1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
"""Knit versionedfile implementation.
19
A knit is a versioned file implementation that supports efficient append only
23
lifeless: the data file is made up of "delta records". each delta record has a delta header
24
that contains; (1) a version id, (2) the size of the delta (in lines), and (3) the digest of
25
the -expanded data- (ie, the delta applied to the parent). the delta also ends with a
26
end-marker; simply "end VERSION"
28
delta can be line or full contents.a
29
... the 8's there are the index number of the annotation.
30
version robertc@robertcollins.net-20051003014215-ee2990904cc4c7ad 7 c7d23b2a5bd6ca00e8e266cec0ec228158ee9f9e
34
8 e.set('executable', 'yes')
36
8 if elt.get('executable') == 'yes':
37
8 ie.executable = True
38
end robertc@robertcollins.net-20051003014215-ee2990904cc4c7ad
42
09:33 < jrydberg> lifeless: each index is made up of a tuple of; version id, options, position, size, parents
43
09:33 < jrydberg> lifeless: the parents are currently dictionary compressed
44
09:33 < jrydberg> lifeless: (meaning it currently does not support ghosts)
45
09:33 < lifeless> right
46
09:33 < jrydberg> lifeless: the position and size is the range in the data file
49
so the index sequence is the dictionary compressed sequence number used
50
in the deltas to provide line annotation
55
# 10:16 < lifeless> make partial index writes safe
56
# 10:16 < lifeless> implement 'knit.check()' like weave.check()
57
# 10:17 < lifeless> record known ghosts so we can detect when they are filled in rather than the current 'reweave
59
# move sha1 out of the content so that join is faster at verifying parents
60
# record content length ?
64
from cStringIO import StringIO
65
from itertools import izip, chain
70
from zlib import Z_DEFAULT_COMPRESSION
73
from bzrlib.lazy_import import lazy_import
74
lazy_import(globals(), """
93
from bzrlib.errors import (
99
KnitDataStreamIncompatible,
102
RevisionAlreadyPresent,
104
from bzrlib.tuned_gzip import GzipFile, bytes_to_gzip
105
from bzrlib.osutils import (
111
from bzrlib.symbol_versioning import DEPRECATED_PARAMETER, deprecated_passed
112
from bzrlib.tsort import topo_sort
115
from bzrlib.versionedfile import VersionedFile, InterVersionedFile
118
# TODO: Split out code specific to this format into an associated object.
120
# TODO: Can we put in some kind of value to check that the index and data
121
# files belong together?
123
# TODO: accommodate binaries, perhaps by storing a byte count
125
# TODO: function to check whole file
127
# TODO: atomically append data, then measure backwards from the cursor
128
# position after writing to work out where it was located. we may need to
129
# bypass python file buffering.
131
DATA_SUFFIX = '.knit'
132
INDEX_SUFFIX = '.kndx'
135
class KnitContent(object):
136
"""Content of a knit version to which deltas can be applied."""
139
"""Return a list of (origin, text) tuples."""
140
return list(self.annotate_iter())
142
def apply_delta(self, delta, new_version_id):
143
"""Apply delta to this object to become new_version_id."""
144
raise NotImplementedError(self.apply_delta)
146
def line_delta_iter(self, new_lines):
147
"""Generate line-based delta from this content to new_lines."""
148
new_texts = new_lines.text()
149
old_texts = self.text()
150
s = patiencediff.PatienceSequenceMatcher(None, old_texts, new_texts)
151
for tag, i1, i2, j1, j2 in s.get_opcodes():
154
# ofrom, oto, length, data
155
yield i1, i2, j2 - j1, new_lines._lines[j1:j2]
157
def line_delta(self, new_lines):
158
return list(self.line_delta_iter(new_lines))
161
def get_line_delta_blocks(knit_delta, source, target):
162
"""Extract SequenceMatcher.get_matching_blocks() from a knit delta"""
163
target_len = len(target)
166
for s_begin, s_end, t_len, new_text in knit_delta:
167
true_n = s_begin - s_pos
170
# knit deltas do not provide reliable info about whether the
171
# last line of a file matches, due to eol handling.
172
if source[s_pos + n -1] != target[t_pos + n -1]:
175
yield s_pos, t_pos, n
176
t_pos += t_len + true_n
178
n = target_len - t_pos
180
if source[s_pos + n -1] != target[t_pos + n -1]:
183
yield s_pos, t_pos, n
184
yield s_pos + (target_len - t_pos), target_len, 0
187
class AnnotatedKnitContent(KnitContent):
188
"""Annotated content."""
190
def __init__(self, lines):
193
def annotate_iter(self):
194
"""Yield tuples of (origin, text) for each content line."""
195
return iter(self._lines)
197
def apply_delta(self, delta, new_version_id):
198
"""Apply delta to this object to become new_version_id."""
201
for start, end, count, delta_lines in delta:
202
lines[offset+start:offset+end] = delta_lines
203
offset = offset + (start - end) + count
205
def strip_last_line_newline(self):
206
line = self._lines[-1][1].rstrip('\n')
207
self._lines[-1] = (self._lines[-1][0], line)
211
return [text for origin, text in self._lines]
212
except ValueError, e:
213
# most commonly (only?) caused by the internal form of the knit
214
# missing annotation information because of a bug - see thread
216
raise KnitCorrupt(self,
217
"line in annotated knit missing annotation information: %s"
221
return AnnotatedKnitContent(self._lines[:])
224
class PlainKnitContent(KnitContent):
225
"""Unannotated content.
227
When annotate[_iter] is called on this content, the same version is reported
228
for all lines. Generally, annotate[_iter] is not useful on PlainKnitContent
232
def __init__(self, lines, version_id):
234
self._version_id = version_id
236
def annotate_iter(self):
237
"""Yield tuples of (origin, text) for each content line."""
238
for line in self._lines:
239
yield self._version_id, line
241
def apply_delta(self, delta, new_version_id):
242
"""Apply delta to this object to become new_version_id."""
245
for start, end, count, delta_lines in delta:
246
lines[offset+start:offset+end] = delta_lines
247
offset = offset + (start - end) + count
248
self._version_id = new_version_id
251
return PlainKnitContent(self._lines[:], self._version_id)
253
def strip_last_line_newline(self):
254
self._lines[-1] = self._lines[-1].rstrip('\n')
260
class KnitAnnotateFactory(object):
261
"""Factory for creating annotated Content objects."""
265
def make(self, lines, version_id):
266
num_lines = len(lines)
267
return AnnotatedKnitContent(zip([version_id] * num_lines, lines))
269
def parse_fulltext(self, content, version_id):
270
"""Convert fulltext to internal representation
272
fulltext content is of the format
273
revid(utf8) plaintext\n
274
internal representation is of the format:
277
# TODO: jam 20070209 The tests expect this to be returned as tuples,
278
# but the code itself doesn't really depend on that.
279
# Figure out a way to not require the overhead of turning the
280
# list back into tuples.
281
lines = [tuple(line.split(' ', 1)) for line in content]
282
return AnnotatedKnitContent(lines)
284
def parse_line_delta_iter(self, lines):
285
return iter(self.parse_line_delta(lines))
287
def parse_line_delta(self, lines, version_id, plain=False):
288
"""Convert a line based delta into internal representation.
290
line delta is in the form of:
291
intstart intend intcount
293
revid(utf8) newline\n
294
internal representation is
295
(start, end, count, [1..count tuples (revid, newline)])
297
:param plain: If True, the lines are returned as a plain
298
list without annotations, not as a list of (origin, content) tuples, i.e.
299
(start, end, count, [1..count newline])
306
def cache_and_return(line):
307
origin, text = line.split(' ', 1)
308
return cache.setdefault(origin, origin), text
310
# walk through the lines parsing.
311
# Note that the plain test is explicitly pulled out of the
312
# loop to minimise any performance impact
315
start, end, count = [int(n) for n in header.split(',')]
316
contents = [next().split(' ', 1)[1] for i in xrange(count)]
317
result.append((start, end, count, contents))
320
start, end, count = [int(n) for n in header.split(',')]
321
contents = [tuple(next().split(' ', 1)) for i in xrange(count)]
322
result.append((start, end, count, contents))
325
def get_fulltext_content(self, lines):
326
"""Extract just the content lines from a fulltext."""
327
return (line.split(' ', 1)[1] for line in lines)
329
def get_linedelta_content(self, lines):
330
"""Extract just the content from a line delta.
332
This doesn't return all of the extra information stored in a delta.
333
Only the actual content lines.
338
header = header.split(',')
339
count = int(header[2])
340
for i in xrange(count):
341
origin, text = next().split(' ', 1)
344
def lower_fulltext(self, content):
345
"""convert a fulltext content record into a serializable form.
347
see parse_fulltext which this inverts.
349
# TODO: jam 20070209 We only do the caching thing to make sure that
350
# the origin is a valid utf-8 line, eventually we could remove it
351
return ['%s %s' % (o, t) for o, t in content._lines]
353
def lower_line_delta(self, delta):
354
"""convert a delta into a serializable form.
356
See parse_line_delta which this inverts.
