15
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
21
from warnings import warn
22
from cStringIO import StringIO
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
25
from inventory import Inventory
26
from trace import mutter, note
27
from tree import Tree, EmptyTree, RevisionTree, WorkingTree
28
from inventory import InventoryEntry, Inventory
29
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, \
30
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
31
joinpath, sha_string, file_kind, local_time_offset, appendpath
32
from store import ImmutableStore
33
from revision import Revision
34
from errors import bailout, BzrError
35
from textui import show_status
36
from diff import diff_trees
38
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
26
from bzrlib.inventory import InventoryEntry
27
from bzrlib.trace import mutter, note
28
from bzrlib.osutils import (isdir, quotefn, compact_date, rand_bytes,
29
rename, splitpath, sha_file, appendpath,
31
from bzrlib.errors import (BzrError, InvalidRevisionNumber, InvalidRevisionId,
32
NoSuchRevision, HistoryMissing, NotBranchError,
33
DivergedBranches, LockError, UnlistableStore,
34
UnlistableBranch, NoSuchFile)
35
from bzrlib.textui import show_status
36
from bzrlib.revision import Revision, validate_revision_id, is_ancestor
37
from bzrlib.delta import compare_trees
38
from bzrlib.tree import EmptyTree, RevisionTree
39
from bzrlib.inventory import Inventory
40
from bzrlib.store import copy_all
41
from bzrlib.store.compressed_text import CompressedTextStore
42
from bzrlib.store.text import TextStore
43
from bzrlib.store.weave import WeaveStore
44
from bzrlib.transport import Transport, get_transport
49
BZR_BRANCH_FORMAT_4 = "Bazaar-NG branch, format 0.0.4\n"
50
BZR_BRANCH_FORMAT_5 = "Bazaar-NG branch, format 5\n"
39
51
## TODO: Maybe include checks for common corruption of newlines, etc?
43
def find_branch_root(f=None):
44
"""Find the branch root enclosing f, or pwd.
46
It is not necessary that f exists.
54
# TODO: Some operations like log might retrieve the same revisions
55
# repeatedly to calculate deltas. We could perhaps have a weakref
56
# cache in memory to make this faster. In general anything can be
57
# cached in memory between lock and unlock operations.
59
def find_branch(*ignored, **ignored_too):
60
# XXX: leave this here for about one release, then remove it
61
raise NotImplementedError('find_branch() is not supported anymore, '
62
'please use one of the new branch constructors')
63
def _relpath(base, path):
64
"""Return path relative to base, or raise exception.
66
The path may be either an absolute path or a path relative to the
67
current working directory.
69
Lifted out of Branch.relpath for ease of testing.
71
os.path.commonprefix (python2.4) has a bad bug that it works just
72
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
73
avoids that problem."""
74
rp = os.path.abspath(path)
78
while len(head) >= len(base):
81
head, tail = os.path.split(head)
85
raise NotBranchError("path %r is not within branch %r" % (rp, base))
90
def find_branch_root(t):
91
"""Find the branch root enclosing the transport's base.
93
t is a Transport object.
95
It is not necessary that the base of t exists.
48
97
Basically we keep looking up until we find the control directory or
52
elif hasattr(os.path, 'realpath'):
53
f = os.path.realpath(f)
55
f = os.path.abspath(f)
98
run into the root. If there isn't one, raises NotBranchError.
60
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
62
head, tail = os.path.split(f)
102
if t.has(bzrlib.BZRDIR):
104
new_t = t.clone('..')
105
if new_t.base == t.base:
64
106
# reached the root, whatever that may be
65
raise BzrError('%r is not in a branch' % orig_f)
107
raise NotBranchError('%s is not in a branch' % orig_base)
70
111
######################################################################
114
class Branch(object):
74
115
"""Branch holding a history of revisions.
77
Base directory of the branch.
118
Base directory/url of the branch.
122
def __init__(self, *ignored, **ignored_too):
123
raise NotImplementedError('The Branch class is abstract')
126
def open_downlevel(base):
127
"""Open a branch which may be of an old format.
129
Only local branches are supported."""
130
return _Branch(get_transport(base), relax_version_check=True)
134
"""Open an existing branch, rooted at 'base' (url)"""
135
t = get_transport(base)
139
def open_containing(url):
140
"""Open an existing branch which contains url.
142
This probes for a branch at url, and searches upwards from there.
144
t = get_transport(url)
145
t = find_branch_root(t)
149
def initialize(base):
150
"""Create a new branch, rooted at 'base' (url)"""
151
t = get_transport(base)
152
return _Branch(t, init=True)
154
def setup_caching(self, cache_root):
155
"""Subclasses that care about caching should override this, and set
156
up cached stores located under cache_root.
160
class _Branch(Branch):
161
"""A branch stored in the actual filesystem.
163
Note that it's "local" in the context of the filesystem; it doesn't
164
really matter if it's on an nfs/smb/afs/coda/... share, as long as
165
it's writable, and can be accessed via the normal filesystem API.
171
If _lock_mode is true, a positive count of the number of times the
175
Lock object from bzrlib.lock.
177
# We actually expect this class to be somewhat short-lived; part of its
178
# purpose is to try to isolate what bits of the branch logic are tied to
179
# filesystem access, so that in a later step, we can extricate them to
180
# a separarte ("storage") class.
184
_inventory_weave = None
81
def __init__(self, base, init=False, find_root=True, lock_mode='w'):
186
# Map some sort of prefix into a namespace
187
# stuff like "revno:10", "revid:", etc.
188
# This should match a prefix with a function which accepts
189
REVISION_NAMESPACES = {}
191
def push_stores(self, branch_to):
192
"""Copy the content of this branches store to branch_to."""
