20
from cStringIO import StringIO
22
23
from bzrlib.trace import mutter, note
23
24
from bzrlib.osutils import isdir, quotefn, compact_date, rand_bytes, \
25
26
sha_file, appendpath, file_kind
26
from bzrlib.errors import BzrError, InvalidRevisionNumber, InvalidRevisionId
28
from bzrlib.errors import (BzrError, InvalidRevisionNumber, InvalidRevisionId,
29
NoSuchRevision, HistoryMissing, NotBranchError,
28
31
from bzrlib.textui import show_status
29
from bzrlib.revision import Revision
30
from bzrlib.xml import unpack_xml
32
from bzrlib.revision import Revision, validate_revision_id
31
33
from bzrlib.delta import compare_trees
32
34
from bzrlib.tree import EmptyTree, RevisionTree
34
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
35
from bzrlib.inventory import Inventory
36
from bzrlib.weavestore import WeaveStore
37
from bzrlib.store import ImmutableStore
42
BZR_BRANCH_FORMAT_4 = "Bazaar-NG branch, format 0.0.4\n"
43
BZR_BRANCH_FORMAT_5 = "Bazaar-NG branch, format 5\n"
35
44
## TODO: Maybe include checks for common corruption of newlines, etc?
38
47
# TODO: Some operations like log might retrieve the same revisions
39
48
# repeatedly to calculate deltas. We could perhaps have a weakref
40
# cache in memory to make this faster.
49
# cache in memory to make this faster. In general anything can be
50
# cached in memory between lock and unlock operations.
52
# TODO: please move the revision-string syntax stuff out of the branch
53
# object; it's clutter
43
56
def find_branch(f, **args):
184
206
self.base = os.path.realpath(base)
185
207
if not isdir(self.controlfilename('.')):
186
from errors import NotBranchError
187
raise NotBranchError("not a bzr branch: %s" % quotefn(base),
188
['use "bzr init" to initialize a new working tree',
189
'current bzr can only operate from top-of-tree'])
192
self.text_store = ImmutableStore(self.controlfilename('text-store'))
193
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
194
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
208
raise NotBranchError('not a bzr branch: %s' % quotefn(base),
209
['use "bzr init" to initialize a '
211
self._check_format(relax_version_check)
212
cfn = self.controlfilename
213
if self._branch_format == 4:
214
self.inventory_store = ImmutableStore(cfn('inventory-store'))
215
self.text_store = ImmutableStore(cfn('text-store'))
216
elif self._branch_format == 5:
217
self.control_weaves = WeaveStore(cfn([]))
218
self.weave_store = WeaveStore(cfn('weaves'))
219
self.revision_store = ImmutableStore(cfn('revision-store'))
197
222
def __str__(self):
297
313
raise BzrError("invalid controlfile mode %r" % mode)
301
315
def _make_control(self):
302
from bzrlib.inventory import Inventory
303
from bzrlib.xml import pack_xml
305
316
os.mkdir(self.controlfilename([]))
306
317
self.controlfile('README', 'w').write(
307
318
"This is a Bazaar-NG control directory.\n"
308
319
"Do not change any files in this directory.\n")
309
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
310
for d in ('text-store', 'inventory-store', 'revision-store'):
320
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT_5)
321
for d in ('text-store', 'revision-store',
311
323
os.mkdir(self.controlfilename(d))
312
for f in ('revision-history', 'merged-patches',
313
'pending-merged-patches', 'branch-name',
324
for f in ('revision-history',
315
327
'pending-merges'):
316
328
self.controlfile(f, 'w').write('')
317
329
mutter('created control directory in ' + self.base)
319
pack_xml(Inventory(gen_root_id()), self.controlfile('inventory','w'))
322
def _check_format(self):
331
# if we want per-tree root ids then this is the place to set
332
# them; they're not needed for now and so ommitted for
334
f = self.controlfile('inventory','w')
335
bzrlib.xml5.serializer_v5.write_inventory(Inventory(), f)
339
def _check_format(self, relax_version_check):
323
340
"""Check this branch format is supported.
325
The current tool only supports the current unstable format.
342
The format level is stored, as an integer, in
343
self._branch_format for code that needs to check it later.
327
345
In the future, we might need different in-memory Branch
328
346
classes to support downlevel branches. But not yet.
