« back to all changes in this revision
Viewing changes to bzrlib/branch.py
- browse files at revision 1418
- compare with another revision
- download diff
- download tarball
- view history from revision 1418
- Committer: Robert Collins
- Date: 2005-10-06 12:14:01 UTC
- mfrom: (1393.1.67)
- Revision ID: robertc@robertcollins.net-20051006121401-ce87bcb93909bbdf
merge martins latest
- files added:
- HACKING
- bzrlib/plugins
- bzrlib/selftest
- bzrlib/store
- bzrlib/transport
- bzrlib/util
- bzrlib/util/effbot
- bzrlib/util/effbot/org
- contrib/emacs
- patches
- tools
- files removed:
- bzrlib/cache.py
- files renamed:
- bzrlib/store.py => bzrlib/store/__init__.py
- elementtree/ => bzrlib/util/elementtree/
- urlgrabber/ => bzrlib/util/urlgrabber/
- contrib/bash/bzr => contrib/bash/bzr.simple
- bzrlib/newinventory.py => contrib/newinventory.py
- files modified:
- .bzrignore
added
removed
bzrlib/branch.py
15
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
16
16
17
17
18
from sets import Set
18
import sys
19
import os
20
import errno
21
from warnings import warn
22
from cStringIO import StringIO
19
23
20
import sys, os, os.path, random, time, sha, sets, types, re, shutil, tempfile
21
import traceback, socket, fnmatch, difflib, time
22
from binascii import hexlify
23
24
24
25
import bzrlib
25
from inventory import Inventory
26
from trace import mutter, note
27
from tree import Tree, EmptyTree, RevisionTree, WorkingTree
28
from inventory import InventoryEntry, Inventory
29
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, \
30
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
31
joinpath, sha_string, file_kind, local_time_offset, appendpath
32
from store import ImmutableStore
33
from revision import Revision
34
from errors import bailout, BzrError
35
from textui import show_status
36
from diff import diff_trees
37
38
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
26
from bzrlib.inventory import InventoryEntry
27
import bzrlib.inventory as inventory
28
from bzrlib.trace import mutter, note
29
from bzrlib.osutils import (isdir, quotefn, compact_date, rand_bytes,
30
rename, splitpath, sha_file, appendpath,
31
file_kind)
32
from bzrlib.errors import (BzrError, InvalidRevisionNumber, InvalidRevisionId,
33
NoSuchRevision, HistoryMissing, NotBranchError,
34
DivergedBranches, LockError, UnlistableStore,
35
UnlistableBranch, NoSuchFile)
36
from bzrlib.textui import show_status
37
from bzrlib.revision import Revision
38
from bzrlib.delta import compare_trees
39
from bzrlib.tree import EmptyTree, RevisionTree
40
from bzrlib.inventory import Inventory
41
from bzrlib.store import copy_all
42
from bzrlib.store.compressed_text import CompressedTextStore
43
from bzrlib.store.text import TextStore
44
from bzrlib.store.weave import WeaveStore
45
from bzrlib.transport import Transport, get_transport
46
import bzrlib.xml5
47
import bzrlib.ui
48
49
50
BZR_BRANCH_FORMAT_4 = "Bazaar-NG branch, format 0.0.4\n"
51
BZR_BRANCH_FORMAT_5 = "Bazaar-NG branch, format 5\n"
39
52
## TODO: Maybe include checks for common corruption of newlines, etc?
40
53
41
54
42
43
def find_branch(f, **args):
44
if f.startswith('http://') or f.startswith('https://'):
45
import remotebranch
46
return remotebranch.RemoteBranch(f, **args)
55
# TODO: Some operations like log might retrieve the same revisions
56
# repeatedly to calculate deltas. We could perhaps have a weakref
57
# cache in memory to make this faster. In general anything can be
58
# cached in memory between lock and unlock operations.
59
60
def find_branch(*ignored, **ignored_too):
61
# XXX: leave this here for about one release, then remove it
62
raise NotImplementedError('find_branch() is not supported anymore, '
63
'please use one of the new branch constructors')
64
def _relpath(base, path):
65
"""Return path relative to base, or raise exception.
66
67
The path may be either an absolute path or a path relative to the
68
current working directory.
69
70
Lifted out of Branch.relpath for ease of testing.
71
72
os.path.commonprefix (python2.4) has a bad bug that it works just
73
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
74
avoids that problem."""
75
rp = os.path.abspath(path)
76
77
s = []
78
head = rp
79
while len(head) >= len(base):
80
if head == base:
81
break
82
head, tail = os.path.split(head)
83
if tail:
84
s.insert(0, tail)
47
85
else:
48
return Branch(f, **args)
86
raise NotBranchError("path %r is not within branch %r" % (rp, base))
87
88
return os.sep.join(s)
49
89
50
90
51
def find_branch_root(f=None):
52
"""Find the branch root enclosing f, or pwd.
53
54
f may be a filename or a URL.
55
56
It is not necessary that f exists.
91
def find_branch_root(t):
92
"""Find the branch root enclosing the transport's base.
93
94
t is a Transport object.
95
96
It is not necessary that the base of t exists.
57
97
58
98
Basically we keep looking up until we find the control directory or
59
run into the root."""
60
if f == None:
61
f = os.getcwd()
62
elif hasattr(os.path, 'realpath'):
63
f = os.path.realpath(f)
64
else:
65
f = os.path.abspath(f)
66
if not os.path.exists(f):
67
raise BzrError('%r does not exist' % f)
68
69
70
orig_f = f
71
99
run into the root. If there isn't one, raises NotBranchError.
100
"""
101
orig_base = t.base
72
102
while True:
73
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
74
return f
75
head, tail = os.path.split(f)
76
if head == f:
103
if t.has(bzrlib.BZRDIR):
104
return t
105
new_t = t.clone('..')
106
if new_t.base == t.base:
77
107
# reached the root, whatever that may be
78
raise BzrError('%r is not in a branch' % orig_f)
79
f = head
80
108
raise NotBranchError('%s is not in a branch' % orig_base)
109
t = new_t
81
110
82
111
83
112
######################################################################
84
113
# branch objects
85
114
86
class Branch:
115
class Branch(object):
87
116
"""Branch holding a history of revisions.
88
117
89
118
base
90
Base directory of the branch.
91
"""
92
_lockmode = None
119
Base directory/url of the branch.
120
"""
121
base = None
122
123
def __init__(self, *ignored, **ignored_too):
124
raise NotImplementedError('The Branch class is abstract')
125
126
@staticmethod
127
def open_downlevel(base):
128
"""Open a branch which may be of an old format.
129
130
Only local branches are supported."""
131
return _Branch(get_transport(base), relax_version_check=True)
132
133
@staticmethod
134
def open(base):
135
"""Open an existing branch, rooted at 'base' (url)"""
136
t = get_transport(base)
137
mutter("trying to open %r with transport %r", base, t)
138
return _Branch(t)
139
140
@staticmethod
141
def open_containing(url):
142
"""Open an existing branch which contains url.
143
144
This probes for a branch at url, and searches upwards from there.
145
"""
146
t = get_transport(url)
147
t = find_branch_root(t)
148
return _Branch(t)
149
150
@staticmethod
151
def initialize(base):
152
"""Create a new branch, rooted at 'base' (url)"""
153
t = get_transport(base)
154
return _Branch(t, init=True)
155
156
def setup_caching(self, cache_root):
157
"""Subclasses that care about caching should override this, and set
158
up cached stores located under cache_root.
159
"""
160
self.cache_root = cache_root
161
162
163
class _Branch(Branch):
164
"""A branch stored in the actual filesystem.
165
166
Note that it's "local" in the context of the filesystem; it doesn't
167
really matter if it's on an nfs/smb/afs/coda/... share, as long as
168
it's writable, and can be accessed via the normal filesystem API.
169
170
_lock_mode
171
None, or 'r' or 'w'
172
173
_lock_count
174
If _lock_mode is true, a positive count of the number of times the
175
lock has been taken.
176
177
_lock
178
Lock object from bzrlib.lock.
179
"""
180
# We actually expect this class to be somewhat short-lived; part of its
181
# purpose is to try to isolate what bits of the branch logic are tied to
182
# filesystem access, so that in a later step, we can extricate them to
183
# a separarte ("storage") class.
184
_lock_mode = None
185
_lock_count = None
186
_lock = None
187
_inventory_weave = None
93
188
94
def __init__(self, base, init=False, find_root=True, lock_mode='w'):
189
# Map some sort of prefix into a namespace
190
# stuff like "revno:10", "revid:", etc.
191
# This should match a prefix with a function which accepts
192
REVISION_NAMESPACES = {}
193
194
def push_stores(self, branch_to):
195
"""Copy the content of this branches store to branch_to."""
