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, chomp, \
30
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
31
joinpath, sha_string, file_kind, local_time_offset
32
from store import ImmutableStore
33
from revision import Revision
34
from errors import bailout
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
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,
32
from bzrlib.errors import (BzrError, InvalidRevisionNumber, InvalidRevisionId,
33
NoSuchRevision, HistoryMissing, NotBranchError,
34
DivergedBranches, LockError, UnlistableStore,
35
UnlistableBranch, NoSuchFile)
36
from bzrlib.textui import show_status
37
from bzrlib.revision import Revision
38
from bzrlib.delta import compare_trees
39
from bzrlib.tree import EmptyTree, RevisionTree
40
from bzrlib.inventory import Inventory
41
from bzrlib.store import copy_all
42
from bzrlib.store.compressed_text import CompressedTextStore
43
from bzrlib.store.text import TextStore
44
from bzrlib.store.weave import WeaveStore
45
from bzrlib.transport import Transport, get_transport
50
BZR_BRANCH_FORMAT_4 = "Bazaar-NG branch, format 0.0.4\n"
51
BZR_BRANCH_FORMAT_5 = "Bazaar-NG branch, format 5\n"
39
52
## TODO: Maybe include checks for common corruption of newlines, etc?
43
def find_branch_root(f=None):
44
"""Find the branch root enclosing f, or pwd.
46
It is not necessary that f exists.
55
# TODO: Some operations like log might retrieve the same revisions
56
# repeatedly to calculate deltas. We could perhaps have a weakref
57
# cache in memory to make this faster. In general anything can be
58
# cached in memory between lock and unlock operations.
60
def find_branch(*ignored, **ignored_too):
61
# XXX: leave this here for about one release, then remove it
62
raise NotImplementedError('find_branch() is not supported anymore, '
63
'please use one of the new branch constructors')
64
def _relpath(base, path):
65
"""Return path relative to base, or raise exception.
67
The path may be either an absolute path or a path relative to the
68
current working directory.
70
Lifted out of Branch.relpath for ease of testing.
72
os.path.commonprefix (python2.4) has a bad bug that it works just
73
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
74
avoids that problem."""
75
rp = os.path.abspath(path)
79
while len(head) >= len(base):
82
head, tail = os.path.split(head)
86
raise NotBranchError("path %r is not within branch %r" % (rp, base))
91
def find_branch_root(t):
92
"""Find the branch root enclosing the transport's base.
94
t is a Transport object.
96
It is not necessary that the base of t exists.
48
98
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)
99
run into the root. If there isn't one, raises NotBranchError.
61
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
63
head, tail = os.path.split(f)
103
if t.has(bzrlib.BZRDIR):
105
new_t = t.clone('..')
106
if new_t.base == t.base:
65
107
# reached the root, whatever that may be
66
bailout('%r is not in a branch' % orig_f)
108
raise NotBranchError('%s is not in a branch' % orig_base)
71
112
######################################################################
115
class Branch(object):
75
116
"""Branch holding a history of revisions.
77
:todo: Perhaps use different stores for different classes of object,
78
so that we can keep track of how much space each one uses,
79
or garbage-collect them.
81
:todo: Add a RemoteBranch subclass. For the basic case of read-only
82
HTTP access this should be very easy by,
83
just redirecting controlfile access into HTTP requests.
84
We would need a RemoteStore working similarly.
86
:todo: Keep the on-disk branch locked while the object exists.
88
:todo: mkdir() method.
90
def __init__(self, base, init=False, find_root=True):
119
Base directory/url of the branch.
123
def __init__(self, *ignored, **ignored_too):
124
raise NotImplementedError('The Branch class is abstract')
127
def open_downlevel(base):
128
"""Open a branch which may be of an old format.
130
Only local branches are supported."""
131
return _Branch(get_transport(base), relax_version_check=True)
135
"""Open an existing branch, rooted at 'base' (url)"""
136
t = get_transport(base)
137
mutter("trying to open %r with transport %r", base, t)
141
def open_containing(url):
142
"""Open an existing branch which contains url.
144
This probes for a branch at url, and searches upwards from there.
146
t = get_transport(url)
147
t = find_branch_root(t)
151
def initialize(base):
152
"""Create a new branch, rooted at 'base' (url)"""
153
t = get_transport(base)
154
return _Branch(t, init=True)
156
def setup_caching(self, cache_root):
157
"""Subclasses that care about caching should override this, and set
158
up cached stores located under cache_root.
160
self.cache_root = cache_root
163
class _Branch(Branch):
164
"""A branch stored in the actual filesystem.
166
Note that it's "local" in the context of the filesystem; it doesn't
167
really matter if it's on an nfs/smb/afs/coda/... share, as long as
168
it's writable, and can be accessed via the normal filesystem API.
174
If _lock_mode is true, a positive count of the number of times the
178
Lock object from bzrlib.lock.
180
# We actually expect this class to be somewhat short-lived; part of its
181
# purpose is to try to isolate what bits of the branch logic are tied to
182
# filesystem access, so that in a later step, we can extricate them to
183
# a separarte ("storage") class.
187
_inventory_weave = None
189
# Map some sort of prefix into a namespace
190
# stuff like "revno:10", "revid:", etc.
191
# This should match a prefix with a function which accepts
192
REVISION_NAMESPACES = {}
194
def push_stores(self, branch_to):
195
"""Copy the content of this branches store to branch_to."""
196
if (self._branch_format != branch_to._branch_format
197
or self._branch_format != 4):
198
from bzrlib.fetch import greedy_fetch
199
mutter("falling back to fetch logic to push between %s(%s) and %s(%s)",
200
self, self._branch_format, branch_to, branch_to._branch_format)
201
greedy_fetch(to_branch=branch_to, from_branch=self,
202
revision=self.last_revision())
205
store_pairs = ((self.text_store, branch_to.text_store),
206
(self.inventory_store, branch_to.inventory_store),
207
(self.revision_store, branch_to.revision_store))
209
for from_store, to_store in store_pairs:
210
copy_all(from_store, to_store)
211
except UnlistableStore:
212
raise UnlistableBranch(from_store)
214
def __init__(self, transport, init=False,
215
relax_version_check=False):
91
216
"""Create new branch object at a particular location.