330
# This ignores newlines so that we can open branches created
331
# on Windows from Linux and so on. I think it might be better
332
# to always make all internal files in unix format.
333
348
fmt = self.controlfile('branch-format', 'r').read()
334
fmt.replace('\r\n', '')
335
if fmt != BZR_BRANCH_FORMAT:
336
raise BzrError('sorry, branch format %r not supported' % fmt,
337
['use a different bzr version',
338
'or remove the .bzr directory and "bzr init" again'])
349
if fmt == BZR_BRANCH_FORMAT_5:
350
self._branch_format = 5
351
elif fmt == BZR_BRANCH_FORMAT_4:
352
self._branch_format = 4
354
if (not relax_version_check
355
and self._branch_format != 5):
356
raise BzrError('sorry, branch format "%s" not supported; '
357
'use a different bzr version, '
358
'or run "bzr upgrade"'
359
% fmt.rstrip('\n\r'))
340
362
def get_root_id(self):
341
363
"""Return the id of this branches root"""
595
610
return self.revision_store[revision_id]
596
611
except IndexError:
597
raise bzrlib.errors.NoSuchRevision(revision_id)
612
raise bzrlib.errors.NoSuchRevision(self, revision_id)
617
def get_revision_xml(self, revision_id):
618
return self.get_revision_xml_file(revision_id).read()
602
621
def get_revision(self, revision_id):
603
622
"""Return the Revision object for a named revision"""
604
xml_file = self.get_revision_xml(revision_id)
623
xml_file = self.get_revision_xml_file(revision_id)
607
r = unpack_xml(Revision, xml_file)
626
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
608
627
except SyntaxError, e:
609
628
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
640
659
def get_revision_sha1(self, revision_id):
641
660
"""Hash the stored value of a revision, and return it."""
642
# In the future, revision entries will be signed. At that
643
# point, it is probably best *not* to include the signature
644
# in the revision hash. Because that lets you re-sign
645
# the revision, (add signatures/remove signatures) and still
646
# have all hash pointers stay consistent.
647
# But for now, just hash the contents.
648
return bzrlib.osutils.sha_file(self.get_revision_xml(revision_id))
651
def get_inventory(self, inventory_id):
652
"""Get Inventory object by hash.
654
TODO: Perhaps for this and similar methods, take a revision
655
parameter which can be either an integer revno or a
657
from bzrlib.inventory import Inventory
658
from bzrlib.xml import unpack_xml
660
return unpack_xml(Inventory, self.inventory_store[inventory_id])
663
def get_inventory_sha1(self, inventory_id):
661
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
664
def _get_ancestry_weave(self):
665
return self.control_weaves.get_weave('ancestry')
668
def get_ancestry(self, revision_id):
669
"""Return a list of revision-ids integrated by a revision.
672
w = self._get_ancestry_weave()
673
return [l[:-1] for l in w.get_iter(w.lookup(revision_id))]
676
def get_inventory_weave(self):
677
return self.control_weaves.get_weave('inventory')
680
def get_inventory(self, revision_id):
681
"""Get Inventory object by hash."""
682
# FIXME: The text gets passed around a lot coming from the weave.
683
f = StringIO(self.get_inventory_xml(revision_id))
684
return bzrlib.xml5.serializer_v5.read_inventory(f)
687
def get_inventory_xml(self, revision_id):
688
"""Get inventory XML as a file object."""
690
assert isinstance(revision_id, basestring), type(revision_id)
691
iw = self.get_inventory_weave()
692
return iw.get_text(iw.lookup(revision_id))
694
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
697
def get_inventory_sha1(self, revision_id):
664
698
"""Return the sha1 hash of the inventory entry
666
return sha_file(self.inventory_store[inventory_id])
700
return self.get_revision(revision_id).inventory_sha1
669
703
def get_revision_inventory(self, revision_id):
670
704
"""Return inventory of a past revision."""
671
# bzr 0.0.6 imposes the constraint that the inventory_id
705
# bzr 0.0.6 and later imposes the constraint that the inventory_id
672
706
# must be the same as its revision, so this is trivial.
673
707
if revision_id == None:
674
from bzrlib.inventory import Inventory
675
708
return Inventory(self.get_root_id())
677
710
return self.get_inventory(revision_id)
680
713
def revision_history(self):
681
"""Return sequence of revision hashes on to this branch.
683
>>> ScratchBranch().revision_history()
714
"""Return sequence of revision hashes on to this branch."""
