« back to all changes in this revision
Viewing changes to bzrlib/commit.py
- browse files at revision 856
- compare with another revision
- download diff
- download tarball
- view history from revision 856
- Committer: Martin Pool
- Date: 2005-07-07 10:31:36 UTC
- Revision ID: mbp@sourcefrog.net-20050707103135-9b4d911d8df6e880
- fix pwk help
- files added:
- bzrlib/meta_store.py
- plugins/changeset
- files removed:
- HACKING
- bzrlib/store
- bzrlib/transport
- bzrlib/util
- contrib/emacs
- files renamed:
- contrib/newinventory.py => bzrlib/newinventory.py
- bzrlib/store/__init__.py => bzrlib/store.py
- bzrlib/util/effbot/ => effbot/
- bzrlib/util/elementtree/ => elementtree/
- bzrlib/plugins/ => plugins/
- bzrlib/selftest/test_weave.py => tools/testweave.py
- bzrlib/util/urlgrabber/ => urlgrabber/
- files modified:
- .bzrignore
- bzr *
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
# TODO: Rather than mashing together the ancestry and storing it back,
48
# perhaps the weave should have single method which does it all in one
49
# go, avoiding a lot of redundant work.
50
51
# TODO: Perhaps give a warning if one of the revisions marked as
52
# merged is already in the ancestry, and then don't record it as a
53
# distinct parent.
54
55
# TODO: If the file is newly merged but unchanged from the version it
56
# merges from, then it should still be reported as newly added
57
# relative to the basis revision.
58
59
60
import os
61
import re
62
import sys
63
import time
64
import pdb
65
66
from binascii import hexlify
67
from cStringIO import StringIO
68
69
from bzrlib.osutils import (local_time_offset,
70
rand_bytes, compact_date,
71
kind_marker, is_inside_any, quotefn,
72
sha_string, sha_strings, sha_file, isdir, isfile,
73
split_lines)
74
from bzrlib.branch import gen_file_id
75
import bzrlib.config
76
from bzrlib.errors import (BzrError, PointlessCommit,
77
HistoryMissing,
78
ConflictsInTree
79
)
80
import bzrlib.gpg as gpg
81
from bzrlib.revision import Revision
82
from bzrlib.trace import mutter, note, warning
83
from bzrlib.xml5 import serializer_v5
84
from bzrlib.inventory import Inventory, ROOT_ID
85
from bzrlib.weave import Weave
86
from bzrlib.weavefile import read_weave, write_weave_v5
87
from bzrlib.atomicfile import AtomicFile
88
89
90
def commit(*args, **kwargs):
91
"""Commit a new revision to a branch.
92
93
Function-style interface for convenience of old callers.
94
95
New code should use the Commit class instead.
96
"""
97
## XXX: Remove this in favor of Branch.commit?
98
Commit().commit(*args, **kwargs)
99
100
101
class NullCommitReporter(object):
102
"""I report on progress of a commit."""
103
104
def snapshot_change(self, change, path):
105
pass
106
107
def completed(self, revno, rev_id):
108
pass
109
110
def deleted(self, file_id):
111
pass
112
113
def escaped(self, escape_count, message):
114
pass
115
116
def missing(self, path):
117
pass
118
119
class ReportCommitToLog(NullCommitReporter):
120
121
def snapshot_change(self, change, path):
122
note("%s %s", change, path)
123
124
def completed(self, revno, rev_id):
125
note('committed r%d {%s}', revno, rev_id)
126
127
def deleted(self, file_id):
128
note('deleted %s', file_id)
129
130
def escaped(self, escape_count, message):
131
note("replaced %d control characters in message", escape_count)
132
133
def missing(self, path):
134
note('missing %s', path)
135
136
class Commit(object):
137
"""Task of committing a new revision.
138
139
This is a MethodObject: it accumulates state as the commit is
140
prepared, and then it is discarded. It doesn't represent
141
historical revisions, just the act of recording a new one.
142
143
missing_ids
144
Modified to hold a list of files that have been deleted from
145
the working directory; these should be removed from the
146
working inventory.
147
"""
148
def __init__(self,
149
reporter=None,
150
config=None):
151
if reporter is not None:
152
self.reporter = reporter
153
else:
154
self.reporter = NullCommitReporter()
155
if config is not None:
156
self.config = config
157
else:
158
self.config = None
159
160
def commit(self,
161
branch, message,
162
timestamp=None,
163
timezone=None,
164
committer=None,
165
specific_files=None,
166
rev_id=None,
167
allow_pointless=True,
168
verbose=False,
169
revprops=None):
170
"""Commit working copy as a new revision.
