1
# Copyright (C) 2005 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22
from bzrlib.trace import mutter, note
23
from bzrlib.osutils import isdir, quotefn, compact_date, rand_bytes, \
25
sha_file, appendpath, file_kind
27
from bzrlib.errors import BzrError, InvalidRevisionNumber, InvalidRevisionId, \
28
DivergedBranches, NotBranchError
29
from bzrlib.textui import show_status
30
from bzrlib.revision import Revision
31
from bzrlib.delta import compare_trees
32
from bzrlib.tree import EmptyTree, RevisionTree
38
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
39
## TODO: Maybe include checks for common corruption of newlines, etc?
42
# TODO: Some operations like log might retrieve the same revisions
43
# repeatedly to calculate deltas. We could perhaps have a weakref
44
# cache in memory to make this faster.
46
# TODO: please move the revision-string syntax stuff out of the branch
47
# object; it's clutter
50
def find_branch(f, **args):
51
from bzrlib.transport import transport
52
from bzrlib.local_transport import LocalTransport
54
# FIXME: This is a hack around transport so that
55
# We can search the local directories for
57
if args.has_key('init') and args['init']:
58
# Don't search if we are init-ing
59
return Branch(t, **args)
60
if isinstance(t, LocalTransport):
61
root = find_branch_root(f)
64
return Branch(t, **args)
66
def _relpath(base, path):
67
"""Return path relative to base, or raise exception.
69
The path may be either an absolute path or a path relative to the
70
current working directory.
72
Lifted out of Branch.relpath for ease of testing.
74
os.path.commonprefix (python2.4) has a bad bug that it works just
75
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
76
avoids that problem."""
77
rp = os.path.abspath(path)
81
while len(head) >= len(base):
84
head, tail = os.path.split(head)
88
raise NotBranchError("path %r is not within branch %r" % (rp, base))
93
def find_branch_root(f=None):
94
"""Find the branch root enclosing f, or pwd.
96
f may be a filename or a URL.
98
It is not necessary that f exists.
100
Basically we keep looking up until we find the control directory or
101
run into the root. If there isn't one, raises NotBranchError.
106
f = os.path.realpath(f)
107
if not os.path.exists(f):
108
raise BzrError('%r does not exist' % f)
114
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
116
head, tail = os.path.split(f)
118
# reached the root, whatever that may be
119
raise NotBranchError('%s is not in a branch' % orig_f)
125
######################################################################
128
class Branch(object):
129
"""Branch holding a history of revisions.
132
Base directory of the branch.
138
If _lock_mode is true, a positive count of the number of times the
142
Lock object from bzrlib.lock.
150
# Map some sort of prefix into a namespace
151
# stuff like "revno:10", "revid:", etc.
152
# This should match a prefix with a function which accepts
153
REVISION_NAMESPACES = {}
155
def __init__(self, transport, init=False):
156
"""Create new branch object at a particular location.
158
transport -- A Transport object, defining how to access files.
159
(If a string, transport.transport() will be used to
160
create a Transport object)
162
init -- If True, create new control files in a previously
163
unversioned directory. If False, the branch must already
166
In the test suite, creation of new trees is tested using the
167
`ScratchBranch` class.
169
if isinstance(transport, basestring):
170
from transport import transport as get_transport
171
transport = get_transport(transport)
173
self._transport = transport
180
return '%s(%r)' % (self.__class__.__name__, self._transport.base)
187
if self._lock_mode or self._lock:
188
from bzrlib.warnings import warn
189
warn("branch %r was not explicitly unlocked" % self)
192
# TODO: It might be best to do this somewhere else,
193
# but it is nice for a Branch object to automatically
194
# cache it's information.
195
# Alternatively, we could have the Transport objects cache requests
196
# See the earlier discussion about how major objects (like Branch)
197
# should never expect their __del__ function to run.
198
if self.cache_root is not None:
199
#from warnings import warn
200
#warn("branch %r auto-cleanup of cache files" % self)
203
shutil.rmtree(self.cache_root)
206
self.cache_root = None
210
return self._transport.base
213
base = property(_get_base)
216
def lock_write(self):
217
# TODO: Upgrade locking to support using a Transport,
218
# and potentially a remote locking protocol
220
if self._lock_mode != 'w':
221
from bzrlib.errors import LockError
222
raise LockError("can't upgrade to a write lock from %r" %
224
self._lock_count += 1
226
self._lock = self._transport.lock_write(
227
self._rel_controlfilename('branch-lock'))
228
self._lock_mode = 'w'
234
assert self._lock_mode in ('r', 'w'), \
235
"invalid lock mode %r" % self._lock_mode
236
self._lock_count += 1
238
self._lock = self._transport.lock_read(
239
self._rel_controlfilename('branch-lock'))
240
self._lock_mode = 'r'
244
if not self._lock_mode:
245
from bzrlib.errors import LockError
246
raise LockError('branch %r is not locked' % (self))
248
if self._lock_count > 1:
249
self._lock_count -= 1
253
self._lock_mode = self._lock_count = None
255
def abspath(self, name):
256
"""Return absolute filename for something in the branch"""
257
return self._transport.abspath(name)
259
def relpath(self, path):
260
"""Return path relative to this branch of something inside it.
262
Raises an error if path is not in this branch."""
263
return self._transport.relpath(path)
266
def _rel_controlfilename(self, file_or_path):
267
if isinstance(file_or_path, basestring):
268
file_or_path = [file_or_path]
269
return [bzrlib.BZRDIR] + file_or_path
271
def controlfilename(self, file_or_path):
272
"""Return location relative to branch."""
