164
if timestamp is None:
165
self.timestamp = time.time()
166
else:
167
self.timestamp = long(timestamp)
168
119
169
if rev_id is None:
120
rev_id = _gen_revision_id(time.time())
121
inv_id = rev_id
122
123
inv_tmp = tempfile.TemporaryFile()
124
pack_xml(new_inv, inv_tmp)
125
inv_tmp.seek(0)
126
branch.inventory_store.add(inv_tmp, inv_id)
127
mutter('new inventory_id is {%s}' % inv_id)
128
129
# We could also just sha hash the inv_tmp file
130
# however, in the case that branch.inventory_store.add()
131
# ever actually does anything special
132
inv_sha1 = branch.get_inventory_sha1(inv_id)
133
134
branch._write_inventory(work_inv)
135
136
if timestamp == None:
137
timestamp = time.time()
138
139
if committer == None:
140
committer = username()
141
142
if timezone == None:
143
timezone = local_time_offset()
144
145
mutter("building commit log message")
146
rev = Revision(timestamp=timestamp,
147
timezone=timezone,
148
committer=committer,
149
message = message,
150
inventory_id=inv_id,
151
inventory_sha1=inv_sha1,
152
revision_id=rev_id)
153
154
rev.parents = []
155
precursor_id = branch.last_patch()
170
self.rev_id = _gen_revision_id(self.branch, self.timestamp)
171
else:
172
self.rev_id = rev_id
173
174
if committer is None:
175
self.committer = username(self.branch)
176
else:
177
assert isinstance(committer, basestring), type(committer)
178
self.committer = committer
179
180
if timezone is None:
181
self.timezone = local_time_offset()
182
else:
183
self.timezone = int(timezone)
184
185
assert isinstance(message, basestring), type(message)
186
self.message = message
187
188
self.branch.lock_write()
189
try:
190
self.work_tree = self.branch.working_tree()
191
self.work_inv = self.work_tree.inventory
192
self.basis_tree = self.branch.basis_tree()
193
self.basis_inv = self.basis_tree.inventory
194
195
self._gather_parents()
196
197
self.new_inv = Inventory()
198
self._store_files()
199
self._report_deletes()
200
201
if not (self.allow_pointless
202
or len(self.parents) != 1
203
or self.new_inv != self.basis_inv):
204
raise PointlessCommit()
205
206
self._record_inventory()
207
self._record_ancestry()
208
self._make_revision()
209
note('committted r%d {%s}', (self.branch.revno() + 1),
210
self.rev_id)
211
self.branch.append_revision(self.rev_id)
212
self.branch.set_pending_merges([])
213
finally:
214
self.branch.unlock()
215
216
217
218
def _record_inventory(self):
219
"""Store the inventory for the new revision."""
220
inv_text = serializer_v5.write_inventory_to_string(self.new_inv)
221
self.inv_sha1 = sha_string(inv_text)
222
self.weave_store.add_text(INVENTORY_FILEID, self.rev_id,
223
split_lines(inv_text), self.parents)
224
225
226
def _record_ancestry(self):
227
"""Append merged revision ancestry to the ancestry file."""
228
w = self.weave_store.get_weave_or_empty(ANCESTRY_FILEID)
229
if self.parents:
230
lines = w.get(w.lookup(self.parents[0]))
231
else:
232
lines = []
233
lines.append(self.rev_id + '\n')
234
parent_idxs = map(w.lookup, self.parents)
235
w.add(self.rev_id, parent_idxs, lines)
236
self.weave_store.put_weave(ANCESTRY_FILEID, w)
237
238
239
def _gather_parents(self):
240
pending_merges = self.branch.pending_merges()
241
self.parents = []
242
precursor_id = self.branch.last_revision()
156
243
if precursor_id:
157
precursor_sha1 = branch.get_revision_sha1(precursor_id)
158
rev.parents.append(RevisionReference(precursor_id, precursor_sha1))
159
for merge_rev in pending_merges:
160
rev.parents.append(RevisionReference(merge_rev))
161
162
rev_tmp = tempfile.TemporaryFile()
163
pack_xml(rev, rev_tmp)
244
self.parents.append(precursor_id)
245
self.parents += pending_merges
246
self.parent_trees = map(self.branch.revision_tree, self.parents)
247
248
249
def _make_revision(self):
250
"""Record a new revision object for this commit."""
