« back to all changes in this revision
Viewing changes to bzrlib/branch.py
- browse files at revision 1417.1.10
- compare with another revision
- download diff
- download tarball
- view history from revision 1417.1.10
- Committer: Robert Collins
- Date: 2005-10-11 01:05:24 UTC
- mto: This revision was merged to the branch mainline in revision 1438.
- Revision ID: robertc@robertcollins.net-20051011010524-e258bc8d051cc9d2
add a cache bound to Transactions, and a precious facility, so that we keep inventory.weave in memory, but can discard weaves for other such files.
- files added:
- .rsyncexclude
- bzrlib/plugins
- bzrlib/selftest
- bzrlib/store
- bzrlib/transport
- bzrlib/util
- bzrlib/util/effbot
- bzrlib/util/effbot/org
- bzrlib/util/elementtree
- bzrlib/util/urlgrabber
- contrib
- contrib/bash
- contrib/emacs
- contrib/zsh
- patches
- tools
- files removed:
- bzrlib/tests.py
- files renamed:
- bzrlib/store.py => bzrlib/store/__init__.py
- files modified:
- .bzrignore
- bzr *
- setup.py *
added
removed
bzrlib/branch.py
1
#! /usr/bin/env python
2
# -*- coding: UTF-8 -*-
1
# Copyright (C) 2005 Canonical Ltd
3
2
4
3
# This program is free software; you can redistribute it and/or modify
5
4
# it under the terms of the GNU General Public License as published by
16
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
16
18
17
19
from sets import Set
18
import sys
19
import os
20
import errno
21
from warnings import warn
22
from cStringIO import StringIO
20
23
21
import sys, os, os.path, random, time, sha, sets, types, re, shutil, tempfile
22
import traceback, socket, fnmatch, difflib, time
23
from binascii import hexlify
24
24
25
25
import bzrlib
26
from inventory import Inventory
27
from trace import mutter, note
28
from tree import Tree, EmptyTree, RevisionTree, WorkingTree
29
from inventory import InventoryEntry, Inventory
30
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, chomp, \
31
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
32
joinpath, sha_string, file_kind, local_time_offset
33
from store import ImmutableStore
34
from revision import Revision
35
from errors import bailout
36
from textui import show_status
37
from diff import diff_trees
38
39
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
import bzrlib.errors as errors
33
from bzrlib.errors import (BzrError, InvalidRevisionNumber, InvalidRevisionId,
34
NoSuchRevision, HistoryMissing, NotBranchError,
35
DivergedBranches, LockError, UnlistableStore,
36
UnlistableBranch, NoSuchFile)
37
from bzrlib.textui import show_status
38
from bzrlib.revision import Revision
39
from bzrlib.delta import compare_trees
40
from bzrlib.tree import EmptyTree, RevisionTree
41
from bzrlib.inventory import Inventory
42
from bzrlib.store import copy_all
43
from bzrlib.store.compressed_text import CompressedTextStore
44
from bzrlib.store.text import TextStore
45
from bzrlib.store.weave import WeaveStore
46
import bzrlib.transactions as transactions
47
from bzrlib.transport import Transport, get_transport
48
import bzrlib.xml5
49
import bzrlib.ui
50
51
52
BZR_BRANCH_FORMAT_4 = "Bazaar-NG branch, format 0.0.4\n"
53
BZR_BRANCH_FORMAT_5 = "Bazaar-NG branch, format 5\n"
54
BZR_BRANCH_FORMAT_6 = "Bazaar-NG branch, format 6\n"
40
55
## TODO: Maybe include checks for common corruption of newlines, etc?
41
56
42
57
43
58
# TODO: Some operations like log might retrieve the same revisions
59
# repeatedly to calculate deltas. We could perhaps have a weakref
60
# cache in memory to make this faster. In general anything can be
61
# cached in memory between lock and unlock operations.
62
63
def find_branch(*ignored, **ignored_too):
64
# XXX: leave this here for about one release, then remove it
65
raise NotImplementedError('find_branch() is not supported anymore, '
66
'please use one of the new branch constructors')
67
def _relpath(base, path):
68
"""Return path relative to base, or raise exception.
69
70
The path may be either an absolute path or a path relative to the
71
current working directory.
72
73
Lifted out of Branch.relpath for ease of testing.
74
75
os.path.commonprefix (python2.4) has a bad bug that it works just
76
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
77
avoids that problem."""
78
rp = os.path.abspath(path)
79
80
s = []
81
head = rp
82
while len(head) >= len(base):
83
if head == base:
84
break
85
head, tail = os.path.split(head)
86
if tail:
87
s.insert(0, tail)
88
else:
89
raise NotBranchError("path %r is not within branch %r" % (rp, base))
90
91
return os.sep.join(s)
92
93
94
def find_branch_root(t):
95
"""Find the branch root enclosing the transport's base.
96
97
t is a Transport object.
98
99
It is not necessary that the base of t exists.
100
101
Basically we keep looking up until we find the control directory or
102
run into the root. If there isn't one, raises NotBranchError.
103
"""
104
orig_base = t.base
105
while True:
106
if t.has(bzrlib.BZRDIR):
107
return t
108
new_t = t.clone('..')
109
if new_t.base == t.base:
110
# reached the root, whatever that may be
111
raise NotBranchError('%s is not in a branch' % orig_base)
112
t = new_t
44
113
45
114
46
115
######################################################################
47
116
# branch objects
48
117
49
class Branch:
118
class Branch(object):
50
119
"""Branch holding a history of revisions.
51
120
52
:todo: Perhaps use different stores for different classes of object,
53
so that we can keep track of how much space each one uses,
54
or garbage-collect them.
55
56
:todo: Add a RemoteBranch subclass. For the basic case of read-only
57
HTTP access this should be very easy by,
58
just redirecting controlfile access into HTTP requests.
59
We would need a RemoteStore working similarly.
60
61
:todo: Keep the on-disk branch locked while the object exists.
62
63
:todo: mkdir() method.
64
"""
65
def __init__(self, base, init=False):
121
base
122
Base directory/url of the branch.
123
"""
124
base = None
125
126
def __init__(self, *ignored, **ignored_too):
127
raise NotImplementedError('The Branch class is abstract')
128
129
@staticmethod
130
def open_downlevel(base):
131
"""Open a branch which may be of an old format.
132
133
Only local branches are supported."""
134
return _Branch(get_transport(base), relax_version_check=True)
135
136
@staticmethod
137
def open(base):
138
"""Open an existing branch, rooted at 'base' (url)"""
139
t = get_transport(base)
140
mutter("trying to open %r with transport %r", base, t)
141
return _Branch(t)
142
143
@staticmethod
144
def open_containing(url):
145
"""Open an existing branch which contains url.
146
147
This probes for a branch at url, and searches upwards from there.
148
"""
149
t = get_transport(url)
150
t = find_branch_root(t)
151
return _Branch(t)
152
153
@staticmethod
154
def initialize(base):
155
"""Create a new branch, rooted at 'base' (url)"""
156
t = get_transport(base)
157
return _Branch(t, init=True)
158
159
def setup_caching(self, cache_root):
160
"""Subclasses that care about caching should override this, and set
161
up cached stores located under cache_root.
162
"""
163
self.cache_root = cache_root
164
165
166
class _Branch(Branch):
167
"""A branch stored in the actual filesystem.
168
169
Note that it's "local" in the context of the filesystem; it doesn't
170
really matter if it's on an nfs/smb/afs/coda/... share, as long as
171
it's writable, and can be accessed via the normal filesystem API.
172
173
_lock_mode
174
None, or 'r' or 'w'
175
176
_lock_count
177
If _lock_mode is true, a positive count of the number of times the
178
lock has been taken.
179
180
_lock
181
Lock object from bzrlib.lock.
182
"""
183
# We actually expect this class to be somewhat short-lived; part of its
184
# purpose is to try to isolate what bits of the branch logic are tied to
185
# filesystem access, so that in a later step, we can extricate them to
186
# a separarte ("storage") class.
187
_lock_mode = None
188
_lock_count = None
189
_lock = None
190
_inventory_weave = None
191
192
# Map some sort of prefix into a namespace
193
# stuff like "revno:10", "revid:", etc.
194
# This should match a prefix with a function which accepts
195
REVISION_NAMESPACES = {}
196
197
def push_stores(self, branch_to):
198
"""Copy the content of this branches store to branch_to."""