171
172
timestamp -- if not None, seconds-since-epoch for a
173
postdated/predated commit.
174
175
specific_files -- If true, commit only those files.
176
177
rev_id -- If set, use this as the new revision id.
178
Useful for test or import commands that need to tightly
179
control what revisions are assigned. If you duplicate
180
a revision id that exists elsewhere it is your own fault.
181
If null (default), a time/random revision id is generated.
182
183
allow_pointless -- If true (default), commit even if nothing
184
has changed and no merges are recorded.
185
186
revprops -- Properties for new revision
187
"""
188
mutter('preparing to commit')
189
190
self.branch = branch
191
self.weave_store = branch.weave_store
192
self.rev_id = rev_id
193
self.specific_files = specific_files
194
self.allow_pointless = allow_pointless
195
self.revprops = revprops
196
197
if timestamp is None:
198
self.timestamp = time.time()
199
else:
200
self.timestamp = long(timestamp)
201
202
if self.config is None:
203
self.config = bzrlib.config.BranchConfig(self.branch)
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
"""Commit working copy as a new revision.
29
30
The basic approach is to add all the file texts into the
31
store, then the inventory, then make a new revision pointing
32
to that inventory and store that.
33
34
This is not quite safe if the working copy changes during the
35
commit; for the moment that is simply not allowed. A better
36
approach is to make a temporary copy of the files before
37
computing their hashes, and then add those hashes in turn to
38
the inventory. This should mean at least that there are no
39
broken hash pointers. There is no way we can get a snapshot
40
of the whole directory at an instant. This would also have to
41
be robust against files disappearing, moving, etc. So the
42
whole thing is a bit hard.
43
44
timestamp -- if not None, seconds-since-epoch for a
45
postdated/predated commit.
46
47
specific_files
48
If true, commit only those files.
49
50
rev_id
51
If set, use this as the new revision id.
52
Useful for test or import commands that need to tightly
53
control what revisions are assigned. If you duplicate
54
a revision id that exists elsewhere it is your own fault.
55
If null (default), a time/random revision id is generated.
56
"""
57
58
import time, tempfile
59
60
from bzrlib.osutils import local_time_offset, username
61
from bzrlib.branch import gen_file_id
62
from bzrlib.errors import BzrError
63
from bzrlib.revision import Revision, RevisionReference
64
from bzrlib.trace import mutter, note
65
from bzrlib.xml import pack_xml
66
67
branch.lock_write()
68
69
try:
70
# First walk over the working inventory; and both update that
71
# and also build a new revision inventory. The revision
72
# inventory needs to hold the text-id, sha1 and size of the
73
# actual file versions committed in the revision. (These are
74
# not present in the working inventory.) We also need to
75
# detect missing/deleted files, and remove them from the
76
# working inventory.
77
78
work_tree = branch.working_tree()
79
work_inv = work_tree.inventory
80
basis = branch.basis_tree()
81
basis_inv = basis.inventory
82
83
if verbose:
84
note('looking for changes...')
85
86
pending_merges = branch.pending_merges()
87
88
missing_ids, new_inv = _gather_commit(branch,
89
work_tree,
90
work_inv,
91
basis_inv,
92
specific_files,
93
verbose)
94
95
for file_id in missing_ids:
96
# Any files that have been deleted are now removed from the
97
# working inventory. Files that were not selected for commit
98
# are left as they were in the working inventory and ommitted
99
# from the revision inventory.
100
101
# have to do this later so we don't mess up the iterator.
102
# since parents may be removed before their children we
103
# have to test.
104
105
# FIXME: There's probably a better way to do this; perhaps
106
# the workingtree should know how to filter itbranch.
107
if work_inv.has_id(file_id):
108
del work_inv[file_id]
109
204
110
205
111
if rev_id is None:
206
self.rev_id = _gen_revision_id(self.config, self.timestamp)
207
else:
208
self.rev_id = rev_id
209
210
if committer is None:
211
self.committer = self.config.username()
212
else:
213
assert isinstance(committer, basestring), type(committer)
214
self.committer = committer
215
216
if timezone is None:
217
self.timezone = local_time_offset()
218
else:
219
self.timezone = int(timezone)
220
221
assert isinstance(message, basestring), type(message)
222
self.message = message
223
self._escape_commit_message()
224
225
self.branch.lock_write()
226
try:
227
self.work_tree = self.branch.working_tree()
228
self.work_inv = self.work_tree.inventory
229
self.basis_tree = self.branch.basis_tree()
230
self.basis_inv = self.basis_tree.inventory
231
232
self._gather_parents()
233
if len(self.parents) > 1 and self.specific_files:
234
raise NotImplementedError('selected-file commit of merges is not supported yet')
235
self._check_parents_present()
236
237
self._remove_deleted()
238
self._populate_new_inv()
239
self._store_snapshot()
240
self._report_deletes()
241
242
if not (self.allow_pointless
243
or len(self.parents) > 1
244
or self.new_inv != self.basis_inv):
245
raise PointlessCommit()
246
247
if len(list(self.work_tree.iter_conflicts()))>0:
248
raise ConflictsInTree
249
250
self._record_inventory()
251
self._make_revision()
252
self.reporter.completed(self.branch.revno()+1, self.rev_id)
253
self.branch.append_revision(self.rev_id)
254
self.branch.set_pending_merges([])
255
finally:
256
self.branch.unlock()
257
258
def _record_inventory(self):
259
"""Store the inventory for the new revision."""