196
if (self._branch_format != branch_to._branch_format
197
or self._branch_format != 4):
198
from bzrlib.fetch import greedy_fetch
199
mutter("falling back to fetch logic to push between %s(%s) and %s(%s)",
200
self, self._branch_format, branch_to, branch_to._branch_format)
201
greedy_fetch(to_branch=branch_to, from_branch=self,
202
revision=self.last_revision())
203
return
204
205
store_pairs = ((self.text_store, branch_to.text_store),
206
(self.inventory_store, branch_to.inventory_store),
207
(self.revision_store, branch_to.revision_store))
208
try:
209
for from_store, to_store in store_pairs:
210
copy_all(from_store, to_store)
211
except UnlistableStore:
212
raise UnlistableBranch(from_store)
213
214
def __init__(self, transport, init=False,
215
relax_version_check=False):
95
216
"""Create new branch object at a particular location.
96
217
97
base -- Base directory for the branch.
218
transport -- A Transport object, defining how to access files.
219
(If a string, transport.transport() will be used to
220
create a Transport object)
98
221
99
222
init -- If True, create new control files in a previously
100
223
unversioned directory. If False, the branch must already
101
224
be versioned.
102
225
103
find_root -- If true and init is false, find the root of the
104
existing branch containing base.
226
relax_version_check -- If true, the usual check for the branch
227
version is not applied. This is intended only for
228
upgrade/recovery type use; it's not guaranteed that
229
all operations will work on old format branches.
105
230
106
231
In the test suite, creation of new trees is tested using the
107
232
`ScratchBranch` class.
108
233
"""
234
assert isinstance(transport, Transport), \
235
"%r is not a Transport" % transport
236
self._transport = transport
109
237
if init:
110
self.base = os.path.realpath(base)
111
238
self._make_control()
112
elif find_root:
113
self.base = find_branch_root(base)
114
else:
115
self.base = os.path.realpath(base)
116
if not isdir(self.controlfilename('.')):
117
bailout("not a bzr branch: %s" % quotefn(base),
118
['use "bzr init" to initialize a new working tree',
119
'current bzr can only operate from top-of-tree'])
120
self._check_format()
121
self.lock(lock_mode)
122
123
self.text_store = ImmutableStore(self.controlfilename('text-store'))
124
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
125
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
126
239
self._check_format(relax_version_check)
240
241
def get_store(name, compressed=True):
242
# FIXME: This approach of assuming stores are all entirely compressed
243
# or entirely uncompressed is tidy, but breaks upgrade from
244
# some existing branches where there's a mixture; we probably
245
# still want the option to look for both.
246
relpath = self._rel_controlfilename(name)
247
if compressed:
248
store = CompressedTextStore(self._transport.clone(relpath))
249
else:
250
store = TextStore(self._transport.clone(relpath))
251
#if self._transport.should_cache():
252
# cache_path = os.path.join(self.cache_root, name)
253
# os.mkdir(cache_path)
254
# store = bzrlib.store.CachedStore(store, cache_path)
255
return store
256
def get_weave(name):
257
relpath = self._rel_controlfilename(name)
258
ws = WeaveStore(self._transport.clone(relpath))
259
if self._transport.should_cache():
260
ws.enable_cache = True
261
return ws
262
263
if self._branch_format == 4:
264
self.inventory_store = get_store('inventory-store')
265
self.text_store = get_store('text-store')
266
self.revision_store = get_store('revision-store')
267
elif self._branch_format == 5:
268
self.control_weaves = get_weave([])
269
self.weave_store = get_weave('weaves')
270
self.revision_store = get_store('revision-store', compressed=False)
127
271
128
272
def __str__(self):
129
return '%s(%r)' % (self.__class__.__name__, self.base)
273
return '%s(%r)' % (self.__class__.__name__, self._transport.base)
130
274
131
275
132
276
__repr__ = __str__
133
277
134
278
135
136
def lock(self, mode='w'):
137
"""Lock the on-disk branch, excluding other processes."""
138
try:
139
import fcntl, errno
140
141
if mode == 'w':
142
lm = fcntl.LOCK_EX
143
om = os.O_WRONLY | os.O_CREAT
144
elif mode == 'r':
145
lm = fcntl.LOCK_SH
146
om = os.O_RDONLY
147
else:
148
raise BzrError("invalid locking mode %r" % mode)
149
279
def __del__(self):
280
if self._lock_mode or self._lock:
281
# XXX: This should show something every time, and be suitable for
282
# headless operation and embedding
283
warn("branch %r was not explicitly unlocked" % self)
284
self._lock.unlock()
285
286
# TODO: It might be best to do this somewhere else,
287
# but it is nice for a Branch object to automatically
288
# cache it's information.
289
# Alternatively, we could have the Transport objects cache requests
290
# See the earlier discussion about how major objects (like Branch)
291
# should never expect their __del__ function to run.
292
if hasattr(self, 'cache_root') and self.cache_root is not None:
150
293
try:
151
lockfile = os.open(self.controlfilename('branch-lock'), om)
152
except OSError, e:
153
if e.errno == errno.ENOENT:
154
# might not exist on branches from <0.0.4
155
self.controlfile('branch-lock', 'w').close()
156
lockfile = os.open(self.controlfilename('branch-lock'), om)
157
else:
158
raise e
159
160
fcntl.lockf(lockfile, lm)
161
def unlock():
162
fcntl.lockf(lockfile, fcntl.LOCK_UN)
163
os.close(lockfile)
164
self._lockmode = None
165
self.unlock = unlock
166
self._lockmode = mode
167
except ImportError:
168
warning("please write a locking method for platform %r" % sys.platform)
169
def unlock():
170
self._lockmode = None
171
self.unlock = unlock
172
self._lockmode = mode
173
174
175
def _need_readlock(self):
176
if self._lockmode not in ['r', 'w']:
177
raise BzrError('need read lock on branch, only have %r' % self._lockmode)
178
179
def _need_writelock(self):
180
if self._lockmode not in ['w']:
181
raise BzrError('need write lock on branch, only have %r' % self._lockmode)
182
294
import shutil
295
shutil.rmtree(self.cache_root)
296
except:
297
pass
298
self.cache_root = None
299
300
def _get_base(self):
301
if self._transport:
302
return self._transport.base
303
return None
304
305
base = property(_get_base)
306
307
308
def lock_write(self):
309
# TODO: Upgrade locking to support using a Transport,
310
# and potentially a remote locking protocol
311
if self._lock_mode:
312
if self._lock_mode != 'w':
313
raise LockError("can't upgrade to a write lock from %r" %
314
self._lock_mode)
315
self._lock_count += 1
316
else:
317
self._lock = self._transport.lock_write(
318
self._rel_controlfilename('branch-lock'))
319
self._lock_mode = 'w'
320
self._lock_count = 1
321
322
323
def lock_read(self):
324
if self._lock_mode:
325
assert self._lock_mode in ('r', 'w'), \
326
"invalid lock mode %r" % self._lock_mode
327
self._lock_count += 1
328
else:
329
self._lock = self._transport.lock_read(
330
self._rel_controlfilename('branch-lock'))
331
self._lock_mode = 'r'
332
self._lock_count = 1
333
334
def unlock(self):
335
if not self._lock_mode:
336
raise LockError('branch %r is not locked' % (self))
337
338
if self._lock_count > 1:
339
self._lock_count -= 1
340
else:
341
self._lock.unlock()
342
self._lock = None
343
self._lock_mode = self._lock_count = None
183
344
184
345
def abspath(self, name):
185
346
"""Return absolute filename for something in the branch"""
186
return os.path.join(self.base, name)
187
347
return self._transport.abspath(name)
188
348
189
349
def relpath(self, path):
190
350
"""Return path relative to this branch of something inside it.
191
351
192
352
Raises an error if path is not in this branch."""
193
rp = os.path.realpath(path)
194
# FIXME: windows
195
if not rp.startswith(self.base):
196
bailout("path %r is not within branch %r" % (rp, self.base))
197
rp = rp[len(self.base):]
198
rp = rp.lstrip(os.sep)
199
return rp
200
353
return self._transport.relpath(path)
354
355
356
def _rel_controlfilename(self, file_or_path):
357
if isinstance(file_or_path, basestring):
358
file_or_path = [file_or_path]
359
return [bzrlib.BZRDIR] + file_or_path
201
360
202
361
def controlfilename(self, file_or_path):
203
362
"""Return location relative to branch."""
204
if isinstance(file_or_path, types.StringTypes):
205
file_or_path = [file_or_path]
206
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
363
return self._transport.abspath(self._rel_controlfilename(file_or_path))
207
364
208
365
209
366
def controlfile(self, file_or_path, mode='r'):
213
370
and binary. binary files are untranslated byte streams. Text
214
371
control files are stored with Unix newlines and in UTF-8, even
215
372
if the platform or locale defaults are different.
373
374
Controlfiles should almost never be opened in write mode but
375
rather should be atomically copied and replaced using atomicfile.
