104
133
class KnitContent(object):
105
134
"""Content of a knit version to which deltas can be applied."""
107
def __init__(self, lines):
110
def annotate_iter(self):
111
"""Yield tuples of (origin, text) for each content line."""
112
for origin, text in self._lines:
115
136
def annotate(self):
116
137
"""Return a list of (origin, text) tuples."""
117
138
return list(self.annotate_iter())
119
140
def line_delta_iter(self, new_lines):
120
141
"""Generate line-based delta from this content to new_lines."""
121
new_texts = [text for origin, text in new_lines._lines]
122
old_texts = [text for origin, text in self._lines]
123
s = SequenceMatcher(None, old_texts, new_texts)
124
for op in s.get_opcodes():
142
new_texts = new_lines.text()
143
old_texts = self.text()
144
s = patiencediff.PatienceSequenceMatcher(None, old_texts, new_texts)
145
for tag, i1, i2, j1, j2 in s.get_opcodes():
127
# ofrom oto length data
128
yield (op[1], op[2], op[4]-op[3], new_lines._lines[op[3]:op[4]])
148
# ofrom, oto, length, data
149
yield i1, i2, j2 - j1, new_lines._lines[j1:j2]
130
151
def line_delta(self, new_lines):
131
152
return list(self.line_delta_iter(new_lines))
155
def get_line_delta_blocks(knit_delta, source, target):
156
"""Extract SequenceMatcher.get_matching_blocks() from a knit delta"""
157
target_len = len(target)
160
for s_begin, s_end, t_len, new_text in knit_delta:
161
true_n = s_begin - s_pos
164
# knit deltas do not provide reliable info about whether the
165
# last line of a file matches, due to eol handling.
166
if source[s_pos + n -1] != target[t_pos + n -1]:
169
yield s_pos, t_pos, n
170
t_pos += t_len + true_n
172
n = target_len - t_pos
174
if source[s_pos + n -1] != target[t_pos + n -1]:
177
yield s_pos, t_pos, n
178
yield s_pos + (target_len - t_pos), target_len, 0
181
class AnnotatedKnitContent(KnitContent):
182
"""Annotated content."""
184
def __init__(self, lines):
187
def annotate_iter(self):
188
"""Yield tuples of (origin, text) for each content line."""
189
return iter(self._lines)
191
def strip_last_line_newline(self):
192
line = self._lines[-1][1].rstrip('\n')
193
self._lines[-1] = (self._lines[-1][0], line)
134
196
return [text for origin, text in self._lines]
137
class _KnitFactory(object):
138
"""Base factory for creating content objects."""
140
def make(self, lines, version):
141
num_lines = len(lines)
142
return KnitContent(zip([version] * num_lines, lines))
145
class KnitAnnotateFactory(_KnitFactory):
199
return AnnotatedKnitContent(self._lines[:])
202
class PlainKnitContent(KnitContent):
203
"""Unannotated content.
205
When annotate[_iter] is called on this content, the same version is reported
206
for all lines. Generally, annotate[_iter] is not useful on PlainKnitContent
210
def __init__(self, lines, version_id):
212
self._version_id = version_id
214
def annotate_iter(self):
215
"""Yield tuples of (origin, text) for each content line."""
216
for line in self._lines:
217
yield self._version_id, line
220
return PlainKnitContent(self._lines[:], self._version_id)
222
def strip_last_line_newline(self):
223
self._lines[-1] = self._lines[-1].rstrip('\n')
229
class KnitAnnotateFactory(object):
146
230
"""Factory for creating annotated Content objects."""
150
def parse_fulltext(self, content, version):
234
def make(self, lines, version_id):
235
num_lines = len(lines)
236
return AnnotatedKnitContent(zip([version_id] * num_lines, lines))
238
def parse_fulltext(self, content, version_id):
151
239
"""Convert fulltext to internal representation
153
241
fulltext content is of the format
155
243
internal representation is of the format:
156
244
(revid, plaintext)
160
origin, text = line.split(' ', 1)
161
lines.append((origin.decode('utf-8'), text))
162
return KnitContent(lines)
246
# TODO: jam 20070209 The tests expect this to be returned as tuples,
247
# but the code itself doesn't really depend on that.
248
# Figure out a way to not require the overhead of turning the
249
# list back into tuples.
250
lines = [tuple(line.split(' ', 1)) for line in content]
251
return AnnotatedKnitContent(lines)
164
253
def parse_line_delta_iter(self, lines):
165
for result_item in self.parse_line_delta[lines]:
254
return iter(self.parse_line_delta(lines))
168
def parse_line_delta(self, lines, version):
256
def parse_line_delta(self, lines, version_id):
169
257
"""Convert a line based delta into internal representation.
171
259
line delta is in the form of:
172
260
intstart intend intcount
174
262
revid(utf8) newline\n
175
internal represnetation is
263
internal representation is
176
264
(start, end, count, [1..count tuples (revid, newline)])
179
267
lines = iter(lines)
180
268
next = lines.next
271
def cache_and_return(line):
272
origin, text = line.split(' ', 1)
273
return cache.setdefault(origin, origin), text
181
275
# walk through the lines parsing.
182
276
for header in lines:
183
277
start, end, count = [int(n) for n in header.split(',')]
278
contents = [tuple(next().split(' ', 1)) for i in xrange(count)]
279
result.append((start, end, count, contents))
282
def get_fulltext_content(self, lines):
283
"""Extract just the content lines from a fulltext."""
284
return (line.split(' ', 1)[1] for line in lines)
286
def get_linedelta_content(self, lines):
287
"""Extract just the content from a line delta.
289
This doesn't return all of the extra information stored in a delta.
290
Only the actual content lines.
295
header = header.split(',')
296
count = int(header[2])
297
for i in xrange(count):
187
298
origin, text = next().split(' ', 1)
189
contents.append((origin.decode('utf-8'), text))
190
result.append((start, end, count, contents))
193
301
def lower_fulltext(self, content):
194
302
"""convert a fulltext content record into a serializable form.
196
304
see parse_fulltext which this inverts.
198
return ['%s %s' % (o.encode('utf-8'), t) for o, t in content._lines]
306
# TODO: jam 20070209 We only do the caching thing to make sure that
307
# the origin is a valid utf-8 line, eventually we could remove it
308
return ['%s %s' % (o, t) for o, t in content._lines]
200
310
def lower_line_delta(self, delta):
201
311
"""convert a delta into a serializable form.
203
313
See parse_line_delta which this inverts.
315
# TODO: jam 20070209 We only do the caching thing to make sure that
316
# the origin is a valid utf-8 line, eventually we could remove it
206
318
for start, end, c, lines in delta:
207
319
out.append('%d,%d,%d\n' % (start, end, c))
208
for origin, text in lines:
209
out.append('%s %s' % (origin.encode('utf-8'), text))
320
out.extend(origin + ' ' + text
321
for origin, text in lines)
213
class KnitPlainFactory(_KnitFactory):
324
def annotate_iter(self, knit, version_id):
325
content = knit._get_content(version_id)
326
return content.annotate_iter()
329
class KnitPlainFactory(object):
214
330
"""Factory for creating plain Content objects."""
216
332
annotated = False
218
def parse_fulltext(self, content, version):
334
def make(self, lines, version_id):
335
return PlainKnitContent(lines, version_id)
337
def parse_fulltext(self, content, version_id):
219
338
"""This parses an unannotated fulltext.
221
340
Note that this is not a noop - the internal representation
222
341
has (versionid, line) - its just a constant versionid.
224
return self.make(content, version)
343
return self.make(content, version_id)
226
def parse_line_delta_iter(self, lines, version):
228
header = lines.pop(0)
345
def parse_line_delta_iter(self, lines, version_id):
347
num_lines = len(lines)
348
while cur < num_lines:
229
351
start, end, c = [int(n) for n in header.split(',')]
230
yield start, end, c, zip([version] * c, lines[:c])
233
def parse_line_delta(self, lines, version):
234
return list(self.parse_line_delta_iter(lines, version))
352
yield start, end, c, lines[cur:cur+c]
355
def parse_line_delta(self, lines, version_id):
356
return list(self.parse_line_delta_iter(lines, version_id))
358
def get_fulltext_content(self, lines):
359
"""Extract just the content lines from a fulltext."""
362
def get_linedelta_content(self, lines):
363
"""Extract just the content from a line delta.
365
This doesn't return all of the extra information stored in a delta.
366
Only the actual content lines.
371
header = header.split(',')
372
count = int(header[2])
373
for i in xrange(count):
236
376
def lower_fulltext(self, content):
237
377
return content.text()
265
407
stored and retrieved.
268
def __init__(self, relpath, transport, file_mode=None, access_mode=None, factory=None,
269
basis_knit=None, delta=True, create=False):
410
def __init__(self, relpath, transport, file_mode=None, access_mode=None,
411
factory=None, basis_knit=DEPRECATED_PARAMETER, delta=True,
412
create=False, create_parent_dir=False, delay_create=False,
413
dir_mode=None, index=None, access_method=None):
270
414
"""Construct a knit at location specified by relpath.
272
416
:param create: If not True, only open an existing knit.
417
:param create_parent_dir: If True, create the parent directory if
418
creating the file fails. (This is used for stores with
419
hash-prefixes that may not exist yet)
420
:param delay_create: The calling code is aware that the knit won't
421
actually be created until the first data is stored.
422
:param index: An index to use for the knit.