688
717
return [l.rstrip('\r\n') for l in
697
726
>>> sb = ScratchBranch(files=['foo', 'foo~'])
698
727
>>> sb.common_ancestor(sb) == (None, None)
700
>>> commit.commit(sb, "Committing first revision", verbose=False)
729
>>> commit.commit(sb, "Committing first revision")
701
730
>>> sb.common_ancestor(sb)[0]
703
732
>>> clone = sb.clone()
704
>>> commit.commit(sb, "Committing second revision", verbose=False)
733
>>> commit.commit(sb, "Committing second revision")
705
734
>>> sb.common_ancestor(sb)[0]
707
736
>>> sb.common_ancestor(clone)[0]
709
>>> commit.commit(clone, "Committing divergent second revision",
738
>>> commit.commit(clone, "Committing divergent second revision")
711
739
>>> sb.common_ancestor(clone)[0]
713
741
>>> sb.common_ancestor(clone) == clone.common_ancestor(sb)
794
826
if stop_revision is None:
795
827
stop_revision = other_len
796
elif stop_revision > other_len:
797
raise NoSuchRevision(self, stop_revision)
829
assert isinstance(stop_revision, int)
830
if stop_revision > other_len:
831
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
799
833
return other_history[self_len:stop_revision]
802
def update_revisions(self, other, stop_revision=None):
803
"""Pull in all new revisions from other branch.
805
>>> from bzrlib.commit import commit
806
>>> bzrlib.trace.silent = True
807
>>> br1 = ScratchBranch(files=['foo', 'bar'])
810
>>> commit(br1, "lala!", rev_id="REVISION-ID-1", verbose=False)
811
>>> br2 = ScratchBranch()
812
>>> br2.update_revisions(br1)
816
>>> br2.revision_history()
818
>>> br2.update_revisions(br1)
822
>>> br1.text_store.total_size() == br2.text_store.total_size()
836
def update_revisions(self, other, stop_revno=None):
837
"""Pull in new perfect-fit revisions.
825
from bzrlib.progress import ProgressBar
829
pb.update('comparing histories')
830
revision_ids = self.missing_revisions(other, stop_revision)
832
if hasattr(other.revision_store, "prefetch"):
833
other.revision_store.prefetch(revision_ids)
834
if hasattr(other.inventory_store, "prefetch"):
835
inventory_ids = [other.get_revision(r).inventory_id
836
for r in revision_ids]
837
other.inventory_store.prefetch(inventory_ids)
842
for rev_id in revision_ids:
844
pb.update('fetching revision', i, len(revision_ids))
845
rev = other.get_revision(rev_id)
846
revisions.append(rev)
847
inv = other.get_inventory(str(rev.inventory_id))
848
for key, entry in inv.iter_entries():
849
if entry.text_id is None:
851
if entry.text_id not in self.text_store:
852
needed_texts.add(entry.text_id)
856
count = self.text_store.copy_multi(other.text_store, needed_texts)
857
print "Added %d texts." % count
858
inventory_ids = [ f.inventory_id for f in revisions ]
859
count = self.inventory_store.copy_multi(other.inventory_store,
861
print "Added %d inventories." % count
862
revision_ids = [ f.revision_id for f in revisions]
863
count = self.revision_store.copy_multi(other.revision_store,
865
for revision_id in revision_ids:
866
self.append_revision(revision_id)
867
print "Added %d revisions." % count
839
from bzrlib.fetch import greedy_fetch
842
stop_revision = other.lookup_revision(stop_revno)
845
greedy_fetch(to_branch=self, from_branch=other,
846
revision=stop_revision)
848
pullable_revs = self.missing_revisions(other, stop_revision)
851
greedy_fetch(to_branch=self,
853
revision=pullable_revs[-1])
854
self.append_revision(*pullable_revs)
870
857
def commit(self, *args, **kw):
871
from bzrlib.commit import commit
872
commit(self, *args, **kw)
858
from bzrlib.commit import Commit
859
Commit().commit(self, *args, **kw)
875
862
def lookup_revision(self, revision):
876
863
"""Return the revision identifier for a given revision information."""
877
revno, info = self.get_revision_info(revision)
864
revno, info = self._get_revision_info(revision)
868
def revision_id_to_revno(self, revision_id):
869
"""Given a revision id, return its revno"""
870
history = self.revision_history()
872
return history.index(revision_id) + 1
874
raise bzrlib.errors.NoSuchRevision(self, revision_id)
880
877
def get_revision_info(self, revision):
881
878
"""Return (revno, revision id) for revision identifier.
885
882
revision can also be a string, in which case it is parsed for something like
886
883
'date:' or 'revid:' etc.
885
revno, rev_id = self._get_revision_info(revision)
887
raise bzrlib.errors.NoSuchRevision(self, revision)
890
def get_rev_id(self, revno, history=None):
891
"""Find the revision id of the specified revno."""
895
history = self.revision_history()
896
elif revno <= 0 or revno > len(history):
897
raise bzrlib.errors.NoSuchRevision(self, revno)
898
return history[revno - 1]
900
def _get_revision_info(self, revision):
901
"""Return (revno, revision id) for revision specifier.
