« back to all changes in this revision
Viewing changes to bzrlib/branch.py
- browse files at revision 294
- compare with another revision
- download diff
- download tarball
- view history from revision 294
- Committer: Martin Pool
- Date: 2005-04-28 07:24:55 UTC
- Revision ID: mbp@sourcefrog.net-20050428072453-7b99afa993a1e549
todo
- files added:
- bzrlib/remotebranch.py
- files removed:
- HACKING
- bzrlib/plugins
- bzrlib/selftest
- bzrlib/store
- bzrlib/transport
- bzrlib/util
- bzrlib/util/effbot
- bzrlib/util/effbot/org
- contrib
- contrib/bash
- contrib/emacs
- contrib/zsh
- patches
- tools
- files renamed:
- contrib/newinventory.py => bzrlib/newinventory.py
- bzrlib/store/__init__.py => bzrlib/store.py
- bzrlib/util/elementtree/ => elementtree/
- bzrlib/util/urlgrabber/ => urlgrabber/
- files modified:
- .bzrignore
- bzr *
- setup.py *
added
removed
bzrlib/branch.py
15
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
16
16
17
17
18
import sys
19
import os
20
import errno
21
from warnings import warn
22
from cStringIO import StringIO
18
from sets import Set
23
19
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
24
23
25
24
import bzrlib
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, abspath)
32
import bzrlib.errors as errors
33
from bzrlib.errors import (BzrError, InvalidRevisionNumber, InvalidRevisionId,
34
NoSuchRevision, HistoryMissing, NotBranchError,
35
DivergedBranches, LockError, UnlistableStore,
36
UnlistableBranch, NoSuchFile, NotVersionedError)
37
from bzrlib.textui import show_status
38
from bzrlib.revision import Revision, is_ancestor, get_intervening_revisions
39
40
from bzrlib.delta import compare_trees
41
from bzrlib.tree import EmptyTree, RevisionTree
42
from bzrlib.inventory import Inventory
43
from bzrlib.store import copy_all
44
from bzrlib.store.compressed_text import CompressedTextStore
45
from bzrlib.store.text import TextStore
46
from bzrlib.store.weave import WeaveStore
47
from bzrlib.testament import Testament
48
import bzrlib.transactions as transactions
49
from bzrlib.transport import Transport, get_transport
50
import bzrlib.xml5
51
import bzrlib.ui
52
53
54
BZR_BRANCH_FORMAT_4 = "Bazaar-NG branch, format 0.0.4\n"
55
BZR_BRANCH_FORMAT_5 = "Bazaar-NG branch, format 5\n"
56
BZR_BRANCH_FORMAT_6 = "Bazaar-NG branch, format 6\n"
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, chomp, \
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"
57
39
## TODO: Maybe include checks for common corruption of newlines, etc?
58
40
59
41
60
# TODO: Some operations like log might retrieve the same revisions
61
# repeatedly to calculate deltas. We could perhaps have a weakref
62
# cache in memory to make this faster. In general anything can be
63
# cached in memory between lock and unlock operations.
64
65
def find_branch(*ignored, **ignored_too):
66
# XXX: leave this here for about one release, then remove it
67
raise NotImplementedError('find_branch() is not supported anymore, '
68
'please use one of the new branch constructors')
69
70
71
def needs_read_lock(unbound):
72
"""Decorate unbound to take out and release a read lock."""
73
def decorated(self, *args, **kwargs):
74
self.lock_read()
75
try:
76
return unbound(self, *args, **kwargs)
77
finally:
78
self.unlock()
79
return decorated
80
81
82
def needs_write_lock(unbound):
83
"""Decorate unbound to take out and release a write lock."""
84
def decorated(self, *args, **kwargs):
85
self.lock_write()
86
try:
87
return unbound(self, *args, **kwargs)
88
finally:
89
self.unlock()
90
return decorated
42
43
def find_branch_root(f=None):
44
"""Find the branch root enclosing f, or pwd.
45
46
It is not necessary that f exists.
47
48
Basically we keep looking up until we find the control directory or
49
run into the root."""
50
if f == None:
51
f = os.getcwd()
52
elif hasattr(os.path, 'realpath'):
53
f = os.path.realpath(f)
54
else:
55
f = os.path.abspath(f)
56
57
orig_f = f
58
59
while True:
60
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
61
return f
62
head, tail = os.path.split(f)
63
if head == f:
64
# reached the root, whatever that may be
65
raise BzrError('%r is not in a branch' % orig_f)
66
f = head
67
68
91
69
92
70
######################################################################
93
71
# branch objects
94
72
95
class Branch(object):
73
class Branch:
96
74
"""Branch holding a history of revisions.
97
75
98
base
99
Base directory/url of the branch.
100
"""
101
base = None
102
103
def __init__(self, *ignored, **ignored_too):
104
raise NotImplementedError('The Branch class is abstract')
105
106
@staticmethod
107
def open_downlevel(base):
108
"""Open a branch which may be of an old format.
109
110
Only local branches are supported."""
111
return _Branch(get_transport(base), relax_version_check=True)
112
113
@staticmethod
114
def open(base):
115
"""Open an existing branch, rooted at 'base' (url)"""
116
t = get_transport(base)
117
mutter("trying to open %r with transport %r", base, t)
118
return _Branch(t)
119
120
@staticmethod
121
def open_containing(url):
122
"""Open an existing branch which contains url.
123
124
This probes for a branch at url, and searches upwards from there.
125
126
Basically we keep looking up until we find the control directory or
127
run into the root. If there isn't one, raises NotBranchError.
128
If there is one, it is returned, along with the unused portion of url.
129
"""
130
t = get_transport(url)
131
while True:
132
try:
133
return _Branch(t), t.relpath(url)
134
except NotBranchError:
135
pass
136
new_t = t.clone('..')
137
if new_t.base == t.base:
138
# reached the root, whatever that may be
139
raise NotBranchError(path=url)
140
t = new_t
141
142
@staticmethod
143
def initialize(base):
144
"""Create a new branch, rooted at 'base' (url)"""
145
t = get_transport(base)
146
return _Branch(t, init=True)
147
148
def setup_caching(self, cache_root):
149
"""Subclasses that care about caching should override this, and set
150
up cached stores located under cache_root.
151
"""
152
self.cache_root = cache_root
153
154
155
class _Branch(Branch):
156
"""A branch stored in the actual filesystem.
157
158
Note that it's "local" in the context of the filesystem; it doesn't
159
really matter if it's on an nfs/smb/afs/coda/... share, as long as
160
it's writable, and can be accessed via the normal filesystem API.
161
162
_lock_mode
163
None, or 'r' or 'w'
164
165
_lock_count
166
If _lock_mode is true, a positive count of the number of times the
167
lock has been taken.
168
169
_lock
170
Lock object from bzrlib.lock.
171
"""
172
# We actually expect this class to be somewhat short-lived; part of its
173
# purpose is to try to isolate what bits of the branch logic are tied to
174
# filesystem access, so that in a later step, we can extricate them to
175
# a separarte ("storage") class.
176
_lock_mode = None
177
_lock_count = None
178
_lock = None
179
_inventory_weave = None
180
181
# Map some sort of prefix into a namespace
182
# stuff like "revno:10", "revid:", etc.
183
# This should match a prefix with a function which accepts
184
REVISION_NAMESPACES = {}
185
186
def push_stores(self, branch_to):
187
"""Copy the content of this branches store to branch_to."""
188
if (self._branch_format != branch_to._branch_format
189
or self._branch_format != 4):
190
from bzrlib.fetch import greedy_fetch
191
mutter("falling back to fetch logic to push between %s(%s) and %s(%s)",
192
self, self._branch_format, branch_to, branch_to._branch_format)
193
greedy_fetch(to_branch=branch_to, from_branch=self,
194
revision=self.last_revision())
195
return
196
197
store_pairs = ((self.text_store, branch_to.text_store),
198
(self.inventory_store, branch_to.inventory_store),
199
(self.revision_store, branch_to.revision_store))
200
try:
201
for from_store, to_store in store_pairs:
202
copy_all(from_store, to_store)
203
except UnlistableStore:
204
raise UnlistableBranch(from_store)
205
206
def __init__(self, transport, init=False,
207
relax_version_check=False):
76
TODO: Perhaps use different stores for different classes of object,
77
so that we can keep track of how much space each one uses,
78
or garbage-collect them.