199
if (self._branch_format != branch_to._branch_format
200
or self._branch_format != 4):
201
from bzrlib.fetch import greedy_fetch
202
mutter("falling back to fetch logic to push between %s(%s) and %s(%s)",
203
self, self._branch_format, branch_to, branch_to._branch_format)
204
greedy_fetch(to_branch=branch_to, from_branch=self,
205
revision=self.last_revision())
206
return
207
208
store_pairs = ((self.text_store, branch_to.text_store),
209
(self.inventory_store, branch_to.inventory_store),
210
(self.revision_store, branch_to.revision_store))
211
try:
212
for from_store, to_store in store_pairs:
213
copy_all(from_store, to_store)
214
except UnlistableStore:
215
raise UnlistableBranch(from_store)
216
217
def __init__(self, transport, init=False,
218
relax_version_check=False):
66
219
"""Create new branch object at a particular location.
67
220
68
:param base: Base directory for the branch.
69
70
:param init: If True, create new control files in a previously
221
transport -- A Transport object, defining how to access files.
222
(If a string, transport.transport() will be used to
223
create a Transport object)
224
225
init -- If True, create new control files in a previously
71
226
unversioned directory. If False, the branch must already
72
227
be versioned.
73
228
229
relax_version_check -- If true, the usual check for the branch
230
version is not applied. This is intended only for
231
upgrade/recovery type use; it's not guaranteed that
232
all operations will work on old format branches.
233
74
234
In the test suite, creation of new trees is tested using the
75
235
`ScratchBranch` class.
76
236
"""
77
self.base = os.path.realpath(base)
237
assert isinstance(transport, Transport), \
238
"%r is not a Transport" % transport
239
self._transport = transport
78
240
if init:
79
241
self._make_control()
80
else:
81
if not isdir(self.controlfilename('.')):
82
bailout("not a bzr branch: %s" % quotefn(base),
83
['use "bzr init" to initialize a new working tree',
84
'current bzr can only operate from top-of-tree'])
85
self._check_format()
86
87
self.text_store = ImmutableStore(self.controlfilename('text-store'))
88
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
89
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
90
242
self._check_format(relax_version_check)
243
244
def get_store(name, compressed=True, prefixed=False):
245
# FIXME: This approach of assuming stores are all entirely compressed
246
# or entirely uncompressed is tidy, but breaks upgrade from
247
# some existing branches where there's a mixture; we probably
248
# still want the option to look for both.
249
relpath = self._rel_controlfilename(name)
250
if compressed:
251
store = CompressedTextStore(self._transport.clone(relpath),
252
prefixed=prefixed)
253
else:
254
store = TextStore(self._transport.clone(relpath),
255
prefixed=prefixed)
256
#if self._transport.should_cache():
257
# cache_path = os.path.join(self.cache_root, name)
258
# os.mkdir(cache_path)
259
# store = bzrlib.store.CachedStore(store, cache_path)
260
return store
261
def get_weave(name, prefixed=False):
262
relpath = self._rel_controlfilename(name)
263
ws = WeaveStore(self._transport.clone(relpath), prefixed=prefixed)
264
if self._transport.should_cache():
265
ws.enable_cache = True
266
return ws
267
268
if self._branch_format == 4:
269
self.inventory_store = get_store('inventory-store')
270
self.text_store = get_store('text-store')
271
self.revision_store = get_store('revision-store')
272
elif self._branch_format == 5:
273
self.control_weaves = get_weave([])
274
self.weave_store = get_weave('weaves')
275
self.revision_store = get_store('revision-store', compressed=False)
276
elif self._branch_format == 6:
277
self.control_weaves = get_weave([])
278
self.weave_store = get_weave('weaves', prefixed=True)
279
self.revision_store = get_store('revision-store', compressed=False,
280
prefixed=True)
281
self._transaction = None
91
282
92
283
def __str__(self):
93
return '%s(%r)' % (self.__class__.__name__, self.base)
284
return '%s(%r)' % (self.__class__.__name__, self._transport.base)
94
285
95
286
96
287
__repr__ = __str__
97
288
98
289
99
def _rel(self, name):
100
"""Return filename relative to branch top"""
101
return os.path.join(self.base, name)
102
290
def __del__(self):
291
if self._lock_mode or self._lock:
292
# XXX: This should show something every time, and be suitable for
293
# headless operation and embedding
294
warn("branch %r was not explicitly unlocked" % self)
295
self._lock.unlock()
296
297
# TODO: It might be best to do this somewhere else,
298
# but it is nice for a Branch object to automatically
299
# cache it's information.
300
# Alternatively, we could have the Transport objects cache requests
301
# See the earlier discussion about how major objects (like Branch)
302
# should never expect their __del__ function to run.
303
if hasattr(self, 'cache_root') and self.cache_root is not None:
304
try:
305
import shutil
306
shutil.rmtree(self.cache_root)
307
except:
308
pass
309
self.cache_root = None
310
311
def _get_base(self):
312
if self._transport:
313
return self._transport.base
314
return None
315
316
base = property(_get_base)
317
318
def _finish_transaction(self):
319
"""Exit the current transaction."""
320
if self._transaction is None:
321
raise errors.LockError('Branch %s is not in a transaction' %
322
self)
323
transaction = self._transaction
324
self._transaction = None
325
transaction.finish()
326
327
def get_transaction(self):
328
"""Return the current active transaction.
329
330
If no transaction is active, this returns a passthrough object
331
for which all data is immedaitely flushed and no caching happens.
332
"""
333
if self._transaction is None:
334
return transactions.PassThroughTransaction()
335
else:
336
return self._transaction
337
338
def _set_transaction(self, new_transaction):
339
"""Set a new active transaction."""
340
if self._transaction is not None:
341
raise errors.LockError('Branch %s is in a transaction already.' %
342
self)
343
self._transaction = new_transaction
344
345
def lock_write(self):
346
mutter("lock write: %s (%s)", self, self._lock_count)
347
# TODO: Upgrade locking to support using a Transport,
348
# and potentially a remote locking protocol
349
if self._lock_mode:
350
if self._lock_mode != 'w':
351
raise LockError("can't upgrade to a write lock from %r" %
352
self._lock_mode)
353
self._lock_count += 1
354
else:
355
self._lock = self._transport.lock_write(
356
self._rel_controlfilename('branch-lock'))
357
self._lock_mode = 'w'
358
self._lock_count = 1
359
self._set_transaction(transactions.PassThroughTransaction())
360
361
362
def lock_read(self):
363
mutter("lock read: %s (%s)", self, self._lock_count)
364
if self._lock_mode:
365
assert self._lock_mode in ('r', 'w'), \
366
"invalid lock mode %r" % self._lock_mode
367
self._lock_count += 1
368
else:
369
self._lock = self._transport.lock_read(
370
self._rel_controlfilename('branch-lock'))
371
self._lock_mode = 'r'
372
self._lock_count = 1
373
self._set_transaction(transactions.ReadOnlyTransaction())
374
# 5K may be excessive, but hey, its a knob.
375
self.get_transaction().set_cache_size(5000)
376
377
def unlock(self):
378
mutter("unlock: %s (%s)", self, self._lock_count)
379
if not self._lock_mode:
380
raise LockError('branch %r is not locked' % (self))
381
382
if self._lock_count > 1:
383
self._lock_count -= 1
384
else:
385
self._finish_transaction()
386
self._lock.unlock()
387
self._lock = None
388
self._lock_mode = self._lock_count = None
389
390
def abspath(self, name):
391
"""Return absolute filename for something in the branch"""
392
return self._transport.abspath(name)
393
394
def relpath(self, path):
395
"""Return path relative to this branch of something inside it.
396
397
Raises an error if path is not in this branch."""
398
return self._transport.relpath(path)
399
400
401
def _rel_controlfilename(self, file_or_path):
402
if isinstance(file_or_path, basestring):
403
file_or_path = [file_or_path]
404
return [bzrlib.BZRDIR] + file_or_path
103
405
104
406
def controlfilename(self, file_or_path):
105
407
"""Return location relative to branch."""
106
if isinstance(file_or_path, types.StringTypes):
107
file_or_path = [file_or_path]
108
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
408
return self._transport.abspath(self._rel_controlfilename(file_or_path))
109
409
110
410
111
411
def controlfile(self, file_or_path, mode='r'):
112
"""Open a control file for this branch"""
113
return file(self.controlfilename(file_or_path), mode)
114
412
"""Open a control file for this branch.