273
return self._transport.abspath(self._rel_controlfilename(file_or_path))
276
def controlfile(self, file_or_path, mode='r'):
277
"""Open a control file for this branch.
279
There are two classes of file in the control directory: text
280
and binary. binary files are untranslated byte streams. Text
281
control files are stored with Unix newlines and in UTF-8, even
282
if the platform or locale defaults are different.
284
Controlfiles should almost never be opened in write mode but
285
rather should be atomically copied and replaced using atomicfile.
289
relpath = self._rel_controlfilename(file_or_path)
290
#TODO: codecs.open() buffers linewise, so it was overloaded with
291
# a much larger buffer, do we need to do the same for getreader/getwriter?
293
return self._transport.get(relpath)
295
raise BzrError("Branch.controlfile(mode='wb') is not supported, use put_controlfiles")
297
return codecs.getreader('utf-8')(self._transport.get(relpath), errors='replace')
299
raise BzrError("Branch.controlfile(mode='w') is not supported, use put_controlfiles")
301
raise BzrError("invalid controlfile mode %r" % mode)
303
def put_controlfile(self, path, f, encode=True):
304
"""Write an entry as a controlfile.
306
:param path: The path to put the file, relative to the .bzr control
308
:param f: A file-like or string object whose contents should be copied.
309
:param encode: If true, encode the contents as utf-8
311
self.put_controlfiles([(path, f)], encode=encode)
313
def put_controlfiles(self, files, encode=True):
314
"""Write several entries as controlfiles.
316
:param files: A list of [(path, file)] pairs, where the path is the directory
317
underneath the bzr control directory
318
:param encode: If true, encode the contents as utf-8
322
for path, f in files:
324
if isinstance(f, basestring):
325
f = f.encode('utf-8', 'replace')
327
f = codecs.getwriter('utf-8')(f, errors='replace')
328
path = self._rel_controlfilename(path)
329
ctrl_files.append((path, f))
330
self._transport.put_multi(ctrl_files)
332
def _make_control(self):
333
from bzrlib.inventory import Inventory
334
from cStringIO import StringIO
336
# Create an empty inventory
338
# if we want per-tree root ids then this is the place to set
339
# them; they're not needed for now and so ommitted for
341
bzrlib.xml.serializer_v4.write_inventory(Inventory(), sio)
343
dirs = [[], 'text-store', 'inventory-store', 'revision-store']
345
"This is a Bazaar-NG control directory.\n"
346
"Do not change any files in this directory.\n"),
347
('branch-format', BZR_BRANCH_FORMAT),
348
('revision-history', ''),
349
('merged-patches', ''),
350
('pending-merged-patches', ''),
353
('pending-merges', ''),
354
('inventory', sio.getvalue())
356
self._transport.mkdir_multi([self._rel_controlfilename(d) for d in dirs])
357
self.put_controlfiles(files)
358
mutter('created control directory in ' + self._transport.base)
360
def _check_format(self):
361
"""Check this branch format is supported.
363
The current tool only supports the current unstable format.
365
In the future, we might need different in-memory Branch
366
classes to support downlevel branches. But not yet.
368
# This ignores newlines so that we can open branches created
369
# on Windows from Linux and so on. I think it might be better
370
# to always make all internal files in unix format.
371
fmt = self.controlfile('branch-format', 'r').read()
372
fmt = fmt.replace('\r\n', '\n')
373
if fmt != BZR_BRANCH_FORMAT:
374
raise BzrError('sorry, branch format %r not supported' % fmt,
375
['use a different bzr version',
376
'or remove the .bzr directory and "bzr init" again'])
378
# We know that the format is the currently supported one.
379
# So create the rest of the entries.
380
from bzrlib.store.compressed_text import CompressedTextStore
382
if self._transport.should_cache():
384
self.cache_root = tempfile.mkdtemp(prefix='bzr-cache')
385
mutter('Branch %r using caching in %r' % (self, self.cache_root))
387
self.cache_root = None
390
relpath = self._rel_controlfilename(name)
391
store = CompressedTextStore(self._transport.clone(relpath))
392
if self._transport.should_cache():
393
from meta_store import CachedStore
394
cache_path = os.path.join(self.cache_root, name)
396
store = CachedStore(store, cache_path)
399
self.text_store = get_store('text-store')
400
self.revision_store = get_store('revision-store')
401
self.inventory_store = get_store('inventory-store')
403
def get_root_id(self):
404
"""Return the id of this branches root"""
405
inv = self.read_working_inventory()
406
return inv.root.file_id
408
def set_root_id(self, file_id):
409
inv = self.read_working_inventory()
410
orig_root_id = inv.root.file_id
411
del inv._byid[inv.root.file_id]
412
inv.root.file_id = file_id
413
inv._byid[inv.root.file_id] = inv.root
416
if entry.parent_id in (None, orig_root_id):
417
entry.parent_id = inv.root.file_id
418
self._write_inventory(inv)
420
def read_working_inventory(self):
421
"""Read the working inventory."""
422
from bzrlib.inventory import Inventory
425
# ElementTree does its own conversion from UTF-8, so open in
427
f = self.controlfile('inventory', 'rb')
428
return bzrlib.xml.serializer_v4.read_inventory(f)
433
def _write_inventory(self, inv):
434
"""Update the working inventory.
436
That is to say, the inventory describing changes underway, that
437
will be committed to the next revision.