79
80
TODO: Add a RemoteBranch subclass. For the basic case of read-only
81
HTTP access this should be very easy by,
82
just redirecting controlfile access into HTTP requests.
83
We would need a RemoteStore working similarly.
84
85
TODO: Keep the on-disk branch locked while the object exists.
86
87
TODO: mkdir() method.
88
"""
89
def __init__(self, base, init=False, find_root=True):
208
90
"""Create new branch object at a particular location.
209
91
210
transport -- A Transport object, defining how to access files.
92
base -- Base directory for the branch.
211
93
212
94
init -- If True, create new control files in a previously
213
95
unversioned directory. If False, the branch must already
214
96
be versioned.
215
97
216
relax_version_check -- If true, the usual check for the branch
217
version is not applied. This is intended only for
218
upgrade/recovery type use; it's not guaranteed that
219
all operations will work on old format branches.
98
find_root -- If true and init is false, find the root of the
99
existing branch containing base.
220
100
221
101
In the test suite, creation of new trees is tested using the
222
102
`ScratchBranch` class.
223
103
"""
224
assert isinstance(transport, Transport), \
225
"%r is not a Transport" % transport
226
self._transport = transport
227
104
if init:
105
self.base = os.path.realpath(base)
228
106
self._make_control()
229
self._check_format(relax_version_check)
230
231
def get_store(name, compressed=True, prefixed=False):
232
# FIXME: This approach of assuming stores are all entirely compressed
233
# or entirely uncompressed is tidy, but breaks upgrade from
234
# some existing branches where there's a mixture; we probably
235
# still want the option to look for both.
236
relpath = self._rel_controlfilename(name)
237
if compressed:
238
store = CompressedTextStore(self._transport.clone(relpath),
239
prefixed=prefixed)
240
else:
241
store = TextStore(self._transport.clone(relpath),
242
prefixed=prefixed)
243
#if self._transport.should_cache():
244
# cache_path = os.path.join(self.cache_root, name)
245
# os.mkdir(cache_path)
246
# store = bzrlib.store.CachedStore(store, cache_path)
247
return store
248
def get_weave(name, prefixed=False):
249
relpath = self._rel_controlfilename(name)
250
ws = WeaveStore(self._transport.clone(relpath), prefixed=prefixed)
251
if self._transport.should_cache():
252
ws.enable_cache = True
253
return ws
254
255
if self._branch_format == 4:
256
self.inventory_store = get_store('inventory-store')
257
self.text_store = get_store('text-store')
258
self.revision_store = get_store('revision-store')
259
elif self._branch_format == 5:
260
self.control_weaves = get_weave([])
261
self.weave_store = get_weave('weaves')
262
self.revision_store = get_store('revision-store', compressed=False)
263
elif self._branch_format == 6:
264
self.control_weaves = get_weave([])
265
self.weave_store = get_weave('weaves', prefixed=True)
266
self.revision_store = get_store('revision-store', compressed=False,
267
prefixed=True)
268
self.revision_store.register_suffix('sig')
269
self._transaction = None
107
elif find_root:
108
self.base = find_branch_root(base)
109
else:
110
self.base = os.path.realpath(base)
111
if not isdir(self.controlfilename('.')):
112
bailout("not a bzr branch: %s" % quotefn(base),
113
['use "bzr init" to initialize a new working tree',
114
'current bzr can only operate from top-of-tree'])
115
self._check_format()
116
117
self.text_store = ImmutableStore(self.controlfilename('text-store'))
118
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
119
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
120
270
121
271
122
def __str__(self):
272
return '%s(%r)' % (self.__class__.__name__, self._transport.base)
123
return '%s(%r)' % (self.__class__.__name__, self.base)
273
124
274
125
275
126
__repr__ = __str__
276
127
277
128
278
def __del__(self):
279
if self._lock_mode or self._lock:
280
# XXX: This should show something every time, and be suitable for
281
# headless operation and embedding
282
warn("branch %r was not explicitly unlocked" % self)
283
self._lock.unlock()
284
285
# TODO: It might be best to do this somewhere else,
286
# but it is nice for a Branch object to automatically
287
# cache it's information.
288
# Alternatively, we could have the Transport objects cache requests
289
# See the earlier discussion about how major objects (like Branch)
290
# should never expect their __del__ function to run.
291
if hasattr(self, 'cache_root') and self.cache_root is not None:
292
try:
293
import shutil
294
shutil.rmtree(self.cache_root)
295
except:
296
pass
297
self.cache_root = None
298
299
def _get_base(self):
300
if self._transport:
301
return self._transport.base
302
return None
303
304
base = property(_get_base, doc="The URL for the root of this branch.")
305
306
def _finish_transaction(self):
307
"""Exit the current transaction."""
308
if self._transaction is None:
309
raise errors.LockError('Branch %s is not in a transaction' %
310
self)
311
transaction = self._transaction
312
self._transaction = None
313
transaction.finish()
314
315
def get_transaction(self):
316
"""Return the current active transaction.
317
318
If no transaction is active, this returns a passthrough object
319
for which all data is immediately flushed and no caching happens.
320
"""
321
if self._transaction is None:
322
return transactions.PassThroughTransaction()
323
else:
324
return self._transaction
325
326
def _set_transaction(self, new_transaction):
327
"""Set a new active transaction."""
328
if self._transaction is not None:
329
raise errors.LockError('Branch %s is in a transaction already.' %
330
self)
331
self._transaction = new_transaction
332
333
def lock_write(self):
334
mutter("lock write: %s (%s)", self, self._lock_count)
335
# TODO: Upgrade locking to support using a Transport,
336
# and potentially a remote locking protocol
337
if self._lock_mode:
338
if self._lock_mode != 'w':
339
raise LockError("can't upgrade to a write lock from %r" %
340
self._lock_mode)
341
self._lock_count += 1
342
else:
343
self._lock = self._transport.lock_write(
344
self._rel_controlfilename('branch-lock'))
345
self._lock_mode = 'w'
346
self._lock_count = 1
347
self._set_transaction(transactions.PassThroughTransaction())
348
349
def lock_read(self):
350
mutter("lock read: %s (%s)", self, self._lock_count)
351
if self._lock_mode:
352
assert self._lock_mode in ('r', 'w'), \
353
"invalid lock mode %r" % self._lock_mode
354
self._lock_count += 1
355
else:
356
self._lock = self._transport.lock_read(
357
self._rel_controlfilename('branch-lock'))
358
self._lock_mode = 'r'
359
self._lock_count = 1
360
self._set_transaction(transactions.ReadOnlyTransaction())
361
# 5K may be excessive, but hey, its a knob.
362
self.get_transaction().set_cache_size(5000)
363
364
def unlock(self):
365
mutter("unlock: %s (%s)", self, self._lock_count)
366
if not self._lock_mode:
367
raise LockError('branch %r is not locked' % (self))
368
369
if self._lock_count > 1:
370
self._lock_count -= 1
371
else:
372
self._finish_transaction()
373
self._lock.unlock()
374
self._lock = None
375
self._lock_mode = self._lock_count = None
376
377
129
def abspath(self, name):
378
"""Return absolute filename for something in the branch
379
380
XXX: Robert Collins 20051017 what is this used for? why is it a branch
381
method and not a tree method.
382
"""
383
return self._transport.abspath(name)
384
385
def _rel_controlfilename(self, file_or_path):
386
if isinstance(file_or_path, basestring):
387
file_or_path = [file_or_path]
388
return [bzrlib.BZRDIR] + file_or_path
130
"""Return absolute filename for something in the branch"""
131
return os.path.join(self.base, name)
132
133
134
def relpath(self, path):
135
"""Return path relative to this branch of something inside it.
136
137
Raises an error if path is not in this branch."""
138
rp = os.path.realpath(path)
139
# FIXME: windows
140
if not rp.startswith(self.base):
141
bailout("path %r is not within branch %r" % (rp, self.base))
142
rp = rp[len(self.base):]
143
rp = rp.lstrip(os.sep)
144
return rp
145
389
146
390
147
def controlfilename(self, file_or_path):
391
148
"""Return location relative to branch."""
392
return self._transport.abspath(self._rel_controlfilename(file_or_path))
149
if isinstance(file_or_path, types.StringTypes):
150
file_or_path = [file_or_path]
151
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
393
152
394
153
395
154
def controlfile(self, file_or_path, mode='r'):
399
158
and binary. binary files are untranslated byte streams. Text
400
159
control files are stored with Unix newlines and in UTF-8, even
401
160
if the platform or locale defaults are different.