413
414
There are two classes of file in the control directory: text
415
and binary. binary files are untranslated byte streams. Text
416
control files are stored with Unix newlines and in UTF-8, even
417
if the platform or locale defaults are different.
418
419
Controlfiles should almost never be opened in write mode but
420
rather should be atomically copied and replaced using atomicfile.
421
"""
422
import codecs
423
424
relpath = self._rel_controlfilename(file_or_path)
425
#TODO: codecs.open() buffers linewise, so it was overloaded with
426
# a much larger buffer, do we need to do the same for getreader/getwriter?
427
if mode == 'rb':
428
return self._transport.get(relpath)
429
elif mode == 'wb':
430
raise BzrError("Branch.controlfile(mode='wb') is not supported, use put_controlfiles")
431
elif mode == 'r':
432
return codecs.getreader('utf-8')(self._transport.get(relpath), errors='replace')
433
elif mode == 'w':
434
raise BzrError("Branch.controlfile(mode='w') is not supported, use put_controlfiles")
435
else:
436
raise BzrError("invalid controlfile mode %r" % mode)
437
438
def put_controlfile(self, path, f, encode=True):
439
"""Write an entry as a controlfile.
440
441
:param path: The path to put the file, relative to the .bzr control
442
directory
443
:param f: A file-like or string object whose contents should be copied.
444
:param encode: If true, encode the contents as utf-8
445
"""
446
self.put_controlfiles([(path, f)], encode=encode)
447
448
def put_controlfiles(self, files, encode=True):
449
"""Write several entries as controlfiles.
450
451
:param files: A list of [(path, file)] pairs, where the path is the directory
452
underneath the bzr control directory
453
:param encode: If true, encode the contents as utf-8
454
"""
455
import codecs
456
ctrl_files = []
457
for path, f in files:
458
if encode:
459
if isinstance(f, basestring):
460
f = f.encode('utf-8', 'replace')
461
else:
462
f = codecs.getwriter('utf-8')(f, errors='replace')
463
path = self._rel_controlfilename(path)
464
ctrl_files.append((path, f))
465
self._transport.put_multi(ctrl_files)
115
466
116
467
def _make_control(self):
117
os.mkdir(self.controlfilename([]))
118
self.controlfile('README', 'w').write(
468
from bzrlib.inventory import Inventory
469
from bzrlib.weavefile import write_weave_v5
470
from bzrlib.weave import Weave
471
472
# Create an empty inventory
473
sio = StringIO()
474
# if we want per-tree root ids then this is the place to set
475
# them; they're not needed for now and so ommitted for
476
# simplicity.
477
bzrlib.xml5.serializer_v5.write_inventory(Inventory(), sio)
478
empty_inv = sio.getvalue()
479
sio = StringIO()
480
bzrlib.weavefile.write_weave_v5(Weave(), sio)
481
empty_weave = sio.getvalue()
482
483
dirs = [[], 'revision-store', 'weaves']
484
files = [('README',
119
485
"This is a Bazaar-NG control directory.\n"
120
"Do not change any files in this directory.")
121
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
122
for d in ('text-store', 'inventory-store', 'revision-store'):
123
os.mkdir(self.controlfilename(d))
124
for f in ('revision-history', 'merged-patches',
125
'pending-merged-patches', 'branch-name'):
126
self.controlfile(f, 'w').write('')
127
mutter('created control directory in ' + self.base)
128
Inventory().write_xml(self.controlfile('inventory','w'))
129
130
131
def _check_format(self):
486
"Do not change any files in this directory.\n"),
487
('branch-format', BZR_BRANCH_FORMAT_6),
488
('revision-history', ''),
489
('branch-name', ''),
490
('branch-lock', ''),
491
('pending-merges', ''),
492
('inventory', empty_inv),
493
('inventory.weave', empty_weave),
494
('ancestry.weave', empty_weave)
495
]
496
cfn = self._rel_controlfilename
497
self._transport.mkdir_multi([cfn(d) for d in dirs])
498
self.put_controlfiles(files)
499
mutter('created control directory in ' + self._transport.base)
500
501
def _check_format(self, relax_version_check):
132
502
"""Check this branch format is supported.
133
503
134
The current tool only supports the current unstable format.
504
The format level is stored, as an integer, in
505
self._branch_format for code that needs to check it later.
135
506
136
507
In the future, we might need different in-memory Branch
137
508
classes to support downlevel branches. But not yet.
138
"""
139
# read in binary mode to detect newline wierdness.
140
fmt = self.controlfile('branch-format', 'rb').read()
141
if fmt != BZR_BRANCH_FORMAT:
142
bailout('sorry, branch format %r not supported' % fmt,
143
['use a different bzr version',
144
'or remove the .bzr directory and "bzr init" again'])
145
509
"""
510
try:
511
fmt = self.controlfile('branch-format', 'r').read()
512
except NoSuchFile:
513
raise NotBranchError(self.base)
514
mutter("got branch format %r", fmt)
515
if fmt == BZR_BRANCH_FORMAT_6:
516
self._branch_format = 6
517
elif fmt == BZR_BRANCH_FORMAT_5:
518
self._branch_format = 5
519
elif fmt == BZR_BRANCH_FORMAT_4:
520
self._branch_format = 4
521
522
if (not relax_version_check
523
and self._branch_format not in (5, 6)):
524
raise errors.UnsupportedFormatError(
525
'sorry, branch format %r not supported' % fmt,
526
['use a different bzr version',
527
'or remove the .bzr directory'
528
' and "bzr init" again'])
529
530
def get_root_id(self):
531
"""Return the id of this branches root"""
532
inv = self.read_working_inventory()
533
return inv.root.file_id
534
535
def set_root_id(self, file_id):
536
inv = self.read_working_inventory()
537
orig_root_id = inv.root.file_id
538
del inv._byid[inv.root.file_id]
539
inv.root.file_id = file_id
540
inv._byid[inv.root.file_id] = inv.root
541
for fid in inv:
542
entry = inv[fid]
543
if entry.parent_id in (None, orig_root_id):
544
entry.parent_id = inv.root.file_id
545
self._write_inventory(inv)
146
546
147
547
def read_working_inventory(self):
148
548
"""Read the working inventory."""
149
before = time.time()
150
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
151
mutter("loaded inventory of %d items in %f"
152
% (len(inv), time.time() - before))
153
return inv
154
549
self.lock_read()
550
try:
551
# ElementTree does its own conversion from UTF-8, so open in
552
# binary.
553
f = self.controlfile('inventory', 'rb')
554
return bzrlib.xml5.serializer_v5.read_inventory(f)
555
finally:
556
self.unlock()
557
155
558
156
559
def _write_inventory(self, inv):
157
560
"""Update the working inventory.
159
562
That is to say, the inventory describing changes underway, that
160
563
will be committed to the next revision.
161
564
"""
162
## TODO: factor out to atomicfile? is rename safe on windows?
163
tmpfname = self.controlfilename('inventory.tmp')
164
tmpf = file(tmpfname, 'w')
165
inv.write_xml(tmpf)
166
tmpf.close()
167
os.rename(tmpfname, self.controlfilename('inventory'))
565
from cStringIO import StringIO
566
self.lock_write()
567
try:
568
sio = StringIO()
569
bzrlib.xml5.serializer_v5.write_inventory(inv, sio)
570
sio.seek(0)
571
# Transport handles atomicity
572
self.put_controlfile('inventory', sio)
573
finally:
574
self.unlock()
575
168
576
mutter('wrote working inventory')
169
170
577
171
578
inventory = property(read_working_inventory, _write_inventory, None,
172
579
"""Inventory for the working copy.""")
173
580
174
175
def add(self, files, verbose=False):
581
def add(self, files, ids=None):
176
582
"""Make files versioned.
177
583
584
Note that the command line normally calls smart_add instead,
585
which can automatically recurse.
586
178
587
This puts the files in the Added state, so that they will be
179
588
recorded by the next commit.
180
589
181
:todo: Perhaps have an option to add the ids even if the files do
182
not (yet) exist.
183
184
:todo: Perhaps return the ids of the files? But then again it
185
is easy to retrieve them if they're needed.
186
187
:todo: Option to specify file id.