424
if deprecated_passed(basis_knit):
425
warnings.warn("KnitVersionedFile.__(): The basis_knit parameter is"
426
" deprecated as of bzr 0.9.",
427
DeprecationWarning, stacklevel=2)
274
428
if access_mode is None:
275
429
access_mode = 'w'
276
430
super(KnitVersionedFile, self).__init__(access_mode)
277
431
assert access_mode in ('r', 'w'), "invalid mode specified %r" % access_mode
278
assert not basis_knit or isinstance(basis_knit, KnitVersionedFile), \
281
432
self.transport = transport
282
433
self.filename = relpath
283
self.basis_knit = basis_knit
284
434
self.factory = factory or KnitAnnotateFactory()
285
435
self.writable = (access_mode == 'w')
286
436
self.delta = delta
288
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
289
access_mode, create=create, file_mode=file_mode)
290
self._data = _KnitData(transport, relpath + DATA_SUFFIX,
291
access_mode, create=create and not len(self), file_mode=file_mode)
438
self._max_delta_chain = 200
441
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
442
access_mode, create=create, file_mode=file_mode,
443
create_parent_dir=create_parent_dir, delay_create=delay_create,
447
if access_method is None:
448
_access = _KnitAccess(transport, relpath + DATA_SUFFIX, file_mode, dir_mode,
449
((create and not len(self)) and delay_create), create_parent_dir)
451
_access = access_method
452
if create and not len(self) and not delay_create:
454
self._data = _KnitData(_access)
293
456
def __repr__(self):
294
457
return '%s(%s)' % (self.__class__.__name__,
295
458
self.transport.abspath(self.filename))
297
def _add_delta(self, version_id, parents, delta_parent, sha1, noeol, delta):
298
"""See VersionedFile._add_delta()."""
299
self._check_add(version_id, []) # should we check the lines ?
300
self._check_versions_present(parents)
304
for parent in parents:
305
if not self.has_version(parent):
306
ghosts.append(parent)
308
present_parents.append(parent)
310
if delta_parent is None:
311
# reconstitute as full text.
312
assert len(delta) == 1 or len(delta) == 0
314
assert delta[0][0] == 0
315
assert delta[0][1] == 0, delta[0][1]
316
return super(KnitVersionedFile, self)._add_delta(version_id,
327
options.append('no-eol')
329
if delta_parent is not None:
330
# determine the current delta chain length.
331
# To speed the extract of texts the delta chain is limited
332
# to a fixed number of deltas. This should minimize both
333
# I/O and the time spend applying deltas.
335
delta_parents = [delta_parent]
337
parent = delta_parents[0]
338
method = self._index.get_method(parent)
339
if method == 'fulltext':
341
delta_parents = self._index.get_parents(parent)
343
if method == 'line-delta':
344
# did not find a fulltext in the delta limit.
345
# just do a normal insertion.
346
return super(KnitVersionedFile, self)._add_delta(version_id,
353
options.append('line-delta')
354
store_lines = self.factory.lower_line_delta(delta)
356
where, size = self._data.add_record(version_id, digest, store_lines)
357
self._index.add_version(version_id, options, where, size, parents)
460
def _check_should_delta(self, first_parents):
461
"""Iterate back through the parent listing, looking for a fulltext.
463
This is used when we want to decide whether to add a delta or a new
464
fulltext. It searches for _max_delta_chain parents. When it finds a
465
fulltext parent, it sees if the total size of the deltas leading up to
466
it is large enough to indicate that we want a new full text anyway.
468
Return True if we should create a new delta, False if we should use a
473
delta_parents = first_parents
474
for count in xrange(self._max_delta_chain):
475
parent = delta_parents[0]
476
method = self._index.get_method(parent)
477
index, pos, size = self._index.get_position(parent)
478
if method == 'fulltext':
482
delta_parents = self._index.get_parents(parent)
484
# We couldn't find a fulltext, so we must create a new one
487
return fulltext_size > delta_size
359
489
def _add_raw_records(self, records, data):
360
490
"""Add all the records 'records' with data pre-joined in 'data'.
380
519
"""See VersionedFile.copy_to()."""
381
520
# copy the current index to a temp index to avoid racing with local
383
transport.put(name + INDEX_SUFFIX + '.tmp', self.transport.get(self._index._filename),)
522
transport.put_file_non_atomic(name + INDEX_SUFFIX + '.tmp',
523
self.transport.get(self._index._filename))
384
524
# copy the data file
385
transport.put(name + DATA_SUFFIX, self._data._open_file())
386
# rename the copied index into place
387
transport.rename(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
525
f = self._data._open_file()
527
transport.put_file(name + DATA_SUFFIX, f)
530
# move the copied index into place
531
transport.move(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
389
533
def create_empty(self, name, transport, mode=None):
390
return KnitVersionedFile(name, transport, factory=self.factory, delta=self.delta, create=True)
534
return KnitVersionedFile(name, transport, factory=self.factory,
535
delta=self.delta, create=True)
392
def _fix_parents(self, version, new_parents):
393
"""Fix the parents list for version.
537
def get_data_stream(self, required_versions):
538
"""Get a data stream for the specified versions.
540
Versions may be returned in any order, not necessarily the order
543
:param required_versions: The exact set of versions to be extracted.
544
Unlike some other knit methods, this is not used to generate a
545
transitive closure, rather it is used precisely as given.
395
This is done by appending a new version to the index
396
with identical data except for the parents list.
397
the parents list must be a superset of the current
547
:returns: format_signature, list of (version, options, length, parents),
400
current_values = self._index._cache[version]
401
assert set(current_values[4]).difference(set(new_parents)) == set()
402
self._index.add_version(version,
550
required_versions = set([osutils.safe_revision_id(v) for v in
552
# we don't care about inclusions, the caller cares.
553
# but we need to setup a list of records to visit.
554
for version_id in required_versions:
555
if not self.has_version(version_id):
556
raise RevisionNotPresent(version_id, self.filename)
557
# Pick the desired versions out of the index in oldest-to-newest order
559
for version_id in self.versions():
560
if version_id in required_versions:
561
version_list.append(version_id)
563
# create the list of version information for the result
564
copy_queue_records = []
566
result_version_list = []
567
for version_id in version_list:
568
options = self._index.get_options(version_id)
569
parents = self._index.get_parents_with_ghosts(version_id)
570
index_memo = self._index.get_position(version_id)
571
copy_queue_records.append((version_id, index_memo))
572
none, data_pos, data_size = index_memo
573
copy_set.add(version_id)
574
# version, options, length, parents
575
result_version_list.append((version_id, options, data_size,
578
# Read the compressed record data.
580
# From here down to the return should really be logic in the returned
581
# callable -- in a class that adapts read_records_iter_raw to read
584
for (version_id, raw_data), \
585
(version_id2, options, _, parents) in \
586
izip(self._data.read_records_iter_raw(copy_queue_records),
587
result_version_list):
588
assert version_id == version_id2, 'logic error, inconsistent results'
589
raw_datum.append(raw_data)
590
pseudo_file = StringIO(''.join(raw_datum))
593
return pseudo_file.read()
595
return pseudo_file.read(length)
596
return (self.get_format_signature(), result_version_list, read)
598
def _extract_blocks(self, version_id, source, target):
599
if self._index.get_method(version_id) != 'line-delta':
601
parent, sha1, noeol, delta = self.get_delta(version_id)
602
return KnitContent.get_line_delta_blocks(delta, source, target)
408
604
def get_delta(self, version_id):
409
605
"""Get a delta for constructing version from some other version."""
410
if not self.has_version(version_id):
411
raise RevisionNotPresent(version_id, self.filename)
606
version_id = osutils.safe_revision_id(version_id)
607
self.check_not_reserved_id(version_id)
413
608
parents = self.get_parents(version_id)
415
610
parent = parents[0]
418
data_pos, data_size = self._index.get_position(version_id)
419
data, sha1 = self._data.read_records(((version_id, data_pos, data_size),))[version_id]
420
version_idx = self._index.lookup(version_id)
613
index_memo = self._index.get_position(version_id)
614
data, sha1 = self._data.read_records(((version_id, index_memo),))[version_id]
421
615
noeol = 'no-eol' in self._index.get_options(version_id)
422
616
if 'fulltext' == self._index.get_method(version_id):
423
new_content = self.factory.parse_fulltext(data, version_idx)
617
new_content = self.factory.parse_fulltext(data, version_id)
424
618
if parent is not None:
425
619
reference_content = self._get_content(parent)
426
620
old_texts = reference_content.text()
429
623
new_texts = new_content.text()
430
delta_seq = SequenceMatcher(None, old_texts, new_texts)
624
delta_seq = patiencediff.PatienceSequenceMatcher(None, old_texts,
431
626
return parent, sha1, noeol, self._make_line_delta(delta_seq, new_content)
433
delta = self.factory.parse_line_delta(data, version_idx)
628
delta = self.factory.parse_line_delta(data, version_id)
434
629
return parent, sha1, noeol, delta
631
def get_format_signature(self):
632
"""See VersionedFile.get_format_signature()."""
633
if self.factory.annotated:
634
annotated_part = "annotated"
636
annotated_part = "plain"
637
return "knit-%s" % (annotated_part,)
436
639
def get_graph_with_ghosts(self):
437
640
"""See VersionedFile.get_graph_with_ghosts()."""
674
def insert_data_stream(self, (format, data_list, reader_callable)):
675
"""Insert knit records from a data stream into this knit.
677
If a version in the stream is already present in this knit, it will not
678
be inserted a second time. It will be checked for consistency with the
679
stored version however, and may cause a KnitCorrupt error to be raised
680
if the data in the stream disagrees with the already stored data.