358
# TODO: jam 20070209 We only do the caching thing to make sure that
359
# the origin is a valid utf-8 line, eventually we could remove it
361
for start, end, c, lines in delta:
362
out.append('%d,%d,%d\n' % (start, end, c))
363
out.extend(origin + ' ' + text
364
for origin, text in lines)
367
def annotate_iter(self, knit, version_id):
368
content = knit._get_content(version_id)
369
return content.annotate_iter()
372
class KnitPlainFactory(object):
373
"""Factory for creating plain Content objects."""
377
def make(self, lines, version_id):
378
return PlainKnitContent(lines, version_id)
380
def parse_fulltext(self, content, version_id):
381
"""This parses an unannotated fulltext.
383
Note that this is not a noop - the internal representation
384
has (versionid, line) - its just a constant versionid.
386
return self.make(content, version_id)
388
def parse_line_delta_iter(self, lines, version_id):
390
num_lines = len(lines)
391
while cur < num_lines:
394
start, end, c = [int(n) for n in header.split(',')]
395
yield start, end, c, lines[cur:cur+c]
398
def parse_line_delta(self, lines, version_id):
399
return list(self.parse_line_delta_iter(lines, version_id))
401
def get_fulltext_content(self, lines):
402
"""Extract just the content lines from a fulltext."""
405
def get_linedelta_content(self, lines):
406
"""Extract just the content from a line delta.
408
This doesn't return all of the extra information stored in a delta.
409
Only the actual content lines.
414
header = header.split(',')
415
count = int(header[2])
416
for i in xrange(count):
419
def lower_fulltext(self, content):
420
return content.text()
422
def lower_line_delta(self, delta):
424
for start, end, c, lines in delta:
425
out.append('%d,%d,%d\n' % (start, end, c))
429
def annotate_iter(self, knit, version_id):
430
return annotate_knit(knit, version_id)
433
def make_empty_knit(transport, relpath):
434
"""Construct a empty knit at the specified location."""
435
k = KnitVersionedFile(transport, relpath, 'w', KnitPlainFactory)
438
class KnitVersionedFile(VersionedFile):
439
"""Weave-like structure with faster random access.
441
A knit stores a number of texts and a summary of the relationships
442
between them. Texts are identified by a string version-id. Texts
443
are normally stored and retrieved as a series of lines, but can
444
also be passed as single strings.
446
Lines are stored with the trailing newline (if any) included, to
447
avoid special cases for files with no final newline. Lines are
448
composed of 8-bit characters, not unicode. The combination of
449
these approaches should mean any 'binary' file can be safely
450
stored and retrieved.
453
def __init__(self, relpath, transport, file_mode=None, access_mode=None,
454
factory=None, delta=True, create=False, create_parent_dir=False,
455
delay_create=False, dir_mode=None, index=None, access_method=None):
456
"""Construct a knit at location specified by relpath.
458
:param create: If not True, only open an existing knit.
459
:param create_parent_dir: If True, create the parent directory if
460
creating the file fails. (This is used for stores with
461
hash-prefixes that may not exist yet)
462
:param delay_create: The calling code is aware that the knit won't
463
actually be created until the first data is stored.
464
:param index: An index to use for the knit.
466
if access_mode is None:
468
super(KnitVersionedFile, self).__init__(access_mode)
469
assert access_mode in ('r', 'w'), "invalid mode specified %r" % access_mode
470
self.transport = transport
471
self.filename = relpath
472
self.factory = factory or KnitAnnotateFactory()
473
self.writable = (access_mode == 'w')
476
self._max_delta_chain = 200
479
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
480
access_mode, create=create, file_mode=file_mode,
481
create_parent_dir=create_parent_dir, delay_create=delay_create,
485
if access_method is None:
486
_access = _KnitAccess(transport, relpath + DATA_SUFFIX, file_mode, dir_mode,
487
((create and not len(self)) and delay_create), create_parent_dir)
489
_access = access_method
490
if create and not len(self) and not delay_create:
492
self._data = _KnitData(_access)
495
return '%s(%s)' % (self.__class__.__name__,
496
self.transport.abspath(self.filename))
498
def _check_should_delta(self, first_parents):
499
"""Iterate back through the parent listing, looking for a fulltext.
501
This is used when we want to decide whether to add a delta or a new
502
fulltext. It searches for _max_delta_chain parents. When it finds a
503
fulltext parent, it sees if the total size of the deltas leading up to
504
it is large enough to indicate that we want a new full text anyway.
506
Return True if we should create a new delta, False if we should use a
511
delta_parents = first_parents
512
for count in xrange(self._max_delta_chain):
513
parent = delta_parents[0]
514
method = self._index.get_method(parent)
515
index, pos, size = self._index.get_position(parent)
516
if method == 'fulltext':
520
delta_parents = self._index.get_parents(parent)
522
# We couldn't find a fulltext, so we must create a new one
525
return fulltext_size > delta_size
527
def _add_raw_records(self, records, data):
528
"""Add all the records 'records' with data pre-joined in 'data'.
530
:param records: A list of tuples(version_id, options, parents, size).
531
:param data: The data for the records. When it is written, the records
532
are adjusted to have pos pointing into data by the sum of
533
the preceding records sizes.
536
raw_record_sizes = [record[3] for record in records]
537
positions = self._data.add_raw_records(raw_record_sizes, data)
540
for (version_id, options, parents, size), access_memo in zip(
542
index_entries.append((version_id, options, access_memo, parents))
543
if self._data._do_cache:
544
self._data._cache[version_id] = data[offset:offset+size]
546
self._index.add_versions(index_entries)
548
def enable_cache(self):
549
"""Start caching data for this knit"""
550
self._data.enable_cache()
552
def clear_cache(self):
553
"""Clear the data cache only."""
554
self._data.clear_cache()
556
def copy_to(self, name, transport):
557
"""See VersionedFile.copy_to()."""
558
# copy the current index to a temp index to avoid racing with local
560
transport.put_file_non_atomic(name + INDEX_SUFFIX + '.tmp',
561
self.transport.get(self._index._filename))
563
f = self._data._open_file()
565
transport.put_file(name + DATA_SUFFIX, f)
568
# move the copied index into place
569
transport.move(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
571
def create_empty(self, name, transport, mode=None):
572
return KnitVersionedFile(name, transport, factory=self.factory,
573
delta=self.delta, create=True)
575
def get_data_stream(self, required_versions):
576
"""Get a data stream for the specified versions.
578
Versions may be returned in any order, not necessarily the order
581
:param required_versions: The exact set of versions to be extracted.
582
Unlike some other knit methods, this is not used to generate a
583
transitive closure, rather it is used precisely as given.
585
:returns: format_signature, list of (version, options, length, parents),
588
if not isinstance(required_versions, set):
589
required_versions = set(required_versions)
590
# we don't care about inclusions, the caller cares.
591
# but we need to setup a list of records to visit.
592
for version_id in required_versions:
593
if not self.has_version(version_id):
594
raise RevisionNotPresent(version_id, self.filename)
595
# Pick the desired versions out of the index in oldest-to-newest order
597
for version_id in self.versions():
598
if version_id in required_versions:
599
version_list.append(version_id)
601
# create the list of version information for the result
602
copy_queue_records = []
604
result_version_list = []
605
for version_id in version_list:
606
options = self._index.get_options(version_id)
607
parents = self._index.get_parents_with_ghosts(version_id)
608
index_memo = self._index.get_position(version_id)
609
copy_queue_records.append((version_id, index_memo))
610
none, data_pos, data_size = index_memo
611
copy_set.add(version_id)
612
# version, options, length, parents
613
result_version_list.append((version_id, options, data_size,
616
# Read the compressed record data.
618
# From here down to the return should really be logic in the returned
619
# callable -- in a class that adapts read_records_iter_raw to read
622
for (version_id, raw_data), \
623
(version_id2, options, _, parents) in \
624
izip(self._data.read_records_iter_raw(copy_queue_records),
625
result_version_list):
626
assert version_id == version_id2, 'logic error, inconsistent results'
627
raw_datum.append(raw_data)
628
pseudo_file = StringIO(''.join(raw_datum))
631
return pseudo_file.read()
633
return pseudo_file.read(length)
634
return (self.get_format_signature(), result_version_list, read)
636
def _extract_blocks(self, version_id, source, target):
637
if self._index.get_method(version_id) != 'line-delta':
639
parent, sha1, noeol, delta = self.get_delta(version_id)
640
return KnitContent.get_line_delta_blocks(delta, source, target)
642
def get_delta(self, version_id):
643
"""Get a delta for constructing version from some other version."""
644
self.check_not_reserved_id(version_id)
645
parents = self.get_parents(version_id)
650
index_memo = self._index.get_position(version_id)
651
data, sha1 = self._data.read_records(((version_id, index_memo),))[version_id]
652
noeol = 'no-eol' in self._index.get_options(version_id)
653
if 'fulltext' == self._index.get_method(version_id):
654
new_content = self.factory.parse_fulltext(data, version_id)
655
if parent is not None:
656
reference_content = self._get_content(parent)
657
old_texts = reference_content.text()
660
new_texts = new_content.text()
661
delta_seq = patiencediff.PatienceSequenceMatcher(None, old_texts,
663
return parent, sha1, noeol, self._make_line_delta(delta_seq, new_content)
665
delta = self.factory.parse_line_delta(data, version_id)
666
return parent, sha1, noeol, delta
668
def get_format_signature(self):
669
"""See VersionedFile.get_format_signature()."""