188
189
:todo: Adding a directory should optionally recurse down and
190
add all non-ignored children. Perhaps do that in a
191
higher-level method.
192
193
>>> b = ScratchBranch(files=['foo'])
194
>>> 'foo' in b.unknowns()
195
True
196
>>> b.show_status()
197
? foo
198
>>> b.add('foo')
199
>>> 'foo' in b.unknowns()
200
False
201
>>> bool(b.inventory.path2id('foo'))
202
True
203
>>> b.show_status()
204
A foo
205
206
>>> b.add('foo')
207
Traceback (most recent call last):
208
...
209
BzrError: ('foo is already versioned', [])
210
211
>>> b.add(['nothere'])
212
Traceback (most recent call last):
213
BzrError: ('cannot add: not a regular file or directory: nothere', [])
590
files
591
List of paths to add, relative to the base of the tree.
592
593
ids
594
If set, use these instead of automatically generated ids.
595
Must be the same length as the list of files, but may
596
contain None for ids that are to be autogenerated.
597
598
TODO: Perhaps have an option to add the ids even if the files do
599
not (yet) exist.
600
601
TODO: Perhaps yield the ids and paths as they're added.
214
602
"""
215
216
603
# TODO: Re-adding a file that is removed in the working copy
217
604
# should probably put it back with the previous ID.
218
if isinstance(files, types.StringTypes):
605
if isinstance(files, basestring):
606
assert(ids is None or isinstance(ids, basestring))
219
607
files = [files]
220
221
inv = self.read_working_inventory()
222
for f in files:
223
if is_control_file(f):
224
bailout("cannot add control file %s" % quotefn(f))
225
226
fp = splitpath(f)
227
228
if len(fp) == 0:
229
bailout("cannot add top-level %r" % f)
230
231
fullpath = os.path.normpath(self._rel(f))
232
233
if isfile(fullpath):
234
kind = 'file'
235
elif isdir(fullpath):
236
kind = 'directory'
237
else:
238
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
239
240
if len(fp) > 1:
241
parent_name = joinpath(fp[:-1])
242
mutter("lookup parent %r" % parent_name)
243
parent_id = inv.path2id(parent_name)
244
if parent_id == None:
245
bailout("cannot add: parent %r is not versioned"
246
% joinpath(fp[:-1]))
247
else:
248
parent_id = None
249
250
file_id = _gen_file_id(fp[-1])
251
inv.add(InventoryEntry(file_id, fp[-1], kind=kind, parent_id=parent_id))
252
if verbose:
253
show_status('A', kind, quotefn(f))
254
255
mutter("add file %s file_id:{%s} kind=%r parent_id={%s}"
256
% (f, file_id, kind, parent_id))
257
self._write_inventory(inv)
258
608
if ids is not None:
609
ids = [ids]
610
611
if ids is None:
612
ids = [None] * len(files)
613
else:
614
assert(len(ids) == len(files))
615
616
self.lock_write()
617
try:
618
inv = self.read_working_inventory()
619
for f,file_id in zip(files, ids):
620
if is_control_file(f):
621
raise BzrError("cannot add control file %s" % quotefn(f))
622
623
fp = splitpath(f)
624
625
if len(fp) == 0:
626
raise BzrError("cannot add top-level %r" % f)
627
628
fullpath = os.path.normpath(self.abspath(f))
629
630
try:
631
kind = file_kind(fullpath)
632
except OSError:
633
# maybe something better?
634
raise BzrError('cannot add: not a regular file, symlink or directory: %s' % quotefn(f))
635
636
if not InventoryEntry.versionable_kind(kind):
637
raise BzrError('cannot add: not a versionable file ('
638
'i.e. regular file, symlink or directory): %s' % quotefn(f))
639
640
if file_id is None:
641
file_id = gen_file_id(f)
642
inv.add_path(f, kind=kind, file_id=file_id)
643
644
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
645
646
self._write_inventory(inv)
647
finally:
648
self.unlock()
649
650
651
def print_file(self, file, revno):
652
"""Print `file` to stdout."""
653
self.lock_read()
654
try:
655
tree = self.revision_tree(self.get_rev_id(revno))
656
# use inventory as it was in that revision
657
file_id = tree.inventory.path2id(file)
658
if not file_id:
659
raise BzrError("%r is not present in revision %s" % (file, revno))
660
tree.print_file(file_id)
661
finally:
662
self.unlock()
259
663
260
664
261
665
def remove(self, files, verbose=False):
263
667
264
668
This does not remove their text. This does not run on
265
669
266
:todo: Refuse to remove modified files unless --force is given?
267
268
>>> b = ScratchBranch(files=['foo'])
269
>>> b.add('foo')
270
>>> b.inventory.has_filename('foo')
271
True
272
>>> b.remove('foo')
273
>>> b.working_tree().has_filename('foo')
274
True
275
>>> b.inventory.has_filename('foo')
276
False
277
278
>>> b = ScratchBranch(files=['foo'])
279
>>> b.add('foo')
280
>>> b.commit('one')
281
>>> b.remove('foo')
282
>>> b.commit('two')
283
>>> b.inventory.has_filename('foo')
284
False
285
>>> b.basis_tree().has_filename('foo')
286
False
287
>>> b.working_tree().has_filename('foo')
288
True
289
290
:todo: Do something useful with directories.
291
292
:todo: Should this remove the text or not? Tough call; not
670
TODO: Refuse to remove modified files unless --force is given?
671
672
TODO: Do something useful with directories.
673
674
TODO: Should this remove the text or not? Tough call; not
293
675
removing may be useful and the user can just use use rm, and
294
676
is the opposite of add. Removing it is consistent with most
295
677
other tools. Maybe an option.
296
678
"""
297
679
## TODO: Normalize names
298
680
## TODO: Remove nested loops; better scalability
299
300
if isinstance(files, types.StringTypes):
681
if isinstance(files, basestring):
301
682
files = [files]
302
303
tree = self.working_tree()
304
inv = tree.inventory
305
306
# do this before any modifications
307
for f in files:
308
fid = inv.path2id(f)
309
if not fid:
310
bailout("cannot remove unversioned file %s" % quotefn(f))
311
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
312
if verbose:
313
# having remove it, it must be either ignored or unknown
314
if tree.is_ignored(f):
315
new_status = 'I'
316
else:
317
new_status = '?'
318
show_status(new_status, inv[fid].kind, quotefn(f))
319
del inv[fid]
320
683
684
self.lock_write()
685
686
try:
687
tree = self.working_tree()
688
inv = tree.inventory
689
690
# do this before any modifications
691
for f in files:
692
fid = inv.path2id(f)
693
if not fid:
694
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
695
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
696
if verbose:
697
# having remove it, it must be either ignored or unknown
698
if tree.is_ignored(f):
699
new_status = 'I'
700
else:
701
new_status = '?'
702
show_status(new_status, inv[fid].kind, quotefn(f))
703
del inv[fid]
704
705
self._write_inventory(inv)
706
finally:
707
self.unlock()
708
709
# FIXME: this doesn't need to be a branch method
710
def set_inventory(self, new_inventory_list):
711
from bzrlib.inventory import Inventory, InventoryEntry
712
inv = Inventory(self.get_root_id())
713
for path, file_id, parent, kind in new_inventory_list:
714
name = os.path.basename(path)
715
if name == "":
716
continue
717
# fixme, there should be a factory function inv,add_??
718
if kind == 'directory':
719
inv.add(inventory.InventoryDirectory(file_id, name, parent))
720
elif kind == 'file':
721
inv.add(inventory.InventoryFile(file_id, name, parent))
722
elif kind == 'symlink':
723
inv.add(inventory.InventoryLink(file_id, name, parent))
724
else:
725
raise BzrError("unknown kind %r" % kind)
321
726
self._write_inventory(inv)
322
727
323
324
728
def unknowns(self):
325
729
"""Return all unknown files.
326
730
340
744
return self.working_tree().unknowns()
341
745
342
746
343
def commit(self, message, timestamp=None, timezone=None,
344
committer=None,
345
verbose=False):
346
"""Commit working copy as a new revision.
347
348
The basic approach is to add all the file texts into the
349
store, then the inventory, then make a new revision pointing
350
to that inventory and store that.