682
:seealso: get_data_stream
684
if format != self.get_format_signature():
685
trace.mutter('incompatible format signature inserting to %r', self)
686
raise KnitDataStreamIncompatible(
687
format, self.get_format_signature())
689
for version_id, options, length, parents in data_list:
690
if self.has_version(version_id):
691
# First check: the list of parents.
692
my_parents = self.get_parents_with_ghosts(version_id)
693
if my_parents != parents:
694
# XXX: KnitCorrupt is not quite the right exception here.
697
'parents list %r from data stream does not match '
698
'already recorded parents %r for %s'
699
% (parents, my_parents, version_id))
701
# Also check the SHA-1 of the fulltext this content will
703
raw_data = reader_callable(length)
704
my_fulltext_sha1 = self.get_sha1(version_id)
705
df, rec = self._data._parse_record_header(version_id, raw_data)
706
stream_fulltext_sha1 = rec[3]
707
if my_fulltext_sha1 != stream_fulltext_sha1:
708
# Actually, we don't know if it's this knit that's corrupt,
709
# or the data stream we're trying to insert.
711
self.filename, 'sha-1 does not match %s' % version_id)
713
self._add_raw_records(
714
[(version_id, options, parents, length)],
715
reader_callable(length))
465
717
def versions(self):
466
718
"""See VersionedFile.versions."""
719
if 'evil' in debug.debug_flags:
720
trace.mutter_callsite(2, "versions scales with size of history")
467
721
return self._index.get_versions()
469
723
def has_version(self, version_id):
470
724
"""See VersionedFile.has_version."""
725
if 'evil' in debug.debug_flags:
726
trace.mutter_callsite(2, "has_version is a LBYL scenario")
727
version_id = osutils.safe_revision_id(version_id)
471
728
return self._index.has_version(version_id)
473
730
__contains__ = has_version
475
732
def _merge_annotations(self, content, parents, parent_texts={},
476
delta=None, annotated=None):
733
delta=None, annotated=None,
734
left_matching_blocks=None):
477
735
"""Merge annotations for content. This is done by comparing
478
736
the annotations based on changed to the text.
738
if left_matching_blocks is not None:
739
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
482
743
for parent_id in parents:
483
744
merge_content = self._get_content(parent_id, parent_texts)
484
seq = SequenceMatcher(None, merge_content.text(), content.text())
485
if delta_seq is None:
486
# setup a delta seq to reuse.
745
if (parent_id == parents[0] and delta_seq is not None):
748
seq = patiencediff.PatienceSequenceMatcher(
749
None, merge_content.text(), content.text())
488
750
for i, j, n in seq.get_matching_blocks():
491
# this appears to copy (origin, text) pairs across to the new
492
# content for any line that matches the last-checked parent.
493
# FIXME: save the sequence control data for delta compression
494
# against the most relevant parent rather than rediffing.
753
# this appears to copy (origin, text) pairs across to the
754
# new content for any line that matches the last-checked
495
756
content._lines[j:j+n] = merge_content._lines[i:i+n]
758
if delta_seq is None:
498
759
reference_content = self._get_content(parents[0], parent_texts)
499
760
new_texts = content.text()
500
761
old_texts = reference_content.text()
501
delta_seq = SequenceMatcher(None, old_texts, new_texts)
762
delta_seq = patiencediff.PatienceSequenceMatcher(
763
None, old_texts, new_texts)
502
764
return self._make_line_delta(delta_seq, content)
504
766
def _make_line_delta(self, delta_seq, new_content):
510
772
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
511
773
return diff_hunks
513
def _get_components(self, version_id):
514
"""Return a list of (version_id, method, data) tuples that
515
makes up version specified by version_id of the knit.
517
The components should be applied in the order of the returned
520
The basis knit will be used to the largest extent possible
521
since it is assumed that accesses to it is faster.
775
def _get_components_positions(self, version_ids):
776
"""Produce a map of position data for the components of versions.
778
This data is intended to be used for retrieving the knit records.
780
A dict of version_id to (method, data_pos, data_size, next) is
782
method is the way referenced data should be applied.
783
data_pos is the position of the data in the knit.
784
data_size is the size of the data in the knit.
785
next is the build-parent of the version, or None for fulltexts.
524
# 4168 calls in 14912, 2289 internal
525
# 4168 in 9711 to read_records
526
# 52554 in 1250 to get_parents
527
# 170166 in 865 to list.append
529
# needed_revisions holds a list of (method, version_id) of
530
# versions that is needed to be fetched to construct the final
531
# version of the file.
533
# basis_revisions is a list of versions that needs to be
534
# fetched but exists in the basis knit.
536
basis = self.basis_knit
543
if basis and basis._index.has_version(cursor):
545
basis_versions.append(cursor)
546
method = picked_knit._index.get_method(cursor)
547
needed_versions.append((method, cursor))
548
if method == 'fulltext':
550
cursor = picked_knit.get_parents(cursor)[0]
555
for comp_id in basis_versions:
556
data_pos, data_size = basis._index.get_data_position(comp_id)
557
records.append((piece_id, data_pos, data_size))
558
components.update(basis._data.read_records(records))
561
for comp_id in [vid for method, vid in needed_versions
562
if vid not in basis_versions]:
563
data_pos, data_size = self._index.get_position(comp_id)
564
records.append((comp_id, data_pos, data_size))
565
components.update(self._data.read_records(records))
567
# get_data_records returns a mapping with the version id as
568
# index and the value as data. The order the components need
569
# to be applied is held by needed_versions (reversed).
571
for method, comp_id in reversed(needed_versions):
572
out.append((comp_id, method, components[comp_id]))
788
for version_id in version_ids:
791
while cursor is not None and cursor not in component_data:
792
method = self._index.get_method(cursor)
793
if method == 'fulltext':
796
next = self.get_parents(cursor)[0]
797
index_memo = self._index.get_position(cursor)
798
component_data[cursor] = (method, index_memo, next)
800
return component_data
576
802
def _get_content(self, version_id, parent_texts={}):
577
803
"""Returns a content object that makes up the specified
579
if not self.has_version(version_id):
580
raise RevisionNotPresent(version_id, self.filename)
582
805
cached_version = parent_texts.get(version_id, None)
583
806
if cached_version is not None:
807
if not self.has_version(version_id):
808
raise RevisionNotPresent(version_id, self.filename)
584
809
return cached_version
586
if self.basis_knit and version_id in self.basis_knit:
587
return self.basis_knit._get_content(version_id)
590
components = self._get_components(version_id)
591
for component_id, method, (data, digest) in components:
592
version_idx = self._index.lookup(component_id)
593
if method == 'fulltext':
594
assert content is None
595
content = self.factory.parse_fulltext(data, version_idx)
596
elif method == 'line-delta':
597
delta = self.factory.parse_line_delta(data, version_idx)
598
content._lines = self._apply_delta(content._lines, delta)
600
if 'no-eol' in self._index.get_options(version_id):
601
line = content._lines[-1][1].rstrip('\n')
602
content._lines[-1] = (content._lines[-1][0], line)
604
# digest here is the digest from the last applied component.
605
if sha_strings(content.text()) != digest:
606
import pdb;pdb.set_trace()
607
raise KnitCorrupt(self.filename, 'sha-1 does not match %s' % version_id)
811
text_map, contents_map = self._get_content_maps([version_id])
812
return contents_map[version_id]
611
814
def _check_versions_present(self, version_ids):
612
815
"""Check that all specified versions are present."""
613
version_ids = set(version_ids)
614
for r in list(version_ids):
615
if self._index.has_version(r):
616
version_ids.remove(r)
618
raise RevisionNotPresent(list(version_ids)[0], self.filename)
816
self._index.check_versions_present(version_ids)
620
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts):
818
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
819
nostore_sha, random_id, check_content):
621
820
"""See VersionedFile.add_lines_with_ghosts()."""
622
self._check_add(version_id, lines)
623
return self._add(version_id, lines[:], parents, self.delta, parent_texts)
821
self._check_add(version_id, lines, random_id, check_content)
822
return self._add(version_id, lines, parents, self.delta,
823
parent_texts, None, nostore_sha)
625
def _add_lines(self, version_id, parents, lines, parent_texts):
825
def _add_lines(self, version_id, parents, lines, parent_texts,
826
left_matching_blocks, nostore_sha, random_id, check_content):
626
827
"""See VersionedFile.add_lines."""
627
self._check_add(version_id, lines)
828
self._check_add(version_id, lines, random_id, check_content)
628
829
self._check_versions_present(parents)
629
return self._add(version_id, lines[:], parents, self.delta, parent_texts)
830
return self._add(version_id, lines[:], parents, self.delta,
831
parent_texts, left_matching_blocks, nostore_sha)
631
def _check_add(self, version_id, lines):
833
def _check_add(self, version_id, lines, random_id, check_content):
632
834
"""check that version_id and lines are safe to add."""
633
assert self.writable, "knit is not opened for write"
634
### FIXME escape. RBC 20060228
635
835
if contains_whitespace(version_id):
636
836
raise InvalidRevisionId(version_id, self.filename)
637
if self.has_version(version_id):
837
self.check_not_reserved_id(version_id)
838
# Technically this could be avoided if we are happy to allow duplicate
839
# id insertion when other things than bzr core insert texts, but it
840
# seems useful for folk using the knit api directly to have some safety
841
# blanket that we can disable.
842
if not random_id and self.has_version(version_id):
638
843
raise RevisionAlreadyPresent(version_id, self.filename)
639
self._check_lines_not_unicode(lines)
640
self._check_lines_are_lines(lines)
845
self._check_lines_not_unicode(lines)
846
self._check_lines_are_lines(lines)
642
def _add(self, version_id, lines, parents, delta, parent_texts):
848
def _add(self, version_id, lines, parents, delta, parent_texts,
849
left_matching_blocks, nostore_sha):
643
850
"""Add a set of lines on top of version specified by parents.