670
if self.factory.annotated:
671
annotated_part = "annotated"
673
annotated_part = "plain"
674
return "knit-%s" % (annotated_part,)
676
def get_graph_with_ghosts(self):
677
"""See VersionedFile.get_graph_with_ghosts()."""
678
graph_items = self._index.get_graph()
679
return dict(graph_items)
681
def get_sha1(self, version_id):
682
return self.get_sha1s([version_id])[0]
684
def get_sha1s(self, version_ids):
685
"""See VersionedFile.get_sha1()."""
686
record_map = self._get_record_map(version_ids)
687
# record entry 2 is the 'digest'.
688
return [record_map[v][2] for v in version_ids]
692
"""See VersionedFile.get_suffixes()."""
693
return [DATA_SUFFIX, INDEX_SUFFIX]
695
def has_ghost(self, version_id):
696
"""True if there is a ghost reference in the file to version_id."""
698
if self.has_version(version_id):
700
# optimisable if needed by memoising the _ghosts set.
701
items = self._index.get_graph()
702
for node, parents in items:
703
for parent in parents:
704
if parent not in self._index._cache:
705
if parent == version_id:
709
def insert_data_stream(self, (format, data_list, reader_callable)):
710
"""Insert knit records from a data stream into this knit.
712
If a version in the stream is already present in this knit, it will not
713
be inserted a second time. It will be checked for consistency with the
714
stored version however, and may cause a KnitCorrupt error to be raised
715
if the data in the stream disagrees with the already stored data.
717
:seealso: get_data_stream
719
if format != self.get_format_signature():
720
trace.mutter('incompatible format signature inserting to %r', self)
721
raise KnitDataStreamIncompatible(
722
format, self.get_format_signature())
724
for version_id, options, length, parents in data_list:
725
if self.has_version(version_id):
726
# First check: the list of parents.
727
my_parents = self.get_parents_with_ghosts(version_id)
728
if my_parents != parents:
729
# XXX: KnitCorrupt is not quite the right exception here.
732
'parents list %r from data stream does not match '
733
'already recorded parents %r for %s'
734
% (parents, my_parents, version_id))
736
# Also check the SHA-1 of the fulltext this content will
738
raw_data = reader_callable(length)
739
my_fulltext_sha1 = self.get_sha1(version_id)
740
df, rec = self._data._parse_record_header(version_id, raw_data)
741
stream_fulltext_sha1 = rec[3]
742
if my_fulltext_sha1 != stream_fulltext_sha1:
743
# Actually, we don't know if it's this knit that's corrupt,
744
# or the data stream we're trying to insert.
746
self.filename, 'sha-1 does not match %s' % version_id)
748
if 'line-delta' in options:
749
# Make sure that this knit record is actually useful: a
750
# line-delta is no use unless we have its parent.
751
# Fetching from a broken repository with this problem
752
# shouldn't break the target repository.
753
if not self._index.has_version(parents[0]):
756
'line-delta from stream references '
757
'missing parent %s' % parents[0])
758
self._add_raw_records(
759
[(version_id, options, parents, length)],
760
reader_callable(length))
763
"""See VersionedFile.versions."""
764
if 'evil' in debug.debug_flags:
765
trace.mutter_callsite(2, "versions scales with size of history")
766
return self._index.get_versions()
768
def has_version(self, version_id):
769
"""See VersionedFile.has_version."""
770
if 'evil' in debug.debug_flags:
771
trace.mutter_callsite(2, "has_version is a LBYL scenario")
772
return self._index.has_version(version_id)
774
__contains__ = has_version
776
def _merge_annotations(self, content, parents, parent_texts={},
777
delta=None, annotated=None,
778
left_matching_blocks=None):
779
"""Merge annotations for content. This is done by comparing
780
the annotations based on changed to the text.
782
if left_matching_blocks is not None:
783
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
787
for parent_id in parents:
788
merge_content = self._get_content(parent_id, parent_texts)
789
if (parent_id == parents[0] and delta_seq is not None):
792
seq = patiencediff.PatienceSequenceMatcher(
793
None, merge_content.text(), content.text())
794
for i, j, n in seq.get_matching_blocks():
797
# this appears to copy (origin, text) pairs across to the
798
# new content for any line that matches the last-checked
800
content._lines[j:j+n] = merge_content._lines[i:i+n]
802
if delta_seq is None:
803
reference_content = self._get_content(parents[0], parent_texts)
804
new_texts = content.text()
805
old_texts = reference_content.text()
806
delta_seq = patiencediff.PatienceSequenceMatcher(
807
None, old_texts, new_texts)
808
return self._make_line_delta(delta_seq, content)
810
def _make_line_delta(self, delta_seq, new_content):
811
"""Generate a line delta from delta_seq and new_content."""
813
for op in delta_seq.get_opcodes():
816
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
819
def _get_components_positions(self, version_ids):
820
"""Produce a map of position data for the components of versions.
822
This data is intended to be used for retrieving the knit records.
824
A dict of version_id to (method, data_pos, data_size, next) is
826
method is the way referenced data should be applied.
827
data_pos is the position of the data in the knit.
828
data_size is the size of the data in the knit.
829
next is the build-parent of the version, or None for fulltexts.
832
for version_id in version_ids:
835
while cursor is not None and cursor not in component_data:
836
method = self._index.get_method(cursor)
837
if method == 'fulltext':
840
next = self.get_parents_with_ghosts(cursor)[0]
841
index_memo = self._index.get_position(cursor)
842
component_data[cursor] = (method, index_memo, next)
844
return component_data
846
def _get_content(self, version_id, parent_texts={}):
847
"""Returns a content object that makes up the specified
849
cached_version = parent_texts.get(version_id, None)
850
if cached_version is not None:
851
if not self.has_version(version_id):
852
raise RevisionNotPresent(version_id, self.filename)
853
return cached_version
855
text_map, contents_map = self._get_content_maps([version_id])
856
return contents_map[version_id]
858
def _check_versions_present(self, version_ids):
859
"""Check that all specified versions are present."""
860
self._index.check_versions_present(version_ids)
862
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
863
nostore_sha, random_id, check_content):
864
"""See VersionedFile.add_lines_with_ghosts()."""
865
self._check_add(version_id, lines, random_id, check_content)
866
return self._add(version_id, lines, parents, self.delta,
867
parent_texts, None, nostore_sha, random_id)
869
def _add_lines(self, version_id, parents, lines, parent_texts,
870
left_matching_blocks, nostore_sha, random_id, check_content):
871
"""See VersionedFile.add_lines."""
872
self._check_add(version_id, lines, random_id, check_content)
873
self._check_versions_present(parents)
874
return self._add(version_id, lines[:], parents, self.delta,
875
parent_texts, left_matching_blocks, nostore_sha, random_id)
877
def _check_add(self, version_id, lines, random_id, check_content):
878
"""check that version_id and lines are safe to add."""
879
if contains_whitespace(version_id):
880
raise InvalidRevisionId(version_id, self.filename)
881
self.check_not_reserved_id(version_id)
882
# Technically this could be avoided if we are happy to allow duplicate
883
# id insertion when other things than bzr core insert texts, but it
884
# seems useful for folk using the knit api directly to have some safety
885
# blanket that we can disable.
886
if not random_id and self.has_version(version_id):
887
raise RevisionAlreadyPresent(version_id, self.filename)
889
self._check_lines_not_unicode(lines)
890
self._check_lines_are_lines(lines)
892
def _add(self, version_id, lines, parents, delta, parent_texts,
893
left_matching_blocks, nostore_sha, random_id):
894
"""Add a set of lines on top of version specified by parents.
896
If delta is true, compress the text as a line-delta against
899
Any versions not present will be converted into ghosts.
901
# first thing, if the content is something we don't need to store, find
903
line_bytes = ''.join(lines)
904
digest = sha_string(line_bytes)
905
if nostore_sha == digest:
906
raise errors.ExistingContent
909
if parent_texts is None:
911
for parent in parents:
912
if self.has_version(parent):
913
present_parents.append(parent)
915
# can only compress against the left most present parent.
917
(len(present_parents) == 0 or
918
present_parents[0] != parents[0])):
921
text_length = len(line_bytes)
924
if lines[-1][-1] != '\n':
925
# copy the contents of lines.
927
options.append('no-eol')
928
lines[-1] = lines[-1] + '\n'
932
# To speed the extract of texts the delta chain is limited
933
# to a fixed number of deltas. This should minimize both
934
# I/O and the time spend applying deltas.
935
delta = self._check_should_delta(present_parents)
937
assert isinstance(version_id, str)
938
content = self.factory.make(lines, version_id)
939
if delta or (self.factory.annotated and len(present_parents) > 0):
940
# Merge annotations from parent texts if needed.
941
delta_hunks = self._merge_annotations(content, present_parents,
942
parent_texts, delta, self.factory.annotated,
943
left_matching_blocks)
946
options.append('line-delta')
947
store_lines = self.factory.lower_line_delta(delta_hunks)
948
size, bytes = self._data._record_to_data(version_id, digest,
951
options.append('fulltext')
952
# isinstance is slower and we have no hierarchy.
953
if self.factory.__class__ == KnitPlainFactory:
954
# Use the already joined bytes saving iteration time in
956
size, bytes = self._data._record_to_data(version_id, digest,
959
# get mixed annotation + content and feed it into the
961
store_lines = self.factory.lower_fulltext(content)
962
size, bytes = self._data._record_to_data(version_id, digest,
965
access_memo = self._data.add_raw_records([size], bytes)[0]
966
self._index.add_versions(
967
((version_id, options, access_memo, parents),),
969
return digest, text_length, content
971
def check(self, progress_bar=None):
972
"""See VersionedFile.check()."""