402
403
Controlfiles should almost never be opened in write mode but
404
rather should be atomically copied and replaced using atomicfile.
405
161
"""
406
import codecs
407
408
relpath = self._rel_controlfilename(file_or_path)
409
#TODO: codecs.open() buffers linewise, so it was overloaded with
410
# a much larger buffer, do we need to do the same for getreader/getwriter?
411
if mode == 'rb':
412
return self._transport.get(relpath)
413
elif mode == 'wb':
414
raise BzrError("Branch.controlfile(mode='wb') is not supported, use put_controlfiles")
415
elif mode == 'r':
416
return codecs.getreader('utf-8')(self._transport.get(relpath), errors='replace')
417
elif mode == 'w':
418
raise BzrError("Branch.controlfile(mode='w') is not supported, use put_controlfiles")
162
163
fn = self.controlfilename(file_or_path)
164
165
if mode == 'rb' or mode == 'wb':
166
return file(fn, mode)
167
elif mode == 'r' or mode == 'w':
168
# open in binary mode anyhow so there's no newline translation;
169
# codecs uses line buffering by default; don't want that.
170
import codecs
171
return codecs.open(fn, mode + 'b', 'utf-8',
172
buffering=60000)
419
173
else:
420
174
raise BzrError("invalid controlfile mode %r" % mode)
421
175
422
def put_controlfile(self, path, f, encode=True):
423
"""Write an entry as a controlfile.
424
425
:param path: The path to put the file, relative to the .bzr control
426
directory
427
:param f: A file-like or string object whose contents should be copied.
428
:param encode: If true, encode the contents as utf-8
429
"""
430
self.put_controlfiles([(path, f)], encode=encode)
431
432
def put_controlfiles(self, files, encode=True):
433
"""Write several entries as controlfiles.
434
435
:param files: A list of [(path, file)] pairs, where the path is the directory
436
underneath the bzr control directory
437
:param encode: If true, encode the contents as utf-8
438
"""
439
import codecs
440
ctrl_files = []
441
for path, f in files:
442
if encode:
443
if isinstance(f, basestring):
444
f = f.encode('utf-8', 'replace')
445
else:
446
f = codecs.getwriter('utf-8')(f, errors='replace')
447
path = self._rel_controlfilename(path)
448
ctrl_files.append((path, f))
449
self._transport.put_multi(ctrl_files)
176
450
177
451
178
def _make_control(self):
452
from bzrlib.inventory import Inventory
453
from bzrlib.weavefile import write_weave_v5
454
from bzrlib.weave import Weave
455
456
# Create an empty inventory
457
sio = StringIO()
458
# if we want per-tree root ids then this is the place to set
459
# them; they're not needed for now and so ommitted for
460
# simplicity.
461
bzrlib.xml5.serializer_v5.write_inventory(Inventory(), sio)
462
empty_inv = sio.getvalue()
463
sio = StringIO()
464
bzrlib.weavefile.write_weave_v5(Weave(), sio)
465
empty_weave = sio.getvalue()
466
467
dirs = [[], 'revision-store', 'weaves']
468
files = [('README',
179
os.mkdir(self.controlfilename([]))
180
self.controlfile('README', 'w').write(
469
181
"This is a Bazaar-NG control directory.\n"
470
"Do not change any files in this directory.\n"),
471
('branch-format', BZR_BRANCH_FORMAT_6),
472
('revision-history', ''),
473
('branch-name', ''),
474
('branch-lock', ''),
475
('pending-merges', ''),
476
('inventory', empty_inv),
477
('inventory.weave', empty_weave),
478
('ancestry.weave', empty_weave)
479
]
480
cfn = self._rel_controlfilename
481
self._transport.mkdir_multi([cfn(d) for d in dirs])
482
self.put_controlfiles(files)
483
mutter('created control directory in ' + self._transport.base)
484
485
def _check_format(self, relax_version_check):
182
"Do not change any files in this directory.")
183
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
184
for d in ('text-store', 'inventory-store', 'revision-store'):
185
os.mkdir(self.controlfilename(d))
186
for f in ('revision-history', 'merged-patches',
187
'pending-merged-patches', 'branch-name'):
188
self.controlfile(f, 'w').write('')
189
mutter('created control directory in ' + self.base)
190
Inventory().write_xml(self.controlfile('inventory','w'))
191
192
193
def _check_format(self):
486
194
"""Check this branch format is supported.
487
195
488
The format level is stored, as an integer, in
489
self._branch_format for code that needs to check it later.
196
The current tool only supports the current unstable format.
490
197
491
198
In the future, we might need different in-memory Branch
492
199
classes to support downlevel branches. But not yet.
493
200
"""
494
try:
495
fmt = self.controlfile('branch-format', 'r').read()
496
except NoSuchFile:
497
raise NotBranchError(path=self.base)
498
mutter("got branch format %r", fmt)
499
if fmt == BZR_BRANCH_FORMAT_6:
500
self._branch_format = 6
501
elif fmt == BZR_BRANCH_FORMAT_5:
502
self._branch_format = 5
503
elif fmt == BZR_BRANCH_FORMAT_4:
504
self._branch_format = 4
505
506
if (not relax_version_check
507
and self._branch_format not in (5, 6)):
508
raise errors.UnsupportedFormatError(
509
'sorry, branch format %r not supported' % fmt,
510
['use a different bzr version',
511
'or remove the .bzr directory'
512
' and "bzr init" again'])
513
514
def get_root_id(self):
515
"""Return the id of this branches root"""
516
inv = self.read_working_inventory()
517
return inv.root.file_id
518
519
def set_root_id(self, file_id):
520
inv = self.read_working_inventory()
521
orig_root_id = inv.root.file_id
522
del inv._byid[inv.root.file_id]
523
inv.root.file_id = file_id
524
inv._byid[inv.root.file_id] = inv.root
525
for fid in inv:
526
entry = inv[fid]
527
if entry.parent_id in (None, orig_root_id):
528
entry.parent_id = inv.root.file_id
529
self._write_inventory(inv)
530
531
@needs_read_lock
201
# This ignores newlines so that we can open branches created
202
# on Windows from Linux and so on. I think it might be better
203
# to always make all internal files in unix format.
204
fmt = self.controlfile('branch-format', 'r').read()
205
fmt.replace('\r\n', '')
206
if fmt != BZR_BRANCH_FORMAT:
207
bailout('sorry, branch format %r not supported' % fmt,
208
['use a different bzr version',
209
'or remove the .bzr directory and "bzr init" again'])
210
211
532
212
def read_working_inventory(self):
533
213
"""Read the working inventory."""
214
before = time.time()
534
215
# ElementTree does its own conversion from UTF-8, so open in
535
216
# binary.
536
f = self.controlfile('inventory', 'rb')
537
return bzrlib.xml5.serializer_v5.read_inventory(f)
538
539
@needs_write_lock
217
inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))
218
mutter("loaded inventory of %d items in %f"
219
% (len(inv), time.time() - before))
220
return inv
221
222
540
223
def _write_inventory(self, inv):
541
224
"""Update the working inventory.
542
225
543
226
That is to say, the inventory describing changes underway, that
544
227
will be committed to the next revision.
545
228
"""
546
from cStringIO import StringIO
547
sio = StringIO()
548
bzrlib.xml5.serializer_v5.write_inventory(inv, sio)
549
sio.seek(0)
550
# Transport handles atomicity
551
self.put_controlfile('inventory', sio)
552
229
## TODO: factor out to atomicfile? is rename safe on windows?
230
## TODO: Maybe some kind of clean/dirty marker on inventory?
231
tmpfname = self.controlfilename('inventory.tmp')
232
tmpf = file(tmpfname, 'wb')
233
inv.write_xml(tmpf)
234
tmpf.close()
235
inv_fname = self.controlfilename('inventory')
236
if sys.platform == 'win32':
237
os.remove(inv_fname)
238
os.rename(tmpfname, inv_fname)
553
239
mutter('wrote working inventory')
554
240
241
555
242
inventory = property(read_working_inventory, _write_inventory, None,
556
243
"""Inventory for the working copy.""")
557
244
558
@needs_write_lock
559
def add(self, files, ids=None):
245
246
def add(self, files, verbose=False):
560
247
"""Make files versioned.