351
352
This is not quite safe if the working copy changes during the
353
commit; for the moment that is simply not allowed. A better
354
approach is to make a temporary copy of the files before
355
computing their hashes, and then add those hashes in turn to
356
the inventory. This should mean at least that there are no
357
broken hash pointers. There is no way we can get a snapshot
358
of the whole directory at an instant. This would also have to
359
be robust against files disappearing, moving, etc. So the
360
whole thing is a bit hard.
361
362
:param timestamp: if not None, seconds-since-epoch for a
363
postdated/predated commit.
364
"""
365
366
## TODO: Show branch names
367
368
# TODO: Don't commit if there are no changes, unless forced?
369
370
# First walk over the working inventory; and both update that
371
# and also build a new revision inventory. The revision
372
# inventory needs to hold the text-id, sha1 and size of the
373
# actual file versions committed in the revision. (These are
374
# not present in the working inventory.) We also need to
375
# detect missing/deleted files, and remove them from the
376
# working inventory.
377
378
work_inv = self.read_working_inventory()
379
inv = Inventory()
380
basis = self.basis_tree()
381
basis_inv = basis.inventory
382
missing_ids = []
383
for path, entry in work_inv.iter_entries():
384
## TODO: Cope with files that have gone missing.
385
386
## TODO: Check that the file kind has not changed from the previous
387
## revision of this file (if any).
388
389
entry = entry.copy()
390
391
p = self._rel(path)
392
file_id = entry.file_id
393
mutter('commit prep file %s, id %r ' % (p, file_id))
394
395
if not os.path.exists(p):
396
mutter(" file is missing, removing from inventory")
397
if verbose:
398
show_status('D', entry.kind, quotefn(path))
399
missing_ids.append(file_id)
400
continue
401
402
# TODO: Handle files that have been deleted
403
404
# TODO: Maybe a special case for empty files? Seems a
405
# waste to store them many times.
406
407
inv.add(entry)
408
409
if basis_inv.has_id(file_id):
410
old_kind = basis_inv[file_id].kind
411
if old_kind != entry.kind:
412
bailout("entry %r changed kind from %r to %r"
413
% (file_id, old_kind, entry.kind))
414
415
if entry.kind == 'directory':
416
if not isdir(p):
417
bailout("%s is entered as directory but not a directory" % quotefn(p))
418
elif entry.kind == 'file':
419
if not isfile(p):
420
bailout("%s is entered as file but is not a file" % quotefn(p))
421
422
content = file(p, 'rb').read()
423
424
entry.text_sha1 = sha_string(content)
425
entry.text_size = len(content)
426
427
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
428
if (old_ie
429
and (old_ie.text_size == entry.text_size)
430
and (old_ie.text_sha1 == entry.text_sha1)):
431
## assert content == basis.get_file(file_id).read()
432
entry.text_id = basis_inv[file_id].text_id
433
mutter(' unchanged from previous text_id {%s}' %
434
entry.text_id)
435
436
else:
437
entry.text_id = _gen_file_id(entry.name)
438
self.text_store.add(content, entry.text_id)
439
mutter(' stored with text_id {%s}' % entry.text_id)
440
if verbose:
441
if not old_ie:
442
state = 'A'
443
elif (old_ie.name == entry.name
444
and old_ie.parent_id == entry.parent_id):
445
state = 'R'
446
else:
447
state = 'M'
448
449
show_status(state, entry.kind, quotefn(path))
450
451
for file_id in missing_ids:
452
# have to do this later so we don't mess up the iterator.
453
# since parents may be removed before their children we
454
# have to test.
455
456
# FIXME: There's probably a better way to do this; perhaps
457
# the workingtree should know how to filter itself.
458
if work_inv.has_id(file_id):
459
del work_inv[file_id]
460
461
462
inv_id = rev_id = _gen_revision_id(time.time())
463
464
inv_tmp = tempfile.TemporaryFile()
465
inv.write_xml(inv_tmp)
466
inv_tmp.seek(0)
467
self.inventory_store.add(inv_tmp, inv_id)
468
mutter('new inventory_id is {%s}' % inv_id)
469
470
self._write_inventory(work_inv)
471
472
if timestamp == None:
473
timestamp = time.time()
474
475
if committer == None:
476
committer = username()
477
478
if timezone == None:
479
timezone = local_time_offset()
480
481
mutter("building commit log message")
482
rev = Revision(timestamp=timestamp,
483
timezone=timezone,
484
committer=committer,
485
precursor = self.last_patch(),
486
message = message,
487
inventory_id=inv_id,
488
revision_id=rev_id)
489
490
rev_tmp = tempfile.TemporaryFile()
491
rev.write_xml(rev_tmp)
492
rev_tmp.seek(0)
493
self.revision_store.add(rev_tmp, rev_id)
494
mutter("new revision_id is {%s}" % rev_id)
495
496
## XXX: Everything up to here can simply be orphaned if we abort
497
## the commit; it will leave junk files behind but that doesn't
498
## matter.
499
500
## TODO: Read back the just-generated changeset, and make sure it
501
## applies and recreates the right state.
502
503
## TODO: Also calculate and store the inventory SHA1
504
mutter("committing patch r%d" % (self.revno() + 1))
505
506
mutter("append to revision-history")
507
self.controlfile('revision-history', 'at').write(rev_id + '\n')
508
509
mutter("done!")
747
def append_revision(self, *revision_ids):
748
for revision_id in revision_ids:
749
mutter("add {%s} to revision-history" % revision_id)
750
self.lock_write()
751
try:
752
rev_history = self.revision_history()
753
rev_history.extend(revision_ids)
754
self.put_controlfile('revision-history', '\n'.join(rev_history))
755
finally:
756
self.unlock()
757
758
def has_revision(self, revision_id):
759
"""True if this branch has a copy of the revision.
760
761
This does not necessarily imply the revision is merge
762
or on the mainline."""
763
return (revision_id is None
764
or revision_id in self.revision_store)
765
766
def get_revision_xml_file(self, revision_id):
767
"""Return XML file object for revision object."""
768
if not revision_id or not isinstance(revision_id, basestring):
769
raise InvalidRevisionId(revision_id)
770
771
self.lock_read()
772
try:
773
try:
774
return self.revision_store[revision_id]
775
except (IndexError, KeyError):
776
raise bzrlib.errors.NoSuchRevision(self, revision_id)
777
finally:
778
self.unlock()
779
780
#deprecated
781
get_revision_xml = get_revision_xml_file
782
783
def get_revision_xml(self, revision_id):
784
return self.get_revision_xml_file(revision_id).read()
510
785
511
786
512
787
def get_revision(self, revision_id):
513
788
"""Return the Revision object for a named revision"""
514
r = Revision.read_xml(self.revision_store[revision_id])
789
xml_file = self.get_revision_xml_file(revision_id)
790
791
try:
792
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
793
except SyntaxError, e:
794
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
795
[revision_id,
796
str(e)])
797
515
798
assert r.revision_id == revision_id
516
799
return r
517
800
518
519
def get_inventory(self, inventory_id):
520
"""Get Inventory object by hash.
521
522
:todo: Perhaps for this and similar methods, take a revision
523
parameter which can be either an integer revno or a
524
string hash."""
525
i = Inventory.read_xml(self.inventory_store[inventory_id])
526
return i
527
801
def get_revision_delta(self, revno):
802
"""Return the delta for one revision.
803
804
The delta is relative to its mainline predecessor, or the
805
empty tree for revision 1.
806
"""
807
assert isinstance(revno, int)
808
rh = self.revision_history()
809
if not (1 <= revno <= len(rh)):
810
raise InvalidRevisionNumber(revno)
811
812
# revno is 1-based; list is 0-based
813
814
new_tree = self.revision_tree(rh[revno-1])
815
if revno == 1:
816
old_tree = EmptyTree()
817
else:
818
old_tree = self.revision_tree(rh[revno-2])
819
820
return compare_trees(old_tree, new_tree)
821
822
def get_revision_sha1(self, revision_id):
823
"""Hash the stored value of a revision, and return it."""
824
# In the future, revision entries will be signed. At that
825
# point, it is probably best *not* to include the signature
826
# in the revision hash. Because that lets you re-sign
827
# the revision, (add signatures/remove signatures) and still
828
# have all hash pointers stay consistent.
829
# But for now, just hash the contents.
830
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
831
832
def get_ancestry(self, revision_id):
833
"""Return a list of revision-ids integrated by a revision.
834
835
This currently returns a list, but the ordering is not guaranteed:
836
treat it as a set.