251
self.rev = Revision(timestamp=self.timestamp,
252
timezone=self.timezone,
253
committer=self.committer,
254
message=self.message,
255
inventory_sha1=self.inv_sha1,
256
revision_id=self.rev_id)
257
self.rev.parents = map(RevisionReference, self.parents)
258
rev_tmp = StringIO()
259
serializer_v5.write_revision(self.rev, rev_tmp)
164
260
rev_tmp.seek(0)
165
branch.revision_store.add(rev_tmp, rev_id)
166
mutter("new revision_id is {%s}" % rev_id)
167
168
## XXX: Everything up to here can simply be orphaned if we abort
169
## the commit; it will leave junk files behind but that doesn't
170
## matter.
171
172
## TODO: Read back the just-generated changeset, and make sure it
173
## applies and recreates the right state.
174
175
## TODO: Also calculate and store the inventory SHA1
176
mutter("committing patch r%d" % (branch.revno() + 1))
177
178
branch.append_revision(rev_id)
179
180
branch.set_pending_merges([])
181
182
if verbose:
183
note("commited r%d" % branch.revno())
184
finally:
185
branch.unlock()
186
187
188
189
def _gen_revision_id(when):
261
self.branch.revision_store.add(rev_tmp, self.rev_id)
262
mutter('new revision_id is {%s}', self.rev_id)
263
264
265
def _remove_deleted(self):
266
"""Remove deleted files from the working inventories.
267
268
This is done prior to taking the working inventory as the
269
basis for the new committed inventory.
270
271
This returns true if any files
272
*that existed in the basis inventory* were deleted.
273
Files that were added and deleted
274
in the working copy don't matter.
275
"""
276
any_deletes = False
277
for file_id in list(iter(self.work_inv)): # snapshot for deletion
278
if not self.work_tree.has_id(file_id):
279
note('missing %s', self.work_inv.id2path(file_id))
280
del self.work_inv[file_id]
281
any_deletes = True
282
if any_deletions:
283
self.branch._write_inventory(self.work_inv)
284
285
286
def _find_file_parents(self, file_id):
287
"""Return the text versions and hashes for all file parents.
288
289
Returned as a map from text version to text sha1.
290
291
This is a set containing the file versions in all parents
292
revisions containing the file. If the file is new, the set
293
will be empty."""
294
r = {}
295
for tree in self.parent_trees:
296
if file_id in tree.inventory:
297
ie = tree.inventory[file_id]
298
assert ie.kind == 'file'
299
assert ie.file_id == file_id
300
if ie.text_version in r:
301
assert r[ie.text_version] == ie.text_sha1
302
else:
303
r[ie.text_version] = ie.text_sha1
304
return r
305
306
307
def _store_files(self):
308
"""Store new texts of modified/added files.
309
310
This is called with new_inv set to a copy of the working
311
inventory, with deleted/removed files already cut out. So
312
this code only needs to deal with setting text versions, and
313
possibly recording new file texts."""
314
for path, new_ie in self.work_inv.iter_entries():
315
file_id = new_ie.file_id
316
mutter('check %s {%s}', path, new_ie.file_id)
317
if self.specific_files:
318
if not is_inside_any(self.specific_files, path):
319
mutter('%s not selected for commit', path)
320
self._carry_file(file_id)
321
continue
322
if new_ie.kind != 'file':
323
self._commit_nonfile(file_id)
324
continue
325
file_parents = self._find_file_parents(file_id)
326
wc_sha1 = self.work_tree.get_file_sha1(file_id)
327
if (len(file_parents) == 1
328
and file_parents.values()[0] == wc_sha1):
329
# not changed or merged
330
self._carry_file(file_id)
331
continue
332
333
mutter('parents of %s are %r', path, file_parents)
334
335
# file is either new, or a file merge; need to record
336
# a new version
337
if len(file_parents) > 1:
338
note('merged %s', path)
339
elif len(file_parents) == 0:
340
note('added %s', path)
341
else:
342
note('modified %s', path)
343
self._commit_file(new_ie, file_id, file_parents)
344
345
346
def _commit_nonfile(self, file_id):
347
self.new_inv.add(self.work_inv[file_id].copy())
348
349
350
def _carry_file(self, file_id):
351
"""Keep a file in the same state as in the basis."""
352
if self.basis_inv.has_id(file_id):
353
self.new_inv.add(self.basis_inv[file_id].copy())
354
355
356
def _report_deletes(self):
357
for file_id in self.basis_inv:
358
if file_id not in self.new_inv:
359
note('deleted %s', self.basis_inv.id2path(file_id))
360
361
362
def _commit_file(self, new_ie, file_id, file_parents):
363
mutter('store new text for {%s} in revision {%s}',
364
file_id, self.rev_id)
365
new_lines = self.work_tree.get_file(file_id).readlines()
366
self._add_text_to_weave(file_id, new_lines, file_parents)
367
new_ie.text_version = self.rev_id
368
new_ie.text_sha1 = sha_strings(new_lines)
369
new_ie.text_size = sum(map(len, new_lines))
370
self.new_inv.add(new_ie)
371
372
373
def _add_text_to_weave(self, file_id, new_lines, parents):
374
if file_id.startswith('__'):
375
raise ValueError('illegal file-id %r for text file' % file_id)
376
self.weave_store.add_text(file_id, self.rev_id, new_lines, parents)
377
378
379
def _gen_revision_id(branch, when):
190
380
"""Return new revision-id."""