974
def _clone_text(self, new_version_id, old_version_id, parents):
975
"""See VersionedFile.clone_text()."""
976
# FIXME RBC 20060228 make fast by only inserting an index with null
978
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
980
def get_lines(self, version_id):
981
"""See VersionedFile.get_lines()."""
982
return self.get_line_list([version_id])[0]
984
def _get_record_map(self, version_ids):
985
"""Produce a dictionary of knit records.
987
The keys are version_ids, the values are tuples of (method, content,
989
method is the way the content should be applied.
990
content is a KnitContent object.
991
digest is the SHA1 digest of this version id after all steps are done
992
next is the build-parent of the version, i.e. the leftmost ancestor.
993
If the method is fulltext, next will be None.
995
position_map = self._get_components_positions(version_ids)
996
# c = component_id, m = method, i_m = index_memo, n = next
997
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
999
for component_id, content, digest in \
1000
self._data.read_records_iter(records):
1001
method, index_memo, next = position_map[component_id]
1002
record_map[component_id] = method, content, digest, next
1006
def get_text(self, version_id):
1007
"""See VersionedFile.get_text"""
1008
return self.get_texts([version_id])[0]
1010
def get_texts(self, version_ids):
1011
return [''.join(l) for l in self.get_line_list(version_ids)]
1013
def get_line_list(self, version_ids):
1014
"""Return the texts of listed versions as a list of strings."""
1015
for version_id in version_ids:
1016
self.check_not_reserved_id(version_id)
1017
text_map, content_map = self._get_content_maps(version_ids)
1018
return [text_map[v] for v in version_ids]
1020
_get_lf_split_line_list = get_line_list
1022
def _get_content_maps(self, version_ids):
1023
"""Produce maps of text and KnitContents
1025
:return: (text_map, content_map) where text_map contains the texts for
1026
the requested versions and content_map contains the KnitContents.
1027
Both dicts take version_ids as their keys.
1029
# FUTURE: This function could be improved for the 'extract many' case
1030
# by tracking each component and only doing the copy when the number of
1031
# children than need to apply delta's to it is > 1 or it is part of the
1033
version_ids = list(version_ids)
1034
multiple_versions = len(version_ids) != 1
1035
record_map = self._get_record_map(version_ids)
1040
for version_id in version_ids:
1043
while cursor is not None:
1044
method, data, digest, next = record_map[cursor]
1045
components.append((cursor, method, data, digest))
1046
if cursor in content_map:
1051
for component_id, method, data, digest in reversed(components):
1052
if component_id in content_map:
1053
content = content_map[component_id]
1055
if method == 'fulltext':
1056
assert content is None
1057
content = self.factory.parse_fulltext(data, version_id)
1058
elif method == 'line-delta':
1059
delta = self.factory.parse_line_delta(data, version_id)
1060
if multiple_versions:
1061
# only doing this when we want multiple versions
1062
# output avoids list copies - which reference and
1063
# dereference many strings.
1064
content = content.copy()
1065
content.apply_delta(delta, version_id)
1066
if multiple_versions:
1067
content_map[component_id] = content
1069
if 'no-eol' in self._index.get_options(version_id):
1070
if multiple_versions:
1071
content = content.copy()
1072
content.strip_last_line_newline()
1073
final_content[version_id] = content
1075
# digest here is the digest from the last applied component.
1076
text = content.text()
1077
actual_sha = sha_strings(text)
1078
if actual_sha != digest:
1079
raise KnitCorrupt(self.filename,
1081
'\n of reconstructed text does not match'
1083
'\n for version %s' %
1084
(actual_sha, digest, version_id))
1085
text_map[version_id] = text
1086
return text_map, final_content
1088
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1090
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1091
if version_ids is None:
1092
version_ids = self.versions()
1094
pb = progress.DummyProgress()
1095
# we don't care about inclusions, the caller cares.
1096
# but we need to setup a list of records to visit.
1097
# we need version_id, position, length
1098
version_id_records = []
1099
requested_versions = set(version_ids)
1100
# filter for available versions
1101
for version_id in requested_versions:
1102
if not self.has_version(version_id):
1103
raise RevisionNotPresent(version_id, self.filename)
1104
# get a in-component-order queue:
1105
for version_id in self.versions():
1106
if version_id in requested_versions:
1107
index_memo = self._index.get_position(version_id)
1108
version_id_records.append((version_id, index_memo))
1110
total = len(version_id_records)
1111
for version_idx, (version_id, data, sha_value) in \
1112
enumerate(self._data.read_records_iter(version_id_records)):
1113
pb.update('Walking content.', version_idx, total)
1114
method = self._index.get_method(version_id)
1116
assert method in ('fulltext', 'line-delta')
1117
if method == 'fulltext':
1118
line_iterator = self.factory.get_fulltext_content(data)
1120
line_iterator = self.factory.get_linedelta_content(data)
1121
# XXX: It might be more efficient to yield (version_id,
1122
# line_iterator) in the future. However for now, this is a simpler
1123
# change to integrate into the rest of the codebase. RBC 20071110
1124
for line in line_iterator:
1125
yield line, version_id
1127
pb.update('Walking content.', total, total)
1129
def iter_parents(self, version_ids):
1130
"""Iterate through the parents for many version ids.
1132
:param version_ids: An iterable yielding version_ids.
1133
:return: An iterator that yields (version_id, parents). Requested
1134
version_ids not present in the versioned file are simply skipped.
1135
The order is undefined, allowing for different optimisations in
1136
the underlying implementation.
1138
return self._index.iter_parents(version_ids)
1140
def num_versions(self):
1141
"""See VersionedFile.num_versions()."""
1142
return self._index.num_versions()
1144
__len__ = num_versions
1146
def annotate_iter(self, version_id):
1147
"""See VersionedFile.annotate_iter."""
1148
return self.factory.annotate_iter(self, version_id)
1150
def get_parents(self, version_id):
1151
"""See VersionedFile.get_parents."""
1154
# 52554 calls in 1264 872 internal down from 3674
1156
return self._index.get_parents(version_id)
1158
raise RevisionNotPresent(version_id, self.filename)
1160
def get_parents_with_ghosts(self, version_id):
1161
"""See VersionedFile.get_parents."""
1163
return self._index.get_parents_with_ghosts(version_id)
1165
raise RevisionNotPresent(version_id, self.filename)
1167
def get_ancestry(self, versions, topo_sorted=True):
1168
"""See VersionedFile.get_ancestry."""
1169
if isinstance(versions, basestring):
1170
versions = [versions]
1173
return self._index.get_ancestry(versions, topo_sorted)
1175
def get_ancestry_with_ghosts(self, versions):
1176
"""See VersionedFile.get_ancestry_with_ghosts."""
1177
if isinstance(versions, basestring):
1178
versions = [versions]
1181
return self._index.get_ancestry_with_ghosts(versions)
1183
def plan_merge(self, ver_a, ver_b):
1184
"""See VersionedFile.plan_merge."""
1185
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1186
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1187
annotated_a = self.annotate(ver_a)
1188
annotated_b = self.annotate(ver_b)
1189
return merge._plan_annotate_merge(annotated_a, annotated_b,
1190
ancestors_a, ancestors_b)
1193
class _KnitComponentFile(object):
1194
"""One of the files used to implement a knit database"""
1196
def __init__(self, transport, filename, mode, file_mode=None,
1197
create_parent_dir=False, dir_mode=None):
1198
self._transport = transport
1199
self._filename = filename
1201
self._file_mode = file_mode
1202
self._dir_mode = dir_mode
1203
self._create_parent_dir = create_parent_dir
1204
self._need_to_create = False
1206
def _full_path(self):
1207
"""Return the full path to this file."""
1208
return self._transport.base + self._filename
1210
def check_header(self, fp):
1211
line = fp.readline()
1213
# An empty file can actually be treated as though the file doesn't
1215
raise errors.NoSuchFile(self._full_path())
1216
if line != self.HEADER:
1217
raise KnitHeaderError(badline=line,
1218
filename=self._transport.abspath(self._filename))
1221
return '%s(%s)' % (self.__class__.__name__, self._filename)
1224
class _KnitIndex(_KnitComponentFile):
1225
"""Manages knit index file.