837
"""
838
if revision_id is None:
839
return [None]
840
w = self.get_inventory_weave()
841
return [None] + map(w.idx_to_name,
842
w.inclusions([w.lookup(revision_id)]))
843
844
def get_inventory_weave(self):
845
return self.control_weaves.get_weave('inventory',
846
self.get_transaction())
847
848
def get_inventory(self, revision_id):
849
"""Get Inventory object by hash."""
850
xml = self.get_inventory_xml(revision_id)
851
return bzrlib.xml5.serializer_v5.read_inventory_from_string(xml)
852
853
def get_inventory_xml(self, revision_id):
854
"""Get inventory XML as a file object."""
855
try:
856
assert isinstance(revision_id, basestring), type(revision_id)
857
iw = self.get_inventory_weave()
858
return iw.get_text(iw.lookup(revision_id))
859
except IndexError:
860
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
861
862
def get_inventory_sha1(self, revision_id):
863
"""Return the sha1 hash of the inventory entry
864
"""
865
return self.get_revision(revision_id).inventory_sha1
528
866
529
867
def get_revision_inventory(self, revision_id):
530
868
"""Return inventory of a past revision."""
869
# TODO: Unify this with get_inventory()
870
# bzr 0.0.6 and later imposes the constraint that the inventory_id
871
# must be the same as its revision, so this is trivial.
531
872
if revision_id == None:
532
return Inventory()
873
return Inventory(self.get_root_id())
533
874
else:
534
return self.get_inventory(self.get_revision(revision_id).inventory_id)
535
875
return self.get_inventory(revision_id)
536
876
537
877
def revision_history(self):
538
"""Return sequence of revision hashes on to this branch.
878
"""Return sequence of revision hashes on to this branch."""
879
self.lock_read()
880
try:
881
return [l.rstrip('\r\n') for l in
882
self.controlfile('revision-history', 'r').readlines()]
883
finally:
884
self.unlock()
539
885
540
>>> ScratchBranch().revision_history()
541
[]
542
"""
543
return [chomp(l) for l in self.controlfile('revision-history').readlines()]
886
def common_ancestor(self, other, self_revno=None, other_revno=None):
887
"""
888
>>> from bzrlib.commit import commit
889
>>> sb = ScratchBranch(files=['foo', 'foo~'])
890
>>> sb.common_ancestor(sb) == (None, None)
891
True
892
>>> commit(sb, "Committing first revision", verbose=False)
893
>>> sb.common_ancestor(sb)[0]
894
1
895
>>> clone = sb.clone()
896
>>> commit(sb, "Committing second revision", verbose=False)
897
>>> sb.common_ancestor(sb)[0]
898
2
899
>>> sb.common_ancestor(clone)[0]
900
1
901
>>> commit(clone, "Committing divergent second revision",
902
... verbose=False)
903
>>> sb.common_ancestor(clone)[0]
904
1
905
>>> sb.common_ancestor(clone) == clone.common_ancestor(sb)
906
True
907
>>> sb.common_ancestor(sb) != clone.common_ancestor(clone)
908
True
909
>>> clone2 = sb.clone()
910
>>> sb.common_ancestor(clone2)[0]
911
2
912
>>> sb.common_ancestor(clone2, self_revno=1)[0]
913
1
914
>>> sb.common_ancestor(clone2, other_revno=1)[0]
915
1
916
"""
917
my_history = self.revision_history()
918
other_history = other.revision_history()
919
if self_revno is None:
920
self_revno = len(my_history)
921
if other_revno is None:
922
other_revno = len(other_history)
923
indices = range(min((self_revno, other_revno)))
924
indices.reverse()
925
for r in indices:
926
if my_history[r] == other_history[r]:
927
return r+1, my_history[r]
928
return None, None
544
929
545
930
546
931
def revno(self):
548
933
549
934
That is equivalent to the number of revisions committed to
550
935
this branch.
551
552
>>> b = ScratchBranch()
553
>>> b.revno()
554
0
555
>>> b.commit('no foo')
556
>>> b.revno()
557
1
558
936
"""
559
937
return len(self.revision_history())
560
938
561
939
562
def last_patch(self):
940
def last_revision(self):
563
941
"""Return last patch hash, or None if no history.
564
565
>>> ScratchBranch().last_patch() == None
566
True
567
942
"""
568
943
ph = self.revision_history()
569
944
if ph:
570
945
return ph[-1]
571
572
573
def lookup_revision(self, revno):
574
"""Return revision hash for revision number."""
946
else:
947
return None
948
949
950
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
951
"""Return a list of new revisions that would perfectly fit.
952
953
If self and other have not diverged, return a list of the revisions
954
present in other, but missing from self.
955
956
>>> from bzrlib.commit import commit
957
>>> bzrlib.trace.silent = True
958
>>> br1 = ScratchBranch()
959
>>> br2 = ScratchBranch()
960
>>> br1.missing_revisions(br2)
961
[]
962
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
963
>>> br1.missing_revisions(br2)
964
[u'REVISION-ID-1']
965
>>> br2.missing_revisions(br1)
966
[]
967
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
968
>>> br1.missing_revisions(br2)
969
[]
970
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
971
>>> br1.missing_revisions(br2)
972
[u'REVISION-ID-2A']
973
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
974
>>> br1.missing_revisions(br2)
975
Traceback (most recent call last):
976
DivergedBranches: These branches have diverged.
977
"""
978
# FIXME: If the branches have diverged, but the latest
979
# revision in this branch is completely merged into the other,
980
# then we should still be able to pull.
981
self_history = self.revision_history()
982
self_len = len(self_history)
983
other_history = other.revision_history()
984
other_len = len(other_history)
985
common_index = min(self_len, other_len) -1
986
if common_index >= 0 and \
987
self_history[common_index] != other_history[common_index]:
988
raise DivergedBranches(self, other)
989
990
if stop_revision is None:
991
stop_revision = other_len
992
else:
993
assert isinstance(stop_revision, int)
994
if stop_revision > other_len:
995
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
996
return other_history[self_len:stop_revision]
997
998
def update_revisions(self, other, stop_revision=None):
999
"""Pull in new perfect-fit revisions."""
1000
from bzrlib.fetch import greedy_fetch
1001
from bzrlib.revision import get_intervening_revisions
1002
if stop_revision is None:
1003
stop_revision = other.last_revision()
1004
greedy_fetch(to_branch=self, from_branch=other,
1005
revision=stop_revision)
1006
pullable_revs = self.missing_revisions(
1007
other, other.revision_id_to_revno(stop_revision))
1008
if pullable_revs:
1009
greedy_fetch(to_branch=self,
1010
from_branch=other,
1011
revision=pullable_revs[-1])
1012
self.append_revision(*pullable_revs)
1013
1014
1015
def commit(self, *args, **kw):
1016
from bzrlib.commit import Commit
1017
Commit().commit(self, *args, **kw)
1018
1019
def revision_id_to_revno(self, revision_id):
1020
"""Given a revision id, return its revno"""
1021
if revision_id is None:
1022
return 0
1023
history = self.revision_history()
1024
try:
1025
return history.index(revision_id) + 1
1026
except ValueError:
1027
raise bzrlib.errors.NoSuchRevision(self, revision_id)
1028
1029
def get_rev_id(self, revno, history=None):
1030
"""Find the revision id of the specified revno."""
575
1031
if revno == 0:
576
1032
return None
577
578
try:
579
# list is 0-based; revisions are 1-based
580
return self.revision_history()[revno-1]
581
except IndexError:
582
bailout("no such revision %s" % revno)
583
1033
if history is None:
1034
history = self.revision_history()
1035
elif revno <= 0 or revno > len(history):
1036
raise bzrlib.errors.NoSuchRevision(self, revno)
1037
return history[revno - 1]
584
1038
585
1039
def revision_tree(self, revision_id):
586
1040
"""Return Tree for a revision on this branch.
587
1041
588
1042
`revision_id` may be None for the null revision, in which case
589
1043
an `EmptyTree` is returned."""
590
1044
# TODO: refactor this to use an existing revision object
1045
# so we don't need to read it in twice.
591
1046
if revision_id == None:
592
1047
return EmptyTree()
593
1048
else:
594
1049
inv = self.get_revision_inventory(revision_id)
595
return RevisionTree(self.text_store, inv)
1050
return RevisionTree(self.weave_store, inv, revision_id)
596
1051
597
1052
598
1053
def working_tree(self):
599
1054
"""Return a `Tree` for the working copy."""