93
:param base: Base directory for the branch.
218
transport -- A Transport object, defining how to access files.
219
(If a string, transport.transport() will be used to
220
create a Transport object)
95
:param init: If True, create new control files in a previously
222
init -- If True, create new control files in a previously
96
223
unversioned directory. If False, the branch must already
99
:param find_root: If true and init is false, find the root of the
100
existing branch containing base.
226
relax_version_check -- If true, the usual check for the branch
227
version is not applied. This is intended only for
228
upgrade/recovery type use; it's not guaranteed that
229
all operations will work on old format branches.
102
231
In the test suite, creation of new trees is tested using the
103
232
`ScratchBranch` class.
234
assert isinstance(transport, Transport), \
235
"%r is not a Transport" % transport
236
self._transport = transport
106
self.base = os.path.realpath(base)
107
238
self._make_control()
109
self.base = find_branch_root(base)
111
self.base = os.path.realpath(base)
112
if not isdir(self.controlfilename('.')):
113
bailout("not a bzr branch: %s" % quotefn(base),
114
['use "bzr init" to initialize a new working tree',
115
'current bzr can only operate from top-of-tree'])
118
self.text_store = ImmutableStore(self.controlfilename('text-store'))
119
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
120
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
239
self._check_format(relax_version_check)
241
def get_store(name, compressed=True):
242
# FIXME: This approach of assuming stores are all entirely compressed
243
# or entirely uncompressed is tidy, but breaks upgrade from
244
# some existing branches where there's a mixture; we probably
245
# still want the option to look for both.
246
relpath = self._rel_controlfilename(name)
248
store = CompressedTextStore(self._transport.clone(relpath))
250
store = TextStore(self._transport.clone(relpath))
251
#if self._transport.should_cache():
252
# cache_path = os.path.join(self.cache_root, name)
253
# os.mkdir(cache_path)
254
# store = bzrlib.store.CachedStore(store, cache_path)
257
relpath = self._rel_controlfilename(name)
258
ws = WeaveStore(self._transport.clone(relpath))
259
if self._transport.should_cache():
260
ws.enable_cache = True
263
if self._branch_format == 4:
264
self.inventory_store = get_store('inventory-store')
265
self.text_store = get_store('text-store')
266
self.revision_store = get_store('revision-store')
267
elif self._branch_format == 5:
268
self.control_weaves = get_weave([])
269
self.weave_store = get_weave('weaves')
270
self.revision_store = get_store('revision-store', compressed=False)
123
272
def __str__(self):
124
return '%s(%r)' % (self.__class__.__name__, self.base)
273
return '%s(%r)' % (self.__class__.__name__, self._transport.base)
127
276
__repr__ = __str__
280
if self._lock_mode or self._lock:
281
# XXX: This should show something every time, and be suitable for
282
# headless operation and embedding
283
warn("branch %r was not explicitly unlocked" % self)
286
# TODO: It might be best to do this somewhere else,
287
# but it is nice for a Branch object to automatically
288
# cache it's information.
289
# Alternatively, we could have the Transport objects cache requests
290
# See the earlier discussion about how major objects (like Branch)
291
# should never expect their __del__ function to run.
292
if hasattr(self, 'cache_root') and self.cache_root is not None:
295
shutil.rmtree(self.cache_root)
298
self.cache_root = None
302
return self._transport.base
305
base = property(_get_base)
308
def lock_write(self):
309
# TODO: Upgrade locking to support using a Transport,
310
# and potentially a remote locking protocol
312
if self._lock_mode != 'w':
313
raise LockError("can't upgrade to a write lock from %r" %
315
self._lock_count += 1
317
self._lock = self._transport.lock_write(
318
self._rel_controlfilename('branch-lock'))
319
self._lock_mode = 'w'
325
assert self._lock_mode in ('r', 'w'), \
326
"invalid lock mode %r" % self._lock_mode
327
self._lock_count += 1
329
self._lock = self._transport.lock_read(
330
self._rel_controlfilename('branch-lock'))
331
self._lock_mode = 'r'
335
if not self._lock_mode:
336
raise LockError('branch %r is not locked' % (self))
338
if self._lock_count > 1:
339
self._lock_count -= 1
343
self._lock_mode = self._lock_count = None
130
345
def abspath(self, name):
131
346
"""Return absolute filename for something in the branch"""
132
return os.path.join(self.base, name)
347
return self._transport.abspath(name)
135
349
def relpath(self, path):
136
350
"""Return path relative to this branch of something inside it.
138
352
Raises an error if path is not in this branch."""
139
rp = os.path.realpath(path)
141
if not rp.startswith(self.base):
142
bailout("path %r is not within branch %r" % (rp, self.base))
143
rp = rp[len(self.base):]
144
rp = rp.lstrip(os.sep)
353
return self._transport.relpath(path)
356
def _rel_controlfilename(self, file_or_path):
357
if isinstance(file_or_path, basestring):
358
file_or_path = [file_or_path]
359
return [bzrlib.BZRDIR] + file_or_path
148
361
def controlfilename(self, file_or_path):
149
362
"""Return location relative to branch."""
150
if isinstance(file_or_path, types.StringTypes):
151
file_or_path = [file_or_path]
152
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
363
return self._transport.abspath(self._rel_controlfilename(file_or_path))
155
366
def controlfile(self, file_or_path, mode='r'):
156
"""Open a control file for this branch"""
157
return file(self.controlfilename(file_or_path), mode)
367
"""Open a control file for this branch.
369
There are two classes of file in the control directory: text
370
and binary. binary files are untranslated byte streams. Text
371
control files are stored with Unix newlines and in UTF-8, even
372
if the platform or locale defaults are different.
374
Controlfiles should almost never be opened in write mode but
375
rather should be atomically copied and replaced using atomicfile.