1227
The index is already kept in memory and read on startup, to enable
1228
fast lookups of revision information. The cursor of the index
1229
file is always pointing to the end, making it easy to append
1232
_cache is a cache for fast mapping from version id to a Index
1235
_history is a cache for fast mapping from indexes to version ids.
1237
The index data format is dictionary compressed when it comes to
1238
parent references; a index entry may only have parents that with a
1239
lover index number. As a result, the index is topological sorted.
1241
Duplicate entries may be written to the index for a single version id
1242
if this is done then the latter one completely replaces the former:
1243
this allows updates to correct version and parent information.
1244
Note that the two entries may share the delta, and that successive
1245
annotations and references MUST point to the first entry.
1247
The index file on disc contains a header, followed by one line per knit
1248
record. The same revision can be present in an index file more than once.
1249
The first occurrence gets assigned a sequence number starting from 0.
1251
The format of a single line is
1252
REVISION_ID FLAGS BYTE_OFFSET LENGTH( PARENT_ID|PARENT_SEQUENCE_ID)* :\n
1253
REVISION_ID is a utf8-encoded revision id
1254
FLAGS is a comma separated list of flags about the record. Values include
1255
no-eol, line-delta, fulltext.
1256
BYTE_OFFSET is the ascii representation of the byte offset in the data file
1257
that the the compressed data starts at.
1258
LENGTH is the ascii representation of the length of the data file.
1259
PARENT_ID a utf-8 revision id prefixed by a '.' that is a parent of
1261
PARENT_SEQUENCE_ID the ascii representation of the sequence number of a
1262
revision id already in the knit that is a parent of REVISION_ID.
1263
The ' :' marker is the end of record marker.
1266
when a write is interrupted to the index file, it will result in a line
1267
that does not end in ' :'. If the ' :' is not present at the end of a line,
1268
or at the end of the file, then the record that is missing it will be
1269
ignored by the parser.
1271
When writing new records to the index file, the data is preceded by '\n'
1272
to ensure that records always start on new lines even if the last write was
1273
interrupted. As a result its normal for the last line in the index to be
1274
missing a trailing newline. One can be added with no harmful effects.
1277
HEADER = "# bzr knit index 8\n"
1279
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1280
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1282
def _cache_version(self, version_id, options, pos, size, parents):
1283
"""Cache a version record in the history array and index cache.
1285
This is inlined into _load_data for performance. KEEP IN SYNC.
1286
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1289
# only want the _history index to reference the 1st index entry
1291
if version_id not in self._cache:
1292
index = len(self._history)
1293
self._history.append(version_id)
1295
index = self._cache[version_id][5]
1296
self._cache[version_id] = (version_id,
1303
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1304
create_parent_dir=False, delay_create=False, dir_mode=None):
1305
_KnitComponentFile.__init__(self, transport, filename, mode,
1306
file_mode=file_mode,
1307
create_parent_dir=create_parent_dir,
1310
# position in _history is the 'official' index for a revision
1311
# but the values may have come from a newer entry.
1312
# so - wc -l of a knit index is != the number of unique names
1316
fp = self._transport.get(self._filename)
1318
# _load_data may raise NoSuchFile if the target knit is
1320
_load_data(self, fp)
1324
if mode != 'w' or not create:
1327
self._need_to_create = True
1329
self._transport.put_bytes_non_atomic(
1330
self._filename, self.HEADER, mode=self._file_mode)
1332
def get_graph(self):
1333
"""Return a list of the node:parents lists from this knit index."""
1334
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1336
def get_ancestry(self, versions, topo_sorted=True):
1337
"""See VersionedFile.get_ancestry."""
1338
# get a graph of all the mentioned versions:
1340
pending = set(versions)
1343
version = pending.pop()
1346
parents = [p for p in cache[version][4] if p in cache]
1348
raise RevisionNotPresent(version, self._filename)
1349
# if not completed and not a ghost
1350
pending.update([p for p in parents if p not in graph])
1351
graph[version] = parents
1354
return topo_sort(graph.items())
1356
def get_ancestry_with_ghosts(self, versions):
1357
"""See VersionedFile.get_ancestry_with_ghosts."""
1358
# get a graph of all the mentioned versions:
1359
self.check_versions_present(versions)
1362
pending = set(versions)
1364
version = pending.pop()
1366
parents = cache[version][4]
1372
pending.update([p for p in parents if p not in graph])
1373
graph[version] = parents
1374
return topo_sort(graph.items())
1376
def iter_parents(self, version_ids):
1377
"""Iterate through the parents for many version ids.
1379
:param version_ids: An iterable yielding version_ids.
1380
:return: An iterator that yields (version_id, parents). Requested
1381
version_ids not present in the versioned file are simply skipped.
1382
The order is undefined, allowing for different optimisations in
1383
the underlying implementation.
1385
for version_id in version_ids:
1387
yield version_id, tuple(self.get_parents(version_id))
1391
def num_versions(self):
1392
return len(self._history)
1394
__len__ = num_versions
1396
def get_versions(self):
1397
"""Get all the versions in the file. not topologically sorted."""
1398
return self._history
1400
def _version_list_to_index(self, versions):
1403
for version in versions:
1404
if version in cache:
1405
# -- inlined lookup() --
1406
result_list.append(str(cache[version][5]))
1407
# -- end lookup () --
1409
result_list.append('.' + version)
1410
return ' '.join(result_list)
1412
def add_version(self, version_id, options, index_memo, parents):
1413
"""Add a version record to the index."""
1414
self.add_versions(((version_id, options, index_memo, parents),))
1416
def add_versions(self, versions, random_id=False):
1417
"""Add multiple versions to the index.
1419
:param versions: a list of tuples:
1420
(version_id, options, pos, size, parents).
1421
:param random_id: If True the ids being added were randomly generated
1422
and no check for existence will be performed.
1425
orig_history = self._history[:]
1426
orig_cache = self._cache.copy()
1429
for version_id, options, (index, pos, size), parents in versions:
1430
line = "\n%s %s %s %s %s :" % (version_id,
1434
self._version_list_to_index(parents))
1435
assert isinstance(line, str), \
1436
'content must be utf-8 encoded: %r' % (line,)
1438
self._cache_version(version_id, options, pos, size, parents)
1439
if not self._need_to_create:
1440
self._transport.append_bytes(self._filename, ''.join(lines))
1443
sio.write(self.HEADER)
1444
sio.writelines(lines)
1446
self._transport.put_file_non_atomic(self._filename, sio,
1447
create_parent_dir=self._create_parent_dir,
1448
mode=self._file_mode,
1449
dir_mode=self._dir_mode)
1450
self._need_to_create = False
1452
# If any problems happen, restore the original values and re-raise
1453
self._history = orig_history
1454
self._cache = orig_cache
1457
def has_version(self, version_id):
1458
"""True if the version is in the index."""
1459
return version_id in self._cache
1461
def get_position(self, version_id):
1462
"""Return details needed to access the version.
1464
.kndx indices do not support split-out data, so return None for the
1467
:return: a tuple (None, data position, size) to hand to the access
1468
logic to get the record.
1470
entry = self._cache[version_id]
1471
return None, entry[2], entry[3]
1473
def get_method(self, version_id):
1474
"""Return compression method of specified version."""
1476
options = self._cache[version_id][1]
1478
raise RevisionNotPresent(version_id, self._filename)
1479
if 'fulltext' in options:
1482
if 'line-delta' not in options:
1483
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1486
def get_options(self, version_id):
1487
"""Return a string represention options.
1491
return self._cache[version_id][1]
1493
def get_parents(self, version_id):
1494
"""Return parents of specified version ignoring ghosts."""
1495
return [parent for parent in self._cache[version_id][4]
1496
if parent in self._cache]
1498
def get_parents_with_ghosts(self, version_id):
1499
"""Return parents of specified version with ghosts."""
1500
return self._cache[version_id][4]
1502
def check_versions_present(self, version_ids):
1503
"""Check that all specified versions are present."""
1505
for version_id in version_ids:
1506
if version_id not in cache:
1507
raise RevisionNotPresent(version_id, self._filename)
1510
class KnitGraphIndex(object):
1511
"""A knit index that builds on GraphIndex."""
1513
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1514
"""Construct a KnitGraphIndex on a graph_index.
1516
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1517
:param deltas: Allow delta-compressed records.
1518
:param add_callback: If not None, allow additions to the index and call
1519
this callback with a list of added GraphIndex nodes:
1520
[(node, value, node_refs), ...]
1521
:param parents: If True, record knits parents, if not do not record
1524
self._graph_index = graph_index
1525
self._deltas = deltas
1526
self._add_callback = add_callback
1527
self._parents = parents
1528
if deltas and not parents:
1529
raise KnitCorrupt(self, "Cannot do delta compression without "
1532
def _get_entries(self, keys, check_present=False):
1533
"""Get the entries for keys.
1535
:param keys: An iterable of index keys, - 1-tuples.
1540
for node in self._graph_index.iter_entries(keys):
1542
found_keys.add(node[1])
1544
# adapt parentless index to the rest of the code.
1545
for node in self._graph_index.iter_entries(keys):
1546
yield node[0], node[1], node[2], ()
1547
found_keys.add(node[1])
1549
missing_keys = keys.difference(found_keys)
1551
raise RevisionNotPresent(missing_keys.pop(), self)
1553
def _present_keys(self, version_ids):
1555
node[1] for node in self._get_entries(version_ids)])
1557
def _parentless_ancestry(self, versions):
1558
"""Honour the get_ancestry API for parentless knit indices."""