439
from cStringIO import StringIO
443
bzrlib.xml.serializer_v4.write_inventory(inv, sio)
445
# Transport handles atomicity
446
self.put_controlfile('inventory', sio)
450
mutter('wrote working inventory')
453
inventory = property(read_working_inventory, _write_inventory, None,
454
"""Inventory for the working copy.""")
457
def add(self, files, ids=None):
458
"""Make files versioned.
460
Note that the command line normally calls smart_add instead,
461
which can automatically recurse.
463
This puts the files in the Added state, so that they will be
464
recorded by the next commit.
467
List of paths to add, relative to the base of the tree.
470
If set, use these instead of automatically generated ids.
471
Must be the same length as the list of files, but may
472
contain None for ids that are to be autogenerated.
474
TODO: Perhaps have an option to add the ids even if the files do
477
TODO: Perhaps yield the ids and paths as they're added.
479
# TODO: Re-adding a file that is removed in the working copy
480
# should probably put it back with the previous ID.
481
if isinstance(files, basestring):
482
assert(ids is None or isinstance(ids, basestring))
488
ids = [None] * len(files)
490
assert(len(ids) == len(files))
494
inv = self.read_working_inventory()
495
for f,file_id in zip(files, ids):
496
if is_control_file(f):
497
raise BzrError("cannot add control file %s" % quotefn(f))
502
raise BzrError("cannot add top-level %r" % f)
504
fullpath = os.path.normpath(self.abspath(f))
507
kind = file_kind(fullpath)
509
# maybe something better?
510
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
512
if kind != 'file' and kind != 'directory':
513
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
516
file_id = gen_file_id(f)
517
inv.add_path(f, kind=kind, file_id=file_id)
519
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
521
self._write_inventory(inv)
526
def print_file(self, file, revno):
527
"""Print `file` to stdout."""
530
tree = self.revision_tree(self.lookup_revision(revno))
531
# use inventory as it was in that revision
532
file_id = tree.inventory.path2id(file)
534
raise BzrError("%r is not present in revision %s" % (file, revno))
535
tree.print_file(file_id)
540
def remove(self, files, verbose=False):
541
"""Mark nominated files for removal from the inventory.
543
This does not remove their text. This does not run on
545
TODO: Refuse to remove modified files unless --force is given?
547
TODO: Do something useful with directories.
549
TODO: Should this remove the text or not? Tough call; not
550
removing may be useful and the user can just use use rm, and
551
is the opposite of add. Removing it is consistent with most
552
other tools. Maybe an option.
554
## TODO: Normalize names
555
## TODO: Remove nested loops; better scalability
556
if isinstance(files, basestring):
562
tree = self.working_tree()
565
# do this before any modifications
569
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
570
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
572
# having remove it, it must be either ignored or unknown
573
if tree.is_ignored(f):
577
show_status(new_status, inv[fid].kind, quotefn(f))
580
self._write_inventory(inv)
585
# FIXME: this doesn't need to be a branch method
586
def set_inventory(self, new_inventory_list):
587
from bzrlib.inventory import Inventory, InventoryEntry
588
inv = Inventory(self.get_root_id())
589
for path, file_id, parent, kind in new_inventory_list:
590
name = os.path.basename(path)
593
inv.add(InventoryEntry(file_id, name, kind, parent))
594
self._write_inventory(inv)
598
"""Return all unknown files.
600
These are files in the working directory that are not versioned or
601
control files or ignored.
603
>>> b = ScratchBranch(files=['foo', 'foo~'])
604
>>> list(b.unknowns())
607
>>> list(b.unknowns())
610
>>> list(b.unknowns())
613
return self.working_tree().unknowns()
616
def append_revision(self, *revision_ids):
617
for revision_id in revision_ids:
618
mutter("add {%s} to revision-history" % revision_id)
620
rev_history = self.revision_history()
621
rev_history.extend(revision_ids)
625
self.put_controlfile('revision-history', '\n'.join(rev_history))
630
def get_revision_xml_file(self, revision_id):
631
"""Return XML file object for revision object."""
632
if not revision_id or not isinstance(revision_id, basestring):
633
raise InvalidRevisionId(revision_id)
638
return self.revision_store[revision_id]
639
except (IndexError, KeyError):
640
raise bzrlib.errors.NoSuchRevision(self, revision_id)
646
get_revision_xml = get_revision_xml_file
649
def get_revision(self, revision_id):
650
"""Return the Revision object for a named revision"""
651
xml_file = self.get_revision_xml_file(revision_id)
654
r = bzrlib.xml.serializer_v4.read_revision(xml_file)
655
except SyntaxError, e:
656
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
660
assert r.revision_id == revision_id
664
def get_revision_delta(self, revno):
665
"""Return the delta for one revision.
667
The delta is relative to its mainline predecessor, or the
668
empty tree for revision 1.