193
if (self._branch_format != branch_to._branch_format
194
or self._branch_format != 4):
195
from bzrlib.fetch import greedy_fetch
196
mutter("falling back to fetch logic to push between %s(%s) and %s(%s)",
197
self, self._branch_format, branch_to, branch_to._branch_format)
198
greedy_fetch(to_branch=branch_to, from_branch=self,
199
revision=self.last_revision())
202
store_pairs = ((self.text_store, branch_to.text_store),
203
(self.inventory_store, branch_to.inventory_store),
204
(self.revision_store, branch_to.revision_store))
206
for from_store, to_store in store_pairs:
207
copy_all(from_store, to_store)
208
except UnlistableStore:
209
raise UnlistableBranch(from_store)
211
def __init__(self, transport, init=False,
212
relax_version_check=False):
82
213
"""Create new branch object at a particular location.
84
base -- Base directory for the branch.
215
transport -- A Transport object, defining how to access files.
216
(If a string, transport.transport() will be used to
217
create a Transport object)
86
219
init -- If True, create new control files in a previously
87
220
unversioned directory. If False, the branch must already
90
find_root -- If true and init is false, find the root of the
91
existing branch containing base.
223
relax_version_check -- If true, the usual check for the branch
224
version is not applied. This is intended only for
225
upgrade/recovery type use; it's not guaranteed that
226
all operations will work on old format branches.
93
228
In the test suite, creation of new trees is tested using the
94
229
`ScratchBranch` class.
231
assert isinstance(transport, Transport), \
232
"%r is not a Transport" % transport
233
self._transport = transport
97
self.base = os.path.realpath(base)
98
235
self._make_control()
100
self.base = find_branch_root(base)
102
self.base = os.path.realpath(base)
103
if not isdir(self.controlfilename('.')):
104
bailout("not a bzr branch: %s" % quotefn(base),
105
['use "bzr init" to initialize a new working tree',
106
'current bzr can only operate from top-of-tree'])
110
self.text_store = ImmutableStore(self.controlfilename('text-store'))
111
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
112
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
236
self._check_format(relax_version_check)
238
def get_store(name, compressed=True):
239
relpath = self._rel_controlfilename(name)
241
store = CompressedTextStore(self._transport.clone(relpath))
243
store = TextStore(self._transport.clone(relpath))
244
if self._transport.should_cache():
245
from meta_store import CachedStore
246
cache_path = os.path.join(self.cache_root, name)
248
store = CachedStore(store, cache_path)
251
relpath = self._rel_controlfilename(name)
252
ws = WeaveStore(self._transport.clone(relpath))
253
if self._transport.should_cache():
254
ws.enable_cache = True
257
if self._branch_format == 4:
258
self.inventory_store = get_store('inventory-store')
259
self.text_store = get_store('text-store')
260
self.revision_store = get_store('revision-store')
261
elif self._branch_format == 5:
262
self.control_weaves = get_weave([])
263
self.weave_store = get_weave('weaves')
264
self.revision_store = get_store('revision-store', compressed=False)
115
266
def __str__(self):
116
return '%s(%r)' % (self.__class__.__name__, self.base)
267
return '%s(%r)' % (self.__class__.__name__, self._transport.base)
119
270
__repr__ = __str__
123
def lock(self, mode='w'):
124
"""Lock the on-disk branch, excluding other processes."""
130
om = os.O_WRONLY | os.O_CREAT
135
raise BzrError("invalid locking mode %r" % mode)
274
if self._lock_mode or self._lock:
275
# XXX: This should show something every time, and be suitable for
276
# headless operation and embedding
277
warn("branch %r was not explicitly unlocked" % self)
280
# TODO: It might be best to do this somewhere else,
281
# but it is nice for a Branch object to automatically
282
# cache it's information.
283
# Alternatively, we could have the Transport objects cache requests
284
# See the earlier discussion about how major objects (like Branch)
285
# should never expect their __del__ function to run.
286
if hasattr(self, 'cache_root') and self.cache_root is not None:
138
lockfile = os.open(self.controlfilename('branch-lock'), om)
140
if e.errno == errno.ENOENT:
141
# might not exist on branches from <0.0.4
142
self.controlfile('branch-lock', 'w').close()
143
lockfile = os.open(self.controlfilename('branch-lock'), om)
147
fcntl.lockf(lockfile, lm)
149
fcntl.lockf(lockfile, fcntl.LOCK_UN)
151
self._lockmode = None
153
self._lockmode = mode
155
warning("please write a locking method for platform %r" % sys.platform)
157
self._lockmode = None
159
self._lockmode = mode
162
def _need_readlock(self):
163
if self._lockmode not in ['r', 'w']:
164
raise BzrError('need read lock on branch, only have %r' % self._lockmode)
166
def _need_writelock(self):
167
if self._lockmode not in ['w']:
168
raise BzrError('need write lock on branch, only have %r' % self._lockmode)
289
shutil.rmtree(self.cache_root)
292
self.cache_root = None
296
return self._transport.base
299
base = property(_get_base)
302
def lock_write(self):
303
# TODO: Upgrade locking to support using a Transport,
304
# and potentially a remote locking protocol
306
if self._lock_mode != 'w':
307
raise LockError("can't upgrade to a write lock from %r" %
309
self._lock_count += 1
311
self._lock = self._transport.lock_write(
312
self._rel_controlfilename('branch-lock'))
313
self._lock_mode = 'w'
319
assert self._lock_mode in ('r', 'w'), \
320
"invalid lock mode %r" % self._lock_mode
321
self._lock_count += 1
323
self._lock = self._transport.lock_read(
324
self._rel_controlfilename('branch-lock'))
325
self._lock_mode = 'r'
329
if not self._lock_mode:
330
raise LockError('branch %r is not locked' % (self))
332
if self._lock_count > 1:
333
self._lock_count -= 1
337
self._lock_mode = self._lock_count = None
171
339
def abspath(self, name):
172
340
"""Return absolute filename for something in the branch"""
173
return os.path.join(self.base, name)
341
return self._transport.abspath(name)
176
343
def relpath(self, path):
177
344
"""Return path relative to this branch of something inside it.
179
346
Raises an error if path is not in this branch."""