1559
wanted_keys = self._version_ids_to_keys(versions)
1560
present_keys = self._present_keys(wanted_keys)
1561
missing = set(wanted_keys).difference(present_keys)
1563
raise RevisionNotPresent(missing.pop(), self)
1564
return list(self._keys_to_version_ids(present_keys))
1566
def get_ancestry(self, versions, topo_sorted=True):
1567
"""See VersionedFile.get_ancestry."""
1568
if not self._parents:
1569
return self._parentless_ancestry(versions)
1570
# XXX: This will do len(history) index calls - perhaps
1571
# it should be altered to be a index core feature?
1572
# get a graph of all the mentioned versions:
1575
versions = self._version_ids_to_keys(versions)
1576
pending = set(versions)
1578
# get all pending nodes
1579
this_iteration = pending
1580
new_nodes = self._get_entries(this_iteration)
1583
for (index, key, value, node_refs) in new_nodes:
1584
# dont ask for ghosties - otherwise
1585
# we we can end up looping with pending
1586
# being entirely ghosted.
1587
graph[key] = [parent for parent in node_refs[0]
1588
if parent not in ghosts]
1590
for parent in graph[key]:
1591
# dont examine known nodes again
1596
ghosts.update(this_iteration.difference(found))
1597
if versions.difference(graph):
1598
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1600
result_keys = topo_sort(graph.items())
1602
result_keys = graph.iterkeys()
1603
return [key[0] for key in result_keys]
1605
def get_ancestry_with_ghosts(self, versions):
1606
"""See VersionedFile.get_ancestry."""
1607
if not self._parents:
1608
return self._parentless_ancestry(versions)
1609
# XXX: This will do len(history) index calls - perhaps
1610
# it should be altered to be a index core feature?
1611
# get a graph of all the mentioned versions:
1613
versions = self._version_ids_to_keys(versions)
1614
pending = set(versions)
1616
# get all pending nodes
1617
this_iteration = pending
1618
new_nodes = self._get_entries(this_iteration)
1620
for (index, key, value, node_refs) in new_nodes:
1621
graph[key] = node_refs[0]
1623
for parent in graph[key]:
1624
# dont examine known nodes again
1628
missing_versions = this_iteration.difference(graph)
1629
missing_needed = versions.intersection(missing_versions)
1631
raise RevisionNotPresent(missing_needed.pop(), self)
1632
for missing_version in missing_versions:
1633
# add a key, no parents
1634
graph[missing_version] = []
1635
pending.discard(missing_version) # don't look for it
1636
result_keys = topo_sort(graph.items())
1637
return [key[0] for key in result_keys]
1639
def get_graph(self):
1640
"""Return a list of the node:parents lists from this knit index."""
1641
if not self._parents:
1642
return [(key, ()) for key in self.get_versions()]
1644
for index, key, value, refs in self._graph_index.iter_all_entries():
1645
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1648
def iter_parents(self, version_ids):
1649
"""Iterate through the parents for many version ids.
1651
:param version_ids: An iterable yielding version_ids.
1652
:return: An iterator that yields (version_id, parents). Requested
1653
version_ids not present in the versioned file are simply skipped.
1654
The order is undefined, allowing for different optimisations in
1655
the underlying implementation.
1658
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1660
present_parents = set()
1661
for node in all_nodes:
1662
all_parents.update(node[3][0])
1663
# any node we are querying must be present
1664
present_parents.add(node[1])
1665
unknown_parents = all_parents.difference(present_parents)
1666
present_parents.update(self._present_keys(unknown_parents))
1667
for node in all_nodes:
1669
for parent in node[3][0]:
1670
if parent in present_parents:
1671
parents.append(parent[0])
1672
yield node[1][0], tuple(parents)
1674
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1675
yield node[1][0], ()
1677
def num_versions(self):
1678
return len(list(self._graph_index.iter_all_entries()))
1680
__len__ = num_versions
1682
def get_versions(self):
1683
"""Get all the versions in the file. not topologically sorted."""
1684
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1686
def has_version(self, version_id):
1687
"""True if the version is in the index."""
1688
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1690
def _keys_to_version_ids(self, keys):
1691
return tuple(key[0] for key in keys)
1693
def get_position(self, version_id):
1694
"""Return details needed to access the version.
1696
:return: a tuple (index, data position, size) to hand to the access
1697
logic to get the record.
1699
node = self._get_node(version_id)
1700
bits = node[2][1:].split(' ')
1701
return node[0], int(bits[0]), int(bits[1])
1703
def get_method(self, version_id):
1704
"""Return compression method of specified version."""
1705
if not self._deltas:
1707
return self._parent_compression(self._get_node(version_id)[3][1])
1709
def _parent_compression(self, reference_list):
1710
# use the second reference list to decide if this is delta'd or not.
1711
if len(reference_list):
1716
def _get_node(self, version_id):
1718
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1720
raise RevisionNotPresent(version_id, self)
1722
def get_options(self, version_id):
1723
"""Return a string represention options.
1727
node = self._get_node(version_id)
1728
if not self._deltas:
1729
options = ['fulltext']
1731
options = [self._parent_compression(node[3][1])]
1732
if node[2][0] == 'N':
1733
options.append('no-eol')
1736
def get_parents(self, version_id):
1737
"""Return parents of specified version ignoring ghosts."""
1738
parents = list(self.iter_parents([version_id]))
1741
raise errors.RevisionNotPresent(version_id, self)
1742
return parents[0][1]
1744
def get_parents_with_ghosts(self, version_id):
1745
"""Return parents of specified version with ghosts."""
1746
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1747
check_present=True))
1748
if not self._parents:
1750
return self._keys_to_version_ids(nodes[0][3][0])
1752
def check_versions_present(self, version_ids):
1753
"""Check that all specified versions are present."""
1754
keys = self._version_ids_to_keys(version_ids)
1755
present = self._present_keys(keys)
1756
missing = keys.difference(present)
1758
raise RevisionNotPresent(missing.pop(), self)
1760
def add_version(self, version_id, options, access_memo, parents):
1761
"""Add a version record to the index."""
1762
return self.add_versions(((version_id, options, access_memo, parents),))
1764
def add_versions(self, versions, random_id=False):
1765
"""Add multiple versions to the index.
1767
This function does not insert data into the Immutable GraphIndex
1768
backing the KnitGraphIndex, instead it prepares data for insertion by
1769
the caller and checks that it is safe to insert then calls
1770
self._add_callback with the prepared GraphIndex nodes.
1772
:param versions: a list of tuples:
1773
(version_id, options, pos, size, parents).
1774
:param random_id: If True the ids being added were randomly generated
1775
and no check for existence will be performed.
1777
if not self._add_callback:
1778
raise errors.ReadOnlyError(self)
1779
# we hope there are no repositories with inconsistent parentage
1784
for (version_id, options, access_memo, parents) in versions:
1785
index, pos, size = access_memo
1786
key = (version_id, )
1787
parents = tuple((parent, ) for parent in parents)
1788
if 'no-eol' in options:
1792
value += "%d %d" % (pos, size)
1793
if not self._deltas:
1794
if 'line-delta' in options:
1795
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1798
if 'line-delta' in options:
1799
node_refs = (parents, (parents[0],))
1801
node_refs = (parents, ())
1803
node_refs = (parents, )
1806
raise KnitCorrupt(self, "attempt to add node with parents "
1807
"in parentless index.")
1809
keys[key] = (value, node_refs)
1811
present_nodes = self._get_entries(keys)
1812
for (index, key, value, node_refs) in present_nodes:
1813
if (value, node_refs) != keys[key]:
1814
raise KnitCorrupt(self, "inconsistent details in add_versions"
1815
": %s %s" % ((value, node_refs), keys[key]))
1819
for key, (value, node_refs) in keys.iteritems():
1820
result.append((key, value, node_refs))
1822
for key, (value, node_refs) in keys.iteritems():
1823
result.append((key, value))
1824
self._add_callback(result)
1826
def _version_ids_to_keys(self, version_ids):
1827
return set((version_id, ) for version_id in version_ids)
1830
class _KnitAccess(object):
1831
"""Access to knit records in a .knit file."""
1833
def __init__(self, transport, filename, _file_mode, _dir_mode,
1834
_need_to_create, _create_parent_dir):
1835
"""Create a _KnitAccess for accessing and inserting data.
1837
:param transport: The transport the .knit is located on.
1838
:param filename: The filename of the .knit.
1840
self._transport = transport
1841
self._filename = filename
1842
self._file_mode = _file_mode
1843
self._dir_mode = _dir_mode
1844
self._need_to_create = _need_to_create
1845
self._create_parent_dir = _create_parent_dir
1847
def add_raw_records(self, sizes, raw_data):
1848
"""Add raw knit bytes to a storage area.
1850
The data is spooled to whereever the access method is storing data.