670
assert isinstance(revno, int)
671
rh = self.revision_history()
672
if not (1 <= revno <= len(rh)):
673
raise InvalidRevisionNumber(revno)
675
# revno is 1-based; list is 0-based
677
new_tree = self.revision_tree(rh[revno-1])
679
old_tree = EmptyTree()
681
old_tree = self.revision_tree(rh[revno-2])
683
return compare_trees(old_tree, new_tree)
686
def get_revisions(self, revision_ids, pb=None):
687
"""Return the Revision object for a set of named revisions"""
688
from bzrlib.revision import Revision
689
from bzrlib.xml import unpack_xml
691
# TODO: We need to decide what to do here
692
# we cannot use a generator with a try/finally, because
693
# you cannot guarantee that the caller will iterate through
695
# in the past, get_inventory wasn't even wrapped in a
696
# try/finally locking block.
697
# We could either lock without the try/finally, or just
698
# not lock at all. We are reading entries that should
700
# I prefer locking with no finally, so that if someone
701
# asks for a list of revisions, but doesn't consume them,
702
# that is their problem, and they will suffer the consequences
704
for xml_file in self.revision_store.get(revision_ids, pb=pb):
706
r = bzrlib.xml.serializer_v4.read_revision(xml_file)
707
except SyntaxError, e:
708
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
714
def get_revision_sha1(self, revision_id):
715
"""Hash the stored value of a revision, and return it."""
716
# In the future, revision entries will be signed. At that
717
# point, it is probably best *not* to include the signature
718
# in the revision hash. Because that lets you re-sign
719
# the revision, (add signatures/remove signatures) and still
720
# have all hash pointers stay consistent.
721
# But for now, just hash the contents.
722
return bzrlib.osutils.sha_file(self.get_revision_xml(revision_id))
725
def get_inventory(self, inventory_id):
726
"""Get Inventory object by hash.
728
TODO: Perhaps for this and similar methods, take a revision
729
parameter which can be either an integer revno or a
732
f = self.get_inventory_xml_file(inventory_id)
733
return bzrlib.xml.serializer_v4.read_inventory(f)
736
def get_inventory_xml(self, inventory_id):
737
"""Get inventory XML as a file object."""
738
# Shouldn't this have a read-lock around it?
739
# As well as some sort of trap for missing ids?
740
return self.inventory_store[inventory_id]
742
get_inventory_xml_file = get_inventory_xml
744
def get_inventories(self, inventory_ids, pb=None, ignore_missing=False):
745
"""Get Inventory objects by id
747
from bzrlib.inventory import Inventory
749
# See the discussion in get_revisions for why
750
# we don't use a try/finally block here
752
for f in self.inventory_store.get(inventory_ids, pb=pb, ignore_missing=ignore_missing):
754
# TODO: Possibly put a try/except around this to handle
755
# read serialization errors
756
r = bzrlib.xml.serializer_v4.read_inventory(f)
761
raise bzrlib.errors.NoSuchRevision(self, revision_id)
764
def get_inventory_sha1(self, inventory_id):
765
"""Return the sha1 hash of the inventory entry
767
return sha_file(self.get_inventory_xml(inventory_id))
770
def get_revision_inventory(self, revision_id):
771
"""Return inventory of a past revision."""
772
# bzr 0.0.6 imposes the constraint that the inventory_id
773
# must be the same as its revision, so this is trivial.
774
if revision_id == None:
775
from bzrlib.inventory import Inventory
776
return Inventory(self.get_root_id())
778
return self.get_inventory(revision_id)
781
def revision_history(self):
782
"""Return sequence of revision hashes on to this branch.
784
>>> ScratchBranch().revision_history()
789
return [l.rstrip('\r\n') for l in
790
self.controlfile('revision-history', 'r').readlines()]
795
def common_ancestor(self, other, self_revno=None, other_revno=None):
797
>>> from bzrlib.commit import commit
798
>>> sb = ScratchBranch(files=['foo', 'foo~'])
799
>>> sb.common_ancestor(sb) == (None, None)
801
>>> commit(sb, "Committing first revision", verbose=False)
802
>>> sb.common_ancestor(sb)[0]
804
>>> clone = sb.clone()
805
>>> commit(sb, "Committing second revision", verbose=False)
806
>>> sb.common_ancestor(sb)[0]
808
>>> sb.common_ancestor(clone)[0]
810
>>> commit(clone, "Committing divergent second revision",
812
>>> sb.common_ancestor(clone)[0]
814
>>> sb.common_ancestor(clone) == clone.common_ancestor(sb)
816
>>> sb.common_ancestor(sb) != clone.common_ancestor(clone)
818
>>> clone2 = sb.clone()
819
>>> sb.common_ancestor(clone2)[0]
821
>>> sb.common_ancestor(clone2, self_revno=1)[0]
823
>>> sb.common_ancestor(clone2, other_revno=1)[0]
826
my_history = self.revision_history()
827
other_history = other.revision_history()
828
if self_revno is None:
829
self_revno = len(my_history)
830
if other_revno is None:
831
other_revno = len(other_history)
832
indices = range(min((self_revno, other_revno)))
835
if my_history[r] == other_history[r]:
836
return r+1, my_history[r]
841
"""Return current revision number for this branch.
843
That is equivalent to the number of revisions committed to
846
return len(self.revision_history())
849
def last_patch(self):
850
"""Return last patch hash, or None if no history.
852
ph = self.revision_history()
859
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
861
If self and other have not diverged, return a list of the revisions
862
present in other, but missing from self.
864
>>> from bzrlib.commit import commit
865
>>> bzrlib.trace.silent = True
866
>>> br1 = ScratchBranch()
867
>>> br2 = ScratchBranch()
868
>>> br1.missing_revisions(br2)
870
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
871
>>> br1.missing_revisions(br2)
873
>>> br2.missing_revisions(br1)
875
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
876
>>> br1.missing_revisions(br2)
878
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
879
>>> br1.missing_revisions(br2)
881
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
882
>>> br1.missing_revisions(br2)
883
Traceback (most recent call last):
884
DivergedBranches: These branches have diverged.