180
rp = os.path.realpath(path)
182
if not rp.startswith(self.base):
183
bailout("path %r is not within branch %r" % (rp, self.base))
184
rp = rp[len(self.base):]
185
rp = rp.lstrip(os.sep)
347
return self._transport.relpath(path)
350
def _rel_controlfilename(self, file_or_path):
351
if isinstance(file_or_path, basestring):
352
file_or_path = [file_or_path]
353
return [bzrlib.BZRDIR] + file_or_path
189
355
def controlfilename(self, file_or_path):
190
356
"""Return location relative to branch."""
191
if isinstance(file_or_path, types.StringTypes):
192
file_or_path = [file_or_path]
193
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
357
return self._transport.abspath(self._rel_controlfilename(file_or_path))
196
360
def controlfile(self, file_or_path, mode='r'):
200
364
and binary. binary files are untranslated byte streams. Text
201
365
control files are stored with Unix newlines and in UTF-8, even
202
366
if the platform or locale defaults are different.
368
Controlfiles should almost never be opened in write mode but
369
rather should be atomically copied and replaced using atomicfile.
205
fn = self.controlfilename(file_or_path)
207
if mode == 'rb' or mode == 'wb':
208
return file(fn, mode)
209
elif mode == 'r' or mode == 'w':
210
# open in binary mode anyhow so there's no newline translation;
211
# codecs uses line buffering by default; don't want that.
213
return codecs.open(fn, mode + 'b', 'utf-8',
373
relpath = self._rel_controlfilename(file_or_path)
374
#TODO: codecs.open() buffers linewise, so it was overloaded with
375
# a much larger buffer, do we need to do the same for getreader/getwriter?
377
return self._transport.get(relpath)
379
raise BzrError("Branch.controlfile(mode='wb') is not supported, use put_controlfiles")
381
return codecs.getreader('utf-8')(self._transport.get(relpath), errors='replace')
383
raise BzrError("Branch.controlfile(mode='w') is not supported, use put_controlfiles")
216
385
raise BzrError("invalid controlfile mode %r" % mode)
387
def put_controlfile(self, path, f, encode=True):
388
"""Write an entry as a controlfile.
390
:param path: The path to put the file, relative to the .bzr control
392
:param f: A file-like or string object whose contents should be copied.
393
:param encode: If true, encode the contents as utf-8
395
self.put_controlfiles([(path, f)], encode=encode)
397
def put_controlfiles(self, files, encode=True):
398
"""Write several entries as controlfiles.
400
:param files: A list of [(path, file)] pairs, where the path is the directory
401
underneath the bzr control directory
402
:param encode: If true, encode the contents as utf-8
406
for path, f in files:
408
if isinstance(f, basestring):
409
f = f.encode('utf-8', 'replace')
411
f = codecs.getwriter('utf-8')(f, errors='replace')
412
path = self._rel_controlfilename(path)
413
ctrl_files.append((path, f))
414
self._transport.put_multi(ctrl_files)
220
416
def _make_control(self):
221
os.mkdir(self.controlfilename([]))
222
self.controlfile('README', 'w').write(
417
from bzrlib.inventory import Inventory
418
from bzrlib.weavefile import write_weave_v5
419
from bzrlib.weave import Weave
421
# Create an empty inventory
423
# if we want per-tree root ids then this is the place to set
424
# them; they're not needed for now and so ommitted for
426
bzrlib.xml5.serializer_v5.write_inventory(Inventory(), sio)
427
empty_inv = sio.getvalue()
429
bzrlib.weavefile.write_weave_v5(Weave(), sio)
430
empty_weave = sio.getvalue()
432
dirs = [[], 'revision-store', 'weaves']
223
434
"This is a Bazaar-NG control directory.\n"
224
"Do not change any files in this directory.")
225
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
226
for d in ('text-store', 'inventory-store', 'revision-store'):
227
os.mkdir(self.controlfilename(d))
228
for f in ('revision-history', 'merged-patches',
229
'pending-merged-patches', 'branch-name',
231
self.controlfile(f, 'w').write('')
232
mutter('created control directory in ' + self.base)
233
Inventory().write_xml(self.controlfile('inventory','w'))
236
def _check_format(self):
435
"Do not change any files in this directory.\n"),
436
('branch-format', BZR_BRANCH_FORMAT_5),
437
('revision-history', ''),
440
('pending-merges', ''),
441
('inventory', empty_inv),
442
('inventory.weave', empty_weave),
443
('ancestry.weave', empty_weave)
445
cfn = self._rel_controlfilename
446
self._transport.mkdir_multi([cfn(d) for d in dirs])
447
self.put_controlfiles(files)
448
mutter('created control directory in ' + self._transport.base)
450
def _check_format(self, relax_version_check):
237
451
"""Check this branch format is supported.
239
The current tool only supports the current unstable format.
453
The format level is stored, as an integer, in
454
self._branch_format for code that needs to check it later.
241
456
In the future, we might need different in-memory Branch
242
457
classes to support downlevel branches. But not yet.
244
# This ignores newlines so that we can open branches created
245
# on Windows from Linux and so on. I think it might be better
246
# to always make all internal files in unix format.
247
fmt = self.controlfile('branch-format', 'r').read()
248
fmt.replace('\r\n', '')
249
if fmt != BZR_BRANCH_FORMAT:
250
bailout('sorry, branch format %r not supported' % fmt,
251
['use a different bzr version',
252
'or remove the .bzr directory and "bzr init" again'])
460
fmt = self.controlfile('branch-format', 'r').read()
462
raise NotBranchError(self.base)
464
if fmt == BZR_BRANCH_FORMAT_5:
465
self._branch_format = 5
466
elif fmt == BZR_BRANCH_FORMAT_4:
467
self._branch_format = 4
469
if (not relax_version_check
470
and self._branch_format != 5):
471
raise BzrError('sorry, branch format %r not supported' % fmt,
472
['use a different bzr version',
473
'or remove the .bzr directory'
474
' and "bzr init" again'])
476
def get_root_id(self):
477
"""Return the id of this branches root"""
478
inv = self.read_working_inventory()
479
return inv.root.file_id
481
def set_root_id(self, file_id):
482
inv = self.read_working_inventory()
483
orig_root_id = inv.root.file_id
484
del inv._byid[inv.root.file_id]
485
inv.root.file_id = file_id
486
inv._byid[inv.root.file_id] = inv.root
489
if entry.parent_id in (None, orig_root_id):
490
entry.parent_id = inv.root.file_id
491
self._write_inventory(inv)
255
493
def read_working_inventory(self):
256
494
"""Read the working inventory."""