645
852
If delta is true, compress the text as a line-delta against
660
867
# +61 0 1918.1800 5.2640 +bzrlib.knit:359(_merge_annotations)
662
869
present_parents = []
664
870
if parent_texts is None:
665
871
parent_texts = {}
666
872
for parent in parents:
667
if not self.has_version(parent):
668
ghosts.append(parent)
873
if self.has_version(parent):
670
874
present_parents.append(parent)
672
if delta and not len(present_parents):
876
# can only compress against the left most present parent.
878
(len(present_parents) == 0 or
879
present_parents[0] != parents[0])):
675
882
digest = sha_strings(lines)
883
if nostore_sha == digest:
884
raise errors.ExistingContent
885
text_length = sum(map(len, lines))
678
888
if lines[-1][-1] != '\n':
889
# copy the contents of lines.
679
891
options.append('no-eol')
680
892
lines[-1] = lines[-1] + '\n'
682
if len(present_parents) and delta:
683
895
# To speed the extract of texts the delta chain is limited
684
896
# to a fixed number of deltas. This should minimize both
685
897
# I/O and the time spend applying deltas.
687
delta_parents = present_parents
689
parent = delta_parents[0]
690
method = self._index.get_method(parent)
691
if method == 'fulltext':
693
delta_parents = self._index.get_parents(parent)
695
if method == 'line-delta':
898
delta = self._check_should_delta(present_parents)
698
lines = self.factory.make(lines, version_id)
900
assert isinstance(version_id, str)
901
content = self.factory.make(lines, version_id)
699
902
if delta or (self.factory.annotated and len(present_parents) > 0):
700
# Merge annotations from parent texts if so is needed.
701
delta_hunks = self._merge_annotations(lines, present_parents, parent_texts,
702
delta, self.factory.annotated)
903
# Merge annotations from parent texts if needed.
904
delta_hunks = self._merge_annotations(content, present_parents,
905
parent_texts, delta, self.factory.annotated,
906
left_matching_blocks)
705
909
options.append('line-delta')
706
910
store_lines = self.factory.lower_line_delta(delta_hunks)
708
912
options.append('fulltext')
709
store_lines = self.factory.lower_fulltext(lines)
913
store_lines = self.factory.lower_fulltext(content)
711
where, size = self._data.add_record(version_id, digest, store_lines)
712
self._index.add_version(version_id, options, where, size, parents)
915
access_memo = self._data.add_record(version_id, digest, store_lines)
916
self._index.add_version(version_id, options, access_memo, parents)
917
return digest, text_length, content
715
919
def check(self, progress_bar=None):
716
920
"""See VersionedFile.check()."""
718
922
def _clone_text(self, new_version_id, old_version_id, parents):
719
923
"""See VersionedFile.clone_text()."""
720
# FIXME RBC 20060228 make fast by only inserting an index with null delta.
924
# FIXME RBC 20060228 make fast by only inserting an index with null
721
926
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
723
928
def get_lines(self, version_id):
724
929
"""See VersionedFile.get_lines()."""
725
return self._get_content(version_id).text()
727
def iter_lines_added_or_present_in_versions(self, version_ids=None):
930
return self.get_line_list([version_id])[0]
932
def _get_record_map(self, version_ids):
933
"""Produce a dictionary of knit records.
935
The keys are version_ids, the values are tuples of (method, content,
937
method is the way the content should be applied.
938
content is a KnitContent object.
939
digest is the SHA1 digest of this version id after all steps are done
940
next is the build-parent of the version, i.e. the leftmost ancestor.
941
If the method is fulltext, next will be None.
943
position_map = self._get_components_positions(version_ids)
944
# c = component_id, m = method, i_m = index_memo, n = next
945
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
947
for component_id, content, digest in \
948
self._data.read_records_iter(records):
949
method, index_memo, next = position_map[component_id]
950
record_map[component_id] = method, content, digest, next
954
def get_text(self, version_id):
955
"""See VersionedFile.get_text"""
956
return self.get_texts([version_id])[0]
958
def get_texts(self, version_ids):
959
return [''.join(l) for l in self.get_line_list(version_ids)]
961
def get_line_list(self, version_ids):
962
"""Return the texts of listed versions as a list of strings."""
963
version_ids = [osutils.safe_revision_id(v) for v in version_ids]
964
for version_id in version_ids:
965
self.check_not_reserved_id(version_id)
966
text_map, content_map = self._get_content_maps(version_ids)
967
return [text_map[v] for v in version_ids]
969
_get_lf_split_line_list = get_line_list
971
def _get_content_maps(self, version_ids):
972
"""Produce maps of text and KnitContents
974
:return: (text_map, content_map) where text_map contains the texts for
975
the requested versions and content_map contains the KnitContents.
976
Both dicts take version_ids as their keys.
978
for version_id in version_ids:
979
if not self.has_version(version_id):
980
raise RevisionNotPresent(version_id, self.filename)
981
record_map = self._get_record_map(version_ids)
986
for version_id in version_ids:
989
while cursor is not None:
990
method, data, digest, next = record_map[cursor]
991
components.append((cursor, method, data, digest))
992
if cursor in content_map:
997
for component_id, method, data, digest in reversed(components):
998
if component_id in content_map:
999
content = content_map[component_id]
1001
if method == 'fulltext':
1002
assert content is None
1003
content = self.factory.parse_fulltext(data, version_id)
1004
elif method == 'line-delta':
1005
delta = self.factory.parse_line_delta(data, version_id)
1006
content = content.copy()
1007
content._lines = self._apply_delta(content._lines,
1009
content_map[component_id] = content
1011
if 'no-eol' in self._index.get_options(version_id):
1012
content = content.copy()
1013
content.strip_last_line_newline()
1014
final_content[version_id] = content
1016
# digest here is the digest from the last applied component.
1017
text = content.text()
1018
if sha_strings(text) != digest:
1019
raise KnitCorrupt(self.filename,
1020
'sha-1 does not match %s' % version_id)
1022
text_map[version_id] = text
1023
return text_map, final_content
1025
def iter_lines_added_or_present_in_versions(self, version_ids=None,
728
1027
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
729
1028
if version_ids is None:
730
1029
version_ids = self.versions()
731
# we dont care about inclusions, the caller cares.
1031
version_ids = [osutils.safe_revision_id(v) for v in version_ids]
1033
pb = progress.DummyProgress()
1034
# we don't care about inclusions, the caller cares.
732
1035
# but we need to setup a list of records to visit.
733
1036
# we need version_id, position, length
734
1037
version_id_records = []
735
requested_versions = list(version_ids)
1038
requested_versions = set(version_ids)
736
1039
# filter for available versions
737
1040
for version_id in requested_versions:
738
1041
if not self.has_version(version_id):
739
1042
raise RevisionNotPresent(version_id, self.filename)
740
1043
# get a in-component-order queue:
742
1044
for version_id in self.versions():
743
1045
if version_id in requested_versions:
744
version_ids.append(version_id)
745
data_pos, length = self._index.get_position(version_id)
746
version_id_records.append((version_id, data_pos, length))
1046
index_memo = self._index.get_position(version_id)
1047
version_id_records.append((version_id, index_memo))
748
pb = bzrlib.ui.ui_factory.nested_progress_bar()
750
1049
total = len(version_id_records)
752
pb.update('Walking content.', count, total)
753
for version_id, data, sha_value in \
754
self._data.read_records_iter(version_id_records):
755
pb.update('Walking content.', count, total)
756
method = self._index.get_method(version_id)
757
version_idx = self._index.lookup(version_id)
758
assert method in ('fulltext', 'line-delta')
759
if method == 'fulltext':
760
content = self.factory.parse_fulltext(data, version_idx)
761
for line in content.text():
764
delta = self.factory.parse_line_delta(data, version_idx)
765
for start, end, count, lines in delta:
766
for origin, line in lines:
769
pb.update('Walking content.', total, total)
772
pb.update('Walking content.', total, total)
1050
for version_idx, (version_id, data, sha_value) in \
1051
enumerate(self._data.read_records_iter(version_id_records)):
1052
pb.update('Walking content.', version_idx, total)
1053
method = self._index.get_method(version_id)
1055
assert method in ('fulltext', 'line-delta')
1056
if method == 'fulltext':
1057
line_iterator = self.factory.get_fulltext_content(data)
1059
line_iterator = self.factory.get_linedelta_content(data)
1060
for line in line_iterator:
1063
pb.update('Walking content.', total, total)
1065
def iter_parents(self, version_ids):
1066
"""Iterate through the parents for many version ids.
1068
:param version_ids: An iterable yielding version_ids.
1069
:return: An iterator that yields (version_id, parents). Requested
1070
version_ids not present in the versioned file are simply skipped.
1071
The order is undefined, allowing for different optimisations in
1072
the underlying implementation.
1074
version_ids = [osutils.safe_revision_id(version_id) for
1075
version_id in version_ids]
1076
return self._index.iter_parents(version_ids)
776
1078
def num_versions(self):
777
1079
"""See VersionedFile.num_versions()."""
