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)
28
30
from bzrlib.textui import show_status
29
from bzrlib.revision import Revision
30
from bzrlib.xml import unpack_xml
31
from bzrlib.revision import Revision, validate_revision_id
31
32
from bzrlib.delta import compare_trees
32
33
from bzrlib.tree import EmptyTree, RevisionTree
34
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
34
from bzrlib.inventory import Inventory
35
from bzrlib.weavestore import WeaveStore
36
from bzrlib.store import ImmutableStore
41
INVENTORY_FILEID = '__inventory'
42
ANCESTRY_FILEID = '__ancestry'
45
BZR_BRANCH_FORMAT_4 = "Bazaar-NG branch, format 0.0.4\n"
46
BZR_BRANCH_FORMAT_5 = "Bazaar-NG branch, format 5\n"
35
47
## TODO: Maybe include checks for common corruption of newlines, etc?
38
50
# TODO: Some operations like log might retrieve the same revisions
39
51
# repeatedly to calculate deltas. We could perhaps have a weakref
40
# cache in memory to make this faster.
52
# cache in memory to make this faster. In general anything can be
53
# cached in memory between lock and unlock operations.
55
# TODO: please move the revision-string syntax stuff out of the branch
56
# object; it's clutter
43
59
def find_branch(f, **args):
297
309
raise BzrError("invalid controlfile mode %r" % mode)
301
311
def _make_control(self):
302
from bzrlib.inventory import Inventory
303
from bzrlib.xml import pack_xml
305
312
os.mkdir(self.controlfilename([]))
306
313
self.controlfile('README', 'w').write(
307
314
"This is a Bazaar-NG control directory.\n"
308
315
"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'):
316
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT_5)
317
for d in ('text-store', 'revision-store',
311
319
os.mkdir(self.controlfilename(d))
312
320
for f in ('revision-history', 'merged-patches',
313
321
'pending-merged-patches', 'branch-name',
316
324
self.controlfile(f, 'w').write('')
317
325
mutter('created control directory in ' + self.base)
319
pack_xml(Inventory(gen_root_id()), self.controlfile('inventory','w'))
327
# if we want per-tree root ids then this is the place to set
328
# them; they're not needed for now and so ommitted for
330
f = self.controlfile('inventory','w')
331
bzrlib.xml5.serializer_v5.write_inventory(Inventory(), f)
322
335
def _check_format(self):
323
336
"""Check this branch format is supported.
325
The current tool only supports the current unstable format.
338
The format level is stored, as an integer, in
339
self._branch_format for code that needs to check it later.
327
341
In the future, we might need different in-memory Branch
328
342
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
344
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'])
345
if fmt == BZR_BRANCH_FORMAT_5:
346
self._branch_format = 5
348
raise BzrError('sorry, branch format "%s" not supported; '
349
'use a different bzr version, '
350
'or run "bzr upgrade", '
351
'or remove the .bzr directory and "bzr init" again'
352
% fmt.rstrip('\n\r'))
340
354
def get_root_id(self):
341
355
"""Return the id of this branches root"""
595
602
return self.revision_store[revision_id]
596
603
except IndexError:
597
raise bzrlib.errors.NoSuchRevision(revision_id)
604
raise bzrlib.errors.NoSuchRevision(self, revision_id)
609
def get_revision_xml(self, revision_id):
610
return self.get_revision_xml_file(revision_id).read()
602
613
def get_revision(self, revision_id):
603
614
"""Return the Revision object for a named revision"""
604
xml_file = self.get_revision_xml(revision_id)
615
xml_file = self.get_revision_xml_file(revision_id)
607
r = unpack_xml(Revision, xml_file)
618
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
608
619
except SyntaxError, e:
609
620
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
640
651
def get_revision_sha1(self, revision_id):
641
652
"""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):
653
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
656
def get_ancestry(self, revision_id):
657
"""Return a list of revision-ids integrated by a revision.
659
w = self.weave_store.get_weave(ANCESTRY_FILEID)
661
return [l[:-1] for l in w.get_iter(w.lookup(revision_id))]
664
def get_inventory_weave(self):
665
return self.weave_store.get_weave(INVENTORY_FILEID)
668
def get_inventory(self, revision_id):
669
"""Get Inventory object by hash."""