257
self._need_readlock()
259
# ElementTree does its own conversion from UTF-8, so open in
261
inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))
262
mutter("loaded inventory of %d items in %f"
263
% (len(inv), time.time() - before))
497
# ElementTree does its own conversion from UTF-8, so open in
499
f = self.controlfile('inventory', 'rb')
500
return bzrlib.xml5.serializer_v5.read_inventory(f)
267
505
def _write_inventory(self, inv):
268
506
"""Update the working inventory.
270
508
That is to say, the inventory describing changes underway, that
271
509
will be committed to the next revision.
273
self._need_writelock()
274
## TODO: factor out to atomicfile? is rename safe on windows?
275
## TODO: Maybe some kind of clean/dirty marker on inventory?
276
tmpfname = self.controlfilename('inventory.tmp')
277
tmpf = file(tmpfname, 'wb')
280
inv_fname = self.controlfilename('inventory')
281
if sys.platform == 'win32':
283
os.rename(tmpfname, inv_fname)
511
from cStringIO import StringIO
515
bzrlib.xml5.serializer_v5.write_inventory(inv, sio)
517
# Transport handles atomicity
518
self.put_controlfile('inventory', sio)
284
522
mutter('wrote working inventory')
287
524
inventory = property(read_working_inventory, _write_inventory, None,
288
525
"""Inventory for the working copy.""")
291
def add(self, files, verbose=False):
527
def add(self, files, ids=None):
292
528
"""Make files versioned.
294
Note that the command line normally calls smart_add instead.
530
Note that the command line normally calls smart_add instead,
531
which can automatically recurse.
296
533
This puts the files in the Added state, so that they will be
297
534
recorded by the next commit.
537
List of paths to add, relative to the base of the tree.
540
If set, use these instead of automatically generated ids.
541
Must be the same length as the list of files, but may
542
contain None for ids that are to be autogenerated.
299
544
TODO: Perhaps have an option to add the ids even if the files do
302
TODO: Perhaps return the ids of the files? But then again it
303
is easy to retrieve them if they're needed.
305
TODO: Option to specify file id.
307
TODO: Adding a directory should optionally recurse down and
308
add all non-ignored children. Perhaps do that in a
311
>>> b = ScratchBranch(files=['foo'])
312
>>> 'foo' in b.unknowns()
317
>>> 'foo' in b.unknowns()
319
>>> bool(b.inventory.path2id('foo'))
325
Traceback (most recent call last):
327
BzrError: ('foo is already versioned', [])
329
>>> b.add(['nothere'])
330
Traceback (most recent call last):
331
BzrError: ('cannot add: not a regular file or directory: nothere', [])
547
TODO: Perhaps yield the ids and paths as they're added.
333
self._need_writelock()
335
549
# TODO: Re-adding a file that is removed in the working copy
336
550
# should probably put it back with the previous ID.
337
if isinstance(files, types.StringTypes):
551
if isinstance(files, basestring):
552
assert(ids is None or isinstance(ids, basestring))
340
inv = self.read_working_inventory()
342
if is_control_file(f):
343
bailout("cannot add control file %s" % quotefn(f))
348
bailout("cannot add top-level %r" % f)
350
fullpath = os.path.normpath(self.abspath(f))
353
kind = file_kind(fullpath)
355
# maybe something better?
356
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
358
if kind != 'file' and kind != 'directory':
359
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
361
file_id = gen_file_id(f)
362
inv.add_path(f, kind=kind, file_id=file_id)
365
show_status('A', kind, quotefn(f))
367
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
369
self._write_inventory(inv)
558
ids = [None] * len(files)
560
assert(len(ids) == len(files))
564
inv = self.read_working_inventory()
565
for f,file_id in zip(files, ids):
566
if is_control_file(f):
567
raise BzrError("cannot add control file %s" % quotefn(f))
572
raise BzrError("cannot add top-level %r" % f)
574
fullpath = os.path.normpath(self.abspath(f))
577
kind = file_kind(fullpath)
579
# maybe something better?
580
raise BzrError('cannot add: not a regular file, symlink or directory: %s' % quotefn(f))
582
if not InventoryEntry.versionable_kind(kind):
583
raise BzrError('cannot add: not a versionable file ('
584
'i.e. regular file, symlink or directory): %s' % quotefn(f))
587
file_id = gen_file_id(f)
588
inv.add_path(f, kind=kind, file_id=file_id)
590
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
592
self._write_inventory(inv)
372
597
def print_file(self, file, revno):
373
598
"""Print `file` to stdout."""
374
self._need_readlock()
375
tree = self.revision_tree(self.lookup_revision(revno))
376
# use inventory as it was in that revision
377
file_id = tree.inventory.path2id(file)
379
bailout("%r is not present in revision %d" % (file, revno))
380
tree.print_file(file_id)
601
tree = self.revision_tree(self.get_rev_id(revno))
602
# use inventory as it was in that revision
603
file_id = tree.inventory.path2id(file)
605
raise BzrError("%r is not present in revision %s" % (file, revno))
606
tree.print_file(file_id)
383
611
def remove(self, files, verbose=False):
384
612
"""Mark nominated files for removal from the inventory.
463
682
return self.working_tree().unknowns()
466
def commit(self, message, timestamp=None, timezone=None,
469
"""Commit working copy as a new revision.
471
The basic approach is to add all the file texts into the
472
store, then the inventory, then make a new revision pointing
473
to that inventory and store that.
475
This is not quite safe if the working copy changes during the
476
commit; for the moment that is simply not allowed. A better
477
approach is to make a temporary copy of the files before
478
computing their hashes, and then add those hashes in turn to
479
the inventory. This should mean at least that there are no
480
broken hash pointers. There is no way we can get a snapshot
481
of the whole directory at an instant. This would also have to
482
be robust against files disappearing, moving, etc. So the
483
whole thing is a bit hard.