561
248
562
Note that the command line normally calls smart_add instead,
563
which can automatically recurse.
249
Note that the command line normally calls smart_add instead.
564
250
565
251
This puts the files in the Added state, so that they will be
566
252
recorded by the next commit.
567
253
568
files
569
List of paths to add, relative to the base of the tree.
570
571
ids
572
If set, use these instead of automatically generated ids.
573
Must be the same length as the list of files, but may
574
contain None for ids that are to be autogenerated.
575
576
254
TODO: Perhaps have an option to add the ids even if the files do
577
not (yet) exist.
578
579
TODO: Perhaps yield the ids and paths as they're added.
255
not (yet) exist.
256
257
TODO: Perhaps return the ids of the files? But then again it
258
is easy to retrieve them if they're needed.
259
260
TODO: Option to specify file id.
261
262
TODO: Adding a directory should optionally recurse down and
263
add all non-ignored children. Perhaps do that in a
264
higher-level method.
265
266
>>> b = ScratchBranch(files=['foo'])
267
>>> 'foo' in b.unknowns()
268
True
269
>>> b.show_status()
270
? foo
271
>>> b.add('foo')
272
>>> 'foo' in b.unknowns()
273
False
274
>>> bool(b.inventory.path2id('foo'))
275
True
276
>>> b.show_status()
277
A foo
278
279
>>> b.add('foo')
280
Traceback (most recent call last):
281
...
282
BzrError: ('foo is already versioned', [])
283
284
>>> b.add(['nothere'])
285
Traceback (most recent call last):
286
BzrError: ('cannot add: not a regular file or directory: nothere', [])
580
287
"""
288
581
289
# TODO: Re-adding a file that is removed in the working copy
582
290
# should probably put it back with the previous ID.
583
if isinstance(files, basestring):
584
assert(ids is None or isinstance(ids, basestring))
291
if isinstance(files, types.StringTypes):
585
292
files = [files]
586
if ids is not None:
587
ids = [ids]
588
589
if ids is None:
590
ids = [None] * len(files)
591
else:
592
assert(len(ids) == len(files))
593
293
594
294
inv = self.read_working_inventory()
595
for f,file_id in zip(files, ids):
295
for f in files:
596
296
if is_control_file(f):
597
raise BzrError("cannot add control file %s" % quotefn(f))
297
bailout("cannot add control file %s" % quotefn(f))
598
298
599
299
fp = splitpath(f)
600
300
601
301
if len(fp) == 0:
602
raise BzrError("cannot add top-level %r" % f)
603
302
bailout("cannot add top-level %r" % f)
303
604
304
fullpath = os.path.normpath(self.abspath(f))
605
305
606
306
try:
607
307
kind = file_kind(fullpath)
608
308
except OSError:
609
309
# maybe something better?
610
raise BzrError('cannot add: not a regular file, symlink or directory: %s' % quotefn(f))
611
612
if not InventoryEntry.versionable_kind(kind):
613
raise BzrError('cannot add: not a versionable file ('
614
'i.e. regular file, symlink or directory): %s' % quotefn(f))
615
616
if file_id is None:
617
file_id = gen_file_id(f)
310
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
311
312
if kind != 'file' and kind != 'directory':
313
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
314
315
file_id = gen_file_id(f)
618
316
inv.add_path(f, kind=kind, file_id=file_id)
619
317
318
if verbose:
319
show_status('A', kind, quotefn(f))
320
620
321
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
621
322
622
323
self._write_inventory(inv)
623
324
624
@needs_read_lock
325
625
326
def print_file(self, file, revno):
626
327
"""Print `file` to stdout."""
627
tree = self.revision_tree(self.get_rev_id(revno))
328
tree = self.revision_tree(self.lookup_revision(revno))
628
329
# use inventory as it was in that revision
629
330
file_id = tree.inventory.path2id(file)
630
331
if not file_id:
631
raise BzrError("%r is not present in revision %s" % (file, revno))
332
bailout("%r is not present in revision %d" % (file, revno))
632
333
tree.print_file(file_id)
633
634
# FIXME: this doesn't need to be a branch method
635
def set_inventory(self, new_inventory_list):
636
from bzrlib.inventory import Inventory, InventoryEntry
637
inv = Inventory(self.get_root_id())
638
for path, file_id, parent, kind in new_inventory_list:
639
name = os.path.basename(path)
640
if name == "":
641
continue
642
# fixme, there should be a factory function inv,add_??
643
if kind == 'directory':
644
inv.add(inventory.InventoryDirectory(file_id, name, parent))
645
elif kind == 'file':
646
inv.add(inventory.InventoryFile(file_id, name, parent))
647
elif kind == 'symlink':
648
inv.add(inventory.InventoryLink(file_id, name, parent))
649
else:
650
raise BzrError("unknown kind %r" % kind)
334
335
336
def remove(self, files, verbose=False):
337
"""Mark nominated files for removal from the inventory.
338
339
This does not remove their text. This does not run on
340
341
TODO: Refuse to remove modified files unless --force is given?
342
343
>>> b = ScratchBranch(files=['foo'])
344
>>> b.add('foo')
345
>>> b.inventory.has_filename('foo')
346
True
347
>>> b.remove('foo')
348
>>> b.working_tree().has_filename('foo')
349
True
350
>>> b.inventory.has_filename('foo')
351
False
352
353
>>> b = ScratchBranch(files=['foo'])
354
>>> b.add('foo')
355
>>> b.commit('one')
356
>>> b.remove('foo')
357
>>> b.commit('two')
358
>>> b.inventory.has_filename('foo')
359
False
360
>>> b.basis_tree().has_filename('foo')
361
False
362
>>> b.working_tree().has_filename('foo')
363
True
364
365
TODO: Do something useful with directories.
366
367
TODO: Should this remove the text or not? Tough call; not
368
removing may be useful and the user can just use use rm, and
369
is the opposite of add. Removing it is consistent with most
370
other tools. Maybe an option.
371
"""
372
## TODO: Normalize names
373
## TODO: Remove nested loops; better scalability
374
375
if isinstance(files, types.StringTypes):
376
files = [files]
377
378
tree = self.working_tree()
379
inv = tree.inventory
380
381
# do this before any modifications
382
for f in files:
383
fid = inv.path2id(f)
384
if not fid:
385
bailout("cannot remove unversioned file %s" % quotefn(f))
386
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
387
if verbose:
388
# having remove it, it must be either ignored or unknown
389
if tree.is_ignored(f):
390
new_status = 'I'
391
else:
392
new_status = '?'
393
show_status(new_status, inv[fid].kind, quotefn(f))
394
del inv[fid]
395
651
396
self._write_inventory(inv)
652
397
398
653
399
def unknowns(self):
654
400
"""Return all unknown files.
655
401
656
402
These are files in the working directory that are not versioned or
657
403
control files or ignored.
658
404
659
>>> from bzrlib.workingtree import WorkingTree
660
405
>>> b = ScratchBranch(files=['foo', 'foo~'])
661
>>> map(str, b.unknowns())
406
>>> list(b.unknowns())
662
407
['foo']
663
408
>>> b.add('foo')
664
409
>>> list(b.unknowns())
665
410
[]
666
>>> WorkingTree(b.base, b).remove('foo')
411
>>> b.remove('foo')
667
412
>>> list(b.unknowns())
668
413
['foo']
669
414
"""
670
415
return self.working_tree().unknowns()
671
416
672
@needs_write_lock
673
def append_revision(self, *revision_ids):
674
for revision_id in revision_ids:
675
mutter("add {%s} to revision-history" % revision_id)
417
418
def commit(self, message, timestamp=None, timezone=None,
419
committer=None,
420
verbose=False):
421
"""Commit working copy as a new revision.
422
423
The basic approach is to add all the file texts into the
424
store, then the inventory, then make a new revision pointing
425
to that inventory and store that.
426
427
This is not quite safe if the working copy changes during the
428
commit; for the moment that is simply not allowed. A better
429
approach is to make a temporary copy of the files before
430
computing their hashes, and then add those hashes in turn to
431
the inventory. This should mean at least that there are no
432
broken hash pointers. There is no way we can get a snapshot
433
of the whole directory at an instant. This would also have to
434
be robust against files disappearing, moving, etc. So the
435
whole thing is a bit hard.
436
437
timestamp -- if not None, seconds-since-epoch for a
438
postdated/predated commit.