260
inv_text = serializer_v5.write_inventory_to_string(self.new_inv)
261
self.inv_sha1 = sha_string(inv_text)
262
s = self.branch.control_weaves
263
s.add_text('inventory', self.rev_id,
264
split_lines(inv_text), self.present_parents,
265
self.branch.get_transaction())
266
267
def _escape_commit_message(self):
268
"""Replace xml-incompatible control characters."""
269
# Python strings can include characters that can't be
270
# represented in well-formed XML; escape characters that
271
# aren't listed in the XML specification
272
# (http://www.w3.org/TR/REC-xml/#NT-Char).
273
if isinstance(self.message, unicode):
274
char_pattern = u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]'
275
else:
276
# Use a regular 'str' as pattern to avoid having re.subn
277
# return 'unicode' results.
278
char_pattern = '[^x09\x0A\x0D\x20-\xFF]'
279
self.message, escape_count = re.subn(
280
char_pattern,
281
lambda match: match.group(0).encode('unicode_escape'),
282
self.message)
283
if escape_count:
284
self.reporter.escaped(escape_count, self.message)
285
286
def _gather_parents(self):
287
"""Record the parents of a merge for merge detection."""
288
pending_merges = self.branch.pending_merges()
289
self.parents = []
290
self.parent_invs = []
291
self.present_parents = []
292
precursor_id = self.branch.last_revision()
112
rev_id = _gen_revision_id(time.time())
113
inv_id = rev_id
114
115
inv_tmp = tempfile.TemporaryFile()
116
pack_xml(new_inv, inv_tmp)
117
inv_tmp.seek(0)
118
branch.inventory_store.add(inv_tmp, inv_id)
119
mutter('new inventory_id is {%s}' % inv_id)
120
121
# We could also just sha hash the inv_tmp file
122
# however, in the case that branch.inventory_store.add()
123
# ever actually does anything special
124
inv_sha1 = branch.get_inventory_sha1(inv_id)
125
126
branch._write_inventory(work_inv)
127
128
if timestamp == None:
129
timestamp = time.time()
130
131
if committer == None:
132
committer = username()
133
134
if timezone == None:
135
timezone = local_time_offset()
136
137
mutter("building commit log message")
138
rev = Revision(timestamp=timestamp,
139
timezone=timezone,
140
committer=committer,
141
message = message,
142
inventory_id=inv_id,
143
inventory_sha1=inv_sha1,
144
revision_id=rev_id)
145
146
rev.parents = []
147
precursor_id = branch.last_patch()
293
148
if precursor_id:
294
self.parents.append(precursor_id)
295
self.parents += pending_merges
296
for revision in self.parents:
297
if self.branch.has_revision(revision):
298
self.parent_invs.append(self.branch.get_inventory(revision))
299
self.present_parents.append(revision)
149
precursor_sha1 = branch.get_revision_sha1(precursor_id)
150
rev.parents.append(RevisionReference(precursor_id, precursor_sha1))
151
for merge_rev in pending_merges:
152
rev.parents.append(RevisionReference(merge_rev))
300
153
301
def _check_parents_present(self):
302
for parent_id in self.parents:
303
mutter('commit parent revision {%s}', parent_id)
304
if not self.branch.has_revision(parent_id):
305
if parent_id == self.branch.last_revision():
306
warning("parent is missing %r", parent_id)
307
raise HistoryMissing(self.branch, 'revision', parent_id)
308
else:
309
mutter("commit will ghost revision %r", parent_id)
310
311
def _make_revision(self):
312
"""Record a new revision object for this commit."""