485
timestamp -- if not None, seconds-since-epoch for a
486
postdated/predated commit.
488
self._need_writelock()
490
## TODO: Show branch names
492
# TODO: Don't commit if there are no changes, unless forced?
494
# First walk over the working inventory; and both update that
495
# and also build a new revision inventory. The revision
496
# inventory needs to hold the text-id, sha1 and size of the
497
# actual file versions committed in the revision. (These are
498
# not present in the working inventory.) We also need to
499
# detect missing/deleted files, and remove them from the
502
work_inv = self.read_working_inventory()
504
basis = self.basis_tree()
505
basis_inv = basis.inventory
507
for path, entry in work_inv.iter_entries():
508
## TODO: Cope with files that have gone missing.
510
## TODO: Check that the file kind has not changed from the previous
511
## revision of this file (if any).
515
p = self.abspath(path)
516
file_id = entry.file_id
517
mutter('commit prep file %s, id %r ' % (p, file_id))
519
if not os.path.exists(p):
520
mutter(" file is missing, removing from inventory")
522
show_status('D', entry.kind, quotefn(path))
523
missing_ids.append(file_id)
526
# TODO: Handle files that have been deleted
528
# TODO: Maybe a special case for empty files? Seems a
529
# waste to store them many times.
533
if basis_inv.has_id(file_id):
534
old_kind = basis_inv[file_id].kind
535
if old_kind != entry.kind:
536
bailout("entry %r changed kind from %r to %r"
537
% (file_id, old_kind, entry.kind))
539
if entry.kind == 'directory':
541
bailout("%s is entered as directory but not a directory" % quotefn(p))
542
elif entry.kind == 'file':
544
bailout("%s is entered as file but is not a file" % quotefn(p))
546
content = file(p, 'rb').read()
548
entry.text_sha1 = sha_string(content)
549
entry.text_size = len(content)
551
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
553
and (old_ie.text_size == entry.text_size)
554
and (old_ie.text_sha1 == entry.text_sha1)):
555
## assert content == basis.get_file(file_id).read()
556
entry.text_id = basis_inv[file_id].text_id
557
mutter(' unchanged from previous text_id {%s}' %
561
entry.text_id = gen_file_id(entry.name)
562
self.text_store.add(content, entry.text_id)
563
mutter(' stored with text_id {%s}' % entry.text_id)
567
elif (old_ie.name == entry.name
568
and old_ie.parent_id == entry.parent_id):
573
show_status(state, entry.kind, quotefn(path))
575
for file_id in missing_ids:
576
# have to do this later so we don't mess up the iterator.
577
# since parents may be removed before their children we
580
# FIXME: There's probably a better way to do this; perhaps
581
# the workingtree should know how to filter itself.
582
if work_inv.has_id(file_id):
583
del work_inv[file_id]
586
inv_id = rev_id = _gen_revision_id(time.time())
588
inv_tmp = tempfile.TemporaryFile()
589
inv.write_xml(inv_tmp)
591
self.inventory_store.add(inv_tmp, inv_id)
592
mutter('new inventory_id is {%s}' % inv_id)
594
self._write_inventory(work_inv)
596
if timestamp == None:
597
timestamp = time.time()
599
if committer == None:
600
committer = username()
603
timezone = local_time_offset()
605
mutter("building commit log message")
606
rev = Revision(timestamp=timestamp,
609
precursor = self.last_patch(),
614
rev_tmp = tempfile.TemporaryFile()
615
rev.write_xml(rev_tmp)
617
self.revision_store.add(rev_tmp, rev_id)
618
mutter("new revision_id is {%s}" % rev_id)
620
## XXX: Everything up to here can simply be orphaned if we abort
621
## the commit; it will leave junk files behind but that doesn't
624
## TODO: Read back the just-generated changeset, and make sure it
625
## applies and recreates the right state.
627
## TODO: Also calculate and store the inventory SHA1
628
mutter("committing patch r%d" % (self.revno() + 1))
631
self.append_revision(rev_id)
634
note("commited r%d" % self.revno())
637
def append_revision(self, revision_id):
638
mutter("add {%s} to revision-history" % revision_id)
639
rev_history = self.revision_history()
641
tmprhname = self.controlfilename('revision-history.tmp')
642
rhname = self.controlfilename('revision-history')
644
f = file(tmprhname, 'wt')
645
rev_history.append(revision_id)
646
f.write('\n'.join(rev_history))
650
if sys.platform == 'win32':
652
os.rename(tmprhname, rhname)
685
def append_revision(self, *revision_ids):
686
for revision_id in revision_ids:
687
mutter("add {%s} to revision-history" % revision_id)
690
rev_history = self.revision_history()
691
rev_history.extend(revision_ids)
692
self.put_controlfile('revision-history', '\n'.join(rev_history))
696
def has_revision(self, revision_id):
697
"""True if this branch has a copy of the revision.
699
This does not necessarily imply the revision is merge
700
or on the mainline."""
701
return (revision_id is None
702
or revision_id in self.revision_store)
704
def get_revision_xml_file(self, revision_id):
705
"""Return XML file object for revision object."""
706
if not revision_id or not isinstance(revision_id, basestring):
707
raise InvalidRevisionId(revision_id)
712
return self.revision_store[revision_id]
713
except (IndexError, KeyError):
714
raise bzrlib.errors.NoSuchRevision(self, revision_id)
719
get_revision_xml = get_revision_xml_file
721
def get_revision_xml(self, revision_id):
722
return self.get_revision_xml_file(revision_id).read()
656
725
def get_revision(self, revision_id):
657
726
"""Return the Revision object for a named revision"""
658
self._need_readlock()
659
r = Revision.read_xml(self.revision_store[revision_id])
727
xml_file = self.get_revision_xml_file(revision_id)
730
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
731
except SyntaxError, e:
732
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
660
736
assert r.revision_id == revision_id
664
def get_inventory(self, inventory_id):
665
"""Get Inventory object by hash.