439
"""
440
441
## TODO: Show branch names
442
443
# TODO: Don't commit if there are no changes, unless forced?
444
445
# First walk over the working inventory; and both update that
446
# and also build a new revision inventory. The revision
447
# inventory needs to hold the text-id, sha1 and size of the
448
# actual file versions committed in the revision. (These are
449
# not present in the working inventory.) We also need to
450
# detect missing/deleted files, and remove them from the
451
# working inventory.
452
453
work_inv = self.read_working_inventory()
454
inv = Inventory()
455
basis = self.basis_tree()
456
basis_inv = basis.inventory
457
missing_ids = []
458
for path, entry in work_inv.iter_entries():
459
## TODO: Cope with files that have gone missing.
460
461
## TODO: Check that the file kind has not changed from the previous
462
## revision of this file (if any).
463
464
entry = entry.copy()
465
466
p = self.abspath(path)
467
file_id = entry.file_id
468
mutter('commit prep file %s, id %r ' % (p, file_id))
469
470
if not os.path.exists(p):
471
mutter(" file is missing, removing from inventory")
472
if verbose:
473
show_status('D', entry.kind, quotefn(path))
474
missing_ids.append(file_id)
475
continue
476
477
# TODO: Handle files that have been deleted
478
479
# TODO: Maybe a special case for empty files? Seems a
480
# waste to store them many times.
481
482
inv.add(entry)
483
484
if basis_inv.has_id(file_id):
485
old_kind = basis_inv[file_id].kind
486
if old_kind != entry.kind:
487
bailout("entry %r changed kind from %r to %r"
488
% (file_id, old_kind, entry.kind))
489
490
if entry.kind == 'directory':
491
if not isdir(p):
492
bailout("%s is entered as directory but not a directory" % quotefn(p))
493
elif entry.kind == 'file':
494
if not isfile(p):
495
bailout("%s is entered as file but is not a file" % quotefn(p))
496
497
content = file(p, 'rb').read()
498
499
entry.text_sha1 = sha_string(content)
500
entry.text_size = len(content)
501
502
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
503
if (old_ie
504
and (old_ie.text_size == entry.text_size)
505
and (old_ie.text_sha1 == entry.text_sha1)):
506
## assert content == basis.get_file(file_id).read()
507
entry.text_id = basis_inv[file_id].text_id
508
mutter(' unchanged from previous text_id {%s}' %
509
entry.text_id)
510
511
else:
512
entry.text_id = gen_file_id(entry.name)
513
self.text_store.add(content, entry.text_id)
514
mutter(' stored with text_id {%s}' % entry.text_id)
515
if verbose:
516
if not old_ie:
517
state = 'A'
518
elif (old_ie.name == entry.name
519
and old_ie.parent_id == entry.parent_id):
520
state = 'M'
521
else:
522
state = 'R'
523
524
show_status(state, entry.kind, quotefn(path))
525
526
for file_id in missing_ids:
527
# have to do this later so we don't mess up the iterator.
528
# since parents may be removed before their children we
529
# have to test.
530
531
# FIXME: There's probably a better way to do this; perhaps
532
# the workingtree should know how to filter itself.
533
if work_inv.has_id(file_id):
534
del work_inv[file_id]
535
536
537
inv_id = rev_id = _gen_revision_id(time.time())
538
539
inv_tmp = tempfile.TemporaryFile()
540
inv.write_xml(inv_tmp)
541
inv_tmp.seek(0)
542
self.inventory_store.add(inv_tmp, inv_id)
543
mutter('new inventory_id is {%s}' % inv_id)
544
545
self._write_inventory(work_inv)
546
547
if timestamp == None:
548
timestamp = time.time()
549
550
if committer == None:
551
committer = username()
552
553
if timezone == None:
554
timezone = local_time_offset()
555
556
mutter("building commit log message")
557
rev = Revision(timestamp=timestamp,
558
timezone=timezone,
559
committer=committer,
560
precursor = self.last_patch(),
561
message = message,
562
inventory_id=inv_id,
563
revision_id=rev_id)
564
565
rev_tmp = tempfile.TemporaryFile()
566
rev.write_xml(rev_tmp)
567
rev_tmp.seek(0)
568
self.revision_store.add(rev_tmp, rev_id)
569
mutter("new revision_id is {%s}" % rev_id)
570
571
## XXX: Everything up to here can simply be orphaned if we abort
572
## the commit; it will leave junk files behind but that doesn't
573
## matter.
574
575
## TODO: Read back the just-generated changeset, and make sure it
576
## applies and recreates the right state.
577
578
## TODO: Also calculate and store the inventory SHA1
579
mutter("committing patch r%d" % (self.revno() + 1))
580
581
582
self.append_revision(rev_id)
583
584
if verbose:
585
note("commited r%d" % self.revno())
586
587
588
def append_revision(self, revision_id):
589
mutter("add {%s} to revision-history" % revision_id)
676
590
rev_history = self.revision_history()
677
rev_history.extend(revision_ids)
678
self.put_controlfile('revision-history', '\n'.join(rev_history))
679
680
def has_revision(self, revision_id):
681
"""True if this branch has a copy of the revision.
682
683
This does not necessarily imply the revision is merge
684
or on the mainline."""
685
return (revision_id is None
686
or self.revision_store.has_id(revision_id))
687
688
@needs_read_lock
689
def get_revision_xml_file(self, revision_id):
690
"""Return XML file object for revision object."""
691
if not revision_id or not isinstance(revision_id, basestring):
692
raise InvalidRevisionId(revision_id)
693
try:
694
return self.revision_store.get(revision_id)
695
except (IndexError, KeyError):
696
raise bzrlib.errors.NoSuchRevision(self, revision_id)
697
698
#deprecated
699
get_revision_xml = get_revision_xml_file
700
701
def get_revision_xml(self, revision_id):
702
return self.get_revision_xml_file(revision_id).read()
591
592
tmprhname = self.controlfilename('revision-history.tmp')
593
rhname = self.controlfilename('revision-history')
594
595
f = file(tmprhname, 'wt')
596
rev_history.append(revision_id)
597
f.write('\n'.join(rev_history))
598
f.write('\n')
599
f.close()
600
601
if sys.platform == 'win32':
602
os.remove(rhname)
603
os.rename(tmprhname, rhname)
604
703
605
704
606
705
607
def get_revision(self, revision_id):
706
608
"""Return the Revision object for a named revision"""
707
xml_file = self.get_revision_xml_file(revision_id)
708
709
try:
710
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
711
except SyntaxError, e:
712
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
713
[revision_id,
714
str(e)])
715
609
r = Revision.read_xml(self.revision_store[revision_id])
716
610
assert r.revision_id == revision_id
717
611
return r
718
612
719
def get_revision_delta(self, revno):
720
"""Return the delta for one revision.
721
722
The delta is relative to its mainline predecessor, or the
723
empty tree for revision 1.
724
"""
725
assert isinstance(revno, int)
726
rh = self.revision_history()
727
if not (1 <= revno <= len(rh)):
728
raise InvalidRevisionNumber(revno)
729
730
# revno is 1-based; list is 0-based
731
732
new_tree = self.revision_tree(rh[revno-1])
733
if revno == 1:
734
old_tree = EmptyTree()
735
else:
736
old_tree = self.revision_tree(rh[revno-2])
737
738
return compare_trees(old_tree, new_tree)
739
740
def get_revision_sha1(self, revision_id):
741
"""Hash the stored value of a revision, and return it."""
742
# In the future, revision entries will be signed. At that
743
# point, it is probably best *not* to include the signature
744
# in the revision hash. Because that lets you re-sign
745
# the revision, (add signatures/remove signatures) and still
746
# have all hash pointers stay consistent.
747
# But for now, just hash the contents.
748
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
749
750
def get_ancestry(self, revision_id):
751
"""Return a list of revision-ids integrated by a revision.
752
753
This currently returns a list, but the ordering is not guaranteed:
754
treat it as a set.
755
"""
756
if revision_id is None:
757
return [None]
758
w = self.get_inventory_weave()
759
return [None] + map(w.idx_to_name,
760
w.inclusions([w.lookup(revision_id)]))
761
762
def get_inventory_weave(self):
763
return self.control_weaves.get_weave('inventory',
764
self.get_transaction())
765
766
def get_inventory(self, revision_id):
767
"""Get Inventory object by hash."""