886
self_history = self.revision_history()
887
self_len = len(self_history)
888
other_history = other.revision_history()
889
other_len = len(other_history)
890
common_index = min(self_len, other_len) -1
891
if common_index >= 0 and \
892
self_history[common_index] != other_history[common_index]:
893
raise DivergedBranches(self, other)
895
if stop_revision is None:
896
stop_revision = other_len
897
elif stop_revision > other_len:
898
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
900
return other_history[self_len:stop_revision]
903
def update_revisions(self, other, stop_revision=None):
904
"""Pull in all new revisions from other branch.
906
from bzrlib.fetch import greedy_fetch
907
from bzrlib.revision import get_intervening_revisions
909
pb = bzrlib.ui.ui_factory.progress_bar()
910
pb.update('comparing histories')
911
if stop_revision is None:
912
other_revision = other.last_patch()
914
other_revision = other.lookup_revision(stop_revision)
915
count = greedy_fetch(self, other, other_revision, pb)[0]
917
revision_ids = self.missing_revisions(other, stop_revision)
918
except DivergedBranches, e:
920
revision_ids = get_intervening_revisions(self.last_patch(),
921
other_revision, self)
922
assert self.last_patch() not in revision_ids
923
except bzrlib.errors.NotAncestor:
926
self.append_revision(*revision_ids)
929
def install_revisions(self, other, revision_ids, pb):
930
# We are going to iterate this many times, so make sure
931
# that it is a list, and not a generator
932
revision_ids = list(revision_ids)
933
if hasattr(other.revision_store, "prefetch"):
934
other.revision_store.prefetch(revision_ids)
935
if hasattr(other.inventory_store, "prefetch"):
936
other.inventory_store.prefetch(inventory_ids)
939
pb = bzrlib.ui.ui_factory.progress_bar()
941
# This entire next section is generally done
942
# with either generators, or bulk updates
943
inventories = other.get_inventories(revision_ids, ignore_missing=True)
947
good_revisions = set()
948
for i, (inv, rev_id) in enumerate(zip(inventories, revision_ids)):
949
pb.update('fetching revision', i+1, len(revision_ids))
951
# We don't really need to get the revision here, because
952
# the only thing we needed was the inventory_id, which now
953
# is (by design) identical to the revision_id
955
# rev = other.get_revision(rev_id)
956
# except bzrlib.errors.NoSuchRevision:
957
# failures.add(rev_id)
964
good_revisions.add(rev_id)
967
for key, entry in inv.iter_entries():
968
if entry.text_id is None:
970
text_ids.append(entry.text_id)
972
has_ids = self.text_store.has(text_ids)
973
for has, text_id in zip(has_ids, text_ids):
975
needed_texts.add(text_id)
979
count, cp_fail = self.text_store.copy_multi(other.text_store,
981
#print "Added %d texts." % count
982
count, cp_fail = self.inventory_store.copy_multi(other.inventory_store,
984
#print "Added %d inventories." % count
985
count, cp_fail = self.revision_store.copy_multi(other.revision_store,
988
assert len(cp_fail) == 0
989
return count, failures
992
def commit(self, *args, **kw):
993
from bzrlib.commit import commit
994
commit(self, *args, **kw)
997
def lookup_revision(self, revision):
998
"""Return the revision identifier for a given revision information."""
999
revno, info = self._get_revision_info(revision)
1003
def revision_id_to_revno(self, revision_id):
1004
"""Given a revision id, return its revno"""
1005
history = self.revision_history()
1007
return history.index(revision_id) + 1
1009
raise bzrlib.errors.NoSuchRevision(self, revision_id)
1012
def get_revision_info(self, revision):
1013
"""Return (revno, revision id) for revision identifier.
1015
revision can be an integer, in which case it is assumed to be revno (though
1016
this will translate negative values into positive ones)
1017
revision can also be a string, in which case it is parsed for something like
1018
'date:' or 'revid:' etc.
1020
revno, rev_id = self._get_revision_info(revision)
1022
raise bzrlib.errors.NoSuchRevision(self, revision)
1023
return revno, rev_id
1025
def get_rev_id(self, revno, history=None):
1026
"""Find the revision id of the specified revno."""
1030
history = self.revision_history()
1031
elif revno <= 0 or revno > len(history):
1032
raise bzrlib.errors.NoSuchRevision(self, revno)
1033
return history[revno - 1]
1035
def _get_revision_info(self, revision):
1036
"""Return (revno, revision id) for revision specifier.
1038
revision can be an integer, in which case it is assumed to be revno
1039
(though this will translate negative values into positive ones)
1040
revision can also be a string, in which case it is parsed for something
1041
like 'date:' or 'revid:' etc.
1043
A revid is always returned. If it is None, the specifier referred to
1044
the null revision. If the revid does not occur in the revision
1045
history, revno will be None.