667
TODO: Perhaps for this and similar methods, take a revision
668
parameter which can be either an integer revno or a
670
self._need_readlock()
671
i = Inventory.read_xml(self.inventory_store[inventory_id])
739
def get_revision_delta(self, revno):
740
"""Return the delta for one revision.
742
The delta is relative to its mainline predecessor, or the
743
empty tree for revision 1.
745
assert isinstance(revno, int)
746
rh = self.revision_history()
747
if not (1 <= revno <= len(rh)):
748
raise InvalidRevisionNumber(revno)
750
# revno is 1-based; list is 0-based
752
new_tree = self.revision_tree(rh[revno-1])
754
old_tree = EmptyTree()
756
old_tree = self.revision_tree(rh[revno-2])
758
return compare_trees(old_tree, new_tree)
760
def get_revision_sha1(self, revision_id):
761
"""Hash the stored value of a revision, and return it."""
762
# In the future, revision entries will be signed. At that
763
# point, it is probably best *not* to include the signature
764
# in the revision hash. Because that lets you re-sign
765
# the revision, (add signatures/remove signatures) and still
766
# have all hash pointers stay consistent.
767
# But for now, just hash the contents.
768
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
770
def _get_ancestry_weave(self):
771
return self.control_weaves.get_weave('ancestry')
773
def get_ancestry(self, revision_id):
774
"""Return a list of revision-ids integrated by a revision.
777
if revision_id is None:
779
w = self._get_ancestry_weave()
780
return [None] + [l[:-1] for l in w.get_iter(w.lookup(revision_id))]
782
def get_inventory_weave(self):
783
return self.control_weaves.get_weave('inventory')
785
def get_inventory(self, revision_id):
786
"""Get Inventory object by hash."""
787
xml = self.get_inventory_xml(revision_id)
788
return bzrlib.xml5.serializer_v5.read_inventory_from_string(xml)
790
def get_inventory_xml(self, revision_id):
791
"""Get inventory XML as a file object."""
793
assert isinstance(revision_id, basestring), type(revision_id)
794
iw = self.get_inventory_weave()
795
return iw.get_text(iw.lookup(revision_id))
797
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
799
def get_inventory_sha1(self, revision_id):
800
"""Return the sha1 hash of the inventory entry
802
return self.get_revision(revision_id).inventory_sha1
675
804
def get_revision_inventory(self, revision_id):
676
805
"""Return inventory of a past revision."""
677
self._need_readlock()
806
# TODO: Unify this with get_inventory()
807
# bzr 0.0.6 and later imposes the constraint that the inventory_id
808
# must be the same as its revision, so this is trivial.
678
809
if revision_id == None:
810
return Inventory(self.get_root_id())
681
return self.get_inventory(self.get_revision(revision_id).inventory_id)
812
return self.get_inventory(revision_id)
684
814
def revision_history(self):
685
"""Return sequence of revision hashes on to this branch.
687
>>> ScratchBranch().revision_history()
690
self._need_readlock()
691
return [l.rstrip('\r\n') for l in self.controlfile('revision-history', 'r').readlines()]
694
def enum_history(self, direction):
695
"""Return (revno, revision_id) for history of branch.
698
'forward' is from earliest to latest
699
'reverse' is from latest to earliest
701
rh = self.revision_history()
702
if direction == 'forward':
707
elif direction == 'reverse':
713
raise BzrError('invalid history direction %r' % direction)
815
"""Return sequence of revision hashes on to this branch."""
818
return [l.rstrip('\r\n') for l in
819
self.controlfile('revision-history', 'r').readlines()]
823
def common_ancestor(self, other, self_revno=None, other_revno=None):
825
>>> from bzrlib.commit import commit
826
>>> sb = ScratchBranch(files=['foo', 'foo~'])
827
>>> sb.common_ancestor(sb) == (None, None)
829
>>> commit(sb, "Committing first revision", verbose=False)
830
>>> sb.common_ancestor(sb)[0]
832
>>> clone = sb.clone()
833
>>> commit(sb, "Committing second revision", verbose=False)
834
>>> sb.common_ancestor(sb)[0]
836
>>> sb.common_ancestor(clone)[0]
838
>>> commit(clone, "Committing divergent second revision",
840
>>> sb.common_ancestor(clone)[0]
842
>>> sb.common_ancestor(clone) == clone.common_ancestor(sb)
844
>>> sb.common_ancestor(sb) != clone.common_ancestor(clone)
846
>>> clone2 = sb.clone()
847
>>> sb.common_ancestor(clone2)[0]
849
>>> sb.common_ancestor(clone2, self_revno=1)[0]
851
>>> sb.common_ancestor(clone2, other_revno=1)[0]
854
my_history = self.revision_history()
855
other_history = other.revision_history()
856
if self_revno is None:
857
self_revno = len(my_history)
858
if other_revno is None:
859
other_revno = len(other_history)
860
indices = range(min((self_revno, other_revno)))
863
if my_history[r] == other_history[r]:
864
return r+1, my_history[r]
719
871
That is equivalent to the number of revisions committed to
722
>>> b = ScratchBranch()
725
>>> b.commit('no foo')
729
874
return len(self.revision_history())
732
def last_patch(self):
877
def last_revision(self):
733
878
"""Return last patch hash, or None if no history.
735
>>> ScratchBranch().last_patch() == None
738
880
ph = self.revision_history()
887
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
888
"""Return a list of new revisions that would perfectly fit.
745
def lookup_revision(self, revno):
746
"""Return revision hash for revision number."""
890
If self and other have not diverged, return a list of the revisions
891
present in other, but missing from self.
893
>>> from bzrlib.commit import commit
894
>>> bzrlib.trace.silent = True
895
>>> br1 = ScratchBranch()
896
>>> br2 = ScratchBranch()
897
>>> br1.missing_revisions(br2)
899
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
900
>>> br1.missing_revisions(br2)
902
>>> br2.missing_revisions(br1)
904
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
905
>>> br1.missing_revisions(br2)
907
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
908
>>> br1.missing_revisions(br2)
910
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
911
>>> br1.missing_revisions(br2)
912
Traceback (most recent call last):
913
DivergedBranches: These branches have diverged.