768
xml = self.get_inventory_xml(revision_id)
769
return bzrlib.xml5.serializer_v5.read_inventory_from_string(xml)
770
771
def get_inventory_xml(self, revision_id):
772
"""Get inventory XML as a file object."""
773
try:
774
assert isinstance(revision_id, basestring), type(revision_id)
775
iw = self.get_inventory_weave()
776
return iw.get_text(iw.lookup(revision_id))
777
except IndexError:
778
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
779
780
def get_inventory_sha1(self, revision_id):
781
"""Return the sha1 hash of the inventory entry
782
"""
783
return self.get_revision(revision_id).inventory_sha1
613
614
def get_inventory(self, inventory_id):
615
"""Get Inventory object by hash.
616
617
TODO: Perhaps for this and similar methods, take a revision
618
parameter which can be either an integer revno or a
619
string hash."""
620
i = Inventory.read_xml(self.inventory_store[inventory_id])
621
return i
622
784
623
785
624
def get_revision_inventory(self, revision_id):
786
625
"""Return inventory of a past revision."""
787
# TODO: Unify this with get_inventory()
788
# bzr 0.0.6 and later imposes the constraint that the inventory_id
789
# must be the same as its revision, so this is trivial.
790
626
if revision_id == None:
791
return Inventory(self.get_root_id())
627
return Inventory()
792
628
else:
793
return self.get_inventory(revision_id)
794
795
@needs_read_lock
629
return self.get_inventory(self.get_revision(revision_id).inventory_id)
630
631
796
632
def revision_history(self):
797
"""Return sequence of revision hashes on to this branch."""
798
transaction = self.get_transaction()
799
history = transaction.map.find_revision_history()
800
if history is not None:
801
mutter("cache hit for revision-history in %s", self)
802
return list(history)
803
history = [l.rstrip('\r\n') for l in
804
self.controlfile('revision-history', 'r').readlines()]
805
transaction.map.add_revision_history(history)
806
# this call is disabled because revision_history is
807
# not really an object yet, and the transaction is for objects.
808
# transaction.register_clean(history, precious=True)
809
return list(history)
633
"""Return sequence of revision hashes on to this branch.
634
635
>>> ScratchBranch().revision_history()
636
[]
637
"""
638
return [chomp(l) for l in self.controlfile('revision-history', 'r').readlines()]
639
810
640
811
641
def revno(self):
812
642
"""Return current revision number for this branch.
813
643
814
644
That is equivalent to the number of revisions committed to
815
645
this branch.
646
647
>>> b = ScratchBranch()
648
>>> b.revno()
649
0
650
>>> b.commit('no foo')
651
>>> b.revno()
652
1
816
653
"""
817
654
return len(self.revision_history())
818
655
819
def last_revision(self):
656
657
def last_patch(self):
820
658
"""Return last patch hash, or None if no history.
659
660
>>> ScratchBranch().last_patch() == None
661
True
821
662
"""
822
663
ph = self.revision_history()
823
664
if ph:
824
665
return ph[-1]
825
666
else:
826
667
return None
827
828
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
829
"""Return a list of new revisions that would perfectly fit.
830
831
If self and other have not diverged, return a list of the revisions
832
present in other, but missing from self.
833
834
>>> from bzrlib.commit import commit
835
>>> bzrlib.trace.silent = True
836
>>> br1 = ScratchBranch()
837
>>> br2 = ScratchBranch()
838
>>> br1.missing_revisions(br2)
839
[]
840
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
841
>>> br1.missing_revisions(br2)
842
[u'REVISION-ID-1']
843
>>> br2.missing_revisions(br1)
844
[]
845
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
846
>>> br1.missing_revisions(br2)
847
[]
848
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
849
>>> br1.missing_revisions(br2)
850
[u'REVISION-ID-2A']
851
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
852
>>> br1.missing_revisions(br2)
853
Traceback (most recent call last):
854
DivergedBranches: These branches have diverged.
855
"""
856
self_history = self.revision_history()
857
self_len = len(self_history)
858
other_history = other.revision_history()
859
other_len = len(other_history)
860
common_index = min(self_len, other_len) -1
861
if common_index >= 0 and \
862
self_history[common_index] != other_history[common_index]:
863
raise DivergedBranches(self, other)
864
865
if stop_revision is None:
866
stop_revision = other_len
867
else:
868
assert isinstance(stop_revision, int)
869
if stop_revision > other_len:
870
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
871
return other_history[self_len:stop_revision]
872
873
def update_revisions(self, other, stop_revision=None):
874
"""Pull in new perfect-fit revisions."""
875
# FIXME: If the branches have diverged, but the latest
876
# revision in this branch is completely merged into the other,
877
# then we should still be able to pull.
878
from bzrlib.fetch import greedy_fetch
879
if stop_revision is None:
880
stop_revision = other.last_revision()
881
### Should this be checking is_ancestor instead of revision_history?
882
if (stop_revision is not None and
883
stop_revision in self.revision_history()):
884
return
885
greedy_fetch(to_branch=self, from_branch=other,
886
revision=stop_revision)
887
pullable_revs = self.pullable_revisions(other, stop_revision)
888
if len(pullable_revs) > 0:
889
self.append_revision(*pullable_revs)
890
891
def pullable_revisions(self, other, stop_revision):
892
other_revno = other.revision_id_to_revno(stop_revision)
893
try:
894
return self.missing_revisions(other, other_revno)
895
except DivergedBranches, e:
896
try:
897
pullable_revs = get_intervening_revisions(self.last_revision(),
898
stop_revision, self)
899
assert self.last_revision() not in pullable_revs
900
return pullable_revs
901
except bzrlib.errors.NotAncestor:
902
if is_ancestor(self.last_revision(), stop_revision, self):
903
return []
904
else:
905
raise e
906
907
def commit(self, *args, **kw):
908
from bzrlib.commit import Commit
909
Commit().commit(self, *args, **kw)
910
911
def revision_id_to_revno(self, revision_id):
912
"""Given a revision id, return its revno"""
913
if revision_id is None:
914
return 0
915
history = self.revision_history()
916
try:
917
return history.index(revision_id) + 1
918
except ValueError:
919
raise bzrlib.errors.NoSuchRevision(self, revision_id)
920
921
def get_rev_id(self, revno, history=None):
922
"""Find the revision id of the specified revno."""
668
669
670
def lookup_revision(self, revno):
671
"""Return revision hash for revision number."""
923
672
if revno == 0:
924
673
return None
925
if history is None:
926
history = self.revision_history()
927
elif revno <= 0 or revno > len(history):
928
raise bzrlib.errors.NoSuchRevision(self, revno)
929
return history[revno - 1]
674
675
try:
676
# list is 0-based; revisions are 1-based
677
return self.revision_history()[revno-1]
678
except IndexError:
679
raise BzrError("no such revision %s" % revno)
680
930
681
931
682
def revision_tree(self, revision_id):
932
683
"""Return Tree for a revision on this branch.
933
684
934
685
`revision_id` may be None for the null revision, in which case
935
686
an `EmptyTree` is returned."""
936
# TODO: refactor this to use an existing revision object
937
# so we don't need to read it in twice.
687
938
688
if revision_id == None:
939
689
return EmptyTree()
940
690
else:
941
691
inv = self.get_revision_inventory(revision_id)
942
return RevisionTree(self.weave_store, inv, revision_id)
692
return RevisionTree(self.text_store, inv)
693
943
694
944
695
def working_tree(self):
945
696
"""Return a `Tree` for the working copy."""
946
from bzrlib.workingtree import WorkingTree
947
# TODO: In the future, perhaps WorkingTree should utilize Transport
948
# RobertCollins 20051003 - I don't think it should - working trees are
949
# much more complex to keep consistent than our careful .bzr subset.
950
# instead, we should say that working trees are local only, and optimise
951
# for that.
952
return WorkingTree(self.base, branch=self)
697
return WorkingTree(self.base, self.read_working_inventory())
953
698
954
699
955
700
def basis_tree(self):
956
701
"""Return `Tree` object for last revision.
957
702
958
703
If there are no revisions yet, return an `EmptyTree`.