379
relpath = self._rel_controlfilename(file_or_path)
380
#TODO: codecs.open() buffers linewise, so it was overloaded with
381
# a much larger buffer, do we need to do the same for getreader/getwriter?
383
return self._transport.get(relpath)
385
raise BzrError("Branch.controlfile(mode='wb') is not supported, use put_controlfiles")
387
return codecs.getreader('utf-8')(self._transport.get(relpath), errors='replace')
389
raise BzrError("Branch.controlfile(mode='w') is not supported, use put_controlfiles")
391
raise BzrError("invalid controlfile mode %r" % mode)
393
def put_controlfile(self, path, f, encode=True):
394
"""Write an entry as a controlfile.
396
:param path: The path to put the file, relative to the .bzr control
398
:param f: A file-like or string object whose contents should be copied.
399
:param encode: If true, encode the contents as utf-8
401
self.put_controlfiles([(path, f)], encode=encode)
403
def put_controlfiles(self, files, encode=True):
404
"""Write several entries as controlfiles.
406
:param files: A list of [(path, file)] pairs, where the path is the directory
407
underneath the bzr control directory
408
:param encode: If true, encode the contents as utf-8
412
for path, f in files:
414
if isinstance(f, basestring):
415
f = f.encode('utf-8', 'replace')
417
f = codecs.getwriter('utf-8')(f, errors='replace')
418
path = self._rel_controlfilename(path)
419
ctrl_files.append((path, f))
420
self._transport.put_multi(ctrl_files)
160
422
def _make_control(self):
161
os.mkdir(self.controlfilename([]))
162
self.controlfile('README', 'w').write(
423
from bzrlib.inventory import Inventory
424
from bzrlib.weavefile import write_weave_v5
425
from bzrlib.weave import Weave
427
# Create an empty inventory
429
# if we want per-tree root ids then this is the place to set
430
# them; they're not needed for now and so ommitted for
432
bzrlib.xml5.serializer_v5.write_inventory(Inventory(), sio)
433
empty_inv = sio.getvalue()
435
bzrlib.weavefile.write_weave_v5(Weave(), sio)
436
empty_weave = sio.getvalue()
438
dirs = [[], 'revision-store', 'weaves']
163
440
"This is a Bazaar-NG control directory.\n"
164
"Do not change any files in this directory.")
165
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
166
for d in ('text-store', 'inventory-store', 'revision-store'):
167
os.mkdir(self.controlfilename(d))
168
for f in ('revision-history', 'merged-patches',
169
'pending-merged-patches', 'branch-name'):
170
self.controlfile(f, 'w').write('')
171
mutter('created control directory in ' + self.base)
172
Inventory().write_xml(self.controlfile('inventory','w'))
175
def _check_format(self):
441
"Do not change any files in this directory.\n"),
442
('branch-format', BZR_BRANCH_FORMAT_5),
443
('revision-history', ''),
446
('pending-merges', ''),
447
('inventory', empty_inv),
448
('inventory.weave', empty_weave),
449
('ancestry.weave', empty_weave)
451
cfn = self._rel_controlfilename
452
self._transport.mkdir_multi([cfn(d) for d in dirs])
453
self.put_controlfiles(files)
454
mutter('created control directory in ' + self._transport.base)
456
def _check_format(self, relax_version_check):
176
457
"""Check this branch format is supported.
178
The current tool only supports the current unstable format.
459
The format level is stored, as an integer, in
460
self._branch_format for code that needs to check it later.
180
462
In the future, we might need different in-memory Branch
181
463
classes to support downlevel branches. But not yet.
183
# read in binary mode to detect newline wierdness.
184
fmt = self.controlfile('branch-format', 'rb').read()
185
if fmt != BZR_BRANCH_FORMAT:
186
bailout('sorry, branch format %r not supported' % fmt,
187
['use a different bzr version',
188
'or remove the .bzr directory and "bzr init" again'])
466
fmt = self.controlfile('branch-format', 'r').read()
468
raise NotBranchError(self.base)
469
mutter("got branch format %r", fmt)
470
if fmt == BZR_BRANCH_FORMAT_5:
471
self._branch_format = 5
472
elif fmt == BZR_BRANCH_FORMAT_4:
473
self._branch_format = 4
475
if (not relax_version_check
476
and self._branch_format != 5):
477
raise BzrError('sorry, branch format %r not supported' % fmt,
478
['use a different bzr version',
479
'or remove the .bzr directory'
480
' and "bzr init" again'])
482
def get_root_id(self):
483
"""Return the id of this branches root"""
484
inv = self.read_working_inventory()
485
return inv.root.file_id
487
def set_root_id(self, file_id):
488
inv = self.read_working_inventory()
489
orig_root_id = inv.root.file_id
490
del inv._byid[inv.root.file_id]
491
inv.root.file_id = file_id
492
inv._byid[inv.root.file_id] = inv.root
495
if entry.parent_id in (None, orig_root_id):
496
entry.parent_id = inv.root.file_id
497
self._write_inventory(inv)
191
499
def read_working_inventory(self):
192
500
"""Read the working inventory."""
194
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
195
mutter("loaded inventory of %d items in %f"
196
% (len(inv), time.time() - before))
503
# ElementTree does its own conversion from UTF-8, so open in
505
f = self.controlfile('inventory', 'rb')
506
return bzrlib.xml5.serializer_v5.read_inventory(f)
200
511
def _write_inventory(self, inv):
201
512
"""Update the working inventory.
203
514
That is to say, the inventory describing changes underway, that
204
515
will be committed to the next revision.
206
## TODO: factor out to atomicfile? is rename safe on windows?
207
## TODO: Maybe some kind of clean/dirty marker on inventory?
208
tmpfname = self.controlfilename('inventory.tmp')
209
tmpf = file(tmpfname, 'w')
212
os.rename(tmpfname, self.controlfilename('inventory'))
517
from cStringIO import StringIO
521
bzrlib.xml5.serializer_v5.write_inventory(inv, sio)
523
# Transport handles atomicity
524
self.put_controlfile('inventory', sio)
213
528
mutter('wrote working inventory')
216
530
inventory = property(read_working_inventory, _write_inventory, None,
217
531
"""Inventory for the working copy.""")
