69
71
from copy import copy
70
72
from cStringIO import StringIO
73
from bzrlib.lazy_import import lazy_import
74
lazy_import(globals(), """
75
from bzrlib import tsort
77
78
from bzrlib import (
81
from bzrlib.trace import mutter
81
82
from bzrlib.errors import (WeaveError, WeaveFormatError, WeaveParentMismatch,
82
83
RevisionAlreadyPresent,
83
84
RevisionNotPresent,
84
UnavailableRepresentation,
85
WeaveRevisionAlreadyPresent,
86
WeaveRevisionNotPresent,
86
from bzrlib.osutils import dirname, sha, sha_strings, split_lines
88
import bzrlib.errors as errors
89
from bzrlib.osutils import sha_strings
87
90
import bzrlib.patiencediff
88
from bzrlib.revision import NULL_REVISION
89
from bzrlib.symbol_versioning import *
90
from bzrlib.trace import mutter
91
from bzrlib.versionedfile import (
91
from bzrlib.tsort import topo_sort
92
from bzrlib.versionedfile import VersionedFile, InterVersionedFile
98
93
from bzrlib.weavefile import _read_weave_v5, write_weave_v5
101
class WeaveContentFactory(ContentFactory):
102
"""Content factory for streaming from weaves.
104
:seealso ContentFactory:
107
def __init__(self, version, weave):
108
"""Create a WeaveContentFactory for version from weave."""
109
ContentFactory.__init__(self)
110
self.sha1 = weave.get_sha1s([version])[version]
111
self.key = (version,)
112
parents = weave.get_parent_map([version])[version]
113
self.parents = tuple((parent,) for parent in parents)
114
self.storage_kind = 'fulltext'
117
def get_bytes_as(self, storage_kind):
118
if storage_kind == 'fulltext':
119
return self._weave.get_text(self.key[-1])
120
elif storage_kind == 'chunked':
121
return self._weave.get_lines(self.key[-1])
123
raise UnavailableRepresentation(self.key, storage_kind, 'fulltext')
126
96
class Weave(VersionedFile):
127
97
"""weave - versioned text file storage.
129
99
A Weave manages versions of line-based text files, keeping track
130
100
of the originating version for each line.
297
246
__contains__ = has_version
299
def get_record_stream(self, versions, ordering, include_delta_closure):
300
"""Get a stream of records for versions.
302
:param versions: The versions to include. Each version is a tuple
304
:param ordering: Either 'unordered' or 'topological'. A topologically
305
sorted stream has compression parents strictly before their
307
:param include_delta_closure: If True then the closure across any
308
compression parents will be included (in the opaque data).
309
:return: An iterator of ContentFactory objects, each of which is only
310
valid until the iterator is advanced.
312
versions = [version[-1] for version in versions]
313
if ordering == 'topological':
314
parents = self.get_parent_map(versions)
315
new_versions = tsort.topo_sort(parents)
316
new_versions.extend(set(versions).difference(set(parents)))
317
versions = new_versions
318
elif ordering == 'groupcompress':
319
parents = self.get_parent_map(versions)
320
new_versions = sort_groupcompress(parents)
321
new_versions.extend(set(versions).difference(set(parents)))
322
versions = new_versions
323
for version in versions:
325
yield WeaveContentFactory(version, self)
327
yield AbsentContentFactory((version,))
329
248
def get_parent_map(self, version_ids):
330
249
"""See VersionedFile.get_parent_map."""
332
251
for version_id in version_ids:
333
if version_id == NULL_REVISION:
338
map(self._idx_to_name,
339
self._parents[self._lookup(version_id)]))
340
except RevisionNotPresent:
342
result[version_id] = parents
253
result[version_id] = tuple(
254
map(self._idx_to_name, self._parents[self._lookup(version_id)]))
255
except RevisionNotPresent:
345
259
def get_parents_with_ghosts(self, version_id):
346
260
raise NotImplementedError(self.get_parents_with_ghosts)
348
def insert_record_stream(self, stream):
349
"""Insert a record stream into this versioned file.
351
:param stream: A stream of records to insert.