216
376
"""
217
218
fn = self.controlfilename(file_or_path)
219
220
if mode == 'rb' or mode == 'wb':
221
return file(fn, mode)
222
elif mode == 'r' or mode == 'w':
223
# open in binary mode anyhow so there's no newline translation;
224
# codecs uses line buffering by default; don't want that.
225
import codecs
226
return codecs.open(fn, mode + 'b', 'utf-8',
227
buffering=60000)
377
import codecs
378
379
relpath = self._rel_controlfilename(file_or_path)
380
#TODO: codecs.open() buffers linewise, so it was overloaded with
381
# a much larger buffer, do we need to do the same for getreader/getwriter?
382
if mode == 'rb':
383
return self._transport.get(relpath)
384
elif mode == 'wb':
385
raise BzrError("Branch.controlfile(mode='wb') is not supported, use put_controlfiles")
386
elif mode == 'r':
387
return codecs.getreader('utf-8')(self._transport.get(relpath), errors='replace')
388
elif mode == 'w':
389
raise BzrError("Branch.controlfile(mode='w') is not supported, use put_controlfiles")
228
390
else:
229
391
raise BzrError("invalid controlfile mode %r" % mode)
230
392
231
393
def put_controlfile(self, path, f, encode=True):
394
"""Write an entry as a controlfile.
395
396
:param path: The path to put the file, relative to the .bzr control
397
directory
398
:param f: A file-like or string object whose contents should be copied.
399
:param encode: If true, encode the contents as utf-8
400
"""
401
self.put_controlfiles([(path, f)], encode=encode)
402
403
def put_controlfiles(self, files, encode=True):
404
"""Write several entries as controlfiles.
405
406
:param files: A list of [(path, file)] pairs, where the path is the directory
407
underneath the bzr control directory
408
:param encode: If true, encode the contents as utf-8
409
"""
410
import codecs
411
ctrl_files = []
412
for path, f in files:
413
if encode:
414
if isinstance(f, basestring):
415
f = f.encode('utf-8', 'replace')
416
else:
417
f = codecs.getwriter('utf-8')(f, errors='replace')
418
path = self._rel_controlfilename(path)
419
ctrl_files.append((path, f))
420
self._transport.put_multi(ctrl_files)
232
421
233
422
def _make_control(self):
234
os.mkdir(self.controlfilename([]))
235
self.controlfile('README', 'w').write(
423
from bzrlib.inventory import Inventory
424
from bzrlib.weavefile import write_weave_v5
425
from bzrlib.weave import Weave
426
427
# Create an empty inventory
428
sio = StringIO()
429
# if we want per-tree root ids then this is the place to set
430
# them; they're not needed for now and so ommitted for
431
# simplicity.
432
bzrlib.xml5.serializer_v5.write_inventory(Inventory(), sio)
433
empty_inv = sio.getvalue()
434
sio = StringIO()
435
bzrlib.weavefile.write_weave_v5(Weave(), sio)
436
empty_weave = sio.getvalue()
437
438
dirs = [[], 'revision-store', 'weaves']
439
files = [('README',
236
440
"This is a Bazaar-NG control directory.\n"
237
"Do not change any files in this directory.")
238
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
239
for d in ('text-store', 'inventory-store', 'revision-store'):
240
os.mkdir(self.controlfilename(d))
241
for f in ('revision-history', 'merged-patches',
242
'pending-merged-patches', 'branch-name',
243
'branch-lock'):
244
self.controlfile(f, 'w').write('')
245
mutter('created control directory in ' + self.base)
246
Inventory().write_xml(self.controlfile('inventory','w'))
247
248
249
def _check_format(self):
441
"Do not change any files in this directory.\n"),
442
('branch-format', BZR_BRANCH_FORMAT_5),
443
('revision-history', ''),
444
('branch-name', ''),
445
('branch-lock', ''),
446
('pending-merges', ''),
447
('inventory', empty_inv),
448
('inventory.weave', empty_weave),
449
('ancestry.weave', empty_weave)
450
]
451
cfn = self._rel_controlfilename
452
self._transport.mkdir_multi([cfn(d) for d in dirs])
453
self.put_controlfiles(files)
454
mutter('created control directory in ' + self._transport.base)
455
456
def _check_format(self, relax_version_check):
250
457
"""Check this branch format is supported.
251
458
252
The current tool only supports the current unstable format.
459
The format level is stored, as an integer, in
460
self._branch_format for code that needs to check it later.
253
461
254
462
In the future, we might need different in-memory Branch
255
463
classes to support downlevel branches. But not yet.
256
464
"""
257
# This ignores newlines so that we can open branches created
258
# on Windows from Linux and so on. I think it might be better
259
# to always make all internal files in unix format.
260
fmt = self.controlfile('branch-format', 'r').read()
261
fmt.replace('\r\n', '')
262
if fmt != BZR_BRANCH_FORMAT:
263
bailout('sorry, branch format %r not supported' % fmt,
264
['use a different bzr version',
265
'or remove the .bzr directory and "bzr init" again'])
266
465
try:
466
fmt = self.controlfile('branch-format', 'r').read()
467
except NoSuchFile:
468
raise NotBranchError(self.base)
469
mutter("got branch format %r", fmt)
470
if fmt == BZR_BRANCH_FORMAT_5:
471
self._branch_format = 5
472
elif fmt == BZR_BRANCH_FORMAT_4:
473
self._branch_format = 4
474
475
if (not relax_version_check
476
and self._branch_format != 5):
477
raise BzrError('sorry, branch format %r not supported' % fmt,
478
['use a different bzr version',
479
'or remove the .bzr directory'
480
' and "bzr init" again'])
481
482
def get_root_id(self):
483
"""Return the id of this branches root"""
484
inv = self.read_working_inventory()
485
return inv.root.file_id
486
487
def set_root_id(self, file_id):
488
inv = self.read_working_inventory()
489
orig_root_id = inv.root.file_id
490
del inv._byid[inv.root.file_id]
491
inv.root.file_id = file_id
492
inv._byid[inv.root.file_id] = inv.root
493
for fid in inv:
494
entry = inv[fid]
495
if entry.parent_id in (None, orig_root_id):
496
entry.parent_id = inv.root.file_id
497
self._write_inventory(inv)
267
498
268
499
def read_working_inventory(self):
269
500
"""Read the working inventory."""
270
self._need_readlock()
271
before = time.time()
272
# ElementTree does its own conversion from UTF-8, so open in
273
# binary.
274
inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))
275
mutter("loaded inventory of %d items in %f"
276
% (len(inv), time.time() - before))
277
return inv
278
501
self.lock_read()
502
try:
503
# ElementTree does its own conversion from UTF-8, so open in
504
# binary.
505
f = self.controlfile('inventory', 'rb')
506
return bzrlib.xml5.serializer_v5.read_inventory(f)
507
finally:
508
self.unlock()
509
279
510
280
511
def _write_inventory(self, inv):
281
512
"""Update the working inventory.
283
514
That is to say, the inventory describing changes underway, that
284
515
will be committed to the next revision.
285
516
"""
286
self._need_writelock()
287
## TODO: factor out to atomicfile? is rename safe on windows?
288
## TODO: Maybe some kind of clean/dirty marker on inventory?
289
tmpfname = self.controlfilename('inventory.tmp')
290
tmpf = file(tmpfname, 'wb')
291
inv.write_xml(tmpf)
292
tmpf.close()
293
inv_fname = self.controlfilename('inventory')
294
if sys.platform == 'win32':
295
os.remove(inv_fname)
296
os.rename(tmpfname, inv_fname)
517
from cStringIO import StringIO
518
self.lock_write()
519
try:
520
sio = StringIO()
521
bzrlib.xml5.serializer_v5.write_inventory(inv, sio)
522
sio.seek(0)
523
# Transport handles atomicity
524
self.put_controlfile('inventory', sio)
525
finally:
526
self.unlock()
527
297
528
mutter('wrote working inventory')
298
299
529
300
530
inventory = property(read_working_inventory, _write_inventory, None,
301
531
"""Inventory for the working copy.""")
302
532
303
304
def add(self, files, verbose=False):
533
def add(self, files, ids=None):
305
534
"""Make files versioned.
306
535
307
Note that the command line normally calls smart_add instead.
536
Note that the command line normally calls smart_add instead,
537
which can automatically recurse.
308
538
309
539
This puts the files in the Added state, so that they will be
310
540
recorded by the next commit.
311
541
542
files
543
List of paths to add, relative to the base of the tree.
544
545
ids
546
If set, use these instead of automatically generated ids.
547
Must be the same length as the list of files, but may
548
contain None for ids that are to be autogenerated.
549
312
550
TODO: Perhaps have an option to add the ids even if the files do
313
not (yet) exist.
314
315
TODO: Perhaps return the ids of the files? But then again it
316
is easy to retrieve them if they're needed.
317
318
TODO: Option to specify file id.
319
320
TODO: Adding a directory should optionally recurse down and
321
add all non-ignored children. Perhaps do that in a
322
higher-level method.