1852
:param sizes: An iterable containing the size of each raw data segment.
1853
:param raw_data: A bytestring containing the data.
1854
:return: A list of memos to retrieve the record later. Each memo is a
1855
tuple - (index, pos, length), where the index field is always None
1856
for the .knit access method.
1858
assert type(raw_data) == str, \
1859
'data must be plain bytes was %s' % type(raw_data)
1860
if not self._need_to_create:
1861
base = self._transport.append_bytes(self._filename, raw_data)
1863
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1864
create_parent_dir=self._create_parent_dir,
1865
mode=self._file_mode,
1866
dir_mode=self._dir_mode)
1867
self._need_to_create = False
1871
result.append((None, base, size))
1876
"""IFF this data access has its own storage area, initialise it.
1880
self._transport.put_bytes_non_atomic(self._filename, '',
1881
mode=self._file_mode)
1883
def open_file(self):
1884
"""IFF this data access can be represented as a single file, open it.
1886
For knits that are not mapped to a single file on disk this will
1889
:return: None or a file handle.
1892
return self._transport.get(self._filename)
1897
def get_raw_records(self, memos_for_retrieval):
1898
"""Get the raw bytes for a records.
1900
:param memos_for_retrieval: An iterable containing the (index, pos,
1901
length) memo for retrieving the bytes. The .knit method ignores
1902
the index as there is always only a single file.
1903
:return: An iterator over the bytes of the records.
1905
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1906
for pos, data in self._transport.readv(self._filename, read_vector):
1910
class _PackAccess(object):
1911
"""Access to knit records via a collection of packs."""
1913
def __init__(self, index_to_packs, writer=None):
1914
"""Create a _PackAccess object.
1916
:param index_to_packs: A dict mapping index objects to the transport
1917
and file names for obtaining data.
1918
:param writer: A tuple (pack.ContainerWriter, write_index) which
1919
contains the pack to write, and the index that reads from it will
1923
self.container_writer = writer[0]
1924
self.write_index = writer[1]
1926
self.container_writer = None
1927
self.write_index = None
1928
self.indices = index_to_packs
1930
def add_raw_records(self, sizes, raw_data):
1931
"""Add raw knit bytes to a storage area.
1933
The data is spooled to the container writer in one bytes-record per
1936
:param sizes: An iterable containing the size of each raw data segment.
1937
:param raw_data: A bytestring containing the data.
1938
:return: A list of memos to retrieve the record later. Each memo is a
1939
tuple - (index, pos, length), where the index field is the
1940
write_index object supplied to the PackAccess object.
1942
assert type(raw_data) == str, \
1943
'data must be plain bytes was %s' % type(raw_data)
1947
p_offset, p_length = self.container_writer.add_bytes_record(
1948
raw_data[offset:offset+size], [])
1950
result.append((self.write_index, p_offset, p_length))
1954
"""Pack based knits do not get individually created."""
1956
def get_raw_records(self, memos_for_retrieval):
1957
"""Get the raw bytes for a records.
1959
:param memos_for_retrieval: An iterable containing the (index, pos,
1960
length) memo for retrieving the bytes. The Pack access method
1961
looks up the pack to use for a given record in its index_to_pack
1963
:return: An iterator over the bytes of the records.
1965
# first pass, group into same-index requests
1967
current_index = None
1968
for (index, offset, length) in memos_for_retrieval:
1969
if current_index == index:
1970
current_list.append((offset, length))
1972
if current_index is not None:
1973
request_lists.append((current_index, current_list))
1974
current_index = index
1975
current_list = [(offset, length)]
1976
# handle the last entry
1977
if current_index is not None:
1978
request_lists.append((current_index, current_list))
1979
for index, offsets in request_lists:
1980
transport, path = self.indices[index]
1981
reader = pack.make_readv_reader(transport, path, offsets)
1982
for names, read_func in reader.iter_records():
1983
yield read_func(None)
1985
def open_file(self):
1986
"""Pack based knits have no single file."""
1989
def set_writer(self, writer, index, (transport, packname)):
1990
"""Set a writer to use for adding data."""
1991
if index is not None:
1992
self.indices[index] = (transport, packname)
1993
self.container_writer = writer
1994
self.write_index = index
1997
class _KnitData(object):
1998
"""Manage extraction of data from a KnitAccess, caching and decompressing.
2000
The KnitData class provides the logic for parsing and using knit records,
2001
making use of an access method for the low level read and write operations.
2004
def __init__(self, access):
2005
"""Create a KnitData object.
2007
:param access: The access method to use. Access methods such as
2008
_KnitAccess manage the insertion of raw records and the subsequent
2009
retrieval of the same.
2011
self._access = access
2012
self._checked = False
2013
# TODO: jam 20060713 conceptually, this could spill to disk
2014
# if the cached size gets larger than a certain amount
2015
# but it complicates the model a bit, so for now just use
2016
# a simple dictionary
2018
self._do_cache = False
2020
def enable_cache(self):
2021
"""Enable caching of reads."""
2022
self._do_cache = True
2024
def clear_cache(self):
2025
"""Clear the record cache."""
2026
self._do_cache = False
2029
def _open_file(self):
2030
return self._access.open_file()
2032
def _record_to_data(self, version_id, digest, lines, dense_lines=None):
2033
"""Convert version_id, digest, lines into a raw data block.
2035
:param dense_lines: The bytes of lines but in a denser form. For
2036
instance, if lines is a list of 1000 bytestrings each ending in \n,
2037
dense_lines may be a list with one line in it, containing all the
2038
1000's lines and their \n's. Using dense_lines if it is already
2039
known is a win because the string join to create bytes in this
2040
function spends less time resizing the final string.
2041
:return: (len, a StringIO instance with the raw data ready to read.)
2043
# Note: using a string copy here increases memory pressure with e.g.
2044
# ISO's, but it is about 3 seconds faster on a 1.2Ghz intel machine
2045
# when doing the initial commit of a mozilla tree. RBC 20070921
2046
bytes = ''.join(chain(
2047
["version %s %d %s\n" % (version_id,
2050
dense_lines or lines,
2051
["end %s\n" % version_id]))
2052
assert bytes.__class__ == str
2053
compressed_bytes = bytes_to_gzip(bytes)
2054
return len(compressed_bytes), compressed_bytes
2056
def add_raw_records(self, sizes, raw_data):
2057
"""Append a prepared record to the data file.
2059
:param sizes: An iterable containing the size of each raw data segment.
2060
:param raw_data: A bytestring containing the data.
2061
:return: a list of index data for the way the data was stored.
2062
See the access method add_raw_records documentation for more
2065
return self._access.add_raw_records(sizes, raw_data)
2067
def _parse_record_header(self, version_id, raw_data):
2068
"""Parse a record header for consistency.
2070
:return: the header and the decompressor stream.
2071
as (stream, header_record)
2073
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
2075
rec = self._check_header(version_id, df.readline())
2076
except Exception, e:
2077
raise KnitCorrupt(self._access,
2078
"While reading {%s} got %s(%s)"
2079
% (version_id, e.__class__.__name__, str(e)))
2082
def _check_header(self, version_id, line):
2085
raise KnitCorrupt(self._access,
2086
'unexpected number of elements in record header')
2087
if rec[1] != version_id:
2088
raise KnitCorrupt(self._access,
2089
'unexpected version, wanted %r, got %r'
2090
% (version_id, rec[1]))
2093
def _parse_record(self, version_id, data):
2095
# 4168 calls in 2880 217 internal
2096
# 4168 calls to _parse_record_header in 2121
2097
# 4168 calls to readlines in 330
2098
df = GzipFile(mode='rb', fileobj=StringIO(data))
2101
record_contents = df.readlines()
2102
except Exception, e:
2103
raise KnitCorrupt(self._access,
2104
"While reading {%s} got %s(%s)"
2105
% (version_id, e.__class__.__name__, str(e)))
2106
header = record_contents.pop(0)
2107
rec = self._check_header(version_id, header)
2109
last_line = record_contents.pop()
2110
if len(record_contents) != int(rec[2]):
2111
raise KnitCorrupt(self._access,
2112
'incorrect number of lines %s != %s'
2114
% (len(record_contents), int(rec[2]),
2116
if last_line != 'end %s\n' % rec[1]:
2117
raise KnitCorrupt(self._access,
2118
'unexpected version end line %r, wanted %r'
2119
% (last_line, version_id))
2121
return record_contents, rec[3]
2123
def read_records_iter_raw(self, records):
2124
"""Read text records from data file and yield raw data.
2126
This unpacks enough of the text record to validate the id is
2127
as expected but thats all.
2129
# setup an iterator of the external records:
2130
# uses readv so nice and fast we hope.
2132
# grab the disk data needed.
2134
# Don't check _cache if it is empty
2135
needed_offsets = [index_memo for version_id, index_memo
2137
if version_id not in self._cache]
2139
needed_offsets = [index_memo for version_id, index_memo
2142
raw_records = self._access.get_raw_records(needed_offsets)
2144
for version_id, index_memo in records:
2145
if version_id in self._cache:
2146
# This data has already been validated
2147
data = self._cache[version_id]
2149
data = raw_records.next()
2151
self._cache[version_id] = data
2153
# validate the header
2154
df, rec = self._parse_record_header(version_id, data)
2156
yield version_id, data
2158
def read_records_iter(self, records):
2159
"""Read text records from data file and yield result.