353
:seealso VersionedFile.get_record_stream:
356
for record in stream:
357
# Raise an error when a record is missing.
358
if record.storage_kind == 'absent':
359
raise RevisionNotPresent([record.key[0]], self)
360
# adapt to non-tuple interface
361
parents = [parent[0] for parent in record.parents]
362
if (record.storage_kind == 'fulltext'
363
or record.storage_kind == 'chunked'):
364
self.add_lines(record.key[0], parents,
365
osutils.chunks_to_lines(record.get_bytes_as('chunked')))
367
adapter_key = record.storage_kind, 'fulltext'
369
adapter = adapters[adapter_key]
371
adapter_factory = adapter_registry.get(adapter_key)
372
adapter = adapter_factory(self)
373
adapters[adapter_key] = adapter
374
lines = split_lines(adapter.get_bytes(record))
376
self.add_lines(record.key[0], parents, lines)
377
except RevisionAlreadyPresent:
380
262
def _check_repeated_add(self, name, parents, text, sha1):
381
263
"""Check that a duplicated add is OK.
398
280
def _add(self, version_id, lines, parents, sha1=None, nostore_sha=None):
399
281
"""Add a single text on top of the weave.
401
283
Returns the index number of the newly added version.
404
286
Symbolic name for this version.
405
287
(Typically the revision-id of the revision that added it.)
406
If None, a name will be allocated based on the hash. (sha1:SHAHASH)
409
290
List or set of direct parent version numbers.
412
293
Sequence of lines to be added in the new version.
414
295
:param nostore_sha: See VersionedFile.add_lines.
297
assert isinstance(version_id, basestring)
416
298
self._check_lines_not_unicode(lines)
417
299
self._check_lines_are_lines(lines)
419
301
sha1 = sha_strings(lines)
420
302
if sha1 == nostore_sha:
421
303
raise errors.ExistingContent
422
if version_id is None:
423
version_id = "sha1:" + sha1
424
304
if version_id in self._name_map:
425
305
return self._check_repeated_add(version_id, parents, lines, sha1)
557
448
def _compatible_parents(self, my_parents, other_parents):
558
449
"""During join check that other_parents are joinable with my_parents.
560
Joinable is defined as 'is a subset of' - supersets may require
451
Joinable is defined as 'is a subset of' - supersets may require
561
452
regeneration of diffs, but subsets do not.
563
454
return len(other_parents.difference(my_parents)) == 0
565
def annotate(self, version_id):
566
"""Return a list of (version-id, line) tuples for version_id.
456
def annotate_iter(self, version_id):
457
"""Yield list of (version-id, line) pairs for the specified version.
568
459
The index indicates when the line originated in the weave."""
569
460
incls = [self._lookup(version_id)]
570
return [(self._idx_to_name(origin), text) for origin, lineno, text in
571
self._extract(incls)]
461
for origin, lineno, text in self._extract(incls):
462
yield self._idx_to_name(origin), text
573
464
def iter_lines_added_or_present_in_versions(self, version_ids=None,
699
598
# we're still spending ~1/4 of the method in isinstance though.
700
599
# so lets hard code the acceptable string classes we expect:
701
600
# 449 0 1202.9420 786.2930 bzrlib.weave:556(_extract)
702
# +71352 0 377.5560 377.5560 +<method 'append' of 'list'
601
# +71352 0 377.5560 377.5560 +<method 'append' of 'list'
704
603
# yay, down to ~1/4 the initial extract time, and our inline time
705
604
# has shrunk again, with isinstance no longer dominating.
706
605
# tweaking the stack inclusion test to use a set gives:
707
606
# 449 0 1122.8030 713.0080 bzrlib.weave:556(_extract)
708
# +71352 0 354.9980 354.9980 +<method 'append' of 'list'
607
# +71352 0 354.9980 354.9980 +<method 'append' of 'list'
710
609
# - a 5% win, or possibly just noise. However with large istacks that
711
610
# 'in' test could dominate, so I'm leaving this change in place -
712
611
# when its fast enough to consider profiling big datasets we can review.
717
616
for l in self._weave:
718
617
if l.__class__ == tuple:
725
625
iset.remove(istack.pop())
727
627
if v in included:
730
632
if v in included:
733
raise AssertionError()
636
assert l.__class__ in (str, unicode)
735
637
if isactive is None:
736
638
isactive = (not dset) and istack and (istack[-1] in included)
848
750
# no lines outside of insertion blocks, that deletions are
849
751
# properly paired, etc.
753
def _join(self, other, pb, msg, version_ids, ignore_missing):
754
"""Worker routine for join()."""