1048
if revision is None:
1051
try:# Convert to int if possible
1052
revision = int(revision)
1055
revs = self.revision_history()
1056
if isinstance(revision, int):
1058
revno = len(revs) + revision + 1
1061
rev_id = self.get_rev_id(revno, revs)
1062
elif isinstance(revision, basestring):
1063
for prefix, func in Branch.REVISION_NAMESPACES.iteritems():
1064
if revision.startswith(prefix):
1065
result = func(self, revs, revision)
1067
revno, rev_id = result
1070
rev_id = self.get_rev_id(revno, revs)
1073
raise BzrError('No namespace registered for string: %r' %
1076
raise TypeError('Unhandled revision type %s' % revision)
1080
raise bzrlib.errors.NoSuchRevision(self, revision)
1081
return revno, rev_id
1083
def _namespace_revno(self, revs, revision):
1084
"""Lookup a revision by revision number"""
1085
assert revision.startswith('revno:')
1087
return (int(revision[6:]),)
1090
REVISION_NAMESPACES['revno:'] = _namespace_revno
1092
def _namespace_revid(self, revs, revision):
1093
assert revision.startswith('revid:')
1094
rev_id = revision[len('revid:'):]
1096
return revs.index(rev_id) + 1, rev_id
1099
REVISION_NAMESPACES['revid:'] = _namespace_revid
1101
def _namespace_last(self, revs, revision):
1102
assert revision.startswith('last:')
1104
offset = int(revision[5:])
1109
raise BzrError('You must supply a positive value for --revision last:XXX')
1110
return (len(revs) - offset + 1,)
1111
REVISION_NAMESPACES['last:'] = _namespace_last
1113
def _namespace_tag(self, revs, revision):
1114
assert revision.startswith('tag:')
1115
raise BzrError('tag: namespace registered, but not implemented.')
1116
REVISION_NAMESPACES['tag:'] = _namespace_tag
1118
def _namespace_date(self, revs, revision):
1119
assert revision.startswith('date:')
1121
# Spec for date revisions:
1123
# value can be 'yesterday', 'today', 'tomorrow' or a YYYY-MM-DD string.
1124
# it can also start with a '+/-/='. '+' says match the first
1125
# entry after the given date. '-' is match the first entry before the date
1126
# '=' is match the first entry after, but still on the given date.
1128
# +2005-05-12 says find the first matching entry after May 12th, 2005 at 0:00
1129
# -2005-05-12 says find the first matching entry before May 12th, 2005 at 0:00
1130
# =2005-05-12 says find the first match after May 12th, 2005 at 0:00 but before
1131
# May 13th, 2005 at 0:00
1133
# So the proper way of saying 'give me all entries for today' is:
1134
# -r {date:+today}:{date:-tomorrow}
1135
# The default is '=' when not supplied
1138
if val[:1] in ('+', '-', '='):
1139
match_style = val[:1]
1142
today = datetime.datetime.today().replace(hour=0,minute=0,second=0,microsecond=0)
1143
if val.lower() == 'yesterday':
1144
dt = today - datetime.timedelta(days=1)
1145
elif val.lower() == 'today':
1147
elif val.lower() == 'tomorrow':
1148
dt = today + datetime.timedelta(days=1)
1151
# This should be done outside the function to avoid recompiling it.
1152
_date_re = re.compile(
1153
r'(?P<date>(?P<year>\d\d\d\d)-(?P<month>\d\d)-(?P<day>\d\d))?'
1155
r'(?P<time>(?P<hour>\d\d):(?P<minute>\d\d)(:(?P<second>\d\d))?)?'
1157
m = _date_re.match(val)
1158
if not m or (not m.group('date') and not m.group('time')):
1159
raise BzrError('Invalid revision date %r' % revision)
1162
year, month, day = int(m.group('year')), int(m.group('month')), int(m.group('day'))
1164
year, month, day = today.year, today.month, today.day
1166
hour = int(m.group('hour'))
1167
minute = int(m.group('minute'))
1168
if m.group('second'):
1169
second = int(m.group('second'))
1173
hour, minute, second = 0,0,0
1175
dt = datetime.datetime(year=year, month=month, day=day,
1176
hour=hour, minute=minute, second=second)
1180
if match_style == '-':
1182
elif match_style == '=':
1183
last = dt + datetime.timedelta(days=1)
1186
for i in range(len(revs)-1, -1, -1):
1187
r = self.get_revision(revs[i])
1188
# TODO: Handle timezone.
1189
dt = datetime.datetime.fromtimestamp(r.timestamp)
1190
if first >= dt and (last is None or dt >= last):
1193
for i in range(len(revs)):
1194
r = self.get_revision(revs[i])
1195
# TODO: Handle timezone.
1196
dt = datetime.datetime.fromtimestamp(r.timestamp)
1197
if first <= dt and (last is None or dt <= last):
1199
REVISION_NAMESPACES['date:'] = _namespace_date
1202
def _namespace_ancestor(self, revs, revision):
1203
from revision import common_ancestor, MultipleRevisionSources
1204
other_branch = find_branch(_trim_namespace('ancestor', revision))
1205
revision_a = self.last_patch()
1206
revision_b = other_branch.last_patch()
1207
for r, b in ((revision_a, self), (revision_b, other_branch)):
1209
raise bzrlib.errors.NoCommits(b)
1210
revision_source = MultipleRevisionSources(self, other_branch)
1211
result = common_ancestor(revision_a, revision_b, revision_source)
1213
revno = self.revision_id_to_revno(result)
1214
except bzrlib.errors.NoSuchRevision:
1219
REVISION_NAMESPACES['ancestor:'] = _namespace_ancestor
1221
def revision_tree(self, revision_id):
1222
"""Return Tree for a revision on this branch.
1224
`revision_id` may be None for the null revision, in which case
1225
an `EmptyTree` is returned."""
1226
# TODO: refactor this to use an existing revision object
1227
# so we don't need to read it in twice.
1228
if revision_id == None:
1231
inv = self.get_revision_inventory(revision_id)
1232
return RevisionTree(self.text_store, inv)
1235
def working_tree(self):
1236
"""Return a `Tree` for the working copy."""