704
705
>>> b = ScratchBranch(files=['foo'])
706
>>> b.basis_tree().has_filename('foo')
707
False
708
>>> b.working_tree().has_filename('foo')
709
True
710
>>> b.add('foo')
711
>>> b.commit('add foo')
712
>>> b.basis_tree().has_filename('foo')
713
True
959
714
"""
960
return self.revision_tree(self.last_revision())
961
962
@needs_write_lock
715
r = self.last_patch()
716
if r == None:
717
return EmptyTree()
718
else:
719
return RevisionTree(self.text_store, self.get_revision_inventory(r))
720
721
722
723
def write_log(self, show_timezone='original', verbose=False):
724
"""Write out human-readable log of commits to this branch
725
726
utc -- If true, show dates in universal time, not local time."""
727
## TODO: Option to choose either original, utc or local timezone
728
revno = 1
729
precursor = None
730
for p in self.revision_history():
731
print '-' * 40
732
print 'revno:', revno
733
## TODO: Show hash if --id is given.
734
##print 'revision-hash:', p
735
rev = self.get_revision(p)
736
print 'committer:', rev.committer
737
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0,
738
show_timezone))
739
740
## opportunistic consistency check, same as check_patch_chaining
741
if rev.precursor != precursor:
742
bailout("mismatched precursor!")
743
744
print 'message:'
745
if not rev.message:
746
print ' (no message)'
747
else:
748
for l in rev.message.split('\n'):
749
print ' ' + l
750
751
if verbose == True and precursor != None:
752
print 'changed files:'
753
tree = self.revision_tree(p)
754
prevtree = self.revision_tree(precursor)
755
756
for file_state, fid, old_name, new_name, kind in \
757
diff_trees(prevtree, tree, ):
758
if file_state == 'A' or file_state == 'M':
759
show_status(file_state, kind, new_name)
760
elif file_state == 'D':
761
show_status(file_state, kind, old_name)
762
elif file_state == 'R':
763
show_status(file_state, kind,
764
old_name + ' => ' + new_name)
765
766
revno += 1
767
precursor = p
768
769
963
770
def rename_one(self, from_rel, to_rel):
964
"""Rename one file.
965
966
This can change the directory or the filename or both.
967
"""
968
771
tree = self.working_tree()
969
772
inv = tree.inventory
970
773
if not tree.has_filename(from_rel):
971
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
774
bailout("can't rename: old working file %r does not exist" % from_rel)
972
775
if tree.has_filename(to_rel):
973
raise BzrError("can't rename: new working file %r already exists" % to_rel)
974
776
bailout("can't rename: new working file %r already exists" % to_rel)
777
975
778
file_id = inv.path2id(from_rel)
976
779
if file_id == None:
977
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
780
bailout("can't rename: old name %r is not versioned" % from_rel)
978
781
979
782
if inv.path2id(to_rel):
980
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
783
bailout("can't rename: new name %r is already versioned" % to_rel)
981
784
982
785
to_dir, to_tail = os.path.split(to_rel)
983
786
to_dir_id = inv.path2id(to_dir)
984
787
if to_dir_id == None and to_dir != '':
985
raise BzrError("can't determine destination directory id for %r" % to_dir)
788
bailout("can't determine destination directory id for %r" % to_dir)
986
789
987
790
mutter("rename_one:")
988
791
mutter(" file_id {%s}" % file_id)
990
793
mutter(" to_rel %r" % to_rel)
991
794
mutter(" to_dir %r" % to_dir)
992
795
mutter(" to_dir_id {%s}" % to_dir_id)
993
796
994
797
inv.rename(file_id, to_dir_id, to_tail)
995
798
799
print "%s => %s" % (from_rel, to_rel)
800
996
801
from_abs = self.abspath(from_rel)
997
802
to_abs = self.abspath(to_rel)
998
803
try:
999
rename(from_abs, to_abs)
804
os.rename(from_abs, to_abs)
1000
805
except OSError, e:
1001
raise BzrError("failed to rename %r to %r: %s"
806
bailout("failed to rename %r to %r: %s"
1002
807
% (from_abs, to_abs, e[1]),
1003
808
["rename rolled back"])
1004
809
1005
810
self._write_inventory(inv)
1006
1007
@needs_write_lock
811
812
813
1008
814
def move(self, from_paths, to_name):
1009
815
"""Rename files.
1010
816
1015
821
1016
822
Note that to_name is only the last component of the new name;
1017
823
this doesn't change the directory.
1018
1019
This returns a list of (from_path, to_path) pairs for each
1020
entry that is moved.
1021
824
"""
1022
result = []
1023
825
## TODO: Option to move IDs only
1024
826
assert not isinstance(from_paths, basestring)
1025
827
tree = self.working_tree()
1026
828
inv = tree.inventory
1027
829
to_abs = self.abspath(to_name)
1028
830
if not isdir(to_abs):
1029
raise BzrError("destination %r is not a directory" % to_abs)
831
bailout("destination %r is not a directory" % to_abs)
1030
832
if not tree.has_filename(to_name):
1031
raise BzrError("destination %r not in working directory" % to_abs)
833
bailout("destination %r not in working directory" % to_abs)
1032
834
to_dir_id = inv.path2id(to_name)
1033
835
if to_dir_id == None and to_name != '':
1034
raise BzrError("destination %r is not a versioned directory" % to_name)
836
bailout("destination %r is not a versioned directory" % to_name)
1035
837
to_dir_ie = inv[to_dir_id]
1036
838
if to_dir_ie.kind not in ('directory', 'root_directory'):
1037
raise BzrError("destination %r is not a directory" % to_abs)
839
bailout("destination %r is not a directory" % to_abs)
1038
840
1039
to_idpath = inv.get_idpath(to_dir_id)
841
to_idpath = Set(inv.get_idpath(to_dir_id))
1040
842
1041
843
for f in from_paths:
1042
844
if not tree.has_filename(f):
1043
raise BzrError("%r does not exist in working tree" % f)
845
bailout("%r does not exist in working tree" % f)
1044
846
f_id = inv.path2id(f)
1045
847
if f_id == None:
1046
raise BzrError("%r is not versioned" % f)
848
bailout("%r is not versioned" % f)
1047
849
name_tail = splitpath(f)[-1]
1048
850
dest_path = appendpath(to_name, name_tail)
1049
851
if tree.has_filename(dest_path):
1050
raise BzrError("destination %r already exists" % dest_path)
852
bailout("destination %r already exists" % dest_path)
1051
853
if f_id in to_idpath:
1052
raise BzrError("can't move %r to a subdirectory of itself" % f)
854
bailout("can't move %r to a subdirectory of itself" % f)
1053
855
1054
856
# OK, so there's a race here, it's possible that someone will
1055
857
# create a file in this interval and then the rename might be
1058
860
for f in from_paths:
1059
861
name_tail = splitpath(f)[-1]
1060
862
dest_path = appendpath(to_name, name_tail)
1061
result.append((f, dest_path))
863
print "%s => %s" % (f, dest_path)
1062
864
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1063
865
try:
1064
rename(self.abspath(f), self.abspath(dest_path))
866
os.rename(self.abspath(f), self.abspath(dest_path))
1065
867
except OSError, e:
1066
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
868
bailout("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1067
869
["rename rolled back"])
1068
870
1069
871
self._write_inventory(inv)
1070
return result
1071
1072
1073
def revert(self, filenames, old_tree=None, backups=True):
1074
"""Restore selected files to the versions from a previous tree.
1075
1076
backups
1077
If true (default) backups are made of files before
1078
they're renamed.
1079
"""
1080
from bzrlib.atomicfile import AtomicFile
1081
from bzrlib.osutils import backup_file
1082
1083
inv = self.read_working_inventory()
1084
if old_tree is None:
1085
old_tree = self.basis_tree()
1086
old_inv = old_tree.inventory
1087
1088
nids = []
1089
for fn in filenames:
1090
file_id = inv.path2id(fn)
1091
if not file_id:
1092
raise NotVersionedError(path=fn)
1093
if not old_inv.has_id(file_id):
1094
raise BzrError("file not present in old tree", fn, file_id)
1095
nids.append((fn, file_id))
1096
1097
# TODO: Rename back if it was previously at a different location
1098
1099
# TODO: If given a directory, restore the entire contents from
1100
# the previous version.
1101
1102
# TODO: Make a backup to a temporary file.
1103
1104
# TODO: If the file previously didn't exist, delete it?
1105
for fn, file_id in nids:
1106
backup_file(fn)
1107
1108
f = AtomicFile(fn, 'wb')
1109
try:
1110
f.write(old_tree.get_file(file_id).read())
1111
f.commit()
1112
finally:
1113
f.close()
1114
1115
1116
def pending_merges(self):
1117
"""Return a list of pending merges.