313
self.rev = Revision(timestamp=self.timestamp,
314
timezone=self.timezone,
315
committer=self.committer,
316
message=self.message,
317
inventory_sha1=self.inv_sha1,
318
revision_id=self.rev_id,
319
properties=self.revprops)
320
self.rev.parent_ids = self.parents
321
rev_tmp = StringIO()
322
serializer_v5.write_revision(self.rev, rev_tmp)
154
rev_tmp = tempfile.TemporaryFile()
155
pack_xml(rev, rev_tmp)
323
156
rev_tmp.seek(0)
324
self.branch.revision_store.add(rev_tmp, self.rev_id)
325
if self.config.signature_needed():
326
self.branch.sign_revision(self.rev_id, gpg.GPGStrategy(self.config))
327
mutter('new revision_id is {%s}', self.rev_id)
328
329
def _remove_deleted(self):
330
"""Remove deleted files from the working inventories.
331
332
This is done prior to taking the working inventory as the
333
basis for the new committed inventory.
334
335
This returns true if any files
336
*that existed in the basis inventory* were deleted.
337
Files that were added and deleted
338
in the working copy don't matter.
339
"""
340
specific = self.specific_files
341
deleted_ids = []
342
for path, ie in self.work_inv.iter_entries():
343
if specific and not is_inside_any(specific, path):
344
continue
345
if not self.work_tree.has_filename(path):
346
self.reporter.missing(path)
347
deleted_ids.append((path, ie.file_id))
348
if deleted_ids:
349
deleted_ids.sort(reverse=True)
350
for path, file_id in deleted_ids:
351
del self.work_inv[file_id]
352
self.branch._write_inventory(self.work_inv)
353
354
def _store_snapshot(self):
355
"""Pass over inventory and record a snapshot.
356
357
Entries get a new revision when they are modified in
358
any way, which includes a merge with a new set of
359
parents that have the same entry.
360
"""
361
# XXX: Need to think more here about when the user has
362
# made a specific decision on a particular value -- c.f.
363
# mark-merge.
364
for path, ie in self.new_inv.iter_entries():
365
previous_entries = ie.find_previous_heads(
366
self.parent_invs,
367
self.weave_store.get_weave_or_empty(ie.file_id,
368
self.branch.get_transaction()))
369
if ie.revision is None:
370
change = ie.snapshot(self.rev_id, path, previous_entries,
371
self.work_tree, self.weave_store,
372
self.branch.get_transaction())
373
else:
374
change = "unchanged"
375
self.reporter.snapshot_change(change, path)
376
377
def _populate_new_inv(self):
378
"""Build revision inventory.
379
380
This creates a new empty inventory. Depending on
381
which files are selected for commit, and what is present in the
382
current tree, the new inventory is populated. inventory entries
383
which are candidates for modification have their revision set to
384
None; inventory entries that are carried over untouched have their
385
revision set to their prior value.
386
"""
387
mutter("Selecting files for commit with filter %s", self.specific_files)
388
self.new_inv = Inventory()
389
for path, new_ie in self.work_inv.iter_entries():
390
file_id = new_ie.file_id
391
mutter('check %s {%s}', path, new_ie.file_id)
392
if self.specific_files:
393
if not is_inside_any(self.specific_files, path):
394
mutter('%s not selected for commit', path)
395
self._carry_entry(file_id)
396
continue
397
else:
398
# this is selected, ensure its parents are too.
399
parent_id = new_ie.parent_id
400
while parent_id != ROOT_ID:
401
if not self.new_inv.has_id(parent_id):
402
ie = self._select_entry(self.work_inv[parent_id])
403
mutter('%s selected for commit because of %s',
404
self.new_inv.id2path(parent_id), path)
405
406
ie = self.new_inv[parent_id]
407
if ie.revision is not None:
408
ie.revision = None
409
mutter('%s selected for commit because of %s',
410
self.new_inv.id2path(parent_id), path)
411
parent_id = ie.parent_id
412
mutter('%s selected for commit', path)
413
self._select_entry(new_ie)
414
415
def _select_entry(self, new_ie):
416
"""Make new_ie be considered for committing."""
417
ie = new_ie.copy()
418
ie.revision = None
419
self.new_inv.add(ie)
420
return ie
421
422
def _carry_entry(self, file_id):
423
"""Carry the file unchanged from the basis revision."""