600
return WorkingTree(self.base, self.read_working_inventory())
1055
from bzrlib.workingtree import WorkingTree
1056
# TODO: In the future, WorkingTree should utilize Transport
1057
# RobertCollins 20051003 - I don't think it should - working trees are
1058
# much more complex to keep consistent than our careful .bzr subset.
1059
# instead, we should say that working trees are local only, and optimise
1060
# for that.
1061
return WorkingTree(self._transport.base, self.read_working_inventory())
601
1062
602
1063
603
1064
def basis_tree(self):
604
1065
"""Return `Tree` object for last revision.
605
1066
606
1067
If there are no revisions yet, return an `EmptyTree`.
607
608
>>> b = ScratchBranch(files=['foo'])
609
>>> b.basis_tree().has_filename('foo')
610
False
611
>>> b.working_tree().has_filename('foo')
612
True
613
>>> b.add('foo')
614
>>> b.commit('add foo')
615
>>> b.basis_tree().has_filename('foo')
616
True
617
"""
618
r = self.last_patch()
619
if r == None:
620
return EmptyTree()
621
else:
622
return RevisionTree(self.text_store, self.get_revision_inventory(r))
623
624
625
626
def write_log(self, show_timezone='original'):
627
"""Write out human-readable log of commits to this branch
628
629
:param utc: If true, show dates in universal time, not local time."""
630
## TODO: Option to choose either original, utc or local timezone
631
revno = 1
632
precursor = None
633
for p in self.revision_history():
634
print '-' * 40
635
print 'revno:', revno
636
## TODO: Show hash if --id is given.
637
##print 'revision-hash:', p
638
rev = self.get_revision(p)
639
print 'committer:', rev.committer
640
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0,
641
show_timezone))
642
643
## opportunistic consistency check, same as check_patch_chaining
644
if rev.precursor != precursor:
645
bailout("mismatched precursor!")
646
647
print 'message:'
648
if not rev.message:
649
print ' (no message)'
650
else:
651
for l in rev.message.split('\n'):
652
print ' ' + l
653
654
revno += 1
655
precursor = p
656
657
658
659
def show_status(branch, show_all=False):
660
"""Display single-line status for non-ignored working files.
661
662
The list is show sorted in order by file name.
663
664
>>> b = ScratchBranch(files=['foo', 'foo~'])
665
>>> b.show_status()
666
? foo
667
>>> b.add('foo')
668
>>> b.show_status()
669
A foo
670
>>> b.commit("add foo")
671
>>> b.show_status()
672
>>> os.unlink(b._rel('foo'))
673
>>> b.show_status()
674
D foo
675
676
677
:todo: Get state for single files.
678
679
:todo: Perhaps show a slash at the end of directory names.
680
681
"""
682
683
# We have to build everything into a list first so that it can
684
# sorted by name, incorporating all the different sources.
685
686
# FIXME: Rather than getting things in random order and then sorting,
687
# just step through in order.
688
689
# Interesting case: the old ID for a file has been removed,
690
# but a new file has been created under that name.
691
692
old = branch.basis_tree()
693
old_inv = old.inventory
694
new = branch.working_tree()
695
new_inv = new.inventory
696
697
for fs, fid, oldname, newname, kind in diff_trees(old, new):
698
if fs == 'R':
699
show_status(fs, kind,
700
oldname + ' => ' + newname)
701
elif fs == 'A' or fs == 'M':
702
show_status(fs, kind, newname)
703
elif fs == 'D':
704
show_status(fs, kind, oldname)
705
elif fs == '.':
706
if show_all:
707
show_status(fs, kind, newname)
708
elif fs == 'I':
709
if show_all:
710
show_status(fs, kind, newname)
711
elif fs == '?':
712
show_status(fs, kind, newname)
713
else:
714
bailout("wierd file state %r" % ((fs, fid),))
715
716
717
718
class ScratchBranch(Branch):
1068
"""
1069
return self.revision_tree(self.last_revision())
1070
1071
1072
def rename_one(self, from_rel, to_rel):
1073
"""Rename one file.
1074
1075
This can change the directory or the filename or both.
1076
"""
1077
self.lock_write()
1078
try:
1079
tree = self.working_tree()
1080
inv = tree.inventory
1081
if not tree.has_filename(from_rel):
1082
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
1083
if tree.has_filename(to_rel):
1084
raise BzrError("can't rename: new working file %r already exists" % to_rel)
1085
1086
file_id = inv.path2id(from_rel)
1087
if file_id == None:
1088
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
1089
1090
if inv.path2id(to_rel):
1091
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
1092
1093
to_dir, to_tail = os.path.split(to_rel)
1094
to_dir_id = inv.path2id(to_dir)
1095
if to_dir_id == None and to_dir != '':
1096
raise BzrError("can't determine destination directory id for %r" % to_dir)
1097
1098
mutter("rename_one:")
1099
mutter(" file_id {%s}" % file_id)
1100
mutter(" from_rel %r" % from_rel)
1101
mutter(" to_rel %r" % to_rel)
1102
mutter(" to_dir %r" % to_dir)
1103
mutter(" to_dir_id {%s}" % to_dir_id)
1104
1105
inv.rename(file_id, to_dir_id, to_tail)
1106
1107
from_abs = self.abspath(from_rel)
1108
to_abs = self.abspath(to_rel)
1109
try:
1110
rename(from_abs, to_abs)
1111
except OSError, e:
1112
raise BzrError("failed to rename %r to %r: %s"
1113
% (from_abs, to_abs, e[1]),
1114
["rename rolled back"])
1115
1116
self._write_inventory(inv)
1117
finally:
1118
self.unlock()
1119
1120
1121
def move(self, from_paths, to_name):
1122
"""Rename files.
1123
1124
to_name must exist as a versioned directory.
1125
1126
If to_name exists and is a directory, the files are moved into
1127
it, keeping their old names. If it is a directory,
1128
1129
Note that to_name is only the last component of the new name;
1130
this doesn't change the directory.
1131
1132
This returns a list of (from_path, to_path) pairs for each
1133
entry that is moved.
1134
"""
1135
result = []
1136
self.lock_write()
1137
try:
1138
## TODO: Option to move IDs only
1139
assert not isinstance(from_paths, basestring)
1140
tree = self.working_tree()
1141
inv = tree.inventory
1142
to_abs = self.abspath(to_name)
1143
if not isdir(to_abs):
1144
raise BzrError("destination %r is not a directory" % to_abs)
1145
if not tree.has_filename(to_name):
1146
raise BzrError("destination %r not in working directory" % to_abs)
1147
to_dir_id = inv.path2id(to_name)
1148
if to_dir_id == None and to_name != '':
1149
raise BzrError("destination %r is not a versioned directory" % to_name)
1150
to_dir_ie = inv[to_dir_id]
1151
if to_dir_ie.kind not in ('directory', 'root_directory'):
1152
raise BzrError("destination %r is not a directory" % to_abs)
1153
1154
to_idpath = inv.get_idpath(to_dir_id)
1155
1156
for f in from_paths:
1157
if not tree.has_filename(f):
1158
raise BzrError("%r does not exist in working tree" % f)
1159
f_id = inv.path2id(f)
1160
if f_id == None:
1161
raise BzrError("%r is not versioned" % f)
1162
name_tail = splitpath(f)[-1]
1163
dest_path = appendpath(to_name, name_tail)
1164
if tree.has_filename(dest_path):
1165
raise BzrError("destination %r already exists" % dest_path)
1166
if f_id in to_idpath:
1167
raise BzrError("can't move %r to a subdirectory of itself" % f)
1168
1169
# OK, so there's a race here, it's possible that someone will
1170
# create a file in this interval and then the rename might be
1171
# left half-done. But we should have caught most problems.
1172
1173
for f in from_paths:
1174
name_tail = splitpath(f)[-1]
1175
dest_path = appendpath(to_name, name_tail)
1176
result.append((f, dest_path))
1177
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1178
try:
1179
rename(self.abspath(f), self.abspath(dest_path))
1180
except OSError, e:
1181
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1182
["rename rolled back"])
1183
1184
self._write_inventory(inv)
1185
finally:
1186
self.unlock()
1187
1188
return result
1189
1190
1191
def revert(self, filenames, old_tree=None, backups=True):
1192
"""Restore selected files to the versions from a previous tree.