220
def add(self, files, verbose=False):
533
def add(self, files, ids=None):
221
534
"""Make files versioned.
536
Note that the command line normally calls smart_add instead,
537
which can automatically recurse.
223
539
This puts the files in the Added state, so that they will be
224
540
recorded by the next commit.
226
:todo: Perhaps have an option to add the ids even if the files do
229
:todo: Perhaps return the ids of the files? But then again it
230
is easy to retrieve them if they're needed.
232
:todo: Option to specify file id.
234
:todo: Adding a directory should optionally recurse down and
235
add all non-ignored children. Perhaps do that in a
238
>>> b = ScratchBranch(files=['foo'])
239
>>> 'foo' in b.unknowns()
244
>>> 'foo' in b.unknowns()
246
>>> bool(b.inventory.path2id('foo'))
252
Traceback (most recent call last):
254
BzrError: ('foo is already versioned', [])
256
>>> b.add(['nothere'])
257
Traceback (most recent call last):
258
BzrError: ('cannot add: not a regular file or directory: nothere', [])
543
List of paths to add, relative to the base of the tree.
546
If set, use these instead of automatically generated ids.
547
Must be the same length as the list of files, but may
548
contain None for ids that are to be autogenerated.
550
TODO: Perhaps have an option to add the ids even if the files do
553
TODO: Perhaps yield the ids and paths as they're added.
261
555
# TODO: Re-adding a file that is removed in the working copy
262
556
# should probably put it back with the previous ID.
263
if isinstance(files, types.StringTypes):
557
if isinstance(files, basestring):
558
assert(ids is None or isinstance(ids, basestring))
266
inv = self.read_working_inventory()
268
if is_control_file(f):
269
bailout("cannot add control file %s" % quotefn(f))
274
bailout("cannot add top-level %r" % f)
276
fullpath = os.path.normpath(self.abspath(f))
279
kind = file_kind(fullpath)
281
# maybe something better?
282
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
284
if kind != 'file' and kind != 'directory':
285
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
287
file_id = gen_file_id(f)
288
inv.add_path(f, kind=kind, file_id=file_id)
291
show_status('A', kind, quotefn(f))
293
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
295
self._write_inventory(inv)
564
ids = [None] * len(files)
566
assert(len(ids) == len(files))
570
inv = self.read_working_inventory()
571
for f,file_id in zip(files, ids):
572
if is_control_file(f):
573
raise BzrError("cannot add control file %s" % quotefn(f))
578
raise BzrError("cannot add top-level %r" % f)
580
fullpath = os.path.normpath(self.abspath(f))
583
kind = file_kind(fullpath)
585
# maybe something better?
586
raise BzrError('cannot add: not a regular file, symlink or directory: %s' % quotefn(f))
588
if not InventoryEntry.versionable_kind(kind):
589
raise BzrError('cannot add: not a versionable file ('
590
'i.e. regular file, symlink or directory): %s' % quotefn(f))
593
file_id = gen_file_id(f)
594
inv.add_path(f, kind=kind, file_id=file_id)
596
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
598
self._write_inventory(inv)
603
def print_file(self, file, revno):
604
"""Print `file` to stdout."""
607
tree = self.revision_tree(self.get_rev_id(revno))
608
# use inventory as it was in that revision
609
file_id = tree.inventory.path2id(file)
611
raise BzrError("%r is not present in revision %s" % (file, revno))
612
tree.print_file(file_id)
299
617
def remove(self, files, verbose=False):
378
696
return self.working_tree().unknowns()
381
def commit(self, message, timestamp=None, timezone=None,
384
"""Commit working copy as a new revision.
386
The basic approach is to add all the file texts into the
387
store, then the inventory, then make a new revision pointing
388
to that inventory and store that.
390
This is not quite safe if the working copy changes during the
391
commit; for the moment that is simply not allowed. A better
392
approach is to make a temporary copy of the files before
393
computing their hashes, and then add those hashes in turn to
394
the inventory. This should mean at least that there are no
395
broken hash pointers. There is no way we can get a snapshot
396
of the whole directory at an instant. This would also have to
397
be robust against files disappearing, moving, etc. So the
398
whole thing is a bit hard.
400
:param timestamp: if not None, seconds-since-epoch for a
401
postdated/predated commit.
404
## TODO: Show branch names
406
# TODO: Don't commit if there are no changes, unless forced?
408
# First walk over the working inventory; and both update that
409
# and also build a new revision inventory. The revision
410
# inventory needs to hold the text-id, sha1 and size of the
411
# actual file versions committed in the revision. (These are
412
# not present in the working inventory.) We also need to
413
# detect missing/deleted files, and remove them from the
416
work_inv = self.read_working_inventory()
418
basis = self.basis_tree()
419
basis_inv = basis.inventory
421
for path, entry in work_inv.iter_entries():
422
## TODO: Cope with files that have gone missing.
424
## TODO: Check that the file kind has not changed from the previous
425
## revision of this file (if any).
429
p = self.abspath(path)
430
file_id = entry.file_id
431
mutter('commit prep file %s, id %r ' % (p, file_id))
433
if not os.path.exists(p):
434
mutter(" file is missing, removing from inventory")
436
show_status('D', entry.kind, quotefn(path))
437
missing_ids.append(file_id)
