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 1198
- compare with another revision
- download diff
- download tarball
- view history from revision 1198
- Committer: Martin Pool
- Date: 2005-09-08 11:17:45 UTC
- Revision ID: mbp@sourcefrog.net-20050908111745-b36aa1b62bb3c73b
- move away newinventory module which isn't actually used
- files added:
- HACKING
- bzrlib/util
- files removed:
- plugins/changeset
- files renamed:
- plugins/ => bzrlib/plugins/
- 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
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.
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
30
31
import os
32
import sys
33
import time
34
import tempfile
35
import sha
36
37
from binascii import hexlify
38
from cStringIO import StringIO
39
40
from bzrlib.osutils import (local_time_offset, username,
41
rand_bytes, compact_date, user_email,
42
kind_marker, is_inside_any, quotefn,
43
sha_string, sha_file, isdir, isfile)
44
from bzrlib.branch import gen_file_id
45
from bzrlib.errors import BzrError, PointlessCommit
46
from bzrlib.revision import Revision, RevisionReference
47
from bzrlib.trace import mutter, note
48
from bzrlib.xml5 import serializer_v5
49
from bzrlib.inventory import Inventory
50
from bzrlib.delta import compare_trees
51
from bzrlib.weave import Weave
52
from bzrlib.weavefile import read_weave, write_weave_v5
53
from bzrlib.atomicfile import AtomicFile
54
55
56
class NullCommitReporter(object):
57
"""I report on progress of a commit."""
58
def added(self, path):
59
pass
60
61
def removed(self, path):
62
pass
63
64
def renamed(self, old_path, new_path):
65
pass
66
67
68
class ReportCommitToLog(NullCommitReporter):
69
def added(self, path):
70
note('added %s', path)
71
72
def removed(self, path):
73
note('removed %s', path)
74
75
def renamed(self, old_path, new_path):
76
note('renamed %s => %s', old_path, new_path)
77
78
79
class Commit(object):
80
"""Task of committing a new revision.
81
82
This is a MethodObject: it accumulates state as the commit is
83
prepared, and then it is discarded. It doesn't represent
84
historical revisions, just the act of recording a new one.
85
86
missing_ids
87
Modified to hold a list of files that have been deleted from
88
the working directory; these should be removed from the
89
working inventory.
60
90
"""
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
119
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)
91
def __init__(self,
92
reporter=None):
93
if reporter is not None:
94
self.reporter = reporter
95
else:
96
self.reporter = NullCommitReporter()
97
98
99
def commit(self,
100
branch, message,
101
timestamp=None,
102
timezone=None,
103
committer=None,
104
specific_files=None,
105
rev_id=None,
106
allow_pointless=True):
107
"""Commit working copy as a new revision.
108
109
The basic approach is to add all the file texts into the
110
store, then the inventory, then make a new revision pointing
111
to that inventory and store that.
112
113
This is not quite safe if the working copy changes during the
114
commit; for the moment that is simply not allowed. A better
115
approach is to make a temporary copy of the files before
116
computing their hashes, and then add those hashes in turn to
117
the inventory. This should mean at least that there are no
118
broken hash pointers. There is no way we can get a snapshot
119
of the whole directory at an instant. This would also have to
120
be robust against files disappearing, moving, etc. So the
121
whole thing is a bit hard.
122
123
This raises PointlessCommit if there are no changes, no new merges,
124
and allow_pointless is false.
125
126
timestamp -- if not None, seconds-since-epoch for a
127
postdated/predated commit.
128
129
specific_files
130
If true, commit only those files.
131
132
rev_id
133
If set, use this as the new revision id.
134
Useful for test or import commands that need to tightly
135
control what revisions are assigned. If you duplicate
136
a revision id that exists elsewhere it is your own fault.
137
If null (default), a time/random revision id is generated.
138
"""
139
140
self.branch = branch
141
self.branch.lock_write()
142
self.rev_id = rev_id
143
self.specific_files = specific_files
144
self.allow_pointless = allow_pointless
145
146
if timestamp is None:
147
self.timestamp = time.time()
148
else:
149
self.timestamp = long(timestamp)
150
151
if committer is None:
152
self.committer = username(self.branch)
153
else:
154
assert isinstance(committer, basestring), type(committer)
155
self.committer = committer
156
157
if timezone is None:
158
self.timezone = local_time_offset()
159
else:
160
self.timezone = int(timezone)
161
162
assert isinstance(message, basestring), type(message)
163
self.message = message
164
165
try:
166
# First walk over the working inventory; and both update that
167
# and also build a new revision inventory. The revision
168
# inventory needs to hold the text-id, sha1 and size of the
169
# actual file versions committed in the revision. (These are
170
# not present in the working inventory.) We also need to
171
# detect missing/deleted files, and remove them from the
172
# working inventory.
173
174
self.work_tree = self.branch.working_tree()
175
self.work_inv = self.work_tree.inventory
176
self.basis_tree = self.branch.basis_tree()
177
self.basis_inv = self.basis_tree.inventory
178
179
self.pending_merges = self.branch.pending_merges()
180
181
if self.rev_id is None:
182
self.rev_id = _gen_revision_id(self.branch, time.time())
183
184
# todo: update hashcache
185
self.delta = compare_trees(self.basis_tree, self.work_tree,
186
specific_files=self.specific_files)
187
188
if not (self.delta.has_changed()
189
or self.allow_pointless
190
or self.pending_merges):
191
raise PointlessCommit()
192
193
self.new_inv = self.basis_inv.copy()
194
195
self.delta.show(sys.stdout)
196
197
self._remove_deleted()
198
self._store_files()
199
200
self.branch._write_inventory(self.work_inv)
201
self._record_inventory()
202
203
self._make_revision()
204
note('committted r%d {%s}', (self.branch.revno() + 1),
205
self.rev_id)
206
self.branch.append_revision(self.rev_id)
207
self.branch.set_pending_merges([])
208
finally:
209
self.branch.unlock()
210
211
212
def _record_inventory(self):
213
inv_tmp = StringIO()
214
serializer_v5.write_inventory(self.new_inv, inv_tmp)
215
self.inv_sha1 = sha_string(inv_tmp.getvalue())
125
216
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()
217
self.branch.inventory_store.add(inv_tmp, self.rev_id)
218
219
220
def _make_revision(self):
221
"""Record a new revision object for this commit."""
222
self.rev = Revision(timestamp=self.timestamp,
223
timezone=self.timezone,
224
committer=self.committer,
225
message=self.message,
226
inventory_sha1=self.inv_sha1,
227
revision_id=self.rev_id)
228
229
self.rev.parents = []
230
precursor_id = self.branch.last_patch()
156
231
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))
232
self.rev.parents.append(RevisionReference(precursor_id))
233
for merge_rev in self.pending_merges:
234
rev.parents.append(RevisionReference(merge_rev))
161
235
162
236
rev_tmp = tempfile.TemporaryFile()
163
pack_xml(rev, rev_tmp)
237
serializer_v5.write_revision(self.rev, rev_tmp)
164
238
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):
239
self.branch.revision_store.add(rev_tmp, self.rev_id)
240
mutter('new revision_id is {%s}', self.rev_id)
241
242
243
def _remove_deleted(self):
244
"""Remove deleted files from the working and stored inventories."""
245
for path, id, kind in self.delta.removed:
246
if self.work_inv.has_id(id):
247
del self.work_inv[id]
248
if self.new_inv.has_id(id):
249
del self.new_inv[id]
250
251
252
253
def _store_files(self):
254
"""Store new texts of modified/added files."""
255
for path, id, kind in self.delta.modified:
256
if kind != 'file':
257
continue
258
self._store_file_text(id)
259
260
for path, id, kind in self.delta.added:
261
if kind != 'file':
262
continue
263
self._store_file_text(id)
264
265
for old_path, new_path, id, kind, text_modified in self.delta.renamed:
266
if kind != 'file':
267
continue
268
if not text_modified:
269
continue
270
self._store_file_text(id)
271
272
273
def _store_file_text(self, file_id):
274
"""Store updated text for one modified or added file."""
275
note('store new text for {%s} in revision {%s}', id, self.rev_id)
276
new_lines = self.work_tree.get_file(file_id).readlines()
277
self._add_text_to_weave(file_id, new_lines)
278
# update or add an entry
279
if file_id in self.new_inv:
280
ie = self.new_inv[file_id]
281
assert ie.file_id == file_id
282
else:
283
ie = self.work_inv[file_id].copy()
284
self.new_inv.add(ie)
285
assert ie.kind == 'file'
286
# make a new inventory entry for this file, using whatever
287
# it had in the working copy, plus details on the new text
288
ie.text_sha1 = _sha_strings(new_lines)
289
ie.text_size = sum(map(len, new_lines))
290
ie.text_version = self.rev_id
291
ie.entry_version = self.rev_id
292
293
294
def _add_text_to_weave(self, file_id, new_lines):
295
weave_fn = self.branch.controlfilename(['weaves', file_id+'.weave'])
296
if os.path.exists(weave_fn):
297
w = read_weave(file(weave_fn, 'rb'))
298
else:
299
w = Weave()
300
# XXX: Should set the appropriate parents by looking for this file_id
301
# in all revision parents
302
w.add(self.rev_id, [], new_lines)
303
af = AtomicFile(weave_fn)
304
try:
305
write_weave_v5(w, af)
306
af.commit()
307
finally:
308
af.close()
309
310
311
def _gen_revision_id(branch, when):
190
312
"""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))
313
s = '%s-%s-' % (user_email(branch), compact_date(when))
195
314
s += hexlify(rand_bytes(8))
196
315
return s
197
316
198
317
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
318
def _sha_strings(strings):
319
"""Return the sha-1 of concatenation of strings"""
320
s = sha.new()
321
map(s.update, strings)
322
return s.hexdigest()
Loggerhead is a web-based interface for Breezy
Version: 2.0.2.dev0