323
324
>>> b = ScratchBranch(files=['foo'])
325
>>> 'foo' in b.unknowns()
326
True
327
>>> b.show_status()
328
? foo
329
>>> b.add('foo')
330
>>> 'foo' in b.unknowns()
331
False
332
>>> bool(b.inventory.path2id('foo'))
333
True
334
>>> b.show_status()
335
A foo
336
337
>>> b.add('foo')
338
Traceback (most recent call last):
339
...
340
BzrError: ('foo is already versioned', [])
341
342
>>> b.add(['nothere'])
343
Traceback (most recent call last):
344
BzrError: ('cannot add: not a regular file or directory: nothere', [])
551
not (yet) exist.
552
553
TODO: Perhaps yield the ids and paths as they're added.
345
554
"""
346
self._need_writelock()
347
348
555
# TODO: Re-adding a file that is removed in the working copy
349
556
# should probably put it back with the previous ID.
350
if isinstance(files, types.StringTypes):
557
if isinstance(files, basestring):
558
assert(ids is None or isinstance(ids, basestring))
351
559
files = [files]
352
353
inv = self.read_working_inventory()
354
for f in files:
355
if is_control_file(f):
356
bailout("cannot add control file %s" % quotefn(f))
357
358
fp = splitpath(f)
359
360
if len(fp) == 0:
361
bailout("cannot add top-level %r" % f)
362
363
fullpath = os.path.normpath(self.abspath(f))
364
365
try:
366
kind = file_kind(fullpath)
367
except OSError:
368
# maybe something better?
369
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
370
371
if kind != 'file' and kind != 'directory':
372
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
373
374
file_id = gen_file_id(f)
375
inv.add_path(f, kind=kind, file_id=file_id)
376
377
if verbose:
378
show_status('A', kind, quotefn(f))
379
380
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
381
382
self._write_inventory(inv)
383
560
if ids is not None:
561
ids = [ids]
562
563
if ids is None:
564
ids = [None] * len(files)
565
else:
566
assert(len(ids) == len(files))
567
568
self.lock_write()
569
try:
570
inv = self.read_working_inventory()
571
for f,file_id in zip(files, ids):
572
if is_control_file(f):
573
raise BzrError("cannot add control file %s" % quotefn(f))
574
575
fp = splitpath(f)
576
577
if len(fp) == 0:
578
raise BzrError("cannot add top-level %r" % f)
579
580
fullpath = os.path.normpath(self.abspath(f))
581
582
try:
583
kind = file_kind(fullpath)
584
except OSError:
585
# maybe something better?
586
raise BzrError('cannot add: not a regular file, symlink or directory: %s' % quotefn(f))
587
588
if not InventoryEntry.versionable_kind(kind):
589
raise BzrError('cannot add: not a versionable file ('
590
'i.e. regular file, symlink or directory): %s' % quotefn(f))
591
592
if file_id is None:
593
file_id = gen_file_id(f)
594
inv.add_path(f, kind=kind, file_id=file_id)
595
596
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
597
598
self._write_inventory(inv)
599
finally:
600
self.unlock()
601
384
602
385
603
def print_file(self, file, revno):
386
604
"""Print `file` to stdout."""
387
self._need_readlock()
388
tree = self.revision_tree(self.lookup_revision(revno))
389
# use inventory as it was in that revision
390
file_id = tree.inventory.path2id(file)
391
if not file_id:
392
bailout("%r is not present in revision %d" % (file, revno))
393
tree.print_file(file_id)
394
605
self.lock_read()
606
try:
607
tree = self.revision_tree(self.get_rev_id(revno))
608
# use inventory as it was in that revision
609
file_id = tree.inventory.path2id(file)
610
if not file_id:
611
raise BzrError("%r is not present in revision %s" % (file, revno))
612
tree.print_file(file_id)
613
finally:
614
self.unlock()
615
395
616
396
617
def remove(self, files, verbose=False):
397
618
"""Mark nominated files for removal from the inventory.
400
621
401
622
TODO: Refuse to remove modified files unless --force is given?
402
623
403
>>> b = ScratchBranch(files=['foo'])
404
>>> b.add('foo')
405
>>> b.inventory.has_filename('foo')
406
True
407
>>> b.remove('foo')
408
>>> b.working_tree().has_filename('foo')
409
True
410
>>> b.inventory.has_filename('foo')
411
False
412
413
>>> b = ScratchBranch(files=['foo'])
414
>>> b.add('foo')
415
>>> b.commit('one')
416
>>> b.remove('foo')
417
>>> b.commit('two')
418
>>> b.inventory.has_filename('foo')
419
False
420
>>> b.basis_tree().has_filename('foo')
421
False
422
>>> b.working_tree().has_filename('foo')
423
True
424
425
624
TODO: Do something useful with directories.
426
625
427
626
TODO: Should this remove the text or not? Tough call; not
431
630
"""
432
631
## TODO: Normalize names
433
632
## TODO: Remove nested loops; better scalability
434
self._need_writelock()
435
436
if isinstance(files, types.StringTypes):
633
if isinstance(files, basestring):
437
634
files = [files]
438
439
tree = self.working_tree()
440
inv = tree.inventory
441
442
# do this before any modifications
443
for f in files:
444
fid = inv.path2id(f)
445
if not fid:
446
bailout("cannot remove unversioned file %s" % quotefn(f))
447
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
448
if verbose:
449
# having remove it, it must be either ignored or unknown
450
if tree.is_ignored(f):
451
new_status = 'I'
452
else:
453
new_status = '?'
454
show_status(new_status, inv[fid].kind, quotefn(f))
455
del inv[fid]
456
635
636
self.lock_write()
637
638
try:
639
tree = self.working_tree()
640
inv = tree.inventory
641
642
# do this before any modifications
643
for f in files:
644
fid = inv.path2id(f)
645
if not fid:
646
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
647
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
648
if verbose:
649
# having remove it, it must be either ignored or unknown
650
if tree.is_ignored(f):
651
new_status = 'I'
652
else:
653
new_status = '?'
654
show_status(new_status, inv[fid].kind, quotefn(f))
655
del inv[fid]
656
657
self._write_inventory(inv)
658
finally:
659
self.unlock()
660
661
# FIXME: this doesn't need to be a branch method
662
def set_inventory(self, new_inventory_list):
663
from bzrlib.inventory import Inventory, InventoryEntry
664
inv = Inventory(self.get_root_id())
665
for path, file_id, parent, kind in new_inventory_list:
666
name = os.path.basename(path)
667
if name == "":
668
continue
669
# fixme, there should be a factory function inv,add_??
670
if kind == 'directory':
671
inv.add(inventory.InventoryDirectory(file_id, name, parent))
672
elif kind == 'file':
673
inv.add(inventory.InventoryFile(file_id, name, parent))
674
elif kind == 'symlink':
675
inv.add(inventory.InventoryLink(file_id, name, parent))
676
else:
677
raise BzrError("unknown kind %r" % kind)
457
678
self._write_inventory(inv)
458
679
459
460
680
def unknowns(self):
461
681
"""Return all unknown files.
462
682
476
696
return self.working_tree().unknowns()
477
697
478
698
479
def commit(self, message, timestamp=None, timezone=None,
480
committer=None,
481
verbose=False):
482
"""Commit working copy as a new revision.
483
484
The basic approach is to add all the file texts into the
485
store, then the inventory, then make a new revision pointing
486
to that inventory and store that.
487
488
This is not quite safe if the working copy changes during the
489
commit; for the moment that is simply not allowed. A better
490
approach is to make a temporary copy of the files before
491
computing their hashes, and then add those hashes in turn to
492
the inventory. This should mean at least that there are no
493
broken hash pointers. There is no way we can get a snapshot
494
of the whole directory at an instant. This would also have to
495
be robust against files disappearing, moving, etc. So the
496
whole thing is a bit hard.
497
498
timestamp -- if not None, seconds-since-epoch for a
499
postdated/predated commit.