440
# TODO: Handle files that have been deleted
442
# TODO: Maybe a special case for empty files? Seems a
443
# waste to store them many times.
447
if basis_inv.has_id(file_id):
448
old_kind = basis_inv[file_id].kind
449
if old_kind != entry.kind:
450
bailout("entry %r changed kind from %r to %r"
451
% (file_id, old_kind, entry.kind))
453
if entry.kind == 'directory':
455
bailout("%s is entered as directory but not a directory" % quotefn(p))
456
elif entry.kind == 'file':
458
bailout("%s is entered as file but is not a file" % quotefn(p))
460
content = file(p, 'rb').read()
462
entry.text_sha1 = sha_string(content)
463
entry.text_size = len(content)
465
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
467
and (old_ie.text_size == entry.text_size)
468
and (old_ie.text_sha1 == entry.text_sha1)):
469
## assert content == basis.get_file(file_id).read()
470
entry.text_id = basis_inv[file_id].text_id
471
mutter(' unchanged from previous text_id {%s}' %
475
entry.text_id = gen_file_id(entry.name)
476
self.text_store.add(content, entry.text_id)
477
mutter(' stored with text_id {%s}' % entry.text_id)
481
elif (old_ie.name == entry.name
482
and old_ie.parent_id == entry.parent_id):
487
show_status(state, entry.kind, quotefn(path))
489
for file_id in missing_ids:
490
# have to do this later so we don't mess up the iterator.
491
# since parents may be removed before their children we
494
# FIXME: There's probably a better way to do this; perhaps
495
# the workingtree should know how to filter itself.
496
if work_inv.has_id(file_id):
497
del work_inv[file_id]
500
inv_id = rev_id = _gen_revision_id(time.time())
502
inv_tmp = tempfile.TemporaryFile()
503
inv.write_xml(inv_tmp)
505
self.inventory_store.add(inv_tmp, inv_id)
506
mutter('new inventory_id is {%s}' % inv_id)
508
self._write_inventory(work_inv)
510
if timestamp == None:
511
timestamp = time.time()
513
if committer == None:
514
committer = username()
517
timezone = local_time_offset()
519
mutter("building commit log message")
520
rev = Revision(timestamp=timestamp,
523
precursor = self.last_patch(),
528
rev_tmp = tempfile.TemporaryFile()
529
rev.write_xml(rev_tmp)
531
self.revision_store.add(rev_tmp, rev_id)
532
mutter("new revision_id is {%s}" % rev_id)
534
## XXX: Everything up to here can simply be orphaned if we abort
535
## the commit; it will leave junk files behind but that doesn't
538
## TODO: Read back the just-generated changeset, and make sure it
539
## applies and recreates the right state.
541
## TODO: Also calculate and store the inventory SHA1
542
mutter("committing patch r%d" % (self.revno() + 1))
544
mutter("append to revision-history")
545
f = self.controlfile('revision-history', 'at')
546
f.write(rev_id + '\n')
550
note("commited r%d" % self.revno())
699
def append_revision(self, *revision_ids):
700
for revision_id in revision_ids:
701
mutter("add {%s} to revision-history" % revision_id)
704
rev_history = self.revision_history()
705
rev_history.extend(revision_ids)
706
self.put_controlfile('revision-history', '\n'.join(rev_history))
710
def has_revision(self, revision_id):
711
"""True if this branch has a copy of the revision.
713
This does not necessarily imply the revision is merge
714
or on the mainline."""
715
return (revision_id is None
716
or revision_id in self.revision_store)
718
def get_revision_xml_file(self, revision_id):
719
"""Return XML file object for revision object."""
720
if not revision_id or not isinstance(revision_id, basestring):
721
raise InvalidRevisionId(revision_id)
726
return self.revision_store[revision_id]
727
except (IndexError, KeyError):
728
raise bzrlib.errors.NoSuchRevision(self, revision_id)
733
get_revision_xml = get_revision_xml_file
735
def get_revision_xml(self, revision_id):
736
return self.get_revision_xml_file(revision_id).read()
553
739
def get_revision(self, revision_id):
554
740
"""Return the Revision object for a named revision"""
555
r = Revision.read_xml(self.revision_store[revision_id])
741
xml_file = self.get_revision_xml_file(revision_id)
744
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
745
except SyntaxError, e:
746
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
556
750
assert r.revision_id == revision_id
560
def get_inventory(self, inventory_id):
561
"""Get Inventory object by hash.
563
:todo: Perhaps for this and similar methods, take a revision
564
parameter which can be either an integer revno or a
566
i = Inventory.read_xml(self.inventory_store[inventory_id])
753
def get_revision_delta(self, revno):
754
"""Return the delta for one revision.
756
The delta is relative to its mainline predecessor, or the
757
empty tree for revision 1.
759
assert isinstance(revno, int)
760
rh = self.revision_history()
761
if not (1 <= revno <= len(rh)):
762
raise InvalidRevisionNumber(revno)
764
# revno is 1-based; list is 0-based
766
new_tree = self.revision_tree(rh[revno-1])
768
old_tree = EmptyTree()
770
old_tree = self.revision_tree(rh[revno-2])
772
return compare_trees(old_tree, new_tree)
774
def get_revision_sha1(self, revision_id):
775
"""Hash the stored value of a revision, and return it."""
776
# In the future, revision entries will be signed. At that
777
# point, it is probably best *not* to include the signature
778
# in the revision hash. Because that lets you re-sign
779
# the revision, (add signatures/remove signatures) and still
780
# have all hash pointers stay consistent.
781
# But for now, just hash the contents.
782
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
784
def get_ancestry(self, revision_id):
785
"""Return a list of revision-ids integrated by a revision.
787
This currently returns a list, but the ordering is not guaranteed:
790
if revision_id is None:
792
w = self.control_weaves.get_weave('inventory')
793
return [None] + map(w.idx_to_name,
794
w.inclusions([w.lookup(revision_id)]))
796
def get_inventory_weave(self):
797
return self.control_weaves.get_weave('inventory')
799
def get_inventory(self, revision_id):
800
"""Get Inventory object by hash."""
801
xml = self.get_inventory_xml(revision_id)
802
return bzrlib.xml5.serializer_v5.read_inventory_from_string(xml)
804
def get_inventory_xml(self, revision_id):
805
"""Get inventory XML as a file object."""