755
if not other.versions():
756
return # nothing to update, easy
759
# versions is never none, InterWeave checks this.
762
# two loops so that we do not change ourselves before verifying it
764
# work through in index order to make sure we get all dependencies
767
# get the selected versions only that are in other.versions.
768
version_ids = set(other.versions()).intersection(set(version_ids))
769
# pull in the referenced graph.
770
version_ids = other.get_ancestry(version_ids)
771
pending_parents = other.get_parent_map(version_ids)
772
pending_graph = pending_parents.items()
773
if len(pending_graph) != len(version_ids):
774
raise RevisionNotPresent(
775
set(version_ids) - set(pending_parents.keys()), self)
776
for name in topo_sort(pending_graph):
777
other_idx = other._name_map[name]
778
# returns True if we have it, False if we need it.
779
if not self._check_version_consistent(other, other_idx, name):
780
names_to_join.append((other_idx, name))
788
for other_idx, name in names_to_join:
789
# TODO: If all the parents of the other version are already
790
# present then we can avoid some work by just taking the delta
791
# and adjusting the offsets.
792
new_parents = self._imported_parents(other, other_idx)
793
sha1 = other._sha1s[other_idx]
798
pb.update(msg, merged, len(names_to_join))
800
lines = other.get_lines(other_idx)
801
self._add(name, lines, new_parents, sha1)
803
mutter("merged = %d, processed = %d, file_id=%s; deltat=%d"%(
804
merged, processed, self._weave_name, time.time()-time0))
851
806
def _imported_parents(self, other, other_idx):
852
807
"""Return list of parents in self corresponding to indexes in other."""
946
905
transport.put_file(name + WeaveFile.WEAVE_SUFFIX, sio, self._filemode)
907
def create_empty(self, name, transport, filemode=None):
908
return WeaveFile(name, transport, filemode, create=True)
949
911
"""Save the weave."""
950
912
self._check_write_ok()
952
914
write_weave_v5(self, sio)
954
bytes = sio.getvalue()
955
path = self._weave_name + WeaveFile.WEAVE_SUFFIX
957
self._transport.put_bytes(path, bytes, self._filemode)
958
except errors.NoSuchFile:
959
self._transport.mkdir(dirname(path))
960
self._transport.put_bytes(path, bytes, self._filemode)
916
self._transport.put_file(self._weave_name + WeaveFile.WEAVE_SUFFIX,
963
921
def get_suffixes():
964
922
"""See VersionedFile.get_suffixes()."""
965
923
return [WeaveFile.WEAVE_SUFFIX]
967
def insert_record_stream(self, stream):
968
super(WeaveFile, self).insert_record_stream(stream)
925
def join(self, other, pb=None, msg=None, version_ids=None,
926
ignore_missing=False):
927
"""Join other into self and save."""
928
super(WeaveFile, self).join(other, pb, msg, version_ids, ignore_missing)
972
932
def _reweave(wa, wb, pb=None, msg=None):
973
933
"""Combine two weaves and return the result.
975
This works even if a revision R has different parents in
935
This works even if a revision R has different parents in
976
936
wa and wb. In the resulting weave all the parents are given.
978
This is done by just building up a new weave, maintaining ordering
938
This is done by just building up a new weave, maintaining ordering
979
939
of the versions in the two inputs. More efficient approaches
980
might be possible but it should only be necessary to do
981
this operation rarely, when a new previously ghost version is
940
might be possible but it should only be necessary to do
941
this operation rarely, when a new previously ghost version is
984
944
:param pb: An optional progress bar, indicating how far done we are
1029
988
p = combined.setdefault(name, set())
1030
989
p.update(map(weave._idx_to_name, weave._parents[idx]))
994
"""Show the weave's table-of-contents"""
995
print '%6s %50s %10s %10s' % ('ver', 'name', 'sha1', 'parents')
996
for i in (6, 50, 10, 10):
999
for i in range(w.num_versions()):
1002
parent_str = ' '.join(map(str, w._parents[i]))
1003
print '%6d %-50.50s %10.10s %s' % (i, name, sha1, parent_str)
1007
def weave_stats(weave_file, pb):
1008
from bzrlib.weavefile import read_weave
1010
wf = file(weave_file, 'rb')
1012
# FIXME: doesn't work on pipes
1013
weave_size = wf.tell()
1017
for i in range(vers):
1018
pb.update('checking sizes', i, vers)
1019
for origin, lineno, line in w._extract([i]):
1024
print 'versions %9d' % vers
1025
print 'weave file %9d bytes' % weave_size
1026
print 'total contents %9d bytes' % total
1027
print 'compression ratio %9.2fx' % (float(total) / float(weave_size))
1030
print 'average size %9d bytes' % avg
1031
print 'relative size %9.2fx' % (float(weave_size) / float(avg))
1035
print """bzr weave tool
1037
Experimental tool for weave algorithm.