1237
from bzrlib.workingtree import WorkingTree
1238
# TODO: In the future, WorkingTree should utilize Transport
1239
return WorkingTree(self._transport.base, self.read_working_inventory())
1242
def basis_tree(self):
1243
"""Return `Tree` object for last revision.
1245
If there are no revisions yet, return an `EmptyTree`.
1247
r = self.last_patch()
1251
return RevisionTree(self.text_store, self.get_revision_inventory(r))
1255
def rename_one(self, from_rel, to_rel):
1258
This can change the directory or the filename or both.
1262
tree = self.working_tree()
1263
inv = tree.inventory
1264
if not tree.has_filename(from_rel):
1265
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
1266
if tree.has_filename(to_rel):
1267
raise BzrError("can't rename: new working file %r already exists" % to_rel)
1269
file_id = inv.path2id(from_rel)
1271
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
1273
if inv.path2id(to_rel):
1274
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
1276
to_dir, to_tail = os.path.split(to_rel)
1277
to_dir_id = inv.path2id(to_dir)
1278
if to_dir_id == None and to_dir != '':
1279
raise BzrError("can't determine destination directory id for %r" % to_dir)
1281
mutter("rename_one:")
1282
mutter(" file_id {%s}" % file_id)
1283
mutter(" from_rel %r" % from_rel)
1284
mutter(" to_rel %r" % to_rel)
1285
mutter(" to_dir %r" % to_dir)
1286
mutter(" to_dir_id {%s}" % to_dir_id)
1288
inv.rename(file_id, to_dir_id, to_tail)
1290
from_abs = self.abspath(from_rel)
1291
to_abs = self.abspath(to_rel)
1293
os.rename(from_abs, to_abs)
1295
raise BzrError("failed to rename %r to %r: %s"
1296
% (from_abs, to_abs, e[1]),
1297
["rename rolled back"])
1299
self._write_inventory(inv)
1304
def move(self, from_paths, to_name):
1307
to_name must exist as a versioned directory.
1309
If to_name exists and is a directory, the files are moved into
1310
it, keeping their old names. If it is a directory,
1312
Note that to_name is only the last component of the new name;
1313
this doesn't change the directory.
1315
This returns a list of (from_path, to_path) pairs for each
1316
entry that is moved.
1321
## TODO: Option to move IDs only
1322
assert not isinstance(from_paths, basestring)
1323
tree = self.working_tree()
1324
inv = tree.inventory
1325
to_abs = self.abspath(to_name)
1326
if not isdir(to_abs):
1327
raise BzrError("destination %r is not a directory" % to_abs)
1328
if not tree.has_filename(to_name):
1329
raise BzrError("destination %r not in working directory" % to_abs)
1330
to_dir_id = inv.path2id(to_name)
1331
if to_dir_id == None and to_name != '':
1332
raise BzrError("destination %r is not a versioned directory" % to_name)
1333
to_dir_ie = inv[to_dir_id]
1334
if to_dir_ie.kind not in ('directory', 'root_directory'):
1335
raise BzrError("destination %r is not a directory" % to_abs)
1337
to_idpath = inv.get_idpath(to_dir_id)
1339
for f in from_paths:
1340
if not tree.has_filename(f):
1341
raise BzrError("%r does not exist in working tree" % f)
1342
f_id = inv.path2id(f)
1344
raise BzrError("%r is not versioned" % f)
1345
name_tail = splitpath(f)[-1]
1346
dest_path = appendpath(to_name, name_tail)
1347
if tree.has_filename(dest_path):
1348
raise BzrError("destination %r already exists" % dest_path)
1349
if f_id in to_idpath:
1350
raise BzrError("can't move %r to a subdirectory of itself" % f)
1352
# OK, so there's a race here, it's possible that someone will
1353
# create a file in this interval and then the rename might be
1354
# left half-done. But we should have caught most problems.
1356
for f in from_paths:
1357
name_tail = splitpath(f)[-1]
1358
dest_path = appendpath(to_name, name_tail)
1359
result.append((f, dest_path))
1360
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1362
os.rename(self.abspath(f), self.abspath(dest_path))
1364
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1365
["rename rolled back"])
1367
self._write_inventory(inv)
1374
def revert(self, filenames, old_tree=None, backups=True):
1375
"""Restore selected files to the versions from a previous tree.
1378
If true (default) backups are made of files before
1381
from bzrlib.errors import NotVersionedError, BzrError
1382
from bzrlib.atomicfile import AtomicFile
1383
from bzrlib.osutils import backup_file
1385
inv = self.read_working_inventory()
1386
if old_tree is None:
1387
old_tree = self.basis_tree()
1388
old_inv = old_tree.inventory
1391
for fn in filenames:
1392
file_id = inv.path2id(fn)
1394
raise NotVersionedError("not a versioned file", fn)
1395
if not old_inv.has_id(file_id):
1396
raise BzrError("file not present in old tree", fn, file_id)
1397
nids.append((fn, file_id))
1399
# TODO: Rename back if it was previously at a different location
1401
# TODO: If given a directory, restore the entire contents from
1402
# the previous version.
1404
# TODO: Make a backup to a temporary file.
1406
# TODO: If the file previously didn't exist, delete it?
1407
for fn, file_id in nids:
1410
f = AtomicFile(fn, 'wb')
1412
f.write(old_tree.get_file(file_id).read())
1418
def pending_merges(self):
1419
"""Return a list of pending merges.