424
if self.basis_inv.has_id(file_id):
425
self.new_inv.add(self.basis_inv[file_id].copy())
426
427
def _report_deletes(self):
428
for file_id in self.basis_inv:
429
if file_id not in self.new_inv:
430
self.reporter.deleted(self.basis_inv.id2path(file_id))
431
432
def _gen_revision_id(config, when):
157
branch.revision_store.add(rev_tmp, rev_id)
158
mutter("new revision_id is {%s}" % rev_id)
159
160
## XXX: Everything up to here can simply be orphaned if we abort
161
## the commit; it will leave junk files behind but that doesn't
162
## matter.
163
164
## TODO: Read back the just-generated changeset, and make sure it
165
## applies and recreates the right state.
166
167
## TODO: Also calculate and store the inventory SHA1
168
mutter("committing patch r%d" % (branch.revno() + 1))
169
170
branch.append_revision(rev_id)
171
172
branch.set_pending_merges([])
173
174
if verbose:
175
note("commited r%d" % branch.revno())
176
finally:
177
branch.unlock()
178
179
180
181
def _gen_revision_id(when):
433
182
"""Return new revision-id."""
434
s = '%s-%s-' % (config.user_email(), compact_date(when))
183
from binascii import hexlify
184
from osutils import rand_bytes, compact_date, user_email
185
186
s = '%s-%s-' % (user_email(), compact_date(when))
435
187
s += hexlify(rand_bytes(8))
436
188
return s
189
190
191
def _gather_commit(branch, work_tree, work_inv, basis_inv, specific_files,
192
verbose):
193
"""Build inventory preparatory to commit.
194
195
This adds any changed files into the text store, and sets their
196
test-id, sha and size in the returned inventory appropriately.
197
198
missing_ids
199
Modified to hold a list of files that have been deleted from
200
the working directory; these should be removed from the
201
working inventory.
202
"""
203
from bzrlib.inventory import Inventory
204
from osutils import isdir, isfile, sha_string, quotefn, \
205
local_time_offset, username, kind_marker, is_inside_any
206
207
from branch import gen_file_id
208
from errors import BzrError
209
from revision import Revision
210
from bzrlib.trace import mutter, note
211
212
inv = Inventory()
213
missing_ids = []
214
215
for path, entry in work_inv.iter_entries():
216
## TODO: Check that the file kind has not changed from the previous
217
## revision of this file (if any).
218
219
p = branch.abspath(path)
220
file_id = entry.file_id
221
mutter('commit prep file %s, id %r ' % (p, file_id))
222
223
if specific_files and not is_inside_any(specific_files, path):
224
if basis_inv.has_id(file_id):
225
# carry over with previous state
226
inv.add(basis_inv[file_id].copy())
227
else:
228
# omit this from committed inventory
229
pass
230
continue
231
232
if not work_tree.has_id(file_id):
233
if verbose:
234
print('deleted %s%s' % (path, kind_marker(entry.kind)))
235
mutter(" file is missing, removing from inventory")
236
missing_ids.append(file_id)
237
continue
238
239
# this is present in the new inventory; may be new, modified or
240
# unchanged.
241
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
242
243
entry = entry.copy()
244
inv.add(entry)
245
246
if old_ie:
247
old_kind = old_ie.kind
248
if old_kind != entry.kind:
249
raise BzrError("entry %r changed kind from %r to %r"
250
% (file_id, old_kind, entry.kind))
251
252
if entry.kind == 'directory':
253
if not isdir(p):
254
raise BzrError("%s is entered as directory but not a directory"
255
% quotefn(p))
256
elif entry.kind == 'file':
257
if not isfile(p):
258
raise BzrError("%s is entered as file but is not a file" % quotefn(p))
259
260
new_sha1 = work_tree.get_file_sha1(file_id)
261
262
if (old_ie
263
and old_ie.text_sha1 == new_sha1):
264
## assert content == basis.get_file(file_id).read()
265
entry.text_id = old_ie.text_id
266
entry.text_sha1 = new_sha1
267
entry.text_size = old_ie.text_size
268
mutter(' unchanged from previous text_id {%s}' %
269
entry.text_id)
270
else:
271
content = file(p, 'rb').read()
272
273
# calculate the sha again, just in case the file contents
274
# changed since we updated the cache
275
entry.text_sha1 = sha_string(content)
276
entry.text_size = len(content)
277
278
entry.text_id = gen_file_id(entry.name)
279
branch.text_store.add(content, entry.text_id)
280
mutter(' stored with text_id {%s}' % entry.text_id)
281
282
if verbose:
283
marked = path + kind_marker(entry.kind)
284
if not old_ie:
285
print 'added', marked
286
elif old_ie == entry:
287
pass # unchanged
288
elif (old_ie.name == entry.name
289
and old_ie.parent_id == entry.parent_id):
290
print 'modified', marked
291
else:
292
print 'renamed', marked
293
294
return missing_ids, inv
295
296
Loggerhead is a web-based interface for Breezy
Version: 2.0.1