915
# FIXME: If the branches have diverged, but the latest
916
# revision in this branch is completely merged into the other,
917
# then we should still be able to pull.
918
self_history = self.revision_history()
919
self_len = len(self_history)
920
other_history = other.revision_history()
921
other_len = len(other_history)
922
common_index = min(self_len, other_len) -1
923
if common_index >= 0 and \
924
self_history[common_index] != other_history[common_index]:
925
raise DivergedBranches(self, other)
927
if stop_revision is None:
928
stop_revision = other_len
930
assert isinstance(stop_revision, int)
931
if stop_revision > other_len:
932
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
933
return other_history[self_len:stop_revision]
935
def update_revisions(self, other, stop_revision=None):
936
"""Pull in new perfect-fit revisions."""
937
from bzrlib.fetch import greedy_fetch
938
from bzrlib.revision import get_intervening_revisions
939
if stop_revision is None:
940
stop_revision = other.last_revision()
941
greedy_fetch(to_branch=self, from_branch=other,
942
revision=stop_revision)
943
pullable_revs = self.missing_revisions(
944
other, other.revision_id_to_revno(stop_revision))
946
greedy_fetch(to_branch=self,
948
revision=pullable_revs[-1])
949
self.append_revision(*pullable_revs)
952
def commit(self, *args, **kw):
953
from bzrlib.commit import Commit
954
Commit().commit(self, *args, **kw)
956
def revision_id_to_revno(self, revision_id):
957
"""Given a revision id, return its revno"""
958
if revision_id is None:
960
history = self.revision_history()
962
return history.index(revision_id) + 1
964
raise bzrlib.errors.NoSuchRevision(self, revision_id)
966
def get_rev_id(self, revno, history=None):
967
"""Find the revision id of the specified revno."""
751
# list is 0-based; revisions are 1-based
752
return self.revision_history()[revno-1]
754
raise BzrError("no such revision %s" % revno)
971
history = self.revision_history()
972
elif revno <= 0 or revno > len(history):
973
raise bzrlib.errors.NoSuchRevision(self, revno)
974
return history[revno - 1]
757
976
def revision_tree(self, revision_id):
758
977
"""Return Tree for a revision on this branch.
760
979
`revision_id` may be None for the null revision, in which case
761
980
an `EmptyTree` is returned."""
762
self._need_readlock()
981
# TODO: refactor this to use an existing revision object
982
# so we don't need to read it in twice.
763
983
if revision_id == None:
764
984
return EmptyTree()
766
986
inv = self.get_revision_inventory(revision_id)
767
return RevisionTree(self.text_store, inv)
987
return RevisionTree(self.weave_store, inv, revision_id)
770
990
def working_tree(self):
771
991
"""Return a `Tree` for the working copy."""
772
return WorkingTree(self.base, self.read_working_inventory())
992
from bzrlib.workingtree import WorkingTree
993
# TODO: In the future, WorkingTree should utilize Transport
994
# RobertCollins 20051003 - I don't think it should - working trees are
995
# much more complex to keep consistent than our careful .bzr subset.
996
# instead, we should say that working trees are local only, and optimise
998
return WorkingTree(self._transport.base, self.read_working_inventory())
775
1001
def basis_tree(self):
776
1002
"""Return `Tree` object for last revision.
778
1004
If there are no revisions yet, return an `EmptyTree`.
780
>>> b = ScratchBranch(files=['foo'])
781
>>> b.basis_tree().has_filename('foo')
783
>>> b.working_tree().has_filename('foo')
786
>>> b.commit('add foo')
787
>>> b.basis_tree().has_filename('foo')
790
r = self.last_patch()
794
return RevisionTree(self.text_store, self.get_revision_inventory(r))
1006
return self.revision_tree(self.last_revision())
798
1009
def rename_one(self, from_rel, to_rel):
855
1066
Note that to_name is only the last component of the new name;
856
1067
this doesn't change the directory.
858
self._need_writelock()
859
## TODO: Option to move IDs only
860
assert not isinstance(from_paths, basestring)
861
tree = self.working_tree()
863
to_abs = self.abspath(to_name)
864
if not isdir(to_abs):
865
bailout("destination %r is not a directory" % to_abs)
866
if not tree.has_filename(to_name):
867
bailout("destination %r not in working directory" % to_abs)
868
to_dir_id = inv.path2id(to_name)
869
if to_dir_id == None and to_name != '':
870
bailout("destination %r is not a versioned directory" % to_name)
871
to_dir_ie = inv[to_dir_id]
872
if to_dir_ie.kind not in ('directory', 'root_directory'):
873
bailout("destination %r is not a directory" % to_abs)
875
to_idpath = Set(inv.get_idpath(to_dir_id))
878
if not tree.has_filename(f):
879
bailout("%r does not exist in working tree" % f)
880
f_id = inv.path2id(f)
882
bailout("%r is not versioned" % f)
883
name_tail = splitpath(f)[-1]
884
dest_path = appendpath(to_name, name_tail)
885
if tree.has_filename(dest_path):
886
bailout("destination %r already exists" % dest_path)
887
if f_id in to_idpath:
888
bailout("can't move %r to a subdirectory of itself" % f)
890
# OK, so there's a race here, it's possible that someone will
891
# create a file in this interval and then the rename might be
892
# left half-done. But we should have caught most problems.
895
name_tail = splitpath(f)[-1]
896
dest_path = appendpath(to_name, name_tail)
897
print "%s => %s" % (f, dest_path)
898
inv.rename(inv.path2id(f), to_dir_id, name_tail)
900
os.rename(self.abspath(f), self.abspath(dest_path))
902
bailout("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
903
["rename rolled back"])
905
self._write_inventory(inv)
909
def show_status(self, show_all=False, file_list=None):
910
"""Display single-line status for non-ignored working files.
912
The list is show sorted in order by file name.