670
# FIXME: The text gets passed around a lot coming from the weave.
671
f = StringIO(self.get_inventory_xml(revision_id))
672
return bzrlib.xml5.serializer_v5.read_inventory(f)
675
def get_inventory_xml(self, revision_id):
676
"""Get inventory XML as a file object."""
678
assert isinstance(revision_id, basestring), type(revision_id)
679
iw = self.get_inventory_weave()
680
return iw.get_text(iw.lookup(revision_id))
682
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
685
def get_inventory_sha1(self, revision_id):
664
686
"""Return the sha1 hash of the inventory entry
666
return sha_file(self.inventory_store[inventory_id])
688
return self.get_revision(revision_id).inventory_sha1
669
691
def get_revision_inventory(self, revision_id):
670
692
"""Return inventory of a past revision."""
671
# bzr 0.0.6 imposes the constraint that the inventory_id
693
# bzr 0.0.6 and later imposes the constraint that the inventory_id
672
694
# must be the same as its revision, so this is trivial.
673
695
if revision_id == None:
674
from bzrlib.inventory import Inventory
675
696
return Inventory(self.get_root_id())
677
698
return self.get_inventory(revision_id)
697
718
>>> sb = ScratchBranch(files=['foo', 'foo~'])
698
719
>>> sb.common_ancestor(sb) == (None, None)
700
>>> commit.commit(sb, "Committing first revision", verbose=False)
721
>>> commit.commit(sb, "Committing first revision")
701
722
>>> sb.common_ancestor(sb)[0]
703
724
>>> clone = sb.clone()
704
>>> commit.commit(sb, "Committing second revision", verbose=False)
725
>>> commit.commit(sb, "Committing second revision")
705
726
>>> sb.common_ancestor(sb)[0]
707
728
>>> sb.common_ancestor(clone)[0]
709
>>> commit.commit(clone, "Committing divergent second revision",
730
>>> commit.commit(clone, "Committing divergent second revision")
711
731
>>> sb.common_ancestor(clone)[0]
713
733
>>> sb.common_ancestor(clone) == clone.common_ancestor(sb)
794
818
if stop_revision is None:
795
819
stop_revision = other_len
796
elif stop_revision > other_len:
797
raise NoSuchRevision(self, stop_revision)
821
assert isinstance(stop_revision, int)
822
if stop_revision > other_len:
823
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
799
825
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()
828
def update_revisions(self, other, stop_revno=None):
829
"""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
831
from bzrlib.fetch import greedy_fetch
834
stop_revision = other.lookup_revision(stop_revno)
837
greedy_fetch(to_branch=self, from_branch=other,
838
revision=stop_revision)
840
pullable_revs = self.missing_revisions(other, stop_revision)
843
greedy_fetch(to_branch=self,
845
revision=pullable_revs[-1])
846
self.append_revision(*pullable_revs)
870
849
def commit(self, *args, **kw):
871
from bzrlib.commit import commit
872
commit(self, *args, **kw)
850
from bzrlib.commit import Commit
851
Commit().commit(self, *args, **kw)
875
854
def lookup_revision(self, revision):
876
855
"""Return the revision identifier for a given revision information."""
877
revno, info = self.get_revision_info(revision)
856
revno, info = self._get_revision_info(revision)
860
def revision_id_to_revno(self, revision_id):
861
"""Given a revision id, return its revno"""
862
history = self.revision_history()
864
return history.index(revision_id) + 1
866
raise bzrlib.errors.NoSuchRevision(self, revision_id)
880
869
def get_revision_info(self, revision):
881
870
"""Return (revno, revision id) for revision identifier.
885
874
revision can also be a string, in which case it is parsed for something like
886
875
'date:' or 'revid:' etc.
877
revno, rev_id = self._get_revision_info(revision)
879
raise bzrlib.errors.NoSuchRevision(self, revision)
882
def get_rev_id(self, revno, history=None):
883
"""Find the revision id of the specified revno."""
887
history = self.revision_history()
888
elif revno <= 0 or revno > len(history):
889
raise bzrlib.errors.NoSuchRevision(self, revno)
890
return history[revno - 1]
892
def _get_revision_info(self, revision):
893
"""Return (revno, revision id) for revision specifier.
895
revision can be an integer, in which case it is assumed to be revno
896
(though this will translate negative values into positive ones)
897
revision can also be a string, in which case it is parsed for something
898
like 'date:' or 'revid:' etc.