903
revision can be an integer, in which case it is assumed to be revno
904
(though this will translate negative values into positive ones)
905
revision can also be a string, in which case it is parsed for something
906
like 'date:' or 'revid:' etc.
908
A revid is always returned. If it is None, the specifier referred to
909
the null revision. If the revid does not occur in the revision
910
history, revno will be None.
888
913
if revision is None:
895
920
revs = self.revision_history()
896
921
if isinstance(revision, int):
899
# Mabye we should do this first, but we don't need it if revision == 0
901
923
revno = len(revs) + revision + 1
926
rev_id = self.get_rev_id(revno, revs)
904
927
elif isinstance(revision, basestring):
905
928
for prefix, func in Branch.REVISION_NAMESPACES.iteritems():
906
929
if revision.startswith(prefix):
907
revno = func(self, revs, revision)
930
result = func(self, revs, revision)
932
revno, rev_id = result
935
rev_id = self.get_rev_id(revno, revs)
910
raise BzrError('No namespace registered for string: %r' % revision)
938
raise BzrError('No namespace registered for string: %r' %
941
raise TypeError('Unhandled revision type %s' % revision)
912
if revno is None or revno <= 0 or revno > len(revs):
913
raise BzrError("no such revision %s" % revision)
914
return revno, revs[revno-1]
945
raise bzrlib.errors.NoSuchRevision(self, revision)
916
948
def _namespace_revno(self, revs, revision):
917
949
"""Lookup a revision by revision number"""
918
950
assert revision.startswith('revno:')
920
return int(revision[6:])
952
return (int(revision[6:]),)
921
953
except ValueError:
923
955
REVISION_NAMESPACES['revno:'] = _namespace_revno
925
957
def _namespace_revid(self, revs, revision):
926
958
assert revision.startswith('revid:')
959
rev_id = revision[len('revid:'):]
928
return revs.index(revision[6:]) + 1
961
return revs.index(rev_id) + 1, rev_id
929
962
except ValueError:
931
964
REVISION_NAMESPACES['revid:'] = _namespace_revid
933
966
def _namespace_last(self, revs, revision):
1298
def get_parent(self):
1299
"""Return the parent location of the branch.
1301
This is the default location for push/pull/missing. The usual
1302
pattern is that the user can override it by specifying a
1306
_locs = ['parent', 'pull', 'x-pull']
1309
return self.controlfile(l, 'r').read().strip('\n')
1311
if e.errno != errno.ENOENT:
1316
def set_parent(self, url):
1317
# TODO: Maybe delete old location files?
1318
from bzrlib.atomicfile import AtomicFile
1321
f = AtomicFile(self.controlfilename('parent'))
1330
def check_revno(self, revno):
1332
Check whether a revno corresponds to any revision.
1333
Zero (the NULL revision) is considered valid.
1336
self.check_real_revno(revno)
1338
def check_real_revno(self, revno):
1340
Check whether a revno corresponds to a real revision.
1341
Zero (the NULL revision) is considered invalid
1343
if revno < 1 or revno > self.revno():
1344
raise InvalidRevisionNumber(revno)
1268
1349
class ScratchBranch(Branch):
1269
1350
"""Special test class: a branch that cleans up after itself.
1386
1469
"""Return a new tree-root file id."""
1387
1470
return gen_file_id('TREE_ROOT')
1473
def pull_loc(branch):
1474
# TODO: Should perhaps just make attribute be 'base' in
1475
# RemoteBranch and Branch?
1476
if hasattr(branch, "baseurl"):
1477
return branch.baseurl
1482
def copy_branch(branch_from, to_location, revision=None):
1483
"""Copy branch_from into the existing directory to_location.
1486
If not None, only revisions up to this point will be copied.
1487
The head of the new branch will be that revision. Can be a
1491
The name of a local directory that exists but is empty.
1493
# TODO: This could be done *much* more efficiently by just copying
1494
# all the whole weaves and revisions, rather than getting one
1495
# revision at a time.
1496
from bzrlib.merge import merge
1497
from bzrlib.branch import Branch
1499
assert isinstance(branch_from, Branch)
1500
assert isinstance(to_location, basestring)
1502
br_to = Branch(to_location, init=True)
1503
br_to.set_root_id(branch_from.get_root_id())
1504
if revision is None:
1507
revno, rev_id = branch_from.get_revision_info(revision)
1508
br_to.update_revisions(branch_from, stop_revno=revno)
1509
merge((to_location, -1), (to_location, 0), this_dir=to_location,
1510
check_clean=False, ignore_zero=True)
1512
from_location = pull_loc(branch_from)
1513
br_to.set_parent(pull_loc(branch_from))