1421
These are revisions that have been merged into the working
1422
directory but not yet committed.
1424
cfn = self._rel_controlfilename('pending-merges')
1425
if not self._transport.has(cfn):
1428
for l in self.controlfile('pending-merges', 'r').readlines():
1429
p.append(l.rstrip('\n'))
1433
def add_pending_merge(self, *revision_ids):
1434
from bzrlib.revision import validate_revision_id
1436
for rev_id in revision_ids:
1437
validate_revision_id(rev_id)
1439
p = self.pending_merges()
1441
for rev_id in revision_ids:
1447
self.set_pending_merges(p)
1449
def set_pending_merges(self, rev_list):
1452
self.put_controlfile('pending-merges', '\n'.join(rev_list))
1457
def get_parent(self):
1458
"""Return the parent location of the branch.
1460
This is the default location for push/pull/missing. The usual
1461
pattern is that the user can override it by specifying a
1465
_locs = ['parent', 'pull', 'x-pull']
1468
return self.controlfile(l, 'r').read().strip('\n')
1470
if e.errno != errno.ENOENT:
1475
def set_parent(self, url):
1476
# TODO: Maybe delete old location files?
1477
from bzrlib.atomicfile import AtomicFile
1480
f = AtomicFile(self.controlfilename('parent'))
1489
def check_revno(self, revno):
1491
Check whether a revno corresponds to any revision.
1492
Zero (the NULL revision) is considered valid.
1495
self.check_real_revno(revno)
1497
def check_real_revno(self, revno):
1499
Check whether a revno corresponds to a real revision.
1500
Zero (the NULL revision) is considered invalid
1502
if revno < 1 or revno > self.revno():
1503
raise InvalidRevisionNumber(revno)
1508
class ScratchBranch(Branch):
1509
"""Special test class: a branch that cleans up after itself.
1511
>>> b = ScratchBranch()
1519
def __init__(self, files=[], dirs=[], base=None):
1520
"""Make a test branch.
1522
This creates a temporary directory and runs init-tree in it.
1524
If any files are listed, they are created in the working copy.
1526
from tempfile import mkdtemp
1531
Branch.__init__(self, base, init=init)
1533
self._transport.mkdir(d)
1536
self._transport.put(f, 'content of %s' % f)
1541
>>> orig = ScratchBranch(files=["file1", "file2"])
1542
>>> clone = orig.clone()
1543
>>> os.path.samefile(orig.base, clone.base)
1545
>>> os.path.isfile(os.path.join(clone.base, "file1"))
1548
from shutil import copytree
1549
from tempfile import mkdtemp
1552
copytree(self.base, base, symlinks=True)
1553
return ScratchBranch(base=base)
1561
"""Destroy the test branch, removing the scratch directory."""
1562
from shutil import rmtree
1565
mutter("delete ScratchBranch %s" % self.base)
1568
# Work around for shutil.rmtree failing on Windows when
1569
# readonly files are encountered
1570
mutter("hit exception in destroying ScratchBranch: %s" % e)
1571
for root, dirs, files in os.walk(self.base, topdown=False):
1573
os.chmod(os.path.join(root, name), 0700)
1575
self._transport = None
1579
######################################################################
1583
def is_control_file(filename):
1584
## FIXME: better check
1585
filename = os.path.normpath(filename)
1586
while filename != '':
1587
head, tail = os.path.split(filename)
1588
## mutter('check %r for control file' % ((head, tail), ))
1589
if tail == bzrlib.BZRDIR:
1591
if filename == head:
1598
def gen_file_id(name):
1599
"""Return new file id.
1601
This should probably generate proper UUIDs, but for the moment we
1602
cope with just randomness because running uuidgen every time is
1605
from binascii import hexlify
1606
from time import time
1608
# get last component
1609
idx = name.rfind('/')
1611
name = name[idx+1 : ]
1612
idx = name.rfind('\\')
1614
name = name[idx+1 : ]
1616
# make it not a hidden file
1617
name = name.lstrip('.')
1619
# remove any wierd characters; we don't escape them but rather
1620
# just pull them out
1621
name = re.sub(r'[^\w.]', '', name)
1623
s = hexlify(rand_bytes(8))
1624
return '-'.join((name, compact_date(time()), s))
1628
"""Return a new tree-root file id."""
1629
return gen_file_id('TREE_ROOT')
1632
def copy_branch(branch_from, to_location, revision=None):
1633
"""Copy branch_from into the existing directory to_location.
1636
If not None, only revisions up to this point will be copied.
1637
The head of the new branch will be that revision.
1640
The name of a local directory that exists but is empty.
1642
from bzrlib.merge import merge
1644
assert isinstance(branch_from, Branch)
1645
assert isinstance(to_location, basestring)
1647
br_to = Branch(to_location, init=True)
1648
br_to.set_root_id(branch_from.get_root_id())
1649
if revision is None:
1650
revno = branch_from.revno()
1652
revno, rev_id = branch_from.get_revision_info(revision)
1653
br_to.update_revisions(branch_from, stop_revision=revno)
1654
merge((to_location, -1), (to_location, 0), this_dir=to_location,
1655
check_clean=False, ignore_zero=True)
1656
br_to.set_parent(branch_from.base)
1659
def _trim_namespace(namespace, spec):
1660
full_namespace = namespace + ':'
1661
assert spec.startswith(full_namespace)
1662
return spec[len(full_namespace):]