500
"""
501
self._need_writelock()
502
503
## TODO: Show branch names
504
505
# TODO: Don't commit if there are no changes, unless forced?
506
507
# First walk over the working inventory; and both update that
508
# and also build a new revision inventory. The revision
509
# inventory needs to hold the text-id, sha1 and size of the
510
# actual file versions committed in the revision. (These are
511
# not present in the working inventory.) We also need to
512
# detect missing/deleted files, and remove them from the
513
# working inventory.
514
515
work_inv = self.read_working_inventory()
516
inv = Inventory()
517
basis = self.basis_tree()
518
basis_inv = basis.inventory
519
missing_ids = []
520
for path, entry in work_inv.iter_entries():
521
## TODO: Cope with files that have gone missing.
522
523
## TODO: Check that the file kind has not changed from the previous
524
## revision of this file (if any).
525
526
entry = entry.copy()
527
528
p = self.abspath(path)
529
file_id = entry.file_id
530
mutter('commit prep file %s, id %r ' % (p, file_id))
531
532
if not os.path.exists(p):
533
mutter(" file is missing, removing from inventory")
534
if verbose:
535
show_status('D', entry.kind, quotefn(path))
536
missing_ids.append(file_id)
537
continue
538
539
# TODO: Handle files that have been deleted
540
541
# TODO: Maybe a special case for empty files? Seems a
542
# waste to store them many times.
543
544
inv.add(entry)
545
546
if basis_inv.has_id(file_id):
547
old_kind = basis_inv[file_id].kind
548
if old_kind != entry.kind:
549
bailout("entry %r changed kind from %r to %r"
550
% (file_id, old_kind, entry.kind))
551
552
if entry.kind == 'directory':
553
if not isdir(p):
554
bailout("%s is entered as directory but not a directory" % quotefn(p))
555
elif entry.kind == 'file':
556
if not isfile(p):
557
bailout("%s is entered as file but is not a file" % quotefn(p))
558
559
content = file(p, 'rb').read()
560
561
entry.text_sha1 = sha_string(content)
562
entry.text_size = len(content)
563
564
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
565
if (old_ie
566
and (old_ie.text_size == entry.text_size)
567
and (old_ie.text_sha1 == entry.text_sha1)):
568
## assert content == basis.get_file(file_id).read()
569
entry.text_id = basis_inv[file_id].text_id
570
mutter(' unchanged from previous text_id {%s}' %
571
entry.text_id)
572
573
else:
574
entry.text_id = gen_file_id(entry.name)
575
self.text_store.add(content, entry.text_id)
576
mutter(' stored with text_id {%s}' % entry.text_id)
577
if verbose:
578
if not old_ie:
579
state = 'A'
580
elif (old_ie.name == entry.name
581
and old_ie.parent_id == entry.parent_id):
582
state = 'M'
583
else:
584
state = 'R'
585
586
show_status(state, entry.kind, quotefn(path))
587
588
for file_id in missing_ids:
589
# have to do this later so we don't mess up the iterator.
590
# since parents may be removed before their children we
591
# have to test.
592
593
# FIXME: There's probably a better way to do this; perhaps
594
# the workingtree should know how to filter itself.
595
if work_inv.has_id(file_id):
596
del work_inv[file_id]
597
598
599
inv_id = rev_id = _gen_revision_id(time.time())
600
601
inv_tmp = tempfile.TemporaryFile()
602
inv.write_xml(inv_tmp)
603
inv_tmp.seek(0)
604
self.inventory_store.add(inv_tmp, inv_id)
605
mutter('new inventory_id is {%s}' % inv_id)
606
607
self._write_inventory(work_inv)
608
609
if timestamp == None:
610
timestamp = time.time()
611
612
if committer == None:
613
committer = username()
614
615
if timezone == None:
616
timezone = local_time_offset()
617
618
mutter("building commit log message")
619
rev = Revision(timestamp=timestamp,
620
timezone=timezone,
621
committer=committer,
622
precursor = self.last_patch(),
623
message = message,
624
inventory_id=inv_id,
625
revision_id=rev_id)
626
627
rev_tmp = tempfile.TemporaryFile()
628
rev.write_xml(rev_tmp)
629
rev_tmp.seek(0)
630
self.revision_store.add(rev_tmp, rev_id)
631
mutter("new revision_id is {%s}" % rev_id)
632
633
## XXX: Everything up to here can simply be orphaned if we abort
634
## the commit; it will leave junk files behind but that doesn't
635
## matter.
636
637
## TODO: Read back the just-generated changeset, and make sure it
638
## applies and recreates the right state.
639
640
## TODO: Also calculate and store the inventory SHA1
641
mutter("committing patch r%d" % (self.revno() + 1))
642
643
644
self.append_revision(rev_id)
645
646
if verbose:
647
note("commited r%d" % self.revno())
648
649
650
def append_revision(self, revision_id):
651
mutter("add {%s} to revision-history" % revision_id)
652
rev_history = self.revision_history()
653
654
tmprhname = self.controlfilename('revision-history.tmp')
655
rhname = self.controlfilename('revision-history')
656
657
f = file(tmprhname, 'wt')
658
rev_history.append(revision_id)
659
f.write('\n'.join(rev_history))
660
f.write('\n')
661
f.close()
662
663
if sys.platform == 'win32':
664
os.remove(rhname)
665
os.rename(tmprhname, rhname)
666
699
def append_revision(self, *revision_ids):
700
for revision_id in revision_ids:
701
mutter("add {%s} to revision-history" % revision_id)
702
self.lock_write()
703
try:
704
rev_history = self.revision_history()
705
rev_history.extend(revision_ids)
706
self.put_controlfile('revision-history', '\n'.join(rev_history))
707
finally:
708
self.unlock()
709
710
def has_revision(self, revision_id):
711
"""True if this branch has a copy of the revision.
712
713
This does not necessarily imply the revision is merge
714
or on the mainline."""
715
return (revision_id is None
716
or revision_id in self.revision_store)
717
718
def get_revision_xml_file(self, revision_id):
719
"""Return XML file object for revision object."""
720
if not revision_id or not isinstance(revision_id, basestring):
721
raise InvalidRevisionId(revision_id)
722
723
self.lock_read()
724
try:
725
try:
726
return self.revision_store[revision_id]
727
except (IndexError, KeyError):
728
raise bzrlib.errors.NoSuchRevision(self, revision_id)
729
finally:
730
self.unlock()
731
732
#deprecated
733
get_revision_xml = get_revision_xml_file
734
735
def get_revision_xml(self, revision_id):
736
return self.get_revision_xml_file(revision_id).read()
667
737
668
738
669
739
def get_revision(self, revision_id):
670
740
"""Return the Revision object for a named revision"""
671
self._need_readlock()
672
r = Revision.read_xml(self.revision_store[revision_id])
741
xml_file = self.get_revision_xml_file(revision_id)
742
743
try:
744
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
745
except SyntaxError, e:
746
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
747
[revision_id,
748
str(e)])
749
673
750
assert r.revision_id == revision_id
674
751
return r
675
752
676
677
def get_inventory(self, inventory_id):
678
"""Get Inventory object by hash.
679
680
TODO: Perhaps for this and similar methods, take a revision
681
parameter which can be either an integer revno or a
682
string hash."""
683
self._need_readlock()
684
i = Inventory.read_xml(self.inventory_store[inventory_id])
685
return i
686
753
def get_revision_delta(self, revno):
754
"""Return the delta for one revision.
755
756
The delta is relative to its mainline predecessor, or the
757
empty tree for revision 1.
758
"""
759
assert isinstance(revno, int)
760
rh = self.revision_history()
761
if not (1 <= revno <= len(rh)):
762
raise InvalidRevisionNumber(revno)
763
764
# revno is 1-based; list is 0-based
765
766
new_tree = self.revision_tree(rh[revno-1])
767
if revno == 1:
768
old_tree = EmptyTree()
769
else:
770
old_tree = self.revision_tree(rh[revno-2])
771
772
return compare_trees(old_tree, new_tree)
773
774
def get_revision_sha1(self, revision_id):
775
"""Hash the stored value of a revision, and return it."""
776
# In the future, revision entries will be signed. At that
777
# point, it is probably best *not* to include the signature
778
# in the revision hash. Because that lets you re-sign
779
# the revision, (add signatures/remove signatures) and still
780
# have all hash pointers stay consistent.
781
# But for now, just hash the contents.
782
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
783
784
def get_ancestry(self, revision_id):
785
"""Return a list of revision-ids integrated by a revision.
786
787
This currently returns a list, but the ordering is not guaranteed:
788
treat it as a set.
789
"""
790
if revision_id is None:
791
return [None]
792
w = self.control_weaves.get_weave('inventory')
793
return [None] + map(w.idx_to_name,
794
w.inclusions([w.lookup(revision_id)]))
795
796
def get_inventory_weave(self):
797
return self.control_weaves.get_weave('inventory')
798
799
def get_inventory(self, revision_id):
800
"""Get Inventory object by hash."""
801
xml = self.get_inventory_xml(revision_id)
802
return bzrlib.xml5.serializer_v5.read_inventory_from_string(xml)
803
804
def get_inventory_xml(self, revision_id):
805
"""Get inventory XML as a file object."""
806
try:
807
assert isinstance(revision_id, basestring), type(revision_id)
808
iw = self.get_inventory_weave()
809
return iw.get_text(iw.lookup(revision_id))
810
except IndexError:
811
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
812
813
def get_inventory_sha1(self, revision_id):
814
"""Return the sha1 hash of the inventory entry
815
"""
816
return self.get_revision(revision_id).inventory_sha1
687
817
688
818
def get_revision_inventory(self, revision_id):
689
819
"""Return inventory of a past revision."""