1118
1119
These are revisions that have been merged into the working
1120
directory but not yet committed.
1121
"""
1122
cfn = self._rel_controlfilename('pending-merges')
1123
if not self._transport.has(cfn):
1124
return []
1125
p = []
1126
for l in self.controlfile('pending-merges', 'r').readlines():
1127
p.append(l.rstrip('\n'))
1128
return p
1129
1130
1131
def add_pending_merge(self, *revision_ids):
1132
# TODO: Perhaps should check at this point that the
1133
# history of the revision is actually present?
1134
p = self.pending_merges()
1135
updated = False
1136
for rev_id in revision_ids:
1137
if rev_id in p:
1138
continue
1139
p.append(rev_id)
1140
updated = True
1141
if updated:
1142
self.set_pending_merges(p)
1143
1144
@needs_write_lock
1145
def set_pending_merges(self, rev_list):
1146
self.put_controlfile('pending-merges', '\n'.join(rev_list))
1147
1148
def get_parent(self):
1149
"""Return the parent location of the branch.
1150
1151
This is the default location for push/pull/missing. The usual
1152
pattern is that the user can override it by specifying a
1153
location.
1154
"""
1155
import errno
1156
_locs = ['parent', 'pull', 'x-pull']
1157
for l in _locs:
1158
try:
1159
return self.controlfile(l, 'r').read().strip('\n')
1160
except IOError, e:
1161
if e.errno != errno.ENOENT:
1162
raise
1163
return None
1164
1165
@needs_write_lock
1166
def set_parent(self, url):
1167
# TODO: Maybe delete old location files?
1168
from bzrlib.atomicfile import AtomicFile
1169
f = AtomicFile(self.controlfilename('parent'))
1170
try:
1171
f.write(url + '\n')
1172
f.commit()
1173
finally:
1174
f.close()
1175
1176
def check_revno(self, revno):
1177
"""\
1178
Check whether a revno corresponds to any revision.
1179
Zero (the NULL revision) is considered valid.
1180
"""
1181
if revno != 0:
1182
self.check_real_revno(revno)
1183
1184
def check_real_revno(self, revno):
1185
"""\
1186
Check whether a revno corresponds to a real revision.
1187
Zero (the NULL revision) is considered invalid
1188
"""
1189
if revno < 1 or revno > self.revno():
1190
raise InvalidRevisionNumber(revno)
1191
1192
def sign_revision(self, revision_id, gpg_strategy):
1193
plaintext = Testament.from_revision(self, revision_id).as_short_text()
1194
self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1195
1196
@needs_write_lock
1197
def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1198
self.revision_store.add(StringIO(gpg_strategy.sign(plaintext)),
1199
revision_id, "sig")
1200
1201
1202
class ScratchBranch(_Branch):
872
873
874
875
def show_status(self, show_all=False):
876
"""Display single-line status for non-ignored working files.
877
878
The list is show sorted in order by file name.
879
880
>>> b = ScratchBranch(files=['foo', 'foo~'])
881
>>> b.show_status()
882
? foo
883
>>> b.add('foo')
884
>>> b.show_status()
885
A foo
886
>>> b.commit("add foo")
887
>>> b.show_status()
888
>>> os.unlink(b.abspath('foo'))
889
>>> b.show_status()
890
D foo
891
892
TODO: Get state for single files.
893
"""
894
895
# We have to build everything into a list first so that it can
896
# sorted by name, incorporating all the different sources.
897
898
# FIXME: Rather than getting things in random order and then sorting,
899
# just step through in order.
900
901
# Interesting case: the old ID for a file has been removed,
902
# but a new file has been created under that name.
903
904
old = self.basis_tree()
905
new = self.working_tree()
906
907
for fs, fid, oldname, newname, kind in diff_trees(old, new):
908
if fs == 'R':
909
show_status(fs, kind,
910
oldname + ' => ' + newname)
911
elif fs == 'A' or fs == 'M':
912
show_status(fs, kind, newname)
913
elif fs == 'D':
914
show_status(fs, kind, oldname)
915
elif fs == '.':
916
if show_all:
917
show_status(fs, kind, newname)
918
elif fs == 'I':
919
if show_all:
920
show_status(fs, kind, newname)
921
elif fs == '?':
922
show_status(fs, kind, newname)
923
else:
924
bailout("weird file state %r" % ((fs, fid),))
925
926
927
928
class ScratchBranch(Branch):
1203
929
"""Special test class: a branch that cleans up after itself.
1204
930
1205
931
>>> b = ScratchBranch()
1206
932
>>> isdir(b.base)
1207
933
True
1208
934
>>> bd = b.base
1209
>>> b._transport.__del__()
935
>>> del b
1210
936
>>> isdir(bd)
1211
937
False
1212
938
"""
1213
1214
def __init__(self, files=[], dirs=[], transport=None):
939
def __init__(self, files=[], dirs=[]):
1215
940
"""Make a test branch.
1216
941
1217
942
This creates a temporary directory and runs init-tree in it.
1218
943
1219
944
If any files are listed, they are created in the working copy.
1220
945
"""
1221
if transport is None:
1222
transport = bzrlib.transport.local.ScratchTransport()
1223
super(ScratchBranch, self).__init__(transport, init=True)
1224
else:
1225
super(ScratchBranch, self).__init__(transport)
1226
946
Branch.__init__(self, tempfile.mkdtemp(), init=True)
1227
947
for d in dirs:
1228
self._transport.mkdir(d)
948
os.mkdir(self.abspath(d))
1229
949
1230
950
for f in files:
1231
self._transport.put(f, 'content of %s' % f)
1232
1233
1234
def clone(self):
1235
"""
1236
>>> orig = ScratchBranch(files=["file1", "file2"])
1237
>>> clone = orig.clone()
1238
>>> if os.name != 'nt':
1239
... os.path.samefile(orig.base, clone.base)
1240
... else:
1241
... orig.base == clone.base
1242
...
1243
False
1244
>>> os.path.isfile(os.path.join(clone.base, "file1"))
1245
True
1246
"""
1247
from shutil import copytree
1248
from tempfile import mkdtemp
1249
base = mkdtemp()
1250
os.rmdir(base)
1251
copytree(self.base, base, symlinks=True)
1252
return ScratchBranch(
1253
transport=bzrlib.transport.local.ScratchTransport(base))
951
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
952
953
954
def __del__(self):
955
"""Destroy the test branch, removing the scratch directory."""
956
try:
957
shutil.rmtree(self.base)
958
except OSError:
959
# Work around for shutil.rmtree failing on Windows when
960
# readonly files are encountered
961
for root, dirs, files in os.walk(self.base, topdown=False):
962
for name in files:
963
os.chmod(os.path.join(root, name), 0700)
964
shutil.rmtree(self.base)
965
1254
966
1255
967
1256
968
######################################################################
1272
984
1273
985
1274
986
987
def _gen_revision_id(when):
988
"""Return new revision-id."""
989
s = '%s-%s-' % (user_email(), compact_date(when))
990
s += hexlify(rand_bytes(8))
991
return s
992
993
1275
994
def gen_file_id(name):
1276
995
"""Return new file id.
1277
996
1278
997
This should probably generate proper UUIDs, but for the moment we
1279
998
cope with just randomness because running uuidgen every time is
1280
999
slow."""
1281
import re
1282
from binascii import hexlify
1283
from time import time
1284
1285
# get last component
1286
1000
idx = name.rfind('/')
1287
1001
if idx != -1:
1288
1002
name = name[idx+1 : ]
1290
1004
if idx != -1:
1291
1005
name = name[idx+1 : ]
1292
1006
1293
# make it not a hidden file
1294
1007
name = name.lstrip('.')
1295
1008
1296
# remove any wierd characters; we don't escape them but rather
1297
# just pull them out
1298
name = re.sub(r'[^\w.]', '', name)
1299
1300
1009
s = hexlify(rand_bytes(8))
1301
return '-'.join((name, compact_date(time()), s))
1302
1303
1304
def gen_root_id():
1305
"""Return a new tree-root file id."""
1306
return gen_file_id('TREE_ROOT')
1307
1308
1010
return '-'.join((name, compact_date(time.time()), s))
Loggerhead is a web-based interface for Breezy
Version: 2.0.1