1193
1194
backups
1195
If true (default) backups are made of files before
1196
they're renamed.
1197
"""
1198
from bzrlib.errors import NotVersionedError, BzrError
1199
from bzrlib.atomicfile import AtomicFile
1200
from bzrlib.osutils import backup_file
1201
1202
inv = self.read_working_inventory()
1203
if old_tree is None:
1204
old_tree = self.basis_tree()
1205
old_inv = old_tree.inventory
1206
1207
nids = []
1208
for fn in filenames:
1209
file_id = inv.path2id(fn)
1210
if not file_id:
1211
raise NotVersionedError("not a versioned file", fn)
1212
if not old_inv.has_id(file_id):
1213
raise BzrError("file not present in old tree", fn, file_id)
1214
nids.append((fn, file_id))
1215
1216
# TODO: Rename back if it was previously at a different location
1217
1218
# TODO: If given a directory, restore the entire contents from
1219
# the previous version.
1220
1221
# TODO: Make a backup to a temporary file.
1222
1223
# TODO: If the file previously didn't exist, delete it?
1224
for fn, file_id in nids:
1225
backup_file(fn)
1226
1227
f = AtomicFile(fn, 'wb')
1228
try:
1229
f.write(old_tree.get_file(file_id).read())
1230
f.commit()
1231
finally:
1232
f.close()
1233
1234
1235
def pending_merges(self):
1236
"""Return a list of pending merges.
1237
1238
These are revisions that have been merged into the working
1239
directory but not yet committed.
1240
"""
1241
cfn = self._rel_controlfilename('pending-merges')
1242
if not self._transport.has(cfn):
1243
return []
1244
p = []
1245
for l in self.controlfile('pending-merges', 'r').readlines():
1246
p.append(l.rstrip('\n'))
1247
return p
1248
1249
1250
def add_pending_merge(self, *revision_ids):
1251
# TODO: Perhaps should check at this point that the
1252
# history of the revision is actually present?
1253
p = self.pending_merges()
1254
updated = False
1255
for rev_id in revision_ids:
1256
if rev_id in p:
1257
continue
1258
p.append(rev_id)
1259
updated = True
1260
if updated:
1261
self.set_pending_merges(p)
1262
1263
def set_pending_merges(self, rev_list):
1264
self.lock_write()
1265
try:
1266
self.put_controlfile('pending-merges', '\n'.join(rev_list))
1267
finally:
1268
self.unlock()
1269
1270
1271
def get_parent(self):
1272
"""Return the parent location of the branch.
1273
1274
This is the default location for push/pull/missing. The usual
1275
pattern is that the user can override it by specifying a
1276
location.
1277
"""
1278
import errno
1279
_locs = ['parent', 'pull', 'x-pull']
1280
for l in _locs:
1281
try:
1282
return self.controlfile(l, 'r').read().strip('\n')
1283
except IOError, e:
1284
if e.errno != errno.ENOENT:
1285
raise
1286
return None
1287
1288
1289
def set_parent(self, url):
1290
# TODO: Maybe delete old location files?
1291
from bzrlib.atomicfile import AtomicFile
1292
self.lock_write()
1293
try:
1294
f = AtomicFile(self.controlfilename('parent'))
1295
try:
1296
f.write(url + '\n')
1297
f.commit()
1298
finally:
1299
f.close()
1300
finally:
1301
self.unlock()
1302
1303
def check_revno(self, revno):
1304
"""\
1305
Check whether a revno corresponds to any revision.
1306
Zero (the NULL revision) is considered valid.
1307
"""
1308
if revno != 0:
1309
self.check_real_revno(revno)
1310
1311
def check_real_revno(self, revno):
1312
"""\
1313
Check whether a revno corresponds to a real revision.
1314
Zero (the NULL revision) is considered invalid
1315
"""
1316
if revno < 1 or revno > self.revno():
1317
raise InvalidRevisionNumber(revno)
1318
1319
1320
1321
1322
1323
class ScratchBranch(_Branch):
719
1324
"""Special test class: a branch that cleans up after itself.
720
1325
721
1326
>>> b = ScratchBranch()
722
1327
>>> isdir(b.base)
723
1328
True
724
1329
>>> bd = b.base
725
>>> del b
1330
>>> b.destroy()
726
1331
>>> isdir(bd)
727
1332
False
728
1333
"""
729
def __init__(self, files = []):
1334
def __init__(self, files=[], dirs=[], base=None):
730
1335
"""Make a test branch.
731
1336
732
1337
This creates a temporary directory and runs init-tree in it.
733
1338
734
1339
If any files are listed, they are created in the working copy.
735
1340
"""
736
Branch.__init__(self, tempfile.mkdtemp(), init=True)
1341
from tempfile import mkdtemp
1342
init = False
1343
if base is None:
1344
base = mkdtemp()
1345
init = True
1346
if isinstance(base, basestring):
1347
base = get_transport(base)
1348
_Branch.__init__(self, base, init=init)
1349
for d in dirs:
1350
self._transport.mkdir(d)
1351
737
1352
for f in files:
738
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
739
1353
self._transport.put(f, 'content of %s' % f)
1354
1355
1356
def clone(self):
1357
"""
1358
>>> orig = ScratchBranch(files=["file1", "file2"])
1359
>>> clone = orig.clone()
1360
>>> if os.name != 'nt':
1361
... os.path.samefile(orig.base, clone.base)
1362
... else:
1363
... orig.base == clone.base
1364
...
1365
False
1366
>>> os.path.isfile(os.path.join(clone.base, "file1"))
1367
True
1368
"""
1369
from shutil import copytree
1370
from tempfile import mkdtemp
1371
base = mkdtemp()
1372
os.rmdir(base)
1373
copytree(self.base, base, symlinks=True)
1374
return ScratchBranch(base=base)
740
1375
741
1376
def __del__(self):
1377
self.destroy()
1378
1379
def destroy(self):
742
1380
"""Destroy the test branch, removing the scratch directory."""
743
shutil.rmtree(self.base)
1381
from shutil import rmtree
1382
try:
1383
if self.base:
1384
mutter("delete ScratchBranch %s" % self.base)
1385
rmtree(self.base)
1386
except OSError, e:
1387
# Work around for shutil.rmtree failing on Windows when
1388
# readonly files are encountered
1389
mutter("hit exception in destroying ScratchBranch: %s" % e)
1390
for root, dirs, files in os.walk(self.base, topdown=False):
1391
for name in files:
1392
os.chmod(os.path.join(root, name), 0700)
1393
rmtree(self.base)
1394
self._transport = None
744
1395
745
1396
746
1397
756
1407
## mutter('check %r for control file' % ((head, tail), ))
757
1408
if tail == bzrlib.BZRDIR:
758
1409
return True
1410
if filename == head:
1411
break
759
1412
filename = head
760
1413
return False
761
1414
762
1415
763
1416
764
def _gen_revision_id(when):
765
"""Return new revision-id."""
766
s = '%s-%s-' % (user_email(), compact_date(when))
767
s += hexlify(rand_bytes(8))
768
return s
769
770
771
def _gen_file_id(name):
1417
def gen_file_id(name):
772
1418
"""Return new file id.
773
1419
774
1420
This should probably generate proper UUIDs, but for the moment we
775
1421
cope with just randomness because running uuidgen every time is
776
1422
slow."""
777
assert '/' not in name
778
while name[0] == '.':
779
name = name[1:]
1423
import re
1424
from binascii import hexlify
1425
from time import time
1426
1427
# get last component
1428
idx = name.rfind('/')
1429
if idx != -1:
1430
name = name[idx+1 : ]
1431
idx = name.rfind('\\')
1432
if idx != -1:
1433
name = name[idx+1 : ]
1434
1435
# make it not a hidden file
1436
name = name.lstrip('.')
1437
1438
# remove any wierd characters; we don't escape them but rather
1439
# just pull them out
1440
name = re.sub(r'[^\w.]', '', name)
1441
780
1442
s = hexlify(rand_bytes(8))
781
return '-'.join((name, compact_date(time.time()), s))
1443
return '-'.join((name, compact_date(time()), s))
1444
1445
1446
def gen_root_id():
1447
"""Return a new tree-root file id."""
1448
return gen_file_id('TREE_ROOT')
782
1449
783
1450
Loggerhead is a web-based interface for Breezy
Version: 2.0.1