690
self._need_readlock()
820
# TODO: Unify this with get_inventory()
821
# bzr 0.0.6 and later imposes the constraint that the inventory_id
822
# must be the same as its revision, so this is trivial.
691
823
if revision_id == None:
692
return Inventory()
824
return Inventory(self.get_root_id())
693
825
else:
694
return self.get_inventory(self.get_revision(revision_id).inventory_id)
695
826
return self.get_inventory(revision_id)
696
827
697
828
def revision_history(self):
698
"""Return sequence of revision hashes on to this branch.
699
700
>>> ScratchBranch().revision_history()
701
[]
702
"""
703
self._need_readlock()
704
return [l.rstrip('\r\n') for l in self.controlfile('revision-history', 'r').readlines()]
705
706
707
def enum_history(self, direction):
708
"""Return (revno, revision_id) for history of branch.
709
710
direction
711
'forward' is from earliest to latest
712
'reverse' is from latest to earliest
713
"""
714
rh = self.revision_history()
715
if direction == 'forward':
716
i = 1
717
for rid in rh:
718
yield i, rid
719
i += 1
720
elif direction == 'reverse':
721
i = len(rh)
722
while i > 0:
723
yield i, rh[i-1]
724
i -= 1
725
else:
726
raise BzrError('invalid history direction %r' % direction)
829
"""Return sequence of revision hashes on to this branch."""
830
self.lock_read()
831
try:
832
return [l.rstrip('\r\n') for l in
833
self.controlfile('revision-history', 'r').readlines()]
834
finally:
835
self.unlock()
836
837
def common_ancestor(self, other, self_revno=None, other_revno=None):
838
"""
839
>>> from bzrlib.commit import commit
840
>>> sb = ScratchBranch(files=['foo', 'foo~'])
841
>>> sb.common_ancestor(sb) == (None, None)
842
True
843
>>> commit(sb, "Committing first revision", verbose=False)
844
>>> sb.common_ancestor(sb)[0]
845
1
846
>>> clone = sb.clone()
847
>>> commit(sb, "Committing second revision", verbose=False)
848
>>> sb.common_ancestor(sb)[0]
849
2
850
>>> sb.common_ancestor(clone)[0]
851
1
852
>>> commit(clone, "Committing divergent second revision",
853
... verbose=False)
854
>>> sb.common_ancestor(clone)[0]
855
1
856
>>> sb.common_ancestor(clone) == clone.common_ancestor(sb)
857
True
858
>>> sb.common_ancestor(sb) != clone.common_ancestor(clone)
859
True
860
>>> clone2 = sb.clone()
861
>>> sb.common_ancestor(clone2)[0]
862
2
863
>>> sb.common_ancestor(clone2, self_revno=1)[0]
864
1
865
>>> sb.common_ancestor(clone2, other_revno=1)[0]
866
1
867
"""
868
my_history = self.revision_history()
869
other_history = other.revision_history()
870
if self_revno is None:
871
self_revno = len(my_history)
872
if other_revno is None:
873
other_revno = len(other_history)
874
indices = range(min((self_revno, other_revno)))
875
indices.reverse()
876
for r in indices:
877
if my_history[r] == other_history[r]:
878
return r+1, my_history[r]
879
return None, None
727
880
728
881
729
882
def revno(self):
731
884
732
885
That is equivalent to the number of revisions committed to
733
886
this branch.
734
735
>>> b = ScratchBranch()
736
>>> b.revno()
737
0
738
>>> b.commit('no foo')
739
>>> b.revno()
740
1
741
887
"""
742
888
return len(self.revision_history())
743
889
744
890
745
def last_patch(self):
891
def last_revision(self):
746
892
"""Return last patch hash, or None if no history.
747
748
>>> ScratchBranch().last_patch() == None
749
True
750
893
"""
751
894
ph = self.revision_history()
752
895
if ph:
753
896
return ph[-1]
754
897
else:
755
898
return None
899
900
901
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
902
"""Return a list of new revisions that would perfectly fit.
756
903
757
758
def lookup_revision(self, revno):
759
"""Return revision hash for revision number."""
904
If self and other have not diverged, return a list of the revisions
905
present in other, but missing from self.
906
907
>>> from bzrlib.commit import commit
908
>>> bzrlib.trace.silent = True
909
>>> br1 = ScratchBranch()
910
>>> br2 = ScratchBranch()
911
>>> br1.missing_revisions(br2)
912
[]
913
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
914
>>> br1.missing_revisions(br2)
915
[u'REVISION-ID-1']
916
>>> br2.missing_revisions(br1)
917
[]
918
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
919
>>> br1.missing_revisions(br2)
920
[]
921
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
922
>>> br1.missing_revisions(br2)
923
[u'REVISION-ID-2A']
924
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
925
>>> br1.missing_revisions(br2)
926
Traceback (most recent call last):
927
DivergedBranches: These branches have diverged.
928
"""
929
# FIXME: If the branches have diverged, but the latest
930
# revision in this branch is completely merged into the other,
931
# then we should still be able to pull.
932
self_history = self.revision_history()
933
self_len = len(self_history)
934
other_history = other.revision_history()
935
other_len = len(other_history)
936
common_index = min(self_len, other_len) -1
937
if common_index >= 0 and \
938
self_history[common_index] != other_history[common_index]:
939
raise DivergedBranches(self, other)
940
941
if stop_revision is None:
942
stop_revision = other_len
943
else:
944
assert isinstance(stop_revision, int)
945
if stop_revision > other_len:
946
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
947
return other_history[self_len:stop_revision]
948
949
def update_revisions(self, other, stop_revision=None):
950
"""Pull in new perfect-fit revisions."""
951
from bzrlib.fetch import greedy_fetch
952
from bzrlib.revision import get_intervening_revisions
953
if stop_revision is None:
954
stop_revision = other.last_revision()
955
greedy_fetch(to_branch=self, from_branch=other,
956
revision=stop_revision)
957
pullable_revs = self.missing_revisions(
958
other, other.revision_id_to_revno(stop_revision))
959
if pullable_revs:
960
greedy_fetch(to_branch=self,
961
from_branch=other,
962
revision=pullable_revs[-1])
963
self.append_revision(*pullable_revs)
964
965
966
def commit(self, *args, **kw):
967
from bzrlib.commit import Commit
968
Commit().commit(self, *args, **kw)
969
970
def revision_id_to_revno(self, revision_id):
971
"""Given a revision id, return its revno"""
972
if revision_id is None:
973
return 0
974
history = self.revision_history()
975
try:
976
return history.index(revision_id) + 1
977
except ValueError:
978
raise bzrlib.errors.NoSuchRevision(self, revision_id)
979
980
def get_rev_id(self, revno, history=None):
981
"""Find the revision id of the specified revno."""
760
982
if revno == 0:
761
983
return None
762
763
try:
764
# list is 0-based; revisions are 1-based
765
return self.revision_history()[revno-1]
766
except IndexError:
767
raise BzrError("no such revision %s" % revno)
768
984
if history is None:
985
history = self.revision_history()
986
elif revno <= 0 or revno > len(history):
987
raise bzrlib.errors.NoSuchRevision(self, revno)
988
return history[revno - 1]
769
989
770
990
def revision_tree(self, revision_id):
771
991
"""Return Tree for a revision on this branch.
772
992
773
993
`revision_id` may be None for the null revision, in which case
774
994
an `EmptyTree` is returned."""
775
self._need_readlock()
995
# TODO: refactor this to use an existing revision object
996
# so we don't need to read it in twice.
776
997
if revision_id == None:
777
998
return EmptyTree()
778
999
else:
779
1000
inv = self.get_revision_inventory(revision_id)
780
return RevisionTree(self.text_store, inv)
1001
return RevisionTree(self.weave_store, inv, revision_id)
781
1002
782
1003
783
1004
def working_tree(self):
784
1005
"""Return a `Tree` for the working copy."""
785
return WorkingTree(self.base, self.read_working_inventory())
1006
from bzrlib.workingtree import WorkingTree
1007
# TODO: In the future, WorkingTree should utilize Transport
1008
# RobertCollins 20051003 - I don't think it should - working trees are
1009
# much more complex to keep consistent than our careful .bzr subset.
1010
# instead, we should say that working trees are local only, and optimise
1011
# for that.
1012
return WorkingTree(self._transport.base, self.read_working_inventory())
786
1013
787
1014
788
1015
def basis_tree(self):
789
1016
"""Return `Tree` object for last revision.
790
1017
791
1018
If there are no revisions yet, return an `EmptyTree`.
792
793
>>> b = ScratchBranch(files=['foo'])
794
>>> b.basis_tree().has_filename('foo')
795
False
796
>>> b.working_tree().has_filename('foo')
797
True
798
>>> b.add('foo')
799
>>> b.commit('add foo')
800
>>> b.basis_tree().has_filename('foo')
801
True
802
1019
"""
803
r = self.last_patch()
804
if r == None:
805
return EmptyTree()
806
else:
807
return RevisionTree(self.text_store, self.get_revision_inventory(r))
808
1020
return self.revision_tree(self.last_revision())
809
1021
810
1022
811
1023
def rename_one(self, from_rel, to_rel):
813
1025
814
1026
This can change the directory or the filename or both.