778
1080
return self._index.num_versions()
817
1120
versions = [versions]
818
1121
if not versions:
820
self._check_versions_present(versions)
1123
versions = [osutils.safe_revision_id(v) for v in versions]
821
1124
return self._index.get_ancestry_with_ghosts(versions)
823
#@deprecated_method(zero_eight)
824
def walk(self, version_ids):
825
"""See VersionedFile.walk."""
826
# We take the short path here, and extract all relevant texts
827
# and put them in a weave and let that do all the work. Far
828
# from optimal, but is much simpler.
829
# FIXME RB 20060228 this really is inefficient!
830
from bzrlib.weave import Weave
832
w = Weave(self.filename)
833
ancestry = self.get_ancestry(version_ids)
834
sorted_graph = topo_sort(self._index.get_graph())
835
version_list = [vid for vid in sorted_graph if vid in ancestry]
837
for version_id in version_list:
838
lines = self.get_lines(version_id)
839
w.add_lines(version_id, self.get_parents(version_id), lines)
841
for lineno, insert_id, dset, line in w.walk(version_ids):
842
yield lineno, insert_id, dset, line
844
1126
def plan_merge(self, ver_a, ver_b):
845
1127
"""See VersionedFile.plan_merge."""
846
ancestors_b = set(self.get_ancestry(ver_b))
847
def status_a(revision, text):
848
if revision in ancestors_b:
849
return 'killed-b', text
1128
ver_a = osutils.safe_revision_id(ver_a)
1129
ver_b = osutils.safe_revision_id(ver_b)
1130
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
853
ancestors_a = set(self.get_ancestry(ver_a))
854
def status_b(revision, text):
855
if revision in ancestors_a:
856
return 'killed-a', text
1132
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
860
1133
annotated_a = self.annotate(ver_a)
861
1134
annotated_b = self.annotate(ver_b)
862
plain_a = [t for (a, t) in annotated_a]
863
plain_b = [t for (a, t) in annotated_b]
864
blocks = SequenceMatcher(None, plain_a, plain_b).get_matching_blocks()
867
for ai, bi, l in blocks:
868
# process all mismatched sections
869
# (last mismatched section is handled because blocks always
870
# includes a 0-length last block)
871
for revision, text in annotated_a[a_cur:ai]:
872
yield status_a(revision, text)
873
for revision, text in annotated_b[b_cur:bi]:
874
yield status_b(revision, text)
876
# and now the matched section
879
for text_a, text_b in zip(plain_a[ai:a_cur], plain_b[bi:b_cur]):
880
assert text_a == text_b
881
yield "unchanged", text_a
1135
return merge._plan_annotate_merge(annotated_a, annotated_b,
1136
ancestors_a, ancestors_b)
884
1139
class _KnitComponentFile(object):
885
1140
"""One of the files used to implement a knit database"""
887
def __init__(self, transport, filename, mode, file_mode=None):
1142
def __init__(self, transport, filename, mode, file_mode=None,
1143
create_parent_dir=False, dir_mode=None):
888
1144
self._transport = transport
889
1145
self._filename = filename
890
1146
self._mode = mode
891
self._file_mode=file_mode
1147
self._file_mode = file_mode
1148
self._dir_mode = dir_mode
1149
self._create_parent_dir = create_parent_dir
1150
self._need_to_create = False
893
def write_header(self):
894
if self._transport.append(self._filename, StringIO(self.HEADER),
895
mode=self._file_mode):
896
raise KnitCorrupt(self._filename, 'misaligned after writing header')
1152
def _full_path(self):
1153
"""Return the full path to this file."""
1154
return self._transport.base + self._filename
898
1156
def check_header(self, fp):
899
1157
line = fp.readline()
1159
# An empty file can actually be treated as though the file doesn't
1161
raise errors.NoSuchFile(self._full_path())
900
1162
if line != self.HEADER:
901
raise KnitHeaderError(badline=line)
904
"""Commit is a nop."""
1163
raise KnitHeaderError(badline=line,
1164
filename=self._transport.abspath(self._filename))
906
1166
def __repr__(self):
907
1167
return '%s(%s)' % (self.__class__.__name__, self._filename)
979
1239
self._history.append(version_id)
981
1241
index = self._cache[version_id][5]
982
self._cache[version_id] = (version_id,
1242
self._cache[version_id] = (version_id,
989
def __init__(self, transport, filename, mode, create=False, file_mode=None):
990
_KnitComponentFile.__init__(self, transport, filename, mode, file_mode)
1249
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1250
create_parent_dir=False, delay_create=False, dir_mode=None):
1251
_KnitComponentFile.__init__(self, transport, filename, mode,
1252
file_mode=file_mode,
1253
create_parent_dir=create_parent_dir,
991
1255
self._cache = {}
992
1256
# position in _history is the 'official' index for a revision
993
1257
# but the values may have come from a newer entry.
994
# so - wc -l of a knit index is != the number of uniqe names
1258
# so - wc -l of a knit index is != the number of unique names
996
1260
self._history = []
997
pb = bzrlib.ui.ui_factory.nested_progress_bar()
1262
fp = self._transport.get(self._filename)
1002
pb.update('read knit index', count, total)
1003
fp = self._transport.get(self._filename)
1004
self.check_header(fp)
1005
# readlines reads the whole file at once:
1006
# bad for transports like http, good for local disk
1007
# we save 60 ms doing this one change (
1008
# from calling readline each time to calling
1010
# probably what we want for nice behaviour on
1011
# http is a incremental readlines that yields, or
1012
# a check for local vs non local indexes,
1013
for l in fp.readlines():
1015
if len(rec) < 5 or rec[-1] != ':':
1017
# FIXME: in the future we should determine if its a
1018
# short write - and ignore it
1019
# or a different failure, and raise. RBC 20060407
1023
#pb.update('read knit index', count, total)
1024
# See self._parse_parents
1026
for value in rec[4:-1]:
1028
# uncompressed reference
1029
parents.append(value[1:])
1031
# this is 15/4000ms faster than isinstance,
1033
# this function is called thousands of times a
1034
# second so small variations add up.
1035
assert value.__class__ is str
1036
parents.append(self._history[int(value)])
1037
# end self._parse_parents
1038
# self._cache_version(rec[0],
1039
# rec[1].split(','),
1043
# --- self._cache_version
1044
# only want the _history index to reference the 1st
1045
# index entry for version_id
1047
if version_id not in self._cache:
1048
index = len(self._history)
1049
self._history.append(version_id)
1051
index = self._cache[version_id][5]
1052
self._cache[version_id] = (version_id,
1058
# --- self._cache_version
1059
except NoSuchFile, e:
1060
if mode != 'w' or not create:
1064
pb.update('read knit index', total, total)
1067
def _parse_parents(self, compressed_parents):
1068
"""convert a list of string parent values into version ids.
1070
ints are looked up in the index.
1071
.FOO values are ghosts and converted in to FOO.
1073
NOTE: the function is retained here for clarity, and for possible
1074
use in partial index reads. However bulk processing now has
1075
it inlined in __init__ for inner-loop optimisation.
1078
for value in compressed_parents:
1079
if value[-1] == '.':
1080
# uncompressed reference
1081
result.append(value[1:])
1264
# _load_data may raise NoSuchFile if the target knit is
1266
_load_data(self, fp)
1270
if mode != 'w' or not create:
1273
self._need_to_create = True
1083
# this is 15/4000ms faster than isinstance,
1084
# this function is called thousands of times a
1085
# second so small variations add up.
1086
assert value.__class__ is str
1087
result.append(self._history[int(value)])
1275
self._transport.put_bytes_non_atomic(
1276
self._filename, self.HEADER, mode=self._file_mode)
1090
1278
def get_graph(self):
1092
for version_id, index in self._cache.iteritems():
1093
graph.append((version_id, index[4]))
1279
"""Return a list of the node:parents lists from this knit index."""
1280
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1096
def get_ancestry(self, versions):
1282
def get_ancestry(self, versions, topo_sorted=True):
1097
1283
"""See VersionedFile.get_ancestry."""
1098
1284
# get a graph of all the mentioned versions:
1100
1286
pending = set(versions)
1102
1289
version = pending.pop()
1103
parents = self._cache[version][4]
1104
# got the parents ok
1106
parents = [parent for parent in parents if parent in self._cache]
1107
for parent in parents:
1108
# if not completed and not a ghost
1109
if parent not in graph:
1292
parents = [p for p in cache[version][4] if p in cache]
1294
raise RevisionNotPresent(version, self._filename)
1295
# if not completed and not a ghost
1296
pending.update([p for p in parents if p not in graph])
1111
1297
graph[version] = parents
1112
1300
return topo_sort(graph.items())
1114
1302
def get_ancestry_with_ghosts(self, versions):
1115
1303
"""See VersionedFile.get_ancestry_with_ghosts."""
1116
1304
# get a graph of all the mentioned versions:
1305
self.check_versions_present(versions)
1118
1308
pending = set(versions)
1120
1310
version = pending.pop()
1122
parents = self._cache[version][4]
1312
parents = cache[version][4]
1123
1313
except KeyError:
1124
1314
# ghost, fake it
1125
1315
graph[version] = []
1128
# got the parents ok
1129
for parent in parents:
1130
if parent not in graph:
1318
pending.update([p for p in parents if p not in graph])
1132
1319
graph[version] = parents
1133
1320
return topo_sort(graph.items())
1322
def iter_parents(self, version_ids):
1323
"""Iterate through the parents for many version ids.
1325
:param version_ids: An iterable yielding version_ids.
1326
:return: An iterator that yields (version_id, parents). Requested
1327
version_ids not present in the versioned file are simply skipped.