914
>>> b = ScratchBranch(files=['foo', 'foo~'])
920
>>> b.commit("add foo")
922
>>> os.unlink(b.abspath('foo'))
926
self._need_readlock()
928
# We have to build everything into a list first so that it can
929
# sorted by name, incorporating all the different sources.
931
# FIXME: Rather than getting things in random order and then sorting,
932
# just step through in order.
934
# Interesting case: the old ID for a file has been removed,
935
# but a new file has been created under that name.
937
old = self.basis_tree()
938
new = self.working_tree()
940
items = diff_trees(old, new)
941
# We want to filter out only if any file was provided in the file_list.
942
if isinstance(file_list, list) and len(file_list):
943
items = [item for item in items if item[3] in file_list]
945
for fs, fid, oldname, newname, kind in items:
947
show_status(fs, kind,
948
oldname + ' => ' + newname)
949
elif fs == 'A' or fs == 'M':
950
show_status(fs, kind, newname)
952
show_status(fs, kind, oldname)
955
show_status(fs, kind, newname)
958
show_status(fs, kind, newname)
960
show_status(fs, kind, newname)
962
bailout("weird file state %r" % ((fs, fid),))
966
class ScratchBranch(Branch):
1069
This returns a list of (from_path, to_path) pairs for each
1070
entry that is moved.
1075
## TODO: Option to move IDs only
1076
assert not isinstance(from_paths, basestring)
1077
tree = self.working_tree()
1078
inv = tree.inventory
1079
to_abs = self.abspath(to_name)
1080
if not isdir(to_abs):
1081
raise BzrError("destination %r is not a directory" % to_abs)
1082
if not tree.has_filename(to_name):
1083
raise BzrError("destination %r not in working directory" % to_abs)
1084
to_dir_id = inv.path2id(to_name)
1085
if to_dir_id == None and to_name != '':
1086
raise BzrError("destination %r is not a versioned directory" % to_name)
1087
to_dir_ie = inv[to_dir_id]
1088
if to_dir_ie.kind not in ('directory', 'root_directory'):
1089
raise BzrError("destination %r is not a directory" % to_abs)
1091
to_idpath = inv.get_idpath(to_dir_id)
1093
for f in from_paths:
1094
if not tree.has_filename(f):
1095
raise BzrError("%r does not exist in working tree" % f)
1096
f_id = inv.path2id(f)
1098
raise BzrError("%r is not versioned" % f)
1099
name_tail = splitpath(f)[-1]
1100
dest_path = appendpath(to_name, name_tail)
1101
if tree.has_filename(dest_path):
1102
raise BzrError("destination %r already exists" % dest_path)
1103
if f_id in to_idpath:
1104
raise BzrError("can't move %r to a subdirectory of itself" % f)
1106
# OK, so there's a race here, it's possible that someone will
1107
# create a file in this interval and then the rename might be
1108
# left half-done. But we should have caught most problems.
1110
for f in from_paths:
1111
name_tail = splitpath(f)[-1]
1112
dest_path = appendpath(to_name, name_tail)
1113
result.append((f, dest_path))
1114
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1116
rename(self.abspath(f), self.abspath(dest_path))
1118
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1119
["rename rolled back"])
1121
self._write_inventory(inv)
1128
def revert(self, filenames, old_tree=None, backups=True):
1129
"""Restore selected files to the versions from a previous tree.
1132
If true (default) backups are made of files before
1135
from bzrlib.errors import NotVersionedError, BzrError
1136
from bzrlib.atomicfile import AtomicFile
1137
from bzrlib.osutils import backup_file
1139
inv = self.read_working_inventory()
1140
if old_tree is None:
1141
old_tree = self.basis_tree()
1142
old_inv = old_tree.inventory
1145
for fn in filenames:
1146
file_id = inv.path2id(fn)
1148
raise NotVersionedError("not a versioned file", fn)
1149
if not old_inv.has_id(file_id):
1150
raise BzrError("file not present in old tree", fn, file_id)
1151
nids.append((fn, file_id))
1153
# TODO: Rename back if it was previously at a different location
1155
# TODO: If given a directory, restore the entire contents from
1156
# the previous version.
1158
# TODO: Make a backup to a temporary file.
1160
# TODO: If the file previously didn't exist, delete it?
1161
for fn, file_id in nids:
1164
f = AtomicFile(fn, 'wb')
1166
f.write(old_tree.get_file(file_id).read())
1172
def pending_merges(self):
1173
"""Return a list of pending merges.
1175
These are revisions that have been merged into the working
1176
directory but not yet committed.
1178
cfn = self._rel_controlfilename('pending-merges')
1179
if not self._transport.has(cfn):
1182
for l in self.controlfile('pending-merges', 'r').readlines():
1183
p.append(l.rstrip('\n'))
1187
def add_pending_merge(self, *revision_ids):
1188
# TODO: Perhaps should check at this point that the
1189
# history of the revision is actually present?
1190
for rev_id in revision_ids:
1191
validate_revision_id(rev_id)
1193
p = self.pending_merges()
1195
for rev_id in revision_ids:
1201
self.set_pending_merges(p)
1203
def set_pending_merges(self, rev_list):
1206
self.put_controlfile('pending-merges', '\n'.join(rev_list))
1211
def get_parent(self):
1212
"""Return the parent location of the branch.
1214
This is the default location for push/pull/missing. The usual
1215
pattern is that the user can override it by specifying a
1219
_locs = ['parent', 'pull', 'x-pull']
1222
return self.controlfile(l, 'r').read().strip('\n')
1224
if e.errno != errno.ENOENT:
1229
def set_parent(self, url):
1230
# TODO: Maybe delete old location files?
1231
from bzrlib.atomicfile import AtomicFile
1234
f = AtomicFile(self.controlfilename('parent'))
1243
def check_revno(self, revno):
1245
Check whether a revno corresponds to any revision.
1246
Zero (the NULL revision) is considered valid.
1249
self.check_real_revno(revno)
1251
def check_real_revno(self, revno):
1253
Check whether a revno corresponds to a real revision.
1254
Zero (the NULL revision) is considered invalid
1256
if revno < 1 or revno > self.revno():
1257
raise InvalidRevisionNumber(revno)
1263
class ScratchBranch(_Branch):
967
1264
"""Special test class: a branch that cleans up after itself.
969
1266
>>> b = ScratchBranch()