815
1027
"""
816
self._need_writelock()
817
tree = self.working_tree()
818
inv = tree.inventory
819
if not tree.has_filename(from_rel):
820
bailout("can't rename: old working file %r does not exist" % from_rel)
821
if tree.has_filename(to_rel):
822
bailout("can't rename: new working file %r already exists" % to_rel)
823
824
file_id = inv.path2id(from_rel)
825
if file_id == None:
826
bailout("can't rename: old name %r is not versioned" % from_rel)
827
828
if inv.path2id(to_rel):
829
bailout("can't rename: new name %r is already versioned" % to_rel)
830
831
to_dir, to_tail = os.path.split(to_rel)
832
to_dir_id = inv.path2id(to_dir)
833
if to_dir_id == None and to_dir != '':
834
bailout("can't determine destination directory id for %r" % to_dir)
835
836
mutter("rename_one:")
837
mutter(" file_id {%s}" % file_id)
838
mutter(" from_rel %r" % from_rel)
839
mutter(" to_rel %r" % to_rel)
840
mutter(" to_dir %r" % to_dir)
841
mutter(" to_dir_id {%s}" % to_dir_id)
842
843
inv.rename(file_id, to_dir_id, to_tail)
844
845
print "%s => %s" % (from_rel, to_rel)
846
847
from_abs = self.abspath(from_rel)
848
to_abs = self.abspath(to_rel)
1028
self.lock_write()
849
1029
try:
850
os.rename(from_abs, to_abs)
851
except OSError, e:
852
bailout("failed to rename %r to %r: %s"
853
% (from_abs, to_abs, e[1]),
854
["rename rolled back"])
855
856
self._write_inventory(inv)
857
1030
tree = self.working_tree()
1031
inv = tree.inventory
1032
if not tree.has_filename(from_rel):
1033
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
1034
if tree.has_filename(to_rel):
1035
raise BzrError("can't rename: new working file %r already exists" % to_rel)
1036
1037
file_id = inv.path2id(from_rel)
1038
if file_id == None:
1039
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
1040
1041
if inv.path2id(to_rel):
1042
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
1043
1044
to_dir, to_tail = os.path.split(to_rel)
1045
to_dir_id = inv.path2id(to_dir)
1046
if to_dir_id == None and to_dir != '':
1047
raise BzrError("can't determine destination directory id for %r" % to_dir)
1048
1049
mutter("rename_one:")
1050
mutter(" file_id {%s}" % file_id)
1051
mutter(" from_rel %r" % from_rel)
1052
mutter(" to_rel %r" % to_rel)
1053
mutter(" to_dir %r" % to_dir)
1054
mutter(" to_dir_id {%s}" % to_dir_id)
1055
1056
inv.rename(file_id, to_dir_id, to_tail)
1057
1058
from_abs = self.abspath(from_rel)
1059
to_abs = self.abspath(to_rel)
1060
try:
1061
rename(from_abs, to_abs)
1062
except OSError, e:
1063
raise BzrError("failed to rename %r to %r: %s"
1064
% (from_abs, to_abs, e[1]),
1065
["rename rolled back"])
1066
1067
self._write_inventory(inv)
1068
finally:
1069
self.unlock()
858
1070
859
1071
860
1072
def move(self, from_paths, to_name):
867
1079
868
1080
Note that to_name is only the last component of the new name;
869
1081
this doesn't change the directory.
870
"""
871
self._need_writelock()
872
## TODO: Option to move IDs only
873
assert not isinstance(from_paths, basestring)
874
tree = self.working_tree()
875
inv = tree.inventory
876
to_abs = self.abspath(to_name)
877
if not isdir(to_abs):
878
bailout("destination %r is not a directory" % to_abs)
879
if not tree.has_filename(to_name):
880
bailout("destination %r not in working directory" % to_abs)
881
to_dir_id = inv.path2id(to_name)
882
if to_dir_id == None and to_name != '':
883
bailout("destination %r is not a versioned directory" % to_name)
884
to_dir_ie = inv[to_dir_id]
885
if to_dir_ie.kind not in ('directory', 'root_directory'):
886
bailout("destination %r is not a directory" % to_abs)
887
888
to_idpath = Set(inv.get_idpath(to_dir_id))
889
890
for f in from_paths:
891
if not tree.has_filename(f):
892
bailout("%r does not exist in working tree" % f)
893
f_id = inv.path2id(f)
894
if f_id == None:
895
bailout("%r is not versioned" % f)
896
name_tail = splitpath(f)[-1]
897
dest_path = appendpath(to_name, name_tail)
898
if tree.has_filename(dest_path):
899
bailout("destination %r already exists" % dest_path)
900
if f_id in to_idpath:
901
bailout("can't move %r to a subdirectory of itself" % f)
902
903
# OK, so there's a race here, it's possible that someone will
904
# create a file in this interval and then the rename might be
905
# left half-done. But we should have caught most problems.
906
907
for f in from_paths:
908
name_tail = splitpath(f)[-1]
909
dest_path = appendpath(to_name, name_tail)
910
print "%s => %s" % (f, dest_path)
911
inv.rename(inv.path2id(f), to_dir_id, name_tail)
912
try:
913
os.rename(self.abspath(f), self.abspath(dest_path))
914
except OSError, e:
915
bailout("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
916
["rename rolled back"])
917
918
self._write_inventory(inv)
919
920
921
922
def show_status(self, show_all=False, file_list=None):
923
"""Display single-line status for non-ignored working files.
924
925
The list is show sorted in order by file name.
926
927
>>> b = ScratchBranch(files=['foo', 'foo~'])
928
>>> b.show_status()
929
? foo
930
>>> b.add('foo')
931
>>> b.show_status()
932
A foo
933
>>> b.commit("add foo")
934
>>> b.show_status()
935
>>> os.unlink(b.abspath('foo'))
936
>>> b.show_status()
937
D foo
938
"""
939
self._need_readlock()
940
941
# We have to build everything into a list first so that it can
942
# sorted by name, incorporating all the different sources.
943
944
# FIXME: Rather than getting things in random order and then sorting,
945
# just step through in order.
946
947
# Interesting case: the old ID for a file has been removed,
948
# but a new file has been created under that name.
949
950
old = self.basis_tree()
951
new = self.working_tree()
952
953
items = diff_trees(old, new)
954
# We want to filter out only if any file was provided in the file_list.
955
if isinstance(file_list, list) and len(file_list):
956
items = [item for item in items if item[3] in file_list]
957
958
for fs, fid, oldname, newname, kind in items:
959
if fs == 'R':
960
show_status(fs, kind,
961
oldname + ' => ' + newname)
962
elif fs == 'A' or fs == 'M':
963
show_status(fs, kind, newname)
964
elif fs == 'D':
965
show_status(fs, kind, oldname)
966
elif fs == '.':
967
if show_all:
968
show_status(fs, kind, newname)
969
elif fs == 'I':
970
if show_all:
971
show_status(fs, kind, newname)
972
elif fs == '?':
973
show_status(fs, kind, newname)
974
else:
975
bailout("weird file state %r" % ((fs, fid),))
976
977
978
979
class ScratchBranch(Branch):
1082
1083
This returns a list of (from_path, to_path) pairs for each
1084
entry that is moved.
1085
"""
1086
result = []
1087
self.lock_write()
1088
try:
1089
## TODO: Option to move IDs only
1090
assert not isinstance(from_paths, basestring)
1091
tree = self.working_tree()
1092
inv = tree.inventory
1093
to_abs = self.abspath(to_name)
1094
if not isdir(to_abs):
1095
raise BzrError("destination %r is not a directory" % to_abs)
1096
if not tree.has_filename(to_name):
1097
raise BzrError("destination %r not in working directory" % to_abs)
1098
to_dir_id = inv.path2id(to_name)
1099
if to_dir_id == None and to_name != '':
1100
raise BzrError("destination %r is not a versioned directory" % to_name)
1101
to_dir_ie = inv[to_dir_id]
1102
if to_dir_ie.kind not in ('directory', 'root_directory'):
1103
raise BzrError("destination %r is not a directory" % to_abs)
1104
1105
to_idpath = inv.get_idpath(to_dir_id)
1106
1107
for f in from_paths:
1108
if not tree.has_filename(f):
1109
raise BzrError("%r does not exist in working tree" % f)
1110
f_id = inv.path2id(f)
1111
if f_id == None:
1112
raise BzrError("%r is not versioned" % f)
1113
name_tail = splitpath(f)[-1]
1114
dest_path = appendpath(to_name, name_tail)
1115
if tree.has_filename(dest_path):
1116
raise BzrError("destination %r already exists" % dest_path)
1117
if f_id in to_idpath:
1118
raise BzrError("can't move %r to a subdirectory of itself" % f)
1119
1120
# OK, so there's a race here, it's possible that someone will
1121
# create a file in this interval and then the rename might be
1122
# left half-done. But we should have caught most problems.
1123
1124
for f in from_paths:
1125
name_tail = splitpath(f)[-1]
1126
dest_path = appendpath(to_name, name_tail)
1127
result.append((f, dest_path))
1128
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1129
try:
1130
rename(self.abspath(f), self.abspath(dest_path))
1131
except OSError, e:
1132
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1133
["rename rolled back"])
1134
1135
self._write_inventory(inv)
1136
finally:
1137
self.unlock()
1138
1139
return result
1140
1141
1142
def revert(self, filenames, old_tree=None, backups=True):
1143
"""Restore selected files to the versions from a previous tree.