807
assert isinstance(revision_id, basestring), type(revision_id)
808
iw = self.get_inventory_weave()
809
return iw.get_text(iw.lookup(revision_id))
811
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
813
def get_inventory_sha1(self, revision_id):
814
"""Return the sha1 hash of the inventory entry
816
return self.get_revision(revision_id).inventory_sha1
570
818
def get_revision_inventory(self, revision_id):
571
819
"""Return inventory of a past revision."""
820
# TODO: Unify this with get_inventory()
821
# bzr 0.0.6 and later imposes the constraint that the inventory_id
822
# must be the same as its revision, so this is trivial.
572
823
if revision_id == None:
824
return Inventory(self.get_root_id())
575
return self.get_inventory(self.get_revision(revision_id).inventory_id)
826
return self.get_inventory(revision_id)
578
828
def revision_history(self):
579
"""Return sequence of revision hashes on to this branch.
829
"""Return sequence of revision hashes on to this branch."""
832
return [l.rstrip('\r\n') for l in
833
self.controlfile('revision-history', 'r').readlines()]
581
>>> ScratchBranch().revision_history()
584
return [chomp(l) for l in self.controlfile('revision-history').readlines()]
837
def common_ancestor(self, other, self_revno=None, other_revno=None):
839
>>> from bzrlib.commit import commit
840
>>> sb = ScratchBranch(files=['foo', 'foo~'])
841
>>> sb.common_ancestor(sb) == (None, None)
843
>>> commit(sb, "Committing first revision", verbose=False)
844
>>> sb.common_ancestor(sb)[0]
846
>>> clone = sb.clone()
847
>>> commit(sb, "Committing second revision", verbose=False)
848
>>> sb.common_ancestor(sb)[0]
850
>>> sb.common_ancestor(clone)[0]
852
>>> commit(clone, "Committing divergent second revision",
854
>>> sb.common_ancestor(clone)[0]
856
>>> sb.common_ancestor(clone) == clone.common_ancestor(sb)
858
>>> sb.common_ancestor(sb) != clone.common_ancestor(clone)
860
>>> clone2 = sb.clone()
861
>>> sb.common_ancestor(clone2)[0]
863
>>> sb.common_ancestor(clone2, self_revno=1)[0]
865
>>> sb.common_ancestor(clone2, other_revno=1)[0]
868
my_history = self.revision_history()
869
other_history = other.revision_history()
870
if self_revno is None:
871
self_revno = len(my_history)
872
if other_revno is None:
873
other_revno = len(other_history)
874
indices = range(min((self_revno, other_revno)))
877
if my_history[r] == other_history[r]:
878
return r+1, my_history[r]
590
885
That is equivalent to the number of revisions committed to
593
>>> b = ScratchBranch()
596
>>> b.commit('no foo')
600
888
return len(self.revision_history())
603
def last_patch(self):
891
def last_revision(self):
604
892
"""Return last patch hash, or None if no history.
606
>>> ScratchBranch().last_patch() == None
609
894
ph = self.revision_history()
614
def lookup_revision(self, revno):
615
"""Return revision hash for revision number."""
901
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
902
"""Return a list of new revisions that would perfectly fit.
904
If self and other have not diverged, return a list of the revisions
905
present in other, but missing from self.
907
>>> from bzrlib.commit import commit
908
>>> bzrlib.trace.silent = True
909
>>> br1 = ScratchBranch()
910
>>> br2 = ScratchBranch()
911
>>> br1.missing_revisions(br2)
913
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
914
>>> br1.missing_revisions(br2)
916
>>> br2.missing_revisions(br1)
918
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
919
>>> br1.missing_revisions(br2)
921
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
922
>>> br1.missing_revisions(br2)
924
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
925
>>> br1.missing_revisions(br2)
926
Traceback (most recent call last):
927
DivergedBranches: These branches have diverged.
929
# FIXME: If the branches have diverged, but the latest
930
# revision in this branch is completely merged into the other,
931
# then we should still be able to pull.
932
self_history = self.revision_history()
933
self_len = len(self_history)
934
other_history = other.revision_history()
935
other_len = len(other_history)
936
common_index = min(self_len, other_len) -1
937
if common_index >= 0 and \
938
self_history[common_index] != other_history[common_index]:
939
raise DivergedBranches(self, other)
941
if stop_revision is None:
942
stop_revision = other_len
944
assert isinstance(stop_revision, int)
945
if stop_revision > other_len:
946
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
947
return other_history[self_len:stop_revision]
949
def update_revisions(self, other, stop_revision=None):
950
"""Pull in new perfect-fit revisions."""
951
from bzrlib.fetch import greedy_fetch
952
from bzrlib.revision import get_intervening_revisions
953
if stop_revision is None:
954
stop_revision = other.last_revision()
955
greedy_fetch(to_branch=self, from_branch=other,
956
revision=stop_revision)
957
pullable_revs = self.missing_revisions(
958
other, other.revision_id_to_revno(stop_revision))
960
greedy_fetch(to_branch=self,
962
revision=pullable_revs[-1])
963
self.append_revision(*pullable_revs)
966
def commit(self, *args, **kw):
967
from bzrlib.commit import Commit
968
Commit().commit(self, *args, **kw)
970
def revision_id_to_revno(self, revision_id):
971
"""Given a revision id, return its revno"""
972
if revision_id is None:
974
history = self.revision_history()
976
return history.index(revision_id) + 1
978
raise bzrlib.errors.NoSuchRevision(self, revision_id)
980
def get_rev_id(self, revno, history=None):
981
"""Find the revision id of the specified revno."""
620
# list is 0-based; revisions are 1-based
621
return self.revision_history()[revno-1]
623
bailout("no such revision %s" % revno)
985
history = self.revision_history()
986
elif revno <= 0 or revno > len(history):
987
raise bzrlib.errors.NoSuchRevision(self, revno)
988
return history[revno - 1]
626
990
def revision_tree(self, revision_id):
627
991
"""Return Tree for a revision on this branch.
629
993
`revision_id` may be None for the null revision, in which case
630
994
an `EmptyTree` is returned."""