1328
The order is undefined, allowing for different optimisations in
1329
the underlying implementation.
1331
for version_id in version_ids:
1333
yield version_id, tuple(self.get_parents(version_id))
1135
1337
def num_versions(self):
1136
1338
return len(self._history)
1138
1340
__len__ = num_versions
1140
1342
def get_versions(self):
1343
"""Get all the versions in the file. not topologically sorted."""
1141
1344
return self._history
1143
def idx_to_name(self, idx):
1144
return self._history[idx]
1146
def lookup(self, version_id):
1147
assert version_id in self._cache
1148
return self._cache[version_id][5]
1150
1346
def _version_list_to_index(self, versions):
1151
1347
result_list = []
1152
1349
for version in versions:
1153
if version in self._cache:
1350
if version in cache:
1154
1351
# -- inlined lookup() --
1155
result_list.append(str(self._cache[version][5]))
1352
result_list.append(str(cache[version][5]))
1156
1353
# -- end lookup () --
1158
result_list.append('.' + version.encode('utf-8'))
1355
result_list.append('.' + version)
1159
1356
return ' '.join(result_list)
1161
def add_version(self, version_id, options, pos, size, parents):
1358
def add_version(self, version_id, options, index_memo, parents):
1162
1359
"""Add a version record to the index."""
1163
self.add_versions(((version_id, options, pos, size, parents),))
1360
self.add_versions(((version_id, options, index_memo, parents),))
1165
1362
def add_versions(self, versions):
1166
1363
"""Add multiple versions to the index.
1213
1437
if parent in self._cache]
1215
1439
def get_parents_with_ghosts(self, version_id):
1216
"""Return parents of specified version wth ghosts."""
1440
"""Return parents of specified version with ghosts."""
1217
1441
return self._cache[version_id][4]
1219
1443
def check_versions_present(self, version_ids):
1220
1444
"""Check that all specified versions are present."""
1221
version_ids = set(version_ids)
1222
for version_id in list(version_ids):
1223
if version_id in self._cache:
1224
version_ids.remove(version_id)
1226
raise RevisionNotPresent(list(version_ids)[0], self.filename)
1229
class _KnitData(_KnitComponentFile):
1230
"""Contents of the knit data file"""
1232
HEADER = "# bzr knit data 8\n"
1234
def __init__(self, transport, filename, mode, create=False, file_mode=None):
1235
_KnitComponentFile.__init__(self, transport, filename, mode)
1446
for version_id in version_ids:
1447
if version_id not in cache:
1448
raise RevisionNotPresent(version_id, self._filename)
1451
class KnitGraphIndex(object):
1452
"""A knit index that builds on GraphIndex."""
1454
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1455
"""Construct a KnitGraphIndex on a graph_index.
1457
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1458
:param deltas: Allow delta-compressed records.
1459
:param add_callback: If not None, allow additions to the index and call
1460
this callback with a list of added GraphIndex nodes:
1461
[(node, value, node_refs), ...]
1462
:param parents: If True, record knits parents, if not do not record
1465
self._graph_index = graph_index
1466
self._deltas = deltas
1467
self._add_callback = add_callback
1468
self._parents = parents
1469
if deltas and not parents:
1470
raise KnitCorrupt(self, "Cannot do delta compression without "
1473
def _get_entries(self, keys, check_present=False):
1474
"""Get the entries for keys.
1476
:param keys: An iterable of index keys, - 1-tuples.
1481
for node in self._graph_index.iter_entries(keys):
1483
found_keys.add(node[1])
1485
# adapt parentless index to the rest of the code.
1486
for node in self._graph_index.iter_entries(keys):
1487
yield node[0], node[1], node[2], ()
1488
found_keys.add(node[1])
1490
missing_keys = keys.difference(found_keys)
1492
raise RevisionNotPresent(missing_keys.pop(), self)
1494
def _present_keys(self, version_ids):
1496
node[1] for node in self._get_entries(version_ids)])
1498
def _parentless_ancestry(self, versions):
1499
"""Honour the get_ancestry API for parentless knit indices."""
1500
wanted_keys = self._version_ids_to_keys(versions)
1501
present_keys = self._present_keys(wanted_keys)
1502
missing = set(wanted_keys).difference(present_keys)
1504
raise RevisionNotPresent(missing.pop(), self)
1505
return list(self._keys_to_version_ids(present_keys))
1507
def get_ancestry(self, versions, topo_sorted=True):
1508
"""See VersionedFile.get_ancestry."""
1509
if not self._parents:
1510
return self._parentless_ancestry(versions)
1511
# XXX: This will do len(history) index calls - perhaps
1512
# it should be altered to be a index core feature?
1513
# get a graph of all the mentioned versions:
1516
versions = self._version_ids_to_keys(versions)
1517
pending = set(versions)
1519
# get all pending nodes
1520
this_iteration = pending
1521
new_nodes = self._get_entries(this_iteration)
1524
for (index, key, value, node_refs) in new_nodes:
1525
# dont ask for ghosties - otherwise
1526
# we we can end up looping with pending
1527
# being entirely ghosted.
1528
graph[key] = [parent for parent in node_refs[0]
1529
if parent not in ghosts]
1531
for parent in graph[key]:
1532
# dont examine known nodes again
1537
ghosts.update(this_iteration.difference(found))
1538
if versions.difference(graph):
1539
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1541
result_keys = topo_sort(graph.items())
1543
result_keys = graph.iterkeys()
1544
return [key[0] for key in result_keys]
1546
def get_ancestry_with_ghosts(self, versions):
1547
"""See VersionedFile.get_ancestry."""
1548
if not self._parents:
1549
return self._parentless_ancestry(versions)
1550
# XXX: This will do len(history) index calls - perhaps
1551
# it should be altered to be a index core feature?
1552
# get a graph of all the mentioned versions:
1554
versions = self._version_ids_to_keys(versions)
1555
pending = set(versions)
1557
# get all pending nodes
1558
this_iteration = pending
1559
new_nodes = self._get_entries(this_iteration)
1561
for (index, key, value, node_refs) in new_nodes:
1562
graph[key] = node_refs[0]
1564
for parent in graph[key]:
1565
# dont examine known nodes again
1569
missing_versions = this_iteration.difference(graph)
1570
missing_needed = versions.intersection(missing_versions)
1572
raise RevisionNotPresent(missing_needed.pop(), self)
1573
for missing_version in missing_versions:
1574
# add a key, no parents
1575
graph[missing_version] = []
1576
pending.discard(missing_version) # don't look for it
1577
result_keys = topo_sort(graph.items())
1578
return [key[0] for key in result_keys]
1580
def get_graph(self):
1581
"""Return a list of the node:parents lists from this knit index."""
1582
if not self._parents:
1583
return [(key, ()) for key in self.get_versions()]
1585
for index, key, value, refs in self._graph_index.iter_all_entries():
1586
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1589
def iter_parents(self, version_ids):
1590
"""Iterate through the parents for many version ids.
1592
:param version_ids: An iterable yielding version_ids.
1593
:return: An iterator that yields (version_id, parents). Requested
1594
version_ids not present in the versioned file are simply skipped.
1595
The order is undefined, allowing for different optimisations in
1596
the underlying implementation.
1599
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1601
present_parents = set()
1602
for node in all_nodes:
1603
all_parents.update(node[3][0])
1604
# any node we are querying must be present
1605
present_parents.add(node[1])
1606
unknown_parents = all_parents.difference(present_parents)
1607
present_parents.update(self._present_keys(unknown_parents))
1608
for node in all_nodes:
1610
for parent in node[3][0]:
1611
if parent in present_parents:
1612
parents.append(parent[0])
1613
yield node[1][0], tuple(parents)
1615
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1616
yield node[1][0], ()
1618
def num_versions(self):
1619
return len(list(self._graph_index.iter_all_entries()))
1621
__len__ = num_versions
1623
def get_versions(self):
1624
"""Get all the versions in the file. not topologically sorted."""
1625
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1627
def has_version(self, version_id):
1628
"""True if the version is in the index."""
1629
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1631
def _keys_to_version_ids(self, keys):
1632
return tuple(key[0] for key in keys)
1634
def get_position(self, version_id):
1635
"""Return details needed to access the version.
1637
:return: a tuple (index, data position, size) to hand to the access
1638
logic to get the record.
1640
node = self._get_node(version_id)
1641
bits = node[2][1:].split(' ')
1642
return node[0], int(bits[0]), int(bits[1])
1644
def get_method(self, version_id):
1645
"""Return compression method of specified version."""
1646
if not self._deltas:
1648
return self._parent_compression(self._get_node(version_id)[3][1])
1650
def _parent_compression(self, reference_list):
1651
# use the second reference list to decide if this is delta'd or not.
1652
if len(reference_list):
1657
def _get_node(self, version_id):
1658
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1660
def get_options(self, version_id):
1661
"""Return a string represention options.
1665
node = self._get_node(version_id)
1666
if not self._deltas:
1667
options = ['fulltext']
1669
options = [self._parent_compression(node[3][1])]
1670
if node[2][0] == 'N':
1671
options.append('no-eol')
1674
def get_parents(self, version_id):
1675
"""Return parents of specified version ignoring ghosts."""
1676
parents = list(self.iter_parents([version_id]))
1679
raise errors.RevisionNotPresent(version_id, self)
1680
return parents[0][1]
1682
def get_parents_with_ghosts(self, version_id):
1683
"""Return parents of specified version with ghosts."""
1684
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1685
check_present=True))
1686
if not self._parents:
1688
return self._keys_to_version_ids(nodes[0][3][0])
1690
def check_versions_present(self, version_ids):
1691
"""Check that all specified versions are present."""