900
A revid is always returned. If it is None, the specifier referred to
901
the null revision. If the revid does not occur in the revision
902
history, revno will be None.
888
905
if revision is None:
895
912
revs = self.revision_history()
896
913
if isinstance(revision, int):
899
# Mabye we should do this first, but we don't need it if revision == 0
901
915
revno = len(revs) + revision + 1
918
rev_id = self.get_rev_id(revno, revs)
904
919
elif isinstance(revision, basestring):
905
920
for prefix, func in Branch.REVISION_NAMESPACES.iteritems():
906
921
if revision.startswith(prefix):
907
revno = func(self, revs, revision)
922
result = func(self, revs, revision)
924
revno, rev_id = result
927
rev_id = self.get_rev_id(revno, revs)
910
raise BzrError('No namespace registered for string: %r' % revision)
930
raise BzrError('No namespace registered for string: %r' %
933
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]
937
raise bzrlib.errors.NoSuchRevision(self, revision)
916
940
def _namespace_revno(self, revs, revision):
917
941
"""Lookup a revision by revision number"""
918
942
assert revision.startswith('revno:')
920
return int(revision[6:])
944
return (int(revision[6:]),)
921
945
except ValueError:
923
947
REVISION_NAMESPACES['revno:'] = _namespace_revno
925
949
def _namespace_revid(self, revs, revision):
926
950
assert revision.startswith('revid:')
951
rev_id = revision[len('revid:'):]
928
return revs.index(revision[6:]) + 1
953
return revs.index(rev_id) + 1, rev_id
929
954
except ValueError:
931
956
REVISION_NAMESPACES['revid:'] = _namespace_revid
933
958
def _namespace_last(self, revs, revision):
1290
def get_parent(self):
1291
"""Return the parent location of the branch.
1293
This is the default location for push/pull/missing. The usual
1294
pattern is that the user can override it by specifying a
1298
_locs = ['parent', 'pull', 'x-pull']
1301
return self.controlfile(l, 'r').read().strip('\n')
1303
if e.errno != errno.ENOENT:
1308
def set_parent(self, url):
1309
# TODO: Maybe delete old location files?
1310
from bzrlib.atomicfile import AtomicFile
1313
f = AtomicFile(self.controlfilename('parent'))
1322
def check_revno(self, revno):
1324
Check whether a revno corresponds to any revision.
1325
Zero (the NULL revision) is considered valid.
1328
self.check_real_revno(revno)
1330
def check_real_revno(self, revno):
1332
Check whether a revno corresponds to a real revision.
1333
Zero (the NULL revision) is considered invalid
1335
if revno < 1 or revno > self.revno():
1336
raise InvalidRevisionNumber(revno)
1268
1341
class ScratchBranch(Branch):
1269
1342
"""Special test class: a branch that cleans up after itself.
1386
1461
"""Return a new tree-root file id."""
1387
1462
return gen_file_id('TREE_ROOT')
1465
def pull_loc(branch):
1466
# TODO: Should perhaps just make attribute be 'base' in
1467
# RemoteBranch and Branch?
1468
if hasattr(branch, "baseurl"):
1469
return branch.baseurl
1474
def copy_branch(branch_from, to_location, revision=None):
1475
"""Copy branch_from into the existing directory to_location.
1478
If not None, only revisions up to this point will be copied.
1479
The head of the new branch will be that revision. Can be a
1483
The name of a local directory that exists but is empty.
1485
# TODO: This could be done *much* more efficiently by just copying
1486
# all the whole weaves and revisions, rather than getting one
1487
# revision at a time.
1488
from bzrlib.merge import merge
1489
from bzrlib.branch import Branch
1491
assert isinstance(branch_from, Branch)
1492
assert isinstance(to_location, basestring)
1494
br_to = Branch(to_location, init=True)
1495
br_to.set_root_id(branch_from.get_root_id())
1496
if revision is None:
1499
revno, rev_id = branch_from.get_revision_info(revision)
1500
br_to.update_revisions(branch_from, stop_revno=revno)
1501
merge((to_location, -1), (to_location, 0), this_dir=to_location,
1502
check_clean=False, ignore_zero=True)
1504
from_location = pull_loc(branch_from)
1505
br_to.set_parent(pull_loc(branch_from))