2161
The result will be returned in whatever is the fastest to read.
2162
Not by the order requested. Also, multiple requests for the same
2163
record will only yield 1 response.
2164
:param records: A list of (version_id, pos, len) entries
2165
:return: Yields (version_id, contents, digest) in the order
2166
read, not the order requested
2172
# Skip records we have alread seen
2173
yielded_records = set()
2174
needed_records = set()
2175
for record in records:
2176
if record[0] in self._cache:
2177
if record[0] in yielded_records:
2179
yielded_records.add(record[0])
2180
data = self._cache[record[0]]
2181
content, digest = self._parse_record(record[0], data)
2182
yield (record[0], content, digest)
2184
needed_records.add(record)
2185
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2187
needed_records = sorted(set(records), key=operator.itemgetter(1))
2189
if not needed_records:
2192
# The transport optimizes the fetching as well
2193
# (ie, reads continuous ranges.)
2194
raw_data = self._access.get_raw_records(
2195
[index_memo for version_id, index_memo in needed_records])
2197
for (version_id, index_memo), data in \
2198
izip(iter(needed_records), raw_data):
2199
content, digest = self._parse_record(version_id, data)
2201
self._cache[version_id] = data
2202
yield version_id, content, digest
2204
def read_records(self, records):
2205
"""Read records into a dictionary."""
2207
for record_id, content, digest in \
2208
self.read_records_iter(records):
2209
components[record_id] = (content, digest)
2213
class InterKnit(InterVersionedFile):
2214
"""Optimised code paths for knit to knit operations."""
2216
_matching_file_from_factory = KnitVersionedFile
2217
_matching_file_to_factory = KnitVersionedFile
2220
def is_compatible(source, target):
2221
"""Be compatible with knits. """
2223
return (isinstance(source, KnitVersionedFile) and
2224
isinstance(target, KnitVersionedFile))
2225
except AttributeError:
2228
def _copy_texts(self, pb, msg, version_ids, ignore_missing=False):
2229
"""Copy texts to the target by extracting and adding them one by one.
2231
see join() for the parameter definitions.
2233
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2234
graph = self.source.get_graph(version_ids)
2235
order = topo_sort(graph.items())
2237
def size_of_content(content):
2238
return sum(len(line) for line in content.text())
2239
# Cache at most 10MB of parent texts
2240
parent_cache = lru_cache.LRUSizeCache(max_size=10*1024*1024,
2241
compute_size=size_of_content)
2242
# TODO: jam 20071116 It would be nice to have a streaming interface to
2243
# get multiple texts from a source. The source could be smarter
2244
# about how it handled intermediate stages.
2245
# get_line_list() or make_mpdiffs() seem like a possibility, but
2246
# at the moment they extract all full texts into memory, which
2247
# causes us to store more than our 3x fulltext goal.
2248
# Repository.iter_files_bytes() may be another possibility
2249
to_process = [version for version in order
2250
if version not in self.target]
2251
total = len(to_process)
2252
pb = ui.ui_factory.nested_progress_bar()
2254
for index, version in enumerate(to_process):
2255
pb.update('Converting versioned data', index, total)
2256
sha1, num_bytes, parent_text = self.target.add_lines(version,
2257
self.source.get_parents(version),
2258
self.source.get_lines(version),
2259
parent_texts=parent_cache)
2260
parent_cache[version] = parent_text
2265
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2266
"""See InterVersionedFile.join."""
2267
assert isinstance(self.source, KnitVersionedFile)
2268
assert isinstance(self.target, KnitVersionedFile)
2270
# If the source and target are mismatched w.r.t. annotations vs
2271
# plain, the data needs to be converted accordingly
2272
if self.source.factory.annotated == self.target.factory.annotated:
2274
elif self.source.factory.annotated:
2275
converter = self._anno_to_plain_converter
2277
# We're converting from a plain to an annotated knit. Copy them
2278
# across by full texts.
2279
return self._copy_texts(pb, msg, version_ids, ignore_missing)
2281
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2285
pb = ui.ui_factory.nested_progress_bar()
2287
version_ids = list(version_ids)
2288
if None in version_ids:
2289
version_ids.remove(None)
2291
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2292
this_versions = set(self.target._index.get_versions())
2293
# XXX: For efficiency we should not look at the whole index,
2294
# we only need to consider the referenced revisions - they
2295
# must all be present, or the method must be full-text.
2296
# TODO, RBC 20070919
2297
needed_versions = self.source_ancestry - this_versions
2299
if not needed_versions:
2301
full_list = topo_sort(self.source.get_graph())
2303
version_list = [i for i in full_list if (not self.target.has_version(i)
2304
and i in needed_versions)]
2308
copy_queue_records = []
2310
for version_id in version_list:
2311
options = self.source._index.get_options(version_id)
2312
parents = self.source._index.get_parents_with_ghosts(version_id)
2313
# check that its will be a consistent copy:
2314
for parent in parents:
2315
# if source has the parent, we must :
2316
# * already have it or
2317
# * have it scheduled already
2318
# otherwise we don't care
2319
assert (self.target.has_version(parent) or
2320
parent in copy_set or
2321
not self.source.has_version(parent))
2322
index_memo = self.source._index.get_position(version_id)
2323
copy_queue_records.append((version_id, index_memo))
2324
copy_queue.append((version_id, options, parents))
2325
copy_set.add(version_id)
2327
# data suck the join:
2329
total = len(version_list)
2332
for (version_id, raw_data), \
2333
(version_id2, options, parents) in \
2334
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2336
assert version_id == version_id2, 'logic error, inconsistent results'
2338
pb.update("Joining knit", count, total)
2340
size, raw_data = converter(raw_data, version_id, options,
2343
size = len(raw_data)
2344
raw_records.append((version_id, options, parents, size))
2345
raw_datum.append(raw_data)
2346
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2351
def _anno_to_plain_converter(self, raw_data, version_id, options,
2353
"""Convert annotated content to plain content."""
2354
data, digest = self.source._data._parse_record(version_id, raw_data)
2355
if 'fulltext' in options:
2356
content = self.source.factory.parse_fulltext(data, version_id)
2357
lines = self.target.factory.lower_fulltext(content)
2359
delta = self.source.factory.parse_line_delta(data, version_id,
2361
lines = self.target.factory.lower_line_delta(delta)
2362
return self.target._data._record_to_data(version_id, digest, lines)
2365
InterVersionedFile.register_optimiser(InterKnit)
2368
class WeaveToKnit(InterVersionedFile):
2369
"""Optimised code paths for weave to knit operations."""
2371
_matching_file_from_factory = bzrlib.weave.WeaveFile
2372
_matching_file_to_factory = KnitVersionedFile
2375
def is_compatible(source, target):
2376
"""Be compatible with weaves to knits."""
2378
return (isinstance(source, bzrlib.weave.Weave) and
2379
isinstance(target, KnitVersionedFile))
2380
except AttributeError:
2383
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2384
"""See InterVersionedFile.join."""
2385
assert isinstance(self.source, bzrlib.weave.Weave)
2386
assert isinstance(self.target, KnitVersionedFile)
2388
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2393
pb = ui.ui_factory.nested_progress_bar()
2395
version_ids = list(version_ids)
2397
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2398
this_versions = set(self.target._index.get_versions())
2399
needed_versions = self.source_ancestry - this_versions
2401
if not needed_versions:
2403
full_list = topo_sort(self.source.get_graph())
2405
version_list = [i for i in full_list if (not self.target.has_version(i)
2406
and i in needed_versions)]
2410
total = len(version_list)
2411
for version_id in version_list:
2412
pb.update("Converting to knit", count, total)
2413
parents = self.source.get_parents(version_id)
2414
# check that its will be a consistent copy:
2415
for parent in parents:
2416
# if source has the parent, we must already have it
2417
assert (self.target.has_version(parent))
2418
self.target.add_lines(
2419
version_id, parents, self.source.get_lines(version_id))
2426
InterVersionedFile.register_optimiser(WeaveToKnit)
2429
# Deprecated, use PatienceSequenceMatcher instead
2430
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2433
def annotate_knit(knit, revision_id):
2434
"""Annotate a knit with no cached annotations.
2436
This implementation is for knits with no cached annotations.
2437
It will work for knits with cached annotations, but this is not
2440
ancestry = knit.get_ancestry(revision_id)
2441
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2443
for candidate in ancestry:
2444
if candidate in annotations:
2446
parents = knit.get_parents(candidate)
2447
if len(parents) == 0:
2449
elif knit._index.get_method(candidate) != 'line-delta':
2452
parent, sha1, noeol, delta = knit.get_delta(candidate)
2453
blocks = KnitContent.get_line_delta_blocks(delta,
2454
fulltext[parents[0]], fulltext[candidate])
2455
annotations[candidate] = list(annotate.reannotate([annotations[p]
2456
for p in parents], fulltext[candidate], candidate, blocks))
2457
return iter(annotations[revision_id])
2461
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2463
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