1692
keys = self._version_ids_to_keys(version_ids)
1693
present = self._present_keys(keys)
1694
missing = keys.difference(present)
1696
raise RevisionNotPresent(missing.pop(), self)
1698
def add_version(self, version_id, options, access_memo, parents):
1699
"""Add a version record to the index."""
1700
return self.add_versions(((version_id, options, access_memo, parents),))
1702
def add_versions(self, versions):
1703
"""Add multiple versions to the index.
1705
This function does not insert data into the Immutable GraphIndex
1706
backing the KnitGraphIndex, instead it prepares data for insertion by
1707
the caller and checks that it is safe to insert then calls
1708
self._add_callback with the prepared GraphIndex nodes.
1710
:param versions: a list of tuples:
1711
(version_id, options, pos, size, parents).
1713
if not self._add_callback:
1714
raise errors.ReadOnlyError(self)
1715
# we hope there are no repositories with inconsistent parentage
1720
for (version_id, options, access_memo, parents) in versions:
1721
index, pos, size = access_memo
1722
key = (version_id, )
1723
parents = tuple((parent, ) for parent in parents)
1724
if 'no-eol' in options:
1728
value += "%d %d" % (pos, size)
1729
if not self._deltas:
1730
if 'line-delta' in options:
1731
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1734
if 'line-delta' in options:
1735
node_refs = (parents, (parents[0],))
1737
node_refs = (parents, ())
1739
node_refs = (parents, )
1742
raise KnitCorrupt(self, "attempt to add node with parents "
1743
"in parentless index.")
1745
keys[key] = (value, node_refs)
1746
present_nodes = self._get_entries(keys)
1747
for (index, key, value, node_refs) in present_nodes:
1748
if (value, node_refs) != keys[key]:
1749
raise KnitCorrupt(self, "inconsistent details in add_versions"
1750
": %s %s" % ((value, node_refs), keys[key]))
1754
for key, (value, node_refs) in keys.iteritems():
1755
result.append((key, value, node_refs))
1757
for key, (value, node_refs) in keys.iteritems():
1758
result.append((key, value))
1759
self._add_callback(result)
1761
def _version_ids_to_keys(self, version_ids):
1762
return set((version_id, ) for version_id in version_ids)
1765
class _KnitAccess(object):
1766
"""Access to knit records in a .knit file."""
1768
def __init__(self, transport, filename, _file_mode, _dir_mode,
1769
_need_to_create, _create_parent_dir):
1770
"""Create a _KnitAccess for accessing and inserting data.
1772
:param transport: The transport the .knit is located on.
1773
:param filename: The filename of the .knit.
1775
self._transport = transport
1776
self._filename = filename
1777
self._file_mode = _file_mode
1778
self._dir_mode = _dir_mode
1779
self._need_to_create = _need_to_create
1780
self._create_parent_dir = _create_parent_dir
1782
def add_raw_records(self, sizes, raw_data):
1783
"""Add raw knit bytes to a storage area.
1785
The data is spooled to whereever the access method is storing data.
1787
:param sizes: An iterable containing the size of each raw data segment.
1788
:param raw_data: A bytestring containing the data.
1789
:return: A list of memos to retrieve the record later. Each memo is a
1790
tuple - (index, pos, length), where the index field is always None
1791
for the .knit access method.
1793
assert type(raw_data) == str, \
1794
'data must be plain bytes was %s' % type(raw_data)
1795
if not self._need_to_create:
1796
base = self._transport.append_bytes(self._filename, raw_data)
1798
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1799
create_parent_dir=self._create_parent_dir,
1800
mode=self._file_mode,
1801
dir_mode=self._dir_mode)
1802
self._need_to_create = False
1806
result.append((None, base, size))
1811
"""IFF this data access has its own storage area, initialise it.
1815
self._transport.put_bytes_non_atomic(self._filename, '',
1816
mode=self._file_mode)
1818
def open_file(self):
1819
"""IFF this data access can be represented as a single file, open it.
1821
For knits that are not mapped to a single file on disk this will
1824
:return: None or a file handle.
1827
return self._transport.get(self._filename)
1832
def get_raw_records(self, memos_for_retrieval):
1833
"""Get the raw bytes for a records.
1835
:param memos_for_retrieval: An iterable containing the (index, pos,
1836
length) memo for retrieving the bytes. The .knit method ignores
1837
the index as there is always only a single file.
1838
:return: An iterator over the bytes of the records.
1840
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1841
for pos, data in self._transport.readv(self._filename, read_vector):
1845
class _PackAccess(object):
1846
"""Access to knit records via a collection of packs."""
1848
def __init__(self, index_to_packs, writer=None):
1849
"""Create a _PackAccess object.
1851
:param index_to_packs: A dict mapping index objects to the transport
1852
and file names for obtaining data.
1853
:param writer: A tuple (pack.ContainerWriter, write_index) which
1854
contains the pack to write, and the index that reads from it will
1858
self.container_writer = writer[0]
1859
self.write_index = writer[1]
1861
self.container_writer = None
1862
self.write_index = None
1863
self.indices = index_to_packs
1865
def add_raw_records(self, sizes, raw_data):
1866
"""Add raw knit bytes to a storage area.
1868
The data is spooled to the container writer in one bytes-record per
1871
:param sizes: An iterable containing the size of each raw data segment.
1872
:param raw_data: A bytestring containing the data.
1873
:return: A list of memos to retrieve the record later. Each memo is a
1874
tuple - (index, pos, length), where the index field is the
1875
write_index object supplied to the PackAccess object.
1877
assert type(raw_data) == str, \
1878
'data must be plain bytes was %s' % type(raw_data)
1882
p_offset, p_length = self.container_writer.add_bytes_record(
1883
raw_data[offset:offset+size], [])
1885
result.append((self.write_index, p_offset, p_length))
1889
"""Pack based knits do not get individually created."""
1891
def get_raw_records(self, memos_for_retrieval):
1892
"""Get the raw bytes for a records.
1894
:param memos_for_retrieval: An iterable containing the (index, pos,
1895
length) memo for retrieving the bytes. The Pack access method
1896
looks up the pack to use for a given record in its index_to_pack
1898
:return: An iterator over the bytes of the records.
1900
# first pass, group into same-index requests
1902
current_index = None
1903
for (index, offset, length) in memos_for_retrieval:
1904
if current_index == index:
1905
current_list.append((offset, length))
1907
if current_index is not None:
1908
request_lists.append((current_index, current_list))
1909
current_index = index
1910
current_list = [(offset, length)]
1911
# handle the last entry
1912
if current_index is not None:
1913
request_lists.append((current_index, current_list))
1914
for index, offsets in request_lists:
1915
transport, path = self.indices[index]
1916
reader = pack.make_readv_reader(transport, path, offsets)
1917
for names, read_func in reader.iter_records():
1918
yield read_func(None)
1920
def open_file(self):
1921
"""Pack based knits have no single file."""
1924
def set_writer(self, writer, index, (transport, packname)):
1925
"""Set a writer to use for adding data."""
1926
self.indices[index] = (transport, packname)
1927
self.container_writer = writer
1928
self.write_index = index
1931
class _KnitData(object):
1932
"""Manage extraction of data from a KnitAccess, caching and decompressing.
1934
The KnitData class provides the logic for parsing and using knit records,
1935
making use of an access method for the low level read and write operations.
1938
def __init__(self, access):
1939
"""Create a KnitData object.
1941
:param access: The access method to use. Access methods such as
1942
_KnitAccess manage the insertion of raw records and the subsequent
1943
retrieval of the same.
1945
self._access = access
1237
1946
self._checked = False
1239
self._transport.put(self._filename, StringIO(''), mode=file_mode)
1947
# TODO: jam 20060713 conceptually, this could spill to disk
1948
# if the cached size gets larger than a certain amount
1949
# but it complicates the model a bit, so for now just use
1950
# a simple dictionary
1952
self._do_cache = False
1954
def enable_cache(self):
1955
"""Enable caching of reads."""
1956
self._do_cache = True
1242
1958
def clear_cache(self):
1243
1959
"""Clear the record cache."""
1960
self._do_cache = False
1246
1963
def _open_file(self):
1247
if self._file is None:
1249
self._file = self._transport.get(self._filename)
1964
return self._access.open_file()
1254
1966
def _record_to_data(self, version_id, digest, lines):
1255
1967
"""Convert version_id, digest, lines into a raw data block.
1325
2069
This unpacks enough of the text record to validate the id is
1326
2070
as expected but thats all.
1328
It will actively recompress currently cached records on the
1329
basis that that is cheaper than I/O activity.
1332
for version_id, pos, size in records:
1333
if version_id not in self._records:
1334
needed_records.append((version_id, pos, size))
1336
2072
# setup an iterator of the external records:
1337
2073
# uses readv so nice and fast we hope.
1338
if len(needed_records):
1339
2075
# grab the disk data needed.
1340
raw_records = self._transport.readv(self._filename,
1341
[(pos, size) for version_id, pos, size in needed_records])
1343
for version_id, pos, size in records:
1344
if version_id in self._records:
1345
# compress a new version
1346
size, sio = self._record_to_data(version_id,
1347
self._records[version_id][0],
1348
self._records[version_id][1])
1349
yield version_id, sio.getvalue()
1351
pos, data = raw_records.next()
2077
# Don't check _cache if it is empty
2078
needed_offsets = [index_memo for version_id, index_memo
2080
if version_id not in self._cache]
2082
needed_offsets = [index_memo for version_id, index_memo
2085
raw_records = self._access.get_raw_records(needed_offsets)
2087
for version_id, index_memo in records:
2088
if version_id in self._cache:
2089
# This data has already been validated
2090
data = self._cache[version_id]
2092
data = raw_records.next()
2094
self._cache[version_id] = data
1352
2096
# validate the header
1353
2097
df, rec = self._parse_record_header(version_id, data)
1355
yield version_id, data
2099
yield version_id, data
1358
2101
def read_records_iter(self, records):
1359
2102
"""Read text records from data file and yield result.