1040
weave init WEAVEFILE
1041
Create an empty weave file
1042
weave get WEAVEFILE VERSION
1043
Write out specified version.
1044
weave check WEAVEFILE
1045
Check consistency of all versions.
1047
Display table of contents.
1048
weave add WEAVEFILE NAME [BASE...] < NEWTEXT
1049
Add NEWTEXT, with specified parent versions.
1050
weave annotate WEAVEFILE VERSION
1051
Display origin of each line.
1052
weave merge WEAVEFILE VERSION1 VERSION2 > OUT
1053
Auto-merge two versions and display conflicts.
1054
weave diff WEAVEFILE VERSION1 VERSION2
1055
Show differences between two versions.
1059
% weave init foo.weave
1061
% weave add foo.weave ver0 < foo.txt
1064
(create updated version)
1066
% weave get foo.weave 0 | diff -u - foo.txt
1067
% weave add foo.weave ver1 0 < foo.txt
1070
% weave get foo.weave 0 > foo.txt (create forked version)
1072
% weave add foo.weave ver2 0 < foo.txt
1075
% weave merge foo.weave 1 2 > foo.txt (merge them)
1076
% vi foo.txt (resolve conflicts)
1077
% weave add foo.weave merged 1 2 < foo.txt (commit merged version)
1089
# in case we're run directly from the subdirectory
1090
sys.path.append('..')
1092
from bzrlib.weavefile import write_weave, read_weave
1093
from bzrlib.progress import ProgressBar
1108
return read_weave(file(argv[2], 'rb'))
1114
# at the moment, based on everything in the file
1116
parents = map(int, argv[4:])
1117
lines = sys.stdin.readlines()
1118
ver = w.add(name, parents, lines)
1119
write_weave(w, file(argv[2], 'wb'))
1120
print 'added version %r %d' % (name, ver)
1123
if os.path.exists(fn):
1124
raise IOError("file exists")
1126
write_weave(w, file(fn, 'wb'))
1127
elif cmd == 'get': # get one version
1129
sys.stdout.writelines(w.get_iter(int(argv[3])))
1134
v1, v2 = map(int, argv[3:5])
1137
diff_gen = bzrlib.patiencediff.unified_diff(lines1, lines2,
1138
'%s version %d' % (fn, v1),
1139
'%s version %d' % (fn, v2))
1140
sys.stdout.writelines(diff_gen)
1142
elif cmd == 'annotate':
1144
# newline is added to all lines regardless; too hard to get
1145
# reasonable formatting otherwise
1147
for origin, text in w.annotate(int(argv[3])):
1148
text = text.rstrip('\r\n')
1150
print ' | %s' % (text)
1152
print '%5d | %s' % (origin, text)
1158
elif cmd == 'stats':
1159
weave_stats(argv[2], ProgressBar())
1161
elif cmd == 'check':
1166
print '%d versions ok' % w.num_versions()
1168
elif cmd == 'inclusions':
1170
print ' '.join(map(str, w.inclusions([int(argv[3])])))
1172
elif cmd == 'parents':
1174
print ' '.join(map(str, w._parents[int(argv[3])]))
1176
elif cmd == 'plan-merge':
1177
# replaced by 'bzr weave-plan-merge'
1179
for state, line in w.plan_merge(int(argv[3]), int(argv[4])):
1181
print '%14s | %s' % (state, line),
1182
elif cmd == 'merge':
1183
# replaced by 'bzr weave-merge-text'
1185
p = w.plan_merge(int(argv[3]), int(argv[4]))
1186
sys.stdout.writelines(w.weave_merge(p))
1188
raise ValueError('unknown command %r' % cmd)
1191
if __name__ == '__main__':
1193
sys.exit(main(sys.argv))
1196
class InterWeave(InterVersionedFile):
1197
"""Optimised code paths for weave to weave operations."""
1199
_matching_file_from_factory = staticmethod(WeaveFile)
1200
_matching_file_to_factory = staticmethod(WeaveFile)
1203
def is_compatible(source, target):
1204
"""Be compatible with weaves."""
1206
return (isinstance(source, Weave) and
1207
isinstance(target, Weave))
1208
except AttributeError:
1211
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
1212
"""See InterVersionedFile.join."""
1213
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
1214
if self.target.versions() == [] and version_ids is None:
1215
self.target._copy_weave_content(self.source)
1217
self.target._join(self.source, pb, msg, version_ids, ignore_missing)
1220
InterVersionedFile.register_optimiser(InterWeave)