191
from binascii import hexlify
192
from osutils import rand_bytes, compact_date, user_email
193
194
s = '%s-%s-' % (user_email(), compact_date(when))
381
s = '%s-%s-' % (user_email(branch), compact_date(when))
195
382
s += hexlify(rand_bytes(8))
196
383
return s
197
384
198
385
199
def _gather_commit(branch, work_tree, work_inv, basis_inv, specific_files,
200
verbose):
201
"""Build inventory preparatory to commit.
202
203
Returns missing_ids, new_inv, any_changes.
204
205
This adds any changed files into the text store, and sets their
206
test-id, sha and size in the returned inventory appropriately.
207
208
missing_ids
209
Modified to hold a list of files that have been deleted from
210
the working directory; these should be removed from the
211
working inventory.
212
"""
213
from bzrlib.inventory import Inventory
214
from bzrlib.osutils import isdir, isfile, sha_string, quotefn, \
215
local_time_offset, username, kind_marker, is_inside_any
216
217
from bzrlib.branch import gen_file_id
218
from bzrlib.errors import BzrError
219
from bzrlib.revision import Revision
220
from bzrlib.trace import mutter, note
221
222
any_changes = False
223
inv = Inventory(work_inv.root.file_id)
224
missing_ids = []
225
226
for path, entry in work_inv.iter_entries():
227
## TODO: Check that the file kind has not changed from the previous
228
## revision of this file (if any).
229
230
p = branch.abspath(path)
231
file_id = entry.file_id
232
mutter('commit prep file %s, id %r ' % (p, file_id))
233
234
if specific_files and not is_inside_any(specific_files, path):
235
mutter(' skipping file excluded from commit')
236
if basis_inv.has_id(file_id):
237
# carry over with previous state
238
inv.add(basis_inv[file_id].copy())
239
else:
240
# omit this from committed inventory
241
pass
242
continue
243
244
if not work_tree.has_id(file_id):
245
if verbose:
246
print('deleted %s%s' % (path, kind_marker(entry.kind)))
247
any_changes = True
248
mutter(" file is missing, removing from inventory")
249
missing_ids.append(file_id)
250
continue
251
252
# this is present in the new inventory; may be new, modified or
253
# unchanged.
254
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
255
256
entry = entry.copy()
257
inv.add(entry)
258
259
if old_ie:
260
old_kind = old_ie.kind
261
if old_kind != entry.kind:
262
raise BzrError("entry %r changed kind from %r to %r"
263
% (file_id, old_kind, entry.kind))
264
265
if entry.kind == 'directory':
266
if not isdir(p):
267
raise BzrError("%s is entered as directory but not a directory"
268
% quotefn(p))
269
elif entry.kind == 'file':
270
if not isfile(p):
271
raise BzrError("%s is entered as file but is not a file" % quotefn(p))
272
273
new_sha1 = work_tree.get_file_sha1(file_id)
274
275
if (old_ie
276
and old_ie.text_sha1 == new_sha1):
277
## assert content == basis.get_file(file_id).read()
278
entry.text_id = old_ie.text_id
279
entry.text_sha1 = new_sha1
280
entry.text_size = old_ie.text_size
281
mutter(' unchanged from previous text_id {%s}' %
282
entry.text_id)
283
else:
284
content = file(p, 'rb').read()
285
286
# calculate the sha again, just in case the file contents
287
# changed since we updated the cache
288
entry.text_sha1 = sha_string(content)
289
entry.text_size = len(content)
290
291
entry.text_id = gen_file_id(entry.name)
292
branch.text_store.add(content, entry.text_id)
293
mutter(' stored with text_id {%s}' % entry.text_id)
294
295
if verbose:
296
marked = path + kind_marker(entry.kind)
297
if not old_ie:
298
print 'added', marked
299
any_changes = True
300
elif old_ie == entry:
301
pass # unchanged
302
elif (old_ie.name == entry.name
303
and old_ie.parent_id == entry.parent_id):
304
print 'modified', marked
305
any_changes = True
306
else:
307
print 'renamed', marked
308
any_changes = True
309
310
return missing_ids, inv, any_changes
311
312
386
387
Loggerhead is a web-based interface for Breezy
Version: 2.0.2.dev0