995
# TODO: refactor this to use an existing revision object
996
# so we don't need to read it in twice.
632
997
if revision_id == None:
633
998
return EmptyTree()
635
1000
inv = self.get_revision_inventory(revision_id)
636
return RevisionTree(self.text_store, inv)
1001
return RevisionTree(self.weave_store, inv, revision_id)
639
1004
def working_tree(self):
640
1005
"""Return a `Tree` for the working copy."""
641
return WorkingTree(self.base, self.read_working_inventory())
1006
from bzrlib.workingtree import WorkingTree
1007
# TODO: In the future, WorkingTree should utilize Transport
1008
# RobertCollins 20051003 - I don't think it should - working trees are
1009
# much more complex to keep consistent than our careful .bzr subset.
1010
# instead, we should say that working trees are local only, and optimise
1012
return WorkingTree(self._transport.base, self.read_working_inventory())
644
1015
def basis_tree(self):
645
1016
"""Return `Tree` object for last revision.
647
1018
If there are no revisions yet, return an `EmptyTree`.
649
>>> b = ScratchBranch(files=['foo'])
650
>>> b.basis_tree().has_filename('foo')
652
>>> b.working_tree().has_filename('foo')
655
>>> b.commit('add foo')
656
>>> b.basis_tree().has_filename('foo')
659
r = self.last_patch()
663
return RevisionTree(self.text_store, self.get_revision_inventory(r))
667
def write_log(self, show_timezone='original'):
668
"""Write out human-readable log of commits to this branch
670
:param utc: If true, show dates in universal time, not local time."""
671
## TODO: Option to choose either original, utc or local timezone
674
for p in self.revision_history():
676
print 'revno:', revno
677
## TODO: Show hash if --id is given.
678
##print 'revision-hash:', p
679
rev = self.get_revision(p)
680
print 'committer:', rev.committer
681
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0,
684
## opportunistic consistency check, same as check_patch_chaining
685
if rev.precursor != precursor:
686
bailout("mismatched precursor!")
690
print ' (no message)'
692
for l in rev.message.split('\n'):
700
def show_status(branch, show_all=False):
701
"""Display single-line status for non-ignored working files.
703
The list is show sorted in order by file name.
705
>>> b = ScratchBranch(files=['foo', 'foo~'])
711
>>> b.commit("add foo")
713
>>> os.unlink(b.abspath('foo'))
718
:todo: Get state for single files.
720
:todo: Perhaps show a slash at the end of directory names.
724
# We have to build everything into a list first so that it can
725
# sorted by name, incorporating all the different sources.
727
# FIXME: Rather than getting things in random order and then sorting,
728
# just step through in order.
730
# Interesting case: the old ID for a file has been removed,
731
# but a new file has been created under that name.
733
old = branch.basis_tree()
734
old_inv = old.inventory
735
new = branch.working_tree()
736
new_inv = new.inventory
738
for fs, fid, oldname, newname, kind in diff_trees(old, new):
740
show_status(fs, kind,
741
oldname + ' => ' + newname)
742
elif fs == 'A' or fs == 'M':
743
show_status(fs, kind, newname)
745
show_status(fs, kind, oldname)
748
show_status(fs, kind, newname)
751
show_status(fs, kind, newname)
753
show_status(fs, kind, newname)
755
bailout("wierd file state %r" % ((fs, fid),))
759
class ScratchBranch(Branch):
1020
return self.revision_tree(self.last_revision())
1023
def rename_one(self, from_rel, to_rel):
1026
This can change the directory or the filename or both.
1030
tree = self.working_tree()
1031
inv = tree.inventory
1032
if not tree.has_filename(from_rel):
1033
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
1034
if tree.has_filename(to_rel):
1035
raise BzrError("can't rename: new working file %r already exists" % to_rel)
1037
file_id = inv.path2id(from_rel)
1039
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
1041
if inv.path2id(to_rel):
1042
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
1044
to_dir, to_tail = os.path.split(to_rel)
1045
to_dir_id = inv.path2id(to_dir)
1046
if to_dir_id == None and to_dir != '':
1047
raise BzrError("can't determine destination directory id for %r" % to_dir)
1049
mutter("rename_one:")
1050
mutter(" file_id {%s}" % file_id)
1051
mutter(" from_rel %r" % from_rel)
1052
mutter(" to_rel %r" % to_rel)
1053
mutter(" to_dir %r" % to_dir)
1054
mutter(" to_dir_id {%s}" % to_dir_id)
1056
inv.rename(file_id, to_dir_id, to_tail)
1058
from_abs = self.abspath(from_rel)
1059
to_abs = self.abspath(to_rel)
1061
rename(from_abs, to_abs)
1063
raise BzrError("failed to rename %r to %r: %s"
1064
% (from_abs, to_abs, e[1]),
1065
["rename rolled back"])
1067
self._write_inventory(inv)
1072
def move(self, from_paths, to_name):
1075
to_name must exist as a versioned directory.
1077
If to_name exists and is a directory, the files are moved into
1078
it, keeping their old names. If it is a directory,
1080
Note that to_name is only the last component of the new name;
1081
this doesn't change the directory.
1083
This returns a list of (from_path, to_path) pairs for each
1084
entry that is moved.