1144
1145
backups
1146
If true (default) backups are made of files before
1147
they're renamed.
1148
"""
1149
from bzrlib.errors import NotVersionedError, BzrError
1150
from bzrlib.atomicfile import AtomicFile
1151
from bzrlib.osutils import backup_file
1152
1153
inv = self.read_working_inventory()
1154
if old_tree is None:
1155
old_tree = self.basis_tree()
1156
old_inv = old_tree.inventory
1157
1158
nids = []
1159
for fn in filenames:
1160
file_id = inv.path2id(fn)
1161
if not file_id:
1162
raise NotVersionedError("not a versioned file", fn)
1163
if not old_inv.has_id(file_id):
1164
raise BzrError("file not present in old tree", fn, file_id)
1165
nids.append((fn, file_id))
1166
1167
# TODO: Rename back if it was previously at a different location
1168
1169
# TODO: If given a directory, restore the entire contents from
1170
# the previous version.
1171
1172
# TODO: Make a backup to a temporary file.
1173
1174
# TODO: If the file previously didn't exist, delete it?
1175
for fn, file_id in nids:
1176
backup_file(fn)
1177
1178
f = AtomicFile(fn, 'wb')
1179
try:
1180
f.write(old_tree.get_file(file_id).read())
1181
f.commit()
1182
finally:
1183
f.close()
1184
1185
1186
def pending_merges(self):
1187
"""Return a list of pending merges.
1188
1189
These are revisions that have been merged into the working
1190
directory but not yet committed.
1191
"""
1192
cfn = self._rel_controlfilename('pending-merges')
1193
if not self._transport.has(cfn):
1194
return []
1195
p = []
1196
for l in self.controlfile('pending-merges', 'r').readlines():
1197
p.append(l.rstrip('\n'))
1198
return p
1199
1200
1201
def add_pending_merge(self, *revision_ids):
1202
# TODO: Perhaps should check at this point that the
1203
# history of the revision is actually present?
1204
p = self.pending_merges()
1205
updated = False
1206
for rev_id in revision_ids:
1207
if rev_id in p:
1208
continue
1209
p.append(rev_id)
1210
updated = True
1211
if updated:
1212
self.set_pending_merges(p)
1213
1214
def set_pending_merges(self, rev_list):
1215
self.lock_write()
1216
try:
1217
self.put_controlfile('pending-merges', '\n'.join(rev_list))
1218
finally:
1219
self.unlock()
1220
1221
1222
def get_parent(self):
1223
"""Return the parent location of the branch.
1224
1225
This is the default location for push/pull/missing. The usual
1226
pattern is that the user can override it by specifying a
1227
location.
1228
"""
1229
import errno
1230
_locs = ['parent', 'pull', 'x-pull']
1231
for l in _locs:
1232
try:
1233
return self.controlfile(l, 'r').read().strip('\n')
1234
except IOError, e:
1235
if e.errno != errno.ENOENT:
1236
raise
1237
return None
1238
1239
1240
def set_parent(self, url):
1241
# TODO: Maybe delete old location files?
1242
from bzrlib.atomicfile import AtomicFile
1243
self.lock_write()
1244
try:
1245
f = AtomicFile(self.controlfilename('parent'))
1246
try:
1247
f.write(url + '\n')
1248
f.commit()
1249
finally:
1250
f.close()
1251
finally:
1252
self.unlock()
1253
1254
def check_revno(self, revno):
1255
"""\
1256
Check whether a revno corresponds to any revision.
1257
Zero (the NULL revision) is considered valid.
1258
"""
1259
if revno != 0:
1260
self.check_real_revno(revno)
1261
1262
def check_real_revno(self, revno):
1263
"""\
1264
Check whether a revno corresponds to a real revision.
1265
Zero (the NULL revision) is considered invalid
1266
"""
1267
if revno < 1 or revno > self.revno():
1268
raise InvalidRevisionNumber(revno)
1269
1270
1271
1272
1273
1274
class ScratchBranch(_Branch):
980
1275
"""Special test class: a branch that cleans up after itself.
981
1276
982
1277
>>> b = ScratchBranch()
987
1282
>>> isdir(bd)
988
1283
False
989
1284
"""
990
def __init__(self, files=[], dirs=[]):
1285
def __init__(self, files=[], dirs=[], base=None):
991
1286
"""Make a test branch.
992
1287
993
1288
This creates a temporary directory and runs init-tree in it.
994
1289
995
1290
If any files are listed, they are created in the working copy.
996
1291
"""
997
Branch.__init__(self, tempfile.mkdtemp(), init=True)
1292
from tempfile import mkdtemp
1293
init = False
1294
if base is None:
1295
base = mkdtemp()
1296
init = True
1297
if isinstance(base, basestring):
1298
base = get_transport(base)
1299
_Branch.__init__(self, base, init=init)
998
1300
for d in dirs:
999
os.mkdir(self.abspath(d))
1301
self._transport.mkdir(d)
1000
1302
1001
1303
for f in files:
1002
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
1003
1304
self._transport.put(f, 'content of %s' % f)
1305
1306
1307
def clone(self):
1308
"""
1309
>>> orig = ScratchBranch(files=["file1", "file2"])
1310
>>> clone = orig.clone()
1311
>>> if os.name != 'nt':
1312
... os.path.samefile(orig.base, clone.base)
1313
... else:
1314
... orig.base == clone.base
1315
...
1316
False
1317
>>> os.path.isfile(os.path.join(clone.base, "file1"))
1318
True
1319
"""
1320
from shutil import copytree
1321
from tempfile import mkdtemp
1322
base = mkdtemp()
1323
os.rmdir(base)
1324
copytree(self.base, base, symlinks=True)
1325
return ScratchBranch(base=base)
1004
1326
1005
1327
def __del__(self):
1006
1328
self.destroy()
1007
1329
1008
1330
def destroy(self):
1009
1331
"""Destroy the test branch, removing the scratch directory."""
1332
from shutil import rmtree
1010
1333
try:
1011
mutter("delete ScratchBranch %s" % self.base)
1012
shutil.rmtree(self.base)
1334
if self.base:
1335
mutter("delete ScratchBranch %s" % self.base)
1336
rmtree(self.base)
1013
1337
except OSError, e:
1014
1338
# Work around for shutil.rmtree failing on Windows when
1015
1339
# readonly files are encountered
1017
1341
for root, dirs, files in os.walk(self.base, topdown=False):
1018
1342
for name in files:
1019
1343
os.chmod(os.path.join(root, name), 0700)
1020
shutil.rmtree(self.base)
1021
self.base = None
1344
rmtree(self.base)
1345
self._transport = None
1022
1346
1023
1347
1024
1348
1041
1365
1042
1366
1043
1367
1044
def _gen_revision_id(when):
1045
"""Return new revision-id."""
1046
s = '%s-%s-' % (user_email(), compact_date(when))
1047
s += hexlify(rand_bytes(8))
1048
return s
1049
1050
1051
1368
def gen_file_id(name):
1052
1369
"""Return new file id.
1053
1370
1054
1371
This should probably generate proper UUIDs, but for the moment we
1055
1372
cope with just randomness because running uuidgen every time is
1056
1373
slow."""
1374
import re
1375
from binascii import hexlify
1376
from time import time
1377
1378
# get last component
1057
1379
idx = name.rfind('/')
1058
1380
if idx != -1:
1059
1381
name = name[idx+1 : ]
1061
1383
if idx != -1:
1062
1384
name = name[idx+1 : ]
1063
1385
1386
# make it not a hidden file
1064
1387
name = name.lstrip('.')
1065
1388
1389
# remove any wierd characters; we don't escape them but rather
1390
# just pull them out
1391
name = re.sub(r'[^\w.]', '', name)
1392
1066
1393
s = hexlify(rand_bytes(8))
1067
return '-'.join((name, compact_date(time.time()), s))
1394
return '-'.join((name, compact_date(time()), s))
1395
1396
1397
def gen_root_id():
1398
"""Return a new tree-root file id."""
1399
return gen_file_id('TREE_ROOT')
1400
1401
Loggerhead is a web-based interface for Breezy
Version: 2.0.1