1361
Each passed record is a tuple of (version_id, pos, len) and
1362
will be read in the given order. Yields (version_id,
2104
The result will be returned in whatever is the fastest to read.
2105
Not by the order requested. Also, multiple requests for the same
2106
record will only yield 1 response.
2107
:param records: A list of (version_id, pos, len) entries
2108
:return: Yields (version_id, contents, digest) in the order
2109
read, not the order requested
1366
# 60890 calls for 4168 extractions in 5045, 683 internal.
1367
# 4168 calls to readv in 1411
1368
# 4168 calls to parse_record in 2880
1371
for version_id, pos, size in records:
1372
if version_id not in self._records:
1373
needed_records.append((version_id, pos, size))
1375
if len(needed_records):
1376
# We take it that the transport optimizes the fetching as good
1377
# as possible (ie, reads continous ranges.)
1378
response = self._transport.readv(self._filename,
1379
[(pos, size) for version_id, pos, size in needed_records])
1381
for (record_id, pos, size), (pos, data) in izip(iter(needed_records), response):
1382
content, digest = self._parse_record(record_id, data)
1383
self._records[record_id] = (digest, content)
1385
for version_id, pos, size in records:
1386
yield version_id, list(self._records[version_id][1]), self._records[version_id][0]
2115
# Skip records we have alread seen
2116
yielded_records = set()
2117
needed_records = set()
2118
for record in records:
2119
if record[0] in self._cache:
2120
if record[0] in yielded_records:
2122
yielded_records.add(record[0])
2123
data = self._cache[record[0]]
2124
content, digest = self._parse_record(record[0], data)
2125
yield (record[0], content, digest)
2127
needed_records.add(record)
2128
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2130
needed_records = sorted(set(records), key=operator.itemgetter(1))
2132
if not needed_records:
2135
# The transport optimizes the fetching as well
2136
# (ie, reads continuous ranges.)
2137
raw_data = self._access.get_raw_records(
2138
[index_memo for version_id, index_memo in needed_records])
2140
for (version_id, index_memo), data in \
2141
izip(iter(needed_records), raw_data):
2142
content, digest = self._parse_record(version_id, data)
2144
self._cache[version_id] = data
2145
yield version_id, content, digest
1388
2147
def read_records(self, records):
1389
2148
"""Read records into a dictionary."""
1390
2149
components = {}
1391
for record_id, content, digest in self.read_records_iter(records):
2150
for record_id, content, digest in \
2151
self.read_records_iter(records):
1392
2152
components[record_id] = (content, digest)
1393
2153
return components
1603
2304
InterVersionedFile.register_optimiser(WeaveToKnit)
1606
class SequenceMatcher(difflib.SequenceMatcher):
1607
"""Knit tuned sequence matcher.
1609
This is based on profiling of difflib which indicated some improvements
1610
for our usage pattern.
2307
# Deprecated, use PatienceSequenceMatcher instead
2308
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2311
def annotate_knit(knit, revision_id):
2312
"""Annotate a knit with no cached annotations.
2314
This implementation is for knits with no cached annotations.
2315
It will work for knits with cached annotations, but this is not
1613
def find_longest_match(self, alo, ahi, blo, bhi):
1614
"""Find longest matching block in a[alo:ahi] and b[blo:bhi].
1616
If isjunk is not defined:
1618
Return (i,j,k) such that a[i:i+k] is equal to b[j:j+k], where
1619
alo <= i <= i+k <= ahi
1620
blo <= j <= j+k <= bhi
1621
and for all (i',j',k') meeting those conditions,
1624
and if i == i', j <= j'
1626
In other words, of all maximal matching blocks, return one that
1627
starts earliest in a, and of all those maximal matching blocks that
1628
start earliest in a, return the one that starts earliest in b.
1630
>>> s = SequenceMatcher(None, " abcd", "abcd abcd")
1631
>>> s.find_longest_match(0, 5, 0, 9)
1634
If isjunk is defined, first the longest matching block is
1635
determined as above, but with the additional restriction that no
1636
junk element appears in the block. Then that block is extended as
1637
far as possible by matching (only) junk elements on both sides. So
1638
the resulting block never matches on junk except as identical junk
1639
happens to be adjacent to an "interesting" match.
1641
Here's the same example as before, but considering blanks to be
1642
junk. That prevents " abcd" from matching the " abcd" at the tail
1643
end of the second sequence directly. Instead only the "abcd" can
1644
match, and matches the leftmost "abcd" in the second sequence:
1646
>>> s = SequenceMatcher(lambda x: x==" ", " abcd", "abcd abcd")
1647
>>> s.find_longest_match(0, 5, 0, 9)
1650
If no blocks match, return (alo, blo, 0).
1652
>>> s = SequenceMatcher(None, "ab", "c")
1653
>>> s.find_longest_match(0, 2, 0, 1)
1657
# CAUTION: stripping common prefix or suffix would be incorrect.
1661
# Longest matching block is "ab", but if common prefix is
1662
# stripped, it's "a" (tied with "b"). UNIX(tm) diff does so
1663
# strip, so ends up claiming that ab is changed to acab by
1664
# inserting "ca" in the middle. That's minimal but unintuitive:
1665
# "it's obvious" that someone inserted "ac" at the front.
1666
# Windiff ends up at the same place as diff, but by pairing up
1667
# the unique 'b's and then matching the first two 'a's.
1669
a, b, b2j, isbjunk = self.a, self.b, self.b2j, self.isbjunk
1670
besti, bestj, bestsize = alo, blo, 0
1671
# find longest junk-free match
1672
# during an iteration of the loop, j2len[j] = length of longest
1673
# junk-free match ending with a[i-1] and b[j]
1677
for i in xrange(alo, ahi):
1678
# look at all instances of a[i] in b; note that because
1679
# b2j has no junk keys, the loop is skipped if a[i] is junk
1680
j2lenget = j2len.get
1683
# changing b2j.get(a[i], nothing) to a try:Keyerror pair produced the
1684
# following improvement
1685
# 704 0 4650.5320 2620.7410 bzrlib.knit:1336(find_longest_match)
1686
# +326674 0 1655.1210 1655.1210 +<method 'get' of 'dict' objects>
1687
# +76519 0 374.6700 374.6700 +<method 'has_key' of 'dict' objects>
1689
# 704 0 3733.2820 2209.6520 bzrlib.knit:1336(find_longest_match)
1690
# +211400 0 1147.3520 1147.3520 +<method 'get' of 'dict' objects>
1691
# +76519 0 376.2780 376.2780 +<method 'has_key' of 'dict' objects>
1703
k = newj2len[j] = 1 + j2lenget(-1 + j, 0)
1705
besti, bestj, bestsize = 1 + i-k, 1 + j-k, k
1708
# Extend the best by non-junk elements on each end. In particular,
1709
# "popular" non-junk elements aren't in b2j, which greatly speeds
1710
# the inner loop above, but also means "the best" match so far
1711
# doesn't contain any junk *or* popular non-junk elements.
1712
while besti > alo and bestj > blo and \
1713
not isbjunk(b[bestj-1]) and \
1714
a[besti-1] == b[bestj-1]:
1715
besti, bestj, bestsize = besti-1, bestj-1, bestsize+1
1716
while besti+bestsize < ahi and bestj+bestsize < bhi and \
1717
not isbjunk(b[bestj+bestsize]) and \
1718
a[besti+bestsize] == b[bestj+bestsize]:
1721
# Now that we have a wholly interesting match (albeit possibly
1722
# empty!), we may as well suck up the matching junk on each
1723
# side of it too. Can't think of a good reason not to, and it
1724
# saves post-processing the (possibly considerable) expense of
1725
# figuring out what to do with it. In the case of an empty
1726
# interesting match, this is clearly the right thing to do,
1727
# because no other kind of match is possible in the regions.
1728
while besti > alo and bestj > blo and \
1729
isbjunk(b[bestj-1]) and \
1730
a[besti-1] == b[bestj-1]:
1731
besti, bestj, bestsize = besti-1, bestj-1, bestsize+1
1732
while besti+bestsize < ahi and bestj+bestsize < bhi and \
1733
isbjunk(b[bestj+bestsize]) and \
1734
a[besti+bestsize] == b[bestj+bestsize]:
1735
bestsize = bestsize + 1
1737
return besti, bestj, bestsize
2318
ancestry = knit.get_ancestry(revision_id)
2319
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2321
for candidate in ancestry:
2322
if candidate in annotations:
2324
parents = knit.get_parents(candidate)
2325
if len(parents) == 0:
2327
elif knit._index.get_method(candidate) != 'line-delta':
2330
parent, sha1, noeol, delta = knit.get_delta(candidate)
2331
blocks = KnitContent.get_line_delta_blocks(delta,
2332
fulltext[parents[0]], fulltext[candidate])
2333
annotations[candidate] = list(annotate.reannotate([annotations[p]
2334
for p in parents], fulltext[candidate], candidate, blocks))
2335
return iter(annotations[revision_id])
2339
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2341
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