1089
## TODO: Option to move IDs only
1090
assert not isinstance(from_paths, basestring)
1091
tree = self.working_tree()
1092
inv = tree.inventory
1093
to_abs = self.abspath(to_name)
1094
if not isdir(to_abs):
1095
raise BzrError("destination %r is not a directory" % to_abs)
1096
if not tree.has_filename(to_name):
1097
raise BzrError("destination %r not in working directory" % to_abs)
1098
to_dir_id = inv.path2id(to_name)
1099
if to_dir_id == None and to_name != '':
1100
raise BzrError("destination %r is not a versioned directory" % to_name)
1101
to_dir_ie = inv[to_dir_id]
1102
if to_dir_ie.kind not in ('directory', 'root_directory'):
1103
raise BzrError("destination %r is not a directory" % to_abs)
1105
to_idpath = inv.get_idpath(to_dir_id)
1107
for f in from_paths:
1108
if not tree.has_filename(f):
1109
raise BzrError("%r does not exist in working tree" % f)
1110
f_id = inv.path2id(f)
1112
raise BzrError("%r is not versioned" % f)
1113
name_tail = splitpath(f)[-1]
1114
dest_path = appendpath(to_name, name_tail)
1115
if tree.has_filename(dest_path):
1116
raise BzrError("destination %r already exists" % dest_path)
1117
if f_id in to_idpath:
1118
raise BzrError("can't move %r to a subdirectory of itself" % f)
1120
# OK, so there's a race here, it's possible that someone will
1121
# create a file in this interval and then the rename might be
1122
# left half-done. But we should have caught most problems.
1124
for f in from_paths:
1125
name_tail = splitpath(f)[-1]
1126
dest_path = appendpath(to_name, name_tail)
1127
result.append((f, dest_path))
1128
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1130
rename(self.abspath(f), self.abspath(dest_path))
1132
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1133
["rename rolled back"])
1135
self._write_inventory(inv)
1142
def revert(self, filenames, old_tree=None, backups=True):
1143
"""Restore selected files to the versions from a previous tree.
1146
If true (default) backups are made of files before
1149
from bzrlib.errors import NotVersionedError, BzrError
1150
from bzrlib.atomicfile import AtomicFile
1151
from bzrlib.osutils import backup_file
1153
inv = self.read_working_inventory()
1154
if old_tree is None:
1155
old_tree = self.basis_tree()
1156
old_inv = old_tree.inventory
1159
for fn in filenames:
1160
file_id = inv.path2id(fn)
1162
raise NotVersionedError("not a versioned file", fn)
1163
if not old_inv.has_id(file_id):
1164
raise BzrError("file not present in old tree", fn, file_id)
1165
nids.append((fn, file_id))
1167
# TODO: Rename back if it was previously at a different location
1169
# TODO: If given a directory, restore the entire contents from
1170
# the previous version.
1172
# TODO: Make a backup to a temporary file.
1174
# TODO: If the file previously didn't exist, delete it?
1175
for fn, file_id in nids:
1178
f = AtomicFile(fn, 'wb')
1180
f.write(old_tree.get_file(file_id).read())
1186
def pending_merges(self):
1187
"""Return a list of pending merges.
1189
These are revisions that have been merged into the working
1190
directory but not yet committed.
1192
cfn = self._rel_controlfilename('pending-merges')
1193
if not self._transport.has(cfn):
1196
for l in self.controlfile('pending-merges', 'r').readlines():
1197
p.append(l.rstrip('\n'))
1201
def add_pending_merge(self, *revision_ids):
1202
# TODO: Perhaps should check at this point that the
1203
# history of the revision is actually present?
1204
p = self.pending_merges()
1206
for rev_id in revision_ids:
1212
self.set_pending_merges(p)
1214
def set_pending_merges(self, rev_list):
1217
self.put_controlfile('pending-merges', '\n'.join(rev_list))
1222
def get_parent(self):
1223
"""Return the parent location of the branch.
1225
This is the default location for push/pull/missing. The usual
1226
pattern is that the user can override it by specifying a
1230
_locs = ['parent', 'pull', 'x-pull']
1233
return self.controlfile(l, 'r').read().strip('\n')
1235
if e.errno != errno.ENOENT:
1240
def set_parent(self, url):
1241
# TODO: Maybe delete old location files?
1242
from bzrlib.atomicfile import AtomicFile
1245
f = AtomicFile(self.controlfilename('parent'))
1254
def check_revno(self, revno):
1256
Check whether a revno corresponds to any revision.
1257
Zero (the NULL revision) is considered valid.
1260
self.check_real_revno(revno)
1262
def check_real_revno(self, revno):
1264
Check whether a revno corresponds to a real revision.
1265
Zero (the NULL revision) is considered invalid
1267
if revno < 1 or revno > self.revno():
1268
raise InvalidRevisionNumber(revno)
1274
class ScratchBranch(_Branch):
760
1275
"""Special test class: a branch that cleans up after itself.
762
1277
>>> b = ScratchBranch()
763
1278
>>> isdir(b.base)
770
def __init__(self, files=[], dirs=[]):
1285
def __init__(self, files=[], dirs=[], base=None):
771
1286
"""Make a test branch.
773
1288
This creates a temporary directory and runs init-tree in it.
775
1290
If any files are listed, they are created in the working copy.
777
Branch.__init__(self, tempfile.mkdtemp(), init=True)
1292
from tempfile import mkdtemp
1297
if isinstance(base, basestring):
1298
base = get_transport(base)
1299
_Branch.__init__(self, base, init=init)
779
os.mkdir(self.abspath(d))
1301
self._transport.mkdir(d)
782
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
1304
self._transport.put(f, 'content of %s' % f)
1309
>>> orig = ScratchBranch(files=["file1", "file2"])
1310
>>> clone = orig.clone()
1311
>>> if os.name != 'nt':
1312
... os.path.samefile(orig.base, clone.base)
1314
... orig.base == clone.base
1317
>>> os.path.isfile(os.path.join(clone.base, "file1"))
1320
from shutil import copytree
1321
from tempfile import mkdtemp
1324
copytree(self.base, base, symlinks=True)
1325
return ScratchBranch(base=base)
785
1327
def __del__(self):
786
1331
"""Destroy the test branch, removing the scratch directory."""
787
shutil.rmtree(self.base)
1332
from shutil import rmtree
1335
mutter("delete ScratchBranch %s" % self.base)
1338
# Work around for shutil.rmtree failing on Windows when
1339
# readonly files are encountered
1340
mutter("hit exception in destroying ScratchBranch: %s" % e)
1341
for root, dirs, files in os.walk(self.base, topdown=False):
1343
os.chmod(os.path.join(root, name), 0700)
1345
self._transport = None